Characterization of Discontinuous CoarseningReaction Products in INCONEL� Alloy 740H� FusionWelds

DANIEL H. BECHETTI, JOHN N. DUPONT, MASASHI WATANABE,and JOHN J. DE BARBADILLO

Characterization of c¢ coarsened zones (CZs) in alloy 740H fusion welds via a variety of electronmicroscopy techniques was conducted. The effects of solute partitioning during nonequilibriumsolidification on the amount of strengthening precipitates along the grain boundaries wereevaluated via electron-probe microanalysis and scanning electron microscopy. Electronbackscatter diffraction was used to present evidence for the preferential growth of CZs towardregions of lower c¢ content, even if growth in that direction increases grain boundary area.Scanning electron microscopy and image analysis were used to quantify the propensity for CZsto develop along certain segments of the grain boundaries, as governed by the local variations inc¢ content. Scanning transmission electron microscopy with X-ray energy-dispersive spectrom-etry (XEDS) was used to assess the compositions of the matrix and precipitate phases within theCZs and to quantify the segregation of alloying components to the reaction front.Thermodynamic and kinetic modeling were used to compare calculated and experimentalcompositions. The work presented here provides new insight into the progression of thediscontinuous coarsening (DC) reaction in a complex engineering alloy.

DOI: 10.1007/s11661-016-3952-2� The Minerals, Metals & Materials Society and ASM International 2017

I. INTRODUCTION

GRAIN boundary sliding/grain boundary migration(GBM)-assisted discontinuous coarsening (DC) of the c¢phase has been identified as the mechanism of grainboundary c¢ coarsened zone (CZ) formation inINCONEL alloy 740H* fusion weld creep rupture

specimens.[1] The presence of these CZs has been linkedto a weld strength reduction in creep of about 30 pct,and their development has been correlated with opera-tional and processing conditions such as creep strain,composition, and postweld heat treatment.[2]

Discontinuous precipitation (DP) and DC reactionshave been identified and studied in the literature formany decades, but most of the in-depth characterizationof the reaction products has been performed on simplealloy systems (e.g., binaries and ternaries); well-an-nealed, strain-free microstructures; or bicrystals.[3–8]

Regions consistent with DP/DC have been identifiedin more complex systems, such as NIMONIC** 80A,[9]

but characterization of these regions has not beenexplored in great detail. Furthermore, a practicalsolution for the mitigation of these reactions in suchcomplex engineering alloys is not available.As demonstrated by previous work, discontinuous

phase transformations continue to be a significantproblem, with real consequences to the performance ofmaterials for the energy, automotive, and aerospaceindustries.[1,2,4,6,8,10] Unfortunately, nickel-based super-alloys, such as alloy 740H, are very complex systemswhose discontinuous phase transformation characteris-tics are difficult to describe using the available DP/DCtheories, which were developed for much simpler binaryand ternary systems. Some of the complicating factorshampering such an analysis include the presence of morethan 10 alloying components, compositional inhomo-geneities from weld metal solidification, continuousprecipitation of multiple phases, and long-termhigh-temperature plastic deformation during creep.However, because colonies of DC have been shown toaffect the creep performance of alloy 740H weld metal,in-depth characterization of these regions to understandwhy they form and how to mitigate them in this systemis warranted. Thus, the main objective of this work is to

DANIEL H. BECHETTI, JOHN N. DUPONT, and MASASHIWATANABE are with the Department of Materials Science andEngineering, Lehigh University, Bethlehem, PA 18015. Contact e-mail:dbechetti@gmail.com JOHN J. DE BARBADILLO is with theProduct and Process Development Department, Special MetalsCorporation, Huntington, WV 25705.

Manuscript submitted July 19, 2016.Article published online February 2, 2017

* INCONEL and 740H are registered trademarks of Special MetalsCorporation, Huntington, WV.

** NIMONIC is a registered trademark of Special Metals Corpo-ration, Huntington, WV.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 48A, APRIL 2017—1727

characterize the growth behavior of DC colonies withrespect to the local microstructural variations in alloy740H weld metal. Insight into the effect of microstruc-tural conditions imposed by the fusion welding process(e.g., microsegregation) on their development will bepursued. Also, an attempt will be made to identifywhether certain alloying additions promote the DCreaction by compositionally characterizing thematrix-matrix and matrix-precipitate interfaces in theDC reaction zone. By addressing these objectives, newinsights into the evolution of discontinuous reactionproducts in complex systems will be gained.

II. EXPERIMENTAL PROCEDURE

Creep testing and microstructural characterizationpresented in this work were performed on automatedsingle-pass autogenous bead-on-plate gas tungsten arc(GTA) welds produced at Lehigh University and amultipass manual GTA butt weld produced at SpecialMetals Corporation. The parameters used to fabricatethese welds are given in Table I. Cross sectioning of thesingle-pass weld revealed that it penetrated approxi-mately 20 pct through the thickness of the base plate.The compositions of the investigated alloys are given inTable II. The single-pass welds were fabricated on heats2127JY (alloy 740H) and HV1701 (a Nb-free variant ofalloy 740H), and the multipass weld was fabricatedusing heats HV1278 (alloy 740H wrought base metal)and HV1219 (alloy 740H filler metal).

Cross-weld specimens in the form of 10 9 10 9 75mm (H 9 W 9 L) bars with threaded ends wereextracted from the aforementioned welds for interruptedcreep and creep-rupture testing. The matrix for thistesting is shown in Table III. As given in Tables II andIII, most specimens were tested in the direct-agedcondition (aged at 1073 K (800 �C) for 4 hours and aircooled immediately after welding), while one sample wasgiven a 1373 K (1100 �C) for 4 hours homogenization

heat treatment in air using a Fisher Isotemp 550 mufflefurnace after welding. After homogenization, this spec-imen was water quenched and then given the aforemen-tioned aging treatment. Creep testing was performed at1123 K (850 �C) in a Gleeble 3500 thermal–mechanicalsimulator. Specimens were heated to the test tempera-ture at a rate of 2.75 �C/s and soaked at temperature for30 seconds prior to applying the stress. All creep testswere constant load, with an initial applied stress of 100MPa, calculated using the specimens’ initial cross-sec-tional area. Minimum-contact stainless steel hot gripswere used, resulting in a 1113 K £ T £ 1123 K(840 �C £ T £ 850 �C) hot zone approximately 20-mmwide, as measured by secondary thermocouples duringtesting. All sample surfaces were ground to a 600-gritfinish using SiC metallographic paper before testing.After rupture or test interruption at the times indi-

cated in Table III, the specimens were longitudinallycross sectioned, and regions of interest 250 lm by 500lm were marked using 300-g Vickers microhardnessindents. The specimens were prepared using standardmetallographic techniques and electrolytically etched at6 V in a solution of 20 mL H3PO4 and 150 mL H2SO4

saturated with CrO3. Secondary electron (SE) imagescovering the entirety of the 250 lm by 500 lm regions ofinterest were collected using a Hitachi 4300SE/N Schot-tky field emission scanning electron microscope (SEM)at an operating voltage of 20 keV. Measurements of thelength of grain boundaries covered by c¢ CZs in eachregion of interest were made using Image J. Creep voidsand cracks are predisposed to occur on grain boundariesthat contain CZs.[1,2] However, these features were notincluded as part of this analysis, because it could not beknown with certainty whether the boundaries thatcontained cracks or voids at the time of analysis hadpreviously contained CZs. Additional analysis includedelectron backscatter diffraction (EBSD) mapping ofseveral CZs, using the same microscope and operatingconditions in combination with energy-dispersive X-rayanalysis (EDAX) optical interface module (OIM) EBSD

Table I. Parameters for the Welds Characterized in This Study

Process Single-Pass Automated GTAW Multipass Manual GTAW

Current (A) 250 198Voltage (V) 12 ± 0.1 11Torch speed (mm/s) 2 0.9 (root)

3.4 (fill)Wire feed rate (mm/s) N/A 26 (root)

19 (fill)Shielding gas 100 pct Ar 75/25 Ar/HeElectrode 4-mm-diameter W-2 pct Th 3.2-mm-diameter W-2 pct ThJoint geometry bead on 12-mm-thick plate 60 deg included angle V groove

15-mm-thick plate4.8-mm root opening

Preweld heat treatment 1393 K (1120 �C), 1 h water quench1073 K (800 �C), 4 h air cool

1393 K (1120 �C), 1 hwater quench1073 K (800 �C), 4 hair cool

Postweld heat treatment 1073 K (800 �C), 4 h air coolor1373 K (1100 �C), 4 h water quench1073 K (800 �C), 4 h air cool

1073 K (800 �C), 4 hair cool

1728—VOLUME 48A, APRIL 2017 METALLURGICAL AND MATERIALS TRANSACTIONS A

collection software. Postprocessing of the EBSD datawas performed using EDAX OIM Data Analysissoftware and involved generation of pseudo-color mapswith a 3.5 deg point-to-point tolerance to accuratelyreconstruct the microstructure while also removing theeffects of local orientation variations due to the highdegree of deformation induced during creep.

In addition, micron-scale compositional analysis inthe area of the grain boundaries in as-solidified alloy740H weld metal was performed using wavelengthdispersive spectrometry (WDS) in a JEOL� JXA-8900R

electron microprobe operated at 15 keV with a probecurrent of 49.65 ± 0.04 nA. Mn, Fe, Ni, and Co WDScounts were collected using a LiF spectrometer. Nb, Ti,and Cr were collected using a pentaerythritol spectrom-eter. Si and Al were collected using a thallium acidphthalate spectrometer. The collection times for the

elements listed previously (in the order listed) were 20,20, 50, 50, 60, 50, 30, 60, and 60 seconds. ZAFcorrection using the Armstrong/Love–Scott modeldefined by Armstrong[11] was used to convert the rawcount data to elemental concentrations, and calculationsof the peak-intensity-based 95 pct confidence intervalfor each measurement were performed. Elemental totalswere consistently around 98 pct due to the exclusion ofminor alloying elements from the analysis.Further advanced microstructural characterization

was performed on several samples, denoted in Table III,using focused ion beam (FIB) and scanning transmis-sion electron microscopy (STEM) techniques. An FEIScios dual-beam FIB with a Ga ion source operated at30 keV was used to extract and thin cross sections ofgrain boundaries containing CZs to electron trans-parency. After FIB thinning, the specimens were Ar ionmilled in a Fischione NanoMill 1040 at an acceleratingvoltage of 900 eV for 20 minutes to remove Ga iondamage. Microstructural analysis via imaging andnanoscale compositional analysis using X-ray

Table II. Compositions of Alloys Investigated in This Study, as Measured via Wet Chemical/OES Analysis

Heat Number Ni Cr Co Nb Ti Al Mo Fe Si C Mn V W Zr Ta P Cu S

2127JY 50.08 24.60 20.19 1.51 1.36 1.34 0.01 0.24 0.14 0.05 0.29 0.009 0.043 0.023 0.008 0.007 0.03 <0.001HV1278 49.17 24.35 20.08 1.53 1.45 1.28 0.53 1.07 0.20 0.05 0.30 0.007 0.008 0.020 <0.001 0.002 <0.001 —HV1219 50.20 23.90 19.40 1.52 1.28 1.31 0.54 1.10 0.22 0.05 0.29 0.006 0.013 0.018 <0.001 0.002 0.11 0.001HV1701 51.54 24.81 20.32 0.01 1.42 1.43 0.036 0.07 0.16 0.04 0.002 0.007 0.043 — 0.008 0.004 0.01 0.001

Values given in weight percent.

Table III. Test Matrix for Interrupted Creep and Creep Rupture Specimens Investigated in This Study

Heat Weld TypePostweld HeatTreatment

Prestrain(Pct)

AppliedStress (MPa)

Time (h)R = Rupture

2127JY single-passautogenous GTAW

direct age* 0 0 0**, 1, 5, 10, 318

100 1, 5, 10, 25, 50,318**, 392 (R)

5 ± 0.5 0 1, 5, 10, 80

100 1, 5, 10, 25, 50, 80 (R)

1373 K(1100 �C),4 hwater quenchage*

0 100 326** (R)

single-pass autogenousGTAW, through-thickness weld

direct age* 479 (R)

HV1278 (base metal) +HV1219 (filler metal)

multipass GMAW direct age* 97 (R)

HV1701 Nb-free, single-pass GTAW direct age* 66 (R)

* Standard age: 1073 K (800 �C), 4 h, and air cool.** Specimens analyzed via STEM.

� JEOL is a trademark of JEOL, Tokyo.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 48A, APRIL 2017—1729

energy-dispersive spectrometry (XEDS) of the thinnedspecimens was carried out using a JEOL ARM 200CFSTEM operated at 200 keV. The probe current used forimaging and X-ray collection was either 220 or 303 pA,and the manufacturer-reported XEDS spatial resolutionwas 0.5 nm. EDS line scans were collected using a stepsize of either 0.5 or 1 nm with a dwell time of either 15 or20 seconds per point, and area scans were collected witha 300-second dwell time. Variations in probe currentand dwell time between scans were accounted for byquantifying the raw X-ray data using the iterativef-factor method developed by Watanabe and Wil-liams.[12] This allows the quantified X-ray results fromdifferent scans to be directly compared. X-ray counts forall elements were based on their respective Ka peaks.

III. RESULTS AND DISCUSSION

A. Grain Boundary Configurations in As-Solidified WeldMetal

During nonequilibrium solidification, solute parti-tioning creates local variations in composition through-out the solidification substructure. When columnardendrites form as part of this substructure, dendriteswithin the same grain will grow parallel to each other. Ina solidifying polycrystalline material where grains growepitaxially into the liquid from unmelted parent grains(e.g., weld metal solidification), the orientation of thedendrites in a grain is determined by the crystallographyof its parent grain. Therefore, it is expected thatepitaxial nonequilibrium solidification from a nontex-tured, random polycrystal will result in grains whosedendrites are randomly oriented with respect to those inneighboring grains. It follows, then, that when two suchrandomly oriented solidifying grains meet to form agrain boundary, a combination of compositional vari-ation within each grain and slight grain boundarymovement during cooling after solidification will pro-duce large variations in cross-boundary composition. Inmost cases (because composition across a dendritegenerally changes gradually until the interdendriticregion is reached[13,14]), three types of cross-boundarycomposition profiles will be present. There will be areasof (1) near constant composition where two dendritecores lie across the boundary from each other (core-coreor ‘‘CC’’ configuration), (2) near constant compositionwhere two interdendritic regions meet across the bound-ary from each other (interdendritic-interdendritic or‘‘II’’ configuration), and (3) steeply varied compositionwhere a dendrite core and an interdendritic region lieacross the grain boundary from each other (interden-dritic-core or ‘‘IC’’ configuration). Examples of thesecases are given in Figure 1, which presents a backscat-tered electron image from a grain boundary in as-solid-ified alloy 740H, and Figure 2, which presents electronmicroprobe traces across the three boundary configura-tions indicated in Figure 1. Figure 2 focuses on the Al,Ti, and Nb content across the boundary in each of theaforementioned local boundary configurations, becausethese are the elements that contribute most strongly to c¢

precipitation in alloy 740H. As anticipated, the case inwhich a dendrite core intersects an interdendritic regionat the grain boundary produces a sharp gradient inas-solidified Ti and Nb content across the boundary.The variation in Al is not very large because Al does notsegregate highly during solidification.[13,15]

The consequences of this local variation in c¢-formercontent in alloy 740H weld metal manifest duringsubsolidus cooling through the austenite+ c¢ phase fieldand during subsequent age hardening. c¢ is known toprecipitate quickly (even at the high cooling ratesexperienced after welding) in many systems, includingalloy 740H2. Thus, because areas enriched in the c¢formers will have a higher c¢ solvus temperature, someprecipitation of c¢ during weld metal cooling aftersolidification will occur, beginning in the interdendriticregions and concluding in the dendrite cores.[13] Thedifference in c¢ solvus between solute-rich interdendriticregions and solute-depleted dendrite cores can be up toseveral hundred degrees,[13] so significant growth of thefirst-formed c¢ (in the interdendritic regions) may occurduring cooling. The result of this is an as-solidifiedmicrostructure that contains a gradient in both c¢ phasefraction (because of the differences in local composi-tions) and average size (because of growth after precip-itation). This microstructure will persist after aging, asshown in Figure 3. Note that the c¢ phase fraction isrelatively high on each side of the boundary for the I–Icase and relatively low on each side of the boundary forthe C–C case. The c¢ phase fraction varies significantlyacross the boundary for the I–C condition. Thesedifferent microstructural conditions may favor the typeof grain boundary DC reactions observed in alloy 740H.The propensity for CZ formation in alloy 740H duringcreep has been related to creep strain accommodation bygrain boundary migration (GBM).[2] Therefore, differ-ences in c¢ size and phase fraction across a grainboundary, which translate to differences in particle

Fig. 1—Backscattered electron image of a typical grain boundary inan alloy 740H fusion weld. The dark area is a grain enclosed by thedenoted grain boundary. Three variants of local concentration acrossthe boundary are indicated by varying image contrast. I = interden-dritic region (high contrast due to solute enrichment), and C = den-drite core (low contrast due to solute depletion).

1730—VOLUME 48A, APRIL 2017 METALLURGICAL AND MATERIALS TRANSACTIONS A

pinning effects, may promote enhanced boundarymobility in the direction of lower c¢ content. This,coupled with the inherent driving force for coarseningwhen small precipitates in a dendrite core are in theimmediate vicinity of larger precipitates across a knownshort-circuit diffusion path such as a grain boundary,creates a microstructural condition that is supportive ofdirectional coarsening reactions like DC.

B. Correlation of CZ Development to Grain BoundaryConfiguration

As part of the analysis used in prior work to quantifythe amount of grain boundary CZ coverage in alloy

Fig. 2—Electron microprobe traces across the three local grainboundary configurations in alloy 740H fusion welds after solidifica-tion. I = interdendritic region (enriched in solute), and C = den-drite core (depleted of solute).

Fig. 3—Scanning electron micrographs of the three local grainboundary configurations in alloy 740H fusion welds after precipita-tion heat treatment at 1073 K (800 �C) for 4 h. I = interdendriticregion (enriched in c¢), and C = dendrite core (depleted of c¢). Thelarge particle at the lower right of the upper image is a carbide thatformed at the end of weld metal solidification due to solute enrich-ment.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 48A, APRIL 2017—1731

740H creep specimens,[2] the length of CZs occurringalong each of the three types of boundary configurationsdescribed previously was determined for 20 specimens.These values were then normalized to the total fractionsof CC, IC, and II boundaries (with and without CZs) inone specimen. The specimen used for normalizationcontained 4.5 mm of total grain boundary length, whichis believed to be enough boundary length to provide asufficiently representative sample. The total measuredfractions of each boundary type in this sample were CC= 0.22, IC = 0.30, and II = 0.48. The variation in thesefractions, and specifically the observation that the IIconfiguration is most common, is consistent with thecharacteristics of grain growth during weld metalsolidification. As discussed previously, differences inthe orientation of neighboring grains leads to the threetypes of boundary configurations. It does not imply,however, that the fraction of each boundary type will beequal, because competitive grain growth during solidi-fication occurs in the direction of the largest temperaturegradient and, therefore, will follow the moving heatsource (i.e., the welding arc). This results in changes inthe physical direction of grain growth to align the grainsalong the length of the weld.[16] Consequently, when twoneighboring grains’ substructures run parallel, the grainboundary that forms between them is most likely to formfrom solute-rich liquid, thus producing an II boundaryconfiguration. The CC and IC configurations would notbe favored during this regime of solidification, so theirpresence in significant fractions is attributed to samplingareas of nonparallel grain growth.

The normalized fraction of grain boundaries in eachconfiguration that contained CZs across the 20 interro-gated samples was then determined by

fCZi ¼ LCZi

fnormi � LGB½1�

where fCZi is the fraction of grain boundaries inconfiguration i containing CZs, LCZ

i is the total length

of boundaries in configuration i that exhibit CZs, fnormi isthe total fraction of grain boundaries in configuration i(with and without CZs), and LGB is the total grainboundary length in all analyzed samples. The results ofthis analysis are presented in Figure 4. As shown, thereis a clear preference for CZ growth along boundarysections in the IC configuration. Approximately 25 pctof these boundaries were shown to contain CZs, whereasonly around 10 pct of boundaries in each of the CC andII conditions contained CZs. Previous work has assertedthat the increased susceptibility of alloy 740H weldmetal to DC (as compared to wrought alloy 740H) stemsfrom the microstructural gradients that develop in theweld metal, as described previously. Specifically, it wasargued that the local differences in c¢ content result inlocal variations in the grain boundary mobility, which inturn increase the ability of grain boundaries to migrateto accommodate creep strain. This GBM was thenthought to serve as the source of moving grain bound-aries required for the DC reaction.[2] The currentobservation of increased propensity for CZ formationalong IC boundary segments is, therefore, a logicalextension of that argument, insomuch as the boundarysegments with the sharpest microstructural gradients arethe segments that are likely to experience the most strainlocalization during creep and migrate most easilybecause of the reduced c¢ fraction on the core side ofthe boundary. Further discussion of the GBM tenden-cies during DC in alloy 740H is presented in Sec-tion III–C. CC and II boundary segments are alsosubject to CZ development because of creep strainlocalization, but the localization of strain along thosesegments likely results from only the slowing dislocationmotion by the crystallographic boundary. Thus, thedriving force for GBM along those segments is expectedto be lower than along IC segments, and the measuredfraction of CZs along CC and II segments is corre-spondingly lower than along IC segments.It is acknowledged that there is some inherent

subjectivity involved in the determination of whether aparticular region is closer in c¢ phase fraction to adendrite core or interdendritic region, and that thisjudgment is made qualitatively. However, the obviousdifferences in c¢ content throughout the substructure(Figure 3) and the quantity of grain boundaries ana-lyzed (nearly 10 cm) appear to be large enough to justifyidentifying individual regions as consistent with aninterdendritic region or a dendrite core.

C. Correlation of CZ Growth Direction to GrainBoundary Configuration

Evidence in Section III–B led to the conclusion thatthe sharp change in microstructure across IC boundariespromotes the DC reaction, in general, on IC boundarysegments. In this section, electron microscopy tech-niques are used to address how those same microstruc-tural gradients affect the motion of grain boundaries ineach configuration during DC. Prior work has demon-strated the asymmetry of the CZs in alloy 740H abouttheir respective grain boundaries and shown that, atshort creep times, they grow preferentially into one of

Fig. 4—Quantification of the propensity for CZ formation alongsegments of each of the three identified boundary configurations.The total length of the grain boundaries analyzed is approximately10 cm. The data are normalized to the fractions of each type ofboundary configuration in one specimen.

1732—VOLUME 48A, APRIL 2017 METALLURGICAL AND MATERIALS TRANSACTIONS A

the grains adjacent to the boundary.[1] Further evidenceof this, with specific attention paid to the localmicrostructures adjacent to the grain boundaries, ispresented in Figure 5. The images shown are SE imagestaken from alloy 740H weld metal interrupted creepspecimens overlaid with EBSD maps collected from thearea adjacent to the CZs. Figures 5(a) and (b) demon-strate the typical growth direction for CZs occurringalong IC portions of the grain boundaries in thesespecimens. The CZs were observed to primarily growinto the side of the boundary containing a dendrite core(and, therefore, a lower volume fraction of c¢).

Previous work has asserted that the observed differ-ences in CZ content of alloy 740H base metal,direct-aged weld metal, and homogenization heat-treated weld metal after creep are related to the relativeease of and propensity for creep strain accommodationby GBM.[2] Figure 5 is consistent with this theory, as it

demonstrates a preference for CZ growth along ICboundary segments in the direction of least microstruc-tural resistance (i.e., increased mobility), despite anapparent increase in grain boundary area caused by thatgrowth. In addition, the preference for DC growth intothe dendrite core side of IC boundaries is consistent withthe strain localization that would be seen in boundarysegments of this configuration. In the CC and IIconfigurations, dislocation motion on either side of thegrain boundary during creep should progress relativelyuniformly because of the relatively uniform precipitatephase fraction on either side of the boundary. Somestrain localization is likely to occur in these boundarysegments due to the increased energy required to movedislocations across the boundary itself, so a drivingforce for boundary migration (and, therefore, DC)would be present along these segments, in agreementwith the experimentally observed development of CZ

Fig. 5—Scanning electron micrographs of CZs in alloy 740H fusion weld creep samples: (a) and (d) crept for 50 h and (b) and (c) crept for 318h. Both samples were crept at 1123 K (850 �C) with a 100 MPa applied stress. Micrographs are overlaid with EBSD maps of the CZs. White ar-rows indicate local substructure features and black arrows indicate the CZ growth direction.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 48A, APRIL 2017—1733

along CC segments (e.g., Figure 5(c)) and II segments(e.g., Figure 5(d)). In the IC configuration, dislocationmovement from the core side of the boundary to theinterdendritic side would be slowed by the boundaryitself and by the increased volume fraction of c¢ on theinterdendritic side. This would result in a much higherdegree of strain localization along IC boundary seg-ments, which is likely to enhance the driving force forGBM to accommodate the creep strain in these areas.Then, when the colonies begin to grow, they would growin a manner that seeks to provide a recovery mechanismfor the most highly strained locations (i.e., the dendritecore side). As mentioned previously, this growth direc-tion offers the added benefit of increased boundarymobility because of reduced particle drag.

A mechanism that defines the growth direction ofdiscontinuous reactions as being specifically related tolocal precipitate content variations and strain accumula-tion has not been rigorously addressed in the literature.However, it recently has been demonstrated that theoverall development of DC colonies in alloy 740H isstrain dependent and does not occur during stress-freeaging.[2] Also, in a related example, Hillert described DPas a type of local recrystallization reaction whereby acold-worked portion of a parent grain near a grainboundary is consumed by strain-relieving boundarymotion.[17] While that theory was developed for DP, itis plausible to extend its application to DC when there isan external source (creep) of deformation. If that sourceof deformation acts upon a grain boundary where adiscontinuous change in precipitate content is present, itis conceivable that it may result in a discontinuous changein strain accumulation that would preferentially drive theGBM into the more highly strained side of the boundary.

CZs were also present on grain boundary segments inthe II and CC conditions, as shown in Figures 5(c) and(d). Because the local microstructures (and, therefore,strain accumulation) on either side of these types ofboundary segments are expected to be similar, a growthdirection preference in accordance with the theorydescribed previously would be difficult to assign andmay simply be a matter of chance relating to theorientation of the original grain boundary precipitatesor crystallographic orientation of the grains. Thispossibility has been discussed in other work regardingdiscontinuous reactions.[18,19] CZ growth on theseboundaries, however, does not disqualify the idea thatthe direction of CZ growth on IC boundary segments isstrongly influenced by the local microstructure.

It is acknowledged that other driving forces unrelatedto the local microstructural constituents, such as localgrain boundary energy, initial grain boundary curva-ture, or a chemical driving force, may influence thedirection of grain boundary motion during DC. Theinfluence of grain boundary misorientation angle/con-figuration/energy has been discussed in the literature asa driving force for discontinuous reactions, and severalcorrelations have been observed.[7,8,20,21] However, thesestudies have generally focused on equiaxed microstruc-tures or bicrystals, so that grain boundary characteris-tics can be readily determined. In weld metalmicrostructures, the grain boundaries are highly curved

in three dimensions, which results in continuouslyvariable grain boundary properties and makes statisti-cally significant local measurement of such propertiesprohibitively difficult without extensive use of, forexample, serial sectioning techniques. Such analysiswas not deemed feasible in this work, so no conclusionscan be drawn regarding the effect of grain boundarygeometry/energy on the driving force for DC growth.Initial grain boundary curvature does not appear to be amajor determining factor in the choice of GBM direc-tion, because, as indicated in Figure 5, the boundariesroutinely moved in a direction that increased grainboundary area. In addition, previous work has demon-strated that CZs did not develop in alloy 740H weldmetal that was isothermally aged at 1123 K (850 �C)without the application of stress.[2] This is inconsistentwith a mechanism of DC that relies on significantstraightening of the highly curved weld metal grainboundaries. Finally, Hillert has described a mechanismof GBM that is driven by the discontinuous concentra-tion gradients set up in the nanometers ahead of thereaction front during a discontinuous phase transfor-mation.[17] In this case, it is conceivable that themicroscale composition changes due to solidificationcould provide the significant chemical driving force forpreferential DC growth direction by creating a discon-tinuous change in both matrix and precipitate compo-sition across an IC boundary. However, CZs have beenshown to develop along II and CC boundaries, wherethe microscale concentration gradients are small tonegligible. This points to a lack of a large chemicaldriving force and the importance of an alternate,strain-related one.

D. STEM Analysis of CZs

A large body of work regarding discontinuous phasetransformations is available in the literature, and thesephenomena have been studied since at least the1950s.[3,14,19,20,22–26] Much of the early work on thesereactions involved simple binary and ternary sys-tems[4,19,24,25,27]; additional seminal studies on discontin-uous reactions in particle-strengthened systems wereperformed later,[20,22,28] and fairly recent studies haveaimed at addressing these transformations in morecomplex alloy systems in great detail.[8,9,29,30] Despitethis abundance of information, no consensus has beenreached as to which elements or combinations of ele-ments are likely to promote discontinuous phase trans-formations. While it is acknowledged that aninhomogeneous, highly complex alloy, such as 740H, isnot an ideal system on which to perform a systematicanalysis of the effects of alloying additions on DC, it wassurmised that use of currently available advanced ana-lytical electronmicroscopy tools could shed some light onthe role of various alloying elements in discontinuoustransformations within an engineering alloy.The three specimens investigated using STEM in this

study are highlighted in Table III. One sample wastested in the direct-aged condition and one sample washomogenized at 1373 K (1100 �C) and aged beforetesting. Both of these specimens developed around 40

1734—VOLUME 48A, APRIL 2017 METALLURGICAL AND MATERIALS TRANSACTIONS A

pct grain boundary CZ coverage during testing, andthey had similar rupture/interruption times. Thus,comparisons of the chemical composition of CZsformed from a starting homogeneous and heterogeneousmicrostructure could be made. In addition to these twospecimens, an as-aged alloy 740H sample was analyzedto investigate the concentration profiles within the DPcolonies that were observed to form during direct agingof the weld.

Figure 6 shows an SEM image of the region ofinterest in the specimen that was direct aged afterwelding and crept for 318 hours, along with a low-mag-nification bright-field STEM image of the resultantcross-sectional sample. As shown, the specimen containsa CZ with a few large c¢, a two-phase grain with adispersion of small c¢, and the grain boundary (DCreaction front) that divides the two regions. The regionsof interest denoted in the STEM image show the

Fig. 6—(Left) Scanning electron micrograph of the CZ of interest in an alloy 740H weld metal creep specimen that was given a direct age afterwelding and crept at 1123 K (850 �C) for 318 h with a 100 MPa applied stress. (Right) Bright-field TEM micrograph of the specimen extractedfrom this region of interest. The locations of XEDS acquisitions are indicated.

Fig. 7—(a) Nb profile from the XEDS line scan across the CZ reaction front indicated in Fig. 6. (b) Higher magnification bright-field STEMmicrograph of the region of interest, showing the grain boundary width at the location of the scan. The dashed line is the c–c¢ interface in theCZ.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 48A, APRIL 2017—1735

locations of XEDS analysis. Included are (a) a line scanacross the reaction front and (b) through (e) area scansof the c and c¢ in both grains. As shown in the STEMimage, the grain boundary exhibits significant twistingalong its length. As a result, tilting of the specimen wasnecessary to achieve minimum interface width for theline scan by bringing the investigated interface as closeas possible to parallel with the electron beam. Reportederror values/bars are for a 95 pct confidence interval.The validity of the data was assessed by calculating theminimum detectability limit (minimum mass fraction(MMF)) for the elements of interest. MMF values wereon the order of 0.20 ± 0.10 wt pct for elements otherthan Nb and on the order of 0.30 ± 0.20 wt pct for Nbbecause of the lower peak to background ratio for theNb Ka peak at 16.6 keV. Figure 7(a) shows the XEDSdata for Nb from the cross-boundary line scan in thisspecimen, and Table IV presents the area scan data.

The line scan appears to indicate a slight enrichment ofNb at the boundary in this specimen. This enrichmentmanifests over approximately 5 nm, which correspondswell to the measured width of the boundary at the pointthe scan was taken (4.5 nm, as shown in the highermagnification bright-field image of the boundary inFigure 7(b)). This was as narrow as the boundary couldbe made within the tilting limits of the microscope. Noneof the other alloying elements demonstrated a significantchange at the boundary except for Ni, which was slightlydepleted to accommodate the extra Nb. The results givenin Table IV do not indicate a significant difference in thematrix or c¢ compositions between the CZ and bottomgrain, with the exception of Si, which appears to be moreconcentrated in the CZ c and c¢. The measured

composition of c¢ in theCZ is approximatelyNi3.02(Al0.47-Ti0.37Nb0.15), and the measured composition of c¢ in thebottom grain is approximately Ni3.07(Al0.46Ti0.38Nb0.16).These Al:Ti:Nb ratios differ from that observed by Evanset al. in alloy 740.[31] This is not unexpected, however, assignificant changes to the nominal alloy compositionwere made to produce the alloy 740H variant.[15] Bycomparing the c¢ compositions to the neighboring matrixcompositions in both regions, the partition coefficients ofthe c¢ formers into c¢ can be calculated. These coefficientsare tabulated in Table V and indicate that in both the CZand bottom grain, Ti partitions most strongly to theprecipitates, followed by Nb and then Al. Additionally,Ti is shown to partition slightly more to the c¢ within thebottom grain than the CZ.The measured values were also compared to thermo-

dynamic and kinetic calculations. First, the as-solidifiedweld metal concentration profiles were calculated usingthe Scheil solidification module within Thermo-Calc(McMurray, PA), in combination with the TTNi7database.[32,33] Details about the inputs to this simula-tion and the resulting austenite concentration profilesare provided elsewhere.[15] The austenite compositionprofiles, which spanned 10 lm and represent onehalf-dendrite, were then imported into the DICTRAsoftware package, and the TTNi7 thermodynamicdatabase and MOB2 kinetic database[32–34] were usedto simulate a 1073 K (800 �C) for 4 hours direct agetreatment followed by 1123 K (850 �C) for 318 hours.The resulting austenite and c¢ concentration profiles forNi, Al, Ti, and Nb are shown in Figure 8 and summa-rized in Table VI. After several hundred hours at 1123K (850 �C), the compositions of austenite and c¢ are

Table IV. XEDS Data for Area Scans (300-Second Dwell) in Alloy 740H Weld Specimens That Were Crept at 1123 K (850 �C)with 100 MPa Applied Stress; Deviations are the 95 Pct Confidence Interval

Postweld HeatTreatment Location Ni Cr Co Al Ti Nb Fe Si

Direct age* c¢ in CZ 68 ± 1 1.9 ± 0.1 6.5 ± 0.2 11 ± 1 8.4 ± 0.3 3.4 ± 0.1 0.39 ± 0.02 0.75 ± 0.03c in CZ 45 ± 1 28 ± 1 21 ± 1 2.3 ± 0.1 0.99 ± 0.03 0.57 ± 0.03 0.59 ± 0.02 1.2 ± 0.1c¢ in bottom grain 68 ± 1 2.1 ± 0.1 6.6 ± 0.2 10 ± 1 8.5 ± 0.3 3.6 ± 0.1 0.40 ± 0.02 0.16 ± 0.01c in bottom grain 46 ± 1 29 ± 1 21 ± 1 2.2 ± 0.1 0.97 ± 0.03 0.63 ± 0.03 0.61 ± 0.02 0.33 ± 0.02

Homogenizationand age**

c¢ in CZ 68 ± 1 2.1 ± 0.1 6.6 ± 0.2 10 ± 1 8.6 ± 0.3 3.3 ± 0.1 0.40 ± 0.02 0.31 ± 0.02c in CZ 46 ± 1 29 ± 1 20 ± 1 2.2 ± 0.1 1.1 ± 0.1 0.62 ± 0.03 0.61 ± 0.02 0.48 ± 0.02c¢ in left grain 68 ± 1 2.2 ± 0.1 6.7 ± 0.2 11 ± 1 8.6 ± 0.3 3.1 ± 0.1 0.43 ± 0.02 0.28 ± 0.02c in left grain 46 ± 1 28 ± 1 20 ± 1 2.9 ± 0.1 1.0 ± 0.1 0.57 ± 0.03 0.57 ± 0.02 0.86 ± 0.04

Values given in atomic percent.* Direct age: 1073 K (800 �C), 4 h, and air cool.** Homogenization and age: 1373 K (1100 �C), 4 h, water quench, 1073 K (800 �C), 4 h, and air cool.

Table V. Partition Coefficients for the c¢ Forming Elements, Calculated from the XEDS Measurements Given in Tables IV and VI

Postweld Heat Treatment Location Ni Al Ti Nb

Direct age* CZ 1.51 ± 0.03 4.78 ± 0.10 8.48 ± 0.05 5.96 ± 0.06bottom grain 1.48 ± 0.03 4.55 ± 0.11 8.76 ± 0.05 5.71 ± 0.06

Homogenization and age** CZ 1.48 ± 0.03 4.55 ± 0.11 7.82 ± 0.10 5.32 ± 0.04left grain 1.48 ± 0.03 3.79 ± 0.10 8.60 ± 0.11 5.44 ± 0.06

* Direct age: 1073 K (800 �C), 4 h, and air cool.** Homogenization and age: 1373 K (1100 �C), 4 h, water quench, 1073 K (800 �C), 4 h, and air cool.

1736—VOLUME 48A, APRIL 2017 METALLURGICAL AND MATERIALS TRANSACTIONS A

predicted to be nearly uniform throughout the weldmetal microstructure, in agreement with the experimen-tal data. Note that this solute redistribution is onlypredicted to eliminate the concentration gradients in thesystem, not the c¢ phase fraction gradients, which clearlypersist, as shown in Figures 5(b) and (c). Comparing themeasured and calculated compositions (Tables IV andVI) reveals that the solid-state solute redistributionpredicted by the modeling has occurred during creep. Inaddition, Thermo-Calc was used to calculate the equi-librium compositions of c and c¢ for the nominal alloy740H composition set at 1123 K (850 �C). These values,which are given in Table VI, are also very close to theexperimental and DICTRA-calculated values, indicatingthat the solid-state solute redistribution that hasoccurred during creep has nearly reached equilibrium.It is recognized that the kinetics of this solute redistri-bution will be significantly slower within the plannedoperating temperature range of the Advanced Ultrasu-percritical (A-USC) coal-fired power plants where alloy740H will find significant use [973 K to 1033 K (700 �Cto 760 �C)]. Therefore, the same DICTRA calculationwas performed at these lower temperatures. Figure 9

presents the calculated time evolution of the concentra-tion of Nb in c and c¢ at 1033 K (760 �C) and 973 K (700�C). The concentration of Nb in c¢ was chosen as thebasis for comparison, because it is the slowest diffusingc¢ forming element. As expected, less homogenization ofthe matrix and precipitate compositions occurs at thelower temperature, but significant reductions in thestarting composition gradients are predicted for bothtemperatures. This indicates that the type of soluteredistribution observed experimentally in acceleratedcreep tests at 1123 K (850 �C) is likely to also occurwithin the A-USC temperature range for times relevantto currently completed creep tests of alloy 740H fusionwelds (on the order of 10,000 hours)[1,10] and the designlife of A-USC boiler components (100,000 hours).[35]

Finally, the significantly higher dislocation density inthe bottom grain of Figure 6 should be noted. Whilemany of these dislocations are associated with the c¢precipitates and are likely interfacial dislocations, anetwork of dislocations through the matrix in the bottomgrain is also present. This is in contrast to the CZ, wheresome c¢ interfacial dislocations are present but there arevery few dislocations in the c¢ matrix. This observation

Fig. 8—DICTRA-calculated composition profiles for (a) c¢ and (b) c in an alloy 740H fusion weld after nonequilibrium solidification, aging at1073 K (800 �C) for 4 h, and isothermal exposure during creep at 1123 K (850 �C) for 318 h.

Table VI. Calculated Compositions for c and c¢ in Alloy 740H Weld Specimens That Were Crept at 1123 K (850 �C)

Postweld HeatTreatment Composition Type Ni Cr Co Al Ti Nb Fe Si

Direct age* DICTRA, c¢ 67 ± 1 1.9 ± 0.5 7.6 ± 0.2 11 ± 0 9.0 ± 0.6 3.5 ± 0.7 0.03 ± 0 0.10 ± 0DICTRA, c 46 ± 1 30 ± 1 21 ± 1 1.7 ± 0.1 0.45 ± 0.04 0.18 ± 0.05 0.27 ± 0 0.29 ± 0

Homogenizationand age**

DICTRA, c¢ 67 ± 1 1.9 ± 0 7.6 ± 0.1 11 ± 1 8.9 ± 0.2 3.7 ± 0 0.03 ± 0 0.10 ± 0DICTRA, c 46 ± 1 30 ± 1 21 ± 1 1.7 ± 0.1 0.44 ± 0 0.19 ± 0 0.27 ± 0 0.29 ± 0

N/A Thermo-Calc, c¢ 67 1.9 7.6 11 8.9 4.0 0.03 0.10Thermo-Calc, c 45 30 21 1.6 0.43 0.22 0.27 0.31

DICTRA-Calculated Values Are the Average of the 75 Grid Points in the Modeled System, and Deviations Are the 95 Pct Confidence Interval;Thermo-Calc–Calculated Values Are Single Point Equilibria of Nominal Alloy 740H Composition at 1123 K (850 �C) (all values are in atomicpercent).

*Direct age: 1073 K (800 �C), 4 h, and air cool.**Homogenization and age: 1373 K (1100 �C), 4 h, water quench, 1073 K (800 �C), 4 h, and air cool.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 48A, APRIL 2017—1737

lends credence to the previous statements regarding CZformation by DC as a mechanism for local recoveryprocess for grain boundary strain accumulation.

An SEM image and accompanying low-magnificationbright-field STEM image of the region of interest andextracted cross-sectional TEM specimen from the creepsample that was homogenized at 1373 K (1100 �C) for 4hours and then aged before being crept to rupture (326hours) are given in Figure 10. The grain boundary/CZreaction front in this specimen runs from the top left tothe bottom right of the STEM image, with the CZoccupying the region above the reaction front and thegrain with fine c¢ occupying the lower region. It has beenhighlighted in this figure for clarity. The regions ofinterest for EDS analysis are again indicated and includethe same types of measurements discussed for theprevious specimen. Figure 11 presents the EDS datafor Ti and Nb from a line scan taken across the CZreaction front. Much more enrichment of Nb at the

grain boundary is observed in this sample, as is anapparent enrichment of Ti at the boundary. Figure 11also shows the location of this scan at a highermagnification and demonstrates that the apparent widthof the boundary for this scan was significantly smallerthan in the preceding specimen. This ultimately leads tothe sharper, more well-defined peaks in the EDS data.As in the case of the previous sample, no other elementswere found to vary significantly at the boundary, exceptfor Ni. Table IV shows the area scan data for the matrixand c¢ within the CZ and left grain in this specimen.Again, the compositions of the austenite and c¢ arenearly identical between the CZ and left grain. Themeasured composition of c¢ in the CZ is approximatelyNi3.11(Al0.46Ti0.39Nb0.15), and the measured compositionof c¢ in the bottom grain is approximately Ni3.00(Al0.48-Ti0.38Nb0.13). The partitioning behavior of the c¢ formersin this specimen is given in Table V and indicates that Tialso most strongly partitions to c¢ in this specimen.

Fig. 9—DICTRA-calculated composition profiles for Nb in c¢ and c in an alloy 740H fusion weld after nonequilibrium solidification, aging at1073 K (800 �C) for 4 h, and isothermal exposure during creep. (a) and (b) Simulated creep exposure at 1033 K (760 �C) and (c) and (d) 973 K(700 �C).

1738—VOLUME 48A, APRIL 2017 METALLURGICAL AND MATERIALS TRANSACTIONS A

However, partitioning of the c¢ to the precipitates in thissample, in general, is smaller in magnitude than in thepreviously discussed specimen.

Thermodynamic and kinetic calculations for thehomogenized specimen began with the same as-solidifiedweld metal concentration profiles discussed previously.After importing them into DICTRA, the TTNi7 andMOB2 databases were used to simulate a 1373 K (1100�C) for 4 hours homogenization treatment, a 1073 K (800�C) for 4 hours aging treatment, and a 1123 K (850 �C)for 326 hours isothermal hold. The resulting austeniteand c¢ concentration profiles are shown in Figure 12 andsummarized in Table VI. Once again, the compositionsof austenite and c¢ are predicted to be uniform through-out the weld metal microstructure and close to thepredicted equilibrium compositions. The experimentalmeasurements given in Table IV confirm the accuracy ofthese calculations. It is important to again recognize thateven though the matrix and c¢ compositions for both thedirect-aged and homogenized weld metal substructuresappear to have reached their equilibrium values duringcreep, the final microstructures of these samples are notthe same. Postweld homogenization will eliminate themicrosegregation present immediately after solidifica-tion, thereby resulting in a uniform c¢ size and phasefraction throughout the microstructure before creep. Asdescribed previously, the c¢ size and phase fractiongradients that arise from the microsegregation willpersist in the direct-aged sample even after equilibriumphase compositions have been reached during creep. Theimportance of these microstructural gradients to DC andcreep rupture life of alloy 740H fusion welds has beendiscussed previously and in earlier work.[2]

Figure 13 shows an SEM image and the resulting TEMspecimen extracted from a colony of discontinuous

precipitation in as-aged alloy 740H weld metal. Thegrain boundary/CZ reaction front is along the left side ofthe DP colony in the STEM image, which runs from thetop to bottom in the center of the sample. The region ofinterest for an EDS line scan is indicated. The measuredconcentration of Nb along this line scan is given inFigure 14 and shows, once again, enrichment ofNb at theinterface. In this specimen, the boundary width where theline scan was acquired was 2 nm, as indicated inFigure 14(b). Besides slight Ni depletion at the boundary,no other elements varied significantly. In addition,discontinuous concentration profiles across the bound-ary, which are generally considered to be a distinguishingfeature of DP,[19,29] were not detected. It is possible thatsuch discontinuous composition changes may havehomogenized during the 4-hour hold at 1073 K (800�C). Evidence of this has been observed inNi-Cu, whereina discontinuous concentration profile left in the wake of adiscontinuous phase transformation was shown to relaxduring thermal treatment at 1023 K (750 �C) for severalhours.[36] Prior work on this alloy system has alsopresented modeling results that suggest that relaxationof the discontinuous concentration profiles during ther-mal exposure at 1073 K (800 �C) is plausible.[1]The presence of Nb segregation to the DP/DC

reaction front in each of these samples implies threepossibilities: (1) it promotes the discontinuous reactions(e.g., by increasing the grain boundary vacancy concen-tration and enhancing mobility[37]), (2) it retards thediscontinuous reactions (e.g., by solute drag that reducesboundary mobility), or (3) it has no effect on thepropensity for discontinuous reactions but is therate-limiting species that must diffuse along the bound-ary because of its large atomic radius. Some studies haveindicated that the addition of Nb (or, more generally,

Fig. 10—(Left) Scanning electron micrograph of the CZ of interest in an alloy 740H weld metal creep specimen that was homogenized at 1373 K(1100 �C) for 4 h, aged at 1073 K (800 �C) for 4 h, and crept at 1123 K (850 �C) for 326 h with a 100 MPa applied stress. (Right) Bright-fieldTEM micrograph of the specimen extracted from this region of interest. The locations of XEDS acquisitions are indicated.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 48A, APRIL 2017—1739

any strong boundary segregant) slows discontinuousreactions via solute drag on the boundary, lowering thegrain boundary energy, lowering the grain boundarymobility, increasing the activation energy for boundarymigration, increasing the activation energy for boundarydiffusion, or occupying the precipitate nucleationsites.[19,38,39] However, these studies have focused onternary systems or engineering alloys that did not havecontinuous precipitation that involves Nb, do not havethe types of microstructural gradients discussed previ-ously, or were not deformed in creep. In addition, otherstudies on ternary systems have noted an increase in thepropensity for discontinuous transformations when astrong boundary segregant (e.g., Ga in Ag-Cu and Al inCu-Be) is introduced. Williams and Butler[19] acknowl-edge that the consistent documentation of segregantsthat increase the propensity for discontinuous reactionsrefutes the argument that such additions would unilat-erally decrease the occurrence of DP/DC, unless anadditional variable that has not been rigorouslyaddressed is affecting the alloys’ response. It is, therefore,

unclear how well the conclusions from these studies canbe relied upon/extended to this system.Alternatively, some insight may be gained from com-

parisons of the microstructural stability of alloy 740Hwith other A-USC alloys. Prior work has assessed thepropensity for grain boundary c¢-denuded zone forma-tion in several alloys that are targeted for use in A-USCapplications and concluded that alloys NIMONIC 263and INCONEL 740 form grain boundary denuded zones,while HAYNES 282� does not. Neither alloy 263 nor 740

was shown to form denuded zones via discontinuousphase transformations. Instead, they formed via stabi-lization of phases that precipitated at the expense of thelocal c¢.[10] Typical compositions of alloys 263, 282, and740 are given in Table VII. By considering the observa-tions made here and in Reference 10, it can be noted that

Fig. 11—(a) Higher magnification bright-field STEM micrograph of the region of interest, showing the grain boundary width at the location ofthe scan; (b) Ti profile from the XEDS line scan across the CZ reaction front indicated in Fig. 10; and (c) Nb profile from the XEDS line scanacross the CZ reaction front indicated in Fig. 10.

� HAYNES and 282 are registered trademarks of Haynes Interna-tional.

1740—VOLUME 48A, APRIL 2017 METALLURGICAL AND MATERIALS TRANSACTIONS A

the only A-USC alloy composition sets that are prone toDP and DC are those that contain Nb and have a highenough Al:Ti ratio to suppress g and G-phase formation.It should also be noted that the DP/DC-prone alloys alsocontain significantly less Mo than those that did notexhibit discontinuous transformations. From the mech-anisms described previously (solute-vacancy interac-tion/solute drag) and the observed propensity for Mo tosegregate to the grain boundaries in high-Ni alloys,[40] it is

possible that Mo may also affect DP/DC in the alloy740H system, but no specific investigation into Mo wasperformed in this work. While the information gleanedfrom these comparisons may serve to narrow the list ofalloying elements that contribute significantly to the DP/DC susceptibility of alloy 740H, unfortunately, there isnot enough documented information to firmly state thespecific influence of these elements or the effect of Nbsegregation at this time.

Fig. 12—DICTRA-calculated composition profiles for (a) c¢ and (b) c in an alloy 740H fusion weld after nonequilibrium solidification, homoge-nization at 1373 K (1100 �C) for 4 h, aging at 1073 K (800 �C) for 4 h, and isothermal exposure during creep at 1123 K (850 �C) for 326 h.

Fig. 13—(Left) Scanning electron micrograph of the CZ of interest in an alloy 740H weld metal creep specimen that was aged at 1073 K (800�C) for 4 h. (Right) Bright-field TEM micrograph of the specimen extracted from this region of interest. The locations of XEDS acquisitions areindicated.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 48A, APRIL 2017—1741

IV. CONCLUSIONS

In-depth microstructural analysis and quantificationof c¢ CZs in alloy 740H fusion weld creep specimenswere conducted. The results of this work show thefollowing.

1. The local variations in composition along theas-solidified weld metal grain boundaries lead tomicrostructural heterogeneities that have significanteffects on the evolution of CZs in the weld metal.These variations greatly enhance the general sus-ceptibility of the weld metal to CZ formation ascompared to wrought alloy 740H.

2. CZ colonies on boundary segments where a den-drite core and an interdendritic region meet growpreferentially into the dendrite core side of theboundary. It is postulated that this is a result ofincreased recovery of strain accumulation andenhanced grain boundary mobility due to reducedparticle pinning in that direction. Other possibledriving forces for this apparent preference, with theexception of local configurational grain boundaryenergy, have been ruled out.

3. Preference for CZ colony growth along IC bound-ary segments was demonstrated for a sample size of

20 specimens encompassing nearly 10 cm of ana-lyzed boundaries. This preference was also a resultof enhanced creep strain localization along thesesegments because of the sharp change in precipitatephase fraction.

4. Enrichment of Nb at a CZ reaction front wasobserved in a specimen that was given a direct agebefore creep and in a specimen that was homogenizedand then aged before creep.Niobium enrichment wasalso detected at the discontinuous precipitationreaction front in the direct-aged alloy 740H weldmetal. Enrichment of Ti at the DC reaction front inthe homogenized specimen was also observed.

5. Measured matrix and precipitate compositions inthe crept specimens correlated well with predictionsfrom kinetic calculations and equilibrium calcula-tions for austenite and c¢ using the nominal alloy740H composition set.

6. Titanium most strongly partitioned to the c¢ pre-cipitates in crept specimens, followed by Nb andthen Al.

7. The specific influence, if any, of reaction front Nbsegregation on DP/DC in alloy 740H could not beascertained from the work performed here orcomparisons to prior work on other A-USC alloys.

Fig. 14—(a) Nb profile from the XEDS line scan across the CZ reaction front indicated in Fig. 13. (b) Higher magnification bright-field STEMmicrograph of the region of interest, showing the grain boundary width at the location of the scan.

Table VII. Typical As-Deposited Weld Metal Compositions of A-USC Candidate Alloys

Alloy Ni Cr Co Al Ti Nb Fe Mn Mo Si C S B P

NIMONIC 263 bal 21 20 0.52 2.2 0 0.37 0.21 5.9 0.08 0.06 <0.002 <0.002 <0.005INCONEL 740 bal 24 20 0.98 1.8 2.0 0.46 0.26 0.51 0.51 0.03 <0.001 0.004 <0.005HAYNES 282 bal 19 10 1.5 2.2 0 0.68 0.05 8.6 <0.05 0.06 <0.002 0.005 0.002

Values are given in weight percent.

1742—VOLUME 48A, APRIL 2017 METALLURGICAL AND MATERIALS TRANSACTIONS A

ACKNOWLEDGMENTS

The authors gratefully acknowledge the financial sup-port of the NSF I/UCRC Manufacturing and MaterialsJoining Innovation Center (Ma2JIC) under ContractNo. IIP-1034703. They also acknowledge the financialsupport provided by Special Metals Corporation andThermo-Calc software. Special thanks for the technicaldiscussions are given to John Shingledecker and JohnSiefert, Electric Power Research Institute (Charlotte,NC) and Paul Mason and Kevin Wu (Thermo-Calc).

REFERENCES1. D.H. Bechetti, J.N. DuPont, J.J. deBarbadillo, B.A. Baker, and

M. Watanabe: Metall. Mater. Trans. A, 2015, vol. 46A,pp. 739–55.

2. D.H. Bechetti, J.N. DuPont, and J.J. deBarbadillo: LehighUniversity, Bethlehem, PA, unpublished research, 2016.

3. Y. Brechet and C. Hutchinson: in Solid State Physics: Advances inResearch in Applications, vol. 60, H. Ehrenreich and F. Spaepen,eds., Elsevier, Oxford, United Kingdom, 2006, pp. 230–31.

4. J.D. Livingston and J.W. Cahn: Acta Metall., 1974, vol. 22,pp. 495–503.

5. R.A. Fournelle: Acta Metall., 1979, vol. 27, pp. 1147–55.6. S. Mitao and L.A. Bendersky: Acta Mater., 1997, vol. 45,

pp. 4475–89.7. R. Monzen, H. Shigehara, and K. Kita: J. Mater. Sci., 2000,

vol. 35, pp. 5839–43.8. H. Azzeddine, S. Abdessameud, B. Alili, Z. Boumerzoug, and D.

Bradai: Bull. Mater. Sci., 2011, vol. 34, pp. 1471–76.9. W.E. Voice and R.G. Faulkner: J. Mater. Sci., 1987, vol. 22,

pp. 4221–32.10. D.H. Bechetti, J.N. DuPont, S.A. Siefert, and J.P. Shingledecker:

Lehigh University, Bethlehem, PA, unpublished research, 2016.11. J.T. Armstrong: Proc. 23rd Ann. Conf. of the Microbeam Analysis

Society, Milwaukee, WI, Aug. 8–12, 1988; Microbeam Analysis,San Francisco Press, San Francisco, CA, 1988, pp. 239–46.

12. M. Watanabe and D.B. Williams: J. Microsc., 2006, vol. 221,pp. 89–109.

13. J.N. DuPont, J.C. Lippold, and S.D. Kiser: Welding Metallurgyand Weldability of Nickel-Base Alloys, 1st ed., Wiley, Hoboken,NJ, 2011.

14. D.A. Porter and K.E. Easterling: Phase Transformations in Metalsand Alloys, 2nd ed., Taylor and Francis Group, Boca Raton, FL,1992.

15. D.H. Bechetti, J.N. Dupont, J.J. deBarbadillo, and B.A. Baker:Metall. Mater. Trans. A, 2014, vol. 45A, pp. 3051–63.

16. S. Kou: Welding Metallurgy, 2nd ed., Wiley, Hoboken, NJ, 2003.17. M. Hillert: Scripta Metall., 1983, vol. 17, pp. 237–40.18. M. Hillert and R. Lagneborg: J. Mater. Sci., 1971, vol. 6,

pp. 208–12.19. D.B. Williams and E.P. Butler: Int. Mater. Rev., 1981, vol. 26,

pp. 153–83.20. A.W. Funkenbusch:Metall. Trans. A, 1983, vol. 14A, pp. 1283–92.21. S. Hirth and G. Gottstein: Acta Mater., 1998, vol. 46, pp. 3975–84.22. A.W. Funkenbusch, T.A. Stephenson, G. McCarthy, and H.

Fraser: Metall. Mater. Trans. A, 1985, vol. 16A, pp. 11–16.23. A. Tavassoli and G. Colombe: Metall. Trans. A, 1978, vol. 9A,

pp. 1203–11.24. J. Petermann and E. Hornbogen: Z. Met., 1968, vol. 59,

pp. 814–22.25. D. Turnbull: Acta Metall., 1955, vol. 3, pp. 55–63.26. K. Tu and D. Turnbull: Acta Metall., 1967, vol. 15, pp. 369–76.27. J.W. Cahn: Acta Metall., 1959, vol. 7, pp. 18–28.28. D.B. Williams and J.W. Edington: Acta Metall., 1976, vol. 24,

pp. 323–32.29. I. Manna, S.K. Pabi, and W. Gust: Int. Mater. Rev., 2001, vol. 46,

pp. 53–91.30. J.D. Nystrom, T.M. Pollock, W.H. Murphy, and A. Garg: Metall.

Mater. Trans. A, 1997, vol. 28A, pp. 2443–52.31. N.D. Evans, P.J. Maziasz, R.W. Swindeman, and G.D. Smith:

Scripta Mater., 2004, vol. 51, pp. 503–07.32. J.O. Andersson, T. Helander, L. Hoglund, P.F. Shi, and B.

Sundman: Calphad, 2002, vol. 26, pp. 273–312.33. ThermoTech TTNi7 Ni-Based Superalloys Thermodynamic Data-

base Version 7, ThermoTech, Ltd., Guildford, United Kingdom,2006.

34. Thermo-Calc Software MOB2 TCS Alloy Mobility Database,Thermo-Calc Software AB, Stockholm, Sweden.

35. G. Smith and L. Shoemaker: Adv. Mater. Process., 2004, vol. 162,pp. 23–26.

36. C.Y. Ma, E. Rabkin, W. Gust, and S.E. Hsu: Acta Metall. Mater.,1995, vol. 43, pp. 3113–24.

37. W.W. Mullins: Acta Metall., 1956, vol. 4, pp. 421–32.38. E. Hornbogen: Metall. Trans., 1972, vol. 3, pp. 2717–27.39. R.D. Knutsen, C.I. Lang, and J.A. Basson: Acta Mater., 2004,

vol. 52, pp. 2407–17.40. D.H. Ping, Y.F. Gu, C.Y. Cui, and H. Harada: Mater. Sci. Eng.

A, 2007, vol. 456, pp. 99–102.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 48A, APRIL 2017—1743