1 Introduction

Selective laser melting (SLM/LBPF) is a laser-based powder-bed additive manufacturing procedure which allows fabricating parts and components in a layer-by-layer approach from the pre-alloyed or pre-mixed material powder [1]. With the use of additive manufacturing (AM) methods, there is almost no limitation for the geometry and material selection. Different parameters affect the quality and thermomechanical properties of the SLM-fabricated parts. Part of these parameters are related to the laser source, such as the laser type, scanning speed, scanning strategy, laser power, and hatch spacing, and the rest are the parameters related to the powder and build conditions, such as the properties of powder particles (e.g., shape and size distribution), layer thickness, build chamber temperature and oxygen level, build orientation, and support types [2,3,4,5,6,7]. Only an optimized combination of all these parameters leads to a successful fabrication with a high level of density and acceptable thermomechanical properties.

Recently, the SLM technique has been optimized and developed [8,9,10,11,12,13] for fabricating special alloys, such as NiTi, in which the conventional methods of fabrication are challenging [14, 15]. Due to the coupled thermomechanical properties, NiTi as a shape memory alloy (SMA) appears in two solid-state phases, i.e., austenite and martensite. The reversible phase transformation between these phases provides the material with unique characteristics, such as shape memory effect (SME) and superelasticity (SE) which enables NiTi to recover a large amount of deformation upon unloading (in SE) or heating up (in SME) [16, 17]. Besides, valuable functional properties, good biocompatibility, high corrosion resistance, and excellent damping properties make NiTi a promising candidate for a wide range of applications including aerospace, automotive, and biomedical [18,19,20,21,22,23,24,25] applications. Successful SLM fabrication of NiTi with acceptable superelastic and shape memory properties enables the design and development of novel functional parts, e.g., stiffness-modulated bone implants [26, 27] as well as sensors and actuators [28, 29].

Using the unique mechanical properties of NiTi and the design freedom provided by the additive manufacturing, our group has introduced a new generation of bone fixation implants for improving the bone reconstruction surgeries [26, 27, 30,31,32,33]. The conventional bone fixation implants provide a high level of stiffness in comparison to the adjacent bone tissue which leads to long-term complications. Although these high stiff bone fixation implants provide a high level of immobilization immediately after the surgery after the bone healing, the stiffness mismatch between the implant and the bone tissue causes stress shielding and bone resorption. The excessive bone resorption may lead to bone or implant failure which necessitates a revision surgery [31]. As a solution, our group introduced a new design methodology that enables stiffness modulation for bone fixation plates based on the patient’s need and therefore minimizes the stress shielding effect. Through finite element simulations, we have shown that by introducing an engineered level and type of porosity to the geometry of the bone plates, one is able to achieve implants with only the required level of stiffness (also referred to as stiffness matching). The complex-shaped porous implants can only be fabricated via additive manufacturing [27]. Although we have been able to fabricate realistic stiffness-modulated implants with the use of AM and implementing our methodology, it was shown that the AM-fabricated NiTi implants may contain microsize pore/voids and have a rough surface due to the nature of the SLM method that adversely affects the performance. Figure 1 shows a sample of the designed bone fixation plates that contain a high level of porosity (46% porosity). The fine porous structures do not allow the use of conventional methods, such as mechanical polishing and hot isostatic pressing (HIP) for improving the surfaces and reducing the microsize pores/voids, respectively. Therefore, other alternatives for these purposes should be investigated. More information on the design of these bone plates can be found elsewhere [27].

Fig. 1
figure 1

A sample of the porous bone fixation plates with modulated stiffness. Region A and cross section B are magnified for a better demonstration (red surfaces in the cross section demonstrate internal surfaces)

Full size image

In general, static mechanical properties of the SLM-fabricated parts are comparable with the conventionally fabricated parts and in some cases may be even superior [34,35,36]. Although SLM provides numerous advantages and design freedoms, surface quality, geometrical accuracy, fatigue resistance, and relative density of the parts may be inferior to conventional technologies [37,38,39,40]. By optimizing the process parameters, some research groups have successfully fabricated NiTi samples with a high level of relative densities (i.e., above 99%) demonstrating SE and SME [41,42,43,44,45]. However, achieving a 100% relative density by eliminating internal micropores/voids and profoundly improving the mechanical properties, such as fatigue behavior, has not yet been obtained. In addition, SLM-fabricated parts generally have rough surfaces which also adversely affect the fatigue life of the components. On the other hand, in many applications including bone implants, a NiTi component undergoes repetitive cyclic loading and is supposed to last up to five million cycles in the body that restricts the use of SLM-fabricated parts [46, 47]. Therefore, reducing the internal voids/pores, increasing the relative density, and improving the surface quality of the SLM-fabricated NiTi parts are of high interest and are being studied by different groups. In addition to optimizing the SLM process parameters, some post-processing techniques such as annealing, aging, electropolishing, chemical polishing, and HIP have been used to homogenize and customize the mechanical properties, improve the surface quality, reduce the internal porosities of the parts, and eventually improve the fatigue performance [48,49,50,51,52,53,54].

In this article, we investigated an improvement of the SLM process, by introducing selective laser remelting, to improve the relative density and surface roughness of the AM-fabricated NiTi parts such as the bone fixation plates. Remelting is an in situ strategy during the SLM fabrication, in which the laser rescans each layer after the solidification and before the deposition of the powder for the next layer. Similar to the main scan, remelting has different process parameters related to the laser, such as laser power, scanning speed, and hatch spacing. Remelting has been employed to improve the densification, surface roughness, and grain refinement of the other alloys [55,56,57]. It has been shown that due to the higher cooling rate during the remelting than the main laser scanning of the powder bed, finer grains can be achieved [58]. It is well-known that finer grain sizes can prevent the crack nucleation and improve the fatigue life of the parts [59]. Besides, some studies have concluded that remelting can increase the corrosion resistance and hardness and decrease the cracks and internal pores [60, 61]. Gutsmann et al. [62] examined various remelting procedures to study the effect of remelting on the thermomechanical properties of Cu-Al-Ni-Mn as a shape memory alloy. They reported that although the mechanical properties of the parts did not significantly change, the transformation temperatures (TTs) changed depending on the remelting process parameters. Yasa et al. [63] studied the effect of different remelting process parameters, including the number of remelting paths, scanning speed, laser power, and scanning space on the properties of Ti-6Al-4V and stainless steel 316. They showed that the remelting process could increase the surface quality of the parts, and the most influential factor to achieve the highest density for the parts was the scanning speed of the remelting process. On the one hand, Zhang et al. [64] concluded that the remelting could not lead to a big difference on the microstructure and grain sizes of the AlSi10Mg samples. They even reported that the remelting process could reduce the density of the parts by enlarging the melt pool boundaries and increasing the pores at the bottom of the melt pool, which can be a source of early fracture. In some studies, researchers reported that the high cooling rate and thermal gradients during the remelting procedure can produce residual stresses and make the parts more textured; however, after the second or third time of remelting, the thermal gradient decreases and residual stress can reduce [65]. Shiomi et al. [66] showed that residual stress could be reduced by 55% in the top layer of the parts by selecting sufficiently high energy densities during the remelting procedure. Ali et al. [67] showed that although remelting with 150% energy density of the main SLM process reduces the residual stress of Ti64 samples by ~ 30%, it caused premature failure of the sample during mechanical testing because of the formation of a thick oxide layer during the remelting.

In this work, for the first time and as an essential step for fabricating functional NiTi parts via SLM, we have studied the effect of in situ remelting during the SLM fabrication of NiTi parts. To this end and through a series of experiments, we have evaluated the effect of remelting on the thermomechanical properties, density, and surface roughness of the SLM-fabricated NiTi specimens. Eventually, an optimized set of remelting process parameters for the successful fabrication of NiTi porous parts is reported. This study should also pave the way for the application of SLM-fabricated NiTi in other realistic applications.

2 Materials and methods

The starting NiTi ingot had a composition of Ni50.8Ti at.%. The ingot was gas atomized by TLS Technik GmbH (Bitterfeld, Germany) and sieved to produce powder with a distribution of 15–65 μm. A selective laser melting (SLM) machine (ProX 200, 3D Systems, Rock Hill, USA) with the maximum laser power of 300 W was used for printing specimens. Parts were fabricated on grid support structures in an argon atmosphere from fresh (unused) powder. The chamber was purged with argon during fabrication and O2 level was kept below 600 ppm. CAD dimensions (4 × 4 × 10 mm cuboids) were identical for all the parts with the parameters listed in Table 1. The process parameters, as shown in Fig. 2, were selected in a way to cover a relatively wide range of process parameters (PPs). Control part, also called the as-built sample (#1), was fabricated using our standard process parameters obtained from previous studies [68] and in fabricating it, similar to the conventional SLM, laser melting happened only once for each layer. For parts #2 through #8, each layer was exposed to a second laser pass, which was rotated 90° as depicted in Fig. 3. As for other remelting PPs, the hatch spacing (h) and layer thickness (t) were kept constant for all the parts, while laser power (P) and laser speed (v) were changed. P and v values for the second melting (i.e., remelting) were chosen in a way to obtain 20–80% of the primary energy density (E = P/v.h.t), which was 83 J/mm3. Scanning speed versus the laser power, as well as the energy ratio for the remelting samples, is also shown in Fig. 2. Densities were measured by the Archimedes method by a Mettler Toledo (Toledo, USA) density-determination kit. Differential scanning calorimetry (DSC) was employed to find transformation temperatures using a Q20 DSC (TA Instruments, New Castle, USA) with a heating/cooling rate of 10 K/min. Samples were cycled between − 50 and 80 °C. Surface roughness and roughness maps were measured by a Keyence VHX 6000 (Keyence, Osaka, Japan) and a non-contact surface measurement system by Talysurf® (Leicester, UK). The FEI Quanta 3D FEG environmental scanning electron microscope (Thermo Fisher Scientific, Waltham, USA) was also used for high-resolution imaging. A 25-kN TestResources (Shakopee, USA) mechanical testing machine equipped with hydraulic actuators was used to perform compression tests. A Nikon XT H 225 X-ray and CT inspection 225-kV-microfocus system with 3 μm focal spot size (Brighton, MI, USA) was employed to 3D-scan the specimens and to quantitatively assess the internal pores/voids of the parts.

Table 1 Process parameters used for fabrication
Full size table
Fig. 2
figure 2

The laser scanning speed versus the laser power for the remelting samples. The number of the samples as well as the energy ratio with respect to the original process parameters (sample 1) is also shown for each sample

Full size image
Fig. 3
figure 3

Schematics of regular fabrication versus remelting strategy where each layer was melted twice with a second laser pass

Full size image

3 Results

3.1 SEM evaluation

Material properties and PPs influence the shape, size, wettability, and stability of the melt pools and surface quality of the parts [69,70,71]. As the first step, the surfaces of the samples were evaluated via SEM to analyze the effect of remelting. Figure 4 shows SEM images of the top surface of as-fabricated as well as remelted SLM parts with different PPs. Since the main SLM PPs are the same for all the parts, in the next sections of the paper, we refer to the remelting PPs as the PPs.

Fig. 4
figure 4

Top surface SEM micrographs of as-built and remelted samples. The energy ratio with respect to the original process parameters (Eratio) for the samples are as follows: (Eratio)1 = 0, (Eratio)2 = 31%, (Eratio)3 = 36%, (Eratio)4 = 78%, (Eratio)5 = 21%, (Eratio)6 = 44%, (Eratio)7 = 52%, (Eratio)8 = 50% (yellow arrows show the ballings in parts #1, #2, and #5)

Full size image

The as-built condition (sample#1, Fig. 4.1) provides a non-smooth surface that includes several regions containing the balling effect (shown by the arrows), un-melted powder, and spattered particles, whereas the remelted samples, generally, show much more stable and flatter melt pool (Fig. 4.2 to .8). In the as-built sample (with no remelting), the melt pool forms and flows within a layer of powder that is not deposited uniformly over the substrate (because of the defects of the previous layer). Since the particle sizes were in the range of 15–63 μm, a layer thickness of 30 μm results in a non-uniform powder distribution on the underlying bed, which leads to different energy absorptions in different locations around the laser spot and, therefore, melt pools with different sizes form [72]. Also, the unsmooth underlying solidified surface in the regular SLM process results in a non-uniform powder deposition. Moreover, the higher energy absorptivity in the powder form of the material, in comparison to the bulk form, leads to a higher absorbed energy and the concave shape of melt pools [73]. In general, when a melt pool continuously moves forward toward forming a relatively long scanning track, it experiences some level of instability at some points that leads to the balling effect (shown with the arrows on Fig. 4) and a non-smooth surface, consequently [74, 75]. This effect was not fully disappeared in the remelting specimens with low energy densities, such as samples #2 and #5 (Fig. 4.2 and Fig. 4.5). All these factors play key roles in the unsmooth surface that is formed in the as-built sample or prior to the remelting in the remelted samples. In the remelted samples, the solid substrate provides a more uniform surface with a lower energy absorption. Wetting the melt pool to the underlying material leads to a flat melt track, which makes a smooth surface for the next layer deposition. Therefore, a more stable and flatter melt pool can be seen for remelted scans.

Samples #2 and #5 (Fig. 4.2 and Fig. 4.5) have lower energy density in remelting process in comparison to the other specimens (energy ration is less than 31%) and the average width of the melt pool tracks are 42 μm and 33 μm, respectively. While the hatch spacing for the fabrication is 80 μm, therefore, there is no overlap between remelted tracks in these two samples (Fig. 4.2 and Fig. 4.5). As a result, because of not having an overlap, these samples were not homogeneously and fully remelted. In addition, as it was mentioned earlier, the smoothness of the layer that new powder will be deposited on, affects the new layer smoothness quality. Therefore, in samples #2 and #5 (Fig. 4.2 and Fig. 4.5), due to the lack of sufficient energy density, some unstable regions can be observed, where the remelting tracks pass through the concave shape of the underlying melt pools. Therefore, these two sets of PPs are not efficient parameters for the remelting process.

For samples #3, #4, #6, #7, and #8 with higher remelting energy density (Fig. 4.3, .4, .6, .7, and .8), stable and flat melt pool tracks with a sufficient overlap are achieved. For instance, sample #4 (Fig. 4.4) with the energy ratio of 78% and sample #7 (Fig. 4.7) with the energy ratio of 52% (both with energy ratios higher than 50% in comparison to the as-built sample) have shown the best surface quality in the SEM images. As a result, the smooth surfaces achieved in the aforementioned four samples with high energy densities resulted in a proper powder feeding and fewer surface defects. Since these trends occurred in every layer, the remelting process also could reduce the porosity and improve the density, which is evaluated in the next sections.

3.2 Surface roughness evaluation

Surface roughness is one of the key factors in representing the quality of SLM parts. As it was observed in the SEM analysis, the remelting technique can have a significant effect on the quality of the top surface. Non-contact profilometry was performed to assess the effect of remelting on the surface texture of the SLM parts. Figure 5 shows the texture maps of the side surface for each sample. The colors show the height of each point with respect to the color bar. Areal field parameters were calculated and tabulated in a table in Fig. 6. Among these values, 1D and 2D arithmetical mean height—Ra and Sa, respectively—are mostly adopted [76] to evaluate roughness and can be calculated by:

$$ \mathrm{Sa}=\frac{1}{A}{\iint}_A\left|Z\left(x,y\right)\right| dxdy $$
(1)
$$ \mathrm{Ra}=\frac{1}{l}{\int}_l\left|Z(x)\right| dx $$
(2)

where A and l are the target area and the target length on the sample, respectively, and Z is the height of each point. Ra is the arithmetical mean deviation of a line and Sa is the extension of Ra to a surface. Basically, Sa shows the difference in height of each point compared to the arithmetical mean of the whole surface, and Ra shows the difference in height of each point compared to the arithmetical mean of the line. As it can be seen in Fig. 6, surface roughness of the side surfaces slightly changed followed by the remelting process, except in case of samples #5 and #7 in which the change in side surface roughness was not negligible. The average Sa values for samples #2, #3, #4, #6, and #8 were measured as 0.0170 ± 0.002 mm which is comparable to the Sa value of 0.0172 for the as-built sample (#1). On the other hand, there was a 41% improvement in case of sample #7 remelted with a laser power of 87.5 W and a scan speed of 800 mm/s, while a 50% deterioration in case of sample #5 which had the lowest remelting energy density. This trend can also be seen for the other texture parameters such as Sq, Sp, and Sz values in Fig. 6, which show the root mean square height, maximum peak height, and maximum height, respectively.

Fig. 5
figure 5

Surface roughness maps from the side of the samples (parallel to the building direction). (Eratio)1 = 0, (Eratio)2 = 31%, (Eratio)3 = 36%, (Eratio)4 = 78%, (Eratio)5 = 21%, (Eratio)6 = 44%, (Eratio)7 = 52%, (Eratio)8 = 50%

Full size image
Fig. 6
figure 6

(Left) Areal field parameters for side of the samples. (Right) Variation graph showing Sa values

Full size image

Figure 7 shows the surface roughness of the top surfaces measured for the as-built (#1) as well as the remelted samples. As it can be seen in this figure—and was earlier observed using the SEM analysis—the top surface significantly smoothened and flattened in the case of remelting that could be up to two times the as-built sample (samples #7 and #6). Although side surfaces were almost not affected by the remelting process, improvements on the top surface are expected to be an indication of the improvement of the internal pores/voids and the relative density of the SLM parts. No remarkable correlation could be found between energy density and surface roughness. Melt pool overlap (hatch spacing), scanning speed, and laser power are the important parameters affecting the morphology and wettability of the melt pool [77, 78]. Stable and flat melt pool leads to a smoother surface [79]. To further investigate the internal features of the part, X-ray computed tomography (CT) and density measurements were used as described in the following section.

Fig. 7
figure 7

Roughness of the tope surface of the remelted as well as the as-built samples, demonstrated on (a) based on the sample number and (b) based on the process parameters. (Eratio)1 = 0, (Eratio)2 = 31%, (Eratio)3 = 36%, (Eratio)4 = 78%, (Eratio)5 = 21%, (Eratio)6 = 44%, (Eratio)7 = 52%, (Eratio)8 = 50%

Full size image

3.3 X-ray computed tomography

There are various types of defects in additive manufacturing: cracks, delamination, lack of fusion, and porosity [80]. Based on our previous studies, the optimized process parameters resulted in crack-free samples with densities of over 99% [81]. However, as it is shown in optical micrographs in Fig. 8, there are still some microsize pores with both spherical and irregular shapes in the parts. Formation of such pores is attributed to different mechanisms that are mainly the keyhole effect, lack of fusion (LOF), and the trapped gas [82,83,84]. As the result of excessive energy densities, the melt pools could be deeper and in some spots gas could entrap, form porosities, and cause improper closing of a keyhole. Fig. 8A and Fig. 8A1 show the example of entrapped gas porosities with spherical shapes in one SLM sample. In the keyhole mode, due to the excessive energy, relatively deep melt pools (with the depth-to-width ratio of more than 0.5) form that can melt up to 8 times than a normal melt pool [85, 86]. On the other hand, when the energy is not enough, it is possible to form LOF areas and incomplete solidification during the fabrication. Except for entrapped gas and keyhole porosities with spherical shapes, other kinds of defects mostly have an irregular shape. Figure 8B shows defects with irregular shapes which can be formed on the boundary of the melt pools due to the lack of fusion. In addition, the location of the defects can be different based on their type. A combination of these pores could be formed during the fabrication in some regions, even with a sufficient energy density in microsizes. From the semi-circular shape and size of the melt pools, it is evident that keyholing may not the cause of the spherical pores. Thus, LOF and gas trapped are the main reasons for the porosities of the parts in this study. In order to quantify and visualize the pores/voids at the microscale, X-ray computed tomography scan (CT scan) was utilized and the results for samples #4, #8 and #1 are reported in Fig. 9b. Using CT scan, one is able to detect defects larger than the resolution (here it is 0.03 mm) in the scanned area and measure the sphericity of them. Geometry and the size of the defects together help in distinguishing the types of the defects and analyzing them. Defect sphericity (Φ) is defined as a dimensionless value to quantify the irregularity of the defect shape. The value of sphericity is equal to the ratio of the surface of the defect to the surface of a sphere with the same volume as the defect, and it can be in the ranges between 0 and 1. A sphericity value of 1 represents a perfect sphere, whereas a value close to 0 represents a highly irregular shape defect. Sphericity (Φ) for each defect can be calculated using Eq. 3, in which V is the volume of the defect and A is the area of the defect [87]:

$$ \varPhi =\frac{4\pi }{{\left(4\pi /3\right)}^{2/3}}.\frac{V^{2/3}}{A} $$
(3)
Fig. 8
figure 8

Pore morphology along and perpendicular to building direction (BD) of AM NiTi with regular processing (#1). Yellow arrows show spherical pores, while black arrows show irregular shape pores. It should be noted that the curved black lines are melt pool boundaries, which were burnt during etching and are not defects

Full size image
Fig. 9
figure 9

Average of void diameter and sphericity for samples #1 (as-built with Eratio = 0), #4 (Eratio = 78%), and #8 (Eratio = 50 % ) based on the X-ray CT results

Full size image

Entrapped gas porosities usually have smaller dimensions and they do not have sharp edges. In fact, they exhibit the shape of a sphere and have higher sphericity values close to 1. LOF regions on the other hand are relatively larger and have irregular shapes with sharp edges. For this type of defects due to the irregular shapes, the sphericity values are closer to 0. Based on the results, no general trend can be found based on the average void diameter or average sphericity values for the remelted samples in comparison with the as-built one (Fig. 9a). However, as it is shown in Fig. 9b, there are less irregular-shaped defects in samples #4 and #8 in comparison to sample #1. It could be concluded that remelting process was able to remove LOF defects, while no significant change can be observed in the entrapped gas porosities. It should be noted that the presence of LOF defects is more detrimental for mechanical and fatigue performance. The reason is that the sharp edges of the LOF pores lead to higher stress concentration and act as crack initiation sites. As a solution to minimize the chance for LOF formation, it is recommended to increase the energy density. However, increasing energy density may also affect the part composition which is not desired, while the remelting process may offer a reliable approach to eliminate LOF defects without increasing the input energy and the chance of keyhole defect formation. The reduction in pores also plays an important role in the density of the parts, which is discussed in the next section.

3.4 Density evaluation

The densities of the as-built and remelted parts with different remelting PPs are provided in Fig. 10. The energy ratio indicates the ratio of remelting energy density to the regular energy density (i.e., as-built). Also, the density change points out the percentage change of the remelted parts’ density with respect to the as-built density. Generally, all remelting processes lead to density improvement, which could be deducted by evaluating the SEM images and the surface roughness of the top surfaces. However, no systematic relationship between energy density of remelting processes and density improvement could be detected. Unremarkable correlation between energy input and SLM parts’ density was also reported for Cu-Al-Ni-Mn shape memory alloys [62]. While Yasa et al. [56] shows a direct relationship between rescanning laser energy input and the density of the parts for Ti-6Al-4V alloy, when it comes to remelting process, the energy input cannot be the only factor and melt pool overlap, laser power, and scanning speed individually also play key roles.

Fig. 10
figure 10

Density measurements of the SLM and remelted parts with different remelting process parameters. Change in density for each sample is also shown on the scan speed versus laser power diagram of the remelted samples

Full size image

There are two possible mechanisms leading to higher density and reducing the micropores/voids after the remelting process: (i) uniform powder distribution in each layer and (ii) porosity reduction. Having a uniform powder distribution in each layer is the main reason to reduce the defects in the SLM parts. The smooth surface in each layer facilitates a proper and uniform powder deposition, resulting in smaller defects and higher density, consequently. As it was shown, the surface roughness is also improved in all remelted samples in comparison with the as-built part. The unstable and concave melt pools for as-built sample are shown in SEM images (Fig. 4), while more stable and smoothed melt pool tracks are achieved in the remelted specimens. The schematic of melt pool morphology is presented for the two cases of as-built and remelting condition in Fig. 11. The concave shape of melt pool in the as-built sample provided a wavy surface which results in high surface roughness and poor powder deposition. The flat melt pools achieved from the remelting process provide a smoother surface and more uniform powder distribution. Figure 12 shows the top surface roughness and density improvement on the same graph. In order to have a better comparison, the samples are sorted from high to low surface roughness. The general trend shows an increase in density, as the surface roughness decreases, which confirms the first mechanism. However, there is a sharp reduction in density (encircled with a dashed line) in which the density stops further increasing. The density reduction happens in sample #4, which has the highest energy ratio (78 %), and it can be explained by the second mechanism. In the second mechanism, due to the lower energy input of the remelting process in comparison with the regular PPs, a shallower melt pool is achieved. Such a shallow melt pool only can remelt a portion of the previously melted track and fill the pores, which are close to the upper surface of the melted track. As it was shown in the optical microcopy images in Fig. 8, the irregular pores are formed where two melt pools intersect, while the entrapped gas (spherical) pores are formed in bottom section of the melt pool because of the high solidification rate. The shallow rescanning melt pools can only remelt the regions in the top portion of the previous layer, and therefore, only the porosities in these regions will be removed. In this case, rescanning mostly reduces the irregular shape pores and the spherical pores are not affected significantly. The second mechanism is also confirmed by the X-ray CT results in which rescanning mostly removed the pores with irregular shapes, while no significant change can be seen in the entrapped gas porosities. The X-ray CT results show that the average size of pores in sample #4 decreased compared to the as-built part. However, there is no significant change in the number of the pores. Although the remelting process removed the irregular pores, the high energy input in sample #4 might result in entrapped gas pore formation.

Fig. 11
figure 11

Schematic presentation of melt pool morphology: (a) regular SLM melt pool tracks, (b) remelting melt pool tracks

Full size image
Fig. 12
figure 12

Top surface roughness as well as the change in density of the samples. For all the remelting samples, surface roughness decreases (improves) and density increases. (Eratio)1 = 0, (Eratio)2 = 31%, (Eratio)3 = 36%, (Eratio)4 = 78%, (Eratio)5 = 21%, (Eratio)6 = 44%, (Eratio)7 = 52%, (Eratio)8 = 50%

Full size image

3.5 Transformation behavior

Transformation temperatures (TTs) are important properties of SMAs that determine the solid-solid phase transformation. DSC analyses were performed to examine the effect of remelting process on TTs of AM NiTi (Fig. 13a). Based on DSC results, all the remelting samples showed relatively similar behavior and transformation temperatures, except in case of sample #4 in which the TTs were slightly deviated toward higher temperatures. There are multiple factors that might have affected the TTs such as composition and precipitation [88,89,90,91]; hence, based on DSC graphs, it could be said that the effect of remelting on the mentioned factors is negligible. In SLM, mainly Ni evaporation that occurs due to its lower boiling temperature was reported as the main mechanism of changing TTs. Further, the power and energy density were reported to have significant contribution in changing the Ni content by controlling the melt pool temperature in SLM of NiTi alloys [92, 93]. Based on the evidences, these two factors for the remelting samples were low enough that could not heat up the previously melted layer above the boiling point and probably no compositional changes were occurred compared to the as-built part. However, for sample #4, which had the highest remelting energy density (almost 80% of the regular energy density), TTs were shifted 15 °C toward higher temperatures that could be attributed to some Ni evaporation in this case. Energy-dispersive X-ray spectroscopy (EDS) was performed to determine the Ni/Ti ratio as a factor that shows the change in TTs. Multiple areas were scanned for composition on the top surfaces of each sample and the average is plotted in Fig. 13b. As it can be seen in this figure, the average Ni/Ti is 0.971 ± 0.002 for all the samples, consistent with the minimal changes of TTs of both SLM and remelted samples. It should be noted that the slight change in #4 TTs could not be captured via EDS since this variation can be within the accuracy limit of EDS [79].

Fig. 13
figure 13

a DSC thermograph of SLM NiTi samples showing stress-free transformation behavior of all the samples. #4 TTs were shifted approx. ΔT = 15 °C. b Austenite peak temperature and Ni/Ti ratio variation. (Eratio)1 = 0, (Eratio)2 = 31%, (Eratio)3 = 36%, (Eratio)4 = 78%, (Eratio)5 = 21%, (Eratio)6 = 44%, (Eratio)7 = 52%, (Eratio)8 = 50%

Full size image

3.6 Mechanical properties

As the first step to investigate the mechanical properties of the as-built and remelted samples, uniaxial monotonic compression tests were performed on the specimens at room temperature. Due to the close TTs for all the samples, the specimens were all in the same phase during the experiments. As it is shown in Fig. 14, the stress-strain plots for remelted and as-built samples show similar critical stress and plateau levels. Based on the results from the mechanical tests, the performed remelting procedures in this study did not significantly affect the compression behavior of the remelted samples and the reason is mainly related to the fact that with the primary melting of the powder, the previous layers were remelted as well. Thus, the primary laser melting of the following layers may cancel the effect of remelting laser on the microstructure due to the fact that the molten pools of the remelting step are expected to be shallower. Therefore, although transformation temperature and compression response of the samples were almost identical, due to the improvements in the relative density and reducing the internal defects, it is expected to see different tensile and fatigue response for the remelted parts [87]. Therefore, as the next step and in a follow-on study, our group will investigate potential improvements of the monotonic and cyclic tensile properties of SLM-fabricated NiTi with the remelting procedure. There is also another point that authors think should be mentioned regarding the difference in the compression behavior of the SLM NiTi parts of this study and previously reported data by our group for the same process parameters [81]. The reason is related to the use of recycled powder, which can affect the properties of the final parts. This is also another ongoing study to establish printing with consistent properties by evaluating the effect of powder recycling on the mechanical properties of the parts. As discussed, the remelting process successfully eliminates the LOF defects, which have a detrimental effect on the fatigue life.

Fig. 14
figure 14

Compression stress versus the strain for the uniaxial compression test of the samples performed at room temperature and up to a constant level of strain. (Eratio)1 = 0, (Eratio)2 = 31%, (Eratio)3 = 36%, (Eratio)4 = 78%, (Eratio)5 = 21%, (Eratio)6 = 44%, (Eratio)7 = 52%, (Eratio)8 = 50%

Full size image

4 Summary and conclusions

SLM has been recognized as a promising potential method for manufacturing functional NiTi components with complex geometries for different applications. However, SLM-fabricated components generally have rough surfaces and may contain microvoids, cracks, and unmelted regions that adversely affect the thermomechanical properties of the parts. So far different modifications to the SLM process as well as post-processing procedures have been investigated to improve these challenges for the NiTi parts. However in this paper, for the first time, the remelting approach—which is an in situ procedure during the SLM fabrication where the laser remelts each layer after the main scan and prior to the powder deposition for the next layer—is investigated as a modification to the SLM to improve the surface roughness and quality of the NiTi fabricated parts. To this end, by altering the laser power and scanning speed, different remelting procedures with a wide range of energy densities were applied to the optimized NiTi SLM process and cubic samples were fabricated. Different characterization methods, including SEM analysis, full surface roughness measurements, micro CT scans, and density measurements as well as DSC and compression tests, were utilized to evaluate the remelted samples. The conclusions and findings of this study can be summarized as below:

  • The results showed that the density and surface roughness of the parts fabricated with remelting procedure could be improved up to ~ 1% and ~ 50%, respectively. The energy density for the remelting process in this study was not high enough to trigger any significant change, such as changes in transformation temperatures.

  • Remelting procedure was able to improve the surface roughness and density and to reduce the number of irregular shape voids.

  • Based on the X-ray CT scans, it could be concluded that the entrapped gas porosities with spherical shapes remain in the parts after the remelting procedure.

  • Although these modifications after the remelting, as expected, did not significantly affect the compression response, they could have a significant impact on the uniaxial tensile and especially tensile fatigue behavior of the SLM-fabricated parts. Consequently, the authors consider in continuing research on the effect of remelting on the tensile and fatigue analysis as the next step on the AM NiTi parts.