In situ TEM Characterization of Microstructure Evolution and Mechanical Behavior of the 3D-Printed Inconel 718 Exposed to High Temperature

Abstract This in situ transmission electron microscopy work presents a nanoscale characterization of the microstructural evolution in 3D-printed Inconel 718 (IN718) while exposed to elevated temperature and an associated change in the mechanical property under tensile loading. Here, we utilized a specially designed specimen shape that enables tensile testing of nano-sized thin films without off-plane deformations. Additionally, it allows a seamless transition from the in situ heating to tensile experiment using the same specimen, which enables a direct correlation of the microstructure and the mechanical property of the sample. The method was successfully used to observe the residual stress relaxation and the formation of incoherent γ′ precipitates when temperature was increased to 700°C. The subsequent in situ tensile test revealed that the exposure of the as-printed IN718 to a high temperature without full heat treatment (solutionizing and double aging) leads to loss of ductility.


Introduction
Additive manufacturing (AM), commonly referred to as 3D printing, is expected to be a game-changer for the manufacturing industry (Aydinöz et al., 2016). The advancement of AM processes such as laser powder-bed fusion (LPBF) enables direct 3D printing of metal alloys. The LPBF method uses a scanning laser to melt or sinter (cohesive bonding) a powder-bed layer-by-layer to create complex and intricate three-dimensional geometries . This technology, compared to conventional manufacturing methods, has advantages such as design flexibility, thus evading the limiting factors such as production quantity, lead-time, and dimensions (Wang et al., 2012;Aydinöz et al., 2016;Trosch et al., 2016;Kuo et al., 2017b).
Several studies have been conducted to compare the mechanical properties of materials produced by LPBF and the traditional production methods involving casting and thermo-mechanical working (Wang et al., 2012;Zhang et al., 2015;Trosch et al., 2016;Kuo et al., 2017a). The distinct melting/solidification dynamics of LPBF is one of the biggest challenges in successfully implementing the AM technology. LPBF involves rapid heating and cooling, and together with the small laser spot size, it leads to the formation of nonequilibrium phases with characteristic microstructures (Kelly & Kampe, 2004;Mumtaz et al., 2008;Thijs et al., 2010;Song et al., 2015;Li et al., 2016). Additionally, the microstructure of the LPBF-produced alloys can vary within the built sample (Baufeld et al., 2010;Thijs et al., 2015) because of the variation in the cooling rate from the rapid cooling of the melt pool at the top part to the slowcooled region at the lower part (or the region processed at the earlier stage). It is also highly dependent on the build direction. Columnar grains are commonly observed along the build direction in the LPBF alloys (Amato et al., 2012;Vilaro et al., 2012;Mostafa et al., 2017) causing anisotropy in the mechanical properties (Zhao et al., 2008;Frazier, 2014;Song et al., 2015;Strößner et al., 2015;Deev et al., 2016). To achieve the desirable mechanical properties and thus meet the design specifications, the control of the microstructure by adequate solution heat treatment and aging is crucial (Zhou et al., 2019). There are many parameters that affect the microstructure and, consequently, the mechanical properties of the LPBF-produced alloys. This includes, but is not limited to, laser power and spot sizes, scan speed and direction, hatch spacing, and slice thickness, to name a few. By changing these parameters, the microstructure and the mechanical properties can be altered even for the same alloy composition. And the postproduction heat treatment gives additional parameters leading to the fabrication of materials with a wide range of ductility and mechanical strengths (Ding et al., 2016). It is extremely challenging to optimize the entire processing route in order to achieve the target properties. Understanding the fundamental mechanism of the microstructure development and accompanying mechanical properties are essential for the efficient development of the LPBF processes. However, the knowledge and experience from conventionally manufactured alloys may not be directly applicable to LPBF alloys due to the difference in the microstructure.
There have been numerous studies comparing LPBF and the traditional production methods, such as casting and thermomechanical working (Wang et al., 2012;Zhang et al., 2015;Trosch et al., 2016;Kuo et al., 2017a), and investigating the effect of the heat treatment on the microstructures and the mechanical properties of the LPBF alloys (Song et al., 2012(Song et al., , 2014Vrancken et al., 2012;Chlebus et al., 2015;Li et al., 2015Li et al., , 2016Zhang et al., 2015;Ter Haar et al., 2016;Tucho et al., 2017;Zhou et al., 2019). Components manufactured using LPBF can achieve mechanical properties close to those of the conventional techniques or even better. However, mechanical properties of the as-printed alloys may not satisfy the requirement of the highly demanding applications because of their characteristic microstructures. For example, due to the rapid heating and cooling cycles of the LPBF process, solute partitioning is prevalent, leading to micro-segregation and subsequent formation of brittle intermetallics such as Laves phase (Tucho et al., 2017). Here, postproduction heat treatment is the key to enhancing the uniform elemental distribution and improving mechanical properties by promoting the precipitation of strengthening phases while eliminating the segregation of undesired phases Tucho et al., 2017).
Such postproduction heat-treatment studies have been well established for the bulk samples produced by conventional techniques, but limited information is available for powder-bed AM-processed samples Sing et al., 2020;Tsai et al., 2020). The LPBF-produced alloy may undergo a different structural evolution because its initial characteristic microstructure is different from those produced by the conventional routes. It is reported that during laser manufacturing of the IN718, the alloying elements, especially niobium (Nb) and molybdenum (Mo), tend to segregate in the inter-dendritic/cellular region compared to the conventionally produced alloys (Tucho et al., 2017). An enormous number of research works have been documented in the study of heat treatment and its relation to the deformation behavior in metallic structures. In the literature, several experimental and theoretical investigations account for analyzing the mechanical behavior of the bulk sample (de Formanoir et al., 2019;Hilaire et al., 2019;Sing et al., 2020). Meanwhile, there are only a handful of studies on the LPBFprinted alloys at the submicron scale Shen et al., 2018;Bär et al., 2019).
Since the LPBF alloy has unique microstructures that vary within the material, fine-scale analysis is necessary to fully understand their mechanical properties. Here, in situ transmission electron microscopy (TEM) and scanning electron microscopy (SEM) experiments have been widely used to capture the deformation and the fracture mechanics in the materials at nano-and atomic scales (Rao et al., 2003;Haque & Saif, 2004;Greer & Nix, 2006;Kiener & Minor, 2011;Kushima et al., 2012;Imrich et al., 2015;. For example, compressive tests of additively-manufactured micro-pillars with several microns of diameter can be performed inside an SEM to analyze the force-deformation responses (Daryadel et al., 2019). Additionally, nanoindentation combined with SEM has been employed to study the mechanical properties of the LPBF-produced alloys and their spatial variations associated with the difference in the microstructures (Legros et al., 2011). The technique allows the quantitative evaluation of the mechanical properties of the very fine micro-phases found in the LPBF.
On the other hand, in situ TEM provides direct observations of the intrinsic deformation mechanisms and their relationships with the microstructures at higher spatial resolutions. It allows the characterization of the nanostructures within the microphases and analysis of their effects on the mechanical properties through real-time observation of the dynamic processes during mechanical deformations. The in situ TEM techniques have been used to investigate structural changes in materials during heating Wang & Haque, 2015;Rao et al., 2017;Callisti et al., 2018;Topping et al., 2018;Albu et al., 2020) as well as deformation mechanisms at nano-and atomic scales (Haque & Saif, 2004;Kiener & Minor, 2011;. Various in situ TEM mechanical testing methods were developed in the past decade, including nanocompression on focused ion beam (FIB) microfabricated pillars (Shan et al., 2008), bending tests using electric field-induced resonance (Wang et al., 2008), and microelectromechanical systems (MEMS) based on lithography for performing in situ TEM straining experiments (Wang & Haque, 2015). MEMS fabrication technology of a freestanding tensile specimen with the force sensor and actuation mechanism for the force/displacement measurement shows distinct capabilities (Haque & Saif, 2004). The applications of these techniques have been explored with tremendous interest. For example, Minor et al. investigated the onset of plasticity using a miniature capacitive load-displacement transducer equipped on the TEM holder (Minor et al., 2006), and Liu et al. in their work revealed the effect of the dislocation network formed during selective laser melting on the dislocation motion by in situ TEM analysis using a straining holder .
These capabilities of the in situ TEM/SEM are important for establishing a fundamental understanding of the LPBF-produced alloys. In fact, several recent studies employed these techniques to study the AM process and the materials. The rapid solidification process of polycrystalline aluminum was observed by TEM using an ultrahigh frame-rate CCD camera with single-electron detection sensitivity Zweiacker et al., 2016). Microstructure evolutions during the postprocess heating of the AM materials were investigated by in situ TEM for Ni-based superalloy (Deshpande et al., 2020), Ti-alloy (Li et al., 2021), and Al-alloy (Albu et al., 2020). For the mechanical properties of AM-produced materials, the effect of the microstructure on the strength (Wu et al., 2016;Liu et al., 2018;Lin et al., 2019) and fatigue behavior  was studied by in situ TEM/SEM. However, performing in situ heating and mechanical tests on the exact same specimen to directly link the structural evolution and the mechanical property is still a challenging task.

Experimental Methodology
In this work, a unique in situ TEM procedure was developed to evaluate the microstructure evolution of LPBF-produced metal alloys when exposed to a high temperature and its effect on the mechanical property. This real-time experiment provides important insights regarding the evolution of the microstructure at different heating conditions and how the associated changes in the microstructures affect the mechanical properties. One of the biggest challenges during the in situ testing is maintaining the mechanical integrity because nanoscale samples are so thin. A significant amount of stress accumulates while thinning down using a FIB, which makes the sample intrinsically more fragile. This challenge is further complicated by leveling the thin testing sample onto the tensioning mechanism to ensure that the pure tension is applied to the sample. Otherwise, other forces such as torsion or bending may occur in addition to the tension. The fundamental principle is to measure the displacement of the tensile sample accurately . Additionally, even a slight misalignment in the specimen orientation and loading direction can result in unaccounted flexural stresses on the specimen (Haque & Saif, 2002). Therefore, in order to seamlessly observe the change in the microstructure during heating and to conduct the mechanical test on the exact same sample inside the TEM, we designed a push-to-pull TEM sample configuration as shown in Figure 1. With the two ribs at the ends of the thin center part, off-plane motions such as bending and twisting of the film are restricted. When a pushing force is applied on the protruding part on one of the ribs with the thinned root, the rib bends away from the support on the other side of the center lamella, and the pushing force is converted into the tensile load on the center of the specimen. Because the off-plane motions are constrained by the ribs, a tensile deformation of the observation area can be achieved without undesired bending and twisting. Here, a notch was introduced as a stress concentration site to nucleate a crack and initiate the fracture, since it is very difficult to observe a specific feature at high magnification (Wang & Haque, 2015). A finite element analysis was conducted to confirm that the offset of the force does not affect the stress distribution at the crack tip (see Supplementary Material for details).
A specimen block with dimensions of 5 μm × 2 μm × 2 μm was cut from a LPBF-produced alloy and shaped by FIB using a FEI 200 TEM FIB to the push-to-pull geometry as shown in Figure 1. The detailed procedure is explained in Supplementary Figure S4. FIB current was carefully controlled to minimize the beam damage that may affect the mechanical property of the specimen (Kiener et al., 2007). The sample was heated inside the TEM using a heating holder (Gatan Model 652) for residual stress relaxation; the heated sample was then transferred to a 3D-nanomanipulation holder (Hummingbird Scientific) for tensile testing. A tungsten probe was used to apply the quasi-static displacement on the freestanding end of the FIB-prepared end of the specimen.
IN718 fabricated by LPBF was used as the subject of the study. The pre-alloyed IN718 gas-atomized powder with a diameter ranging from 10 to 45 μm was built into cubic samples by scanning laser manufacturing (SLM) 125HL LPBF unit. A laser beam of 200 W at a scanning speed of 900 mm/s in an argon atmosphere was used to fuse the powder particles. The details are recorded elsewhere (Zhou et al., 2019). First, an in situ heating experiment was conducted to confirm that the microstructure evolution can be captured with our setup. A thin TEM specimen was prepared using FIB. The specimen was mounted on the in situ heating holder, and the temperature was raised inside the TEM to analyze the nucleation and the growth of precipitates (γ ′ , γ ′′ ). First, the temperature was increased from room temperature (∼20°C) to 600°C at a heating rate of 50°C/min. The increase in temperature caused a sudden drift due to thermal expansion. After the drift was stabilized for 20 min, the temperature was raised to 700°C at a slower rate of 10°C/min.

Results and Discussion
Microstructural Evolution under Heating Figure 2 shows the structural evolution of precipitates while heating (see also Supplementary Movie 1). The as-printed IN718 sample had a significant amount of residual stress due to fast heating/ cooling processes in SLM. As the sample was heated, the stress was relieved (ii-iv) and the strain contrast was no longer visible at 640°C (iv). At 680°C, small precipitates began to nucleate in the matrix as indicated by the arrowheads (v). As the temperature increased, the precipitates continued to nucleate and grow introducing local strains in the material as shown in (vi). Electron diffraction patterns before and after the heating experiment were analyzed to confirm that the precipitates were γ ′ and γ ′′ phases (see Supplementary Figs. 6 and 7). Furthermore, increasing the temperature to 800°C (at a rate of 10°C/min) resulted in the formation of a needle-like δ-phase (see Supplementary Fig. 5(d)). However, no Laves phases were found in our study, which agrees with Amato et al. (2012). γ ′′ phase is metastable which transforms into δ-phase with similar composition and acicular structure at high temperature. This acicular δ-phase conversion reaction is accelerated around 800°C and would be highly undesirable because of the δ-phase embrittlement. The amount of δ-phase that formed in the sample was comparatively low because the IN718 alloy studied in this work has Nb in the range of 4.75-5.50 wt%, while it should be more than 7% for significant distributions of δ-phase (Zhang et al., 2010;Kuo et al., 2017aKuo et al., , 2017b. The above observations indicate that the microstructure evolution of the IN718 at high temperature can be captured within a thin TEM specimen used in this work.

Deformation and Fracture Behavior
To analyze the effect of the microstructure evolution on the mechanical properties, two specimens were cut from a single grain in as-printed IN718 within a few micrometers of each other and shaped into the push-to-pull samples as shown in Figure 1. Here, the tungsten probe was pushed against the sample at a displacement rate of 1.0 nm/s to apply tension to the thin observation area. Figure 3 shows the fracture process observed for the IN718 sample before heating (see also Supplementary  Movies 3 and 4). Tensile deformation was successfully applied to the observation area, and the fracture process was captured. As the tension was applied to the specimen, the notch started to open wider (Figs. 3a and 3b). When further strain was imposed, a crack slowly nucleated from the surface as indicated by the arrowhead (Fig. 3c). Then, the crack started to open and propagated through the specimen (Figs. 3d-3f). At the latter stage of the deformation just before the fracture (Figs. 3g-3i), necking was observed, indicating a ductile fracture. After heating to 800°C in a same manner as described above (direct heating in TEM), the fracture process of the specimen showed completely different behavior as shown in Figure 4 (see also Supplementary Movies 5 and 6). A crack initiated from the notch at much smaller strain compared to the as-printed sample (Figs. 4a-4c). Once the crack was nucleated, it propagated rapidly across the specimen (Figs. 4d-4f) and led to the final fracture (Fig. 4g). The specimen showed no necking after the fracture (Fig. 4h), and the higher magnification images of the rough fracture surface indicated that the process was brittle.
As a comparison, a push-to-pull tensile specimen was cut from the "fully heat-treated" IN718 (Zhou et al., 2019). It was solution heat-treated at 1,065°C for 1 h followed by the two-step aging at 720°C for 8 h and 650°C for 7 h, with γ ′ /γ ′′ precipitates coherent with the γ matrix. This sample, hereafter, is referred to as fully heat treated (FHT). As shown in Figure 5, the deformation behavior was different from the as-printed and directly heated specimen as discussed above. At the initiation stage, a network of nanosized cracks was formed near the notch, forming 5-10 nm grains in the region marked by the dashed square (Fig. 5b). The detailed view of the crack nucleation site is shown in Supplementary  Figure 8. The size of the grains matches well with the coherent Fig. 2. Temperature change and the corresponding microstructure evolution of SLM-produced IN718 during in situ heating experiment. The sample morphology at room temperature before heating is shown in (i) as a reference. The residual stress in the as-produced sample was relaxed (ii, iii) and completely removed at 640°C (iv). At 660°C, precipitates started to nucleate as indicated by the arrowheads (v) accompanying local stresses as can be seen from the strain contrast. More precipitates nucleated and grew as the temperature was increased (vi, vii). When the sample was held at 700°C, the precipitates were uniformly distributed in the sample and the stress had been relaxed (viii).  γ ′ and γ ′′ precipitates. As the tension continued, the crack propagated rapidly (Figs. 5c-5e). However, it did not lead to a complete fracture as in the directly heated specimen. Instead, it showed necking before the fracture as indicated by the arrowheads (Figs. 5f-5h). These results demonstrate the important correlation between heat treatment, microstructural constituents (i.e., γ ′ / γ ′′ precipitates and δ-phase), and mechanical properties (i.e., strength, ductility, and fracture toughness). To quantify the ductility of the LPBF-produced IN718 before heating (i.e., as-built), after direct heating, and the proper two-stage heat treatment, the relationship between the crack length and the notch opening was analyzed. The initial notch size l 0 is indicated by the solid lines as shown in Figures 3a, 4a, and 5a. The features in the samples were tracked to determine the evolution of the notch size and the crack length. Figure 6 shows the relationship between the crack length and the notch opening, Δl = l−l 0 . (Relationships between crack length, notch opening, and the displacement of the indentor are shown in Supplementary Fig. S9.) The crack developed slowly for the as-printed sample, and the length increased linearly as the notch opened. The directly heated sample showed a rapid increase in the crack length at a much smaller notch opening that corresponded to the brittle fracture. On the other hand, the crack length increased slowly for the FHT sample during its propagation, showing a ductile behavior. While both directly heated and FHT samples improved their strength due to the precipitation hardening, only the latter specimen with coherent precipitates maintained its toughness. Since the specimen was prepared from a single grain, the samples tested were single crystal. Therefore, the effect of needle-like δ-phase segregation at the grain boundary that restricts the dislocation motion (Trosch et al., 2016;Kuo et al., 2017a) was not present in the as-printed IN718 specimen. This enables dislocations to propagate freely and allows easy gliding of the crystallographic planes leading to the ductile deformation (see Supplementary  Fig. S10). On the other hand, the abrupt fracture at the lower strain in the sample after direct heating to 800°C is attributed to the improper heat treatment, which caused a segregated formation of precipitates leading to several nucleation sites for catastrophic failure (Zhou et al., 2019). Here, the embrittling precipitates formed in the specimen after heating causes dislocation pinning, which restricts the sliding of the crystallographic planes and causes brittle fracture. A similar ductile-to-brittle transition from the as-printed IN718 alloy after heating was reported for the bulk specimen (Popovich et al., 2017). The reason attributed to this behavior was the large amount of needle-like δ precipitates and undissolved Laves phases. Our observation demonstrated that the matrix grain itself also loses ductility as incoherent γ ′ / γ ′′ phases precipitate.

Conclusion
Great attention has been paid to LPBF production and postproduction heat treatment parameters to produce desired microstructure and eliminate pores. Our observation confirms that as-printed IN718 is less suited for high-temperature applications, and that a simple postprint heating (e.g., direct heating) cannot ensure its mechanical properties. However, further improvements may be possible by tailoring the nanostructures within the grains. For example, the hardness of the LPBF-produced IN718 was improved by the precipitation of coherent γ ′ and γ ′′ precipitates following the full heat treatment (Zhang et al., 2010;Kuo et al., 2017b;Zhou et al., 2019). An effectively homogenizing solution treatment is necessary for subsequent precipitation hardening to properly take place while preventing the microsegregation of incoherent precipitates within the grains that lead to the embrittlement of LPBF-produced IN718. Although this work focused on LPBF-produced IN718 alloy and evaluated the relationships between microstructural evolution and the mechanical properties, the unique technique developed in this work is not restricted to the field of LPBF, but can be widely used to study the mechanical  behavior of materials, including ceramics, biomaterials, and semiconductors, where the relationship among processing, nano/ micro-structure, and mechanical properties elucidates the performance of engineered components and systems.