The Effects of Chemistry Variations in NewNickel-Based Superalloys for Industrial Gas TurbineApplicationsSABIN SULZER , MAGNUS HASSELQVIST, HIDEYUKI MURAKAMI,PAUL BAGOT, MICHAEL MOODY, and ROGER REEDIndustrial gas turbines (IGT) require novel single-crystal superalloys with demonstrablysuperior corrosion resistance to those used for aerospace applications and thus higher Crcontents. Multi-scale modeling approaches are aiding in the design of new alloy grades;however, the CALPHAD databases on which these rely remain unproven in this compositionregime. A set of trial nickel-based superalloys for IGT blades is investigated, with carefullydesigned chemistries which isolate the influence of individual additions. Results from anextensive experimental characterization campaign are compared with CALPHAD predictions.Insights gained from this study are used to derive guidelines for optimized gas turbine alloydesign and to gauge the reliability of the CALPHAD -7Ó The Author(s) 2020I.INTRODUCTIONNICKEL-BASED superalloys have emerged as thematerials of choice for high-temperature industrial gasturbine (IGT) applications due to their unique combination of resistance to loading under static, fatigue, andcreep conditions, as well as to environmental degradation.[1] Materials tailored to particular gas turbinespecifications are required, as a simple adaptation ormodification of aeroengine alloys cannot yield optimalresults due to distinct design factors including weightrestriction, operating time, fuel quality, and cyclicduty.[2]In Figure 1, a proposed design space for new,improved IGT alloys is compared with existing commercial compositions. The latter tend to either possess(1) a high Al/Cr ratio and exhibit good creep properties,but weak corrosion resistance, having been developedfor aerospace applications, or (2) a low Al/Cr ratio,SABIN SULZER is with the Hilti Corporation, Feldkircherstraße100, 9494, Schaan, Liechtenstein and also with the Department ofMaterials, Univerity of Oxford, Parks Road, Oxford, OX1 3PH, UK.Contact e-mail: [email protected] MAGNUS HASSELQVISTis with the Turbine Group, Siemens Industrial Turbomachinery AB,61283 Finspång, Sweden. HIDEYUKI MURAKAMI is with theResearch Centre for Structural Materials, National Institute forMaterials Science, 1-2-1 Sengen, Tsukuba, 305-0047, Japan and alsowith the Department of Nanoscience and Nanoengineering, WasedaUniversity, 3-4-1 Okubo, Shinjuku-ku, Tokyo, 169-8555, Japan.PAUL BAGOT, MICHAEL MOODY and ROGER REED arewith the Department of Materials, University of Oxford.Manuscript submitted March 8, 2020.Article published online June 22, 20204902—VOLUME 51A, SEPTEMBER 2020resulting in IGT environment suitability, but at the costof lower strength. ‘Classical’ IGT alloys like IN-792[3]and Mar-M247[4] were developed more than 45 yearsago when turbine inlet temperature levels were muchlower. As such, only few compositions satisfy bothrequirements and populate the envisioned area above13.2 at. pct Cr and 8.5 at. pct Al.A recent foray into this space was made with thedevelopment of the single-crystal alloy eSTAL15CC.[8,9] These alloys contain 17.8 at. pct Crfor high hot corrosion resistance and also provideexcellent oxidation resistance due to their ability to forma continuous Al2O3 surface layer at high temperatures.This enables both high turbine inlet temperatures andhigh fuel flexibility, while minimizing the risk ofcomponent failure should protective thermal barriercoatings crack. STAL15SX possesses sufficient creepstrength for use in IGT first-stage blades, whileSTAL15CC is aimed at hot stator components.There is an obvious possibility to design furtherderivatives at lower Cr and higher Al contents. Thiswould increase creep and oxidation resistance, whileretaining Cr levels above the commonplace threshold of13.5 at. pct for sufficient hot corrosion resistance.[10]Within this scope, multi-scale modeling approacheshave been successfully developed and applied to designtailored nickel-based superalloys for turbine blade[11–13]and turbine disk applications.[13,14] However, an important caveat is that several key input parameters arederived from thermodynamic CALPHAD methods,which in turn depend on commercial parameterMETALLURGICAL AND MATERIALS TRANSACTIONS A

the design point of 14.2 at. pct Cr and 12 at. pct Al.Four compositions were designed to assess the influenceof adding 0.3 at. pct Hf (SX2), 5.2 at. pct Fe (SX3), and1 at. pct Si (SX6) and of increasing the Co content from5 to 10 at. pct (SX5). Finally, a composition with lowerAl and Ta levels and containing 2.9 at. pct Ti (SX4) wasalso included as a ‘combination’ between the base alloyand IN-792.B. Differential Scanning CalorimetryFig. 1—Comparison of the proposed design space for new IGTsuperalloys (green rounded rectangle) with the general compositionalspace occupied by Re-containing aerospace alloys (red roundedrectangle). This dataset was compiled from literature published asjournal articles or patents; therefore, slight differences to versionssold commercially may occur. Alloys containing more than 9 at. pctCr are labeled accordingly (Color figure online).databases. The reliability and validity of modelingresults are dictated by the accuracy of these databasesand any significant errors will give rise to false conclusions regarding optimal alloy compositions. This issue isespecially acute for new regions of the alloy designspace, such as the one considered here, since experimental data will be scarce or missing altogether. Thus, itis imperative to improve and expand the performance ofCALPHAD tools and to gain a better understanding ofthe influence of chemistry on thermochemical andphysical alloy properties and on microstructure.The present work aims to remedy this situation byinvestigating a set of trial IGT alloys with carefullydesigned chemistries, which isolate the influence ofindividual additions. Results from an extensive experimental characterization campaign are compared withCALPHAD predictions using the Thermo-Calc software package. Insights gained from this study are usedto derive guidelines for optimized IGT alloy design andto improve the reliability of CALPHAD techniques.II.EXPERIMENTAL METHODA. MaterialA set of six trial alloys, denoted in the following asSX1 to SX6, was pre-alloyed either from pure elements(Al, Cr, Co, Fe, Ni, Si, Ti) or from binary Ni-X alloysfor refractory metals (Ni-Hf, Ni-Mo, Ni-Ta, Ni-W).Batches of 120 g were weighed, arc-melted under Arusing an Edmund Bühler AM 500 arc melter, andpoured into water-cooled copper molds. Nominalchemical compositions in at. pct are presented inTable I. The base composition of the investigated alloyseries, denoted as SX1, was derived from STAL15SX toMETALLURGICAL AND MATERIALS TRANSACTIONS ADifferential scanning calorimetry (DSC) testing wascarried out using a Netzsch DSC 404 F1 Pegasus.Measurements were carried out at heating/cooling ratesof 10 C/min under a high purity Ar atmosphere at apurge flow rate of 40 mL/min. PtRh crucibles with PtRhlids and thin-walled Al2O3 liners were used in combination with a high-accuracy sample carrier system withtype S thermocouples. Disk specimens of 3 mm diameterand 1 mm thickness were manufactured by wire-guidedelectro-discharge machining (EDM). Both circular surfaces were polished down to 4000 grit silicon carbidegrinding paper to achieve good thermal contact with thesensor. Each specimen was ultrasonically cleaned inethanol and weighed before the test.For each test, two heating/cooling cycles were performed between 700 C and 1450 C to assess theinfluence of prior melting/solidification and of reducedsegregation on the second cycle. To equalize thetemperature in the system, constant temperature segments of 15 minutes duration were added between rampsteps. Following the ASTM method, three measurements were carried out as part of each test: (1) a baselinemeasurement with an empty sample pan, (2) a calibration measurement with a synthetic sapphire disk ofknown mass and heat capacity, and (3) a samplemeasurement with the actual superalloy disk. Thereference pan remained empty during all three runs.Finally, changes in alloy heat capacity, Cp , withtemperature, T, were calculated using the NetzschProteus analysis program.C. Heat Treatment and Chemical AnalysisPreliminary DSC studies on as-received material gavean indication of the c0 solvus and solidus temperaturesfor each alloy. Using this information, solutioning heattreatments were carried out to reduce microsegregationbetween dendrite core and interdendritic regions, whileavoiding incipient melting. Samples were placed in apre-heated furnace at 1170 C. The temperature wasthen slowly ramped up at a constant rate of 0.1 C/minup to 20 C below the expected solidus, following themethod proposed by Hegde et al.[15] After reaching thepeak temperature, specimens were cooled down at10 C/min to 850 C and were then removed from thefurnace and air-cooled. Subsequent primary and secondary aging heat treatments were chosen at 1120 C for4 hours and 845 C for 24 hours, both completed byair-cooling.VOLUME 51A, SEPTEMBER 2020—4903

Table I.Chemical Compositions of the Alloys in Fully Heat-Treated ConditionChemical Composition (At. Pct)AlloyMethodNiSX1 63.563.862.959.258.858.664.466.264.8SX2 ( Hf)SX3 ( Fe)SX4 ( Ti)SX5 ( Co)SX6 ( 3Nominal values are compared with measurements from ICP-OES and EPMA analyses. Note that elements not expected in an alloy were omittedfor EPMA.The chemistries of fully heat-treated specimens werechecked at an independent laboratory using inductivelycoupled plasma optical emission spectrometry(ICP-OES) for metallic elements and combustion analysis for C and S impurities. To further assess theremaining degree of microsegregation, chemistries werealso checked on polished specimens by electron probemicroanalysis (EPMA) using a JEOL JXA 8800 Superprobe equipped with four wavelength-dispersive X-rayspectrometers (WDS) for parallel detection of elements.Line scans of 200 points were carried out at a spacing of10 lm, a probe diameter of 10 lm, an acceleratingvoltage of 25 keV, a beam current of 32 nA, and a dwelltime for each point of 100 ms. The diffracting crystalsused for WDS were LIF for Ni, Co, W, Ta, Hf, and Fe,PETJ for Cr, Mo, and Ti, and TAP for Al and Si. X-raycounts were converted to concentration values usingstandard correction procedures and signals from pureelement standards. Average compositions obtained fromICP-OES and EPMA are shown in Table I below thenominal values.D. Electron MicroscopyMetallographic preparation was carried out on aStruers Tegramin-25 system down to 1 lm diamondsuspension. An optimal finish was achieved by polishingfor 5 minutes with a 1 to 1 mixture of Struers OP-S 0.04lm colloidal silica suspension and water on a StruersMD-Chem polishing cloth. Specimens were cleaned inan ultrasonic bath with acetone and then with ethanol.Backscatter electron (BSE) images of fully heat-treatedmicrostructures were obtained using a Zeiss Merlin fieldemission gun (FEG) SEM operating at an acceleratingvoltage of 10 keV and a probe current of 500 pA. Theworking distance was set at 8.5 mm to yield optimal4904—VOLUME 51A, SEPTEMBER 2020energy-dispersive X-ray spectroscopy (EDX) resultsusing an Oxford Instruments X-MaxN silicon driftdetector with a size of 150 mm2. EDX data wereprocessed using the TruMap function of the OxfordInstruments AZtec 3.3 software package to eliminateartifacts, correct background and peak overlaps, andenhance compositional variations.E. Atom Probe TomographyMatchsticks of 0:5 0:5 2 mm3 were manufacturedby EDM from fully heat-treated material. Sharp tips foratom probe tomography (APT) with a radius below 100nm were produced in a two-stage electro-polishingprocess. The matchsticks were first coarse-polished ina solution of 10 vol pct perchloric acid in acetic acid at avoltage of 12 to 20 VDC and then fine-polished in asolution of 2 vol pct perchloric acid in 2-butoxyethanolat 8 to 12 VDC. APT data were collected using aCAMECA local electrode atom probe (LEAP) 5000 XRat a specimen base temperature of 50 K and an ambientgauge pressure of less than 10-10 Pa. Samples were run inlaser-assisted mode with an ultraviolet laser wavelengthof 355 nm, a pulse energy of 50 pJ, a pulse frequency of125 kHz, and a constant ion detection rate of 0.5 pct. 3Dreconstructions were generated using the CAMECAIVAS 3.8.2 software package. Peak overlaps of multipleelemental species were deconvolved using the standardIVAS decomposition parameters.F. CALPHAD PredictionsThroughout the present study, experimental results arecompared with predictions made using version 2019b ofthe Thermo-Calc software package and two associateddatabases for nickel-based superalloys—TCNi8, providedMETALLURGICAL AND MATERIALS TRANSACTIONS A

by Thermo-Calc Software AB, and TTNi8, provided byThermoTech Ltd. The chemical compositions measuredby ICP-OES were used as inputs. Phase diagrams weregenerated between 600 C and 1400 C. Equilibriumphase compositions were calculated at the final agingtemperature of 845 C. In both cases, a simplifiedthree-phase system consisting of c, c0 , and liquid at hightemperatures was assumed and all other phases weresuspended. SEM results support this assumption, as onlyinsignificant levels of small MC carbides and no topologically close-packed (TCP) phases were observed after heattreatment.III.RESULTSA. Alloy MicrostructureChemical compositions determined by ICP-OES andEPMA are in good overall agreement with nominalvalues. Trace levels of Fe were detected, but no otherimpurities are present. Furthermore, EPMA line scansshow little variation across the samples, implying goodhomogenization and low microsegregation after the heattreatment. Nonetheless, some discrepancies can benoted; the levels of Cr in SX2 and SX6, Co in SX6, Win SX2, and Ta in SX6 are somewhat lower thanintended, while the levels of Ta in SX2 and Fe in SX3are higher.The microstructures shown in Figure 2 are characteristic of IGT superalloys and are composed of cuboidal,secondary c0 precipitates with side lengths on the orderof 1 lm and spherical, tertiary c0 particles with diameterson the order of 10 nm, embedded in the c matrix.Secondary particles form after the homogenization stepduring cooling from around 1260 C, while tertiary onesprecipitate during the subsequent aging heat treatmentsat 1120 C and 845 C. Secondary precipitatemorphology varies between the alloys and appears morecuboidal for SX2, SX3, SX5, and SX6 and morerounded for SX1 and SX4. The substitution of Ta forTi in SX4 results in a marked reduction in contrastbetween c and c0 , as the heavy Ta makes precipitatesappear brighter in BSE micrographs.Particle size distributions were analyzed using theimage processing software package MIPAR. Cross-sectional area fractions comprising both secondary andtertiary c0 were measured for each BSE micrograph.These can be considered equal to c0 volume fractionsfrom a stereological standpoint as long as the precipitates do not have a preferred orientation.[16] Valuesobtained by averaging the results from at least fivemicrographs are shown in Table II, along with Thermo-Calc predictions for equilibrium conditions at845 C.EDX maps of elements added to the base compositionof SX1 are presented in Figure 3. Hf and Ti are found topartition preferentially to c0 , whereas Fe and Co areenriched in c. EDX results for Si are inconclusive, inpart due to peak overlaps with the Ta largely present inc0 , as shown for the SX1 reference.B. Phase Transition Temperatures1. Interpreting Cp -T curves from DSCCurves of heat capacity illustrated in Figure 4 revealseveral transitions with increasing temperature. Between700 C and 1100 C, there is a gradual increase in Cp dueto coarsening of the c0 phase.[17, 18] This is followed by agradual dissolution of c0 into the c matrix, whichculminates in a peak at around 1200 C. Once allprecipitates have dissolved, a single-phase plateauregime is observed up to about 1300 C, at which pointmelting initiates and gives rise to a strong endothermicpeak. The melting stage concludes at around 1350 CFig. 2—SEM micrographs of the trial IGT alloys in fully heat-treated condition. The illustrated scale bar applies to all images.METALLURGICAL AND MATERIALS TRANSACTIONS AVOLUME 51A, SEPTEMBER 2020—4905

and Cp values return to similar levels to the beginning ofthe test.Significant differences are observed between heating/cooling curves given the dynamic nature of the DSCtechnique. During heating, transitions take place athigher temperatures due to instrument lag. Conversely,data from cooling steps are significantly affected byundercooling effects which suppress transitions andcause them to occur more abruptly at lower temperatures.[19,20] Consequently, eutectic reactions, which produce a weaker signal, are only detected during cooling,with distinct peaks appearing at 1240 C for SX2 and at1255 C for SX4. Within the scope of comparingexperimental data with ideal CALPHAD predictions,undercooling effects can lead to significant discrepanciesof up to 20 C.[19,21] This is especially true for alloyslacking grain boundary elements like C or B, whichwould otherwise reduce undercooling by precipitatingcarbides or borides as nucleation points.[20] Finally,differences can also be noted between the first andsecond heating/cooling steps. Owing to homogenizationof the material during the first cycle, the magnitude ofendothermic/exothermic peaks during dissolution/precipitation decreases, as seen in Figure 5.Table II. Comparison of Average c0 Volume FractionsEstimated from SEM Micrographs with Thermo-CalcPredictionsc0 Volume Fraction 554.4 .757.859.551.062.656.551.0Interpreting Cp -T curves to extract transition temperatures remains a matter of debate given the sluggishnature of solid state transitions. Sponseller[19] recommends taking the c0 solvus as the ‘recovery point’ in aCp -T curve, at which the signal returns to the baselineduring heating or departs from it during cooling. Theargument for this choice is that the location of theendothermic/exothermic peak is not only influenced bythe actual location of the c0 solvus, but also bymicrosegregation between dendrite cores and interdendritic regions and by heating/cooling rate effects. However, Burton[21] argues that peak temperatures are moreindicative of the solvus, as the locations of ‘recoverypoints’ will be affected by instrument lag and by thefinite sample size. In the present work, this latterinterpretation is preferred and the c0 solvus is reportedas the observed endothermic peak of dissolution duringthe second heating. In like manner, following Questedet al.,[18] the liquidus temperature is reported as theendothermic melting peak. Finally, following Chapman[20] and Quested et al.,[18] the solidus temperature isdetermined by extrapolating the Cp -T baseline until itmeets the slope of the front of the endothermic meltingpeak during second heating. These temperatures wereevaluated using the built-in ‘peak’ and ‘onset’ functionsof the Netzsch Proteus analysis program.2. Effects of chemistry on the heat treatment windowExperimentally determined transition temperaturesare compared with Thermo-Calc predictions inTable III. Both databases give accurate estimates ofliquidus temperatures, with differences of 10 C or less.DSC results tend to lie between TTNi8 values, whichslightly underestimate the liquidus, and TCNi8 values,which slightly overestimate it. Solidus predictions showless agreement, with both databases largely underestimating experimental results. Discrepancies on the orderof 10 C to 20 C can be attributed to heating rate effectsand to the challenging curve analysis when determiningthe ‘onset’ of melting. Overall, solidus values fromFig. 3—EDX maps showing the preferential partitioning of the additions in alloys SX2–SX6. The illustrated scale bar applies to all images.4906—VOLUME 51A, SEPTEMBER 2020METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 4—(a) through (f) Calculated changes in heat capacity, Cp , with temperature, T, for the six trial alloys. Two heating/cooling stages werecarried out in each DSC test. Insets show enlarged views of the heat treatment window sections.METALLURGICAL AND MATERIALS TRANSACTIONS AVOLUME 51A, SEPTEMBER 2020—4907

TCNi8 are closer to DSC data. Conversely, TTNi8predictions for the c0 solvus show better agreement withexperiments. TCNi8 strongly overestimates the solvus,even though non-equilibrium DSC results from heatingFig. 5—Comparison of heat treatment window sections obtainedduring (a) the first and (b) the second heating stage.stages would be expected to yield higher values. Theheat treatment window (HTW), obtained as the difference between solidus and c0 solvus, is a key alloy designfeature.[11] It provides a measure for the potential ofpost-casting heat treatments, which should ideally allowfor a full homogenization in the single-phase c regimewithout any incipient melting. The best estimates ofexperimental HTWs are obtained by combining c0 solvuspredictions from TTNi8 and solidus predictions fromTCNi8.Effects of chemistry changes on HTWs are presentedin Figure 5 using data from the first and second heatingstages. Comparisons are made to Thermo-Calc phasediagrams in Figure 6, using TTNi8 and TCNi8, respectively, to estimate changes in c0 and liquid molarfractions with temperature.The addition of 0.34 at. pct Hf in SX2 significantlydecreases the solidus temperature, while having littleimpact on the solvus. This is in good agreement withprevious reports of Hf acting as a strong melting-pointdepressant and restricting the HTW.[19,21,22] Nonetheless, the same reports also indicate that Hf additions upto around 0.3 at. pct raise the solvus temperature due tothe preferential partitioning of Hf to the c0 phase,[21,23]also observed in the EDX maps of Figure 3. Thisdiscrepancy can be attributed to the lower level of W inSX2 compared with the nominal composition, as W isalso known to raise the c0 solvus.[19,24] It should be notedthat Thermo-Calc correctly captures the reduced HTWas well as the tapered slope for the remaining liquidfraction during the final stages of solidification, asobserved in Figure 6. Studies on hot tearing susceptibility of superalloys have shown that adding Hfincreases the time available for liquid feeding and forstress relaxation, thus improving castability.[25,26]A similar HTW reduction is observed for alloy SX4,in which 1.5 at. pct Al and 0.65 at. pct Ta are substitutedwith 2.9 at. pct Ti. Differential thermal analysis work bySponseller[19] showed that Ta and Ti are more potentthan Al at raising the solvus and lowering the solidusand liquidus temperatures. A comparison of model IGTalloys with carefully controlled chemistries revealed thatthe liquidus decreases at a rate of 6.6 C/at. pct Al, 13.3 C/at. pct Ti, and 17.3 C/at. pct Ta,respectively,[19] in agreement with the present resultsfor SX4. Thermo-Calc correctly predicts the reducedTable III. Comparison of DSC Experimental Results with Thermo-Calc Modeling Predictions for c0 Solvus, Solidus, and LiquidusTemperaturesc0 Solvus ( C)Solidus ( C)Liquidus ( C)Heat Treatment Window ( 112412216010014613561 17487010969107521401209613812064152113123149Heat treatment windows are calculated using values from DSC, TCNi8, TTNi8, and a combination of the two databases, taking the c0 solvus fromTTNi8 and the solidus from TCNi8.4908—VOLUME 51A, SEPTEMBER 2020METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 6—Thermo-Calc predictions for the influence of temperature on (a) c0 volume fraction using the TTNi8 database and (c) liquid molarfraction using the TCNi8 database. The images in (b) and (d) present enlarged views of fractions below 10 pct.solidus and a higher c0 volume fraction up to around1000 C; however, this value falls rapidly at highertemperatures and results in an incorrect, lower solvusprediction compared with SX1.The third chemical modification expected to restrictthe HTW is the additional 1.05 at. pct Si in SX6. Anearly study by Miner[27] showed that Si has a similareffect to Ti by lowering the solidus and liquidus andraising the solvus. This finding was confirmed by severallater studies.[5,28,29] While the solidus does decrease forSX6, the solvus is also reduced, which can be related tothe lower measured Al content in this alloy. Thesefindings are in agreement with Thermo-Calc predictionsin Figure 6.A substantial widening of the HTW is obtained byincreasing the Co content from 4.9 to 10.3 at. pct in SX5.The effects of Co in superalloys were studied at lengthMETALLURGICAL AND MATERIALS TRANSACTIONS Aduring the 1980s following critical supply shortages andrevealed a complex influence of Co on alloy microstructure and properties.[30] For example, the addition of 5.1at. pct Co to a Co-free variation of alloy René N6significantly raised both the solvus and the solidus, butmaintained the same HTW.[19] Further additions up to12.7 at. pct and 20.4 at. pct Co slightly lowered thesolidus, but strongly reduced the solvus, resulting in awider HTW. Similar results were obtained for Udimet700 containing between 0 and 16 at. pct Co; however, thesolvus showed a continuous decrease with increasing Cocontent.[31–33] These same trends were found for alloysMar-M247 and NASAIR 100 when increasing the Colevel from 0 to 10 at. pct[24,34] and for a set of modelNi-Cr-Co-Al-Ti alloys containing between 0 and 57 at.pct Co.[35] Thermo-Calc predictions are in line with theseliterature results and with the DSC data for SX5.VOLUME 51A, SEPTEMBER 2020—4909

Fig. 7—3D reconstruction of an APT dataset of around 21 millionions for the base alloy SX1. For clarity, only a fraction of the Cr/Al,Co/Ta, and Mo/W atoms are shown and Ni atoms are omittedaltogether. The c/c0 interface is shown in gray at aniso-concentration of 20 at. pct Cr.An even stronger effect of depressing the solvus isobserved for SX3 after 6.5 at. pct Ni is substituted withFe. As the solidus is only slightly lowered, the overallHTW expands by almost 40 C. In commercial nickeland nickel-iron-based superalloys, Fe is either (1)completely avoided, as it promotes the precipitation ofdeleterious TCP phases and it lowers high-temperaturestrength, or (2) it is added in significantly higherquantities in low-cost, high-toughness materials forlow-temperature applications. The only commercialcomposition similar to SX3 is the wrought alloy Haynes214, which contains 3 at. pct Fe and which has beendesigned for components operating in carburizing and/or oxidizing environments.[36] Hence, DSC data forcomparable materials with around 5 at. pct Fe arelacking. Nonetheless, Thermo-Calc correctly reproducesthe strong decrease in solvus, although it overestimatesthe lower solidus in Figure 6.C. APT Analysis of Alloy MicrochemistryFigure 7 illustrates an exemplary 3D APT datasetcollected for the base alloy SX1. The c matrix and a c0precipitate are identified unambiguously as a result ofthe preferential partitioning of Cr, Co, and Mo to c andof Al and Ta to c0 . To highlight the c/c0 interface, aniso-concentration surface was generated at 20 at. pct Cr.This cut-off value is used consistently for all six alloys.1. Phase compositions and element partitioningPhase compositions were estimated by generatingcylindrical regions of interest at distances of at least 5nm from the c/c0 interfaces. Results are shown inTable IV, where they are compared with Thermo-Calcpredictions for equilibrium conditions at 845 C. Usingthis data, partitioningare calculated and are 0 coefficients c cin Table V and in Figure 8.presented as log KiCALPHAD predictions and experimental resultsshow a reasonable level of agreement. Some generalc0 ctrends emerge when considering the values of Ki foreach element i in Table V. Thermo-Calc results consisc0 cc0 cc0 cc0 ctently underestimate KNi , KMo , KW , and KAl , whilec0 coverestimating KCr . These differences could be relatedto phases not having attained their equilibrium compositions after heat treatment[37–39] and/or to changes in4910—VOLUME 51A, SEPTEMBER 2020chemistry during the final cooling after secondary agingat 845 C. However, considering the discrepanciesobserved between Thermo-Calc phase diagrams andexperimental DSC results and the general lack of datafor this region of the alloy design space, it is notsurprising that phase compositions differ. The underestimated preferential partitioning of Ni and Al to the c0phase also explains why Thermo-Calc estimates of the c0volume fraction in Table II are lower than those basedon SEM micrographs. Whether TCNi8 or TTNi8 willprovide more accurate estimates depends on both theparticular alloy and the element considered. Overall,TTNi8 values are closer to experim

Mar 08, 2020 · Differential scanning calorimetry (DSC) testing was carried out using a Netzsch DSC 404 F1 Pegasus. Measurements were carried out at heating/cooling rates of 10 C/min under a high purity Ar atmosphere at a purge flow rate of 40 mL/min. PtRh crucibles with PtRh