Effect of solution treatment on the microstructure, phase transformation behavior and functional properties of NiTiNb ternary shape memory alloys fabricated via laser powder bed fusion in-situ alloying

Post-heat treatment is commonly employed to improve the microstructural homogeneity and enhance the mechanical performances of the additively manufactured metallic materials. In this work, a ternary (NiTi)91Nb9 (at.%) shape memory alloy was produced by laser powder bed fusion (L-PBF) using pre-alloyed NiTi and elemental Nb powders. The effect of solution treatment on the microstructure, phase transformation behavior and mechanical/functional performances was investigated. The in-situ alloyed (NiTi)91Nb9 alloy exhibits a submicron cellular-dendritic structure surrounding the supersaturated B2-NiTi matrix. Upon high-temperature (1273 K) solution treatment, Nb-rich precipitates were precipitated from the supersaturated matrix. The fragmentation and spheroidization of the NiTi/Nb eutectics occurred during solution treatment, leading to a morphological transition from mesh-like into rod-like and sphere-like. Coarsening of the β-Nb phases occurred with increasing holding time. The martensite transformation temperature increases after solution treatment, mainly attributed to: (i) reduced lattice distortion due to the Nb expulsion from the supersaturated B2-NiTi, and (ii) the Ti expulsion from the β-Nb phases that lowers the ratio Ni/Ti in the B2-NiTi matrix, which resulted from the microstructure changes from non-equilibrium to equilibrium state. The thermal hysteresis of the solutionized alloys is around 145 K after 20% pre-deformation, which is comparable to the conventional NiTiNb alloys. A short-term solution treatment (i.e. at 1 273 K for 30 min) enhances the ductility and strength of the as-printed specimen, with the increase of fracture stress from (613 ± 19) MPa to (781 ± 20) MPa and the increase of fracture strain from (7.6 ± 0.1)% to (9.5 ± 0.4)%. Both the as-printed and solutionized samples exhibit good tensile shape memory effects with recovery rates >90%. This work suggests that post-process heat treatment is essential to optimize the microstructure and improve the mechanical performances of the L-PBF in-situ alloyed parts.

Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Introduction
NiTi-based shape memory alloys (SMAs) are attracting growing attention in biomedical, aerospace, and automotive industries due to their unique shape memory effect and superelasticity [1][2][3][4].Adding ternary elements could alter the transformation characteristics of binary NiTi alloys and expand their applications, especially in some cases that are incompatible with the binary NiTi alloys [3][4][5].For example, NiTiHf alloy has been developed for the applications at the temperatures above 373 K [4].Among the NiTi-based ternary SMAs, NiTiNb alloy exhibits a wide thermal hysteresis (>130 K) and excellent mechanical properties, which has been widely applied in sealing and coupling fields [3,5].Moreover, the excellent biocompatibility and non-toxicity of Nb also provide the potential for the biomedical applications of NiTiNb alloys [6][7][8][9].
However, traditional metallurgical techniques (e.g.casting, powder metallurgy) encounter challenges in fabricating NiTiNb alloys.The difficulties in casting lie in: (i) the compositional segregation caused by the large melting point difference between Ni (1726 K), Ti (1941 K) and Nb (2741 K) [10,11], and (ii) the formation of unfavorable impurities (e.g.Ti 4 Ni 2 O x ) caused by the high Ti affinity with oxygen [12].Powder metallurgy faces obstacles in controlling the porosity characteristics (shape and size) and fabricating complex parts [11,13].Additionally, poor machinability and weldability make it challenging to prepare complex NiTi-based structures through conventional processing methods [14,15].
Laser powder bed fusion (L-PBF), a layer-wise additive manufacturing technology, is capable of producing complex NiTi parts, such as scaffolds [16][17][18][19].L-PBF could also act as an in-situ alloying (ISA) method when using blended powders, which has been applied in the development of Ti- [20,21], Albased [22,23] and high entropy alloys [24].Compared with L-PBF of pre-alloyed powders, ISA enables flexible compositional and microstructural design, as well as shortened lead times.In the last decade, there have been many researches on ISA of binary NiTi SMAs via L-PBF using elemental Ni and Ti powders [25][26][27][28].The focus of these studies was on the effect of L-PBF process parameter on the densification behavior, phase composition, microstructure, phase transformation (PT) behavior and functional performances of the in-situ alloyed (ISA-ed) NiTi alloys.It has been reported that the ISA-ed NiTi alloys show a trade-off between microstructural heterogeneity and keyhole porosity when optimizing the process parameters [25].Moreover, the strong exothermic reaction of the Ni-Ti powder mixture will disturb the melt-pool, resulting in difficulty in obtaining the NiTi phase but producing unwanted secondary phases (e.g.Ni 3 Ti, Ti 2 Ni) [25,28].To avoid these issues, the ISA of ternary NiTi-X SMAs was mainly carried out via L-PBF using the pre-alloyed NiTi powders as the base materials and using the ternary element powders as additives, such as NiTi with HfH 2 [29], NiTi with Fe [30] and NiTi with Nb [8,[31][32][33].This ensures the formation of the NiTi phase, and the ISA-ed NiTi-based SMAs exhibit typical PT behavior and stress-strain response [8,[30][31][32].However, the intrinsic problems of the ISA-ed NiTi-based SMAs, such as microstructural inhomogeneity and the presence of unmelted powders [8,29,31,32], result in the poor mechanical/functional performances, particularly under tension.
Post-heat treatment has been commonly adopted to enhance compositional uniformity and modulate the microstructure, and thus stabilizes the service performance of the additively manufactured parts [21][22][23][24][34][35][36].Previous studies have shown that the mechanical/functional performances of the LPBF-fabricated binary NiTi alloys could be improved by modulating the distribution of the secondary phases (e.g.Ni 4 Ti 3 , Ti 2 Ni) through post-heat treatment [37][38][39][40][41].For example, Li et al [38] and Lu et al [39] reported that the mechanical and functional performances of the LPBF-fabricated Tirich NiTi SMAs could be improved largely by homogenization of Ti 2 Ni precipitates, achieved by high-temperature treatment.It was also reported that solution and aging treatment could provide homogeneous precipitation of Ni 4 Ti 3 nanoprecipitates, thus improving the superelastic stability of the LPBFfabricated Ni-rich NiTi SMAs [37,40,41].However, as a pseudo-binary eutectic system [5], the mechanical/functional performances of NiTiNb alloys are dominated by the morphology, size and distribution of eutectic structures.It has been reported that the morphology of eutectic structures of the LPBF-fabricated eutectic-type alloys (e.g.Al-12Si, Al-5.7Ni [42,43]) is quite distinct from that of conventionally prepared alloys, because of the rapid solidification characteristics and complex thermal histories associated with L-PBF.The existing heat treatment regimes for conventionally manufactured alloys are not always compatible with the additively manufactured alloys due to the differences in initial microstructure [42,44].There is a need to understand the influence of post-heat treatment on the LPBF-fabricated NiTiNb alloys, and to establish the inter-relationship between the microstructure, PT behavior and mechanical/functional performances.
In this work, a systematic investigation was conducted on the effect of solution treatment on the microstructure, PT behavior and mechanical/functional performances of the (NiTi) 91 Nb 9 (at.%)alloy produced by L-PBF using prealloyed NiTi and elemental Nb powders.Targeting the classical Ni 47 Ti 44 Nb 9 alloy, a high Nb content of 9 at.% was selected for ISA to obtain a comparable thermal hysteresis.The mechanical/functional performances of (NiTi) 91 Nb 9 alloys were improved by heat treatment.As a result, NiTiNb alloys with good tensile performances and wide hysteresis were obtained.The mechanisms on the evolution of microstructure, transformation behavior and mechanical performances after solution treatment were discussed.This study provides insights into the development of high-performance NiTi-based SMAs through L-PBF ISA and post-heat treatment.The integration of material synthesis and net-shaping of complex geometries may create new opportunities to prepare novel NiTibased smart structures.

Feedstock powders
Pre-alloyed Ni 50.6 Ti 49.4 (at.%) powders and high purity (>99.9%)elemental Nb powders were used in this work.The NiTi and Nb powder mixtures, with Nb content of 9 at.%, were mechanically mixed in a tumbler mixer for 6 h. Figure 1

L-PBF process and post-heat treatment
A commercial L-PBF system (Renishaw RenAM 500E) was utilized to fabricate the (NiTi) 91 Nb 9 parts.The L-PBF machine was equipped with a 500 W fiber laser, and the beam size is 80 µm.All the specimens were produced on a NiTi substrate (78 mm × 78 mm) without pre-heating.The L-PBF process was protected with high purity (>99.999%)Ar gas, and the O content of the build chamber was reduced to <20 ppm before processing.The chessboard scan strategy was used with a rotation of 67 • between layers (figure 1(d)).The field size of each island is 5 mm × 5 mm.
The previously optimized L-PBF process parameter, laser power (P) = 200 W, scanning speed (v) = 800 mm s −1 , hatch spacing (h) = 80 µm and layer thickness (t) = 30 µm, was used to produce dense (NiTi) 91 Nb 9 parts.Small specimens (6 mm × 8 mm × 8 mm) were made for microstructural observation (figure 1(d)).Cylinders with dimensions of Φ2 mm × 2.5 mm were prepared to measure the thermal hysteresis.Blocks with sizes of 60 mm × 10 mm × 10 mm were made to test the tensile performances and shape memory effects.The blocks were first machined into slices of 2 mm thickness through electrical discharge machining.Subsequently, these slices were machined into dog-boneshaped tensile specimens (gauge length: 15 mm, width: 3 mm).
Solution treatment was performed in a tube furnace at 1273 K (below the NiTi-Nb quasi-binary eutectic temperature, figure 1(e)) under Ar atmosphere, followed by water quenching.Different holding time was set as 15, 30, 60 and 120 min (figure 1(f)) to study the effect of post-heat treatment on the microstructure, PT behavior and mechanical/functional performances of the LPBF-fabricated (NiTi) 91 Nb 9 alloy.Hereafter, the solutionized samples are denoted as S15, S30, S60 and S120 samples, respectively.

Characterization techniques
The microstructure was characterized using scanning electron microscopy (SEM, JEOL JSM-7800) under backscattered electron (BSE) mode.The area fraction and number of the β-Nb phases and Ti-rich particles were determined from five BSE images for each sample using ImageJ ® [45] software.The elemental distribution was examined through electron probe microanalysis (EPMA, JEOL JXA-8530F PLUS).Transmission electron microscope (TEM) characterizations were performed in an FEI Talos F200X microscope operated at 200 kV.The specimens for TEM characterization were made in a Gatan 695 precision ion polishing instrument.The grain morphology and crystallographic orientation were investigated through electron backscatter diffraction (EBSD) in a JEOL JSM-7800 microscopy equipped with an Oxford NordlysMax3 system.Differential scanning calorimetry (DSC, NETZSCH DSC 3500 Sirius) was used to characterize the PT behavior at a heating/cooling rate of 10 K min −1 .Tensile performances and shape memory effects were tested in an MTS testing machine (mode E44.304) with a strain rate of 1.67 × 10 -4 s -1 .A home-made heating/cooling chamber was

Microstructure
Figures 2(a-i)-(a-v) show the BSE micrographs of the asprinted (NiTi) 91 Nb 9 alloy.A heterogeneous microstructure is observed, which consists of bright unmelted Nb particles, bright Nb-rich swirls, dark Nb-lean swirls and gray NiTi (Nb) matrix (figure 2(a-i)).The presence of the unmelted Nb particles is caused by the large melting point difference between NiTi (1 583 K) and Nb (2 741 K).A transition zone exists around the Nb particle (figure 2(a-ii)), which resulted from the compositional gradient caused by the diffusion of Nb.This also provides a strong metallurgical bond between NiTi matrix and the unmelted Nb particle.The Nb-rich dendritic structure was formed in the Nb-rich swirls (the inset in figures 2(a-i)), while the Nb-lean swirls mainly consisted of the B2-NiTi phase without Nb-rich phase (figure S1 in supplement).Figures 2(a-iii)-(a-v) show that the NiTi/Nb eutectics decorate the primary NiTi phase in a cellular-dendritic morphology.Depending on the thermal history, two distinct regions can be distinguished: (i) the fine eutectic (FE) region in the melt pool (MP) center, and (ii) the coarse eutectic (CE) region at the MP boundaries.The cell sizes in the FE and CE regions are 200 nm-500 nm and 1 µm-2 µm, respectively.The size of the eutectic cell is much smaller than that of the conventional NiTiNb alloy [5,12], since the cooling rate during L-PBF (10 6 K s −1 -10 8 K s −1 [46]) is several orders of magnitude higher than that in conventional casting (10 0 K s −1 -10 2 K s −1 ) [47].Also, columnar grains that grow from the boundary toward the center of the MP (i.e.parallel to the local heat flow direction) are observed in the FE region (figure 2(a-iii)).Some Nb-rich precipitates are observed within the cell (marked by the red arrow in figure 2(a-iv) and (a-v)).These precipitates were precipitated from the primary B2-NiTi phase due to the repetitive thermal cycling during L-PBF [48].Figures 2(a-iv) and (a-v) also show that some black Ti 4 Ni 2 O x particles (marked by the blue arrow) were formed at the cell boundaries.Detailed analysis of these particles using TEM is shown in figures 5 and 6.The following scenario of precipitation of the tiny Nb-rich precipitates is suggested: (i) the rapid cooling associated with L-PBF causes a solute trapping effect [42], which leads to a high Nb solubility in B2-NiTi that exceeds the maximum solubility under equilibrium conditions; (ii) upon solution treatment, the excess Nb atoms will precipitate out of the supersaturated B2-NiTi matrix.Similar observations have been reported in solution treatment of the L-PBF made Al-Si alloys [49].Figures 2(b-iv) and (b-v) also indicate that more Ti 4 Ni 2 O x phases were formed in the S15 sample as compared with the as-printed sample.
Figures 2(c 2(d-v).Coarsening of spherical β-Nb particles was mainly caused by the coalescence of adjacent β-Nb particles, as indicated by the joining neck between two neighboring β-Nb particles in figure 2(dv) (marked by the black arrow), as well as Ostwald ripening [42,50].It is worth nothing that the amount of spherical β-Nb particle in the FE region is increased as the holding time increases from 60 min to 120 min (figure 3(c)).This is because many rod-like β-Nb phases in the FE region transformed into spherical particles with prolonged holding time.Figures 2(c  Figure 4 shows the distribution of Nb, Ti and Ni, characterized by EPMA, in the as-printed and solutionized (NiTi) 91 Nb 9 alloys.The EPMA maps of the sample S30 are shown in figure S5 in supplement.Figure 4(a) shows that the distribution of Nb in the as-printed sample is heterogeneous, and many Nbrich and Nb-lean swirls can be observed.The compositional heterogeneity is mainly attributed to: (i) uneven physical mixing (e.g.Marangoni convection and recoil pressure [52,53]), which is exacerbated by the presence of unmelted Nb particles, and (ii) insufficient chemical diffusion due to the rapid cooling (10 6 K s −1 -10 8 K s −1 [46]) during L-PBF.After solution treatment, the Nb-rich swirls can still be observed (figures 4(b)-(d)), but their substructure has changed notably.Figure S4   in the FE region, the cell boundary in the CE region is composed of a lamellar NiTi/Nb eutectic structure, similar to the transition zone around the unmelted Nb particle (figure 2(aii)).Some particles enriched in Ti and O (i.e.black particles in figure 2) are also observed at the cell boundaries.Figure 5(m) also indicates the segregation of O at the cell boundaries in the CE region, which may be due to the high oxygen affinity of Nb [55].
Figure 6 shows the TEM results of the sample solutionized at 1273 K for 60 min (i.e.sample S60).In the FE region (figures 6(a)-(e)), the NiTi/Nb eutectics in the grain interior changed from a cellular-dendritic structure to rod-like and sphere-like morphology after solution treatment.The NiTi/Nb eutectics at the GBs preferentially spheroidized into large β-Nb particles with size up to 500 nm.The following scenario is suggested: (i) at the terminal stage of solidification, lowmelting-point NiTi/Nb eutectics are segregated along the GBs in form of liquid films; (ii) Nb is enriched at the GBs after solidification; (iii) the local enrichment of Nb leads to preferential spheroidization upon solution treatment.In addition, some Ti-rich particles are detected both at the GBs and in the grain interior (figure 6(b)).In the CE region (figures 6(f)-(j)), the eutectic cell was fully spheroidized after solution treatment.The enrichment of O element in β-Nb particles is observed (figure 6(j)).TEM-EDS point analysis (table S1 in supplement) shows that the oxygen content in O-rich β-Nb particles is relatively low (<1 at.%), and is not sufficient to form Nb oxides [56].Also, the Nb-rich precipitates with size of ∼100 nm are observed.
Figure 6(k) shows the HRTEM image of a rod-like β-Nb in the FE region.The corresponding FFT and inverse FFT patterns (figures 6(l) and (m)) indicate that the β-Nb phase and the B2 matrix have a semi-coherent interface with the orientation relationship of {110} β -Nb //{110} B2 and <111> β -Nb //<111> B2 .This is consistent with previous reports in conventional NiTiNb alloys [57,58].Figure 6(n) indicates that the large spheroidized β-Nb particle is incoherent with the B2 matrix.The loss of interface coherency is mainly caused by the large size of the spheroidized β-Nb phase, where the lattice misfit becomes too large to maintain the coherent interface.x phase is stable even at the Ni-rich composition [59].The formation of these particles may be due to the oxygen pickup during L-PBF processing or the oxidation of the original NiTi and Nb powders [48].Similar observation has been reported in L-PBF of the binary NiTi alloys [60,61].
TEM-EDS line scan analysis was conducted to analyze the composition changes of the matrix and existing phases before and after solution treatment (figure 7).The chemical composition measured by TEM-EDS point analysis is also given in table S2 in supplement.For the as-printed alloy, the Nb content of the matrix is around 5.0 at.% (figure 7(a)).After solution treatment for 60 min, the Nb content of the matrix near the rodlike β-Nb phase in the FE region (figure 7(b)) drops to ca. 3.0 at.%. Figures 7(c) and (d) show that the Nb content of the matrix near the large spheroidized β-Nb phases (at the GBs in the FE region or the cell boundary the CE region) is reduced to ca. 3.2 at.%.This is consistent with the TEM-EDS point analysis (table S2).The decrease in the Nb content of the B2 matrix is because the extended Nb solubility in B2-NiTi matrix (caused by solute trapping effects [42] under rapid solidification) is balanced to near-equilibrium levels due to sufficient chemical diffusion during high-temperature solution treatment.Similar observations have been reported in other alloys (e.g.Al-12Si [42], Al-Sc-Zr [62], Al-Mg-Sc-Zr [63] alloys).
Figure 7 also shows that the composition of the β-Nb phase has changed after solution treatment.For the as-printed (NiTi) 91 Nb 9 alloy (figure 7(a)), the Nb content of the β-Nb phase in the lamellar NiTi/Nb eutectics (i.e.cell boundary) in the CE region is around 60 at.%, which is lower than that in conventionally prepared NiTiNb alloys (ca.80 at.% [5,58]).The β-Nb phase (i.e.cell boundary) in the FE region also contains around 60 at.%Nb, as shown in table S2 in supplement.The solubility of Ti in the β-Nb phase is much higher than that of Ni (figure 7(a), table S2).According to the Ti-Nb [64] and Ni-Nb [65] phase diagram, Nb is infinitely soluble with Ti but has a limited solubility with Ni at high temperatures.Upon rapid cooling (cooling rate of 10 6 K s −1 -10 8 K s −1 [46]), more Ti may be retained in the β-Nb phase.S2).This is because the solubilities of Ti and Ni revert to an equilibrium state during spheroidization of the β-Nb phase due to diffusion.Moreover, the Ti solubility in the β-Nb phase decreased Figure 8 shows the EBSD maps and GB distribution maps of the as-printed and solutionized (NiTi) 91 Nb 9 alloys.The linear-intercept method (figure S7 in supplement) was used to measure the average grain size.Figure 8(a) shows the grain morphology of the as-printed alloy.Fine equiaxed grains with size of 1 µm-5 µm are located at the MP boundary, and large columnar grains with size of 10 µm-60 µm exist at the MP center.The fraction of equiaxed and columnar grains is around 47% and 53%, respectively (figure 8(i)).The average grain size of the as-printed sample is around 8.8 µm.
Figures 8(b)-(d) and S8 in supplement indicate that the grain morphology of the solutionized samples is essentially the same as the as-printed sample.Also, the grain size distribution remains almost unchanged with increasing holding time (figure 8(i)).For instance, the fraction of equiaxed grains remains constant at ∼48% with increasing holding time from 15 min to 120 min.No obvious grain growth is observed, and the average grain size is around 9.0 µm (figure 8(i)) even after a long-term solution treatment (120 min).Similar observations have been reported in solution treatment of the LPBF-fabricated Al-Si alloys [49].Figures 8(e)-(h) show that many low-angle GBs are observed in the as-printed and solutionized specimens, distributed mainly inside the columnar grains.The histogram in figure 8(j) indicates that the proportion of low-angle GBs versus high-angle GBs is essentially unaffected after high-temperature solution treatment.Figure S9 in supplement shows that the as-printed sample shows a weak <111> B2 //BD fiber texture, and the texture remains essentially unchanged after solution treatment.The high thermal stability of grain morphology is attributed to: (i) no enough driving force for recrystallization [49], and (ii) restricted grain growth caused by the large amount of β-Nb phase and Ti 4 Ni 2 O x particle pinning the GBs, according to the Zener pinning effect [66] (see details in figure S10 in supplement).The above results indicate that the solution treatment has less effect on the grain morphology.Therefore, the change in eutectic structure (figure 2), rather than grain morphology, plays a dominant role in affecting the mechanical properties.
Figure 9 shows the XRD spectrums, collected at room temperature (RT), of the as-printed and solutionized (NiTi) 91 Nb 9 alloys.All the samples are composed of the B2 austenite phase and β-Nb phase at RT (figure 9(a)).Figure 9(b) shows an enlarged view of the (110) B2 peak.The (110) B2 peak of the as-printed specimen is detected at 2θ of 42.22 • , while that of the solutionized sample shifts to higher angles (∼42.30• ).The shift of diffraction peaks suggests a reduced lattice distortion after solution treatment, mainly due to the expulsion of Nb from the supersaturated matrix (figure 7), as well as the release of internal stress [44,67].Figure 9(c) shows an enlarged view of the (110) β -Nb peak.The volume fraction of β-Nb phase was estimated by the intensity of diffraction peaks [39,68].The intensity of the (110) β -Nb peak in the S15 specimen is higher as compared with the as-printed specimen, indicating an overall increase in the amount of β-Nb phase.This is caused by the expulsion of Nb atoms from the supersaturated matrix, favoring the formation of the β-Nb phase (figure 2(b-v)).The intensity of the (110) β -Nb peak remains essentially constant with increasing holding time.This indicates that increasing holding time has a less effect on the amount of the β-Nb phase, although the morphology of the β-Nb phase changes (figure 2).

Phase transformation behavior
Figure 10(a) shows the DSC curves of the as-printed and solutionized (NiTi) 91 Nb 9 alloys.The martensite transformation peak temperature (M p ) and austenite transformation peak temperature (A p ) are summarized in figure 10(b).The thermal hysteresis (T hys ), which is defined as the temperature difference between A p and M p (i.e.A p -M p ), and transformation heat Figure 10(a) shows that all the samples exhibit a one-stage A↔M transformation.The B2 structured austenite transforms into the B19 ′ structured martensite phase during cooling, and the B19 ′ martensite reverts to the B2 austenite during heating.This is consistent with conventional NiTiNb alloys [3,69,70].The martensite transformation temperature (MTT) of the as-printed alloy is very low.For example, the M p temperature is 171 K, which is much lower than that of the original NiTi powders (∼244 K).After solution treatment for 15 min, the MTTs increase remarkably.The M p temperature is increased to 245 K (figure 10(b)).With increasing holding time to 60 min and 120 min, the MTTs are essentially unchanged, and the M p temperature remains constant at (246 ± 1) K.The detailed mechanism of the MTT variation will be discussed in section 4.2. Figure 10(a) also indicates that the transformation interval (i.e., the width of the transformation peaks) of the solutionized samples is much smaller than that of the as-printed sample, indicating that the microstructural homogeneity is improved after solution treatment [71,72].This is in agreement with the EPMA results in figure 4. Figure 10(c) indicates that the solutionized samples have larger transformation heat than the as-printed part.The ∆H M-A of the as-printed sample is 5.8 J g −1 , and the ∆H M-A of the S15 sample is increased to 18.2 J g −1 .In addition, solution treatment leads to a slight reduction in the intrinsic T hys , from 73 K in the as-printed sample to around 56 K in the solutionized samples.
It is known that pre-deformation could stabilize the martensite phase in conventional NiTiNb alloy, leading to an expansion of the thermal hysteresis, which enables the components to be stored and transported at ambient temperatures [3,70,73].To investigate the deformationinduced expansion of hysteresis in the LPBF-fabricated (NiTi) 91 Nb 9 alloys, cylindrical specimens were compressed by 20% and then subjected to two consecutive DSC cycles.The results are shown in figure 10(d-i).The solutionized alloys were deformed at 298 K (RT).The as-printed alloy was deformed at 218 K (the lowest temperature attainable by the cooling system) to ensure the occurrence of stress-induced martensitic (SIM) transformation.The forward and reverse transformation peak temperatures after pre-deformation are denoted as M p ′ and A p ′ , respectively.After pre-deformation, all the samples undergo an increase in A p temperature but a decrease in M p temperature during the first thermal cycle (figures 10(d)-(h)).In the second cycle, the A p temperature returns to a level comparable to the undeformed state, whereas M p hardly changes.This phenomenon, often called 'martensite stabilization', has been frequently reported in conventional NiTiNb alloys [3,70,73].Figure 10(i) indicates that the amplitude of hysteresis expansion is decreased after solution treatment at 1273 K for 15 min, and remains essentially constant with holding time.The thermal hysteresis after pre-deformation (A p ′ -M p ′ ) is 168 K for the as-printed sample and around 145 K for the solutionized samples.This is comparable to that of the conventional Ni 47 Ti 44 Nb 9 alloy (130 K-170 K [3,69,70,73]).

Mechanical and functional properties
Figure 11(a) shows the tensile curves of the as-printed and solutionized (NiTi) 91 Nb 9 alloys.The solutionized samples were tested at 10 K below M f (223 K), while the as-printed sample was deformed at 218 K (the lowest temperature of the cooling system).The plateau stress (σ p , which is determined using the tangent method), fracture stress (σ f ) and fracture strain (ε f ) are summarized in figure 11(b).
All the samples exhibit a typical stress-strain behavior of the NiTi-based SMAs (figure 11(a)), which is featured with a stress plateau.Depending on the initial state before loading, the type of stress plateau can be distinguished [1].The as-printed specimen undergoes SIM transformation, while the solutionized samples undergo SIM reorientation. Figure 11(a) also indicates that the σ p of the solutionized samples remains essentially unchanged with increasing holding time, consistent with the DSC results (figure 10(a)).
The as-printed specimen shows a fracture strain of (7.6 ± 0.1)% and a fracture stress of (613 ± 19) MPa (figure 11(b)).The tensile properties benefit from short-term solution treatment (e.g. 15 min and 30 min).A good combination of fracture strain ((9.5 ± 0.4)%) and fracture stress ((781 ± 20) MPa) is obtained in S30 sample.The mechanical performances are comparable to those of conventional casted NiTiNb alloys but lower than those in the wrought state (see details in figure S12 in supplement).To the best of authors' knowledge, the tensile ductility of 9.5% is superior to the LPBF-fabricated NiTiNb alloys reported so far [8,32,33,74].The enhanced mechanical performances are mainly attributed to the transition of the NiTi/Nb eutectics from meshlike into the rod-like and sphere-like morphology (figure 2).As compared with mesh-like β-Nb, rod-like and sphere-like β-Nb phases are more favorable for toughening because of the following reasons: (i) the matrix around the rod-like and spherelike β-Nb has a more favorable twinning orientation than that inside the mesh-like β-Nb, which favors the release of strain energy through SIM [8]; (ii) the rod-like and sphere-like β-Nb have a better ability to store dislocations [8]; (iii) the rodlike and sphere-like β-Nb are favorable for maintaining the continuity of the matrix.However, the ductility is deteriorated with increasing holding time to 60 min and 120 min.The fracture strains of the S60 and S120 samples are (7.2 ± 0.6)% and (6.6 ± 0.6)%, respectively.According to figures 2 and 3, both the ductile β-Nb phase and brittle Ti 4 Ni 2 O x phase coarsened  after long-term solution treatment.The previous study showed that the coarsening of the ductile particles did not seriously impair the ductility [42].Therefore, the presence of large Ti 4 Ni 2 O x phases, especially at GBs, is the major cause for the deterioration of the mechanical performances after long-term solution treatment.
Figure 12 shows the fracture morphologies of the asprinted, S30 and S120 samples.Quasi-cleavage facets small dimples (∼1 µm) are observed in the as-printed specimen (figures 12(a) and (b)), indicating a mixed ductile-brittle fracture mode.Figures 12(d)-(i) show that the S30 and S120 samples also exhibit mixed failure mechanism.The reasons are considered as: (i) no obvious change in grain morphology (figure 8) between the different samples; (ii) high volume fraction of β-Nb phase remaining the FE region after long-term solution treatment (figure 2).Figures 12(c), (f) and (i) show the comparison of the fracture morphology near the unmelted Nb particles before and after solution treatment.The Nb particles in the as-printed specimen were torn in half, indicating a strong bonding with the matrix.However, the particles in the solutionized specimen were completely detached from the matrix, suggesting a debonding of the unmelted Nb particles with matrix.This is due to the weakening of the metallurgical bond between the Nb particle and Overall, there are three factors that affect the mechanical performances of the solutionized NiTiNb alloys: (i) the transition of the NiTi/Nb eutectics from mesh-like to rod-like and sphere-like, (ii) the increase of the amount and size of the Ti 4 Ni 2 O x particles, and (iii) the debonding of the unmelted Nb particles with matrix.Factor (i) enhances, while factors (ii) and (iii) impair the mechanical properties.
Figure 12 indicates that the unmelted Nb particles lose metallurgical bonding with the matrix, despite of the solution treatment duration from 15 min to 120 min.As a result, the contribution of factor (iii) on mechanical properties is essentially the same for all the solutionized samples.Therefore, the change in mechanical properties of the solutionized NiTiNb alloys is a result of the competition between the factor (i) and (ii).
After a short-term solution treatment (e.g. 15 min, 30 min), the area fraction of Ti 4 Ni 2 O x phases is relatively low (figure 3), and the contribution of factor (ii) is at a low level.Therefore, the influence of the factor (i) is dominant.With increasing holding time, the NiTi/Nb eutectics transformed from meshlike to rod-like and sphere-like.The ductility of the alloy gradually increases, and reaches a maximum value at the holding time of 30 min.After long-term solution treatment (e.g. 120 min), more Ti 4 Ni 2 O x phases were generated probably due to the increased oxidation (figure 3).The factor (ii) plays a stronger role than the factor (i), and dominates the change in mechanical properties.As a result, the ductility of the alloy gradually decreases with prolonging holding time from 30 min to 120 min.
Figures 13(a)-(e) show the shape memory behavior of the as-printed and solutionized (NiTi) 91 Nb 9 alloys.The specimen was subjected to a pre-strain of 5% at low temperatures (same as the tensile tests), and then unloading to a stress of 5 MPa.Subsequently, the specimen was heated up to 373 K under a constant stress of 5 MPa.The shape recovery temperature (T r ) is determined using the tangent method (figure S13 in supplement).The shape recovery rate (η) is defined as the ratio between the total recoverable strain (ε rec ) and the total pre-strain (ε tot , i.e. 5%).The η and T r are summarized in figure 13(f).
Both the as-printed and solutionized parts show good shape memory effect with η > 90%.This is comparable to that of the conventionally manufactured NiTiNb alloys [75].Although a long-term solution treatment (e.g. 120 min) compromises the ductility of the material (figure 11), it does not worsen the shape memory effect.This is probably due that plastic activity is not introduced, since the stress at the strain of 5% (around 250 MPa) is lower than the critical stress for dislocation slip.Figure 13(f) also shows that the T r increases after solution treatment and then remains essentially unchanged with holding time, in agreement with the DSC results (figure 10(a)).The shape memory effect of LPBF-fabricated NiTiNb alloys is comparable to that of binary NiTi alloys prepared by L-PBF (η > 90% under pre-strain of 5%-8%) [39,61,76], which indicates that Nb addition does not undermine the shape memory properties.

Effect of solution treatment on microstructures
L-PBF usually generates a non-equilibrium microstructure caused by the large thermal gradients (G) and high cooling rates ( Ṫ), notably surpassing those typical of conventional metallurgical processes (e.g.casting).ISA with powder mixture further exacerbates the compositional and microstructural heterogeneity.Figures 2-6 indicate that solution treatment not only improves the compositional homogeneity but also alters the microstructure evolution.Since the morphology of the eutectic structure plays a decisive role in the mechanical performances of eutectic-type alloys [42,77,78], it is essential to understand the effect of solution treatment on the microstructure.
The solidification microstructure of the LPBF-fabricated metallic materials depends highly on the undercooling ∆T ahead of solid/liquid (S/L) interface [46,79].For pure metals, solidification occurs with a stable planar interface.For alloys, the solute-driven constitutional subcooling ∆T CS and cooling rate-induced thermal undercooling ∆T t will generate [80,81], leading to the destabilization of the solidification front.This causes a transition from planar to cellular/dendritic growth.It is suggested that NiTi/Nb cellular-dendritic structure was formed by the solute segregation ahead of the cellular/dendritic interface at a high Nb content of 9 at.%(figure 2).In comparison to the conventional casting process with relatively low Ṫ (10 0 K s −1 -10 2 K s −1 [47]), the extremely high Ṫ (10 6 K s −1 -10 8 K s −1 [46]) during L-PBF limits the chemical diffusion, which affects the microstructure in three aspects: (i) formation of supersaturated primary B2-NiTi matrix, where the Nb solubility in B2-NiTi exceeds the maximum equilibrium solubility due to the solute trapping effect [42]; (ii) refinement of the eutectic cells compared to the conventional casted alloys, since the cell size is inversely proportional to the cooling rate [46]; (iii) Ti enrichment in β-Nb phases due to rapid cooling of the Nb-rich melts (regions that form the β-Nb phase after cooling) with high Ti solubility.The generation of CE cells at the MP boundary is attributed to: (i) a lower at the MP boundaries than that in the MP center [61,82,83], and (ii) the interlayer remelting that promotes the growth of the cells.
Solution treatment provides sufficient chemical diffusion, resulting in the microstructure changes from non-equilibrium to equilibrium state.As schematically shown in figure 14, two main effects are involved upon solution treatment.First, the extended Nb solubility in primary B2-NiTi was balanced to near-equilibrium levels (figure 7).The Nb expulsion from the supersaturated B2-NiTi phase led to the formation of tiny Nb-rich precipitates (figures 2(b-v) and 14(b-ii)).Second, the fragmentation, spheroidization and coarsening of the NiTi/Nb eutectics occurred.The fragmentation of the NiTi/Nb eutectics mainly occurred during short-term solution treatment (figures 14(a-ii) and (b-ii)), and spheroidization and coarsening occurred with increasing holding time (figures 14(a-iii) and (b-iii)).
The morphological transition of the NiTi/Nb eutectics can be explained by the Gibbs-Thompson effect [50,84].The theory assumes that particles with different curvatures have different chemical potentials, and thus different solute solubility in the matrix phase.The solute concentration of the matrix (C r ) at the surface of a spherical particle with radius r can be expressed as [50]: where C e is the matrix concentration at a flat interface (i.e. a particle with infinite radius), γ is the interfacial energy between the matrix and particle, Ω is the molar volume of the particle, R B is the Universal gas constant and T is the temperature.
According to equation (1), C r is inversely related to r.A small r will generate a large C r , whereas a large r will generate a small C r .For the cellular-dendritic structure (e.g.NiTi/Nb eutectics), the curvatures at different positions are different.Here we consider that the cellular structure consists of the nodes (position A in figures 14(b)-(i)) and the cell boundaries away from the nodes (position B in figure 14(b-i)).The nodes of the cells have large curvature (small r), while the cell boundaries away from the nodes have small curvature (very large r) and are a nearly flat interface.This results in a large C r near the nodes (position A) and a small C r near the cell boundaries (position B).During high-temperature solution treatment, the difference in solute concentration induces a diffusive flux of atoms from the nodes (position A) to the cell boundaries (position B).As a result, fragmentation of the NiTi/Nb eutectics occurs at the nodes with the larger curvatures, and the morphology of the eutectics changes from mesh-like to rod-like (figures 14(a-ii) and (b-ii)).Afterwards, spheroidization of the rod-like phase (i.e., the fragmented NiTi/Nb eutectics) occurs.The ends of the rod-like phase have larger curvature than the middle, and thus the edges will preferentially decompose.With increasing holding time, the size of the rod-like phase is decreased, and the rod-like phase gradually transforms into spherical particles with the close curvatures (figures 14(a-iii) and (b-iii)).
It is noteworthy that the degree of spheroidization of NiTi/Nb eutectics in the FE and CE regions is different, mainly due to the distinct substructure of the cell boundaries (figure 5).In the CE region, the cell boundary consists of a lamellar NiTi/Nb eutectic substructure (figures 5(i)-(m)).This structure is thermally unstable and decomposes rapidly during high-temperature heat treatment, which has been frequently reported in the eutectic alloys [85,86].As a result, complete spheroidization occurred in the CE region after solution treatment for 60 min (figure 2(c-v)).Similarly, the NiTi/Nb eutectics in the local Nb-rich regions (i.e., Nb-rich swirls, transition zone around the unmelted Nb particle, and columnar GBs in the FE region) were all preferentially spheroidized because of their instability at elevated temperatures (figure 2).However, the cell boundary in the FE region is fully β-Nb phase (figure 5(g)).Due to the relatively high thermal stability, only partial spheroidization occurred, and many rod-like β-Nb phases were still retained after solution treatment for 120 min (figure 14(a-iii)).
With prolonging holding time, coarsening of the initially spheroidized β-Nb particles and Nb-rich precipitates (formed in L-PBF processing or after solution treatment) occurred (figures 14(a-iii) and (b-iii)).According to figures 2 and 5, most of the particles are less than 100 nm in size.Due to the large interfacial energy, these small particles are thermodynamically unstable and have a tendency to spontaneously coarsen into large particles with small interfacial energy [50].According to equation (1), the C r of small particles is higher than that of large particles.The solute concentration gradients in the matrix cause smaller particles to dissolve and larger ones to grow, known as Ostwald ripening [50].In addition, the coalescence of adjacent β-Nb particles (indicated by the joining neck between two neighboring β-Nb particles in figure 2(d-v)) facilitates coarsening, which has also been reported in the coarsening of eutectic Si particles [42,87].The Ti 4 Ni 2 O x phases also coarsened with increasing holding time (figures 3(e), 14(a-iii) and (b-iii)).The presence of large Ti 4 Ni 2 O x phases, especially at GBs, is the main cause for the deterioration of the mechanical performances after long-term solution treatment (figure 11).

Phase transformation behavior
Figure 10(a) indicates that the MTT of the as-printed (NiTi) 91 Nb 9 alloy is very low (e.g.M p = 171 K), which is much lower than that of the original powders (M p = 244 K, figure 1). Figure S14 in supplement indicates that the MTT of the LPBF-fabricated binary NiTi alloy (e.g.M p = 294 K) is higher than that of the original powders.According to previous studies [71,88], the rise in MTT of the LPBF-fabricated binary NiTi is mainly caused by the preferential Ni evaporation during L-PBF processing.Therefore, the Nb-related effects contribute to the decrease in MTT of the as-printed (NiTi) 91 Nb 9 alloy.
The Nb-related effects affecting the MTTs of the LPBFfabricated NiTiNb alloys could be divided into four categories [31] (figure 15(a)): (i) substitution of Ti atoms by Nb, resulting in a decrease in valence electron concentration [12], which increases the MTTs; (ii) presence of Nb-rich phases (i.e.eutectic β-Nb, Nb-rich precipitates), which suppresses the martensite transformation; (iii) Ti enrichment in the β-Nb phase, causing an increase in ratio Ni/Ti of the matrix, which decreases the MTTs; (iv) lattice distortion caused by Nb solid solution in the supersaturated matrix (due to solute trapping effect), which suppresses the martensite transformation.After a short-term solution treatment (e.g. 15 min), the MTT of the solutionized alloy is increased largely (M p = 245 K, figure 10(a)).The MTT value is close to that of the casted Ni 46 Ti 45 Nb 9 alloy (M p of ∼250 K) [12].This is mainly caused by the microstructure changes from nonequilibrium to equilibrium state, and thus the contribution of the abovementioned factors changes quickly (figure 15(b)).The expulsion of Nb atoms from the supersaturated matrix reduces the lattice distortion of the matrix (which increases the MTTs), but increases the amount of the Nb-rich precipitates (which decreases the MTTs).The expulsion of Ti atoms from the β-Nb phase (i.e. the decrease of Ti solubility in β-Nb phase, figure 7) during spheroidization lowers the ratio Ni/Ti in the matrix, which increases the MTTs.The amount of the Ti 4 Ni 2 O x phase is increased (figure 3), which decreases the MTTs.Overall, the total rise in MTT after solution treatment for 15 min is mainly attributed to: (i) the reduced lattice distortion due to the Nb expulsion from the supersaturated B2-NiTi, and (ii) the expulsion of Ti from the β-Nb phases that lowers the ratio Ni/Ti of the matrix.
With increasing holding time, the MTTs of the solutionized alloys remain essentially unchanged (figure 10(a)).It is supposed that the solution of Nb atoms in the NiTi matrix is in a stable state after high-temperature solution treatment for 15 min or longer.The role of the related factors (i.e., lattice distortion caused by Nb solid solution in the supersaturated matrix, substitution of Ti atoms by Nb) on the MTT variation remains essentially unaffected with holding time.As a result, the stable MTT is the combined effect of three remaining factors (figure 15(b)).First, prolonging holding time leads to the increased spheroidization of the rod-like β-Nb in the FE region (figure 2).The expulsion of Ti atoms from the β-Nb will reduce the ratio Ni/Ti of the matrix, thus increasing the MTTs.Second, coarsening of β-Nb phases reduces the volume fraction of small particles and increases the amount of large particle (figures 2 and 3).Large β-Nb particles are incoherent with the B2-NiTi matrix (figure 6(n)) and have a weak influence on the martensite transformation [11].This tends to increase the MTTs.Third, the content of the Ti 4 Ni 2 O x phases is increased with increasing holding time (figure 3(d)), which decreases the MTTs.It is suggested that the combined effects of the Nb-related effects and presence of Ti 4 Ni 2 O x phases lead to stable MTTs with increasing holding time (figure 15(b)).To verify this conclusion, a higher temperature (i.e. 1 323 K) solution treatment for different holding time was performed.The DSC curves are shown in figures 15(c) and (d).With the increase of holding time, M p temperature gradually increases instead of remaining constant.A higher temperature (1323 K) solution treatment will intensify the spheroidization of the rod-like β-Nb in the FE region and accelerate the coarsening of spherical β-Nb particles.This will weaken the role of the related factors (i.e.Ti enrichment in the β-Nb phase and presence of Nb-rich phases).As a result, the MTTs increase with holding time.This also demonstrates that the variation of MTT with holding time is dynamically compensated by multiple factors during solution treatment at 1273 K.It is worth nothing that the thermal hysteresis of the solutionized alloys is around 145 K after 20% pre-deformation (figure 10), comparable to that of the conventional Ni 47 Ti 44 Nb 9 alloy (130 K-170 K [3,69,70,73]).Together with the good mechanical and functional properties (figures 11 and 13), this work demonstrates the feasibility of developing NiTi-X alloys through L-PBF ISA and post-heat treatment.

Conclusions
In this work, a (NiTi) 91 Nb 9 alloy was in-situ synthesized by L-PBF using pre-alloyed NiTi and elemental Nb powders.The effect of solution treatment on the microstructure, PT behavior and mechanical/functional performances of the (NiTi) 91 Nb 9 alloy was investigated.The results suggest that L-PBF ISA combined with post-heat treatment is a viable metallurgical route for fabricating NiTiNb ternary alloys, and offers potential for developing high-performance NiTi-based structures with complex geometries.The main conclusions are summarized as follows: (1) A submicron cellular-dendritic structure surrounding the supersaturated B2-NiTi matrix was formed in the as-printed (NiTi) 91 Nb 9 alloy.Upon solution treatment, Nb-rich precipitates were precipitated from the supersaturated matrix.The fragmentation and spheroidization of the NiTi/Nb eutectics occurred, and the eutectic networks transformed into rod-like and sphere-like morphology.The β-Nb particles and Nb-rich precipitates coarsened with increasing holding time.The decrease in interfacial energy is the driving force for the morphological evolution of the NiTi/Nb eutectics.(2) The MTT increases after solution treatment.This is mainly caused by the microstructure changes from nonequilibrium to equilibrium state, including: (i) reduced lattice distortion due to the Nb expulsion from the supersaturated B2-NiTi, and (ii) the Ti expulsion from the β-Nb phases that lowers the ratio Ni/Ti in the matrix.The thermal hysteresis in the solutionized alloys remains at around 145 K after 20% pre-deformation.Regulation of the transformation characteristics (e.g.MTT and thermal hysteresis) of the NiTiNb alloys can be achieved by adjusting the chemical composition.(3) The alloy solution treated for 30 min shows good tensile properties with a fracture stress of (781 ± 20) MPa and a fracture strain of (9.5 ± 0.4)%.The improvement in mechanical performances compared to the asprinted alloy is due to the morphological transition of the NiTi/Nb eutectics from mesh-like to rod-like and sphere-like.All the as-printed and solutionized samples exhibit good tensile shape memory effects with recovery rates >90%.Further improvement of the mechanical performances of the LPBF-fabricated NiTiNb alloys may be achieved through hot isostatic pressing.
(a) shows that both NiTi and Nb powders have good sphericity, and Nb powders are evenly distributed in the NiTi-Nb powder mixtures.The particle sizes of both the NiTi and Nb powders are ranging from 15 µm to 60 µm.The D 50 of NiTi, Nb and (NiTi) 91 Nb 9 powders is 35.2 µm, 37.5 µm and 36.1 µm, respectively (figure 1(b)).The PT behavior of the NiTi and (NiTi) 91 Nb 9 powders is shown in figure 1(c).Two transformation peaks are observed during both the forward and reverse martensitic transformation, which indicates an inhomogeneous composition of the NiTi powders.

Figure 1 .
Figure 1.Powder information, scanning strategy and heat treatment regime of L-PBF fabricated NiTiNb: (a) backscattered electron (BSE) micrograph of the (NiTi) 91 Nb 9 powder mixture; (b) particle size distribution of the binary NiTi, Nb and (NiTi) 91 Nb 9 powders; (c) DSC curves of the NiTi and (NiTi) 91 Nb 9 powders; (d) schematic illustration of the chessboard scanning strategy and the LPBF-fabricated parts; (e) NiTi-Nb pseudo-binary equilibrium phase diagram, which is calculated using the Thermo-Calc software with TCNi8: Ni-Alloys v8.2 database; (f) heat treatment route.

Figure 2 (
b) shows the microstructure of the NiTiNb sample after solutionized at 1273 K for 15 min (i.e.sample S15).Some Nb-rich and Nb-lean swirls are still observed (figure 2(b-i)).High-magnification BSE images (figure 2(b-ii)-(b-v)) show obvious changes in the morphology of the NiTi/Nb eutectics.The evolution of NiTi/Nb eutectics can be divided into the following three categories: (i) complete spheroidization of the eutectics into large spherical particles in the local Nbrich region (i.e.Nb-rich swirls, transition zone around the unmelted Nb particle, columnar grain boundaries (GBs) in the FE region (marked by the green arrows in figure 2(b-i)-(biii)); (ii) fragmentation of the mesh-like eutectics into rod-like

Figure 2 .
Figure 2. BSE micrographs of the (a) as-printed (NiTi) 91 Nb 9 alloy, (b) S15 sample, (c) S60 sample and (d) S120 sample.The last two rows show the enlarged view of the microstructure of the fine eutectic (FE) and coarse eutectic (CE) regions, respectively.
) and (d) show the BSE images of the samples solution treated for 60 (i.e.sample S60) and 120 min (i.e.sample S120).The BSE micrographs of the sample S30 are given in figure S2 in supplement.To quantitatively investigate the changes of the morphology and size of β-Nb phases and Ti 4 Ni 2 O x phases with increasing holding time, image analysis was performed using ImageJ ® [45] software (figure 3, see details in figure S3 in supplement).As the holding time increases, the area fraction of the rod-like β-Nb phase in the FE

Figure 3 .
Figure 3. Area fraction of the spherical β-Nb particles (including the initially spheroidized β-Nb particles, Nb-rich precipitates) and the rod-like β-Nb phases in (a) fine eutectic (FE) region and (b) coarse eutectic (CE) region versus holding time.(c) The number of spherical β-Nb particles versus holding time.(d) Area fraction and (e) average size of Ti 4 Ni 2 Ox phases versus holding time.Image analysis was performed based on BSE images using ImageJ software (see details in figure S3 in supplement).
) and (d) also indicate that more Ti 4 Ni 2 O x phases were generated in the FE and CE regions, probably due to the increased oxidation during long-term solution treatment[51].These particles also coarsened after long-term solution treatment (figures 3(d)and (e)), due to the coalescence of adjacent particles and Ostwald ripening (figure2(d-v)).For instance, the area fraction of Ti 4 Ni 2 O x phases in the CE region is around 1.5% after solution treatment for 15 min and is increased to ca. 2.3% when the holding time is 120 min (figure3(d)).Also, the average size of Ti 4 Ni 2 O x phases in the CE region increases from (136 ± 9) nm to (219 ± 22) nm with increasing holding time from 15 min to 120 min (figure 3(e)).It is worth noting that Ti 4 Ni 2 O x phases at GBs in the FE region were coarsened more obviously than those in the grain interior, e.g. the size of the Ti 4 Ni 2 O x phases at GBs of S60 sample is ca.220 nm, whereas that inside the grains is ca. 100 nm (figure 2(c-iv)).The growth of the Ti 4 Ni 2 O x phases and the increase of their

Figure 4 .
Figure 4. Electron probe microanalysis (EPMA) images of the (a) as-printed (NiTi) 91 Nb 9 alloy, (b) S15 sample, (c) S60 sample and (d) S120 sample.The EPMA scale bar represents in wt.%.The Nb distribution maps in the second row were processed by setting the unmelted Nb particles to black color and displaying the Nb content in a range from 5 wt.% to 30 wt.%.
in supplement indicates the morphological transition in the Nbrich swirls from micron-scale dendritic structure (in the asprinted sample) to nanoscale spherical particles (in the solutionized sample).The elimination of the dendritic segregation indicates that the compositional heterogeneity is alleviated due to sufficient chemical diffusion during solution treatment.The TEM images of the as-printed (NiTi) 91 Nb 9 alloy are shown in figure 5. Figures 5(a)-(h) show the microstructure of the FEs in the FE region.Some precipitates with size of 50 nm-100 nm are detected inside the cell (marked by the yellow arrow in figures 5(a) and (b)).Dislocations are observed inside the cell and near the precipitates (figures 5(c)and (d)), resulting from the accumulation of thermal stress during L-PBF process [48, 54].The selected-area electron diffraction pattern in figure 5(f) indicates the B2 structure of the matrix.The HRTEM image (figure 5(g)) and the corresponding fast Fourier transform (FFT) pattern (figure 5(h)) indicate that the cellular boundary in the FE region is β-Nb phase.Figures 5(i)-(m) show the high-angle annular dark field scanning transmission electron microscopy (STEM) image and STEM-EDS maps of the CE region.The cell size is larger and the cell wall is thicker as compared with that in the FE region.Instead of the fully β-Nb structure of the cell boundary

Figure 5 .
Figure 5. TEM images of the as-printed (NiTi) 91 Nb 9 alloy: (a), (b) low magnification bright field (BF) image and the corresponding dark field (DF) image of the fine eutectic (FE) region; (c)-(e) high magnification BF images of the FE region; (f) SAED pattern of the red circle in (e); (g) high-resolution TEM (HRTEM) image of the yellow rectangle in (e); (h) the corresponding FFT pattern of (g); (i)-(m) HAADF-STEM image and STEM-EDS maps of the cell in the coarse eutectic (CE) region.
Figure 6(o) indicates that the Tirich particle has a Ti 2 Ni type structure.Due to the enrichment of O element (figure 6(e)), these particles are identified as Ti 4 Ni 2 O x phases.The Ti 4 Ni 2 O x phase has nearly the same

Figure 6 .
Figure 6.TEM micrographs of the sample solutionized at 1273 K for 60 min (i.e., sample S60): (a)-(e) HAADF and STEM-EDS maps of the fine eutectic (FE) region; (f)-(j) HAADF and STEM-EDS maps of the coarse eutectic (CE) region; (k) high-resolution TEM (HRTEM) image of a rod-like β-Nb phase in the FE region; (l) fast Fourier transform (FFT) pattern of (k); (m) the IFFT pattern of the green square in (k); (n) HRTEM of the interface between a large spheroidized β-Nb particle and B2 matrix in the CE region; (o) HRTEM image of a Ti-rich particle.

Figure 7 (
b) and table S2 indicate that the Nb content of the rod-like β-Nb phase in the FE region remains essentially unchanged (∼60 at.%) after solution treatment.However, the Nb content of the large spheroidized β-Nb particles (both at the columnar GBs in the FE region and the cell boundaries in the CE region) is increased to around 80 at.% (figures 7(c) and (d), table

Figure 7 .
Figure 7. TEM line scan results showing the Ti, Ni and Nb content across (a) the cell boundary in the CE region for as-printed sample, (b) the cell boundary in the FE region of the S60 sample, (c) a large spheroidized β-Nb particle at GBs in the FE region of the S60 sample, and (d) a large spheroidized β-Nb particle in the CE region of the S60 sample.The locations for the TEM-EDS line scans are given in figure S6 in supplement.

Figure 8 .
Figure 8. EBSD results of the as-printed and solutionized (NiTi) 91 Nb 9 alloys: (a)-(d) EBSD orientation maps; (e)-(h) grain boundary distribution maps; (i) the histogram of EBSD analysis for grain size distribution; (j) the histogram of EBSD analysis for misorientation.LAGB and HAGB indicate the low-angle and high-angle grain boundary, respectively.

Figure 9 .
Figure 9. Room temperature XRD patterns of L-PBF fabricated (NiTi) 91 Nb 9 alloys: (a) at as-printed and solutionized state; (b) the enlarged view of the blue rectangle in (a); (c) the enlarged view of the red rectangle in (a).The XRD patterns are collected on the surface parallel to the building direction.

Figure 10 .
Figure 10.Phase transformation behavior of the as-printed and solutionized (NiTi) 91 Nb 9 alloys: (a) DSC curves; (b) change in Mp and Ap holding time; (c) thermal hysteresis (T hys , Ap-Mp) and ∆H M-A versus holding time; (d)-(h) DSC curves after pre-deformation, where the dashed line is the DSC curve for the sample without deformation; (i) variation of Mp ′ , Ap ′ and Ap ′ -Mp ′ with holding time.For the as-printed sample, the value of Mp temperature is used as the Mp ′ temperature in (i), since the forward A→M transformation peak after deformation is very broad.

Figure 11 .
Figure 11.Tensile results of the as-printed and solutionized (NiTi) 91 Nb 9 alloys: (a) stress-strain curves, where the as-printed part was tested at 218 K, while the solutionized samples were tested at 223 K; (b) variation of the plateau stress (σp), fracture stress (σ f ) and fracture strain (ε f ) versus holding time.The tensile testing for each specimen was repeated three times (see details in figure S11 in supplement).

Figure 14 .
Figure 14.Schematic illustration of the microstructure evolution of (a) the fine eutectic region and (b) the coarse eutectic region for the LPBF-fabricated (NiTi) 91 Nb 9 alloy before and after solution treatment.

Figure 15 .
Figure 15.The mechanism of martensite transformation temperature (MTT) variation: (a) schematic illustration showing the MTT changes versus Nb content for the LPBF-fabricated (NiTi) 91 Nb 9 alloy; (b) schematic illustration showing the MTTs variation versus holding time, where the contribution of factors that increase (red region) and decrease (blue region) the MTTs is schematically shown; (c) DSC curves of the (NiTi) 91 Nb 9 alloys solutionized at 1323 K for different holding time (15 min-120 min); (d) change in Mp versus holding time.