Effect of heat treatment on the microstructural evolution of a nickel-based superalloy additive-manufactured by laser powder bed fusion

Abstract

Elemental segregation is a ubiquitous phenomenon in additive-manufactured (AM) parts due to solute rejection and redistribution during the solidification process. Using electron microscopy, in situ synchrotron X-ray scattering and diffraction, and thermodynamic modeling, we reveal that in an AM nickel-based superalloy, Inconel 625, stress-relief heat treatment leads to the growth of unwanted δ-phase precipitates on a time scale much faster than that in wrought alloys (minutes versus tens to hundreds of hours). The root cause for this behavior is the elemental segregation that results in local compositions of AM alloys outside the bounds of the allowable range set for wrought alloys. In situ small angle scattering experiments reveal that platelet-shaped δ phase precipitates grow continuously and preferentially along their lateral dimensions during stress-relief heat treatment, while the thickness dimension reaches a plateau very quickly. In situ XRD experiments reveal that nucleation and growth of δ-phase precipitates occur within 5 min during stress-relief heat treatment, indicating a low nucleation barrier and a short incubation time. An activation energy for the growth of δ phase was found to be (131.04 ± 0.69) kJ mol⁻¹. We further demonstrate that a subsequent homogenization heat treatment can effectively homogenize the AM alloy and remove the deleterious δ phase. The combined experimental and modeling methodology in this work can be extended to elucidate the phase evolution during heat treatments in a broad range of AM materials.

1. Introduction

Additive manufacturing (AM) of metals is based on a layer-by-layer additive process, in contrast to traditional manufacturing processes that often require labor-intensive and cost-prohibitive subtraction or forming [1]. Most existing AM technologies make use of powder-based or wire-based feedstock materials that are selectively melted and solidified at high heating and cooling rates, with various heating sources such as lasers, electron beams, and plasma arcs [2–5]. AM technologies provide great flexibility in manufacturing parts with complex geometrical shapes and can significantly reduce manufacturing lead times and associated cost. Thus, AM is fast becoming an attractive option for the fabrication of increasingly complex and high-valued metal components in aerospace, oil & gas, automobile, electronics, and biomedical industries. It is worth noting that in these demanding applications, it is not sufficient that the AM parts have the desired shape. They must also have consistent and reliable material properties that are competitive with those of their wrought counterparts [6,7].

Despite the great promise that AM materials have clearly demonstrated, from the standpoint of fundamental materials science and engineering, significant challenges and knowledge gaps remain. Some of these challenges are rooted in the AM process itself. During AM, the line-by-line scanning of a high-powered laser or electron beam is by nature a highly volatile and nonequilibrium process [5,8]. This process is characterized by short interaction times between powder and laser/electron beams with localized high heat inputs. A given material volume is subject to a complex thermal cycle that includes rapid heating above the melting temperature of the material due to energy deposition from the heat source, rapid cooling and solidification of the molten material after the source moves away, and subsequent repeated re-heating and re-cooling in the layer-by-layer fusion process. Thus, the resulting material often has a high level of residual stress, heterogeneous metastable microstructures, and nonequilibrium elemental compositions or phase distributions [5,8–10]. These characteristics not only make it difficult to produce AM parts with reproducible and acceptable materials properties, but also presents challenges to model and predict the microstructures and compositions of AM materials in ways that can accelerate the design and discovery of AM alloys suitable for applications with desired materials performance [2,11]. Specifically, these characteristics of the AM process negatively impact the mechanical properties of AM parts [6]. To fulfill the promise of AM materials, these challenges must be addressed.

A common approach to mitigate such challenges is through a series of post-build heat treatments to reduce internal stresses, increase materials density, finalize component shape and finish, and perhaps most importantly, develop the required phase and microstructure landscape that leads to the desired materials properties [12–14]. Among these, stress-relief and homogenization are two essential types of heat treatment. On the one hand, stress-relief is routinely used to reduce the residual stress that is built up in the thermal cycles, which, if untreated, can lead to geometrical distortion of the AM components, microscopic cracks that reduce mechanical strength, and even part failure at lower-than-designed applied stress level [15]. On the other hand, homogenization heat treatments act to homogenize the elemental distribution and potentially dissolve deleterious phases present in the material, thus providing a path for producing materials and parts with predictable and reproducible microstructure and materials properties [16].

Despite the distinct benefits of heat treatments, they may also bring unexpected side effects. It has been shown that in stress-relief and hot isostatic pressing (HIP) heat treatments of AM-built materials, unexpected intermetallic and carbide phases can form. For example, Idell et al. showed that δ-phase precipitates can grow during a 1066 °C isothermal anneal of an AM-built ATI 718Plus³ alloy produced by direct metal laser sintering [17]. Similarly, work by Mostafa et al. on Inconel 718 material built by selective laser melting demonstrated that a HIP treatment at 1160 °C and 100 MPa for 4 h promoted the formation of δ phase and MC carbides along the grain boundaries [18]. The homogenization heat treatment, which normally occurs at a higher temperature to allow dissolution of deleterious phases, can also promote grain growth, which adversely affects the hardness, yield strength, tensile strength, fatigue strength, and impact strength of these alloys. Hence, a comprehensive understanding of the effect of heat treatment on the microstructure and phase evolution of AM alloys is required.

So far, this important technical issue has only drawn limited attention, and much of the progress was achieved using ex situ probes such as lab-based X-ray diffraction (XRD) and local probes such as scanning electron microscopy (SEM) or transmission electron microscopy (TEM). AM materials are known to inherently possess compositional and microstructural inhomogeneity across a range of length scales, which makes it difficult to directly compare results acquired from different sample volumes. To elucidate the kinetics of the phase evolution of AM alloys during heat treatment and reveal their thermodynamics, it is important to conduct investigations using in situ, statistically representative measurement techniques.

Inconel 625 (IN625) is a nickel-based superalloy strengthened by solid-solution hardening of Nb and Mo in a Ni-Cr matrix [19,20]. Because IN625 has a combination of high yield strength, fatigue strength, and excellent corrosion resistance in aggressive environments, it has found widespread applications in the aerospace, marine, and nuclear industries where complex shapes are often required. At the same time, the machinability of IN625 is poor, making IN625 an excellent candidate for AM. Meanwhile, IN625, as a simple solid-solution alloy, serves as an excellent model system for a detailed understanding of the effect that the AM process has on alloy performance. Coupled with the availability of IN625 feedstock powder, extensive experimental and modelling work has been conducted on AM IN625 [3,14,21–27].

IN625, while designed as a solid-solution strengthening alloy, is sensitive to precipitation of intermetallic phases such as Ni₃M γ” phase, Ni₃M δ phase, and Ni₂(Cr, Mo) Laves phase as well as MC primary carbides and M₆C and M₂₃C₆ secondary carbides [28,29]. A large amount of literature exists on the heat-treatment-induced phase evolution of wrought IN625 [30–33]. Similarly, the effect of post-weld heat treatment on the microstructure of weld IN625 has been studied by several groups [34–36]. It is worth pointing out that while welding and additive manufacturing both feature melting and solidification processes, the resulting microstructures of as-weld and as-built AM components are expected to be different due to different thermal gradients and cooling rates (hundreds K/s for welding [37,38] and hundreds of thousand K/s for L-PBF [39]). At the same time, limited work has been reported on AM IN625. Notably, Lass et al. discovered a copious amount of deleterious δ phase precipitates following the manufacturer-recommended stress-relief heat treatment (1 h at 870 °C) [23]. In contrast, precipitation of δ phase precipitates in wrought material is normally not expected until after tens to hundreds of hours [32,33].

In this paper, we present a detailed study of phase evolution of AM IN625 during stress-relief heat treatment within a temperature range identified by Lass et al. [23]. Our approach combines in situ X-ray scattering and diffraction, ex situ electron microscopy and X-ray diffraction, and thermodynamic modeling. In particular, we make use of a synchrotron-based X-ray scattering and diffraction probe that has been recently developed to simultaneously monitor both the evolution in precipitate morphology and changes in crystal structure [40]. The phase evolution and the kinetics information are vital for a proper understanding of the microstructure across all relevant length scales with the ultimate goal of achieving the comprehensive material design and optimization of AM IN625, as well as other AM materials.

2. Materials and methods

2.1. Manufacturing of materials

Fifteen-millimeter cubes of IN625 were additively manufactured from virgin IN625 powder by the Measurement Science for Additive Manufacturing program at the National Institute of Standards and Technology (NIST), Gaithersburg, U.S.A., using an EOS M270 laser-sintering powder-bed fusion instrument (EOS GmbH, Munich, Germany). The manufacturer-certified chemical composition of the raw EOS IN625 powder used in this study and the corresponding allowable composition ranges (from the ASTM Standard for Additive Manufacturing Nickel Alloy UNS N06625) are listed in Table 1 [22], which shows that the mass fractions of all elements are within the standard ranges for IN625. More details about the composition can be found elsewhere [22]. A Nd:YAG laser was operated at 195 W. The scanning speed was 800 mm/s. The hatching distance was 100 μm. The melt-pool width ranged between 105 μm and 115 μm. The solidification velocity is location-specific and ranges between 1 mm/s to 30 mm/s, according a recent finite element analysis of a model Ni-Nb binary alloy using the same set of build parameters [39]. An “all core” build pattern was used, where the laser scan pattern was identical at the external surfaces and in the interior of the samples. After the build, the test cubes were removed from the base plate by electro-discharge machining in the as-built condition, i.e., without a stress-relief heat treatment. Specimens were cut and mechanically polished to remove surface abnormalities, following a standard metallographic procedure.

Table 1.

Measured composition of the IN625 feedstock powders used in this work and the allowable range of composition of IN625. More details can be found in Ref. [22].

Element	Standard Range (mass fraction)	Measured (mass fraction)
Ni	Balance	Balance
Cr	20.0%–23.0%	20.7%
Mo	8.0%–10.0%	8.83%
Nb	3.15%–4.15%	3.75%
Fe	5.0% maximum	0.72%
Ti	0.4% maximum	0.35%
Al	0.4% maximum	0.28%
Co	1.0% maximum	0.18%
Si	0.5% maximum	0.13%
Mn	0.5% maximum	0.03%
C	0.1% maximum	0.01%

2.2. Ex situ scanning electron microscopy (SEM)

We characterized the microstructure and spatial variation in elemental composition of the as-built and heat-treated samples using a JEOL S-7100F (JEOL, Ltd., Akishima, Tokyo, Japan) field emission SEM. This instrument is equipped with an Oxford X-MAX^N (Oxford Instruments Plc., Abingdon, UK) energy dispersive X-ray spectrometry (EDS) detector, which allows elemental composition of the specimen to be analyzed. It is also equipped with Oxford NordlysMax² electron backscattered diffraction (EBSD) detectors, allowing microstructural aspects such as grain morphology, texture, and crystallographic mapping to be explored. We operated the SEM at 15 kV for secondary electron (SE) imaging and at 20 kV for backscattered electron (BSE) imaging. The EBSD data were acquired with a step size of 3.0 μm. The SEM specimens were oriented in the plane that is perpendicular to the base plate (along the build direction), which allows microstructural information related to the dendritic and interdendritic regions to be acquired.

2.3. Ex situ synchrotron X-ray diffraction (XRD)

High-resolution synchrotron XRD experiments were conducted at the dedicated powder XRD beamline 11-BM-B at the Advanced Photon Source (APS), Argonne National Laboratory [41]. The monochromatic X-ray energy was 30 keV (wavelength, λ = 0.414554 Å). We acquired XRD data in a 2θ (diffraction angle) range of 0.5°–28°, with a step size of 0.001° and dwell time per step of 0.1 s. This 2θ range translates to a q range of 0.132265 Å⁻¹ to 7.333376 Å⁻¹, with q = 4π sin(θ)/λ. The unique combination of high X-ray flux, angular resolution, and sensitivity allows the lattice parameters to be accurately determined and weak phases to be detected.

We performed XRD experiments on as-built, IN625 specimens heat-treated at 870 °C for 8 h, and 870 °C for 1 h followed by 1150 °C for 1 h. The specimens were encapsulated in an evacuated ampoule during the heat treatments, and were subsequently thinned to approximate dimensions of 200 μm × 200 μm × 10 mm. During data collection, the samples spun rapidly in the beam. The phase identification was conducted using GSAS-II [42] following a Le Bail refinement. Due to sample texturing and compositional complexity, full Rietveld analysis was not conducted.

2.4. In situ synchrotron small angle X-ray scattering and X-ray diffraction

In-situ synchrotron ultra-small angle X-ray scattering (USAXS), small-angle X-ray scattering (SAXS), and XRD experiments were conducted at the USAXS facility at the APS [43,44]. The X-ray energy was 21 keV (λ = 0.5904 Å). The flux density was ≈ 10¹³ photons/s/mm².

We used combined USAXS and SAXS to monitor the changes in the morphology of the precipitates during growth. The USAXS instrument makes use of Bonse-Hart-type double-crystal optics and extends the scattering vector, q, range of SAXS down to 1 × 10⁻⁴ Å⁻¹. For USAXS measurements, the beam size was 0.8 mm × 0.8 mm. To provide better signal-to-noise at high q, we supplemented the USAXS instrument with a PILATUS 100K detector (Model: 100K-S, Dectris, Baden, Switzerland) [45] in a conventional pinhole small-angle scattering geometry. The SAXS q values were calibrated using a AgBe calibration standard. We used a beam size of 0.8 mm × 0.2 mm for SAXS measurements, seeking the best possible counting statistics while matching the detector resolution. The combined accessible q range for USAXS and SAXS is 1 × 10⁻⁴ Å⁻¹ to 1.5 Å⁻¹, and the combined dynamic range in linear intensity response exceeds 10 orders of magnitude.

To evaluate the changes in the atomic structures of the precipitates and matrix in IN625 during the heat treatments, we made use of a modified PILATUS 300 KW detector to perform area-detector based XRD experiments in a q range between 1.4 Å⁻¹ and 6.8 Å⁻¹. We calibrated the q values and sample-to-detector distances using NIST standard reference 660a (LaB₆: lanthanum hexaboride) [46]. We used a beam size of 0.8 mm × 0.2 mm for the in situ XRD measurements.

For the in situ experiments, the as-built specimens were mechanically polished to ≈50 μm in thickness. The in situ heat-treatment experiments were conducted using a Linkam 1500 thermal stage (Linkam Scientific Instruments Ltd., Tadworth, UK). The thermal stage was calibrated using a 99.999% purity Pt foil sample. The temperature uncertainty depends on the set temperature, and is estimated to be 1 °C in the temperature range used for this study. We performed three sets of in situ experiments to follow the formation of the precipitates at isothermal heat-treatment temperatures of 800 °C, 835 °C, and 870 °C, respectively. Among these temperatures, 870 °C is the manufacturer-recommended heat-treatment temperature, and 800 °C is the suggested alternative stress-relief heat-treatment temperature by Lass et al. [23]. The heating rate from room temperature to the target temperatures was set at 200 °C per min.

For each heat-treatment series, we conducted the combined measurements in a repeated sequence of USAXS, SAXS, and XRD, with individual scan time set at 90 s, 30 s, and 60 s, respectively. Including the time required for motor motions, each set of measurements took ≈5 min. For each isothermal heat-treatment condition, we conducted the in situ experiment for ≈10 h, the maximum amount of time allowed by beamtime allocation.

2.5. Thermodynamic calculations

We calculated the phase diagram of IN625 using the software Thermo-Calc (v2015b) [47] and the commercial thermodynamic database TCNI8 developed for Ni-rich alloys [48]. To elucidate the observed precipitation kinetics in the current experiment, we have calculated the precipitation kinetics using the TC-PRISMA precipitation software code [49–51]. TC-PRISMA applies Langer-Schwartz theory [52] and Kampmann-Wagner numerical analysis [53,54] to simulate the nucleation and growth of precipitates in multicom-ponent and multiphase alloy systems, where the time-dependent evolution of the precipitate volume fraction over the entire course of the heat treatment is predicted by solving a set of rate equations.

3. Results and discussion

3.1. As-built AM IN625

Fig. 1 shows the EBSD microstructural and crystallographic characterization of AM IN625. Fig. 1(a) shows the grain and sub-grain microstructure, with pronounced grain elongation along the build direction, which is characteristic of the directional build process of laser-based AM and has been repeatedly observed in various AM alloys [14,17,18,55]. Wrought IN625, in contrast, has an equiaxed grain structure [30]. Additionally, we observed that within individual grains, the color associated with crystallographic orientations is not always uniform, which indicates that a range of crystallographic orientations may be present. This is consistent with the high local stresses and high dislocation densities found in these materials [17]. We further quantitatively analyzed the grain microstructure. For statistical purposes, we defined an individual domain using the following set of criteria: (1) the misorientation between neighboring grain needs to be greater than 15°; (2) each grain needs to contain at least 5 data points, i.e., dimensions along either the major or minor direction of the elongated domains need to be equal to or greater than 15 μm; and (3) the grain does not reside at the border of the EBSD image shown in Fig. 1(a). The analyzed grain-size histogram is shown in Fig. 1(b). Here, the grain size is defined as the average of the lengths of the major and minor axes of each grain. We identified a unimodal grain size distribution with an average grain size of 57.1 μm and an average aspect ratio of 4.4. This grain size is significantly larger than the 12.8 μm average grain size identified in a study of wrought IN625 [30]. A texture analysis of the EBSD data, shown in Fig. 1(c), demonstrates that the as-built AM IN625 had a weak to moderate cubic texture, which is denoted by the strong off-axis peak in the {200} map and the four weak off-axis peaks in the {111} map. Strong texture in {200} has been frequently observed in as-built AM nickel-based alloys, which is related to the grain formation during repeated heating and cooling, and was also found to be strongly dependent on the power of the heat source [3,17,18].

Fig. 1.

EBSD results of as-built AM IN625 alloy. (a) An image of the microstructure with color superimposed based on EBSD demonstrates an elongated grain microstructure along the build direction. The inverse pole figure is included as a legend to enable mapping between color and crystallographic orientation. (b). The relative grain size histogram in percentage (η_i/η_total) shows a unimodal grain-size distribution with an average grain size of 57.1 μm and an average aspect ratio of 4.4. Here η_i is the number of grains in each size bin, and η_total is the total number of grains used in this analysis. (c) Texture analysis reveals that the as-built AM IN625 alloy has a weak to moderate cubic texture.

Fig. 2(a) presents the synchrotron XRD data for as-built AM IN625. The peaks are noticeably asymmetric. From a strict peak-fitting point of view, this dataset is best described by two crystal structures with identical symmetry (F m −3 m) but slightly different lattice parameters – 3.6009 (2) Å and 3.6196 (3) Å, respectively (all lattice parameters reported in this paper are summarized in Table (2)). The formation of solid-solution face-centered cubic (FCC) structures is consistent with previous results of as-built AM IN625 [23,56], indicating that the crystal symmetry of the starting powder material was preserved during the rapid AM process. The peak asymmetry can be explained by two distinct mechanisms, localized elastic strains due to the dislocation distributions [57,58] and the compositional gradients within the dendritic microstructure which produce a corresponding variation in lattice spacing. Both mechanisms likely contribute, but attempting to separate these effects is beyond the scope of the present paper.

Fig. 2.

(a): Synchrotron XRD data demonstrates that as-built AM IN625 demonstrates peak asymmetry that can be characterized by two face-centered cubic structures with lattice parameters of 3.6009 (2) Å and 3.6196 (3) Å, respectively. This peak asymmetry is likely due to a combination of local elastic strains within the dislocation structures and elemental microsegregation. (b) High-resolution synchrotron XRD of AM IN625 after a heat treatment of 8 h at 870 °C. Three phases are present: FCC matrix, orthorhombic δ phase, and FCC MC carbide. The lattice parameters of these phases are listed in Table 2.

Table 2.

Lattice parameters for the phases identified in this study. All measurements were conducted at room temperature. Both lattices of matrix and MC carbide are FCC. The lattice of δ phase is orthorhombic.

	Matrix lattice parameter 1 (Å)	Matrix lattice parameter 2 (Å)	δ phase lattice parameter (Å)	δ phase lattice parameter (Å)	δ phase lattice parameter (Å)	MC carbide lattice parameter (Å)
As-built, AM IN625	3.6009 (2)	3.6196 (3)
870 °C for 8 h	3.5993 (1)		5.0913 (5)	4.2310 (8)	4.4797 (14)	4.0863 (2)
Homogenized	3.6160 (1)

To examine the spatial elemental distribution in AM IN625, we performed EDS mapping of an as-built IN625 specimen. Typical elemental maps of Ni, Cr, Nb and Mo for the same area of sample are shown in Fig. 3. A segregated dendritic microstructure is observed. Such microstructure is closely tied to the rapid and nonequilibrium solidification conditions inherent to the AM process [60,61]. We observed local enrichment of Nb and Mo and corresponding deficits of Ni and Cr in the interdendritic regions. The trends within the dendrites were the opposite. Such elemental microsegregations occur due to solute rejection and redistribution during a solidification process, a process, not surprisingly, also broadly identified in the welding literature [62,63]. Furthermore, Fig. 3 shows that the microsegregation is not homogeneous across the probed sample area. This behavior is consistent with previous experimentally observed and theoretically predicted microstructure heterogeneities [64,65]. Here, we surmise that the extent of microsegregation is related to the complex heating (reheating) profile at a local level.

Fig. 3.

EDS compositional mapping of (a) Ni, (b) Cr, (c) Nb and (d) Mo of the as-built AM IN625.

We performed quantitative analysis of the extent of micro-segregation by using an EDS line scan with a step size of 2 μm and a dwell time of 300 s over ≈200 μm, as highlighted by the arrow in Fig. 3(c). The calibrated, spatially-resolved mass fractions of two heavy elements (Mo and Nb) are shown in Fig. 4. The Nb and Mo mass fractions were measured between 2.81% and 5.84% and 9.35% and 10.85%, respectively. Fig. 4 shows that these mass fractions are not confined to bounds set forth by the chemically allowable values of IN625. Similar fluctuation was also observed for Ni and Cr. We emphasize that these results were acquired using a 2 μm beam size, hence the measured mass fractions were the average value within the probed volume. Our previous synchrotron experiment, by analyzing the lengths of X-ray streaks in the two-dimensional SAXS patterns, indirectly demonstrated that the atomic concentration profile within the microsegregated interdendritic regions has a nominal width on the order of 10 nm [24]. This result was further validated by both thermodynamic and phase-field modeling. A CALPHAD-based solidification modeling effort was carried out using both DICTRA software and a Scheil-Gulliver model, where the temperature profile was supplied by a finite-element thermal model [22]. Similarly, two dimensional and three dimensional phase-field models were studied using the same finite-element thermal model [66]. All of these models predict that most of the microsegregation occurs within a 10 nm region, and that the upper bound of local Mo and Nb mass fractions are predicted at as high as 20 mass % and 29 mass %, respectively. Such observations strongly suggest that, due to microsegregation, local composition of as-built IN625 alloy does not consistently meet the alloy specifications and must be further processed via heat treatments so that the expected performance requirements can be achieved. It is also important to point out that the existing experimental evidence, while consistent with modeling predictions, is nevertheless indirect. Direct evidence using techniques that can resolve compositional changes with a nm resolution is needed to validate predictions made by thermodynamic and phase field models.

Fig. 4.

EDS line profiles of Nb and Mo for as-built IN625 across a distance of 200 μm. The solid straight lines show the upper and lower bounds of the allowed mass fractions of Nb and Mo for IN625, as given in Table 1. The line profiles demonstrate that the elemental microsegregation results in significant local departures of the chemistry of AM IN625 from the specified values of IN625.

Previous residual stress studies of as-built IN625 reveal the macroscopic residual stress can exceed 750 MPa [23,67]. Heat treatment is required to relieve the residual stress. Thus, it is important to monitor the microstructural and phase evolution of AM IN625 under such heat-treatment conditions.

3.2. Stress-relieved AM IN625: Ex situ studies

Fig. 2(b) shows an XRD profile of AM IN625 after a heat treatment of 870 °C for 8 h. Comparing with the XRD profile of as-built IN625 shown in Fig. 2(a), additional peaks emerged from this heat treatment. A careful analysis identified that the additional peaks belong to two types of precipitates, an orthorhombic δ phase with lattice parameters of 5.0913 (5) Å, 4.2310 (8) Å, and 4.4797 (14) Å and an FCC MC carbide phase with lattice parameter of 4.0863 (2) Å. The crystallographic symmetries of the matrix phase, δ phase, and MC carbide are consistent with previous findings in wrought IN625 alloys. The δ phase lattice parameters are generally smaller than the reported lattice parameters of Ni₃Nb (5.07 Å, 4.23 Å and 4.57 Å, respectively) [68]. Previous EDS results of the stress-relieved IN625 show that δ phase precipitates are rich in both Nb and Mo [23], suggesting a compositional deviation is a likely cause for the lattice-parameter mismatch. Nanoscale precipitates in an alloy can have a strained lattice, which may also contribute to the reduced lattice parameters observed here. Microscopic evidence shows that the MC carbide is rich in Nb and Mo, and we tentatively assign it as (Nb, Mo)C. Furthermore, the FCC matrix of the stress-relieved IN625 has a lattice parameter of 3.5993 (1) Å, which is smaller than the lattice parameters of the as-built IN625 (3.6009 (2) Å and 3.6196 (3) Å). This result indicates that heavy atoms such as Nb and Mo are removed from the matrix solid solution to form the precipitates, in agreement with the SEM observation of enriched Mo and Nb content in the δ phase precipitates.

We examined the microstructure of AM IN625 heat-treated at 870 °C for 0.5 h, 1 h, 4 h, and 8 h, respectively. The SEM micrographs are shown in Fig. 5. Evidently, precipitates formed in the sample, and their volume fraction increased with increasing heat-treatment duration. Particularly, it is readily observed in the micrograph acquired after 0.5 h that initially, the precipitates preferentially formed within interdendritic regions, where Nb and Mo were enriched. The appearance of most precipitates was needle-like. From a geometrical point of view, needle-shaped projections serve as strong evidence for a precipitate morphology of two-dimensional planar objects, i.e., disks. This is consistent with previous reports where δ precipitates were identified as platelet-shaped [23,28,30]. Additionally, we also observed much smaller globular precipitates, which is consistent with the reported size and shape of MC carbides in IN625 [28,69–71]. The presence of MC carbides is also confirmed by TEM (data not shown). At the same time, TEM data did not show γ′ precipitates, which were identified in as-welded IN625 [35]. We again emphasize that the observed local number density of δ phase is not uniform. This can be attributed to the nonuniform chemical composition, as highlighted by Fig. 3, which in turn gives rise to a localized thermodynamic driving force for the nucleation and growth of the precipitates. While our observation is specific to AM IN625, this statement of nonuniform precipitation is expected to hold for many other AM alloys.

Fig. 5.

SEM micrographs of AM IN625 after 0.5 h, 1 h, 4 h, and 8 h heat treatment at 870 °C clearly show the formation of δ phase precipitates. Particularly, the micrograph acquired after 0.5 h clearly demonstrates the initial precipitation of δ phase occurs within the interdendritic regions. The samples were etched via immersion in aqua regia for 10–60 s to reveal the microstructure.

3.3. Stress-relieved AM IN625: In situ studies

While the ex situ SEM and XRD clearly reveal the phases present in the heat-treated AM IN625 alloy, they nonetheless were acquired with different sample volumes. To overcome this limitation, we performed in situ synchrotron X-ray scattering and diffraction studies to investigate the time- and temperature-dependent phase evolution within the same sample volume. Such in situ techniques have the distinct advantage of quantitively capturing the phase-evolution kinetics within the same macroscopically significant and statistically meaningful volume, thus providing precise thermodynamic parameters necessary to model the time-dependent evolution of materials properties, particularly the mechanical strengths.

Fig. 6 shows the complete set of SAXS data acquired from an as-built AM IN625 sample during an isothermal heat treatment at 870 °C for ≈10 h (613 min). The USAXS data form the main plot, and the higher-q SAXS data are displayed as an inset. The scattering curves are color-coded according to their time of acquisition from the start of the experiment, following the color scale shown by the arrow in Fig. 6. The SAXS data exhibit the following characteristics. First, the presence of initial low-q scattering between 1 × 10⁻⁴ Å⁻¹ and 1 × 10⁻³ Å⁻¹ indicates that scattering inhomogeneities with sizes greater than the instrumental detection limit (a few micrometers) exist. This low-q scattering of power-law type, due to its very low q range, is infrequently observed in SAXS studies of alloys due to q-range limitations. Hence its contribution is often neglected in the subsequent SAXS analysis [72,73]. For this work, we associate this scattering component with grain-associated scattering, similar to recent SAXS studies of a model Ni-based superalloy [74] and of aluminum alloy 2024 [40]. Importantly, a recent study found that grain growth of IN625 at the stress-relief heat-treatment temperatures (temperature < 900 °C) is minimal [75], in part due to the roles that MC carbide and δ phase precipitates play in promoting grain growth resistance via a grain-boundary pinning effect [71]. Thus, we can reasonably assume that the low-q power law associated with the heat treatment does not change, and that the scattering profile prior to the heat treatment can be used as the scattering baseline. Second, across the entire q range, we observed a monotonic increase in the scattering intensity as a function of time, as highlighted by the arrow direction. Such increase in intensity is known to be associated with nucleation and growth of precipitates in alloys [76,77]. Fig. 6 indicates that the rate of precipitation was initially rapid, then it gradually slowed down. Notably, the SAXS data shown in the inset simply extend the high q power-law in the USAXS data, suggesting lack of growth of nm-scale precipitates during the heat treatment.

Fig. 6.

A complete set of in situ SAXS data acquired from an AM IN625 sample during isothermal heat treatment at 870 °C for ≈10 h (613 min). The main plot shows the USAXS data, and the inset shows the SAXS data. The scattering curves are color-coded by acquisition time. This dataset demonstrates a continuous precipitate growth.

We performed detailed SAXS analyses to investigate the morphological evolution of the precipitates. Data reduction and analysis were conducted using the small angle scattering analysis software Indra and Irena, developed in the Igor Pro programming environment [78]. An ideal SAXS analysis makes use of microscopic evidence as the starting point and constructs a scattering model assuming scattering objects that are consistent with the observed sample characteristics [17,40]. Alternatively, when spherical form factor and isotropic structure factor of the scattering entities can be assumed, recently developed Bayesian-MaxEnt analysis approach can recover the precipitate size distribution without assuming the functional form of the distribution [74]. These assumptions cannot be made about δ-phase precipitates in AM IN625. The fundamental challenges are best explained by Fig. 5. While the morphology of δ phase precipitates is simple, their spatial distribution is not. The precipitates exhibit strong preferred orientations, local high number density, and spatial nonuniformity. Models that take into consideration some of these characteristics exist [79], although substantial assumptions such as strong global preferred orientations often have to be made and 2-dimensional scattering patterns are required. Because of these challenges, instead, we opted to adopt an analysis approach analogous to the unified small-angle scattering analysis method [80,81], where characteristic length scales of different scattering levels in a complex scattering system can be reasonably extracted.

An illustration of this analysis is shown in Fig. 7(a). Here, the SAXS dataset was the last set of the in situ series, and was acquired at 613 min into the isothermal heat treatment at 870 °C. For this analysis, we assumed that the changes in the scattering profiles were mainly due to the morphological changes in δ phase precipitates, as SEM micrographs in Fig. 5 clearly demonstrate. This assumption is further supported by the in situ XRD data reported later in this paper. We assumed that the scattering baseline, acquired for the same sample volume at room temperature prior to isothermal heat treatment, does not change appreciably through the heat treatment. We also assumed that two scattering levels exist in the precipitate scattering, which correspond to the lateral dimension and the thickness of δ phase precipitates, respectively. As demonstrated by Fig. 7(a), this model describes the data well. While this model is simple, we believe that it captures the essential elements of the physical precipitation events in AM IN625.

Fig. 7.

(a) An illustration of the SAXS model. These data were acquired at 613 min during the isothermal heat treatment of AM IN625 at 870 °C. The overall fit, depicted by the red solid line, is the sum of the scattering baseline and two scattering levels related to the diameter and thickness of the δ phase precipitates. (b) Time-dependent evolution of average diameter and thickness of the δ phase at 870 °C. The diameter shows a monotonic increase as heat treatment proceeds, while the thickness shows a weak dependence on time. The inset demonstrates that the thickness plateaus at (51.7 ± 5.2) nm. The uncertainties in this Figure and hereafter represent one-standard deviation.

As an example, Fig. 7(b) shows the evolution of the average thickness and diameter of δ phase precipitates at 870 °C, extracted from the analyses of all in situ SAXS datasets. Evidently, the thickness and diameter have strikingly different behaviors. The diameter of δ phase precipitates, after an initial rapid rise, continued to increase gradually as time increased. The average thickness, on the other hand, had a weak dependence on time and appeared to plateau at (51.7 ± 5.2) nm. This again is consistent with microscopic observations, where the changes in δ phase thickness over time were found to be subtle. The USAXS dimensionalities of the δ phase precipitates are also in good agreement with the values identified from the SEM results (after 8 h, diameter at (815 ± 157) nm and thickness at (48 ± 6) nm). Previous TEM micrographs of δ phase in AM IN625 heat-treated at 870 °C for 1 h demonstrated that the precipitates are highly crystalline and that the orientation relationship between the FCC matrix and δ phase precipitates follows {111}_FCC//(100)_δ and <11̄0>_FCC //[100]_δ, a result also identified in an earlier TEM study of δ phase in 718 alloy [23,82]. In particular, it was reported that in each close-packed plane, three variants of δ phase can form with their long axes aligned with the close-packed directions in the FCC matrix. From this observation, we infer that it is this required crystallographic orientation of δ phase with the FCC matrix that determines the growth direction.

The simultaneously acquired in situ XRD data provide a direct window to probe the growth kinetics of the crystalline precipitates, owing to the precise phase identification demonstrated in Fig. 2(b). Fig. 8 shows the evolution of the XRD data from the same in situ experiment conducted at 870 °C. Fig. 8(a) shows the XRD data of the as-built IN625, acquired prior to the isothermal heat treatment, where only the FCC matrix peaks were observed. The XRD data acquired at 870 °C, shown in Fig. 8(b), show a monotonic increase in the peak intensities associated with the δ phase precipitates, and a monotonic decrease in the peak intensities associated with the FCC matrix. The post heat-treatment data, acquired after cooling to room temperature, can be indexed by FCC peaks and δ phase peaks alone, with MC carbide peaks, identified in ex situ high-resolution XRD experiments (Fig. 2(b)), noticeably missing. This may originate from the relatively small sample volume in the in situ XRD measurement (0.8 mm × 0.2 mm × 50 μm). Additionally, the elemental microsegregation may also contribute to a low number of MC carbide within the probed volume, leading to XRD peaks of MC carbide not being observed.

Fig. 8.

In situ XRD data acquired during heat treatment of AM IN625 at 870 °C. (a) XRD data of as-built IN625 acquired at room temperature only show peaks of FCC matrix. (b) evolution of in situ XRD data during isothermal heat treatment shows that additional peaks (phases) emerge. The inset shows two closely-located peaks of δ phase, with the {211} reflection being the strongest peak of the δ phase XRD spectrum. The data acquisition time follows the arrowed color scale. (c) XRD data after the heat treatment, acquired at room temperature.

Furthermore, we emphasize that across the temperature range reported in this paper (800 °C–870 °C), the first XRD measurement of the in situ series (acquired within 5 min after the start of the in situ heat treatment) always shows weak XRD peaks characteristic of the δ phase. Such peaks were not observed in the XRD patterns of the as-built alloys, indicating that the δ phase was below the detection limit of the highly sensitive XRD experiments, whose sensitivity arises from the high X-ray flux density of the synchrotron and the single photon-counting performance of the area detector. Thus, we conclude that the incubation time, necessary for the nucleation of the δ phase, is less than 5 min, the detection limit of our study, in the temperature range 800 °C–870 °C. This result indicates that the nucleation barrier for the δ phase precipitation is low, most likely due to a high dislocation density commonly identified in AM materials which provides heterogeneities nucleation sites for the precipitates [59]. This observation also shows that the growth of δ-phase precipitate occurs at a rapid rate as their diffraction signals became detectable very shortly after the isothermal heat-treatment temperatures were reached.

We analyzed the δ-phase peak profiles to extract the growth kinetics. As the inset of Fig. 8(b) highlights, the δ {211} peak is the most prominent peak in the observed XRD pattern, and is also the most intense peak in the simulated δ-phase stick pattern shown in Fig. 2(b). The time-dependent integrated peak intensities of the δ {211} reflection at 800 °C, 835 °C, and 870 °C are shown in Fig. 9(a), (b), and (c), respectively. The kinetics demonstrate a strong dependence on temperature – the higher the temperature, the faster the growth kinetics. We modeled the development of the integrated peak intensity as a function of time following a simple exponential-decay function I = I₀ – C exp (−κ fit), where t is time, κ is a temperature-dependent rate, I₀ is the asymptotic term as t → ∞. The least-squares fits, the solid lines in Fig. 9, show good agreement with the observed kinetics. The fitting results are summarized in Table 3. Within the one-standard-deviation un-certainties, we found that I₀ and C are equivalent, which again indicates that, within our instrumental resolution, no δ phase was present at the start of the isothermal heat treatments. We also note that we explored using the Austin-Rickett kinetic rate equation that was found successful in describing diffusion-controlled precipitation kinetics of many alloy systems [83]. For the kinetic data reported in this paper, a simple exponential decay function presents better fitting than the Austin-Rickett equation. We surmise that the compositional heterogeneity may lead to a distribution of localized precipitation rates, for which a smearing effect renders a generic exponential decay function a better description of the kinetics.

Fig. 9.

Temperature-dependent δ phase growth kinetics as detected by in situ XRD experiments at (a) 800 °C, (b) 835 °C, and (c) 870 °C, respectively. The solid red lines show least-squares fits to the data using an exponential decay model described in the main text. The resulting kinetic rates are reported in Table 3.

Table 3.

A summary of the isothermal precipitation growth rates related to the δ phase precipitates at different temperatures.

Temperature (°C)	Temperature (K)	κ (min−1)
800 ± 1	1073.15 ± 1	0.00060 ± 0.00009
835 ± 1	1108.15 ± 1	0.00174 ± 0.00005
870 ± 1	1143.15 ± 1	0.00480 ± 0.00011

We further followed the evolution of the δ phase peak positions and widths as a function of time at these temperatures (data not shown). We found that the peak positions remain stable, indicating a stable atomic structure of δ phase precipitates after they form. This is in contrast to an in situ study of δ phase precipitation in AM 718 + alloy, where a gradual reduction in the lattice parameters was observed, attributed to stress relaxation in the sample during a heat treatment [17]. We also found that the peak width decreases with increasing time, consistent with a continuous growth of δ phase precipitates as revealed by the SAXS analysis.

We also examined the evolution of the atomic structure of the FCC matrix. As an example, the lattice parameter variation at 870 °C is shown in Fig. 10. We observed a monotonic reduction in the lattice parameter, which is inversely correlated with the growth of δ phase precipitates as illustrated in Figs. 7 and 9. This result is expected, due to the formation of δ phase precipitates requiring removal of heavy elements such as Nb and Mo from the matrix, thus reducing the size of the lattice in the matrix [31]. The effect of this depletion on the mechanical strength has not yet been explored. While IN625 is designed to be strengthened via solid-solution hardening, suggesting that removal of solutes may lead to a reduction in strength, it is reported that formation of precipitates serve to increase the strength and reduce the ductility of wrought IN625 [28]. Careful investigation of the effect of precipitation on the mechanical strength in AM IN625 is required.

Fig. 10.

Time-dependent evolution of lattice parameter of the FCC matrix, acquired from an in situ XRD experiment at 870 °C. A monotonic decrease in the matrix lattice parameter is observed, which suggests that heavy elements such as Nb and Mo are gradually depleted from the solid-solution matrix during the formation of the δ phase precipitates.

Using the temperature-dependent kinetic rates in Table 3, we performed an Arrhenius analysis, which is shown in Fig. 11. Through this analysis, we calculated an activation energy for the growth of δ-phase precipitates in AM IN625 at (131.04 ± 0.69) kJ mol⁻¹, which translates to (1.359 ± 0.007) eV/atom. This energy is on the same scale of activation energies for the growth of γ′ and γ” phases in wrought Inconel 718 alloy [84], which suggests that our calculated activation energy is indeed related to the growth of δ-phase precipitates. We note that our calculated activation energy is significantly lower than the activation energy of Nb diffusion in Ni, identified at (202.6 ± 4.7) kJ mol⁻¹ [85], indicating that the growth of the δ phase precipitates is not limited by long-range Nb diffusion. Interestingly, the SEM results in Fig. 5 clearly demonstrate that the initial formation of the δ-phase precipitates resides at the inter-dendritic region, where the mass fraction of Nb is high, hence the long-range diffusion of Nb is not required for the formation of δ phase. Again, this points to the compositional heterogeneity as the root cause that facilitates a rapid precipitation of the δ-phase precipitates.

Fig. 11.

An Arrhenius analysis of the temperature-dependent kinetic rates shows that the activation energy of δ phase precipitate growth in AM IN625 is (131.04 ± 0.69) kJ mol⁻¹.

3.4. Thermodynamic calculations

We used thermodynamic calculations to better understand the experimental results, particularly those pertaining to precipitate evolution. As Figs. 3 and 4 demonstrate, the elemental segregation causes the local composition of the AM material to significantly deviate from the nominal values of its wrought counterpart. Thus, thermodynamic analysis over a range of compositions is required.

Fig. 12 shows the calculated equilibrium isopleth phase diagram over a range of Nb mass fraction from 3.0% to 10.0%. In this calculation, the composition in Table 1 was used. When varying the Nb mass fraction, we replaced Nb with Ni while keeping the mass fractions of other elements constant. The dashed line highlights the Nb mass fraction of the stock powder at 3.75%, as reported in Table 1. It is readily seen that for the temperature range investigated, the phase diagram predicts possible phases of FCC matrix, MC carbide, M₂₃C₆ carbide, ^δ phase, and σ phase. All were experimentally observed other than the σ phase and M₂₃C₆ carbide, which may be due to the slower kinetics of these more complex phases. For example, previous XRD studies of the long-term aging of IN625 in the temperature range between 650 °C and 870 °C for up to 46,000 h also failed to identify the presence of intermetallic σ phase [33], which indicates that σ phase may be a very slowly developing phase even though it is thermodynamically possible. That δ is an equilibrium phase for the IN625 composition in this temperature interval is consistent with previous studies of wrought IN625 alloys after prolonged (hundreds of hours) heat treatments [20,32,33]. It is also noted that δ is stable for the whole Nb range (3%–10%) at temperature below the FCC + MC phase region (purple line) shown in Fig. 12. Hence, from a thermodynamic point of view, δ is expected to form both for the nominal composition and locally in segregated areas. Nevertheless, the equilibrium phase diagram cannot predict when and how the δ phase forms at these Nb concentrations. To elucidate the precipitation sequence and time scale, we performed TC-PRISMA precipitation simulations.

Fig. 12.

Calculated equilibrium isopleth phase diagram for IN625 alloy over a range of Nb mass fraction from 3.0% to 10.0%. The dashed line shows the average Nb fraction in the IN625 feedstock powder.

For the TC-PRISMA simulations, many parameters, such as composition, nucleation site types, interfacial energies, and diffusion coefficients, can affect the predicted precipitation kinetics. Our goal is to compare the precipitation rate dependence on the composition when everything else is equal. We assumed that the heterogenous nucleation events in AM IN625 occur on dislocations. Without loss of generality, we assumed a TC-PRIMSA default nucleation site density of 2.2 × 10²² m⁻³, which translates to a dislocation density of 5.0 ×10¹² m⁻². We included δ phase, γ” phase, MC carbide, and σ phase, i.e., equilibrium phases predicted by the phase diagram at 870 °C plus γ” phase. All the precipitates were assumed to be spherical in the simulations. The interfacial energies were calculated by TC-PRISMA based on the extended Becker model imbedded in TC-PRISMA [86] and parameterized by general interfacial energy considerations between precipitate phases and matrix phase. The interfacial energies were 0.06 J/m² for MC, 0.05 J/m² for δ, 0.01 J/m² for γ”, and 0.14 J/m² for σ. Here, we note that γ” phase is a metastable phase that can precipitate in Ni-based super alloys such as IN625, Inconel 718 or ATI 718+. It has a tetragonal (D0₂₂) crystal lattice and possesses a disc-shaped morphology. In wrought alloys, it is known that upon prolonged heat treatment at intermediate temperatures, the metastable γ” phase eventually transforms to the equilibrium δ phase [28].

Fig. 13 shows the simulated growth kinetics of these constituent phases at 870 °C in terms of volume fraction as a function of time for two compositions. The first composition is based on the nominal composition of the powder simplified to include only Ni, Cr, Mo, Nb, Fe and C (Ni-20.70Cr-9.00Mo-4.00Nb-0.72Fe-0.05C). In other words, this simulation aims to predict the precipitation behavior of the wrought alloy. Fig. 13(a) shows that at this composition, MC carbide forms immediately. The formation of γ” phase starts at approximately 1 h and reaches its peak volume fraction near 30 h. The formation of the δ phase starts at the peak volume fraction of γ” and its increase in volume fraction clearly comes at the expense of γ” phase being consumed, indicating a direct transformation from γ” to δ at this temperature. Fig. 13(b) shows the simulated growth kinetics for a segregated composition Ni-20.20Nb-11.00Mo-8.70Nb-0.05C in mass fraction as predicted by DICTRA solidification simulations for AM IN625 [22]. A comparison with the case of nominal composition (Fig. 13(a)) reveals a strikingly different kinetic landscape. While the transformation sequence persists, both γ” phase and δ phase emerge much sooner when compared with their wrought counterparts. The predicted peak volume fraction of γ” phase occurs at ≈ 1 min. The growth of δ phase starts at a similar time scale. We emphasize that the simulation presented in Fig. 13(b) only corresponds to one composition. AM IN625, as Fig. 4 indicates, has a wide range of local compositions. Hence, the realistic transformation kinetic time scale is expected to be a weighted sum of the calculated kinetic time scales according to different compositions, which may at least partially explain the discrepancy between our measured time scale (1/κ in Table 3) of ≈3.5 h at 870 °C and the corresponding simulated time scale of ≈0.5 h. Furthermore, the mathematical models in TC-PRISMA include numerous assumptions and unknown parameters and exact agreement with the experimentally observed phase evolution in complex AM materials cannot be expected.

Fig. 13.

TC-PRISMA simulations of the precipitation kinetics using (a) compositions of starting IN625 feedstock powder and (b) a typical composition identified in AM IN625 alloy where Nb concentration exceeds the specified upper limit in IN625.

3.5. Homogenization heat treatment

While the metastable γ” phase serves as a strengthening phase, transformation from γ” to δ phase is known to reduce the strength of γ” strengthened alloys [26,87]. Furthermore, δ phase, developed along the grain boundaries, can significantly reduce the fracture toughness and ductility of IN625 [23,88]. It is, therefore, of interest to eliminate the deleterious δ phase using a homogenization heat treatment.

As an initial attempt, we conducted a second heat treatment at 1150 °C for 1 h for a specimen that was first subjected to an 870 °C for 1 h heat treatment to promote the formation of δ phase. The synchrotron XRD result of the alloy after the heat treatments is shown in Fig. 14. Here, we identified a single-phase FCC structure, demonstrating complete dissolution of δ phase precipitates. The FCC lattice parameter is 3.6160 (1) Å, greater than the matrix lattice parameter of 870 °C for 8 h sample at 3.5993 (1) Å. This result suggests that the heavy elements that make up the δ phase have become part of the solid solution of the homogenized matrix. This result is in agreement with our previous SEM and SAXS observations, where we observed an homogenized elemental distribution after a 1150 °C for 1 h heat treatment [24].

Fig. 14.

High-resolution synchrotron XRD of AM IN625 after a homogenization heat treatment of 1 h at 1150 °C, preceded by a stress-relief heat treatment of 1 h at 870 °C. All observed peaks are associated with the FCC matrix with a lattice parameter of 3.6160 (1) Å.

It is also worth noting that within the instrumental detection limit, we did not observe characteristic XRD peaks associated with the MC carbides in the XRD data for the homogenized IN625 sample. This result is consistent with the prediction of the equilibrium phase diagram as shown in Fig. 12, where only FCC is expected to be present at 1150 °C.

Finally, we underscore that a homogenization heat treatment, while effective at removing the δ phase, also promotes grain growth, hence potentially adversely affecting the mechanical strength of the homogenized alloy. A more in-depth study of the homogenization heat treatment of AM IN625 is currently under way, and will be reported separately.

4. Conclusions

In this paper, we performed a detailed analysis of the phase landscape and precipitation kinetics of AM IN625 alloy at temperatures relevant to stress-relief heat treatments and highlighted the role of elemental segregation, ubiquitous to AM process, on microstructural evolution. This often-overlooked topic has critical implications for the evolution of microstructure and mechanical performance of AM alloys in general. Our approach combines ex situ electron microscopy and high-resolution and high-sensitivity synchrotron X-ray diffraction, in situ synchrotron X-ray scattering and diffraction, and thermodynamic modeling, which allows us not only to monitor the precipitation kinetics but also to understand and validate the fundamental thermodynamics of the observed phenomena.

We found that elemental segregation in AM materials is the root cause for the unusual precipitation behavior in AM IN625. Heavy elements such as Nb and Mo in as-built IN625 are locally segregated in interdendritic regions, where the mass fraction of Nb and Mo are noticeably beyond the bounds set by the allowable range for wrought IN625. This result suggests that while the starting powder feedstock may have the designed chemical composition, localized departure from the nominal composition in AM materials is expected and its consequence for the materials properties must be considered.

Following heat treatment, we observed the inhomogeneous formation of δ phase precipitates and MC carbides, with initial formation of δ phase located in the interdendritic regions. Analysis of in situ SAXS data reveals a preferential growth of platelet-like δ phase precipitates along their lateral (long) dimension while the thickness (short) dimension quickly reaches a plateau. The growth of δ phase precipitation during heat treatment is continuous and temperature-dependent, with an activation energy for growth of (131.04 ± 0.69) kJ mol⁻¹. The incubation time for the precipitation of δ-phase precipitates is shorter than 5 min suggesting a low nucleation barrier. This incubation time in AM IN625 is at least two orders of magnitude smaller than that in wrought IN625. The experimental observations are confirmed by the calculated equilibrium phase diagram and TC-PRISMA based precipitation simulations. In particular, TC-PRISMA simulations unequivocally demonstrate that local elemental segregation is responsible for the observed fast precipitation of the deleterious δ phase precipitates. We further demonstrate that, as an initial attempt, a homogenization heat treatment can effectively remove the δ phase precipitates as well as homogenize the AM IN625 alloy.

Finally, we emphasize that while the results reported in this paper are specific to AM IN625 alone, we expect that the general conclusions hold for many other types of AM alloys. Elemental segregation in as-built AM alloys is inherent to AM technologies, which leads to the local composition of AM parts not being within the specifications set for their wrought counterparts. As our study clearly demonstrates, this can lead to formation of unwanted precipitates on a time scale much faster than in wrought alloys. Alloy-specific strategies must be developed to mitigate this effect, either through experimental exploration or numerical modeling.