Optimising Additive Manufacturing of NiTi and NiMnGa Shape Memory Alloys: A Review

Abstract

NiTi and NiMnGa stand out as prime thermal and magnetic shape memory alloys (SMAs), possessing a superior shape memory effect (SME) and superelasticity (SE). These alloys have crucial current and potential future applications across industries. Additive manufacturing (AM) offers a transformative approach to fabricating these materials into complex geometries; however, the quest to create integral additively manufactured structures with reliable thermal or magnetic shape memory properties remains a recent and fast-emerging research frontier. This article provides a comprehensive review on (i) the intricate principles giving rise to the thermal SME and SE in NiTi, and the magnetic SME in NiMnGa alloys, emphasising their specific relevance in the realm of AM, and (ii) the latest developments, recent findings, and ongoing challenges in the AM of NiTi- and NiMnGa-based SMAs, including their functional lattice structures. Based on this review, for the first time, novel, empirically derived AM process design maps tailored to maximise SME and SE in laser powder bed fusion- and directed-energy deposition-processed NiTi structures are proposed. Similarly, promising avenues to resolve the key challenges regarding the AM of NiMnGa magnetic SMAs are suggested. This article concludes by outlining the most promising future research directions shaping the trajectory of AM of these SMAs.

1. Introduction

Shape memory effect (SME) is a unique phenomenon governed by reversible and diffusionless microstructural alterations, allowing shape memory alloys (SMAs) to recover their original, pre-deformed shape upon the application of an external stimulus. Superelasticity (SE), a related behaviour, enables large, reversible strains to be created solely via the removal of the applied mechanical stress. The most promising SMAs are NiTi and NiMnGa, which respond to thermal and magnetic stimuli, respectively. While conventional elasticity typically accommodates less than 1% strain [1], SME and SE can achieve recoverable strains of up to ~11% in NiTi [2,3] and ~12% in NiMnGa [4,5,6]. The thermal SME and SE exhibited by NiTi are driven by a reversible phase transformation from a high-symmetry and temperature B2 phase to a low-symmetry martensitic B19’ phase, and, in some cases, consist of intermediate R and/or B19 phases. Despite extensive research, the detailed mechanisms of this transformation and the role of the intermediate phases remain topics of ongoing debate [7,8,9,10,11,12,13,14,15].

Unlike the SME in NiTi, which is thermally induced, the SME in NiMnGa alloys can be activated both thermally [16] and magnetically [4]. For the sake of simplicity and to reflect the significant recent research interest, only the magnetic SME of NiMnGa alloys will be considered in this study. This property is driven by the mobility of the twin boundaries within the magnetically modulated martensitic phases [4,17], which are formed upon cooling from a high-symmetry ${L2}_1$ phase [18]. The magnetocrystalline anisotropy of the martensite crystals allows their twin boundaries to move when a magnetic field with sufficient strength is applied, leading to a macroscopically realised SME.

Despite these excellent functional properties, the superelasticity of NiTi and the brittle nature of NiMnGa alloys make them difficult to manufacture and process into desirable shapes via conventional routes (e.g., machining or rolling) [19,20]. Meanwhile, the layer-by-layer process of additive manufacturing (AM) provides a promising avenue for making these materials into complex geometries, thereby offering greater design freedom. Furthermore, AM-enabled lattices and auxetic structures can be designed specifically to improve the SME or SE properties of the final part [21]. These structures can potentially enable other functional properties, such as a negative Poisson’s ratio [22] or an altered stiffness [23], to be established while retaining the SME or SE of NiTi or NiMnGa. AM provides easier control over microstructural features such as crystallographic texture [24,25] and defects [26], which are critical for the SME in these alloys. However, the high cooling rates, complex thermal profile caused by thermal cycling, and the resulting residual stress still pose significant challenges to AM of NiTi and NiMnGa alloys [27,28]. In addition, the elemental evaporation (Ni in NiTi and Mn in NiMnGa) caused by the high melt pool temperature and, in some cases (i.e., electron beam melting), the need for a vacuum, can lead to the loss or compromise of SME and SE in these alloys.

This paper summarises the recent insights into the principles giving rise to thermal SME and SE in NiTi and magnetic SME in NiMnGa alloys and discusses their importance specifically in relation to their AM. Important findings and challenges involved in the design and the AM of NiTi and NiMnGa shape memory bulk and lattice structures are discussed. Studies on conventionally fabricated NiMnGa alloys are critically analysed in the context of proposing potential avenues for resolving the current challenges regarding their AM. Additionally, the effect of the process parameters of relevant AM techniques on the SME and SE of NiTi is summarised and novel, empirically derived AM process design tools (maps) are proposed.

2. Driving Principles of Shape Memory Effect and Superelasticity

The SME and SE functional properties of SMAs are generally driven by a reversible and diffusion-less martensitic transformation or martensite reorientation, which can be induced magnetically, thermally, or mechanically, leading to magnetic and thermal SMEs and SEs, respectively [29].

2.1. NiTi

2.1.1. Thermal SME in NiTi

The NiTi alloy is the most used and studied thermal SMA due to its superior shape memory and superelastic properties, as well as its good strength, ductility, and corrosion resistance [30]. The fast cooling of the equi-atomic NiTi alloy from the high-temperature B2 phase field (Austenite phase with a CsCl structure) to room temperature leads to the formation of a twinned B19’ martensite phase (a monoclinic distortion of the B19 phase) (Figure 1b) [31]. However, the addition of ternary elements such as Fe or Cu may lead to the formation of intermediate R and/or B19 phases [3]. Mechanical deformation of the twinned B19’ martensite leads to its reorientation and de-twinning, which accommodates a certain amount of the applied strain (ε). Subsequent heating of the alloy to above the Austenite Finish temperature ($A_f$) re-stabilises the B2 phase. This transformation recovers the strain that was previously accommodated by the martensite de-twinning process, which is macroscopically realised and gives rise to a thermal SME. Once the material cools down to below the Martensite Finish temperature ($M_f$), the twinned martensitic microstructure returns and the cycle repeats. This process is known as one-way SME.

Alternatively, NiTi can exhibit a two-way SME, which allows the material to be trained for two distinct and pre-defined shapes and exhibit them at two different temperatures (Figure 1a). Several methods have been reported for training NiTi to exhibit a two-way shape memory behaviour (

Table 1. Common methods for training shape memory NiTi to exhibit a two-way shape memory (TWSM) behaviour [32,33].

#	Training Method	Description
1	Over-deformation of martensite	Deform beyond the shape memory strain (SMS) limit below $M_f$ to set the first shape. Heat above the $A_f$, recover the strain, and set the second shape.
2	Shape memory (SM) cycling	Deform within the SMS limit below $M_f$. Heat to above $A_f$ to recover the undeformed shape. Repeat until TWSM is observed without the need for mechanical deformation.
3	Superelastic (SE) cycling	Cyclically load and unload the material above $A_f$, but within the temperature rate suitable for superelasticity. Sufficient cycles will result in a TWSM upon heating and cooling within a fraction of the applied strain values.
4	SME + SE cycling	Deform above $A_f$ to induce a certain amount of stress-induced martensite, then cool to below $M_f$ while constraining the applied strain, followed by heating to above $A_f$ to recover the strain. Repeat these steps until two-way shape memory is observed.
5	Mechanically constrained temperature cycling	Deform below $M_f$ within the SMS limit and constrain the strain. Heat to above $A_f$ and release the mechanical strain. Repeat until TWSM is observed.

) [32,33]. These training procedures all include a thermally or mechanically induced martensitic transformation, as well as the anisotropic formation of dislocations within the parent B2 (austenite) phase [32,34,35,36,37]. In addition to the typical plasticity of the parent phase, the internal stresses caused by the orientation mismatch amongst the preferentially oriented martensite variants of the adjacent parent grains also contribute to the formation of dislocations. The consistent iterative application of the uniaxial deformation performed during the training procedures leads to the formation of crystallographically anisotropic dislocations. This promotes the preferential formation of certain martensite variants, which drives the two-way SME in NiTi [32,34,35,36,37,38,39]. Hence, it is essential to consider the iterative deformations required for the training procedures when designing NiTi shape memory structures for AM [40]. The practicality of these methods depends on the application and the difficulties in constraining the two “memorised” geometries. Generally, due to the near-net shape fabrication process, significant deformation is not desired when training AM-fabricated parts. Additionally, additively manufactured parts often have complex geometries, which might make it difficult to constrain them mechanically at the different training temperatures. Hence, methods 1 and 2 are perhaps the most suitable for additively manufactured parts, as they require minimal deformation and geometrical constraining, given that the deformation required for these training methods is considered during the part design stage.

2.1.2. Superelasticity in NiTi

Unlike SME, SE requires the presence of a metastable austenite (B2) phase at the operating temperature [5], which can be achieved through careful stabilisation of the austenite, thereby reducing $M_s$ to below the operating temperature. In the case of NiTi, this is usually achieved by increasing the nickel (Ni) content to 51 at.% (from 50 at.% for SME) [41]. SE in NiTi is typically controlled by the nominal Ni content in the NiTi matrix, microstructure, texture, loading condition, and martensite variant selection. Factors such as heat treatments (annealing and/or ageing), thermal cycling, and the substitution of a third element can influence the martensitic transformation temperatures (MTTs) and ultimately the SE or SME modes of NiTi [31].

Equiatomic or near-equiatomic NiTi may exhibit a number of intermetallic phases, such as ${\mathrm{T}\mathrm{i}}_3{\mathrm{N}\mathrm{i}}_4$, ${\mathrm{T}\mathrm{i}}_2\mathrm{N}\mathrm{i},$ and $\mathrm{T}\mathrm{i}{\mathrm{N}\mathrm{i}}_3$, depending on its chemistry and thermomechanical history [42]. Figure 2a,b show examples of how NiTi, ${\mathrm{T}\mathrm{i}}_2\mathrm{N}\mathrm{i},$ and ${\mathrm{T}\mathrm{i}}_3{\mathrm{N}\mathrm{i}}_4$ manifest themselves within the microstructure of additively manufactured NiTi alloys. Thermodynamic studies have revealed that the higher driving force of ${\mathrm{T}\mathrm{i}}_2\mathrm{N}\mathrm{i}$ and $\mathrm{T}\mathrm{i}{\mathrm{N}\mathrm{i}}_3$ phases relative to that of the NiTi (Austenite) phase makes their formation unavoidable [42,43]. The formation of these phases is typically codependent. For instance, the precipitation of ${\mathrm{T}\mathrm{i}}_2\mathrm{N}\mathrm{i}$ leads to Ni enrichment in the matrix, thereby leading to the formation of Ni-rich intermetallics (i.e., $\mathrm{T}\mathrm{i}{\mathrm{N}\mathrm{i}}_3$ and ${\mathrm{T}\mathrm{i}}_3{\mathrm{N}\mathrm{i}}_4$ phases) [44,45]. Furthermore, recent studies suggest that these precipitates tend to form along the boundary or in the centre of the melt pool, as shown by the black and blue arrows, respectively, in Figure 2f [46]. These regions are assumed to accommodate higher number of solute atoms than the solubility limit and hence promote local precipitation [46]. Amongst these intermetallic phases, the lenticular-shaped ${\mathrm{T}\mathrm{i}}_3{\mathrm{N}\mathrm{i}}_4$ phase, shown in Figure 2b [47], has been shown to improve the strength and shape memory properties of NiTi due to its pinning effect on dislocation motion [42,48,49]. However, the size and density of these precipitates play a crucial role in their capacity to enhance these properties. Ageing temperature and time are critical in controlling the size, density [42], and homogeneity [49] of these precipitates. Recent studies have revealed that ageing a Ti-50.9 at.% Ni alloy at 573 °K for less than 360 ks led to ${\mathrm{T}\mathrm{i}}_3{\mathrm{N}\mathrm{i}}_4$ precipitation with an average size of <15 nm, which abruptly increased the yield strength (YS) and superelastic recoverable strain [48]. Increasing the duration of ageing was shown to increase only the precipitate density while leaving its size unchanged [40]. However, further increasing the average precipitate size via increasing the ageing temperature lowered the YS and the recoverable strain [40]. In addition to their size and density, the diffusion-driven decomposition of a ${\mathrm{T}\mathrm{i}}_3{\mathrm{N}\mathrm{i}}_4$ phase into $\mathrm{T}\mathrm{i}{\mathrm{N}\mathrm{i}}_3$, via ${\mathrm{T}\mathrm{i}}_3{\mathrm{N}\mathrm{i}}_4$ → ${\mathrm{T}\mathrm{i}}_2{\mathrm{N}\mathrm{i}}_3$ → $\mathrm{T}\mathrm{i}{\mathrm{N}\mathrm{i}}_3$, should also be considered when selecting the ageing temperature and duration [42,50,51]. Another important phase that can precipitate within additively manufactured NiTi specimens is the long, strip-like ${\mathrm{T}\mathrm{i}}_4{\mathrm{N}\mathrm{i}}_2{\mathrm{O}}_{\mathrm{x}}$ oxide intermetallic. When formed at grain boundaries, as shown in Figure 2c, this can substantially deteriorate the fatigue life of NiTi alloys. In fact, recent studies have identified this phase as the primary reason for the limited superelasticity and early fracture of additively manufactured NiTi alloys during superelastic cycling [52].

SE in NiTi is achieved when an external mechanical stress stabilises the B19’ martensite phase, leading to the B2-to-B19’ transformation. When the stress is relieved, the B2 phase is stabilised, and the reverse transformation takes place, the large strain accommodated by this transformation is recovered and SE is realised. The MTTs and how they compare with the operational temperature (i.e., the temperature of the intended application), the texture, and the loading condition all play important roles in SE properties. For instance, the crystallographic and diffusionless nature of this transformation leads to the crystallographic anisotropy of the resultant transformation strain (ε) [54,55,56]. Therefore, texture and loading condition need to be taken into consideration when designing NiTi structures for AM in the context of SE. A theoretical transformation strain with respect to the crystallographic orientation of the parent phase is shown in Figure 1c [31]. Another important factor to consider when it comes to SE is residual stress, which is commonly found in as-fabricated AM structures. As previously discussed, in SE NiTi, an external stress can stabilise the B19’ martensite and drive the B2-to-B19’ transformation. Hence, excessive residual stress can lead to premature stabilisation of the martensite and thereby deteriorate SE.

2.2. NiMnGa

2.2.1. Phase Evolution in NiMnGa

In addition to magnetoresistance and magnetocaloric properties, NiMnGa alloys exhibit both a thermal and magnetic SME and SE, due to a thermoelastic martensitic transformation, during which the high-symmetry ${L2}_1$ phase transforms into a low-order martensite phase upon cooling [4,57,58]. Similarly to NiTi, the martensitic transformation temperatures in NiMnGa magnetic SMAs are highly dependent on the chemical composition, particularly the Mn and Ga contents [59], which also determine the type or the crystal structure of the martensite phase [59,60,61]. Jin et al. [59] proposed an empirically derived formula for estimating the martensite start ($M_s$) temperature of NiMnGa magnetic SMAs based on their Mn and Ga atomic concentrations:

$$M_s(°\mathrm{K})=1960-(2110\times \mathrm{Mn})-(4920\times \mathrm{Ga}),$$

(1)

where the Mn and Ga concentrations are in the range of 20–35 at.% and 18–27 at.%, respectively.

The crystal structure of the ${L2}_1$ phase is best described as four interpenetrating FCC sublattices which constitute a Strukturbericht-type structure [62]. Depending on the composition and the thermal history, the NiMnGa alloys can exhibit both modulated and non-modulated (NM) martensitic phases (Table 2). The 5M and 7M are modulated martensite phases, named based on their long periodic stacking order, where the number refers to the shuffled layers and the letter M denotes monoclinic crystal structure in Ramsdel notation [63,64]. The 5M martensite can be described as an approximate tetragonal structure in the parent cubic coordinate system that follows the ${(110)\left[1\overline{1}0\right]}_{{L2}_1}$ periodic dislocations with five cyclic ${\left(110\right)}_{{L2}_1}$ plane stackings. On the other hand, the 7M martensite is an orthorhombic structure in the parent cubic coordinate system and has a modulation period of seven ${\left(110\right)}_{{L2}_1}$ plane stackings. Hence, the modulation of the 5M and 7M martensite phases arises from the periodical shuffling of the {110} planes along the $\left[1\overline{1}0\right]$ direction (Figure 3a). The NM martensite has a tetragonal structure with a c/a > 1 [4,65,66]. The crystal structures of the parent ${L2}_1$ phase and those of the martensite phases, as well as their lattice correspondence, are illustrated in Figure 3a. Even though the tetragonal and orthorhombic layered structures of the 5M and 7M martensite phases are five and seven planes long, respectively, these numbers are often doubled (i.e., 10M and 14M) when denoting these phases due to the ordering of the ${L2}_1$ structure, consisting of two subcells.

To summarise, the phase evolution from high to low temperatures in NiMnGa alloys is generally in the following sequence: high-symmetry ${L2}_1$ phase, 5M, 7M, and NM phases [67,68,69]. During this phase evolution, a range of different martensite variants are formed, and the resultant lattice strains are self-accommodated via twinning, giving rise to coherent planar twin boundaries [70]. Recent studies [71] have revealed that in single-crystalline specimens, upon cooling, the ${L2}_1$ phase first transforms into a martensite twin, which is subsequently detwinned into a single-martensite variant. At higher Ni contents or through the introduction of additional alloying elements, γ or γ′ phases may also precipitate in these alloys [72,73].

Table 2. Different martensitic phases that exist in NiMnGa magnetic shape memory alloys.

Martensite Phases	Description	${\mathit{\sigma }}_{\mathit{m}\mathit{a}\mathit{g}}$ [MPa]	${\mathit{\sigma }}_{\mathit{t}\mathit{w}}$ [MPa]	${\mathit{K}}_{\mathbf{\mu }}$$[{10}^5$$\mathbf{J}/{\mathbf{m}}^3$]	Reference
5M/10M	Modulated pseudo-tetragonal lattice; three twin variants	2.8	0.5–4	>6	[63,74,75]
7M/14M	Modulated orthorhombic lattice; six twin variants	1.6	3–5	0.47	[63,74]
NM	Non-modulated L10 tetragonal lattice; three twin variants	0.7	17–25	None	[63,74]

Table 2. Different martensitic phases that exist in NiMnGa magnetic shape memory alloys.

Martensite Phases	Description	${\mathit{\sigma }}_{\mathit{m}\mathit{a}\mathit{g}}$ [MPa]	${\mathit{\sigma }}_{\mathit{t}\mathit{w}}$ [MPa]	${\mathit{K}}_{\mathbf{\mu }}$$[{10}^5$$\mathbf{J}/{\mathbf{m}}^3$]	Reference
5M/10M	Modulated pseudo-tetragonal lattice; three twin variants	2.8	0.5–4	>6	[63,74,75]
7M/14M	Modulated orthorhombic lattice; six twin variants	1.6	3–5	0.47	[63,74]
NM	Non-modulated L10 tetragonal lattice; three twin variants	0.7	17–25	None	[63,74]

2.2.2. Magnetic Field-Induced Shape Memory Effect in NiMnGa

The magnetic SME is one of the diverse range of features offered by Heusler alloys, which first attracted significant attention when Friedrich Heusler discovered the possibility of developing ferromagnetic alloys consisting entirely of non-ferromagnetic elements [62]. The strain recovery in NiMnGa magnetic SMAs is induced by a magnetic field and therefore offers a faster response than thermally activated SMAs at significant frequencies of up to 1 KHz for over millions of cycles [76,77,78,79,80]. Hence, magnetic SMAs have excellent potential for devices such as microactuators [81,82,83], strain sensors [84], energy conversion devices [85,86], and microfluidic pumps [87,88,89,90].

Metallurgical principles of magnetic SME in NiMnGa—The introduction of an external magnetic field (H-field) with sufficient energy can lead to a macroscopic shape change, known as the magnetic field-induced strain (MFIS), via two different mechanisms [91,92,93]: transformation of the ${L2}_1$ phase to one of the martensite phases, and the magnetically induced martensite reorientation (MIR) or martensite variant reorientation (MVR) process. The MVR process results in the motion of the martensite twin boundaries, leading to the growth of specific martensite variants whose easy magnetisation crystallographic axis aligns with the external magnetic field [93,94,95]. This is illustrated schematically in Figure 3b. However, the first mechanism demands a significantly stronger magnetic field, which is unfeasible in practice [96]. Hence, the magnetocrystalline anisotropy of the martensite phases in NiMnGa alloys, allowing for the MVR process, is key to their magnetic shape memory properties [4,97,98,99]. Specifically, the c axis is an easy magnetization axis for both the 5M and 7M martensite phases. However, the NM martensite has no easy magnetisation axis; instead, the basal plane is considered an easy magnetization plane [4]. For single-crystalline or highly textured specimens consisting of a single- or multiple similarly oriented martensite variants, an applied magnetic field can be set up to align with their easy magnetization axis, leading to a high MFIS. However, for randomly oriented polycrystalline specimens, the magnetic field-induced motion of a martensite twin boundary can lead to destructive interference with another (i.e., they “cancel each other out”), leading to no or a low net MFIS value [70,71].

MFIS for NiMnGa—The maximum strain induced by twin boundary motion in a single crystal can be determined as ${\epsilon }_0$ = 1 − c/a, where c and a are the lattice constants [100,101]. Hence, the maximum theoretical MFIS calculated for the 5M, 7M, and the non-modulated martensite phases was 6.0%, 9.4%, and 20%, respectively [4,102]. In the case of the 5M martensite, the Type 1 and Type 2 twin boundaries were discovered: Type 1 has a rotational twinning plane and an irrational shear direction while Type 2 has an irrational twinning plane and a rotational shear direction [103,104]. Moreover, the maximum MFISs achieved experimentally for single-crystal samples of NiMnGa with 5M and 7M compositions were 6% and 10%, respectively [101,105,106,107,108]. However, the largest MFIS value that was experimentally measured for the NM martensite in single-crystalline NiMnGa-based alloys was about 12% [109]. MFIS can be practically realised only if the magnetic stress (${\sigma }_{mag}$) is greater than the twinning stress (${\sigma }_{\mathrm{t}\mathrm{w}}$). Magnetic stress is experienced by the material when placed in an anisotropic magnetic field whose value depends on the magnetic field strength and the crystal structure of the material [110,111]:

$${\sigma }_{\mathrm{m}\mathrm{a}\mathrm{g}}={\mathrm{K}}_{\mu }/(1-\mathrm{c}/\mathrm{a}),$$

(2)

where ${\mathrm{K}}_{\mu }$ is the magnetocrystalline anisotropy of the material. Hence, the large ${\sigma }_{\mathrm{t}\mathrm{w}}$ of the NM phase, as shown in Table 2, is believed to be the primary reason for the relatively low experimentally measured MFIS values. Similarly, the ${\sigma }_{\mathrm{t}\mathrm{w}}$ for the Type 1 and 2 twin boundaries that occur in 5M martensite was typically 1 MPa and from 0.05 to 0.3 MPa, respectively, making the Type 2 twin boundaries significantly more mobile [103,104]. In addition to an external magnetic field, the application of mechanical stress can initiate and drive the MVR process, giving rise to a plateau in the resultant stress–strain curve [112]. Moreover, the stress and/or the magnetic field intensity required for the commencement of the MVR process can be reduced via training procedures involving the application of a low cyclic compressive stress [112].

Determining factors for MFIS—The realisation of a large MFIS is highly dependent on the mobility of twin boundaries and the presence of defects that hinder it. High angle grain boundaries are perhaps the most dominant type of defects preventing twin boundary mobility; therefore, large MFIS values are typically exclusive to single-crystal magnetic SMAs [81,100]. However, producing single-crystal samples of these alloys is difficult and industrially unfeasible. Recent studies have indicated that the introduction of pores can significantly reduce grain boundary constraints and result in MFIS values as high as 8.7% for NiMnGa polycrystalline foams [113,114].

Similarly, the induction of a strong crystallographic texture has been shown to lead to an MFIS value of about 4% [115]. As both the induction of a controlled porosity and a strong crystallographic texture are strengths of AM, these microstructural features should be aimed for during the AM part and process parameter design stages. Additionally, grain growth and obtaining bamboo-grained NiMnGa structures are also known to enhance twin boundary mobility [81]. Moreover, the modulation nature and the crystal structure of the martensite phase are governed by the valence electron density, or the electron-to-atom ratio (e/a) [4,63]. Figure 3c shows the relationship between the e/a ratio and the stability of the different martensite phases, as well as their respective $T_a$ and $T_m$ temperatures. The $T_a$ and $T_m$ temperatures are defined as the average of the $A_s$ and $A_f$ and the $M_s$ and $M_f$ temperatures, respectively. The e/a ratio also influences the Curie temperature ($T_c$), as shown in Figure 3c. The e/a ratio of 7.7 was shown to be the critical threshold below which the $M_s$ of the alloy is lower than the $T_c$, indicating that the martensitic transformation occurs at a lower magnetic moment than the paramagnetic-to-ferromagnetic transformation [18,60]. These important factors, their relevance to different laser-based AM methods, and some potential methods for controlling them during the AM process will be elaborated on in future sections of this article.

3. Additive Manufacturing of Shape Memory Alloys

3.1. NiTi

Despite the outstanding properties of NiTi, the difficulties in smelting, casting, composition control, formability, and machinability, and its high cost, have limited its industrial applications to simple structures such as tubes, rods, plates, and wires [70,116,117,118,119]. AM, on the other hand, is believed to be a superior alternative method for the fabrication of more complex NiTi structures and a means of broadening their industrial applicability [116,117,118,119,120,121,122]. To date, research on the AM of NiTi shape memory and superelastic parts has mostly focused on Powder Bed Fusion (PBF)-based techniques and, to a lesser extent, on the Directed Energy Deposition (DED) AM methods [28,123,124,125].

3.1.1. Powder Bed Fusion-Based Additive Manufacturing of NiTi

PBF-based AM technologies consist of laser powder bed fusion (LPBF), selective electron beam melting (SEBM), and binder jetting (BJ). In LPBF and SEBM processes, parts are fabricated via a melting and re-solidification procedure, whereas, in BJ, powders are sintered in the solid state to form the final structure. The functional properties (SME and SE) of NiTi are highly dependent on the MTTs, which can be significantly influenced by the chemistry and/or stress state of the final part. These factors should be considered at all stages of the fabrication process, including the geometric design, feedstock material selection and characterisation, printing parameters, and post-heat treatment.

Powder feedstock—The impurity levels in the feedstock material (powder, wire, etc.) are a critical factor in determining the MTTs of the final printed product, and therefore the SME and SE [28]. Oxygen and carbon are the most common and detrimental impurities that may be present in NiTi feedstock [126]. Excessive carbon and oxygen can lead to the formation of phases such as TiC and ${\mathrm{Ti}}_4{\mathrm{Ni}}_{2}O$, respectively, which enriches the matrix with Ni, thereby lowering the MTTs [127]. Moreover, it has also been shown that an excessive impurity content can significantly deteriorate the fatigue properties of NiTi [128]. In the case of NiTi powder production, gas atomisation has been shown to result in the fewest impurities and, amongst the different gas atomisation techniques, the electrode induction inert gas atomisation (EIGA) process seems to be the most promising [28,129,130]. The primary drawback of the EIGA method is that it produces a relatively large range of particle sizes and non-spherical particles, which necessitates the use of sieving and fractioning processes [28,131,132]. Apart from the feedstock material, the AM atmosphere can be another source of impurities, particularly in LPBF, despite the use of high-purity argon [28,133].

Process parameters—Typically, a high relative density and low levels of impurity and nickel evaporation are important competing factors during PBF-based AM processes, optimising the volumetric energy density (${\mathrm{E}}_{\mathrm{v}}$), which is defined as follows [28,133]:

$${\mathrm{E}}_{\mathrm{v}}=\mathrm{P}/\left(\mathrm{v.h.t}\right),$$

(3)

where P is the laser/electron beam power, v is the scanning velocity, h is the hatch spacing, and t is the layer thickness. The energy density should be sufficiently high to obtain a fully dense part but sufficiently low to minimise the development of impurities [28,118] and nickel evaporation [133,134,135,136]. It has been reported that an energy density of 200 J/mm³ can yield fully dense NiTi parts while keeping the impurity levels within the acceptable level specified in ASTM F2063-05 [28] for laser-based processes. Nevertheless, recent studies have shown that, at an energy density sufficient to achieve a fully dense part, a Ni loss of approximately 0.4–0.5 at.% is anticipated during the LPBF process [137].

However, when controlling the SE and/or SME properties, only considering the energy density is not sufficient [136], and each individual constituent parameter of the energy density (i.e., P, v, h, and t) needs to be evaluated independently. These include any parameter that affects the nickel content, microstructure, and texture. For example, hatch spacing can influence the formation of the ${\left[002\right]}_{B2}$ texture [138], which not only affects the SE but also the superelastic anisotropy [136]. Similarly, the scanning speed, hatch spacing, and laser power influence the severity of nickel evaporation, in descending order of importance [139]. We summarised the influences of process parameters on the shape memory and superelastic recoverable strain (${\epsilon }_{rec}$) of LPBF-processed NiTi based on empirical investigations in the relevant literature, as illustrated in Figure 4 [136,139,140,141,142,143,144,145]. All the data points were taken from samples printed with the feedstock powder compositions ranging from Ti-50.6Ni at.% to Ti-50.8Ni at.%, and a layer thickness of 30 µm, and measured in the as-printed state. Each point corresponds to a shape memory or superelastic NiTi sample fabricated via the associated process parameters and is colour-coded in accordance with the total measured ${\epsilon }_{rec}$. The ${\epsilon }_{rec}$ corresponds to the addition of elastic and superelastic strain recovery for superelastic specimens and the elastic and shape memory strain recovery for the shape memory specimens. The regions corresponding to the relatively high ${\epsilon }_{rec}$ values are highlighted in red, while the blue-highlighted regions show the process parameters that resulted in a relatively low ${\epsilon }_{rec}$. It should be noted that all the data points correspond to samples with a high relative density (>96%); hence, porosity is not a contributing factor.

Processing window definition—As clearly illustrated, the shape memory and superelastic ${\epsilon }_{rec}$ exhibited by LPBF-printed NiTi are highly dependant on individual process parameters. For instance, as shown in Figure 4a, ${\epsilon }_{rec}$ is diminished when the laser power and scanning speed are simultaneously and excessively increased or decreased. This indicates that there is an optimal intermediate range for these quantities. Furthermore, the figure illustrates that keeping the scanning speed constant at 600 mm/s and changing the laser power from 80 W to 200 W does not have a significant influence on ${\epsilon }_{rec}$. On the other hand, keeping the laser power constant (e.g., at 150 W) while altering the scanning speed even slightly can have a much more severe influence on ${\epsilon }_{rec}$, indicating the higher sensitivity of ${\epsilon }_{rec}$ to scanning speed relative to laser power. This finding is in line with a recent study [146], which also demonstrated that scanning speed has a particularly strong influence on the shape memory properties of LPBF-fabricated NiTi components. However, the overall sensitivity of ${\epsilon }_{rec}$ to both laser power and scanning speed is highly dependant on the hatch spacing, as clearly shown in Figure 4c,d. Both these figures suggest a reduction in ${\epsilon }_{rec}$ at excessively high hatch spacing values. Figure 4b shows the combined influence of laser power, scanning speed, and hatch spacing on ${\epsilon }_{rec}$. As illustrated, a combination of high laser power, hatch spacing, and scanning speed results in a low ${\epsilon }_{rec}$. However, the processing window that results in significant ${\epsilon }_{rec}$ values (larger than 6.5%) can be proposed and summarised as follows:

100 W < laser power < 200 W;

520 mm/s < scanning speed < 850 mm/s;

60 µm < hatch spacing < 130 µm.

The interdependency of the process parameters and the complexity of the factors that influence SE and SME, such as chemistry and MTTs, texture and residual stress, and even the part geometry [147], make it difficult to speculate on the potential underlying principles giving rise to these trends. These maps, empirically derived and proposed for the first time in the present study, are intended to form the foundation of an AM process design tool, specifically for NiTi components with an SME and SE. While all data points were obtained from samples with a consistent feedstock composition (Ti-50.6–50.8Ni at.%), a fixed layer thickness of 30 µm, and a relative density exceeding 96%, the predictive capability of these maps remains limited at this stage. Other influential factors, such as gas flow dynamics, build size, and machine-specific parameters, were not considered in the current version. As such, these maps should be interpreted as a preliminary framework that will require further refinement as more diverse and comprehensive datasets become available.

3.1.2. Directed Energy Deposition

The DED techniques either keep the heat source and the nozzles (i.e., printing head) in static X-Y coordinates and move the build plate in accordance with the computer-aided designed (CAD) 3D model, or move the printing head while keeping the bed stationary [148]. Furthermore, unlike PBF-based AM techniques, where the powder bed moves in the Z axis to realise a 3D structure, the DED techniques progress through Z layers by raising the printing head. Typical DED systems use powder or wire as feedstock, and laser (i.e., Laser Engineered Net Shaping (LENS)), electron beam (i.e., wire-feed electron beam additive manufacturing (WEBAM)) and electric arc (i.e., Wire-Arc Additive Manufacturing (WAAM)) as heating sources.

DED technologies allow for the injection of powders/powder combinations with specific user-defined stoichiometries for in situ alloy formation. In the case of NiTi, in situ adjustments of the nickel content across the printed part allow for different MTTs and therefore, functionally gradient shape memory or superelastic structures, leading to desirable properties such as multi-stage shape memory actuation [149]. However, this adjustment should be carried out carefully as the effect of nickel content variations on the MTTs is severe (about 10 °C for every 0.1 at.% variation [150]) and could also lead to the formation of undesirable intermetallics. Both pre-alloyed [151] and elemental [152] powders have been used to fabricate NiTi parts via DED techniques. Similarly to LPBF and SEBM processes, nickel evaporation has been reported for parts manufactured via DED techniques, particularly when elemental powders are used [152]. Typically, a larger amount of nickel is injected to compensate for the nickel evaporation and to obtain the desired Ni/Ti ratio [148]. Unlike LPBF, where critical parameters such as the combination of laser power and laser effective diameter are usually pre-defined, LENS, as well as most other DED techniques, are free-form methods of AM [153,154] that provide greater control over the process.

On the other hand, the materials produced by DED techniques normally have larger melt pool sizes as a consequence of the application of coarser powders (45 to 150 µm) and wire feedstocks, together with other process parameters (larger laser spot size, a larger layer thickness and hatch distance, etc.). These can cause significant differences in the thermal profile within the melt pool and the thermal history of the entire build, which can, in turn, affect the chemistry (Ni/Ti ratio), microstructure, and texture of the printed materials, and ultimately the SE and SME of the NiTi parts. Recent studies reported equivalent elastic moduli and critical and yield stresses exhibited by the DED- and PBF-based AM-fabricated NiTi. However, a greater degree of the anisotropy of the microstructure and the resultant shape memory properties of DED-fabricated NiTi was identified compared to their PBF-based AM-fabricated counterparts [154,155]. These superelastic and shape memory anisotropies are of great importance, particularly when it comes to the AM of smart structures, as they will allow for different actuation behaviours, both spatially throughout the structure and along the different orientations.

As reviewed in Section 2.1, the MTTs and their relation to the operating temperature are determining factors for the SE and SME mode selection in NiTi. Generally, the extent to which the MTTs are influenced by the different DED methods seems to differ noticeably. For instance, Dutkiewics et al. [156] measured the MTTs of the raw powder and wire feedstock used for the LENS and SEBM processes, respectively, and compared them to those exhibited by the parts after AM. The ${\mathrm{A}}_{\mathrm{s}}$ of the LENS- and SEBM-fabricated parts decreased by 41.4 °C and increased by 69.3 °C and ${\mathrm{A}}_{\mathrm{f}}$ decreased by 34.8 °C and increased by 26.7 °C, respectively, compared to the input feedstock [156]. However, the decrease in the MTTs during the LENS process reported in this study contradicts several other studies [157,158,159], where the LENS process had an increasing effect on these temperatures. This inconsistency seems to be driven by the competing effects of nickel evaporation and the effects of Ti-rich precipitates and oxides (e.g., ${\mathrm{T}\mathrm{i}}_4{\mathrm{N}\mathrm{i}}_2\mathrm{O}$) on the MTTs. Figure 5 shows the empirically derived relationship we observed between the LENS process parameters and the shape memory and superelastic ${\epsilon }_{\mathrm{r}\mathrm{e}\mathrm{c}}$ of bulk NiTi specimens [151,156,160,161].

The insufficient available data on the shape memory and superelastic ${\epsilon }_{rec}$ exhibited by the LENS-fabricated NiTi compared to the LPBF-fabricated ones do not allow for a clear specification of high- and low-${\epsilon }_{rec}$ regions on the maps. However, similar to Figure 4, the LENS process design maps shown in Figure 5 form a foundation that requires further development once sufficient data are available.

As the LENS technique allows for easy adjustments to the laser spot diameter by simply altering the lens position relative to the laser focal point and the working plane, the hatch spacing previously used in Figure 4 for LPBF is replaced with the laser spot diameter in Figure 5. The composition of the powder and the layer thickness used to fabricate the NiTi specimens corresponding to all data points in Figure 5 fell within the Ti (50.1–50.8) Ni at.% and 0.15–0.3 mm ranges, respectively. Furthermore, similar to Figure 4, the ${\epsilon }_{rec}$ used in Figure 5 was measured in the as-printed state without any post-processing and all the samples exhibited high relative densities. The significant influence of laser spot diameter on the shape memory and superelastic properties can be clearly observed in Figure 5c,d. Both these figures suggest an inverse correlation between the ${\epsilon }_{rec}$ and laser spot diameter, with the highest ${\epsilon }_{rec}$ achieved at a laser spot diameter of 0.5 mm. Assuming that the laser–powder interaction is sufficiently consistent amongst the data points, an increase in the laser spot size enlarges the melt pool, altering the thermal profile during the solidification process. This may influence critical factors such as Ni evaporation, or potentially lead to a certain degree of chemical segregation, thereby deteriorating the SE and SME properties of the final specimen. Similarly, the detrimental effect of excessive laser power is demonstrated in Figure 5a, and may be attributed to Ni evaporation and intermetallic precipitation in subsequent layers during the printing process.

One of the drawbacks of powder-based DED techniques, especially those using elemental powders as input, is the extremely low powder-to-part yield [162]. WAAM, on the other hand, has recently been shown to be an effective DED method for fabricating NiTi parts due to its high deposition rates and lower overall costs [163]. However, the poor thermal control offered by WAAM, particularly the gas metal arc welding (GMAW) method, typically result in relatively undesirable mechanical properties, such as fracture strains as low as 6.2% [164]. In order to enhance the thermal control of WAAM, a cold metal transfer (CMT) technology was incorporated into GMAW, which increased the fracture strain of the as-printed NiTi to approximately 13.6% [165]. Furthermore, the gas tungsten arc welding (GTAW) WAAM method, which uses a non-consumable tungsten electrode, was used to enhance the accuracy of the formation; this study also illustrated the high sensitivity of the fracture strain to the deposition current [165,166,167,168]. The implementation of an ultra-high-frequency pulsed heat source in GTAW led to a stable superelasticity, greater fracture strain (~17%), good forming accuracy, and grain refinement [169,170,171,172]. Since NiTi alloys are heavily influenced by the total heat input of the process [165], it is important to understand the heat and mass flow mechanisms involved in the WAAM process. However, the exact relationship between the interactions between the heat source and the wire, the thermal gradient, and the resultant microstructure and properties of the final part has yet to be clarified. This has encouraged several recent studies to focus on developing physics-based numerical computational fluid dynamics (CFD) models [173,174].

The effect of substrate temperature during the WAAM process on intermetallic precipitation, crystallographic orientation, MTTs, and shape memory/martensitic transformation has also attracted noticeable attention recently. Wang et al. [167] studied the effect of substrate temperature on the microstructure and properties of WAAM-fabricated Ni-rich NiTi. This study demonstrated that increasing the substrate temperature from 150 °C to 350 °C increased the average B2 grain size by almost 32 µm, coarsened and increased the weight fraction of ${\mathrm{Ni}}_4{\mathrm{Ti}}_3$ intermetallics, and increased $M_s$, $M_f$, $A_s$, and $A_f$ temperatures by 17.6 °C, 7.1 °C, 17.7 °C, and 8.3 °C, respectively. Moreover, increasing the substrate temperature was shown to result in a reduction in total elongation and yield and ultimate strengths; however, the effect on the crystallographic texture and superelastic recoverable strain was found to be minimal [167]. The results clearly illustrate the substantial sensitivity of NiTi to the thermal profile and cooling rates associated with the AM process.

3.2. Magnetic Shape Memory NiMnGa Alloys

The AM of NiMnGa magnetic SMAs is still a relatively new concept, and the majority of studies are focused on binder-based processes [24,115,175,176,177,178,179,180,181]. However, the number of recent investigations focusing on the LPBF and laser-based DED of these alloys is increasing rapidly [182,183,184].

Table 3. LPBF and DED process parameters used for fabricating NiMnGa specimens and their relative densities. S#: sample number; v: scanning speed; t: layer thickness; h: hatch distance; P: laser power; $E_v$: volumetric energy density; S. Type: sample type.

AM Instrument	S#	v (mm/s)	t (mm)	h (mm)	P (W)	${\mathit{E}}_{\mathit{v}}$$(\mathbf{J}/{\mathbf{m}\mathbf{m}}^3$)	Density (%)	S. Type	Ref.
Laser Powder Bed Fusion
Mlab Cusing, Concept Laser	1	500	0.025	0.098	30	24.5	89.6 ± 0.6	$\mathrm{Cuboids}—10 \times 10 \times 5 {\mathrm{m}\mathrm{m}}^3$	[184]
	2	600			30	20.4	78.8 ± 0.6
	3	700			30	17.5	75.2 ± 0.7
	4	500			35	28.6	84.1 ± 0.6
	5	600			35	23.8	82.6 ± 0.6
	6	700			35	20.4	80.5 ± 0.6
	7	500			40	32.7	87.1 ± 0.6
	8	600			40	27.2	85.2 ± 0.6
	9	700			40	23.3	84.9 ± 0.6
Concept Laser M2 Cusing	10	200	0.025	No hatching	70	N/A	97.3 *	$\mathrm{Lattices}—8 \times$$8 \times$$8 {\mathrm{m}\mathrm{m}}^3$	[188]
	11	450			70		98.9 *
	12	600			70		98.9 *
	13	200			100		94.0 *
	14	450			100		97.4 *
	15	600			100		97.9 *
SLM Solutions	16	250	0.025	No hatching	37.5	N/A	N/A	$\mathrm{Thin} \mathrm{walls}—30-50 \times 5–6 \times 0.24–0.28 {\mathrm{m}\mathrm{m}}^3$	[189]
Modified EOSINT M-Series	17	200–600	0.06	No hatching	80–200	N/A	N/A	$\mathrm{Thin} \mathrm{walls}—0.16 \times 0.77 \times 5 {\mathrm{m}\mathrm{m}}^3$	[190]
In-House Built	18	300	0.05	0.1	200	133.3	96.8	$\mathrm{Cuboids}—7 \times 7 \times 1 {\mathrm{m}\mathrm{m}}^3$	[27]
	19	500				80.0	94.9
	20	700				57.1	93.5
Concept Laser M2 Cusing	21	3000	0.025	0.09	200	29.6	80.14	$\mathrm{Cuboids}—7 \times 7 \times 4 {\mathrm{m}\mathrm{m}}^3$	[191]
	22	3000		0.09	300	44.4	76.89
	23	3000		0.045	250	74.1	82.21
	24	3000		0.015	100	88.9	76.98
	25	3000		0.045	150	44.4	83.06
	26	2000		0.09	150	33.3	79.15
	27	2000		0.045	200	88.9	N/A
	28	2000		0.09	200	44.4	N/A
	29	3000		0.09	300	44.4	N/A
	30	2000		0.09	250	55.6	N/A
In-House Built	31	125	0.06	0.05	50	133.3	75.2 ± 4.8	$\mathrm{Cuboids}—3.5 \times 3.5 \times 3.5 {\mathrm{m}\mathrm{m}}^3$	[183]
	32	50		0.05	85	566.7	86.2 ± 2.8
	33	200		0.05	85	141.7	87.7 ± 1.6
	34	300		0.05	100	111.1	79.4 ± 2.1
	35	125		0.05	120	320.0	95.2 ± 0.9
	36	150		0.05	150	333.3	92.7 ± 2.1
	37	450		0.05	150	111.1	93.7 ± 0.4
	38	300		0.05	200	222.2	96.2 ± 0.5
	39	50		0.075	50	222.2	81.0 ± 3.2
	40	200		0.075	50	55.6	72.6 ± 5.2
	41	125		0.075	85	151.1	91.2 ± 1.5
	42	150		0.075	100	148.1	94.9 ± 0.6
	43	450		0.075	100	49.4	83.1 ± 3.7
	44	50		0.075	120	533.3	92.6 ± 0.5
	45	200		0.075	120	133.3	97.8 ± 0.3
	46	300		0.075	150	111.1	91.8 ± 2.4
	47	150		0.075	200	296.3	93.7 ± 0.9
	48	450		0.075	200	98.8	98.7 ± 0.5
	49	125		0.1	50	66.7	76.0 ± 2.7
	50	50		0.1	85	283.3	93.5 ± 0.9
	51	200		0.1	85	70.8	93.9 ± 2.3
	52	300		0.1	100	55.6	84.2 ± 1.8
	53	125		0.1	120	160.0	96.8 ± 0.6
	54	150		0.1	150	166.7	94.2 ± 1.1
	55	450		0.1	150	55.6	94.8 ± 1.1
	56	300		0.1	200	111.1	98.5 ± 0.3
In-House Built	57	750	0.06	0.1	200	44.4	98.6 ± 0.33	$\mathrm{Cuboids}—5 \times 5 \times 5 {\mathrm{m}\mathrm{m}}^3$	[81]
Mlab Cusing, Concept Laser	58	500	0.025	0.098	45	36.7	91.1 ± 0.6	$\mathrm{Cuboids}—10 \times 10 \times 5 {\mathrm{m}\mathrm{m}}^3$	[20]
	59	600			45	30.6	88.2 ± 0.6
	60	700			45	26.2	85.5 ± 0.6
	61	500			50	40.8	93.8 ± 0.6
	62	600			50	34.0	90.8 ± 0.6
	63	700			50	29.2	86.5 ± 0.7
	64	500			55	44.9	95.9 ± 0.6
	65	600			55	37.4	93.7 ± 0.6
	66	700			55	32.1	91.5 ± 0.6
SLM Realizer 50	67	250	0.025	N/A	37.5	N/A	N/A	$\mathrm{Thin} \mathrm{walls}—6 \times 3 \times 0.5 {\mathrm{m}\mathrm{m}}^3$	[192]
Laser Based Directed Energy Deposition
Laser-Based DED—LENS	68	2.5	0.25	No hatching	350	N/A	N/A	Five deposition layers	[186]
Laser-Based DED	69	10.8	N/A	No hatching	300	N/A	97.71 ± 0.08	$\mathrm{Thin} \mathrm{walls}—30 \times 2.5 \times 15 {\mathrm{m}\mathrm{m}}^3$—single track layers; bidirectional scan strategy	[193]
	70				400		98.68 ± 0.07
	71				500		97.29 ± 0.12
	72				600		94.88 ± 0.14
	73				700		N/A #
	74				300		96.92 ± 0.13	$\mathrm{Thin} \mathrm{walls}—30 \times 2.5 \times 15 {\mathrm{m}\mathrm{m}}^3$—single track layers; unidirectional scan strategy
	75				400		98.44 ± 0.11
	76				500		95.15 ± 0.13
	77				600		92.31 ± 0.26
	78				700		N/A #

* As the printed specimens were lattices, the internal strut density was considered. ^# 700 W of laser power, for both scanning strategies, led to significant over-melting and the detachment of specimens from the substrate. N/A stands for Not Available.

summarises the process parameters and the resultant relative densities of 78 specimens recently fabricated via LBPF and laser based-DED methods, investigated across 12 studies. Overall, these studies indicate the great ability of these AM methods to produce high-density polycrystalline NiMnGa parts with great compositional and geometrical repeatability and consistency [81]. Typically, LPBF-manufactured NiMnGa exhibits a twinned martensite microstructure with weak magnetic anisotropy [183], where the 5M and 7M martensites coexist. Depending on the final composition of the fabricated part, specifically the Mn content, post-fabrication homogenisation heat treatment at about 1040 °C can stabilise the 7M martensite [81]. Only a few studies have reported twins spanning grain boundaries in the as-printed state [181,185,186], and only one study reported an MFIS of only 0.01% in BJ-processed samples after thermo-magneto-mechanical training [179]. However, a substantial MFIS of 5.8% was observed for a single crystal of an LPBF-fabricated polycrystalline NiMnGa sample containing a 5M martensitic structure [187]. In this section, the important process parameters of the PBF-based AM and DED routes and their influences on the microstructure and magnetic shape memory properties of NiMnGa alloys will be discussed.

3.2.1. Laser Powder Bed Fusion

The LPBF of NiMnGa magnetic SMAs is believed to be a promising avenue for generating magnetic actuators with complex geometries, further expanding their practicality and applicability. Recent studies demonstrate the great ability of LPBF to produce NiMnGa polycrystalline parts with high relative densities [27,81,183,184,192]. However, as mentioned previously, the maximum MFIS exhibited by AM NiMnGa polycrystalline parts is limited to 0.01%. This limitation is primarily driven by a range of factors, including internal defects caused by the AM process [194], small grains, and the lack of a crystallographic texture [81,183,187]. As summarised in Section 2.2.2, twin mobility is the key controlling factor for the magneto-structural properties of NiMnGa. With a lower ${\sigma }_{\mathrm{t}\mathrm{w}}$ and faster twin boundary velocities, better twin mobility and better magneto-structural properties can be obtained. The residual stresses and atomic disorder caused by the non-equilibrium conditions of the AM process significantly influence the magneto-structural properties of NiMnGa [81,187]. For instance, the ${\sigma }_{\mathrm{t}\mathrm{w}}$ of the Type 1 and Type 2 twin boundaries of an LPBF-fabricated 5M specimen were measured to be 1.4 and 0.6 MPa, respectively, which are about 1.5 to 2 times greater than those of conventionally fabricated single-crystalline specimens [195]. Similarly, the maximum Type 1 and 2 twin boundary velocities were measured to be 1.5 and 24 m/s, respectively, which are about half of the respective values for conventionally grown single-crystalline specimens [195]. The specimen used to measure the twin boundary mobility in this study was a single crystal cut out of a larger LPBF-fabricated sample. This indicates that grain boundaries are not the only defects that hinder twin boundary mobility in AM-fabricated specimens. In fact, defects such as the porosity formed due to Mn evaporation and the formation of Mn and Mn oxides on pore walls during the LPBF process are believed to be major contributors.

Elemental Evaporation and Its Effect on Magneto-Structural Properties

The evaporation of manganese is a major issue in the AM of NiMnGa magnetic SMAs and is intensified with increasing energy density [27,183,184,192]. Like NiTi, the MTTs of NiMnGa magnetic SMAs have been shown to be extremely compositionally sensitive. For example, previous studies have reported that a Mn variation as small as 0.1 at.% can shift the MTTs by 5 °C [59,196]. The reduction in Mn as a result of evaporation leads to alterations in the martensite type from 5M to 7M, and eventually to the non-modulated martensite, increase in the MTTs and a decrease in the Curie temperature [187].

On the other hand, the compositional dependency of MTTs can also be exploited by a controlled selective Mn evaporation process via in situ adjustments to the energy density during the AM process, which could potentially allow for parts with multi-stage actuation. Figure 5b and Figure 6a illustrate how the energy density influences Mn evaporation and, therefore, the 5M and 7M martensite lattice parameters, MTTs, and the Curie temperature ($T_c$), based on data gathered from [187]. The initial Mn content of the original powder was 30.7 at.%. As shown, increasing the energy density results in an increase in Mn evaporation, which undesirably decreases the difference between the MTTs and $T_c$. Within the 5M martensite region, increasing the energy density results in a slight decrease in the a lattice parameter and a slight increase in the c lattice parameter, leading to an increase in the c/a ratio (Figure 6b). On the contrary, an increase in the energy density in the 7M region decreases the c/a ratio. Hence, according to ${\epsilon }_0$ = 1 − c/a, minimising the energy density while in the 5M region and maximising it while in the 7M region may be beneficial for increasing the MFIS. Within the range of energy densities considered in this study [187], on average, the Mn content decreases by 1 at.% per 40 J/mm³ increase in energy density.

The influence of a much broader range of laser energy densities on the Mn and Ga evaporation, as well as the relative densities of the corresponding samples, based on the data gathered from [183], is demonstrated in Figure 6c. As illustrated, the relative densities measured for the corresponding samples did not fall below 73% for the entire range of energy densities investigated. As shown, the energy density values below 100 J/mm³ led to a negligible Ga loss, a Mn loss of less than 2 at.%, and relative densities above 98.5%. the LPBF of the NiMnGa alloys with a lower energy density, in the range of 17.49 to 31.65 J/mm³, was also investigated [184], and resulted in a noticeably lower Mn evaporation rate and a lower relative density. Interestingly, the density was further reduced after homogenisation and ordering heat treatments at 1000 °C for 95 h and 800 °C for 24 h. This was believed to be caused by gas porosity and pore-coarsening during heat treatment [184,197,198]. Despite the low energy density, the samples did not exhibit any sign of severe chemical inhomogeneity. The open pores within some samples resembled a bamboo-like microstructure and contained twins that span across individual grains. A systematic investigation by Laitinena et al. [183] indicated that an energy density of about 75 J/mm³ can produce a sample with more than 98% relative density while keeping the Mn evaporation as low as about 1.1 at.%. Future research should focus on the development of dense NiMnGa parts with minimal Mn evaporation, followed by the introduction of a controlled porosity (e.g., lattice structures) [182,183,199] designed to maximise twin boundary mobility. If modelled and controlled properly, this could be a way to enhance the porosity of AM-fabricated NiMnGa alloys that plays an important role in the relaxation of grain boundary constraints and potentially leads to a greater MFIS.

Twin Boundary Mobility

The twin boundary mobility of AM-processed NiMnGa magnetic SMAs and how this is influenced by the AM process is another crucial parameter to consider to achieve a magnetic SME. For the 5M martensite, the type of twin boundaries (i.e., Type 1 or Type 2) significantly affects the twinning stress and, ultimately, the twin boundary mobility. As discussed previously, the twinning stresses exhibited by the Type 1 and 2 twin boundaries in conventionally grown single-crystal specimens are about 1 MPa and 0.05–0.3 MPa, respectively [103,104]. Figure 7a shows a single-crystal specimen that was cut out of an LPBF-fabricated 5M NiMnGa sample, exhibiting a clear twin boundary motion and a giant MFIS of 5.8% after the application of a 0.8 T magnetic field in two perpendicular directions [195]. This study revealed that the twinning stresses for Type 1 and 2 twin boundaries within the LPBF-fabricated specimen are, respectively, about 0.4 and 0.3 MPa higher than those exhibited by conventionally fabricated single crystals [195]. This indicates that for a given magnetic field strength, the activation of the twins for LPBF-fabricated NiMnGa magnetic SMAs is more difficult. Additionally, Figure 7b,c show the applied magnetic field pulse alongside the resultant twin boundary displacement and velocity as a function of time for the Type 1 and 2 twin boundaries, respectively [195]. As shown, at saturating field conditions, the Type 2 boundary is displaced about seven times further than the Type 1 boundary. Additionally, at saturating conditions, the maximum velocity exhibited by the Type 2 boundary is about 12 times higher than that of the Type 1 boundary. These observations clearly indicate the superior mobility of Type 2 twin boundaries; however, controlling the type of twin boundaries in AM-fabricated NiMnGa parts remains a challenge.

3.2.2. Directed Energy Deposition

DED techniques offer significant technological benefits and spark great scientific interest in the fabrication of NiMnGa magnetic SMAs. For example, under certain conditions, DED can produce grains with a favourable grain morphology (columnar) and crystallographic texture [200,201,202,203], which have been reported to allow for MFIS along their long crystallographic axes [204]. However, despite these potential benefits, the number of studies covering this topic in the literature is minimal. DED-fabricated small NiMnGa samples were found to have significant microstructural inhomogeneity along the build direction, with a mixed grain microstructure consisting of fully dendritic regions, regions with partially dendritic microstructures, and pieces of the primary dendrite arms [186]. These inhomogeneities resulted in broad ranges of MTTs, which were also observed in as-cast and untreated (prior to homogenisation and ordering treatments) specimens [205]. Homogenisation at 1000 °C for 24 h followed by an ordering treatment at 700 °C for 12 h promoted recrystallisation and grain growth, increased $T_c$, as expected [206,207] by 4 °C, increased the saturation magnetization by 10%, and reduced ferromagnetic hysteresis. Similarly, Laitinen et al. [193] utilised laser-based DED to fabricate 30 $\times$ 2.5 $\times$ 15 ${\mathrm{m}\mathrm{m}}^3$ specimens using single tracks for each layer, resulting in high relative densities (>97.5%) and a consistent Mn loss of about 0.7–0.9 at.% across all specimens. The process parameters used in this study are summarised in Table 3 (samples 69 to 78) and, as indicated, two batches of specimens were produced using unidirectional and bidirectional scanning strategies. Unidirectional melting produced samples with better geometric accuracy, whereas bidirectional melting promoted large columnar prior ${L2}_1$ crystals, on the order of several millimetres, with a strong <100> crystallographic texture along the build direction. Although promising, the large grains and strong texture alone were not sufficient to result in an MFIS due to the grain boundary constraints.

3.2.3. Effect of Quaternary Elements

As discussed in Section 3.2.1 and Section 3.2.2, NiMnGa magnetic SMAs produced via LPBF-based AM and laser-based DED techniques experience both material- and process-related issues, such as severe brittleness, Mn evaporation, a small grain size, and low twin boundary mobility. A few recent studies focused on the modifying effect of small additions of a quaternary element on these alloys’ mechanical and functional properties. Even though these studies investigate samples fabricated via conventional methods (i.e., casting), their findings are important and may potentially be pathways to the successful AM of NiMnGa-based magnetic SMA parts. It is important to note that, to date, no systematic studies have explored the effect of quaternary element additions on NiMnGa magnetic SMAs fabricated via LPBF or laser-based DED additive manufacturing processes. The findings presented in this section are therefore based on conventional processing studies. While these results cannot be directly extrapolated to AM-processed parts, they offer valuable mechanistic insights and hypotheses for further investigation. Importantly, recent publications [208,209] have highlighted the potential of Co-doped NiMnGa alloys for additive manufacturing, indicating that such compositions may improve the printability and performance of AM-fabricated magnetic SMA components. These observations support the idea that quaternary alloying strategies remain a compelling direction for future AM-specific research.

Enhancement of Ductility and Strength

NiMnGa magnetic SMA parts suffer from severe intrinsic brittleness as well as a relatively low strength and toughness, which limit the ability of the material to accommodate the residual stresses developed during the AM process, leading to severe cracking, and ultimately, fabrication failure [27,81,188]. Even in the case of successful fabrication, this drawback limits their practicality and industrial uptake. Small additions of alloying elements such as Fe [73], Ta [210], Cu [73], Y [211], and Co [212] have been shown to increase the ductility of NiMnGa alloys while preserving their functional properties. The addition of only 0.5 at.% Ta to a Ni-25Mn-21Ga at.% alloy increases the compressive strength and fracture strain by approximate factors of 5 and 1.7, respectively [210]. Similarly, the addition of up to 3 at.% of Y to a Ni-28Mn-22Ga at.% alloy refines the ${L2}_1$ crystals, thereby increasing the compressive strength by about 3.4 times and enhancing the ductility by about 33% [211]. Yang et al. [73] studied the effect of substituting Mn with Fe, Cu, and Co in a Ni-25Mn-19Ga at.% alloy, reporting that the ductility improvement of the elements can be ranked as follows: Fe > Co > Cu. However, it should be noted that the addition of these elements can result in the formation of γ and γ′ phases, which hinder the reorientation of martensite, thereby deteriorating the magnetic and/or thermal shape memory properties to some extent.

The enhancement of the ductility of NiMnGa magnetic SMAs via the addition of a quaternary element is critical for improving their AM compatibility. Future studies are encouraged to explore the AM techniques that allow for the fabrication of compositionally gradient parts with NiMnGa magnetic SMA sections supported by NiMnGaX (X = additional alloying elements) ductile sections. Figure 8a–e demonstrate the enhancement of ductility and strength as a result of Ta, Co, Y, Fe/Co, and Cu additions, respectively. The Ta- and Y-containing alloys were tested via compression (Figure 8a,c), while the Co-, Fe-, and Cu-containing alloys were tested via tensile tests (Figure 8b,d). As shown in Figure 8a,e, Ta and Cu have the strongest effect on enhancing the ductility and toughness. The dotted arrow within the inset of Figure 8e shows the recovered strain of the Cu-containing alloy after heating it to 800 °C, indicating the presence of SME [213]. Furthermore, despite the addition of 8 at.% Cu, this alloy retained a single NM martensite-phase microstructure [213].

Crystal Size

The size of the ${L2}_1$ crystals is a determining factor in achieving a bamboo-grained structure, which possesses the best magnetic SME. As discussed previously, a bamboo-grained structure is achieved when the ${L2}_1$ crystals and the martensite plates within them span the dimensions of the part. In the case of porous or lattice structures, a bamboo-grained structure is achieved when the prior ${L2}_1$ crystals span the diameter or the thickness of the ligaments or struts. Hence, larger ${L2}_1$ crystals are typically preferred, and their growth typically takes place via post-AM homogenisation treatments at temperatures close to the alloy’s melting point for relatively long periods of time (24 or 48 h) [187,188]. The effects of the addition of Co, Bi, and Y on the prior ${L2}_1$ crystal size, and that of the internal martensite crystals in homogenised and ordered NiMnGa alloys, are shown in Figure 9. It can be observed that the substitution of 1.5 at.% of Mn with Co in a Ni-28.5Mn-21.7Ga at.% alloy increased the prior ${L2}_1$ average crystal size from 0.777 mm to 1.98 mm [209]. It should be noted that the crystal size for both alloys was measured after performing homogenisation and ordering heat treatments at 1040 °C and 800 °C for 24 h and 4 h, respectively. Importantly, the 5M martensite was preserved, and no γ phase was detected up to a Co addition of 2 at.% [209]. On the other hand, the addition of Y and Bi to a Ni-28Mn-22Ga at.% alloy tends to reduce the average ${L2}_1$ crystal size [211,214].

MTTs and the Curie Temperature

The addition of a small amount of a quaternary element can modify the stability of martensite phases in NiMnGa alloys, and thereby alter the MTTs and the Curie temperature. This is important to consider when adding these elements to enhance the ductility of these alloys or alter their crystal size. In addition to the magnetic SME, the NiMnGa alloys with appropriate stoichiometry, or after the addition of a quaternary element, can also exhibit high-temperature thermal shape memory and superelastic properties [215,216]. The AM of these alloys has not been investigated yet. The change in the MTTs (∆MTT) for NiMnGa alloys with the addition of a few different quaternary elements is summarised and demonstrated in Figure 10. As the effect of the Ta addition is significantly greater than that of the other elements, the Ta values were plotted separately for clarity. It should be noted that the change in the MTTs per at.% addition of each quaternary element was extracted from separate studies and linearly extrapolated using the range of elemental concentration covered in this particular study. Hence, the effect of these elements on the MTTs may differ from that shown in Figure 10 at concentrations that fall outside of those covered in the respective studies. As shown, Bi seems to be the only element that reduces the $A_f$ while increasing the $M_s$ and $M_f$, thereby reducing the transformation hysteresis. However, in the relevant study [214], a maximum amount of only 0.05 at.% of Bi was considered; therefore, future studies should investigate a broader range of added Bi. Additionally, as Mn evaporation remains a significant issue for the AM of these alloys, future research should investigate potential elements that can partially replace the Mn while maintaining the same martensite stability and the same MTTs. A sufficient replacement of Mn with other appropriate elements may potentially suppress or minimise Mn evaporation.

4. Additive Manufacturing of Shape Memory Lattice Structures

One of the primary advantages of AM is its ability to develop parts with complex geometries. Renowned examples of such geometries are lattice and cellular structures [221,222]. The fabrication of these structures is typically feasible only using AM; in addition to mass reduction, they allow for the structural design for specific properties, such as ultrahigh [223] or negative [224] stiffness, zero shear modulus [225], structural superelasticity [5], negative Poisson’s ratio (NPR) [226], or osseointegration [227], to be optimised. Moreover, combining the benefits offered by the AM-fabricated lattice structures with shape memory properties is predicted to give rise to new technological horizons. However, to realise the full potential of such shape memory-based lattice structures, important metallurgical and mechanical design-related factors should be considered. This section will discuss the recent findings and important considerations for the design and the AM of NiTi and NiMnGa shape memory lattice structures.

4.1. NiTi Lattice Structures

Most microstructural studies on the AM of NiTi in the literature are focused on the bulk specimens [136,154,221,228,229], and only a few investigate thin or lattice structures [121,221,230]. Generally, the AM of thin parts limits the processability window and intensifies the sensitivity of the final part to thermal gradient-related effects such as residual stress and thermal distortions. This is particularly challenging for NiTi parts as such processing issues could drastically alter the martensitic transformation characteristics of the final part [231,232]. For instance, Biffi et al. [221] showed that the $A_s$, $A_f$, $M_s$, and $M_f$ exhibited by LPBF-printed NiTi lattice specimens were 14 °C, 46 °C, 33 °C, and 34 °C higher than their bulk counterparts fabricated with the same energy density of 127 J/mm³, respectively. Considering that no significant difference between the overall compositions of the lattice and bulk samples was detected, it is believed that this difference in MTTs was a direct effect of the greater residual stress within the lattice structures. A greater inhomogeneity of the crystallographic texture of the tetragonal diamond-like lattice NiTi structures compared to that of the bulk samples, while exhibiting a maximum recoverable strain of 4.5%, was also identified [221].

Most of the recent literature on simple NiTi lattice structures, such as simple cubic [121,233,234], body-centred cubic [121,235], or octahedral structures [236], is focused on how different mechanical properties, including SME and SE, are influenced by process parameters, porosity, or design parameters such as strut dimensions. However, these structures are not specifically designed for superelasticity or shape memory properties. In some cases, their limited allowable strain can deleteriously affect their superelastic or shape memory recoverable strain [237]. This has sparked a recent trend in the superelastic AM research community of aiming to develop new, often bioinspired, structures that are specifically designed for shape memory and superelastic properties and energy absorption applications [21,237,238,239,240,241].

Table 4. Recent AM-fabricated NiTi lattice/cellular structures alongside their measured specific energy absorption (SEA) and strain recovery ratio (RR).

Structure No.	2D	3D	SEA (J/g)	RR (%)	Reference
1			1.6	93.6 (SE)/98.5 (SE + SME)	[21]
2			2.8	93.7 (SE)/98.7 (SE + SME)	[21]
3			6.3	91.2 (SE)/98.7 (SE + SME)	[21]
4			1.53	94.7 (SE)/99.7 (SE + SME)	[21]
5			1.85	99.17	[237]
6			1.25	99.78	[237]
7			N/A	96.17	[238]
8			N/A	95.5	[238]
9			N/A	93.93	[238]
10	Not Available		N/A	$~$74%	[239]

lists ten different types of AM-fabricated NiTi lattice and cellular structures whose superelastic and shape memory properties were recently investigated. These NiTi structures are primarily designed for energy absorption applications; therefore, after performing cyclic compression tests, their specific energy absorption (SEA) and strain recovery ratio (RR) were calculated. The SEA parameter is defined as follows [21]:

$$\mathrm{SEA}={\displaystyle \frac{{\int }_0^{{\mathrm{S}}_{\mathrm{e}\mathrm{f}}}\mathrm{F}\left(\mathrm{S}\right)\mathrm{d}\mathrm{S}}{\mathrm{M}}}$$

(4)

where S is the displacement exhibited by the structure during the compression test, M is the total mass of the specimen, and ${\mathrm{S}}_{\mathrm{e}\mathrm{f}}$ is the compression displacement when the energy absorption coefficient f, defined below, reaches its maximum:

$$f={\displaystyle \frac{{\int }_0^{\mathrm{S}}\mathrm{F}\left(\mathrm{S}\right)\mathrm{d}\mathrm{s}}{{\mathrm{F}}_{\mathrm{m}\mathrm{a}\mathrm{x}}}}$$

(5)

where ${\mathrm{F}}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ is the maximum force that the structure can bear during the compression process. Similarly, the RR parameter is defined as the ratio between the total applied compressive strain and the recovered strain after unloading. The RR (SE) only takes the elastic and superelastic strain recovery into account, while the RR (SE + SME) considers the additional strain recovery via SME after heating. These properties indicate the material’s ability to absorb mechanical energy both elastically (RR) and plastically (SEA).

$~$Bioinspired structures—Yu et al. [21] investigated the bionic structures (Structure No. 1 to 4 in Table 4), which were inspired by a combination of honeycomb structures and the columnar fibre structures found in the elytra of the ghost shovel-shaped beetle. This study revealed that Structure No. 3 exhibited a significant SEA, at 294%, 126%, and 314% greater than that exhibited by Structures No. 1, 2, and 4, respectively. Additionally, relative to other structures, it exhibited a greater peak force and a noticeably smaller load fluctuation (fractures), which makes it much more suitable for energy absorption applications. Furthermore, the total superelastic (after unloading) and shape memory (after heating) strain recovery exhibited by Structures No. 1 to 4 after five compression cycles was found to be 98.5%, 98.7%, 98.7%, and 99.7%, respectively. Similarly, Structures No. 5 and 6 were inspired by the frustule of the Campylodiscus diatom, whose unit cell includes an external frame structure and an internal support structure [237]. Each of these structures contained four distinct angles; selectively changing their radii from traditional sharp angles (TSA) to bionic arc angles (BAA) gave rise to a quantity described as the ratio of TSAs to BAAs, denoted as ξ. Sun et al. [237] designed four unit cells with distinct ξ values, and Structures No. 5 and 6 exhibited the two most extreme ξ values of 6⁄0 and 0⁄6, respectively. As the number of BAAs increased (i.e., ξ changed from 6⁄0 to 0⁄6), the compressive maximum first peak force decreased by 31.5%, the elastic modulus decreased by 34.2% and the shape memory properties first deteriorated then improved. These studies clearly indicate the importance of structural design and controlling the stress concentration regions of AM-fabricated NiTi lattice structures.

A group of LPBF-fabricated NiTi fractal structures with geometries that mimic the architectures of human bone and porosities ranging from 25.93% to 95.64% (three examples of which are Structures No. 7, 8, and 9 in Table 4) were explored for biomedical and energy absorption applications [238]. These fractal structures exhibited great superelasticity and an almost complete strain recovery at up to 3% of the applied compressive strain. Interestingly, the fractal structures did not exhibit any sign of structurally driven deformation, which is conventionally observed as a segmented plateau in the stress–strain curves of metal foams [242,243]. This indicates that their deformation mode is driven predominantly by the intrinsic material properties and not the structure. The experiments indicate that increasing the fractal orders or the fractal dimensions consistently decreased the superelastic RR and the SEA. This was attributed to the higher number of high-stress concentration regions near the sharp edges of multilevel fractal pores. Structures No. 3 and 6 in Table 4 are the lattice structures that exhibited the largest SEA and RR, respectively.

Honeycomb structures are another group of nature-inspired structures that provide significant stiffness for out-of-plane directed loads while providing great in-plane flexibility, which is of immense benefit for shape memory or superelastic actuation [244,245,246]. Additionally, the vertical-wall nature of honeycomb and honeycomb-based lattice structures, and their therefore low dependency on support structures, enhances their additive manufacturability. Recent studies have revealed that LPBF-fabricated NiTi honeycomb-inspired structures exhibit stable superelastic recoverable strains as high as 30% [247]. One important aspect of these structures if aiming to realise their full superelastic potential is their stress concentration points, which can be minimised by eliminating non-continuous curvatures [247,248,249].

Triply Periodic Minimal Surface (TPMS) lattice—Triply Periodic Minimal Surface (TPMS) lattice designs are also promising candidates for AM-fabricated NiTi structures due to their uniform radius of curvature and their smooth outer surface, allowing for a homogenous stress distribution under loading [239,250,251,252]. Additionally, TPMS structures exhibit a high stiffness-to-weight ratio and good compatibility with AM [253,254,255]. Amongst the different NiTi TPMS structures, the gyroid-type design has been shown to exhibit superior fatigue properties [236], which is essential for superelastic or shape memory applications. The gyroid TPMS (Structure No. 10 in Table 4) exhibited stable superelasticity with an almost complete strain recovery at an applied strain of up to 4% [239]. The geometrical design of AM-fabricated NiTi TPMS gyroid structures can be modified to further enhance their strength, fatigue, superelastic, and shape memory properties. Chen et al. [256] studied the effect of inducing porosity gradients in LPBF-fabricated TPMS gyroid NiTi structures on their strength and fatigue properties. This study suggested that the induction of a porosity gradient perpendicular to the build direction (Y-GCS) results in higher strength and better fatigue properties than those exhibited by samples with uniform porosity (U-GCS). The failure mechanism of the Y-GCS and U-GCS specimens at low fatigue loadings was found to be a distinct layer-by-layer collapse while a higher fatigue loading led to the formation of ~45° oriented shear bands and ultimately cracking along the sample diagonal. However, the induction of a porosity gradient parallel to the build direction led to layer-by-layer failure at all stress levels. This alteration in the structurally driven deformation mechanisms gives rise to the dependency of the mechanical properties of such structures on their unit cell design parameters, such as the orders of hierarchy of their porosity [256], as well as their unit cell size and volume fraction [257].

To summarise, the AM of superelastic and shape memory NiTi lattice structures has attracted substantial attention from the research community; however, the fundamentals of their structural and process design parameters still require more in-depth exploration. Numerical simulations should be developed for the topological optimisation of these lattice structures with special consideration of stress concentrations and martensitic transformation characteristic stresses and temperatures. Furthermore, as superelastic NiTi lattice structures are expected to undergo large loading cycles in practice, future research should dedicate further attention to the failure mechanisms and improving their fatigue properties.

4.2. NiMnGa Lattice Structures

Compared to NiTi, the research conducted on AM-fabricated NiMnGa magnetic SMA lattice structures has been extremely limited, with the majority of the available literature focused on BJ [178,258]. This is attributed mainly to the intrinsic brittle nature of NiMnGa and the significant residual stresses generated during the laser-beam AM processes, which pose significant challenges to the printability of NiMnGa-based lattice structures. Unlike NiTi, the most important benefit of developing NiMnGa structures with a controlled porosity is that the constraints imposed on twin boundary motion by the grain boundaries are minimised. Additionally, as NiMnGa alloys are generally brittle at room temperature [178,259], incorporating energy absorption properties into the porous structure is also critical.

Some examples of BJ-fabricated NiMnGa lattice structures are shown in Figure 11a–e. The trestle-like NiMnGa structure (Figure 11a) was fabricated using a pre-alloyed ball-milled powder. The microstructure contained twinned 5M martensite with different twin variants extending across the sintered surfaces, as illustrated in Figure 11b [178]. The more intricate NiMnGa micro-truss shown in Figure 11c was fabricated via a novel extrusion-based AM technique using liquid inks consisting of Ni, Mn, and Ga powders and a polymer binder. The binder was then removed via an annealing process, followed by a sintering process to produce the final NiMnGa micro-truss structure. The process resulted in a hierarchy of pore sizes, giving rise to a bamboo-grained structure, as shown in Figure 11d [185]. The larger pores were realised via the initial truss design (450 µm channels) and the medium-sized pores were produced via the melting and diffusion of Ga particles (50–100 µm pores), while incomplete sintering led to the formation of the smaller ones (5–25 µm) [185]. Moreover, as previously discussed, lattice structures can be designed to induce specific properties that were not intrinsically exhibited by the bulk material. For instance, the controlled porosity and structural design of the gyroid lattice structure shown in Figure 11e enhanced the ductility of the intrinsically brittle NiMnGa to about 14% [258].

Milleret et al. [188] investigated the LPBF fabrication of 7M NiMnGa lattice structures, shown in Figure 11f, with a strut thickness of 200 µm and an internal strut density of ~99%. Performing homogenisation heat treatment at 1060 °C for 12 h, followed by an ordering treatment at 800 °C for 4 h, led to the formation of bamboo-grained structures for a few of the struts, as shown in Figure 11f.

Despite the great success of these methods in fabricating NiMnGa lattice structures with a hierarchical porosity and bamboo-grained microstructure, no MFIS has been reported so far. Given the geometric nature of the lattice structures, future research is encouraged to investigate the use of NiMnGa hierarchical lattice structures for energy absorption applications via PBF-based AM. Post-processing procedures could be applied to induce further orders of porosity. Numerical simulations should also be performed to understand the possibilities of the simultaneous control of grain growth and porosity in an in situ manner with the aim of producing bamboo-grained structures in the as-printed NiMnGa lattice structures.

5. Summary and Future Perspectives

Despite the excellent SE and SME offered by NiTi and NiMnGa SMAs, the SE of NiTi and the brittle nature of NiMnGa pose significant challenges in processing them into desirable and intricate geometries via conventional routes. Fortunately, AM offers a promising alternative for fabricating these materials into complex geometries. However, in addition to the SE- or SME-focused structural design, the success of the AM of these alloys necessitates an in-depth understanding of the unique thermal conditions of AM processes, determining the chemistry, microstructure, the stress state, and ultimately the SE and SME of the final part. This review provides an unbiased feasibility and optimisation analysis of the AM of NiTi and NiMnGa SMAs, with critical insights summarised as follows:

(1) Recent insights into the driving mechanisms of SME and SE in NiTi and NiMnGa reveal the multifaceted nature of their successful AM, demanding simultaneous control over the feed stock (powder and/or wire) quality, part design, training procedures, AM process parameters, AM atmosphere control, and post-processing methods and procedures.

(2) Developing balanced AM processing parameters to achieve high density, minimising the Ni evaporation, and gaining control over the crystallographic texture, the MTTs, and the microstructural homogeneity of the final part constitute the current challenges of the AM of NiTi.

(3) Laser-based PBF and DED are the most common AM techniques used for fabricating NiTi parts. However, due to its relatively higher deposition rates and lower overall costs, WAAM is becoming an increasingly popular alternative method. The current research on WAAM-fabrication of NiTi is mainly focused on studying the associated thermal profiles and interactions between the heat source and the wire. These studies are primarily aimed at gaining control over the microstructure, defects, and the overall properties and, to a lesser extent, improving the SME and SE performance.

(4) Even though considering the energy density alone is generally sufficient for achieving high density in LPBF- and DED-fabricated NiTi, our analysis indicates the importance of taking individual process parameters into account for achieving SME and SE. Hence, for the first time, novel empirically derived process design maps, based on individual process parameters, were developed specifically to maximise the SME and SE of laser PBF- and DED-fabricated NiTi. These preliminary maps are intended to form the basis of a continuously evolving process design tool.

(5) Twin boundary mobility driven by the alignment of an external magnetic field with the easy magnetisation axes of martensite crystals is essential for the magnetic SME in NiMnGa alloys. However, grain boundaries are potent inhibitors of twin boundary motion. Therefore, the current efforts regarding the AM of NiMnGa magnetic SMAs are primarily focused on obtaining a bamboo-grained structure and enhancing their crystallographic texture to improve their twin boundary mobility. Mn evaporation during the AM of NiMnGa alloys poses another significant challenge, leading to undesirable chemistry alteration and pores that can act as additional twin boundary motion inhibitors. These considerations, as well as the intrinsic brittleness of NiMnGa, lowering their tolerance to residual stress, form the major challenges in their AM.

(6) Considering the key challenges of the AM of NiMnGa, doping with elements such as Cu, Co, Y, Fe, and Ta was proposed as a means of forming a bamboo-grained structure, suppressing the Mn evaporation, and enhancing the ductility of these alloys, while potentially preserving their magnetic SME.

(7) Laser-based PBF and DED have been shown to fabricate NiMnGa specimens whose individual crystals exhibit significant MFIS values; however, they have yet to produce specimens with a noticeable MFIS in bulk format. A simultaneous in situ process control of porosity and grain size aiming to achieve a bamboo-grained structure during the AM process is believed to be a yet-to-be-realised pinnacle in this field of research.

(8) The recent research on the AM of NiTi lattice structures is headed towards the design of an often bio-inspired hierarchical porosity specifically aimed at maximising their SME and SE. Apart from the AM process parameter optimisation to ensure a desirable and homogenous microstructure, low residual stress, and suitable MTTs, the success of the AM of NiTi shape memory or superelastic lattice structures calls for careful part design. The design should focus on the distribution of stress across the part to avoid undesirable pre-mature martensite stabilisation at stress concentration points during service. Future research on NiTi lattice design is encouraged to take shape memory training procedures into consideration to allow for AM-fabricated NiTi lattices to undergo training and exhibit a two-way SME.

(9) The AM of NiMnGa lattice structures is a more recent field of research and is primarily aimed at controlling the ligament/strut size to obtain a bamboo-grained structure. While the available literature on the laser-based PBF or DED of NiMnGa lattice structures is limited, BJ-fabricated lattices have been shown to contain bamboo-structured ligaments. However, a noticeable MFIS has yet to be achieved.