Weld Integrity and Performance, 06593G Copyright 1997 ASM International

All rights reserved.

www.asminternational.org

Chapter 1

Weld Solidification*

JOINING METALLURGY has assumed an In this article, a general introduction is pro-

even greater role in the fabrication of metals vided on key welding variables including solidi-
within the last few decades, largelybecause of the fication, microstructure, and causes and reme-
development of new alloys with tremendously dies of common welding flaws.
increased strength and toughness. Therefore, a
working knowledge of metallurgy is essential to Solidification Behavior
understanding current engineering structures
and, in particular, the mechanisms that control The integrity and performance of a weld is, to a
weldment performance. Fundamental to joining large extent, controlled by the solidification be-
metallurgy are the microstructures of a weld havior of the weld metal or fusion zone. It con-
joint, which determine the mechanical proper- trols weld-metal microstructure, grain structure,
ties, and welding variables such as weld thermal inclusion distribution, porosity, hot-cracking be-
cycle, chemical reactions in the molten pool, al- havior, and, ultimately, weld-metal properties.
loying, flux composition, and contaminants, Current knowledge of weld pool solidification
which significantly affect the weld and heat-af- is an active area of research, as fundamental ad-
fected zone (HAZ) microstructures. vances in solidification mechanics for castings
Metallurgically, a fusion weld consists of three and ingots have been applied to weldments (Ref
major zones, namely the fusion zone, the un- 1-4). Fundamental solidification mechanics de-
melted heat affected zone (HAZ) adjacent to the veloped primarily for cast metals have been suc-
fusion zone, and the unaffected base metal as cessfully applied to the solidification of welds.
shown in Fig. 1. In alloys, there also is a fourth However, major differences do exist between
region surrounding the weld pool consisting of a casting and welding including:
partially melted or liquated zone, where the peak
temperatures experienced by the weldment fall Dynamic nature of welding process
between the liquidus and the solidus. Unknown pool shape
It is well established that solidification behav- Epitaxial growth
ior in the fusion zone controls the size and shape Variations in temperature gradient and
of grains, the extent of segregation, and the distri- growth rates within the pool
bution of inclusions and defects such as porosity
and hot cracks. Since the properties and integrity In solidification mechanics, the important pa-
of the weld metal depend on the solidification rameters that influence microstructure are tem-
behavior and the resulting microstructural char- perature gradient G, growth rate (R), undercool-
acteristics, understanding weld pool solidifica- ing (11T), and alloy composition (Co). In the weld
tion behavior is essential. pool, G, R, and l1T vary considerably from region

"This Chapter has been adapted from "Principles of Joining Metallurgy" by Jack Devletian and William Wood in Met-
alsHandbook, Volume 6, 1983,P 22-49 and "Principles of Weld Solidification and Microstructures" by S. David and J.
Vitek in International Trends in Welding Science andTechnology, ASM, 1993,P 147-157
4 / Introduction

HEAT
SOURCE

FUSION
ZONE

Fig. 1 Schematic diagram showing the three zones within a weldment

FUSION ZONE

(LARGER Rs
SMALL GLl

BOND PLANAR
BASE METAL

Fig. 2 Schematic drawing of structural variation of weld microstructure across fusion zone

to region and process to process. Therefore, the available for refining the weld-metal grain struc-
microstructure that develops in the weld metal ture. Since grain growth is epitaxial in nature, the
varies noticeably from region to region. Figure 2 size of the grains in the base metal controls the
shows schematically variations in weld-metal final grain size to a limited extent.
microstructure influenced by temperature gradi- The weld pool shape also influences the fusion
ent and growth rate. zone grain structure. For example, in an ellipti-
cally or circularly shaped weld puddle, not only
Weld Grain Structure does the magnitude of the maximum thermal
The final weld-metal grain structure has con- gradient change continuously from the fusion
siderable practical significance. It has been line to the weld centerline but also the direction
shown that a coarse columnar structure is sus- of this gradient changes. Since the average
ceptible to hot cracking, whereas a finer grain growth direction during solidification of a weld
structure, a distorted columnar structure, or an pool is approximately normal to the solid/liquid
equiaxed structure is more resistant to hot crack- interface, which is also along the maximum tem-
ing. In addition, the fusion zone grain structure perature gradient, a given grain will not be favor-
has a strong influence on the mechanical proper- ably oriented during the entire solidification
ties of the weld. To improve both the mechanical process since the direction of this gradient
behavior and the hot cracking resistance of weld changes. Therefore, many of the grains at the
metal, efforts have concentrated on refining the fusion line that are initially of unfavorable orien-
fusion zone grain structure. Several means are tations may become more favorably orientated
 Weld Solidification / 5

,
,-' '-- ---'''"''-..-1-_-'

""
1()2 L..-_---'--_ _L..-_---'--_ _L..-_---'--_ _..L-_---'--_---'
10-6 10-4 10-2
Growth velocity (R l, rnIs

Fig. 3 G-R diagram for different solidification processes with columnar growth structures. The G and R values are indicative only and
represent orders of magnitude. For welding and laser treatment only one process is represented with conditions prevailing during
the solidification cycle. Source: Ref 4

before they are completely eliminated and thus one of the easy growth directions coincides with
they may survive and continue to grow towards the heat flow direction in welds.
the centerline, resulting in a finer fusion zone Figure 4 shows both epitaxial growth and a
grain structure. Such observations are typical in columnar grain structure in an iridium alloy
high energy beam welds or arc welds made at weld. The preferred grain selection process that
low welding speeds. For a tear drop shaped pud- occurs during welding promotes a columnar
dle, there is almost an invariant direction of maxi- grain structure. The weld-metal grain structure
mum thermal gradient at all points on the pool can be refined by making changes in the process
edge from the fusion boundary to the weld cen- variables and the use of external devices such as
terline. This will promote large columnar grain magnetic arc oscillator and vibrators. Although
growth as a result of the competitive grain selec- not done in practice, inoculants maybe used as in
tion process since only a few grains survive and castings to refine the weld-metal grain structure.
grow uninterrupted toward the center. The result Unlike in castings, the natural occurrence of a
is a coarse columnar fusion zone structure. columnar-to-equiaxed transition in the grain
Figure 3 shows general growth conditions in structure of the weld is not very common. How-
various solidification processes which all lead to ever, depending on the alloy system, such a tran-
columnar growth where the liquid temperature sition has been observed and is described by
gradient G > O. In welding, temperature gradi- n.w, Kerr (Ref 5).
ents are typically of the order of 105 Kim and An understanding of the development of grain
growth velocities in the range of 10-3 to 10-1 ml s. structure in the fusion zone of polycrystalline
With the introduction of laser welding tech- welds is limited. This is because details of the
niques there is a trend to increase the welding growth selection process and the three-dimen-
speed which also increases the temperature gra- sional pool shape are obscured by the multitude
dient. As will be shown later, this trend may lead of grains and crystal orientations that are present
to problems in steel welding if the same alloys are in polycrystalline materials.
used (Ref 4). Epitaxial Growth. The outstanding difference
Weld-metal grain structure is predominantly between the solidification of a casting and that of
controlled by the base-metal grain structure and a weld (aside from the relative size and cooling
the welding conditions. Initial growth occurs epi- rates) is the phenomenon of epitaxial growth in
taxially at the partially melted grains. Both welds. In castings, formation of solid crystals
growth crystallography and thermal conditions from the melt requires heterogeneous nucleation
can strongly influence the development of grain of solid particles, principally on the mold walls,
structure in the weld metal by favoring a strong followed by grain growth.
grain-growth selection process. Growth crystal- For weld solidification, there is no nucleation
lography will influence grain growth by favoring barrier, and solidification occurs spontaneously
growth along the easy growth direction. For cu- by epitaxial growth. The nucleation event in
bic metals, the easy growth directions are <100>. welds is eliminated during the initial stages of
The conditions are even more favorable when solidification because of the mechanism of epi-
6 / Introduction

Fig. 4 Epitaxial and columnar growth near fusion lin e in iridium alloy electron beam w eld. Source: Ref 1

taxial growth wherein atoms from the molten or

weld pool are rapidly deposited on preexisting
lattice sites in the adjacent solid base metal. 6.'[,
':'S a r.esult, the structure and crystallographic GJR < D: for cellular or dendritic solidification
orientation of the HAZ grains at the weld inter-
face continue into the weld fusion zone. In fact, (Eq2)
the exact location of the weld interface is very
where GL is the thermal gradient in the liquid at the
difficult to determine in any weld deposited on
solidification front, 6.To is the equilibritun solidifi-
pure metals using matching filler metal. Even
cation range at the alloy composition (Co), DL is the
microstructural features, such as annealing twins
solute diffusion coefficient in liquid, and R is the
located in the HAZ weld joints, will continue to
solidification front growth rate.
grow epitaxially into the weld during solidifica-
Equations 1 and 2 define the conditions for the
tion .
development of various solidification substruc-
Epitaxial growth is always the case during
hires in welds. This concept of constitutional su-
autogenous welding. For nonautogenous proc-
percooling is sufficient to understand the devel-
esses in which a filler metal is used, epitaxial
opment of microstructures and the influence of
growth may still occur. Nomnatching filler met-
process parameters on these microstructures in
a.ls will also solidify epitaxially, particularly if the
welds.
filler metal and base metal have the same crystal
structure upon solidification, e.g., welding Weld Pool Shepe
Monel (fcc) with nickel (fcc) filler metal.
However, the more classical case of heteroge- The macroscopic shape of the weld is an impor-
neous nucleation also applies in this case. In tant factor that influences not only weld integrity
welds, growth of the solid progresses from the but also grain size and the microstructure of the
unmelted grains in the base metal, and the stabil- fusion zone. Therefore, it is critical to understand
ity of the solid/liquid interface is critical in deter- the dynamics of the weld pool development and
mining the microstructural characteristics of the its geometry.
weld metal. The thermal conditions in the imme- The weld-pool geometry is determined mainly
diate vicinity of the interface determine whether by the thermal conditions in and near the weld
the growth occurs by planar, cellular, or dendritic pool and the nature of the fluid flow.
g~owth. Composition gradients and thermal gra- In general, the weld pool volume is controlled
dients ahead of the interface are of primary im- to a large extent by the welding parameters. The
portance. The effect of solute and thermal gradi- volume is directly proportional to the arc current
ents on the solidification front can be described (heat input) and inversely proportional to the
by .the concept of constitutional supercooling cri- welding speed. In addition, the speed of the mov-
tenon (Ref 6), which can be mathematically ing heat source has an influence on the overall
st ated as : bead shape. For arc welding processes, the pud-
dle shape changes from elliptical to tear drop
shaped as the welding speed increases. Corre-
6.'[,
Gill? > D O [orplane front solidification (Eq I) spondingly, the isotherms also change in shape.
L For high energy density processes such as elec-
 Weld Solidification /7

Location
of arc

- Welding direction

(a)

~y-1
Location
of arc

Welding direction

(b) A

------ y -----~_1

Location
of arc

Welding direction

(c) A

Fig.5 Comparisonof weld pool shapes. Travelspeeds: (a)slow, (b)intermediate,(e)fast

tron beam or laser welding, the thermal gradients better oriented for continued growth. The <001>
are steeper and as a result the puddles are circu- and <0010> are the generally favored directions
lar at lower speeds, becoming more elongated for crystal growth in cubic (fccand bcc) and hex-
and elliptical in shape as the welding speed in- agonal (hcp) metals, respectively. In fcc metals,
creases before eventually becoming tear drop for example, the <001> most favored direction
shaped at high speeds. leads each solidifying grain because the four
For example, if a single-phase metal is gas close-packed {Ill} planes symmetrically located
tungsten arc welded at a low velocity, the weld around the <001> axes require the greatest time
pool is elliptical (nearly circular), as shown in Fig. to solidify and, therefore, serve both to drag and
5(a).The columnar grains grow in the direction of guide the growth of solidifying grains.
the thermal gradient produced by the moving The shape of the weld pool tends to become
heat source (arc). The grains grow epitaxially more elongated with increasing welding speed.
from the base metal toward the arc. Because the In Fig. 5(b), the direction of maximum tempera-
direction of maximum temperature gradient is ture gradient is perpendicular to the weld inter-
constantly changing from approximately 90 to 0
face at positions A and B, but because the weld
the weld interface at position A to nearly parallel pool is trailing a greater distance behind the arc,
to the weld axis at position B, the grains must the temperature gradient at position B is no
grow from position A and continuously turn to- longer strongly directed toward the electrode.
ward the position of the moving arc. The process Therefore, the columnar grains do not turn as
of "competitive growth" provides a means much as in the case of a nearly circular weld pool.
whereby grains less favorably oriented for Finally, the weld takes on a teardrop shape at
growth are pinched off or crowded out by grains the fast welding speeds that are usually encoun-
8 / Introduction

Volume
element

Solute-enriched
zone
~
Q>
a.
a. Solid
o
U 0: L

(b) (c) Distance

1300
CoK
L
C*L
1200
~
u,
?st-
~ <. ~Vc.: Actual

-
o temperature

,:l
i
e
1100
s:\
C*
L+o:
--/
~/
Q>
I
E1000 I e ~~ \
Liquidus
~ ,:l
o:~ ~
Q>
temperature
900 a. Constitutional
I E Solid
Q> -- supercooled

800 !
AI Co 5 10 15 20 Cu
t- 0:
---- zone

(a) Copper, wt% (d) Distance

Fig. 6 Solidification of dendrites in a weld. (a) Solidification of 3% Cu-AI alloy by the growth of dendrites, (b). (c) Solute redistribution
occurring ahead of the solid/liquid interface. (d) Constitutional supercooling develops when the actual temperature of liquid in
the copper-rich zone is greater than the liquidus temperature.

tered in commercial welding practice. The weld vection in the weld pool is driven mainly by
pool is elongated so far behind the welding arc buoyancy, electromagnetic, and surface tension
that the directions of the maximum temperature forces. Significant work has concentrated on the
gradient at position A and B in Fig. 5(c) have convective heat transfer and, in particular, the
changed only slightly. As a result, the grains effect of the spatial variation of surface tension
grow from the base metal and converge abruptly (surface tension gradient) on fluid flow and weld
at the centerline of the weld with little change in penetration (Ref 11-14).
direction. Welds that solidify in a teardrop shape A fundamental limitation to a realistic predic-
have the poorest resistance to centerline hot tion of the weld-pool shape is the lack of neces-
cracking because low-melting impurities and sary thermophysical data. Unless a comprehen-
other low-melting constituents tend to segregate sive data base on thermophysical properties and
at the centerline. Unfortunately, this solidifica- mechanical behavior of solids at very high tem-
tion geometry occurs most frequently in com- peratures (including the mushy zone) during so-
mercial welding applications, because high heat lidification is established, use of computational
input and fast travel speeds produce the most models for a realistic prediction of microstruc-
cost-effective method of welding. ture, defects, and stresses will be stifled. In addi-
Quantitative Modeling. Computational mod- tion, verification of the models using critical ex-
eling is a powerful tool for understanding the periments is essential.
development of weld-pool geometry. In recent
years, significant advances have been made to Solute Segregation
model and understand the development of pool Solute redistribution during weld-pool solidifi-
shape in welds (Ref 7-10). Some of these models cation is an important phenomenon that can sig-
can address coupled conduction and convection nificantly affect weldability and, in particular,
heat-transfer problems to predict weld-pool ge- hot-cracking behavior, weld microstructure, and
ometry and weld penetration. Of the various properties. As alloys solidify, extensive solute
modes of heat transfer, convection plays a critical redistribution occurs, resulting in segregation of
role in determining the weld penetration. Con- the alloying elements that constitute the alloy.
 Weld Solidification /9

Segregation on a small scale (few microns) is This equation was modified by Brody and
referred to as microsegregation, and large-scale Flemings (Ref 16)by allowing for diffusion in the
segregation (hundreds of microns) is known as solid during solidification. Kurz and Clyne (Ref
macrosegregation. 17) have further refined Brody and Fleming's
Macrosegregation can occur in welds, but it is model to account for extensive solid diffusion.
not very common. In welds, it is primarily due to These models can be easily adapted to analyze
a sudden change in the welding parameters and microsegregation in welds.
the resultant change in growth velocity. A com- Example: Solute Redistribution in the Den-
mon manifestation of macrosegregation in welds drites of a 3%Cu-97%AI Composition. Consider
is banding. Another form of macrosegregation in an alloy of composition Coequal to 3%Cu-97%AI
welds is the formation of a solute-enriched region in Fig. 6(a) so that per Eq 3:
along the centerline of the weld, which could
promote the formation of centerline hot-crack- K= G/ct = 1.7%/10% = 0.17
ing. However, evidence or proof of this phe-
nomenon is lacking or not entirely clear. The first metal to solidify will contain only:
Microsegregation is characterized by a compo-
sitional difference between the cores and periph- CoK = (3) 0.17 = 0.51% Cu
eries of individual cells and cellular dendrites. while the last liquid to solidify between cells or
Cells are microscopic pencil-shaped protrusions cellular dendrites is rich in copper:
of solid metal that freeze ahead of the solid-liquid
interface in the weld. Cellular dendrites are more ColK = 3/0.17 = 17.6% Cu
developed than cells and appear to have a "tree-
like" shape; the main stalk is called the "primary These values represent the short transients at
dendrite arm," and the orthogonal branches are the start and finish of solidification of a cell or
called the "secondary dendrite arms." The cores cellular dendrite. As the cell or dendrite grows in
of the cells and dendrite arms have a higher soli- the weld, a dynamic equilibrium is achieved be-
dus temperature and contain less solute than the tween the newly forming solid of composition, Co
intercellular and interdendritic regions. In actual = 3% Cu, and the copper-rich liquid containing a
welding practice, cellular or dendritic mi- maximum of Col K 17.6% Cu, at the solid/liquid
crosegregation is virtually impossible to avoid interface as shown in Fig. 6(b) and (c).Ifthe actual
unless the metal being welded is a pure element. temperature distribution ahead of the solid/liq-
Generally, the important parameters control- uid interface is less than the liquidus tempera-
ling the cellular or cellular dendritic substruc- ture, constitutional supercooling occurs (Fig. 6d).
tures in welds are: Supercooling means that the solute-enriched liq-
uid ahead of the solid-liquid interface has been
The equilibrium partition ratio, K, which is an cooled below its equilibrium freezing tempera-
index of the segregation potential of an alloy: ture, and constitutional indicates that the super-
cooling originated from an enrichment in compo-
K= C;/Ci. (Eq3)
sition rather than temperature.
Microsegregationresults when the copper-rich
where C~ is the solute content of the solid at the
liquid at the solid/liquid interface solidifies be-
solid-liquid interface and ct is the solute content of
tween the cellular dendrites. The interdendritic
the liquid at the solid-liquid interface;
regions are so segregated with copper (solute)
The alloy composition itself, Co that a small amount of eutectic (0'.8)is frequently
The temperature gradient, G, in the liquid at observed. Eutectic structures can only occur
the weld interface, and when the composition of solidifying metal ex-
The growth rate, R, or velocity of the interface ceeds the maximum solid solubility.
Effect of Temperature Gradient. Whether or
Considering the solute redistribution in the in- not a planar, cellular, or dendritic substructure
terdendritic regions, also known as the mushy occurs upon solidification is largely determined
zone, it may be adequate to apply the solidifica- by G and R (Fig. 7) which control the amount of
tion models for microsegregation originally for- constitutional supercooling. If a weld is depos-
mulated for castings. The solute profile in the ited at a constant travel speed, R becomes fixed.
solid is given by the Scheil equation (Ref 15): By inducing an extremely steep temperature gra-
dient Gl (Fig. 8a), no constitutional supercooling
(Eq4) occurs and the solidified weld-metal grain struc-
ture is planar.
where C~ is the solid composition, Cois the initial When the gradient is decreased slightly to G2
alloy composition, Fs is the volume fraction solid, (Fig. 8b), any protuberance of solid metal on the
and k is the equilibrium partition coefficient. interface will grow faster than the remaining flat
10 {Introduction

G. R = Constant

Cooling rate, F =G R
Equiaxed
F2
dendritic

t
e
.,
E Planar
15
l!!
Cl
~

~.,
a.
E
i'2

..

_
I Solidification rate, R _
,,'-j;
I
I
I ..

" .. Planar
I
-., ,
,--.-,'--
I
..

10-1 10
Temperature gradient in melt (G), Klmm
Planar Cellular Dendritic

Fig. 7 Variation of weld microstructure as function oftemperature gradient, growth rate, and combinations ofthese variables (GR, G/I?J.
Source: Ref 1 and 3

interface because the solid is growing into super- tered in welds. In practice, cellular and cellular
cooled liquid; that is, the solid protuberance ex- dendritic substructures are most frequently ob-
ists at a temperature below that of the liquidus served in welds. The difference between the cel-
for that alloy. As a result, a cellular substructure lular dendritic and columnar dendritic structures
develops in each epitaxially grown grain. The is related to the length of the constitutionally
liquid ahead and alongside each cell contains supercooled zone ahead of the solid-liquid inter-
greater solute content than the cell core. face. This zone is typically much smaller for cel-
If the value of the temperature gradient is de- lular dendritic than for columnar dendritic solidi-
creased further to G3 (Fig. Bc), constitutional su- fication. Therefore, each grain will contain many
percooling becomes so extensive that secondary cellular dendrites, whereas only one columnar
arms form and cellular dendritic growth is ob- dendrite occupies one grain. Unfortunately, it is
served. The greatest degree of microsegregation very difficult to control G and R independently in
occurs during columnar dendritic solidification, welding practice. As a general rule, a fast weld-
while no measurable segregation is encountered ing speed (R) will produce a steep G.The relative
in planar growth. Whether planar, cellular, or values of G and R, however, determine the solidi-
cellular dendritic, growth is always anisotropic. fication morphology for a given alloy of fixed Co
Investigators have found that these solidifica- andK.
tion substructures can be characterized by the Solidification Rate. While G/R controls the
combined parameter G/R. Figure 9 shows that a mode of solidification, the weld cooling rate, in
large value of G/R combined with a very dilute terms of the parameter GR (solidification rate in
alloy will result in a planar solidification struc- units of OF{s), determines both the size and spac-
ture, while a low G/R and high solute concentra- ing of cells and dendrites. Flemings and others
tion will produce a heavily segregated columnar have demonstrated that the effect of solidifica-
dendritic structure. Both columnar dendritic and tion rate on the dendrite arm spacing (d) is:
equiaxed dendritic structures, although com-
mon in large castings, are not frequently encoun- d = a(GRrn (Eq5)
 Weld Solidification / 11

Weld

HAgJ"rt~"

,lQ. {b] ~ I"~ ~

"- Liquidus
temperature
e
a
III

8.
E
Solid L
Q)
I-

(d) Distance

Fig. 8 Effectof thermal gradient on mode of solidification in welds for constant growth rate. (a) Steep G J planar growth. (b) Intermediate
G1 cellular growth. (c) Small G3 cellular dendritic growth. (d) Solidification of the weld

where a is a constant and n is approximately lf2 for

primary arms and between 1J:l and lf2 for secon-
dary dendrite arms. The dendrite arm spacing of Equiaxed
stainless steel in an electroslag weld is often several dendritic
hundred times greater than that found in a rapidly
cooled laser weld.
Solute Banding. The phenomenon of solute
banding occurs to some degree in all alloy welds.
The formation of ripples on the weld surface and
solute banding within the weld are both caused
by the discontinuous nature of weld-metal solidi-
fication and occurs in manual as well as in auto-
matic welds where the travel speed is mechani-
cally constant. During weld-metal solidification, Planar
however, R fluctuates cyclically above and below
a mean value of growth rate that is determined by Solidification parameter, Gh/fi
the weld travel speed. Fluctuations in R result in
not only ripple formation, but also solute band- Fig. 9 Dependence of mode of solidification on GIR parameter
ing. Because an abrupt increase in R causes a for different solute concentrations (Co)
reduction in the amount of solute that can be held
in the solute-enriched liquid (Fig. 15c), excess
solute is dumped and appears as a solute-rich Microstructure of the Weld and
band. Similarly, a sudden decrease in R produces Heat-Affected Zone
a solute-poor band. Solute banding lines are very
helpful in welding research because they always Through the process of epitaxial growth, the
outline the weld-pool shape at a given instant initial columnar grain width of the fusion zone is
during solidification. For example, the form fac- determined by the size of the base-metal grains
tor (ratio of width to depth of weld pool), which adjacent to the weld interface. Because the peak
is so important in electroslag welding (ESW),can HAZ temperature increases with decreasing dis-
be easily measured metallographically using sol- tance from the weld interface and grain growth is
ute band lines. a function of temperature, the maximum grain
12 / Introduction

size in the HAZ always occurs along the weld Because the grain growth region of the HAZ has
interface. It is this maximum grain size that is experienced peak temperatures approaching the
transmitted into the weld fusion zone. solidus of the base metal, coarse grains develop in
Grain Size. The relationship used to calculate accordance with Eq 6. The grain-refined region of
the grain size in the HAZ is: ~e HAZ has been thermally cycled only briefly
mto the low-temperature portion of the austenite
D - Do = be-Q/2RTp (t')" (Eq6) region, resulting in significant grain refinement.
This grain-refining reaction occurs by the nuclea-
where D is final grain diameter; Dois the original tion of new grains each time the Al and A3lines are
grain diameter; e is the natural base for logarithms; crossed, either upon heating or weld cooling. The
Tp is the peak temperature, which would be the ~eneral structure of a steelweld will always appear
solidus temperature at the fusion line; t' is the time fine-grained when compared to similar welds de-
at temperature; Q is the activation energy for grain posited on single-phase metals, such as pure
growth of the alloy; R is the universal gas constant; nickel, copper, a brass, and ferritic stainless steel.
and band n are constants determined by the mate- The grain size distribution in precipitation-
rials. Both temperature and time at temperature hardening alloys-which include maraging
produce grain growth in the HAZ, and as stated steels; precipitation-hardening stainless steels;
previously, the maximum grain size always occurs 2xxx, 6xxx, and 7xxx series aluminum alloys; co-
immediately adjacent to the weld interface. Be- balt- or nickel-based superalloys; copper, tita-
cause all welds experience the same spectrum of nium, and magnesium alloys; and many others-
peak temperature from To to the solidus tempera- is generally similar to that of the single-phase
ture, the only significant variable in Eq 6 is the alloys. The majority of precipitation-hardening
residence time, t'. As the cooling rate decreases, alloys develop coarse grain structures in both the
residence time increases, substantially coarsening
weld and HAZ, and the small amount of second-
the maximum HAZ grain size. phase transformation is insufficient to produce
The process of competitive grain growth may any grain refinement. For example, welding and
cause further lateral growth of the weld-metal
slow cooling a typical nickel-based superalloy
grain size. Columnar grains emanating from the
containing small additions of titanium and alu-
HAZ continue to widen as they grow into the
weld fusion zone. minum result in a coarse-grained weld and HAZ
The maximum columnar grain width in the structure with small amounts of Ni3(Al,Ti)phase
weld metal is limited only by the physical size of along the y (nickel solid solution) grain bounda-
the weld bead and the arc energy input. For ex- ries. If the weld cooling rate is fast, as in an
ample, it is virtually impossible for a gas tungsten electron beam weld, the Ni3(AI,Ti) does not form
arc weld deposited on a coarse-grained copper at all on cooling, but remains in a supersaturated
casting to exhibit any distinguishable HAZ be- solid solution. Subsequent aging treatments only
cause the residence time is insufficient to cause precipitate Ni3(AI,Ti) as microscopic particles
noticeable grain growth; for example, D in Eq 6 is throughout the weld and the HAZ.
insignificantly greater than Do. Furthermore, be- Multiple-Pass Welds. Grains of single-phase
ca~se t~e large base-metal grains that grow epi-
metals continue to grow without obstruction
taxially mto the weld must squeeze into a bead of through each succeeding weld pass of multiple-
limited volume, lateral growth of the columnar pass welds, until all of the required passes are
grains in the weld is not possible. complete. Such interpass epitaxial growth leads
When cold-worked alloys are welded, the HAZ to coarse columnar grain structures and extreme
experiences both recrystallization and grain anisotropy of mechanical properties. Peening or
growth reactions. The hardness and strength cold working each weld pass prior to deposition
properties of the recrystallized HAZ lose the of the subsequent pass helps mitigate the prob-
benefits derived by cold working, and joint lem. The peening action sufficiently cold works
strength approaches that of an annealed alloy. the columnar grains of a freshly deposited pass to
Although weld-metal properties can always be cause development of a refined or recrystallized
controlled by judicious alloying, HAZ properties grain structure in the new HAZ of this weld pass.
can only be controlled by regulating heat input or Through epitaxial growth, these refined grains
by changing the base-metal composition. grow into the weld. By interpass peening, colum-
In steels and other metals that undergo allo- nar grains are restricted to growth only within
tropic phase transformation, the HAZ is conven- each weld pass, thus greatly reducing the overall
iently divided into two regions: grain size and anisotropy of multiple-pass welds
of single-phase alloys. Peening is not recom-
The grain growth region, which lies adjacent mended by most welding codes for the first and
to the weld interface last (surface) passes because of the likelihood of
The grain-refined region, which is farther fracturing the first pass and heavily distorting the
away from the weld interface surface of the last pass.