Figure 2 shows the morphologies of SMT under different LEDs. As shown in Figure 2A, when LED was less than 100 J/mm³, and the SMT appeared discontinuous because the powder was not sufficiently melted and the liquid phase did not flow sufficiently. Therefore, the concave-convex phenomenon occurred at the edges of the track. When LED was between 100 J/mm³ and 200 J/mm³, the SMT was continuous without obvious water ridges on the surface as shown in Figure 2B. A common feature shared by these two zones was that a large number of pore defects were observed on the surface. This could be related to the gas between powder, which got enclosed by the liquid metal as the laser beam passed by, then escaped from the surface of liquid metal, and finally left pores behind during the solidification of metal since the liquid metal acquired high viscosity when LED was low. When LED was between 200 J/mm³ and 500 J/mm³, the SMT was continuous with evenly water ridges, good contour straightness, smooth edge, little adhesive powder, and high densification, as shown in Figure 2C. This was regarded as the best processing window for 316 L SLM- process. It is worth mentioning that there were almost no pore defects on the track surface, which was related to the high temperature and low viscosity of liquid metal. The pores formed during the gas escapement from the liquid metal can be recovered before the solidification of the liquid metal. When LED was greater than 500 J/mm³, the surface of molten track still showed uniform water ridges, but a lot of splash droplets can be seen around it as shown in Figure 2D. The laser beam with high LED heated the metal to a temperature higher than the boiling point, which led to splash of liquid metal and finally spherical particles around tracks formed. These splashed droplets had high possibility to weld on the surface, which was a drawback for the forming accuracy not only because of the increased surface roughness, but also because of the possible damage to the scraper which led to uneven distribution of raw powder.

The SEM images in Figures 3A–F show that the depth of the molten pool increased sharply and the keyhole phenomenon occurred with LEDs greater than 500 J/mm³. This was mainly due to the high liquid temperature caused by the evaporation phenomenon, which increased the laser absorption in the central region and allowed the heat to transmit further. As shown in Figures 3B, D, hollow pores can be seen in the profile of the molten pool when the LEDs increased to 917 J/mm³ and 1090 J/mm³. This is definitely unacceptable for additive manufacturing, since the appearance of the hollow pores will result in low density and consequently poor mechanical properties. When LEDs were between 100 J/mm³ and 500 J/mm³, there was no keyhole in the molten pool, and the wetting angle of the molten pool was less than 45° as shown in Figures 3G–L, which was conducive to the metallurgical lapping between the tracks. When LED was less than 100 J/mm³, the wetting angle of the molten pool was more than 70°, and there was a tendency for balling as shown in Figures 3K, L. This molten pool was not conducive to the metallurgical lapping between the tracks. It was found that an optimized range of LEDs, e.g., between 200 and 500 J/mm³, was the best for the SLM process of 316 L, which led to the beautiful appearance of single molten track and appropriate molten pool structure based on the result of Figures 2, 3.

In order to better understand the phenomena observed in above study and to explore the mechanism, a simulation for the SLM processing was done with different LEDs. As shown in Figure 4A, at a relatively low value of LED, e.g., 92 J/mm³, gasses were traced between the powder and substrate in the liquid metal, which then turned into holes after scanning of laser beam, and the surface of the track was uneven and concave-convex. This was mainly because the insufficient LED resulted in relatively low temperature, high viscosity, and low flow velocity of liquid metal, resulting in the survival of bubbles and poor surface morphology. In some cases, the LED was so small that it was insufficient to penetrate the substrate, and the substrate was even not melted, explaining the concave-convex surface at lower LED. When the LED was set at 229 J/mm³ as shown in Figure 4B, no pores appeared in the molten pool and the surface of the molten track was smooth. This was due to the sufficient LED, which could penetrate into the powder bed. Therefore, a considerable liquid with a larger flow velocity could be formed, which easily migrated towards the edge of the pool and subsequently solidified with a smooth surface. When the LED was 611 J/mm³, the peak temperature of the liquid phase exceeded the vaporization temperature of 316 L. The metal gas increased the absorption of energy, creating a high-temperature-high-pressure region localized under the focus of laser beam, which led to the blowing of both powder and liquid metal, and finally the keyhole formed under the laser beam. Based on the above experimental and simulation results, the following conclusion can be drawn: During the SLM process, a reasonable LED should be selected to avoid the phenomenon of pore defects in the molten pool caused by too low energy and the splashing of the droplets caused by excessive energy.

Effects of LED on the widths and depths of SMT was shown in Figure 5. The Figure 5A shows how the width and depth of SMT are defined. The width of SMT increased from 60 μm to 200 μm with the increase of LED from 66 J/mm³ to 500 J/mm³, and then tended to stabilize at 200 μm even when LED was higher than 500 J/mm³. However, the depth of SMT increased almost linearly from 20 μm to 450 μm when LED was increased from 66 J/mm³ to 1019 J/mm³ without a trend of saturation for LED used in this study. In fact, the width and depth of the melt pool depend on both the laser power and the scanning speed. In the SLM-process, the peak temperature of the material heated by the Gaussian laser heat source is related to the laser power and the scanning speed, as higher laser power or lower scanning speed can increase the peak temperature, and higher peak temperature can lead to a deeper molten pool. Meanwhile, the width of the melt pool is mainly influenced by the scanning speed. It is mainly related to the dwell time of the laser at a fixed point, which increases as the scanning speed decreases; melting the irradiated powder should require a sufficient dwell time, which eventually leads to a wider molten pool. There has a correspondence between laser energy density and scanning speed and laser power: L E D = 4 P / p i * V * D 2 (1a) where P (W) is the laser power, V (mm/s) is the scanning speed, and D (mm) is the laser beam diameter. Therefore, with the increase of LED, the melt pool will be deeper and wider. K is the ratio of depth to width. The K value for each sample is also listed in Figure 5B. As mentioned above, K value increases with increasing LED. From the above studies, the liquid phase involved in the metallurgy of the SMT consisted of two parts, the first part was the solidified body of the previous layer or layers, and the other part was the powder particle of the current layer captured by the laser beam. Since the width of SMT tends to be stable, it is reasonable to assume that the volume of powder captured by the laser beam could have an extreme value for a certain beam diameter. The results can be used to design the hatch space.

Figure 6A shows a typical microstructure of 316 L molten pool at a laser power of 100 W and scanning speed of 100 mm/s. It is mainly composed of equiaxed cellulars and columnar dendrites. The magnified microstructure of the different regions of the molten pool boundary is shown in Figures 6B–E, respectively, which were different from one another. However, they all contained columnar dendrites, which were perpendicular to the molten pool boundary. In the central region of the sample, the growth direction of the columnar grains was close to the deposition direction, as shown in Figure 6E. On the other hand, at the edge of the sample, the growth direction of the columnar grains slightly deviated from the deposition direction since the temperature gradient was perpendicular to the molten pool boundary as Figures 6C, D shows. Figure 6B shows that the cellulars were mostly pentagonal in shape and the grain size was small. There were cellulars at the top, center and boundary of the molten pool, which were distributed randomly (Zhang et al., 2018). However, most of the cellulars were located at the top and in the center of the molten pool rather than the boundary.

The microstructure was ascribed to the mechanism of heat transfer during SLM. For SMT printing, heat was first transferred across the boundary of the molten pool to the substrate or solidified part, and then dissipated from the substrate, contributing to the direction of crystallization perpendicular to the boundary of the molten pool. This was similar to the solidification process during casting (Zhao et al., 2019). The nucleation near the boundary of the molten pool was inhomogeneous as the solidified part provided the substrate for the epitaxial growth of the subsequent melt, which was the main axis for faster growth. Therefore, grains with crystal orientation perpendicular to the boundary of the molten pool had priority for growth, whereas growth in other directions was inhibited by heat transfer. Parts farther from the boundary of the molten pool were composed of cellular grains because the temperature gradient decreased gradually and so did the solidification rate, leading to the transformation of columnar grains into equiaxed grains and the formation of cellular grains (Ares et al., 2002). This was similar to the transformation of columnar grains into equiaxed grains during casting (Burden and Hunt, 1975). In addition, inside the molten pool, the fast cooling of the melting led to the tension gradient of the surface, which was larger than the viscous force. As a result, the melt became convective and another capillary force known as Marangoni convection, which led to the columnar grains from the boundary to the interior of the molten pool broke into many fine grains near the center of the molten pool (Smith, 1966). Therefore, it indicated that the Marangoni convection contributed to the formation of cellular substructures inside the molten pool.

Figures 7A–C shows the microstructure of the whole molten pool under different LEDs. It showed that under high LED, the columnar and cellular grains appeared alternately, whereas with the decrease of LED, the cellular grains decreased and the columnar grains almost occupied the whole molten pool. Moreover, with the decrease of LED, the columnar grains appeared much thicker. When LED was low, the columnar grains could even grow across the entire molten pool, from the boundary to the top. In addition, as LED decreased, the number of cellulars became smaller and even disappeared, while the length and width of the column crystals gradually increased. This phenomenon may be due to the fact that when LED was low, the temperature gradient decreased, weakening the Marangoni convection phenomenon and allowing the columnar crystals to grow continuously. On the above basis, the size variation of the cellular grains was further studied. Figure 8 shows the cellular grains information at a magnification of 1500 times. It can be seen that with the decrease of LED, the size of cellular tends to decrease, and its shape changes from a pentagon to an ellipse. In Figure 9A, the cellular grains from Figure 8C were extracted using Image-Pro Plus and the average size of all grains was calculated. Figure 9B shows the size distribution of cellular grains ranging from 0.43 to 1.70 μm under different LEDs. The advantage of this study is that it provides a data basis for in situ microstructure regulation of SLM.