Influence of laser printing mode on thermal behaviors, forming characteristics, and microstructure evolutions of additive manufactured Ti6Al4V alloy

Ti6Al4V parts fabricated by laser powder bed fusion (LPBF) technology have been widely used in such fields as aerospace, automotive and medical implants. This study investigates the effect of laser scanning modes on thermal behaviors, forming characteristics, and microstructural evolutions of LPBF-fabricated Ti6Al4V parts. The numerical simulations on the temperature field provide a theoretical explanation of various surface morphologies, surface roughness values and relative densities of samples. The rotation between adjacent layers diminishes the large directional thermal stress generated by X scanning or Y scanning, the Ti6Al4V samples using XY scanning and Island scanning present smooth surface and higher relative density (> 99.0%). The observed staggered martensite within the columnar β grains is due to the epitaxial solidification across the deposited layers with 90° or 37° rotation. The martensite growth of LPBF-processed Ti6Al4V components using X scanning has a similar inclination along the building direction and presents anisotropic characteristics. These findings provide new inspirations for achieving high-performance titanium alloy components with specific microstructure by LPBF technique using a proper laser scanning mode.


Introduction
As an advanced powder-bed manufacturing method of additive manufacturing technologies, laser powder bed fusion (LPBF) can fabricate complex-shaped metallic components with metallic powders [1].Among the metallic powders, Ti6Al4V alloy is the most popular titanium alloy.It has been widely used in different industries such as aerospace, automotive and biomedical industries because of its high special strength, superior corrosion resistance and excellent biocompatibility [2,3].Although LPBF technology can reduce the production cost of titanium alloy parts due to the high utilization of

Powder materials and LPBF process
All the samples tested in this work were fabricated by a LPBF equipment consisting of a highly stable IPG YLR-500 ytterbium fibre laser with a spot size of 70 μm.As for the starting material, gasatomized Ti6Al4V powder with a maximum and mean diameter of 70.93 μm and 26.54 μm was smoothly spread on a 150 mm × 150 mm × 40 mm Ti6Al4V substrate (Fig. 1).Based on our previous study [5], the main processing parameters, including laser power = 250 W, scanning speed = 800 mm/s, layer thickness = 50 μm, and hatch spacing = 50 μm were employed for the LPBF of Ti6Al4V powder.As shown in Fig. 1b and 1c, all the rectangular samples were fabricated by four different laser scanning modes: X scanning mode (bidirectional scanning along the short edge, i.e., X-axis), Y scanning mode (bidirectional scanning along the long line, i.e., Y-axis), XY scanning mode (bidirectional scanning with 90° rotation and island scanning mode (island scanning with an increment of 37° rotation) in an argon atmosphere.

Materials characterization
The surface roughness of the as-built Ti6Al4V samples was measured by Surftest SJ-210 tester (Mitutoyo, Japan) according to the international standard ISO 4287:1997.Then, these samples were ground by waterproof abrasive papers with various degrees of roughness and polished by colloidal silica suspension.Following the corrosion of the surface by the solution composing of HF, HNO3 and distilled water with a volume ratio of 1:3:50 for 40 s, the horizontal and vertical planes of the samples were characterized by PMG3 optical microscope (OM, Olympus Corporation, Japan).The upper surfaces of samples were imaged by an FEI Quanta 200 scanning electron microscope (SEM).The grain boundary, crystallographic orientations, and phase composition were analyzed by electron backscattered diffraction (EBSD).

Numerical simulations
Finite element analysis (FEA) simulations were performed to study the distinct thermal behaviors of the LPBF-processed Ti6Al4V alloy by ANSYS.While actual LPBF-processed samples are expected to comprise 60 layers, it is hard to perform simulation considering the large amounts of computation and high computational cost.Thus, a smaller size model has been built in this study.Fig. 2 shows the geometry of the model, which has a solid substrate with a dimension of 2.10 mm × 0.84 mm × 0.20 mm and two Ti6Al4V powder layers with each layer thickness of 0.05 mm.For the improvement of the simulation precision and efficiency, the Ti6Al4V substrate was coarsely meshed using the coarsen tetrahedron elements and the powder bed was meshed by the Solid70 hexahedron elements with the dimensions of 17.5 μm × 17.5μm × 25.0 μm.The laser heat source with a spot size of 70 μm possesses a character of Gaussian distribution.In this simulation, the moving laser spot has specific scanning paths based on the four experimental scanning strategies in Fig. 1b.It should be pointed out that the XY scanning was carried out by scanning along the X axis in the first layer and the Y axis in the second layer.Laser power of 250 W and scanning speed of 800 mm/s were set for LPBF process.The details of thermal physical parameters of Ti6Al4V, governing equations and other basic setups can be found in our previous research [8].
Fig. 2 The three-dimensional finite element model and simulation schematic of LPBF process.

Thermal behaviors
The effects of four common laser printing modes on temperature field were simulated by the finite element model as shown in Fig. 3.When the laser spot moved along the X-axis on the second layer, it can be observed that the colour variance in the Y-Z plane is faster than that in the X-Z plane (Fig. 3a).
In other words, the local temperature difference paralleling to the direction of laser scanning is higher than that perpendicular to the direction of laser scanning.Further comparing the maximum temperature in the centre point of Ti6Al4V layer as depicted in Fig. 3b, it can be found that the maximum temperature of X scanning is lower than that of Y scanning.This is due to a stronger heat accumulation caused by the longer scanning time scales of a longer scan vector compared with X scanning.However, as for the scale of small islands, shorter scan vectors in a single island obtained the highest temperature at centre point A attributed to the strong residual heat effect in the situation of quick-return laser.This is in agreement with Masoomi's study that the shortest scanning length/time in an island gives rise to most residual heat and thus the highest temperature [9].Therefore, the residual heat effect has a great influence on the maximum temperature of melted Ti6Al4V alloy.Wunderlich et al. [10] demonstrated that the relationship between surface tension and temperature of melted Ti6Al4V meets a negative correlation according to the equation of surface tension: σ (T) = 1.52-5.5×10 - (T-1655) (N/m).It can be deduced that the scanning tracks of X scanning with the lowest top temperature will obtain maximum surface tension (1.278 N/m).Fig. 3c depicts the largest temperature gradient of X scanning combining with the largest surface tension may have a great impact on the surface morphology of LPBF-processed Ti6Al4V alloy.The thermal gradient (< 5 × 10 7 ℃/μm) of Y scanning and XY scanning both are relatively lower than other two strategies.Fig. 3d depicts the temperature variations of different laser scanning modes during LPBF process.It can be observed that all the temperatures of the initial point on the top center of the molten pool rapidly decrease, and uprush again when the laser moves back to the sintered Ti6Al4V material.Temperature-time curves of Y scanning and XY scanning show almost a similar tendency in accord with the results in Fig. 3b and 3c, a tiny curve deviation is derived from the different scanning length/time in the first Ti6Al4V layer.By comparing the second crests affected by the second scan vector, it can be seen that only the second peak value of island scanning is over the melting point of Ti6Al4V materials due to the ultra-short scanning length/time.Thus, island scanning is good for reheating samples immediately and prolongs liquid lifetime.Generally, this diverse temperature history during LPBF has a significant effect on the microstructure of Ti6Al4V alloy [11].

Forming characteristics
Fig. 4a illustrates that all the surfaces of four different scanning modes appear slightly adhered powder.
Owing to the temperature gradient of above 5.4×10 7 ℃/μm and accompanying high surface tension for X scanning and Island scanning in Fig. 3c, the slight balling with small spheres about 140 μm is observed in red circles.To better analyze the surface characteristics of as-built samples with different scanning modes, surface profile in Fig. 4b and roughness data in Table 1 are used to further discuss the influence of scanning modes.It can be seen that the surface profile experienced significant fluctuations for X scanning and the Rz value of 112.690 is larger than others.This is because the Ti6Al4V powder absorbed laser energy and got the lowest temperature of 2150 ℃ and highest temperature gradient of 6.4×10 7 ℃/μm (Fig. 3b and 3c), which brings about the highest surface tension under X scanning.Therefore, as shown in Fig. 4a, a rippling effect occurs and shows an undulating surface with the largest height difference between peak and valley because of largest surface tension yielding a shear force on the molten pool surface.This force primarily stems from a surficial temperature difference between the moving laser beam and the solidifying region.Unlike the small overlapping zone between the scan tracks of X scanning, Y scanning induced closer cylindrical tracks with too large overlapping zones.This is owing to the higher temperature of melted material makes a wider molten pool, so the values of Ra and Rq have a certain degree of decreasing.However, the excessive overlapping gives rise to uneven surfaces consisting of open pores and many micro-humps.This instability surface will be detrimental for connecting adjacent layers and final mechanical property of LPBF-fabricated Ti6Al4V samples.Besides, the rotation between adjacent layers is good for diminishing the large directional thermal stress generated by the X scanning or Y scanning with no rotation [12].The measurement of the surface roughness along the white lines confirms this potential reason that thermal stress decreasing is helpful in improving the surface finish.The specific performance is that Ra values (the blue straight lines in Fig. 4b) of XY scanning and Island scanning are lower than the other two.By contrasting the Ra, Rq and Rz of XY and Island scanning modes, much finer surface finish of XY scanning can be elucidated.Ali's [7] research have demonstrated that XY scanning strategy results in the lowest residual stress for LPBF-processed Ti6Al4V samples compared with other strategies including Island scanning.Given the minimum residual heat stress, it is thus plausible that the sample under XY scanning is easy to achieve a flat surface.To further analyze the effect of laser scanning mode on densification behaviors, relative density and corresponding OM images are applied to elucidate the role of scanning mode.As shown in Fig. 5, it can be observed that the relative density increases along with the decreasing size of pores and finally shows a value of 99.8% for Island scanning samples.This is because the laser beam moved forth and back very quickly and yielded a strong residual heat effect, highest maximum temperature and thus a prolonged liquid time during laser melting of Ti6Al4V (Fig. 3).The mean relative densities of Island scanning and XY scanning both reaching up to 99% is due to the rotation between layers leading to diminished residual stress and smooth surface.Thus, they present only microvoids or even fully dense image.In addition, the decreasing relative density of Ti6Al4V samples fabricated by LPBF using X scanning and Y scanning can be attributed to the fact that directional thermal stresses are more prominent compared with these thermal stresses of other two strategies.This is because 90° or 37° rotation between LPBF layers significantly reduces the stress directivity.Therefore, the samples without layer rotation present poor surface finish and low density.Noticeably, the samples under Y scanning present the minimum density among the four common strategies.This lowest relative density with a slightly bigger error margin suggests the poor interlayer connection leading to the large and irregular void in the size of about 100 μm, which is attributed to surface micro-humps and open pores as shown in Fig. 4.These defects of large voids inside the as-built Ti6Al4V components might be the source of cracking and be detrimental to the mechanical property for the samples using Y scanning.

Microstructure evolutions
Microstructure analysis on LPBF-fabricated Ti6Al4V samples with different scanning modes using optical microscopy is presented in Fig. 6.It depicts that the grain boundaries of prior β phase are vertical in the frontal and lateral plane, while they shape irregular block regions in the upper plane.
The columnar grains of prior β phase in the X-Z plane and Y-Z plane were elongated parallel to the building direction because of the epitaxial growth and large temperature gradient along the building orientation, which is in good agreement with previous studies [4].As indicated in Fig. 6a, the width of columnar grain in frontal plane perpendicular to the scanning direction is larger than that in lateral plane paralleling to the scanning vector.In contrast, the corresponding coarsening microstructure in X-Z plane of samples under Y scanning is due to the laser scanning along the Y axis (Fig. 6b).As discussed in the literature [13], cooling rates higher than 410 ℃/s contribute to typical martensitic transformation for Ti6Al4V.As depicted in Fig. 3d, it is believed that martensitic microstructure within the parent β phase can grow along the building direction because the high cooling rate of LPBF process is much higher than that required for martensitic transformation in Ti6Al4V alloy.Therefore, α' martensite can be observed in all samples irrespective of scanning mode.It is worth noting that the acicular α' martensite didn't precipitate along the boundaries of β grain but at a certain inclined angle in the parent β grain.Unlike the X scanning and Y scanning, two black ovals in Fig. 6c and 6d present the staggered martensite influenced by the rotations between deposited layers.As discussed above, the EBSD analysis on the Y-Z plane of the specimens by X scanning and Island scanning also indicated that the acicular martensite was formed within the prior β grain boundaries defined by the black lines, i.e., high-angle grain boundaries with misorientations of 15°-45° are exactly the boundaries of parent β grains (Fig. 7).This corresponds with reported research that reconstructed β grains in LPBF-processed Ti6Al4V parts have high angle misorientation of grain boundaries [14].Furthermore, the presence of these columnar β grains is because Ti6Al4V grew epitaxially across the deposition layers when the heat conducted away vertically, which is in agreement with Roberts's finite element analysis [11] and consistent with the vertical β grains paralleling the building direction as shown in Fig. 6.Concerning acicular martensite growth, green colour around white martensite indicates the low-angle grain boundaries of martensite as depicts in misorientation map.Comparing the orientation map of these two samples, it could be seen that martensite growth in samples using X scanning has a similar inclination along the building direction and presents anisotropic characteristics (Fig. 7a).This inclination martensite texture is mainly characterized by the same colour (i.e.crystal orientation) within each β grains.Daymond et al. [15] pointed out that the similar orientation of martensite along the Z direction is owing to the grains inherited from columnar β grains during martensitic transformation.But differing from the α′ textures with similar inclination to the growth direction dominated in each parent β grains, Fig. 7b depicts that the random growth of martensite microstructure attributed to the 37° rotation of direction imposed by the Island scanning, which changed the heat flux and thermal distribution of molten pool for each layer.No packets of martensite share the same crystallographic orientation according to the various colours in each β grain.It can be deduced that the rotation angle of scanning strategy has a strong influence on the solidification of Ti6Al4V and final microstructure [16].0° rotation of X scanning and 37° rotation of Island scanning result in the inclination and isotropy of the martensite respectively.Thus, laser printing mode can regulate desired crystallographic textures of as-built components characterized by isotropic or anisotropic microstructures.Fig. 7 (a) Orientation map and grain boundary misorientation map of the YZ-plane in an LPBFprocessed specimen with X scanning, (b) orientation map and grain boundary misorientation map of the YZ-plane in a LPBF-processed specimen with island scanning.Scale bars = 100 μm.For a better understanding of microstructure evolutions of LPBF-processed Ti6Al4V specimens, high-resolution EBSD was adopted to study in detail the phase composition and grain size of samples using X scanning and Island scanning.Fig. 8 shows the phase fraction and lath width among the martensite of two samples.It can be observed that the large amount of hexagonal closed-packed (hcp) phase (α', α) and small amount of body-centred cubic (bcc) phase (β) are respectively illustrated in red and yellow colour.Generally, α' martensite usually exhibits an aspect ratio of 10:1 or greater according to the description of martensite microstructures of titanium alloy in Aerospace Standard (SAE AS1814-2016).But many laths in Fig. 8a and 8b present an aspect ratio around 5:1, which indicates the conceivable α phase in red regions.Moreover, phase transformation and FEA simulations are combined to confirm the phase composition.As an allotropic alloy, Ti6Al4V might be transformed fully into the β phase field above the β transus temperature (~ 950°C) during laser heating and then precipitated into α' martensite below this critical temperature [14].The track-by-track and layer-bylayer LPBF process of Ti6Al4V powder involves remelting of adjacent tracks and the deposited layers, just like the heat treatment transforms the initially formed α' martensite into α+β phase [17].Finally, the β phase is mainly detected between the boundary of lamellar structure, which is consistent with the previous research [18].Phase fraction of β phase indexed in the pattern is about 2.13% for X scanning and 6.46% for Island scanning.According to the numerical simulations in Fig. 3d, a constant high temperature above β transus was kept during the initial 20 ms period.Thus, the Ti6Al4V samples under Island scanning experienced a stronger heat input, demonstrating that the reduced scan vector length in every individual island can induce higher maximum temperature and lower cooling rate.On the one hand, the different cooling rates affect the increased intrinsic heat treatment of Ti6Al4V alloy with longer heating periods.Intensified remelting of the overlapping portion of tracks promoted the energy transferred to underlying layers and brought about the as-built samples with no pores, rather than some white spots (i.e.voids) in Fig. 8a.On the other hand, the cooling rate has a great effect on the amount of β phase and grain size of α laths during solidification [17,19,20].It is believed that when using the Island scanning, the average temperature and liquid lifetime of Ti6Al4V molten pool melted with decreased cooling rate are larger than that of other three scanning modes.This improves the diffusional α'→ α+β phase transformation forming a coexistence with α and β phase aside acicular martensite abbreviated as α'+(α+β), and also elevates the size of α laths and the amount of β phase more than two times.For this reason, the samples present an increasing width of α laths from 0.41 μm to 1.06 μm (Fig. 8c).For the sample fabricated using Island scanning, the coarser microstructure at lower cooling rate experienced a severe competition of grain nucleation and grain growth.In contrast, the refined grain under higher cooling rate provoked more nucleation sites for samples fabricated by X scanning.According to the Hall-Petch equation [21]: ΔσHall-Petch = k/d1/2, where k is a constant about material and d is the grain size.The finer grains inside LPBF-processed Ti6Al4V samples produce more grain boundaries impeding dislocation movement and contribute to Hall-Petch strengthening of as-built Ti6Al4V components.Thus, in addition to the martensite deposition increasing the ductility and strength of the Ti6Al4V alloy [22], grain refinement is also good for the strengthening of LPBFprocessed Ti6Al4V specimens under increasing cooling rate.When LPBF-processed Ti6Al4V specimens are applied in biomedical field [23], results indicate the potential application of These LPBF-processed Ti6Al4V specimens with good mechanical performance could be used as implant at human body temperature.

Conclusion
This work aims to investigate the influence of laser scanning mode on LPBF-fabricated Ti6Al4V alloy, including thermal behaviors, forming characteristics, and microstructure evolution.FEA simulations were carried out to study the temperature evolution during LPBF of Ti6Al4V material and provide a theoretical explanation for the experimental results.The samples using X scanning obtained the lowest temperature, highest temperature gradient (6.4×107℃/μm) and largest surface tension (1.278 N/m), resulting in a poor surface with a rippling effect and balling effect, and thus a low density.The rotation between adjacent layers diminishes the large directional thermal stress generated by X scanning or Y scanning, the Ti6Al4V samples using XY scanning and Island scanning present smooth surface and higher relative density (> 99.0%).During layer-by-layer deposition process, the rotation along the building direction brings about the staggered martensite within the prior columnar β grains because of the epitaxial growth behavior.The martensite growth of LPBF-processed Ti6Al4V components using X scanning has a similar inclination along the building direction and presents anisotropic characteristics.In addition to the fine-grain strengthening by martensite decomposition, it is plausible that α'+(α+β) microstructure can achieve a good match of strength, ductile and toughness by LPBF under a special printing mode.

Fig. 1 (
Fig. 1 (a) LPBF process of Ti6Al4V powder by LPBF-150 equipment, (b) schematic diagrams of four different scanning strategies, (c) the SEM images of surface morphologies of as-built LPBF Ti6Al4V samples.The scale bars in (a) and (c) are 100 μm and 2 mm, respectively.

Fig. 3 (
Fig. 3 (a) The temperature counters of samples under the X scanning, (b) the temperature distribution and (c) temperature gradient of samples under four scanning modes along the X-axis near the centre point A; (d) temperature-time curves of samples under four scanning modes.

Fig. 4 (
Fig. 4 (a) Surface morphologies of as-built Ti6Al4V alloy with respect to four scanning modes: X scanning, Y scanning, XY scanning, and Island scanning from left to right, (b) surface roughness tests along the lines in (a).The scale bars in (a) are 1 mm.

Fig. 5
Fig. 5 Relative densities of four different scanning modes of LPBF-processed Ti6Al4V samples with laser power of 250 W, scan speed of 800 mm/s.The scale bars inset are 100 μm.

Fig. 6
Fig. 6 Three-dimensional OM images of LPBF-processed Ti6Al4V samples with different scanning modes: (a) X scanning, (b) Y scanning, (c) XY scanning, and (d) Island scanning.Scale bars = 100 μm.As discussed above, the EBSD analysis on the Y-Z plane of the specimens by X scanning and Island scanning also indicated that the acicular martensite was formed within the prior β grain boundaries defined by the black lines, i.e., high-angle grain boundaries with misorientations of 15°-45° are exactly the boundaries of parent β grains (Fig.7).This corresponds with reported research that reconstructed β grains in LPBF-processed Ti6Al4V parts have high angle misorientation of grain boundaries[14].Furthermore, the presence of these columnar β grains is because Ti6Al4V grew epitaxially across the deposition layers when the heat conducted away vertically, which is in agreement with Roberts's finite element analysis[11] and consistent with the vertical β grains paralleling the building direction as shown in Fig.6.Concerning acicular martensite growth, green colour around white martensite indicates the low-angle grain boundaries of martensite as depicts in misorientation map.Comparing the orientation map of these two samples, it could be seen that martensite growth in samples using X scanning has a similar inclination along the building direction and presents anisotropic characteristics (Fig.7a).This inclination martensite texture is mainly characterized by the same colour (i.e.crystal orientation) within each β grains.Daymond et al.[15] pointed out that the similar orientation of martensite along the Z direction is owing to the grains inherited from columnar β grains during martensitic transformation.But differing from the α′ textures with similar inclination to the growth direction dominated in each parent β grains, Fig.7bdepicts that the random growth of martensite microstructure attributed to the 37° rotation of direction imposed by the Island scanning, which changed the heat flux and thermal distribution of molten pool for each layer.No packets of martensite share the same crystallographic orientation according to the various colours in each β grain.It can be deduced that the rotation angle of scanning strategy has a strong influence on the solidification of Ti6Al4V and final microstructure[16].0° rotation of X scanning and 37° rotation of Island scanning result in the inclination and isotropy of the martensite respectively.Thus, laser printing mode can regulate desired crystallographic textures of as-built components characterized by isotropic or anisotropic microstructures.

Fig. 8
Fig. 8 EBSD analysis for the SLM-processed specimens: phase maps confirming uniform distribution of α phase (red) and β phase (yellow) in as-fabricated specimens with X scanning (a), and Island scanning (b), the α lath width and β phase fraction in fabricated specimens under two different scanning strategies (c).Scale bars = 5 μm.

Table 1 Ti6Al4V
surface roughness with respect to scanning modes.Arithmetical mean deviation (Ra), root mean square deviation (Rq) and mean difference between peak and valley (Rz) of the profile in Fig.