Laser powder bed fusion additive manufacturing of NiTi shape memory alloys: a review

NiTi alloys have drawn significant attentions in biomedical and aerospace fields due to their unique shape memory effect (SME), superelasticity (SE), damping characteristics, high corrosion resistance, and good biocompatibility. Because of the unsatisfying processabilities and manufacturing requirements of complex NiTi components, additive manufacturing technology, especially laser powder bed fusion (LPBF), is appropriate for fabricating NiTi products. This paper comprehensively summarizes recent research on the NiTi alloys fabricated by LPBF, including printability, microstructural characteristics, phase transformation behaviors, lattice structures, and applications. Process parameters and microstructural features mainly influence the printability of LPBF-processed NiTi alloys. The phase transformation behaviors between austenite and martensite phases, phase transformation temperatures, and an overview of the influencing factors are summarized in this paper. This paper provides a comprehensive review of the mechanical properties with unique strain-stress responses, which comprise tensile mechanical properties, thermomechanical properties (e.g. critical stress to induce martensitic transformation, thermo-recoverable strain, and SE strain), damping properties and hardness. Moreover, several common structures (e.g. a negative Poisson’s ratio structure and a diamond-like structure) are considered, and the corresponding studies are summarized. It illustrates the various fields of application, including biological scaffolds, shock absorbers, and driving devices. In the end, the paper concludes with the main achievements from the recent studies and puts forward the limitations and development tendencies in the future.


Introduction
In recent years, shape memory alloys (SMAs) have drawn much attention for their unique functional features (e.g. shape memory effect (SME), superelasticity (SE)) [1][2][3]. At present, there are dozens of SMAs, the most relevant being NiTi, Cubased (e.g. Cu-Zn-Al, Cu-Al-Ni, Cu-Sn, Cu-Mn) [4,5] and Fe-based alloys (e.g. Fe-Mn-Si) [6]. Although Fe-based SMAs and Cu-based SMAs have lower costs, the low corrosion resistance, poor stability, and poor thermal-mechanical performance also limit their applications. NiTi alloy, also known as Nitinol, is the most commonly-employed SMA in the industrial and biomedical fields due to its superior SME and SE performance, good damping capacity, exceptional corrosion resistance, and biocompatibility [7]. The NiTi SMAs exhibit recoverable shape induced by external physical fields thanks to the thermoelastic martensitic transformation. The mutual transformation between the martensite and austenite phases causes the functional properties (e.g. SME, SE) of SMAs. The microstructure, phase transformation behavior, and mechanical property are all significant in recognizing the features of NiTi alloy [8,9]. The unique characteristics have enabled the extensive application of NiTi products in various fields, such as actuators [10], electrical devices [11], biomedical [12], and aerospace [13].
However, due to the strong work-hardening and high tool wear, it is rather challenging to prepare NiTi SMA devices, especially those with complex geometries. Casting and powder metallurgy (PM) technologies are commonly utilized to prepare NiTi products. But the high temperature melting process associated with the casting process is easy to introduce impurities such as carbon and oxygen and form TiC and Ti 4 Ni 2 O x secondary phases, lowering the mechanical and functional performance [14]. During melting, micro and macro segregations are easily formed in the NiTi products [15]. Moreover, unfavorable second phases are usually found in the as-cast NiTi products, adversely impacting mechanical properties [13]. The NiTi alloys prepared by the PM technology generally have inferior mechanical properties [3]. In some PM processes (e.g. conventional sintering, self-propagating high-temperature synthesis), the shape and size of pores are inabilities to control, resulting in the limitation of usage [16][17][18]. Additionally, casting and PM methods encounter limitations in fabricating complex parts owing to the need for mold preparation prior to the forming process [3]. Thus, suitable manufacturing technology is imperative to solve the traditional limitations, including work-hardening, regulation, and control of microstructures, mechanical properties, and different applications of complex parts.
In the past decades [8,[19][20][21], additive manufacturing technology (AM) has been highly anticipated for investigators and engineers to prepare advanced complex components due to the virtually unlimited design freedom and near-netshape production capability. Laser powder bed fusion (LPBF) is regarded as a significant part of AM to rapidly fabricate highly individualized metallic components on demand, such as thin-walled structures and lattice-like shapes without the need for tooling and molds [22,23]. This technique utilizes high-energy laser beam selectively melts metal powder layer by layer according to CAD data until the final part is finished [24]. The LPBF process features extremely rapid melting/solidification and periodic thermal cycles, which endows the as-fabricated specimens with unique microstructure and performance that differ from conventional techniques [25]. The features of the LPBF-fabricated parts can be summarized as follows [26][27][28]: (i) Non-equilibrium microstructure with high supersaturation and metastable or unsteady phase; (ii) Many process parameters that affect the heat and mass transfer process, such as laser power (P), scanning speed (v), hatch spacing (h) and layer thickness (t); (iii) high level of residual stress resulting in low machinability; (iv) low toughness and low fatigue performance. Thus, it is worth of investigating the microstructure and performance of NiTi samples fabricated by LPBF.
Recently, a large number of research focused on LPBFfabricated NiTi alloys. Several reviews reported the related research progresses. For example, Wang et al [29] summarized the unique microstructure, phase transformation behaviors, and mechanical properties of LPBF-fabricated NiTi alloys. Sabahi et al [30] reviewed some NiTi SMAs for biomedical applications, including lattice structures, modulus, and the Ni release. Farber et al [31] concentrated on the factors influencing the phase transformation temperatures (TTs) and some application issues.
The above reports introduced just a small amount of the characteristics or applications of LPBF-fabricated NiTi alloys briefly. A systematic review of the recent research progresses including the control of microstructures and mechanical properties, and development tendencies of LPBF-fabricated NiTi alloys are scarce and imperative. In this paper, we comprehensively conclude the effect of process parameters on printability, microstructural characteristics, phase transformation behaviors, mechanical properties, lattice structures, and some application scenarios of LPBF-fabricated NiTi alloys. Furthermore, some technical challenges, limitations, and development tendencies not mentioned before are put forward in this paper.

Relative densities of different process parameter combinations
The main process parameters include v, P, h, and t in LPBF. And an equation generally calculates the volumetric energy density (E v ): E v = P/(v × h × t) to evaluate the impact of laser energy on the materials. Relative density, which represents an aggregative indicator of porosity and cracks, is an important parameter for evaluating the printability of LPBF parts. The improper parameters can be attributed to microcracks formation usually observed by secondary electron microscopy (SEM). The process parameters profoundly influence defect formation, densification behavior, manufacturing fidelity, and impurity pick-up [8]. Thus, suitable E v and process parameter combinations for high relative density and low-impurity contents are crucial.
In the LPBF process, the similar energy density with different P and v combinations leads to discrepancies in printability, microstructure, phase TTs and mechanical properties. The TTs and evaporation of Ni elements can be altered by scanning speed, while porosity and grain size are affected by laser power [32]. The values of E v used in dense NiTi samples are listed in table 1, and the maximum E v is 222.2 J·mm −3 . Generally speaking, most values of E v were below 100 J·mm −3 .
When the values of E v were on the same level, the parameters of dense samples can be roughly divided into two process parameter combinations: relatively high laser power with high scanning speed (HP) and low laser power with low scanning speed (LP) with an indistinct boundary [33,34]. As shown in table 1, the statistics of the optimum process parameters for different NiTi chemical components and LPBF equipment are listed.
The process parameters typically use HP combination for Ni-rich and near-equiatomic materials. Xue et al [2] determined a proper keyholing criterion to eliminate the porosity near defect boundaries and obtained the dense samples by controlling the laser power, scanning speed, and volumetric energy density (near-equiatomic (Ni 50

Prediction models for process parameter combinations
Although the above traditional experimental methods can confirm satisfactory process parameters, the additional human resources and economic costs are still enormous. Nowadays, various prediction models and optimization algorithms are used for searching and predicting the optimal parameters in AM. Combining experiments and machine learning classification techniques for process parameter design is an effective method of determining process windows. Mahmoudi et al [36] showed that the linear energy density (E l ) was a more precise process design parameter than E v for identifying satisfactory printability. The non-defective samples (Ni 50.9 Ti at.%) were manufactured when E l reached 0.5 J mm −1 . And E v was more relevant in controlling the transformation behavior and TT. An artificial neural network (ANN) was another prediction tool for estimating the optimal operational parameters. The ANN model predicted NiTi samples' strain recovery ratio and TT. The results showed that the multi-layer perceptron models represented reasonable rates of correlation of coefficients between the modified multi-layer perceptron code and experimental data [37].

Microstructure of LPBF NiTi
The mechanical properties and functional behaviors (SME and SE) of NiTi SMAs depend on the microstructural characteristics, including morphology, phase, and texture, which manipulate the phase transformation behaviors during deformation process [44]. The austenite phase (stable body-centered cubic (BCC) B2 phase) and the martensite phase (unstable monoclinic cubic B19' phase) are two main phases in NiTi samples, which exhibit different effects on mechanical properties. Grain size, average grain diameter, and orientation of grains in NiTi alloys significantly influence strength, toughness, and thermomechanical response. Thus, controlling phases, crystallization and grain growth during the LPBF is crucial to regulating the microstructure of the NiTi part with different compositions [38,46,47]. The process parameters significantly influence the phases of Ni-rich and Tirich materials, respectively. For Ni-rich materials, the matrix phase of as-fabricated NiTi samples is usually the B2 austenite phase which is preferentially obtained at room temperature because of the higher cooling rate in the LPBF process. Yu et al [48] discovered higher P would restrict the formation of B19' phase and stabilize B2 phase in Ni 55.8 Ti wt.% samples. Yang et al [49] fabricated the Ni-rich (Ni 55.98 Ti wt.%) samples and reported that the content of martensite phase reduced with the increasing of v. Especially when the v reached a certain value, the martensite would be inconspicuous or even vanished [21,38,49,50]. Qin et al [51] reported that the increase of h from 80 to 110 µm restrained the formation of B19 ′ phase for Ni 55.98 Ti (at.%) samples. In Ti-rich (Ni 49.4 Ti at.%) samples, - [35] Lu et al [52] drew a similar conclusion and calculated the phase fraction of B19' martensite decreased from 99% to 78% with the decrease of E v from 292 to 155 J·mm −3 . Thus, the B19' martensite phase content decreases with the decreasing of energy input, and the increasing of v and h. Based on the preparation of dense LPBF-NiTi parts, the phase compositions are difficult to control because of the complex process parameters. Some research made an attempt on the challenge. When the room temperature is higher than M s at some scanning speeds, the phase composition of NiTi samples (Ni 50.6 Ti at.%) has only an austenite phase without any martensite phase [32]. And the elevated phase TTs will increase the content of B19 ′ phase. Thus, the initial phase constitution is mainly influenced by the M s, which was affected by the chemical composition [49]. Yu et al [53] built a prediction model between the process parameters and phase TTs. Further, they obtained B19' NiTi, B2 NiTi and a composite gradient NiTi by the prediction model, achieving the simultaneous improvement of the strength and toughness.

Influence of precipitate phases.
Some secondary phase and oxide precipitates can also be discovered in LPBF NiTi samples. Ti 2 Ni phase, which surrounds the NiTi matrix phase, is a typical unfavorable phase [50,54,55]. As shown in figures 1(a1)-(a3), the content of TiNi phase decreased and the content of Ti 2 Ni phase increased as the energy input increased due to the loss of Ni [55]. The Ti 2 Ni phase has some adverse effects on the performance: promoting cracks formation along the continuous Ti 2 Ni phase, and disturbing the TiNi matrix chemistry and phase transformation behavior [35]. The reactions in the equilibrium state between Ti and Ni are as follows [55]: the Gibbs free energy change (∆G) of all the reactions is negative, and the reactions can occur spontaneously.  [55]. Reprinted from [55], Copyright (2020), with permission from Elsevier. (b) The optical micrographs of the anisotropic grain structure of the Ni 55 And O could diffuse along the grain boundaries so that the Ti 4 Ni 2 O x oxide precipitates could be observed in the austenitic B2 matrix. The Ti-rich phase depletes the Ti element from the matrix, inhibiting Ni element loss by covering the molten pool. Thus it could be observed that the TT shifted [56]. Therefore, decreasing input energy suitably with lower O content is favorable for the LPBF NiTi samples with lower contents of Ti 2 Ni and secondary oxides. Ni 4 Ti 3 precipitate in the B2 matrix is a metastable phase as well as an essential phase, which usually precipitates in Ni-rich NiTi alloys. Ni 4 Ti 3 precipitated phase also affects the mechanical properties, martensitic transformation path, temperature, and SE performances in NiTi SMA [57,58]. The Ni 4 Ti 3 phase changed from dispersion to aggregation, and the particle sizes increased gradually with the increase of laser energy density, as shown in figures 2(a1)-(a6) [41]. The Ni 4 Ti 3 precipitates profoundly affect morphologies of the B19' variants and thermally-induced martensitic transformation [59]. Uniform dispersion of Ni 4 Ti 3 precipitates benefits martensite nucleation and precipitation hardening [60]. The Ni 4 Ti 3 precipitate can affect the formation of high-density dislocation, leading to the formation of stabilized B19' martensite [42,61]. Thus a stable tensile recovery strain was obtained by the saturation of dislocation, which was attributed to the dislocation stacking around the Ni 4 Ti 3 precipitate or cut-through the lenticular Ni 4 Ti 3 precipitate in figures 2(b1)-(b3) [42].

Grain sizes and grain growth
During the LPBF process, higher energy input causes molten pool deeper and broader, which conducts a higher temperature gradient between the edge and interior of the molten pool and non-equilibrium rapid solidification. Due to the higher cooling rate of the molten pool edges, the ultrafine cellular grains tend to form on the edge of the molten pool and fine dendrite grains tend to form in the interior [52,62,63]. The grains of as-fabricated Ni 50.8 Ti (at.%) alloys showed a square shape with a side length of 100-140 µm close to the h. The second phases or smaller grains were formed at the rectangular grain boundaries [64,65]. As the laser power gets lower, the grain size of the Ni 55.96 Ti (wt.%) sample was refined with a narrower grain size distribution. On the contrary, columnar grains tended to form when the laser power gradually increased. And the regular columnar grains were interrupted by the increasing amount of pores with increasing energy density, as shown in figures 1(b1)-(b4) [38]. Using the alternate process parameters of LP and HP, Wang et al [66] obtained the 'layerstructured' NiTi samples with alternate large and small-sized grains. Lu et al [52] reported the Ni 49.4 Ti (at.%) samples fabricated by 222 J·mm −3 had finer cellular grains and predominant dendritic grains compared to the samples manufactured by 292 J·mm −3 . Nigito et al [67] revealed that the LPBF-NiTi alloys could be considered as 'Metallic Self Micro-Mold Casting' of 'Self Micro-Molding', which exhibited complex mltiscale microstructures with patches, micro-lamellae and nanocells.
To widen the function of LPBF-NiTi, the controllability of microstructure is significant to promote the widespread application as smart components further. The grains in the NiTi alloys usually grow along with the direction of heat dissipation. The location of the laser spot in the molten pool has the highest temperature, causing the grains growing from the boundary to the center [68]. Adjusting scanning strategy is an effective method to alter the direction of heat dissipation and grain growth. Xiong et al [68] manufactured NiTi parts with zigzag grain boundary without unidirectional columnar grains by using a scanning strategy named stripe rotation, which combined the intraformational laser scanning length of 4 mm and scanning direction of 67 • hatch rotation. As shown in figures 1(c1)-(c3), Yang et al [49] fabricated the Ni 55.98 Ti (wt.%) samples and showed that the distribution of crystallographic orientation tended to be more discrete by increasing v, resulting in weak texture. In figures 1(d1) and (d2), Safdel and Elbestawi [69] manufactured the Ni-rich (Ni 55.7 Ti wt.%) samples using two set parameters with the same h (80 µm), and represented that the textures along <001> parallel to the building direction (BD) of the sample fabricated by 1250 mm·s −1 were weaker than those of the sample manufactured by 750 mm·s −1 . Shayesteh Moghaddam et al [70] reported that the Ni-rich (Ni 50 [71] illustrated that the microstructure was related to the critical h (120 µm), which was derived from the synergistic effect of thermal stress and in situ reheating. Epitaxial grain growth and in situ recrystallization were enhanced when h was below this value. On the contrary, irregular grains formed and dislocations induced by thermal stresses decreased. Suitable scanning strategy, lower v, and lower h are beneficial to the stronger texture in the NiTi parts.
Except for the regulation of texture, gradient microstructures, which was helpful for the graded functionality and the controllability of microstructure, have been reported in recent years. Yang et al [72] designed a repetitive laser processing strategy to manufacture functionally graded NiTi alloy. And the result showed that the microstructure gradient varied with the increasing grain size and increasing amount of B19 ′ phase, accompanying with the gradient variations of property and function. Nematollahi et al [73] designed a LPBF-NiTi part with two different microstructures and functional properties (SME and SE) through two sets of process parameters. This functionally graded system achieved different deformation modes at the same temperature, which was meaningful to tailor the properties of the parts.

Phase transformation paths
During cooling and heating processes, the SME and SE of NiTi alloys are based on the existence of a high-temperature austenite phase, a low-temperature martensite phase and phases transformation, which is in connection with phase TTs (M s : martensite start temperature, M f : martensite finish temperature, A s : austenite start temperature, A f : austenite finish temperature). As shown in figure 3, the SME process includes: firstly, when the temperature was below M f , B19' twinned martensite deformed elastically and transformed into detwinned martensite with the increased loading, and NiTi alloys keep detwinned martensite after unloading. Then affter subsequent heating to a temperature higher than A f , inverse martensitic transformation (B19 ′ -B2) occurs in the NiTi alloys and restores the initial shape. Finally, when the temperature drops to M s during the cooling process, the cubic B2 phase transforms back to the initial B19' twinned martensite phase (B2-B19'). The martensite phase transformation finishes after the temperature is below M f , and the residual deformation vanishes in the subsequent cooling process. The SE process is as follows: When the temperature is higher than A f , the austenite phase transforms into detwins martensite (B2-B19 ′ ) at a certain stress level, and the martensite transforms back to austenite (B19 ′ -B2) after unloading, resulting in shape recovery ultimately [8,52,74,75].
In some conditions, including thermal-mechanical cycle, heat treatment, or adding other components (e.g. Fe, Co), Rphase transformation can appear during the process of B2-B19' and overlap with the B2-B19' phase transformation (B2-R-B19') [76,77]. There were two exothermic peaks during the cooling process: the first phase transformation (from the austenite phase to the metastable R-phase) and the second phase transformation (from the R-phase to the martensite phase). The reasons were that the nucleation barriers of the R-phase were lower than that of the martensite phase and excessive Ni resulted from local inhomogeneous [55]. And R-phase transformation is generally in connection with the precipitation of the metastable Ni 4 Ti 3 [78]. The Ni 4 Ti 3 precipitate converts the phase transformation path in Ni-rich NiTi alloys from B2-B19 ′ to B2-R-B19 ′ . Zhou et al [58] built a quantitative phase field model to calculate the spatial variation in the elastic interaction energy between Ni 4 Ti 3 precipitate and R or B19' phase. The aim was to identify nucleation sites and preferred transformation paths for martensitic phases. The reason for R-phase formation before B19 ′ phase formation is stress-induced spatial variation and direct elastic interaction in concentration near Ni 4 Ti 3 precipitates [61]. Lu et al [52] indicated that the high-density dislocations along grain boundaries and interiors of ultrafine grains were beneficial to the formation of R phase.
The phase transformation range widens due to the inhibition effect of defects in the specimens. Some dislocations were generated during the LPBF solidification process [33,79]. While the NiTi samples have a certain number of dislocation boundaries or residual stresses, the phase transformation and the growth of martensite plates can be hindered, leading to more stabilized martensite phases [34,[80][81][82]. And phase transformations cannot occur in the retained stabilized B19 ′ phase during the heating and cooling process. The accumulated dislocations cut through and interacted with lenticular Ni 4 Ti 3 nanoprecipitates, which enhanced the strength and SE, leading to residual strain retained in the samples [42].

Phase
TTs of NiTi alloys are sensitive to chemical composition. The phase TTs mainly depended on the ratio of Ni to Ti and decreased with Ni content increased [83]. Because the boiling point of Ni (2913 • C) is lower than that of Ti (3287 • C), at elevated temperatures, Ni has a high equilibrium vapor pressure than Ti [29,34,84]. Specifically, the M s decreases by approximately 10 • C caused by just a 0.1 (at.%) increase in Ni content [9,[85][86][87][88]. The TTs are disparate between LPBF-fabricated NiTi specimens and the starting materials on account of evaporation of Ni. The decrease of Ni content caused the M s to rise and increased the width of thermal hysteresis during the martensitic transformation [88]. The process parameters and precipitate phases mainly influence the Ni content in the matrix and phase TTs.
The evaporation of Ni element is an inherent characteristic in LPBF-NiTi alloy, directly influencing local chemistry, material properties, microstructures, functionalities, and phase TTs. Thus, the compensation of Ni or the evaporation content of Ni is crucial for the manufacturing of LPBF-NiTi alloy. Ranaiefar et al [89] established a differential evaporation model to predict and control locationspecific chemistry for LPBF-NiTi, explaining the multi-layer design, the inherent melting pool overlap and chemistry propagation.

4.2.2.
Influence of process parameters. The process parameters have different influences on the loss of Ni in the matrix, resulting in different phase TTs, as shown in figure 4(a) [28,45,52]. The TTs showed a remarkable correlation with the E v , as shown in figure 4(b) [36]. It could be explained by the fact that higher E v led to the temperature of molten pools increasing for a more extended period, resulting in the higher evaporate rate of Ni element [39]. Parameter P affects the depth of the molten pool, and v can change the width of molten pool, while h influences the overlap area between two adjacent laser tracks. Thus, it could be realized that the decreased width, the increased depth of molten pool, and the larger overlap area by altering the process parameters resulted in a lower loss of Ni element and increased TTs [48]. Yu et al [48] reported that t significantly influenced the latent heat of transformation, resulting in variational phase TTs. The change of t led the materials to lacking long-range order, resulting in the loss of crystallographic reversibility and the destruction of phase transform characteristics.
It is remarkable that the Ni loss also depended on exposure time which calculated by point distance divided by v, as shown in figures 4(c1) and (c2) [32,48,90]. The evaporated Ni increases with the increasing exposure time, indicating a time-dependent Ni-evaporation. The phenomenon could be explained by the formation of oxides overhead the molten pool impeding the Ni evaporation because of the rapid solidification. Another reason was that the Ni loss in the molten pool surface region was challenging to be wholly compensated by diffusion or Marangoni convection [38,91].
LPBF process is a layer-by-layer process, including multiple thermal cycles and rapid solidification. Melting strategies, including single and multiple melting, influence the chemical composition of final fabricated material. Microstructure inhomogeneity is beneficial to wide TTs and hardly detectable transformation peaks [9,92]. Multiple remelting strategy enhances the chemical homogeneity, and the amount of Ni evaporation increases with the increasing number of melt cycles [43].
The control of phase TTs is a challenging work due to complex LPBF process parameters. Thus, the establishment of phase TTs prediction model is helpful for regulating the phase TTs, as well as the improvement of mechanical and thermomechanical response. Further, the control of phase TTs have a significant impact for different applications. Shi et al [93] established a quadratic regression model between process parameters and phase TTs via response surface methodology, and verified the average error percentage between the predicted and experimental results was within the acceptable range.  [35]. Reprinted from [35], Copyright (2019), with permission from Elsevier. (b) The relationship between the energy density and Ms [48]. Reprinted from [48], Copyright (2021), with permission from Elsevier. (c) DSC curves of NiTi samples fabricated by LPBF with the fixed hatch spacing of 110 mm and fixed layer thickness of 25 µm: (c1) fixed laser power of 60 W and the scanning speeds varied from 300 to 480 mm·s −1 , (c2) fixed laser power of 95 W and the scanning speeds varied from 475 to 850 mm·s −1 [99]. Reprinted from [99], Copyright (2019), with permission from Elsevier. Phase transformation behavior of NiTi samples with (d) different scanning speeds, (e) different hatch spacings, and (f) different laser powers when other processing parameters were fixed [32]. Reprinted from [32], Copyright (2020), with permission from Elsevier. (g) DSC curves of the powder, samples in different orientations, and corresponding phase TTs [95]. Reprinted from [95], Copyright (2021), with permission from Elsevier. (h) The variation of phase TTs under different cooling conditions [97]. Reprinted from [97], Copyright (2012), with permission from Elsevier.

Influence of precipitate phases.
However, sometimes the TTs changed a lot even though at the same energy density, especially under the higher energy density level, as shown in figures 4(d)-(f) [32]. The reason was that the influence on the phase TTs of precipitate phases in the NiTi samples was also an important factor [90]. The formations of Ti-rich secondary phases induced by impurities (e.g. O, C) and Tirich precipitate could change the Ni content of main matrix, further influencing the phase TTs. For example, the increase of O content in the samples (Ni 50 Ti at.%) which went through heat treatment (solution annealing), led to the formation of Ti 4 Ni 2 O x [94]. The formation of secondary phases in the Nirich NiTi alloys can influence the phase TTs. It could be observed that the TTs significantly shifted to lower temperatures in DSC curves with increasing the C and O content. The samples (Ni 50.8 Ti at.%) had higher content impurities compared to horizontal samples, which generated a lower level of TTs in figure 4(g) [95]. The secondary phases could reduce Ti content in the TiNi matrix and decrease the TTs [38,88]. This phenomenon could be explained by a mechanism named 'conductive vaporization'; the titanium oxide covering layer prevented the more volatile Ni from escaping. In the bottom region of HP samples, initial excess oxygen is beneficial to generating Ti-rich oxides. There are empirical equations [35,50,96] to express the relationship between M s and the impurity element content (C, O):  [50]. But it was worth noting that the existence of large amounts of Ti 2 Ni in Ti-rich NiTi alloys could not change the TTs [88]. The cooling rate is also a significant factor influencing the phase TTs. As shown in figure 4(h), the Ni 49.1 Ti (at.%) samples cooled in the furnace had the highest A f and the lowest M f [97]. During the LP process, the reactivity of NiTi powders with oxygen is inferior at lower temperatures, which is suitable for the formation of Ni-rich precipitates. The supersaturated Ni element diffused faster in the NiTi matrix at high temperatures, which promoted the nucleation and the growth of Ni 4 Ti 3 precipitate [42]. And Ni 4 Ti 3 precipitate, which is enabled to generate by a longer response time, could modify the local Niconcentration of surrounding matrix [59,98]. The formation and coarsening of Ni-rich precipitates, which extract excess Ni from the matrix, raise the TTs and impede the propagation of the reversible phase transformation front [79]. On account of the martensitic stability at room temperature, the samples (Ni 55.4 Ti wt.%) fabricated at LP parameters had a pronounced thermal memory response. In contrast, the samples manufactured at HP parameters were mainly composed of austenitic phase [28]. Thus, the bottom region of HP samples had the lowest phase TTs while the top region of LP samples had the highest phase TTs with higher levels of martensite phases [28].

Strain-stress response
The stress-strain curve of NiTi alloys is divided into four stages, as shown in figure 5(a): stage I is the physically elastic deformation stage of twinned martensite phase or austenite phase; stage II is the stress-induced martensitic transformation or the detwinning/variant reorientation stage of martensite phase; stage III is elastic deformation of reoriented martensite or continuous martensite reorientation; and the last stage is the plastic deformation stage of reorientation martensite phase [41,52,100]. The critical stress to induce martensitic transformation (σ SIM ) is usually determined by the tangent method [101,102] (the point of two tangent lines intersect in the stress-strain curves between stage I and stage II). In recent years, as shown in table 2, there has been some research about the tensile mechanical properties of LPBF fabricated NiTi samples with different material compositions.
The NiTi alloys come through the stress-induced martensitic transformation or the martensite reorientation processes before plastic deformation stage, and can bear a large deformation at the low-stress level. In figure 5(a), the excellent elongation up to 15.6% was fabricated with a unique stripe rotation scanning strategy, as mentioned before [68]. As shown in figures 5(b1) and (b2), the ductility values were significantly different: the samples (Ni 50.8 Ti at.%) showed ductility levels of around 9%, while the samples (Ni 50.1 Ti at.%) exhibited better ductility up to 16% [2]. It was because the former expressed mostly austenite phase due to the low M s temperatures, while the latter expressed martensite phase at room temperature. And the austenite phase consists of a relatively stable BCC crystal structure, while the martensite phase deforms easily and has better ductility because of the unstable monoclinic cubic crystal structure. The stress plateau region (between 150 to 200 MPa) of the former includes stress-induced martensitic transformation and martensite reorientation, and stress plateau region (about 150 MPa) of the latter is self-accommodated martensite variants reorientation [2].
The anisotropic tensile properties of the samples, which were fabricated with two different scanning strategies, were evaluated in three orthogonal building orientations [103]. As shown in figure 5(c1), the sample loaded perpendicular to the BD showed higher ultimate tensile strength (UTS) and elongation than the samples in other orientations. Remarkably, the UTS and elongation of the sample, where the loading direction was parallel to the BD, were only about half of the former. At the same time, the scanning strategy had almost no influence on the tensile properties. The reason was that the binding force between continuous layers had significant anisotropy, which is far higher parallel to the BD [103].  [52]. Reprinted from [52], Copyright (2019), with permission from Elsevier. First, the unsatisfactory mechanical properties are attributed to localized defects in the interface between the layers. The defects include the transgranular fracture surface with fine microcracks and dimples, the voids and impurities, and the regions of numerous unmelted powders concentrating on the edges of samples, as shown in figures 5(c2) and (c3) [29,103,104]. Second, due to the columnar grains growing throughout the layers, the cracks can propagate on a flat plane [68,105,106]. In addition, the low O content in the building chamber is suitable for good ductility of as-fabricated NiTi. The presence of O will influence the grain boundary characteristic that existed in quasi-amorphous layers, which destroyed the ductility of NiTi [32,107,108].
Thus, the foundation of high UTS is that the as-fabricated samples have no visible defects, including microcracks and porosities at a low O content environment, as shown in figures 5(d1)-(d3) [52]. On account of the mixture of fine dendritic structure and ultrafine cellular structure, grain refinement strengthening is another reason for favorable mechanical properties [109,110]. And the aggregative Ni 4 Ti 3 phase at grain boundaries impeded the formation and motion of dislocations, which were in favor of enhancing the strength and ductility [111,112]. According to the Orwan mechanism [110], the dislocations form in the dendritic grains pinned by the nanoprecipitates, enhancing the strength and ductility.

Critical stress to induce martensitic transformation.
The unique thermal and mechanical memories include SME and SE. They are based on the forward martensitic transformation and the reverse martensitic transformation [3,8,45,75].
As shown in figure 6(a), σ SIM is the stress level at which the initial twinned martensite in NiTi begins detwinning, representing the functional properties from the viewpoint of stress. And the recoverable mechanical strain is another value for evaluating the functional properties from the viewpoint of strain [49]. The maximum applied stress before NiTi samples occurred in a large transformation [41]. And the σ SIM , which depends highly on the TT at certain temperatures, could be explained according to the Clausius-Clapeyron equation [48,102,114]: where σ critical stress, M s martensitic transformation starting temperature, ∆S the entropy per unit volume during transformation, ε the linear transformation strain, ∆H the enthalpy per unit volume during transformation, and T 0 the phase equilibrium temperature of matrix and martensite phase. The equation gives a linear relationship between the σ SIM and the M s , which means σ SIM increases with the decreasing M s at constant testing temperature. Thus, the variation of Ms temperature and initial phase constitution are two crucial factors for the evolution of functional properties [49]. Some Ti-rich secondary phases could depress martensitic transformation and decrease M s , causing the σ SIM to increase. Farjam et al [115] reported the σ SIM of different geometries and sizes. The results showed that the smallest samples possessed the highest σ SIM while the largest samples had the lowest σ SIM . The reason is that different size samples undergo different thermal histories. During the LPBF process, the cooling rates of external regions are higher, resulting in less time for grains to grow and finer grains. The σ SIM gradually decreases, and the stress hysteresis increases as the number of cycles increases.
The σ SIM decreases when the laser energy input increases. For example, the σ SIM increases as the scanning speed increases, as shown in figure 6(b), resulting in higher mechanical recoverable strain, lower M s , and lower volume fraction of the B19' phase [49,50,99]. When the applied stress exceeds the stress for plastic deformation, martensite occurs in plastic deformation. Thus, the applied stress level is crucial for the recovery deformation of NiTi parts [41]. When the thermomechanical load (i.e. applied stress, strain, or temperature) exceeds the maximum recoverable strain, the residual strain will increase gradually after unloading or heating. The reason is lattice defects generated by tensile deformation exist in the microstructure, which influences the SE and SME [116][117][118]. As shown in figure 6(c), the deformation twinning in oriented B19' martensite leads to permanent deformation. It restrains the recoverability of transformation strains when the applied stress level exceeds the martensite yield limit, which depends on the austenitic microstructure [119,120]. But restricting the applied stress in the first cycle is unhelpful because permanently unrecoverable strains accumulate in mechanical cycles [121,122].

Superelasticity response.
The content of Ni has a significant influence on the thermomechanical response. Generally speaking, Ni-rich samples exhibited obvious SE, while Ti-rich samples performed SME at room temperature, as shown in figure 6(d) [9,78,123]. The samples fabricated by LP condition usually exhibited higher SE recoverable strain. For example, as shown in figure 6(e), in the single loadingunloading compressive testing, the sample in the LP condition showed a better superelastic response, and the highest energy level sample exhibited the best recovery ratio. Meanwhile, the σ SIM of LP samples was obviously higher than those of HP samples [39]. The samples (Ni 50.8 Ti at.%) fabricated by different energy levels exhibited different strain recoveries, while the σ SIM were almost similar [39]. As shown in figure 6(f), the σ SIM and stress hysteresis decreased with the increasing number of cycles, and the irrecoverable strain accumulated due to the martensite plastic deformation. After 10 cycles, the stabilized tensile superelastic response of the LP sample was higher than others, and the recovery strain reached 5.5% [39].
When different processing parameters were used to fabricate the samples at the same energy level, they demonstrated distinct superelastic responses in terms of σ SIM , hardening, and hysteresis, although they exhibited a similar recovery strain [39]. The maximum recoverable and peak strain increase with the number of cycles, achieving stability eventually. At the same time, austenite and martensite elastic moduli, transformation stresses, and dissipation energy decrease [124,125]. And the SE obtained by a constitutive model, which was based on the minor strains and microplane theory, was consistent with the experiment [125]. The total strain reached 9.28% of the samples (Ni 55.98 Ti wt.%) during the cyclic compression tests under a fixed load of 600 MPa, while the recoverable strain reached 3% [51]. The irrecoverable strain increased by the laser power increasing obviously, and the recoverable strain decreased accordingly [51].
The as-fabricated samples (Ni 50.1 Ti at.%) in LP condition showed a negligible strain recovery after unloading at room temperature and a significant strain recovery after heating. In contrast, the samples in HP condition exhibited a large recoverable deformation during unloading and no apparent  [49]. Reprinted from [49], Copyright (2019), with permission from Elsevier. (c) Lattice defects in the microstructure of the NiTi wire deformed at different strains (0%, 10%, 12%, 14%, 20%, 53%) [116]. Reprinted from [116], Copyright (2019), with permission from Elsevier. (d) The deformation recovery process of LPBF samples with different Ni contents [78]. Reprinted from [78], Copyright (2021), with permission from Elsevier. (e) Superelastic response of the LPBF Ni 50.8 Ti samples B1-B4 at LP parameters and the samples C1-C4 at HP parameters, (f) Cycling response of the samples at the lowest energy level [39]. Reprinted from [39], Copyright (2018), with permission from Elsevier.  [33] Reprinted from [33], Copyright (2020), with permission from Elsevier. recovery after heating. This phenomenon is attributed to different TTs. Austenitic phase transformation completed below room temperature of HP samples with low TTs, resulting in the obvious PE at room temperature. In contrast, as shown in figure 7(a), the LP samples, consisting of mainly martensite phase, occurred the orientated growth of the internally twinned martensite during loading and reversed martensitic transformation after heating [45,74]. Some aged samples (Ni 51.4 Ti at.%) exhibited perfect SE with an almost full recoverable strain during manifold cycles (about 4.6% at the first cycle), while the as-fabricated and solution-treated samples showed some residual strain after first unloading as shown in figure 7(b) [33].
Achieving higher tensile SE response is a major difficulty of LPBF-NiTi SMA, which have been researching for a long time and making considerable progress in recent years. Liu et al [126] showed that high density of dislocations could promote the formation of nanocrystalline and amorphoslike phases, which was beneficial to the enhancement of SE response with a stable pseudoelasticity of 5.8%. Shi et al [127] researched the impact of different crystal orientations on superelastic response, showing that the sample with a strong (0 0 1) texture possessed the highest recoverable superelastic strain of 7.91 ± 0.14%.
During the loading process, there is no permanent plastic deformation by dislocation slid in the aged samples, in which the stress hysteresis is the lowest. The dislocation strain field, generated in the first loading-unloading cycle, is beneficial to stabilizing stress-induced martensitic variants, resulting in the martensitic variants being difficult to transform back to austenite during unloading [128]. The dislocation and residual martensite phases generate beneficial internal stress for subsequent phase transformation, decreasing the σ SIM in the follow-up cycle tests [129,130]. As a consequence, as shown in figures 7(c)-(e), after several cycles, the as-fabricated and solution-treated samples showed fully reversible strains, indicating that the stress-induced martensitic transformation, the σ SIM , and the stress hysteresis tended to be stable [33]. As for the aged samples, the nanoscale fine coherent precipitates, which hindered the dislocation slip and enhanced the precipitation hardening, were dispersed in the matrix [9,65]. Thus, the increased yield strength restrains the generation of dislocation during the stress-induced martensitic transformation. The low-stress hysteresis would lead to low energy dissipation during the phase transformation and reverse transformation process [80,129]. In summary, the high yield strength and the low-stress hysteresis are in favor of stable superelastic response [33].

Shape memory effect.
Cyclic tensile tests and the subsequent heating processes investigate the SME responses of samples (Ni 50.1 Ti at.%). As shown in figure 8(a), in different BDs, the horizontal samples exhibited the highest recoverable strain of 3.83% in the first cycle, while the vertical samples demonstrated the lowest strain recovery significantly [103]. The increasing number of cycles stabilizes the corresponding recovery response, which could be attributed to the fewer microstructural defects and the loading direction perpendicular to the BD. Some samples (Ni 50.1 Ti at.%) showed a recoverable strain up to 4% after 6% deformation in incremental loading tests, as shown in figures 8(b) and (c), while the samples (Ni 50.8 Ti at.%) exhibited a recoverable strain of 6% after 8% deformation [2]. The shape memory stability of samples (Ni 55.92 Ti wt.%) with different process parameters was evaluated by heating at A f + 30 • C after ten cycles [48]. The sample of the most excellent shape memory stability exhibited the maximum recoverable ratio of 76.1% and recoverable strain of 3.95% after heating [48]. The slipped dislocations resulted in an inevitable irrecoverable strain after heating. The BD plays a significant role in functional properties. As shown in figure 8(d), Nematollahi et al [95] revealed the deformation anisotropy by researching the influence of BD on the shape memory properties. The samples of 0 • and 90 • showed a texture of (001), while the samples of 45 • showed a texture of (110) along the loading direction. The former expressed a more than 5% compressive recovery strain, and the latter exhibited a 3.95% stabilized recovery strain at the first few cycles under 800 MPa of compressive load.
An isobaric thermal cyclic test is used to evaluate the thermomechanical properties of LPBF NiTi alloys by a dynamic mechanical analyzer. The process is that the samples are applied by a constant stress isothermally at a temperature above A f and thermally cycle between this temperature and a temperature below M f . The applied stress level increases gradually after a complete thermal cycle until the maximum transformation strain is obtained [32,65]. Under low-stress levels, there is a large temperature window in the forward transformation, while martensite transforms back to R-phase and austenite with a small temperature window during the heating process. It is evident that R-phase transformation is stable, and R s changes inconspicuously with stress while M s increases linearly. When M s increases above the R s , R-phase transformation could not be observed and phase transformation turned into single-step transformation in figure 8(e) [131]. The phase TTs gradually rise, and the recoverable strain increases as the applied stress increases because of the phase transformation during mechanical tests. As shown in figures 8(f) and (g), the slope of strain-temperature curve in the cooling process was lower than that of the heating process under the lower applied stress level [7,65]. In combination with the DSC results, the phenomenon could be explained by the large temperature difference between R s and M f during the cooling process, while the temperature difference between R f and A s was small during the heating process [65,132]. The recovery rate, the plateau stress, and the actuation temperature have a monotonous relationship with scanning speed, hatch spacing, or laser power. Establishing a quantitative relationship between the SME and process parameters is helpful in fabricating functional graded NiTi structures [32].

Damping properties and hardness
It is well known that NiTi alloys have high damping capacity related to austenite-martensite phase boundary motion, twin boundary motion, stress-induced phase transformation, and temperature-induced phase transformation [133]. The damping factor peak (tan δ) is affected by phase transformation properties, transformation kinetics, and microstructures. The high damping capacity is attributed to high internal friction during the martensitic transformation and increased by the formation of stress-induced martensite [134][135][136].
The damping properties of NiTi parts are used to preserve sensitive equipment and instrumentations from structural vibrations and environmental noise [137][138][139]. During the LPBF process, the samples fabricated at HP parameters possessed more dislocations and thermal defects which impeded the phase transformation and the growth of martensite plate [34,140]. At the same time, the LP samples exhibited a more intense martensitic transformation, favoring a high density of mobile interfaces. And it was helpful to improve energy dissipation resulting in higher damping capacity, as shown in figures 9(a) and (b) [28]. Wang et al [66,90] fabricated layer-structured NiTi samples (Ni 55.7 Ti wt.%) alternately using two sets of process parameters. At a certain temperature range, the samples expressed different transformation behavior in different layers, resulting in  [7]. Reprinted from [7], Copyright (2018), with permission from Elsevier. austenite and martensite alternating phases. As shown in figures 9(c)-(e), the samples exhibited a high damping performance at low (1 Hz) and high (90 kHz) oscillation frequencies during the martensitic transformation [66,90]. It was suitable for fabricating functionally graded or composite microstructures.
The applications of NiTi damper involve low and high strain amplitudes under isothermal conditions [141]. At low strain amplitudes, internal friction influenced damping capacity, which mainly existed in the martensite phase. At high strain amplitudes, NiTi parts would dissipate the amount of energy because of the characteristic hysteretic mechanical cycles. The austenite phase showed higher damping capacity at 4%-8% strain amplitudes in figures 9(f1) and (f2) [142,143]. Thus, the damping ratio depends on the phase ingredient, meaning that the temperature influences the damping properties [143,144]. When the testing temperature was lower than A s , the heterogeneous lattices of martensite in the samples (Ni 55.96 Ti wt.%) created twin boundaries between the variants by mechanical stresses, resulting in the creation of energy dissipation and large damping characteristics. When the testing temperature was above A f , there was no existence of twin boundaries in the homogeneous cubic austenite lattice without obvious damping. As for the martensite and austenite phases coexisting condition, the stress-induced phase transformation would lead to irreversible energy conversion. Thus, the best damping property was found at A s temperature [145]. The damping by thermo-elastic promotion was suitable at higher frequencies, while the damping based on defects, twin boundaries, and grain boundaries had to be considered at lower frequencies [139,146].
Hardness is also an important mechanical property indicator, impacting the material applications' range. The hardness of NiTi alloy is complicated and highly depends on the existing phase and the temperature. The hardness of austenite phase is significantly larger than that of martensite phase [147]. Thus stress-induced martensitic transformation or martensite variant reorientation and dislocation have a great influence on the Vickers hardness [65,131]. The atomic percentage had a significant effect on the Vickers hardness, which sharply increased from 372.4 HV (Ni 50.73 Ti at.%) to 724.5 HV (Ni 51.27 Ti at.%) with the Ni content increasing [78]. It was because that the distance between precipitate and particles decreased [78]. Saedi et al [65] reported the influence of aging processes on the variation of Vickers hardness. The Vickers hardness increased gradually and dropped drastically when the aging temperature exceeded 450 • C. The reason was increased precipitates, inter-particle distance, and the higher aging temperature. The regularities of Vickers hardness values distribution followed that the highest hardness value of 560 HV was closer to the center of the sample while the lowest value of 235 HV was located near the surfaces. This phenomenon could explain by the central region of dense samples having fewer cracks and pores, while the surface region had a higher number of defects [148].

Lattice structures and applications of LPBF NiTi
Nowadays, one of the most potential application prospects of additive manufacturing NiTi materials is lattice structure design, which expected superior performance of high specific strength, lightweight, improved damping properties, and higher deformation recovery. An ingenious design could integrate extra functionalities geometrically dependent into the AM products [149].
Li et al [150] built a Negative Poisson's Ratio NiTi-based structure, which could use for multi-functional reusable armor, as shown in figure 10(a). It was observed that higher ductility in the lattice structures underwent the homogenization treatment, which reduced the fraction of Ti 2 Ni intermetallic. Due to the rapid solidification accumulating residual stresses, it could be observed that the phase TTs of diamond-like geometry lattice structures were higher than those of the bulk samples and the lattice structures existing as a part of stressinduced martensite. And in the lattice structures, grains were oriented along preferential directions rather than the fully austenitic bulk samples [149].
Except for the strut-type structures, the triply periodic minimal surfaces (TPMSs), which had some better properties, had been researched extensively. For example, as shown in figure 10(b), the static mechanical properties and fatigue resistance properties of the cellular gyroid and sheet gyroid structures were higher than traditional octahedron beam lattice structures [151]. Yang et al [152] utilized laser selective melting technology to fabricate TPMS structures ( figure 10(c)) and studied the effects of volume fraction and unit cell size of gyroid structures on mechanical properties and shape recovery behavior, found that mechanical properties showed positive dependences on the volume fraction but negative dependences on the unit cell size; while both the volume fraction and unit cell size had little influence on the martensitic TTs as well as the total shape recovery ratio of Ni-Ti Gyroid TPMS lattice structures prepared by LPBF.
The manufacturing fidelity of LPBF lattice structures is a challenge for the achievement of structure properties and functions. Some scholars focused on Micro LPBF (µ-LPBF), which presented enormous potential in fabricating complex metallic components. Khademzadeh [153] fabricated singlephase austenitic NiTi dense and porous materials with the introduction of µ-LPBF, decreasing the deviation of NiTi lattice structures with respect to the predesigned models. The NiTi parts fabricated by µ-LPBF have unique melting pool morphology with fan-shaped grains in the middle and nearvertical grains on side shoulders, consisting of dislocations of comparable volume density and smaller Ni 4 Ti 3 precipitates compared to the traditional LPBF NiTi [154].

NiTi-based biological scaffolds
NiTi SMAs are a kind of special biomaterials which have stable SME and super elasticity, excellent corrosion and wear resistance, low elastic modulus, large deformation recovery rate and biocompatibility. Therefore, they are widely used in dental and biomedical fields, such as orthodontic fixators, bone substitutes, and cardiovascular equipment. Traditional porous biological scaffolds use materials with high stiffness, such as titanium alloy, stainless steel, etc, whose modulus is much higher than human bone [155,156]. When they are implanted into the damaged position, it is easy to cause stress shielding and make the damaged position difficult to repair [157,158]. Secondly, these large modulus metal materials are difficult to resist large deformation. When the human body is subjected to an external load, it is easy to deform and difficult to recover. The low elastic modulus and large deformation recovery rate of NiTi SMA make it an ideal candidate material for metal bone replacement. However, single-bulk NiTi materials are difficult to meet the various physical and biological performance requirements of biomaterials, such as strength, permeability, biocompatibility, etc.
In recent years, metal porous biomaterials have clarified the relationship between structure and performance [23,159,160]. The geometric and topological properties of porous biomaterials can regulate the mechanical properties  [150]. Reprinted from [150], Copyright (2016), with permission from Elsevier. (b) Three types of TPMS structures and the samples manufactured by LPBF [151]. Reprinted from [151], Copyright (2017), with permission from Elsevier. (c) Schematics and samples of Gyroid surface, skeletal Gyroid unit cell, and skeletal Gyroid TPMS lattice structure with different volume fractions and unit cell sizes [152]. Reprinted from [152], Copyright (2022), with permission from Elsevier. of biological scaffolds, adjust their biological mass transport properties and affect the transportation of nutrients and metabolites. Compared with bulk materials, porous biomaterials have significant advantages in performance regulation. Using NiTi SMA to design porous scaffolds with preciselycontrolled interconnected structures would provide a unique possibility to reduce the equivalent stiffness, thereby increasing flexibility, effectively reducing the stress shielding effect and the risk of aseptic implant loosening, as well as increasing the surface area to improve cell oxygenation and implant bone integration. Traditional manufacturing for preparing the porous NiTi alloys includes conventional element sintering, mechanical alloying, spark plasma sintering, self-propagating high temperature synthesis, metal injection molding, hot isostatic pressing, electro-assisted powder metallurgical, direct foaming, etc.
Generally speaking, typical porous structures include two types of periodic regular porous structures composed of many periodically repeated element geometries, namely beambased elements [66,161,162] and plate/shell-based elements [163][164][165], such as topologically ordered lattice/microlattice structure and TPMS structure. In addition, periodic porous scaffolds have also been studied, such as anisotropic microlattice [166,167] and deformed microstructures [168]. As shown in figures 11(a)-(d), Habijan et al [169] evaluated whether the NiTi lattice structure can be used as a scaffold for human mesenchymal stem cells, designed compact and porous scaffolds with variable porosity and surface area, and studied the effect of structural parameters on biocompatibility. Dadbakhsh et al [170] researched the mechanical behavior and shape memory responses of the different volume fractions of octahedron-shaped porous scaffolds. And the results showed that HP samples had a large geometrical mismatch with the original design and a higher volume fraction with thickened struts, which were beneficial to load-bearing performance against directional loading. As shown in figures 11(e) and (f), Lu et al [171] studied the microstructure, shape memory properties, and in vitro biocompatibility of porous TPMS-based NiTi scaffolds fabricated via LPBF. The above studies on the combination of NiTi and porous structure are limited to the existing topological structures, i.e. lattice structures and TPMS structures, and do not consider the special shape recovery and hyperelastic mechanical properties of NiTi alloy. How to choose a suitable topology for NiTi biological porous scaffold design is still a challenging topic, which lies in the effective combination of the characteristics of low modulus and Figure 11. Biocompatibilities of LPBF NiTi scaffolds evaluated by cell growths. Quantification of (a) interleukin-6, (b) interleukin-6, (c) vascular endothelial growth factor, and (d) the Nickel release of different specimens [169]. (e) The optical density values of the porous LPBF NiTi scaffolds, bulk NiTi samples and control group, (f) cell adhesion rate of LPBF porous NiTi scaffolds and bulk NiTi sample for different culture times [171]. Reprinted from [169], Copyright (2013), with permission from Elsevier. Reprinted from [171], Copyright (2021), with permission from Elsevier. high recovery deformation and topological geometry, the constraint of shape topology by additive manufacturing process, and the regulation of various mechanical and mass transport properties.
Bionic design is the artificial construction of components by imitating the geometric characteristics, performance attributes and functional properties of natural substances from the micro-scale to macro scale. It has potential application value in the field of biological scaffold design. For example, Feng et al [172] designed and prepared biomimetic materials with lotus root-like structures that can significantly improve in vitro cell attachment and proliferation as well as promote in vivo osteogenesis. Wang et al [173] fabricated a cell-container-like scaffold composed of bioceramics containing four pore structures, including triangles, squares, parallelograms, and rectangles. Inspired by the fractal nature of bone scaffold, Zhao et al [174] designed Menger sponge-like fractal geometries created by the fractal iteration algorithm, then fabricated a series of Menger sponge-like fractal geometries with different hierarchically architected topologies, involving three ranges of fractal pore sizes using NiTi alloy. The results showed that the NiTi Menger sponge with fractal topology, nanostructured microstructure, highly controllable elastic modulus and large recoverable deformation present a promising candidate for metallic implants. Therefore, we believe that constructing NiTi biological porous scaffolds by imitating the multi-level structural design in the biological world can give full play to the hyperelasticity and shape memory characteristics of NiTi alloy.
However, the current research still focuses on the supporting effect of NiTi's special mechanical response on the mechanical properties of scaffolds. A suitable biological scaffold also needs to consider biological properties such as matter transport and antibacterial performance. Subsequent studies should focus on the effects of NiTi's shape memory deformation and recovery process on other physical and biological properties except for mechanical properties. For example, a technical challenge of the applications of NiTi-based biological scaffold is the release of Ni ions, poor corrosion performance, and biological intertness. There were some research about the NiTi-based scaffolds to overcoming the challenge through the coating preparation. Guo et al [175] reported that the preparation of graphene oxide coating on the NiTi SMAs benefited to the corrosion resistance, inhibiting the release of Ni ions, as well as the facilitation of the adhesion, growth, and proliferation of osteoblasts. Similarly, they prepared a bioactive dicalcium phosphate dihydrate (DCPD) coating on the LPBF-NiTi alloys via electrodeposition to improve the corrosion resistance and biocompatibility of the LPBF-NiTi [176]. The DCPD coating offered a good interface and microenvironment for the growth and adhesion of osteoblasts by in vitro cell experiments. Wu et al [177] prepared a bioactive coating rich in Ca and P elements by plasma electrolytic oxidation (PEO), Figure 12. Potential applications (energy absorpation, and actuator) of NiTi structures. (a) A NiTi honeycomb sample before/after compression and after heating recovery, (b) percentages of different strain components as functions of the global strain, (c) evolution of the shape-recovery rate under 15 deformation-recovery cycles, (d) schematic of the repeated landing of a landing leg when filled with NiTi architected materials as the energy-absorbing material [179]. Reprinted with permission from [179]. Copyright (2021) American Chemical Society. (e) Placement of hinges and pairs of SMA actuators in airfoil architecture to maximize deformability and to efficiently adjust the shape, and the morphing wing cross-section with one of the actuation pairs [182]. Reproduced from [181]. © IOP Publishing Ltd All rights reserved. (f) Two basic designs of the composite actuators with two hinges [185]. Reprinted from [185], Copyright (2021), with permission from Elsevier. and demonstrated that PEO process could effectively inhibit the release of Ni ions and improve the corrosion resistance and biological activity.

Shock absorbers and driving devices
SMA is a widely used intelligent material. In addition to SME and SE, it also has large damping hysteresis characteristics and can achieve stiffness change through geometric design. Therefore, SMA is a potential shock absorber application material.
In the field of aeronautics, there are vibration problems in the process of aircraft taking off, flying and landing. A vibration damper is effective engineering equipment used to suppress the vibration of machines and structures. Liu et al [178] proposed a new type of shock absorber by combining the advantages of piezoelectric materials and SMA materials. They studied the influence of number and arrangement of SMA wires on the vibration reduction effect of the structure, and experimentally verified the influence of the piezoelectric ceramic control strategy on the vibration reduction effect of the structure. As shown in figures 12(a)-(c), Xiong et al [179] proposed a damage-tolerant architecture by a novel constituent material deformation design strategy, which can solve the problem of engineering mechanical instability manifested as the emergence of localized deformation bands and collapse of strength. As shown in figure 12(d), they depicted a lander application with multiple landing legs that were filled with NiTi architected materials as the energy-absorbing material. When the natural frequency of the tuned shock absorber is equal to the forced frequency, the tuned shock absorber offsets the force exerted on the main system by allowing its own vibration. The common design method of the tuned shock absorber is to add variable stiffness elements to absorb the vibration energy of the main structure. Shaw and Wang [180] used two tunable vibration absorbers composed of SMA to reduce the vibration of a platform structure. They achieved an average of 12.69 dB vibration suppression for resonant excitation of the entire platform structure using the designed SMA tunable vibration absorber. In addition to complex mechanical systems, the SMA shock absorber can effectively reduce the transverse vibration amplitude of the rotating system. Braga et al [181] dealt with the passive control of mechanical vibrations of rotating machines using wires of SMA and verified that the temperature of the SMA wire could be employed to control the vibration control system.
In the process of deformation and recovery, SMA material would produce a large mechanical force and geometrical displacement, which can be used to drive the control system. Therefore, SMA is often used in driving devices. Compared with traditional drives, SMA drives operate silently, have a smaller scale, require a lower driving voltage, and can realize a higher power density.
Recently, with the developed concept of 4D printing and intelligent materials, uncrewed aerial vehicles and intelligent deformable wings have been paid close attention to unprecedentedly. In addition to mechanical programming to customize motion rules, deformable wings also need the support of intelligent materials. It is easy to think of using SMA to manufacture wings and using the force and bending moment generated by SMA in the heating and cooling states to control the movement of wings, thus having an intelligent tunable effect. The configuration of the target shape was obtained through SMA driving devices, and the preset shape was obtained in different flight stages, then realizing the shape control of flexible and deformable aeronautical structures as shown in figure 12(e) [182]. SMA driver operates under a high cycle load and is subject to strict space constraints. The pulley system can increase the flexibility of space constraints, convert linear motion into rotary motion, and use mechanical advantages to realize load or stroke amplification. For example, the SMA pulley system can linearly drive the opening area of the exhaust nozzle of a jet engine to realize the amplification effect of load and stroke [183]. SMA actuator is also used in the control system of a flexible robot. The flexible mode and joint variables are modeled as fast and slow, respectively. The torque of the joint motor controls the slow subsystem, and the fast subsystem is actively controlled by the SMA actuator, which can ensure the uniform index stability of the controller [184].
In addition to the single material driver, other scholars have embedded SMA wire into the polymer matrix to manufacture SMA-SMP composite driver. Specifically, the SMA wire is embedded into the hinge structure to provide bending driving force, and the resistance wire is embedded into the SMP layer to change the temperature and bending stiffness of the driver hinge through Joule heating. The phase change temperature of SMA wire and SMP material changes according to the design to achieve large bending deformation shape preservation and shape recovery in figure 12(f) [185]. At the micro-scale, a novel shape memory actuator based on nano platinum film can realize bending deformation under voltage induction. The actuator can construct electrically responsive microrobot components, such as origami 3D structures, deformed metamaterials, and mechanical memory components [186].

Limitations and development tendency
There has been much research on LPBF-manufactured NiTi alloys in recent years. The process parameters and material components of NiTi alloys greatly influence the formation of defects, densification, element nickel evaporation, manufacturing fidelity, and impurity pick-up. These works focus on the effect of process parameters and material components on discrepancies in microstructure, phase TTs, and mechanical properties of the LPBF-manufactured NiTi alloys. NiTi products can integrate extra functionalities geometrically dependent on the structure configurations, such as negative Poisson's ratio structures, diamond-like lattice structures, cellular gyroid, and sheet gyroid structures. The low elastic modulus and large deformation recovery rate of NiTi lattice structures make it an ideal candidate structural material for metal bone implants. NiTi SMA has potential applications in the shock absorber because of the large damping hysteresis characteristics and can further drive the control system due to the manipulative ability of the deformation recovery. Thus, the design of lattice-based structural components for LPBF-manufactured NiTi alloys has unparalleled application in industrial engineering for brakes and drives.
Despite such a significant number of research, some limitations need to be breakthrough and a lot of work remains to be done to obtain higher performance and more comprehensive applications. We summarize existing limitations and possible development tendencies to push the limits.
1. Because of the complex characteristic of LPBF and NiTi alloys, the controllability of microstructures is significant to expand its applications. Although there is some research about the phase compositions influenced by the process parameters and precipitate phases, the initial phase compositions are also difficult to confirm. The prediction model between the process parameters and compositions is a possible method to solve the problem. However, a large amount of experiments and the universality of model remain to be resolved. In addition, the regulation of texture and gradient microstructures are beneficial to achieve some unique deformation modes, which need to be developed. 2. During the LPBF process, the evaporation of Ni element is unavoidable, which has great influence on the phase TTs.
To control the process stability, we need to clarify further the relationship between the LPBF process parameters and the Ni evaporation. Thus, a systematic approach using design of experiment to establish dynamic relation between the process parameters and the evaporation content is vital to control the local chemistry, phase TTs, microstructures, material properties, as well as functionalities. Furthermore, a general algorithm, accompanied by the input of process parameters and the presumptive output of the ratio of Ni to Ti, phase TTs, and even the recovery response, maybe is meaningful for controlling the phase transformation behavior. 3. Except for the traditional experiment methods, the novel NiTi component for LPBF (e.g. high-temperature NiTi alloy, fatigue-resisting NiTi alloy, and wear-resisting NiTi alloy, etc) with the help of some powder design methods (e.g. thermodynamic and kinetic calculation, phase-field simulations, high-throughput calculation, machine learning and so on) possess great necessity and development potential. 4. Some researchers reported that the optimal process parameters for dense samples were suitable for the porous structures. But there are some geometric deviations between the structure models and the samples, and some attempts to improve the manufacturing fidelity via µ-LPBF. For traditional LPBF, the process parameters maybe devoted to the porous structures to reduce the geometric deviations are in great need of researching and discerning. And the method to decrease the deviation of NiTi lattice structures with the predesigned models is yet to be explored. 5. The LPBF-NiTi lattice structures in some reports can be applied with other materials. So we need to design structures that can take maximum advantage of the SME and SE properties of NiTi alloys, considering the maximum recoverable strain, failure strain and stress, phase TTs, and so on. Different positions of NiTi structures fabricated with varying parameters of process or different material components can achieve a disparate level of recoverable deformation, realizing the application of different scenarios. Taking into account the combination of different pore shapes, pore sizes, strut shapes and strut sizes to design some bio-mimetic structures is a promising direction. 6. The current research about lattice structures focused on the stress-strain response, lacking of the attention of phase transformation behavior, phase composition, texture, and thermomechanical response. The forming process of lattice structures have great differences from the bulk sample, such as the melting and heat dissipation process. The relevant research are beneficial to the application of LPBF-NiTi alloys. 7. There are two examples of applications, including biomedical scaffolds due to the low elastic modulus and biocompatibility and aerospace braces due to the prominent damping hysteresis characteristics. But there are some restrictions in these applications (e.g. the toxicity and ion release, corrosion performance, etc). Some research reported that coating preparation was an effective method, using complicated preparation technologies. We still need to overcome other shortcoming of NiTi porous structures in the applications through further research. In the future, more work is required to embroad the application sphere, including but not limited to the application of industrial, automobile, deep-sea exploration, etc.

Summary
In summary, LPBF NiTi alloys have been drawing much attention on account of their unique performance in recent years. We review the current state and recent achievements of LPBF NiTi alloys, including the effect of process parameters on printability, microstructure, phase transformation behavior, mechanical properties, and some lattice structures and applications. Although some limitations and shortcomings need further investigation, the unique characteristics of LPBF NiTi alloys make them rapidly develop and achieve application in many fields like aerospace, national defense, biomedicine, and others.