The Spectra of the Doubly and Triply Ionized Rare Earths

G. H. Dieke and H. M. Crosswhite

The present status of our knowledge of the structure of the spectra of the doubly and triply ionized
spectra of the rare earths is derived partly from experimental data of the emission spectra of the free
ions which provide the energy level scheme in great detail but are difficult and laborious to analyze.
For the lower levels knowledge of the structure comes from the crystal absorption and fluoresence
spectra. In all cases approximate theoretical calculations of the energies are essential.

1. Introduction carried out or planned for the immediate future at

The rare earths, in common with the actinides, have Johns Hopkins University.
the most complicated spectra of any of the elements.
This is because the incomplete 4f-shell produces a very 11. Experimental Techniques
large number of low-lying levels. The transitions be-
The free-ion spectra are obtained from a controlled
tween these give a many-line spectrum without any
spark in a rare-gas atmosphere at reduced pressure.'
apparent regularities. Even multiplets, which are
Figure 1 shows the development of the spectrum of
such prominent features of the spectra of the transition
gadolinium as the circuit parameters are changed.
elements, are in general unrecognizable. For these
The 3-A spectrum is that of a dc arc, and even at this
reasons the rare-earth spectra have long been neg-
low-excitation lines of the doubly ionized ion appear.
lected. Because of the growing importance of the
The other spectra are of unidirectional pulsed dis-
rare earths in recent years, their spectra have received
charges, the duration varying from about 2 msec for
renewed attention. In the present article we shall
100 A to 100 osec for 1500 A. Above 100 A there is
restrict ourselves to the spectra of the doubly and triply
no trace of the second (singly ionized) spectrum, and
ionized elements for several reasons.
Most atomic spectra are obtained by vaporizing the as the excitation is increased the third also gradually
element, exciting the vapor by an electric discharge, gives way to the fourth. In no case is the spectrum
completely pure for a given ionization stage, but by
and thus obtaining the emission spectrum of the free
comparing one which contains only third and fourth
ions or neutral atoms. The triply ionized rare earths,
and to a lesser extent also the doubly ionized ones, can
("hot spark") with one containing only second and
be incorporated into crystal lattices where they retain
third ("mild spark") the fourth spectrum lines may be
identified. The lower excitation of an arc or micro-
their sharp energy levels only slightly modified by the
wave discharge is also needed to eliminate the first
internal crystal field. This makes it possible to obtain
and second spectra, but for our purposes it is not
some of the energy levels, particularly the important
necessary to distinguish between the two. It is im-
low ones, from the absorption and fluorescence spectra
of crystals containing these rare earths.
portant for this separation that the pulsed discharge
not be allowed to oscillate as this will superimpose a
An extensive program of the study of the spectra of
lower excitation discharge on the high and make the
doubly and triply ionized rare earths is under way at
suppression of the second spectrum difficult.
Johns Hopkins University and both the emissionspectra
Figure 2 shows a comparison of hot and mild sparks
of the free ions and the crystal spectra are investigated.
for various wavelength regions of the neodymium spec-
One of the chief purposes of this study is the com-
parison of the levels of the free rare-earth ion with those
trum. In the upper section Nd iv lines predominate for
of the same ion in a crystal lattice, in order to be able
the hot spark (A) although some very weak Nd III
may be found. A slightly longer wavelength region
to obtain an exact evaluation of the influence of the
(center) is very rich in Nd iii. Even longer wave-
crystal field. The present paper is a report of the work
lengths (bottom) show mainly ii and in the mild
The authors are at the Johns Hopkins University, Baltimore spark (B) and very little in the hot. This separation
18, Maryland. by wavelength is also very significant for the analysis
Received 13 January 1963.
The work described in this paper has been carried out with and is discussed further in Sec. IV.
partial support of the U.S. Atomic Energy Commission and of Accurate wavelength measurements are of the great-
the Office of Aerospace Research of the U.S. Air-Force. est importance as the possibility for finding spurious

July 1963/ Vol. 2, No. 7 / APPLIED OPTICS 675

 through a properly oriented crystal immersed in liquid
helium. The Zeeman effect is often important, so
the dewar vessel must be constructed so that it can fit
between the pole pieces of a magnet. At low enough
temperature the wavenumbers of the absorption lines
directly represent the position of the excited energy
_- l levels. The electric crystal field produces a Stark
effect which splits all levels into a maximum of 2J + 1
components for an even number of electrons and J +
_/ components for an odd number. The number and
spacing of the components depends on the symmetry
2290 2300 *2310 A and intensity of the crystal field. In most cases the
total splitting of a free-ion level is of the order of a few
Fig. 1. Stages of ionizati, n. Different excitation conditions hundred cm-', in general small compared to the spacing
in the gadolinium spectrurn. Lines of Gd iI, iII, and iv are
indicated by corresponding sets of dots. The 3-A spectrum betweenadjacentmultipletcomponents.
is that of a de are, the others are pulsed, critically damped A sharp-line fluorescence spectrum is obtained, con-
d iseharges. taining in some cases several hundred sharp lines,4
by illuminating the crystal by an intense light source.
Nd:
J~f0: When the coupling between the electronic level and
the crystal lattice is large the excitation energy is
A
dissipated before fluorescence can take place. This is
26 0 90 2700 o 2'0 30 the case for most hydrated salts5 except for the central
rare earths from Sm to Dy. In other lattices all rare
d TIE earths fluoresce strongly but not from all levels. The
A ! |liill
-- ii
j*ll1 ItIIF
| , IlJtI
IlgllI[
,~i
ll
l 1,
anhydrous chlorides and bromides, the garnets and the
oxides6 are among the crystals that show general fluores-
2950 60 70 80 9o 3000 cence. Figure 3 illustrates this for a crystal of Y203
containing 1% erbium. From theoretical considera-
AM tions and previous experience we expect to find a group
B 1 (called the A group) of excited levels at about 10,000
3750 60 70 80 so
Fig. 2. A-900-A peak current discharge (hot spark); B-50-A
peak (mild spark). All spectrum regions were taken simul- ABS.
taneously. The short lines on the edges are from the Fe calibra-
tion tube. A(A)
Z PA A)
9600

regularities is proportional to (AV)2,where A

uncertainty in the frequency measurements.
is the
The
A J
use of large gratings is therefore most appropriate.
In our work a 17.8 cm wide, 6.64 m f.l. concave grating
with 1200 lines/mm is used for most of the work, in the
first order for the longer wavelengths (above 5000 A),
9700
in the second or third orders for the shorter ones, which
makes it possible to obtain wavenumbers accurate
to about 0.01 cm-' for most of the spectrum above
2000 A. The use of an interferometer would not con-
tribute much because, particularly in the hot spark, the
lines are not sharp enough for utilizing the greater in-
herent accuracy of the interferometer. The fact that
thousands of lines must be measured in each spectrum 9800 4740
makes the use of modern automatic measuring tech-
niques imperative. 2
For the crystal absorption spectral light from a con-
Fig. 3. Absorption fluorescence spectra of 1 % Er' + in
tinuous source, usually a high-pressure mercury or Y203. Left: Absorption from Z group to A Group 42 -
xenon lamp, a tungsten ribbon lamp or a zirconium arc 4IsI/2). Right: Fluorescence from P group to A group (P, 2 -
depending on the wavelength region, are passed 411/) See Fig. 7.

676 APPLIED OPTICS / Vol. 2, No. 7 / July 1963

 I 40rXIO 3 cm-l

Nd El PrII
2
P

4f 2 6s B C

80
2D 4f 26p

4f 5d2

2G

40 I
2
F
B-C

Odd Even Even Odd Odd Even Even Odd

Fig. 4. Block diagram of three-electron configurations Nd iv and Pr iII. Arrows indicate principal transitions. Parity is
indicated by "odd" or "even."

cm-l above the ground state which would correspond of the ground state, for which the absorption spectrum
to the single free ion 4 I11/2
level (see Sec. III, especially is not in an accessible region, must be confirmed by
Fig. 7). As the number of electrons for Er'+ is odd, additional fluorescing levels.
and as the local crystal field symmetry about this ion In complicated fluorescence spectra, single levels
is low, we expect to find the Stark-split level to consist can be excited by monochromatic illumination, and this
of J + /2 = 6 components. The ground state 4/2 greatly helps in the analysis of such spectra. The
will be split into 8, but if the crystal is cold enough, fluorescence spectra are extremely sensitive for very
only the lowest of these will be occupied. Absorption small amounts of impurities, and great care must be
transitions occur between this lowest level and any taken to identify such impurity lines which occur in
higher level unless forbidden by selection rules. All six even the purest available materials. 7
appear in the left-hand spectrum of Fig. 3. Fluores- For the crystal spectra the task of obtaining the
cence from the lowest Stark component of 2P3/2 (P energy level system from the observed wavelengths
group), shown on the right, confirms all six levels. of the absorption and fluorescence lines is in general
Very low levels, such as the higher Stark components direct and simple. For the free-ion emission spectra

July 1963 / Vol. 2, No. 7 / APPLIED OPTICS 677

50 r X 03 cm-I
Ce I 2
La I
5d

40
Ld 4 f3
F5d6s
Even

2
5 d 6s
4f6s6p
I

2
f 6p
2
4f5d6p | 5d 7s I
30_ 2
5d 4f

2
Sd6sp 1s5d 6p

20- 2
f 5d
5d6s6p 6s 6

4f 6s

I0 4f5d6s 4f5d
2

5d2 6s

I SdF-,rI
0 l

Even
l

Odd Odd Even Odd Even Even Even Odd Odd Odd

Fig. 5. The same for Ce ii and La i.

this is a much more elaborate and difficult problem. their principal quantum numbers. For the fourth
It can be solved much more expeditiously when one spectra the 4f orbits always have the lowest energy,
knows approximately what to expect. In fact, a satis- then come 5d, 6s, and 6p. This means that without
factory analysis of these complicated spectra would be exception the 4fn configuration is lowest, then come
virtually impossible without a thorough theoretical 4fn15d, 4f1-'6s, and 4f'-16p, with several other con-
study of the structure of the energy levels. Such a figurations coming in with energies near that of 4 fn-
study is based on the general theory of atomic struc- 6p. Figures 4 and 5 show for ions with three ex-
ture, as for instance set forth in the book by Condon ternal electrons how this situation changes as we go to.
and Shortley 8 or those by Slater,' systematized for the the lower stages of ionization.
complicated rare ion cases by Racah.' 0 The specific For the third spectra the order of the configurations
situation for the rare earths has been clarified by a is still essentially the same as for the fourth, but the
comparison with the empirical results so far obtained beginning of the rare-earth series 4 fn and 4fn-15d are
from the crystal and free-ion spectra. nearly coincident and 4fn-15d and 4 f'-16s cross over
near Dy (see Fig. 6). For the second and first spectra,
111. Level Structure of the Rare-Earth Ions as Fig. 5 shows, the situation is very confused, and
It has been known since the early days of the Bohr this makes the analysis of these spectra much more
theory that the 4f, 5d, and 6s electrons have very nearly complicated than that of the higher ionization stages.
the same energy for the rare earths. This is true It is, of course, not sufficient to know merely the
especially for the neutral and singly ionized atoms. position of the centers of the configurations as indi-
This means that configurations like 4fl, 4f1-15d, cated in Fig. 6. We must know their total width and
4f-16s, Af'- 2 6s2 , etc., may overlap and create an the arrangement of the individual energy levels. The
extremely complicated set of low-lying energy levels method for doing this can be briefly indicated as follows.
which it is difficult to disentangle. Fortunately, the The outer electrons are regarded first to be in a central
situation is much clearer for the divalent ions and even field provided by the nucleus screened by the 54 elec-
more so for the trivalent ones. The reason for this is trons of the completed xenon-like shell. For the rare
that because of the higher nuclear charge and the con- earths the assumption that this central field is a
solidation of the inner shells the screening is more per- Coulomb field determined by one parameter, the ef-
fect and the levels appear more nearly in the order of fective nuclear charge Z*, leads to practically as good

678 APPLIED OPTICS / Vol. 2, No. 7 / July 1963

results as the more general assumption of a nonrestricted yond possibility of doubt. For the higher energies
central field and simplifies the calculations materially. the situation is often so complicated that positive
After first considering the outer electrons as com- identifications cannot be made without a great deal of
pletely independent, their electrostatic repulsion and further study. Figure 8 gives a summary of the results
their spin-orbit interactions are introduced by the obtained so far. Table I lists the parameters F2 and
potentials v derived from these data.
H = (/rji) and H = IMsili)
,expressing the electrostatic repulsion between the outer
Table I. Numerical Values of the Parameters F2 and ¢
electrons and the spin-orbit interaction, and the in-
2
fluence of these perturbing potentials on the energy Empirical Theoretical'
and wave functions calculated. In performing these F2 i-/F, F2 v
calculations interactions between states of only one Ce - 640 - 453 830
configuration are taken into consideration, but no Pr 305.4 729.5 2.39 477 980
further restrictions are made. For the higher stages 317.4 756.5 - - -
Nd 337 876 2.60 498 1130
of ionization, other configurations of the same parity Pm 346.0 1070 3.09 - -
are far away for the lowest configurations and we make Sm 370 1200 3.2 532 1480
no fatal mistake by leaving out interconfiguration in- Eu 401 1320 - - -
teractions, although they are by no means negligible. Gd 405.4 1581.1 3.9 565 -
The systematics of handling the very large number Tb 434 1705 - - -
Dy 420 1900 4.5 589 2310
of individual states have been developed by Racah. i Ho 450.1 2163 4.8 - -
It is evident that only states with the same L and S Er 429 2445 5.7 616 2830
can interact through HI, only those with the same J Tm 451.1 2656 5.89 - -
through HI. This and other symmetry considerations Yb - 2882.9 - 642 3400
greatly limit the number of matrix elements of HI and
HI that have to be calculated and restrict the order
of the interaction matrix that has to be diagonalized
to the number of different states with the same J
occurring in the configuration. The situation is
simplest for the 4 fn configurations. The calculations,
which are somewhat tedious, but not unduly so for a
modern computer, are carried out numerically as a
function of the spin-orbit parameter and contain a
scaling parameter F2 which is determined by the
effective charge Z*. The two numerical parameters
F2 and are obtained by comparing the calculated set
of levels with some well-identified empirical ones.
La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
This has been carried out for all the crystal spectra 100) X
I I I I I I a-->---1 1

6&1.
. '

004~~!
cl__---
of the trivalent rare earths, and Fig. 7 gives an example
80.
of the agreement that is reached. The 29 lowest of R+.+
the 41 possible levels of the 4fP configuration of Er'+ 60 6p
i
are shown.2 It is clear that for the first few multiplets 40
there cannot be any possible doubt about the identi- 20 1-
fication. For the higher ones crystal splittings also 6s
0
have been used for identifying the levels. It is seen
that in the crowded regions the mere position of a level -20 i Sd
i
is not enough for identification. Here the levels are -40 i
very sensitive for the particular choice of the para- -60
I i
I i
meters, and the fact that the calculations are only ap- I
proximate ones puts a limit to the accuracy one can -80.
I
expect. -100
I
The situation is similar for the other trivalent rare -120 -
i
'earths."-23 The references give in each case first
Fig. 6. Upper: Relative positions of the centers of the low
the papers which give the most recent empirical data configurations in third spectra (doubly ionized). The n symbols
and then the ones containing the calculations. In all refer to the optically active electron; e.g., 6s means 4fn-1 6s.
cases the first few multiplets have been identified be- Lower: Same for fourth spectra (triply ionized).

July 1963 / Vol. 2, No. 7 / APPLIEQ QPTICS 679

I
 t 's
40 cJm- U- 9z2
* I 3 cm-1 U-

3&8 ,PI
6I

T 4
-__ ,Ia
36& C7, X_ - 2D' ,

0 'P.
N-

_ _ _
34 W- - R- 2
6P Lz
0- TU
= - o- M'o ..--- 'K3,
A 2
A 72
SD -D3
32[ 0- S P 2
_- - - - 2
R-
Q
M_
30[ I-
P-
0 -
'F
-3
4D I
F- N- 0 2G7,I ID
L- I/I
28 L-^2 0'- K3 4 E- 2M
N 'H6
N 2K - 0
L_
_ - - O H-
D 5L m "I
_ G-
K L _ I2
p2 N- F
26 a D3 J _
L 7
5G,
M- E _
'p I0
2 H 9,2
L
K - D D3 K _-W
-0-
24 I I-
4G 112 _
I 2p
P
'p H-

22 I
- - - DD
F C--2
15, I-6 H -5 1

---
2 G4G ____
_ _ _5G4
_ E-
F 9 HG G-7/ 4
D_
C 4G
20
F- 050
E-W /I B,
a - I 4 4F
E"1' -
E.---
S2
18
2G 7
_ - -2~ A-0
SK
7
5F
16 3K F,
-

B 4S -- - -4
14 6F 22
…2 W --
2 H 92,, 5I
- - - - 2 A- % '4
SMW 5/ "2
12
R 32 A-1~ 2
'12,, C -H -
1 0-
G
10 4 97t

A 6 "H,
8 0 7'2
-T/
W_94
'F 5I1 R 512 ,5H

- 4 4I 5 3t2/2 Y - %2
_ _.
6 U "F
7F -4

----- 7 V 32 7F
-
- 2
X_ t "I- Y
-- -st -' - 3
4 _1\

---- - -W4 Yi1 Z

-4

2 _ . _-

0 -. _ _.. __ Z _ Z Z- Z _ _
2 3 6 7 8 7 6 3 2
F512
5/2
H
4
4I
9/2
5I4
I
H5 / 2
/
F0
°6
S F. H52
5
5I
8'2
415/2 H6 F772,

Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb
Fig. 8. Observed energy levels of the trivalent rare-earth ions.

680jAPPLIED OPTICS / Vol. 2, No. 7 / July 1963

 The energy levels obtained in this way are, of course, any high degree of accuracy. Comparison of the same
those of the ion in a particular crystal lattice. Using levels in different lattices shows6 that there may be
the center of the Stark components eliminates the in- shifts of several hundred cm-' and we can trust the
fluence of the crystal field in first order. This, how- levels obtained from the crystal spectra to agree with
ever, is not sufficient to obtain the free-ion levels with the free-ion levels at most to this extent.
The configurations where not all outer electrons are
in 4f orbits, such as 4f"'-5d or 4f'-16p, can be dealt
with in similar fashion although the calculations are
x I 000cm-'
somewhat more involved as more parameters enter into
the calculations. Figure 9 shows the situation for the
40 - _u
-
=
_
,,
2I~~~~~-
4f235d configuration of Ybs+, and again it is seen that
the comparison of the empirical levels with the calcu-
lated set gives a positive identification.' 4
Knowledge of the position and extent of the two
lowest configurations 4ff and 4f 0-15d is particularly
important as these two configurations may be expected
to furnish the strongest lines.
15 __________ 20,,
Figures 10 and 11 give a survey of these configurations
-NU _
\ou
2G9
2P~~~~~~~,,2
for the third and fourth spectra, respectively, obtained
4D 2 from preliminary empirical data and very approximate
calculations. Because of the particular couplings in-
volved we must in general expect the strongest lines to
30 _
be transitions from the bottom of one configuration to
2K,
RI___ == ',2 the bottom of the other one, from the middle to the
middle, etc. We see that for the fourth spectra all
-_
…2.,,-_______-_-_2
-_ 2H
5d -* 4f transitions must lie in the vacuum ultraviolet
and shift from 2000 A for Ceiv to 1000 A for Ybiv.
2 The 6p - d transitions remain at about 1300 A for
'2 all fourth spectra. The 6p -- 6s transitions, how-
ever, lie in the more easily accessible part of the ultra-
_ 20~~~~~~~~~~~~~~~~~~2
-F- ~- - - 2H
violet and shift from 2500-2800 A for Ceiv to 1700-
2200 A for Yb iv.
The situation for the third spectra is considerably
F different. For Laiii the 4f and d levels must prac-
'~~~~~-H,,
E ~~~~~~~~~~~~~S'
tically coincide and the transitions between them lie

-D- - - - - - - - - -F
112
YbH1
4f 5d
-Experiment
x 1000 cm- ----Calculation
-- 6s Levels
54-
- - I , . .-- /~~~~~
52 /
\\ F1, 2
\ I\

I0 -A - I, 50
------- I
48
2F
%]
46 ,

-Y _______ 4Ii,
44
'2

42
/- [F21j \

40 ------- -
_ '
,------- I
is I \ 7',21] 3,2,

34 ------- /

0 MZ _4 1-4I,

'12
1 2 3 4 5
0 BS. CALC. J
Fig. 7. Observed and calculated levels of Er'+. Fig. 9. 4f 35d configuration of Yb +.

July 1963 / Vol. 2, No. 7 / APPLIED OPTICS 681

I
5 0 FE in 103 Cm-l
-
Ix\700
Is

800

'D| 900
IS
I 00- I 000
G1
D

1(

:1

I 1500
S
2
F 0

50 -r1
0 H 2000

I
I
Is
0
c3

3000
0 6
I I
- 4000
SD
5000
IF
6
F
IH5F F 2
10000
F

0
0aCe Pr
S US
S

La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb
Fig. 10. Schematic diagram of the 4fn (white) and 4f-'5d (black) configurations of the divalent rare-eoxrth ions.

in the far infrared. For this reason they have not yet of parts of opposite parity in the wave functions.2 6
been found. They move apart with increasing Z, and These transitions have f numbers of the order of 10-
the strong ad - 4f lines lie in the visible for Pr iii and and would be completely blotted out if they occurred
near 2000 A for Ybiii. Figure 10 shows that because in the region of the allowed 5d -> 4f transitions. We
of the varying width of the configurations the situation have seen that for the trivalent ions the latter lie
for the intermediate elements may be irregular. For usually in the vacuum ultraviolet which leaves the
Gdiii for instance the strong lines must again be ex- visible and near ultraviolet free for the crystal spectra.
pected in the infrared, and a search for them (Calla- Figure 10 shows that for the divalent ions the d -
han2 5) has revealed that they certainly cannot lie in the
4f transitions usually reach into the visible which
visible or ultraviolet. For the third spectra also the
leaves much less room for the crystal spectra, and we
transitions between the other low configurations lie
must expect these chiefly in the infrared. Sm and Eu
mostly in the visible or accessible ultraviolet where
precise wavelength measurements can be made more have long been known to form divalent ions. The
easily. fluorescence spectrum of Smi+ is known in the ex-
We may now come back to the crystal spectra of the treme red and infrared 7 and resembles that of Eu3+ with
divalent ions. The sharp-lined crystal spectra in the all the energies roughly reduced by 20%. Something
visible and adjacent regions of the trivalent and di- is also known of the spectrum of Eu 2 +. Recently it
valent ions are due to normally forbidden transitions has been found by Kiss27 and others that in crystals
between levels of the 4f 6 configuration. These are that contain trivalent rare-earth ions these can be
made possible, except for rare cases of magnetic dipole converted into divalent ions by irradiation with
radiation, by the admixture through the crystal field gamma rays. This makes it possible to obtain. the

682 APPLIED OPTICS / Vol, 2, No. 7 / July 1963

lowest levels of all the divalent rare-earth ions from crystal lines somewhat, Fig. 11 shows that we cannot
crystal spectra data. expect to go much further without coming in conflict
with the strong continuous absorption to the 4f'-'5d
IV. The Interpretation of the Free-Ion Spectra levels which may reach farther to the visible than the
The energy levels derived from the crystal spectra are figure indicates because of the crystal field broadening
obtained without ambiguity and their interpretation of the 5d levels. Also other constituents of the crystal
is now possible without difficulties. The data so ob- lattice may absorb in the far ultraviolet. In order to
tained are of necessity very limited, as they cover only obtain a more complete picture of the energy levels, it
the lower portion of the 4fn configurations. While fur- is necessary to go to the free-ion emission spectra.
ther studies in the ultraviolet may extend the known These as we have seen earlier are very complex

X
600
Is

E(in 103cm-')
150 _
700

I.6
800

0D 900

100 1 000
2
F

6G

'S

26
1 500

-o
0 0
EI U 2G
2000
501-
.0

pq
_'P 3000
4000
0 5 000
C -0
2H 'F

-0- F

3F 2F

US
S E 1HI

Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er TM Yb Lu
Fig. 11. The same for trivalent ions.

July 1963 / Vol. 2, No. 7 / APPLIED OPTICS 683

 Table II. The Four Lowest Configurations of the Trivalent Rare-Earth Ions
Number of levels
Ground Allowed
R2+ R3+ as state 4.fn 4fn-'5d 4f-16s 4fn- 1 6p Sum transitions
La 0 is, 1 - - - 1 -
La Ce 1 'F15 2, 2 2 1 2 7 8
Ce Pr 2 3H4 13 20 4 12 49 324
Pr Nd 3 4I9/2 41 107 24 69 241 5393
Nd Pm 4 5I4 107 386 82 242 817 -
Pm Sm 5 GH 5 /2 198 977 208 611 1994 306,604
Sm Eu 6 7Fo 295 1878 396 1168 3737 -
Eu Gd 7 'S7/2 327 2725 576 1095 4723 -
Gd Tb 8 7F6 295 3006 654 1928 5883 -
Tb Dy 9 GH,15 2 198 2725 576 1095 4594 -
D)y Ho 10 4I8 107 1878 396 1168 3549 -
Ho Er 11 41,6/2 41 977 208 611 1837 -
Er Tm 12 3H6 13 386 82 242 723 -
2
Tm Yb 13 F7 /2 2 107 24 69 202 3773
Yb Lu 14 iS, 1 20 4 12 37 217

We have set ourselves the task of obtaining the four to the spectrum. Under these circumstances one must
lowest configurations 4f", 4f"-1Sd, 4f'-16s, and 4f 5 -'6p proceed with the analysis with a careful plan in order
of the third and fourth spectra as completely as pos- not to be involved in a task that is too time-consuming
sible. Experience with the spectra which have so far for the results to be worth the effort.
been analyzed24' 25 28 has shown that the electron in a In the first place it is essential to have as complete
6s or 6p orbit (and even d for the heavier elements) a wavelength table as possible with wavenumbers of
is only loosely coupled to the core, which makes the the highest possible accuracy. The second step is to
so-called Jj coupling scheme proposed by Racah2 9 establish an empirical set of energy levels. Most of
most appropriate. Strong transitions occur only when the labor of this step can be delegated to a computer.
the core does not change and the orbital quantum num- If nothing is known about the spectrum, the computer
ber of the excited electron changes by 4-1. This has can be instructed to compute all wavenumber differ-
a marked influence on the appearance of the spectrum. ences between the lines and list those that occur re-
As the Al transitions are not much affected by the par- peatedly within certain limits of accuracy. Many of
ticular core value, there will be one such transition at these will be accidental coincidences, and in this step
approximately the same wavelength for each choice the accuracy of the wavenumbers is the decisive factor.
of core. For instance for the one-electron Laiii spec- If the tolerances are narrow, the chances for accidental
trum the 6p3/2 -- s and 6p,/2 - 6s transitions occur at coincidences become less. By examining all the signi-
3172 and 3517 A, respectively.' 0 In Ceiii the cor- ficant differences, an empirical energy level diagram
responding 4f6p -* 4f6s transitions occur as groups can be built up, the reliability and completeness of
centered near 3100 and 3400 A; for Priii at 3000 and which depends on the extent and accuracy of the wave-
3350 A. For Nd ii we can, therefore, expect that the length table.
strong group of lines between 2900 and 3050 A will As stated earlier, this process can be carried out in
belong to 4f'6p3/2 4f36s and those between 3200 and principle without any previous knowledge whatsoever
3350 A to 4f36pl/2 - 4f36s. In a similar way the strong
group between 2050 and 2400 A may be associated
with 6p -* 5dtransitions.
The number of levels in these configurations may be
quite considerable, as Table II shows. The last col-
umn (for some elements) gives the number of lines that
one would obtain if the parity and J-selection rules are 0M

obeyed, as well as Al = i 1 rule for the electron that

changes orbits. No other selection rules are used for
this calculation, but many of the allowed lines may be
so weak as to be unobservable. Nevertheless the
table shows that the number of lines from these four E F' H
configurations alone may be very large, and there are Fig. 12. Hyperfine structure of holmium lines. A-D, Ho ii
of course many other configurations that contribute lines; E-G, Ho iii lines; H, Ho iv line in crystal spectra.

684 APPLIED OPTICS / Vol. 2, No. 7 / July 1963

 of the structure of the spectrum. The procedure will, Another important aid for the analysis is the presence
however, be greatly simplified if one knows what to of hyperfine structure. This is simplest when there is
expect. Since the energy differences between the only a single odd mass number isotope, as is the case
lowest 4fn levels of the fourth spectra and some of the for Pr, Tb, Ho, and Tm, particularly with Pr and Ho
third are known approximately from the crystal because of the large value of the nuclear magnetic
spectra, the computer needs to look in its first attempt moment. The magnitude of the hyperfine structure
only for differences within a few hundred cm-' of these indicates whether or not an s-electron is present and
values. The 4 f -16 levels occur in pairs, the splitting can give other information about the nature of the con-
of which can also be computed with considerable pre- figuration. Figure 12 shows a few examples of resolved
cision. The calculated energy levels will then limit hyperfine structure in the spectrum of holmium. The
further the range of differences that have to be ex- number of the components gives the value of J if it is
amined with a great saving in effort and computer below the nuclear spin. The intensities and relative
time. This search is further aided if one can limit it spacing of the components indicate the J values in gen-
to a restricted wavelength region as discussed above. eral, and other important information on the nature
Careful studies of excitation variation such as shown of the levels can be obtained from the hyperfine
in Fig. 1 are also helpful in reducing the number of structure.
lines to be handled. Not only is it possible to specify Obtaining similar information for other elements
the ionization stage, but also the approximate excita- would require separated isotopes. For these, isotope
tion level for a given ion. This allows the upper elec- shifts also could be used for the identification of con-
tron configuration from which the particular transition figurations. So far it has not been necessary to use
occurs to be identified. If enough care is taken, in- separated isotopes, but later they may be needed to
formation may be obtained even about the position of give enough information to decide doubtful cases.
the level within the configuration. If necessary, The most valuable and important aid for the analysis
electronic time-decay techniques may be used.3 ' After of these complicated spectra has, however, been the
successive efforts of this kind a fairly extensive set of comparison of the empirical levels with the calculated
empirical levels can be constructed. ones and this, in general, leads to a positive and un-
The next step is the interpretation of the levels, ambiguous identification of the levels. While matrix
which means assigning quantum numbers to them and elements for the fn configurations are now available, in
correlating them with the levels to be expected theo- general the fn l, ones are not. The computation of
retically. The most important of the quantum num- these is a straightforward, but very tedious process.
bers is the angular momentum J. This often can be Here again, we may make use of high-speed com-
established from the selection rule J = 0, -41 which puters, and these matrices are being compiled as needed.
must be satisfied by all lines of the free-ion spectrum.
Under favorable circumstances this establishes relative
J values of all the levels, which can be turned into V. Results So Far Obtained
absolute values if the J for one level is known. This
is always the case for at least the lowest level which is While the initial incentive for the investigation was
known by Hund's rule. (Highest L of highest multi- the desire to know the 4J" configurations of the free-ion
plicity, lowest J of this multiplet before Gd iv, highest levels of the trivalent rare earths in order to have a
J after Gd iv.) basis from which to evaluate the influence of the crystal
For the crystal spectra, the Zeeman effect is useful
field, most results have been obtained so far with the
spectra of the divalent ions.
in providing additional important information, but
The trivalent 4fP levels can only be obtained through
with these higher ionization stages of the free ion the
the allowed 5d -- 4f transitions which lie in the vacuum
situation is not so favorable. The spectra are very
ultraviolet, and so do the 6p - 5d transitions which can
crowded and the lines somewhat broadened by the be used to confirm the d configuration. At the
violent discharge conditions needed to produce them. present time a new large vacuum spectrograph for our
Furthermore, the strong J coupling makes all 6p + laboratory is under construction which is to be ready
6s transitions for instance look very similar. Obtain- for use during the summer of 1963. In view of this it
ing reliable Zeeman effects under such conditions would is undesirable to use the much less accurate vacuum
add greatly to the burden of wavelength measure- ultraviolet wavelengths now available to us. For
ments and analysis. We have found that the other this reason so far only parts of the Pr iv and Yb iv
methods for the analysis described here give such spectra have been analyzed.
definite and positive results at least for the low con- The situation is much more favorable for the third
figurations that there has been no need for Zeeman spectra, as the most important lines lie in the more easily
effect data. accessible region of the spectrum. Moreover, the

July 1963 / Vol. 2, No. 7 / APPLIED OPTICS 685

crystal spectra of the divalent ions also have become 5. G. H. Dieke and L. A. Hall, J. Chem. Phys. 27, 465 (1957).
of increasing interest recently. 6. G. H. Dieke, Proc. Flirst International Conferenceon Para-
magnetic Resonance, Jerusalem, Israel (1962).
At the present time the goal of obtaining the four 7. G. H. Dieke and R. Sarup, J. Chem. Phys. 36, 371 (1962).
lowest configurations has been substantially reached 8. E. U. Condon and G. H. Shortley, The Theory of Atomic
by Sugar28 for Priii and by Bryant2 4 for Ybiii. Less Spectra (Cambridge University Press, Cambridge, 1951).
complete results have been obtained so far for Gd iI} by 9. J. C. Slater, Quantum Theory of Atomic Structure (McGraw-
Callahan.2 5 For this ion the configurations are par- Hill, NewkYork, 1960), two vols.
10. G. Racah, Phys. Rev. 61, 186 (1942); 62, 438 (1942); 36,
ticularly unwieldy, and it cannot be expected that the
367 (1942); 76, 1352 (1949).
four ground configurations which contain together 11. Cet+: R. Lang, Can. J. Res. A13, 1 (1935); A14, 127
nearly 6000 levels can be obtained with anything even (1936).
remotely approaching completeness. While the parts 12. Er3+: F. L. Varsanyi and G. H. Dieke, J. Chem. Phys.
of the 475d, 46s, and 476p configurations based on 36, 2951 (1962); G. H. Dieke and S. Singh, ibid. 35, 555
the ground term S of 4f' are known, nothing is known, (1961); E. H. Carlson and H. M. Crosswhite, ref. 13;
B. G. Wybourne, ref. 13.
yet about the 4 configuration. As Fig. 10 indicates 13. Nd3+: E. H. Carlson and G. H. Dieke, J. Chem. Phys.
for Gdiii, the 4 and 475d configurations must. be 34, 1602 (1961); E. H. Carlson and H. M. Crosswhite,
expected very close together, so that the principal lines Johns Hopkins University Spectroscopy Report 19 (1960);
connecting them must be expected in the infrared. B. G. Wybourne, J. Chem. Phys. 32, 639 (1960); 34, 279
(1961).
This is apparently true, for none of these lines has
14. Pm3+: J. G. Conway and J. B. Gruber, J. Chem. Phys.
been found in the visible or the easily accessible parts 32, 1586 (1960); M. H. Crozier and W. A. Runciman, ibid.
of the infrared. 35,1392(1961).
Work on the third spectra of the followingelements 15. Sm3+: M. S. Magno and G. H. Dieke, J. Chem. Phys.
is in various stages of progress in our laboratory. 37, 2354 (1962); J. D. Axe and G. H. Dieke, ibid. 37,
2364 (1962); B. G. Wybourne, ibid. 36, 2295 (1962).
NdiII (K. V. Narasimham), Smiii (F. L. Varsanyi),
16. Eu3+: L. G. DeShazer and G. H. Dieke, J. Chem. Phys.
Hoiii (J. H. McElaney), where the large hyperfine (to be published); G. S. Ofelt, ibid.
structure is very valuable, Er ii (J. Becher), and Tm. 17. Gd3 +: A. H. Piksis and G. H. Dieke, J. Chem. Phys. (to
iii (G. Gompertz). Work on the remaining elements be published); W. A. Runciman, J. Chem. Phys. 30, 1632
will be taken up as soon as an opportunity arises. (1959).
Because of the very large number of lines involved 18. Tb3+: K. S. Thomas, S. Singh, and G. H. Dieke, J. Chem.
Phys. (to be published); G. S. Ofelt, ref. 16.
progress is necessarily slow even when all modern aids 19. Dy3+: H. M. Crosswhite and G. H. Dieke, ref. 4; J. D.
are available. Axe and G. H. Dieke, ref. 15; B. G. Wybourne, ref. 15.
It should be emphasized that the preceding article 20. Ho3 +: G. H. Dieke and B. C. Pandey (to be published);
is not a general account of the spectra of the doubly M. H. Crozier and W. A. Runciman, ref. 14.
and triply ionized rare earths which would have been 21. Pr3 +: G. H. Dieke and R. Sarup, J. Chem. Phys. 29,
741 (1958); J. S. Margolis, ibid. 35, 1367 (1961).
impossible in the allotted space, but essentially a. 22. Tm3 +: J. B. Gruber and J. G. Conway, J. Chem. Phys.
outline of the work being carried out in this field at 32,1178(1960); 32,1531(1960).
Johns Hopkins University. 23. Yb3+: G. H. Dieke and H. M. Crosswhite, J. Opt. Soc.
Am. 46, 885 (1956).
24. B. W. Bryant (to be published).
References 25. W. R. Callahan, J. Opt. Soc. Am. (to be published).
1. G. H. Dieke, H. M. Crosswhite, and B. Dunn, J. Opt. Soc. 26. G. S. Ofelt, J. Chem. Phys. 37, 511 (1962); B. R. Judd,
Am. 51, 820 ( 1961). Phys. Rev. 127,750 (1962).
2. G. H. Dieke, D. Dimock, and H. M. Crosswhite, J. Opt. 27. Z. J. Kiss, Phys. Rev. 127, 718 (1962).
Soc. Am. 46, 456 (1956). 28. J. Sugar (to be published).
3. G. H. Dieke and L. Heroux, Phys. Rev. 103, 1227 (1956), 29. G. Racah, J. Opt. Soc. Am. 50, 408 (1960).
and subsequent papers on crystal spectra. 30. H. N. Russell and W. F. Meggers, J. Res. Natl. Bur. Std.
4. See for instance E. H. Carlson and G. H. Dieke, J. Chem. 9, 625(1932).
Phys. 29, 229 (1958); 34, 1602 (1961); H. M. Crosswhite 31. W. A. Hovis, Jr., J. Opt. Soc. Am. 52, 649 (1962).
and G. H. Dieke, ibid. 35, 1535 (1961); G. H. Dieke and 32. A. J. Freeman and R. E. Watson, Phys. Rev. 127, 2058
S: Singh, ibid. 35, 555 (1961). (1962).

Mary Warga, feature editor for this issue, Erik Ingelstam

Stockholm, P. R. Wakeling OSA and Hans W. Horeman The
Netherlands during the International Commission for Optics
meeting in Munich last August.

686 APPLIED OPTICS / Vol. 2, No. 7 / July 1963