Ceramic Electron Devices: Metallurgical Research and Development

0 REPORT 12
METALLURGICAL RESEARCH AND DEVELOPMENT .

M FOR
CERAMIC ELECTRON DEVICES
I, FINAL REPORT
fr 1, 1 JULY 1962 - 30 JUNE 1965
TR-66-1 , * .
I" ~ ~~~R/E-66-115 . '" .
Prepared For
U.S. ARMY ELECTRONICS COMMAND

Fort Monmouth, New Jersey 07703
D. A. TASK NUMBER OST - 7776 -11 -017 -38

CONTRACT NUMBER DA 36- 039 SC- 90903
CLEARI NGH OUSE

FOR FEDERAL SCIENXirY AND
TE-CHNIcAL 1NF0'-1MA,7[ONk
Hardao op mMirofiOhe'
DIVISION OF VARIAN
301 Industrial Way
San Carios, California
DISTRIBUTION OF THIS DOCUMENT IS UNLIMITED

NOTICES
DISCLAIMERS
) /The findings in this report are not to be construed as

an official Department of the Army position, unless so
S- / designated by other authorized documents.
The citation of trade names and nai.ies of manufac-

turers in this report is not to be construed as official
Government indorsement or approval of commercial
proxducts or services referenced herein.
DISPOSITION
Destroy this report when it is no longer needed. Do

not return it to ihe originator.
S- - -- Ya
IReport No. 12
METALLURGICAL RESEARCH & DEVELOPMENT
rFOR
CERAMIC ELECTRON DEVICES
Final Report
Prepared for
U. S. ARMY ELECTRONICS COMMAND
FT MONMOUTH, NEW JERSEY 07703
D.A TASK NO. OST-7900-21-223-15
CONTRACT NO. DA 36-039 SC-90903
EIMAC, DIVISION OF VARIAN ASSOCIATES

301 Industrial Way
San Carlos, California
TR-66-1
R/E 66-115
D
, DISTRIBUTION OF THIS DOCUMENT IS UNLIMITED
REPORT NO. 12
FINAL REPORT
METALLURGICAL RESEARCH & DEVELOPMENT
FOR CERAMIC ELECTRON DEVICES
1 July 1962 - 30 June 1965
OBJECT OF STUDY:
To examine the basic mechanisms of adherence in ceramic-
to-metal seals and the problems connected with the use of
such seals in electron tubes. The rf properties of the
seal shall receive particular attention.
By: L. Reed
W. Wade
S. Vogel
R. McRae
C. Barnes
This research is a part of project DEFENDER

sponsored by the Advanced Research Projects
Agency, Department of Defense, under ARPA
Order No. 318-62 Project Code No. 7300, and
was conducted under the technical guidance
of the U. S. Army Electronics Command,
Fort Monmouth, New Jersey, 07703.
CONTRACT NO. DA 36-039 SC-90903 Tp e-

DA:TASK NO. OST-7900-21-223-15
L, Ranager
Research &Sp. Engineering
EIMAC, DIVISION OF VARIAN ASSOCIATES

Final Report
January 1966
DISTRIBUTTON OF THIS DOCUMENT IS UNLIMITED
w• mir m m( 1n • m m ,pP -.--. Im - -m m mum m m n mm •mp - - m l mmmmm

TABLE OF CONTENTS
1ECTION PAGE NO
1.0 PURPOSE 1
1.1 Purpose of Contract 1

1.2 Objective of Program 1
1.3 Tasks 1
2,0 ABSTRACT 3
3 C PUBITCAT IONS. LECTURES, REPORTS AND 5
CONFERENCES
4,C) TASK I SEAL PARAMETER STUDY 8
4,1 General Ceramic Considerations 10

4.2 Special Ceramics 15
4.2.1 Introduction 15
4.2.2 Microstructure Consider- 19
ations
4.2. 3 The Glassy Phase 34
4.3 Commercial Bodies- 101
Microstructure Considerations
4.4 Exploratory Metallizin9 Studies 107

4.4.2 The Paint 114
4.4.3 Firing Conditions 1.14
4.4.4 Sintering of the Metalli- 114
zing
4.4.5 Solution of Molybdenum 122
4.4.6 Adherence
4.4.7 Chemical Reduction 123
4.4.8 Glassy Phase Inter- 129
diffusion
4.4.9 Composition of the 138
Glassy Phase
4.4.10 Seal Strength versus 145
Metal 1izing Temperature
4.4.1] Seal Strength as a 147
Function of Brazing
4.4.12 initial Metallizing Tests 150
4.4.13 Comments on Sealing 158
Mechanisms
-7
SFCTI ON PAGE NO.
4.5 Studies Using the Special Bodies 159
4.5.1 Introduction of Seal 159

Parameter Study
4.5.2 Seal Parameter Studies 162
4.5.3 Supplementary Studies 200
4.6 Shear Test Study 220
4.6.1 Sample Geometry 220

4.6.2 Testing Methods 220
4.6.3 Test Results and 222
Discussion
4.7 References - Task I 227
5,0 TASK II, BERYLLIA AND QUARTZ STUDY 231
5.1 Beryllia Ceramics 231

5.2 Fused Quartz-to-Metal Seals 241
5.3 References - Task II 243
6.0 TASK III - ELECTRICAL STUDIES 244
6.1 General Introduction 244
6.1.1 Sources of Electrical 244

Loss
6.1.2 Electrical Properties as 245
a Function of Micro-
structure and Chemistry
6.1.3 Program of 248
Experimental Studies
6.2 d.c. Electrical Measurements 249
6.2.1 Temperature -oefficient of 249
d.c. Resistivity of
Metallizing
6.2.2 Effect of Wetting Layer 277
(Overplate) on the
Resistivity of the
Metallizing
OWN.-
SECTION PAGE NO.
6.3 rf CONDUCTION LOSS MEASUREMENTS 285

6.3.2 Selected Measurement 287
Technique
6.3.3 Sample Prepa3ration 287
6.3.4 Equipment and Experi- 291
mental Procudures
6,3.5 Equipment V'h-ckout 300
6o3o6 Results 306
6.3. 7 Discussion of Results 316
6.4 Dielectric Losses 350
A 4.1 Introduction 350

6.4.2 Sample Preparation and 353
Experimental Set-up
6.4.3 Experimental Procedure 359
and Results
6.4.4 Discussion 366
6.4.5 Comparison of Various 380
Metallizings on +99%
Alumina Bodies
6.5 High Power Measurements 390
6.5.1 Introduct ion 390

6.5.2 Calorimetric Test Cell 391
6.5.3 Cold Testing 397
6.5.4 Calorimetric Instrumentation 398
6.5.5 Calorimetric Measurements 404
6.6 References - Task i1 429
7.0 SUMMARY 430
7.1 Introduction 430

7.2 Basic Seal Considerations 430
7.3 Supporting Studies 433
7.4 Electrical Studies 439
8.0 CONCLUSIONS "2
9.0 RECOMMENDAT 'ONS 444
SECTION PAGE NO.
10.0 ACKNOWLEDGEMENTS 445
11.0 SUBCONTRACTS 446
12.0 IDENTIFICATION OF KEY TECHNICAL PERSONNEL 447
13.0 APPENDIX I -Factors Influential in Bonding 449

Ceramics to Metals
14.0 APPENDIX Il Detailed Table of Contents 454
- - owl V *
LIST OF FIGURES
FIGURE NO. PAGE NO,
4.01 Tensile test piece B 13
4.02 Tensile test piece A (CLM-15) 14
4.03 Revised pseudo-binary phase equilibrium 17

diagram of the CaO!/SiO 2 -AI20 3 system
derived from the ternary CaO-SiO 2 -Al 2 0 3
diagram
4.04 Revised pseudo-binary phase equilibrium 18
diagram of the 2 StO2/CaO-A. 2 0 3 system
derived from the ternary CaO-SiO 2-A1 2 0 3
diagram
4.05 Model of two spherical particles, radius R, 20

showing sintering by volume diffusion,
indicating the flux of vacancies (and
counter-flow of material) from a lens of
radius of curvature R, to an internal
grain boundary.
4.06 Surf3ce tension forces, F,due to a pressure 23

-2 mg acting as a solid-melt-
r
solid interface causing increased solubility
of the solid. r is the radius of curvature
indicited.
4.07 Particle size distribution of Alcoa A-14 25

alumina before processing.
4.08 Average griin size of 941, 942, 992 and 26
Body G 3s a function of firing temperature.
Time at temperature was three hours.
4.09 Miciistructure of 94i eramic (400X) 29
4.10 Microstructure of 942 ceramic (400X) 30
4.11 Microstructure of 992 ceramic (400X) 31

4.12 Modulus of rupture of the 941. 942 and 35
I 992 bodies
II
FIGURE NO. PAGE NO.
4.13 Atomic moiel of glass structure after 36
Zachariasen).
414 Electron photomicrographs ;3500X of 45
polished surfaces cf special alumina
bodies (a) 941 1700*1, .b 941 1650 ),
Ic 941 .1600', and )d941 C1550.
,The numbers in parentheses are the body
firing temperatures in °cz
4.15 Electron photomicrographs (3500X) of polished 46
surfaces of special alumina bodies
(a 942 (1750' .bv 942 1700)
.c) 942 1653', and id^- 942 (1600).
The numbers in parentheses are the body
firing, temperatures in O°l
4.16 Electron photcmicrographs '3500X) of 47
polished surfaces of special alumina bodies
(a) 992 (1800f. (b' 992 '1750), and
(c) 992 U1700. The numbers in parentheses
are the bcdy ftring temperatures in °C.
4.17 Schematic presentation of a typical glassy 50
phase region of a high alumina ceramic
showing the effects of polishing the
surface.
4.18 Electron photomicrograph of 941 alumina 56
ceramic at 2150OX magnification showing the
glassy phase between the grain boundaries
of alu'nina particles.
4.19 Electron photomicrograph of 942 ceramic at 57
21,50OX magnification showing the glassy
phase spreading over the alumina crystal
and in the grain boundaries between the
alumina ceramic.
4.20 Electron photomicrograph of 992 body at 58
21,500X magnification showing the glassy
pLase.,
4.21 Glass-sapphire corrosion samples with 1:1 64
SiO2/CaO glass containing no Al2 0 initially.
Samples were reacted at (a) 150oc (b) 1675oC
(c) 1750°C. The glass is the wide band across
the center of each photograph.
.... x II E U I
FIGURE NO.. PAGE NO.

SiO2/CaO glass containing 20% A1 2 03
initially. Samples were reacted at
(al 3500 0 C and ib) 1675 0 C. The glass is
in the band across the center of each
photograph (400X.

SiO2/CaO glass containing 40% A12 03
initiilly. Samples were reacted at
%a) 15000C and (b., 1675 0 C. The glass is
in the wide band across the center of the
top photograph and in the lower two-thirds
of the bottom photograph, (400X)
4.24 Viscosity versus A1 2 03 content for melts 74

with 1:1 SiC2 /CaO ratio. Dash lines
indicate probable viscosity of super-
cooled melt at indicated temperature.
4.25 Viscosity versus A1 2 03 content for melts 75

with 2:1 SiO2/CaO ratio. Dash lines
indicate probable viscosity of supercooled
melt at indicated temperature.
4.26 Self-diffusion values for oxyqen 1" 79

(1),silicon1 i (2) aluminum l- 13),and
calcium,' :4) for melt C and aluminum'3 (5)
for melt B. together with values of
"convective diffusion" for melts C (6)
and D (7) derived from alumina solution in
these melts 1 5 . Also included are esti-
mated Ca2 values for melt B (8) obtained
by extrapol.ating data contained in
reference 30. (See text for key to melts.)
4.27 Coefficient of thermal expansion versus a4

A1 2 0 3 content of three glass compositions.
4.28 Schematic diagram of contact angle experi- 86

ment and cross-section of sessile drop.
4.29 Schematic cross-section of furnace used 90

in wetting angle study.
4.30 Contact angle versus dewpoint of 1:1 93

5i0 2 /CaO melts on sapphire.
• -nnu -•
FIGURE NO. PAGE NO.
SiO 2 /CaO melts on sapphire.
SiO 2 /CaO melts on molybdenum.

SiO 2 /CaO melts on molybdenum.
4.36 Ceramographic cross-section of Body A 102
showing tabular alumina crystal form
and the glassy phase situated between
the grains. (400X . 1
4.37 Electron photomicrograph of Body A 103
showing the glassy phase in the
boundaries between the alumina grains.
(24,200X)
i
4.38 Electron photomicrogra b of Body A after 104
a thermal etch at II00-C in dry hydrogen.
(18,400X)
4.39 Electron photomicrogragh of Body A after i05
a thermal etch at 1100 C in wet
hydrogen. (28,OO0X)
I
4.40 Electron photomicrograph of Body A after 106
a thermal etch at 1100 0 C in air.
(28,000 X)
4.41 The microstructures of three +99% alumina 108
bodies investigated in the seal parameter
and electrical studies.
4.42 Particle size distribution for P-1 112
metallizing paint obtained by using a
Coulter counter.
4.43 Particle size distribution for P-4M 113
metallizing paint obtained by using a
Coulter counter.
. 1- Il -Xl!l;" Il I F. 1 I-Mp,
"M . i,, "W ' jW WM
FIGURE NO. PAGE NO.
4.44 Density versus composition of molybdenum- 118
glass composites sintered at 1425 C.
4.45 Density versus composition of molybdenum- L19
glass composites sintered at 1600 C.
4.46 Analysis of gases evolved from a metallizing 125
paint in the form of a layer 1" x 1" by
10 mils thick on alumina ceramic, under non-
equilibrium conditions.
4.47 Percent weight loss of P-1 metallizing on t26
Body A as a function of unfired sprayed
weight of coating.
4.48 Unfired sprayed weight of metallizing 127
paints P-I and P-2 on Body A versus
microscopically measured fired thickness.
Supersedes curve of third quarterly report.
Firing time was 1/2 hour at 1425 0 C.
4.49 P-2 metallizing on Body A sintered at 1425°C 128
for 1/2 hour in a H2 /3N2 , + 100 0 C dewpoint
atmosphere. The copper plate, copper braze
to cupro-nickel can be seen to infiltrate
the porous molybdenum sponge.
4.50 P-1 metallizing on Body A. Sinter fired 130
at 1425 0 C for 1/2 hour in a H 2 /3N2 , + 1000F
I dewpoint atmosphere. Close inspection of
the glassy interfacial layer between the
metallizing and ceramic reveals its crystal-
line nature. Copper plate, Au-Cu-Ni braze
to 70/30 cupro-nickel at 1040°C/l0 minutes.
4.51 Photomicrographs of P-1 on Body A showing
relationship between glassy interaction zone
thick.ss and metallizing thickness (400X).
4.52 Electron probe microanalysis trace scan
across Body A, P-1 metallizing processed at
1425 0 C, 1/2 hour,H 2 /3N2 atmosphere,+80°F
dew point.
4.53 Electron probe micro-analysis trace for 133
silicon across entire seal region giving
evidence of glassy phase migration.
FIGURE NO. PAGE NO.
34
4.54 Interpreted electro, probe microanalysis s.
trace averaged over four scans of P-i
metallizing on Body A.
4.55 Interpreted electron probe microanalysis 737

trace averaged over four scans of P-I
metallizing on Body A. Continuation of
Fig. 4.54.
4.56 Electron photomicrograph of metallizing 144

P-I on Body A (nicoro braze) metallized
at 1425 0 C for 1/2 hour. (8000X)
4.57 Seal tensile strength with P-1 metallizing 1.46
on Body A versus metallizing sintering
temperature.
4.58 P-2 metallizing on Body E, 99% alumina. .52

Metallized at 1425 0 C for 1/2 hour.
4.59 Top: Body H, 3 evaporated molybdenum -.

60
metallizing, copper braze, P-1 metal-
lizing - Body A.
Bottom: Body H, lx(, evaporated molybdenum

metallizing copper braze 70/30 cupro-
nickel metal member.
4.60 Lucalox* cylinder (top) and sapphire disc
(bottom) assembled for leak check and
tensile testing.
4.61 Plots of tensile strength of ceramic-to- 77
metal seals on 941 alumina bodies originally
fiied at 1550 0 C, 16000 C, 16500 C, and 17000C
versus metallizing temperature.
4.62 Plots of tensile strength of ceramic-to- I18
metal seals on 942 alumina bodies originally
fired at 16000 C, 1650 0 C, 17000 C, and 1750 0 C
*General Electric Co. Trade Name.
-. ,r, i,
FIGURE NO. PAGE NO.
4.63 Plots of tensile strength of ceramic-to- .79

metal seals on 992 alumina bodies
originally fired at 1700 0 C, 1750 0 C,
and 1800 0 C versus metallizing temperature.
4.64 Plots of peel strength bf ceramic-to- 180
metal seals brazed with copper on 941
alumina bodies originallg fired at 1550 0 C,
16000 C, 1650 0 C, and 1700 C versus
metallizing temperature.
4.65 Plots of peel strength of ceramic-to- 181
alumina bodies originally fired at
16000C, 16500 C, 1700 0 C and 1750°C
j versus metallizing temperature.
alumina bodies originallg fired at
17000 C, 1750 0 C, and 1800 C versus
metallizing temperature.
metal seals brazed with CuAg on 941
1550 0 C, 16000 C, 1650 0 C, and 1700 0 C
alumina bodies originally fired 8t
16000 C, 1650 0 C, 17000 C, and 1750 C
4.69 Plots of peel strengths of ceramic-to- 185
17000 C, 1750 0 C, 1800 0 C versus metal-
lizing temperature.
4.70 Photomicrographs of ceramic-to-. 89
metal seals on alumina body 941 (1550)
metallized with P-4M at 1300 0 C (top)
and 1800 0 C (bottom) showing the
structural changes which occur when
the ceramic is metallized at very high
temperatures. (400X)
* 4 "saw
FIGURE NO. PAGE NO.
4.71 Photomicrographs of ceramic-to-metal 191
seals on alumina body 942 (1650) (top)
and 992 (1800) (bottom) both metallized
with P-4M at 1800 0 C. Interlocking grain
structure may be seen at the interface
between the molybdenum metallizing and
the alumina surface. (400X)
seals on alumina body 942 (1700) made
with P-4M metallizing sintered at 1425 C
(top) and 1600 0 C (bottom). The glassy
phase migrated into the metallizing when
sintered at 1600 0 C, but not at 1425 0 C.
(4oOX)
seals on Lucalox* (top) and alumina ceramic
941 (1600) (bottom) both metallized with
P-4M at 1800 0 C. The molybdenum metallizing
was etched for 4 sec. with a solution of
potassium ferricyanide and sodium hydroxide.
(80oX)
4.74 Peel strength versus secondary crystallite 209
level in 941 ceramic after simulated braze
cycles in dry hydrogen at 1100 0 C for
various times. Kovar (K) or cupro-nickel
(CN) peel tabs subsequently brazed at
800 0 C with Cu-Ag.
4.75 Secondary crystallite level and peal 210
strength of 941 ceramic after simulated
braze cycles in dry hydrogen at 1100 0 C
for various times. Kovar geel tabs
subsequently brazed at 800 C with Cu-Ag.
cycles in wet hydrogen at 11000 C. Kovar
peel tabs subsequently brazed at 800 0 C using
Cu-Ag. Numbers indicate time of simulated
braze cycle.

FIGURE NO. PAGE NO.
4.77 Secondary crystallite level and peel 212
strength of 941 ceramic as a function of
the length of time of simulated braze
cycles at 1100 0 C in wet hydrogen. Kovar
peel tabs subsequently brazed at 8000 C
with Cu-Ag.

cycles in dry hydrogen at 11000 C for
various times. Kovar (K) on cupro-nickel
(CN) peel tabs subsequently brazed at 8000 C
with Cu-Ag.
4.79 Secondary crystallite level and peel 214
strength of 942 ceramic after simulated
braze cycles in dry hydrogen at 1100 C for
various times. Kovar peel tabs subsequently
brazed at 8000 C with Cu-Ag.
cycles in wet hydrogen at 11000 C. KNvar
peel tabs subsequently brazed at 800 C u3ing
Cu-Ag. Numbers indicate time of simulated
braze cycle.
4.81 Secondary crystallite level and peel strength 21A
of 942 ceramic as a function of the length
of time of simulated braze cycles at 1100 0 C
in wet hydrogen. Kovar peel tabs subsequently
brazed at 8000 C with Cu-Ag.
4.82 Shear test assembly consisting of two 221
cylinders joined with an O.D.-I.D. braze.
4.83 Concentric cylinder shear test specimen in ?23
loading position in Instron testing machine.
5.01 Microstructure of a 96% beryllia ceramic 232
(400X) which shows similar characteristics
to a 96% alumina ceramic.
5.02 Mici structure cross-section of a ceramic 233
metal seal of P-1 metallizing paint on a
96% beryllia ceramic (400X).
FIGURE NO. PAGE NO.
5.03 Photomicrographs of P-3 metallizing 234
sintered at 1325 0 C (top) and 1425 0 C
(bottom) on beryllia body BB. Cu-Au 60/40
braze, copper braze washer (400X).
5.04 Electron microprobe trace of seal made with 23(
metallizing paint P-3 on beryllia body BA.
5.05 Electron microprobe trace of seal made with 237
metallizing paint P-3 on beryllia body BB.
5.06 Photomicrograph of fused quartz-to-metal 242
seal. The thickness of the deposited
molybdenum layer is near the limit of
resolution of the microscope. (400X)
6.01 Schematic representation of Fig. 4.50 247

illustrating the electrical path length
and width through the metallizing.
6.02 Measurement of thermal coefficient of 251
resistivity in metallizing on ceramic.
Test sample strapped to ceramic heater.
6.03 Circuitry for measurement of thermal 253
coefficient of resistivity.
6.04 Resistance versus temperature of test 254
sample No. 10 (metallizing P-3 on Body A).
6.05 Measured resistance ratio RT/R o of three 255
samples as a function of temperature.
6.06 Measured resistance ratio vs temperature of 261
test samples. 1) molybdenum strip, O.0OQ'"
thick, unannealed. 2) molybdenum strip,
0.001" thick, unannealed. 3) molybdenum
strip, 0.001" thick, annealed. 13) P4-M
on Body A. 14) P4-M on Body E ceramic.
26) molybdenum, evaporated on Body E ceramic.
27) molybdenue evaporated on Body R ceramic,
fired at 1425 C for 1/2 hour in wet hydrogen-
nitrogen atmosphere.
FIGURE NO. PAGE NO.
6.07 Measured resistance ratios vs temperature of 262
test samples. 1) molybdenum strip, 0.002",
unannealed. 4) P-1 on Body A ceramic.
5) P-i on Body A ceramic, thickness .educed
by machining. 6) P-i on Body A ceramic,
partly leached with nitric acid. 18) P-i
on Body A ceramic, fired at 16000 C.
test samples. 1) molybdenum strip, 0.002",
unannealed. 7) paint P-1 on Body E ceramic.
8) paint P-i on Body E, thickness slightly
reduced by machining. 9) paint P-1 on
Body E thickness further reduced. 22) P-i
(modified with 1/6 molybdenum content of
regular P-1 paint) on Body E ceramic.
test samples. 1) molybdenum strip, 0.002"
thick, unannealed. 10) paint P-3 on Body A
ceramic. 11) paint P-3 on Body A ceramic,
thickness reduced by machining. 12) paint
P-3 on Body E ceramic.
6.10 Measured resistance ratio vs temperature of 265
test samples. 7) paint P-1 on ceramic
Body E. 24) paint P-1 on ceramic, slightly
electropolished. 25) paint P-1 on ceramic
Body E, heavily electropolished.
test samples. 4) paint P-i on ceramic
Body A. 15) paint Ti-li on ceramic.
16) paint Ti-li on ceramic Body A, slightly
machined. 17) paint Ti-l on ceramic Body A,
leached with nitric acid. 19) paint Ti-li
on ceramic Body A, heavily leached with nitric
acid. 28) paint Ti-li on ceramic Body E,
oxidized in air. 29) paint Ti-il on ceramic
Body E, heavily oxidized in air.
6.12 Photomicrograph (400X) of test samples. 267
Metallizings in a), b), c) have been over-
plated after the measurements.
wpm,'~~!P*W -"P"M --
FrGURE NO. PAGE NO.
6.13 Photomicrographs (400X) of test samples. 268
Metallizings have been overplated after
the measurements.
6.14 Photomicrographs (400X) of test samples. 269
Metallizings in a), b), c) have been over-
plated after the measurements.
6.15 Interaction of Ni wetting layer with P-1 280
metallizing on Body A ceramic. Relative
resistance as a function of processing
steps. Data obtained from four samples
of Group I, Table 6-4.
6.16 Interaction of Ni wetting layer with P-1 281
metallizing on Body A. Relative resistance
as a function of processing steps. Data
obtained from four samples of Group 2,
Table 6-4.
6.17 Interaction of Ni wetting layer with P-i 282
metallizing on Body A. Relative resistance
as a function of processing steps. Data
obtained from four samples of Group 3,
Table 6-4.
6.18 Interaction of Cu wetting layer with P-i 283
metallizing on Body A ceramic. Relative
resistance as a function of processing steps.
6.19 Relative resistance of metallized and over- 284
plated Body A substrates as a function of
sinterfiring. (1) 0.3 mil metallizing, 0.07
mil Ni-plate; (2) 0.3 mil metallizing, 0.02
mil Cu-plate.
6.20 RF conductivity test assembly. Showing layered 1.89
structure of an iron barrier layer seal and
showing back up ceramic used to reduce seal
stresses.
C.21 Block diagram of admittance bridge equipment 292

for measuring I R conduction losses in
metallizing.
FrGURE NO. PAGE NO.
6.22 Close-up of admittance bridge showing 293
Magic-T and matched resonators.
Ceramic samples in foreground.
6.23 Admittance bridge equipment. 294
6.24 Admittance bridge test -ssonator. 295
6.25 Plot of micrometer reading versus pi: lished 302

resistivity of the six calibration assemblies
(see text). Selected data from actual seals
included for reference.
6.26 Plot of relationship between concentricity 304
of ceramic-metal seal in test cavity,
&TIR. and change in measured cavity loss
(in mils of insertion of lossy material
into standard cavity) as sample holder is
rotated.
6.27 Qo of standard cavity used in conduction 305
loss measuremen" plotted as a function of
micrometer reading. Selected data giving
the various Q's of the test cavity with
various assemblies including the standard
are included for reference.
6.28 Skin depth at 9.15 Gc versus electrical 318
resistivity.
6.29 Photomicrographs (400X) of sample #226, a 3ze
seal with active metals evaporated or. a
polished ceramic (top), and sample #228
with copper evaporated on a polished
ceramic (bottom).
t
6.30 Photomicroqraphs of P-2 metallizing fired 3**9
at 1425 0 C (top) and 1550 0 C (bottom).
Subsequently copper plated. Fe barrier
layer, Cu-Ag braze (400X).
6.31 Conductive loss data for P-2 metallizing 330
on Body A. Metallizing sintered at 1425 0 C
unless otherwise noted.
&
' .
L' .,.' ' -- . .. , . . ,- : '" .. . _- .-. .. .. .. . . .
FIGURE NO. PAGE NO.
6.32 Conductive loss data for P-I on Body A. 333
Metallizing sintered at 1425 0 C.
6.33 Photomicrographs of Cu-Ag braze penetration 334
through Fe barrier layer into Cu (top) and
Ni (bottom) metallizing overplate (400X).
6.34 Microprobe analysis of iron barrier layer 335
seal. Data was averaged.
6.35 Microprobe analysis of iron barrier layer 336
seal described in text showing silver scan
across seal in the region of a grain boundary
discontinuity in the iron barrier layer.
6.36 Photomicrograph of conduction loss sample 337
#93 (P-i, Body A, Cu plate, Fe barrier
layer, Cu-Ag braze). (850X).
6.37 Microprobe analysis of Mo and Cu in 338
conduction loss sample #93 (P-1 Body A,
Cu plate Fe barrier layer, Cu-Ag braze).
Ag and Ni were below limit of detectability.
Scans were parallel to metallizing --
ceramic interface 0.0005 in. into
metallizing.
6.38 Schematic cross-section of ceramic-to- 340
metal seal with iron-nickel barrier layer
(top); photomicrograph (400X) of
sample #207 (bottom).
6.39 Plot of 12 R losses at 9.15 Gc versus 341
metallizing thickness for P-i on Body A
with overplate of copper or nickel. All
specimens were prepared with a barrier layer
and a CuAg eutectic braze and 70/30 Cu-Ni
metal member (see text).
6.40 Photomicrograph (400X) of sectioned sample 342
#205, showing failure of barrier layer and
detachment of portion of metallizing. See
Table 6-8 for data.
6.41 Photomicrographs (400X) of sectioned sample 343
#205 showing disruption of iron barrier
lhyer (dark) and partial dissolution of
copper overplate. See Table 6-8 for data.
FIGURE NO. PAGE NO.
6.42 Photomicrographs (400X) of samples #200 (top) 345
with 53 units of relative loss, #201 (center)
with 66 units of relative loss, #202 (bottom)
with 88 units of relative loss. See Table
6-8 for data.
6.43 Photomicrographs (400X) of sample #224 (top) 346
and sample #225 (bottom).
6.44 12 R (conduction) losses versus metallizing 3-49
thickness for P-1 on Lucalox' with overplate
of copper or nickel. The specimens were
prepared with a barrier layer and.a CuAg
eutectic braze (see text).
6.45 Resonator for dielectric loss measurements. 352
6.46 Redesigned dielectric loss cavity. 354
6.47 Block diagram of "Q" measurement for di- 356

electric loss data at ceramic-metal seal
interface.
6.48 Block diagram of "Q" measurement for di- 357
electric loss data at ceramic-to-metal
seal interface.
6.49 Typical data plot and Q calculation for 360
dielectric loss measurement.
6.50 Dielectric loss factor for selected ceramic 368

f bodies versus frequency. Values are from
manufacturers' data sheets.
6.51 Dielectric loss for alumina bodies versus 369
frequency.
6.52 The dielectric test cavity Q as a function of 370
ceramic disc thickness.
6.53 Loss factor at 2 Gc versus dielectric loss test

cavity Qo plotted from manufacturers loss
figures and test data on Body A, Body H and
Linde Sapphire. The line is drawn through the
values obtained during the second test series.
The Qo intercepts of three selected data points
are indicated for reference only (see text).

FIGURE NO. PAGE NO,
6.54 Photomicrggraph of Boly H test sample (400X) 3/5

#95 (500 A Ti, 5,000 A Mo evaporated, Cu
plated to 0.17 mil)
6.55 Reduction in cavity 0 resulting from noted 316
metallizing paint only on 0.060" thick
Body A ceramic discs. Test freauency 2 Gc.
Sintering time 1/2 hour at 1425 C unless
noted. Single surface metallized unless
noted.
6.56 Reduction in cavity Q resulting from noted 317
metallizing paint only on 0.060" thick
Body H ceramics 99.5% A120. Test frequency
2 Gc. Sintering time 1/2 hour at 1425 0 C
unless noted. Single surface metallized
unless noted.
6.57 Photomicrographs of Body H test samples 381
(400X). Top: #81A (1.6 mil of P-1 metalliz-
ing). Bottom: #82A (1.7 mil of P-1
metaliizing, leached).
6.58 Photomicrograph of Body H test sample 382

(400X) #78 S2 (0.5 ai.l of P-1 metallizing,
leached and sanded off).
6.59 Photomicrographs of P-1 (top) and P-7 (bottom) 383
on Linde sapphire. Dielectric loss samples
sintered one-half hour at 1425 0 C (400X).
6.60 Photographs of sectioned dielectric loss 385
specimens (0.200 in. wide by 0.060 in. thick)
showing P-1 metallizing diffusion into Body
992 (1750) at 1425 0 C with increasing time
(lox).
6.61 Photomicrographs of P-7 metallizing reaction 386
with Body 992 (1750) at 1425 0 C for increas-
ing times (400X).
6.62 Dielectric loss test cavity Qo with various 387
ceramic samples in gap. The unmetallized and
metallized samples were fired at 1425 0 C for
1/2 hour.
FIGURE NO. PAGE NO.
6.63 Dielectric loss test civity Qo versus 1425 0 C 388
firing time. Ceramic was Body 992 (1750)
in all cases.
6.64 Five-channel waveguide test cell used in 392
second series of high power calorimetric
loss measurement.
6.65 Cross-section of waveguide test cell shown 393
in Fig. 6.64 with three test seals.
6.66 Schematic diagram of test set-up for second 401
series of high power calorimetric loss
measurements shown also in Figs. 6.70,
6.71 and 6.72.
6.67 Schematic diagram of set-up used to 402
calibrate thermopiles for the second series
of high power calorimetric loss measure-
ments.
6.68 Photograph of thermopile calibration equip- 403
ment diagrammed in Fig. 6.67.
6.69 Block diagram of microwave circuitry used 405
in second series of high power calorimetric
loss measurements.
6.70 Test set-up for second series of high 407
power calorimeter loss measurements at
8kmc.
6.71 Close-up of test cells and thermopiles in 408
test set-up for second series of high power
calorimetric loss measurements.
6.72 Close-up of test set-up for second series 409
of high power calorimetric loss measure-
ments showing signal gen--'.. (8).M),
amplifier, and klystron at . r.
6.73 Power loss in dielectric-to-metal seal section 41.
of waveguide test cell vs metallizing thickness
as a fraction of transmitted power.
6.74 Sectioned seals in test cells No. 3 and No. 6. 419
FIGURE NO, PAGE NO.
6.75 Electron microprobe analyses of seals in 421
test cell No. 3 (top) and No. 6 (bottom).
6.76 Top: Body I ceramic with P-1 metallizing, 425
thin nickel plating and Cu-Ag braze.
Bottom: Body I ceramic with P-1 metallizing,
copper plating, and Cu-Ag braze showing
part of a braze gap.
7.01 Schematic representation of a metallized 432
layer on a) single crystal sapphire
b) a 99.5% alumina body c) a 94%
alumina body.
7.02 Diagrammatic representation of drum peel 434
and tab peel test fixturing.
7.03 a) Schematic diagram of seal showing 435
laminar construction (idealized).
b) Application of tensile stress to
laminate results in effective opposing
of compressive stresses in the lower-
modulus, nonmetallic phases, which
partially offsets the tensile stress.
The lower the modulus for a given material
strength (of the reaction zone material)
the greater the seal strength.
c) Application of shear stress is equally
felt by all layers of laminate and weakest
layer determines the seal strength.
7.04 Ceramic-to-metal seal, using Body E, showing 436
parting in the weakest layer of the seal.
Silver-copper braze alloy has penetrated
the unplated Kovar giving the weakest zone.
LIST OF TABLES
TABLE NO. PAGE NO.
4- 1 Nominal Composition in Weight Percent Li
of Ceramic Bodies Referred to in Text
4- 2 Manufacture of Special Ceramic Bodies 12
4- 3 Fired Densities versus Firing Temperature 2

(3 hours) for Special Bodies
4- 4 Glass Structure as a Function of Network 37
Modifier Ion
4- 5 Co-ordination Number and Bond Strength 39
of Oxides
4- 6 Anion Complexes in a CaO-Si0 2 -A120 3 Melt 41
of the 1:1 CaO/Si0 2 Composition and the
Correlation with Melt Viscosity (at
15000 C)
4- 7 Anion Complexes in a CaO-SiO 2 -Al201 Melt 42

of the 1:2 CaO/SiO 2 Composition and the
Correlation with Melt Viscosity
(at 15000 C)
4- 8 Weight Percent Glass Phase from 43

Photomicrograph Measurements
4- 9 Crystal Species Present in Ceramic 44

Bodies under Equilibrium Conditions
4-10 Summary of X-ray Diffraction Data on
Special Alumina Bodies (Interior-
Powder S.mples)
4-11 Some Properties of the Glassy Phase of 55
Special Alumina Bodies
4-12 Summary of X-ray Diffraction Data on 59
Special Alumina Bodies (Surface
Condition)
4-13 Electron Probe Microanalyses of Glass 61
Layer of Sapphire-Glass Sandwiches
- -,--- . - -,
TABLE NO. PAGE NO.
4-14 Average Results for Analyses of Glass 69
Zone of Glass-Sapphire Reaction
Couples
4-15 Contaminants in Phase 2 of Sample #343 72
4-16 Correlation of Sag-Bar Results on the 76

Special Ceramics with Viscosity and
Chemical Composition
4-17 Percent Thermal Expansion of Crystal .
Species from 25 C after Floyd3 6 .
4-18 Thermal Expansion Coefficients of c3

Glasses
4-19 Wettability of Seal Components at 88
1550-1600 0 C
4-20 Solubility of Metal in Melt 89
4-21 Wetting Angles of Glasses on Molybdenum 92

and Sapphire
4-22 Nominal Composition in Weight Percent 109
of Metallizing Paints Referred to in
Text
4-23 Composition of Metallizing Paints 110
Discussed in Text
4-24 Sintering of Molybdenum Powders in an 116
H2/3N2 atmosphere, 300 Dewpoint
4-25 Sintered Densities of Glass-Molybdenum 11/
Composites
4-26 Molybdenum Concentrations in Glasses
Melted at Temperature for 1/2 Hour in
Molybdenum Boats as Determined by
Spectrographic Analysis
4-27 Surface Appearance of Metallized Surfaces 139

After Acid Leaching Treatments
4-28 Summary of X-Ray Diffraction Analysis i40
TABLE NO. PAGE NO.
4-29 Relative Amounts of Anorthite Present 141
in Metallizing Layer (Peak Height) and
Occurrence of Anorthite Main Peak
4-30 Spectrographic Analysis of Leached 14;,
Metallized Surface Indicating the
Strongest FeO.A12 03 Concentration 2 - 2B
4-31 Spectrographic Analysis Data on Leachates 1.43

and Swabs from Acid Leached Metallizings
P-1 and P-2 on Body A
4-32 Tensile Test Results psi on Body A lfk
Metal-Ceramic Seals Brazed with Copper,
under Various FurLace Conditions
4-33 Peel Test Results (Pounds) on Body A 149
Metal-Ceramic Seals
4-34 Tensile Strength Results for Pure 151
Refractory Metal and Metal-Oxide Paints
on 94%, 97% and 99% Alumina Ceramics.
Copper Braze - No Washer
4-35 Tensile Test Data of the Special Paints 155

on Body A (94% Alumina) and Body H
(99.5% Alumina) Using a Copper Braze
and 20 Mil, 70-30 Cupro-Nickel
Metal Washer
4-36 Spectrographic Analysis of P-4M 163
Metallizing Paint
4-37 Metallizing Scratch Hardness 165
4-38 Metallizing Resistance, Ohms 167
4-39 Seal Leak Check 1/0
4-40 Seal Tensile Strength, psi I1.
4-41 Seal Peel Strength, lbs 173
4-42 Tensile Strength of Seals to Polished 192

and Unpolished Lucalox of Three Grain
Sizes
I
TABLE NO. PAGE NO.
4-43 Tab Peel Results 201
4-44 Secondary Phases Found in Ceramic- 205

Metallizing Interface by X-Ray
Diffraction Analysis
4-45 Peel Strength of Seals to Heat Treated 207 J
Ceramics Kovar Unless Otherwise
Indicated as Cupro-Nickel (cn)
4-46 Secondary Phases Present in Ceramic- 208 |
Metallizing Interface
4-47 Mechanical Properties of Two Frequently 219
Used Metal Members
4-48 Shear Test Results Obtained from 22.e j
Concentric Cylinder Specimen
Geometry
5- 1 Chemical Composition of Beryllia 5
Ceramics
5- 2 Beryllia Tensile Test Results I 39
5- 3 Beryllia Tensile Test Results II 240
6- 1 Measured Thermal Coefficients of dc ki "

Resistivity
6- 2 Metallizing Dimensions and Resistances 0
of Samples in Table 6-1
6- 3 Computed Thermal Coefficients of 476
Electrical Resistivity
6- 4 Thicknesses of Metallizing and Plating .78
on Representative Test Samples.
Measured Resistances, Surface Resistances
and Deducted Resistances
6- 5 Conduction Loss Data at 9.15 Oc on 301

Body A Test Pieces Glued to Noted
Sample Holder
6- 6 Series 1, Conduction Loss Data at 9.15 Oc 307

on Body A Test Pieces Sealed to Cupro-
Nickel (70-30) with Copper Braze
- -~ ______IW O
TABLE NO. PAGE NO.
6- 7 Series 2, Conduction Loss Data at 308
9.15 Gc on Body A Test Pieces with
Ag-Cu Braze Sealed to Cupro-Nickel
(70-30)
6- 8 Series 3, Barrier Layer Seals 311
6- 9 Series 3, Loss in Standard Ceramic-to- 312
Metal Seals
6-10 Series 3, Evaporation Metallized Seals 313
6-11 Non-Metallic "Seals" 315
6-12 Room Temperature Resistivity of 325
Selected Copper, Nickel and
Molybdenum Alloys
6-13 Room Temperature Resistivity of 347
Selected Braze Materials
6-14 Series 1, Dielectric Loss Data on 361

Body A (94% A12 0 ) as Received and When
Painted with Notid Metallizing
6-15 Series 1. Dielectric Loss Data on 36k
Body H (99.5% A12 0) as Received and
When Coated with N~ted Metallizing
6-16 Series 2, Dielectric Loss Data at 2Gc 364

on Body H (99.5% Al 0 ) and Sapphire
with Noted Metallizna. Sintered
! 1/2 Hour at 1425 0 C. No Plating.
Ceramic Discs Nominally 0.060 inches
Thick
6-17 Dielectric Loss Data at 2 Gc on 365
Body 992 (1750) with Noted Metallizing.
Sintering Time at 1425 0 C Noted.
No Plating. Ceramic Discs Nominally
0.060" Thick
6-18 Series 3. Dielectric Loss Data at 2Gc 367

on Body H Test Samples with Noted
Metallizing and Subsequent Treatment
W- - - -
I
TABLE NO. PAGE NO.
6-19 Dielectric Loss Increase in Body H 373
(99.5% A1 2 03 ) Resulting froma
Application of P-3 (g), SiO 2 , TiO 2 ,
MnO 2 Glass, to One Surface
6-20 Physical Characteristics of Test Cells 3q
6-21 Test Cell VSWR Measurement Results

6-22 High Power rf Loss Measurement
Results
I
I
I
:;. __ _
1.0 PURPOSE
1.1 Purpose of Contract
The purpose of the contract was to investigate
the factors contributing to radio frequency power losses in
ceramic-to-metal seals and to examine the basic mechanisms
of adherence in these seals and the problems connected with
their use in electron tubes. Data obtained by other workers
gave evidence that heating of the seal region, as distinct
from the main portion of the window occurs. This heating is
due to rf interactions with the materials in the seal region.
Therefore, the specific objectives of this contract were
to investigate the factors contributing to rf losses in the
seal region and to gain a more fundamental understanding of
the various phenomena which combine to effect a mechanically
strong, vacuum tight seal with desirable electrical proper-
ties.
This ccntract is part of the window investigations
being conducted under Project DEFENDER, a program involving
the development of superpower microwave tubes. Other
window work under Project DEFENDER included Contract DA
36-039 SC-90818 with Eitel-McCullough on multipactor
suppression.
1.2 Objectives of Program
The objectives of the program were the following:
a) To gain a fundamental understanding of the pertinent
parameters involved in effecting a ceramic-to-metal seal,
b) Measurement of the rf losses of ceramic-to-metal seals
in order to learn more about theio -ource and nature and
their relationship to the metallurgical properties of such
seals, C) Evaluation of research data accumulated during
the proram, to provide a basis for the fabrication of seals,
with improved mechanical and electrical characteristics, and
model seals especially suitable for further fundamental
studies.
To achieve the objectives of the program, the

following tasks were pursued:
Task I - Investigation of the Pertinent
Parameters of the Formation of
Ceramic-to-Metal Seals
1
- - ....--
Seal parameter studies included work on the
following factors which affect seal properties:
a) Diffusion rates and viscosity of the components
of the seal.
b) Sintering mechanisms of metallizing particles.
c) Corrosion of alumina by seal components.
d) Surface tension relationships among the differ-
ent phases of the seal.
e) Thermal expansion of the various phases
present in the seal.
Task 2 - Preparation of Seals for Beryllia and
Fused Quartz
Beryllia-to-metal and fused quartz-to-metal seals
were prepared utilizing the information and experience
gained in these studies for the improvement of seals and
sealing techniques.
Task 3 - Radio Frequency Studies Designed to
Obtain a Quantitative Understanding of
Seal Losses
Radio frequency work included studies of the effect
of temperature on the seal losses and determination of how
and to what extent the metal, braze, metallizing, and
ceramic contribute to these losses.
2
2.0 ABSTRACT
Fifteen special 94% and 99% alumina ceramics in
the system SiO2-CaO-Al 203 with a Si02/CaO ratio of 1:1 and
2:1 were fabricated and their behavior characterized.
A combination of phase diagram interpolations, density and
strength determinations, optical and electron microscope
planemetric and crystal growth studies, X-ray diffraction
and electron microprobe studies, and chemical analysis of
their glassy phases, established that equilibrium was very
nearly reached at their maturing temperature. Cooling
tended to freeze in the high temperature state although
crystal phases such as anorthite, gehlenite and calcium
hexaluminate were partially precipitated in some cases.
It was tentatively established that the rate

determining mechanism of alumina solution both during the
initial processing of the ceramic and during its subse-
quent reaction during metallizing with refractory metal
paints containing oxide additions, was the chemical
reaction rate.
The composition of the bulk and surface grain
boundary phases of the ceramic was used to determine the
viscosity of its melt phase. This, in turn, was used to
predict the occurrence of highly electrical conductive,
high strength, vacuum tight, ceramic-to-metal seals when
metallized with a pure metal paint. Devitrification,
thermal expansion, sintering, surface energy, dissolution,
diffusion, crystal size, microscope, X-ray diffraction,
electron microprobe and spectrographic studies were also
used to predict the occurrence of high and low strength
vacuum tight ceramic-to-metal seals on the twelve complex
metallizing paint systems investigated.
Lucalox*, sapphire, beryllia and fused quartz

vacuum tight seals were made and their sealing mechanisms
were examined as were those of several commercial alumina
ceramics.
It was established that chemical and/or semi-

conducting bond mechanisms were responsible for the actual
sealing process of refractory metal to the oxide substrate.
*General Electric Company trade name for a 99.75% alumina

body.
0m. • •
For this bonding to occur either the metal phase had to
be deposited on the oxide substrate by vapor phase
processes or the oxide phases had to move by viscous flow
processes (e.g., glassy phase migration) in order that
the initial atom to atom contact could be made.
The electrical studies included dc resistance

measurements, low power rf conduction and dielectric loss
studies, and high power loss studies. It was shown that
the metallizing layer is made up of metallic, semi-conductive
and insulating layers. Conduction losses are minimized by
using thin metallizing and a copper braze. Dielectric
losses are only important for +99% aluminas and they
account for a quarter of the total metallizing losses.
The total metallizing losses are about 25% of the total
window losses which, in turn, are about 0.3% of the
transmitted power at 10 Gc.
In addition to usage on high power klystrons.

the low loss metallizing development in this study can
profitably be used on planar triodes and reflex klystrons.
4
3.0 PUBLICATIONS. LECTURES, REPORTS AND CONFERENCES
The following conferences were held and reports

presented during the final quarter tc discuss progress and
to disseminate information gained during the course of the
contract. All previous publications lectures, reports,
and conferences were listed in the eleven preceding
quarterly reports.
On May 6th, L. Reed, Eitel-McCullough, Inc.,

presented a review of the complete program on ceramic-to-
metal seals at the USAECOM, Fort Monmouth New Jersey.
On May 18th a meeting was held at EIMAC. San Carlos,

California. to discuss the factors influential in bonding
ceramics to metals. Present at the meeting were other
contractors who were contributing to the ARPA sponsored
project DEFENDER, as well as a representative from USAECOM.
They were Stanley Leefe. USAECOM7 0. Heil, Heil Scientific
Laboratories: J. White, Sperry Electronic Tube Division;
S. Colgate. Florida University; L. Reed, EIMAC; W. Wade,
EIMAC. The conclusions of that meeting are presented in
Appendix I of this report.
On June 15th L. Reed of Eitel-McCullough, Inc.,

presented a summary of the program to Col. Hill of ARPA at
the Pentagon. Washington D. Co Also .cesent were
P. M Lally and A. Saharian of Sperry-Gainsville (Sperry
Electronic Tube Division.), L. Feinstein of Stanford
Research Institute, L. Kaplan, G. Wurthmann, I. Reingold
and Miss B. Malley of USAECOM. and W. Wade of Eitel-
McCullough. Inc.
On October 20th L. Reed of Eitel-McCullough

resented the paper "The Electrical Behavior of Ceramic-
to-Metal Seals" by S. Vogel. W. Wade, R. McRae and L. Reed,
at the International Electron Devices Meeting in Washington.
D. C.
PAPERS AND PUBLICATIONS FOR CONTRACT NO. DA 36-039 SC-90903
1. L. Reed, "Micrcstructure of Output Windows", Paper pre-

sented at meeting on window problems at Varian Associ-
ates, Palo Alto, California, February 6-7, 1963.
2. L. Reed, "General Discussion of the Role of the Electron

Probe Microanalyzer in the Investigation of Ceramic-to-
Metal Sealing" , Paper presented to Signal Corps personnel
at Fort Monmouth, N. J., sponsored by L. L T' pioi-' of
USA".: February 7, 1964.
3. L. Reed and W. Wade, "Discussion of Metallurgical Research

and Development for Ceramic Electron Devices", Paper
presented to students in the School of Mineral Technology,
University of California. Berkeley, California, February II,
1964.
4. L. Reed, 'Survey of the Factors Affecting Ceramic-to-Metal
Sealing", Paper presented at the Fall Meeting of the
Electronics Division of the American Ceramic Society,
Philadelphia, Pennsylvania, September 16-18, 1964,
5. L. Reed, "Progress in Ceramics and Ceramic Seals", Paper

presented at the Seventh National Conference on Tube
Techniques, New York, N. Y., September 28-30, 1964.
6. R. McRae, C. Barnes and L. Reed, "rf Losses in Ceramic-to-

Metal Seals", Paper presented at the Seventh National
Conference on Tube Techniques, New York, N. Y., September
28-30, 1964.
7. S. Vogel, "Calorimetric Measurements of Radio Frequency

Losses in Dielectric-to-Metal Seals": Paper presented
in a Technical Seminar airanged by L FK,"'aji at Fort
Monmouth, N. J., October 2, 1964.
8. L. Reed, "Report on the Metallizing Workshop in Phila-

delphia", Paper presented at the Pacific Coast Regional
Meeting of the American Ceramic Society, San Francisco,
California, October 29, 1964.
9. L. Reed and R. McRae. "Evaporated Metallizing on Ceramics',

American Ceramic Society Bulletin, vol. 44 .). 12-13,
(1965).
6
10. L. Reed, S. Vogel, W. Wade and R. McRae, "The Electrical
Behavior of Ceramic-to-Metal Seals", Paper presented
at the International Electron Devices Meeting, Washing-
ton, D C., October 20, 1965.
7
4.0 TASK I SEAL PARAMETER STUDY
A metallized ceramic made by a conventional

refractory metal metallizing process* is composed of a sin-
tered molybdenum porous sponge which has been infiltrated
by a solidified oxide melt during a high temperature
sintering process. On cooling the solidified melt phase,
sometimes referred to as the "glassy phase". bonds both to
the ceramic and the porous sponge to form a strong vacuum
tight metallizing layer. The seal melt phase is formed
from the grain boundary phase(s) of the ceramic and/or
non-metallic addition(s) to the metallizing paint.
A complete ceramic-to-metal seal is made by

brazing this metallized ceramic to a metal member A
wetting layer which can be plated on is sometimes employed
to facilitate a good wetting action by the braze. Many
attempts have been made to further characterize the ceramic-
to-metal sealing process, but the interrelation of the many
phenomena involved, which include chemical reaction,
viscosity, diffusion, sintering, surface energy (wetting),
have made the task very difficult.
As a first step toward understanding these

phenomena, a comprehensive interpretative analysis of
ceramic-to-metal sealing technology was presented in the
1st, 2nd, 3rd and 4th Quarterly Reports. This review is not
repeated in this final summary report.
Concurrent with and following the above analysis,

experimental work was carried out on both "simple" and
complex matallizing sy tems.
The simple system consisted of a single phase

molybdenum metal paint metallized onto well-characterized
one or three component "special ceramic bodies". It was
hoped to be able to treat the subject from a quantitative
scientific viewpoint. In practice. the interrelation of
the many phenomena involved highlighted the difficulties of
attaining such a goal within the modest effort allowed for
this task.
On the other hand, it is also important to obtain

a qualitative understanding of commercial complex metalliz-
ing systems. The success achieved in analyzing a complex
*Unless speclf.Lcally mentioned discussion of metallizing

in this report refers to this process.
m - ___m_ _____ _____ ___

system by carrying out selected "exploratory studies" was
gratifying.
The following sections present the pertinent
observations and experiments conducted concerning the
various components constituting a ceramic-to-metal seal;
the ceramic, the metallizing, the wetting and brazing layer,
and the metal member. The investigations are brought into
focus by considerations of the total seal.
Both commercially obtainable and specially made
ceramic bodies were studied extensively throughout this
program. A major effort was madei ;:o -'ezt..r.zet-e- grain
boundary phase of the ceramic- Experiments were conducted
to determine if the grain boundary phase has attained
equilibrium at temperature. The properties of the grain
boundary phase such as its chemical composition, viscosity,
diffusion, surface tension, thermal expansion were determined.
This information is of value in interpreting the data obtained
in the subsequent studies.
Key "exploratory studies" on all aspects of the

metallizing process are rext reported. The most significant
key experiments outlined here concern the use of the electron
microprobe to follow the movement of the various ionic species
in the seal volume,
Finally the "special bodies" are used to

systematically examine the complex seal process. The
statistical nature of the strength property of a seal is
very evident in an examination of this data as also are
opposing trends of various properties- Thus for example,
the increased stress induced in a Kovar-to-cerAmic seal
brazed at 1110 0 C rather than 800 0 C can be ofL-et by the
partial devitrification of the glassy layer,but its complete
devitrification augments the critical stress situation
Some attempt is made in the final section to

develop a shea. test for testing ceramic-to-metal seals.
Much work on the measurement and analysis of stresses in
ceramic-to-metal seals needs to be done and the present work
should only be considered a first step in this direction.
I - ~q w -..-
4.1 General Ceramic Considerations
Both commercially obtairtable and specially made
ceramics were investigated on this program.
Commercially obtainable ceramics are complex in
formulation due to the varied raw materials used. Further-
more. their processing is proprietary so that a rigorous
discussion of the end product cannot be attempted, While
much of the present study was conducted with commercial
ceramics, special bodies were also fabricated to overcome
the above objections,
The chemical composition of the ceramics considered

on the program is given in Table 4-1. The shapes of the
special bodies are given in Table 4-2.
The special bodies were manufactured after a
careful review was made of typical commercial bodies,
including those listed in Table 4-1, which seemed to give
rise to the widest variation in metallizing behavior.
The microstructure and physical properties of the special
ceramics were Lnvestigated with particular attention being
paid to the so-called "glassy phase" of the ceramics as
distinct from the alumina crystal phase and the pore volume.
The special 99% body and one 94% body have
SiO 2 /CaO in the ratio 2!1. A second special 94% body has
the CaO/SiO 2 ratio 1:1 A total of four hundred tensile
test pieces were made of each body together with suitable
specimens for thermal stress analysis and rf loss measure-
ments. The tensile test pieces were used for peel test
measuremenys also; they conform to the Sperry design
Fig. 4.01. The thermal stress cylinders are I" x 1-7/8'
x 0.125" wall. The ends are flat to 0.2 mil.' The thermal
sag bars are 6' t- 0.1" long by l" ± wide by 0.18" ± 0.005"
thick. The low power rf loss specimens are in the form of
flat rings 1,203" OD x 0.703" [D x 0.100" high and disc
shapes 0.200" diameter x 0.60" thick.
The commercial bodies had tensile and peel test
pieces made from them which conformed to the ASTM CLM-15
design Fig. 4.02 2 The rf loss specimens are of the
*The cylinders were fabricated for use in the event that a

future study on ceramic microstructure would be programmed.
10
- '~ ~ ~ ~ -'
TABLE 4-1
Nominal Composition in Weight Percent* of Ceramic Bodies

Referred to in Text
Network (+) (+) (+)
Intermediate Network Former Network 1odifierMO
Body A1 20 3 Si0 2 ZrO 2 CaO HgO
Special Bodies
941** 94 3 - 3 -
942"* 94 4 - 2 -
992** 99 0.6 - 0.3 -
Lucalox + 99.75 - - 0.25
Sapphire 100 - - -
Commercial Bodies
A 94 4 0.5 1 0.8
97 1.5 - 1.2 -
E 99 0.6 - 0.15 0.15
G 94 4 - 1 1
H 99.5 0.25 - 0.05 0.25
I 95.5 2.0 - 0.5 2.0
'Best estimate from spectrographic analysis.
"Firing temperature in °C usually included in parentheses after

composition code, i.e.. 992 (1730).
*See section 4.2.3.1.

++General Electric Co. Trade Name.
11
TABLE 4-2
Manufacture of Special Ceramic Bodies
Firing Temperature °C
_X 1550 1600 1650 1700 1750
99% Body
Tensile
Test Pcs. 200 - 50 50 50 50
Thermal Stress
Cylinders 10 - 4 4 4 4
ri Discs 20 - 10 10
10 10
f Rods 10 - 5 5 5 5
94%/ Bodies
Tensile
Test Pcs. 150 50 50 50 50 50*
'hermal Stress
Cvi inders 10 4 4 4 4 4*
rf Rings 20 10 10 10 10 10
rf Discs 10 5 5 5 5 5
Thermal Sag
Bars 4 1 1 1 1 1"
Notes: Time at temperature is three hours.
X was 1750 0 C for 992. 1700°C for 942 and 1650°C for 941.
*Substitute 1500 0 C firing for the 1750 0 C firing for the

941 body.
12
1.5 0't 0015"
moo- 0625"± 0 005"

- - 0402"t 0.005"
0.52 32FA± BRAZED S 4URFAC

ES0
0.52 0.0a075"
TAN 90 .7
1 0 j3d 0
I Figure 4.01: Tensile test piece B
-13-
-1.115" ±010"
- 0625" ±0.005"
- - - 0402' t0.005'
GROUND FACE
METALLIZED FLAT TO 0.00025"
SURFACE
PARALLEL TO
7040' REFI ± 0.001
0.6 T- 8 -\lFLAT
REF
Figure 4.02: Tensile test piece A (CLM-15)
-14-
'-1 7 -
%joietrv ulven above and also in the form of sheets 25 and
50 mil thick x 0.680" ' 1.360" for hiqh power window tests
and bars 0.300" x 0.5uj" x 6.0" for dc1:ctrlcai
measurement tests.
Further details on the above and other special

will be found under the appropriate section later
:':,.:''S
in the text.
4.2 Special Ceramics
4.2.1 Introduction
The basic study portion of the program dictate-3

that the characteristics of the ceramic be well known.
With this in mind, the three component CaO-SiO2-Al 2 () 3
;;ystcm was selected and ceramics manutactured within this
system according to the schedule laid down in Table 4-2.
Summarizing. the following compositions were

selected for study:
1. 94% alumina CaO/SiO 2 ratio 1:1, desiqnated

941.
2. 94% alumina CaO/SiO 2 ratio 1:2, designated
942.
3. 99% alumina CaO/SiO 2 ratio 1:2, designated
992.
4. 99.75% alumina, Lucalox*, 0.25% MgO.
5. 100% alumina single crystal sapphire.
The rationale'behind the selection of the above

bodies is as follows: The 94% compositions were selected
on two counts.
i. more easily detectable concentration

gradients could be detected using the
electron microprobe and dielectric loss
measuring equipment than would be the case
for the 992 alumina.
2. the study is oriented to basic sealing

mechanisms of all high grade aluminas.
15
The trend in high power rf windows is to use
the high aluminas. +97%; hence a reason for the inclusions
of the latter three bodies.
Sapphire, however was not primarily included

for study on the basis of being a material of use in
electron tubes but rather to provide useful information
on sealing and rf loss mechanisms.
Lucalox* was chosen for similar reasons as it was

considered to be the simplest cecamic available being
glass free and pore free- Lucalox* does have useful output
window characteristics. The 0 25% of magnesia is combined
with alumina as MqO.A1 20 3 spinel in the grain boundaries of
the alumina grain, Three average grain sizes were obtained
5- 20 t and 40.(, so that the effect of grain size could
be studied independently of other complicating factors.
Most commercial ceramics contain four or more

oxides; the selection of a three-oxide system allows quanti-
tative predictions to be made with regard to the amount of
glassy phase and crystalline phases present if the phase
diagram is known. Representation of selected compositions
on a two-dimensional phase diagram may be easily made.
Figures 4.03 and 4.04 show such cuts through the ternary
CaO-SiO 2 -A1 2 0 3 phase diagram as compiled b Osborn and Muon'
with later revisions by Gentile and Foster Suppoyt'np
viscosity surface tension " density diffusion
and chemical reactivity datal" - 1-6 =re iva.lEle for the
CaO-SiO 2 -A1 2 0 3 system
In addition to the considerations discussed above

the selection of CaO over MgO and A1 2 0 3 over BeO was based
on the fact that the light elements Mg and Be are below the
quantitative detection level for electron probe micro-
analysis with the equipment available tu the project.
Beryllia ceramics are considered in Task IT of this
report )
The above range of ceramics enables us to inter-

polate results for ceramics in the range 94% - 100O alumina
and also enables extrapolations to be made within the system.
General concepts should also be applicable to more complex
diverse situations This latter statement assumes that the
characteristics of the ceramic being used are known. This
is not always the case.
16
I
- -- ' . -.-. _ _ .,,- _ . "q , '"i
0
N 00
V 0
\'~~~ 0 _
I0
N 4.
+
i.I +~~ U. ______ o
0 a0
J.
+ Z0 +
0 ,000
\
0
4)
3""
N0*
N 00 W N
0 r
\+ N
+00 0
@0n -e
04O
00 Nto 0
0 0 F')0
CL+
'00
5% *I 0.
00 4c cg
NOO
0 :3 + +
•- ooI 0 OU
VHI\ M
0- a + 0
o+
0: + o
N I?
000
NO 0
Figure 4.03: Revised pseudo-binary phase equilibrium diagram of the CaO/SiO 2-A1 2 0 3 system
derived from the ternary CaO-SiO.2 -AI 20 3 diagram.
-17-
00
'-i4
P.-0
oo
+ N
0
N 0
I 01
0o 0
.'
0
~. _V." CID
+0 +
000
C3 N ,I *1
o
&gf U I-V N N
O Vi- 0
- -00 ,1-
00 00
_j +
0.00+
CL 0 +! 0 + -EO
0+N ~00-1
00
-I-+
0 0 0 0 0 0
Figure 4.04. Revised pseudo-binary phase equilibrium diagram of the 2 SiO 2 /CaO-AI20 3 system
derived from the ternary CaO-SiO 2-AI 20 3 diagram.
18-
4.2.2 Microstructure Considerations
Alumina ceramics consist of

i. A grain boundary phase(s)
ii. Predominantly an alumina crystalline phase
iii. A pore volume.
The grain boundary phase(s) will, in the case of

the 941 and 942 ceramics, consist of CaO,SiO 2 and A1 2 0 3
components. At the temperature at which the ceramics are
made, these components will react to form a molten phase.
This "melt" phase will e"entually reach chemical equilibrium
and be saturated with aluminal5. If chemical equilibrium
is attained, the amount of alumina in the melt will correspond
to the amount shown on the liquidus line at the temperature
indicated, this temperature being the temperature at which
the special bodies are fired (made or matured) for 3 hours,
see Table 4-2.
On cooling under equilibrium conditions, i.e.,

very slowly, the melt phase will deposit alumina on existing
grains and then completely crystallize to give the equilibrium
species noted in the appropriate phase diagram, Fig. 4.03
or 4.04. The phases which crystallize will depend on the
initial composition of the body.
In actual ceramic practice, ceramic bodies are

seldom cooled so slowly as to allow the system to maintain
equilibrium and to permit the melt phase to completely
crystallize. If the cooling rate is very rapid or the melt
is very viscous (high in silica), the melt will supercool
to a glass, if the cooling rate is somewhat slower, or the
melt is a less viscous fluid, the melt will cool to a
mixture of crystals and glass. In both cases, some alumina
will deposit on the existing alumina grains, thereby alter-
ing the melt composition. When cooling under non-equilibrium
conditions, the phase distribution of the system departs
from that described in Fig. 4.04 because liquid phase (glass)
is retained to room temperature. The extent of this departure
from equilibrium is determined by the cooling rate.
When the ceramic is reheated to a temperature

approaching the melting point, but not to the liquidus
point, of the glass phase in the grain boundaries, further
crystallization termed devitrification will occur. The
species which devitrify may or may not be the equilibrium
phases, shown in Fig. 4.03 and 4.04.
19
PARTICLE,
RADIUS R
VACANCY FLUX
RADIUS OF CURVATURE
OF LENS p
MATERIAL FLUX
GRAIN BOUNDARY
BETWEEN PARTICLES
Figure 4.05: Model of two spherical particles, radius

R, showing sintering by volume diffusion, indi-
cating the flux of vacancies (and counter-flow of material)
from a lens of radius of curv-
ature R, to an internal grain boundary.
-20-
If the ceramic is reheated to a temperature
above the solidus, the grain boundary phase(s) will start
to melt, taking some alumina into solution. When the
liquidus temperature is exceeded (above the original
firing temperiture), the grain boundary phase is completely
molten and will take further alumina into solution. This
situation was actually an experimental fact in the case of
ceramics which were metallized at 18000 C.
The grain boundary phases in alumina ceramics

is soiewhat loosely referred to as the "glassy phase".
We will use this time-honored term in this report with the
understanding that it is only a convenient term used to
describe the grain boundary phases discussed above.
The extent of the pore volume is determined by
many factors and their elimination is considered below,
together with the factors causing crystal growth.
The size of the alumina crystals depends on
1. the firing temperature

2. the firing time
3. the particle size and size distribution of
the alumina calcine
4. the type and quantity of flux used
5. the cooling rate.
Alumina ceramics that contain 100% alumina or
form no liquid phases rely on solid state sintering
processes to cause densification and crystal growth.
Coblel' has summarized the work done on the

initial sintering of alumina based on volume diffusion
mechanisms. The model considered is two spheres touching
each other, Fig. 4.05. Basically, the Kelvin equation
applies. 3
kT
where /C = Change in vacancy concentration
between the curved lenses formed at
point of contact of the spheres
and the bulk of the material
21
CO = Vacancy concentration of flat surface
Ysg Surface energy
S = Vacancy volume
k = Boltzmann's constant
T = Absolute temperature
/ = Radius of curvature of lens,

This vacancy concentration gives rise ti a
material diffusion flux and can be described by Ficks
law of diffusion. Applying Ficks equations, 'e can
obtain the quantity of matter transferred -ror one point
in the system to another, and hence obtain a sintering or
densification rate for the process.
A description of the kinetics of the comilete

sintering procees using a more complicated geometric model 8 ,
for the final st tes of sintering has also een ttempteu- 19
The concepts discutsed above are again employed.
Grain growth proceeds simultaneously with the
densification process. The driving force for zhe process
is the difference in surface energy between the fine
grained material and larger grain sized product resulting
from the decreasC in grain boundary area and total
boundary energy2- .
Three overlapping stages of densification can
be observed in bodies containing a liquid phase -
Phase 1. Initial re-arrangement of the

structure caused by viscous flow
of the melt phase, eue to surface
energy forces.
Phase 2. Solution and transport of alumina
through the liquid phase as discussed
immediately below.
Phase 3. Finally, solid state sintering as
discussed above.
The densification cf a ceramic and the growth
of a primary crystal phase in the presence of a secondary
liquid phase as a functikn of ircrease time, temperature
22
- E--. - - -. W
GAS
I Melt
Figure 4.06: Surface tension forces, F, due to a pressure -11=-2YV mg acting as a solid-melt-solid
interface causing increased solubility of the solid. r is the radius of curvature indicated.
-23--
~. - - r .= • r . .V
and decreasing particle size has been outlined 21 . Liquid-
solid surface tension forces tend to draw the particles
of alumina together, Fig. 4.06. A particle-particle
pressure is generated. This pressure increases the
chemical potential of the alumina in the immediate
vicinity. The alumina goes into solution and if the
system is in chemical equilibrium will redeposit on other
alumina grains that are at a lower chemical potential.
Densification results. In general a change in particle
shape is involved in this process. This process slows
as a solid skeleton of the primary crystal phase forms.
Further densification will then largely take place by i
solid state sintering process.
Crystal growth will accompany Phase 2 especially

if the amount of liquid phase involved is largc; this
process is termed ta Type I activated solid-liquid-solid
diffusion grain growth process.* Later grain growth
processes will take place by a second type + (Type Ill of
normal solid-liquid-solid diffusion piocess together with
a solid-solid transport of material across grain boundaries.
The driving force for both of these latter processes is
the ditference in energy between the fine grained material
and the larger grain sized product. Grain growth via a
liquid grain boundary will be expected to be slower than
across a solid-solid grain boundary.
The raLe of deposition in both Type I and

Type II processes is determined by the rate of formation
of lattice steps on the alumina crystal surface and by
the rate of migration of these steps across the surface.
The steps are due to dislocations, surface nucleation
and crystal imperfections due to impurities on the -lumina
crystal surface.
The special alumina bodies, 941, 942. 3nd 99R,

are made from Alcoa A-14 alumina with a grain size distri-
bution prior to processing as shown in Fig. 4.07. Pure
SiO 2 powder and pure CaCO3 were used as fluxes.
The bodies were fired for three hours at

selected temperatures. The grain sizes of the bodies are
plotted in Fig. 4.08, and the fired densities are given
*The driving force for growth will b,' the same as for
Phase 2 densification, i.e., interfacial particle-particle
pressure generated by surface tension forces,
,By the authors
24
-71
Go-
20*
40
2 3 4 5 4 7S 9IO 20 30 400 60 700O
DeANETER (mia~,
Figure 4.07: Particie *.Ize distribution of Alcoa A.i 4 aluminaO before processing.
-2S -
0C
1550 1600 '650 1?00 1750 1800

100
50
C
0
.~10-
U
.C 5
E
o body 992
* body 941
M body 942
X body G
5.5 5.4 5., 5.2 5.1 5.0 4A 4.6

4
!xlO
T
Figure 4.08: Average grain size of 941, 942, 992, and Body G as a function of firing temperature.
Time at temperature was three hours.
-26-
in Table 4-3. The grain size of Body G (94% alumina,
4% SiO 2 , 1% CaO, 0.8%A MgO) is also plotted on the same
graph. The photomicrographs of the bodies are shown in
Figs. 4.09, 4.10 and 4.11. The primary alumina crystal
phase and the "glassy phase" existing in the grain
boundaries can be clearly seen.
The 992 ceramic contains npproximately 2% of

liquid phase at the firing temperatur.e. This is insuffi-
cient for a rearrangement process to occur so that initial
densification will occur by a Phase 2 solution-diffusion-
precipitation process, and by a solid-solid transport of
material, Phase 3. Densification is largely complete by
1700 0 C. However, grain growth in this system is at a
minimum below 1700 0 C because -
i. The presence of sub-micron pores in the

solid-solid grain boundaries "peg"
their movement.
ii. The presence of a small amount of grain

boundary liquid phase means that a
solution-diffusion-precipitation bridging
process must occur. This process is
slower than transport of material across
a solid interface.
Above 1700 0 C,grain bouiary porosity is largely

eliminated within the 3-hour time-temperature limit and
many solid-solid bridging contacts have formed so that
grain growth proceeds rapidly.
Densification in the 941 and 942 systems

proceeds via overlapping Phase 1, 2 and 3 mechanisms.
Examination of their microstructures and the density data
shows that Phase I is cuaiplete and Phase 2 is essentially
complete by 1550 0 C in the case of the 941 ceramic, and
1600 0 C in the case of the more viscous flux phase of the
942 ceramic. The subsequent small increase in density
above this temperature is due to solid state diffusion
processes.
Grain growth in the ceramic proceeds in three

predominant but overlapping stages.
i. A large relative increase in grain growth

is observed during the Phase II densifica-
tion process due to the Type I activated
solid-liquid-solid diffusion process
27
0) r-4 m (4)I i
4)
.r4
mK u
F-4 0 b
10 0 r- N
U) 0
u KN
4u U r
0 z
0 NA N' r- r.D WD VA 4-
C-I m4
m c4 (4 0 0m4 C
0 0 r=F-H(
0 0
-0 LA < co co -4 r -q (
.1-0I I * %D*
lra u- I- m4 v r 0 wm
$Z4 Ir.
4) 0 Q~l 0 m 0ma ) m )
W0 0 A 0 mO
a m
1 u LnL rn ( 0u 0 M10 'J T CNC'N
UV4) LO w)(l )
0 0
0
) 0
0
4 0- I 0 N 0 0
I IP- PI PI MI- -
U 54)()u (
.9o (a"0c0
0rn o 0-
W. (
IL M) OD0.4 r 4
U-4 0 )
rV 5
0 0r 0
UF- 4)
4J 48
1SS 0 C 160C
1650 0C 1700 0C
Figure 4.09: Microstructure of 941 ceramic (400X).
- 29 -
15s00C 1600 0C
1650 0C 1700 0C
Figure 4. 10: Microstructure of 942 ceramic (400X).
- 30 -
1650 0C 1700 0C
1750 0C 1800L'C
Figure 4.11: Microstructure of 992 ceramic (400X).
- 31 -
ii. Then slow grain growth proceeds by a
Type II solid-li id-solid diffusion
process as in the 941 and 942 bodies
fired at 1600OC-1700 0 C
iii. Fast grain growth then proceeds as the
role of the solid-solid diffusion process
increases in importance, as is uniquely
noted for the 942 body, fired at 1750 0 C.
The sag bar results, 4.2.3.3.1, confirm the
presence of a substantial solid crystal skeleton for this
ceramic. If the above theories of densification and grain
growth are correct, it implies that a very close approach
to the equilibrium amount of alumina in the melt phase
will occur above 1500 0 C for the 941 body and above 1600 0 C
for the 942 body.
The average grain growth in the latter two

stages is experimentally determined to be identical for
both the 941 anid 942 bodies but the range of crystal size
for the 941 ceramic is greater than for the 942 ceramic.
This may be because two different grain growth mechanisms
are predominant. The diffusion transport of material by
either Type I or Type II solid-liquid-solid diffusion
process will be faster in the lower silica content melt
phase. This would indicate that the solid-solid diffusion
process has proceeded further in the 942 ceramic due to
exclusion of the glass phase from the grain boundaries due
to a high or solid-melt interfacial energy.*
An alternative explanation is possible if crystal

growth in these bodies via the solution-diffusion-precipitation
processes is predominant in both cases. Crystal growth must
be controlled by a composition insensitive crystal solution
or precipitation rate rather than by the rate of diffusion
transport, as the rate of diffusion is known to be compo-
sition sensitive.
Studies carried out on melt-sapphire couples

(Section 4.2.3.2.6) 3how that the dissolution rate of
alumina grains by the flux in ceramic bodies could be
controlled by the interfacial chemical reaction until
equilibrium conditions are reached, this will also be
composition sensitive. However, subsequent solutions will
*This is borne out by the 1800 0 C results taken from

metallizing data.
32
occur by:
i. surface tension forces generating pressure
at alumina-melt-alumina interfaces as
previously explained, Type I
ii. preferential solution of grains having a
small particle size and hence a larger
solubility than the larger grains,
Type II.
Since the interfacial energies in both the 941
and 942 systems are approximately equal, at least in wet
H2 /3N2 (Section 4.2.3.4), the rate of solution in both
systems will be equal, i.e., it will be composition
ir~ensitive.
Precipitation will occur on the alumina lattice

steps which are energetically most stable. This will be
the same for both systems since the impurity level in the
alumina is constant, the process is also composition
insensitive to a first approximation.
Crys al growth in the last stage, the solid
state grain growth process, can be expected to be equiva-
lent for both systems since only alumina-to-alumina
contact is involved. In this region, crystal growth is
in fact similar to that for the 992 body.
Much more experimentation is needed to provide

a clearer picture of these processes.
The larger grain size of the 94% aluminas
fluxed with CaO and SiO 2 as compared to the body fluxed
with CaO, MgO and SiO 2 , Body C is int-eqting (Fig. 4.08).
It is known that the rearrangentent and solution-precipitation
densification processes are complete by 150OC 2 2 . A sub-
micron spinel layer MgO.Al203 may form on the alumina
grains. This would hinder any solid state or solid-liquid-
solid mode of grain growth. In addition MgO in solution
in the Al2 03 is known to hinder grain growth. For example,
the process variations underlying the manufacture of
Lucalox* have been investigated.23 The addition of traces
of magnesia to form a solid solution of MgO in the A12 0 3
rEsults in elimination of exaggerated grain growth. Larger
33
amounts, up to 0.25% are beneficial in reducinq the normal
grain growth by alumina grain boundary "pegging' with the
finely dispersed MgO.AI20 3 spinel piiise formed in the
grain boundaries. MgO is also found to have a beneficial
effect on limiting grain growth in the initial sintering
stage of the alumina which results in a higher end point
density. Although solid state grain growth only becomes
predominant at the higher temperatures (within the times
normally encountered in the maturing of most ceramIc
bodies) it is important throughout the process in bodies
containing as much as 94% of alumina.
The plot of the Modulus of Rupture of three

special bodies is given in Fig. 4.12. As the fired density
increases, the strength of the body will increase bucaisc
porosity is eliminated. However, the strength of ceramics
decreases as its crystal size increases ana the amount of
"glassy phase" increases. Thus the Modulus of Rupture
displays a maximum at a certain temperature for each body
corresponding to the lowest temperature at which a maximum
density is achieved.
4.2.3 The Glassy Phase
4.2.3.1 Molecular 46tructure
A brief interpretive review of the structure of

the glassy state is des4 able in order to appreciate thu
material considered in the remainder of Section 4.2.
Glass is a supercooled liquid possessing no

long range atomic order in the liquid or "solid" .Atc.
This property is due to "network forming" ions which coordi-
nate ith oxygen anions to furm a three dimensional net-
work. Fused silica, SiGO, itself is a glass.
The silicon ion is considered to be tetrahederally

coordinated with four oxygen ions in a three dimensional
array, Fig. 4.13. to form an anionic network structure.
The Si-O distance, 1.62 A (at room temperature) is smaller
than the sum of the normal radii of the Si 4 + and 02- ions.
The oxygen ion linking two silicon neighbors does not lie
on the line joining these ions but is displaced to 140 °
indicating a compromise between purely ionic bonding with
an anyle of 1800 and a purely homopolar ;-nqle of 1090.
From this the Si-O bond is assumed to be 5Y. ionized at
room temperature.
34
I \
46--
V
I
44-I I
941
42
Il
II
40-'
wI I
38
0. 36,/\
.S
624
/\
A
32 1 x 9
304
26 1
26 .
1500 1550 1600 1650 1700 1750 1800

FIRING TEMP (OC)
Figure 4,12: Modulus of rupture of the 941, 942 and 992 bodies.
-35-
h.
0
3
S
C
00
~
-
C)
.- U.-
.~; I
N
E
U
1
C)
C
0
U,
to
+ m
-c
'0
NU
I-
0
'0
0
1~
h.
0 -
3 U
L.
Sc
U,
O'~o -~
.0
U,
U,
'U
0)
*-~ mE
0
* ~0
0
E
U
* E
*
(0
CS,
5
SW
.- ~ 00
~,
~- '~.a
Sc
4-S
.- 0
0
U'
-36-
As metal ions are added as oxides to silica to
form, for example, Na20.SiO 2 ,MgO.SiO 2 , CaO.SiO 2 , tiie network
is broken up. The Mnr+ ion coordinatcq wit,, oxygen ions
which are only bound to o:ie silicon ion as in Fig. 4.13.
Tie Mlcbr ion will coordinate with more than four oxygens;
in addition. its M-O single (chemical) bcnd energy is less
than the Si-O bond energy. This type of ion is called a
network modifier.
The following table indicates typical silicate

anions which will form as MO is added to the melt.
TABLE 4-4
Glass Structure As a Function of Network Modifier Ion
Composition / Anion Structure

SiO 2 SiO 2 2 Network of silicate
anions
MO.2SiO 2 or M 2 + , Si 2 0 5 2-2.5 Sheet of silicate

anions
MO.SiO 2 or M2 +,SiO3 2.5-3 Chain and ring

silicate anions
2MO.SiO 2 or M 2 +,SiO 4 3-4 Island silicate

anions
The table is expressed in O/Si ratios.
In addition to network forming and network

modifying cations, a third type of cation exists that
contributes to glass structure called a intermediate ion.
For example, A1 + when coordinated with four oxygens is
a network former. When six-fold co -linated with oxygen
ions it is a network modifier. This occurs because the
Al/O radius ratio of 0.43 is so close to the critical value
of 0.414 for the transition from octahedral six fold
(6 A13 + ) to tetrahedral four fold coordination (4 Al 3 +)that
3+
the ion can occur in either or both states. When 4A,
coordina ed it can replace Si 4 + in the silicate network;
when 6 A1 + it modifies the network. The readily polarized
oxygen ion suffers considerable distortion in the former
37
.. . . . ,J,,r
case. A measure of this polarization is provided by
comparing the A1-O distance in 6-fold and 4-fold coordination
which is 1.9 A and 1.7 A respectively.
An intermediate ion is incapable of forming
a network of its own ions, for instance alumina, A1 2 0 3 is
fully crystallized, whereas silica, SiO 2 ,may exist as a
glass. (However, glasses composed largely of CaO, A1 2 03
and also BeO, A1 2 03 have been observed to form).
Table 4-5 gives a list of ion network formers

and modifiers as discussed above.
The simple picture of glass structure presented

thus far is not completely so. Some evidence has accumu-
lated which shows that glass does have a structure. The
properties of glass are also dependent on the previous
history of the material.
The glass phases that we are concerned with

in alumina ceramics are composed of an alumina-silicate
anionic network with modifying cations such as Ca2+,
Mg 2 +. Cu 2 + , Mn 2 + in the interstices. The role of ions
such as A1 3 +, Fe 3 +, Ti 4 +, Zr4 +, are as network intermedi-
ates, i.e., they can exist in the silicate network to form
an alumina-silicate network, and in the network interstices,
depending on the coordination number assumed by the
cation.
When a three dimensional silicate (or alumina-
silicate) network is not formed, e.g., O/Si 2.5, Table 4.4,
the glass is unstable and will devitrify (crystallize)
under prolonged heating in a temperature range suitable
for nuclei formation and crystal growth. Conversely,
under conditions of severe quenching, devitrification may
not occur at the point 2MO.SiO 2. This is attributed to
divalent cations, especially Be and Mg, bridging isolated
silica tetrahedra. In general, however, when O/Si.>. 4 the
melt will crystallize on cooling.
The CaO-SiO 2 -A1 2 03 glass melts considered in

this report will have some properties of the room temperature
glassy state, but may also be expected to display other
characteristics because of the high temperatures involved,
(1200°C-20000 C). The glass melt will also have an
equilibrium structure depending on the time at temperature
and rate of cooling from this temperature.
38
TABLE 4-5
Co-ordination Number and Bond Strength of Oxides

(Taken from K. H. Sun J. Am. Ceramn. Soc., 30, 277 (1947)
Dissocia- Co- Single-

tion Energy ordina- Bond
per MOx tion Strength
M in MOx Valence (kcal/mole) Number (kcal/mole)
Glass formers B 3 356 3 119
Si 4 424 4 106
Ge 4 431 4 108
Al 3 402-317 4 101-79
B 3 356 4 89
P 5 442 4 111-88
V 5 449 4 112-90
As 5 349 4 87-70
Sb 5 339 4 85-68
Zr 4 485 6 81
Intermediates Ti 4 435 6 73
Zn 2 144 2 72
Pb 2 145 2 73
Al 3 317-402 6 53-67
Th 4 516 8 64
Be 2 250 4 63
Zr 4 485 8 61
Cd 2 119 2 60
Modifiers Sc 3 362 6 60
La 3 406 7 58
Y 3 399 8 50
Sn 4 278 6 46
Ga 3 267 6 45
In 3 259 6 43
Th 4 516 12 43
Pb 4 232 6 39
Mg 2 222 C 37
Li 1 144 4 36
Pb 2 145 4 36
Zn 2 144 4 36
Ba 2 260 8 33
Ca 2 257 8 32
Sr 2 256 6 32
Cd 2 119 4 30
Na 1 120 6 20
Cd 2 119 6 20
K 1 115 9 13
Rb 1 115 10 12
Hg 2 68 6 11
Cs 1 114 12 10
39
I
Tables 4-6 and 4-7 give the molecular structures

that can exist in silicate melts based on the network
theory. We have anticipated the viscosity, section 4.2.3.3.1,
by inserting viscosity data in these tables. As expected.
as the alumina silicate anions become more complex, the
viscosity increases. However, since the entry of the
3
4 A1 + into the network is dependent on the A120 3 /CaO ratio
being less than one, with A1 2 03/CaO. compositions greater
than one, the viscosity will decrease because the network
structure is loosening with the introduction of 6 Al 3 into
network modifier positions. This ratio is exceeded at the AI2
470% level for the 1:1 Si0 2 /CaO ratio and 37% for the 2:1 wt.
Si0 2 /CaO atio. Many of the physical and electzical
properties in the Ca0-Si0 2 -Al 2 03 system behave similarly, 0
i.e. the direction of the property under consideration
changes as the A1 2 03 /CaO ratio passes through one. The
subject of glass structure is considered in some detail 5
ir the 3rd Quarterly Report and elsewhere1 5 .
20
4.2.3.2 Composition and Amount
Glass flow out from the ceramic into the porous

molybdenum metallizing sinter sponge is one phenomena 25
responsible for seal strength, It is important to know 30
the mineralogical and chemical composition and the amount
of the "glassy phase" of the special ceramics in order to
predict their metallizing behavior.
3
The alumina microstructure considerations
discussed earler, Section 4.2.2, indicate that if one 40
accepts the theories of sintering and grain growth which
are forwarded, equilibrium is reached when the grain growth
process becomes complete due to a Type 2 solid-liquid-solid
I
50
and a solid-solid material transport process This is
above 1550 0 C for the 941 ceramic and 1600 0 C for the 942 I 60
ceramic. i
Various techniques were use,! in order to
further characterize the glass phase. They included a study 100
of existing phase diagrams, planimetric measurements of
optical and electron photomicrographs of selected areas
of the ceramics, x-ray diffraction analysis of the ceramics,
electron microprobe scans across the ceramic, chemical
analysis of the glassy phase and chemical reaction rate
studies on glass-sapphire couples were also investigated.
40
I
I.
TABLE 4-6
Anion Complexes in a CaO-SiO2-A1 2 0 3 Melt
of the 1:1 CaO/Si0 2 Composition
and the Correlation with Melt Viscosity (at 1500 0 C)
A1 2 03 Mole Viscosity
wt. % % O/Al,Si 0/Si Al/Si Ions Poise
0 0 3 3 0 (Si 3 0 9 ) -6,Ca2+ ; rings and 2.0
chains
5 - - - 4.5
20 22 2.46 3.9 .55 n(Sil. 3 3A 2 +;

6 60 5 )27nCa 9.5
(sheet)
25 12
30 33 2.25 4.5 1.0 n( Si 0 . 5 A10 . 5 02.25) , 14
Ca2 network
35 15
40 43 2.1 5.7 1.5 ( Si-Al 07 )+A 15
2+
Ca
50 54 2.0 6.5 2.3 Forms unstable glasses on
cooling composed of alumino-
60 63 1.85 8.0 3.4 silicate anions and six fold
A1 3 + ions and Ca 2 + ions in
modifier positions
100 100 1.5 e-" Al3+ + 02 melt
41
TABLE 4-7
Anion Complexes in a CaO-SiO 2 -A12 03 Melt

of the 1:2 CaO/SiO 2 Composition and the Correlation
with Melt Viscosity (at 1500 0 C)
A120 3 viscosity
wt. % 0/Al,Si O/Si Al/Si Ions Poise
0 2.5 2.5 0 n(Si 2 0 5 )-2,Ca2 +; sheet --
structure
5 37
20 2.15 3.25 .5 Al 0 )-.66n 65
Sn(Si ' 33 2.15
66
Ca2+network structure
25 80
30 2.05 3.7 .8 n(Si 54 ,A 4402.5 6n 100 1

Ca2 +; network structure
40 1.93 4.4 1.3 (Si,Al 07)- , Al , Ca2+; I
forms unstable glasses on
cooling
100 1.5 00O> A13++0- 2 alumina melt
I
I
I
I
I
42
4.2.3.2.1 Equilibrium Phase Diagram Studies
If the system is in equilibrium, then it is
possible to predict the amount and composition of the various
phases from the CaO-SiO 2 -A12 03 phase equilibrium diagram.
The binary cuts given in Figs. 4.03 and 4.04 can give the
amount of melt phase existing at the solidus and liquidus.
This has been calculated for the 941, 942 and 992 bodies
and is given in Table 4-8, columns 5 and 6. The equilibrium
crystal species present are given in Table 4-9.
TABLE 4-8
Weight Percent Glass Phase from Photomicrograph
Measurements
(Values obtained from electron and light
photomicrographs)
Ceramic Weight Percent Equilibrium Amounts of Melt

Body Firing Glass from Photos At Firing At Invariant
Type Temp.OC Electron Light Temp. Point, 14050"
1 2 3 4 5 6
941 1700 7.5 17.7 14.2 10.3

941 1650 7.3 n.d. 12.4 10.3
941 1600 6.7 .2 11.6 10.3
941 1550 6.6 n.d. 11.1 10.3
942 1750 8.4 n.d. 14.7 10.0
942 1700 8.4 13.3 12.1 10.0
942 1650 7.0 n.d. 10.9 10.0
942 1600 7.4 n.d. 10.3 10.0
992 1800 1.3 5.4 3.0 1.66
992 1750 1.2 5.35 2.5 1.66
992 1700 1.25 n.d. 2.0 1.66
While it is probable that chemical equilibrium is

attained at the firing temperature of these bodies, experi-
mental proof was sought to confirm this possibility. Also
it is important to know whether this high temperature state
is "frozen in" on cooling or whether it tends to adjust to
a lower temperature equilibrium situation, The remaining
sections of 4.2.3.2 attempt to answer these questions. It
is shown that equilibrium at temperature is approached and
that the system does tend to adjust to a lower temperature
equilibrium on cooling. The latter fact is supported by
the sag bar studies reported in section 4.2.3.3.1.
43
TABLE 4-9
Crystal Species Present in Ceramic Bodies
under Equilibrium Condition
Ceramic Primary Secondary

941 A12 03 CaO.6A1 2 03*, CaO.A1 2 03 .2 Si0 2 **
942 Al203 CaO.A120 . 2 SiO 2 **
992 Al203 CaO.Al203. 2 SiO **
*Calcium hexa-aluminate
**Anorthite
4.2.3.2.2 Optical and Electron Microscope Studies

Measurements of the amounts of secondary or
glassy phase in the special alumina ceramic bodies using
the line-intercept fraction method were made using a series
of electron photomicrographs of the special bodies listed
in Table 4-8. Pictures of six different areas of a polished
sample of each body were made in order to provide a
representative sampling of the materials. A magnification
of 3500X was used because it was a good compromise between
the high magnification required for good resolution of
phase boundaries and the amount of area required for a
representative sampling of the surface.
Examples of areas of each sample are presented
in Figs. 4.14, 4.15 and 4.16. The areas occupied by
corundum crystals (o(Al203) and the glassy phase are easily
distinguishable in each potograph.
The sample preparation technique is described

in detail in the Sixth (pp. 59-60) and Seventh (p. 28)
Quarterly Reports.
The line intercept fraction measurements were

made by placing a sheet of tracing graph paper over each
photomicrograph which was taped to a lighted tracing board.
The photomicrographs were 24 cm. by 21 cm. and the graph
paper was aligned so that 23 vertical lines (one every
centimeter) crossed each photograph. Each line was marked
to the nearest 1/2 millimeter to indicate the portions
44
. . • m • OWN
a bO
Ids'
Figure 4.14: Electron photomicrographs (3500X) of polished surfaces of special alumina bodies
(a)941 (1700), (b)941 (1650), (c)941 (1600), and (d)941 (1550), The numbers in
parentheses are the body firing temperatures in 'C.
Figure 4.15.. Electron phote-ncrographs (3500X) of polished surfaces of special alumina bojies
(a) 942 (1750), (b) 942 ii700 (c)942 'Pifl. and (d) 942 "1600), 1he nujmbers in
parentheses are the body firing temperatures in C.
-46-
C
Figure 4.16, Ele~tron photomicrographs (3SGOX) cf polished surfaces of spcial alumina bodies
(a)992 01000), (b) 992 (1750). and (c) 992 (1700), Thm numbers in paonthe.-es or*
the body firing temperatures in 'C.
-47 -
traversing corundum crystals, glassy phase regions and
voids. The fraction glass in each photomicrograph was
calculated by dividing the total length of the line
traversing glassy regions by the total length of lines
traversing corundum crystals and glass regions. The numbers
thus obtained were line-intercept fraction glass and were
equal to the volume fraction of glass in the bodies with
zero percent porosity. The weight percent of glass was
calculated assuming a specific gravity of 4.0 gm/Icc for
corundum and 2.6-2.8 gm/cc for the glass. 8,24
Similar measurements were made using light

photomicrographs (100OX) taken with a metallograph. Data
were not obtained for ail bodies, but representative
samples of each composition were measured.
The results obtained from both electron and

light photomicrographs are presented in Table 4-8 for
comparison. The difference in the two sets of values is
most noticeable and is discussed in detail below. Values
obtained from light photomicrographs were roughly twice
as large as electron photomicrograph values for the same
bodies.
The fifth and sixth columns in Teble 4-8 list the

quantities of secondary or melt phase which would be present
if equilibrium were reached in each special ceramic body
at each of two temperatures. The fifth is the temperature
at which the ceramic was fired, and the sixth is the
temperature of the invariant point, i.e. the temperature
at which all remaining mel +- crystallizes on cooling under
equilibrium conditions. The amounts of melt phase present
in a body at each temperature represent the maximum and
minimum amounts, respectively, that one would expect to
find in each body. For example, body 941 (1700) should
contain 14.2 weight percent melt at its f2ring temperature
of 1700 0 C if equil*brium is reached. If equilibrium is
maintained during cooling, the body would contain 10.3
percent melt at the invariant point, 14050 C, just before
the secondary crystaliine phase (3) began to precipitate.
These values were calculated for each body using the
"lever rule"* and the psuedo-.binary phase equilibrium
diagrams in Figs. 4.03 and 4.04.
*The lever rule is a means of determining the amount of

each phase present at any point within a two phase region
of a binary phase equilibrium diagram. The point is
located by the composition and temperature of the system.
48
In order for these aiumina bodies to maintain
equilibrium during cooling, alumina must precipitate from
the melt to continue the growth of the corandum crystal
phase. However. as is subsequently discussed in section
4.2.3.3.1, the viscosity of the melt phase increases
rapidly as the temperature drops below 1450 0 C so the
diffusion of A1+ 3 and Al complexed ions necessary for
continued precipitation and growth of the primary corundum
crystal phase is so reduced that the system only sluggishly
approaches equilibrium. Rapid cooling thus traps some
aiumina in the melt and holds it while the melt is super-
cooled to the glassy state. Because of the additional
alumina in the melt, che quantity of melt is greater than
it would be if equilibrium ccnditions were maintained.
Therefore, the amount of melt measured in Body 941 1700)
should be somewhere between 14.2 percent, the maximum
amount that could be present at 1700 0 C and 10.3 percent
the minimum amount that would be present at 1405°C, the
invariant point. (Even if some calcium hexaluminate and
anorthite secondary crystalline phases were able to
nucleate and grow below the invariant point temperature,
they would do so within the "glassy areas" and would be
counted as glass when line-intercept techniques are used
to interpret the photomicrographs).
The fact that all values obtained by line-

intercept-fraction measurements were outside the possible
range of values as discussed above was disappointing, but
can be explained. The most likely source of error was the
surface relief of the samples and the difference this causes
in the two types of photomicrographs used in the study.
Measurements made on electron photomicrographs consistently
gave values below the theoretical minimum amounts but
the values generally fell in descending order indicating a
strong relationship to the firing temperatures. Measure-
ments made on light photomicrographs gave values above
the theoretical maximum, but again showed good correlation
to the firing temperatures.
Relief in the sample surface is practically

impossible to eliminate because it results from the fact that
the corundum czystbel p i much
ALS I 1a dr Lba1 the gl1assy
phase. Consequently, any polishing operation depresses
the softer glassy areas with respect to the rest of the
surface. Reference to Fig. 4.17 will help to explain how
surface relief could have produced the errors in the glass
phase measurements. The upper sketch in Fig. 4.17 shows
a typical glassy region surrounded by corundum crystals in
the surface of a ceramic sample. The surfaces of the
49
Corundum(0: A12 03)
Corundum
Corundum Corundum
Top view of glassy region in alumina surface.
-A
Corundum Corundum
Cross section through glassy region in a!umina surface.

Dash lines indicate contours before polishing.
Figure 4.17: Schematic presentation of a typical glassy phase region of a high alumina ceramic
showing the effects of polishing the surface.
- 50 -
I
" "'q-'...
. ...
" ... .. ...
_- _:at_ i v"* .,e--,,-- '
I I I I * Y
corundum crystals all lie generally in the same plane,
while the glassy region is slightly depressed and its
surface lies in a plane below that of the corundum. The
edges of the corundum crystals are rounded during polish-
ing and slope down to meet the glass surface.
When a shadowed carbon replica is made of this

surface feature, contours are faithfully reproduced so
that the subsequent electron photomicrograph of the replica
displays all aspects of the feature in perfect focus.
The exact location of the interface between the glass
surface and the sloping corundum crystal can be seen. Cn
a line intercept measurement the extent ot the glassy
region would then be marked as the distance "C' in the
lower sketch of Fig. 4.17.
A photomicrograph of the same area taken with
a conventional Metallograph at 800X does not bring the
entire feature into focus because the depth of focus is
usually less than the depth of the feature. This type
of photomicrograph is normally taken with the upper
surface (corundum crystals) in focus and the voids or
glassy regions slightly out of focus. A line intercept
measurement of the feature would then include the distance
."A" as the glassy region, because the actual region would
be fuzzy and the true boundary difficult to locate.
The correct or original extent of the glassy

region as indicated in Table 4-8 and the previous discussion,
is somewhere between the extreme values obtained from the
two types of photomicrograph. This should be the distance
"B" shown in Fig. 4.17 which indicates the extent of the
glassy region before it was worn down preferentially and
befire the sharp edges of the corundum crystals were
rounded by polishing. The dotted lines indicate the
original glass-crystal interface. The possibility of
estimating the original interface location has been con-
sidered, but this could be misleading and no more accurate
than averaging the values obtained from measurements on the
two types of photomicrograph. In most cases such averages
are within the limits set by equilibrium calculations.
However, for our purposes, it is desirable to locate
accurately the correct value for the amount of glass within
the range of possible values, not just to determine that it
is within the range.
Tb, location of this value within the range is
very important because it must be used in an application of
the "lever rule" and phase equilibrium diagrams to determine
51
______________,____ -
the amount of A1 2 03 dissolved in the melt. The melt
composition (and temperature) determines it-- viscosity,
which in turn is a major factor in determining metallizing
behavior characteristics of the melt such as mass flow and
ionic diffusion rates. When several steps are required to
arrive at the desired result as in this case, the accuracy
of each step must be optimized to avoid compounding errors.
I
Because of the nature of the CaO-A1 2 03 -Si02
system, an error in determing the amount of glass or melt
phase results in a agnified error in determining the A1 2 03
content of that phase. While the amount of melt varies
over an allowable span of approximately 4% (Table 4-8).,
the amount of dissolved A1 2 0 which causes the change in
I
the amount of melt, varies o~er a span of about 20 percent.
Consequently, an error of 10 percent in determining the
amount of melt would produce an error of 40 percent in
computing the percentage of A1 2 03 in the melt. This
error would not have a great effect on viscosities
determined for 1:1 Si0 2 /CaO-Al 20 3 melts, but would have
importance in determining the viscosities of 2:1
Si0 2/Ca0-AlA O 3 melts, especially at lower temperatures.
Therefore t e-A1 2 03 contents of melts must be accurately
determined if subsequent determin : ions of melt viscosity
and diffusion coefficients are to be accurate..
In conclusion, it should be noted the quality

of the results (accurate to within a few percent) is
comparable to those obtained by other workers 2 5 . As the
results are not good enough for the present purpose (which
requires values to within a few tenths of a percent for use
in subsequent calculations of the glassy phase composition),
similar calculations by other workers zhould be treated with
reserve.
I'
4.2.3.2.3 Mineralogical Analysis
The equilibrium phases present in the ceramics

as given by the equilibrium phase diagrams were shown in
Table 4-9. In order for these phases to crystallize and no
other, the alumina content of the melt phase must be a
minimum of 42% for the 941 body and 37% for the 942 and 992
bodies.
The three series of bodies were crushed to a

powder form for X-ray diffraction analysis. The -esults
are presented in Table 4-10 and are discussed below in
II
conjunction with the phase equilibrium diagrams, Figs. 4.03
and 4.04. In the terminology used in regard to the X-ray
data, a trace is 2% or more, a minor phase is 4% or more.
52
_ - .~ -- 0*
a) W)
0. 044 4-
+ 0rE-*-'q
'4UIVU' 4-) 4.)
N a)
U) 4.)
)
4) 4)W W)
0 . 40
.4 4J 4J
a)1
Q) a) fO' 4.J 4J 4J4J (D
C4U0 X iii0 14~$4W14~
(aI' "4) 0 0 0 0
4.)
4.)
r'4 fu :j
0 0un + Ic4- 4J'-
'-4 4) ) 4.) (a r4
IU1
fu U m 141 1i 0 I u X
4. )
44' (a (a 0 '4
tU 4 r., 0 0
0 .
E-4 F4 )
4j~ > :
u e4 4.)
P14
'(e1 a) a) (aU
14 0 C)4. 4.)"4
144 a IV 'U0-
44 a) a)U .4'4
+ C:- IUF-4 9 'Upq4

w a) 4) *'4 to .04to (a-r4. 4. '
0' m F-4 U g 94 0
0 a) '4 ) 0U $'40 0
0 P-?0
0
00
$4- r4 F- -. - )4
P" 4 "4 m N N NN N N N N N2
4)'
53
-o~ -omo W- "woo.., -

The 941 body is examined first. The 941-1550
underfired body, Table 4-11, Fig. 4.09, Fig. 4.14, shows the
presence of a non-equilibrium phase - Gehlenite
(2CaO.SiO2 .Al2 03 ) co-existing with the two equilibrium
phases, indicating that the alumina content of the glass
phase of this ceramic is in the range 37-43% giving complete
melt formation at a maximum temperature of 1450 0 C and a
minimum of 1400 0 C.
The 1600 0 C firing gives the equilibrium phases
expected, indicating an alumina content in the range
42-480% having a melt range from 1450-16000 C. The anorthite
phase is the last phase to crystallize in this system, and
its presence in only a trace amount indicates that the melt
tends to supercool to give some glass chase.
Based on the foregoing, it is logical to assume
that the melt from both the 1650 0 C and 1700 0 C firing super-
cools from the firing temperatures with the precipitation
of some alumina, giving an alumina content in the glass
phase of 42-52% and 42-57% ranges respectively.
An examination of the 1425 0 C refire is instruc-

tive. Two crystal phases are resorbed in body 941-1550 0 C
which further tie down the concentration of alumina in the
melt phase to 39-41%, the melt supercools to a glass and
primary CaO.6A12 03 crystals. The anorthite is reabsorbed
in the 941-1600 body as 1425 0 C is above the crystallization
temperature of 1405 C (Fig. 4.03) and some of the CaO.6A1 2 03
is resorbed giving a melt which supercools to a mixture
of glass and CaO.6A1 20 3 .
The 941-1650 0 C continues t8 present a super-
cooled glass phase, while the 941-1700 C shows some tendency
towards devitrification as may be expected of a higher
alumina content in the glass phase. Table 4-11 presents
the derived data discussed above.
The X-ray data on the 942 and 992 ceramics may
be analyzed similarly and the range of alumina content in
the glass phase obtained similarly. this data is presented
in Table 4-11.
The electron microscope photographs, Figs. 4.15,
4.16 and 4.17 do not reveal the presence of crystalline
phases because of polishing difficulties.
An examination of the surface condition of the

bodies is also of interest. This is shown in Table 4-12.
54
TABLE 4-11
Some Properties of the Glassy Phase

of Special Alumina Bodies
Alumina Equili- Presence of Presence of

Content brium Glass as Major Glass After
Firing of Melt Melt Grain Boundary 1425 0 C
Ceramic Temp. Phase Range "as Received" Firing
941 1550 39-41 + 1400-1450 doubtful, C* yes C

1600 42-48 1450-1600 yes, C yes-more, C
1650 42-52 1450-1650 yes - yes
1700 42-57 1450-1700 yes - yes C
942 1550 <37- + 1547 no C no C

1600 <37- + 1547-1600 no C no C
1650 37-44 1547-1650 yes - doubtful C
1700 37-50 1547-1700 yes - doubtful C
1750 37-58 1547-1750 doubtful C no C
992 1650 37-44 1547-1650 yes - yes

1700 37-50 1547-1700 yes - yes
1750 37-58 1547-1750 yes - yes C
1800 37-66 1547-1800 yes C yes C
*C denotes plesence of crystalline phase
+definitely non-equilibrium
55
- ~ ~ ~ ~ ~ ~ ~ os 7 ._7________________
I
I
I
I
I
I
I
I
I
I
I
Figure 4.18: Electron photomicrogroph of 941 alumina ceram" at 21.SOOX magnification showing j
tlie glassy phase between the grain boundaries of alumina particles.
I
-54-
___ - -w -
Figure 4.19: Electron photomicrograph of 942 ceramic at 21 .50X magnification showing the
glassy phas. spreading over the alumina crystal and in the grain boundaries
between the alumina ceramic.
Figure 4.20: EI-ctron photomicro~graph of 992 bodi at 21 .500X magn4~.cation showing the
-'S8-
4.' 4' 4.J 4.' 4.'
.9.4 '9 . q .4 04 9
.4.) 4.' 4.) 4. 4.'
ut
0
'1
2 4) 4 4.) .4. 4.) 4.' 4.

0) 94 09 m $4 $4 $94 .Iw I
m : o0 0 0 0 0
.9,4
r-4
.r.4 4) Q) )4
U41) 41' 41) 41
4) 41' 41 41' 41
0 mU W4 '4 I I'II4 I I I I4
-r94 tiv0 000
N~ 04.' 4I
"-4 P4 'U E-
I IV V 4)
r. 4)
4 444) 4
144 .9
'444,4
a5 V"4 "4 I
i00
O-A a 0000 0 0 0 0
0 0~ 000
r 0 ~ f
00
0 f% 'a 0
cA 00in
F~. 00 0
r- in 0 IA 0 I 0 %D 0- % 0o
04 04
P 0-4 P4 4 "-4 9-' "4 "-4 "4 P4 P4 P4 4
ca
US
59
I
w - ~ ' -~. ~ - "- ~ - .- -- ~ ..
The 941 body shows the absence of CaO.6AI20 . noted in the
center of the body showing that the faster Iate of surface
cooling induces more glass to be formed than in the interior
of the body, The formation of gehlenite as the most pre-
dominate phase on the surface in the "as fired" material at
the lower temperature indicates that the crystallization
occurred as a devitrification product during cooling.
The lack of crystallinity of the "as received"

surface of the 942 and 992 bodies indicates a glass phase.
This glass phase may be completely devitrified to ancrthite
on reheating to 1425 0 C. The presence or absence of the
crystal phases in the boundaries was confirmed by the
electron microscope as is typically shown in Figs. 4.18,
4.19 And 4.20.*
The differences in the composition and amounts

of glass present in the three ceramics and also the ground
(originally the body interior) versus un-ground "as received
surface) differences in composition will be of significance
when discussing the behavior of the metallized ceramics.
The data obtained using the X-ray/phase diagram
technique i' considered to be of significant value.
4.2.3.2.4 Electron-Microprobe Analysis
Historically the first technique employed on

the program to determine alumina content was the electron
microprobe. The basic problem encountered with the micro-
probe was one of resolution, see the Sixth Quarterly Report,
pp. 50-59. The grain boundary widths of the ceramic are of
the order of one micron which is comparable in size with the
diameter of the beam. However, selected areas of the higher
temperature fired ceramics contained glassy areas 2 to 3
microns in diameter. Even so, due to secondary X-ray
fluorescence effects and selective differential absorption
of Ca K-alpha and Si K-alpha radiation by overhanging
alumina crystals and the evaporated aluminum film covering
the surface, erroneous reading7 were obtained; the Si0 2 /CaO
ratios being variable and averaging much too low.
it was regretfully decided to discontinue this

investigation and to use the microprube for tc-ks more suited
to its capabilities.
*Space precludes the presentation of the more than 100

photographs obtained on the ceramics showing the 'as
received" or "heat treated" surfaces.
60
4.2.3.2.5 Chemical Analysis
The chemical analysis technique, using in the

preliminary series of experiments a HCl solvent and later a
HF solvent, (9th Quarterly Report, ppo, 21-22) showed promise.
The experimental procedure required that the

ceramic sample first be crushed to minus 60 mesh. A dried
and weighed sample of the pulverized sample was then leached
in warm 1:1 HCl for at least 24 hours. The remaining
corundum crystallites were filtered and weighed. Steps were
taken to remove iron introduced in the crushing step. This
consisted of washing the sample in a solution which dissolved
the iron without attacking the glass or crystalline phases
of the ceramic. (The resistance of corundum crystals to
attack by HCl was determined by exposing a sample with one
micron average particle size to the acid under the same
conditions of time, temperature and concentration that were
used for the leaching step. Six weight percent of the
corundum was dissolved.)
The alumina body 941 (1700) was used in all

solution work up to this time. Two gram samples crushed
below 60 mesh were soaked in 100 ml of 1:1 HCI for lengths
of time ranging from one to 32 hours. At the end of the
soaking period the solvent and insoluble material were
separated by filtering, then the insoluble material was dried
and weighed. The filtrate was analyzed for Si, Ca, Al, and
Fe.
Invariably the relationship between the time and

the amount of the sample dissolved was the reverse of what
it should have been. That is, the samples which were exposed
to the acid solution for longer times had more rather than
less insoluble material retained on the filter paper.
Subsequent chemical analysis of the filtrate showed much
less silicon in relation to the amounts of aluminum and
calcium that there should have been. The silica was
precipitating as silica gel and being retained with the
insoluble fraction. The amount of soluble material observed
in the several experiments varied from 7 to 10.5 weight per-
cent. This was less than it would have been if the silica
had not precipitated, but was close enough to the calculated
range of 10.3 to 14.2 percent glassy phase to expect reasonable
accuracy from the technique if the proper solvent is used.
In order to overcome the difficulty of silicate

precipitation which was encountered using only HCI, a solvent
coataining HF and H 2 SO4 in 4:1 ratio was prepared. The HF
61
dissolves silica and reacts to form SiF4 (a gas) and water.
This type of reaction removes siliccn from the glass struc.-
ture and leaves an amorphous mass that is easily dissolved.
Here again, the solubility of the Crystalline A1 2 03 in the
solvent must be accurately determined because it is large
enough to affect the accuracy of the results, and therefore
some correction of the data for this factor is necessary.
The procedure is similar to that used with HCl.

The insoluble material is separated and weighed and the solute
is analyzed for Ca, Al and any metal ion which may be
introduced by crushing the sample. SiF 4 is driven off by
repeatedly evaporating the sample to dryness and repeating
the solution step with fresh acids.
The results again indicated that no greater

accuracy could be obtained than with other techniques. It
was seen that further improvements in the accuracy of this
method would require an inordinate amount of effort in
relation to the overall scope of the program. It was there-
I
fore decided to divert this effort to more productive areas
of the program.
4.2.3.2.6 Reaction Rate Studies
It had formerly been assumed that ceramic

systems were very sluggish in their approach to equilibrium,
but at the high temperatures employed to manufacture alumina
ceramics, the fluid glassy phases present in alumina ceramics
may enable a close approach to equilibrium as can be seen
[
from the X-ray data presented in 4.2.3.2.3. Also, work by
Barrett and Reed 2 5 showed that corrosion of alumina
(sapphire) by large volumes of CaO.SiO 2 -Al 03 glass was
initially controlled by the interfacial chimical reaction
which changed to a diffusion controlled mechanism with a
build up of product in a Nernst layer of the order of
100 microns thick. Thus it is reasonable to expect an
equilibrium condition to occur over the 1 micron glassy
phase thickness of the grain boundary of alumina ceramics.
In order to confirm this, a series of glass-

sapphire sandwiches was made by placing minus 325 mesh,
1:1 Si02/CaO glass containing 0%, 20% or 40% Al20 3 between
sapphire discs. The discs were about 0.035 inc es thick
and had an optical polish on the side in contact with the
glass. Each unit was fired to one of three temperatures,
15000 C, 1675 0 C, or 1750 0 C in air for three hours and then
cooled rapidly.
62
.~~~I , - IORM? T 71,

In each case, a portion of the sapphire was
dissolved by the glass, depending on the reaction tempera-
tures and the amount of A1 2 03 placed in the glass initially.
The amount ard distribution of the A1 2 03 in the glass layer
will depend upon the rates of solution and/or diffusion of
A1203 through the glass as well as the initial A1 2 0 3 content
of the glass.
Tr:he samples were prepared for microprobe
examination by mounting and polishing them metallurgically.
Photomicrocraphs of the polished samples appear in Figs. 4.21,
4.22 and 4.23. ihe preferential attack of the glass phase on
the sapphire revealed the hexagonal crystal habit of the
sapphire. This observation proved that under the conditions
of the present experiment the rate of attack was chemically
controlled, at least during the initial stages of the
reaction. The sample containing 40% A1 2 0 3 fired at 1750 0 C
could not be recovered because it stuck to the setter and
shattered when it was removed. The glass containing 40%
A1 2 03 heated at both 1500 0 C and 1675 C appears to have
partially crystallized. Possibly these are the only samples
in which the glass became saturated with A1 2 03 .
It was initially planned to determine the amount

of alumina present by using the technique of weighing the
sapphire discs before and after corrosion by a given amount
of glass. 1 5 This approach failed in this case because only
the I mil edge of the glass was exposed. Thus the reacted
glass melt could not be dissolved by either HCI or HF
without dissolving considerable quantities of the alumina
substrates.*
The sectioned and polished couples were submitted

for electron microprobe examination. The results of the
initial analysis are reported in Table 4-13. The results
indicated that equilibrium was not attained. Moreover,
anomalies in the data such as less alumina dissolved at
a higher temperature were surprising. Detailed considera-
tion of these .3sults is presented in the Eighth Quarterly
Report, pp. 29-32.
Further microprobe studies were made of glass-

sapphire reaction couples #341, 342 and 343 toward the end
of the program. The glass layer between the sapphire discs
was about 15 microns thick. The concentration of alumina
was the same across the layer (within experimental accuracy:
4 micron resolution) and had increased to more than 30 weight
*A hot furnace removal technique for one of the discs might

solve this problem.
63
C
Figure 4.2 1: Glass-sapphire corrosion samples with 1: 1 SiO,./CaO glass containing no A120 3
initially. Samples were reacted at (a)1500'C (b) 1675'C (c) 1750*C. The glass
is the wide band across the center of each photograph.
-64-
-" -77 I 4
a
Figure 4.22: Glass-sapphire corrosion samples with 1:1 SiO;/CaO glass containing 20% A1I20
initially. Samples were reacted at (a) 1500*C and (b) 1675*C. The glass is In the
band across the center of each photograph (400X).
*-..-Y6.w
.'~ .,. ---
a
Piguro 4.23: Git, -apphlr* eoerosion samples with 1:1 0Siog/CaO glass containing 40% A60Oa
initially. Sam"Ples were reacted at (a)1500 C and (b) 1675*C. The glasn is in the
band across the contor of th top photograph and in the lower two-thirds of the
bottom photograph. (400X)
66-~
TABLE 4-13
Electron Probe Microanalyses of Glass Layer

of Sapphire-Glass Sandwiches
Percent Alumina
Sample Initial Glass Reaction in Glass after
Number Composition Temperature Reaction
336 1:lSi0 2/CaO,OA120 3 Not reacted* 0

337 " 15000 C 27.6
338 1675 0 C 21.5
339 1750 0 C 24.1
340 l:lSiO 2/CaO,20%AI 2 03 Not reacted 18.7

341 " 1500 0 C 28.7
342 o 1675 0 C 31.3
343 o 1750 0 C 21.1
344 :1:Si0 2 /CaO,40%A1 2 03 Not reacted n.d.

345 " 15000 C 44.4
346 t 1675 0 C 43.1
347 1750oC
1 n.d.
*Fused 10 secs 1550 0 C
67
• .. •~~~ • •- - - --. -.
percent in all cases.
The accuracy of the original microprobe analyses
can be considered to be only * 10 percent because they were
made relative to analyses of pure samples of the elements
being analyzed. The second analyses had an accuracy better
than 1 5 percent because the probe data were corrected for
characteristic fluorescence absorption, mass absorption,
and by the atomic number factor in a computerized program
of successive approximations.
The results obtained in the second microprobe
analysis appear in Table 4-14.
The alumina values are in all cases higher than
those obtained in the first probe work. In addition, it
was discovered that the glass in sample #343 was actually a
two-phase system, thus accounting for the discrepancy
reported in Table 4-13. This was first discovered with
the microprobe, then verified visually by very careful
microscopic examination. Sample #343 was not an intact
sandwich as were #342 and #341, because it fractured when
it was quenched. However, several glass pockets up to 3 2m
in thickness adhered to one sapphire disc, and these were
satisfactory for analysis.
Microscopic examination revealed that samples #341

and 342 appeared much alike, both exhibiting a one-phase
glass between the alumina sandwich. The absence of an
alumina concentration gradient across the glass layer is
interesting. It indicated that the alumina level of the
glass was reaction rate controlled (dissolution of A12 0 3 )
rather than diffusion rate controlled. Resolution of
the probe was such that it was safe to say that the alumina
concentrations in th2 glass layers of the samples were
level at the values in Table 4-14 to within two microns
of either interface with the sapphire sandwich.
This conclusion supports the work of Reed and
Barrett'5 . They snowed that the initial vigorous chemical
reaction occurring at the melt-sapphire interface would
disrupt any concentration gradient that might tend to form.
Dissolution would be rate controlled by the chemical reaction
as long as insufficient product wai dissolved to form a
Nernst diffusion layer of sufficient thickness such that
it could not be totally disrupte6 by the vigorous surface
reaction. When an amount of product sufficient to give a
layer greater than 50 microns thick had accumulated,
68
..... ... II III , ~ ~ ~all"-

100-
v 00
tw Ln n N N
0A O* 4)
u9 *4 >~ 40
m~ c4
9: 0
0 La L
p4 N
Nl 4) D O N 0) c-4c:
to e(o-i Pq 4)0
U) 4 ( 1- CO0
0~CV -4 w N
ON4
04
44) E-4 -O0 0n
0- 04 )
a 00 N O 0-
s r-4' 0I ; 40 u 4 0
L
fu en0% 41
No V m~4 r
Lo5)
A~ I 0
SkC N Ln OD
01. V-4 0 0 942
92
0
~n M' M fn 0%4
a P
to0
'44 4 A L n1
"4)
14
0o 940 O f-D 0
*9 040 0-
0 -4 fn4 m
F4 "
I0.4
SE
P4 "
NO
4 1
69
clieuical-diffuuion find finally diffasion coriLrol WOUld take
over with a build up of a product layer extending up to
400 microns into the melt. Whereas in Reed and Barrett's
work, the amount and dimensions of the melt phase were to
all intents and purposes a semi-infinite system, the present
work is a finite system with a 15 micron thickness dimensior
on the glassy phase. It is obvious, therefore, that the
continuing vigorous chemical reaction will at all times
destroy the tendency for a concentration gradient to form.
A deduction of paramount imporance imaybe made as a result
of these observations, i.e.,Lhe alc e o:i will Dfso most
certainly hold true for the case of alumina dissolution by
the fluxes during the firing of the special alumina bodies
and the great majority of commercial ceramic bodies as
the critical dimension of the grain boundary is of the order
of one micron.
It was interesting to note that no relationship

existed between the end alumina concentration and the
reaction temperature. The alumina concentration having
increased by the same amount in both samples #341 and 342
to 13 - 15%. The finding confirmed the earlier work in
.iis and the other CaO.SiO2-Al O systems investigated,
as shown in Table 4-13. an in i ates that chemical
equilibrium is not attained (see prior sections).
It would be expected that the reaction rate

of the forward reaction
dt YsCs
would increase as a function of temperature according to

the equation
K = Koe - AH/RT
where K is the forward rate reaction constant, KO is a
constant, R is the gas constant and T is the absolute
temperature: dq/dt is the rate of the forward reaction,
C s is the saturated alumina concentration, which also
increases with temperature, c is the concentration of
alumina in the melt and 6s ang Ym are their respective
activities in the melt.
Based on the present data one must assume that
either K or I- mCm decreases with temperature increase
""its
70
or that both are constant. The net result is that LH,
the temperature coefficient of the reaction, is zero.
It has been shown that Y = 1 for six fold co-

ordinated alumina over the range of saturation concentration
of interest to us (3rd Quarterly Report) and that .'
increases with rise of temperature over the region 0% to 50%

alumina. It is j possible that this factor could explain
the present results.
Other published data however. indicates that

the reaction rate does increase with temperature with an
activation energy in excess of 50 K cals.
It would be expected that a closer approach

to an equilibrium condition would be attained in the grain
boundaries of the special alumina ceramics.
A complete sandwich was not available for

simple #343, so that we cannot be sure that the C:.1O has been
,!!mved from the system by volatilization in the '12 /3N 2
.-
atmosphere. if this is the case however, then we are left
with a very viscous melt high in silica and alumina. Alternatively,
Roy2 6 has shown that this region of the phase diagram can
produce a glass in glass separation. It is therefore
possible that this had occurred in sample #343.
Phase I corresponds to the mullite composition,

2 A-12 0 3 .Si0 2 . The possible removal of the CaO by volatili-
zation has placed the composition in a region of the
CaO-SiO2 -Al 0 ternary phase diagram where mullite will be
the first phase to crystallize. It will n4nlf.eate and grow
from the alumina interface. The microprobe data shows that
Phase I does indeed extend from the alumina interface into
the glassy Phase II. X-ray analysis of the couple is
required to establish whether the qlass-in-glass hypotheses
or the mullite crystallization theory is correct.
If the CaO has not been volatillzed. then a

region rich in CaO.6AI2 03 must exist. It is possible that
this spalled from the saAple during its preparation. A
s"ctrographic scan of the sample #343 revealed contaiina-
tion by several other elements which amounted to 5 to 10%
of the sample. These are listed with their approximate
concentrations in Table 4-15.
It is possible that this contamination affected

the X-ray absorption factors as well as upsetting the total
analysis. The source of this contamination was most likely
71
TABLE 4-15
Contaminants in Phase II of Sample #343
Element As Oxides
Fe .6 to .9% Fe2 3 .9 to 1.3

Ti 1 to 2 TiO 2 1.7 to 3.3
K I to 2 K2 0 1.2 to 2.4
Cb 2 Cb 2 0 5 .7
Zr .2 ZrO 2 .9
P .2 P2 05 .5
the polishing operation in which the glass-sapphire picked

up the above elements which were left from a previously
polished sample.
It should be noted that X-ray analysis of the

941 bodies only confirmed the presence of calcium-containing
compounds such as CaO.6 A1 2 03 or anorthite, thus indicating
that the behavior of the couple is not duplicated in the
ceramic body itself in this respect.
Further evaluation of the above microprobe

results is not justified until furtber microprobe studies
backed up by electron microscope work and X-ray diffraction
work can be programmed.
4.2.3.3 Properties of the Glassy Phase
4.2.3.3.1 Viscosity
Alumina flows plastically above 1000 0 C. At

1000°C the flow is imperceptible, but at 1800 0 C long single
crystal rods can be bent into U-shaped geometries. The
"glassy phase" of polycrystalline alumina ceramics is,
however, several orders of magnitude more fluid than the
alumina crystal at the same temperature. The amount and
type of the glassy phase, therefore, controls the tempera-
ture at which the ceramic will sag and the temperature at
which the glassy phase is fluid enough to flow out of the
grain boundaries of the ceramic onto its surface and into
a metallizing sintered porous sponge.
72
Figures 4.24 and 4,25 show the viscosity of the
grain boundary phases in the special ceramics 941, ,iFig. 4.24)
942 and 992 (both Fig. 4.25). based on the best values
available in the literature. The maximum viscosity in the
curves lies approximately at 470/ and 37%. alumina respectively.
The sag bars listed in Table 4-16 were used to
indirectly determine the relative viscosity of the glassy
phase of the 941 and 942 series of ceramics at 1425 0 C.
The bars were placed in a furnace under simulated
metaliizing conditions such that they had unsupported span
of 5 inches over the center section of the 6-inch bar. When
the furnace temperature approached the softening point of
the glass phase, the bars sagged under their own weight.
The extent of the ':sag" under these conditions is determined
by:
a. Temperature
b. The time at temperature
.c. The composition of the glass phase
d. The amount of the glass phase
e. Presence of a continuous crystal network
structure.
The composition and amount of the glass phase are the only
variables in the test results discussed in thia report.
The composition of the glass phase is different
in the 941 and 942 bodies because of the deliberately
introduced differences in the silica-to-calcia ratio.
Since silica is a network former it tends to make a glass
more viscous at a given temperature. Conversely, calcia
is a network modifier and it will tend to make glass more
fluid under similar conditions. Alumina occupies an inter-
mediate position, i.e., it can either be a network former
or a network modifier depending on the calcia-alumina ratio
in the glass phase. Thus at low alumina percentages the
viscosity will increase with alumina content i.e., it is
acting as a network former, whereas at higher alumina
percentages, the viscosity will decrease, i.e., it is
behaving as a network modifier. This subject is discussed
in more detail in the section on structure, 4.2.3.1.
From the above discussion, we can see that the

behavior of the 941 and 942 bodies will be different under
the sag-bar test conditions describea above. The actual
results of the test after a sag time of one hour at 1425 0 C
are shown in Table 4-16, colutan 3. Referring to Figs. 4.03
73
1001
90-
80-
Viscosity vs A1203 Content S102 /CoOn 1/1
70-
a60- TheOrecticag Maximm inviscosity
: 50-I
>'
30- 14500C
20-
116009C
010 20 30 40 50 60 70 80
Wt. %A20
Figure 4.24: Viscosity versus A120 3 content for melts with 1:1 SiO 2 /CaO ratio. Dash lines indicate
probable viscosity of melt supercooled to a glass at indicated temperature.J
-74-
110 1400*C
/
100 1450"C
90
80
Viscosity vs Al 03 Content S0 2 / cOo . 2/ ,
70-- Theoreticol Maximum in Viscosity
' /SO" I
:60 I
15000C
50I
050-
40-I
30-
20 1600C
10 .I
- I"Xcooc
ISOO
0 I I I - I I I
0 10 20 30 40 50 60 70 80
Wt. % Al 2 0 3
Figure 4.25: Viscosity versus A120 3 content for melts with 2:1 Si0 2/CaO ratio. Dash lines indicate
probable viscosity of supercooled melt at indicated temperature.
-75-
TABLE 4-16
Correlation of 1425 0 C Sag-Bar Results
on the Special Ceramics with Viscosity and Chemical
Composition (as determined from
phase equilibria calculations)
(I) (2) (3) (4) (5)

Body Weight Per Equilibrium
Type Cent Chemical 1425 C Viscosity Alumina
and Firing Composition Sag Poise % in
Temp. C SiO 2 CaO A1 2 0 3 Inches 1800°C Glass Phase
(a) 1b)
941-1500 3 3 94 - -
941-1550 3 3 94 0.092 2.3 45.6 - 42
941-1600 3 3 94 0.059 2.0 48.1 - 42
941-1650 3 3 94 0.033 1.9 52.0 - 42
941-1700 3 3 94 0.014 1.7 57.8 - 42
942-1500 4 2 94 - - -
942-1600 4 2 94 0.022 4.0 46.3 - 37
942-1650 4 2 94 0.016 3.3 51.0 - 37
942-1700 4 2 94 0.012 2.7 56.3 - 37
942-1750 4 2 94 0.012 1.9 62.6 - 37
ka) at temperature
(b) between entectic point (14000 C for 941, 1512 0 C for 942) and
room temperature
76
and 4.04 and to Table 4-10 and 4-11, it can be seen that the
942 "glassy phase" will be very largelg, if not completely,
composed of anorthite crystals at 1425 C.
As expected, the less viscous glass/melt phase in
the 941 ceramic, as opposed to the more viscous and smaller
quantity of glass phase which exists in the 942 ceramic,
caused more slump in the 941 ceramic..
The decrease in sag of both series of bodies

with increase of their initial firing temperature shows that
the viscosity of the melt phase in these bodies is greater for
the bodies made at the higher temperatures.
The viscosity data, Figs. 4.24 and 4_25, con-
sidered together with that data given in Table 4-11 shows
that the alumina content of the 941 glassy phase must
progressively increase to a maximum in the region of 40-47%
for the 941 series of bodies. Thus, assuming equilibrium
at the manufacture temperature (as was done to obtain column 5
in Table 4-11), some but not all of the alumina must precipi-
tate from the melt on cooling; the bodies fired at. the higher
temperatures retaining more alumina in the glass phase.
Little precipitated alumina will redissolve on heating to the
1425 0 C sag bar test temperature as this nears the solidus
line on the phase diagram, Fig. 4.03.
A similar situation occurs in the 942 series of
bodies. Again it can be assumed that if alumina equilibrium
is attained during the manufacture of the body, alumina will
precipitate on cooling prior to the grain boundary phase super-
cooling to a glassy phase. This glassy phase then almost
completely devitrifies to anorthite (and alumina) on reheating
to the sag bar temperature of 1425 0 C. Table 4-10. Thus
although the viscosity trend in the 942 body is similar to
that of the 941 body, the variation of the sag amount over
the series is less because the amount of glassy phase
present will be very small.
It is noted that the sag of 942-1700 and 942-1750
is the same. Three explanations or combinations thereof are
possible to account for this fact -
(i) The amount of glass phase is relatively less,
see Table 4-8
(ii) The amount of alumina in the glass phase
straddles the viscosity maximum at 37%,
Fig. 4.24, and could, for example, be 35%
77
and 39%. This would result from a departure
from equilibrium as discussed in the text
of Section 4.2.2. This departure from
equilibrium is likely to be more pronounced
in the 942 than in the 941 body because of
the greater viscosity of the melt phase.
(iii) Figures 4.08 and 4.10 show that the mechanism
of final grain growth in the 942-1750
ceramic is due to solid-solid alumina
diffusion. The formation of a substantial
rigid skeleton structure in the 942 ceramic
can provide an alternative explanation for
the levelling out of the sag bar 'sag".
4.2.3.3.2 Ionic Diffusion
While viscous flow provides the mechanism for

mass flow of the melt, ionic diffusion of its various ionic
species provides a mechanism for selective migration of
cations such as Ca 2 and anions such as Si7A6 + Fig. 4.26
shows the relative rates of diffusion of eSi 4+ ,
02- and Ca 2 + in the CaO.SiO 2 -A12 03 system. From this data,
it appears that diffusion of Si4 +, A13+ and 02- as silicate
and alumina silicate anions does not occur exclusively as
the migration rate of the 02- ion is much higher than that
of the Al and Si ions.
The melts cited in Fig. 4.26 all have a SiO 2 /CaO

ratio of 1:1. Melt B contains 11% alumina C contains
20% and D contains 40% alumina. It i. to ve expected from
this data that the higher the SiO 2 /CaO ratio for a given
alumina content, the lower will be the diffusion rate.
During the formation of the alumina special

bodies, the alumina grains are reacting with the calcium-
alumino-silicate melt phase that has formed. Reed and
Barrettl 5 have shown that the bulk redction between a
single or polycrystalline alumina surface and an alumino-
silicate glass is initially chemically reaction controlled.
A graduated product layer of thickness, called the Nernst
layer, is gradually built up. The reaction then becomes
diffusion controlled, for the alumina corroded product must
diffuse through this layer. It was shown that diffusion
coeffic ents for alumina may be calculated from the
experimental data. This is presented in Fig. 4.26.
78
Temperature OC
,3500 14000 1450" 1500. 15501 600
g Ix II
IxlO '
E A
U0
C
.0
6
5 10 4
1I
A12
6.4 6. 6.2 6.1 &0 5.9 5.6 5.7 5.6 5.5 5.4 5.3 5.2
I+X10 4
Figure 4.26: Self-diffusion values for oxygen" (1), silicon"3 (2), aluminum" (3), and calcium" (4) for
melt C and aluminum' s (5)for melt B, together with values of "convective diffusion" for
melts C (6) and D (7)derived from alumina solution In these melts". Also included are
estimated Ca2+ values for melt B(8)obtained by extrapolating data contained in reference
30. (See text for key to melts.)
-79-
m z~q~ .=-. .... - ... ,J. . ,, .. ' L- - w -

mm - F .... - .. . ... .
The data presented in Section 4.2.3.2.6 shows
tiat it is probable that diffusion control does not take
over in the case of alumina body sinter firing, at least
prior to equilibrium conditions being established. Data
subsequently to be presented also indicates that diffusion
control is not the factor determining alumina solution
between the non-metallic component of the metallizing and
the alumina phase of the ceraminc.
Diffusion phenomena does become impcrtant in

complex metallizing processes where the ions in the non-
metallic portion of the metallizing countcr diffuse with
the ions of the glassy grain boundary phases.
4.2.3.3.3 Thermal Expansion
The thermal expansion of the glassy phase

formed in the 941 and 942 system below the annealing
temperature* of their glass component can give a picture
of the forms of stress existing in alumina ceramics. The
data may also be used to determine the stress existing on
a pure metal metallizing layer sintered onto the ceramic
surface.
In this study, preliminary thermal expansion

measurements of eleven glasses were made. The thermal
expansion of several crystalline phases which have bign
found in high-alumina bodies are shown in Table 4-17 o .
Thermal expansion coefficients of eleven glass

compositions, fused silica, cast molybdenum. and sintered
molybdenum were determined with a Brinkman Dilatometer.
Samples were in the form of 0.375" diameter rods, approxi-
mately one inch long. Sample preparation was discussed on
page 23 of the Tenth Quarterly Report.
The Brinkman apparatus is designed to hold the

expansion sample horizontally in a controlled atmosphere
hcating chamber. The sample is held in a high alumina tube,
and its dimensional changes are sensed by a lightlv spring-
loaded alumina rod. The rod is held in very low-friction
bearings and is coupled to a transducer which converts
mechanical displacement into an electrical signal. The
displacement thus read is the difference between the
expansion of the sample and the system (the alumina tube
and rods). It is, therefore, necessary to standardize the
*The expansion of the melt phase cag be calculated from the

density data of Barrett and Thomasa.
80
0
41
0 N 4
.4X
4.))
o m'
fn 414
4 0 LALAL NL
0 U1 cf
ul.4 0 0 . CD OD M n W. to
IL)4 4M U0'*02
002
.4 0 0 D c1 OD LA ON u
0 C 0 ~ 4)
"V4 m-
Iv C
4U UEU
'C 0 ~ 0
002 NL 0
02 0n 4NN - N
64 N x L,>1
(n xi 0 o ( 00
'-k aO U-
'-0-40 04 N 0#
9 00 A49-0 o4r .94 04 t
14 V0 c 0 0
811
system with a known sample such as fused silica. The sample
chamber was provided with a flow of argon gas to prevent
any chemical reaction such as oxidization of the molybdenum
samples.
The procedures employed to obtain the thermal

expansion data are described as follows:
Two samples of each glass composition were run.

Each of the two was run twice to give a total of four runs
per composition. The reported results are therefore averages
of four values. The runs were exten-ed to 800°C or 9000 C.
depending on the softening point of the glass.
The furnace heating cycle was progi.ammed for a
heating rate of 60 C/min. and controlled automatically. The
electrical signal from the transducer was fed to a chart
I
recorder which traced the expansion difference between the
system and the
similar traces
sample. By comparison of this trace to
made determined.
on fused silica, the true expansion
I
of the samples was
The results appear in Table 4-18 and Fig. 4-27.

These are average coefficients of thermal expansion
calculated from chart-recorded data. The glass compositions
were determined by wet chemical analysis. The difference I
between the compositions as determined by direct analysis
and the nominal batch compositions resulted from the
solution of additional alumina into the glasses with high
liquidus temperatures, which were melted in al,;mina
crucib~es
The results plotted
in Fig. 4.27 show
some
definite trends. The relative magnitude of the expansion
coefficient of each glass was primArily determined by its
silica content as was expected. The glass with a 11 I
SiO 2 to CaO ratio 13A) had a high coefficient compAred to
the glass with a 3:1 ratio 'A when both contained similar
amounts of Al 0 . The addition of Ai 2 O 3 initially caused
a reduction oi 1xpansion coefficients as the Al 3 ions
entered the glass structure in tetrahedral or network-
former positions. Ihen the A1 2 0 3 /CaO mole ratio reached
1:1, additional Al ions began to occupy octahedral or
network modifier positions which caused an increase in
expansion coefficients. The similarity of the form of the
data with that of the viscosity data is obvious.
82
7- mom, ,,Z
TABLE 4-18
Thermal Expansion Coefficients of Glasses
Thermal Ex-
pansion Mole Ratio % A1 2 03
Glass Coefficient 6 Si0 2 /CaO
No. in'in/OCxl06 Nominal Actual Nominal Actual
100-8000C
3A 6.78 1:1 1.04:1 15 16.1
10 5.41 1:1 0.85:1 25 22.1
1 6.28 1:1 1.05:1 30 30.1
11 4.82 1:1 0.90:1 40 37.4
i 2 4.60 111 1.08:1 50 52.3
1 6A 5.47 2:1 1.93:1 15 14.3
4 4.26 2:1 2.2:1 30 37.4
1 5 4.30 2:1 1.69:1 45 42.4
9A 4.54 3:1 2.66:1 15 29.3
7 3.56 3:1 2.64:1 30 31.4
8 4.59 3:1 2.62:1 40 39.7
Sintered Mo 5.95
I Cast Mo 7.38
I Fused Si02 0.624
8
83
00
cli 0
0C
00
E'
'E
0
0
0 0
W4-
0
cu 0
0, a
N *1
0 %
0 0
cb 0
g~IO 0U.
000004 00 uvC
3 oot'901X
usud3IW41A03110:En
--- 84-
Some systematic error may have been introduced
by the use of alumina tubes in thL system. since the
accuracy of the results depended upon the calibration of
the system.
Preliminary empirical calculations had indicated
a higher expansion for the 1:1 glass. 8 It was hoped that
the thre. Ilasses would span the range, greater than
alumina, between alumina and molybdenum and less than
molybdenum. This difference in turn would lead to differences
in metallizing strength due to different stress situations.
As can be seen from Fig. 427 based on the probable
alumina contents of the systems being between 37-47%. the
thermal expansions are similar and lie below that of molyb-
denum. This is the most favored situation, i.e., the glass
phase will be in compression both in the grain boundaries
of the ceramic itself and in the metallizing layer.
In order to obtain a range of high expansion

glasses one must evidently synthesize glasses containing
the glassy phase of the 941 ceramic plus selected elements
that could be incorporated into the non-metallic portion
of metallizing points.
4.2.3.3.4 Contact Angle Studies

The surface energies of the glass phase in
relation to the alumina ceramic and the sintered molybdenum
metallizing layer are of primary concern to a basic
metallizing study. If adequate wetting and adhesion does
not occur a vacuum tight seal cannot be made.
Thic section presents the results of a pre-.
liminary series of experiments and later experiments which
used the same glasses deployed in the thermal expansion
studies. The section concludes with a brief interpretation
of and discussion of the results, further discussion being
reserved for a later section on adherence mechanisms.
in both series of experiments the wetting is

described in terms of the contact angle that the melt makes
with a solid plaque at the metallizing temperature. The
lower the contact angle the better the wetting.
The arrangement used in the first series is

shown in Fig. 4.28 together with a diagramatic cross-section
of a sessile drop on the plaque. An adequate description
of the method used -- the sessile drop method -- may be
found in the literaturi.2 9.
85
S 0
0 EE
.su
au 0E ~ a,
E* . 0
0E0
I5
E
Sw
z~ 0* 1 cc.
iCO 2uo CL-

x 4)0
-0 go 0
0u, . .20
W)CG d ic 0
- c 0
0 u
o .. j )~.
CbC
4
U))
VI CP 4
a-u~ 4)
CP C)
u E
0 0
0 11 0
4-
4)
-86-
Substrates examined were molybdenum, tungsten
and sapphire. The main melts examined were of the anorthite
(CaO.2SiO2.Al20 3 ) and gehlenite (2CaO.SiO 2 .A12 03 ) composi-
tion. The melts were made by compounding the raw oxides
and melting them in an alumina crucible (1\RR,99.9/ alumina)
to establish homogeneity. Impurities such as Fe and Ti
were added to similar melts to discover the effects of trace
elements in the glassy phase of the ceramic on the wetting
of the substrate plaque. A silica free eutectic melt
CaO 41.5% MgO 6.7y/. AI 2 0 5 51.81(2.?13450 C) was also
investig3ted. The experiments were carried out at 1550 C-
1600 0 C in a hydrogen and hydrogen plus nitrogen atmosphere.
'The results are given in Table 4-19.
The surface tension studies under wet H2 and
S2/3N conditions at +lOOCF dewpoint were repeated in an
effort to obtain consistent results for the contact angle
on molybdenum and tungsten- The contact angle varied
from 300 to 00, i.e. complete spreading. In one case
bubbles of gas appeared on top of the liquid sessile drop
at temperature. It is thought that the crystal orientation,
chemi-absorbed surface layers and outgassing of the metal
substrate affects the metal surface condition in a wet
atmosphere. This alters ' and i- This variability
in wetting will affect the meTahlizingSbeha-ior of the
metallizing system.
A spectrographic analysis of the melts after

being melted on plaques shows a variable amount of metal in
the melt phase as given in Table 4-20.
In the second series of experiments, the contact

angles of nine different glasses on molybdenum and on
sapphire were measured. Sessile drops were photocraphed
at temperature, and the contact angles were measured on the
photographs. All data were taken in an atmosphere of 3:1
nitrogen to hydrogen by volume, at 1600°C except (as noted
in Table 4-21)where the liquidus temperatures of the glasses
were higher, and at dew points of 0°F, 200 F, 40°F and 800F.
To overcome the difficulties experienced with

the first contact angle measurements, a new test furnace
was employed.
The test furnace, a sketch of which appears in

Fig. 4-29, was a vertical electric resistance type with a
tungsten tube heating element. Holes in the heating element,
the two molybdenum heat shields, and a viewing port in the
water-cooled outer shell permitted observation of the sample
87
I(n Ln co fH (V Ln N 0 0
~ZN
N N N-
rL
N4 4
wOj
.4. 0
9 N
-4 +
0 0 N I4
0 z
e
t'U) LA) r-4 LA) 0
z Nr NN
)I N N
0 CN
0 K~ LA) iLro LA) 0D 0' 4

0..ON MH-- N r- rn N
CV M
E-4 HOUcf
9D N LA) H-
0 r4'
o1 0
0 N C4 0-- N
0 04V
N N
0 00 f4 0 0 0 u
4.) 4. ) 4.) 4.) 4 E-1 - 4 ) .4.) 4-) -U Cl)
- pq P4 F4 rq " W~Iz -H u'ri,4 0' -H(
4.) 4.) 0J 0 4) 4.) 4.) 4) u u
H- 0 0 r. .0 0 ' 0' 0 0O: 0 4J
0 0 + 04 W t0r~d
H
4.) V
-) O HP44
88
TAB3LE 4-20
Solubility of Metal in Melt
Melt Metal Solubility Atmosphere
Anorthite, top of drop Mo Mo: 0.002% Wet, H2 + N2

Anorthite, bottom of drop Mo Mo: 0.000% 4
Anorthite, before test -- Mo: 0.000% I
MgO, CaO, A12 03 Mo Mo: 2.0% i

Eutectic sessile drop
Anorthite W W: 0.1%
89
WATER ~TO GAS OR VAC.
VIEWING POWER
PORT SUPPLYJ
CAMERA,
T i
HEAT
SHIELDS --,,"
J -- '
6-- ' - = "
W RESISTANCE HEATING ELEMENT--
i
Figure 4.29: Schematic cross-section of furnace used in wetting angle study.
-90 -
at temperature, Photographs were made through the viewing
port and the sample temperature was read with an optical
pyrometer to an accuracy of ± 100 C. A 10 KW power supply
was designed to supply over 700 amps at low voltage to
reach operating temperature.
A constant weight of the glass sample (100 mgm

on molybdenum and 50 mgm on sapphire) was lightly compacted
into a pellet and placed on the substrate which was in turn
placed on a pedestal located in the center of the hot zone
of the furnace. The furnace was closed and sealed, and the
atmosphere changed by repeatedly evacuating the chamber
and backfilling with 3N2 /H 2 gas The sample was heated to
1600eC (or higher when necessary, in approximately 20 minutes,
taking care to expel all gas entrapped in the glass just
above the fusion temperature. All furnace conditions
except dew point were held constant for the duration of
the ran. Because the furnace
contained no porous
ceramic
insulatic-1, the dew point could be varied between -100°F
and +100°F fairly rapidly. This was not pcssible in the
first series of experiments, The contact angle was allowed
to stabilize at 0 F dew point. then the sample was photo-
graphed. The dew point was then increased to 200 F, the
sample allowed to stabilize and again was ghotographed.
This was repeatcd at D.P.'s of 40 F and 80 F, then the D.P.
was lowered to less than 25 F to observe the reversing
effect on the contact angle.
The contact angles were measured on enlarged

photographs with good resolution and have an accuracy of
about 50, but reproducibility was no better than ± 150.
These results are presented in Table 4-21 and Figs. 4.30
through 4.35.
Spectrographic analyses of glass beads 1, 6A

and 9A whose contact angles had been measured on molybdenum
disClosed trace amounts of molybdenum (0.008%, 0.005% and
0.002% respectively. This evidence Confirmed suspicions
that the glass did dissolve small amounts of molybdenum.
Adherence of the sessile drops to the sapphire

substrates was excellent and it was fair to good on
molybdenum. Drops could not be chipped from the sapphire
plates, and when removed by leaching in HF, the sapphire
surface beneath the drop was noticeably eroded. In most
cases, the molybdenum substrates had to be bent through
more than 50 to pop the glass drop off the surface. They
could not be removed by finger pressure alone.
91
I
I
TABLE 4-21
Wetting Angles of Glasses on Molybdenum and Sapphire
Glass Number** Liquidus Angle at Indicated Dewpoint
on Temp. 0 F 20 F 40 F 80 F Less than 25 F
Molybdenum 0 0 0 0
1 1350°C - 46 14 Ii 24
1 1350 780 56 45 24 60
2 1465 25 16 - 6 -
3A 1250 - 45 10 4 -
3A 1250 63 50 42 12 46
4 1605 48 24 20 20 -
4 1605 68 56 44 32 64
5 1625 16 6 4 - -
5 1625 45 12 5 - 28
6A 1325 48 45 20 6
7 1550 62 5" 48 45 38
8 1550 30 (13) 21 20 57
9A 1400 66 52 46 - 64
Sub-
Saair Plane*
trate [
1 9 1350 210 . ..
2 28 1465 18 - 11 9 -
3A 17 1250 15 12 10 9 -
4 27 1605 24 21 18 11 -
5 19 1625 18 16 15 10 15
6A 28 1325 26 24 14 - 16 I j
7 17 1550 30 28 26 23 23
8 27 1550 27 24 24 24 20
9A 28 1400 22 17 17 10 -
"Angle between C-axis and plane of surface. See pp. 26 & 27,
11th Quarterly Report.

**Glass Corrositions, Weight Percent
No. CaO Si0 2 A1 20 3

1 33.8 36.2 30
2 24.2 25.8 50
3A 41.0 44.0 15
4 22.3 47.7 30
5 17.5 37.5 45
6A 27.0 58.0 15
7 16.6 53.4 30
8 14.2 45.8 40
9A 20.2 64.8 15
92
I
• -. . - :,, . ...- : : m .,. i I I .r l 'i '. ,,,.,, ,,,
~~44-444
+,-4-'14 1
4 +
~~-~ +2~ 4Tt4~-~- .
t t-4- - . t- . .*-. 144 T I.44
4- 4 +*-$
+* .
+ - --
4 . 4*~4* 4 4- *- 9-.. . 44-.t
4444*.-.-4* 44- .9* *.-. -.-
I 1j.I
foI -T
4-m4 J*WLN
lil I * ~ ' +4
+ 4- -- 4
4-~ ~~ ~4-~4.
~ ~- ~ ~~4 -....
4**4~4. -"4, 4- -4
40
-
S4i
-I L-- -
4 .. . ..
4 4 I
-7:r
.
44+..
I I F -4 4- 1
T. .. . . . . *4
-14-4+
t ~ - - - - - -
N -
1- 94 - -
I. T I
. . .
T; 44
T T.
+ q4.
4A
a m T
4 j0
I T
0.4-
44-,
4i~~ T 4 -i
T- .4. .* .. . -- t ' 4 -*~ Cs
-S... . . .~~-. . -
T T4* *
-~ *.t4. 4.- -.. .. *~---.*-. *. to
-. 5.~~~~~~~ *. i . S 0 *S..*4.f
AtC*
05.o
4 4. s4.-4 4
j ~4
4-~ ~ - ~ 4 4--+ - ..
.444 4 4.-.- . +
44--"-- 4 I
,,,~~ - 4 +*4 -4+--.

-'.
-- 5
+ 4- E
+C
-im 0
4 4-4 t)
00
4+4
t4 k
444-
C
0 f-
+LL 0
IT -T I
*- 0 04 0- 000u
-4
4 i- u TIl I I
-96--
+-0- .4
+ +
4-+ -~ - ~ E
+ 0
I In
-+ r4-.
+ -. 4-4E
-
T4 -4
I- 4+ +.
t 1 4 1~1-IF-4~
I4 I- I0
4 0
UJ
44-H
+ ~0 0
IF F
444.
'tt LL
0- 0 0 0 00
3+4 .~.LO
-97
44 .- 15I
++
lH-o
44
+ 1- 4.4 -+--+I -+ - -
41.1,
Ci- 4- 4-
044
24 8
44-
4-4-
c
0~~~9 ----- T
The general tttjids of the contact angle variation
for the preiimina- and final series of results agreed
very well except in the one case where the same melt was
repeated, the anorthite composition 2:1 SiO 2 .CaO, 37%
-'1203 . The second series of results showed a greater
variation with dew point on tie molybdenum substrates.
C
0 With all threeSi 2 /Ca0 ratios, the contact angle
-o on molybdenum plaques initially increas7 with alumina
0- content and then decreases, the intermediate alumina
E content showing the highest contact angle. (There is also
0 a slight overall trend for the contact angle to increase
for all four dew points as the SiO 2 /CaO ratio increases.)
E From Young's equation -
0
U
3 ,,sg - )S1
O Cos 49
Olg
0 L .nce-g is constant an increase in contact

angle must be due to an increase in6'Ig and/or sl and
vice versa.
Barrett and Thomas 8 found that the surface

tension ( Ig' of CaO.SiO 2 .Al2 0 melts increased with
increasing ayumina and calcia cnte.t. Probably the
> increase in alumina (and calcia) reduces the screening of
-n the melt cations. This explains the initial increase in
6contact angle on initially inczeasing the alumina content
t of the three systems. The decrease of contact angle at
C the higher alumina percentage in each melt series must be
o due to a continuous decrease in sl. Thus ?lg and Ysl
are acting in opposition to each Uwher.
Water vapor lowers the surface tension of

soda-lime glasses as -OH groups can be absorbed to increase
the screening of the silicon ions. It follc s that the
contact angle on molybdenum plaques will also be lowered,
as is the case for thr four dew points studied. This
implies that Ysg - ?sl does not change; however, if
this quantity increases, the contact angle will also be
lowered. Both 'sg and .fsl can br, expected to decrease
as a layer of -OH ions will cover the molybdenum. However,
Y.s will be more affected than sg as the -OH surface
presented to the melt phase will be energetically similar
to the surface ions in the melt while sg is initially so
high that the surface energy changes are relatively small
percentages.
99
The low apparent* contact angle on alumina
plaques can be explained on the basis that bls is
essentially zero as a vigorous chemical reaction occurs
between the melt and the alumina phase which produces a
saturated layer of alumina in the melt phase in thermo-
dynamic equilibrium with the solid alumina phase. The
slight change of contact angle with dew point must in
this case be due solely to the change in Ys / ig" An
interpretation of the effect of A1 2 03 content at a given
SiO 2 /CaO ratio is not possible because in all cases it is
probable that a saturated layer of alumina exists next to
the alumina interface (see diffusion section).
The addition of minor amounts of variable

valance ions, in oxides such as TiOx, which can change
their valance states, Ti 4 + - Ti 3 + - Ti 2 +, as the dew
point is lowered, produces a constant contact angle regard-
less of dew point, whereas the addition of variable
valance ions that are in their lowest valance state both
in wet and dry atmospheres (Fe2 +) does not influence the
general trend with dew noint. It is thought that
(i) the melt reacts with the glass phase,
Mo 4 (Ti 4+ + 202-) *. Mo2 + + Ti 2+ + 202-
thus reducing 'is to a low value, similar

to the sapphire-glass situation. Proportion-
ing of oxygen between such systems has been
noted elsewhere.
(ii) the reduction of T 4 + ion-- Ti 3 +--,- Ti 2 + ion

produces an equivalent oxygen-hydroxal
screening action over the entire dew point
range so the Y lg is constant.
(iii) Ysg is so high that small changes such as

induced by dew point are not significant.
The type of bonding produced on solidification

which may be chemical or semi-conducting in nature is
discussed later. (See Appendix I.)
*The true angle will be different due to solution of the

alumina, leaving a circular segment in the alumina.
100
4.3 Commercial Bodies - Microstructure
Considerations
The commercial ceramics investigated are shown

in Table 4-1.
A detailed discussion of Body A is presented

elsewhere;2 8 '1 summary of its pertinent properties is
presented below.
Figure 4.36 shows a 400X ceramographic cross-

section of this body which illustrates the tabular alumina
crystal form and the glissy phase located between the grains.
The ceramic has an 8 micron average grain size with no
grains above 40 microns.
Based on equilibrium phase diagram considerations

and planometric measurements, its glassy phase composition
was determined to be 50% alumina by weight. The glassy
phase was synthesized and the powdered glass placed on
alumina and molybdenum plaques. The glass was sufficiently
mobile to form sessile drops at 1425°C in a wet H 2/3N 2
atmosphere; at 1600 0 C sessile drops on molybdenum varied
from 200 in a wet atmosphere to 600 in a dry atmosphere.
An electron photomicrograph of this ceramic, which has
been thermally etched at 1425 0 C, Fig. 4.37, clearly shows
the surface meniscus of the once fluid glassy hase residing
in theL grdin boundaries between the alumina grains. A
further thermal etch at 1100 0 C in dry hydrogen, in wet
hydrogen, or in air produces a devitrification product with
a greater crystal size, 1 to 5 microns in the latter cases,
Figs. 4.38, 4.39, 4.40. No glassy phase devitrification
is noted on specimens etched at 950°C or below. Body G
is similar in composition and behavior to Body A. It has
already been discussed in this report. These two bodies are
typified by the 942 special ceramic.
Body C possesses a larger grain size and much

more fluid glassy phase. 2 8' 3 0 The more fluid glassy phase is
due to the lower OiO 2 /MO ratio of the glass. In this case,
primary a. rthite crystallizes from the glass phase during
a 1425 0 C thermal etch. After a 1100 0 C etch the associated
glassy phase devitrifies to give secondary anorthite and
gehlen~te crystals. This body has been discussed in detail
in the 2nd Ouarterly Report, Appendix I. This body is
typified by the 941 ceramic body.
101
.j ~ ' -f
;M. ).t
-W
SW~
A *,
S*
Figure 4.36: Ceramographic cross section of Body A showing tabular alumina crystal form and the
glassy phase situated between the grairs. (400X)
-- 102 -
o-A
b S
R '"
.4.
.,
.4 J,,, '
.5 . -
S -
a:ii
. .
.,- . - . e,, . 2 " -. "t
..
' ,- S .
1:. '- ' " ""
// '. , / // (;
' # i/4
~x~y Ashowng
Figue 437:Eletro phtomirogaphot he lesy pasein te bundrie bewee
- lS
the lumna rain. 14,2OX) 9
A 4*?
Jn *
I0 j
4p4
Figure 4,38 Electron photomicrograph of Body A after a thermal etch at I1100 C in dry hydrogen.
(I 8,400X)
.- 14 -
/ 4
of ,
ir
49
Figvre 4 39. E~eciron photomicrograph of Body A after a thermal etch at 1100 C in' wet hydrogen
'8 ,0OO
105
NN
,404
AL.
Figure 4.40: Electron photomicrograph of Body A after a thermal etch at 11000C in air. (28,OOOX)
-106-
The microstructure of the three +99% alumina
bodies investigated in the seal parameter and electrical
studies shows an equiaxed crystal habit, Fig. 4.41, because
of the mode of sintering. No distinct glassy phase can be
seen in their grain boundaries under the optical micro-
scope. (Sintering occurs largely by solid state processes
as opposed to liquid-solid sintering mechanisms for the
lower alumina content bodies.
4.4 Exploratory Metallizing Studies
4.4.1 Introduction
In the first four Quarterly Reports on this

program, a summation of the physics and chemistry of the
metallizing process was given based on an interpretation of
the scientific literature directly or indirectly relating
to the sealing field.
Exploratory studies were carried out in order to

lend substance to this interpretption and to form a basis
for later studies with the special bodies. This section
presents these studies in a logical sequence. They stand
on their own, but, of course, for further details the
reader is referred to the above reports. The interesting
general result that has been obtained from these studies
is that despite its complexity, considerable understanding
can be obtained of an actual commercial ceramic-metallizing
system.
The metallizing paints listed in Tables 4-22

and 4-23 were formulated with several purposes in mind.
The main purposes may be listed as follows:
1. P-1 and P-3 are paints which empirically

-have proven of value for many metallizing
40 applications, so that it was decided to
examine them in some detail. Also SiO 2
and A1 2 0 could not be determined by the
microprobe technique early in this program,
whereas TiO 2 (a network intermediate) which
would diffuse in a cow-mIrable fashion,
could be detected.
2. P-2, P-4, P-5, P-6, P-11, P-12 are single
phase paints which provide simple metallizing
systems capable of analyses by the techniques
used on this program.
107
! b,.
99% Alumina Body Thermally Etched at
1425°C for /2 Hour Showing Primary
Corundum - Alumina plus 2% Gloss
A. Phase at Grain Boundaries. Ave Porosity
._ 5%, Ave Grain Size 13.5,4, Max 601L Min
2.5L Body E
o:" " ," -,. -****e o..
.a .P " . Am-- a-
k, - ' .y .. .'9. oy Size 11.9p , Max 55tL, Min 2.8/A,

Grain
995 As for Fig 5,7 ooiy v

Bod
Groin Size 8.51L, Max 45*, Min 2.51.

" • -"Body F
Figure 4.41: The microstructures of three +99% 9lum5in in the seal par a meter
bodies investighted
and electrical studies.
--I08-
HH Vo
4.) (U .0( E(
a)I VU) HV
0 r'-4 0 Q)c>v
ri ~(a rq 4 ri H r
U I ~ rU I U) to 4-'
o 0 0 (a 0H U)U
4- ra VJ to 4)0>
0 0 (U ) 00+ + H
V4 En' A *>~>)~ + +
4J 04 tpn~-
< 0
4 LO0 Sq~ 0H >U 0'.E
Q) $4 a) a)c r4 -4 -
'4-4 (U > :: H.-i > Ul : LA V
a i u LA
E o
a) 0 0 H t'0 -4N 0 4J D Qw Q1 a) 0) O
Wr $4
.. -1> H r -- 1 04 04 to
a4a nU)) -
14
CN
rqN N NCCN n N CN
N 0 1 0 0 000r-o H 0 U)
.. r4 4 -iH.,I * N
%J N ~
- ' -4 U)
1:: E-1 E- N- (UU
N -) 4-) .L) m I'D C.4 0 in
U) +n 0 0N rq L) U
0++ +U E N +
00010 0 0 0 (
0 0 (Un( (U_
r-U C9 CzN U -
a) CI, N,04 0
04 ) 00 -N It 0 Nc 00
Q- ) (f4
(n-HC1
a40 (a
4- 44U
d0) r-- O
0 0CO 0 00 0 0000 0 0 0 in)
-1 Cl) N CN 0 CNJN N N UA a
a) zH CN:1
41)
H m 0
0 (U 0
*H- 4.) mOO N 00 00 00 0 0 0
to U
00 L
V)
4-) 0) 0 0 00 0 OCOcODO 00 OC)0O a) 0
-4. H H HH H H
U)
E
(
044
E) U)
0 4.v
H- 4-i U) 0
(U 0) 00 0 00 IS00 0 + Q)0 U)
0 .- v0O
4-I 0 (UI
fu H 4
9 HNN
- U' m (V 11 LAw r- co 0 P"4~ 41 UU) 0) H.
LA
tH 4- <H I
(U i - * *i +11 r4a~~-9
04. 04 134l0--
1l.4 alCl4l 04a 2 4 a 4 4a 4 a 4 +E-
109
4J Q)0
4.) u z U uO E-
wI 0 w E4 2: N r4 P E- -4
0 -4 0 a) r.*4 .
u 0 u r
0 0 4 4 U U S * S u u u
4- 4
o) u.
))(
-4 U - 44.
OH -'-
U) .4~) 44 U U z 4 4 4Q
Q) 4-) ~UO 04UU
U) (o r4 a) ) l
rf0 z 4 U -T 4 4 4 p4 04
u E (a (a -H Q
U) 0 4.J
U) u Fuz<rEuzzu
U) 0 00
4.)
Hr- .,I I UF4
(a c- -4 4
It ' 134 a I I I H r4 H r- -4 -4 r- S
(a 0 1I z
i0 0 0
4)f *H 4-i
H- u I E Cl) U) U
H/H IU IU2 Iiii ,U)U .4f 5-Ii

a) :
u
u
4.4 (a
(N U)
0 I00 (N I .
41) 'o0 (N 0 (N40V
.r4I r4 0 .,q
W) ~N LOEO)- U)
o .r.4
0 0
. H0C
(N &0Un %0
4.) H NC 0 uQ
84 (n W'U
mrqr4r4r 4 V-
c
r. 0 9 9 9 H
E 0 00 0 1 M 4
0
rz
~~ Nla i((
110
3. P-7 was formulated as an electrical!y low
lcss paint. It was thought that the Ba 2 +
ions in the melt phase would provide a
nearly lossless condition in the silica
network at radio frequencies. This hypothesis
appears to be justified.
4. P-8 was formulated as a medium loss paint,
It also provides a simple system containing
MnO, a common additive to metallizing mixes.
Yt seals well to sapphire and the phase
relationships of the MnO-SiO 2-AI 20 3 system
are well known so that it can be used to
obtain a further understanding of metallizing
mechanisms.
5. P-8 and P-9 were formulated because of the

extensive knowledge available concerning
the melt phase obtained above. Also, the
metallizing paint and the special ceramics
still form a three component CaO-SiO 2 -
A1 2 03 system amenable to detailed analysis.
The paints were ground in a ceramic ball mill,

unless otherwise noted, with binders and solvents for
48 hours. Particle size distributions for P-I and P-4M
paints are given in Figs. 4.42 and 4.43. These distributions
were obtained by using a Coulter counter.* This instrument
works on the principle that a change in electrical properties
of an orifice submerging in a suitable solvent occurs when
a particle passes through the orifice; the size of the
particle passing is related to the magnitude of the
electrical change recorded. The run shown in Fig. 4.42
was run on an 18 micron aperture in 4% sodium pyrophosphate
in water using Darvan #7** as a dispersing agent prior to
introduction in the electrolyte. When dealing with
materials of different densities in the same sample, the
Coulter counter does not produce a true weight percent
curve but rather a volume percent curve. This is due to
the fact that the Coulter sensing zone completely ignores
*Coulter Electronics Inc., Chicago, Ill.
**R. T. Vanderbilt Co., New York, N. Y.
111
.. K °)
- 112 - CII
__ ___ __ 0I
.. h.
__ __ -1- 1
tV
f-W
f,~ fv -hf ..... IAII
:iit :ii t" 1 T
4-1- -~ T t H .t4
1 4~rl4T I
-i-
90 K ' t - 44
I R-.T
4 4 l
+I -
soI-
Fiur
Paricl4.43:i ditiufnfrillealzngpitotiedb sn outrcut Iiz
IT
-+4- i 4 -413-
31
density and is sensitive only to particle vrlume.
This technique was used because it was felt that

in the complex metallizing paints under study, normal
sedimentation techniques would give erroneous readings.
The data shown for P-1 paint and P-4M paint is typical of
the particle sizes to be expected for the oxide and metal
based paints under study.
4.4.2 The -dint
A metallizing paint consists of a refractory

metal or metal oxide component, which reduces to a metal
in the metallizing operation, to which is sometimes added
a nonmetallic component. This "pigment" is suspended in an
organic lacquer. A typical mixture contains a methyl ethyl
ketone base lacquer, a grinding resin, and a thickening
agent. The considerations involved in the manufacture of
a metallizing paint are outlined in the Third Quarterly
Report, pages 5-13. The paints manufactured on this
program are listed in Table 4-22.
The spectrograph and/or typical analysis of
some of the paint components are given in Table 4-23.
4.4.3 Firing Conditions
The furnace used for metallizing consisted of

a two-inch diameter, three-foot long alumina tube which
was wound with molybdenum tape and which was surrounded by
pure alumina insulating powder. Water cooled preheat and
cooling chambers mated with the ends of the alumina tube.
The furnace was capable of 18250C oleration and could be
run in a wet or dry reducing atmosphere - or air as required.
An atmosphere of three pares of nitrogen to one

part of hydrogen with a dewpoint of 30 C was used to sinter
the metallizing paint onto the ceramic at temperatures
within the range 1100 0 C - 1800 0 C.
4.4.4 Sintering of the Metallizing
The extent to which P-4M paint sinters if placed

on an inert substrate is given in Table 4-24. The molybdenum
-325 Fansteel* powder was not ball milled for this study.
Ball milling the material would have two effects, firstly
*Fansteel Metallurgical Corporation, North Chicago, Ill.

114
TABLE- 4-24
Sintering of Molybdenum Powders

in a 11
2 73N 2 Atmosphere, 30 0 C Dewpoint
Compacting
Temperature Time Pressure Fired Density
°C Min. Tons/in 2 g/cc
1425 30 1 3.76* _t 0.03
30 2 4.40* 0.03
30 6 5.24* ± 0.03
1520 20 6 5.3 ± 0.1

33 6 5.55 - 0.1
67 6 5.65 + 0.1
*Also unfired density
Density of Molybdenum is 10.2 gm/cc
115
it would reduce the particle size of the aggregates and
enhance sintering. Secondly, it would I.ntrcduce impurities
and affect sintering.
The compacts at both _emperacures were

mechanically strong. The data indicates that a volume
diffuson densification prucess is initiated between
1425 0 C - 1500°C; below this temperature a surface diffusion
mechanism predominates with no net densification increase 3 2
We can see that even at moderately hiqh pressing pressures
the porosity is 50%. Thus a metallizing mix using this
molybdenum powder would be expected to be infiltrated
extensively by the melt phase, even under these compacting
pressures (a necessary condition for a good bond) and which
was experimentally confirmed when non-reactive, well-
characterized, glass mixtures were used.
Composites of hypothetical molybdenum and glass

metallizing paints were ball milled, compacted, and
sintered at 1425°C and 1600 0 C for 1/2 hour in the usual
wet metallizing atmosphere. The density of each composite
was then measured to dctermine the effect of the glass
addition on the additional consolidation of the metallizing
composite.
P-4M paint was used as the zero glass standard.

Two glass compositions were added, a 1:1 SiO 2/CaO (mole
ratio) glass with 20% alumina, and a 2:1 SiO 2 /CaO glass
with 16X alumina. These compositions represented the
eutectics on their respective pseudo-binary phase diagrams,
r'igs. 4.03 and 4.04.
The glasses were added in one of three amounts,

10%, 20% and 40% by weight to make the glass-molybdenum
mrxtures. Additional compacts were made only of glass in
order to locate the other end of the density versus
composition curve.
The results are tabulated in Table 4-25 and

plotted in Figs. 4.44 and 4.45. The two figures show the
results of sintering at 1425 0 C and 1600 0 C respectively.
The solid curve in each figure is for compacts containing
1:1 SiO2/CaO glass, and the dash curve is for compacts
containing 2:1 SiO 2 /CaO glass.
The curves marked "A" indicate the hypothetical

densities of composites if the two phases, theoretically
dense metal and glass, were mechanically mixed and formed
no voids. The curves marked "B" indicate the hypothetical
116
TABLE 4-25
Sintered Densities of Glass-Molybdenum Composites
Paint Comp. Sintered Density gm/cc

1425 0 C 1600 0 C
P-4M 5.86 6.88
90Mo+l%l:i 7.58 7.19
80Mo+20%1:1 6.19 6.13
60Mo+400/1:1 4.37 4.36
100%1:1 2.2 -
90Mo+10%2:1 5.57 4.21
80Mo+20%2:1 5.35 5.19
60Mo+40%2:1 3.91 n.d.
100% 2:1 2.4
117
44
-4 t - 1 t
44-
__~~ 44- 7
I- f
litW I+~us
444:Densty
Figueerss t4125C
cmpostio ofmolbdenm-gasscomosits snted
1 8 ...
1.- ...---
t -J
00
H0
44
4 4-
I-4-
Es
in
is)
T.4-
119-
densities of composites if the glass phase and the porous
metal phase (considered as closed-pore particles which
cannot be infiltrated by the glass phase) are mechanically
mixed and form no further voids.
The sintered density of P-4M compacts (5.86 gm/cc

sintered at 1425 0 C, 6.88 gm/cc sintered at 1600°C) was
naturally lees than theoretical (10.2 gm/cc) because
sintering was incomplete at these relatively low tempera-
tures and considerable porosity remained. However, ball
milling of the molybdenum powder before compacting and
sintering did result in greater densities than those
given in Table 4-25 for unground metal powder.
The solid curves in Figs. 4.44 and 4.45
indicate that compacts (or metallizing paint) containing
more than ten weight percent 1:1 SiO 2 /CaO glass very
closely approach hypothetical, voidless density (Curve A).
The dash curves indicate that lower densities were obtained
in compacts made with 2:1 SiO 2 /CaO glass because this more
viscous melt was not able to penetrate the pores of the
molybdenum particles as thoroughly as the relatively
fluid 1:1 Si02/CaO melt. However, all but one density
value for the sintered composites fell between the
minimum and maximum hypothetical densities, Curves A and
B, thereby proving that the glass melt did penetrate the
molybdenum particles to a considerable extent.
Based on the experimental data, it appears that

a non-metallic volume of 30-40% (10% by weight corresponds
to 33% by volume) is suitable to produce a dense metallizing
layer. More data, especially in the 0-20% glass addition
range, is needed before the data can be analyzed further,
and the results applied quantitatively to the sintering
kinetics taking place in the metallizing layer.
4.4.5 Solution of Molybdenum
One complication occurring in an actual metal-

lizing paint when the non-metallic phase is added as raw
CaO, A1 2 03 and SiO 2 is the possibility of initial solution
of the molybdenum in the melt. Typical amounts are given
in Table 4-26. The contact angle experiments reported in
Section 4.2.3.3.4 indicated that if the glass were formed
prior to being placed in contact with molybdenum, then
solution was limited to less than 1.vu ppm, the one exception
being the case of the CaO-MgO-AI20 3 eutectic glass. The
evidence strongly points therefore to an initial reaction
occurring between the metal oxides and/or carbonates and
the molybdenum before the oxide reacts and is neutralized
by the silica.
120
TABLE 4-26
Molybdenum Concentrations in Glasses

Melted at Temperature for 1/2 Hour in Molybdenum Boats
as Determined
by Spectrographic Analysis
Comp. Initial Soak Percent

No. Weight Percent Temp. Molybdenum
CaO A1 2 0 3 Si0 2
1 33.8 30 36.2 1600 0.007
2 24.1 50 25.9 1700 1.0
3 14.5 70 15.5 1800 0.3
3A 41.0 15 44.0 1500 0.01
4 22.3 30 47.7 1600 0.2
5 17.5 45 37.5 1700 1.0
6A 27.1 15 58.4 1500 1.0
9A 20.2 15 64.8 1500 1.0
121
A further investigation was made to determine
whether the molybdenum was actually dissolved or was
present as a finely dispersed separzt2 phase at elevated
temperatures. At 40OX on a polisheu section of a glass
sample, numerous metallic particles approximately 1 to 15
microns in diameter were observed suspended in the glass.
These particles indicated that molybdenum was dissolved
in the glass but that a much faster quench rate would be
necessary to keep the molybdenum in solution.
The mix -- 47% BaO-53% SiO 2 -- showed darkening

of the melt due to solution of the molybdenum during the
melt formation process at temperature and its precipitation
on cooling. Remelting the glass six times resulted in
clarification of the melt due to the Mo in solution at
temperature precipitating out onto the wall of the molyb-
denum crucible because of the insolubility of Mo in a
silicate glass.
The influence of the 'solution of molybdenum"

on the metallizing sintering by a comparison of a paint
consisting of raw oxides plus molybdenum versus a glass
plus molybdenum was not directly investigated. A CaO-
A120 - Mo paint was made and sintered, however. A soft
metailizing resulted at 1425 0 C and 1600 0 C indicating that
sintering was hindered. No explanation is advanced for
this phenomena; one observation is that the melt is very
fluid. At 1800 0 C excess melt is absorbed into the alumina
ceramic leaving a dense metallizing. Suc> a metallizing is
useful for situations where the presence of silica is
undesirable, e.g., in the presence of alkali me~als in
thermionic convertors, an6 reactor liquid metal loops.
4.4.6 Adherence
The melt phase must bond to both the alumina
ceramic and to the molybdenum metal phase if it is to
form a strong seal. Data for the strength of the bond of
sessile drops on alumina .nd molybdenum substrates has
been given above*and was indicated to be good.
During the melting of the various oxide
components in molybdenum boats prior to incorporating them
into the paints listed in Table 4.26, their relative
adherence to the molybdenum boat walls were noted. The
EaO.SiO 2 glass did not adhere at all, whereas the 62%
MnO-38% SiO 2 melt crept up the walls and adhered so
strongly that the boat had to be shattered and the small
pieces of molybdenum had to be ground away from the
* Section 4.2.3.3.4
122
solidified melt. Obviously a strong chemical bond had
formed. Contact angle results on molybdenum and alurina
substrates employing this melt, MnO-TiO 2 or MnO-SiO 2 -TiO 2 ,
have been reported elsewhere; they are almost zero.
Two CLM-15 tensile test pieces of Body A were

painted with the non-metallic phase CaO/SiO 2 -20% A1 2 03 and
were used to sandwich a sapphire disc. The assembly was
heated to 1425 0 C. On cooling, a vacuum tight seal with a
strength of 7,000 psi was formed.
Next, two CLM-15 tensile test pieces of Body A

were butted together and heated to 1750 0 C. Again a tensile
bond strength of 5,000-10,000 psi developed due to migration
of the glassy phase of the ceramic across the joint. A
similar test with Body H (99.5% A1 2 0 3 ) does not give a
joint due to the absence of a glass phase in the ceramic.
It was rioted that in the surface tension
experiments reported earlier that the melts showed good
adherence to the refractory metal. Test assemblies were
therefore made up as follows: ceramic-melt-molybdenum-
metal-melt-ceramic, which were fired in a metallizing
cycle. The assemblies pulled with a strength of 2,000-
5,000 psi. This latter assembly is equivalent to a
metallizing system where the metal and melt phases have
separated into discrete layers. In all the cases cited
above, the joints were vacuum tight.
Detailed investigation of adherence mechanisms,
van der Waals bonding, semi-conduction bonding, chemical
bonding, mechanical interlocking would take us beyond the
scope of this study. It should be noted, however, that
such a study is required to provide the corner stone to
our store of knowledge with regard to metallizing.
4.4.7 Chemical Reduction
During the metallizing process, the furnace

gases react with the metallizing layer if it contains
reactive reducible oxides. An indication of these types of
reaction products for P-i metallizing was provided by
carrying out a gas analysis experiment in a combustion
furnace. The gas mix was fed into the furnace tube at the
rate of 25 cc a minute and the products were fed into a
123
Burrell Chromatograph . These results are shown in Fig. 4.46.
A dewpoint of 70 F for the hydrogen was maintained during
the run. Nitrogen was not used as it was desired to
follow the nitrogen evolution from the nitrocellulose
lacquer. The oxygen and nitrogen peaks are not differenti-
ated. The main gas evolution is due to breakdown and
reaction of the nitro-cellulose with the metallizing
oxides and the wet hydrogen atmosphere.
The components of P-i metallizing paint were

heated separately at 1425 0 C for one-half hour and then
weighed. It was found that MoO 3 was reduced to the metal
while the MnO 2 and TiO 2 were reduced to TiOl.8 and MnO 0 . 9 6 .
These weight losses will be very sensitive to the dewpoint
of the furnace atmosphere.
In connection with the electrical studies,

section 7.3.2., weight loss measurements were made of various
weights of metallizing paints P-I and P-2 on 1" diameter
discs of Body A, Fig. 4.47. These discs were also cross-
sectioned after firing at 1425 0 C for 1/2 hour to give a
relationship between the weight of metallizing paint
applied and the thickness of the sintered layer, Fig. 4.48.
The weight loss in P-2 metallizing was 36%
which corresponded to 33-1/3% loss of 02, as MoO 3 is
reduced to Mo and 2-3% binder loss. No glass flow out is
noted in this system at 1425 0 C and the layer is only lightly
sintered, as is indicated in the cross-section, Fig. 4.49.
(Increasing the metallizing temperature for

firing P-2 on Body A from 1425 0 C to 1550 0 C resulted in
extensive migration of the glassy phase of the ceramic into
the seal area.)
The weight loss in P-i metallizing decreased
progressively from 35.1% to 31.5%+indicating a decreasing
reduction in a component in the metallizing and/or glassy
phase.
Assuming reduction of Moo3 to Mo, MnO 2 to MnO,
for the reduction of TiO 2 to TiO, we find total weight loss
of 29.7%. If TiO 2 remains unreduced, it gives a weight loss
of 28.9%. Further reduction to Ti gives a loss of 30.5%.
*Burrell & Kromo-tog Model K-2. See "Instruments for Gas

& Vapor Chromotography", Catalog 84, Burrell Corporation,
2223 Fifth Avenue, Pittsburg 19, Pa.
+As the metalliring thickness increased
124
-3
600- -2
Temperature -
time curve
500 -
I 400-,- -0
;300-
0
o200-
100- 0
200 400 600 st; 1600 1200 1400 400

Toffows OC
Figure 4.46: Analysis of gases evolved from a metallizing paint in th, form of a layer 1'* x I" by 10
milt thick on alumina cramic, under non-equilibrium conditions.
- 123-
3611
00
35
33
20 40 so
/c)
UNFIRED WEIGNI (moh
Figure 4.47:. Percent weight loss of P-1 mnetallizing on Body Aasea function of unfired Wpayed weight
of coating.
-1 26-
Wafre wow~ wowgh ve POW OM06ec Ihlsbmm
@3
10 to 30 40
Figure 4.48: Unfired sprayed weight of mtallizing points P-1 and P-2 on SoyA versus fired
microscpically measured thickness. Supersedes curve of third quarterly report.
Firing t was %ihour at 14250C.
127-
Figure 4-49: P.2 metall zing on Body A sintered at 1425'C for 16s hour
in a Ht 3NI. +1O0 Cdowpolnt
atmosphetre. The copper platit, copper braze to cupro-riickel can be seen to infiltrate te
porous molybdenum sponge.
-129 -
The difference between these calculated values
and the experimental values would give a binder loss of 1%
to 2.3% based on the lowest loss in weight values of 31.5%,
Fig. 4.47. Actual amount of bin- er added is 2%. A simple
conclusion to draw is that the MnO 2 is reduced below the
MnO level to contribute at least a further 2% total weight
loss. This would correspond to a formula MnO 0 5 - However,
the loss decreases with metallizing thickness, indicating
that the reaction is controlled by reaction and inter-
diffusion with the grain boundary phases of the alumina
ceramic. The electron microprobe data presented in Figs.
4.52, 4.53 and 4.54 shows considerable migration of the
glassy phase into the metallizing layer. It is probable,
therefore, that some of these constituents will be reduced.
For instance, the presenc~e of an intermetallic compound
M-nxTiy was suspected in the case of Body C wi. h P-1 based
on X-ray diffraction data obtained by examination of a
fractured metallized surface: thermodynamically the least
stable oxide is SiO 2 , which, while stable in reducing
atmospheres in the pure state, may not be so in the situation
under study.
Fig. 4.50 also shows that migration of the glass

phase into the metallizing occurs. As th. .hickness of the
metallizing increases, however, a perfectly linear increase
in the amount of glass flow out does not occur (as the
amount of glass flow-out is proportional to the depth to
which Kn3 and TiO 2 can penetrate into the ceramic and lower
the viscosity of the glassy phase) thus the thickness
increase will show the trend away from linearity indicated
in Fig. 4.48. Fig. 4.50 shows a typical seal in this
system, while Fig. 4.51 shows cross-sections of various
thicknesses of metallizing. This latter figure illustrates
very clearly the increase in the amount of glass flow out,
and the resulting glassy interaction zone, as the amount of
metallizing with its associated fluxes is increased.
4.4.8 Glassy Phase Interdiffusion

A seal similar to that shown in rig. 4.50
utilizing a copper rather than a Nicoro (Au-Cu-Ni) braze 'as
subjected to the electron microprobe analysis. The results
are given in Fig. 4.52.
The distribution curves of titanium and manganese

are very similar. Manganese is detected further into the
ceramic than titanium. (There is a larger amount of manganese
present in the original P-1 paint.) From the seal interface
the manganese diffusion into the ceramic, through the glass
129
- 1300.
METALLIZING GLASSY
IRW1IACTION
ZONE
Figure 4.51: Photomicrographs of P-Ion Body A showing relationship between glassy interaction zone
thickness and metallizing thickness (400X).
- 31 --
4
Electron Probe Microanalysis of Metal-Ceromic Seal Scan*4C
Ceramic Seal metal Cu.- N, .

100% -
8 )%A lum i umI-

J4 'o as Ai lt ... .
2001 sI0
%. 20%-
6%, L
3% Calciuu
200'
20%
0%Silicon z
. I 74%
2ell
44%
2 Titanium
450 PL 200*1A 20%

10% Manganese
600#& 2001,
0M
400%/6-1 30%
|
tI
Nickel 0% ,% NO /
0%. Oiffusion
4-Cu.
Cu.- Ni, Washer Ii
Cu. C..
Copper 40%-- Plate Braze I
Molybdenum 40%,olI No Mo. diffusion in braze

or ceramic detected
I
Microns 60 60 40 20 0 20 40 60 80 100 120 140 160 ISO 200
Figure 4.52: Electron probe microanalysis trace scan across Body A, P-) metallizing processed at
1425°C, 1/2 hour, H2/3N, atmosphere, +80'F dew point. I_
-- 132- Ij
. . ... _ ,_ _. _ .. . - _ .. .
-- --
'-... ----- , -
*... +- --- - ....
_ _ -- _ _ _- I -~~
_- a I__
• :-1 - - - ' _ ,i- - ---- '--- ., ___ - - '
.... -_ ~ -
-------- 4.
[ --
" - ,."a . .... -_
-- _---- -------- - - :-
4-.. . .- - ) -
- -_-- I ----_ ____ _ ____
_ _ _ _II
- -
- - ------- -t- - -- - .. . " -
.... .. .. *f "___- _ . a
.. . -.. ._... .- . . .... , ,

_"/ =
_ _ -v 7_ 1 -
T1----- -*-.
I 4L
- - 4--_T- t-,
_ _ Ito-~ _ *-- ----

Figure 4.53: Electron probe micro-analysis 'race for silicon across entire seal region giving evidence of
glassy phase migration.
. .... __ W I f7"
-4----- -
-- - - ---
" . -4- - -- :
- ------
_ _ i-lica,
-- - ii ....
-
. .. ......
-- - . ,,-- --------4_ -- + __. .......... -I
- . .. - - _ -i j i / _ _ _ - - - -- - * '
--- -- * - - - -- ' - - - - - r- -...-- "...-.-
__
... .
...... ________ ____
--
_. __ _.t
~----f-". I _'__
'rz L
....- - _-- .... !" -- - -- ,.- + -
-- :m M-
... r -I ,II'~ -- ___:I __ " ____ - -
- .. . . . .----.....------.--
-- -.. -- I "-" = __ ...... _--
- ---------- --------- "- 1 -
-- --- '-- ....

-! .. .. - -- - ,-. . . '-, - ..--- , -- - -T --
--________'____...._
- " - , : ______ I -- --- 4" ' _" -__ _.

.. ___. .__ .. .... . . .... - -- - 4--..-
. .___ -
. . .__..
. J---
-- +--- __ _-:- =...-- - . ..
- ---:: .. .. _-_....... _-=,__.. -- - --------- =---- -I --- ; :- -
-4:-- 1
M -k
-:...:: - _-[ - - _ - _ .. I
' :: --... .. _-"-:.
.. __..... = ";
-
-
:- --
_-__:
'_!...
_-
.. -
.
--
... _._
, __- _ _-
_ _-_
: --
4
_ _
-,9.---
• .• i- _-
-,-- -- -- ------.-
--- ... •..

.. .... .. -i "----
.... .
. .. ... mI.. m1~- ..... - __ __
-- -i -
---. - . . . ------- 4 . . -4. .. -
--.- - - ........- - ;-----.-- t-- ---- 4--- -is ,4
_:- --.- -,-- -- --- --, :~ - - -- - .... _ _-----
.. - -- I--" A.. .
---- 4- ..- .- r. -L-T--.
. ---- - -- - -
____- .-- . ...-i-; .. ---.. - ... - - -T----

. " .. ... 1 - . ..--- . . . --.
---- . .. .
9. ...... -- . ....- -------- _ __ -

_ _ _ _ _. - t.-..........
to--7- I_27
AZ I-4-
0 - - I& 0
- - - + ------- --..-.- - ~ -. - _ _ _ _-
ago t-.---~--...-- -~-

.. nce ot
---.--..----
.4t. -
. .. . ......... .. ..
.... , - - - -- - - - -
.. . . ,__._n-- _ _ - -
-... .----
.. . - - --.......
-:: ---- -- ... ---- * -e r -~
"-----
" - '" --- -
P-I Metalizing -- - - ... "- ..
- A
_ _ _-- •i • - T-i .i
.- ..-. .. , . . -- -"
"... 7 . ._
.ml
.F
__ __
_ _ _ __ A. . . . . .. . ...
.i
..
. ...
. .
--- -. .
-- c- • ---- - .. _..__- .. -
....
i t --- 1- - - - m -- ' , i -
77 :- ------ -.t--. ____ '.
.- _ - -- 4 --- --- a-.... .....--

-- r,........_f:"- 1LL -......-.---
r .. --... i. -" -..
-. -- - - - - .. .. ....- _ . --- 4-- "•
_. - 4133 ____________.
c ::t3 .............
a0
r ! 0 i5
° I
0 A
0~0
8 4--
40 Ica
;C !
0- rr)
0 'to0I 6.
4 6.
# ,S
) 0 00 0 f
-- -Cdlo--
1U
-k4._ _
phase, diminishes at about 600 microns, and the titanium,
at about 450 microns.
The molybdenum remains within the 45 micron seal,

and shows a rather uniform distribution. None of the
metallizing elements show a detectable diffusion into the
metal braze.
The copper analyses indicate that copper
penetrates the molybdenum for about six microns. However,
this distance is exaggerated by the irregular interface and
the limited resolution of the probe. Nickel from the cupro-
nickel sealing washer has diffused into the copper braze,
giving a residual concentration gradient.
Silicon has a distribution curve similar to
titanium. However, since silicon originates from the inter-
granular glassy phase of the ceramic, a deficiency of
silicon in the 150 micron area of the ceramic adjacent to
the seal is apparent.
A more detailed study was made of the glassy

phase migration and the resulting silicon deficiency.
Fig. 4.53 shows such a trace for silicon across the entire
seal region. It is obvious that a considerable amount of
silica (as the major component of the glassy phase) has
flowed from near the ceramic surface into the metallizing
sponge. Silicon peaks further out in the body of the ceramic
indicate glassy phase regions.
The calcium distribution curve (Fig. 4.52) al.o
suggests this same deficiency. The calcium curve indicates
a pos-ible association with molybdenum in the seal area.
In a second trace, the calcium curve showed an inverse
relationship to silicon, aluminum, and titanium in some
parts of the seal. The existence of a calcium molybdate
complex is tentatively suggested by this dati and gains
some support from the molybdenum solution studies discussed
earlier. Other possibilities exist.
Aluminum exists within the seal anI appears to

be associated with silicon. Quantitative analyses of several
points along the seal gave an average value of 3-3/4 % A1 2 03
within the central part of the seal.
The apparent irregularity in concentration

gradient is explainable in terms of the phase that the probe
is traversing, whether alumina or "glass". A quantitative
estimate of the concentration gradients existing is presented
135
00 0 0 0 0 0 0
0 to 00 0oY
-0
\4 ,
0 0
O 00 2
00
_
0
LCL
4-
00
0C
0~ Cc
h.g
"0
4)
~ic - -4)W
"-..
00
0 0
0. 2
CIL
CLC
0.
UU
0 0 0a0
~137
to 1% at the molybdenum ceramic interface. The copper
varies from 70% in the cupro-nickel,to 85% at the molybdenum-
braze interface,to 65% at the molybdenum-ceramic interface.
4.4.9 Composition of the Glassy Phase
The crystal structure of the "glassy" phase of

P-1 and P-2 metallizings was investigated by X-ray diffrac.-
tion (X-R-D) and spectrographic techniques. Metallized
1" diameter ceramic discs of Body A were acid leached
according to the following schedule.
Sample No. Metallizing Heat Treatment
1 P-2 metallizing on Body A ceramic

2 P-1 metallizing on Body A ceramic
3 P-2 metallizing on Body A, reheated
to 1100 0 C in dry furnace (-800 F)
4 P-1 metallizing on Body A, reheated
to 1100 0 C in dry furnace (-800 F)
5 pure molybdenum strip (control).
These samples were treated in the follow1.ig

manner:
A) as processed - X-R-D on surface

B) leached with 100 ml of nitric acid for 1 hour
1. liquor decanted - 5 ml. spectrographed
2. sample dried; X-R-D on fragile surface
C) as for b, base material removed with wet swab
1. swab material - spectrographed
2. sample dried; X-R-D on surface
D) leached with 60 ml of 50/50 HF and HNO 3 for
15 minutes
1. liquor decanted and 5 ml spectrographed
2. base material swabbed from surface and
swab material spectrographed
E) sample leached with 20% IIIl for 15 minutes
1. liquor decanted and 5 ml spectrographed
2. sample dried - X-R-D on surface.
The appearance of the surface after each treatment

is given in Table 4-27. The X-R-D results are presented in
Table 4-28, and the relative amounts of CaO2SiO 2.A1 2 03
(anorthite) as determined from peak heights is presented in
Table 4-29.
138
I
TABLE 4-27
Surface Appearance of Metallized Surfaces

after Acid Leaching Treatments
1 2 3 4 5
Treatment (P-2). (P-i) P-2 P-I Mo
Nitric acid dark brown light brown, black black

brown*
Swabbed grey with brown grey with black

loose materi- brown brown
al removed flecks fl.ecks
HF/HNO 3 off white green- white- charcoal-

grey grey grey
*Took longest to react.
I
I
139
TABLE 4-28
Summary of X-Ray Diffraction Analysis
Treatment
Sample A B-2 c-2 E-3
1 Mo -A1 2 03 .<-A1 2 0 3 -A1 2 0 3

Mo Mo--
CaO.2SiO 2 .A1 2 0 3 --
2 MIfr-A1 203 Mo20 A203

2MMoMA10 Mo
FeO.A1 2 0 3 * FeO.A1 2 0 3 * FeO.A1 2 0 3 *
3 M0 2 C A1 2 0 3
- *-Al 2 0j- -A1 2 0 3
MO Mo Mo Anorthite
CaO.2SiO 2 .A1 2 0 3 CaO.2SiO 2 A1 2 O 3
4 M0 2 C N-A1 203 ;-A1 203 -A1 2 0 3

MO Mo Mo
FeO.A12 0 3 * FeO.A1 2 0 3 * FeO.A1 2 0 3
CaO.2SiO 2 .A1 2 0 3 CaO.2SiO 2 .A1 2 0 3 CaO.2SiO 2 A1 2 O 3
Mo (trace)
5 Mo, orientated
*Identified later as spiriel phase with only a trace of iron.
140
TABLE 4-29
Relative Amounts of Anorthite Present in Metallizing

Layer (Peak Height) and
Occurrence of Anorthite Main Peak
Treatment
Sample A B-2 C-2 E-3
1 -- 3.215 (8)
3 -- 3.205 (22) 3.20 (10) 3.20 (3)
4 -- 3.205 (20) 3.20 (10) 3.20 (2)
The appearance of FeO.A12 03 (hercynite) phase in

a system, where the only iron present was a trace (0.15%)
in the ceramic was doubted. The migration rate of iron
from the glass phase of the ceramic would have had to be
very great. Spectrographic analysis, Table 4-3 . of
sample 2-2B and a blank ceramic confirmed the doubt. Also
X-ray fluorescence showed no iron on the surface of the
sample. The possibility therefore arose that the phase
was a modified MnO.AI2 03 spinel with traces of iron in
the lattice. To test this thesis a paint consisting of
MnO.SiO 2 was made up in eutectic portions (P-8g) and a
blob placed on a sapphire disc. The eutectic melts at 1250 C.
The experiment was perfcrmed at 1425 0 C in

H2/3N2 + 100 0 F dewpoint. The time at temperature was
1/2 hour. The melt on cooling consisted of light brown
crystals and a manganese alumino-silicate glass on the
sapphire substrate. The crystals and glass were subjected
to X-R-D analysis. A pattern identical with the hercynite
pattern was again found. The iron content of the sample
was less than 0.1%. The indications are that, under the
metallizing furnace conditions the usual pattern of
MnO.Al2 O3 spinel (galaxite) is modified to give a pattern
very close to hercynite. The crystal structure analysis
was not pursued further. The formation of a molybdenum
carbide during the dry atmosphere furnace run is also of
interest.
141
TABLE 4-30
Spectrographic Analysis of Leached Metallized Surface
Indicating the Strongest FeO.A1 2 0 3 Concentration 2-2B
2-2B Blank
Dark Surface White Surface
Aluminum 46. % Aluminum 50. %

Silicon 2.4 Silicon 1.3
Calcium 0.18 Calcium 0.12
Magnesium 1.1 Magnesium 0.95
Titanium 1.1 Titanium 0.0086
Zirconium 0.44 Zirconium 0.17
Manganese 0.81 Manganese 0.0078

Molybdenum 1.4 Molybdenum Not detected*
Boron 0.021 Boron 0.035
Iron 0.028 Iron 0.046

Copper 0.0041 Copper Not detected**
Other elements nil Gallium 0.013
Zinc 0.024
Nickel 0.0012
Chromium 0.00032
Other elements nil
*less than 0.002

**less than 0.00003
II
142
TABLE 4-31
Spectrographic Analysis Data on Leachates and Swabs

from Acid Leached Metallizings P-I and P-2
on Body A
Sample
Treatment Al Si Mg Ca Fe Zr Mo Mn Ti
1 B-1 .005 .02 .002 T .006 -- .4 .001 --

2 (see .01 .02 .003 .005 .006 -- .4 .02 --
section
4.4.9)
3 .008 .02 .002 .006 .01 -- .6 .001 --
4 T .02 T -- .007 -- .5 .002 --
5 .002 .01 .001 -- T -- 4.0 .001 --
1 C-I .004 .01 T T -- .01 .04 --.

2 .05 .02 .02 .25 .008 .02 .5 .04 .02
3 .05 .02 .02 .02 .001 .01 .05 -- --
4 .02 .1 .001 .005 -- .006 .006 .006 .004

5 .005 .01 .02 .005 .004 -- 2.0 T --
1 D-1 P.C. T .15 .3 .4 .8 .2 .008 .2

2 P.C. T .3 .5 .5 1.5 1.0 .05 .5
3 P.C. T .05 .15 .5 2.0 .04 .008 .4
4 P.C. .08 .6 1.0 1.0 1.5 2.0 .3 1.0
1 D-2 .002 .001 .007 .002 .005 .004 --

2 .003 .002 .008 .003 .005 -- -- .004 --
3 .005 .002 .02 .04 .004 -- .004 --
4 .005 .003 .008 .003 .01 -- .004 .003 --
Blank
on swab .002 T .02 .007 .006 -- --
1 E-1 .2 .02 .2 3 .05 .01 .4 .002 .002

2 .2 .02 .2 .3 .03 .01 .003 .2 .01
3 .2 .1 .2 .4 .06 .008 .001 --
4 .2 .02 .2 .5 .02 .008 .04 .0? .002
143
VID IN REPLICA
GLASS
ACTION ZONE
MOLBDENUM
DIFUSONPRODUCT
REACTION ZONE
tAnorthite I CRYSTALS
Gaexite j
ALUMINA
Figure 4.56: Electron photomnicrograph of metallizing P-1 on Body A (nicoro braze) metallized at
1425'C for 1/ hour. (SOQOX)
-144 -
The main spectrographic analysis data is present
in Table 4-31. Use can be made of this data in assisting
in the determination of the placement of ions in the
crystal species.
The P-1 data is in error somewhat as a transfer

to a glass beaker took place during the processing. The
HF solution originally in a polyethylene beaker was
evaporated to dryness in a platinum dish. It was reconsti-
tuted in a glass beaker to 100 cc with water. A boron,
barium, sodium and vanadium (trace) content was noted in
the spectrograph analysis.
In an attempt to visually isolate the crystal
phases in the ceramic to metal seal, low power electron
microscope scans (X8,000) were made of P-1 metallizing on
Body A. A typical cross-section is shown in Fig. 4.56.
Significantly new details over those revealed by the
metallurgical microscope are shown. A definite keying of
the corroded alumina crystals into the metallized layer is
again noted with a "glassy phase" tightly adhering to the
alumina interface. Sub-micron sized anorthite and galaxite
can also be seen in the interfacial layer. A crystalline
diffusion product can be seen in the alumina grain
boundaries.
see how 4.4

now secti.on
Note: We incanthis the tools
all thus far canand
combine
experiments performed
to give the reader a good picture of the "make-up" of the
metallizing layer. With this picture in mind one can now turn
to experiments designed to correlate the properties thus far
discussed with the vacuum tightness, strength and reliajility
of ceramic-to-metal seals.
4.4.10 Seal Strength Versus Metallizing Temperature

The P-1 paint was painted onto Body A CLM-15
tensile test pieces and sinter fired at various temperatures.
The tensile test results are shown in Fig. 4.57. A bell
shaped curve was obtained, all samples being vacuum tight
withothe exception of some of the 1100 0 C fired series. The
1100 C firing temperature was repeated with -
1) P-1 paint ball milled in an alumina mill using

steel balls, for 50 and 100 hours.
2) P-1 paint ball milled in a steel mill using

steel balls for similaL cimes.
145
20
18
16 x
2K 14
S
ocs
4-
0-
1100 1200 1300 1400 1500 I600

Metallizing Sintering Temperature
Figure 4.57: Seal tensile strength with P-I metallizing on Body A versus metallizing sintering
temperature.
--1O6--
. .. . -- ... ' .- ": , NI- L_ I . . il~l .. -- ,' -- i n P

In all cases, the metallizing was soft and vacuum tight
brazes were not obtained. The addition of 1% copper oxide
did not consistently produce hard metallizings or vacuum
tight brazes even when iron was introduced into the paint
via the steel balls; results contrary to expectations
occurred. Further work in this area is indicated in order
to resolve the inconsistency.
The effect of Cu-Fe on aiding sintering has

been previously discussed. (Third Quarterly Report,
pp. 62-66.)
Evidently in the simple P-1 system, 1200 0 C is

required to give sufficient mobility to the glass phase to
allow adequate adhesion to the molybdenum phase ia either
a chemical or semi-conducting bonding mechanism.
4.4.11 Seal Strength as a Function of Brazing

Conditions
The effects of different brazing cycles,

braing temperatures and furnace atmospheres on the
strength of the P-1 and P-2 metallizing was also investi-
gated. The tensile results are shown in Table 4-32.
In addition peel test samples were run on

Body A with P-1 and P-2 me allizings following a modified
procedure outlined by Cole 3. The circular band of the
CLM-15 test pieces were metallized with a roller gun
technique to give a fired thickness -f one mil, and bonded
to nickel plated 15 mil Kovar. Runs were made in the retort
under wet and dry conditions with copper braze (1100 0 C)
and also under wet conditions with Cu-Ag autectic braze
(8100 C). The results are shown in Table 4-33. The "pull"
in pounds is recorded continuously on a recorder attached
Ito an Instron* tensile tester. Thus the average plus
the maximum and minimum "pulls" can be readily determined.
These results of Table 4-32 indicated that a

"glassy layer" devitrification process due to furnace cycle
and atmosphere variables may be responsible for the strength
of the seal in the case of P-2 paint as P-1 is not affected
by such changes. The results of Table 4-33 indicate
different stress levels in the seal due to different
brazing temperatures and cooling rates are responsible for
variable seal strength.
*Instron Engineering Corp., Canton, Mass.
147
TABLE 4-32
Tensile Test Results PSI on Body A Metal-Ceramic Seals*

Brazed with Copper,
under Various Furnace Conditions
Paint Metal Member WL** WR** DR**
P-1 70/30,Cu/Ni 11,820 -- 11,460

P-2 70/30,Cu/Ni 8,46n -- 13,740
P-2 -- 8,518 9,698 10,085
P-2 ...... 10,950
P-2 ...... 10,170
*Average of five, all vacuum tight.

**WL is wet tunnel furnace +60°F quick heat and cool.
WR is a wet retort furnace +fn°F with a slow heat and cool.
DR is a dry retort furnace -80°F with a slow heat and cool.
148
i
I
TABLE 4-33
Peel Test Results (Pounds) on Body A

Metal-Ceramic Seals*
Paint Braze Plate WR** DR***
P-1 Copper Copper 2 4
P-2 Copper Copper 2 4
P-1 Cu-Ag Nickel 18
P-2 Cu-Ag Nickel 16
*Average of 5
**Wet retort furnace (+60°F dewpoint)
***Dry retort furnace (-80°F dewpoint)
1,49
Ow- _. "Now-~ww"-'-
4.4.12 Initial Metallizing Tests
Two series of metallizing experiments were run

using the paints listed in Table 4-22 in order to "prove
out" the correct manufacture of the paints, i.e., no cracks
or blisters after drying and metaliizing, and to see if the
predictions made for the behavior of these paints would be
verified in practice. Compositions of ceramics appear in
Table 4-1.
In the first series of experiments, the main

purpose was to obtain a qualitative relationship for the
strength of bonds to be expected between the three principal
refractory metals (Mo, W, Re) metallized onto three grades
of alumina (94%, 97% and 99% A1 2 03 ). All the seals were
vacuum-tight.
P-2 paint was sintered on Body A (94% alumina)

and Body E (99% alumina) at 1550 0 C and 14250"-. P-4 metal-
lizing paint (pure molybdenum) was also sintered on at
1425 0 C. The tensile test pieces were copter brazed in a
retort furnace using wet hydrogen at 1l00C.
In addition, since Body C (97% alumina) was

investigated extensively in Appendix A (Second Quarterly
Report), P-2 was also met llized on it at 1425 0 C.
Two other special paints, P-11 (W03 ) and P-12

(Re) were investigated in a preliminary fashion on Body A
and Body C. These results are also given in Table 4-34.
The microstructure of P-2 metallizing on the

99% alumina ceramic body E, Fig. 4.58, shows the same open
structure as P-2 on Body A, noted earlier in Fig. 4.49.
The lower metallizing tensile strength on Body E as compared
to Body A reflects the difference in their gloss phase
content at the interface, causing a variable "flypaper
effect". No visible migration of glass phase into the
metallizing is observed for P-2 on Body E.
The higher firing temperature of P-2 on Body A

gives a higher strength because of glass flow out from this
ceramic which increases the area of interfacial bonding.
No such effect is noted for the 99% alumina body, indicating
that the small amount of glass phase from the body and
impurities in the metallizing paint makes a constant but
minor contribution to total seal strength in this case
(over the temperature range studied).
150
TABLE 4-34
Tensile Strength Results for P,-e Refractory Metal

and Metal-Oxide Paints on 94%, 97%, and 990 Alumina Ceramics
Copper Braze - No Washer
Paint Body 1425 0 C 1550 0 C

P-2 Body A 10,000 15,500 + (min.)
Body E 6,000 6,000
Body C 10,000*
P-4 Body A 6,500
Body E 1,000
P-11 Body A 10,000
P-12 Body A 8,500
Body C 9,300
Body A 94% A1 2 03
Body C 97% A1 2 03
Body E 99% A1 2 0 3
* Peel test value was 25 pounds with Cu-Ag braze, Ni plate,

15 mil Kovar.
+Neck brakes.
151
POTTING COMPOUND
METALIZING (2nd coat)
~- METALIZING
CERAMIC
Figure 4.58: P-2 metallizing on Body E,99% alumina. Metallized at 14251C for 1/2 hour.
-152 -
The lower strength of P-4 metallizing on Body A
and Body E in comparison with P-2 metallizing is due to the
substitution of the coarse grained molybdenum particles for
the MoO 3 . The possibility of chemical or mechanical
reaction of the MoO 3 with the alumina phase via the vapor
or liquid phase is present.
Further, considering the data for P-2 and P-4,

it is probable that the major adhesion mechanism on Body E
with molybdenum oxide metallizing is due to the chemical
and/or physical reaction between A1 2 03 which occurs some
time in the firing cycle.
The low strengths obtained with P-2 and P-4 on

Body E indicate that multi-component metallizing paints such
as P-7 to P-10 are required to develop a high strength seal
on +99% alumina ceramics.
The strength of the bond of P-2 metallizing on

Body C, a 97.5% alumina ceramic, is accounted for by the
fact that much greater flow out of the glass phase of the
ceramic occurs than for Body A at the same metallizing
temperature,1425 0 C. This is due to the lower viscosity of
the 1/1 over the 1/2 alkaline earth/silica flux composition
of Body C over Body A. This factor more than compensates
for the fact that there is only a third a,7 much "glassy
phase" present in Body C as in Body A. Aiso the thermal
expansion of the glass phase is a closer match to molyb-
denum, Fig. 4.27.
The fact that W0 3 and Re, as well as Mo and MoO 3

provided good bonds to Body A in spite of (1) the low self
sintering capability of these metal powders at 1429 0 C and
(2) the small amount of glass phase flow out from the ceramic,
indicates that adherence of isolated particles Lf these
refractury metals to the glassy phase initially present on
the surface of the ceramic ("the flypaper effect") is an
important phenomena in ceramic metallizing adherence. The
data are insufficient in quantity and variety to detail the
effect of the variation in expansion coefficients of the
refractory metals on the strength of the bond (W = 4.3 x 10-6
in/in/°C, Mo = 5.5 x 10-6 in/in, Re = 6.5 x 10-6 in/in/°C).
Considering the type of layer formed in this case, the strength
of the bond should not be affected to a major degree by the
small differences in expansion coefficients noted abcve,
since all are below the coefficient for alumina, i.e. the
interfacial cermet layer is in compression, a favorable
situation: the Rhenium metal paint is, however, nearer
153
the alumina expansion coefficient which means that the
composite layer, the glass-metal layer, while still in
compression will be a better match than the molybdenum
glass layer; further the 1:1 glass of Body C does give an
additional small increase in strength due to the better
match as noted above.
The paints listed in Table 4-22 plus the two

additional paints* were painted onto Bodies A, 94% Al 03t
and H, 99.5% Al2 03 . Two firing temperatures were empioyed,
1425 0 C and 15508C. The usual atmosphere of 75% nitrogen,
25% hydrogen and a dewpoint of 100OF were used. The
sintering time was 1/2 hour at temperature. The tensile
test results of these paints on both bodies following a
copper braze with a 20 mil 70-30 cupro-nickel sealing ring,
are shown in Table 4-35.
It is not the intention to engage in a detailed

evaluation of the results contained in Table 4-35, as the
systems are too complex to permit very meaningful analysis.
However, a brief discussion is in order. To reiterate,
the main reason for running these tests was to obtain a
preliminary idea of the behavior of these metallizing paints
before commiting them to the limited number of tensile test
pieces of the special bodies that were available. Therefore,
a limited sample size of only two test pieces was used in
each test.
P-1 Metallizing Paint
The strength of this metallizing paint on the

94% ceramic is apparently little affected by temperature,
indicating a wide sintering range for this metallizing as
has been noted earlier. Taking these results in conjunction
with the results for the 99.5% alumina, one can see that
the manganese oxide and titanium dioxide react more with
the alumina and "glassy" phase at higher temperatures and
provide a superior bond for the 99.5% alumina at the higher
temperature.
The glassy phase from the 94% alumina body flows

out into the molybdenum sponge at 1425 0 C and even more at
*P-2M is a molybdenum oxide paint made in a high alumina ball

mill using molybdenum balls. Analysis of this paint shows
virtually no ball mill oxide contamination. To remove the
last traces of contamination, a paint P-2MR was formulated
using a rubber-lined ball mill and molybdenum balls.
154
TABLE 4-35
Tensile Test Data of the Special Paints on Body A

(94% Alumina) and Body H (99.5% Alumina)
Using a Copper Braze and 20 mil, 70-30
Cupro-Nickel Metal Washer
Paint 1 4 2 5 0 C Bod A 1550C 14 2 50 CBod H15 500 C
P-1 15,900 15,525 9,150 12,775
P-2 15,575 17,250 3,180 2,120
P-2M 9,100 5,210 3,300 2,710
P-2MR 7,520 20,100 4,520 6,170
P-4 3,315 7,100 1,142* 938
P-5 957 13,700 1,540 1,939
P-6 17,500 11,720 975* 1,300
P-7 2,265 5,700 2,375 7,100
P-8 16,400 15,175 15,425 14,275
P-9 6,125 15,775 13,525 12,325
P-10 11,600 11,300 5,140* 8,160
P-li 11,450 16,100 4,260 10,500
P-12 1,700 + -- 1,965 2,830
+Cupro-Nickel braze attacked rhenium.
*One leaker.
155
1550 0 C to give a high-strength bond. However, the use of
the P-2M and P-2MR paint significantly lowers these tensile
strength results at the 1425 0 C level, indicating that the
impurities in P-2 are partially responsible for the high
metallizing strength recorded. It is apparent from a
comparison of the tensile strength data on the 99.5%
alumina ceramic with that of the 94% alumina ceramic that
the flow out of the glassy phase of the ceramic is the most
important factor contributing to bonding strength.
No data have been obtained using this paint on

alumina ceramics. Data obtained using BeO ceramics are
reported later.
Metallizing Paints P-4, P-5 and P-6*

These three paints are considered together as
they are all molybdenum metal paints. Consider the 94%
ceramic first. The paints used all give a general increase
in strength with metallizing temperature. This is expected
because the molybdenum sinters better at higher temperatures
and a glass phase flows out of the ceramic more readily at
higher temperatures. Another interesting observation is
that there is a general trend for the metallizing strength
to increase as the particle size of the molybdenum decreases.
This factor is complicated in the case of P-6 by minor
impurities of nickel, iron, and chromium present in the
molybdenum metal. These impurities are generally considered
to aid sintering.
Surveying the situation with the 99.5% alumina,
one sees very low metallizing strengths, indicating that
flow-out of glass phase from the ceramic is not occurring.
The slightly higher strengths recorded for the P-2 paints
under these conditions are possibly due to the tendency of
the molybdenum trioxide to volatilize and deposit a reduced
layer of strongly adherent molybdenum on the surface of the
alumina grains.
*P-4 Mo metal -325 mesh Fansteel

P-5 Mo metal -325 mesh
70% less than 10 microns,
5 micron average, G.E.
P-6 Mo metal 70% less than 5 microns,
1.65 micron avg., Sylvania.
156
9. -=- 77
This metallizing paint is of interest to us

because of the low dielectric loss characteristics displayed
by it. The tensile strengths in all cases are rather low.
This was expected, because it was observed during the manu-
facture of the glass that the glass shrank away from the
molybdenum crucible, indicating that the glass had a very
much higher thermal coefficient of expansion than the
molybdenum itself. This would be deleterious in a metallizing
paint oxide phase. It is, however, noted that the strength
of the paints increases with temperature, due presumably,
to an increase in alumina content in the paint, which in
turn will reduce the thermal expansion of the glassy phase
of the paint.
This paint shows consistently high strengths on

both ceramics at both temperatures. It was noted in the
manufacture of the manganese silicate glass that it adhered
very strongly to the walls of the molybdenum crucible on
cooling. The crucible had to be destroyed in order to remove
the glass.
Metallizing Paints P-9 and P-10
These paints are useful in the study of complex

paints on the special bodies that we are considering because
they enable the system being studied to remain a three-
component metal oxide system. One leaker is reported in
this system, but it may be due to a bad braze. The metalliz-
ing strengths are not exceptionally high on the present
ceramics, but since the main objective of this project was
to obtain insight into mechanisms rather than high bond
strengths, this variation in strengths is to be welcomed
rather than regretted.
Metallizing Paints P-il and P-12
The relatively high strengths of the P-i1

metallizing paint are unexpected, because tungsten might
not be expected to sinter as well as molybdenum at these
temperatures. This indicates that another factor, previously
alluded to, is at work in this case. This factor, for want
of a better name, we call the "flypaper effect". Obviously,
tungsten is able to take better advantage of this effect
than is molybdenum, because of the lower coefficient of
expansion of tungsten in comparison to molybdenum. The fine
157
particle size of the decomposed tungsten oxide applies to
advantage here, because small particles can cling to
isolated patches of glass phase very readily.
P-12 shows the same characteristics as P-4 as
might be expected from a metal paint. However, a high strength
result was obtained when no cupro-nickel washer was used
(Table 4-34). The lower strength obtained in Table 4-35
may be due to attack of the Rhenium by the nickel absorL ,d
in the braze from the cupro-nickel brazing washer. Unfortu-
nately, the results for the 1550 0 C temperature are not
available. It will be expected that a significant increase
in strength would occur in this case because of glass flow
out and consolidation of the metal phase.
Following the above line of reasoning furthec,

we can see that except where we have a paint which i,
consolidated by a glass phase flow out, we have a degradation
of the seal in the second series of experiments due to
either
1) attack of the metallizing by the nickel in soiutio i
or
2) stresses induced by the cupro-nickel washer.
The first factor is considered the most potent.
4.4.13 Comments on Sealing Mechanisms
Evidence accumulated thus far shows several

mechanisms of material movement that can lead to adherence,
it being understood that the actual bond,whether chemical,
mechanical or semi-conducting,cannot be affected until the
metal and oxide phases are brought into near atomic contact.
In practice several of these movements may occur in one
system. These movements are -
1) "Glassy Phase" flow out from the ceramic
2) Metallizing melt phase wetting of the metal ari
ceramic
3) The "flypaper" sticking of the metal phase to the
highly viscous phase on the surface of the ceramic.
Photomicrographs of the surface of ground ceramics
shows the presence of this glass phase.
158
4) Melting and/or vaporization of MoO 3 or W0 3
onto the surface of the ceramic and its sub-
sequent reduction to the metal, leaving a
large area transition oxide chemical bond at
the interface.
5) Solid state reaction between the refractory

metal and the ceramic giving again a point-to-
point Mo-Mo-O-Al-O-Al chemical bond.
Some pertinent results of a parallel company

sponsored program 3 4 may be mentioned at this point to supply
further evidence of the existence of mechanism 4 and 5.
Molybdenum was evaporated onto a ceramic surface and found to
give vacuum tight seals of strengths approaching 20,000 psi.
It is considered that the mobile molybdenum atoms are able
to accommodate themselves intimately to the contours of
the alumina surface which results in enough atomic inter-
action to form a strong bond, Fig. 4.59. It is probable
that a transition Mo - Mo-O-Al-O layer exists at the inter-
face.
Since mechanism 4 approximates the evaporation

process, a vacuum tight seal may be expected to be obtained,
whereas the solid-solid widely dispersed point contact of
mechanism 5 could not be expected to give a useful bond.
The next section gives a more detailed study of
these eftects.
4.5 Studies Using the Special Bodies
4.5.1 Introduction of Seal larameter Study
The series of alumina ceramics used covered a

range of glass phase contents from about 20 volume percent
to zero percent. The purpose of such a selection was to
observe, among othe; things, the role which the glass phase
plays in the sealing mechanism and to study the influence
of such variables as the metallizing temperature.
Sapphire is the simplest ceramic that can be

studi . On this program it was metallized with the single
component paints, P-2M and P-4M as well as the more complex
paints P-1 and P-7.
The corrosion of sapphire by silicate melts has

been discussed in earlier reports and has been reported
elsewhere in the li.terature.14,15,16 This work shows that
159
Figue 459:Top: Body H, 3s evaporated molybdenum metallizinig, copper braze, P-1 metallizing,
Bottom: Body H, lps. evaporated molybdenum metalliring copper braze 701.30 cupro-I
nickel metal member.
- 160-
M.f
the solution of alumina is initially controlled by a
chemical reaction at the interface which subsequently
changes to a diffusion controlled reaction. Sapphire-
melt couple work reported earlier in Section 4.2.3.2.6 of
this report indicates that in narrow bi-facial zones such
as the 2-10 micron "glassy" layer of metallizing the reaction
remains chemically controlled indefinitely.
Contact angle studies of various melts on
sapphire have been reported earlier in this report. It
was shown that in both wet and dry hydrogen/nitrogen atmos.-
pheres that the wetting and adherence was good. The adher-
ence of the silicate glasses to molybdenum substrates was
also good; this result leads us to expect that a metallizing
system composed of sapphire-silicate melt-molybdenum would
give a good seal combination. Conversely, it has been shown
earlier 3 5 that while the adherence of a MnO-TiO 2 melt on a
sapphire plaque is good with a low contact angle, its
adherence is poor on a molybdenum plaque; hence a combination
sapphire-(MnO-TiO 2 )-molybdenum may not give a good seal
combination.
Lucalox* is the simplest polycrystalline ceramic

that can be studied. It was obtained in three grain sizes
to allow the effect of grain size in the absence of a
glassy phase on the metillizing tc be determined.
The remarks made above for sapphire will also

hold for Lucalox!
The three bodies, 992, 942, and 941 will behave

differently ti sapphire as "glassy phase" flow out from thc
ceramic will occur. Thus the MnO-TiO2 phase in P-1 metallx;-
ing will become a .nO-TiO-CaO-A1 O 3 -SiO 2 phase which can 1.
expected to show good adherence ito sapphire and) to molyb-
denum and give a good seal combination.
Limits were set on the scope of this basic study

in order to attain the other goals of the program, tl'ere-
fore, it was decided to devote the majority of the time to
the study of a simple metallizing system,P-4M, while carrin;,
out exploratory studies using the more complex paint systems
listed in Table 4-22.
*General Electric Co., Tradc N3me.
161
4.5.2 Seal Parameter Studies
4.5.2.1 Experimental Procedure and Results
4.5.2.1.1 Sample Preparation
The first series of tests consisted of metallizing

the five types of ceramics (941, 942, 992, Lucalox* and
sapphire) using P-4M paint at four different metallizing
temperatures (13000, 14250, 16000, and 1800 0 C) while keeping
the metallizing paint composition, furnace atmosphere, and
plate-braze-metal member assembly constant. The 941, 942,
992 and Lucalox* bodies were differentiated further accord-
ing to firing temperature and grain size, so that a total of
15 different bodies were studied. The variables under study
with these bodies were the amount and composition (and
resulting viscosity) of the glass phase and the alumina
crystal size.
The tests used to evaluate the metallized

specimens included (1) an electrical dc resistance test,
(2) a scratch test following the sintering step to determine
the quality of the metallizing, (3) leak check, tensile and
peel tests to determine the mechanical properties of the
seal, and (4) metallurgical examination of the seal to
observe such processes as glass phase migration, sintering
of the molybdenum metallizing, and braze-metallizing inter-
action. P-4M has a particle size extending from the
5 micron ranr-e down to the sub-micron range as shown in
Fig. 4.43.
In order to avoid any contamination of the

paint during processing, the mix was ground in a molybdenum
lined ball mill using cylindrical molybdenum grinding media.
The mill charge consisted of minus 325 mesh molybdenum
metal, lacquer, and resin which were mixed in a 1:1 solid/
liquid ratio by weight. This was ball milled for ipproxi-
mately 16 hours to get a #7 grind (6 to 12 micron maximum
particle size range). Table 4-36 lists the results of
spectrographic analyses of Fansteel** molybdenum powder
before milling and of P-4M paint after milling. The accuracy
of the analyses is .± a factor of 2, so it is apparent from
the data that contamination by the ball mill was negligible
within analytical limits.
*General Electric Co., Trade Name.

**Fansteel Metallurgical Corp., North Chicago, Illinois.
162
TABLE 4-36
Spectrographic Analysis of P-4M Metallizing Paint
Type 380 (Fansteel*) P-4M Paint

Element (Before Milling) (After Milling)
1 Mo Principal constituent in both
Si 0.1 0.15
Al 0.02 0.02
Ca - T
Mg T 0.005
Mn T T
Fe 0.03 0.02
I Co T 0.001
I Ni 0.01 0.006
Cu 0.003 0.005
I
Note: Values are i " facLor of 2.
I
*Fansteel Metallurgical Corp., North Chicago, Ill.
1
163
The silica trace in the molybdenura is regrettable,
a purer molybdenum was not found until late ir the program.
The use of P-2M for experiments where an absence of silicon
was critical circumvented possible mechanism confusion,
however.
The metallizing paint was applied to the test

specimens with a hand roller (striping tool). The paint
thickness was monitored by weighing each piece before
application, and after the paint was dried. The paint was
dried under heat lamps at approximately 100 F.
The metallizing was sintered in a molybdenum-

wound alumina tube furnace at the four temperatures mentioned
above. The hefting and cooling rates approximated a normal
production furnace schedule and the soak time was one-half
hour at peak temperature. The atmosphere was 75% nitrogen,
25% hydrogen, with a 100OF dewpoint.
4.5.2.1.2 Scratch Test
In order to determine the relative hardness of

the metallizing, the scratch test was performed. It
consisted of scratching the sintered metallizing with a
tungsten probe while observing the results under a 16X
stereoscopic microscope. P-1 metallizing on Body A sintered
at 1425 0 C was used as a standard. An arbitrary scale of
1 (soft) to 5 (hard) was employed. The standard was
assigned a hardness of 5. The data are presented in Table
4-37.
Metallizing hardness increased with the sinter-

ing temperature. On sonte bodies, such as 941, hard metallizing
was produced at lower temperatures than on other bodies.
Metallizing sintered on Lucalox~and sapphire at 1600 0 C was
ranked as "soft" (2 and 1, respectively) to indicate poor
adherence. The molybdenum particles were well sintered to
each other, but the bond to Lucalox*was weak, and to sapphire
was almost non-existent. The metal layer was easily peeled
off of the sapphire specimens. These results demonstrate the
importance of atomic contact via the "glassy phase" of the
ceramic in the development of the metallizing bond. Body 941
presumably contained a relatively large amount of melt with
good mobility, while the Lucalox*bodies contain no melt
phase, (0.2% MgO, 99.8% A1203; first melt forms above
1900 0 C), and the sapphire, o! course, is a single crystal.
164
TABLE 4-37
Metallizing Scratch Hardness
Ceramic
Body Firing Crystal Metallizing Temperature
Type Temp. 0C Size, 13000C 1425°C 16000C 18000C
microns
941 1550 5.6 1 3 4 4
941 1600 8.4 1 3 4 4
941 1650 9.8 2 4 4 5
941 1700 10.4 3 5 5 5
942 1600 8.1 1 2 3 4

942 1650 10.3 1 2 3 4
942 1700 12.2 1 2 3 4
942 1750 22.1 1 2 4 5
992 1700 7.4 1 2 3 4

992 1750 25.0 L 2 3 5
992 1800 31.0 1 2 3 5
Lucalox*unknown 5 n.d. n.d. 2 3

Lucalox' unknown 20 n.d. n.d. 2 3
Lucalox unknown 40 n.d. n.d. 2 4
Sapphire single crystal n.d. n.d. 1 1

(flaked (flaked
off) off)
F.;ale: 1 (soft) to 5 (hard: P-1 on Body A @ 14250C)
1. Soft, powdery, most particles removed without effort.
2. Soft. Some particles rL.,loved without effort.
3. Medium. Some particles removed with moderate effort.
4. Hard, metallic luster. Few particles removed.
5. Hard, metallic luster. No particles removed.
165
4.5.2.1.3 Resistance Measurement
A very simple resistance measurement was made on
each metallized body as an additional means of determining
the degree of sintering or densification of the metallizin7
at each sintering temperature. The measurements were made on
drum peel test samples on which the metallizing is applied
to the large outer rim of the sample. A cut was made through
the metallizing layer at one point, so the conductive path
was around the entire circumference of the sample. Electri-
cal contacts were made with copper probes on either side of
the cut.
The measurements were made with a Wheatstone
bridge and with a Simpson meter calibrated with 1.0 ohm and
0.1 ohm resistors. The accuracy of the measurements is
limited by the quality of the contact made by the probes,
so either instrument gave satisfactory results. All the
reported data were obtained using the Simpson meter, but a
portion was verified with the Wheatstone bridge to be sure
of the accuracy of the results.
The resistance data are presented in Table 4-38.
A brief examination of the table shows that the resistance
decreases with higher metallizing sintering temperatures,
as was expected. However, the data for metallizing sintered
at 1300 0 C varies somewh.t from one body to another. This
variation may be linked to the action of the migrating melt
phase of the ceramics. At this temperature, the melt may
help pull the metal particles together, and thus be the
predominant mechanism in consolidating the metallized
layer. The variations at the low sintering temperature
were probably related to the viscosity and amount of glass
in each ceramic body. At the higher temperatures, the
actual sintering of molybdenum particles becomes the
predominant mechanism and proceeds at about the same rate
on all bodies.
4.5.2.1.4 Assembly Procedure

Tensile and peel test assemblies were made in
essentially the same geometries as in previous work. The
minor exceptions are the Lucalox* and sapphire bodies which
are in the form of small cylinders and flat discs respectively.
These shapes were assembled for leak and tensile tests as
illustrated in Fig. 4.60. The entire seal package consisted
of ceramic or sapphire, P-4M metallizing, 0.0003 in. copper
plate, 0.004 in. copper braze, and 0.020 in. 70/30 copper-
nickel washer. They were brazed in wet hydrogen at 1100 0 C
for 3 to 5 minutes.

166
TABLE 4-38
I Metallizing Resistance, ohms
Ceramic Metallizing Temperature

Body Firing Crystal
Type Temp. °C Size, .j 1300 0 C 1425 0 C 16000C 18000C
941 1550 5.6 1.8 0.5 0.4 0.5

941 1600 8.4 1.9 0.5 0.4 0.5
941 1650 9.8 1.8 0.7 0.4 0.4
1 941 1700 10.4 1.9 0.5 0.4 0.4
942 1600 8.1 6.0 0.8 0.5 0.6

942 1650 10.3 2.8 0.8 0.5 0.4
942 1700 12.2 2.0 1.0 0.4 0.4
942 1750 22.1 1.9 0.8 0.5 0.5
992 1700 7.4 1.9 0.9 0.5 0.5

992 1750 25.0 3.6 1.4 0.5 0.6
I 992 1800 31.0 4.4 0.4 0.4 0.4
Lucalox*: Sample geometry was not suitable for resistance

measurements.
I
I
I
I
I
I,
167
Cu-Ni washer
Seals- Lucolox cylinder
Cu-Ni washer
Seals Sapphire disc
Figure 4.60. Lucalox* cylinder (top) and sapphire disc (bottom) assembled for leok check and
tensile testing.
A full series of peel test assemblies using
Cu-Ag braze as well as copper braze was run for the
purpose of comparing different seal stress conditions. All
peel tests were made with an annealed Kovar 15 mil, nickel
clad strip. Peel test samples to be Cu-Ag brazed were
nickel plated. The plating was sintered for 10 minutes at
800 0 C in wet hydrogen. Cu-Ag brazes were made by heating
to 810 0 C in wet hydrogen for less than five minutes. Peel
strengths with Kovar and Cu-Ag braze have previously always
been higher than with Kovar and copper braze. (Page 76,
Third Quarterly Report and pp. 47-51, Fourth Quarterly
Report.)
4.5.2.1.5 Leak Check
All tensile test assemblies were given a leak

check after brazing to see if a vacuum tight seal had been
formed. This check was made by pulling a vacuum of 10-5
Torr in the center hole of the assembly while directing
a stream of helium over the seal area. A helium mass
spectrometer, coupled to the vacuum system, triggered an
alarm if helium penetrated the seal. This eui nt could
de ect any leak greater than 10-10 atmos. cm /sec.
Leak check results appear in Table 4-39. Data

for the 94% alumina bodies (941 and 942) show that a vacuum
tight seal was formed when the bodies were metallized at
1425 0 C or above. This results is closely associated with
the degree of consolidation of the metallized layer which
was indicated by the scratch and resistance test results.
With the 99% alumina bodies (992), somewhat higher metalliz-
ing temperatures were required to effect a vacuum tight
seal. Since the amount of glassy phase present in the
bodies was the c ! major difference between the 94% and
99% alumina crai,..cs, the importance of the glass migration
mechanism was again apparent.
4.5.2.1.6 Tensile Test
Seal assemblies were tested for tensile strength

on an Instron testing machine using a constant crosshead
speed of one inch/min.* The tensile test results obtained
appear in Table 4-40. The general increase in tensile
strength with increased metallizing temperature ixuobvious.
*The effect of the crosshead speed on the results was examined

by testing some asse' 'lies at a speed of 0.1 in/min and was
found to be insignificant.
169
rq N NJP4 N N N N r4HH -
4i-
k H N ~N(N NH (-4 MN NN H
4) -
aM' qt'4
4. 41HA (N NN N F- N Nr F-
4P4
* * C M
11 (NN C% NNN N.~u
14.043f E C)
0
ocu
moot, . 0 0 '* *% - f%OD c 4) C)
.0 0 "W40 ~ON
P4' #4U4 V4
1 00 '5+
'-4 '-.x N '
"X
0000
04*
0- "1
>1 ofto4
170
000U.) ourj 2u z zz
o kn C L* 0 r- 1- 00 r-o0 N r r-
onO r..uti)r Lf0' tO~N n o(3 r-

0 r- 0 co %L m i- .tl .- r- unr
r-4 f-414 F-4 -4 Hr-
0000 0000 ZZE z 0 0
::I L'D000 tfl 0 tr0 un 0 Lf un 0r-

0 -4q Lri m Fl-4 00 r-I r-oric (N Ln C^ V-.i
0n-4 r-~4 003 CN rl D N. V.) r- (N F-4
ID~ 10 t- r- S ODSHr- M m * cS *
--4 r-4,H.-4 r44
E4~ ~V
U) 0
41 -14 E-4S
e -4 0 0 Ln tf-0 iLn ~ t u0
tou Ln OD -4 > MLl ~Or-o0 t O
$.4 .41 (N m~~~
N~' M~U
en r- 1c cu
4J * %1. -. % *.4 - S *. 5 '4H -4 .. 4
i
r- D lq CCC NDI N F--,
-'-4(0.N
-4
F4 w4
1-4) r-4
E-4 2 410 tu
0 0
0 ~r0~~ 0L A0 0 -0 Aj
0) 0 NQr0 0r r0 0 (a f 4).4-4
0 100'
0 Ti 3. w'0
3 Aj
J
e4,* %T y 0- m m to u 4
41.
N
>1 0- 4 P3N-4e 0O
N -- 4 41
1-4~~4 C' 0-44
to u Aj4
U LI
a 0 0000 0000 000 ON r-4 1 4)u

'e4-e %n 0 L0U0 00A00f 00f0 c c C c -4 to
%00
UheC r, %, %OD
F r I Go j.X. x x 4 I,
EU-44 4 4 4 4 -44 -4 *4 F4 f4 CC c C Co I"C~a
000 0 *..e h
* -4 o-4 -4 N N NrN N NN Cam 4 f" C
-4tv 40404 uuu 40
CS
0% 000 0O'0m
171
The variation in strength between different ceramic bodies
is also quite evident.
While general trends are reliable, a high

confidence level in the actual value for any one point on
the graphs would necessitate a much larger sample than
two or three, because tensile and peel strength of seals is
a statistical property.
4.5.2.1.7 Peel Test
The Instron was also used for the drum peel

test. The peel test sample was held in one fixture by a
steel pin placed through the hole in the center of the
ceramic. One end of the metal tab which had been brazed
on the shoulder of the sample was clamped in the other
fixture. As the fixtures were movel apart, the metal tab
was stripped away from the ceramic, which was free to
rotate about the pin. The force (in pounds) required to
sustain the peeling action was recorded on a strip chart.
The results of peel tests appear in Table 4-41.

The strength of some samples with weak bonds which were
easily pulled by hand were reported as less than 0.5 pound.
Specimens brazed with copper were compared with those with
Cu-Ag brazes to determine the effect of the 290 0 C differ-
ence in brazing temperature. The mismatch in thermal
expansion between Kovar and alumina increases considerably
above 8000C and was expected to have an effect on the
strength of the samples with copper brazes. The effect
of this mismatch is most apparent on samples metallized
at 1300 0 C, but some of the other results present a more
confusing picture.
4.5.2.1.8 Microstructure Examination

All the samples tested were cross-sectioned and
examined. Typical microstructures are reported later in
this section. (Figs. 4.70 through 4.73.)
4.5.2.2 Analysis of Physical Property Data
The tables contain a great amount of useful

data, but these data are difficult to analyze because they
reflect the influence of numerous variables which come
into play in the sealing process. These variables were
discussed in great detail in the Second, Third and Fourth
Quarterly Reports as the factors which directly determine
the properties of ceramic-to-metal seals. The values of
172
O LnflID
ONt c3N0 0 G O Cl ON
o0 r- c,4r-4 r- c,4 tn OD -4 .- U Ct(
0w H*00
kDI n
m0*m C -4C e _
S+
r- in~ m ;t N' Ln M- r4 a'r
(nq Nm k('IefOD(~ r-4 0
NO .
U) 4'-HD~~! 0
4.ca * a a * . . . . . .
0)N
4 ~ m-inNoo'
0 -- nL L ,% n
U) ?-I C) 0 OD *-
000 0* r-4*
H 4.) aLA
(1) 0H - - - - -
C) HC.1 0 0a00* 0 4aH
01 n r-I
Nnrr r-r-r- '-'
0k OD0 D0 4c -) ri .- 4
00
a 9-C0o )00 00 0 m4
(9-HP 0000 0000 0HO H N
w~~5 ooor -r -r 0
(U r- r)r- ) 0-
0. >4 ) ) U)m
E.-4 ' ) N U HC +f~
4)U
tJU2 HH(J ('173

these factors are determined in turn by certain parameters
which can be measured and controlled. The properties of
the seal are thus determined by a three-link chain of
physio-chemical processes. The first link in the chain
consists of the independent variables or seal parameters.
These are the physical or chemical factors that can be
measured and controlled and include: the composition and
microstructure of the ceramic body; the composition, micro-
structure and amount of metallizing; the metallizing
sintering conditions including temperature, time and
atmosphere; the amount, composition and sintering atmos-
phere, sintering time and sintering temperature of the
plating; the braze material, amount, time, temperature and
atmosphere during brazing; and the type, thickness and
microstructure of the metal member. The second link
consists of the dependent variables whose values cannot be
directly measured or controlled during processing, so they
must be manipulated indirectly through the independent
variables. The dependent variables include; the amount,
composition, viscosity, and interfacial energies of the
ceramic melt phase; the ceramic crystallite size; the
porosity, or degree of sintering of the metallizing; the
interfacial energies, thickness, and porosity of the plating;
the thickness, composition (alloy), and the interfacial
energies of the braze-metal member; and the thermal expan-
sion of all phases. The third link consists of the actual
sealing mechanisms which are a combination of phenomena
resulting from the interaction of the physio-chemical
conditions of the second link. The sealing mechanisms
include: the migration of the ceramic melt phase into the
pores of the metallizing followed by super-cooling to form
a continuous bridging glass phase adbering to the crystal
phases of both ceramic and metallizing; adherence and
limited alloying of the plating, braze, and metal member;
slight yielding and elastic deformation of some layers to
accommodate differences in their thermal expansion; and
possibly, partial devitrification of the glass phase for
stress-keying and crack-termination.
In this experimental program, the independent

variables, (the first link) were directly controlled,
measured, and varied, while the effectiveness of the
sealing mechanisms (the third link) was measured by means
of several physical tests. However, there was no direct
control or measurement of the dependent variables (the
second link) during processing. For this reason it is
difficult to give simple explanations of the data. Seal
data are presented in a manner which tends to associate them
174
with the independent variables (metallizing temperature,
ceramic body types, etc.) although it is not proper to
claim direct cause-and-effect relationships between the
two, because the independent variables must act on and
through the dependent variables in the second link in order
to produce the seal. It is the purpose of supplementary
experiments to show the relationships between the independent
variables and dependent variables, then the relationships
between the dependent variables and the sealing mechanisms
will become more apparent. Therefore, it must be remembered
that the associations discussed here are only indirect and
are a result of the complex interactions of the numerous
dependent variables.
The data on metallizing scratch hardness and

metallizing reuistance in Table 4-37 and Table 4-38
describe in a qualitative manner tfe effectiveness of a
portion of the sealing mechanisms. These mechanisms
include the sintering of the metallizing layer, migration
of the ceramic melt phase into the metallizing, and
adherence of the metallizing to the ceramic. The hardness
data give an indication of melt migration and adhesion,
while the resistance data indicate the degree of sintering
and consolidation in the metallizing layer. Several
trends are apparent in the tables. The degree of consoli-
dation, melt migration and adherence generally increased
at high metallizing temperatures and were greater on
ceramics which had more melt phase or a more fluid melt
phase. The ceramic body firing temperature is known to
affect at least three of its properties: crystallite
size, amount of melt phase and composition of melt phase.
These factors influence such dependent variables as the
melt viscosity, interfacial energies, and capillary forces,
which in turn control the sealing mechanisms.
The major difference between 941 and 942 bodies

was that metallizing on 941 was consolidated at lower
temperatures than on 942. This difference is attributed
to the difference in the viscosity of the melt phase of
each body, 941 melts being less viscous than 942 melts as
a result of the 1:1 Si0 2 /CaO ratio in the former which
produces a more fluid melt. There was very little differ-
ence in the hardness of metallizing on 942 and 992 ceramics.
The major difference between these ceramics was the amount
of melt phase. The 942 bodies had about five times as
much melt as 992. A comparison of tlese two groups of data
seems to indicate that the viscosity of the melt phase and
the ceramic crystallite size were more important factors
than the amount of melt. The average crystallite size and
175
melt phase composition of bodies 942 (1750) and 992 (1750)
were very similar, while the amount of melt in each was
quite different, but the metallizing hardness on both bodies
was nearly equal at all metallizing temperatures. This
indicates that once a minimum amount of melt phase has
migrated, there is no further benefit from additional melt.
The results on Lucalox* and sapphire were interesting
because both reprerent materials which have no melt phase
at metallizing temperature, but differ in that Lucalox*
is a polycrystalline material, (99.8Y A1 2 0 3 ) while sapphire
is a single crystal of AI 2 0 . There was moderate adherence
to sapphire. The bonding may have resulted from a surface
area effect or weak chemical bonding. The results on
sapphire showed little or no evidence of chemical bonding.
On the other hand, there was no real correlation between
the crystal size of the Lucalox* and the hardness of its
metallizing to show the effect of surface area. Inter-
crystalline spinel (MgO.A12 03 ) in the Lucalox* may have
provided the adhesive link.
The data on seal vacuum quality, tensile

strength, and peel strength in Tables 4-39, 4-40 and 4-41
respectively describe the combined effectiveness of all
the sealing mechanisms. The tensile test data for the 941,
942 and 992 special alumina bodies are plotted against
metallizing temperature in Figs. 4.61, 4.62 and 4.63.
Peel strength data for the same ceramics (brazed with Cu
and CuAg) also plotted against metallizing temperature
appear in Figs. 4.64 through 4.69.
Plots of seal tensile strength on body 941 in
Figure 4.61 rise rapidly from 1300 0 C to 1425 0 C, peak out
at 1600 0 C, then drop or level out at 1800 0 C. The most
prominent factor in these results was probably the fluidity
of the melt phase in the respective ceramics at the various
metallizing temperatures. The melt phase of the 941 bodies
was consistently more fluid than either 942 or 992.
Consequently, strong seals, which are produced in part by melt
phase migration, were developed at lower temperatures on
941 than on 942 or 992 bodies (Figs. 4.52 and 4.63). In
these latter two cases, the strength continued to increase
as the metallizing temperature was increased to 1800 0 C.
On all three bodies there was a greater spread of data at
1800 0 C than at lower temperatures. This may have resulted
from the fact that the metallizing temperature was higher
176
14 - Tensile
Strength
Metallizing Temperature
13 j
Body 9 41
121
I 1700
0
1010
to
180
0. 6
of /1' /so 0\10
160 9410550)- -
I.- 94101600)
20 941(1650)-------
941 (1700)--
1300 45 1600 1S00
Metallizing Temperature,0C
Figure 4.61: Plots of tensile strength of ceramic-to-metal seals on 941 alumina bodies orginally fired
at 15,50"C, 1600"C, 1650"C, and 1700"C versus metallizing temperature.
177 -
14 Tensile Strength 0,
13 Metallizing Temperature / //
Body 942
12 -
II// + ?0
10700
10 " / /
I
b. 6342 16600
940650
55
4 10
942 1600)
3 942 (1650)
942 (1700)----
2 1?09420(750)
01300 1425 4100 1900
Metallizing Tempetrwue ,*C
Figure 4.62: Plots of tensile strength of ceramic-to-metal seals on 942 alumina bodies originally fired
at 1600"C, 1650-C, 1700°C, and 1750*C versus metallizing temperature.
- 17 -
4 Tensile Strength
i vs
13 Metallizing Temperature
Body 992
12
Io
* 9
x
/-
I IS
-3 ,*..-- 992 (100)--

992 0750)- --
- WO
-. 992(100)
0 U
// 1370 14U 60 O
Figure 4.63: Plot& of tensile strength of ceramic-to-metal seals on 992 alumina bodies originally fired
at 17000 C, 1750*C, and I 800C-19 versus metallizing temperature.
15
Peel Strength
vI
14 Metallizing Temperature /
941 - Cu braze
13 /
/
/
12 //
/
/
II
/
10.
9 /\og
/1/5
7o
-~
2
4 1700 9 4 1 01
5 0 01- - -
, 0/ 941 (1650)---
,3// 941 10 10)
0
1044 1600 1(60-
Mlq Tomptoleoe:
Figure 4.64: Plots of peel strength of ceramic-to-motal seals brazed with copper on 941 alumina bodio
originally fired at 1550°C, 1600°C, 1650'C, ard 1700C versus metallizing temperature
Pel Strength
Ve
i4- Metallizing Temperatwe
942- Cu braze
13
12
I'
10-
I.
A
'a 6
-9421 ).-----
942(1m5)--
942 07001
000,942 0750)--
0 30 1425 00
1111
meolwa 719m,11 96C
Figure 4.65: Plots of peel strength of ceramlct4o.metal seals brazed wih copper on 942 alumina bodies
originally fired at 1600C. 1650C, 17W*C and 1750*C versus metalliing telnperature.
It -
15 -POOl Strangth
vs
14 - Metalkizng Temperature
*92-Cu brase
13
12
10
4
99201700)-
3 992(1750) -
2(1800)-.-.
2
-- - - - - - - - -
1300 14n5 160 wo00
Figure 4,66:, Plots of peel strength of cernmk-bofrtaI $#ll brazed with Copper an 992 alumina bodies~
originally fired at 17W00C. 1750*C. end I SO-t versus meotallizing temperature.
-162 -
iS Pee Slrenqth
vs
( 941 -Cu A,
braze
13
//00
94050 ------
I"
II
01300 4Gm10
M*k# TOMPr.ks,*C
FiPgure 4.67: Plots of peel strength of ceromic-to-otal seals brazed with :uAg on 941 alumr bodies
originally fired at 11550t, 16001C. 1650'C, and 17000 C versu rntallizing tL orature.
51 Peel Strength
V9i
14-Mtli'n Metaizing Temperature
942-Cu Ag braze
/
13
I
12 i
I
I
I
il /
II I
/
Io I
10 I
I
I
I
I
52 I
I
4 -Im // I
/ 42 ( 6 0
15 I I
3 - ,600
S-9420700 942 (1650)---
2 - ROO-IM 9420(750)
0I
0942 12600)
Metolling Temperoahe,0C
Figure 4.68. Plots of peel strength of ceramic-to-metal seals brazed with CuAg on 942 alumina bodies
originally fired at 16001C, 1650*C, 1700 0 C, and 17501C versus metallizing temperature.
- 184-
P". Skrngrh
I4 WS
MOt~iN" Twmperoknr
992-Cu Ag braze
13
12
S1
9;
I7
41 992700) -
01300 1425 1600 10
I Metdafir TemperatweeC
IFigure 4.69: Plots of peel strength of ceramic-to-metal seals brazed with CuAg on 992 -alumina bodies
originally fired at 1700 0 C, 1750 0C, 1800*C versus metallizing temperature.
WI
than most of the body firing temperatures and caused major
changes in the ceramic microstructure as the bodies
attempted to reach equilibrium at the higher temperature.
Some grain growth occurs, especially in 941 bodies, and
more alumina is dissolved into the melt phase. During the
subsequent copper braze cycle to 1100 0 C, some of that
alumina can recrystallize and either strengthen or weaken
the seal depending on the amount, grain size, and distri-
bution of the crystal phase. The two 941 bodies whose
seal tensile strength dropped the most at 1800 0 C (16500 C
and 1700 0 C firing temperatures, Fig. 4.61) also produced
some leaky seals at 1800 0 C (Table 4-39). This is further
evidence of excessive devitrification of the glassy phase
which probably occurred during the 1100 0 C braze cycle.
Peel strengths of copper braze seals on
ceramics 941, 942 and 992 are plotted against metallizing
temperatures in Figs. 4.64, 4.65 and 4.66 respectively.
In Fig. 4.64 the plots of peel strength of bodies 941
(1650) and 941 (1700) peak at 1425 0 C, much lower than the
other two plots. This could again be related to the amount
and mobility of the melt phase of the bodies. The tempera-
ture, 1425 0 C, may be near a point of critical viscosity
for the melt,which could make secondary factors such as
ceramic crystal size and interfacial energies more prominent
in affecting the mobility of the melt. The wide spread of
peel strength values at the 1800 0 C metallizing temperature
may again reflect the phenomenon of glass phase devitrifica-
tion,which was discussed previously in connection with the
tensile strength data.
Peel strength values for seals on bodies 942
and 992, Figs. 4.65 and 4.66, were generally much lower
than on 941. The properties of the melt phase may also
have been significant here. The 2:1 SiO 2 /CaO melts are
more viscous than 1:1 SiO 2 /CaO melts and thus have less
mobility and less tendency to devitrify. These factors may
have caused the difference in strength. Consistently low
values on 992 could be caused by insufficient melt to
produce a good bond with the metallizing, although this
did not appear to be the case in the tensile tests.
Peel strengths of 810 0 C copper-silver braze
seals on ceramics 941, 942 and 992 are plotted against
metallizing temperatures in Figs. 4.67, 4.68 and 4.69
respectively. These peel strengths are in the aggregate,
higher and more closely related to the metallizing
temperature than the copper brazed peel strength values.
The major difference in sample preparation between the two
186
mnn.9 n nnn mm m num ~u unu ~ m u m u uu m m. m •m

groups of samples was the brazing temperature: 1100 0 C for
Cu braze versus 8100C for CuAg braze. This made a great
deal of difference in the amount of stress introduced into
the seal by differential thermal expansion of the alumina
ceramic and the Kovar peel tab. The expansion of alumina and
Kovar match fairly well to around 800 0 C, but they diverge
rapidly above that temperature. As a result of this mismatch,
considerable stress is produced in the copper brazed seals
as they cool from the 1100 0 C braze temperature. This accounts
partially for the generally irregular and lower peel strength
of copper brazed peel assemblies. Another important
dependent variable which is affected by the braze tempera-
ture is the viscosity of the glassy phase and its ability
to devitrifv. The 300 0 C difference in braze temperatures
makes a great difference (orders of magnitude) in viscosity
in the range which vitally affects crystallization.
There is less of a difference in the peel

strength values for 941 and 942 bodies with CuAg brazes
than there is for the same bodies with Cu brazes. This
probably results from the lower seal stress conditions due
to expansion mismatch and the diminished tendency for glassy
phase devitrification. The greater tendency for the 941
body to devitrify separates it from 942 only when the
temperature is high enough for that to occur. The CuAg
braze temperature is not high enough to produce this effect.
In fact, an annealing process takes place which tends to
reduce the stress level in the glass. Alumino-silicate
glasses are generally annealed in the 750 °C-8500 C region.
The data may alternately be plotted as a function

of ceramic firing temperature at a given metallizing
temperature or plotted on a three-variable,triangular
diagram. Interesting conclusions may be reached from this
approach. It appears that strength is not related to crystal
size of the ceramic. However, in order to study this effect,
a special study was made of this variable. This study is
reported in a later section.
4.5.2.3 Microstructure Analysis
Many of the tested seals were examined metallurgi-

cally to help explain the results. In each seal examined,
the visual appearance of each component layer, ceramic,
metallizing, plating, braze, and metal member, was examined
in order to establish a relationship between its properties
and the seal strength. The metallizing layer was found to
vary the most in quality and to have the most dominant in-
fluence on seal characteristics of any of the seal components.
187
Pu1ne n n nnmw n 1 n -n - u .. ...

The relationship between each seal component and its effect
on seal strength and vacuum tightness is discussed below.
4.5.2.3.1 The Ceramic
Three general aspects of the ceramic portion of

the seal structure were considered. They were (1) the
structural changes which occurred in the ceramic during the
metallizing sinter fire step, (2) the quality of the
ceramic surface, and (3) the action of the glassy melt
phase of the ceramic.
The most easily recognized change which occurred

in the ceramics when they were exposed to metallizing
temperatures was an increase in average crystallite size.
Two extreme cases are shown in Figure 4.70. The top photo-
micrograph shows a seal to sgecial alumina body 941 (1550)
which was metallized at 1300 C and which has an average
crystallite size of about 6 microns, changed little if any
from what it was before it was metallized. The lower
photomicrograph shows a seal to the same body, but the
metallizing was sintered at 1800 0 C. The average crystallite
size in the second sample is about 12 microns, a considerable
increase from the initial size.
This sort of change was u'ost pronounced in the
941 bodies because the melt phase of these bodies has a
lower viscosity at elevated temperatures tihun either of the
other aluminas and therefore permitted more diffusion of
aluminum and oxygen ions through the melt to support crystal
growth. Little or no growth occurs unless the ceramic is
metallized at a temperature above the initial firing tempera-
ture, because the bodies had approached equilibrium at the
firing temperature, and a higher temperature would be
required to cause a measureable shift from those conditions.
The marked grain growth in 941 bodies also

indicates that a similar change in the A1 2 03 content of the
melt phase is taking place. This change must affect the
melt viscosity and therefore its mobility. The viscosity
would decrease with increasing temperature, but could either
increase or decrease depending on whether the alumina
content of the melt was initially below or above about 45
weight percent. Melt mobility appeared to be a primary
factor determining the extent or ease of melt intrusion
into the porous molybdenum metallizing layer. This was
most evident in seals to 941 bodies which developed good
strength at lower metallizing temperatures compared to the
other bodies studied.
188
~- Cu Braze
Metallizing
~- Alumina
Ceramic
Cu Braze
Metal lizing
Alumina
Ceramic
Figure 4.70: Photomicrographs of ceramic-to-metal seals on alumina body 941 (1550) metallized with
P.4M at 13001C (top) and 1800 C (bottom) showing the structural chan~es which occur
when the ceramic is metallized at very high temperatures. (400X)
-189 --
The average crystal size of 942 and 992 bodies
changed very little when metallized at 1800 0 C. This implied
that the composition of the melt phase was not greatly
altered by the metallizing treatment either. However, there
is evidence that the amount of melt did vary, increasing by
dissolving some crystal phase at higher temperatures and
then decreasing by again precipitating aluminum oxide.
Figure 4.71 (top) is a photomicrograph of a seal
on alumina ceramic 942 (1650). The P-4M metallizing was
sintered on at 1800 0 C. Evidence of growth caused by partial
solution and recrystallization of alumina crystals appears
at the interface between the molybdenum metallizing and the
ceramic. The alumina crystal phase has grown to match the
contours of the under-surface of the molybdenum mass. This
resulted in an interlocking mechanical joint with very high
tensile strength of 14,000 psi.
Figure 4.71 (bottom) shows a seal on alumina
ceramic 992 (1800), also metallized at 1800 0 C. Here again
the alumina and molybdenum have come into intimate contact
forming a strong (11,300 psi), interlocking interface. The
only noticeable difference between the interfaces on the
942 and 992 aluminas is the clarity of the dark line between
the molybdenum and the alumina. This line may be the
remaining portion of glassy phase. It is easily discernible
in the seal on 942 (top), but is absent or unresolved in
the 992 seal (bottom). In the seal in the bottom photo,
where molybdenum and alumina appear to be in direct contact,
it is possible that an A12 03-Mo chemical bond has formed.
To further investigate the possibility of the

formation of a chemical bond between alumina and molybdenum,
seals were made with P-4M metallizing to Lucalox* of three
different nominal grain sizes, 5, 20 and 40 micron. The
tests were performed with both polished and unpolished
Lucalox* surfaces to study the effect of surface finish on
the interlocking grain phenomenon and its contribution to
seal strength. Lucalox* has no glassy phase as found in
941, 942 and 992 alumina ceramics, so the mechanism of bonding
between the metallizing and ceramic must be a chemical bond,
crystal-to-crystal interlocking, adhesion resulting from a
reduction in surface energies, or some combination of the
three.
None of the seals developed strength above
4000 psi, which indicated that whatever mechanism of bonding

190
MW n mum m mn nnr nu n n lum n nnU N U nuunn n n nn un n nn n mmm nm m nn n sunn mn nmumn~nUNN io ~ mn m unn mam n
Cu Braze
it Ma Metallizing
Alumina
Ceramic
Cu Br aze
~*- Aluonina
Ceramc
Figure 4.71: Photomnicrographs of ceramic-to-metal seals on alumina body 942 (1650 (top) and 992
(1800) (bottom) Vth metalized with P-4M at 1800 C. Interlocking grain structure may
be meen at the Interface between the molybdenum metallizing and the alumina surface.
I (400X)
NON" OW
TABLE 4-42
Tensile Strength of Seals to Polished and Unpolished

Lucalox of Three Grain Sizes
Lucalox* Metallizing Strength, psi

Grain Size Temperature Unpolished Polished
5 micron 1600°C 2038 n.d.
5 1800 2138 2200
20 1600 3725 3338
20 1800 2225 1950
40 1600 3388 3350
40 1800 3138 1563
was in effect was either weak by its nature or it was effective

on only a fraction of the total interface. Results obtained
with the 5 micron Lucalox* were about the same value regard-
less of metallizing temperature or surface condition.
Adhesion to 20 micron and 40 micron Lucalox* bodies seemed
to have a slight relationship to these two variables with
the unpolished surface condition and 1600 0C metallizing
temperature having a slight tendency to produce better
adhesion.
The lower metallizing temperature and relatively
irregular surfaces were conditions which would be more
likely to favor a mechanical type of bond or surface energy
adhesion. If a true chemical type of bonding had predomi-
nated, then its strength should have increased with the
higher metallizing temperature which would promote or
accelerate the chemical reaction. However, this was not
the case.
Further adhesion tests were made with P-4M

sintered on polished sapphire discs at 1600 0 C and 18000 C.
At both temperatures only minimal adhesion was achieved.
The molybdenum particles sintered well to each other, forming
a continuous metal sheet, but this sheet curled away from
the sapphire surface and could be removed with little effort.
192
It was apparent that the stresses which develop
in the layer of metallizing as a result of densification
during sintering were great enough to affect the adhesion
which developed between it and the ceramic. These stresses
increase as higher sintering temperatures are used and
would go unchecked on a smooth surface such as polished
sapphire which presents a natural shear plane. On irregular
surfaces with many crystallites jutting into the metallizing
layer, the shearing tendency would be greatly diminished.
Because of this relationship between the stress
in the metallizing layer and the surface condition of the
ceramic, the results always indicated a mechanical type of
adhesion.
The conditions which should have given isolated

evidence of chemical bonding also contributed to the stress
condition, thereby weakening or destroying any chemical bond
which might have formed. Therefore the results of this
particular experiment cannot be considered conclusive in
confirming the mechanical adhesion theory or in repudiating
the chemical bonding theory. In view of this conclusion,
further experiments were programmed to investigate the
chemical bonding mechanism. These are discussed later in
section 4.5.3.
Most of the data still indicates that the glassy
or melt phase of the ceramic is the predominant factor in
the formation of strong and reliable seals. The flowing
action of the melt at the metallizing temperature was
relatively easy to detect in the model metallizing system
used in this study. The P-4M metallizing paint was
completely metallic, containiig no additions which could
form or contribute to a glassy melt. Therefore, any glassy
material which was present in the metallizing after forming
the seal could only come from the ceramic.
Two contrasting conditions are shown in Figure

4.72. Photomicrographs of body 942 (1700) with P-4K
mLtallizing sintered at 1425 0 C and 16000 C show that the
melt phase of the ceramic flowed Anto the molybdenum at
16000 C, but did not do so at 1425 C. The pore spaces of
the metallizing sintered at 1425 0 C are filled with copper
braze material. The pc,te of the 1600°C sintered metalliz-
ing are filled with glassy material which migrated from
the ceramic. The glass thus forms a strong mechanical
link between the ceramic and the porous sintered layer of
metal particles. The tensile strength of the seal metallized
at 1600 0 C was stronger by a factor of two. This difference
193
...... M =wimp
Cu Braze
" '4 it MO Metallizing

with Cu Braze
Intrusion
Alumina
Ceramic
,-- Cu Braze
Mo Metallizing
I-
with Glossy
Phan Intrusion
Akonna
Ceramic
Figure 4.72: Photmirographs of ceramic-to otal wals on alumina body 942 (1700) mde with P-M
metallizing intered at 1425 C top) and 1600"C (bottom). The gomy phase migrated
into the metall;zing when sintered at 1600 C. but not at 1425,C.(400X)
194-
•m • u -• , • - • • w• - - ___ -• w
i m m • •
was attributed co the flow of the glaasy phase material into
the metallizing to connect the metal particles to the
ceramic.
Similar comparisons have been made with numerous
other seals to indicate the predominant influer7- of glassy
melt phase migration on seal strength. This irnt'uence
determined the shape of the curves in Figures '.71, 4.62
and 4.63.
4.5.2.3.2 The Metallizing
Several of the variables of the ceramic must also
be considered metallizing varia 1les because the two are so
closely interrelated. Thus the melt phase may be considered
as part of either, especially in common production metallizing
paints which contain glass-forming oxide additions.
Mechanical interactions produced by surface finish and
interlocking growth of some crystal phases are also factors
which may be associated with both the ceramic and the
metallizing layers.
One process which occurs exclusively within the
metallizing layer is the sintering or consolidation of thdt
layer. In the model systems used in this study, the metalliz-
ing layer was consolidated by one or both of two processes.
The first process was sintering of the fine molybdenum
particles into a porous sponge. On ceramics which exuded no
glassy phase, the metallizing remained as a porous layer
whose density was primarily determined by the sintering
temperature. The second process of consolidating the
metallizing was the action of the melt phase of certain
ceramics. A thin film of melt apparently formed on the
surface of the ceramic, wet the adjacent molybdenum particles,
and spread through the metal layer. As a consequence of
this wetting action, the metal particles were drawn into and
consolidated by the thin layer of melt.
4M S, e amount of metal-to-metal rintering also
%ed occurred when particles came into direct contact under
favorable conditions, but this process was retarded in the
presence of the melt phase. This retardation resulted from
the wetting action of the melt which tended to isolate the
j molybdenum particles.
The photomicrographs in Figure 4.73 illustrate
the two situations. The top photo shows P-4M metallizing
, sintered together to form large separated grains. The molyb-
denum was mildly etched to bring out the fine crystal
195
- **uP*- •
TV W - - -. - n
.~- Cu Broze
A#- Mo Metallizing
:it CuBrze
Ceramic
No Cu Braze
wi~h Glassy
Phase.Intrusion
Alumina
Ceromic
Figure 4.73: Photomicrographs of ceramic-to-metal seals on Lucalox* (top) and alumina ceramic 941
160C, (bottom) both metallized with P-4M at 1800"C. The molybdenum metallizing was
etched for 4 sec, with a solution of potassium ferricyanidle and sodium hydroxide. (800X)
-196-
structure of the grains which resulted from sintering
together many smaller particles. The bottom photo shows
P-4M metallizing sintered rn alumina ceramic 941 (1600)
at 18009C. This ceramic has a fluid melt phase at 1800 0 C
which flows readily and can be seen (dark gray) dispersed
throughout the metallizing. The metallizing has sintered
to some extent, but did not form grains as large as it did
in the absence of glassy melt. Also, many smaller particles
are isolated and floating about freely in the glassy layer.
This sample also received a mild etch to bring out the
molybdenum crystallite structure.- The etchant was composed
of 30 grams of potassium ferr-icyanide and 10 grams of sodium
hydroxide dissolved in 100 ml Cf water.
The tcp photos in Figures 4.70 and 4.72 are
examples of metallizing sintered to a lesser degree than is
desirable. The molybdenum particles are well dispersed arid
of small size, indicating that little or no sintering has
occurred. In both cases, the metallizing was saturated by
copper in the subsequent bTaze operation due to its poorous,
spongy nature.
4.5.2.3.3 The Plating
ce.a..c Nickel was electroplated on the metallized
ceramics for CuAg (BT) brazes, while copper was plated for
-copper brazes. The purpose of the plating is to form a
wetting layer to assure intimate contact between the braze
and metallizing in Order to form a strong and vacuum tight
seal.
Deposition of metal ions from the plating bath
requires two conditions. First, the metallizing must be
conductive so that current may be carried through it from
the contacts to the plating surface., Second, the surface
must be such that electrons may be exchanged between the
metallized surface and the copper or nickel cations in the
bath. The fact that metallized ceramics were plated- success-
fully was sufficient evidence to indicate that these two
conditions existed, but they were not necessarily idealized.
The conductivity of the metailizing varied
somewhat, Table 4-38, but was adequate for plating in all
cases except when the metallizing actually flaked off.
The quality of the surface at which electron
exchange occurred was quite variable. Some metallized
ceramics which had little or no glassy phase in the metalliz-
ing had many tiny patches of ceramic exposed through the
197
metallizing. On others with generous amounts of glass
present, the molybdenum particles at the surface were
exposed in the mannei of many tiny islands in a sea of glass.
In both cases, the surfaces were plated with a uniform
thickness of metal which generally followed the original
surface contour.
Although electron exchange could not occur at

every point on the surface because of tne non-conductive
ceramic and glass areas, there obviously were sufficient
sites to provide the necessary electrons to build up the
plated layer. It was not determined whether or not the
irregular current density distribution caused local concen-
trations of plating at electron exchange sites, because the
subsequent plating sinter-firing produced a more even
eistribution and thus destroyed any preferred distribution.
The uniform thickness may have resulted from the sintering
step, or the plating was initially deposited in that manner.
Electrons could have been conducted laterally through the
plating to overcome the effects of only spot sources of
electrons.
Plating was employed to assure reliable brazes,
but it did not appear to be the most critical step in form-
ing strong vacuum tight seals. No seals were observed in
which failure could be traced to the plating. Mechanical
failures always occurred in the metallizing or the ceramic --
the brittle zones. However, the plating-braze layer is
subjected to considerable stress and must adhere well to
the metallizing and be able to yield to some extent in
order to distribute more evenly the stress on the metallizing
without failing itself.
4.5.2.3.4 The Braze and Metal Member

Both CuAg (BT) and copper brazes were examined
in this study. The most important braze variables as far
as the seal is concerned are the melting temperature, braze
hold time, mechanical properties, and dimensions.
The braze melting temperature determines the

temperature from which stresses accumulate when the ceramic
and metal member have different coefficients of thermal
expansion. This is primarily a design problem and becomes
more critical as dimensions increase.
A second factor arising from the brazing

temperature is the effect this heat treatment has on other
parts of the seal such as the glassy phase of the metallizing
198
and ceramic. Depending on the composition and phase
'elationships or stability of the glass, this heat treatment
may anneal this material or allow it to devitrify to some
degree. This is an important factor because it affects
stress levels and distribution in the brittle seal zone
where failure usually occurs. However, its direct study
;ias been difficult because of the limited ability to control
and measure all the variables contributing to this effect.
Nevertheless, limited experiments were conducted to isolate
this effect and they are discussed later.
The braze hold time or time above the liquidus

temperature of the braze metal determines the extent of
adhesion and alloying between the plating or metal member
and the braze material. Hold time should be just long
enough to allow the braze material to completely wet the
surfaces to be joined and expel any entrapped gas. This
is mainly a function of the mass and heat capacity of the
assembly being brazed. More lengthy hold times caused
various degrees of reaction with the plating, metallizing
and metal member on the samples studied.
The mechanical properties of the braze metal

such as tensile strength and hardness determine the response
of the seal to the stresses which it encounters. The metal
must be strong enough to resist failure of the seal in that
area, yet it must yield under load to equalize the stress
distribution over the seal area. Both brazes used in the
study were satisfactory in this respect and were never
indicated as a source of seal failure. (The "bad brazes"
which were reported were all caused by jigging errors or
overextended braze hold times, so were considered problems
of execution, not design.)
It would be expected that in the case of the
Kovar metal member employed in the peel test that the lower
Cu Ag braze temperature would universally result in higher
seal strengths as the stress due ,e metal member would
be lower. This was not found to be Lhe case, however. The
reason for this result must lie in the tendency of the glassy
phase of the metallizing to devitrify at 1100 0 C. Sometimes
this will increase the strength of the seal, sometimes it
will decrease it, depending on the size and amount of
crystals which are formed.
The effect of a variable stress induced by the

metal member is shown in a later section.
199
4.5.3 Supplementary Studies j
4.5.3.1 Bonding Mechanisms
The mechanisms of ceramic-to-metal seal adherence

have been discussed in previous sections and the electrical
results still to be reported shed further light on the nature
of the bond which broadly is characterized as either chemi-
cal, semi-conducting or mechanical.
Experiments reported in this section concern

themselves with the strength of a metallizing bond on
sapphire formed as a consequence of:
a) Evaporating molybdenum onto the sapphire surface
b) Firing a MoO 3 paint,P-2M,onto the sapphire.
4.5.3.1.1 Evaporated Molybdenum*
Several seals in the form of tab peel assemblies

were made using sapphire bars one-tenth inch square and one
inch long. Evaporated metallizing was applied to one of the
polished 0.1" x 1.0" sides. The metallizing consisted of a
flash of titanium, a 0.8 micron layer of molybdenum, and a
flash of copper which was followed by a 0.5 mil copper
plate. A 15 mil cupro-nickel tab was brazed on with either
copper braze (1120 0 C for 3 minutes) or Cu-Ag (8600 C for
2 min.).*
The metal tab was removed by grasping one end of

it and pulling in a direction perpendicular to the plane
of the metal-sapphire interface. This was accomplished by
holding the substrate in the lower jaws of an Instron
testing machine, the metal tab in the upper jaws, and
pulling apart at the rate of one inch per minute cross-
head speed. The failure strength was recorded continuously
as in the drum peel test described earlier.
The results are presented in Table 4-43 with some
additional tab peel data using beryllia Body BA as the sub-
strate and both evaporated and standard P-1 metallizing.
The strength of most seals was low, but one evaporated seal
on sapphire and all four standard metallized seals were
*See also Table 4-40.
200
TABLE 4-43
Tab Peel Results
Pull Strength, lbs

Substrate Metallizing Braze Max. Min.
Sapphire Evap Mo Cu 0.6 0.4
11Evap Mo Cu 0.9 0.6
ItEvap Mo BT 0.6 0.4
Evap Mo BT b.5 4.5
BeO* Evap Mo Cu 1.25 1.0

so Evap Mo Cu 1.8 0.8
toP-1 Cu 3.6 1.5
toP-i Cu 5.5 1.0
ofEvap Mo BT 0.3 0
ItEvap Mo BT 0 0
ItP-1 BT 2.8 1.8
P-i BT 6.5 3.0
J *Body BA: 99.5% BeO
201
w loq -u--- - -%- 111 4__________I-___

moderately strong. The results at least demonstrated that a
strong evaporated seal can be made to sapphire, although
not on a reliable basis as yet. Improving the reliability
of this seal forming method is a worthy goal of a future
study. *
4.5.3.1.2 P-2M Metallizing on Sapphire

P-2M metallizing paint was prepared from pure
MoO 3 and was ball milled in a molybdenum-lined ball mill
with molybdenum grinding media. It was the same as P-4M
except that P-2M use2 MoO 3 while P-4M used powdered molyb-
denum metal.
P-2M paint was applied to sapphire discs of
sufficient size (0.625 in. 0. D.) to be brazed between the
tensile testing faces of two CLM-15 ceramic test pieces.
The paint was sintered at 16000 C, resulting in a soft,
powdery layer which was not well sintered within itself,
but which did adhere sufficiently to the sapphire so that
it was very difficult to scrape all metal off the surface
of a test piece.
The metallized sapphire discs were then copper
plated and copper brazed between two CLM.-15 test pieces as
shown in Figure 4.60. Half the samples were assembled with
15 mil, 70/30 Cu/Ni washers and half without washers.
Specimens without washers averaged around 1900 psi tensile
strength; those with washers were near 2800 psi. About
half the seals were vacuum tight. (P-1 metallizing normally
gives strengths in the range of 5000-10,OCO psi on sapphire.)
Post-test metallurgical examinations of polished
seal cross-sections (see Figure 4.73a x 800) revealed that
the copper from the plating and braze permeated the voids
of the very porous metallizing and readily penetrated to the
sapphire surface. Molybdenum particles were in intimate
contact and apparently adhered well to the sapphire surface.
*Future seals should employ thicker molybdenum metallizings

in order to resist attack by the cupro-nickel braze phase:
and for a Cu-Ag braze, also a 0.5 rail oarrier nickel plate
(see Figure 6.29 (top) ); in order to prevent penetration of
0- Cu-Ag braze along the copper grain boundaries and sub-
sequent attack on the molybdenum. See 6th Quarterly Report,
Figs. 4.14, 4.15, 4.11 and section 4.5.3, also this report,
Figs. 6.33 and 8.74 (bottom). Brazing with Cu- ,g at 8000 C
rather than 860 C should also be tried.
202
This provided further evidence of a non-glassy phase
bonding mechanism. It is possible that such bonding could
occur by bringing the reduced metal powder into intimate
contact with the clean sapphire surface, but a very likely
contributor to this bonding mechanism is the additional
molybdenum braze or "blush" which is deposited on the sapphire
during the MoO 3 reduction process. The adherence mechanism
of the molybdenum braze is essenially the same as that of
evaporated molybdenum metallizing.
4.5.3.2 Crystal Size Factor
In order to isolate the effect of alumina grain
size, the special ceramics 941 (1650) and 942 (1700) were
re-fired for up to 50 hours at 1600°C; i.e., for 2, 5, 10,
20 and 50 hours respectively. They were then metallized with
P-4M paint at 1600 0 C and brazed with copper to Kovar tabs.
No significant correlation with crystal size
could be determined using the small sample size of two.
(See Section 4.5.2.1.6.)
4.5.3.3 Glassy Phase Devitrification
Several anomalously high results in the strength
tests conducted under Sections 4.5.2.1.6 and 4.5.2.1.7 were
explained on the basis of metallizing glassy phase devitri-
fication phenomena. Additional data was therefore obtained
which helped to define the conditions under which the glassy
phase of the ceramic-metallizing interface tend to precipi-
tate the secondary crystalline phases, anorthite and
gehlenite. An attempt was then made to relate this
precipitation data to seal strength.
Two series of experiments were conducted. The
first series consisted of an examination of the seals made
earlier and reported in Figures 4.64 to 4.69. The second
series was designed to eliminate some of the extraneous
variables present in the first test series.
4.5.3.3.1 First Series
Eleven drum peel samples of both 941 and 942

alumina ceramics representing several body firing tempera-
tures, metallixing sintering temperatures, and two braze
cycles were examined after peel testing to determine the
presence of secondary phases. All metallic deposits were
203
, ,-- •-,,, _- -
removed from the surface by means of a 1:1 warm nitric acid
etch until all chemical action ceased. The surface thus
exposed was given an X-ray diffraction analysis.
The results of the analyses are presented in

Table 4-44. The samples are divided into two groups,
weak and strong, according to the results of the drum peel
test. The second, third, and fourth columns indicate the
thermal history of each sample, and the last column lists
the types and relative amounts of secondary crystalline
phases detected on the formerly metallized surface.
All the strong seals had little or no secondary
crystalline phases present, while nearly all the weak seals
had some amount of anorthite and/or gehlenite present in
the glassy interface layer. The presence of secondary
crystall4.ie phases was generally related to the braze cycle
to which the parts were exposed. This association also held
for seal strength. Copper brazes generally resulted in
I
weak seals and a partially devitrified glassy phase. The
opposite was true of copper-silver braze.J
The difference in devitrification tendencies

can be attributed to the difference in braze tempe.'atures,
but to associate seal strength with the absence or presence
of a secondary crystalline phase on the basis of these data
I
would be premature. The data reported in Table 4-!4 may be
misleading because of the presence of other factors which
also affect the seal strength. One such factor is the
introduction of considerable Lhermal expansion mismatch
between the Kovar peel tab and the ceramic when the assembly
is brazed above the BT braze temperature. Thus the copper
brazed seals have much more internal stress than the BT
brazed seals, and this may account for their generally
weaker coidition. This variable is eliminated in the
second series of experiments in which devitrification is
produced in some samples by a heat treatment separate from
the braze cycle. All samples were brazed with BT, so that
the only major variable was the amount of secondary crystal-
line phase present.
4.5.3.3.2 Second Series
Sets of 941 (1650) and 942 (1700) test pieces

were metallized with P-4M metallizing sintered at 16000 C.
One set (Group A) was kept as a standard and other sets were
subjected to wet and dry hydrogen firings to 11000 C to
simulate extended copper braze cycles. Kovar or cupro-
nickel tabs were brazed on with BT, and peel tests were made.
I
204
TABLE 4-44
Secondary Phases Found in Ceramic-Metallizing Interface

by X-Ray Diffraction Analysis
STRONG SEALS
Type and Amount of
Body Body Firing Metallizing Braze & Secondary Phases
.ape Temp. Temp.2 Temp. Present
941 1550 1800 Cu-1100 0 C None detected
941 1550 1800 BT-810 N.D.
942 1650 1800 Gehlenite: VST
942 1650 1800 N.D.
WEAK SEALS
942 1650 1800 Cu-1100 Anorthite: minor
942 1650 1600 of An.:++,Gehl.:VST
942 1700 1800 N.D.

942 1750 1600 An.::MinorGehl.:
minor
941 1550 1800 An.:VSTGehl.:++
941 1600 1800 BT-810 N.D.
941 1600 1800 Cu-1100 Gehl.:VST
Amount present in Decreasing Order:
Most: ++
Minor
Last: V.S.T. very slight trace
205
40111*m I"___________
All remaining metallic molybdenum was then removed by a warm
nitric acid leach, and the formerly metallized surface was
subjected to X-ray diffraction analysis to identify
crystalline phases other than alpha-alumina.
Peel test results on two such runs are presented
in Table 4-45 and the X-ray data are presented in Table 4 -4 b.
Based on the general trends shown by the entire data,
interpretive plots of peel strength versus amount of crystal
precipitation, and both peel strength and crystal growth
versus time are presented in Figures 4.74 - 4.81.
These X-ray diffraction crystallographic results

may be compared with those presented in Zable 4-12. The
results of Table 4-12 show that the 1425 C fire is able to
devitrify the anorthite phase completely (8.7%) in the 942
body, as it is heated to just 750 C below the liquidus
temperature of this composition. The 1600 0 C fire*is above
the eutectic temperature, 1500 0 C (Figure 4.04), and rapid
cooling results in only partial crystallization of the glass
phase as shown by the zero hold time specimen. Subsequent
reheating to 1100 0 C results in significant anorthite
devitrification.
For the 941 body, Table 4-12 shows no gehlenite

precipitation. The 16000 C fire*and/or the presence of
molybdenum surface nucleation sites has induced some gehlenite
crystallization in the zero hold time sample.
In the case of the dry H2 fire on the 941

ceramics, there is a general relationship between seal
strength and glass phase devitrification up to the 20-minute
time period as shown in Figures 4.74 and 4.75.
The 20 minute time corresponds to maximum

devitrification of the glassy phase with the appearance of
an anorthite phase, Figure 4.75. Beyond this time limit,
it is apparent that ionic flow and/or diffusion of glass
from the interior to the surface is taking place, which
resorbs the gehlenite and anorthite crystallites in the
metallizing layer. This in general results in a loss of
seal strength.
It is noted that the strength r':ults aro, in
general, higher with a cupro-nickel tab than with a Kovar
tab. The cupro-nickel tab must induce less stress into the
seal area.
*in the present test series
206
- - - -, -- _ III I ' --r .. . -

3 0A r-4 r~ 4 'D cILAr '
M
0
0
E 4)
u' 0 LAi 0n cD 0 r-4 CA N 00 ND

,a 0 r4 4 0 40 4* . .
04 (N
190
4.) --- 4 U4 U U t4
LA r4 ~
E-4iM
~4 c-
04
414)
.C C
. 04 '.4 Vc4 -P4 '.4 '.4 ~ -4 Vp4
6 E E 6 E
0 P- tn C.) 0 Co
*)4
to 0
20
0 tIt
rIi
4-)0
u ra)
44 ., 0 41 j
4J 0 M
4
>1
".4
.ri
4 >i ~ ~ .J -
H 44 .7
$4 0) 0I
ri~~- -4C H40
r-H H .
HW
'C)j H.
4-'> 4-
0)4
0 W0
E-4 -
44
w (0 ~ c 0x0.
0 WC 44
208
0 Gehlenite 1' run
O Gehlenite 2rldrun
0 Anorthite
K Kovor
CN Cupro-Nickel
S Standard: no braze cycle
CN (CN) Curve
7---
©t
CN S, K
5- K
CN
K
-(K) Curve
2@ KO3O
a.
7-
CN
1 6
Most Crystal Precipitation Level Least
Figure 4.74: Peel strength versus secondary crystallite level in 941 ceramic after simulated braze
cycles in dry hydrogen at 1I0°C for various times. Kovar (K) or cupro-nickel (CN)
peel tabs subsequently brazed at 800'C with Cu-Ag.
-- 209-
'q~w~
SP~fld Sj
--0-- - - -- - -- - -
- - - - - -
-. -- - - - - -
- -- - -- -- -- - -- - . . . . .
-- -- -- -- --
- - -- - -
6- - - - - - - - - - - -
- - - - - -- - - --
- ---
4)
-t4'-
C4-4-
-- 0
I-t- -4 7- 4
.-
. - 1
- -- - U
7 -~ I~- &
--- -4-4-4- f
-21--
10 _
I-
2~~
Io
7-
6-~
9-
4-
Run #1 0 Run #2
$,,-
7-
3-
I D Gehlenite I~t run

O
o Gehlenite 2nd run

Anorthite
I 2 3 4 5 6 7 8 9
Most Crystal Precipitation Level Least
Figure 4.76: Peel strength versus secondary crystallite level in 941 ceramic after simulated braze
cycles in wet hydrogen at I 100C. Kovar peel tabs subsequently brazed at 800'C
using Cu-Ag. Numbers indicate time of simulated braze cycle.
-- 211-
0 PUnod '44OUSJ$S 1018d
44
0 214
7 -
It 117 I 1 1
4I
4) C
0 ~ 0
A.-
.220
S * 4-
ILE0
-212-
5ps no IIInw i
7-
10 5-
4-
0 Run#l
9 ~ +Run#2
7---
4.
7 5-
C4-
4-4-
++
I - - C**C *C O
4 3 7 a 91
cylsi r Iygna I fo vaiuIO'ties Koa K ncponc N
peel tabs subsequently brazed at 800'0C with Cu-Ag.
- 213 --
WUflOd I'hU.4S IOd
.............
....
4-~
- - - - - - -
---------.- I
I I f
.00
- - ----
-~~~ ---------
-----
- - - - - - - - - - - -
7 - --- I0 C
IiIi
fiill
---- ------ - -------
-- - - - -- -
~
10
• Run#l
9- + Run #2
8-
7-
6+ R +60
Ar e
h. 4
3-- + 5
2Oleo Ru 1
1- 060
I l , I lI I I ,, , I
I 2 3 4 5 6 a' 9
Most Anorthite Crystal Precipitation Level Last
Figure 4.80: Peel strength versus secondary crystt lite level in 942 ceramic after simulated braze
cycles in wet hydrogen at I I001C. Kovar peel tabs subsequently brazed at 8000 C
using Cu-Ag. Numbers indicate time of simulated braze cvcle.
-215-
..-..... IP~ ,-... . .... • . .. - -. . r -

spsmd 6qIUSJ4S 108d
Ii *
* ___T.*_ __
21(D
The results of the wet hydrogen fire cn the 941
metallizing show no direct relationship of peel strength
with glassy phase devitrification, Figure 4.76.
General trends of both properties are noted with
time, however, Figure 4.77, indicatesthat a third uncontrolled
variable is present. This could be related to furnace
conditions as the two runs show different peel strengths
(glass diffusion and flow-out are very sensitive to furnace
atmosphere). Figure 4.77 might be considered to show an
irverse relationship of peel strength with crystal
precipitation.
Again the interesting effect of increasing and

dccreasing gehlenite precipitation is noted, in this case,
always associated with anorthite precipitation.
Inspection of the data on the 942 ceramics

shows an inverse relationship of crystal precipitation
amount with peel strength, the cupro-nickel metal member
again not inducing as much stress into the seal and there-
fore showing no tendency to decrease in strength with
increased crystal precipitation.
Again both in a wet or dry atmosphere crystal

precipitation increases to a maximum as a function of time
and then apparently decreases, Figures 4.78 through 4.81.
The main observatic may be summarized:
1) There is a relationship of glassy phase

devitrification with strength. In one case,
seal strength increases with "devitrite" amount;
in three cases, it tends to decrease. It might
be hypothesized that in the first case, a highly
siliceous glass phase remains, but not in the
other three cases.
2) A qualitative rel.tionship in (1) is obscured by:
(i) other variables
(ii) the small sample size employed,

(strength data is essentially statistical)
(iii) the semi-quantitative nature of the X-ray

data
(iv) zhe small amount of data obtained.
217
3) Glassy phase devitrification is a maximum
between 5-30 minutes after start of the 1100 0 C
soak. After the maximum is reached, the surface
crystal species is resorbed. This resorption
can be due to migration of ionic species from
the interior of the specimen to the highly non-
equilibrium surface grain boundary phase. (See
Tables 4-10 and 4-12 and their accompanyinq
discussion.) This diffused glass phase may
again devitrify after a further long holding time
(180 minutes).
4) A cupro-ni -:kel metal member induces less stress
into the seal area than a Kovar metal member.
4.5.3.4 Effect of Metal Member
All tensile test specimens in the previous stud,'

employed 70-30 cupro-nickel metal member because of its knownr
desirability, e.g., adequate strength and good ductility.
However, a special comparison was made between 70-.0 cupro-
nickel and Kovar (Fe-Ni-Co) to demonstrate the difference
between a "soft" and "hard" metal member. This comparison
was made using special alumina body 992 metallized with P-4M,
a combination known to give relatively weak seals which fail
in the seal area rather than in the ceramic. (Copper braze
was used.) This mode of failure was selected in order to
make a clearer distinction between the effects of a "soft"
and a "hard" metal member on the stress distribution in the
seal area.
Tensile tests for seals with 70-30 cupro-nickel

averaged 6180 psi versus 3885 psi with Kovar. Since all
other sections of these seals were identical, the difference
in strength was attributed to the additional ability of the
cupro-nickel member to yield and relieve high local stress
concentrations. The properties of the two metals which
affect seal strength are compared in Table 4-47.
4.5.3.5 Complex Metallizing Paints
A continuing seal parameter study must consider

,complex metallizing systems. However, this present work
carries us to just the beginning of this subject.
In the electrical section the behavior of several
metallizing paints on commercial ceramics, sapphire, 992 and
Lucalox* is noted, while a prior section has considered many

218
TABLE 4-47
Mechanical Properties cf Two Frequently Used Metal Members
Property 70-10 Cu-Ni Kovar
Tensile strength, annealed 55,000 psi 75,000 psi
Yield Strength*
0.5% offset 18,000 psi 54,700 psi
0.2% offset 14,500 psi 51,500 psi
Hardness, Rockwell B 40 82
13 x 10 - 6
Mean coefficieilt of thermal 16.3 x 10-6
expansion to 1000 0 C
Note:
Al ough the nominal failure strength in reither type of seal
exceeded the yield stress* of its metal member, it is likely
that in i.ighly localized areas tl]e stress was great enough to
cause at least the softer metal to yield. This minute
ceformation allows the load to be transferred to a larger
area. The total load then may be increased until the critical
failure stress is exceeded at some point and followed immedi-
ately by failure of the sample.
*Yield stress is an arbitrarily defined value whose magnitude

is determined by the shape of the stress-strain curve.
Some amount of plastic deformation may occur under stress
much below the yield stress.
219
paints on bodies A, C, G and H.
It may be pointed out here the vastly different

behavior of P-! and P-7 metallizings on sapphire and 992
ceramic. It is noted that the sapphire surface is relatively
not attacked by the P-7 metallizing, showing that the rate
of chemical attack is low. The attack of the MnO-TiO x of the
P-I metallizing along preferential crystal directions of the
sapphire is evident. (Figure 6.59.)
4.6 Shear Test StLad
The inadequacy of the tensile test has been

previously pointed out. 2 It was therefore decided to
initiate a study of shear test geometries as it was considered
that shear is the predominant mode of stress under whic>
failure of ceramic-to-metal seals nccurs in actual tube
practice. Encouraging results have been obtained in this
initial exploratory study.
4.6.1 -ile Geometry
The design chosen for shear testing seals utilized

two ceramic cyl'nders, one of which fit inside the other.
A short length (.050 to .100 in.) of the I.D. of the larger
and the O.D. of the smaller cylinder were metallized and
plated. The two part:3 were then mated and brazed together
to form the assembly illustrated in Figure 4.82.
There are three advantages to this type of sample

geometry. One is that the joint may be stressed in shear with
a conventional compression tester requiring no unusual
grippers or special equipment. Secondly, the specimen is
such that the joint may be leak checked. This is not the
case with the traditional overlapping plate geometry used
for shear testing. A third feature is that all states of
residual stress may be induced in the seal, which would aid
in the design of O.D.-I.D. seals for such applications as pin
seals and feed-thrus.
4.6.2 Testing Methods
Two test loading techniques were investigated.

In one, the specimen was placed with its big end down and a
load was applied to the end of the small cylinder to cause
the small cylinder to slip inside the larger one after
failure of the joint. In another method, the specimen was
inverted with the small cylinder in a hole and the end of the
larger cylinder bearing on the rim of the hole. The load was
220
.085 DIA
I OD.- I.D. BRAZE.
~~1 I[o--.250
Ir
DIA
.050
1.025
.500
-3 75 DIA
Figure 4.82: Shear test assembly consisting of two cylinders joined with an 0.D.- l.D. braze.
- 221 -
applied through a rod to the end of the small cylince" inside
the larger one which caused the smaller cylinder to be pushed
out of the larger cylinder after failure of the joint. It
was thought that the second method would avoid errors in
results which might arise by wedging a braze fillet into the
gap, because this method would push fillets out rather than
wedge them in. However, the first method proved to be better
because fillets were not a problem, and other difficulties
arose in the inverted position which caused the outer ceramic
cylinder to split open in nearly half the samples. Figure 4.83
shows a specimen in the preferred test position.
Materials used in the evaluation of this seal

testing method included the typical braze metals, copper and
BT (copper-silver eutectic), copper or nickel plating, P-1
metallizing, and cylinders of Body G ceramic, Kovar, or
molybdenum. By selecting different combinations of materials
for the inner and outer cylinder, the braze joint could be
placed in any desized state of residual radial stress. A
molybdenum outer cylinder brazed to a ceramic inner cylinder
placed the joint in radial tension. A Kovar outer cylinder
with a ceramic inner cylinder placed the joint in radial
compression. The effect of residual seal stress on the
ultimate seal shear strength could thus be studied by mixing
materials and comparing results.
4.6.3 Test Results and Discussion
The results of tests using several combinations

of cylinder materials, plating, and braze are listed in
Table 4-18. The table lists all materials used in each part
of the seal structure, the residual radial stress which
resulted from the combination, and the yield and failure
stress of the joint. For the purpose of these tests, the
yield stress was considered to be the stress at which plastic
deformation began, or the first deviation from linearity of
the stress-strain curve. This is not the conventional
approach, in which it is defined as that stress at which some
arbitrary amount of plastic deformation occurs (.2% or .5%
offset), but it was suitably reproducible for the purposes
of this work. With some specimens such as the ceramic-
molybdenum combination (group 1) brittle failure occurred
before plastic deformation, so no value for the yield stress
could be obtained.
Yield occurred in all but a few specimens before

failure, indicating deformation of the metallic portion of the
seal occurred before brittle failure of the ceramic or
metallizing. The yield stress varied considerably from one
222
LOADING RAM
~"INNER CYLINDER"
"OUTER" CYLINDER
ifi
Figure 4.83: Concentric cylinder shear t!st specimen in

loading position in Instron testng machine,
-223-
W4 0 0 0 Lro 0 0 CD
wA iA Nl w w m
.H 4J.s C O'D H H- H- a'
a't 1- 9 C~4 (1 (N
4)
9i 0 0 0 0 0
o o H- m~ wO L
wD
OD
>42 H (N N H- (N (N
5I
*'4 H-
~u)
:J
-r 4)
C
W4~r:0
(D ) (3) 0 1 1
a) r4 to -A-j E-4 E-4 U)
Va
C) P E-4 E-4 E-4 H-
O 4.)
0)
'10 C
N
4.) - *H .r 04*4
lz H Qu LA
0)04
Fa ra
4 Q) 4 .. 04 0) 0) C.)1
0
4-) 0
o1 u
0
u u u
0
U) N 4
4.) 0 0 H
(3) rNO N > N 3 H H *

0 0) 0 0) 0)
u) 0)>~
U) N N
+~ ~ U)
0 m1 Ln (3,I
0)4 0
UH 224 N
group of samples to another, but was quite consistent within
groups. This difference was expected to result from differ-
ences in the materials used to make the seal, but the
dimensions of the braze gap seemed to have a much stronger
effect. The residual stress induced in the joint definitely
affected both the yield and the failure stress. In the
case of group 1 samples, the joint may have been stressed
in tension beyond its yield point, so that it failed in
a brittle manner as soon as the shear stress was applied.
No yield was observed in those samples before failure.
The shear failure stress of the test seals was

most directly related to the state of residual stress in the
joint, the dimensions of the braze gap, and the quality
of the braze. The first four test groups best illustrate
the effect of residual stress. The residual stress condi-
tion designed into each group was: (1) high degree of
radial tension, (2) a moderate degree of radial tension,
(3) a moderate degree of radial compression, and (4)
practically no radial stress. The average failure stresses
for these four conditions were 14,090 psi, 19,950 psi,
37,850 psi and 31,175 psi respectively. These results
were qualitatively predicted from theory. The addition
of a tensile stress component to the shear component
contributes to the failure mechanism.
Test groups 4 and 5 illustrate the effect of

braze gap dimensions and braze quality. The material
compositions of the two were identical and braze conditions
were the same, but the braze length (the amount the inner
and outer cylinders were allowed to overlap) was 50 mils
in group 4 and 75 mils in group 5. The longer brazes in
group 5 were not the best quality because braze material
did not flow sufficiently to completely fill the length
of the gap. The difference in yield stress (4: 27,530 psi;
5: 16,380 psi) was more likely due to partial filling of
the braze than any other cause. The braze area used to
calculate the stress in group 5 was greater than the
amount of area actually filled with braze, so the calculated
stress was smaller than the true stress level in each
particle of material.
A comparison of results in groups 4 and 6 shows
some difference in yield and failure stress, but not as
much as might be expected from the different brazes. The
difference that did exist was probably related to the
braze materials, but the similarity may be traced to the
fact that the failure was a brittle failure which usually
occurred in the metallizing-ceramic or metallizina-braze
225
interface. This was confirmed by visual examination of
samples after testing.
Groups 6 and 7 illustrate the reproducibi-ity

of sample preparation and test technique. The two groups
were assembled, brazed, and tested at different times, but
the results from each were very similar.
The shear test method studied has considerable

merit as a developmental or engineering test for seals
because of the variety of data available from it. However,
it may be less useful for basic seal mechanism studies
because of the difficulty in avoding residual stresses
in the joint. (Even when identical materials are used for
the inner and outer cylinders, the gap between Lhem contracts
with falling temperature.) Also, the ultimate state of
stress from an atomistic viewpoint in either this shear
test or a tensile test is that of tension, i.e. the atoms
are pulled apart at failure. Even though the initial
atomic movement in the shear test may be sliding motion
(shear), the gradual necking in the joint caused by
deformation leads to the introduction of a tensile component
to the state of stress. The shear stress is then actually
a vector with a shear (major) component and a tensile (minor)
component. But since this is exactly the condition that
eAists at failure in an application such as a pin seal,
it can be very useful as an engineering test.
226
4.7 REFERENCES - TASK I
1. S. S. Cole, K. W. Larisch, J. E. Inge, and K. H. Styhr, Jr.,

"Metal-to-Ceramic Seal Technology Study", Final Report,
October, 1960, Sperry Gyroscope Company, Contract No.
AF 30(602)2047, pg. 35.
2. "Tentative Method for Tension and Vacuum Testing Metallized

Ceramic Seals", ASTM Designation: F19-61T, 1961 Book of
ASTM Standards, Part 2, American Society for Testing and
Materials, Philadelphia, Penn., 1961, pp. 1509-13.
3. E. F. Osborn and Arnulf Muan, Plate No. 1, "The System
CaO-A1 2 03 -Si0 2 ", in Phase Equilibrium Diagrams of Oxide
Systems, The American Ceramic Society, Inc., Columbus,
Ohio, 1960.
4. A. L. Gentile and W. R. Foster, "Calcium Hexaluminate and

Its Stability Relations in the System CaO-A12 03 -Si0 2 ',
J. Am. Ceram. Soc., vol. 46 (2) 74-6 (1963).
5. E. T. Turkdogan and Patricia M. Bills, "A Critical Review

of Viscosity of CaO-MgO-AI2 0 3-SiO 2 Melts", Am. Ceram.
Soc.. Bull., vol. 39 (11) 682-7 (l16O).
6. Paul Kozakevich, "Viscosity of Lime-Alumina-Silica Melts

between 1600 and 2100°C", Metallurgical Society Confer-
ences, vol. 7, Physical Chemistry_ of Process Metallurgy
Part 1, Ed. George R. St. Pierre, Interscience Pub-
lishers, New York, N. Y., 1961, pp. 97-116.
7. J. S. Machin and Tin Boo Yee, "Viscosity Studies of System

CaO-MgO-Al 2 O-SiO2 :(II), CaO-A12 03 -Si0 2 ", J. Am. Ceram.
Soc., vol. 31 (7) 200-4 (1948).
8. L. R. Barrett and A. G. Thomas, "Surfare Tension and Density

Measurements on Molten Glasses in the CaO-Al 0j-SiO 2 System",
J. Soc. Glass Techaol., vol. 43 179T-190T (1 5 ).
9. T. B. King and P. J. Koros, "Diffusion in Liquid Silicates",

Kinetics of High Temperature Processes, W. D. Kingery, Ed.,
Technology Press of Massachusetts Institute of Technology
and John Wiley & Sons, Inc., New York, N. Y., 1959, pp.
80-85.
10. Helen Towers, Michel Paris, and John Chipman, "Diffusion

of Calcium Ion in Liquid Slag", Am. Inst. Min. Met. Engr.
Trans., vol. 197 1455-58 (1953).
227
"Wee .. "
11. Helen Towers and John Chapman, "Diffusion of Calcium and
Silicon in a Lime-Alumina-Silica Slag", J. Metals,
June, 1957, AIME Trans., vol. 209 (1957).
12. P. J. Koros and T. B. King, "The Self-Diffusion of Oxygen

in a Lime-Silica-Alumina Slag", Based on a DSc. thesis
by P. J. Koros, Mass. Inst. Tech., 19-.
13. John Henderson, Ling Yang, G. Derge, "Self-Diffusion of

Aluminum in CaO-SiO 2 -Al 0 Melts", Trans. of Metallurg.
Soc. AIME, vol. 221 56-J93(196).
14. L. Reed and L. R. Barrett, "The Slagging o. Refractories

I. The Controlling Mechanism in Refract( ies Corrosion",
Trans. Brit. Ceram. Soc., vol. 54 (11) 671 (1955).
15. L. Reed and L. R. Barrett, "The Slagging of R,-fractories

II. The Kinetics of Corrosion", Trans. Brit. C2ram. Soc.,
vol. 63 (10' 5-9-534 (1964).
16. Y. Oishi, A. R. Cooper, W. D. Kingery, "Dissolution in

Ceramic Systems, III: Boundary Layer Cuncentration
Gradients", J. Am. Ceram. Soc., vol. 48 (2) 86 (1965).
17. R. L. Coble, "Initial Sintering of Alumina and Hematite",

J. Am. Ceram. Soc., vol. 41 (2) 55-62 (1958).
18. R. L. Coble, "Sintering Crystalline Solids. I. Intermediate

and Final State Diffusion Models", J. Appl. Phys., vol.
32 (5) 787-792 (1961).
19. R. L. Coble, "Sinteri,,g Crystalline Solids. II. Experimental

Test of Diffusion Models in Powder Compacts", J. Ap.
Phys., vol. 32 (5) 793-799 (1961).
20. W. D. Kingery, Introduction to Ceramics, John Wiley & Sons,

Inc., Nev York, N. Y., 1960, pg. 357.
21. W. D. Kingery, E. Niki, ind M. D. Narasimhan, "Sinterinq

of Oxide and Carbide-Metal Compositions in Presence ot
a Liquid Phase", J. An. Ceram. Soc., vol. 44 (1) 29-35
(1961).
22. L. Reed, "Progress in Ceramic and Ceramic-to-Metal Seals",

Presented at the 7th National Tube Techniques Conference,
September 28-30, 1964, Nev York, N. Y.
228
23. C. A. Bruch, "Sintering Kinetics for the High Density
Alumina Process", Am. Ceram. Soc. Bull., vol. 41 (12)
799-806 (1962).
24. George W. Morey, The Properties of Glass, Reinhold

Publishing Corp., New York, N. Y., 1954.
25. Gerald B. Carrier, "Electron Microscopic Technique for

Determining the Percent Crystallinity of Glass-Ceramic
Materials", J. Am. Ceram. Soc., vol. 47 (8) 365-67 (1964).
26. Rustum Roy, "Note on the Role of Phase Equilibrium Con-

siderations in Crystallization Studies of Glass", Read
at Symposium on Nucleation and Crystallization Phenomena
in Glasses and Melts, Presented by the American Ceramic
Society in Toronto, Canada, April 25, 3961.
27. 1H. R. Wisely, "Investigation for the Development of Ceramic
Bodies for Electron Tubes", Final Report, June, 1962,
General Electric Company, Contract No. AF 19(604)-7410.
28. L. Reed, "Heat Treatment Studies on High Alumina Ceramics",

Reprint from Sixth National Conference on Electron Tube
Techniques", Pergamon Press, New York, N. Y., 1963.
29. Bennett S. Ellefson and Nelson W. Taylor, "Surface

Properties of Fused Salts and Glasses: I. Sessile-Drop
Method for Determining Surface Tension and Density of
Viscous Liquids at High Temperatures", J. Am. Ceram.
Soc., vol. 21 193-213 (1938).
30. L. Reed, R. McRae, C. Barnes, "Metallurgical Research and
Development for Ceramic Electron Devices", Second
Quarterly Report, January, 1963, Eitel-McCullough, Inc.,
Contract No. DA 36-039 SC-90903, Appendix I.
31. R. i. Berg, "Electronic Size Analysis of Sub-Sieve

Particles by Flowing through a Small Liquid Resistor",
Symnosium on Particle Size Measurement, Special
Technical Publication No. 234, American Society for
Testing Materials, Philadelphia, Penn., 1958.
32. W. D. Kingery and M. Berg, "Study of the Initial Stages

of Sintering Solids by Viscous Flow, Evaporation-Con-
densation, and Self-Diffusion", J. Appl. Phys., vol. 26
(10) 1205-12 (1955).
33. S. Cole, Jr. and H. Larisch, "Metal-to-Ceramic Seal
Technology Study", First Technical Note, October, 1959,
Sperry Gyroscope Company, Contract No. AF 30(602)2047,
pg. 52. 229
-- ,r---9--
34. Leonard Reed and R. C. McRae, "Evaporated Metallizing
on Ceramics", Am. Ceram. Soc. Bull., vol. 44 (1) 12
(1965).
35. L. Reed, "Survey of the Factors Affecting Ceramic-to-

Metal Sealing", Paper presented at the Fall Meeting
of the Electronics Division of the American Ceramic
Society, Philadelphia, Penn., September 16-18, 1964.
36. James R. Floyd, "Effect of Secondary Crystalline Phases

on Dielectric Losses in High-Alumina Bodies, "J. Am.,
Cer. Soc. vol. 47 (11) 539-543 (1964).
230
_ 7- - -', fl -.-
5.0 TASK II, BERYLLIA AND QUARTZ STUDY
Beryllia and quartz are important materials used

as output windows in high power tubes. It was therefore
necessary to investigate these materials on this program.
The investigation was confined to the investi-

gation of commercial ceramics metallized with appropriate
metallizing systems.
Beryllia ceramics are similar to alumina

ceramics and come in 94% and 99.5% categories. Their micro-
structure is much like that of alumina ceramics as is shown
in Fig. 5.01. Fused quartz is an entirely different
material than beryllia or alumina, being glassy in nature
rather than crystalline or a mixture of crystalline and
glassy phases.
5.1 Beryllia Ceramics
Fig. 5.02 shows a 96% beryllia ceramic with P-I

metallizing sintered onto it. The bond strength in this
system is about 8000 psi. This combination of materials
represents the first commercial use of beryllia in an
electron tube. (1957.) However, a significant increase
in thermal conductivity can be made if a 99.5% beryllia
ceramic is used in place of a lower berylli:,m oxide content
ceramic. (Figure 2, showing this feature, is contained in
the 2nd Quarterly Report.) Therefore, for this present
study, we have considered 99.5% bcrylliui, oxide ceramics
exclusively. The chemical composition of the two bodies
presently under consideration is given in Table 5-1.
The two 99.5% BeO bozies in the form of CLM:-15

test piece s were metallized at 1425 0 C under the usual
conditions of metallizing, using P-3 metallizing, and
showed good tensile strength. Body BA had an averaje of
7200 psi (neck breaks) and body L3 had an average tensile
strength of 13.300 psi. Body BA leaked, however, whilc
Body BB was vacium-tight. It was noted that Body BA stowedj
extensive penetration of the melt-oyide phase mtallizini
into the beryllia ceramic. The depth of penetration was
approximately three-sixteenths of an inch. Lower metalliin'
temperatures and different paints were next used on Box!o BA
:n order to improve the metallizing characteristics.
Polished sections of Body ,B metallized with P-3

were prepared. PhotomicrogIraphs of metallizinq sintered at
1325 0 C and 1425 0 C appear in Fig. 5.03. One notable feature
231
_____________________.- . * • ~• -
Figure 5.01: Microstructure of a 96% beryllia ceramic (400X) which shows similar characteristics to a
96% alumina ceramic.I
-232 -1
Figure 5.02: Microstructure cross-section of a ceramic metal seal of P-1 metallizing paint on a 96%
beryllia ceramic (400X).
233 .
I P wm p MPT =• v
Figue 503:Phoomirogrphsof -3 etalizng snteed t 1250 (tp) ad 1251 (btto) o
6/40braz, BB.cppeCu-u brze wshe (4OX)

beryliabod
-- 234-
TABLE 5-1
Chemical Composition of Beryllia Ceramics
Weight Per Cent Present
Oxide Body BA Body BB
BeO 99.8 99.49
A1 2 03 0.03 0.23
SiO 2 0.01 0.16
MYO 0.20 0.07
CaO 0.10 0.01
Na.0 0.005 0.03
Fe 2 03 0.01 0.01
of the metallizing in both photomicrographs is the presence

of relatively large crystals in the glassy phase of the
metallizing.
Microprobe analyses were made on seals made

with P-3 metallizing on beryllia bodies BA and BB. This
metallizing had been sintered at 1425 0 C. The data obtained
with the microprobe are presented in Figs. 5.04 and 5.05.
The bodies were analyzed for Mn. Si and Ti. The penetration
of these elements into Body BA wa extremely deep, being
several thousand microns in each case. Penetration into
Body BB was an order of magnitude less, ranging around
100 microns. The reason for this remarkable difference lies
in the difference in microstructure and chemical composition
of the two bodies. The orevalent secondary component in
Body BA is MgO, while the most prevalent secondary
components in Body BB are A120 3 and SiO 2 with some MgO.
At the metallizing temperature, the secondary components
in the beryllia combine with the non-metallic portion of the
metallizing to form a modified melt phase. With very little
235
METALLIZING BERYLLIA
Mn Ti Si LAYER BODY BA
(Somple A-I)
po a eFracture Path
I Metallizing portion of $Sol
1 not probed, somple
fractured through crornic
Z /Extent of Element
Peneftafl from Seal
WJ Into Ceramic
l'-z W---M 4280
4--- " -- Ti 3860

I I~\ \\--Si '-1000
a4.
5 2..
/I'1
t --
/1
0 0 00 I i " "1
40 30 20 10 0 10 20 30 40 50 60 70 80 90 100
DISTANCE MICRONS
Figure 5.04: Electron microprobe trace of seal made with metallizing paint P-3 on beryllia body BA.
-236-
Mp Ti Si METALLIZING BERYLLIA
LAYER BOY B
15 6-
S10"-4"-,i'i Extent of Element

If ,K. Into Ceramic
Penetration from Seal
4, -M Mirons
560
1400 T
52 ---- Si 150
40 30 20 10 0 10 20 30 40 50 60 70 80 90 100
DISTANCE MICRONS
Figure 5.05: Electron microprobe trace of seal made with metallizing paint P-3 on beryllia body BB.
237 -
SiO 2 present in Body BA, the melt phase viscosity is much
luwer than the melt phase of Body BB, which contains much
more than SiO 2 . Th'i low viscosity results in high diffusion
rates for the penet.ation of MnO and TiO 2 into Boy BA.
P-3 metallizing paint was found to be susceptible

to blistering of the molybdenum metallizing. This is due
to evolution of oxygen from titanium compounds at some point
in the metallizing firing schedule. The blistering may be
avoided by using TiH 2 in place of TiO 2 , firing at a lower
temperature, 1300 0 C - 1200 0 C, or by using paints which have
no titania in them as has been done below by using P-8 and
P-13.
Two metallizing paints, P-8* and P-13* were

tested on beryllia bodies BA and BB (see Table 4-221 in
order to develop a metallizing paint suitable for use with
these beryllia bodies. Three metallizing temperatures
were tried, 12000, 13000 and 1400 0 C. Most of the work
was done with Body BA, with Body BB just used in one test
for purposes of comparison.
A scratch test identical to the one described in

Task I was used to measure the hardness of the metallizing.
Scratch test results are presented in Table 5-2. Paint
P-13 produced consistently harder metallizing.
Tensile azserZblies were made with 0.0005 in.

copper plate, 0.004 in. 60/40 Cu-Au braze, and 0.020 in.
copper washer. All BeO pieces were sealed to Body A alumina
pieces and failed on the BeO side of the metal washer
because the alumina had a higher tensile strength than
the beryllia. All assemblies were vacuum tight. Tensile
test results appear in Table 5-3. The best strengths
were obtained with P-13 sintered at 12000 C. The strengths
obtained with F-8 were comparable to P-13 except at the
lowest sintering temperature.
All the assemblies pulled some ceramic or broke
in the neck on the beryllia side when they failed in the
tensile test. P-13 metallizing on an asscmbly of two pieces
of Body A gave a strength of 11,887 Vsi. These two facts
indicate that tl.: weakest reqion of the assemblies which
were tested was the beryllia cerami rather than the seal
*See Table 4-22 for composition.
238
TABLE 5-2
Scratch Tests, Beryllia Metallizing
Metallizing
Temperature, Hardness (1-5)*
Body Type ocP-8 P-13
BB 1300 3 4
BA 1200 2 4
BA 1200 2 ~ 4
BA 1300 2 3+
BA 1300 3 4
BA 1400 2+ 3+
BA 1400 2 3
*Same scale as used in Task 1.
239
• • -. • ~~.....
. . .. . . . . . .r .
TABLE 5-3
Beryllia Tensile Test Results*
Metallizing Tensile Strength,

Temperature, psi
Body Type Oc P-8 P-13
BB 1300 5,505 7,950
BA 1200 4,195 6,825
BA 1200 5,100 7,050
Average 4,647 6,937
BA 1300 6,600 6,775
BA 1300 6,875 6,675
Average 6,737 6,725
BA 1400 6,225 5,910
BA 1400 5, '00 6,625

Average 5,612 6,267
*All assemblies were made with 0.0005 in. copper plate,
0.004 in 60/40 Cu-Au braze, and 0.020 in. copper washf.r.
All BeO pieces were sealed to Body A pieces, and failed on
the BeO side. (P-13 on Body A pulled at 11,887 psi average.)
All assemblies were vacuuwn tight.
240
area. The slight downward trend in strength with higher
metallizing temperatures may be attributed to weakening of
the beryllia ceramic as a result of the heat treatment and
reaction with the metallizing. For this reason alone, the
lower metallizing temperature is preferable.
The work summarized above outlined the develop-
r.ent of successful metallizing systems for use with beryllia
ceramics. The electron microprobe and microphotograph data
provided a valuable insight into the corrective action
required to produce a successful metallizing paint.
5.2 Fused Quartz-to-Metal Seals
An established process for sealing fused quartz

to metal, ,whichwas described by Robert H. Dalton of Corning
Glass Works, was tested1 . The process employs se ling
glass such as General Electric GSCI or Corning CGW No. 7230
to join fused quartz or Vycor to thin (2 mils) platinum
metal members.
Quartz tubing (0.178" O.D. x 0.087" I.D.) was

used in the tests. GSCI glass was ground minus 325 mesh,
mixed with a lacquer vehicle, and applied to the ends of
the auartz tubes to be joined. The glass was fused at
1250C in air for 30 minutes and then cooled. Two glazed
tubes were then butted together with a 2 mil platinum
washer between them. A pressure of 2 psi was applied to
the joint while the assembly was heated to 1350bC in air
for 20 minutes. The pressure was maintained while the
assenibly was cooled. Several vacuum tight seals were
made using this technique, but the strength of the joints
was not determined because of the geometry of the samples.
A one-inch quartz window was also processed and found to be
vacuum tight.
In a second quartz sealing technique, molybdenum

was evaporated onto a quartz disc. The disc was then
sealed to2 a gold metal member by a pressure bonding
technique . A high strength vacuum tight seal resulted.
The seal was sectioned and photomicrographed. The photo
appears in Fig. 5.06. It shows two pieces of quartz
diffusion bonded and sealed to a goid member in the center.
A very thin laer (just under a micron) of molybdenum may
be seen on the quartz members immediately adjacent to the
interface.
Disc windows up to 4 inches in diameter were made
by this process on an EIMAC sponsored program which ran
parallel to this study.
241
GOLD
MOLYBDENUM
QUARTZ
Figure 5.06: Photomicrograph of fused quartz-to-metal seal. The thickness of the deposited
molybdenum layer is near the limit of resolution of the microscope. (400X)
-242-
5.3 REFERENCES - TASK II
1. Robert H. Dalton, "Seiling to Silica and Similar Giasses"

Proceedings of the Eighth Symposium on the Art of
Glassblowing, The American Scientific Glassblowers
Society, Wilmington, Delaware, 1963.
2. 0. Heil, U. S. Patent No. 3,115,957, December 31, 1963.
243
6.0 TASK III - ELECTRICAL STUDIES
6.1 General Introduction
6.1.1 Sources of Electrical Loss
When considering a vacuum tight rf window

structure,which must pass large amounts of power, one
immediately becomes concerned with any losses introduced
by the window. These losses are often of relatively low
magnitude compared to the total power passed, but at high
levels of power even a small loss produces unacceptable
heating effects.
Windows generally have three recognizable

sources of loss, each with distinct characteristics. The
first of these occurs in the ceramic itself and is charac-
terized by a certain dielectric constant and loss tangent.
At the joint between waveguide and ceramic some metallurgical
or mechanical bond must be made. The metal to which the
ceramic is directly bonded forms the second region of
interest. This is often a metallizing paint of relatively
good conductivity whose surface (ceramic side) is irregular.
The third region, distinct in itself, is the non-metallic
"interface" region. This area extends from the metal
boundary into the ceramic. The depth of penetration varies
from a few microns to over 50 mils. It is formed by
diffusion of the components of the metallizing paint into
the glassy phase of the ceramic as shown in Figs. 4.50,
4.51, and 4.52. These components may react chemically with
the glassy phase. The several loss mechanisms in a window
may be summarized as follows:
i) Conduction currents:
In the ceramic there is a finite number of

mobile charge carriers both positive and negative
which in the presence of fields cause conduction
currents to flow. One can define an effective
conductivity of the ceramic which gives rise
to 12 R loss.
In the metal boundary the number of free charge
carriers is many orders of magnitude greater and
hence a given applied field will cause a pro-
portionately greater current to flow.
244
ii) Relaxation losses:
Losses may occur in a dielectric which are

distinct from conduction effects. In certain
materials, an applied electric field will cause
polarization. This might be polarization of a
bound electron with its nucleus, polarization
of several positive and negative ions, etc.
Each type of polarization has its own natural
frequency. As frequency is varied, various
polarization modes are excited. The character-
istics of dielectrics are such that the displace-
ment current flowing when a polarization mode
is excited is not exactly 90° leading the agplied
field, but is some small angle less than 90
This means the displacement current has a
component in phase with this applied field
giving rise to real losses (heat). Very
simply, frictional heating takes place as the
dipoles change orientation in the alternating
field.
iii) Piezoelectric and magnetostrictive effects:
The ceramic material, particularly the interface

region, if piezoelectric or magnetostrictive,
will deform when an electric or magnetic field
is applied. Frictional losses will occur,
thereby creating heat.
iv) Semiconductor effect:
The interface region might form a semiconductor

junction either between itself and the metalliz-
ing material or between areas of different
p-roperties in the interface Losses then would
be a function of current direction.
6.1.2 Electrical Properties as a Function of

Microstructure and Chemistry
It is the purpose of this task to correlate the

losses in the seal with studies of the microstructure
programmed in Task I and additional studies carried out
jointly with the electrical studies.
A microscopic examination of the metallizing
shows that, as mentioned above, the metallizing is made up
245
I
I
of two basic components: (1) a metal portion, and (2) a
non-metallic portion, as shown in Fig. 4.50. The microprobe
data, Fig. 4.52, supports this conclusion.
The dc resistance of the metallic portion is

essentially that of pure molybdenum (or other refractory
metal); i.e. 5.7 x 10-6 ohm-cm. The resistivity of the
non-metallic portion of P-I metallizing can be determined
by painting the non-metallic portion of P-1 metallizing _
only onto various ceramic bodies, including Body A and --
sintering under normal metallizing conditions. With a -
paint thickness of 0.5 mils, the resistivity is greater
than 1 x 1011 ohms per square, indicating that the layer
behaves as an insulator rather than as a semi-conductor at
room temperature. Other compositions such as Mo-MnO-SiO 2-TiO 2
(P-3) and Mo-TiO 2 have also been examined. The non-metallic
portion of P-3 behaves as an insulator also, while the non-
metallic portion of the latter behaves as I semi-condgctor,
and has resistivity in the range of 1 x 10 to 1 x 10 ohms
per square depending on firing conditions and body type. --
Microscopic examination of the metallizing further reveals X
that the dc current through a metallizing layer will follow
a tortuous path, thus accounting for the fact that the layer ><
exhibits a resistance approximately 5 to 50 times greater
than that shown by solid molybdenum.
The rf conduction characteristics of the layer

must be different from that of a solid molybdenum layer
since the electrical path width, in many cases, is less
than the rf skin depth thickness for pure molybdenum
(i.e. 1.15 microns at 10 Gc). Reference to Fig. 6.01 will
clarify this point. This figure shows a schematic path
through the metallizing, and indicates that, no ratter how
thick the metallizing is, some current must be carried in
the overplate and braze for, at certain points in the
continuous molybdenum conduction path, the total width of
the path will always be less than 1.15 microns. The rf
situation is thus as follows. When the metallizing is very
thin, the current is carried almost exclusively within the
overplate and braze region. As the metallizing becomes
thicker, a portion of the current will be carried in the
overplate and braze and a portion in the metallizing itself.
This ratio will reach a fixed point beyond which furthez
increase in moetallizing thickness will result in no more
current being carried in the metallizing. Figui
The ratio may vary, indicating no current being
carried in the metallizing, or all the current being carried
246
/ /
-.... L /11,/ //
.- A-. .
- * -
..
- -na
'I
...
a, a
.
a -"
.
- .-
rf
RR Braze/
/
Meta
I / i~
WIMlbeu sIte 2iMcrn
-247
,.', Non metallic solidified melt phae
Mhemetlizthroug
*Ceramic lg
Figure 6.01: Schematic representation of Figure 4.50 illustrating the electrical path length and width
through the metalliz,~g.
-247 -
in the meta]lizing in extreme cases, depending on the
continuity of the molybdenum component as it is affected by
the sintering time, temperature and other sintering parameterc.
The rf dielectric losses in the seal area will

be a function of the composition and amou: t of the interfacial
glassy layer and the tendency of the non-metallic portion of
the metallizing paint to interdiffuse with the grain boundary
phase of the ceramic.
6.1.3 Program of Experimental Studies
The experimental electrical studies were

programmed in three main areas:
1) dc resistivity measurements of ti.J
(i) metallizing as a function of temperature
(ii) wetting layer as a function of processing

variation
2) low power rf measurements
(i) conduction (12 R) studies
(ii) dielectric studies

3) high power rf studies
From the above discussion, it is obvious that

much valuable information on the nature of current flow in
the metallizing can be obtained from simple &c measurements.
The temperature dependence of the resistivity can define the
contribution of the metallic and non-metallic insulator or
semi-conductor portion of the metallizing to the overall
conduction process. The nickel or copper overplate may
influence the metallizing as . consequence of high temperature
diffusion alloying processes.
Low power rf experiments were then devised in

order to isolate the two main sources of electrical loss so
that the loss variation with paint composition and processing
conditions could be systematically explored.
Lastly, it was desirable to determine the magni-

tude of the losses by a direct measurement technique;
accordingly, high power rf studies were programmed and
a ',ial loss data obtained for a simulated window situation.
248
6.2 dc Electrical Measurements
6.2.1 Temperature Coefficient of dc Resistivity
6.2.1.1 Relation to Total Seal Losses
Elaborate cooling systems have been designed to

prevent high power microwave windows from overheating.
Experiments with high average power demonstrate that a large
heat conductivity of a window material is more important
than a low loss factor. Beryllia ceramics prove, therefore,
in general, better than high alumina ceramics. Most o.ide
ceramics, fortunately, have dielectric constants and ±oss
tangents rising only slowly with increasing temperature,
at least up to 600 0 C. TI..Ls may not be the case at all for
the losses in the seal area, however, because the dielectric
in that region is actually highly contaminated. Microwave
dissipation due to dielectric losses is, therefore, increased,
and one can expect the loss tangent to rise fast with
temperature. TIne other type of loss, due to conduction
currents in the plane of the metallizing may be expected to
depend on the temperature dependence of the metal component
in the metallizing.
By far the simplest method of determining this

characteristic is to measure the direct current conductivity
of the metallizirq on ceramics as a function of temperature.
The dc conductivity is in direct relation to the micriwave
conduction losses as is shown by the following three equa-
tions describing wave propagation in waveguides:
Plost = 2 °(Pin (l) (1)
where
Plost: power dissipated
Pin : power incident

o : attenuation constant (see equation (2)).
1 : length of waveguide section
S,. 2b( )2 (2)
249
where
R : characteristic resistance (see equation (3) ).
-' : intrinsic impedance

a : width of waveguide
b : height of waveguide
A: free space wavelength (m)
R = 10.88 x 10 - 3 I_
107 ohms (3)
dc conductivity.
The microwave power dissipation due to conduction losses in
the metallizing is, other factors being constant, inversely
proportional to the square root of the dc conductivity:
1
Post = Cx C = constant (4)
Equations (2) and (3) refer to the propagating dominant

mode in a rectangular waveguide only, but the general state-
ment in (4) is applicable also to resonance structures such
as microwave windows, supporting propagating and standing
waves.
Results of the dc measurements of the thermal

coefficient of resistivity in the metallizing will indicate
the trend with temperature of the microwave conduction losses
in ceramic-to-metal seals.
6.2.1.2 Experimental Set-Up
The experimental work consisted in the investi-

gation of the thermal coeff ifient of de resistivity for
various kinds of metallizing on two kiids of alumina
ceramics. *
Figure 6.02 is a photograph of a test sample.

A metallized ceramic strip is provided with two heavy copper
contacts brazed to the ends. Two wires for the measuring
current and the detection of the voltage drop are attached.
*Bodies A and E as per Table 4-1.

+Metallizings as per Table 4-22.
250
ONE INCH
Figure 6.0Z: Measurement of thermal coefficient of resistivity in metallizing on ceiramnic. Test sample
strapped to ceramic heater.
The sample is mounted on another ceramic, which has a heater
filament inserted. The temperature is directly measured on
the metallizing surface with two chromel-alumel thermo-
couples, one in the centrr of the sample, the other close
to the end, in order to check in all experiments the even-
ness of the temper tture distribution. The test sample is
surrounded with a heat shield (not shown in the photo),
which helps to establish an even temperature distribution
by reducing radiative losses connected with heat flows and
temperature gradients. Measurements were made up to 800 0 C
in a demountable vacuum system with a pressure lower than
5 x 10-5 Torr. Data were taken during the complete heating
and cooling cycle. Only samples with completely reversible
changes were evaluated.
The electric circuitry is straightforward as

shown in Fig. 6.03. Resistivity is measured from the
voltage drop across the sample produced by an accurately
measured current.
6.2.1.3 Results
The thermal coefficient of resistivity is

defined as the change of resistance of a test sample
relative to the resistance at zero degrees centigrade per
degree centigrade.
This quantity is generally a function of tempera-
ture. Listing only one or two values for specified
temperature ranges means that original measuring curves have
been evaluated with linear approximations. Fig. 6.04 gives
as an example the original measuring curve of simple No. 10
(metallizing P-3 on Body A). Resistance data upon heating
and cooling are marked as dots and circles, respectively.
At 400 0 C. there seems to be a slIqht "kink" in the curve.
Averaging0 therefore, extends over the temperature ranges
0-400C and 400-8000 C. In other cases such "kinks" occur as
low as 300 0 C.
Representinq results as the resistance of test

samples as a function of temperature as in Fig. 6.04 .- ocs
not allow an easy comparison between samples of different
kinds. It is, therefore, desirable to normalize the vagues
of resistance at 0 0 C (extrapolated from the value at 25 C)
and plot the ratio of the resistance at temperature versus
the resistance (normalized) at O°C as a function of tempera-
ture. Fig. 6.05 gives an example which compares the
resistances of samples No. 3 (annealed, 0.001" thick strip
of molybdenum), No. 4 (metall~zing paint P-1 on Body A) and
252
Leeds Northru
ThC ITC
Figure 6.03: Circuitry iGr measurement of thermal coefficient of resistivity.
-- 25J --
.56-
.52-
.49-
.44-
.40-
136
.-
0 .28-
.24"
.20
.16 -
o ,60 2 60 4b, 560

TEMPERATRE 4
Figure 6.04: Resistance versus temperature oi test sample No. '0 (metallizing P-3 on Body A).
2,54
'9 -. m. - . - • - " l
Molybdnum
P-I on Body A
0
0 100 200 300 400 500 6;0 700 600

TEMPERATURE 'C
Figure 6.05: Measured resistance ratio PTI/R. Of three s-ample3 as a functio-n of temperature.
-255-
No. 5 (metallizing paint P-i on Body A machined down).
In this way, the unimportant absolute resistance values are
eliminated, and the slope of the curve can be easily related
to the thermal coefficient of electrical resistivity, c
according to
R = C T + 1 (5)
Ro
Table 6-2 provides back-up information, namely

the lengths, widths, and thicknesses of the metallizing
layers and the resistances measured at room temperature.
Resistivities and resistances per square shown in Table 6-1
were calculated from these data.
Figs. 6.06 through 6.11 contain the measuring

data in graphical form for each comparison of samples.
Figs. 6.12, 6.13 and 6.14 show the cross-section of signifi-
cant metallizing layers mentioned in the text. Numbers on the
curves key to the sample description and the thermal coeffi-
cient of dc resistivity in Tables 6-1 and 6-2.
6.2.1.4 Discussion of Results
In a metallizing layer, molybdenum particles

are partly sintered and suspended in the glassy phase.
Some compositions, especially the ones with titanium dioxide,
suggest that they may behave like semiconductors. The
contribution of the metal portion to electrical conduction
is much larger than the effect of the suspected semiconductor
portion of the metallizing. Measuring results shown in
Fig. 6.05, however, clearly demonstrate how the presence
of a glassy phase with a negative thermal coefficient of
electrical resistivity reduces the measured coefficient of
pure molybdenum (top curve). The metallizing layer adjacent
to the ceramic has a higher glass content. Reducing the
metallizing thickness by machining off surface layers
therefore increases the average glass content. Correspond-
ingly, a still lower thermal coefficient is obtained, as shown
in the lowest curve in Fig. 6.05.
With this introduction a complete evaluation

of all measuring results is reported below, together with
microscopic investigations of the significant test samples.
256
4.> o) U U O U U U U U U
444 0LA i 0t
OLA 0 L 0 0 0Lr o 0 0 0
4) >U 0Or- 0cj ON
0r 0j N in' LA
0 0
0.-H0 Ln o L) O
0r OLn OL 00
LoA-N OD N~ 1 OCD LAn o-0r-
U)re)
~10 M4 I') O CO mi 1
10 1 0 1 1 10 gO
1
1
r- H 10 Gi f OLA LL 00
1
0 0 00 O LA 00 00
(auWO0 N q' q( CN %4 m~ (1m N l N cM N m
a'r L%4 00 3'0) CDm 0.0

S4.4 0 0* LAO0
0 I
E- 0 ~Lt
1u LAIf Nt) * I.
m mm( c4 r4 ~c4r4 c4
0) mY fn C%4
U- C14 C14 04 04
I4 I I I I
z10~00 0 0 0 0 I I
4.) U (UQ.i
13 W 0 0 0
H- H- H- H H H
'Hq W -H
4J 0 V c -4 0 a
) ~ 0"HC
E nG) ODL
L -
>1
0 rqU . .0I'0
>00 '. . %
.
U) *-4I n ~ 0 1 1 1II
H H 0 0 H 0
H 4.) 4J NU 00
H- H H H HH
a) i4) x x x
UCaOr Ho m
x x x
Q) 0 11 N C% mn . H
N H4. N m
(L)- 04 0-4S 0 0) 0)
U Ii $4 4 UV U u
H U4 U) Lo A Ln c)o
SN 0)~ > 0)z
(D :
-IHV r-HV H-" H- 0) ~ r4 4J H r-Ir >1
0) 4)0) 00) 0 (13 ):0 EOV $4 tU to~9 CoA
S EHr-4 v> 41 I
k co
U)-4 (a
E-4 w- U ( 0 (U I 4J a). C
iU )H U)w
04 V 0,
(k :QD
:0 0)) UH u~~- r.
Vo5( H0 HU o g~ ro 9:
0) or: or.0 00) 0) 1
r: r4.C ~ (o g u r
w ~ urz uw "-4Vu~
0
0(a 0 (a 0 0 0 $4 N C .1-4 *rO4-) rq1q-4
0 ~ ~ C: , 6*0 Z 0 n .r ,% (0) 4 >1 00) -r4 1 41 > 4) (o I
U) 0 0 : 0 tu m i4 r- 4J AQ Z H- r 41, 41 w li
00W
U
14
fu 0 +
4J
(Ula,
.j.
co. M
0)lu I I I II I257
4.1
10
>4
U, H
44 -,1 0 0U 0 0) 00 0 0 o000 o 00 o-X _
NL 0 0 00 OO 000000 0
0 -H U
0N L0 UONALt) 0 Lno OLn M 00 0L
000 m Ln Ln r- LAN %F or-
0N 'nr- I 0
410 'I I D r MI N I mi I~ Nl 1
O r-
10 000 0 1
lN 1 1 10 10 1ILO 10 1'0 -
HI
r- 4( 00 0 U, LA Ln LO LA LAn 0 0 ma0
(aU)
M1 ~1
N%~j LAO
q ua) t CN > >1
N N (Ye) e'imn N N ~m
C%4 CNC%N-
S00 , 0
H 0 my
1 4 41
O 00 C4 Ln LAO) kDN rv7 ~ (0 L (a (a
M ODO' 0)Lr- *ur- 0)t m
P 0 C%4 m 4 ( C4jm
(rj C4c*
4 44 C44 M, I H- m
04)k-% cs CN C1' 4 N N N~ (N (N
U Q)W I I I I I I I I
r-04 P4 0 0 0 0 0 0 0 0 I
ri f (a H HHH H H H HH 0
44 (AE X x x x x x x x
$~ 4 rC 0
:3 0- 00 NS (n S 0 In
>1
.
oI- '.0I U . '0
>0~E '0 %.
1 0 0
-H UU 0 0 H- H 0 0 0 1
4)C4J H H- 0H H 0
H- H- H H x A.- AQ.
Ean(a 0 m
O -m (N '.0 N, H- U,
Ln W11 W
U, rH
>1 >1
4J0 0)0
a 40 0o 0 0 44'
U~r U4 - V 0J
0 V 0 >1 > u 0 >4 0 >4
V V 0~ 4-)
O0 H( 0,4 O C
0 0 a)
U) VU~ HO4J >n- HO HD4OD
1 4
w w*- t.o uw0 IU)a
9 4 ( a
0 ~ r4 ~
N04OU 4e, 4J ~ 0U
Ei r. C.: ~ .0~ k~
0) 9- 9 U. ~0n 04 0m 0.
rq f-4.-H ~ 4. .41SrCJ. I4N.1 '
0q U 0 j 1 J 4 4 44 4N4
A4. (0> 0> -4 -, , r , jV , H
HJ 4H Z0 4 %-
- H t 4 44) r-4 0 0
(a C 04 z40H04 4 r4 M4r-4 -4 4
Z 4 4 IDA'
04 E-4 E-1E-
ca~ toH H-
FH v '0I
r4 H- 4 5 8 Oh
H-
.H > U- Li u u Lii . ty. tp
~44 4 -u 00 14 0 0
4.4-ri u u0u %0 -. 00 0 N. N4
0 0 00 0 00 UIL Li u v0 04
4) > 0 41 41~
0.,q LA)
QLA
ON
OLA) OL 0C ~ 0C N a .I~
r
IAN ON Ltlr- 0 ii 1g 41 0
U 4-)i
W120
N LA I q*I ~I NI1 0 00 4) 1
1) 10 10 10 qtLn N 0' 4e mW :r
r-r I, > LA Lo LA LA LAO I %4 N I rq~
('!% 4)
0) H ) ( N0H N~ 10~ --A m~(~M

,C 4-4 4) 0 I 'U
.- N > >1
E-O Orn m'j
r4~j r4N4 0000$
(j (N 04
0(N 4
u Q) ) 1N 1
Liu N 0
Q)r4$
to) 0 H ,
0 0 LA 0
(U 4-0 (N n
C) 1.4
Q Xn 0
~W O 10.J 0 4N0
) a) 4.1 0 0
>1 Nl
44 V 1 N )
.r- 0. 0 'U
> 02 I i H
E I, 0 0 0)A 41
41
u LI H 1. U)
* ~ N
44 -
Ea (N H x<
.r4 E H
rA 4
0 X N..
0 . ,Q 0
Q*6- N 0 (N a)4N
4.1 H U2 N
Ia)~~~ 0)0U 40L
4Jo N4 0 >1) 1 *44 (a H.(

U) 4H 0A, lu (a > H 0 uJ
. 4 CO~ UO N V 'U4 O4 (N N4 H
Ci) ) -H r- 0 Ha>19: .-"4 ,F44" V~ N
NN - 41 4.1
Ln '"4N (a'4 IU
44 r-4 r-
a) H
4) 0. 0.04 0 V1.4
H4
E- U) 4.1 H 4-1 N0 A 0 40 04
0)) r-4 ~ "4) .040w H -
(a
0) N 54 H 040E- -4$
N4 4 'UN4J 10 -H N *044A NI- O'
m O H- >4 "4)'
(a 441L "L eH H1
.4.4
U p >~O 4)'U 00 4
44 P-4 0 0 c 0
0)0
4j. $4 41
to4 I I 1 4I 1
P40 00r l
H- E- 9: 0
404 0 0404 04
I oll MO -,q -Ln

TABLE 6-2
Metallizing Dimensions and Resistances of Samples in Table 6-1
Thickness Resistance
Sample Length Width of Metall. at 250 C
No.* (cm) (cm) (10-3 cm) (ohm) Remarks
1 8.1 - 5.0 0.06 irregular geometry of strip

2 7.6 0.18 2.5 0.12
3 7.1 0.20 2.5 0.10
4 7.0 0.63 4.1 0.12
5 7.2 0.63 0.8 0.66
6 5.7 0.63 1.3 0.84

7 7.1 0.51 1.8 0.49
8
9
5.2
5.3
0.51
0.51
1.1
0.46
0.41
0.62
I
10 6.9 0.63 2.3 0.15
11 6.8 0.63 0.9 0.41

12 7.6 0.51 1.0 0.50
13 7.1 0.63 3.5 0.76
14 7.1 0.63 1.5 1.96
15 7.7 0.63 0.9 0.50
16 7.5 0.63 0.7 0.74

17 7.6 0.63 0.7 0.83
18 7.4 0.63 1.1 0.48
19 6.6 0.63 0.51 31.50
20 7.1 0.63 - 16.3 x 103 irreversible changes of
resistivity
21 6.9 0.63 - 2 x 106 " " "
(4350 C)
22 6.8 0.63 0.6 0.7
23 bare ceramic
24 7.5 0.63 2.2 1.9 (0.4 ohms originally)
25 7.5 0.63 1.7 43.0 " "
26 4.3 0.63 0.94 0.19

27 5.25 0.63 0.63 0.18
28
29
4.1
4.1
0.63
0.63
1.1
irreg.
0.67
40.00
(0.1 ohms originally)
(0.2 ohms originally) j
*For identification see Table 6-1.
260
2
I
I
- . - - 1 I II I I I I | I '- , - V - _~~- -
____~~ _ ___ _ 31 _2
4.5 A
2
4.0
3.5Op0
ip3.0
z
4 #
-2.50
LO
S2.01
1.5
0 100 200 300 400 S00 600 700 BOO

TEMPERATURE (OC)
Figure 6.06: Measured resistance ratio vs temperature of test samples. (1)molybdenum strip, 0.002"
thick, unannealed. (2) molybdenum strip, 0.001" thick, unannealed. (3) molybdenum
strip, 0.001" thick,.aiealed. 013) NA- on Body A. (14) P4-M on Body E ceramic. (26)
molybdenum, evaporated on Body Eceramic. (27) molybdenum evaporated on Body E
ceramic, fired at 1425" C ior '/a hour in wet hydrogen-nitrogen atmosphere.
-261
4.5 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _ _ _ __ _ _ _
4.0
a4
3.5 " l
3.00
0 0 0
I- . 0
we
K"00
~2.0 0
1.5
0 100 200 300 400 500 50 700 Goo

TEMPCAATURE (00
Figure 6.07: Measured resistance ratios vs temperature of test samples. 11) molybdenum strip, 0.002",
unannealed. (4)P-1 on Bod A ceramic. (5)P-1 on Body A ceramic, thickness reduced by
machining. (6) P-I or~ Body A ceramic, partly leached with n ric acid. 018) P. on BoyA
ceramic, fired at 16OOWC.
- 262 -
4.0____
2
3.5
3.0
2.5
N60
hi
1.
hi
0 0 0 3040Sooo WS
i00 ~~TEMPERATURE 1OC) 500 0 C
Figure 6.08: Measured resistance ratios vs temperature of test samples. (1)molybdenum strip, 0.002".
unannealed. (7) paint P-1 on Boy Eceramic. (8) paint P-1 on Body E, thickness slightly
reduced by machining. (9) paint P.1I on Body Ethickness further reduced. 422',P-I (modli-
fied with 1'6 -molybdenum content c f regular P-I paint) on Body Eceamic.
- 263 -
--
. NO.
4.0
3.5 0--
2
3.0
2.5 - -,
w 00
Q C
zIaJ or
2.0.
0 . IA
0 a
£ C
000 00(040 0 00. 0
0 100 200 300 400 500 600 700 goo

TEMPERATURE 1OC)
Figure 6.09: Measured resistance ratios vs tempereture of test samples. (1) molybdenum strip, 0.002,
thick, unannealed. 10) paint P-3 n- i3ody A ceramic. (1 I int P-3 on Body A ceramic,
thickness reduced by machining. (12) point P-3 on Body E ceramic.
---264--
4.0
3.5 24
-7
' 25
3.0
w S2.5
'a.0
00 2.0
0 O 0 0
,0 000 0 0
0 100 200 300 400 .St0 600 ?00 Go00
TEMPERATURE it)
Figure 6. 10: Measured resistance ratio vs temperature of test samples. (7) paint P-I on ceremic Body
E. (24) puint P-I on ceramic, slightly electropolished. (25) paint P-I on ceramic Body E,
heavily electropolished.
-265
4 iS 17
3.00
2.5
0
2.0 _
U)
w
4 ,
Ude
U)
U-
0
0
4
0 '5'
2.0
0 100 200 300 400 500 600 700 900
TEMPERATURE (°C)
Figure 6.11: Measured resistance ratios vs temperature of test samples. (4) paint P-1 on ceramic Body
A. (15) paint Ti-1 I on ceramic. (16) paint Ti-il on ceramic Body A, slightly machined.
(17) paint Ti-ll on ceramic Body A, leached with nitric acid. (19) paint Ti-Il on ceramic
Body A, heavily leached with nitric ecid. (28) paint Ti- 1I on ceramic Body E, oxidized in
air. (29) paint Ti-I 1 on ceramic Body E, heavily oxidized in air.
R,
- 266-
)#26
4- Evaporated molybdanumr
4- Bod~y E
b)027
P WSW4- Molybdenum evaporated on

*-Body Ev fired at 1425*Ct
112 hr- in wet H2 /N 2
c)# 13
on
4- Body A
d)#4
on
4Body A
Figure 6.12: Photomicrographs (400X) of test samples. Metallizings in (a), (b), (c) have been
overplated after the measurements.
-267 -
a)#5
4- P-1 (Machined)
on
4- Body A
b) #6
S4- P-I, partly leached (dark)

on
4- Boy A
c)# 7
on
4-- Body E
4- P-1 (Machlrned)
an
4-"d E
Figure 6.13: Photomicrographs (400X) of test samples. Metallizings l-ave

been overplated after the
measurements.
- 268 -
OW 9
____________4- P-I (Machined)

on
4-- Body E
b)# 10
on
- 4 P-3 (Machined)
on
4-Body A
d) 024
4- P-I (Electropolished)
on
4-Body E
4-1*(Heavily elctmopollshed)
on
4- Body E
Figure 6.14: Photomicrographs (400X) of test samples. Metallizings in (a), (b), (c have been
overpiated after the measurements.
- 269 --
Finer methods of investigation such as X-ray diffraction or
electron diffraction would have to be applied in order to
explain the lower thermal coefficient of the coating either
as an interaction with the ceramic substrate or in terms of
the structure of the evaporated layer.
Curves 13 and 14 were obtained with P4-M
metallizing paint on Body A and E ceramics fired on at 1600 0 C.
This coating has a still lower coefficient than the evapor-
ated metallizing although this paint consists of pure
molybdenum powder and does not introduce any extraneous
melt phase. Firing this paint at 1600 0 C rather than the
normal sintering temperature of 1425 0 C used for the bulk of
the samples, increases the interaction with the ceramic in
order to provide the glass bonding phase. Measured resistivi-
ties are about 10 to 20 times as high as in the evaporated
molybdenum layer. Metallurgical cross-sections indicate
very irregular layers with paint agglomerations (Fig. 6.12c).
There is only a very slight difference in the thermal
coefficient at higher temperatures for the two different
substrates with curve 13 being the lower one as expected.
Ceramic A provides more glass for penetration into the
metallizing, and the thermal coefficient of electrical
resistivity is reduced or the metallizing behaves in a
"less metallic" manner.
In Figure 6.07: Curve 1, obtained with molyb-
denum foil, is repeated for reference. Curve 18, for P-1
on A ceramic fired at 16000 C, is very close to curves 13
and 14 (see Fig. 6.06), and its metallurgical cross-section
has the same characteristic appearance, it is irregular in
thickness and has a number of agglomerations and gaps. The
addition of melt phases in paint P-1 compared with P4-M
does not affect the thermal coefficient of resistivity.
Fired at the standard temperature of 1425 0 C, paint P-1
shows a lower thermal coefficient (curve 4). Reduced metal
and melt phase are well interspersed (Fig. 6.12d). There
are many more interruptions of the metal structure in the
paint, and the melt phase thus contributes more to the
electrical conduction. In spite of this, paint P-1 shows a
lower resistivity than the higher fired pure metallic paint
P4-M both after a 1425 0 C or 1600 0 C firing. $any contact
points between small molybdenum aggregates seem to be more
effective than the fewer contacts between large agglomera-
tions in the paint sintered without the help of penetrating
glass. Previous investigations showed an accumulation of
the melt phase between the metallizing and the ceramic,
suggesting a larger contribution of the melt phase to
electrical conduction in this layer.
270
.. ,. . -= -,-
- I. - - w l - .. T n " ! =- -- i
In sample No. 5, the surface layer of the fired
P-i metallizing is machined off (Fig. 6.13a). While the
resistivity (ohm-cm) of the coating is practically not
affected, the resistance per square is increased by a
factor of five compared with sample No. 4 due to the reduc-
tion of the thickness. Curve 5 indicates a lower thermal
coefficient of electrical resistivity and confirms that the
metallizing nearer to the ceramic behaves "less metallic".
It was thought that leaching of the metallizing paint with
nitric acid (curve 6, dashed) would remove part of the
molybdenum and would also increase the relative contribution
of the melt phase. However, the acid probably affects the
melt phase also. Sample No. 6 shows practically the same
thermal coefficient as sample No. 4, although the resistance
per square is increased by a factor of 2.5. About half of
the original metallizing layer is changed by the leaching
as can be seen in the metallurgical cross-section of
Sample No. 6 in Fig. 6.13b.
In Figure 6.08: Curves 7, 8 and 9 obtained for
metallizing P-1 on Body E ceramic, slightly machined P-1
metallizing, and the same metallizing further reduced in
thickness, respectively, demonstrate also the more glassy
behavior of layers in the metallizing closer to the ceramic.
Changes in the thermal coefficient are smaller than between
curves 4 and 5 in Fig. 6.07 because Body 2 ceramic (99%
alumina) in samples 7, 8 and 9 provides very little melt
phase as compared to Body A (94% alumina). Figures 6.13c
and d, and 6.14a show the degree of machining in the metallurgi-
cal cross-sections of samples 7, 8 and 9. Determinations of
the metallizing thickness are not accurate enought to place
much emphasis on the calculated resistivities. The values
are, however, in the same range as for P-1 on Body A
ceramics. Curve 7 (P-1 on Body E) is lower than curve 4
in Fig. 6.07 (P-1 on Body A). This is surprising at first,
because one would expect the additional glass in the case
of the Body A substrate to cause " smaller thermal coeffici-
ent. On the other hand, it may be the melt penetrating into
the metallizing which eauses a higher degree of sintering of
the reduced molybdenum oxide by pulling the particles together
so that a more metallic conduction results.
The above argument is confirmed by curve 22. It
is still higher than for paint P-1 on Body E (curve 7). al-
though it refers to a modified paint P-1, in which the molyb-
denum oxide content has been reduced to one-sixth of the
standard amount. It is, therefore, not so much the composi-
tion of the paint, but the degree of sintering, which
determines if it behaves more or less metallic. A more
271
metallic conduction is characterized by a higher thermal
coefficient of electrical resistivity.
In Figure 6.09: Paint P-3, contains silica in

addition to the manganese oxide and titania in paint P-i.
The thermal coefficients of paint P-3 on Body A and Body E
are practically the same (curve 10 and curve 12) demonstrat-
ing that the silica addition corrects the deficiency shown
above by P-1 paint on Body E ceramic. Figs. 6.14b and c
show the original and machined metallizing P-3 on Body A.
Machining increased the resistance per square by a factor
of 2.5, proportional to the degrees of thickness. The
thermal coefficient is also reduced when the coating is
machined (compare curves 10 and 11). The resistivity of
the metallizing is slightly lower than that of paint P-I.
In Figure 6.10: Another way of removing

elementary molybdenum from the metallizing was tried in
addition to leaching with nitric acid. Molybdenum was
deplated under conditions used for electropolishing
(1 part H2SO4, 4 parts H 3 PO4 , 4 parts H 2 0, 250 C, 1.2 amps/cm2 )
in order to take off molybdenum exposed on the surface in a
rather regular fashion. Surprisingly, both samples 24 and
25 treated this way had thermal coefficients practically
equal to the original paint (curve 7), although the actual
measured resistances have been increased appreciably in
the procedure (from 0.4 ohms to 1.9 ohms and 40 ohms,
respectively, or from 63 x 10-6 ohm-cm to 350 x 10- 6 ohm-cm
and 6,100 x 10-6 ohm-cm respectively). The metallurgical
cross-sections in Figs. 6.14d and e show the large altera-
tions in the paints. Electropolishing removed a large
amount of the molybdenum, especially in sample 25. However,
it affected the metallizing paint throughout the whole
thickness, leaving molybdenum (white areas) more dispersed
than in the original paint.
In Figure 6.11: The three almost identical

curves 15, 16 and 17 refer to paint Ti-il on ceramic Body A.
A slight machining, which increased the resistance from
4.1 x 10-2 ohms per square to 6.2 x 10-2 ohms per square
(sample 16), as well as leaching with nitric acid (sample 17)
which caused the same increase in the surface resistance,
did not affect the thermal coefficient of resistivity.
However, this paint, Ti-il, on Body A is "less metallic"
compared with paint P-i (sample 4). although the latter has
relatively more glass-forming components. This may be
interpreted as due to the highly semiconductive nature of
the TiOx-containing non-metallic phases. As in other cases,
272
sintering of the molybdenum was enhanced by the glass, which
pulled metallic particles together so that many metallic
contacts were established. A much more severe leaching of
Ti-ll with nitric acid, which increased the surface
resistance from 4.1 x 10-2 ohms per square to 3 ohms per
square produced a coating with negative coefficients of
electrical resistivity (curve 19). This behavior must be
interpreted with reservations. Heavy leaching does not
attack the metallized surface evenly, it is apt to produce
gaps in the conductive path. A metallurgical cross-section
shows an unproportionately large amount of molybdenum left,
which is in contradiction to the relatively high resistance
of the sample. The value of resistivity listed is only an
average and not too meaningful in this particular case,
because the total resistance is practically determined by
a small area, as is the semiconductive behavior.
Yet another method of removing molybdenum
from fired metallizing, for increasing the relative melt
content is to oxidize the samples in air at 540 0 C and
subsequently drive off the formed MoO at 850 0 C in vacuum.
Samples No. 28 and 29 were treated this way to different
degrees, raising the original resistances by factors of
7 and 200, respectively (see Table 6-2). It required the
more severe treatment to achieve a thermal coefficient of
rosistivity shown in Fig. 6.11, curve 29, which is very
low and positive, almost zero, and low and negative, in
varioos temperature ranges.
6.2.1.5 Molybdenum and Melt Phase in Metallizing
as Parallel Conductors
All metallizing formulations, including the
Vure molybdenum paint P-4M and the evaporated molybdenum,
had a lower thermal coefficient of electrical resistivity
than the measured molybdenum foils strapped on the ceramic
heater. Resistivities of all these coatings were low
which indicated that the main conduction mechanism was
metallic. But there must be still another contribution
which reduced the thermal coefficient. An equivalent
circuit representing the situation, therefore, consists
of a conductor of molybdenum with the resistance Rm
shunted with the melt phase, a semiconductor with the
relatively high resistance R9. The total measured resistance
at temperature T is then:
R RmRg/(Rj + Rg) (6)
273
- "-. . , Iq ... . -
In Fii. 6.06: Samples No. 1, 2, 3 consist of
molybdenum foiis strapped on a ceramic heater. They show the
highest thermal coefficients of dc resistivity measured in
all experiments. Nos. 1 and 2 differ slightly, probably
due to different degrees of cold working in the production
of the foils. Annealing (No. 3), however, causes only a
small increase in the thermal coefficient. Its values
(5 x 0- /°C and 6 x 10- 3 /oC, respectively) are in agreement
with some published data for bulk molybdenum, which cover
fairly wide ranges of values. Examples from the literature
are 3.3 x 10-3/UC (250C) and 4.8 x i0- 3 /°C (1000oC) in the
Handbook of Chemistry and Physics, 44th Edition 1961;
4.4 x 0-J7OC (727OC) in W. H. Kohl, Materials and TechniqueE
for Electron Tubes, 1960. Kinks in the curves with a
steeper slope in the higher temperature range indicate an
increase of the averaged coefficient with temperature.
Curves 26 and 2? were obtained for evaporated molybdenum
coatings on Body E ceramics at a substrate temperature of
about 900 0 C.
Sample No. 27 was given an additional firing cycle
of 1/2 hour at 1425 0 C in wet hydrogen-nitrogen atmosphere.
Both thei:mal coefficients are lower than for molybdenum
foil, but there is no appreciable difference between them.
Metallurgical cross--sections in Fig. 6.12a and b indicate
no visible interaction with the ceramic, even at the high
firing temperature. Resistivities and resistances per square
of the evaporated coatings are in the intermediate range
between solid molybdenum and molybdenum metallizing paint.
Firing of the evaporated molybdenum layer reduces the
resistivity by a factor of two, thus indicating that the
electrical properties are atomic and microstructure structure
dependent, e.g. oxygen dissolved interstitially in the
molybdenum lattice causes a change in resistivity.*
The large decrease in resistivity is probably due
to three :'actors:
1. Oxygen dissolved in the lattice.
2. MoO x in the grain boundaries.
3. Interaction of Mo with the A120 3 and grain
boundary phases.
*A study of oxygen solubility and diffusion in presently

underway in the laboratory. Oxygen solubility is being
determined in terms of dc electrical resistivity.
NASA Contract No. NAS3-7626.
274
.9. - - _ __--
In accordance with equation (5) above, the definitions of
the thermal coefficients of electrical resistivity of the
total metallizing, the molybdenum and the melt phase con-
tained in it (o(, C>m , O(g, respectively) re~d as follows:
R/Ro = o(T + 1 (7)

Rm/Rmo = OmT + 1 (8)
Rg/Rgo =(gT+ (9)

The coefficients O( and o( have been measured directly,
while the coefficient c,( Tor the melt phase is to be
determined.
Writing equation (6) for the temperatures T and
T = 0°C one obtains for the total resistance ratio:
R/Ro - (Rm/Rmo) (Rg/o) (1I + o/Rmo) (10)

Rm/Rmo + Rgo/Rmo * Rg/Rgo
or expressed in terms of the thermal coefficients and the

temperature T:
+ 1) ( gT + I) • (1 + Rgo/Rmg ) (1)
O T + I (°(rT
(o<mT + l) + Rgo/Rmo . (0(gT + l)
this equation is to be solved for 0( the thermal coefficient

of electrical resistivity of the melt phase in the metalliz-
ing. There is no way of knowing the resistance ratio Ro/N o
of the melt phase and the molybdenum at 0°C. Therefore,
values are assumed for this ratio covering a wide range,
and values for C~g are calculated with the help of equation (11)
setting the temperature T a 4000 C.
Table 6-3 below shows the computed thermal
coefficients of electrical resistivity of the melt phase in
metallizing for several assumed resistance ratios at 0°C.
Naturally, the thermal coefficient for the glass in the
metallizing comes out to be negative. It is surprisingly
little dependent on the resistance ratio at zero degrees
centigrade.
Curve 19 in Fig. 6.11 indicates a negative

thermal coefficient of -1.2 x 10-3 in the temperature range
of 00 C to 175 0 C for the heavily leached Ti-ll metallizing
and confirms the above concept. It falls into the range
shown feasible in Table 6-3 on the following page.
-75
TABLE 6-3
Computed Thermal Coefficients of E'ecLrical Resistivity
Thermal coefficient of
Resistance Ratio electrical resistivity
Rgo/Rmo of the melt phase, 0(9
10 - 1.03 x 10- 3 /oC
20 - 1.65
50 - 2.13
100 - 2.28
200 - 2.40
500 - 2.46
276
- Am
6.2.2 Effect of Wetting Layer (Overplate) on the
Resistivity of the Metallizing
6.2.2.1 Influence of rf Conduction Lossehs

When the metallizing layer is thin, some of the
rf current will be carried in the overplate and braze
regions; the composition of the overplate is therefore
important. Also sintering of the overplate will cause a
diffusion alloying of the nickel (or copper, or iron) over-
plate into the metallizing with a consequent increase in
resistivity.
Accordingly, these effects were studied in
simple experiments with prolonged and repeated sinter firing
of nickel ane copper plate on a large number of metallized
ceramic rings.
6.2.2.2 Experimental
Standard ceramic rings (2.356 in. O.D.,
1.990 in. I.D., 0.075 in. thick) of Body A were metallized
on one side with P-I metallizing paint and fired at 1425 0 C.
A radial cut produced resistors with a length of 17.5 cm
and a width of 0.51 cm (17.5 " 0.51 = 34.4 squares*).
DC resistances were measured with a bridge and copper probes.
Contact resistances amounted to 0.008 ohms and had to be
considered only on copper plated samples with resistances
as low as 0.123 ohms.
Four groups of four samples each were tested.
Table 6-4 below lists data of one representative sample of
each group. Thicknesses of the metallizing and thick plating
layers were determined from metallurgical cross-sections of
the samples. Irregul-rities were large and, therefore, these
numbers are only for reference, as are the caicul&ted
resistivities. All values were obtained before the sinter
firing of the overplates.
Resistances of the rings were meAsured after

metallizing them. These were then nickel or copper plated
and remeasured. After each -inter firing step. r,sistancez
were again measured. There were four sinter firings at
*Dimensionless ratio of length to width of resistor which is

used in the definition of surface resistivity (ohms/square).
277
TABLE 6-4
Thicknesses of Metallizing and Plating on Representative Test Samples.
i4easured Resistances, Surface Resi3tances and
Deducted Resistances
Sample of Group No. : 1 2 3 4

Thicknesses of
metallizing (mils) 0.3 0.3 0.6 0.5
Resistance (ohm) 1.95 1.67 0.99 0.99
Resistivity of
metallizing (ohm cm) 42xi0 - 6 37xi0 - 6 43xi0 - 6 36xi0 - 6
Surface resistance 5.6x0 -2 5x10- 2 2.9x10- 2 2.8xi0-2

of metallizing
(ohms per square)
Plating nickel nickel nickel copper
Thickness of
plating (mils) 0.07 0.2 0.07 0.3
Total resistance 1.52 0.43 0.76 0.15
(ohms)
Combined resistivity 41.5xi0 - 6 16x10- 6 38xi0 - 6 8.9xl0 - 6
(ohm cm)
Combined surface 4.2xI0 - 2 i.3x10- 2 2.2xi0 - 2 0.4x10- 2
resistance (ohms
per square)
278
"" -- -. ..-
' " .. ~~~~~~~~-
- ----
,t' ~ lI tg e , .,4 ", lle ,
900 0 C for 1/2 hour each, two firings at 900 0 C for 1 hour
followed by three firings at 1000 0 C for 1 hour each, all in
a wet hydrogen atmosphere.
The experiments were designed to study reaction

mechanisms. It is rather unusual for metallized and nickel
plated ceramics in electron tube structures or in micro-
electronic applications to go through extended high tempera-
ture treatment.
6.2.2.3 Results
Figs. 6.15 through 6.18 show resistances of test

samples measured after each processing step starting with
the metallizing and the overplating. The relative resistances
which are plotted are referred to the initial resistances of
the metallized ceramic rings. See the above Table 6-4 for
absolute values.
All curves in Figs. 6.15, 6.16 and 6.17 for
nickel plated samples show a more or less pronounced mini-
mum. The lowest minimum in Fig. 6.15 was, of course,
obtained with the relatively thickest nickel plating. The
resistance went down slightly after the first sintering
step as a result of the consolidation of the plated layer.
The next firing caused a sharp increase of the resistance
which indicated strong interaction between the nickel plating
and metallizing. Note that further firings at 900 0 C lasted
for one hour. After the fifth step at 900 C, the curves
level off, indicating that equilibrium had almost been
reached, A higher firing temperature of 1000 0 C in the
following firings lead again to an increase of resistance
and to a final leveling off of the measured resistance.
The maximum increase of measured resistances was to 195%
referred to the resistances of the metallized ceramic. It
occurred on the sample with the thickest nickel plating which
also had the lowest minimum. The conclusion to be drawn
from this fact is that the reaction was limited by the amount
of nickel available.
Fig. 6.16 confirms the latter statement. As

can be seen from the relatively lower minima, a much thicker
nickel plating was applied. Again, resistances climbed
after the second firing steps. The curves again level off
after the series of 900 0 C firing steps and after the three
subsequent 1000 0 C firings. Referred to the metallizing
resistances, the increases are not as high as in the previous
group of samples, but actually the alloying of nickel and
molybdenum, which leads to a high resistivity coating is
279
2.0 .___
.......... _ __ _ _ _
U)1.0
In
wr
W
>z
0.
0-
METAL N.0 9000_ 9000C 900_00_900_0__IOOC IOOC IO
IZN LT / R 2R. 12H. 12H. IM. R R R R
PRCSIN TP
Fiue61:Itrcino iwetn ae ihP1m tliin nBd ea i.Rltv
reitnea ucino rcsigstp.Dt bandfo orsm lso
gru 1,Tbe64
2.0 -
./
.//
1.0
LJ
so p_ loop_
- _ _
0.5
METAL- Ni 9000C 9000C 9000C 9000C 9000C 900°C IO0O0C 10000C IOOOOC
IZlNG PLATE 1/2 HR. 117 MR. 1/2 HR. 1/2 HR. I HR. I HR. I HR. I MR. I HR.
PROCESSING STEPS
Figure 6.16: Interaction of Ni wetting layer with P-1 metallizing on Body A. Relative resistance as a
function of processing steps. Data obtained from four samples of Group 2, Table 64.
-281-
2.0
1.5
_ -4.
-
U +
0.5
0I
METAL- Ni 9000 C 9000C 9000C 9000C 9000 C 9000C lO00°C IO00'C O00OC
IZING PLATE 1/2 HR. 1/2 HR. 1/2 HR. 1/2 HR. I HR. I HR. I HR. I HR. I HR.
PROCESSING STEPS
Figure 6.17: Interaction of Ni wetting layer with P-1 metallizing on Body A. Relative resistance as a
function of processing steps. Data obtained from four samples of Group 3, Table 6-4.
- 282 -
1.0
to
S0.5 -
U)
U)
Mh
WA
METAL- CU 9000C 9000 C

IZING PLATE 112 HR. 1/2 MR.
PROCESSING STEPS
Figure 6.18: Interaction of Cu wetting layer with P-i metallizing on Body A ceramic. Relative
resistance as a function of processing steps.
- 283 -
7.0
6.0 -6.0
K - x (2)
5.0-
4.0 " XZ -4.0

* K"
Ux
"1
Cu plate
Ir3.0 -
x+--- +1)
2.0- K' +.00' -2.0
.0 0 - 0 135
S-0
1
1.0
-- A (4)
Cu plate
0 ¢
meoll-
me60D
NIcNI I 9600~ 9000oo 9600 10000 100 tooVC
ilng plate hr '
r ;Lh Ih ,hr lhr lhr ,hw ,h,
Processing Steps
Figure 6.19: Relative resistance of metallized and over-plated Body A substrates as a function of
sinterfiring. () 0.3 mil metallizing, 0.07 mil Ni-plate; (2) 0.3 mil metallizing, 0.02 mil
Ni-plate; (3)0.6 mil metallizing, 0.07 mil Ni-plate; (4) 0.5 mil metallizing, 0.3 mil Cu-plate.
284 --..
more obvious if one considers the increase from the minimum.
It amounts to a factor of 7.6, as compared with the
corresponding increase by a factor of 3.1 in the previous
group of samples with thin nickel plating. Again the
larger amount of nickel available provided for the more
pronounced interaction leading to the high resistance.
Results with the next group of samples,
relatively thick metallizing, and thin nickel plating,
again proved the same point. Increases of resistances were
relatively low, the thickest nickel plating causing the
largest increase.
Finally, metallized ceramic rings were over-

.0 plated with copper and sinter fired. See Fig. 6.18 for the
resulting resistances. As expected, the resistances were
initially decreasing and constant after the second sinter
firing. There was no evidence fron the resistance data of
an interaction between copper and molybdenum. From the
data of Table 6-4, column 4, the resistivity of copper in
ohm cm was calculated and found to be 4 x 10-6 fl cm,
which is twice the reported value for that of bulk OFHC
copper.
Data in Figs. 6.15 through 6.18 are summarized
in Fig. 6.19 for one representative sample out of each
.0 group. The scale on the abscissa has been adjusted to
distinguish between the half-hour and the one-hour firing
steps. The break in the curves indicates the switch over
to 1000 0 C sintering. Relative resistances are now referred
to the overplated samples. This way, increases of resistances
show up more clearly; especially curve (2) stands out.
6.3 rf CONDUCTION LOSS MEASUREMENTS
6.3.1 Introduction
The rf losses in ceramic-metal als, since they
will be due to currents in a confined area near the metal
wall, can be described by an equivalent surface resistivity.
By placing a sample of metallized ceramic inside a resonant
cavity, one can measure the effective surface resistivity of
the interface in several ways. One method is to measure the
change of a property associated with the cavity such as Q,
input impedance or power necessary to excite the cavity to a
given level. The cavity can then be calibrated against a
material of known resistivity. A more sensitive method of
measuring these properties is the bridge technique wherein
one can measure the deviation of some property of the cavity
285
I
containing the test ceramic from that of a standard cavity.

The same quantities can be measured by either technique.
An example of each method is found in the
literature. The first method was used by Dehn and Close1
in which they measured the Q of a test cavity containing the
material to be tested. The cavity used was a right circular
I
cylinder operated in the TE11 1 mode. The ceramic samples
were brazed to the interior surface of the removable end
plates of the cavity. The samples were placed near a region
of high magnetic field so as to insure maximum currents
across the ceramic-metal interface and hence
A second requirement on the sample placement maximum loss.
was that it be !
in a region of low electric field.
By noting the frequency of the half-power points

on the response curve of the cavity, the Q may be readily
j
calculated from Af3db - 1
fo Q
where &f3db is the difference in frequency between the half
power points and fo is the resonant frequency of the cavity.
A curve of cavity Q versus end plate resistivity was
calculated and compared with Q measurements with end walls
of known resistivity (materials such as Pb, Sn, Brass).
By the use of this curve a measured Q of an unknown sample
can be readily converted to an equivalent surface resistivity.
The bridge method was used by Ernst and Purl 2 . I
Two right circular cylindrical cavities were used in the
TM0 2 0 mode. The fields are such that a ring shaped sample
with one flat side (the interface) placed against the end of
the cavity can be located in a region of high magnetic field
I
and low electric field. The sample is placed in one cavity
(the test cavity). A lossy material is introduced into the
second cavity and is adjusted by a micrometer screw so that
the loss inserted is some function of micrometer reading.
The two cavities are connected to the two equal arms of a
Magic Tee. When power is red to the third arm, the dif'er-
ence in the input admittance of the two cavities relative
I
to each other is related to the amount of power present at
the fourth arm of the Magic Tee. When the admittances are I
equal, no power appears in the fourth arm. The power out of
arm four is proportional to the square of the difference of
cavity admittances. As in the work of Dehn and Clos,' the
system is calibrated by utilizing samples of known resistivity
and plotting their values versus the loss-inserting-micrometer
reading. When an unknown sample is tested, its effective
interface resistivity is read from the calibration curve.
206
- - - -v
6.3.2 Selected Measurement Technique
In this present research, for the evaluation of

losses in ceramic-to-metal seals due to conduction currents,
an admittance bridge was used similar to that of Ernst and
Purl 2 . By placing a sample of metallized ceramic in a
resonator in a region of high magnetic field and low elec-
tric field, maximum conduction currents will flow along the
seal area while minimum displacement currents will be
present in the ceramic; hence conduction losses will pre-
dominate. The effect of these losses will be to lower the
resonator Q. Looking into a resonator from the input
waveguide, the incident wave sees infinite admittance
except at the resonant frequency. Near resonance the
input admittance to the resonator drops until at resonance
it reaches some low value. The resonant input admittance
of the resonator is a function of the resonator 0, so that
a change in resonator Q can be measured by noting the
change in resonator input admittance.
As noted previously, the admittance bridge

method used in this work compares the relative admittance
of a sample resonator with a standard resonator by utilizing
a Magic Tee. The Magic Tee has the property that if power
is fed into arm three, the power flowing in the fourth arm
is proportional to the square of the difference of the
admittances presented to its two equal arms. Hence two
resonators, identical in all respects, connected to a
Magic Tee's equal arms. at all frequencies will cause zero
power to flow in arm four if powe4 is applied to arm three.
Now, if a small loss is introduced into one of the resonators,
at the resonant frequency, the input admittancen will no
longer be equal. Power will flow in the fourth arm which
is proportional to the square of the difference in input
admittance of the two resonators. If an equal loss is
inserted in the second resonator, the bridge will be brought
back to a balance condition and hence no power will flow
through the fourth arm. By calibrating the loss introduced
into the second resonator, the relative lossiness of the
first resonator can be determined.
6.3.3 Sample Preparation
Three series of seal samples were prepared during

the program.
The initial test sample consisted of a plated

metallized ceramic ring brazed to a 10 mil cupro-nickel disc
which in turn was tin- %'o a copper sample holder. All the
287
exposed metal surfaces were silver plated in order to
minimize rf losses (except in the seal area). Because
the 10 mil disc warped, it was replaced by a 20 mil disc
which utilized a back-up wafer seal as shown in Fig. 6.20.
General details of the seal construction follow.
Specific seal construction is discussed later.
6.3.3.1 The Ceramic
The samples were fabricated from Lucalox* or
Body A in the form of a flat ring 1.207" O.D. x 0.703" I.D.
x 0.100" high.
6.3.3.2 The Metallizing
The metallizing paint compositions P-1 and P-2
were used. The latter paint was sintered at 1550 0 C as well
as 1425 0 C. A spray application gave the most u.Aiform
coating thickness** and was utilized. Copper, nickel,
titanium and molybdenum were also evaporated onto ceramics
to give metallizing films in the range 1-10 microns.
Non-metallic coatings of MnO-TiO 2 were al=z fired onto
substrates.
The amount of coating on actual parts could be

determined by weighing and by measuring with a micrometer
after spraying. In addition, test pieces sprayed at the same
time were cross-sectioned after sinter firing and examined
metallurgically to measure sintered coating thickness and
to detect any thinning at the edges.
The thickness of the metallizing was in general
determined from the calibration curve, Fig. 4.48. This
calibration curve was drawn from data obtained by plotting
the unfired sprayed weight of P-l paint applied to 1" discs
of Body A versus the cross-sectioned and microscopically
measured thickness of the same discs after sinter firing.
6.3.3.3 The Plating
Copper, iron and nickel plating of the metallizing
were evaluated with selected brazing alloys. The copper was
*General Electric Co. trade name.

**Film metallizing would also be satisfactory if the film
thickness could be controlled.
286
Metal izin g
Cu or Ni plate (.002")
Fe plate (.0005")
Cu Ag brodc
.010 Cu. Ni Disc
Standard seal
Bock up ceramic
Body A Ceramic
_ Test Seal (see detail)
Sock up ceramic ring

""b Sample Holder
1462 DIA,
Figure 6.207 6 conductivity test assrombly. Showing layered structure of an iron barrier layer seal and
-Sowing back up ceramic used to rec4uce seal stresses.
289
plated to 1 2 A minimum thickness and the nickel to 25,#.
The nickel was plated and sintered in two "build up" stages.
The plated parts were sintered at 800 0 C or 900 0 C for 1/2 hr.
in hydrogen.
The iron layer was plated on top of either the

copper or nickel layer to a minimum thickness of 12#i and
sintered. It was brazed unsintered in the second series of
tests in order to minimize interdiffusion of the Elated
layers. In a third series it was sintered at 800 C and a
nickel (unsintered) overplate then plated on top. A visual
inspection for blistered plating was made after sintering.
6.3.3.4 The Brazed Assembly
In the first series of tests nickel and copper

plated samples were brazed to 70/30 cupro-nickel discs with
copper braze.
Nickel or copper plated and subsequently iron-

plated samples were brazed with Ag-Cu braze in series two.
An additional nickel plate was inserted in series three.
presentative production type seals were made

such as nickel plated metallizing, Ag-Cu brazed to nickel
plated Kovar; Pnd copper plated metallizing Au-Cu brazed to
copper. Special evaporated metallizing seals were also
brazed.
A direct transition, copper-to-ceramic, seal was

fabricated. This represented an optimized low loss situation.
The ceramic was first electroless copper plated* or vapor
plated just enough to give a conductive surface, then electro-
plated to over 40,a. The strength of the copper-ceramic bond
was quite weak (20 psi) so the plated ceramics in this
instance were simply tinned to the sample holder for the
seal rf conductivity test.
The brazed subassemblies of the type described

were tinned to a solid copper test piece base which is part
of the rf test cavity. Before placing in the cavity, the
assemblies were cleaned and silver plated. The silver was
plated to about 2 0# thickness which is well in excess of the
*Enthone process using "Enplate Cu-400" with Sn activator.
AG90
I
skin depth at test conditions. The silver covered the tinned

joints and covered any material which may have blushed out
during brazing. A cross- ection of a typical assembly was
shown in Fig. 6.20.
6.3.3.5 The Standard Assembly
The standard assembly consisted of a clean

--ramicBody A,glued with a minimum amount of low loss
pulystyrene cement to a gold-plated copper cup holder.
Several standard assemblies were constructed for

the purpose of placing them in the test resonator and the
standard resonator to obtain an initial balance on the
cavities. They were also used in the standard resonator in
the loss measurements to balance out the bulk ceramic effects
in the two cavities.
6.3.4 Equipment and Experimental Procedures
6.3.4.1 Assembly of Apparatus
The block diagram of the test setup used in the

first six quarters is shown in Fig. 6.21. The general
layout of the equipment is shown in Figs. 6.22 and 6.23.
For the third and last series of tests a simplification was
introduced in the test setup by using a Hewlett-Packard
716A power supply for the reflex klystron. This replaced
the repeller supply, beam supply, heater supply, and
sawtooth generator shown in the block diagram of Fig. 6.21.
A cross-sectional drawing of the tunable resonator

which contains the ceramic-metal sample to be tested is shown
in Fig. 6.24. The entire assembly is constructed of OFHC
copper stock. The irner surfaces were fine machined,
polished and then chemically polished to produce as smooth
a surface as possible. The base and top ring were plated
with silver on their common surfaces and hydrogen brazed
together. Visual inspection of the interior of the joint
showed good contact and very little blushing of the silver.
Simultaneous to this operation, a short section of waveguiC:
was silver brazed to the slot milled in the side of the
resonator.
The connection between the resonator body and

the removable top utilizes an rf choke to minimize losses
due to poor electrical contacts and to reduce variations in
readings due to changes in electrical contact. When
machining the top, it was found impossible to turn a good
291
a
0
-1
00
0I I
4d _g 0
W 0
N
000
0 43
-hi-j I-00
la 90-
dCC
CDC
w0 I-
W W
04
I--0hiI
co E
C
0
I- w x cc
J
II IL4
It w 0)
o N c0
m 00cq
k 4L 4 4
04 0. I
40
w
01 x
w f w
J J J~h
1~
U
E
C
1~
0
Li
In
C
C
0
In
4)
I-
U)
£
U
'a
E
C
C
I-
U
0
-c
an
4)
0)
-o
CC
'4.
'4-
o *~
a-In
oE
-C
0
I-
0)
- 293 -
1-Z
294-
w
ci)
w
0
ww
U F
w wI
U)U w w
0 co
z LAJ 0
I4-
o 4 4 0 o U)
44
U))
U) D
CDC
o
z4
I.-29-
choke slot in the top, so the center of the top was made in
the form of a plug, the thickness of which was equal to the
slot depth. It was silver brazed in place and note was
taken of a small fillet formed in the bottom of the groove.
A number of the test piece bases (showr. as

number one in Fig. 6.24) were constructed so that several
ceramic-metal samples could be tested within a short time.
A brazed ceramic-metal assembly was tinned to

the test piece base (as discussed previously) and the
assembly was placed in the cavity. The need for a separate
test piece base for each specimen is not necessary as the
discs from one batch of samples may be un-tinned and new
ones tinned in place.
The standard resonator was constructed nearly

identically to the test resonator. The tuning diaphragm
was omitted and a 0.125-inch hole drilled in the side 900
from the coupling hole. A micrometer screw was bracketed
in front of the hole so that the penetration into the
resonator of a lossy rod could be measured accurately.
This rod provided the loss in the standard resonator with
which to balance the seal loss in the test resonator.
The interior of the tunable resonator and the
standard resonator were gold plated.
Well into the program, when the third test
series samples were mounted in the tunable test resonator,
it was found that they protruded 0.015-inch on the average,
deeper into the test cavity than earlier samples had.
Consequently, it was found that the resonance frequencies
were on average 120 Mc lower than the resonance frequency of
the variable loss standard cavity (9200 Mc). The admittance
bridge measurement, however, requires identical coupling and
identical resonance frequencies of the test and standard
cavities in order to balance the conduction losses of the
test samples with the artificially introduced losses in the
standard cavity. The corresponding change of resonant
frequency had to be compensated for by retuning the standard
resonator.
The retuning required an excessive deflection of
the 20 mil thick cavity bottom. Consequently, the copper
wall of the cavity, already work hardened from previous
tuning cycles, tore after a few measurements. It was then
replaced by a copper disc whch, unfortunately, withstood
only a few tuning attempts before it broke also.
296
A new cavity bottom was then made. The retuned
cavity caused a change in the reference point for the lossy
rod penetrating into the standard resonator. In order to
tie later measurements as closely as possible to the
previous ones, the conduction loss in the ceramic-metal
seals was always compared against the old "#4 gold standard",
a clean ceramic glued with a minimum of low loss cement to
a gold plated cup holder. Remeasuring the samples #83 and
#87 (from the first series of tests) showed that the new loss
indication was, on the average, lower by 22 units.
6.3.4.2 Determination of the Cavity Q
The Q of the resonator was determined by measuring

the VSWR of the line terminated by the resonator. The
Eimac 1075 reflex klystron used in the bridge measurements
was also used here. It was necessary to square wave modu-
late the tube as the VSWR indicator is only sensitive to a
square wave modulated signal. The modulation was achieved
by superimposing a square wave on the dc repeller voltage.
It was possible to drive the tube out of oscillation during
h-.if the cycle and into oscillation during the other half
of the squere wave cycle, hence square wave amplitude
modulating the tube. Since the frequency of oscillation
of the klystron is a function of the repeller voltage, it
is imperative that the tops of the square waves be flat.
A slopL of a few millivolts in the flat portion will modulate
the frequency enough to throw doubts on the validity of any
measurements taken. By carefully choosing the klystron
operating parameters, a frequency deviation of less than
two hundred kilocycles at 9.5 Gc was obtained. This was
deemed sufficient for rough measurements of Q. With the
resonator in place, a plot was made of standing-wave ratio
as a function of frequency. From this plot the unloaded Q
may be determined as outlined by Ginzton 3 . The unloaded Q
of the test resonator (with chemically polisihed inner
surfaces) was measured as 13,700.
It was noticei early in the program that the
copper interiors of the cavities became tarnished several
days after chemical cleaning. In the interest of long term
stability, the resonators were given a thin coating of gold
plate. This would stabilize the Q of the resonators although
lowering it slightly. Upon measurement, the internal Q of the
test resonator was found to have dropped to 11,200. The Q
of the gold plated standard resonator was also measured and
was found to be 12,700 with the effect of the adjustable
loss entirely removed. Although the Q's were lowered appreci-
ably, the resonators were left as they were as results of
297
measurements on ceramic samples determined that the change
in Q was not significant.
6.3.4.3 Operation of rf Admittance Bridge
In order to obtain a complete null condition as

the frequency is swept across the resonant frequency of the
system, the characteristics of each equal arm of the Magic
Tee must be identical. The lengths of the wveguide sections
connecting the Magic Tee to the resonators mu3t be equal as
must the coupling to the resonators. The resonant frequency
and internal Q of the resonators must also be equal. Even
if the admittances presented to the equal arms of ts Tee
are equal, the power flowing from the fourth arm may not be
zero 4 unless the Tee has been matched at the ft quency of
interest.
The above conditions were met as follows: T .e

Magic Tee was ordered from the manufacturer to 'e .,atched
at 9.5 Gc. The test resonator was tuned so that the
resonant frequencif-s of the two cavities coiicide%2,and the
Q of the standard resonator was adjustable so that the Q's
were equal. The coucling of both resonators was adjustable,
so that both couplings were made equal. The lengths of the
waveguide sections varied by less than O.L35 inch. This
had some effect on the completeness of the null because
difficulty was encountered in obtaining a complete null in
the actual setup. However, the accuracy of the measurements
was not significantly affected as will be seen later.
The accuracy of the measurements increases as the

bridge becomes more unbalanced, since the rate of change of
power out of the fourth arm with respect to the difference
in input admittances is proportional to the difference in
input admittances.
The output of the Magic Tee waE detected and
applied to the vertical input of an oscilloscope. As the
bridge became more unbalanced, the oscilloscope vertical
deflection increased. Hence, either a dc blocking voltage
had to be applied, or the vertical amplifier of the oscillo-
scope had to be switched to a less sensitive range. The
first possibility would have added another source of error
and another adjustment to make. The latter suggestion
decreased bridge sensitivity by decreasing change in oscollo-
scope deflection with change in resonator loss. In practice
then, a compromise had to be reLnhed. The reference
unbalance point was adjusted sucl that a change in loss
micrometer setting of one Civision resulted in the greatest
298
change in oscilloscope deflection, 0.4 cm deflection out of
8.0 cm in this case. In the measurements, then, loss was
inserted in the standard resonator until a prescribed un-
balance was reached. The driving power was set to a fixed
reference level at all times.
As originally designed, the loss in the standard

resonator was effected by a one-eighth inch diameter rod of
carbon, inserted in a hole in the side of the resonator.
The depth of penetration was controlled and measured by a
micrometer screw. It was determined that the carbon was not
lossy enough and that it also detuned the resonator. The
carbon rod was replaced by a lucite tube in which a small
roll of lossy paper had been inserted. The loss increased,
but the loss probe insertion necessary was still greater
than was desirable. The hole was increased in size to
0.200 inch and a correspondingly larger lucite tube pre-
pared. The probe insertion necessary to produce adequate
loss was reduced to about 0.100 inch. The lossl paper was
sealed in with plastic cement to prevent changes due to
humidity.
The resonators were originally designed to

resonate at 9.500 Gc. As constructed, they resonated at a
few megacycles above that figure. The detuning effect of
the ceramic test pieces was sizable, bringing the resonant
frequency near 9.200 Gc. The samples had to be prepared
from ceramics of uniform dimensions and were mounted
uniformly from piece to piece as the tunable resonator had
a range of only 15 megacycles. The bridge was assembled
and power applied. The test resonator was adjusted to an
undercoupled state with a bare ceramic in place.
The standard resonator also had a bare ceramic

in place. The standard resonator coupling was adjusted as
follows: The tuning of the test resonator was adjusted
along with the loss and the coupling of the standard
resonator until a null was observed on the oscilloscope.
At this point the coupling of the two resonators was assumed
to be equal.
6.3.4.4 Measurement Procedure
The prepared sample was inserted in the test

resonator. An identical ceramic ring was inserted in the
standard resonator. The test resonator was tuned to
resonance as indicated by a minimum oscilloscope deflection.
The power from the klystron was measured by viewing the
output of a directional coupler on an oscilloscope (see
299
Fig. 6.21) and adjusted to give a pre-determined deflection.
The klystron was adjusted so that the mode peak occurred at
the resonant frequency of the resonators. The micrometer
screw was then adjusted to give a predetermined output from
the output arm of the Tee. After each adjustment of the
micrometer, the tuning was checked to bring both resonators
to the same frequency. After this iterative series of
adjustments was made, the incident power was again checked.
The micrometer reading was then recorded and another sample
inserted.
6.3.5 Equipment Checkout
6.3.5.1 Reproducibility as a Function of Time
Upon completion of the bridge apparatus, during
the second quarter, sample holders without ceramics were
constructed of several materials, the object being to check
the sensitivity of the apparatus. Two gold plated holders
were made as well as a chrome plated one and one with a disc
of 70/30 copper-nickel alloy soldered to the surface of the
copper holder. With a gold holder in the standard resonator
and a gold holder in the test resonator, the bridge was
brought to balance (i.e. to the calibrated off-balance
condition). The other two samples were compared with the
gold one. The chrome plated sample was found to require
an additional insertion of the loss slug of 0.0225 inches.
(Henceforth the insertion will be given in divisions, where
one division equals 0.001 inch. So the above would be 22.5
divisions.) The copper-nickel alloy sample required 31.8
divisions. By making measurements several times over a
space of an afternoon, it was found that the measurements
were repeatable to less than one division accuracy.
Later, at the start of the series II tsts in
the fourth quarter, a somewhat similar program was carried
out. This is reported next.
Six "standards" with a range of resistivities
between 1.4 and 81 ohm centimeters were made in which the
#seal" area consisted of the material noted in Table 6-5.
The balance of the assembly was silver plated and the ceramic
member glued over the seal area with three small spots of
low loss polystyrene* cement. These standards were measured
*Low loss polystyrene Q - Dope (General Electric Co.).
300
TAB LE 6-5
Conduction Loss Data at 9.15 Gc on Body A Test Pieces
Glued to Noted Sample Holder
SAMPLE HOLDER
Electro- 20 mil Micrometer
Surface Bulk plate Tinned Disc Reading Loss
Cu Cu - 646 -4 + +
Ag Cu yes - -19.2
Au Cu yes 625.5 0
Cr Cu yes - 583.3 43.5
CuNi Cu - 70/i0 Cu-Ni 527.3 99.5
Kovar Cu - Kovar 466.11 160.1
Invar Cu Invar 396.8 230.0
++Obtained during the seienth quarter.
in the same manner as brazed assemblies, being compared to

the gold reference assembly. The silver standard had a lower
resistivity and lower loss than the gold reference, so the
A Ag is of the opposite sign from all the other measure-
ments given in Table 6-5.
Using published resistivity values for the
standards, a plot of resistivity versus micrometer reading
is given in Fig. 6.25. Three actual seals subseq-ently
reported are also plotted on Fig. 6.25. From these figures
the sensitivity of the test procedure is seen to be quite
adequate for the present work. It may be noted that the
derived resistivity of the seals. 20-40 x 10-6 ohm-cm, is
in the same range as the dc measurements reported in
Table 6-1, showing excellent correlation of the dc and rf
resistivity measurements.
301
60 lnvor\®
\ r?
60
2 1
XC r Kovor ®
E
--,40
g*92
70-30 CuNi ®
1 *71
20 h85
71Body A, P-1, Ni Plate Cr
85 Bdy A, P-2 (I550),Cu Plate Cu evaporated
92 Body A ,P-1, Cu Plate
400 500 .600

Micrometer Reading (in.)
Figure 6.25: Plot of micrometer reading versus published resistivity of the six calibration assemblies
(See text). Selected data from actual seals included for reference.
..- 302--
6.3.5.2 Standard Test Piecr,
The standard, to which all test samples consistently

were compared, is a raw ceramic of the composition used in
the test piece. Two such standards were prepared. The two
gold-plated sample holders were used, as the gold is insensi-
tive to oxidation effects. Samples of Body A were glued in
place on the sample holder by very small amounts of poly-
styrene cement. Centering was goo(- but not absolute. The
two standards were designated numbers 0 and 00. If the two
standards are identical, then one could expect that a standard
made from any of the bare sample ceramics should have identi-
cal loss properties as numbers 0 and 00. In effect then, a
check on the two standards should indicate if the bare
standard ceramic is cancelling out the bulk ceramic losses
present before metallizing. As it turned out, the plastic
cement used to bond the ceramic to the holder was the
dominating factor in any differences seen between samples 0
and 00.
6.3.5.3 Alignment of Ceramics
An investigation was made to determine the

importance of the test piece being concentric with the
center-line of the resonator. A test piece which has been
eccentrically mounted will exhibit varying degrees of loss
depending on its orientation with respect to the coupling
hole. Tests were run to determine whether the range of loss
encountered is a good indication of sample eccentricity.
The results of this test show that eccentricity can be
estimated from loss data for deviations greater than 0.013
inches off center (Fig. 6.26). The loss varies approximately
3 divisions at this point. For smaller eccentricities,
apparently non-uniformities in the seal or ceramic over-
shadow the effects of centering.
6.3.5.4 Sensitivity of Micrometer Reading with Cavity Q
The micrometer insertion loss to give various

cavity Q's with the standard ceramic in place was determined.
These results are given in Fig. 6.27.
From this data the cavity Q associated with the

various samples listed in Section 6.3.5.1 can be found, as
also can the Q for various ceramic-to-metal seal test
assemblies mentioned in Section 6.3.5.1 (and reported later
in Table 6-11).
It can be seen that the Q of the standard cavity of

11,000 is significantly reduced to 4,000 by the emplacement
of even the low loss silver standard with its attached ceramic.
303
0
C- u
U ~u0
4- 4
CC
E0
0
00
OC
4E E
C.
c7c
Id 0uC
Q 10
- c..
0
E -c
o t
- U)
00 to
304-
5000
4000 2
~Au
3000 - 8
2001 2 l
oo
900
700
600 -
500
400-
300
TEST FREQUENCY 9 15-9 16 Gc
I Silver Standard
2 Copper Nickel (70-30) Standard
200 3. Invor Standard
71 Body A,P-lNI Plate
g92Body A,P-ICu Plate
85 Body A,P-2(1550),Cu Plate
100 i i 1
.700 o600 .500 .400
Micrometer Reading (in.)
Figure 6.27: Q. of standard cavity used in conduction loss measurements plotted as a function of
micrometer reading. Selected data giving the various Q's of the test cavity with var-
ious assemblies including the standard are included for reference.
- 3U5 -
6.3.6 Results
The loss measurements were conducted in three

separate series during the second and third quarters, the
fourth quarter and the sixth quarter respectively, the
results are therefore reported in this sequence.
6.3.6.1 First Series
The initial series of seals consisted of Body A,

P-1 or P-2 metallizing of various thicknesses - sintered
nickel or copper plate - copper braze to cupro-nickel metal
member. As a reference check, electroless copper plate on
Body A with a tin activated surface was tin soldered to a
copper holder - it showed a loss of 48.2 units, sample #60.
The collected data on the tested seals are given

in Table 6-6. The reference specimens were checked at the
beginning and end of the measurement run which was made in a
single 10-hour period. A shift of 3.7 divisions was noted,
which is well within experimental error.
6.3.6.2 Second Series
In order to reduce the I2 R losses exhibited by

the seals in the first series,, a second test series was pro-
grammed. The results are given in Table 6-7. An unsintered
iron barrier layer was used to isolate the copper or nickel
plate from the braze. In this case Cu-Ag braze (M.P. 7800C)
was used. The thickness of the metallizing and plate was
such that the rf currents were carried only in these layers.
An evaporated copper seal tinned to a copper

sample holder was also cvaluated. A loss of 50 units was
obtained. These results are also given in Table 6-7.
The micrometer setting required for balance is

indicated in Tabhe 6-1. Tf more than one reading was taken,
the readings were averaged.
The gold plated reference standard designated Au

in Table 6-5 was measured at the beginning and end of the
test run which was completed in one working day. A drift
from 636.8 to 616.5 (micrometer setting) was noted. As
Fig. 6.27 shows, this is on the high Q end of the Q versus
micrometer reading curve. The slope of the curve in this
region is only about one-third that of the lower portion,
where most of the actual seals were measured. The draft
corresponded, therefore, in terms of precision to 6.7
306
TABLE 6-6
Series 1, Conduction Loss Data at 9.15 Gc on Body A
Test Pieces Sealed to Cupro-Nickel (70-30)
with Copper Braze
Sample Metallizing Thickness*

No. Paint in x 103 Plate+ Loss**
48 P-1 .14 Cu 69.6

44 " .18 Cu 77.1
47 It .26 Cu 79.3
3 .98 Cu 92.1
1 " .98 Cu 90.7
6 1.10 Cu 99.1
5 " 1.20 Cu 99.9
4 1.42 Cu 93.5
19 " 3.17 Cu 98.0
22 " 3.29 Cu 93.6
37 " .17 Ni 120.2

17 " 1.02 Ni 103.8
14 I 1.03 Ni 99.9
39 " 1.96 i I-,8
42 " 2.24 Ni 88.6
52 P-2 .17 Cu 81.2

51 " .19 Cu 97.6
54 " .20 Cu 148.5
53 " .23 Cu 184.6
55 .32 Cu 134.6
12 .61 Cu 83.9
35 " 1.07 Cu 108.3
27 1.12 Cu 136.6
32 1.14 Cu 148.0
58 .18 Ni 122.7
56 .20 Ni 134.6
57 .26 Ni 136.8
8 "86 Ni 146.6
jl .88 Ni 136.3
29 .91 Ni 132.3
30 .93 Ni 141.4
33 " 1.03 Ni 130.8
60 - Cu 48.2
*From calibration Curve #5 (see 3rd Quarterly Report).

**In .001 inches of lossy material required to balance
Impedance Bridge Test Circuit.
+Sintered at 8000 C.
307
TABLE 6-7
Series 2, Conduction Loss Data at 9.15 Gc on Body A
Test Pieces with Ag-Cu Braze Sealed to Cupro-Nickel (70/30)
Plate
Sinter
Sample Metallizing Thickness Temp.
No. Paint in x 10 3 Plate + (Xl00) Loss
64 P-I .12 Ni 8 111.9

65 " .20 " 8 103.1
63 " .22 " 8 102.5
66 " .25 8 109.2
71 " .76 9 100.0
81 " 1.04 I 8 79.7
78 I 1.25 t 8 71.5
67 U .21 Cu 9 101.3
91 I .22 t 9 85.7
92 " .83 i 9 114.1
93 1.46 8 116.3
69 " 1.93 9 120.4
96 P-2 .36 Cu 9 127.7
98 " .78 " 9 122.3
84 P-2(1550)* .52 " 8 83.8
85 11 .86 I 8 64.7
88 t 1.35 " 8 67.6
83 P-2(1550)* .56 Ni 8 72.4
86 " .94 " 8 68.8
87 " 1.56 I 8 65.8
103 Cu (evap) - Cu - 50.4
104 Cu (evap) - - 45.6
+0.002 rnjs thick

then iron plate
*Sintered at 1550 0 C then 1425 0 C.
308
divisions in the region of the measured seals. Cause for
the drift was not resolved. The column desiqnated "loss" is
the difference in micrometer readings between the test
sample being measured and the average of the readings on
the Au standard. This average 626.8 compares favorably with
the 625.5 average determined in a similar fashion with the
same gold reference during Series I (Section 6.3.5.1). The
data for the two runs is therefore comparable with no
further adjustment. The variables being tested are, in
nearly all cases, separated by considerably more than the
6.7 divisions of possible error.
6.3.6.3 Third Series
The test samples were arranged in four main groups:
1. Barrier layer seals

2. Typical ceramic-to-metal seals
3. Low--loss evaporated seals
4. Non-metallic "seals".
Additional barrier layer seals were programmed

because the unsintered iron plated-copper plated samples
were pene rated by the Cu-Ag braze in the second test series.
(The iron plate was not sintered in order to reduce inter-
diffusion of Cu and Fe.) The iron plate was sintered on
at 800°C and an unsintered nickel plate completed the barrier
layer. Cu-Ag braze to a 70/30 Cu-Ni metal member was used.
Lucalox* was also used as a ceramic in addition to Body A
in order to introduce a variable which would influence the
sintering of the metallizing layer.
Detailed processing of the seals is as follows:
P-1 metalli 'ng of various thickness. t, fired

at 142r'OC t-. 30 miivites in wet H 2 and N 2 , 1:3
Cu or Ni overplate, 1 mil, sintered at 900 0 C for
30 minutes in wet H 2
Fe plate, 1 mil, sintered at 800 0 C for 30 minutes
in wet H2
Ni plate, 0.5 mil, no sinter
Cu-Ag braze, 4 mil, flag indication + 30 seconds
hold in wet 112
70/30 Cu-Ni, 20 mil thick.
*General Electric Co. trade name.
309
In order to obtain loss data on "production type" seals,
1. Body A - P-I metallizing - copper plate -

Cu-Au braze - Kovar
2. Body A - P-I metallizing - nickel plate -

Cu-Ag braze - 70/30 cupro-nickel
were programmed.
Several low loss seals also were programmed.

These essentially consisted of evaporated copper, nickel or
molybdenum onto Body A or Lucalox*.
In order to further study loss mechanisms the

loss induced by the non-metallic portions of the metallizings
was also evaluated.
The results of these four series of tests are

given in Tables 6-8, 6-9, 6-10 and 6-11. Asterisks on bome
sample numbers in the tables indicate that photomicrographs
(400X) of these samples are presented in this report and
that the microstructures shown are discussed later.
The -elative loss figures presented in these

tables cannot be directly compared to the data presented in
Tables 6-5 and 6-6 for, as previously mentioned in section
6.3.4.1, the cavi v was reworked for the thir6 series of
tests. New Body A samples and Lucalox* test pieces pro-
truded further into the cavity than the initial samples.
110
TABLE 6-8
Series 3, Barrier Layer Seals
Thickness
Sample No. Ceramic in x 103 Plate Loss**
200* Body A .1 Cu 53
201* .3 66
202* .5 88
203 .7 84
204 i.2 77
205* 2.6 87
206 .1 Ni + 109
207* .2 | 103
208 .3 '32
209 1.3 -2
210 1.6 -
211* Lucalox + .1 Cu 21
212 .3 51
213 .5 62
214 .9 70
215 1.6 50
216 1.6 57
217 .1 Ni + 62
218 .2 46
219 .3 62
220 .8 72
221 1.2 /6
222 2.0 114
+Two layers of 0.5 :il thickness each, both sintered.
*Crnss-sections of these samples are shown in photo-

micrographs.
**Add 22 units to convert to "old units" (Series I and 2).
++General Electric Co. Trade Name.
311
41N
4-)
4- Q)
wo 4.) Q)
(U 41 Va 41 413 4-) V
C: 0
4-) Ev
(a 00
a) cN M, m C:
H4 0 0 c) 0
U) 0lr- 0 a) LO 0 )0
H) e uv 0 tO C
( - + (N 0 +
44
4-) 0 413 0
- (13 "4 C; 413 (U .4
I U)4-) '- U41
0
00) Vu (u V (a
$4U r.1 Q)
V) 0)
1 r4
C.. ) Va 4.) N4 W) V
". -~ 4-) (a H6- 4-1 r
r. ' r-4 l
(U4JC U -r!04i
N4 (Ua -H 0) ty) u r-
ac 0) HU (a4 "
-".4 IU
Up4-) * H- H
rZ 4i U
a.) H -4 41 44 4.. Q)
V NwU) a) tLn -4--
4 t4C
HU H: C) H1 .
4rl H) 4 ri0L (N U)
Ci) ".4 HO) -- 4.to .4.

N 44 * r-IH4J NH (3l
) I .1i
0*-4 (M N H.q (U N z
-'-4 H- . H (a 0 -H H(IN H (U
En 4,J (a~In N4 HE
4 z a, N0 5
-,4 ) . N- ..) %0 r NN
) C4 Q)
0( (U 0 m~)
En N I 44 H 4-I N>
ai) 00 Z 0 IN -HO) : 0 0
N z u4 04u4-
U) -c
4' 0
W) Ur
0 0 0
413
U)
NI
m Ln
( Fa
312
a0)
0
* NN
4J 1 4-)
0'~~C 0)0 0)0
0z U) g.X -4
04- E-4
0r (%J 0
LA$40la E-4 0 $.4E-
4
0 L4-I>1 z ro 0 ro t
0E -P I w. H- Vo Lf, 0V- -
r. 0 4J V 4 0 C I OH ( H4
-LAH
-r H4)) 0 0 0 En 4. 0n
I aC:E 1):5 aCfl04
IV N2 -,4V -4 E c r
* u 0) 0 HN- r
(a m0 41) w02 0 r
CNu a): 4) ) - 0 H 12 a
M - JC 4- 02
n i
LA 0 g0 n
ro $1 -r H Y) CD -- 144 4 , f
Q)O o E) C' r:4.J0 0
.r 0 > :j0 000
H) 0 OHr-4 $4 r-4 z+ E E 0C
400- 0 IQ$4- 0 b u :1u :
U ,a 0 44 0a
-H( H4 9m r4 0 0
0 J (0g 0 0~ 4 * -4 - 6> 0 4iJ
H j 024 .4-4 W 44 (a 49 f a f
r. 41gi
~J
ro~~~~ ~H ~~ J'i U r q Uc9
4-)( IZ Q)-H C'.) 4J4 z H z '. -
W4~ 0 rq -H0 4J -H 4J C - H O -4 V
(U r-4 U)
rii u n 3 u ' - - (1 a V i - t
E-4 P- Vo C 0 toU-H C0 '1 . 14
-t 0 -H LA c ( .H-
04 rc ~ 1- 0 0 0 4-4 4 4~J - *--H4J 4J 44 4. 0 U 44 4J
(o 4j- 0- OD,0 rq - to r.
> r r 00E0-HE -H -.-q H ~0
W r-HHr- -iH O - OD H-H Hr
-- -Ire D r-U)
-r 02H-c VH r-H1
a44
'H -H
44LA
H- '-4 *H E V -H -HI-H H H) 0-H 1-E
4-4 -r4 - S: Q) 44 44 44 -H ,44
rC E
Q) I H- rzH- 0 ,4 E~ o U) H 0
fn r Q2) Q)H40) iLA E -'q a) Q) ) LA H 0) r
0) '-I4Ji4J 4J) C; - H J J 4 J. 0 4J
.r . -H(a fu x ( 0)- 0 (1 ( ( H H0 ( -,4 P4 (aN0 H
$~4 0 0 4w4 1-4 N 014 14 W W N z . N Z
0) mH0 00N 0 . . .- ml 00 0. mr :30 fu
U) 4- P4 04 04Q) 0) )
M-4 * 04 04 0 M01 04 W 04
I - (0 AJ- M 4J 4~J 4.1,00 X 0 M M4J 4J 4 .V M 4jA U
>-> u > (U m(a 0 > > > mm (
w)w 0 Q)HH4 IH ~0 H W Q))H HH4r-IP- a)Hr-I
U >1 - -0 -04 04 M4 m' Mo ~- -- 04 04ev > 04 rvn
w2 a, IN U-
u Vo iN
0) 0 - 0 0~1:0 Q) -4 :30 :1 -- 4
0:j:10)-H- 00 0-H-H 0
Q M EZH-4U L)4 Z UN u u wU z u - mzz uN
CO'-
IC-
04. 0o
OH
0 CN N mmN
(a~ rN N Nl N CN
313
- WJ3W 1'11 11
u
4
4-)
CN 4Jia
. Wa C 4.)
rn : ) U
4-' 0# 0 a
a) = 0 Q,
41 E-4 NVE
z- ZH 0E
Ln 0U Lf H
04 , V -rqII
I H 0t L
VO 0 0 IV fu uH
z H ,r ) 0
0 N* XN 41- x
E1 a) 4-)
(a rcN .i 0T
Vo 0 (3 E -41 0
% u VI
u Q) 0 LAO0 >10
0 En 0 0 H- >
LAO 0 -+ 0 0 iN
NOq 0 w0
O ) 41 ~ 0)04-) b
4-) + 41 : 4J -Hi : 41. Q n
:3 0 ~ 4 0
4.i -4 0
41
m 0 0 Q) S- *4
uM m i-,I U 4.) 4J 5-
n -U4 '-4 '4 U) 0-4
:0N Oc 4-) 0 0 4) 024
*4.j -H 44 $S4 rH 5-4
0)
.4-~ ) 44i a)12
41
pu u 41 ~4.) u ~4) ) Q 4-
r.~~~~~0
) I z 04 c> 0ar
4) -)
U) 4-44J CN - O N -H L) -1 Hi
-H "-E
' 0)i- -"-4 r 0)4-a)H-i r.i 5 rj
VO44 E ra(aH 4~ 0 U 44 -0 a) 0
01) H 0 w U H U) >,
4 V I C,4 -4 Va>4 V-.4 *HVraU Vo-H- 14 Q
to Q) Ha)U WH)z - w QWE 0 0 0
-p4 4 . 0OS4 4-1 Q0PW4) U 4-J
HON(a0).--i Q-"14) OUN 14-H N (N C%4
0
-Hl
05-m
-0
Nz ~44N S-i ~ 440( -i
$4p 41
04O 1 :j 0 0 - 4 0-
C4 4-) 5-4
41 4 W S- N il P Q0) CNO
rQ a) a) V0 0)
M 4- 01 Q (O 40 -0 nJ 4 Va H >
- >0 (u r > (o -H U> M H 0 0
a)H-Itp 0 E-'U 0)Hr- E-40 a)Hr- 0 rc 0
U i-o > 80 - 1: > 8Cj ' P4u
0)
O0H~ 0rH 0 Q))
m u ZUr ~o UU) ~UC H 044-
l+ + + + + 0 l)4l
+ + +1+ +0
~LA LOL 01 N
a) +) + -
NN m c
00 CN NN N+
314
TAbLE 6-11
Non-Metallic "Seals"
Sample Relative
No. Loss Description
Body A, unpolished
P-i metallizing, metallic portion

leached out with 50% nitric acid for
30 minutes at 150OF glued to copper
(silver plated) sample holder with
polystyrene Q-Dope*
236 21 (Paint thickness 0.1 mil)
237 27 (Paint thickness 1.1 mil)
*General Electric Co.
315
6.3.7 Discussion of Results
6.3.7.1 Introduction
The three series of experiments carried out

during the program fall into seven systematic groupings as
outlined below. The discussion is therefore largely con-
ducted in terms of these groupings.
Owing to the geometry of the test cavity, the rf

current will be carried in the metal layers next to the
ceramic ring, (if the ring behaves as a perfect insulator).
In this study, these layers are:
1. A copper sample holder to which is glued* a bare

ceramic (Table 6-5).
2. An electroplated layer of
a. Silver
b. Gold (standard)
c. Chromium
on the copper sample holder to which a bare

ceramic is glued (Table 6-5).
3. Same as for 2a but with a 'lossy' non-0metallic

layer sintered onto the ceramic surface (#236,
#237).
4. A copper evaporated and electroplated (#103, #104)
or electroless plated (#60) metallizing on the
ceramic which is tinned to a copper sample holder.
5. A copper evaporated and electro-plated layer on a

'lossy' ceramic surface whish is glued or tinned
to an electro-silver-plated copper sample holder
(#232, #233, #235).
6. A copper, or nickel, or titanium-molybdenum

evaporation plated layer on a ceramic with additional
plated "barrier" layers which is Cu-Ag brazed k..o a
70/30 cupro-nickel disc which in turn is tinned to
a copper sample holder (#226-#231).
7. A refractory metallizing sintered layer on a

ceramic with various electroplated layers on it,
which is then brazed to a metal disc which in turn
is tinned to a copper sample holder. This group
*Low loss polystyrene Q-Dope (General Electric).

316
constitutes the balance of the samples. The
various subdivisions are:
7.A P-2 metallizing, copper or nickel plate, copper

braze
7.B P-2 metallizing, copper or nickel plate, iron nlate
Cu-Ag braze
7.C P-1 metallizing, copper or nickel plate, copper
braze
7.D P-i metallizing, copper or nickel plate, iron plate,
Cu-Ag braze
7.E P-i metallizing, copper or nickel plate, iron plate,
nickel plate, Cu-Ag braze
7.F P-I metallizing, nickel plate, Cu-Ag braze.
(all these samples use 70/30 cupro-nickel braze

discs, all types are metallized on Ceramic A
while those in 7.E are also metallized on Lucalox*,
7.E.1)
7.G P-1 metallizing; copper plate, Cu-Au braze, nickel

plated Kovar or Cu-Ag braze 70/30 cupro-nickel.
It is proposed to discuss the results in terms

of the above groupings rather than by time sequence as was
used in the results section, since a better understanding of
the 1 2 R loss process may thus be obtained.
6.3.7.2 Skin Depth
The rf currents are carried in the surface layers

or "skin" of a metallic conductor. The "so called" skin
depth is the depth at which the rf current has fallen to a
value of I/e of the current at the surface, or 36.8%. For
the purposes of this study it may be assumed that in a homo-
geneous layer all the rf current is carried in a surface
layer which is four skin depths thick. Reservations concern-
ing a heterogeneous layer volume are discussed below.
The skin depth at 9.15 Gc of various pure metals

and metal alloys is plotted in Fig. 6.28 as a function of
their dc resistivity.
The skin depth of the evaporated or electroless

copper plated layer, #60, #103, #104 can be obtained by

317
4-
.0001 h.
4fto Nickel
.C- 14 Mco Moljbdem,
aprq
*71 Body A, P-1 Nickel plate

4. *5 BodyAP-2(55) Coppe Plate
*92 Body A.P-l Copplate'
20 b 4
DC Resistivity 4j (r- cm x 0
Figure 6.28: Skin depth at 9.15 Gc versus electrical resistivity.
318
taking its rf relative loss figure (or rather the cavity
insertion rod micrometer reading), obtaining its equivalent
dc resistivity from Fig. 6.26, and using this dc resistivity
to obtain the skin depth from Fig. 6.28. The equivalent
dc resistivity obtained in that case, of 12 x 10- 6 ohm-cm,
is higher than the dc resistivity of bulk OFHC copper
1.7 x 10 - 6 ohm-cm, or the value of 4 x 10-6 a/cm obtained
for a plated layer on top of a metallizing layer, (Table
6-4, column 4) as determined by the techniques discussed
earlier in section 6.2.2. This is partially because the
copper-ceramic interface forms a tortuous path along
which the majority of the rf surface currents flow, and
partially because the interface can also contain a varying
arpmunt of semi-conducting oxides to form the heterogeneous
interfacial layer. A skin der .h of 1.9 microns is obtained
from the 12 x 10-6 ohm-cm resistivity which means that the
rf current is carried in the 7.6 microns of copper adjacent
to the ceramic surface (0.3 mils).
By the heterogeneous nature of the volume in

the interfacial region of the metallizings considered in
this section, we mean that the region through which the
current will flow is composed of volumes of gradually and/or
abruptly varying resistivity in addition to layers which
may be mechanically separate. It cannot be expected there-
fore that the simple concept of "skin depth" developed for
homogeneous materials will quantitatively apply in this
instance. An analysis of the heterogeneous effect has not
been attempted in this study and the simple "skin depth"
concept is used. It is possible that some of the data which
cannot be completely explained would yield to this analysis,
however, in turn this would require more experimentation in
order to refine the data.
6.3.7.3 Group I
A Copper Sample Holder to which is Glued a
Bare Ceramic (Table 6-5)
When compared to the "gold standard", a ceramic

glued to a copper test piece base gives, as expected, a
negative loss (Table 6-5) because the dc and rf resistivity
of copper is less than that of gold. Any effect of the glue
joint at the ceramic-metal interface and the "lossiness"
of the ceramic has been compensated for by the initial
balancing of the cavities (6.3.3.5, 6.3.4.3, 6.3.5.3).
319
-'--.--l, -- U. - - -. -... ,' -"-

6.3.7.4 Group II
An Electroplated Layer of a) Silver,
b) Gold (standard) (c) Chromium on the Copper
Sample Holder to which a Bare Ceramic is
Glued (Table 6-5)
As silver is less lossy than gold, as expected,

it shows a negative reading. Chromium is more resistive
than gold and shows a positive micrometer reading. Since
the chromium is silver plated except under the ceramic,
whereas the gold standard is not, the increased lossiness
under the ceramic is somewhat offset by the decreased lossi-
ness in the silver plated region. As the current flow is
maximum under the ceramic and the resistivity of gold and
silver is relatively comparable, this effect will be of a
second order.
Since all the tested assemblies in this report
were silver plated, the above situation holds true through-
out the study.
In retrospect, it would appear that gold plating

throughout would have been advisable (if cost considerations
could be neglected).
6.3.7.5 Group III

An Electroplated Layer of Silver on the Copper
Sample Holder to which the Ceramic is Glued.
The Ceramic had a "lossy" non-metallic layer
sintered onto its surface. (Table 6-11)
The nitric acid treatment has effectively

leached out the molybdenum from the metallizing as the loss
is the same despite the large differences in metallizing
thickness. The resulting loss, 46 "old" units, is there-
fore the contribution of losses due to effects other than
conduction and may be assumed to consibt (. dielectric
losses in the heterogeneous MnO-TiO2-SiO 2 -CaO-MgO-A1 2 03
non-conducting or semi-conducting component of the metalli-
zing (see also sections 6.3.7.6 and 6.3.7.7) as the effect
of the polystyrene glue is cancelled out.
320
- -- -jjjjW ,-
6.3.7.6 Group IV
A Copper Evaporated and Electroplated (#103, #104)
or Electroless Plated (#60) Metallizing on the
Ceramic which is tinned to a copper sample
holder. (Tables 6-6, 6-7)
This group of evaporated or electroless copper

plated ceramics has already been discussed in section 6.3.7.2.
The seals were made as examples of simulated crunch seals
and were expected to display the lowest loss of any seals
that could be made. This expectation was realized but the
loss of 48 units was 52 units higher than initially expected
(see section 6.3.7.3) due to effects such as those discussed
in section 6.3.7.2.
6.3.7.7 Group V
A Copper Evaporated and Electroplated Layer on
a 'Lossy' Ceramic Surface which is Glued
or Tinned to an Electro-Silver-Plated
Copper Sample Holder (#232, #233, #235)
Table 6-10)
This group was synthesized to further study the

effects of the dielectric layer. Pure MnO-TiO powders
were painted and sintered onto the ceramic. The increased
amount of MnO-TiO2 has perhaps increased the dielectric loss*,
but more likely the increased loss is due to:
(i) A combination of factors similar to those

discussed in sections 6.3.7.5 and 6.3.7.6
giving a loss of 75 "old" units for the tinnea
samples.
(ii) In the case of the glued samples a loss of 87 old

units occurs, due to the inferior conductivity
of physically discrete layers as opposed to
metallurgically bonded layers via a tinning
process; i.e., there is a high junction
resistance.
The very high loss of sample #235, 124 old units,

cannot conclusively be explained. It may be due to a very
inferior electrical junction occurring between the copper
*Also, phenomena such as piezolelectric or semi-conduction

mnay be enhanced in this sample.
321
plate and silver plated base or it may be due to phase
changes which occurred in the non-metallic volume, such as
anorthite and MnO.TiO x format-on, after the 1100 0 C wet
hydrogen firing.
6.3.7.8 Group VI
A Copper, or Nickel, or Titanium-molybdenum
Evaporation Plated Layer on a Ceramic with
Additional Plated "Barrier" Layers which is
Cu-Ag Brazed to a 70/30 Cupro-Nickel Disc which
in turn is Tinned to a Copper Sample Holder
(#226-#231) (Table 6-10)
These evaporated metallizings with various

overplates were prepared to obtain loss measurements on
the "simplest" practical metallizing systems which could be
devised.
The copper sample 228# shows a loss of 78 units

because part of the current is carried in the copper, part in
the nickel (6.3.7.2) and part in the interfacial cupro-nickel
diffused zone due to the heat treatment experienced during
the Cu-Ag braze operation. Table 6-12 gives the resistivity
range in this s~stem. In retrospect, it is obvious that a
thick copper plate should have been employed prior to nickel
plating the sample.
The loss of the nickel evaporated and electro-
plated layer at 138 old units is as expected, see Table 6-12,
and furnishes a valuable end point for the data plotted in
succeeding sections.
j22
(323 & 324)
7M*'__
*~~E
TABLE 6-12
Room Temperature Resistivity of Selected Copper,
Nickel and Molybdenum Alloys
Composition Resistivity Reference

(wt %) (ohm cm)
99.99 Cu 1.69 x 1C- 6 b
99.9 Cu 0.1 Ni 2.0 x 10- 6 a
99.0 Cu 1.0 Ni 3.0 x 10 - 6 a
98.0 Cu 2.0 Ni 4.4 x 10 - 6 a
96.9 Cu 3.1 Ni 5.3 x 10 - 6 a
95.7 Cu 4.3 Ni 6.9 x 10-6 a
94.7 Cu 5.3 Ni 8.1 x 10 - 6 a
80.0 Cu 20.0 Ni 24.6 x 10- 6 c
70.0 Cu 30.0 Ni 34.0 x 10- 6 c
55.0 Cu 45.0 Ni 49.0 x 10 - 6 d
40.0 Cu 60.0 Ni 47.0 x 10- 6 a
30.0 Cu 70.0 Ni 43.0 x 10- 6 a
20.0 Cu 80.0 Ni 29.0 x 10- 6 a
99.99 Ni 7.8 x I0-6 b
62 Ni. 30 Mo. 5 Fe 135.0 x 10-6 d
99.95 MO 5.2 x i0- 6 d
a) Goldsmith, A., Waterman, T. E., and Hirschhorn, H. J.,

Handbook of Thermophysical Properties of Solid Materials,
Armour Research Foundation, Macmillan. New York, 1961.
b) Bureau of Standards, from Handbook of Chemistry and Physics,
Chemical Rubber Publishing Company, Ed.-in-Chief Hodgman, C. D.,
1951.
c) Wilkins, R. A., ane Bunn. E. S.. Copper and Copper Bare Alloys,
McGraw-Hill, New York, 1943.
d) Metals Handbook , A. S. M., Editor, L-yran, T., 1948.
325
a--t %"=WNW---r.r
The molybdenum evaporated plated layer is a high
strength vacuum tight layer and it was felt that this would
give a low loss layer. (The dc conductivity work had not
been carried out at this time.) In actual fact, the layer
was quite lossy. This was due to several facts.
(i) The dc conductivity of a pure molybdenum

evaporated layer is half way between that of
pure molybdenum sheet and a regular metallizing
such as P-1, as shown in Table 6-1.
(ii) The formation of a Ti-Mo alloy due to high

temperature interdiffusion processes with much
higher resistivity.
(iii) Factors such as those outlined for copper in

6.3.7.2 come into play, i.e., the formation of
TiOx, and oxygen dissolved in the metal.
The fact hat the loss on Body A utilizing Mo-Ti

layers is in between the copper and nickel situations, at
109 old units, is in agreement with their relative resistivi-
ties, but in view of the complexities listed above is
somewhat fortuitous.
The higher loss of the sample evaporated on

Lucalox* is explained on the basis of the much thinner
layer o.' molybdenum which resulted in the formation of a
high loss Ti-Mo-Cu intermetallic alloy layer. The fact that
Lucalox* is the substrate is not considered significant.
Fig. 6.29 shows metallurgical cross-sections of

the Cu and Mo-Ti samples. The various evaporated, electro-
plated and brazed layers can easily be seen.
326
4-BODY A
(POLISHED)
Cu 4--- Mo FILM
~2~' ' '~'~ ~IIFe PLATE
4-- CuLFIL
V4 Ni PLATE
4--Cu-Ag BRAZE
4-70/30 Cu -Ni
3- -BRZ Cu
6.3.7.9 Group VII
A Refractory Metallizing Sintered I3yer
on a Ceramic with Various Electroplated Layers
on it, which is then brazed to a Metal Disc
which in turn is tinned to a Copper Sample Holder.
This group constitutes the balance of the samples.
The examination of these ceramic-to-metal seal

samples constitutes the large bulk of the work.
Both P-1 and P-2 metallizings (Table 4-23 have

previously been described in detail. Figs. 4.49 and 4.50
show their microstructure. Their microstructure has been
further analyzed with X-ray diffraction, electron micro-
scope and electron microprobe techniques.
P-2 metallizing sintered at 1425 0 C is a porous

sponge of molybdenum particles only partially infiltrated
by the glassy grain boundary phase of the ceramic. After
sintering at 1550 0 C (P-2-1550) it becomes densely sintered
and well infiltrated by the glassy grain boundary phase.
The contrasting microstructures of these two conditions is
shown in Fig. 6.30. The former metallizing is extensively
penetrated by copper plate.
P-1 metallizing sintered at 1425 0 C is a molyb-

denum sintered sponge infiltrated by a reacted semi-conducting
heterogenous mass of MnO-TiOx-CaO-MgO-SiO2-Al 2O 3 . The low
12 R losses that occur in this phase(s) have been discussed
directly above (6.3.7.6, 6.3.7.7). The effect of the layer
on the lowering of the dc coefficient of resistivity has
also been discussed in section 6.2.
6.3.7.9.1 Sub Groups 7.A and 7.B
P-2 Metallizing-Copper Plate
The P-2 (1425 0 C) metallizing which is copper

plated gives data which is random as a function of metalliz-
ing thickness. This situation is shown in Fig. 6.31.
The liquated copper plate which permeates the

metallizing, after brazing is augmented by a copper braze
containing varying amounts of nickel from the cupro-nickel
metal member. This infiltrated metallizing (sub-group 7.A)
gives a high loss Mo-Ni-Cu intermetallic complex on solidifi-
cation. The conduction loss would be expected to be and is
very sensitive to brazing time and brazing temperature Fig
variations.
328
Figure 6.30: Photomicrographs of P-2 metallizing fired at 1425"C (top) and 1550*C (bottom).
Subsequently copper plated, Fe barrier layer, Cu-Ag braze (400X).
- 329 -
190'
* Cu Plate Cu Braze
Curve I - A NI Plote Cu Braze
ISO- Curve 2-0 Cu PIate,Fe, Cu At Braze
Curve 3-0 P-2 (1550),CV PIoteFe,Cu AV Braze
Curve 4-A P-2 (1550),Ni Plate ,Fe,Cu Aq Bree
170
160"
150.
140 Evaporated Nickel

-..-.. NI1Plte Cu Braz7
0I)
. 130
120.
S 10"1
0
90. ,
so. (3)
70 (4)
/
60 I
50 - Eloctroleus Copper
I - Evaporated Copper
40' - 1
.001 .002 .003 .004
P-2 Idetollizing Thickness (inches)
Figure 6.31: Conductive loss data for P-2 metallizing on Body A. Metallizing sintered at 14251C
unless otherwise noted.
-330-
An attempt to reduce this variation (due
to liquation of the copper plating layer) and to give
a low loss layer was made by brazing with Cu-Ag at
810 0 C (versus 11000C for copper braze). An unsintered
iron barrier layer was inserted to prevent the attack of
the Cu-Ag braze on the underlying molybdenum metallizing
which was infiltrated by the copper plate. The small
amount of data obtained (Fig. 6.31, curve 2) indicates
that a low loss situation was not attained. This must
be due to the fact that several percent of copper can
go into solid solution in the molybdenum phase, thus
raising its resistivity.
Sintering of the P-2 metallizing to 1550 0 C

and otherwise identical subsequent processing to that
of the previous paragraph does give a very low loss
situation. A seal with 66 loss units is the lowest
loss displayed by any conventional seal on this program
(Fig. 6.31, curve 3). The initial peak in the loss
curve at the 0.4 mil metallizing thickness is again due
to a portion of the current being carried in the
heterogenous Mo-Cu interfacial boundary region. At
greater thicknesses, the current is carried exclusively
in the pure unpenetrated molybdenum metallizing which
must therefore be lower loss.
P-2 Metallizing-Nickel Plate
The P-2 metallizing, nickel plate situation

is somewhat similar to the initial situation discussed
above, Fig. 6.31, curve 1. The 1425 0 C metallizing allows
the nickel plate to permeate the metallizing and at
the copper brazing temperature of 1000 C and a solid
state Ni-Mo reaction will ensue. This reaction results
in a high loss interfacial film of Mo-Ni intermetallics
forming throughout the metallizing (Table 6-12). Again,
this situation is corrected by the 1550 C sintering
(Fig. 6.31, curve 4). Also, as the metallizing layer
gets thicker, the loss again decreases. i.e. the current
is more and more carried in the metallizing layer
instead of the higher loss Mo-Ni and nickel inter-
facial layers.
331
6.3.7.9.2 Sub Group 7.C
P-I Metallizin.c, Copper or Nickel Overplate,
Copper Brazed
The loss values utilized in these seals are shown
in Fig. 6.32 on curves 1 and 2.
As expected, the losses, utilizing a copper
overplate and braze, progressively increase as more current
is carried in the metallizing layer, while they decrease
in the case of a nickel overplate and copper braze. An
equal loss is experienced in both cases when the current
is carried exclusively in the metallizing layer.
While this simple explanation is fortuitously

convincing for this isolated segment of the data, a more
thorough treatment is required to fully explain the data
as may be seen from succeeding discussions.
6.3.7.9.3 Sub Group 7.D

P-1 Metallizing, Copper or Nickel Overplate,
and Iron Plate, Cu-Ag Braze
While the brazing with copper is feasible with

a cupro-nickel member, it is desirable to braze with Cu-Ag
(or gold bearing brazes) when brazing to copper. Accord-
ingly, an iron plated layer was inserted between the copper
plate and the Cu-Ag braze to prevent attack of the copper
by the Cu-Ag. A loss situation less than that obtained in
Fig. 6.32, curvc 1, was expected, as no nickel-molybdenum
interaction (due to solution of Ni from the cupro-nickel,
see microprobe data Fig. 6.34) could occur.
A much higher loss was, in fact, experienced

(Fig. 6.32, curve 3). Although once again the loss levels
are indicating that the current is carried exclusively in
the thicker metallizing layers, this hiyher loss may partially
be explained by penetration of the Cu-Ag braze through the
Fe barrier layer as shown in Fig. 6.33 (top). The dc
resistivity change alone could not be responsible, however.
(See later Table 6-12.) Ideally, the barrier layer separates
the various layers of the seal as shown in Fig. 6.34, but
as shown above, breakdown of the seal layers does occur
with penetration of the Cu-Ag up to, but not into, the
metallizing layer (Fig. 6.35). r:is fact was established by
carrying out further microprobe studies on seal No. 93,
Fig. 6.36, which are shown in Fig. 6.37.
332
140,
,-Evaporated Nickel
130-
100 A-900C
00
so.
70.- (4)
Curve I - 0 Cu Plate, Cu Braze

50-E loctroless Copper Curve 2 - A~ Ni Plate, Cu Braze
-Evoparated Cop-per Curve 3- 0 Cu Plate Ps Cu Ag Braze
Curve 4- A Ni Plate F,~eCu Ag Braze
.001 .002 .003 .004 .005
P- i Metaliing Thickness (inches)
Figure 6.32. Conductive Ions date for P.) on Body A. Metallizing sintered at 11425TC.
-33---
41
Figure 6.33: Photomicrographs of Cu-Ag braze penetration through Fe barrier layer into Cu (top) and
Ni (bottom) metallizing overplate ROOMK.
334
C¢gm 4 OC"I* e Ce As 11me C a 0666V
"OW
100-C
. -_......
, Ire .--
I,\
a.o ,x
-O 0 0 o 0 sO
O0 0 O 40 '0 ceo tOO
i qJ g mip o- mi~illpu mtNHi Je J

m
~li m uugJm~,IqjiPapn pnei
Figure 6.34: Microprobe analysis of iron barrier layer seat Data was averaged.
--335 ..
F9
Ce Aq @ime Diee Cv Plate MOsOSSM4 Ceremic
go LOVO
to-.
a Aq 40 44 4
do ISO too sea
Figure 6.35- Micropobe analysis of iron barrier layer seal described in text showing silver scan across
sea! in the region of e grain boundary discontinuity in the iron barrier layer.
-- 336
Figure 6.36: Photomicrograph of conduction loss samnple 093 (P-1, Body A, Cu plate, Fe barrier layer.
Cu-Aq braze). (850X).
337
10
0
.
30
20 I I
.005 .010 .015
DISTANCE SCANNED (inches)
Figure 6.37: Microprobe analysis of Mo and Cu in conduction loss sample #93 (P-i Body A, Cu plate
Fe barrier layer, Cu-Ag braze). Ag and Ni were below limit of detectability. Scans were
parallel to metallizing - ceramic interface 0.0005 in. into metallizing.
- 338-
One must conclude that although the current is
carried in the metallizing region either (I) the overplate
and other layers must provide a variable impedance which is
electrically reflected into the metallizing layer, or more
probably (2) the metallizing is altered in some so far un-
detectable fashion by the subsequent processing treatments
such as solid solution of the copper in the molybdenum
phase.
The opposite loss situation is true for the

nickel plate - iron barrier layer series curve 4 in Fig. 6.32.
The nickel plate - molybdenum metallizing interaction zone
sees a maximum of 800 0 C (Cu-Ag braze temperature) as opposed
to the 1100 0 C copper braze temperature of curve 2. Con-
sequently much less high loss Mo-Ni intermetallic is formed
in the metallizing layer and at the interface and lower
losses result. An intermediate loss situation was obtained
in the case of one nickel plated sample sintered at 9000C.
The concept of the influence of seal layers other than the
metallizing directly carrying the rf currents must be
invoked for the two cases under consideration (nickel and
copper overplate) in order to explain the different points
at which the losses appear to level off. (Fig. 6.32, curves
2 and 4.)
6.3.7,9.4 Sub Group 7.E, 7.F

P-I Metallizing, Copper or Nickel Plate,
with or without Iron Plate, Nickel Plate,
Cu-Ag Braze
These data* utilizing seal construction shown

in Fig. 6.38 (top) are plotted in Fig. 6.39 together with
data frcm the previous sub group 7.D.
The use of a Fe-Ni barrier layer was an attempt

to minimize the penetration of the Cu-Ag braze through to
the base plating layer of copper or nickel as is ideally
shown in Fig. 6.38. However, this barrier layer also was
not 100% effective as is shown by Figs. 6.40 and 6.41.
Despite the random penetration of the copper

layer by the Cu-Ag braze, the results are consistent among
themselves and with the previous data, Fig. 6.39, curve 1.
Photo-micrographs of three seals showing metallizing of
different thicknesses with correspondingly increased losses
*Omitting that using a Lucalox (General Electric Co. Trade

Name) base, which is reported later in 6.3.7.9.6.
339
// 0 /30'Y7
- -mCu AM braze
0.5 mil NI plate

1.0milFe plate
1.0 milC o 1 lt
l.Omil0/3
Cuor-Nlae
oP-I eap.ig(aial hikes
-- '-4- 70/30BCu-N
* 4- NIPLATE
- PI METALLIZING
Figure 6.38: Schematic cross-section of ceramic-to-metal seal with iron-nickel barrier layer (top);
photomicrograph (400X) of sample #207 (bottom).
- 340 -
160,
Vaporized Ni
140i'
S\ X Cu -Au( - .
- -0)
1201- -:\o o . .. . ..- 0 .. -. -
mot "-Sintered Twice
, oo .X . k:. ..
. -Pl at
0 Fe
o , Not Sintered
=," ® "(3)
-7-5
60-- Vporized Cu
oI I
40 .001 .002 (.0026)
Metollizing Thickness (inches)
n I mil Nickel overplote ,900°C sinter for 30 mWn.,Fe NI barrier

X 2 " a " 0 , Fe barrer
0 0.5 " 0 N
" " I a - -
* 2 " " 1006C N 0" " Fe barrier

0 2 "Copper N '9000C N I * a
S1 U " N 0 o 0",FeNibarrier
12 R Losses at 9.15 GOC vs. metollizing thickness for P-I on body A

with overplate of copper or nickel, barrier loyor. P" "pper-sliver
eutectic braze.
Figure 6.39: Plot of 12R losses at 9.15 G c versus metallizing thickness for P-1 on Body A with
overplate of copper or nickel. All specimens were prepared with a barrier layer
and a CuAg eutectic braze and 70/3OCuNi metal member (see text).
-341 -
law 70/30 Cu -Ni
Cu - Ag BRAZE
34- Ni PLATE
Fe PLATE
(DISAPPEARED)
P- I METALLIZING
(DETACHED)
(ATTACKED)
4-P-1I METALLIZING
4-BODY A
Figure 6.40: Photomicrograph (400X) of sectioned sample #205, showing failure of barrier layer and
detachment of portion of metallizing. See Table 6-8 for data.
- 342 -
'-70 /30 Cu -Ni
A ~Cu -Ag BRAZE
Ni PLATE
(INTERRUPTED)
*j4b4
_v"7 * .
Cu PLATE
(ATTACKED)
4-P-1 METALLIZING
4-BODYVA
Figure 6.41: Photomicrographs (400X) of sectioned sample #205 showing disruption of iron barrier
layer (dark) and partial dissolution of copper overplate. See Table 6-8 for datm
-- 343 -
are shown in Fig. 6.42. The bottom seal shows extensive
Cu-Ag penetration of the copper layer. The physical heter-
ogeneity of the seals is obvious from these cross-sections.
A contributory factor to the consistency of the data,
despite the variable penetration of the Cu-Ag braze, is due
to two factors:
1. The dc resistance of Cu-Ag is similar to copper,

Table 6-13.
2. At metallizing thicknesses greater than 1 mil
the rf currents are carried almost completely in
the metallizing layer.
When utilizing a nickel overplate on the metalli-
zing, in no ca:e (either in 7.D, 7.E or 7.F) was the nickel
plate penetrated by the Cu-Ag through to the metallizing.
The use of the iron or iron-nickel barrier layer shown in
Fig. 6.38 is therefore superfluous as is shown in Fig. 6.43
(bottom). This being so, the three seals denoted by 0,
X and 3 on Curve 2 all have an intact, 900 0 C sintered,
nickel plate next to the metallizing; and therefore show an
intelligible trend. The losses are between those in which
the nickel plate is sintered on at 8000C and those sintered
at 1100 0 C. The somewhat higher losses for the r3 points are
due to an additional sintering of the nickel plate in the
course of sintering on the iron portion of the iron nickel
barrier layer as would be expected (see Fig. 6.19).
6.3.7.9.5 Sub-Group 7.G

P-1 Metallizing, Copper Plate, Cu-Ay or Cu-Av
Braze, 70/30 Cupro-Nickel or Nickel-Plated
Kovar Respectively (Table 6-9)
Together with Sub-Group 7.F, these represent
conventional seals. Cross-sections of these seals are shown
in Fig. 6.43. Table 6-13 shows -'he resistance of the Cu-Au
braze to be high in comparison to copper or Cu-Ag. Conse-
quently, the losses as shown in Table 6-9 are higher for a
0.5 mil metallizing thickness, but not drastically so.
344
Figur
'
" - _': '"TL ..- . L
I "-L'.. _.- -t" -',- ... ., i- -. -
- b4-70/30OCU -Ni
a. ~ .4.--Cu -Ag BRAZE
Fe PLATE
a ~- Cu PLATE
P-1 METAWLZING
BODY A
or4 70/ 30CU -NI
I. 4- C-AgBRAZE
Fe PLATE
P-1 METALLIZING
BODY A
'a
It
* ~ 4-Cu -A9 BRAZE
Fe PLATE
Cu PLATE
P -I METAL LIZING
gure 6.42: Photomicrographs (400X) of samples *200 (top) with 53 units of relative loss, 0201
(center) with 66 units of relative loss, 0202 (battorni with 88 units of relative loss.
See Tabte6-8 for data. 34
OOYA
4-P-I METALLIZING
* 4- Cu-AuBRAZE
-~ --- 4--Ni PLATE

'W4 vK$WAR
P-1 METALLIZING
Cu - A BRAZE
.4.... 0/30 CU-t4
Figure 6.43: Photomkrophs (400X) of sample 0224 (top and sample 0225 (bottom).
-346
TABLE 6-13
Room Temperature Resistivity of Selected
Braze Materials
Measured Published
Composition Resistivity* Data 5
Material (wt %) (microhm cm) (microhm cm)
OFHC Copper 99.9 + Cu 1.775 1.69
35-65 + 65 Cu, 35 Au 8.65

50-50 + 50 Cu, 50 Au 11.22
80-20 + 20 Cu, 80 Au 13.59
24K gold 99.9 + Au 2.365 2.44
NIORO + 82 Au, 18 Ni 16.35
NICORO + 62 Cu, 35 Au, 3 Ni 12.95
CUSIL + 72 Ag, 28 Cu 2.175
*Determined from measurements on 20-foot lengths of 0.0i0

inch wire.
+From WESGO, Belmont. California.
341
6.3.7.9.6 Sub-Group 7.E.l
P-I Metallizing on Lucalox*, Copper or Nickel
Plate, Cu-Ag Braze (Table 6-8)
The data obtained using Lucalox* as a base given

in Table 6-8 is plotted in Fig. 6.44.
As explained previously, due to the difference

in the loss measuring equipment working point, this data is
not directly translatable to that employing Body A.
In this case,the loss characteristics are

reversed for the nickel and copper overplate situations.
The metallizing is not vacuum tight on Lucalox* and tends to
have micro-fissures and pores. This means that the metalli-
zing is penetrated by the plate.
In the case of the copper plate, the losses

generally follow the usual pattern, being almost identical
to those displayed by curve 1 in Fig. 6.32, although the
dotted line indicates that a more complex electrical inter-
action may occur.
In the nickel plate case, it appears that the

formation of the lossy Ni-Mo intermetallic is dependent on
the amount of molybdenum available to react with the nickel
plate. As the metallizing volume increases, so the losses
increase in an exponential fashion.
The losses would be expected to assume the more

characteristic pattern of Fig. 6.39 if the metallizing were
replaced by a metallizing such as P-3 which forms a vacuum
tight seal on Lucalox*.
348
160-
140"
120-
- Ni Plate
100 0
a0
(2) - 0 -
80
go Plot* 0
.j
so
0 0
=60- 1)'
4I
Ip
O.I
I 2 /0
I
MItwl~Avl Tinn s (Inks)
Figure 6.44: IZR (conduction) losses versus metallizing thickness for P-1 on Lucalox" with overplate of
copper or nickel. The specimens were prepared with a barrier landr d a CuAg eutect;c
braze (see text).
-349 -
6.4 DIELECTRIC LOSSES
6.4.1 Introduction
The property of the ceramic-to-metal seal to be

measured is the loss arising from phenomena other than
conduction currents (due to the presence of mobile charge
carriers) in the interface region. If a ceramic-to-metal
seal is placed in a region of high normal electric field
and little or no magnetic field, the tangential conduction
current flowing should be at a minimum, while the loss
component of the displacement current, herein called
dielectric loss, should be high. Unfortunately, conduction
currents due to the presence of electric fields in material
of finite bulk resistivity can also be expected. The success
of an experiment in separating the two loss mechanisms
hinges on the relative magnitude of the above effects.
By inserting the sample between the posts of a

re-entrant type TM 0 1 0 resonator, the above requirements
should be met. The bulk of the electric field is between
the re-entrant posts of the cavity, while the magnetic
field is a maximum at the outer radius of the resonator and
decreases to zero at the axial center. If a solid cylindri-
cal test sample is placed between the posts, the electric
field will be perpendicular to the interface at the inter-
face if the resonator is symmetric about a plane defined
by the interface. A measurement of resonator Q will
indicate the relative loss of various samples.
The method of measuring the internal Q of the

resonator follows Ginzton 3 . Tne measurement involves taking
readings of the input standing wave ratio of the resonator
as a function of frequency near the resonance point. By a
knowledge of the coupling coefficient, the standing wave
ratio at the 3db power points can be calculated and from
this the Q can easily be found.
The first cavity, and its 0.200" diameter by

4.0" long test pieces, investigated proved to be unsatis-
factory. No results were obtained with this geometry.
Briefly, the problem was dilution of introduced metallized
seal loss effects by a large volume of unaffected ceramic.
The reader is referred to the second and third

quarterly reports for further details.
350
Excerpts from the second and third quarterly
reports and a diagrammatic sketoh of the first cavity
constructed, Fig. 6.45, are given here, however:
"Test specimens for the dielectric loss determina-

tion consist of .200" x 4.0" ceramic rods. The rods are cut
in the middle, ground flat, metallized and plated on each
end, and brazed back together, thus introducing the dielectric
losses of two ceramic-metal seals simultaneously into the
test resonator. The first ceramic test specimens consist
of Body A metallized with P-l, copper plated, and copper
brazed."
"The internal Q of the resonator was measured

as nearly 14,000 with no ceramic in the gap (2424 mc).
Insertion of a bare 94% alumina ceramic rod (0.200" diameter
x 4" long) lowered the Q to around 1300 at 1997 mc.
Subsequent measurements on two pieces metallized and
brazed together on the ends indicated that the additional
loss introduced by the seal was on the same order of magni-
tude as the bulk loss of the ceramic. Variations due to
experimental error masked any change due to introduction of
the seal."
Clearly, to see an effect it was necessary to

reduce the ratio of pure ceramic to contaminated ceramic
(that which had been penetrated by metallizing). A new
cavity was designed which would accommodate disc shaped
samples of 0.200 inch diameter. The posts on this resonator
were made solid instead of hollow since with narrow gap
spacings (compared with gap diameter) the fields in the
hollow post cavity would be non-uniform across the gap, as
well as having transverse components. The disadvantage of
a solid post configuration is that the cavity must be dis-
assembled to remove and replace the ceramic test piece.
In retrospect, in view of similar results with

94% alumina ceramics experienced with the second cavity
which was designed (and which was used for all the dielectric
loss experimental data obtained on this program), it is
possible that the initial cavity made would have proved
adequate for the needs of the program.
The second cavity, the modified test specimen

and the experimental procedure proved satisfactory and are
discussed below.
351
1.HOLE FOR INSERTION OF TEST PIECE
2.RESONATOR TOP
3.RESONATOR BASE
4.RE-ENTRANT POSTS
5. TUNING DIAPHRAGM
6. TUNING JIG
7 TUNING NUT
Figure 6.45: Resonator for dielectric loss measurements.
-352 -
Three series of results were obtained. They may
be categorized as follows:
(i) Initial "prove-in" results
(ii) Systematic experiments designed to establish

certain hypothesis as facts.
(iii) Experiments carried out to refine the data of

(ii), to establish the relative contribution of
conduction losses and to test "low loss" seals.
6.4.2. Sample Preparation and Experimental Set-up
6.4.2.1 The Cavity
The second cavity designed and built is shown in

Fig. 6.46. The top post was designed to be removable so that
the ceramic specimen could be placed in the cavity easily.
The top post was made hollow and saw cuts made along the axis
near the point of contact with the inner surface of the
cavity top. A tapered screw is threaded into the top post
so that when it is screwed in, it forces the walls of the
top post outward tightening the joint between post and top.
The coupling loop can be rotated to allow -djustment of the
coupling.
The cavity was designed for samples 0.060 inch

thick. It is possible, however, to test pieces down to
about 0.010 inch thick, with corresponding lowering of
resonant frequency.
6.4.2.2 Sample Preparation
The geometry of the Jelectric loss test piece is

a 0.200 inch diameter disc. The thickness of the sample must
be between 0.050 and 0.070 inch to keep the resonant fre-
quency of the cavity within the desired limits. Most
measurements were with discs 0.060 inch thick. A few were
made with two stacked 0.030 inch thick discs. Raw ceramics
of 0.050/0.060 and 0.070 inch were used for calibration
purposes.
The dielectric test samples were metallized only.

The plating and braze were omitted since they do not
contribute significa.ntly to this type of loss.
The ceramics used were Body A (94% A1 2 03 ) and

Body H (99.5% A1 2 03 ). The metallizing consisted of P-l,
353
I3
0 I 2 3
Inches
I. Coupling loop 5. Top

2. Type N coaxial connector 6. Ceramic test sample
3. Top post expansion screw 7 Resonator base
4. Expanding top post
Figure 6.46: Redesigned dielectric loss cavity.
-354-
P-2, and P-3 (g) a MnO 2 , TiO 2 , Si0 2 glass. The recorded
thickness of the coatings of Series (1) is from calibration
curve #5, Figure 13, 3rd Quarterly Report, and should be
used as a comparison reference only. They may be modified
by the data presented in Fig. 4.48. The thickness of the
coatings in Series (3) is from direct metallurgical cross-
section examination.
Further sample preparation detail is more
appropriately reported within the individual experimental
series sections.
6.4.2.3 Equipment Set-up and Measurement Accuracy
A block diagram of the system is drawn in Fig.
6.47. Since the measurements are to be made at 2.2 to 2.3
Gc, it is more convenient to use coaxial lines instead of
waveguide. Coupling to the resonator is by means of a small
loop, placed in a region of high magnetic field. By
rotating the loop, the coupling may be varied. A portion
of the power from the Hewlett-Packard Model 616B generator
is sampled and fed to a Hewlett-Packard transfer oscillator
and the frequency measured by a Hewlett-Packard electronic
counter. mhe signal passes through a coaxial slotted line
and thence to the resonator - the standing wave ratio is
read on a Hewlett-Packard standing wave indicator. Phase
data may also he read from the slotted line.
The Hewlett-Packard Model 616B generator after
three hours warm up was adequately stable as a signal
source. Although its tuning rate is high, it was more
convenient to operate than the klystron oscillator which
was used initially.
A new signal generator became available during
the sixth quarter prior to the third series of experiments.
The experimental set-up was thus simplified. (Fig. 6.48.)
The Hewlett-Packard 8614A Signal Generator has
two outputs, an uncalibrated, unmodulated one to provide the
signal for the frequency measurement with an electronic
counter and the main output furnishing the signal square
wave modulated for indication on the VSWR meter.
A 30db directional coupler was inserted to
measure the signal reflected at the cavity and find the
resonance frequency easily and tune in precisely.
355
0
z
0
00
wU
a0 0 -
(0 75 >.
E
E
w
w 4
4w 0
00 0
-4-
-J --
ZWD
0 0
w E
1. 0 4
I-w 0
W IL W
I- -35--
S-BAND
SIGNAL
GENERATOR
6dB
ATTEN.
3kmc
LOW PASS
30 d8BRSA
DIRECTION. IF, ','ECTOR
COUPLER
SLOTTED CRYS TAL
LINE H ETECTR
I
cAVITY INITOR
Figure 6.48: Block diagram of "Q" measurement for dielectric loss data at ceramic-to-metal
seal interface.
-357-
The scatter of data on the same sample in the
same test series (see Series 3, Table 6-18) was attributed
to two facts. The unmodulated output of the signal generator
showed some frequency modulation connected with the square
wave modulation of the measuring signal. On account of
this, the frequency measurement with the electronic counter
was somewhat erratic in some cases as could be detected
from the appearance of the resonance curves. The detrimental
effect of frequency modulation in further work could be
avoided by withdrawing a more attenuated signal from the
source. The second cause for scatter in the loss data and
for variations of the cavity resonance frequency with the
sample reinserted was assumed to be the mounting of the
sample between the two posts of the re-entrant cavity. The
removable, inserted post carrying the pill-shaped sample
had a shoulder to prevent the test sample from gliding to
the side. But there was about 0.010 inch of play; that was
the amount of disalignment by which the sample could be off
center. Unfortunately, the location of the sample between
the posts was very critical.
The results are given in terms of the quality,

f (at 3 db down points) of the test cavity. A large value
indicates low losses of the inserted test samples and vice

versa.
With the low loss ceramics tested, the measured
loss differences resulting from test disc thickness variations
between 0.059 and 0.061 inch were not significant compared
with the metallizing introduced losses. The data therefore
assumes a constant sample thickness of 0.060 inch. In
almost every case, two identical samples were measured and
the Q averaged. Reductions in 0 of less than 10% for the
averaged samples are of questionable significance as dis-
closed by retesting of the same sample over a period of
time (days), and by testing identically treated samples.
A more rigorous "control" system was instituted

in the second series of test runs. All samples, raw and
metallized, were given simultaneous cleaning and firing
procedures, including the 1425 0 C metallizing sinter fire.
It was noted that the results obtained in the second series
of tests showed higher cavity Q's and greater reduction in
Q on metallizing than the firrt test series. The 0 shift
was greater than the 10% considered significant. Day-to-day
correlation by retesting of identical samples during the
second test period still indicated 10% as significant.
however.
358
It was concluded that the second series results
are more accurate.
A typical plot of data for the first test series

and a Q calculation is shown in Fig. 6.49.
6.4.3 Experimental Procedure and Results
6.4.3.1 Series 1
Specimens of Body A and Body H with P-1 and P-2

metallizing were tested. In addition, specimens of Body H
were painted with the non-metallic phase of P-3 consisting
of MnO 2 , TiO 2 and SiO 2 (P-3g). The dielectric loss test
specimens were not plated or brazed after metallizing since
the loss mechanisms being evaluated in this test are not
concerned with the metallic portions of the seal. Data from
the tests on these specimens are given in Tables 6-14 and
6-15. Most tests were on 0.060 inch thick discs. Raw
ceramics of 0.050, 0.060, and 0.070 inch were also tested
and four tests made on stacks of two 0.030 inch specimens.
The data taken on different days was not always

consistent. The cause for this has not been resolved but is
possibly due to wide humidity variations. To compensate for
this the data has been adjusted to numbers relative to the
raw ceramics indicated as Standards 41 and #2 which were run
with each group of tests. Adjustments of 10 to 20% were
required, though in most cases this is still less than the
differences noted between the test variables. Sample #26
gave anamolous readings, but see section 6.4.4.4.1.
Unless noted, the specimens were sintered at
1425 0 C
for 1/2 hour. Certain specimens, noted, were rerun
through the same procedure to further diffuse the non-
metallic phase of the metallizing into the ceramic substrate.
The number of additional sintering runs is indicated.
6.4.3.2 Series 2
The previous tests showed that with the 94%

A1 203 ceramics the additional losses in the test piece
resulting from metallizing are scarcely detectable. The
3amples prepared in this series therefore utilized the lower
loss Body H (99.5% Al O), single crystal sapphire (Linde),
and the special #992 (175;0C) body prepared expressly for
this program.
359
....
. .... -
7 . ..-.. -7
4.'
.. . . . . . . .- .
... ..
isf
COUNTED FREOULNCY (in GO)

Figure 6,49. Typical data plot and Q calculation for dielectric los measurement.
TABLE 6-14
Series 1, Dielectric Loss Data on Body A (94% A1 2 03 )
As Received and When Painted
with Noted Metallizing
Adjusted
Ceramic MetaliLzing Measured
Sample Thickness & Thicknesso Cavity
No. (in. 001 in.) (mil) Q Remarks
19 50.0 bare 1533

23 61.3 bare 1389 Standard #1
27 70.5 bare 1268
22 71.5 bare 1311)
13 61.2 .84 of P-I 1385
15 61.0 .90 of P-i 1401
16 61.1 .90 of P-i 1381
II 61.1 .92 of P-i 1312
10 61.0 .98 of P-i 1408
14 61.1 1.22 of P-i 1326
9-10 30.9 - 30.9 1.04 - 1.02 1243 (2 surfaces
X-X 30.? - 30.? ? - ? 1098 painted; 2 1/2
12 60.9 1.46 i 1128 hours at 1425 0 C
24 61.1 .84 of P-2 1052
30 60.9 .95 of P-2 1177
28 61.3 .98 of P-2 1280
26 61.6 1.00 of P-2 (758) Anamolous data
27 61.0 1.02 of P-2 1240
29 61.2 1.14 of P-2 1270
*From Calibration Curve #5, 3rd Quarterly Report, Figure 13.
361
TABLE 6-15
Series 1, Dielectric Loss Data on Body H (99.5% Al 2 03 )
As Received and when Coated
with Noted Metallizing
Adjusted
Ceramic Metalliziig Measured
Sample Thickness & Thickness* Cavity
No. (in. 001 in.) (mil) Q Remarks
48 50.5 bare 2210

49 50.6 bare 1912
37 59.8 bare 2196 Standard #2
38 59.9 bare 2100
37-38 31.0-30.4 bare 2063 2 pieces
51 70.9 bare 1726
34 61.6 .42 of P-i 1100
33 61.1 .12 of P-I 1115
33-34 30.7-31.6 82 - .54 868 2 surfaces
painted
40 59.9 1.]0 of P-2 1732
39 60.6 1.io of P-2 2094
44 61.3 .08 of P-3(g) 1560-863 2nd set of numbers
43 61.0 .16 of P-3(g) 1570-1082 after additional
35 60.6 .18 of P-3(g) 1398-998 1 1/2 hrs. at
36 60.7 .62 of P-3(g) 1135- 1425 0 C
*From Calibration Curve #5, 3rd Quarterly Report, Figure 13.
362
I C E. I
The test pieces were metallized with P-1, P-2
and P-7. The unfired weight of the applied metallizing was
recorded in each case. Sintering was at 1425 0 C for 1/2 hour
with the exceptions noted below.
The P-I and P-7 metallizing paints on the #992

(1750) body were given additional firings up to 8 hours total
time at 1425 0 C. Raw ceramic control samples were run with
each lot and firing sequence.
Table 6-16 lists the data collected with Body H

and sapphire samples. Table 6-17 lists the data collected
with the 992 (1750) samples.
The measured cavity Q's tended to be higher dur-

ing these measurements than those taken during the previous
series. Lower atmospheric humidity during this test period
may account for the shift. Since the A Q of bare versus
metallized ceramic is used for interpretation in most cases,
the variation of Q is not a reason for concern.
A non-metallized Body H disc was used as a
standard to check day-to-day shift in Q. The presented
data were taken on two days only and the shift was from
o= 2430 to Qo = 2498, or 2.75%, which is well below the
AQ of 100, which is considered as significant.
Two samples of each variable were measured.
Correlation between samples was usually satisfactory. The
amount of blush up the sides of the 0.200" diameter by
0.060" thick disc is relatively' uncontrollable and is the
most probable cause for those discrepancies of up to 25%
Q which did occur.
363
TABLE 6-16
Series 2, Dielectric Loss Data at 2 Gc on Body H (99.5 A1 2 03 )
and Sapphire with Noted Metallizing,
Sintered 1/2 Hour at 1425 0 C, No Plating.
Ceramic Discs Nominally 0.060 In. Thick
Wt. of
Sample Metallizing*
No. Ceramic Metallizing (mg.) Measured Q
1 Body H P-1 2.8 802**

2 " " 3.5 872
3 go P-2 6.7 1795
4 go of 6.3 2045
5 P-7 5.6 1695
6 " 5.4 1520
7 none - 2430
8 " " 2310
45 Sapphire P-1 3.2 2130
46 (Linde) " 3.4 2010
47 " P-2 6.9 2265
48 " 4.0 2115
49 P-7 5.2 2104
50 " 5.0 2240
51 none - 2975
52 i2980
*Actual weight on 0.200" diameter disc.

**Average of 797 and 810.
364
TABLE 6-17
Dielectric Loss Data at 2 Gc on Body 992 (1750)
with Noted Metallizing.
Sintering Time at 1425 0 C Noted. No Plating.
Ceramic Discs Nominally 0.060" Thick
Wt. of
Sample Metallizing* Sintering
No. Metallizing- (m) Time Measured Q
17 none - 1/2 hr. 2560

18 " - 1/2 2675
25 " - 1 2590
43 " - 8 2085
44 " - 8 2585
11 P-I 3.4 1/2 1747
12 " 2.0 1/2 1339**
21 " 3.7 1470
22 " 1.9 1 1261
27 2.9 2 767
28 " 3.9 2 732
40 " 2.6 8 633
13 P-2 5.1 1/2 2118
14 " 6.2 1/2 1831
15 P-7 12.6 1/2 1970
16 " 10.0 1/2 1612***
23 " 7.7 1 1962
24 If 10.1 1 1982
41 " 4.8 3 2137
42 " 6.4 0 1660****
*Actual wt. on 0.200 diameter disc.

**Average of 1320 and 1358.
***Average of 1558 and 1670.
****Average of 1667 and 1653.
365
6.4.3.3 Series 3
Table 6-18 shows the results obtained on the

samples tested in this series. These were measured over a
four-day period with a number of repetitions in order to
check the reproducibility of the data.
6.4.4 Discussion
The losses measured in this test as discussed

previously, section 6.4.1, are created largely by displace-
ment currents. The losses in the entire ceramic, both in
the portion affected by the metallizing ard the unaffected
portion, combine to lower the Q of the test cavity.
Whereas the results were presented in three

series the discussion is given a different but more systematic
treatment. First the ceramic alone is considered, both raw
and modified by the non-metallic phase of a metallizing
paint. Then the losses induced purely by a pure metallic
layer on the ceramic are considered. An attempt is made to
segregate the metallic portion of a conventional metallizing
from the metallized ceramic by leaching with nitric acid.
The remaining altered ceramic is then subjected to various
treatments and the Q of the ceramic measured. This is
followed by a consideration of the above phenomena in
several metallizing paints and body compositions.
6.4.4.1 Ceramic Losses
Figs. 6.50 and 6.51 show the dielectric loss

factor of various bodies used in this study versus frequency
and alumina content respectively. The relative lowering of
the cavity Q with different disc thi.cknesses of Body A
94% alumina and Body H 99.5% alumina are shown in Fig. 6.52.
The increase in loss as a function of cavity Q

using manufacturers loss data for Body A, Body H and Linde
sapphire is determined in Fig. 6.53. From this plot, the
loss of other ceramic bodies either raw or metallized can be
determined. For instance the dielectric loss of 992-1750 is
determined as 6 x 10- at 2 Gc. The loss of altered ceramics
shown on this graph is discussed later.
The three points given for Body H represent the

spread in the experimental values of Q obtained in the three
series of experiments. This spread has been discussed
previously. The results follow the expected trends and so
will not be discussed further.
366
- U * in 7.* U U N
R
o 0 0
o 0
o H 00 0
0 H N0
H-- CNJ C
- 0 0 0,10 0o
ro 0 0rrlr0 L .
U) ) 0) H H CN N H-
04 U 0 0 00 0 .000 000000
5U (a o LOHOq LA Or C) 0000 N Df oNmo O)
(U C) 0 0 M 0c% rH rn 0 0 -
0H~rHYL m
(/ 1 HH H HH 00 Hra -4jc
HHi H
4-~) r
U) C)
C) E
E- 41i
fa a)
m a) C) C) u C)
1.4 U
>i E-1
Va 0 0 4J 0
o 4-'
CQ ro v VaFO
Q) C) a) C) C)
o :3 ror or
C) fu 4.) mU 4-)t
4J 4-J 4)i4J j .4-) .4-
AU) Qi) 0) 4) 11) a) Q)
H
00 ~0 0 '0
La40)(N-3-54. ~53 ~ EE
5. ECv 4
I) U) u 41 N -) 4J 4J 4 4-) 4-J 4) 4-1 '1 N
fa f '1 4j U
fU 4U4 J UrU JU '- 4)4)-W4 'U -W 41
4(a (u (d ( M41 ( 04)( 01 21
N
U) Q)-
4J) -.
H
w.
4-J V
E 3 ::1 u u 04 'o
U C) u u ~*-4 .,4' 14
C 4. 55s 5> 0
HHl 'U'1
m m C 0)4
4.3
A4 Ix 0 0
mo *- (a mUt (0 4-3
4) 4.3AA A H
--
w C: 0 D LAO3 hON 0 0)Or-i n t L &oN a0* o rV 0 C)

W) 10 8. *4 4 .4 . . *.4 8 QOH
(ai (a mm U0
4.3 > > VVr VLA.Ln r
r. HH HH H HH aHCJ% %"'0U
to r4 N~
r- CD NO r-r D0 0ODO N0O NO
367
0
00
-J
Ac lGc I~
FREQUENCY (CPS)I
Figure 6.50: Dielectric loss factor for selected ceramic bodies versus frequency. Values are from
manufacturers' data sheets.I
- 368 -
rM.
0
0
N -
4-x
13Lcao
0 O.hr
-369
200
05a (99
4-00
a 0
1600
20 , 510 610 710

Ceramic disc thickness (in.001 in.)
Figure 6.52: The dielectric test cavity Q as a functic>,i of ceramic disc thickness.
- 370-
10-2
Body A
0-
a+
U-D
00
371
If, lo
6.4.4.2 Altered Ceramic Losses
It was important early in the program to establish

the relative magnitude of dielectric loss introduced by a
non-metallic portion of a commercial paint in order to
determine if the experiment was a sufficiently sensitive
tool to perform the task at hand. Accordingly, P-3(g),
MnO-TiO 2 -SiO 2 , was painted on a raw ceramic disc. The
results are given in Table 6-19. It can be seen that a
significant drop in cavity Q occurred both as a function of
amount of P-3g reacted and with the time of reaction. The
first result is self evident. The second result is ascribed
to the P-3(g) melt partially remaining on the surface of
the ceramic during the first 1/2 hour fire period, and
continuing to diffuse into the grain boundaries of the
ceramic and to react with the alumina phase of the ceramic.
Thus the amount of melt phase increase. as a function of
time and produces additional loss.
The relative consistency of the Q value after

two hours at 1425 0 C indicates that the reaction initially
is diffusion limited but is subsequently amount, limited.
It should bc menLioned that a reduction in

cavity Q would also occur if a glassy phase approximating
the composition of the grain boundary phase of the ceramic,
MgO-CaO-SiO 2-AI20, were painted on. This is apparent from
Figs. 6.51 and 6.53. It is therefore of interest to
determine what additional drop in Q or loss factor is due
to a "lossy" glass phase, such as MnO-Ti0 2 -S-O2 being
introduced into the ceramic.
One milligram of P-3c applied to the 99.5%
ceramic disc (120 milligrams) is approximately 1% of the
weight of the body but the composite unreacted or partially
reacted (1/2 hour at temperature) body will behave more like
a 99% alumina body. After extensive reaction (2 hours) it
will then behave more nearl" like a 98.5% alumina oody as
additional alumina will have entered the melt grain boundary
phase. This would also be true of the latter paint
(MgO-CaO-SiO2-A1 2 0 3 ) if the A12 0 3 were omitted.
The three cases under consideration are outlined

in Table 6-19.
372
•- . " " . --
....- __ - . -
-. --.
u0 (N -4
u 00N o
H41 ro I Co
W 41 -4 ..
11 Q ( x it, (U
u >, -4 0 6 w.40n
(o UlA -4(0
4.4 u (
~ r44J I
L4
U 0 a)
4. (a -4 - >
uw0 -4 r-4J
0. x 4-' .14
41 04 U
E- 0 *U U )
(n 0) au
CV U)
'
(a
u
0 P >
4: 0C1 9:4 U)
do,.4 $N14 (U
LA (N I r-4 a) r-4
0l 0 0 4.) m
m Ln v' 0 r
r- -4 Col 04
C '4 4.4)
0 4.) 0 $
VO E-s clC 04
ON 0 '- U~4t V
(N >. 0 (a@
r.0 0 -.4' > to
.4 -- 4 '0 a C, 4) 0
~
U)4 4 0 a s'a'c
P-4 N.* 0
14 r-4tU4J lo _.4E
(a 4.4 A-44. >0 0 14

C 0 cc E'4. PDQ4
>1 :1~ C
01
4 s 0
4J U c k .4 0-4
C -.4 -4,41 xX41 0.c
P4 A ' 00 0j V N -
-' -4: 0, <
u~ 144 414
"4 ~4'UC 44
C 00 0
0fl E'~ 0U
.4 4: 454c 'oo 4
-'4 (n 6 E 1V
-o4 L: C -4 * 4
v-44
-4 .4
UQU U -4-4A
373
The loss fe.ctor of the equivalent ceramic body
with added CaO-MgO-SiO2-Al0 3 phase is given in Type I.
This is obtained directly irom Fig. 6.51. The corresrconding
losses induced by the MnO-TiO 2 -SiO 2 pha3e is given aS Type II.
This is obtained by taking the measured Q's of the discs
and determining the loss factcx from Fig. 6.53.
It can be seen that the loss factor of the body

is significantly increased over that of an equivalent MgO-CaO-
SiO 2 -A120 body of simi.Lar alumina content. The loss factor
is equivalent to that of a 94% and 90% body for the reaction
times of 1/2 hour and 2 hours respectively.
6.4.4.3 Pure Metallic Layer on Ceramic
Table 6-18 gives the losses induced using

evaporated copper metallizing (Fig. 6.54) or Ti-Mo-Cu
evaporated metallizing on the ceramic. Although the losses
tend to increase over that of the raw ceramic, the increase
is not considered significant. The contribution of any
Cu 20 or TiO formed to increase the dielectric loss is
negligible because of the small volume involved. This may
not be so for the conduction loss case previously considered.
When the copper plate is oxidized, #96, the contribution of
the CuO formed is then significant:
These latter results, #95, #96 also indicate
that I2 R losses are negligible in the present situation.
6.4.4.4 Conventional Metallizings on Body A and Body H
Having obtained a "feel" for the physical

chemistry of the loss process, it is desirable to consider
the effects of metallizing a 94% and 99.5% body with con-
ventional metallizing paints. A typical series of results
(Series I) is presented in Figs. 6.55 and 6.56. The follow-
ing sections consider all three series.
6.4.4.4.1 Pure Metallizing Paint P-2
Continuing the sequence of discussing systems of

increasing complexity let us consider a pure molybdenum
paint, P-2. As P-2 was ground in an alumina ball mill it
picked up 1% of CaO-MgO-SiO 2 -A1 2 03 . The same paint, P-2M,
was also prepared in a molybdenum mill ground with molyb-
denum balls.
374
Figure 6.54: Photomicrograph of Body H test sample (400X) #95 (500 A Ti, 5,000 A Mo evaporated,
Cu plated to 0. 17 mil)
- 375 -
7I
60
500
*
S40- E
E
C -P
.2 - 0'C
0
20 N -
10
FA.
Figure 6.55: Reduction in cavity Q resulting from noted metallizing paint only on 0.060" thick Body
A ceramic discs. Test frequency 2 Gc. Sintering time 1/2 hour at 1425 0 C unless noted.
Single surface metallized unless noted.
-376-
60-
50F
1 044-
C
UU
20-
0
10-U
La. CLI o 0
Figure 6.56: Reduction in cavity Q resulting from noted metallizing paint only on O.Oe60" thick Body
H ceramics 99.5% A12 0 3 . Test frequer.-y 2 Gc. Sintering time 1/2 hour at 1425 0C unless
noted. Single surface metallized unless n, led.
-. 377 -
The two paint batches showed no significant
variation in Q. (Table 6-18, #83, #84, #90.) For our
purposes, therefore, P-2 may be considered as a pure molyb-
denum metallizing paint. Any loss introduced into a system
using this paint must therefore be due to some change
induced into the system by processing, as it has been shown
above that a molybdenum layer evaporated onto the surface
does not induce additional loss. The specimens considered
in all three series #39, #40; #1, #2; #83, #84, #90 gave
lower Q values by 15 ± 5% than the Body H raw ceramic
indicating that some change in the chemistry of the system
was occurring.
The change noted with the 99.5% alumina was also

noted with the P-2 metallizing paint sintered onto Body A,
the 94% alumina body. Again giving 12% lower Q values than
the raw ceramic, the absolute Q values naturally being very
much lower.
The dc conductivity measurements described

earlier had also shown a lowering of the temperature coeffici-
ent of resistance for P-4M metallizing (similar to P-2) as
compared to evaporated metallizings. It is apparent there-
fore that the glass ,Dr glass-metal interface has induced in
it a semi-conducting and/or lossy material. X-ray diffrac-
tion analysis of the glassy phase of the P-2 metallizing on
Body A showed anorthite formation. This condition has been
shown to introduce a high dielectric loss 5 into a ceramic
over a ceramic containing only the undevitrified glassy
phase.
The same situation must hold for Body H also.
The amount of anorthite present is too small to be detected
by X-ray diffraction, however.
The anomalous low Q data noted in Table 6-14

can be explained on the basis of more extensive anorthite
formation.
6.4.4.4.2 Complex Metallizing Paint P-1
Considerable loss would be expected to be intro-

duced into a ceramic body by the MnO and TiOx components of
the metallizing paint. The dc coefficient of resistivity
results, section 6.2, showed that the reacted layer behaved
as a semiconductor, consequently considerable dielectric loss
would be expected to occur in this layer. Because of the
difference in the amount of glass phase present, Body H will
378
have a grain uoundary phase much richer in MnO and TiOx than
will Body A. This, coupled with the fact that the initial Q
of the cavity witl Body H is much higher, will result in a
much higher attenuation for the P-I Body H system.
A lowering in cavity Q of 53 + 5% is noted in
the three series (Tables 6-15, 6-17 and 6-18) for Body H.
The experiment was not sufficiently sensitive to detect
significant differences in loss as a function of the amount
of paint originally sintered on.
A silica rich SiO 2 - (MnO 2 -TiO 2 ) - CaO-MgO-A1 2 0 3
system, P-3g on Body H, is less lossy than P-I on Body H
(Tables 6-1.5, 6-19) despite the five-fold concentration of
the non-metallic phase. This result demonstrates the
modifying effect of a silicate addition on the loss by
removing the lossy MnO, TiOx phase present in the (P-1) -
Body H system. The MnO, AI 2 0 3 phase is retained however.
The loss of the resulting glass phase will be heavily
dependent on the oxidation state of the titanium ion in this
phase. Within the limits of normal metallizing cycles this
is determined primarily by the time and temperature allowed
for diffusion and only secondly by the amount initially
present, as has been shown above. The amount of surface
painted (i.e. painting botlh sides of a disc) will affect the
loss, however, because the volume if diffused material will
alter.
The system P-I on Body A displayed no significant

loss. In order to demonstrate that an increase in loss was
occurring it was necessary to paint both surfaces with P-1
metallizing and/or increase the sintering time for 2-1/2 hours
in order to diffuse the MnO and Ti02 further into the ceramic.
(The microprobe data in Table 4-51 indicated that the bulk
of the MnO and TiO 2 only diffuses 15 mils into the ceramic
in a 1/2 hour sintering at 1425 0 C.)
These latter results might have been predicted
from the results reported earlier in this section. The
dilution of the MnO, TiO2 components of P-i by 14% of grain
boundary silicate phase is a more extreme case than that
noted previously.
This dilution effect coupled with a very low Q

for the raw ceramic Body A indicates that dielectric loss
reduction in metallizing systems utilizing 94% alumina
ceramics is not feasible as the losses introduced by the
metallizing components are of the same order as those
originally present in the ceramic.
379
*.4.4.4.3 Leached P-I Metallizing on Body H
The d electric losses of the raw ceramic and the

separate constituents of the metallizing have been investi-
gated. It remained to be shown that the conduction losses
in an actual sintered and reacted molybdenum layer or any
"induced" effect on the non-metallic "glassy phase" was not
significant under the electrical conditions prevailing for
the "dielectric" experiment. This was accomplished by
leaching out the molybdenum with 50% nitric acid. In some
cases this was followed by sanding the layer with silicon
carbide or alumina. Typical microstructures that resulted
are shown in Figs. 6.57 and 6.58.
The results (±able 6-18) confirmed expectations.

Losses were not reduced by the leaching process. Lielectric
losses were progressively reduced by a continuing A1 2 03
sanding process, because the lossy dielectric was also
progressively removed. It was noted that sanding with silicon
carbide did increase losses. This is because of the 'lossy"
SiC particles which embedded themselves in the ceramic
surface.
6.4.5 Comparison of Various Metallizings
on +99% Alumina Bodies
Having established the loss mechanisms in

metalli~ed ceramics, the studies were extended to additional
low loss ceramics and employing the metallizings P-i and
P-2 noted above and also a low loss metallizing, P-7.
Body H is a 99.5% alumina having a 1:1 SiO2/alkaline

earth ratio whereas 992 (1750) is a 2:1 99% alumina ceramic.
Linde sapphire, of course, is 99.99% alumina with no crystal
grain boundaries.
P-7 is low loss because of the BaO-SiO 2 non-

metallic component which reacts with the alumina phase to
form a low loss glass.
In Fig. 6.59 photomicrographs of P-i and P-7
metallizing s ntered for 1/2 hour at 1425 0 C to Linde sapphire
ire shown. (TIW P-2 sample could not be polished because of
numerous pull-outs in the metallizing.)
_ie reaction is obviously chemically controlled
as denoted by the selective attack on the alumina crystal
phase and can be seen to have progressed much faster with
the MnO-TiO. melt than with the Bao-SiO2 melt.
380
Ie
i gr6.7PhtmcorpsoBoyHtssape 4 X)To. 1A0.6mloP1mealzn)

BotmI9A(. i-o P1mtliig oce)
3I
Figure 6.58: Photomicrogcaph of Body H test sample (dOOX) 078 S2 (0.5 mil of P-.] metallizing,
leached and sanded off).
- 382--
Figure 6,59. Photomicrographs of P.I (top) end P-7 (bottom) on Linde sapphire. Dielectric loss sOmples
sintered one-hl f hour at 1425"C (400X),
i
...
383-.-
I
In the case of reaction of these paints with the
992 (1750) ceramic, increased grain boundary diffusion of
the metallizing melt phase occurs as a function of time as
shown in Fig. 6.60. Also increased chemical reaction occurs
at the surface as a function of time as denoted by more
pronounced etching (selective solution) of the alumina
crystal phase and inc':ease in the amount of the interfacial
melt phase, Fig. 6.61.
It may be noted that the extent of diffusion is

a function of the composition of the grain boundary phase
of the ceramic. In the case of sapphire the diffusion is
zero. The grain boundary phase of Body H is much less
viscous than 992 ceramic so that despite the fact that
there is twice as much grain boundary phase in the 992 ceramic
the diffusing distance of the melt ions of the metallizing
paint can be expected to be much greater in Body H. (See
section 5.0.)
The loss results for sapphire, Body H and 992

ceramic given in Tables6-16 and 6-17 have been plotted in
Fig. 6.62. For sapphire the "lossy" phase is confined to
a narrow interfacial region, so the drop in Q is relatively
small and constant.
The reduction in cavity Q with P-2 paint must

be due in the case of sapphire to aluminum molybdate forma-
tion and to the 1% of MgO-CaO-SiO 2 -Al20 3 introduced during
processing. Additional "glassy phase has migrated into the
metallizing in the case of 992 (1750) and Body H. Anorthite
has been cited previously as the loss producing phase in the
above instance. Solution of molybdenum in the glassy phase
may also result in a loss producing phase.
As expected, the losses introduced by the use

of the P-7 paint is less than the P-1 paint in all three
instances. The increase in loss with either P-1 or P-7
paint is directly related to the extent of metallizing
melt phase migration into the grain boundaries of Body H
or 992 (1750) ceramics.
Further loss data pointing up the influence of

melt phase migration on diffusion is shown in Fig. 6.63.
It had previously been noted that with P-3g the loss
increased in Body H with 2 hours diffusion over 1/2 hour
diffusion. This situation was .lso found to hold true for
the data reported in Tables 6-15 and 6-17 and plotted in
Fig. 6.63. It thus appears that, although further diffusion
is taking place after 2 hours (Figs. 6.60 and 6.61) an
384
/ HR.
2 HR.
8 HR.
Figure 6.60, Photographs of sectioned dielectric loss specimens (0.200 in. wide by 0.060 in. thick)
showing P-1 metallizing diffusion into Body 992 (1750) at 1425'C with Increeslng time
(IlOX).
"w-~~~~~~~~~
V O 385WP 1 rWWvoN ,
1/2 H R.
2 HR.
S HR.
0
Figure 6.61: Photomicrographs of P-7 metallizing reaction with Body 992 (1750) at 1425 C for
increasing times (400X).
- 386 -
300 Spphire
992(0750)
Body H
2000
Body A
1000 -
i I I1
Unmetollized P-2 P-7 P-1

Cermic
Figure 6.62: Dielectric Ios. tc-!t cavity Q. with various ceramic samples in gap. The unmetallized and
metallized samples were fired at 1425°C for 1/2 hour.
-387-
Body 992(0750)
300 1
5 00
Cf ~UNMETALLIZED CONTROLS
2000-
METALLIZED WITH P-7
Sintering Time (Hours)
Figue aviy
661 Qielctrc
vesus1425C
lss tst frin tie. eraic ws Bdy 92 175) i
all cases
-388-
F
equilibrium loss situation is reached after 2 hours,

indicating that the reaction is amount limited.
The losses are determined from Fig. 6.53. It

can be seen that the P-1 loss is an order of magnitude
more than P-7 metallizing.
I
I
i
I
i
I
I
I
I
I
I
389
I
6.5 HIGH POWER MEASUREMENTS
6.5.1 Introduction
Recapitulating, the objecti-.e of Task III is to
investigate the effects of the loss-producing variables of
a ceramic-to-metal seal on the electric losses in radio-
frequency fields and to arrive at the design of low-loss
seals.
In low power rf measurements, losses due to

degradation of the dielectric qualities in the ceramic next
to the seal region were distinguished from losses due to
conduction currents parallel to the seal in the metallizing
and overplate. The results did not indicate the relative
importance of the two types of losses or the relative
contribution of the seal losses to the total dissipation in
a waveguide winlow.
The first set of calorimetric measurements of
ceramic-to-metal seal losses of representative configurations
using a 10 KW,8 Gc rf source were performed in the 9th
quarter of the contract in order to provide absolute loss
data. Those measurements gave us our first indication of
the magnitude of the total losses occurring in a ceramic-
to-metal seal and demonstrated our ability to compare seal
losses with the rf losses in an empty copper waveguide.
The remaining problem was to measure the absolute
amount of dissipation due to dielectric losses and conduc-
tion losses separately in typical seals and to attempt to
minimize such losses. Accordingly, these two modes of seal
loss were separated and measured independently in a second
set of measurements carried out during the llth and 12th
quarters of the contract.
The equipment and its operation discussed in
this report deals mainly with the "set-up" used for the
second series of experiments. This "set-up" is a modifica-
tion of that used in the first series of experiments. The
first "set-up" has been adequately described in the 8th
and 9th Quarterly Reports.
390
6.5.2 Calorimetric Test Cell
The copper test cells used in the first set oi
calorimetric measurements each had three cooling channels.
The center channel cooled the test seal, while the outer
two channels served to isolate the test cell from the rest of
the waveguide and to give an empty waveguide power dissipa-
tion value. A detailed drawing of this test cell appeared in
Fig. 4.01 of the 9th Quarterly Report. Seals tested in this
group were all made using P-1 metallizing paint and 95.5%
alumina substrates (Body I) .025" thick. The metallizing
thickness ranged from about 0.1 mil to 1.7 mil and eith.er
nickel or copper plating was used. All brazes were rnade
with copper-silver eutectic (BT). The waveguide walls to
which the four thin dielctric plates (constituting a
windowette) were brazed were .014 inch thick. The original
waveguide wall had been reduced to this dimension by machin-
ing away part of the outside of the waveguide to provide the
water cooling channels. The water channels were completed
with Lucite cover plates.
In the second set of measurements, the waveguide

test cells were extended to have five cooling channels each
in order to accommodate a larger number of test seals in
each test cell. Fig. 6.64 is a photograph of a completed
unit; its cross-section is shown in Fig. 6.65. Compared
with previous urits, only one other minor change was intro-
duced; the remaining wall thickness of the waveguide i-n the
channels was increased to .024" from the previous .014".
Using the two outer channels for cooling purposes and insula-
tion against heat influx from the flanges, the three inner
sections were available for the placement of test seals.
All were connected in series and the rf energy dissipation in
each "windowette" was measured simultaneously in the set-up
described below in Section 6.5.5.1.
The ceramic used in most of the 2nd series of
seals was a 99.5% alumina, Body H, in the form of 0.050" thick
substrates. Two cells with .050" thick 99.8% EO ceramic,
Body BA, and one with the previously used .025" thick
Body I substrates were also tested. The change in ceramic
was programmed in order that dielectric losses might be
noted. The first set of measurements showed that the loss
of the 96% alumina dielectric swamped the losses introduced
by the metallizing of the body. Metallizing formulae, braze
types and methods of applying each were chosen in order to
verify or disprove various hypotheses on the nature of seal
electrical losses as well as to develop low loss seals. These
choices are listed in Table 6-20.
391
Figure 6.64: Five-channel waveguide test cell used in second series of high power calorimetric
loss measurements.
0
01E
10
0 cr0
00 '0
0 -0i-~~
40 4)
400
IN?
d II
0 '0
WAO mmI rmo

V. 41 to 4J4
0r m 14 c La
.r4 -~4 4J 4 r-44.4 4) 4
0
u 4. r-I4) ?1 C0 .4jV (a
0) *4)to 0 0 41 4) C: )
044 0 11 u m 4V '0 t>.
0) 0)I4 9 4 944r-4 0
M 4) :
4.1 A,~ C:14 A(a V4 V w

(a toW 4 >044
r. > o >t
~ m 0A ) 0 rq0 w w 0
0 A0 0w0 4)W 4w0
:3 $4 i4) U) 124 CAOa 44 M 04-4
00
to I9
$44
4) ~ H IW*94
44 4) s * 0-4
0- r4A0
-0I Wr"4i $4
if r4A
I~ $4Ar4r F04
0 lo-4~ I Dm g0
0tA0.4 Eu% dr 94'0 '4'AOg
r4 "49- "4 4"h0 Ha"4 "4 n

N ~~ ~ Ur4z0u0Z0
~ ?a 4 o ODV 0-
.1 41 I ~ 11*
'0 04m.0aOAO.. .04Oi"
"4 V4 9P 1
0i I 'D 1 0
r4 00.4
10
04 0 4 s- "4 *4 "
"0 4 0 44X 4 P4 3
9 2
uz M U iI
U) U) U)
0 0)
- -w 0 w0
)
4a 4-)ai 4.) . U). J 4
3 rCJ 4 $4. ul to $4U )) 4 14 E)

0 a k) 410 1 41.,,4(a 4 ) (4r 41 r. En 4. .U)
.F 0w
Q) 0 0u r-q41~0 0' WOO 0 WOO 0 0 OO
41 ~O
w .r -I ) 90 - 4 r- r4 0F. -P44 E
U4) EO4 -I
0)Eu w ( u~It u- w uu u Q)u u 0w WCJu
'-4 0) V 4-4>
A A >4 fl r*- 10 ro.4 '
uX 'u.w 1 0 MO 4' 41 W 41 O W r 41 :4
rUU
Et Q Elou tr r 0u w al Fi 0u OHr- E0u
o (0 -1U) $4 $.4 4J -4O
0 4- 4 "H): r4~ H W- W r-
'44 4JNW44 044 N~ H:
- W N44 Q Q VN' 440 W Ir oU 44WW0
wHOq .41 4H Q m~ r4 -H 4J*-r$4 4 4 1 wWNl JH
r-4 - M WU) r-4 (a 0V> V -4 tUm r m -4 ):m v U) t ~m
o r-4 O0 WQ) r- -HO) 4VF40$ . O
0-4
U) U ) U ) a > U) m o(F. cma5U
V O rato
fa-44j y r 4 nU) 414 0 co Q4J.i 04 .000N Oa4 r. 0O
(D .cQ)"--w 0 0 4 0 0 0- -" )$44 -,4W W $4 -r4 HD
-4 14
SE-4Eg wH E 12 r43:M44 ~U44a-4
3: C H :U44
N < ~ <: 4 4 4: <:~
m u u~ u u U u U u U
U)
0
CL-4--4C -r4 a -E ~C r: 0.4 4.14H 1.14.1 )HW Hr VH4H il), H L -
M ~E Q4E 0E
04 0 0 12 oC .SSO E0 d 0 = 1
r-I~ 0 0 0 UHIr-4 0 o 0 0 0 0 0 0 0
04E-4 U U-4U H (UU-4 U -UPUP-4 U in UtjV)
N C ~~0 0404t Z n4
V-4 U) r-4 0- F-4 P0 P44 94 P-4

r-44 -H 'piO
H4 H P4 H4 H eP4 H4
m- c- P-4 E 4E F 4 V9.4 'P4 rOEOm
.9.4 .9. .9. .E.
.41 4 i I I 1 1 * I I > .1I
C) cC64 OC'.~N iQ .H OdH4 r4 CLH
Oe4 3E
CW
E 0>54 0 0 00 01 0 0 0 0
mu >4 0n >Nu1 >Itn >1 ul>tnNo~n >
w -4 0
m 06 m '0
EE -4'U 0-4go0J
1w 1o % 34 00 004 % 0
opI 4 0 0
H N -4 NEv InN
u z
395
I U)I U) 4
V 0H 0 H
4-) (a u41i H (
.
UH i-4 HiU(O
) a HI
94 )):) 'q I n
~H
3 .r, to *r4 U) H- 0 7 .
o vo 0 ) . 4 HIr.( CE1 4) O 0 0
-ri 0 0 )0 0 - O 0 0 4)H- 0 0 r- :3r m0l )g
-P k I
> qUi - 4z> -I.) r- 4d -4 u >Iu~
u0 0 4 4) 40) 40 40 0W
4-)4 -4 u
H *'iU uQ-- 4) uu uk 0 04-0
Q 'u"-
w 4 ' h r 4.)0)~ V~>
4 c 04 4r) )
(n VH-1
3. o rqH
c-1
4
1:)
3
k
- >140 41V0
r4 0)j4A4J0)
0
0
1> r0. -
00 V0u
k>N ( J
o (r O OI (D U. H4
0 04 r4 U Hr) 0 a
'r -'-) Hj V0-14 4) V * -- 0 0
0 4 .I-i4J - $ 4.) 4J
*--4 $4 >i (V
r V '0sr> Al LP 00 aO vr 'a 0) V 0> 0
o VCr.' 13 k
ro :1 Ire H. 0-0 -I 1.40
M) C~ 0> &0IVU
uN9P aN9 4 E*4
-'0~ >WOIN00
Q) ,*o 0 p.e 1- )w4 r 4 ).~
04 rj- 4 $4 0 ~0 -4 4
14 .44r-
-P4 a4 U )tW EnH
UH0 (2En44 En -H4J4.44J
-P Il 1 Z
0
tPa) 4 *I*4
SU H U U
0 0 0 0 UH 9 n
r-4 r-e '.H4 r-

r- u 0-.) 0-J Ene )- rqi f-I H -I 4ri
*- -r4 4H 04- ~ OH OH O O H
1 r 0 0N 10 0 0 0 adE
1- A 1 4 - 00 a 0
a) -lI H H 0 0 04 0r 0 r4iA
rl4U OH4
v-4H
5 00 H 0 0
>r0 (A U)
4J OWI Pcf4 -'6 -4 r-I- *.- *e-4r

H'r- mH rq.~ m H- HJr4
M,- HQ -r .r4:
4-SE- rv*'
HUV 0 fr- if i (d 5 V ~rI
' M >4 ~ 44
0 M 0 H 0 0 04
O!~~~ 0' 0.000 ~0
W~N N) m)
U) U)U ) )U
4J
uo ,HiI H 396
T 0 H >W
Test cell #1 contained the .025" thick Body i
ceramics to provide a check on the previous set of measure-
ments. Test cells #2 through #7 and #10 contained Body H
substrates, and test cells #8 and #9 contained the Body BA
substrates. A study of Table 6-20 shows that some cells
were intended to compare different metallizing preparations
or brazes. Other cells used the same type of seal through-
out, but included three geometric configurations:
(1) seals on all four walls of the waveguide, (2) seals on
the side walls only, and (3) seals on the broad walls only.
Total sial losses measured in (1) could then be broken down
into conduction losses measured in (2) and combined conduc-
tion and dielectric losses measured in (3). A position in
each of several test cells was also left empty so that energy
dissipated per unit length of the empty copper waveguide could
be measured.
6.5.3 Cold Testing
Test seals used in the first set of measurements

were cold tested for the VSWR (voltage standing wave ratio)
and matching frequency., as a function of the frame width,
i.e. the dimension of the ceramic which runs parallel to
the axis of the waveguide. These tests indicated the
dimensions of the frames to be used in the 1st series of
caloremetric tests. A VSWR of just a fraction greater than
unity was obtained.
These values were also of use in determining the

best frame width to be used in the second set of tests.
This procedure was described in Section 4.3 of the 8th and
in Section 4.2 of the 10th Quarterly Reports. Frame widths
selected on the basis of these tests were .830" for Body I,
.680 for Body H, and .710 for Body BA.
The notable difference in frame width between

Body I and Body H was necessitated by the difference in
their thickness, .025" and .050" respectively and their
dielectric constants. The waveguide wavelength of 1.749"
(at 8 kmc) was reduced to 1.360" in the section containing
the .050" thick dielectric frame.
The VSWR of each test cell in the second group
was measured between 7.5 and 8.5 kmc to detect spurious
reflections and resonances set up by interaction between
the seals and thus assure safe operating conditions for the
klystr.n and to select the best operating frequency near 8 kmc.
397
The test cell VSWR was first measured on each
cell individually, then on groups of three or four cells
assembled as they would be later in high power loss measure-
ments. The results of all VSWR measurements made on the
five-channel test cells are presented in Table 6-21. The
resonance frequencies in the second column were the fre-
quencies at which the test cells exhibited the minimum
voltage standing wave ratios as shown in the third column.
These numbers, like those for the VSWR at 7.9 kmc in the
fourth column, were always the upper limits, because the
VSWR may be influenced by reflecting obstacles in the
microwave circuitry other than the dielectric frame.
Other reflections may occur at the termination of the wave-
guide and at waveguide flanges. It was on the basis of the
VSWR measurement results that the klystron operating fre-
quency of 7.9 kmc was selected for the second set of high
power tests.
6.5.4 Calorimetric Instrumentation
6.5.4.1 Principles of Calorimetric Measurements
Calorimetric indication of microwave dissipation
in dielectric-to-metal seals is the direct way of measuring
losses absolutely. It also closely simulates the practical
solution to the problem of cooling high power microwave
windows. Instrumentation is simpler than is that employed
for the low power microwave measurement of seal losses
using cavity techniques. The price paid for this, however,
is the requirement of relatively high microwave power for
the experiment. Using propagating waves in waveguides and
expecting a dissipation in the test structure on the order
of 0.1 percent, a microwave source delivering 10 KW must be
used.
6.5.4.2 Preparation of Calorimetric Measurements

Microwave energy dissipation in the dielectric-
to-metal seals was determined by measuring the temperature
rise of the cooling water upon passing through each of the
cooling channels of the test cell. The instrumentation was
designed to measure reliably a minimum of one watt; refer
to Figure 6.22 for the set-up used in the second set of
measurements. Two differential thermopiles* Model 189A and
one Model 389A were used to indicate temperature differences
in the three central channels of the test cell. The thermo-
pile output voltages were routed through a selector switch
*From Sierra Electronic Corporation

398
TABLE 6-21
Test Cell VSWR Measurement Results
Cell Resonance Minimum VSWR at

No. Frequency VSWR 7.9 Kmc
1 7.57 Kmc 1.00 1.14
2 8.10 1.00 1.08
3 8.22 1.03 1.035
4 7.54 1.00 1.11
5 7.60 1.04 1.14
6 8.32 1.00 1.13
7 7.65 1.04 1.08
8 7.67 1.00 1.06
9 8.01 1.00 1.01
10 7.57 1.04 1.07
1-2-4 7.89 1.035 1.04
3-5-7-10 7,92 1.00 1.03
9-6-8 7.87 1.05 1.06
399
and indicated on a Hewleti-Packard Model 425A dc Micro-
Volt-Ammeter.
The heart of the calorimetric measuring and
cooling system was the constant temperature water container.
This reservoir contained a thermostatically controlled
electric heater, a small mixer, a water cooling coil, and
a pump to circulate the water through the entire test
system. A mixer-buffer container was placed in the water
line between the reservoir and the test cell to prevent
surges of water of a slightly higher or lower temperature
than desired. The water was then routed through one outer
cooling channel of the test cell, through the cold arm of
the first thermopile, through the first test seal cooling
channel, through a short length of lucite tubing containing
a four-watt electric heater, and through the hot arm of the
first thermopile. This sequence was repeated through the
next two test channels, then the water was returned through
the other outer cooling channel to the reservoir. The
small heaters in the water stream could be turned on or off
to permit the operator to make a calibration check on the
thermopiles during operation without otherwise disturbing
the set-up. Water returning to the reservoir flowed into
a 100 ml graduate to provide a means to measure the flow
rate. The period required to fill the graduate was timed
with a stop watch. A switching box was used to energize
each heater separately as well as to direct the output of
each thermopile to the microvolt-ammeter.
All three thermopiles were calibrated in terms
of output voltage per degree C of temperature difference
of the water flowing Lhrough the two branches. The thermo-
piles were calibrated before their use in the high power
measurements using the temperatures and flow rates actually
employea in the tests. There was some minor difference
between the two calibrations, but the preliminary calibra-
tion permitted some calculations during testing. Both
calibrations were made using the set-up shown in the
diagram, Fig. 6.67, and the photograph, Fig. 6.68, which
consists of two constant temperature water circulation
systems.
Water in one system was run through the hot side
of the thermopiles while water in the other was run through
the cold side. The final calibration was made using 32.8 C
as the cooler temperature and varying the temperature of
the warmer water stream from 32.7 C to 33.7 0 C. Using data
thus obtained, the following thermopile outputs were
400
HIGH POWER TEST SET- UP
CALORIMETRIC MEASURING CIRCUIT
Mixer
Constant
Mixer- Temperature
Buffer Water
Container
Pump
€ Heaters Coe
OUTER
COOLING
oo,.~2
,3,AI
TEST SEAL
COOLING
EST SEAL
f"COOLING I
TEST SEAL
COOLING
Io, OUTER
COOLING
Ts
Telt
CHANNEL CHANNEL CHANNEL CHANNEL CHANNEL Cl
H AT
.AT H
F
AT
lIn ]
H
Diffrential
Thermopiles
Heater Thermopile 0 -
Power Selector DC Microvolt -Ammeter
Suty switch H P Model 425A
Switch
Figure 6.66: Schematic diagram of test set-up for second series of high power calorimetric loss
measurements shown also in Figures 6.70, 6.71, and 6.72.
-- 401-
THERMOPILE CALIBRATION
ini
2 L
H 38A
H 189A, CH k
Swi ,Ch
OC icovlt- m__e
OC MWerovolt - Ammeter
N P Model 425A
Figure 6.67: Schematic diagram of set-up used to calibrate thermopiles for the second series of high
power calorimetric loss measurements.
- 402 -
,-,,m-
now
Figure 6.68: Ph~otograph of thermnopile calibration equipimnt diagrammned in Figure 6.67.
- 403 -
calculated:
Model 189A, Differential Thermopile (used in Channel
one): 1.03 mv/degrees C
Model 389A Differential Thermopile (used in Channel
two): 3.16 mv/degrees C
Model 189A, Differential Thermopile (used in Channel
three): 1.02 mv/degrees C.
With water as the coolant, the power, P (in watts), dissipated
in each channel of the test cell is related to the flow
rateF (in ml/sec), and the output voltageV (in millivolts),
of the differential thermopiles through the following
equations:
Channel 1: P = 4.062 x V x F (Model 189A 1 ) (12)
Channel 2: P = 1.325 x V x F (Model 389A) (13)
Channel 3: P = 4.110 x V x F (Model 189A 2 ) (14)
6.5.5 Calorimetric Measurements
6.5.5.1 Experimental Set-up for High Power
Microwave Measurements
Experiments with high power microwaves were
performed in one of Eitel-McCullough's test stations for high
power klystrons. The block diagram in Fig. 6.69 shows the
arrangement. The 10 milliwatt CW signal from a Hewlett-
Packard 6933 Sweep Oscillator set at a frequency of 7.9 kmc
was amplified in a Hewlett-Packard 493A Microw:;e Amplifier
with an output of 20 to 300 milliwatts. The final amplifier
was an Eimac X3050 tube, a tunable 5-cavity klystron ampli-
fier with a C.W. output of 20 kw.
Power was transmitted through all the test cells
(three or four) connected in series and was absorbed and
measured calorimetrically in a water load. Dissipation in
each cell was neglected against the total transmitted
power. The closed loop calorimetric measuring system
described in the previous section (including Fig. 6.66)
was connected Ao the cell under test. Working at a power
level of 10'lkw required cooling of the ot.er parts of the
transmittinq line, the tube-to-waveguide transitions and the
test cells not being measured, in order to protect the
circuitry from overheating. This was accomplished by an
404
HIGH POWER TEST SET-UP MICROWAVE CIRCUIT
hp 693 1
Sweep I10 MILLIWATT
Oscllator
-I
hp 493A Po e
Microwave TO 20 TO 300 MW
Amplifier Supply
II:
Elmac X3050
KlysIron AmpSfer
11 Power
supply
UP TO 20 KW
I
Test Cell Cooling
Test Coll Electric ifferemWoll Closed Loop

Cl eatersNO
Test Cell Meters Motor
iiii
I
T ColCooling
=-
WOW Load-
of high power celorimetric

Figure 6.69: Block diagrm of microwave circuitry used in second seris
toss relsuremont,.
Raw05-
I
independent secondary cooling circuit routed through all

transitions and all five channels of each waveguide cell not
under test.
Detailed views of the test arrangement appear in

Figs. 6.70, 6.71 and 6.72. Fig. 6.70 is an overall view of
the test site with the klystron amplifier in the center-
rear, four test cells with attached cooling water tubes
extending to the right from the klystron, and the three
I
differential thermopiles suspended above the test cells.
The stand in the foreground holds the constant temperature
water container, which is immediately to the right of the
technician's head. The technician's hand is on the dc micro
I
volt-ammeter used to measure thermopile output. On its right
is the panel containing the heater and thermopile selector
switches and the heater ammeter and voltmeter. In front
I
of the switch panel is the stop watch used for timing the
water flow. j
Fig. 6.71 is a close-up of the four waveguide
test cells in Fig. 6.72. The black object in the lower
extreme left corner is the mixer-buffer container from
which water flows to the cell under test. The klystron is
I
at left-center. The three differential thermopiles in the
upper-center are suspended in an inverted position from their
support bracket. The two hoses draped over the thermopiles
I
are part of the secondary cooling system and can be seen
joining hose manifolds on two cells not under test. All
the hoses directly below the thermopiles are part of the
closed loop cooling system for the cell under test. It
should be noted that only two of the three heaters indi-
cated in Fig. 6.66 are present in the photograph. (Lucite
tubes with square lucite end plates.) Heater #1 had been
I
removed for repair and was not replaced until after the
photograph was taken. Wires from the thermopiles and
heaters conne
bottom of t to the switch box just out of view at the
the photograph.
I
Fig. 6.72 is another close-up of the set-up
showing the Hewlett-Packard signal generators in the left
background. The HP 693B Sweep Oscillator is on the bottom
of the stack, and the HP 493A Microwave Amplifier is above
it. The smaller instrument :t extreme top left is a volt-
I
meter. The klystron may again be seen at right behind the
row of test cells. The mixer-buffer container is the black
flask in the left foreground. A thermometer calibrated in
0.01 0 C increments is inserted through the top of the
container to measure input water temperature. Part of the
main water container is at the extreme left.
406
407 4
N.'.-4
AA
I4
By following the water flow in the diagram,
Fig. 6.66, along with c,mparison to and careful study of
Fig. 6.71, a good understanding of the calorimetric measur-
ing system may be gained.
6.5.5.2 Experimental Procedure
Essentially the same procedure was followed in
both groups of calorimetric seal loss measurements. The
first group of measurements was discussed generally on
pages 11 through 13 of the 9th Quarterly Report. Procedures
used in the second group are discussed here in more detail.
Each group of waveguide test cells was assembled
in the same order and orientation in which it had been pre-
viously cold tested for the VSWR. All water and electrical
connections were made as shown in Figs. 6.69 and 6.71.
Water from the closed loop system and the secondary cooling
system was circulated, all leaks and bubbles were eliminated
and a systematic check of all measuring instruments was made.
Experiments prior to measurements made on the
first group of test cells showed that the second order effects
of changes in the closed-loop flow temperature on the
measurements could not be tolerated. Constancy of tempera-
ture to better than 0.10 C was required. Therefore, an
operating temperature 50 C above room temperature was selected,
which was maintained with the thermostatically controlled
heater in the closed-loop system. Additional heat input
from the microwave dissipation was small compared with heat
losses to the environment and to another water cooler in the
main container. Thus, the heat to be measured did not
disturb the temperature equilibrium. Heat losses from the
waveguide, the connecting rubber hoses and the differential
thermopiles were indicated by negative readings of the dc
micro volt-ammeter. Later readings with the microwaves
transmitted were always referred to the negative reading,
also called the adjusted zero level, which was checked before
and after each experiment.
A water flow rate of 10 cc/sec (0.15 gal/min)
was used. It assured effective cooling and fast attainment
of equilibrium temperature and still provided measurable
temperature differences between the input and output of the
cooling channels on the order of 0.50 C.
Several readings were taken at each measuring

point to establish the zero reference condition with no
microwave power transmitted. Each heater was then energized
410
in turn and the readings were taken with approximately four
watts of power supplied as indicated by the ammeter and
voltmeter for the heater power supply. This procedure
provided data for checking the calibration of the thermo-
piles.
The klystron was then put in operation and
mi:-rowave power was transmitted through the waveguide test
clls. The t-ansmitted power was measured with immersion
thermometers and a flowmeter in a calorimetric waterload.
Representative values were: a temperature difference of
9.90 C at a water flow of 3.9 gal/min. at a power trans-
mission of 10.2 kw.
The klystron power was calculated as follows:
Ptube = 264 (T2 -T I ) (Fw) (15)

where: Ptube = tube output power, kilowatts
T2-T 1 = temperature difference of outlet and
inlet water, oC
Fw = water flow rate, gallons per minute
264 = constant including unit conversion
factors.
Klystron output power during each measurement in the second
group of measurements is listed in Column 1, Table 6-22.
The klystron was operated at 10 kw for all test cells and at
5(kw for repeated measurements on four test cells.
In determining the power transmitted through the
test cells, the progressive drop in power along the test
cell line due to absorption in each cell was neglected
agai,.t the total power transmitted. The total power ab-
sorbed by the assembled test cells . sociated flanges
and transitions was estimated (2/3 was measured directly)
at less than 300 watts. The validity of this approximation
was confirmed by placing several cells at different positi-
ons in the test cell line. No differences in power
abdorption were detected.
Several sets of readings were taken at all
measuring points with each heater on at the usual four watts
and again for each test channel with heaters off. Micro-
wave power was again turned off and another set of readings
was taken to recheck the zero power reference values. On
411
some test cells the entire procedure was repeated at half
the usual power level (5 kw) in order to provide a scaling
factor check on the calorimetric measuring system.
Power dissipation in the test cells was cal-
culated using the water flow rate in the cooling channels
(Column 2, Table 6-22), the output voltage developed by
each differential thermopile (Column 3, Table 6-22) and equa-
tions 12, 13 and 14 on page.404. This gave directly the
amount of power dissipated in the seal cooled by each
cooling channel. The water temperature in the mixer-buffer
container (see Fig. 6.66) was monitored to keep it within
the temperature range for which the thermopiles had been
calibrated.
The result of the preceding calculation, the
power loss per channel, was then refined in order to dis-
criminate between the power dissipated in the immediate
seal area from dissipation in the bare waveguide walls on
either side of the seal. This empty waveguide associated
with each channel extended to the center line of the ribs
separating the channel from adjacent channels, as shown in
Fig. 6.65. Since the bare walls were also cooled by the
water in the channel, their contribution had to be
considered. A similar calculation was used to separate
losses occurring on the empty walls where the partial seal
configurations were used. Calculations of this type
were made possible by measuring losses in a section of
empty waveguide. From this result, and knowing the total
length of waveguide cooled by each channel, (Column 6,
Table 6-22) the power loss per unit length of empty wave-
guide was calculated. At least one such empty channel
was present in each of the three groups of test cells in
the second set of measurements in order to provide empty
waveguide power loss data for each group independently.
Power dissipated exclusively in the seal was then related
to the power level of the klystron at the time of testing
and reported as a percent of klystron output power
(Column 5, Table 6-22). Percent klystron power loss per
unit length of seal was also calculated (Column 7, Table
6-22).
Another number called "enhancement of dissipation"
was developed in order to enable comparison of one seal to
another without the confusion of different amounts of loss
in the empty waveguide caused by changes in standing wave
conditions from one test group to the next. This number
put all the results on the same basis. Power loss per
412
centimeter in the seal (Column 7, Table 6-22) was divided by
power loss per centimeter in an empty waveguide (also
Column 7, Table 6-22) run in the same test group. Seal
loss was thus reported as a multiple of the empty wave-
guide loss, or the factor by which the seal enhanced or
increased power dissipation in the waveguide (Column 8,
Table 6-22).
In the several test cells which were designed
for measuring dissipation in the side walls or broad walls
exclusively for one particular seal composition, the ratio
of loss in side walls to broad walls was of primary interest
(Column 9, Table 6-22). The theoretical ratio of conduction
loss in the side walls to the broad walls of the waveguide
was determined by calculation to be 0.26:1. When tiLe ratio
determined experimentally was less than 0.26:1, the
additional loss in the broad walls was attributed to
dielectric loss which can occur only in the broad walls. If
the ratio was greater than 0.26:1, (which would indicate a
negative amount of dielectric loss) it was an indication of
an error in the data or an inconsistency in the seal struc-
ture which upset the ratio of side wall to broad wall
conduction losses.
In order to provide a check on the accuracy of

the calorimetric measuring system during each run, a
measured amount of electrical power (approximately four
watts) was supplied in turn to each of the three heaters in
the calorimetric measuring system (see Fig. 6.66). The
water flow and thermopile output voltage were recorded and
the measured power was calculated from these data. The
measured power was then compared to the power supplied to
the heater to obtain a check on the accuracy of the measuring
system and an estimate of the amount of heat lost from the
water as it flowed through tubes from the cooling channel or
heater to the differential thermopiles.
These values were not presented in the table

because there were so many of them; at least three for
every cooling channel. However, they were always very near
and less than the four watts of electrical power supplied
to the heaters and indicated that the calorimetric measuring
system was operating satisfactorily.
As a further check on the system, several of the
test cells were retested by placing them in a different
position in the test cell sequence. Any differences due to
electrical effects would thus be detected. The measurements
indicated that the measured seal loss was independent of
test cell position.
413
TABLE 6-
HIGH P(OER rf LOSS M
Window Tube Waveguide V

Frae Output Water D
Cell Posi- Seal Description Test Con- Power Flow e
No. tion See Table 6-20 Group fig. Pt,kw FV,
ac/sec
1 1 Body I,P-l,Ni,BT 3 Full* 10.6 10.2

2 Empty 3 Empty* 10.9 10.2
3 Body I,P-lCu,BT 3 Full 10.7 10.2
1 1 Body I,P-1,Ni,BT 3 Full 5.0 10.2

2 Empty 3 Empty 5.0 10.2
3 Body I,P-1,Cu,BT 3 Full 5.0 10.2
1 1 Body I,P-I,NiBT 1 Full 10.55 10.0

2 Empty 1 Empty 10.55 10.0
3 Body I,P-I,Cu,BT 1 Full 10.55 10.0
2 1 Body H,P-1,Cu,BT 1 Full 10.34 8.40
2 Body HP-l,CuBT 1 S.W.* 10.34 8.40

3 Body H,P-l,Cu,BT 1 B.W.* 10.34 8.40
2 1 Body H,P-l,Cu,BT 1 Full 10.01 8.2
2 Body H,P-!,Cu,BT 1 S.W. 10.01 8.2

3 Body H,P-l,Cu,BT 1 B.W. 10.45 8.2
3 1 Body H,P-l(Pol.),Cu,BT 2 Full 10.80 10.5

2 Empty 2 Empty 10.80 10.5
3 Body H,P-I(PoI.),Cu,BT 2 B.W. 10.80 10.5
4 1 Body H,P-7,Cu,BT 1 Full 10.55 10.76

2 Body HP-7,Cu,BT 1 S.W. 10.55 10.76
3 Body H,P-7,Cu,BT 1 B.W. 10.55 10.76
Column Number: (1) (2)

TABLE 6-22
HIGH POWER rf LOSS MEASUREMENT RESULTS

idcw Tube Waveguide Voltage Loss as % Cell % Power Enhanc-
L, Uutout Water Differ- Power Loss in of Tube Length LGst per of Dis...
'- Pow.r Flow entrla Cell Output Consid'rd. cm Seal/w
Po.
t.kw Fwg, V, mV P1 . watts % Pt L, cm %Pt/cm guid
cc/sec
LI* 10.6 10.2 0.41 17.0-1.31=15.69 0.1480 2.11 .0702 3.13

?ty* 10.9 10.2 0.48 6.49 0.0596 2.66 .0224 1.00
L! 10.7 10.2 0.435 18.3-1.32=16.98 0.1588 2.11 .0752 3.36
Li 5.0 10.2 0.180 7.A6-.6=6.86 0.1372 2.11 .0650 2.97

)ty 5.0 10.2 0.215 2.91 0.0583 2.66 .0219 1.0c
Li 5.0 10.2 0.205 8.59-.6=7.99 0.1600 2.11 .0757 3.4C
tl 10.55 10.0 0.418 16.96-1.38=15.58 0.1477 2.11 .0700 2.94

)ty 10.55 i0.0 0.505 6.69 0.0634 2.66 .0238 1.0c
Li 10.55 10.0 0.445 18.30-1.38=16.92 0.1603 2.11 .0760 3.1'
Li 10.34 8.40 0.46 15.7-2.29=13.41 0.1296 1.73 .0749 3.1-
1.* 10.34 8.40 0.73 8.12-5.67=2.45 0.0937 1.73(.332).0413 1.7..

1* 10.34 8.40 0.47 16.2-3.17=13.03 0.1260 1.73 (.736) .0990 4.1
L1 10.01 8.2 0.525 17.5-2.12=15.38 0.1537 1.73 .0888 3.7
I. 10.01 8.2 0.79 8.58-5.56=3.02 0.0302 1.73(.332) .0526 2.2-

1. 10.45 8.2 0.488 16.4-3.21=13.19 0.1261 1.73(.736) .0992 4.1
L1 10.80 10.5 0.70 29.8-2.84=26.96 0.2495 1.73 .1442 5.1.

)ty 10.80 10.5 0.582 8.1 0.0750 2.66 .0282 L.C
q. 10.a0 10.5 0.285 12.3-4.18=8.12 0.0752 1.73(.736) .0592 2.1
Li 10.55 10.76 0.53 23.2-2.34=20.86 0.1976 1.73 .1142 4.6

10.55 10.76 0.767 10.9-5.79=5.11 0.0484 1.73(.332) .0843 3.5
q. 10.55 10.76 0.46 20.3-3.24=17.06 0.1617 1.73(.736) .1271 5..
(1) (2) (3) () (5) (o) (7) (8
B Pagc Nc
EMENT RESULTS
ge Loss as % Cell % Power Enhancemt. Loss Ratio:
-r- Power Loss in of Tube Length Lost per of Dissip. Side Walls
*1 Cell Output Consid-rd. cm Seal/wave- Broad Walls
V Pl. watts % Pt L, cm %Pt/cm guide
17.0-1.31=15.69 0.1480 2.11 .0702 3.13 N.A.

6.49 0.0596 2.66 .0224 1.00 .0046
.0178
18.3-1.32=16.98 0.1588 2.11 .0752 3.36 N.A.
7.46-.6=6.86 0.1372 2.11 .0650 2.97 N.A.
2.91 0.0583 2.66 .0219 1.00 .0045
8.59-.6=7.99 .0174
0.1600 2.11 .0757 3.46 N.A.
16.96-1.38=15.58 0.1477 2.11 .0700 2.94 N.A.
6.69 0.0634 2.66 .0238 1.00 .0049
18.30-1.3R=16.92 .0189
0.1603 2.11 .0760 3.19 N.A.
15.7-2.29=13.41 0.1296 1.73 .0749 3.15 2.45
9.42+3.61
8.12-5.67=2.45 0.0237 1.73(.332).0413 1.73
16.2-3.17=13.03 0.i260 1.73(.736).0990 4.16 0.188
17.5-2.12=15.38 0.1537 1.73 .0888 3.74 3.02
11.62+1.57
8.58-5.56=3.02 0.0302 1.73(.332) .0526 2.21
16.4-3.21=13.19 0.1261 1.73(.736) .0992 4.16 0.229
29.8-2.84=26.96 0.2495 1.73 .1442 5.11 N.A.

8.1 0.0750 2.66 .0282 1.00 .0058
.0224
12.3-4.18=8.12 0.0752 1.73(.736) .0592 2.10 N.A.
23.2-2.34=2C.86 0.1976 1.73 .1142 4.80
10.9-5.79=5.11 0.0484 1.73(.332) .0843 3.54
20.3-3.24=17.06 0.1617 1.73(.736) .1271 5.34 0.3
(3) (O) (7) (8) (9)

Page No.: 414 C
TABLE 6-
HIGH POWER rf LOSS MEX
Window Tube Waveguide VC

Frae Output Water D1
Cell Posi- Seal Description Test Con- Power Flow ei
No. tion See Table 6-20 Group fig. Pt,kw Fs,
cc/sec
1 1 Body I,P-l,Ni,BT 3 Full* 10.6 10.2

2 Empty 3 Empty* 10.9 10.2 C
3 Body I.P-1,CL,BT 3 Full 10.7 10.2
1 Body I,P-lNi,BT 3 Full 5.0 10.2

2 Empty 3 Empty 5.0 10.2
3 Body I,P-1,Cu,BT 3 Full 5.0 10.2
1 1 Body I,P-1,Ni,BT 1 Full 10.55 10.0

2 Empty 1 Empty 10.55 10.0
3 Body I,P-l,Cu,BT 1 Full 10.55 10.0
2 Body H,P-l,Cu,BT 1 S.W.* 10.34 8.40

3 Body H,P-l,Cu,BT 1 B.W.* 10.34 8.40
2 Body H,P-1,Cu,BT 1 S.W. 10.01 8.2

3 Body H,P-l,Cu,BT 1 B.W. 10.45 8.2
3 1 Dody H,P-1(POl.),Cu,BT 2 Full 10.80 10.5

2 Empty 2 Empty 10.80 10.5
3 Body H,P-I(Pol.),Cu,BT 2 B.W. 10.80 10.5
4 1 Body H,P-7,Cu,BT 1 Full 10.55 10.76

2 Body H,P-7,Cu,BT 1 S.W. 10.55 10.76
3 Body H,P-7,Cu,BT 1 B.W. 10.55 10.76
Column Number: (1) (2)

TABLE 6-22
HIGH POWER rf LOSS MEASUREMENT RESULTS

dow Tube Waveguide Voltage Loss as % Cell % Power Enhanc
',.e Output Water Differ- Power Loss in of 'Ptu Length Lost per of Di,
- Power Flow ential Cell Out.put Consid'rd. cm Seal/,
g. Pt,kw FwgI V, mV P1 . watts % Pt L, cm %Pt/cm guip
cc/sec
1* 10.6 10.2 0.41 17.0-1.31=15.69 0.1480 2.11 .0702 3.1:

ty* 10.9 10.2 0.48 6.49 0.0596 2.66 .0224 1.0(
1U.7 10.2 0.436 18.3-1.32=16.98 0.1588 2.11 .0752 3.3(

1 5.0 10.2 0.180 7.46-.6=6.86 0.1372 2.11 .0650 2.9
ty 5.0 10.2 0.215 2.91 0.0583 2.66 .0219 1.0
1 5.0 10.2 0.215 8.59-.6=7.99 0.1600 2.11 .0757 3.4

10.55 10.0 0.418 16.96-1.38=15.58 0.1477 2.11 .0700 2.9
y 10.55 10.0 0.505 6.69 0.0634 2.66 .0238 1.0
10.55 10.0 0.445 18.30-1.38=16.92 0.1603 2.11 .0760 3.1
10.34 8.40 0.46 15.7-2.29=13.41 0.1296 1.73 .0749 3.1
10.34 8.40 0.73 8.12-5.67=2.45 0.0237 1.73 (.332) .0413 i.
10.34 8.40 0.47 16.2-3.17=13.03 0.1260 1.73(.736).0990 4.]
10.01 8.2 0.525 17.5-2.12=15.38 0.1537 1.73 .0888 3.,
10.01 8.2 0.79 8.58-5.56=3.02 0.0302 1.73(.332) .0526 2.;
10.45 8.2 0.488 16.4-3.21=13.19 0.1261 1.73(.736) .0992 4.3
10.80 10.5 0.70 29.8-2.84=26.96 0.2495 1.73 .1442 5.]
y 10.80 10.5 0.582 8.1 0.0750 2.66 .0282 1.C
10.80 10.5 0.285 12.3-4.18=8.12 0.0752 1.73(.736) .0592 2.]

L 10.55 10.76 0.53 23.2-2.34=20.86 0.1976 1.73 .1142 4.E
10.55 10.76 0.767 10.9-5.79=5.11 0.0484 1.73(.332) .0843 3.E
10.55 10.76 0.46 20.3-3.24=17.06 0.1617 1.73(.736) .1271 5.2-
(1) (2) (3) (3) to) (7) (8,
5 Page No
EMENT RESULTS
ige Loss as % Cell % Power Enhancemt. Loss Ratio:
-)-r- Power Loss in of Tube Length Lost per of Dissip. Side Walls
*-il Cell Output Consid'rd. cm Seal/wave- Broad Walls
.I Pi watts % Pt L, cm %Pt/cm guide
17.0-1.31=15.69 0.1480 2.11 .0702 3.13 N.A.

6.49 0.0596 2.66 .0224 1.00 .0046
18.3-1.32=16.98 0.1588 .0178
2.11 .0752 3.36 N.A.
7.46-.6=6.86 0.1372 2.11 .0650 2.97 N.A.
2.91 0.0583 2.66 .0219 1.00 .0045
8.59-.6=7.99 0.1600 .0174
2.11 .0757 3.46 N.A.
3 16.96-1.38=15.58 0.1477 2.11 .0700 2.94 N.A.
6.69 0.0634 2.66 .0238 1.00 00049
18.30-1.38=16.92 0.1603 .0189
2.11 .0760 3.19 N.A.
15.7-2.29=13.41 0.1296 1.73 .0749 3.15 2.45
8.12-5.67=2.45 9.42+3.61
0.0237 1.73(.332).0413 1.73
16.2-3.17=13.03 0.1260 1.73(.736).0990 4.16 0.188
17.5-2.12=15.38 0.1537 1.73 .0888 3.74 3.02
8.58-5.56=3.02 11.62+1.57
0.0302 1.73(.332) .0526 2.21
16.4-3.21=13 19 0.1261 1.73(.736) .0992 4.16 0.229
29.8-2.84=26.96 0.2495 1.73 .1442 5.11 N.A.
8.1 0.0750 2.66 .0282 1.00 .0058
12.3-4.18=8.12 0.0752 .0224
1.73(.736) .0592 2.10 N.A.
23.2,-2.34=20.86 0.1976 1.73 .1142 4.80
10.9-5.79=5.11 0.0484 1.73(.332) .0843 3.54
20.3-3.24=17.06 0.1617 1.73(.736) .1271 5.34 0.3
(o)
M() (7) (8) (9)
Page No.: 414 C
TABLE 6-22
Window Tube Waveguide VoltagE

Frame Output Water Differ-
Cell Posi- Seal Description Test Con- Power Flow ential
No. tion See Table 6-20 Group fig. PtAw Fw, V, mV
cc/sec
5 1 Body H,P-3(20EU),Cu,BT 2 Full 10.8 10.4 0.345
2 Body H,P-3(20EU),Cu,BT 2 S.W. 10.8 10.4 0.525

3 " 2 B.W. 10.8 10.4 0.325
6 1 Body H,Evap,Ti,Mo,Cu,BT 3 Full 10.1 10.1 0.665

2 g 3 S.W. le.1 10.1 0.83
3 3 B.W. 10.2 10.1 0.385
6 1 3 Full 5.24 10.1 0.317

2 3 S.W. 5.24 10.1 0.42
3 3 B.W. 5.24 10.1 0.21
7 1 Body HP-3(Pol.),Cu,BT 2 Full 9.7 10.3 0.295

2 2 Empty 9.7 10.3 0.425
3 2 Full 9.7 10.3 0.380
8 1 Body BA,P-1,Cu,BT 3 Full 10.2 10.2 0.32
2 " 3 S.W. 10.2 10.2 0.55

3 " 3 B.W. 10.2 10.2 0.322
8 1 3 Full 5.25 10.2 0.157
2 "3 S.W. 5.35 10.2 0.275
3 " 3 B.W. 5.35 10.2 0.167
9 1 Body BA,P-3,Cu,BT 3 Full 10.2 10.0 0.95
2 " 3 S.W. 10.2 10.0 0.857
3 3 B.W. 10.2 10.0 0.46
Column Number: (1) (2) (3)

TABLE 6-22 (Ccntinued)
Window Tube Waveguide Voltage Loss as % Cell

Frame Output Water Differ- of Tube Length .I
Test Con- Power Flow Output
ential Power Loss in Cell % Pt Consid'rd
L, cm I
V, mV P1 ,watts
Group fig. PtAw F,
cc/sec
Full 10.8 10.4 0.345 14.6-2.26=12.34 0.1143 1.73

2
10.8 10.4 0.525 7.23-5.50=1.73 0.0160 1.73(.332)

2 S.W.
10.8 10.4 0.325 13.9-3.13=10.77 0.0957 1.73(.736)
2 B.W.
10.1 10.1 0.665 27.3-2.1=25.2 0.2495 1.73

3 Full
10.1 10.1 0.83 11.1-5.20=5.9 0.0584 1.73(.332)
3 S.W.
10.1 0.385 16.0-2.93=13.07 0.1281 1.73(.736)
3 B.W. 10.2
10.1 0.317 13.0-1.09=11.91 0.2275 1.73

3 Full 5.24
5.24 10.1 0.42 5.62-2.70=2.92 0.0557 1.73(.332:
3 S.W.
5.24 10.1 0.21 8.72-1.51=7.21 0.1375 1.73(..36'
3 B.W.
10.3 0.295 12.3-2.03=10.27 0.1059 1.73

2 Full 9.7
10.3 0.425 5.8 0.0598 2.66
2 Empty 9.7
10.3 0.380 16.1-2.03=14.07 0.1450 1.73

2 Full 9.7
10.2 0.32 13.3-1.97=11.33 0.1111 1.80

3 Full 10.2
10.2 10.2 0.55 7.44-5.23=2.21 0.0217 1.80(.332

3 S.W.
10.2 10.2 0.322 13.5-2.82=10.68 0.1048 1.80(.73(
3 B.W.
10.2 0.157 6.5-1.01=5.49 0.1047 1.80

3 Full 5.25
5.35 10.2 0.275 3.7-2.74=0.96 0.0180 1.80(.33-

3 S.W.
5.35 10.2 0.167 7.0-1.47=5.53 0.1034 1.80(.73(
3 B.W.
10.2 10.0 0.95 38.6-1.97=36.63 0.3590 1.80

3 Full
10.0 0.857 11.35-5.23=6.12 0.0600 1.80(.33-
3 S.W. 10.2
10.0 0.46 18.9-2.82=16.08 0.1577 1.80(.73(
3 B.W. 10.2
(5) (6)
jlumn Nurtibe:: (1) (2) (3)
B
:ontinucd)
Loss as % Cell Power Enhancemt. Loss Ratio:

of Tube Length Lost per of Dissip. Side Walls
Power Loss in Cell Output Consid'rd cm Seal/wave- Broad Walls
Plwatts % Pt L, cm %Pt/cm guide
14.6-2.26=12.34 0.1143 1.73 .0661 2.94 1.73

i6.65+4.12
7.23-5.50=1.73 0.0160 1.73(.332) .0279 1.24 0.161
13.9-3.13=10.77 0.0957 1.73(.736) .0783 3.48
27.3-2.1=25.2 0.2495 1.73 .1440 6.43

11.1-5.20=5.9 0.0584 1.73(.332) .1018 4.54 0.45
16.0-2.93=13.07 0.1281 1.73(.736) .1007 4.50
13.0-1.09=11.91 0.2275 1.73 .1314 5.87

5.62-2.70=2.92 0.0557 1.73(.332) .0971 4.340.405
3.72-1.51=7.21 0.1375 1.73(. 36) .1081 4.82
12.3-2.03=10.27 0.1059 1.73 .0611 2.72
5.8 0.0598 2.66 .0225 1.00 .0046
.0179
16.1-2.03=14.07 0.1450 1.73 .0837 3.72 N.A.
L3.3-1.97=11.33 0.1111 1-60 .0617 2.76 2.21

8.50+2.18
7.44-5.23=2.21 0.0217 1.80(.332) .0362 1.62
L3.5-2.82=10.6C, 0.1048 1.80(.736) .0789 3.52 0.207
3.5-1.01=5.49 0.1047 1.80 .0581 2.60 0.96
3.70+1.83
3.7-2.74=0.96 0.0180 l.80(.332) .0300 1.34 0.174
7.0-1.47=5.53 0.1034 1.80(.736) .0780 3.48
38.6-1.97=36.63 0.3590 1.80 .1995 8.91
L1.35-5.23=6.12 0.0600 1.80(.332) .1002 4.48
5.31 0.38
L8.9-2.82=16.08 0.1577 1.80(.736) .1189
(4*) (5) (6) (7) (8) (9)
Page No.: 415
C
6.5.5.3 Results and Discussion
Data obtained in the first set of measurements

on the three-channel test cells were used to calculate the
amount of power dissipated in each seal section. These
results were then plotted (Fig. 6.73) to illustrate the
relationship between power loss in seals and metallizing
thickness for both nickel and copper plating. What was
clear from the data in Fig. 6.73 was that the magnitude of
power dissipated in the seal was determined primarily by
the resistivity of the conductive material in the current-
carrying zone immediately adjacent to the window dielectric.
This is as expectedas dielectric seal losses in a 95.5%
alumina ceramic cannot be distinguished from bulk seal
losses.
6.5.5.3.1 Characteristics of Ceramic-to-Metal Seals

Tested in First Test Series
Table 6-20 lists the characteristics of the

six seals tested in the first series. Waveguide window
seals were simulated with a frame of Body I (96% A1 2 03 )
ceramic plates of 0.025" thickness and 0.830" width
brazed into a rectangular J-band aveguide. The thickness
of P-1 metallizing (75% Mo, 21% MnO, 4% TiO 2 ) was varied
between 0.0001" and 0.0017". One half of the samples
carry copper overplate on the metallizing; the other half
is nickel overplated. The latter variation will allow a
comparison of results of these absolute measurements with
previous relative measurements of conduction losses involving
nickel overplated and copper overplated metallizing in test
samples.
The data on metallizing and plating thicknesses

were confirmed by sectioning samples. Fig. 6.74 shows
photomicrographs (400X) of sections of the seals in test
cell No. 3 (top) and No. 6 (bottom). In the upper picture
the thin metallizing is not easily distinguishable from the
nickel plate and is very irregular. The value of the
thickness given in the table, therefore, is an average.
The nickel overplate is intact. As an interesting feature,
copper can be seen to have precipitated on the interface of
the nickel overplate and copper-silver braze. Radio-
frequency losses are, of course, not affected by that. The
lower photomicrograph in Fig. 6.74 shows the seal in test
cell No. 6 with thick metallizing. The copper overplate has
been partially dissolved by the copper-silver braze.
Approximately 50 percent of the metallizing surface is
covered by copper, and the rest by a copper-silver alloy.
417
0 4
C +
0
+ Copper verplete
a * Nickel overploto
* X Empty woveguS
-J a
01
.001 .002
Metalizing Thickness (inches)
Figure 6.73: Power loss in dielectric.-to-metal seal section of waveguide test cell vs metallizing
thickness as a fraction of transmitted power.
-418-
Sod y I
4-Ceramic
I~TZ W,.- P-1 Metallizing
-Ni Plate
- '~ "4Cu-Aa Braze
Body I
40- Ceramic
IA 40- P-1 Metallizing
~ ~- CU Plate
I Cu-Ag Braze
I ~- OFHC Copper
[ Figure 6.74: Sectioned seals in test cells No. 3 and No. 6.
- 419-
I
As pointed out with many photos of sections in previous

reports, the degree of interaction between various com-
ponents of the dielectric-to-metal seals depends very
critically on details of the sealing procedures, especially
the hold times in the brazing step.
Both sections were scanned with an electron
beam microprobe for the elements Cu, Ag, Ni, Mo, Mn
(Fig. 6.75). Referring again to the top picture in
Fig. 6.74 the seal in test cell no. 3, some of the visible
features were clearly confirmed. These are the attack of
the copper wall by the braze material, and the precipita-
tion of copper from the braze on the n.ckel overplate.
1
The irregularity of the metallizing shows as an overlapping
of the molybdenum and nickel trace in Fig. 6.74 (top).
The width of the molybdenum trace indicates a metallizing
thickness of 0.4 mils. Manganese from the metallizin(,
diffused into the ceramic and was detected 150 microns
(6 mils) inside the ceramic accumulated in the grain
boundaries or glassy phase regions of the alumina. The
seal in test cell No. 6, Fig. 6.74 bottom picture, is
characterized by the thick metallizing, which shows as a
2 mil wide peak on the Mo-scan in Fig. 6.75 (bottom).
Again, manganese is shown to penetrate into the ceramic.
6.5.5.3.2 Test Results
Test data obtained were the dissipation in all
three cooling channels, and dissipation in the central
channel. The difference between the two values is attri-
butable to rf losses in the empty waveguide sections between
the seal and the flanges on either side. Dissipation was
related to the power transmitted and to the length of
waveguide. Fig. 6.73 shows the measured ratio of power
loss and transmitted power per travelled length of wave-
guide containing the seal plotted against the metallizing
thickness. Copper overplating results in losses increasing
with metallizing thickness to a constant value. The samples
with nickel overplating came down from higher losses to this
constant value with the thickness of metallizing increasing.
Point (x) for zero metallizing thickness indicates losses
measured in the empty portions of the waveguide test cells.
Compared with this value, dissipation in the sealed portion
is larger by a factor of about 2.5 for a metallizing thick-
ness of about 0.001". The trend of losses with metallizing
thickness is very similar to curves obtained for relative
conduction losses measured previously and shown in Fig. 4.03
of the 7th Quarterly Report. Conduction losses are, there-
fore, predominant in the total seal dissipation.
420
Trsr CELL
No. 3
ELECTRON Mlc~op.,oai ANALK5YSS

CODE: iA
aq'AZL C
(3 WAzESUIoi
Mn
Ea-r7 RON Mlc~opRoai ANAL Vsls
4) CSAqMC
BAZE
&
Figure 6.75: Electron microprobe analyses of seals In test cell No. 3 (top) and No. 6 (bottom).
-421-
6.5.5.3.3 Discussion of Results
The purpose of this section is to relate the
measured dissipation in the seal of a dielectric frame to
the power losses in the seal of a rectangular slab window
with a thickness of half a wave length. From the trend
of the seal losses with metallizing thickness and from the
pattern of the essentially undisturbed dominant mode, we
may assume that dissipation in the seal is due to conduction
losses caused by the wall currents flowing in the seal
area. The results with the dielectric frame are now to
be applied to the losses in the seal of a solid slab
window.
In general, when an empty waveguid. is filled
completely with an ideal (loss-free) dielectric, wall
conduction currents and the corresponding losses are
increased due to changes in the electric and magnetic
field distribution. Attenuation in J-band waveguide is
increased from 1.5 db/100 feet to 3.3 db/100 feet when
the empty waveguide is filled with a loss-free dielectric
of a dielectric constant of 9.3. In terms of power, this
is an increase of losses by a factor of 2.3. A resonant
window being only half a wave length thick will cause a
smaller increase of wall losses compared with a long
dielectric-filled section because it maintains a different
field distribution. The transverse electric field is
reduced only towards the center area of the window. It is,
therefore, reasonable to assume that the increase of wall
losses will amount to a factor less than 2.3, estimated
here to be 1.5.
A representative value of measured loss in the
seal section is 5.7 x 10- 4 /cm (see Fig. 6.73) of the power
transmitted. A rectangular resonant window being 0.73 cm
thick for transmission of 8 kmc, causes a power loss of
4 x 10-4. This number is to be multiplied by the above
factor of 1.5 to give 6 x 10 - 4 , which represents the
dissipation in the dielectric-to-metal seal of a solid
window.
Numbers available in the Iiterature for total
window dissipation average at 3 x 10- of the power
transmitted. Compared with this number, our results indi-
cate that the dissipation in the seal of a rectangular
slab microwave window can be expected to be about 20 per-
cent of the total heating.
422
The percentage of seal losses is larger for
windows of very low-loss materials and for windows scaled
down in size for use at frequencies above X-band.
The magnitude of loss in the seals is closely
related to the thickness of metallizing used on the ceramic
because this thickness comprised most of the zone in which
the skin current flowed.
When thick metalliziny was used, the conductive
zone consisted entirely of molybdenum metallizing and
whatever plating was used was outside the zone and did not
affect losses. When very thin metalliz*ng was used, the
zone consisted primarily of the plating, nickel or copper,
and the braze, thus the losses resulting from currents
flowing in this zone were determined by the resistivity of
the plating-braze layer. In the case of nickel, the losses
were higher because the resistivity of nickel and the
molybdenum-nickel intermetallic compounds (MoNi4, MoNi 3 ,
MoNi) which were formed were higher than the metallizing
alone. In the case of the copper plated seals, the losses
were low (approaching the loss in an empty waveguide)
because most of the skin currents were carried in the low
resistivity copper or copper-silver braze layer, and no
alloying involving molybdenum and copper or copper-silver
took place.
The high power 2_oss results presented in the
curves of Fig. 6.73 were remarkably similar to the data
obtained using the low power 12 R techniques described in
section 6.3. Such strong agreement in results obtained
from two very different measuring techniques was evidence
that the results were most dependable.
Data obtained in the second set of measurements
on the five-channel test cells were used as in the first
set, but the cells were designed to indicate the relative
magnitude of dielectric ard conductive losses as well as
to compare several seal compositions according to their
relative lossiness as summarized in Table 6-20. The results
of the second group of measurements appear in detail in
Table 6-22.
Test cell #1 was constructed so as to provide a
tie with the loss versus metallizing thickness data which
was obtained in the first set of measurements which is
presented in Fia f. 73. A detailed description of this
seal is in Tabl.. *-22. This test cell was to provide
additional data in the one mil metallizing portion of the
423
... . ...
... . .... -- , .. .. p -, -2"
curve in Fig. 6.73 by comparing nickel plating to copper
plating in full window frames with all other variables such
as one mil metallizing thickness and Cu-Ag braze the same.
The results from runs under three different
combinations of power level and cell orientation (see Table
6-22) all indicated the seal with copper plating had the
higher loss and the loss in both types of seals was higher
than the results from the previous set of tests. It was
expected that the nickel plated seal would have the higher
loss (see Fig. 4.06 9th QuF terly Report) but, as it did
not, metallurgical examinations of the copper plated and the
nickel. plated seals were made in order to seek an explanation
of the results.
Photomicrographs of polished cross-sections of
both seals appear in Fig. 6.76. The scal with nickel plating
appears in the upper photo. The metallizing is about 1.2
mils thick, but the nickel plated layer has been attacked
by the Cu-Ag braze and only a thin layer remains. Nearly
all of the skin currents would flow in the metallizing, so
the 12 R losses would originate there. The seal with copper
plating is shown in the lower photo. The metallizing is
the same thickness as is the other seal. The copper plating
is in better condition than the nickel plate, but there are
many braze gaps. In this seal, the I-4R losses would occur
in the metallizing layer as in the nickel plated seal.
Because of the metallizing thickness involved, the difference
in loss between the two seals was certainly caused by a
factor other than the type of platin_. The higher loss in
the copper plated seal could be an effect of the braze gaps
or might be attributed to a higher resistivity of the
metallizing which could not be determined by a metallurgical
examination. However, the copper plated metallizing did not
polish well because it was very brittle or punky. This
condition could also give the metallizing greater electrical
resistivity and thereby increase the losses in this region.
The reason for both loss values being higher than

comparable seals in the previous series of measuremerts
remained unexplained. It may have resulted from a high VSWR
in the second series of tests as compared to the first, but
this at first did not seem to be the case because loss per
centimeter measured in empty waveguide channels was practically
identical to that in the earlier measurements for all runs.
However, the empty waveguide loss values for the first group
of measurements were obtained from outer cooling channels
whose temperature could have been affected by losses originat-
ing in the flanges at each end of the test cells. This would
424
--- , -- - --. - - .... II ,, _____I

S 4
Figure 6.76:, Tto - P-*:dy I ceramic with P-1 metallizing, thin nickel plating and Cu-Ag braze.
Boitom - Be iy I ceramic with P-1 m"tas'izing, copper plating, a1nd Cv-Ag braze showing
~part of a braze gap.
425 -
result in empty waveguide loss readings which were in error
on the high side. They would then appear equal in magni-
tude to empty waveguide losses, measured in thermally
isolated central channels, in the five-channel test cells
used in the second test series. The latter might, in fact,
be higher than in the three-channel test cells because of
a higher local V6WR in the five-channel cells.
The results from cells 2, 5, 7 and 8 (see
Tables 6-20 and 6-22) were credible and consistent with
each other and with theory. The most useful information
obtained was in regard to the relative amounts of dielectric
and conduction losses. The ratio of conduction loss in the
side walls to total loss in the broad walls (0.19:1 average
of five) was always less than the 0.26:1 side wall to bioad
wall ratio of conduction loss only, which was obtained by
calculation.
Tb' was sufficiently removed from the ratio of

the side wala o broad wall areas (0.45:1) to indicat that
the results were determined by factors other than simple
seal area ratios. The difference between the calculated
conduction loss ratio of 0.26:1 and the measured loss ratios
which averaged 0.19:1 was attributed to the dielectric loss
in the broad walls.
The amount of dielectric loss in the broad walls

was calculated from the results by assuming it to be the
quantity necessary to shift the side wall to broad wall loss
ratio from 0.26:1 to the measured ratios which averaged
0.19:1. For example, in the first run on cell #2, 2.45
watts and 13.03 watts dissipation (from Column 4, Table 6-22)
were measured in the side walls and broad walls resp-:tively
to give a ratio of 2.45/13.03 = 0.188. If the broad wall
loss was only 9.42 watts, the ratio would have been 0.26:1.
Therefore 9.42 watts were attributed to conduction loss in
the broad walls and the additional 3.61 watts were attributed
to dielectric loss. The loss ratio was broken down as follows:
Side Wall Loss = Conduction Loss = 2.45 = 0.188
Broad Wall Loss Conduction+Dielectric Loss 9.42+3.61
The magnitude of dielectric loss measured in

these test cells was about equal to the amount of conduction
loss in the side walls, an average of about 1/4 of the total
loss in the broad walls, or 1/5 of the total loss for the
entize seal. Therefore, conduction loss was the pre-
dominant contributor to the total seal losses, so efforts
426
I
I
J to reduce seal losses should concentrate on minimizing those
conditions which contribute to conduction losses.
This approach may consist of minimizing the

quantity of high resistance material such as molybdenum
and maximizing the low resistance material such as copper.
This concept was tested in cell number 6 using evaporated
molybdenum metallizing (see Table 6-20). The results from
this test cell were unsatisfactory because of separation of
the evaporated copper from the evaporated metallizing which
left numerous voids in the seals. Diversion of skin currents
by these voids caused abnormally high losses in the seals
which could not be considered a true picture of the quality
of this type of seal.
Further analysis of the dc conduction and rf

conduction data presented in this final report indicates
that the premise that a molybdenum evaporated seal will be
low loss is in error. The results obtained are thus not
unexpected on this count also.
The results from cells 3, 4, 6, 9 and 10 were

inconclusive and were apparently affected by some anomaly
or unaccounted for variables. Losses in the seals in these
test cells were unusually high and, in most, the side wall
to broad wall loss ratio was too high to enable calculation
of dielectric losses. The data has been exhaustively re-
checked. No sources of error can be detected. It is
regrettable that a third series of experiments cannot be
programmed to resolve these discrepancies.
The overall high power rf loss data presented

shows -
1) that the seal losses were of the order of 0.1%

of the total rf power transmitted per cm of
seal length (in direction of waveguide axis).
2) the results confirm those presented for the

prior electrical measurements. The lowest loss
seal investigated is thin metallizing which is
copper plated and Cu-Ag brazed (see Fig. 6.73).
3) the dielectric losses for the metallized 99.5%
alumina (Body H) is approximately one fourth of
that due to conduction.
4) Dielectric losses introduced due to metallizing

paint reactions in 94% alumina bodies can be
neglected.
1 427
I
While further analysis of the data may be made
as in other sections, in the interest of brevity, it is left
to the interested reader to consider Table 6-22 in more
depth than has been attempted in this summary.
428
6.6 REFERENCES - TASK III
1. R. A. Dehn and J. N. Close, "Surface Resistivity of

Metals and Dielectric-Metal Joints", Proceedings of
the Fourth National Conference on Tube Techniques,
New York University Press, 1959.
2. E. W. Ernst and 0. T. Purl, "Electrical Losses in
Ceramic-to-Metal Seals", University of Illinois
Technical Report.
3. Edward L. Ginzton, Microwave Measurements, McGraw-Hill
Book Company, Inc., 1957.
4. Carol G. Montgomery, Technique of Microwave Measurements,

McGraw-Hill Book Company, Inc., 1947.
5. James R. Floyd, "Effect of Secondary Crystalline Phases

on Dielectric Losses in High-Alumina Bodies", J. Am.
Ceram. Soc., vol. 47 (11) 539-43 (1964).
429
7.0 SUMMARY
7.1 Introduction
The main purpose of this study was to better

understand the electrical properties of ceramic-to-metal
seals and, if necessary, improve them. An interpretative
analysis of the knowledge concerning ceramic-to-metal seals
was presented in the first four quarterly reports. This
analysis coupled with "probing" key experiments was conducted
as a prelude to intensive experimental work.
The experimental work can be broken down into

two areas -
(1) electrical studies concerning the seal region,

(Task III) and
(2) supporting studies to outline the main parameters
involved in ceramic-to-metal sealing (Tasks I
and II).
The purpose of Tasks I and II was to act as a firm interpre-

tive support to Task III. Although at first these studies
were distinct, by the end of the program they had construc-
tively intermeshed.
In order to present a logical picture of ceramic-

to-metal seal technology, the basic principles* involved are
first reviewed, this is followed by a discussion of the
experimental aspects of the main "supporting studies" and
concluded with a review of the experimental electrical
studies.
7.2 Basic Seal Considerations
Fig. 4.36 shows a schematic micro-structure of a

94% alumina ceramic, Body A (Table 4-1). Two distinct
phases can be seen, the alumina (A12 0 3 ) crystal phase and
the glassy phase surrounding the crystals.
Fig. 4.37 shows the phases in more detail and

Fig. 4.38 shows that the "glassy" phase can undergo modifica-
tion by subsequent heat trea tment, e.g., at 1100 0 C. This
modification involves the precipitation of crystals of
anorthite (CaO ,2SI0 2 .A12 0 3 ) from the glassy phase. This
*A condensed version of the first four quarterly reports.
430
phenomena is termed "devitrification". The glassy phase
amount, composition and devitrification tendency are all
important in determining the manufacture of a reliable
ceramic-to-metal seal.
Fig. 7.01 shows a schematic of the metallizing

systems commonly used for electron tubes. The first system
(A) shows a metallizing paint (see Table 4-22) composed of
molybdenum, silica and manganese, P-8, sintered onto a
sapphire substrate. On heating this pain- to metallizing
temperature (14000 C) in a wet hydrogen-nitrogen atmosphere,
the MnO-SiO2 combines with the alumina of the substrate
and generates a molten mass which draws the particles of
molybdenum together to create a sintered metal sponge. On
cooling, the melt solidifies and forms a chemical and/or
semi-conductive bond both with the molybdenum sintered
sponge and with the alumina substrate which is vacuum
tight. The metallized -eramic is now ready for plating
with a wetting layer and a subsequent braze to a metal member
to complete the ceramic-to-metal seal.
Fig. 7.01 (B) shows the result of the same type

of process on a 99% alumina body and Fig. 7.01 (C) similarly
shows the same on a 94% alumina. In the case of the latter
system, it is possible to use just pure molybdenum powder,
such as P-4M, suspended in a lacquer as the metallizing
paint. At the metallizing temperature of 14000 C, the molyb-
denum particles will sinter together on the ceramic surface
and concurrently the glassy phase of the ceramic will become
fluid enough to flow out from the ceramic and infiltrate the
pores of the molybdenum sponge. On cooling, the glassy
phase will remain strongly adherent to the sintered sponge
and to the alumina ceramic and will form a vacuum tight bond.
As before, this system can then be brazed to the metal
member to complete a vacuum tight seal.
Fig. 4.50 shows a cross-section of a 94% alumina

ceramic seal. On close inspection of the glassy phase
interface of this seal, one can observe crystal species
which have devitrified during the initial cooling or subse-
quent brazing processes. Such physical changes are shown
to affect the performance of the ceramic-to-metal seal,
(section 4.5.3.3).
In order to investigate the physical chemistry

and crystalography of the ceramic-to-metal seals further,
electron microprobe trace analyses of the metallizing have
been conducted. This technique has been supplemented with
431
I *~ ~4$ ~ Jr Molybdenum
(a) 100% Alumina
Melt
~ ~
* .. ~Molybdenum
4P Enclosed pore
-Alumina crystal
Molybdenum
(c) Melt
Alumina
Figure 7.01: Schematic representation of a Mnetallized layer on (a)single crystal sapphire (b)a 99.5%
alumina body (c)a 94% alumina body.
- 432 -
"MR lo O.
spectrographic analysis, X-ray diffraction analysis and with
electron microscope scans of interesting areas. Typical
examples of the results of these techniques are given in
Figs. 4.52, 4.53, 4.56 and Table 4-27.
The mechanical performance of ceramic-to-metal

seals has been determined using the direct tensile test
vehicles shown in Figs. 4.01, 4.02 and the "micro-tensile"
peel test vehicles shown in Fig. 7.02. Figs. 4.83 and 4.84
show a shear test arrangement that was developed, as an
analysis of the stresses at the seal interface indicating
that a shear test was also of interest in evaluating seals.
Fig. 7.03 schematically shows that in the shear
stress situation, each layer of the seal has to support
the full loading imposed upon it while in the tensile test
situation, due to its lower modulus of elasticity, the weak
glassy layer tends to be supported by the adjacent stronger
metallic and ceramic layers.
A cross-section of a torn seal from the drum
peel shear test is shown in Fig. 7.04.
7.3 Supporting Studies

The number of component oxides involved in
ceramic systems are rather small but still sufficiently large
to be confusing. It was therefore decided in the main
experimental phase of this program to investigate a model
ceramic system: the calcium oxide, aluminum oxide, silica
system (CaO-SiO2-Al20 3 ). With this in mind, three bodies
were formulated from the following components:
Si02:CaO / 1:1 - balance 94% A1 2 0 3 - body 941
Si02:Ca O / 2:1 - balance 94% A1 2 0 3 - Body 942
Si0 2 :CaO / 2:1 - balance 99% A1 2 0 3 - Body 992
at five different maturing temperatures. In addition, G. E.

Lucalox (99.75% alumina, 0.25% MgO) and sapphire (100% A1 2 03 )
were investigated, (Table 4-1). The grain size of the
special alumina ceramics was obtained as a function of the
firing temperature of the ceramic, Fig. 4.08.
The CaO-SiO 2 -Al 2 03 phase diagram was intensively

studied and two binary phase diagrams at the joins 1:1
SiO 2/CaO-Al 2 0 3 and 2 Si0 2 /CaO-Al 20 3 were constructed,
Figs. 4.03 and 4.04. From the knowledge of the diffusion-
corrosion phenomena of alumina by calcia-alumina-silica
433
INSTRON
Test Piece
(ASTM CLM-1511
iNSTRON
Tab Pool
Test Piece
Figure 7.02: Diagrammatic representation of drum peel and tab peel test fixturing.
-434-
metal
Braze
-.-__ __ ,- Plate
Metallizing
Reaction zone
Ceramic
Reaction zone
O -- Reaction zone
n_.7
Applied shea stress
Figure 7.03: (a)Schematic diagram of seal showing laminar construction (idealized),

(b) Application of tensile stress to laminate results in effective opposing of compressive
stresses in the lower-modulus, nonmetallic phases, which partially offsets the tensile
stress. The lower the modulus for a given material strength (of the reaction zone material)
the greater the seal strength.
(c)Application of shear stress is equally felt by all layers of laminate and weakest layer
determines the seal strength.
1 -435-
0
ap
*5 N
C 41
0~U
*-0I
-Co
CA 0
C-
.-
00
4w -C
c C
--436-
melts, the rate controlling solution mechanism in micro-
thicknesses of melts in ceramics and metallizing layers
was found to be due to chemical control, Section 4.2.3.2.6.
Micro-structure (grain growth) considerations supplemented
by X-ray diffraction data on the interior and exterior
crystalline condition of the ceramic bodies after various
processing conditions, Figs. 4.10, 4.11, 4.12 indicated
that "near" equilibrium condition is obtained in the
special bodies. Confirmation of this finding was attempted
by estimating the amount of glassy phase present as presented
by electron microscope scans of the micro-structure of the
ceramics. Attempts were also made to obtain precise
chemical analysis of the glassy phase of the ceramic by
electron microprobe and conventional chemical analysis
techniques. From such studies, together with sag bar
studies, Table 4-1, and prior viscosity work on molten
glasses reported in the literature, Figs. 4.24 and 4.25,
the viscosity or fluidity of the glassy phase was found.
This information is of prime usefulness in developing a
quantitative picture of the mechanisms underlying the
metallizing process.
The integrated information obtained above also

enables one to interpret MoR strength data versus firing
temperature of the ceramic and the deformation behavior
of the ceramic under load during metallizing.
Many metallizing paints were formulated to test

our specific mechanical and electrical hypothesis, with
regard to seal behavior: the paints were listed in Table 4-22.
The particle size distribution of the molybdenum metal pait.t
single phase system, P-4M. as analyzed by the Coulter
electronic counter technique, is shown in Fig. 4.43. This
technique has for the first time allowed one to obtain
accurate particle size distribution for a metallizing
composition and is a valuable process control tool.
The paint analyzed above is shown in Fig. 4.72,
after being metallized to a 942 body at two temperature
levels. At the lower metallizing temperature (1425 0 C), as
predicted from the viscosity data garnered above, no flow
out of the glassy phase occurred, while at the higher
temperature (1600°C) complete glassy phase flow-out did
occur. As shown in Figs. 4.62. 4.65, 4.68. this corres-
ponded with a dramatic increase in seal strength and also
with metallized surface harcness. Table 4-36. Similar
results were obtained with the 941 ard 992 ceramic bodies.
437
Experimental scudies were carried out on the
properties of the glassy phase of the special ceramic bodies
in order to obtain additional insight into the metallizinJ
process. Such work included devitrification studies,
sessile drop surface tension measurements, thermal expansion
measurements, metal evaporation-deposition substrate
studies, sintering studies and chemical solution kinetic
concerning the solution of molybdenum in the glassy ph-ase.
The results of these technicies confirmed the general
picture that was developed concerning ceramic-to-metal
seals. For instance, it was found that the contact angle
between the metal member and the glass sessile drop increased
from 150 to 600 as the furnace atmosphere went from the wet
hydrogen-nitrogen mixture (+ 100OF dew point) to a dry
mixture of -100F dew point. This correlates with the well
known fact that it is easier to metallize in a wet atmos-
phere than in a dry atmosphere. It was established that
chemical and/or semi-conducting bond mechanisms were
responsible for the actual joining of the refractory
metal to the oxide substrate. For this bonding to occur,
either the metal phase had to be deposited on the oxide
substrate by vapor phase processes or the oxide phases had
to move by viscous flow processes (e.g. glassy phase
migration) in order that the initial atom-to-atom contact
could be made.
The work on alumina ceramics was extended to

beryllium oxide, Table 5-1, and fused uuartz. Consideration
was also giver to boron nitride (BN). Fig. 5.03 shows a
cross-section of a BeO ceramic-to-metal seal. As an example
of the utility of the microprobe technique, one should
examine the results of traces conducted across two BeO
ceramics that are shown in Fig. 5.04 and 5.05. The marked
difference in penetration of MnO and TiOx from the metalliz-
ing paint (Mo-MnO-TiO2 ) can be completely explained in !
terms of composition and consequent viscosity of the glassy
phase initially present in the ceramic. Fig. D 06 shows a
cross-section of quartz to metal seal which employs the J
technique of evaporated metallizing and pressure sealing,
using evaporated Ti followed by molybdenum and gold as the
metallizing members 6 which is pressure sealed to a gold
metal member at 800 C.
4
!
438
I
7.4 Electrical Studies
The electrical studies were programmed

concurrently with the mechanical studies. These studies
fall into three main areas -
(1) dc resistivity measurements of metallizing as

a function of teml2rature
(2) low power rf measurements which may be sub-

divided into conduction (12 R) studies and
dielectric studies
(3) high power rf studies.
The metallic conduction path through the

metallizing layer is rather tortuous as is shown schemati-
cally in Fig. 6.01. Consequently the resistance of the
metallizing layer is 10 to 100 times greater than that of
a pure molybdenum layer. Such data was used as an index
of the progress of the sintering of the molybdenum, layer,
Table 4-37. It was found that a marked drop in resistance
occurs as the molybdenum metal particles are sintered and
drawn together by emergent glassy phase from the ceramic.
A determination of the coefficient of electrical

resistivity enables one to obtain an insight into the rela-
tive contributions of metallic conduction and non-metallic
semi-conduction behavior to the overall conduction process
occurring in the ceramic-to-metal seal; typical data is
presented in Fig. 6.06. As the amount of non-metallic
phase in the metallizing layer is increased, so the coeffi-
cient of electrical resistivity becomes less positive.
Fig. 6.11 shows that when the metallic portion of the
metallizing is par-ially or completely removed, the tempera-
t ire coefficient of the remaining layer becomes negative,
thus displaying its semi-conducting nature.
It may be noted from Fig. 6.06 that even in the

absence of semi-conduction inducing oxides such as MnO and
Ti02 in the glassy layer, the layer still displays a small
measure of semi-conducting behavior. This can be due to
solution of molybdenum as molybdenum metal and molybdenum
ions through the bulk of the glassy phase as well as to
boundary effects at the glass-metal interface.
An analysis of the results of the dc electrical

studies leads to the development of the theory that a chemical
and/or semi-conducting bond exists at the interface between
439
the metallic and the glassy layers of the metallizing.
While in the case of a glassy insulator phase such as
CaC-SiO 2 -Al20 , chemical bonding may predominate, in the
case of the ahdition of the variable valence ions Mn and Ti
as MnO and TiO 2 to the metallizing phase, it is to be
expected that a semi-conducting bonding mechanism will
become important.
The plated "wetting" layer which is often placed

on top of the metallizing can influence its electrical
properties as is shown in Fig. 6.19. Whereas a copper
layer does not increase the electrical resistivity of the
layer as a function of processing time at high temperatures
(800 0 C - 10000 C), nickel, by its continued diffusion into
and reaction witht the metallizing layer to form Mo-Ti inter-
metallic compounds, raises the resistance and consequent
rf losses in the seal region.
Fig. 6.24 shows a 10 Gc test cavity used to

determine 12 R losses. The ceramic-to-metal seal is placed
in the region of high current density to maximize conduction
losses; using a balance bridge method, the lossiness of the
cavity is compared against a standard cavity in which the
loss can be continuously changed by means of a variable-
length, lossy insert rod joined to a micrometer head.
Typical results are shown in Fig. 6.32. In the case of the
copper wetting layer, used in conjunction with the P-l,
Mo-MnO-TiO 2 , metallizing on Body A, a 94% alumina ceramic,
one can see that as the thickness of the metallizing
increases, the losses increase, for more and more current
is carried in the metallizing. With the use of the nickel
plate wetting layer, the losses decrease as the metallizing
thickness increases. This result is to be completely
expected in view of the dc results described above and shown
in Fig. 6.19.
The low power dielectric loss measurements were

carried out in the test cavity shown in Fig. 6.46. A 60 mil
thick metallized ceramic disc was placed between the pillars
of the re-entrant cavity in the region of high electrical
field intensity in order to maximize the dielectric losses
due to the semi-conducting and insulating components of the (
metallizing layer. The reduction in the quality of the
cavity as determined from VSWR measurements was taken as
the indicator for dielectric loss. Typical quality (Q)
values on various metallizing on ceramics as a function of
various ceramics are shown in Fig. 6.62. Further information
presented in Fig. 6.63 concerns the increase of dielectric
loss as a function of sintering time and by inference,
sintering temperature.
!
440
Dielectric losses may be minimized by eliminating
oxides such as MnO and TiO 2 from the metallizing paint and
substituting low loss glasses such as barium-aluminum-
silicate in the metailizing layer on +99% alumina ceramics.
In the case c! the 94% alumina ceramics, pure molybdenum
could be used and the dielect-ic losses were again minimized
as the glassy phase occurring -. such ceramic!, is low loss,
although it was found that a 94% alumina displayed the same
order of loss regardless of the metallizing.
Having obtained an understanding of the factors

influencing losses in seals, attention was turned to mcasur-
ing losses in ceramic-to-metal seals for the high power
sit *tion. A 10 kw klystron operating at 8 Gc was used to
pass power into a number of waveguide cells containing
"windowettes" of the ceramic-to-metal seals under investiga-
tion. The diagramatic set-up of one such cell showing the
frame-like construction of the windowettes and the water
coolant channels is shown in Fig. 6.65. By using a minimum
of c,.ramic, one is able to measure the electrical losses
arising purely from the effects of introducing the ceramic-
to-metal seals into the waveguide. Fig. 6.70 shows the
general view of the equipment including the calorimetric
set-up, the waveguide assembly, the klystron power source
and the associated electrical equipment used. The tempera-
ture rise recorded was of Lhe order of 10 C and was measured
using differential thermopiles. Fig. 6.73 presents typical
results obtained with this set-up and Table 6-66 gives a
detailed tabulation of all the measurements made. The losses
occurring in ceramic-to-metal seals are of the order of 0.1%
of the power transmitted. It should be noted that the
results on 96% alumina bodies correlate remarkably well with
the low power conduction loss measurements shown in Fig. 6.32,
indicating that dielectric losses were relatively low (as
noted previously).
I
I
44
8.0 CONCLUSIONS
1) It has been shown from the average of ten measurements

that the ceramic-to-metal seal region accounts for
approximately 25% of the total window loss in the
microwave output window and is absolutely of the
order of 0.1% at 10 Gc. At higher frequencies,
this proportion can be expected to increase as the
relative amounts of metallizing and ceramic window
volume increases. In addition to usage on high
power klystrons, the low loss metallizing development
in this study can profitably be used on planar triodes
and reflex klystrons.
2) The dielectric loss for the metallized 99.5% alumina

ceramic is approximately one fourth of that due to
conduct.ion.
3) Dielectric losses introduced due to metallizing

paint reactions in 944 alumina can be neglected.
4) Low "dielectric" loss metallizing may be produced by

using pure molybdenum metallizing on glass containing
ceramics; evaporated molybdenum or BaO-SiO 2 -Mo
metallizing on gla-s-free alumina bodies.
5) Low "conduction" loss metallizing is achieved by

using extremely thin "painted on" metallizing. Thin
evaporated metallizing was not low loss. For amenable I
geometries, a copper to ceramic "crunch" seal is the
best.
6) A semi-quantitative picture of metallizing behavior

which identifies the main parameters, both physical
and electrical, which contribute toward making
acceptable ceramic-to-metal seals has been developed
and is concisely reviewed in the summiary, Section 7.0.
It was established that chemical and/or semi-conducting
bond mechanisms were responsible for the actual join-
ing of the refiactory metal to the oxide substrate.
For this bonding to occur, either the metal phase had
to be deposited on the oxide substrate by vapor phase
processes or the oxide phases had to be moved by
viscous flow processes (e.g., glassy phase migration)
in order that the initial atom-to-atom contact could
be made.
442
I
7) The work on the physical and electrical properties
fused together in such a fashion that they comple-
ment each other and show that electrical techniques
can constitute the most powerful means of investigat-
ing the sealing mechanisms associated with ceramic-
to-metal seals.
8) The study has been mainly concerned with alumina

ceramics but it was proven possible to build on the
concepts developed, to produce high quality beryllia
and quartz seals. It is confidently expected that
an expansion of the technology gained can be made to
include ceramics such as boron nitride which offers
considerable promise as a superior window material.
Boron nitride may very well be considered a 4th
generation window material in succession to glass,
alumina and beryllia sequentially.
9) In spite of the optimistic conclusion statements made
above, a wide gulf still exists between the knowledge

generated to date and the ability to predict the
occurrence of highly reliable seals in tubes from
first principles. This is because although the
principles underlying good ceramic-to-metal seals
are known, the multitude of paints and ceramics
used in the industry permit ar infinite number of
physio-chemical reactions to proceed. These reac-
tions, needless to say, have not all been investi-
gated. It is extremely difficult to predict from
first principles when a glassy phase will or will
not devitrify after reaction with a complex metalliz-
ing paint.
develop It ismetallizing
standard therefore ofprocesses
some importance to
and standard
ceramic bodies for use in the industry with well
known reproducible characteristics.
10) Allied technical areas have benefited, such as heater

technology; for instance, one is now able to gain an
insight into the mysteries causing heater failure
when the polarity of the heater-cathode leads are
reversed, to develop printed heaters, ceramic sub-
strates for cathode use and to further understand the
'measles' effect on ceramic surfaces.
11) Other technological areas have also benefited, such

as those of printed circuit technology, semi-conductor
packaging, high temperature ceramic-to-metal seals
and a general understanding of ceramic body micro-
structure and physical chemistry.
443
9.0 RECOMMENDL IONS
1) A detailed picture of ceramic-to-metal sealing

processes may be obtained by surveying the eleven
quarterly reports and this final report. However,
the interpretive analysis presented in earlier
reports should be modified by subsequent experi-
mental findings. It is therefore suggested that a
monograph on the subject be prepared to update the
analysis. The data presented could be augmented
by several chapters on the metallurgy of the metal
member of the ceramic-to--metal, a subject largely
ignored in the present program. A chapter on stress
analyses could also be included; as could a chapter
on the application of the knowledge gained to semi-
conductor packaging technique and to printed circuit
technology. A special report is in preparation on
the latter subject.
2) Following the thoughts presented in the ninth point

in the conclusions, section 10.0, it is obviously
important to further develop the test procedures
for seals initiated on this program and to refine
stress analysis techniques for predicting seal
failure; areas of effort which need serious
consideration.
3) Boron nitride sealing techniques should be developed;

see point 8.
444
10.0 ACKWOWLDGEKENTS
The efforts of the following persons, whose
contributions enabled the successful completion of the
program, are gratefully acknowledged.
J. Mann, D. Boilard, and W. Stolte for their

assistance in making VSWR measurements on the waveguide test
cells for the high power calorimetric loss measurements
and for their help in setting-up and operating the test
station and klystron where these measurements were made.
J. Thomson, G. Phillips, and R. Hofheins for

the preparation of samples for photomicrographs, electron
photomicrographs, electron probe microanalysis, and X-ray
diffraction analysis, as wei as preparation of test
specimens and assemblies for the seal parameters study and
the low and high power electrical measurements.
H. Mein and F. Burris for the preparation of

evaporated coatings of titanium, molybdenum, and copper on
ceramics and fused quartz and for work on other test
sample. for the seal parameter study.
M. Kosinski for performing the spectrographic

analyses.
D. Harding, D. Enochs, B. Vogel, B. Meek and

M. Wheatley for preparation of curves and illustrations.
E. Sena and L. Martin for preparation of

photographs.
44S
1.0... .. SUBCONTRACTS
The electron probe microanalyses were sub-
contracted from the first through the ninth quarter to the
Centei for Materials Research, Stanford University, and were
carried out by Mr. G. Martin under the administrative direc-
tion of Professor R. Huggins. From the tenth quarter to the
end of the program, electron probe microanalyses were per-
formed at Materials Analysis Company, Palo Alto, by
Mr. R. Wolf.
X-ray diffraction analyses and electron
microscopy were subcontracted to Sloan Research Industries,
Inc., Santa Barbara, and were performed by Dr. R. Sloan
and Mr. W. Gardner.
The special bodies 992, 942 and 941 were
fabricated by Western Gold and Platinum, Belmont, California.
446
12.0 IDENTIFICATION OF KEY TECHNICAL PERSONNEL
The participants in the Eitel-McCullough program
contributed the following time during the period: 1 April
through 30 June, 1965.
fours
L. Reed 184
W. Wade 486
R. Pape 82
J. Thomson 284
G. Phillips 332
H. Meyn 308
447
Robin B. Pape
Research Scientist
Device Research Laboratory
Mr. Pape received his BA degree in Physics from Hofstra

University in 1962. He was employed from 1955 to 1960
in the Electronic Tube Division of Sperry Gyroscope Company
where he was assigned to materials evaluation and tech-
niques investigation for the Tube Techniques Section of
the Tube Research Department. In 1960, he joined Thomas
Electronics as Project Engineer for Microwave Tube Com-
ponent Development and Production. Upon receiving his
degree in 1962, he joined the Magnetron Development
Section at Amperex Electronic Corporation.
In 1963, Mr. Pape joined Eitel-McCullough as a Development

Engineer with the voltage tunable magnetron group of the
Microwave Tube Division.
In 1965, Mr. Pape joined the Device Research Laboratory,

Research Division, as a Research Scientist. His present
responsibilities are in the area of special device de-
velopment.
448
13.0 APPENDIX I
FACTORS INFLUENTIAL IN BONDING CERAMICS TO METALS
(L. Reed)
Lummary of Meeting held May 18, 1965, at
Eitel-McCullough, San Carlos, California.
Those in Attendance were:
S. Leefe, USAECOM, Fort Monmouth, N. J.
0. Heil, Heil Scientific Laboratory
J. 'White, Sperry, Gainesville, Florida
S. Colgate, Florida University Consultant, Sperry
L. Reed, Eitel-McCullough
Ceramics generally are multi-component systems.
For example, in the case of alumina consisting of the crystal
phase, Al203 and a grain boundary glassy phase, formed from
silica, calcia, magnesia and other compounds. The latter
forms a lower temperature melting phase than the ceramic as
a whole.
It is possible for a metal to be placed on such
a surface and for bonding to occur to the metal member and the
ceramic via a flow of glass from the ceramic onto the metal sub-
strate. A common way of achieving intimate bonding is to sub-
stitute metal powder for solid metal sheet in order to provide
for a more intimate contact of the metal with the ceramic.
When such a powder is placed on the ceramic and is heated to
a temperature sufficiert to make the glass phase of the ceramic
mobile (generally over 12000 C) then the glass will flow out
under the action of capillary forces to coat the grains of the
sintered molybdenum sponge.
At lower temperatures, i.e. 1100 0 C, the viscosity
of the glass may be such that it is unable to flow appreciably
449
and it may only plastically deform in the vicinity of the
metal particles to partially coat them, by what may be termed
a "fly paper" effect.
Non-metallic additions are often added to
metallizing mixes to lower the viscosity and increase the
plasticity of the "melt" layer. Both in high temperature

and low temperature, metallizing systems reaction of multi
component metallizing mixes with the ceramic will occur.
In the former system subsequent diffusion of ions such as
Mn2 + will take place up to several 100 microns into the
alumina grain boundaries. In the latter system, the diffusion
will be localized to a range of several microns and will not
be detected necessarily by the electron microprobe techniques.
In both cases, the phenomena so far described
is only a part of the complete bonding mechanism. We have

thus far provided a means for bringing the metal particles
into intimate contact with the non-metallic particles of the
system. The ultimate mechanism of bonding the non-metallic

phase to the metallic phase requires further elaboration.
This bonding may be described in terms of electrical forces.
The bond can be a chemical bond (ionic bond) or may be a
semi-conducting bond, especially when non-stoichiometric
compounds such as MnO, TiO 2 , etc. are added to the metallizing
mix. Both types of bonds will involve the exchange of electrons.
450
it is possible that both chemical bonding and semi-conducting
bonding might be present at the same point in the system, or
different points in the same system, respectively, if the
system is not at equilibrium.
In summary, therefore, we should say that the
necessary conditions for bonding are that there be intimate
contact between the non-metallic and the metallic phase of the
system and that a chemical or semi-cunducting bond be present
to promote satisfactory adherence. Calculations made by Prof.
Colgate indicate that the lower limit of the bonding forces
are several thousand psi at the metal-melt interface.
However, it is felt that these bonding forces are in general
an order of magnitude or greater than the forces which actually
cause seal failure. It is felt that the mechanism leading to
the destruction of conventional type of seals is one of crack
initiation at the surface of thL solidified melt phase,
followed by crack propagation through the solidified melt
layer. It is further felt that the bond can be strengthened
if crack terminators are placed in the path of the crack to
prevent further propagation. Such a crack terminator could

be a crystal species such as galaxite, MnO.A12 03.
The second class of metallizing system is that
which is formed by evaporation, ion sl'ittering or chemical
vapor deposition of molybdenum or other metal onto the ceramic
451
6urface. Once again, it is mandatory that intimate contact
be established between the metal layer and the alumina
substrate. In this case, intimate contact is established
because the molybdenum is deposited atom by atom and will
seek the lowest energy site on the alumina substrate. In
order to achieve uniform deposition over the entire surface,
it is important to move the source or the substrate so that
shadowing does not occur over any region to be deposited.
(This of course is not necessary in the case of chemical
vapor deposition.) It is not necessary for the refractory
metal being deposited to have the same structure as the under-
lying substrate for this epitaxial deposition to occur. It

is most probable that a gradual transition of properties from
the ceramic to the metal will occur, which may or may not
incorporate some oxygen in the transition layer depending upon
the degree of vacuum employed. We feel it is premature at
this point to further discuss the bonding mechanism, but

again it could be some type of chemical or semi-conducIing
bond.
We feel that the meeting between Sperry, Heil
Laboratories and Eimac, chaired by Stanley Leefe (USAEL)
resulted in a most fruitful discussion and that the areas

covered showed that almost complete agreement existed on
the subjects discussed. Furthermore, we feel that the
452
three projects* display just the correct amount of overlap
to give USAEL and the scientific community a basic picture
of ceramic-to-metal sealing, upon which to build a firm
foundation for further understanding of this complex process.
We appreciate the opportunity that has been afforded
us by our coming together, and sincerely hope we will be able
to repeat this in the future as the occasion warrants.
Len Reed, Mgr.,Device Research Lab.

Eitel-McCullough, Inc.
Jof White, Mgr. PMDL (Sperry4
9~("/o-to .
Sam Colga (Florida Univ. Consultant
Sperry)
Oscar Hei. (Heil Laboratories)
*Currently funded at Heil Laboratories, Sperry, Gainesville

and Eimac, Sani Carlos.
453
14.0 APPENDIX II DETAILED
TABLE OF CONTENTS
SECTION PAGE NO.
1.0 PURPOSE 1
1.1 Purpose of Contract 1

1.2 Objective of Program 1
1.3 Tasks 1
2.0 ABSTRACT 3
3.0 PUBLICATIONS, LECTURES, REPORTS AND 5

CONFERENCE S
4.0 TASK I, SEAL PARAMETER STUDY 8
4.1 General Ceramic Considerations 10

4.2 Special Ceramics 15
4.2.2 Microstructure Consider- 19
ations
4.2.3 The Glassy Phase 34
4.2.3.1 Molecular Structure 34
4.2.3.2 Composition and Amount 40
4.2.3.2.1 Equilibrium Phase 43
Diagram Studies
4.2.3.2.2 Optical and Electron 44
Microscope Studies
4.2.3.2.3 Mineralogical Analysis 52
4.2.3.2.4 Electron Microprobe 60
Analysis
4.2.3.2.5 Chemical Analysis 61
4.2.3.2.6 Reaction Rate Studies 62
4.2.3.3 Properties of the 72 II
Glassy Phase
4.2.3.3.1
4.2.3.3.2
4.2.3.3.3
Viscosity
Ionic Diffusion
Thermal Expansion
72
78
80
I
4.2.3.3.4 Contact Angle Studies 85
4.3 Commercial Bodies- i01

Microstructure Considerations
4.4 Exploratory Metallizing Studies 107

4.4.2 The Paint 114
4.4.3 Firing Conditions 114
4.4.4 Sintering of the Aetalli- 114
zing
454
I
SECTION PAGE NO.
4.4.5 Solution of Molybdenum 120

4.4.6 Adherence 122
4.4.7 Chemical Reduction 123
4.4.8 Glassy Phase Inter- 129
diffusion
4.4.9 Composition of the 138
Glassy Phase
4.4.10 Seal Strength versus 145
Metallizing Temperature
4.4.11 Seal Strength as a 147
Function of Brazing
Conditions
4.4.12 Initial Metallizing Tests 15C
4.4.13 Comments on Sealing 158
Mechanisms
4.5 Studies Using the Special Bodies 159
4.5.1 Introduction of Seal 159

Parameter Study
4.5.2 Seal Parameter Studies 162
4.5.2.1 Experimental Procedure 162
and Results
4 5.2.1.1 Sample Preparation 162
4.5.2.1.2. Scratch Test 164
4.5.2.1.3 Resistance Measurement 166
4.5.2.1.4 Assembly Procedure 166
4.5.2.1.5 Leak Check 169
4.5.2.1.6 Tensile Test 169
4.5.2.1.7 Peel Test 172
4.5.2.1.8 Microstructure Examination 172
4.5.2.2 Analysis of Physical 172
II 4.5.2.3
Property Data
Microstructure Analysis 187
4.5.2.3.1 The Ceramic 188
4.5.2.3.2 The Metallizing 195
4.5.2.3.3 The Plating 197
4.5.2.3.4 The Braze and Metal 198
Member
4.5.3 Supplementary Studies 200
4.5.3.1 Bonding Mechanisms 200
4.5.3.1.1 Evaporated Molybdenum 200
4.5.3.1.2 P-2M Metallizing on 202
Sapphire
4.5.3.2 Crystal Size Factor 203
4.5.3.3 Glassy Phase De- 203
vitrification
455
I
SECTION PAGE NO.
4.5.3.3.1 First Series 203

4.5.3.3.2 Second Series 204
4.5.3.4 Effect of Metal Member 218
4.5.3.5 Complex Metallizing Paints 218
4.6 Shear Test Study 220
4.6.1 Sample Geometry 220
4.6.2 Testing Methods 220
4.6.3 Test Results and 222
Discussion
4.7 References - Task I 227
5.0 TASK II, BERYLLIA AND QUARTZ STUDY 231

5.1 Beryllia Ceramics 231
5.2 Fused Quartz-to-Metal Seals 241
5.3 References - Task II 243
6.0 TASK III - ELECTRICAL STUDIES 244
6.1 General Introduction 244
6.1.1 Sources of Electrical 244
Loss
6.1.2 Electrical Properties as 245
a Function of Micro-
structure and Chemistry
6.1.3 Program of 248
Experimental Studies
6.2 d.c. Electrical Measurements 249
6.2.1 Temperature Coefficient of 249
d.c. Resistivity of
Metallizing
6.2.1.1 Relation to Total Seal 249
Losses
6.2.1.2 Experimental Set-up 250
6.2.1.3 Results 252
6.2.1.4 Discussion of Results 256
6.2.1.5 Molybdenum and Melt Phase 273
in Metallizings as
Parallel Conductors
456
.SECTION PAGE NO.
6.2.2 Effect of Wetting Layer 277

(Overplate) en the
Resistivity of the
Metallizing
6.2.2.1 Influence of rf Conduc- 277
tion losses
6.2.2.2 Experimental 277
6.2.2.3 Results 279
6.3 rf CONDUCTION LOSS MEASUREMENTS 285

6.3.2 Selected Measurement 287
Technique
6.3.3 Sample Preparation 287
6.3.3.1 The Ceramic 288
6.3.3.2 The Metallizing 288
6.3.3.3 The Plating 288
6.3.3.4 The Brazed Assembly 290
6.3.3.' The Standard Assembly 291
6.3.4 Equipment and Experi- 291
mental Procedures
6.3.4.1 Assembly of Apparatus 291
6.3.4.2 Determination of the 2S7
Cavity Q
'.3.4.3 Operation of rf Admittance 298
Bridge
6.3.4.4 Measuremen t Procedure 299
6.3.5 Equipment C!cckout 300
6.3.5.1 Reproducibility as a 300
FuncLiOn of Time
6.3.5.2 Standard Test Piece 303
6.3.5.3 Alignment of Ceramics 303
6.3.5.4 Sensitivity of Micro- 303
meter Reading with
Ca ity Q30
6.3.6 Results 306
6.3.6.1 First Series 306
6.3.6.2 Second Series 306
6.3.6.3 Third Series 309
6.3.7 Discussion of Results 316
6.3.7.1 lttroduction 316
6.3.7.2 Skin Depth 317
6.3.7.3 GLoup 1 319
6.3.'.4 Group I 320
6.3.7.5 Group III 320
j 6.3.7.6 Group IV 321
I 457
i
SECTION PAGE NO.
6.3.7.7 Grou" V 321

6.3.7.8 Group VI 322
6.3.7.9 Group VII 328
6.3.7.9.1 Sub-Groups 7.A and 7.B 328
6.3.7.9.2 Sub-Group 7.C 332
6.3.7.9.3 Sub-Group 7.D 332
6.3.7.9.4 Sub-Groups 7.E and 7.F 339
6.3.7.9.5 Sub-Group 7.G 344
6.3.7.9.6 Sub-Group 7.G.1 348
6.4 Dielectric Losses 350

6.4.2 Sample Preparation and 353
Experimental Set-up
6.4.2.1 The Cavity 353
6.4.2.2 Sample Preparation 353
6.4.2.3 Measurement Set-up 355
6.4.3 Experimental Procedure 359
and Results
6.4.3.1 Series 1 359
6.4.3.2 Series 2 359
6.4.3.3 Series 3 366
6.4.4 Discussion 366
6.4.4.1 Ceramic Losses 366
6.4.4.2 A.,tered Ceramic Losses 372
6.4.4.3 Pure Metallic Layer 374
on Ceramic
6.4.4.4 Conventional Metallizings 374
on Body A and Body H
6.4.4.4.1 Pure Metallizing Paint P-2 374
6.4.4.4.2 Complex Metallizing Paint 378
P-1
6.4.4.4.3 Leached P-I Metallizing 380
on Body 11
6C.5 comparison of Various 380
Metallizings on +99,'
Alumina Bodies
6.5 Aligh Power Measurements 390

6.5.2 Caloritmntric Test Cell 391
6.5.3 Cold Testing 397
Instrumentation 3 98
6.5.4 Calorimetric
6.5.4.1 Principles of Calorimetric 39R
Measurements
6.5.4.2 Preparation of Calori- 398
metric Measurements
458
SECTION PAGE NO.
6.5.5 Calorimetric Measurements 404

6.5.5.1 Experimental Set-up for 4n4
High Power Microwave
Measurements
6.5.5.2 Experimental Procedure 410
6.5.5.3 Results and Discussion 417
6.5.5.3.1 Characteristics of 417
CeramiL to-Metal Seals
Tested in First Test
Series
6.5.5.3.2 Test Results, Second Series42C
6.5.5.3.3 Discussion of Results 422
0.6 References - Task III 429
7.0 SUINUMARY 430
7.1 Introduction 430

7.2 Basic Seal Considerations 430
7.3 Supporting Studies 433
7.4 Electricul Studies 439
8.0 CONCLUSIONS 442
9.0 RECOENDATIONS 444
10.0 ACNNO.WLLDiMiNcTS 445
11.0 SULCONTRAC",' 446
12.0 IDENTIFICATION O' KUY TECHNICAL PERSONNEL 447
13.0 APPENDIX I - Factors Influential in Bonding 149

Ceear-.ics to Metals
14.0 APPENDIX II - De-tailed Table of Contents 454
459
UNCLASSIFIED
Security Classification
DOCUMENT CONTROL DATA- R&D
(Security clasification of title, body W* abstatt and indxin4 annotation must be entered heun th. overall report is cb..ifed;
1. ORIGINATING ACIVTY (Corpolate author) 20. REPORT SECURITY C LASSIFCATIONi
Enacp A Division of Varin Associates

San Carlos, Cali oriiia 2b GROUP
r. 3 REPORT TITLE
I N/A
M) A=fUICAL RESEIH & DgV=p

O1 T fOR CERAMIC PLEC'I)N DEVICES
4. DESCRIPTIVE NOTES (Type of report end inclueive date*)
Final Report -1 Jul 62 - 30 Jun 65

S.AUTHOR(S) (Last onae. firot name, initial)
Reed L.; Wade W.; Vogel S.; McRae R.; eines C.
6, RE PORT OATE 78. TOTAL NO. OF PAGES J7b. NO. OF REFS
January 1966 S39f

*a. CONTRACT OR GRANT NO. 9. ORIGINATOR'S REPORT NUMSER(S,
DA 36-039 sc-90903
b. PROJECT NO.
OST-7900-21-223-15 TR-66-1
C. Ib OTHER t PORT NO(S) (Any ohr numb.rs tatmay beaasi.d
d
10. AVA IL ABILITY/LIMITATION NOTICES
Distribution of this document is unlimited.
11. SUPPLEMENTARY NOTES 12. SPONSORING MILITARY ACTIVITY
Sponsored by Advanced Research U. S. Army Electronics Commnd

Projects Agency Fort Nomouth, New Jersey A)MEL-L-TQ
I3. ABSTRACT
See attached sheet
D"D iUm 1473 U--_I_ D _

S.cur.tY Classificatdon
1
13. ABSTRACT
Fifteen special 94% and 99/ alumina ceramics in

the system Si0 2 -CaO-Al 20 3 with a Si0 2/CaO ratio of 1:1 and
2:1 were fabricated and their behavior characterized. A
combination of phase diagram interpolations, density and
strength determinations, optical and electron microscope
planemetric and crystal growth studies, X-ray diffraction
and electron microprobe studies, and chemical analysis of
their glassy phases, established that equilibrium was very
nearly reached at their maturing temperature. Cooling
tended to freeze in the high temperature state although
crystal phas, s such as anorthite, gehlenite and calcium
hexaluminate were partially precipitated in some cases.
It was tentatively established that the rate
determining mechanism of alumina solution both daring the
initial processing of the ceraimic and during its subsequent
reaction during metallizing with refractory metal paints
containing oxide additions, was the chemical reaction rate.
The composition of the bulk and surface grain

boundary phases of the ceramic was used to de.'termine the
viscosity of its melt phase. This, in-turn, was used t.
predict the occurrence of highly electrical conductive,
high strength, vacuum tight, ceramic-tome al seals when
metallized with a pure metal paint.> Devitrification,
thermal expansion, sintering, surface energy, dissolution,
diffusion, crystal size, microscoj:-, X-ray diffraction,
electron microprobe and spenctrographic studies were also
used to predict the occurrence of high and low strength
vacuum tight ceramic-to-metal seals on the twelve complex
metallizing paint systems investigated.
Lucalox* sapphire, beryllia and fused quartz

vacuum tight seals were made and their sealing mechanisms
were examined as were those of several commercial alumina
ceramics.
It was established that chemical and/or semi-

conducting bond mechanisms were responsible for the actual
sealing process of refractory metal to the oxide substrate.
*General Electric Company trade name for a 99.75% alumina

body.
For this bonding to occur either the metal phase had to
be deposited on the oxide substrate by vapor phase
processes or the oxide phases had to move by viscous flow
processes (e.g., glassy phase migration) in order that
the initial atom to atom contact could be made.
The electrical studies included dc resistance

measurements, low power rf conduction and dielectric loss
studies, and high power loss studies. It was shown that
the metallizing layer is made up of metallic, semi-conductive
and insulating layers. Conduction losses are minimized by
using thin metallizing and a copper braze. Dielectric
losses are only important for +99% aluminas and they
account for a quarter of the total metallizing losses.
The total metallizing losses are about 25% of the total
window losses which, in turn, are about 0.3% of the'
transmitted power at 10 Gc.
In addition to usage on high power klystrons,

the low loss metallizing development in this study can
profitably be used on planar triodes and reflex klystrons.
UNLASSIMID
Security Classification
1.' LINK A LINK B LINK C
Seals Ceramic-Metal
Electron Tubes L
rf Seal lobsses
Windows, Microwave
Microprobe
Metallizing Paints
INSTRUCTIONS
1. ORIGINATING ACTIVITY: Enter the name and address 10. AVAILABILITY/LIMITATION NOTICES: Enter any lim-
of the contractor, subcontractor, grantee, Department of De- itations on further dissemination of the report. other than those
fe se activity or other organization (corporate author) issuing imposed by security classification, using standard statements
the report. such as:
2a. REPORT SECUI TY CLASSIFICATION: Enter the over- (1) "Qualified requesters may obtain copies of this
all security classification of the report. Indicate whether eport from DDC.t
"Restricted Data" is included. Marking is to be in accord-
ance with appropriate security regulations. (2) "Foreign announcement and disserrination of this
report by DDC is not authorized."
2b. GROUP: Automatic downgrading is specified in DoD Di-
rective 5200. 10 and Armed Forces Industrial Manual. Enter (3) "U. S. Government agencies may obtain copies of
the group number. Also, when app!icable, show that optional this report directly from DDC. Other qualified DDC
markings have beeui used for Group 3 and Group 4 as author, users shall request through
ized. .1"
3. REPORT TITLE: Enter the complete report title in all (4) "U. S. military agencies may obtain copies of this
capital letters. Titles in all cases should be unclassified, report directly irom DDC. Other qualified users
If a meaningful title cannot be selected without classifica- shall request through
tion, show-title classification in all capitals in parenthesis
immediately following the title.
4. DESCRIPTIVE NOTES: If appropriate, enter the type of (5) "All distribution of this report is controlled. Qual-
report, e.g., interim, progress, summary, annual, or final. ified DDC users shall request through
Give the inclusive dates when a specific reporting period is ."
covered.
AUTr ( rs tIf the report has been furnished to the Office of Technical
5. AUTHOR(S): Enter the nome(s) of outhor(s) as shown on Services, Department of Commerce, for sale to the public, indi.
or in the report. Enter last name, first name, middle initial. cate this fact and enter the price, if known.
If military, show rank and branch of service. The name of
the principal author is an absolute minimum requirement. It. SUPPLEMENTARY NOTES: Use for additional explana-
day, tory notes.
6. REPORT DATE Enter the date of the report as
month, year; or month, year. If more than one date appears 12. SPONSORING MILITARY ACTIVITY: Enter the name of
on the report, use date of publication, the departmental project office or laboratory sponsoring (par
ind for) the research and development. Include address.
7a. TOTAL NUMBER OF PAGES: The total page count
should follow normal pagination procedures, i.e.. enter the 13. ABSTRACT: Enter an at)stract giving a brief and factual
number of pages containing information, summary of the document indicative of the report, even though
it may also appear elsewhere in the body of the technical re.
7b. NUMBER OF REFERENCES Enter the total number of port. If additional space is required, a continuation sheet
references cited in the report, shall be attached.
Ba. CONTRACT OR GRANT NUMBER: If appropriate, enter It is highly desirable that the abstract of classified re-
the applicable number of the contract or grant under which ports be unclassified. Each paragraph of the abstract shall
the report was written. end with an indication of the military security clascificatton
8b, 8, & 8d. PROJECT NUMBER: Enter the appropriate of the information in the paragraph, represented as TS), (S)
military department identification, such as project number, (C). or (U),
subproject number, system numbers, task number, etc. There is no linitation on the length of the abstract. How-
9a. ORIGINATOR'S REPORT NUMBER(S): Enter the offi- ever, the suggested length is from 150 to 221; words.
cial report number by which the document will be identified 14. KEY WORDS- Key words are technically meaningful terms
and controlled by the originating activity. This number must or short phrases that characterize a report and may be used as
be unique to this report. index entries for cataloging the report. Key words must be
9b. OTHER REPORT NUMBER(S): If the report has been selected so that no ecutity classification is required, Iden-
assigned any other report numbers (either by the originator fiers. such as equipment medel designation, trade name. -nill-
tary project code name, georaphic location. may be used as
or by the sponsor), also enter this number(s). key words buit will be followed b' an indication of technical
context. The assignment of links, rules, and weights is
optional.
'
UI SIA81iyCD asiiato
Scurty Cl, sifionh

Ceramic Electron Devices: Metallurgical Research and Development

Uploaded by

Copyright:

Available Formats

Ceramic Electron Devices: Metallurgical Research and Development

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ceramic Electron Devices: Metallurgical Research and Development

Uploaded by

Copyright:

Available Formats

0 REPORT 12

METALLURGICAL RESEARCH AND DEVELOPMENT .

U.S. ARMY ELECTRONICS COMMAND

D. A. TASK NUMBER OST - 7776 -11 -017 -38

CLEARI NGH OUSE

DISTRIBUTION OF THIS DOCUMENT IS UNLIMITED

) /The findings in this report are not to be construed as

The citation of trade names and nai.ies of manufac-

Destroy this report when it is no longer needed. Do

METALLURGICAL RESEARCH & DEVELOPMENT

U. S. ARMY ELECTRONICS COMMAND

FT MONMOUTH, NEW JERSEY 07703

D.A TASK NO. OST-7900-21-223-15

CONTRACT NO. DA 36-039 SC-90903

EIMAC, DIVISION OF VARIAN ASSOCIATES

METALLURGICAL RESEARCH & DEVELOPMENT

FOR CERAMIC ELECTRON DEVICES

1 July 1962 - 30 June 1965

This research is a part of project DEFENDER

CONTRACT NO. DA 36-039 SC-90903 Tp e-

EIMAC, DIVISION OF VARIAN ASSOCIATES

DISTRIBUTTON OF THIS DOCUMENT IS UNLIMITED

w• mir m m( 1n • m m ,pP -.--. Im - -m m mum m m n mm •mp - - m l mmmmm

1.1 Purpose of Contract 1

4,1 General Ceramic Considerations 10

4.4 Exploratory Metallizin9 Studies 107

4.4.1 Introduction 107

4.5 Studies Using the Special Bodies 159

4.5.1 Introduction of Seal 159

4.6 Shear Test Study 220

4.6.1 Sample Geometry 220

4.7 References - Task I 227

5,0 TASK II, BERYLLIA AND QUARTZ STUDY 231

5.1 Beryllia Ceramics 231

6.0 TASK III - ELECTRICAL STUDIES 244

6.1 General Introduction 244

6.1.1 Sources of Electrical 244

6.3 rf CONDUCTION LOSS MEASUREMENTS 285

6.3.1 Introduction 285

6.4 Dielectric Losses 350

A 4.1 Introduction 350

6.5 High Power Measurements 390

6.5.1 Introduct ion 390

6.6 References - Task i1 429

7.0 SUMMARY 430

7.1 Introduction 430

10.0 ACKNOWLEDGEMENTS 445

11.0 SUBCONTRACTS 446

12.0 IDENTIFICATION OF KEY TECHNICAL PERSONNEL 447

13.0 APPENDIX I -Factors Influential in Bonding 449

FIGURE NO. PAGE NO,

4.01 Tensile test piece B 13

4.02 Tensile test piece A (CLM-15) 14

4.03 Revised pseudo-binary phase equilibrium 17

4.05 Model of two spherical particles, radius R, 20

4.06 Surf3ce tension forces, F,due to a pressure 23

4.07 Particle size distribution of Alcoa A-14 25

4.09 Miciistructure of 94i eramic (400X) 29

4.10 Microstructure of 942 ceramic (400X) 30