|ABSTRACT. The solidification and weldability
of experimental superalloys containing systematic variations in Fe, Nb,
Si, and C were studied using differential thermal analysis (DTA), microstructural
characterization techniques, and Varestraint testing. Microstructural evolution
during solidification of DTA samples and gas tungsten arc welds generally
occurred by a three-step solidification process. Solidification initiated
by a primary L
reaction, which enriches the interdendritic liquid in Nb and C until a
eutectic-type L (
+ NbC) reaction occurs over a rather broad temperature range. Solidification
terminated by a second eutectic-type reaction, L (
+ Lave), which occurred over a narrow temperature range. Carbon additions
increased the start temperature of the L
( + NbC) reaction and induced
a concomitant decrease in the temperature interval of the primary L
stage of solidification. As a result, carbon additions often promoted a
significant improvement in solidification cracking resistance. The results
from solidification cracks studies are combined with detailed examinations
of microstructural morphologies in fusion zone solidification cracks to
propose a microstructure morphology classification scheme. The classification
scheme provides a phenomenological explanation of the relation between
alloy composition, solidification temperature range, microstructural morphology
and cracking propensity. This interpretation also leads to an improved
correlation between the solidification temperature range and fusion zone
of cracks during solidification can occur in ingots, castings, and welds.
The phenomenon has been investigated for quite some time and the general
features are understood in a qualitative sense (Ref. 1). Work is also in
progress to quantitatively determine conditions that lead to solidification
cracking (Refs. 2, 3). In general, cracks can form in certain alloy systems
during the terminal stages of solidification when a liquid film is distributed
along grain boundaries and interdendritic regions. At this stage, shrinkage
strains across the partially solidified boundaries can become appreciable.
If the terminal liquid is distributed along the boundaries as a continuous
film, the strains cannot be accommodated and the boundaries separate to
form a crack. In terms of material factors, the solidification temperature
range and morphology of the interfacial liquid that exists at the terminal
stages of solidification are primary factors which control solidification
cracking susceptibility. Solute redistribution plays an important role
in solidification cracking as it affects the solidification temperature
range and amount of terminal liquid. The effect of the solidification temperature
range can be understood in simplified terms by considering its influence
on the size of the solid + liquid (mushy) zone.
|For a fixed temperature gradient in the
mushy zone (constant processing parameters), composition variations that
promote low-temperature eutectic-type reactions at the terminal stages
of solidification will widen the solidification temperature range and generally
aggravate cracking tendency by expanding the crack susceptible mushy zone.
The actual distance a solidification crack propagates through the mushy
zone depends on the distribution of terminal liquid, which exists near
the end of the solid + liquid region (Ref I. 1). The distribution of liquid
near the end of the mushy zone is, in turn, controlled by the amount of
terminal liquid and solid/liquid surface tension. When the amount of terminal
liquid is moderate, between approximately 1 to 10 vol-% (Ref. 4), and/or
the surface tension is low, the liquid tends to wet the boundary and forms
a continuous film. This type of morphology is most detrimental as it interferes
with the formation of solid/solid boundaries, thus reducing the ability
of the material to accommodate shrinkage strains. In contrast, a small
amount of terminal liquid, generally less than approximately 1 vol-% (Ref.
4), which exhibits a high surface tension with the solid, will often exist
as isolated globules and promote solid/solid bridging, thereby reducing
cracking tendency. When the amount of terminal liquid is high, (greater
than approximately 10 vol-%), it can often flow into the cracks and provide
a "crack healing" effect (Refs. 4, 5). For a given alloy system, the solidification
temperature range and amount of terminal liquid are controlled primarily
by composition (processing parameters become important under high cooling
rate conditions where dendrite tip undercooling can be significant, (Ref.
6). Thus, many studies have been aimed at establishing the relation between
solidification cracking susceptibility and alloy composition (Refs. 7-9).
Niobium-strengthened superalloys are used extensively in applications requiring high-temperature strength and resistance to oxidation. Commercial examples include alloys IN625, IN706, IN718, IN903, IN909, and Thermo-Span. These alloys are often used for components that are fabricated by fusion welding. Thus, it is important to establish the relation between alloy composition and cracking tendency in order to avoid fusion zone cracking when these materials are used in critical applications. Previous studies conducted on commercial Nb-bearing superalloys (Refs. 7, 8, 10-12) have sbown that minor variations in Nb, Si, and C have a strong influence on the solidification temperature range, type and amount of secondary phases that form during the terminal stages ofsolidification and result in solidification cracking susceptibility. In general, two types of eutectic-type constituents, /NbC and /Laves, are known to form as the weld metal solidifies. The /Laves constituent is most deleterious in terms of hot cracking as it forms at a lower temperature and thus expands the solidification temperature range. These alloys are also often joined to iron-based alloys such as carbon steels and stainless steels (Refs. 10, 13-15). In these applications,the fusion zone composition can become significantly enriched in Fe and C due to dilution effects, and this composition modification can significantly alter the solidification behavior and associated cracking tendency. For example, Patterson and Milewski evaluated the hot cracking susceptibility of dissimilar metal welds between 304L and IN625 (Ref. 14). The results of that work showed that the diluted fusion zones were more susceptible to cracking than either of the base materials. Subsequent work by Cieslak (Ref. 15) showed that the /Laves constituent forms in the fusion zone of this dissimilar alloy combination. Robitz (Ref. 13) evaluated the solidification cracking of IN625 claddings on C-steel and found that Nb, Si, C, and Fe all increased cracking susceptibility, although no detailed results were reported. DuPont (Ref. 10) showed that solidification cracking in IN625 weld cladding (deposited on C steel) was promoted by formation of the /Laves constituent. Despite the industrial importance of these superalloys and the need to minimize solidification cracking, no general model has been developed for predicting the relationship between alloy composition, solidification behavior (e.g., solidification temperature range and amount of /NbC and /Laves) and resultant weldability. In separate articles, we reported on the development of a solidification model that can be used to predict the solidification temperature range and amounts of /NbC and /Laves constituents as a function of alloy composition (Refs. 16, 17). In this article, the link between solidification behavior and weldability is discussed for a wide range of experimental alloy compositions.
Experimental Alloy Compositions
A four factor, two level set of experimental alloys was designed to simulate commercial compositions of interest to this study. The alloy compositions are summarized in Table 1. The alloys exhibit factorial variations in Fe (in exchange for Ni), Nb, Si, and C at two levels. The high- and low-target levels of Nb, Si, and C are set as follows (all values in wt-%): 2< Nb <5, 0.02< C <0.15, and 0.10 < Si < 0.60. These limits were chosen to represent low- and high-composition values of wrought alloys and filler metals as well as composition limits that can arise in fusion welds made between nickel-based alloys and carbon steels. The nominal Cr content was selected at 20 wt-%, which is typically used in many commercial alloys for good corrosion resistance. Several additional alloys with intermediate C contents (Alloys 1.5, 3.5, 7.5, and 11.5) were also investigated. The alloys were prepared at Sandia National Laboratories by investment casting. All samples were melted and poured in vacuum to produce six bars of each composition with approximate dimensions of 170 x 25 x 6.3 mm. A high-Nb alloy (Ni-1) was also added for supplemental differential thermal analysis measurements. Small cast buttons of this alloy were produced from high-purity powders (> 99.99 wt-%) by gas tungsten arc melting.
ND=Not determined. All values in wt-%.
Differential Thermal Analysis
Differential thermal analysis (DTA) was conducted on a Netzsch STA 409 differential thermal analyzer using 500 to 550 mg samples. The DTA system was calibrated within 2°C of the melting point of pure Ni. Samples were melted and solidified under flowing argon in alumina crucibles using pure Ni as the reference material. The liquidus temperatures were determined by heating samples slowly at a rate of 5°C/min to approximately 10°C above their liquidus temperature.
|The samples were then solidified at a
relatively fast cooling rate of 20°C/min to determine temperatures
of eutectic-type reactions which occur under nonequilibrium solidification
conditions. Reaction temperatures were taken as deviations from the local
baseline. All reported reaction temperatures are an average from at least
three tests, and the order of testing was randomized.
The solidification cracking susceptibility of each alloy was evaluated using the Varestraint test (Refs. 18, 19). The as-cast samples were machined to typical subsize Varestraint specimens (165 x 25 x 3.2 mm). The welds were produced under the following parameters: 95 A, 10 V (2.5 mm arc length), and 3.3 mm/s travel speed with high-purity argon shielding. A 3.2-mm diameter W-2ThO electrode was used with a 60-deg tip angle. In order to acquire adequate statistics, three tests were conducted on each alloy at an augmented strain of 2.5% to simulate highly restrained welds. The maximum crack length (MCL) was used as the indicator of cracking susceptibility and was measured using a light optical microscope at 100X magnification. The level of 2.5% augmented strain was selected on the basis of a large number or tests at varying strain levels (conducted using essentially identical testing equipment, welding conditions and weld sizes) on a range of alloys with similar solidification characteristics and solidification microstructure (Refs. 8, 20, 26). These results have shown that these test conditions, for alloys that solidify over similar temperature ranges and which exhibit MCL values spanning the range observed in the current study, the MCL is generally saturated at the 2.5% strain level. At these strain levels, it is normally expected that cracking will span a major fraction of, if not the entire, length of the mushy zone.
microscopy (LOM) was conducted on Varestraint and DTA samples polished
through 0.04 µm colloidal silica and electrolytically etched in a
10% chromic acid +90% water solution at 3 V. Scanning electron microscopy
was conducted on weld metals (prepared using the same procedure) using
a JEOL 6300 field emission gun scanning electron microscope (FEG-SEM) at
an accelerating voltage of 15kV. The Varestraint samples were mounted in
thermal setting epoxy to provide a planar view of the solidification cracks
which intersected the specimen surface. After mounting, the surfaces were
lightly ground with 600-grit SiC paper to remove a very thin layer of surface
material and provide a flat area in the crack region for subsequent polishing
and etching. Quantitative image analysis (QIA) was conducted on autogenous
GTA welds at locations far removed from solidification cracks. Area fractions
of total eutectic-type contents and, where possible, individual /NbC
and /Laves eutectic-type
constituents were measured along the centerline of each weld with at least
ten SEM photomicrographs. Area fractions were assumed to be equivalent
to volume fractions. Electron probe microanalysis (EPMA-WDS) was conducted
on a JEOL 733 probe at an accelerating voltage of 15kV and beam current
of 20 nA. All EPMA samples were mounted in thermal setting epoxy, polished
flat to a 0.3-µm finish using an alumina slurry, ultrasonically cleaned
in acetone and carbon-coated prior to analysis. The K
lines were used for Fe, Ni, Cr, and Si, while the L
line was used for Nb. Raw data were reduced to weight percentages using
a ZAF algorithm.
|Results and Discussion
behavior of these alloys has been described through microstructual analysis
(Ref. 16) and solidification modeling (Ref. 17) in separate articles. The
important features are summarized below as they provide a necessary framework
for detailed interpretations of the weldability results. Typical DTA cooling
scans for Alloys 2 and 16, which represent the two types of reaction sequences
observed among the experimental alloys, are presented in Fig. 1. The corresponding
microstructures of GTA welds and DTA samples are presented in Fig. 2. The
first type of solidification sequence, which was observed for all alloys
except 2, 3.5, and 4, can be described by a three-step process: 1) Primary
solidification in which the interdendritic liquid becomes enriched in Nb
and C; followed by 2) a eutectic-type L
(+NbC) reaction, which depletes
the interdendritic liquid of C; and 3) termination of solidification by
a second eutectic-type reaction L
( + Laves). The broad peak
associated with the L
( + NbC) reaction indicates
this transformation occurs over a wide temperature range, while the narrow
peak associated with the L
( + Laves) reaction indicates
this step occurs over a small temperature interval (Ref. 16). Identification
of the NbC and Laves phases shown in Fig. 2 was confirmed through EPMA
measurements on coarse phases in DTA samples and backscattered electron
kikuchi patterns collected from finer phases in weld metal samples (Refs.
16, 21). The Ni-based alloys with high C/low Nb (Alloys 2, 3.5, and 4)
represent the only exception to this general solidification sequence. These
alloys did not show any exothermic peaks after the L
( + NbC) reaction and did
not exhibit any of the Laves phase in the as-solidified micro structure.
This indicates that solidification terminated with the L
( + NbC) reaction. The amounts
of total eutectic-type constituents (/NbC
+ /Laves) and, where
possible, individual /NbC
and /Laves constituents,
are summirized in Figs. 3 and 4. The data are arranged to permit comparisons
among Ni-based and Fe-based alloys with similar levels of solute elements
(Nb, Si, and C).
Fig. 2 -- Microstructures of DTA samples
and GTA welds for Alloy 2 and Alloy 16. A -- Alloy 2 DTA sample; B -- Alloy
2 weld; C -- Alloy 16 DTA sample; D -- Alloy 16 weld.
Fig. 4 -- Quantitative image analysis results showing amount and individual /NbC and /Laves constituents.
reactions in these experimental multicomponent alloys and commercial alloys
are similar to those expected in the pure ternary Ni-Nb-C system (Refs.
28, 29). The liquidus projection for this system is shown in Fig. 5. The
liquidus projection exhibits three primary phase fields that are of interest
here, , NbC, and Ni3Nb.
A primary C (graphite) phase field exists at high C contents, which is
not of importance. As noted earlier, additions of Fe, Cr, and Si to the
Ni-Nb system are well known to promote Laves at the expense of Ni3Nb
in commercial superalloys, as well as the experimental alloys utilized
in this work. Thus, by replacing Ni3Nb with Laves, the Ni-Nb-C
liquidus projection can be utilized as a guide in developing a description
of the solidification reactions in these alloys (Ref. 29).
|According to the proposed Ni-Nb-C liquidus
projection, solidification should terminate with the ternary L
( + NbC + Laves) reaction
where the liquidus surface is at a minimum. Under this condition, the Laves
and ternary and NbC
phases should be intermixed in the final microstructure. However, this
type of structure is not observed in any of the alloys that form both the
NbC and Laves phases (e.g., Fig. 2). Instead, the /NbC
and /Laves eutectic-type
constituents are always distinctly separated. This suggests that the liquidus
projection for the alloys in this work is more properly represented by
a class II reaction (Ref. 30). In this case, the local minimum on the liquidus
surface occurs where the line of two-fold saturation separating the
and Laves phases intersects the Ni-Nb side of the diagram. The solidification
sequence for alloys with this type of surface would occur as follows. After
the liquid composition travels along the line of two-fold saturation separating
the and NbC phases
during the L (
+ NbC) reaction. This process continues until the class II reaction is
reached, at which point the NbC stops forming as the L
( + NbC) reaction is replaced
by the L (
+ Laves) reaction. Solidification is completed at the Ni-Nb side of the
diagram by the L (
+ Laves) reaction. This solidification sequence accounts for the two spatially
separate /NbC and /Laves
eutectic-type constituents, which are observed experimentally.
The qualitative description of solidification discussed above has been described in a quantitative sense in separate articles (Refs. 16, 17). In that work, the position of the boundaries separating the , NbC, and Laves phase fields for these experimental alloys was determined through EPMA measurements. The multicomponent alloys were modeled as a ternary system by grouping together the Fe, Ni, and Cr matrix elements to form the "component" of the -Nb-C system. Solute redistribution of Nb and C between the liquid and solid phases was modeled using a previous approach developed by Mehiabian and Flemings (Ref. 31), with modifications made to account for the high diffusion rate of C in the solid. Values for the equilibriurn distribution coefficients of Nb and C, which are needed in the solute redistribution calculations, were determined through a combination of EPMA and DTA techniques. A summary of the modeling results, which are important for interpreting the weldability data, is shown in Fig. 6. Figure 6A shows the calculated primary solidification paths of Alloys 1 through 4 superimposed on the experimentally determined liquidus projection. Note that the nominal alloy composition is given by the start of the solidification path. The alloy number of each path is noted in the figure. The start temperatures of the L ( + NbC) reaction, which represents the intersection of the primary solidification path with the line of two-fold saturation separating and NbC, are noted for several alloys. The start temperature for the L ( + Laves) reaction is also noted. These temperatures were determined through DTA.
The point of intersection between the primary solidification path and line of two-fold saturation is a strong function of C content. Although C additions are intuitively expected to promote the /NbC eutectic-type constituent, this analysis provides a quantitative rationale for the observed behavior. As the nominal C content is increased, the interdendritic liquid becomes more highly enriched in C and the intersection point occurs at higher C contents. As a result, the liquid composition must "travel" a long distance down the L ( + NbC) line of two-fold saturation, forming /NbC as it travels, before the /Laves constituent can possibly form. Carbon additions raise the start temperature of the L ( + NbC) reaction and induce a concomitant decrease in the temperature interval of the primary L stage of solidification. As discussed in more detail below, this effect has important implications with regard to fusion zone solidification cracking response.
Fig. 5 -- Liquidus projection for the
Fig. 6 -- A -- Calculated primary solidification paths of Alloys 1 through 4 superimposed on the experimentally determined liquidus projection; B -- calculated primary solidification paths of Alloys 2 and 10 superimposed on the experimentally determined liquidus projection (Refs. 16, 17).
| The effect of
matrix composition is presented in Fig. 6B. Alloys 2 and 10 have very similar
levels of solute content, but different matrix compositions (Alloy 2 is
Ni-based while Alloy 10 is Fe-based). The matrix composition has a strong
influence on the segregation potential of Nb (Ref. 16) as measured through
the equilibrium distribution coefficient, kNb. The value of
kNb in the Ni-based alloys is equal to 0.46, while kNb
in the Fe-based alloys is only 0.25 (Ref. 16). As kNb is reduced,
Nb segregates more aggessively to the liquid. Thus, at any given level
of C in the liquid, the Fe-based alloys will always possess more Nb. As
a result, the Fe-based alloys will have more liquid remaining after primary
solidification, and the Nb content in that remaining liquid will be higher.
These effects from the reduced kNb value are the primary reasons
for the higher amount of total eutectic-type constituents observed in the
Fe-based alloys -- Fig. 3. The validity of the solidification model was
confirmed by comparing the calculated and measured amounts of /NbC
and /Laves constituents
in the experimental alloys, and reasonable agreement was found (Ref. 17).
Effect of Solidification Temperature Range
Figure 7 shows
the maximum crack length (MCL) of each alloy. (The alloy compositions are
summarized in Table 1). Within the Ni-based alloys, there is a clear separation
in weldability among the low-C alloys (0.017
wt-% C) with relatively poor weldability and the high-C alloys (=0.052
wt-% C), which show very good resistance to solidification cracking. Within
the Fe-based alloys, the addition of C is only beneficial when the Nb content
is low (1.93 wt-%), and
the C level must be above 0.10 wt-% to provide the advantageous effect.
For example, Alloys 11.5 and 12 have essentially identical levels of all
other elements except C (Alloy 12 has 0.079 wt-% C and Alloy 11.5 has 0.116
wt-% C). This small variation in C content leads to a substantial difference
in the MCL values. Within the Fe-based alloys with high Nb (Alloys 13-16),
C has no beneficial effect even at the 0.21 wt-% level.
Fig. 7 -- Maximum crack length of each alloy.
| MCL values for
commercial alloys tested using identical Varestraint equipment, welding
conditions weld sizes and augmented strain are also plotted in Fig. 7.
Previous work (Refs. 8, 20) has shown that the saturated strain for these
alloys is <2.5% under these test conditions. Alloys IN718 and IN625
are Nb-bearing superalloys that have been shown to be fairly susceptible
to solidification cracking (Refs. 7, 8), other Nb-bearing alloys have shown
this relatively high susceptibility as well (Refs. 11, 26). These commercial
alloys have C levels of 0.04
wt-%, which is intermediate to the experimental alloys utilized here. It
is interesting to note that their MCL values fall within the range of the
experimental alloys. This indicates that the experimental alloys accurately
simulate the weldability behavior of comparable commercial superalloys.
The low MCL value of 304 stainless steel is typical, as this alloy is known
to exhibit excellent weldability. The high-C alloys show cracking resistance,
which is comparable to 304 stainless steel. This is a significant improvement
over the experimental and commercial alloys with low-to-moderate C contents.
As discussed in the section on solidification, C content has a significant effect on the primary solidification path and resultant solidification temperature range of these alloys. As the C content is increased, the primary solidification path is driven far into the C-rich side of the solidification surface and intersects the /Nbc line of two-fold saturation at relatively high temperatures near the end of solidification. Thus, C additions decrease the solidification temperature range of the primary L reaction. Considering this effect, together with the general relation between cracking tendency and the solidification temperature range (Ts), a correlation between MCL and Ts might be expected. The solidification temperature range can be determined from DTA measurements. With respect to the upper limit of the solidification temperature range, the weld metal solidifies epitaxially from the base metal so there is no undercooling during solidification (Ref. 32) (i.e. undercooling associated with nucleation of solid). Therefore, the upper limit of the solidification temperature range is well defined by DTA measurements of the on-heating liquidus. With respect to the lower limit of the solidification temperature range, previous work (Refs. 16, 17) showed that the type, quantity and composition of the terminal constituents was essentially identical for the DTA samples and GTA welds of the same bulk composition. Since these terminal constituents have the same composition in either case and one phase of the constituents (the austenite) can grow easily from the primary constituent, it is to be expected that the solidification temperature (range) of the terminal constituents will be similar irrespective of the cooling rate. Moreover, the modeling results from the previous work (Refs. 16, 17, 33) showed that for the cooling rates in the DTA samples and the GTA welds, the partitioning of the two elements, which primarily affect the solidification temperature range (i.e. Nb and C), is similar in both cases. Therefore, the DTA measurements of solidification temperature range provide a reasonable measure of the solidification temperature range for the GTA welds.
|The liquidus temperature of each alloy
is listed in Table 2 along with the temperatures of eutectic-type reactions,
which occur during solidification. All samples, except Alloys 2, 3.5, and
4, exhibited the /Laves
eutectic-type constituent. Thus, the temperature of the L
( + Laves) reaction should
be utilized as the value for the terminal solidus temperature for all alloys
except 2, 3.5, and 4. For the Ni-based alloys, the /Laves
constituent formed in very small amounts, which precluded measurement of
the L (
+ Laves) reaction temperature by DTA. Thus, the temperature for this reaction
was estimated from Alloy Ni-1, which formed a large amount of the /Laves
constituent and produced a large exothermic DTA peak associated with the
+ Laves) reaction. As shown by typical EPMA measurements in Table 3, the /Laves
constituent in Alloy Ni-1 exhibited a composition very similar to that
observed in the Ni-based alloys. Thus, the reaction temperature measured
for Alloy Ni-1 should serve as a good estimate for the terminal solidus
temperatures of the Ni-based alloys, which complete solidification by the
+ Laves) reaction. The L (
+ Laves) reaction temperature could not be directly measured in the Fe-based
alloys with low Nb. The values measured in the high-Nb iron-based alloys
exhibit a fairly narrow range of 13°C. Thus, the average of these values
should provide a good estimate for the low-Nb alloys. For Alloys 2, 3.5,
and 4, the terminal solidus is taken as the value measured for the L
( + NbC) reaction as no /Laves
constituent was observed. Positive identification of secondary phases could
not be made in Alloy 1 or 1.5, so no solidification temperature range is
reported for these alloys.
The /Laves composition measured by EMPA defines the composition of the liquid at the L ( + Laves) reaction (i.e., the eutectic composition) (Refs. 8, 21). Note that the amount of Nb required in the liquid to start the L ( + Laves) reaction is lower in the Fe-based alloys. (Average Nb content is 23.1 wt-% in the Ni-based alloys and 20.4 wt-% in the Fe-based alloys). Thus, the L ( + Laves) reaction will start forming earlier in the solidification process with the Fe-based alloys and lead to a higher amount of the /Laves constituent. This effect, together with the reduced value kNb in the Fe-based alloys, accounts for the higher amount of /Laves observed in the Fe-based alloys.
NA = Not applicable.
ND - Not detected.
(a) Estimated from sample Ni-1.
(b) Estimated from average of Fe base high Nb alloys.
Values in parenthesis are standard deviations calculated from at lteast three measurements.
| Figure 8 plots
the MCL as a function of the solidification temperature range, and no correlation
is readily apparent. Many alloys exhibit a rather wide solidification temperature
range, but show good weldability (low MCL). It should be emphasized that
Cieslak (Ref. 7) was also unable to obtain a correlation between MCL and Ts
in his previous weldability reports on alloy IN625. This lack of correlation
between MCL and Ts
suggests that crack propagation is being affected by the amount and distribution
of solid and liquid phases within the mushy zone. Such potential effects
can be assessed by conducting detailed microstructural characterization
of the eutectic-type constituents in the vicinity of solidification cracks.
These constituents transform from the terminal liquid, which existed in
the grain boundary and interdendritic regions just before solidification
is complete. Thus, the amount and morphology of these constituents provides
a basis for developing an understanding of the distribution of solid and
liquid phases within the mushy zone.
Table 3 - /Laves Average Constituent Compsitions.
Microstructural Evolution and Solidification Cracking
The general solidification process of these alloys was described in the solidification section above. By combining the microstructural characterization results of Varestraint samples with the solidification sequences already discussed, it is possible to extend this general solidification description to identify four individual types of microstructural morphologies and their effect on cracking susceptibility. This proposed classification scheme sheds light on a more appropriate relation between MCL and Ts which could not be resolved by the initial plot in Fig. 8. All of the alloys can be grouped into one of four morphological classification schemes, and Table 4 lists the classification scheme that applies to each alloy.
Table 4 - Summary of Morphological Classification Schemes
| The development
of each microstructural morphology is described schematically in Fig. 9.
At the bottom of each figure, the formation of two neighboring dendrites
within a grain are depicted growing into the trailing edge of the weld
pool. The relation between temperature and phase stability in the mushy
zone is shown by the temperature gradient diagrain in the middle of the
figure, where the dendrite tips are at the liquidus temperature. This point,
given by zero distance behind the pool, represents the boundary between
the liquid weld pool and solid + liquid mushy zone. The distance at which
the actual temperature in the mushy zone reaches the terminal solidus temperature
defines the boundary between the mushy zone and fully solidified weld metal.
The relation between temperature, distance, phase stability, and the solidification
path is given by combining the -Nb-C
liquidus surface with the other two figures. It is important to emphasize
that the schematic illustrations presented in Fig. 9 are not intended to
represent all of the details to microstructural evolution in the mushy
zone nor are they intended to imply that single primary dendrite arms traverse
the entire mushy zone. Rather, they are intended to pictorially describe
the sequence of solidification events and the general distribution of microstructural
constituents as they affect cracking susceptibility.
The Type I microstructure that develops during solidification cracking (Fig. 9A) is found in the Ni-based alloys with Iow Nb and high C that exhibit only the L ( + NbC) reaction (i.e. Alloys 2, 3.5, and 4). A typical SEM photomicrograph from this group is presented in Fig. 10. At Step 1, the dendrite tips form from liquid of nominal composition as the actual temperature intersects the liquidus temperature (assuming negligible tip undercooling). Due to the high C/Nb ratio of these alloys, the interdendritic liquid then becomes highly enriched in C as the solidification path progresses towards the line of two-fold saturation between and NbC. At Step 2, the primary solidification path intersects the line of two-fold saturation at a relatively high temperature. As a result, the /NbC starts forming at a short distance from the dendrite tips. In other words, the primary L solidification process is complete within a relatively small distance behind the advancing weld pool. There is a moderate amount of liquid present at Step 2. This is directly evident from the measured amount of the /NbC constituent, which is in the range of 3.0 to 7.3 vol-% for this class of alloys. Because of the nature ofsolute redistribution (particularly carbon) in these alloys, the remaining liquid is completely consumed along the line of two-fold saturation, so the /Laves constituent does not form (Ref. 17). From this description, it is apparent that the mushy zone in these alloys is expected to be relatively small. Thus, the distance available for crack propagation is also relatively small, and this provides a major contribution to the excellent weldability observed experimentally.
In addition to the small mushy zone, the microstructures of Alloy 2 and 4 exhibit moderate amounts of the /NbC eutectic-type constituent which often back-fills a portion of the crack -- Fig. 10. A smaller amount of /NbC was observed in the solidification cracks in Alloy 1.5. The level of liquid needed for back filling has been reported to be in the range of 6-10 vol-% (Refs. 4, 5), which is very close to the amount of terminal liquid present in Alloys 2 and 4 (5.21.5 and 7.3 1.3). Thus, when strain is imposed on the mushy zone of these alloys, a crack can only propagate a small distance and the leading edge of the crack is healed by back filling. This description accounts for the small MCL values, moderate amounts of /NbC observed within the solidification crack and lack of the /Laves constituent.
| Development of
the Type II microstructure is shown schematically in Fig. 9B. This morphology
is observed in the high-Nb/high C nickel-based alloys and low-Nb/high C
iron-based alloys. The solidification process is similar to Type I, except
that a larger amount of liquid is generally present at Step 2, which is
not completely consumed during the L
( + NbC) reaction. As a
result, the class II reaction point is reached and the /Laves
constituent forms. A typical SEM photomicrograph showing the /NbC
and /Laves morphology
in the hot crack region for this class of alloys is shown in Fig. 11. This
morphology was observed within the solidification cracks in all the Type
II alloys. Reference to Table 2 shows that the L
( + Laves) reaction occurs
at very low temperatures. Thus, the presence of liquid at the /Laves
composition will extend the mushy zone out to larger distances. This has
the potential of increasing the maximum crack length. However, the amount
of /Laves constituent
that forms in alloys of this group is below 2.3 vol-%. At this level, the /NbC
always envelops the /Laves
and keeps it isolated -- Fig. 11. This suggests that the last residual
liquid, from which the /Laves
forms, is also isolated within the mushy zone during crack advancement.
With this type of morphology, the isolated liquid pockets should have little
or no deleterious effect and crack propagation through the entire mushy
zone is therefore unlikely. The Varestraint data tend to support this.
The MCL values for these alloys is in the range of 0.35 to 0.89 mm, which
is similar to the range of 0.41 to 0.64 mm observed in the Type I alloys,
which form no /Laves
and have excellent weldability. This may account for the lack of correlation
and MCL in Fig. 8 and will be discussed in more detail later. Evidence
of back filling was clearly evident in these alloys as well, and an example
of this is provided in Fig. 12. This is not surprising, considering that
these alloys contain 7.3 to 15.2 vol-% of liquid at Step 2.
The amount of total eutectic-type constituent in the Type II alloys, which corresponds to the amount of liquid at Step 2, is always above 7.3 vol-%. It is interesting to note that two alloys which fit into this microstructure type (Alloys 7.5 and 11.5) have almost identical amounts of liquid at Step 2 as Alloy 4, yet Alloy 4 does not form the /Laves constituent. The ability for liquid to remain after the L ( + NbC) reaction, and subsequently transform to /Laves, depends not only on the amount of liquid at Step 2, but also on the relative position of Step 2 and the segregation potential of Nb during the L ( + NbC) reaction. For example, Alloy 7.5 and Alloy 4 have essentially identical amounts of liquid at Step 2 (7.5 vol-%). However, due to its lower C/Nb ratio, the primary solidification path of Alloy 7.5 intersects the /NbC two-fold saturation line closer to the class H reaction point. Thus, the remaining liquid in Alloy 7.5 does not have to travel as far as that in Alloy 4 to satisfy the composition conditions for the L ( + Laves) reaction. Alloy 11.5 also has about 7.5 vol-% liquid at Step 2. In this case, the lower value of kNb in the Fe-based alloy (kNb = 0.25 in Fe-based alloys and kNb = 0.46 in the Ni-based alloys) causes Nb to segregate more aggressively to the liquid so that the /Laves composition can be reached in the liquid.
|| Formation of
the Type III microstructure is shown in Fig. 9C. This morphology is observed
in the iron-based Alloys 12, 14, and 16, The main difference between this
morphology and that of Type II lies within the amount and distribution
of the /Laves constituent,
where the Type III alloys form /Laves
in quantities greater than 2.3 vol-%. As a result, the /Laves
is observed in a continuous network. An example of this microstructural
morphology is shown in Fig. 13. The advantageous effect of C, which was
observed in the Ni-based alloys with high Nb and high C, is lost in the
Fe-based alloys with comparable composition. Again, this effect is attributed
to the higher segregation potential of Nb and lower amount of Nb required
in the liquid to initiate the L
( + Laves) reaction, both
of which translate into larger amounts of terminal liquid that undergoes
the L (
+ Laves) reaction at relatively low temperature. The actual size of the
mushy zone in these alloys is expected to be similar to that for the iron-based
Alloys 10 and 11.5, since each type terminates solidification with the
+ Laves) reaction. However, with the residual liquid existing in a continuous
network until the terminal L
( + Laves) reaction is reached,
crack propagation through most of the mushy zone is more favorable for
alloys that possess the Type III microstructure. Thus, the solidification
temperature range governing crack propagation in these alloys should be
appropriately given by the entire solidification range measured through
DTA. Note that the MCL values observed for these alloys (1.26 to 1.65 mm)
are consistently higher than those observed in the Type II alloys (0.35
to 0.89 mm) where the terminal liquid is isolated. As with the other alloys
that form a large amount of terminal eutectic liquid evidence of back filling
is observed in this microstructure type. However, it is worth noting that
at this applied stain level the amount of back filling is small compared
to the MCL for all of the four microstructure types. For example, as shown
in Fig. 3 Alloy 16 has almost 25 vol-% of eutectic-type constituent (the
highest amount among all the alloys), which is well beyond the values typically
reported to result in back filling (Refs. 4, 5). Although the alloy forms
a large amount of terminal eutectic liquid, inspection of Fig. 13A shows
that back filling occurs over a distance of approximately 0.14 mm, which
is small compared to the MCL value of 1.26 mm. In fact this amount of back
filling is not statistically different than the amount of variation typically
observed when conducting multiple Varestraint tests under identical testing
conditions. Thus, the values reported can be taken as an accurate representation
of the MCL for each alloy without having to make detailed corrections for
Lastly, the evolution of Type IV microstructures are shown in Fig. 9D. All of the low-C alloys fit into this classification category. These alloys exhibit a primary solidification path, which travels very close to the /Nb binary side of the solidification surface. The primary path barely intersects the line of two-fold saturation between and NbC before solidification terminates with the L ( + Laves) reaction. The eutectic-type reactions in this region of the solidification surface occur at low temperatures. Thus, with this solidification path, the mushy zone is relatively large and consists mainly of liquid and primary with small amounts of /NbC and /Laves. The total amount of terminal liquid in these alloys is generally low, between 0.8 and 3.8 vol-%. The only exception to this is Alloys 13 and 15, which form relatively large amounts of /Laves (12.9 vol-%) in a continuous network. The MCL values of all these alloys are high (1.23 to 1.70 mm), suggesting that crack propagation through all or most of the musby zone is likely. A typical example of this morphology is presented in Fig. 14.
Based on these considerations, the terminal solidus temperatures affecting solidification cracking for Type III and Type IV microstructures should be appropriately given by the temperature of the terminal L ( + Laves) reaction. However, assuming crack propagation is difficult through the isolated regions of residual liquid in the Type II microstructures, the temperature range affecting solidification cracking for these alloys would be more realistically represented by using the temperature of the L ( + NbC) reaction as the terminal solidus point. The MCL and Ts data are replotted in Fig. 15 using this effective solidification temperature range, and the trend between cracking susceptibility and the solidification temperature range is now more apparent. The data are grouped into two distinct regions of the plot which separate the low- and high-C alloys. This improved correlation, although somewhat qualitative, serves to support the proposed mechanisms of crack propagation through the various types of microstructures. These results show that cracking susceptibility can be interpreted by knowledge of the solidification temperature range and type/amount of constituent that forms at the terminal stages of solidificaion. These factors are controlled by the solidification path as shown in Fig. 6 and schematically in Fig. 9. The solidification path is, in turn, governed by alloy composition. Thus, it is useful to conclude by summarizing the relations between these factors.
The addition of
C is beneficial as it increases the start temperature of the L
( + NbC) reaction and reduces
the temperature interval of the primary L
stage of solidification. This favors formation of Types I and II microstructures,
which show very good weldability. This beneficial effect can be lost if
more than approximately 2.3 vol-% of the /Laves
constituent forms, at which point the Type III microstructure develops
with a concomitant reduction in cracking resistance. The /Laves
constituent is promoted by additions of Fe, Nb, and Si. In the Ni-based
alloys that have less than 11 wt-% Fe, the segregation potential of Nb
is relatively low and the /Laves
eutectic composition is relatively high. These two effects work together
to reduce the amount of /Laves
that forms, and C additions to the Ni-based alloys are consistently beneficial.
Within the Ni-based alloys, the transition from Type I to Type II microstructures
occurs when the Nb content is increased. The Type III microstructure does
not form in the Ni-based alloys since the /Laves
constituent is always below 2.3 vol-%.
|These two microstructural morphologies
are compared in Fig. 16. When the Fe-based levels have high Nb, a high
amount of /Laves forms
and leads to the Type III microstructure. Finally, for the low-carbon alloys,
the solidification path always travels along the /Nb
binary side of the liquidus surface and leads to the Type IV microstructure,
which have relatively poor weldability. These relations were discussed
qualitatively in this article. Results of solidification modeling presented
in previous articles can be used to predict these interrelations in a quantitative
sense (Refs. 16, 17).
and weldability ofexperimental Nb-bearing superalloys with systematic variations
in Fe, Nb, Si, and C was investigated by differential thermal analysis,
Varestraint testing, and
The authors extend their thanks to Dr. M. J. Cieslak at Sandia National Laboratories for usefull discussions and review of the manuscript. One author (JND) gratefully acknowledges financial support for this research from the American Welding Society Fellowship Award. Preparation of the experimental alloys by B. Damkroger and M. Maguire at Sandia National Laboratory is also greatly appreciated. Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy under Contract DE-ACO4-94AL85000.
1. Borland, J.
C. 1960. Generalized theory of super-solidus cracking in welds (and castings).
British Welding Journal, Vol. 7, pp, 508-512.
| 15. M. J. Cieslak,
C. R. Hills, and Headley, T. J. 1986. Microbeam Analysis, A. D.
Romig and W. F. Chalmers, eds., pp. 69-73.
16. DuPont, J. N., Robino, C. V., Marder, A. R., Notis, M. R., and Michael, J. R. 1997. Solidification of Nb-bearing superalloys, Part I. Reaction sequences. Accepted for publication in Metallurgical and Material Transactions A.
17. DuPont, J. N, Robino, C. V., Marder, A. R. 1997. Solidification of Nb-bearing superalloys, Part II: Pseudo-ternary solidification surfaces. Accepted for publication in Metallurgical and Material Transactions A.
18. Savage, W. F., and Lundin, C. D. 1965. The Varestraint test. Welding Journal 44(10): 433-s to 442-s.
19. Savage, W. F., and Lundin, C. D. 1966. Application of the Varestraint technique to the study of weldability. Welding Journal 45(11): 497-s to 503-s.
20. Cieslak, M. J., Headley, T. J., and Frank, R. B. 1989. The welding metallurgy of custom age 625 PLUS alloy. Welding Journal 68(12): 473-s to 482-s.
21. DuPont, J. N. 1997. Solidification and Welding Metallurgy of Experimental Ni Base and Fe Base Alloys Containing Nb, Si, and C, Ph.D. Thesis, Lehigh University.
22. Blazina, Z., and Trojko, R. 1986. J. Less Common Metals, Vol. 119, pp. 297-305.
23. Wlodek, S. T. 1963. Trans. Am. Soc. Met., Vol. 56, pp. 287-303.
24. Cieslak, M. J., Headley, T. J., Knorovsky, G. A., Romig, A. D., Kollie, T. 1990. A comparison of the solidification behavior of Incoloy 909 and Inconel 718. Metall. Trans. A, Vol. 21A, pp. 479-488.
25. Nakkalil, R. Richards, N. L., Chaturvedi, M. C. 1993. Microstructural characterization of Incoloy 903 weldments. Metall. Trans. A, Vol. 24A, pp. 1169-1179.
26. Robino, C. V., Michael, J. R., Cieslak, M. J. 1996. The solidification and welding metallurgy of Thermo-Span alloy. Accepted for publication in Science and Technology of Welding and Joining.
27. Zhao, Q. H., Gao, Y., Devletian, J. H., McCarthy, J. M., and Wood, W. E. 1992. Microstructural analysis of Ni alloy 625 cladding over carbon steel. Proc. of 3rd Int. Conf. International Trends in Welding Science and Technology. S. A. David and J. M. Vitek, eds., ASM, Materials Park, Ohio, pp. 339-343.
28. Stadelmajer, H. H., and Fiedler, M. L. 1975. The ternary system nickel-niobium-carbon. Z. Metallkd., Vol. 66 (4): 224-225.
29. Radhakrishnan, B., and Thompson, R. G. 1989. Solidification of the nickel-base superalloy 718: a phase diagram approach. Metall. Trans. A, Vol. 20A, pp. 2866-2868.
30. Rhines, F. N. 1956. Phase Diagrams in Metallurgy, R. F. Mehl and M. B. Beaver, eds., McGraw Hill, New York, N.Y., pp. 175-185.
31. Mehrabian, R., and Flemings, M. C. 1970. Macrosegregation in ternary alloys. Metallurgical Transactions A, Vol. 1, pp. 455-464.
32. David, S. A., and Vitek, J. M. 1989. Correlation between solidification parameters and weld microstructures. International Materials Review, Vol. 34, pp. 213-245.
33. DuPont, J. N., Robino, C. V., Marder, A. R. 1997. Modeling solute redistribution and microstructural development in fusion welds of Nb bearing superalloys. Accepted for publication, Acta Materialia.