Microstructure and Crystallographic Texture Evolution of β-Solidifying γ-TiAl Alloy During Single- and Multi-track Exposure via Laser Powder Bed Fusion

The raw powder of β-solidifying γ-TiAl alloy was fabricated using gas atomization (Osaka Titanium Technologies, Japan), and the nominal chemical composition of the powder was determined to be Ti (balance), Al (43.77), Nb (5.96), and Cr (1.22) (at%) through inductively coupled plasma optical emission spectroscopy. As shown in Fig. 1a, the powder is spherical, and the surface exhibits a dendritic morphology. Furthermore, quantitative analysis of the powder size distribution was performed using a laser diffraction-type particle size distribution measuring device (Mastersizer 3000E, Malvern Panalytical, UK). The obtained volume equivalent diameter D₅₀ was 36.0 μm, which is within the range of 10–45 μm (Fig. 1b). A high-circularity powder is expected to result in a high part density owing to its good flowability and high powder layer density.

Single- and multi-track specimens were fabricated using L-PBF (EOS M290, EOS, Germany) on a cast material with the same composition as that of the powder. As shown in Fig. 1c, a 60-μm-thick powder bed was piled up on the casting material, and the used laser power, scanning speed, and hatching space were 180 W, 400–1200 mm/s, and 10, 50 μm, respectively. In the case of the multi-track process, a bidirectional (zigzag) scanning strategy was applied.

The topography of each track was observed via optical microscopy (OM; BX60, Olympus, Japan). For the cross-sectional plane observation, the single-track and multi-track specimens were cut in the transverse direction along the scanning direction and mechanically polished with grinding paper and suspended to the surface of the mirror. The melt pool shape, constituent phase, composition distribution, and crystallographic orientation were determined via field-emission scanning electron microscopy (FE-SEM; JIB-4610F, JEOL, Japan) coupled with energy dispersive X-ray spectroscopy (EDS, X-Max^N, Oxford Instruments, UK) and an electron backscatter diffraction system (EBSD; NordlysMax³, Oxford Instruments, UK). The mechanical properties were evaluated using a nanoindentation tester (ENT-1100a; ELIONIX, Japan). A Berkovich tip was used, and the maximum indentation load and hold time were set to 4.9 mN and 10 s, respectively.

To further understand the microstructural evolution in the β-solidifying γ-TiAl alloys during the L-PBF process, the thermal diffusion behavior and instantaneous temperature transition were estimated via finite element thermal model simulation (COMSOL Multiphysics 6.1, COMSOL Inc., Sweden), following the methodology introduced in the Ref. [30, 31]. In the simulation, the thermal flux of the laser beam was assumed to satisfy a Gaussian distribution[31, 32]. The energy distribution equation can be expressed as follows:where P is the laser powder; R is the radius of the laser spot; r is the distance between the center of the laser spot and the powder bed surface; A is the rate of laser absorption; H is the laser penetration depth; and z is the depth. In addition, to ensure reliability, the simulated size and shape of the melt pool were compared to the actual size. The physical material properties used in the simulation are listed in Table 1.

The microsegregation was simulated via the Thermo-Calc software with TCTI4 and MOBTI4 database. The Scheil–Gulliver model with the solute trapping algorithm, which was built in the Thermo-Calc software, provided an estimation of the solidification in the material according to non-equilibrium conditions, such as the additive manufacturing process. In general, the solidification behavior is sensitive to the solidification rate, which can be calculated from the laser scan speed of a single track by multiplying the angle between the direction of the laser source and the vector normal to the direction of solidification [33]. The calculated solidification rate was approximately 0.56 m/s based on the results obtained from a single track with a laser power of 180 W and scan speed of 800 mm/s.

Figure 2 shows the top surfaces of the single tracks under a fixed scan power and varying scan speed. All process conditions commonly formed cracks in the direction perpendicular to the scanning direction, owing to a large thermal stress below the ductile–brittle transition temperature during rapid solidification [34]. Additionally, as the scan speed increased, the width of the single tracks decreased, and the morphology under moderate scan speed conditions (800 mm/s) showed good continuity and uniformity, whereas low (400 mm/s) or high (1200 mm/s) scan speed conditions induced partial adherence of particles or droplets when the laser power was maintained. The stability of a single track is sensitive to process parameters [35, 36]. The combination of laser power (P) and scan speed (v), the linear energy (P/v), plays a key role in forming consistent and continuous scan tracks, and it should deliver sufficient energy to melt the powders. However, an excessively low scan speed (high energy) generates a melt pool larger than the laser diameter, and the interior of the melt pool is divided into hot and cold regions [37]. The hot region is located under direct laser irradiation, and the existing powder in this section melts completely, whereas the cold region corresponds to the remaining part between the direct laser-irradiated area and melt pool width; therefore, partially melted powders are expected to be attached (400 mm/s). However, a high scan speed (low energy) promotes the instability of the molten track, which causes the splashing of liquid droplets from the surface of the molten track (1200 mm/s) [38, 39].

Figure 3a shows the variation in the melt pool size as a function of the scanning speed under the conditions corresponding to the previous observation. Similarly, the depth (D) and width (W) of the melt pool exhibited a gradually decreasing tendency when the scan speed increased, correlating with the decrease in linear energy. In addition, the melt-pool geometry of the single tracks was classified based on the D/W ratio. With the decreasing scan speed, the melt-pool geometries sequentially changed from keyhole (R > 1) to conduction mode (R close to 0.5). Representative cross-sectional images depending on the melt pool geometry are shown in Fig. 3b. The deposition and remelting zones can be distinguished from the cross-section of the melt pool; both regions were observed to be larger under higher linear energy conditions. In terms of attaining the structural integrity of the final products, the accuracy of the part dimensions may be hindered by an overly large deposition zone, whereas an insufficiently small remelting zone leads to weak cohesion between the underlying and adjacent parts, resulting in a lack of fusion. Therefore, to ensure the structural integrity of the final products, the appropriate process parameters were carefully considered.

For the investigation of the microstructural characteristics, the most stable condition, that is, moderate scan speed (800 mm/s), was selected. Figure 4a and b illustrate the numerically simulated melt pool geometry with the heat flow direction and cross-sectional SEM-BSE image of the melt pool cross-section. Evidently the thermal gradient direction is normal to the melt-pool boundary, and is strongly influenced by the melt-pool shape. Therefore, predominantly ± 45° inclined solidified microstructure developed in the melt pool boundary, while the direction of solidification gradually becomes closer to the vertical at the melt pool center [40].

Furthermore, the SEM-BSE mode provides a contrast image based on segregation in the chemical composition, where the heavy elements exhibit bright contrast, while the lighter elements exhibit dark contrast. Interestingly, a dark contrast of the substructure with irregular morphology was visible, indicating the aggregation of Al at the melt-pool boundaries (Fig. 4c). Indeed, evaluation of the EDS measurements revealed a discrepancy in the chemical composition between the bright and dark contrast regions, as presented in Table 2. A similar local chemical difference phenomenon was consistently observed in rapidly solidified γ-TiAl alloys[14, 16]. Subsequent investigation of phase distribution utilizing the SEM-EBSD (Fig. 4d) revealed that the substrate (casting) part was predominantly composed of the γ phase, while the inside of the melt pool was divided into two regions: (1) the band-like α₂ phase with the blocky morphology in the melt pool boundary (Dark region) and (2) the β phase in the melt pool center (Bright region). The blocky morphology implies a massive transformation occurred at the melt pool boundary and the thickness of the band-like α₂ phase region was approximately 3–20 μm range. These unusual constituents of the microstructure suggest that microstructural evolution is significantly affected by the cooling rate during solidification.

As demonstrated in the pseudo-phase diagram in Fig. 5, the as-cast β-solidifying γ-TiAl alloys undergo numerous phase transformations under the solidification process, the phase transformation sequence is L → L + β → β → α + β → α (Blue line in Fig. 5). Thus, in the substrate (casting) region, the constituted microstructure is represented as the α₂/γ lamellar colonies and segregated β phase [12]. However, the phase transformations in the L-PBF process are quite different. First, the cooling rate variation in a specific region across the melt pool was investigated. Herein, a (Δy, Δz) Cartesian coordinate system is applied to indicate the location in which the cooling rate is measured. The reference point (0, 0) is located at the top center of the melt pool, the Δy is the melt pool width direction from the reference point, and Δz is the variation in the melt pool depth direction, respectively. Commonly, the center of the melt pool shows the highest cooling rate, which gradually decreases with increasing distance from the center of the melt pool (Fig. 6a and b) because the low thermal conductivity of the TiAl alloy induces heat dissipation through the top surface [15]. The decreased cooling rate at the melt pool boundary region provides relatively more time during the solidification, which further induces β to α₂ phase transformation. In contrast, the melt pool center experienced rapid cooling, remaining β phase owing to suppressed phase transformation [41]. In addition, as expected from the high cooling rate, the L-PBF process did not allow sufficient time for the diffusion of the solute. The Scheil-Gulliver solidification simulations predicted microsegregation during rapid solidification. As illustrated in Fig. 6c, the simulation results expected the formation of the β, α₂, and γ phases. Furthermore, the β phase is firstly solidified from the liquid during the solidification, the composition of Al in the liquid gradually increased and reached 49.7% of Al concentration at the initiating stage of the α phase precipitation (Fig. 6d). The different solute diffusivities between the solid and liquid phases led to intensive microsegregation in the cell tip and intercellular regions [15]. Consequently, at the melt pool boundary region, the solidification path is shifted to a higher Al composition, and the massive type of phase transformation is proposed through the single α phase region (Green line in Fig. 5). Meanwhile, the martensitic phase transformation (red line in Fig. 5) occurred at the melt pool center because the region far from the initially nucleated microstructure remained in a perfectly mixed equilibrium liquid state.

The crystallographic texture was investigated based on the microstructural observations. Figure 7a presents the EBSD inverse pole figure (IPF) maps of the corresponding melt pool with previous microstructure observations. In addition, the phase map and IPF maps were obtained at high magnification (Fig. 7b and c). As illustrated in pole figures (PFs) result in the different regions (Fig. 7d), the crystallographic texture of the β phase in the melt pool center (region 1) exhibited ± 45° inclined <100> alignment, which is well matched with the thermal gradient direction represented in the simulated results (Fig. 4a) and generally observed in the cubic based metal alloys [19, 42]. Conversely, the β and α₂ phases at the melt pool boundary interestingly exhibited an identical crystallographic orientation with the substrate (Region 2 and 3). In general, the massive transformation orientation has no specific relationship with the parent and child phases [27]. However, massive epitaxial growth was observed in laser-welded stainless steel when massive growth occurred from a previously existing region of the child phase [28, 29], and the first nucleated massive child grains presented an orientation relationship with the parent phase [18], thus corresponding to the findings of this study.

Figure 8 schematically shows the microstructural evolution and how massive epitaxial grain growth occurred during the L-PBF process based on the results demonstrated in this study. As illustrated in Fig. 8a, when the solidification is initiated from the melt pool boundary, the solidification of the primary β phase results in increasing the concentration of Al composition in the remaining liquid. From the perspective of the crystallographic texture, the β phase, which preferentially grows along the <100> crystal direction, will grow epitaxially with a close alignment to the thermal gradient vector inherited from the seed crystal on the substrate [43, 44]. In addition, the concentration of Al was intensively enriched at the cell tip and intercellular spaces owing to the different diffusivities in the solid and liquid phases, as depicted in Fig. 8b. Afterward, under the continuous cooling state, the α phase nucleated in the interdendritic region following the Burgers orientation relationship, and the crystallographic orientation is arrested during the initial stage of α phase precipitation [29]. Finally, the high Al concentration induces the α phase transformed at the melt pool boundary toward consuming the primary β phase while crystallographic orientation was maintained. However, martensitic transformation occurred in the melt-pool center region, which experienced a higher cooling rate (Fig. 6a and b) than the melt-pool boundary (Fig. 8c). Moreover, there was no chemical disequilibrium in the melt pool center, and the crystallographic texture varied according to the change in the thermal gradient direction.

In the previous observation of the single-track, the melt pool boundary interestingly consisted of the massively transformed α₂ phase and its crystallographic texture inherited from the substrate through epitaxial growth of β phase and subsequent α phase nucleation following Burger orientation relationship. The thickness of the α₂ phase region approximately ranged between 3 and 20 μm. Thus, to investigate the microstructure continuity in terms of α₂ phase selection with the multi-track study, 10 μm (narrow) and 50 μm (wide) hatch spacing conditions were determined. Firstly, the different hatch spacing conditions showed different microstructural evolution tendencies, as illustrated in Fig. 9. The microstructure under the narrow hatching space condition closely resembled the previous single-track microstructure, which consisted of massive α₂ phase at the melt pool boundary. Whereas, the wide hatching space condition formed the mixture of the fine massive α₂ phase and acicular region, owing to the fluctuation of the temperature resulting from the adjacent scan track. In addition, Fig. 10 shows the EBSD phase, IPF maps, and PFs obtained using different hatching spaces under the same process conditions. As inferred from the microstructure observation, the phase of the melt pool consisted of the massive α₂ phase in the melt pool boundary and the martensitic formed β phase region in the melt pool center (Fig. 10a and d). Furthermore, note that the inherited crystallographic orientation from the substrate is continuously maintained in the narrow hatching condition presenting the same α₂ crystallographic orientations with the neighboring regions (Fig. 10b and c). Conversely, although the identical crystallographic texture was observed in the initial stage of the scan track (region 1, 2), the random crystallographic texture was developed (region 3,4) in the wide hatching space condition as the multi-track progressed (Fig. 10e and f), indicating difficulty to achieve epitaxial growth of α₂ phase between the scan tracks with wide hatching space.

Figure 11 schematically illustrates the microstructural development for narrow and wide hatch spacings. The previously observed crystallographic texture in the single-track melt pool represented the single crystalline at the melt pool boundary, which inherited the crystallographic texture from the substrate, and the crystallographic texture in the center of the melt pool continuously rotated along the thermal gradient direction changing the grain orientation, thus resulting polycrystal-like β phase region. However, the narrow hatching condition which was smaller than the epitaxially massive grown α₂ phase region reaching 20 µm as presented in the single melt pool was successful in providing the specifically oriented crystallographic texture to the adjacent track. Such inheritance of the crystallographic texture is well maintained at least within the massive α₂ phase region. However, in the wider hatching space condition, the differently oriented β phase grains act as seed crystals, resulting in the grains being competitively nucleated affected by the thermal cycling from the adjacent scan track and growing in different orientations. Therefore, inheriting a specific crystallographic orientation for the next scan track is difficult, which prevails in random crystallographic textures.

Finally, mechanical properties were evaluated using a nanohardness tester. As shown in Fig. 12a and b, the hardness of the melt pool center (β phase) is commonly higher than the substrate (α₂/γ). In the β-solidifying γ-TiAl alloy, the different crystal structures of each phase are attributed to differences in the hardness, the values are increased in the order of γ, α₂, α₂/γ, and β [45, 46]. Thus, the enhancement of the hardness in the melt pool center region compared to the substrate can be explained by the phase constitution, and the large residual stress resulting from the rapid solidification of the L-PBF process also increases the hardness[47]. Meanwhile, the hardness in the mixture of the fine massive α₂ phase and acicular was significantly higher in the wide hatching space condition. This can be attributed to the influence of small grain size and randomly distributed crystallographic texture. The representative indentation load-depth curves measured at the overlapped microstructure region and the previously observed microstructural characteristics with different hatching space conditions resulted in a distinct hardness behavior (Fig. 12c). Namely, the single crystalline-like α₂ massive region in the narrow hatching space condition low strength, and high ductility can be estimated from the lowest hardness and high penetration depth, respectively. However, wide hatching space conditions may result in high strength and low ductility. Consequently, this study demonstrated that appropriate hatching space conditions should be considered to satisfaction of specific mechanical requirements in the β-solidifying γ-TiAl alloys for the first time in the literature. Moreover, a detailed study of the microstructural evolution and its correlation with the mechanical properties of the bulk specimen is in progress and will be reported elsewhere.