Transition mechanism of melt depth during Laser Powder Bed Fusion using in-situ X-ray and thermal imaging


 This research clarified the transition mechanism of melt depth in Ti powder bed during Laser Powder Bed Fusion process using in-situ X-ray and thermal imaging. A fiber laser beam of 150 W was irradiated on a powder bed at a scan speed of 15 mm/s for 3.5 s in a vacuum chamber. The obtained X-ray images showed a keyhole depth Ld increased immediately after laser irradiation, gradually decreased, and became constant. It also showed a keyhole width Lw increased immediately after laser irradiation and decreased afterward, after that, Lw increased again, and became constant. Furthermore, thermal images that measured the temperature on the powder bed showed the high temperature width Lh gradually increased and become constant. The model of the driving force which pushed the molten droplet was examined by analyzing the volume and scattering speed of the molten droplet. The model indicated the recoil pressure caused by the vaporization of powder metal was a driving force for the molten droplet scattering. The transition mechanism of keyhole depth was considered as follows. The increase of Ld at the beginning is due to the increase of the recoil pressure PT. This is because the decrease of Lw and large quantity of vaporization. Next, the decrease of Ld is due to the decrease in PT. This is because the increase of Lw and decrease of quantity of vaporization. At last, the transition to the constant Ld is caused by stabilization of Lw and Lh followed by stabilization of PT.

decreased, and became constant. It also showed a keyhole width L w increased immediately after laser irradiation and decreased afterward, after that, L w increased again, and became constant. Furthermore, thermal images that measured the temperature on the powder bed showed the high temperature width L h gradually increased and become constant. The model of the driving force which pushed the molten droplet was examined by analyzing the volume and scattering speed of the molten droplet. The model indicated the recoil pressure caused by the vaporization of powder metal was a driving force for the molten droplet scattering. The transition mechanism of keyhole depth was considered as follows. The increase of L d at the beginning is due to the increase of the recoil pressure P T . This is because the decrease of L w and large quantity of vaporization. Next, the decrease of L d is due to the decrease in P T . This is because the increase of L w and decrease of quantity of vaporization. At last, the transition to the constant L d is caused by stabilization of L w and L h followed by stabilization of P T .

Introduction
Additive manufacturing (AM) is a processing technology that is available for manufacturing a hollow and complicated shape, so the further spread of AM is expected in the fields of aerospace [1], automobile [2], and medical [3]. Laser Powder Bed Fusion (LPBF) process, which is a kind of AM, is that a laser scans two-dimensionally and melts a thin layer of powder metal, which is repeated until a 3D shape is fabricated. Now, the rough surface in the product made by the LPBF process affects the drop of the mechanical properties, and additional machining processes to make the surface smooth would be required [4,5]. Previous studies focused on the morphology of the single-track melt pool to clarify the cause of rough surface. Melt depth change affects surface roughness and the bonding between layers. A method to reduce internal defects and improve the surface property is to change the combination of processing parameters. However, it is difficult to find optimal processing condition from a combination of parameters in each material.
Melting behavior influences the relative density and surface property of the product. Therefore, Laser energy density (LED) is widely used for an index to find the most suitable processing condition. LED is expressed as the value that divides laser power by scanning speed, the thickness of the powder bed, and laser diameter [6]. Melting behavior changes by LED. For example, Prashanth et al. revealed that the relative density became 99% at high LED and tensile strength became large with high scan speed and large laser power at constant LED [7]. However, Promoppatum et al. revealed that melting behavior was unstable when LED was superabundant. Although the relationship between LED and the final product has already been revealed, melting behavior between them is unexplained [8]. Therefore, in-situ observation is necessary to figure out the melting behavior in detail.
Previous study revealed that the characteristic melting behavior was observed using either X-ray or thermal imaging: balling [9], spatter [10][11][12], melt pool shape [13,14 ], wettability [9], pore formation [13,15], solidification [10], temperature distribution [16][17][18], keyhole [19]. Based on these, it is expected that the melting behavior in single-track examination can be clarified in detail by combining in-situ simultaneous observation of melting powder by X-ray imaging and temperature distribution measurement by thermal imaging. Then, our group established an analytical method using X-ray and thermal imaging simultaneously for the first time [20]. Wakai et al. classified the variations of the melting behavior with time at one-point irradiation using X-ray and thermal imaging. As a result, it became clear that a keyhole was formed just after irradiation start and the expansion of the keyhole depth is settled by the generation of the molten metal, and then molten metal rose by evaporation of powder metal.
Chimura et al. observed keyhole shape by in-situ X-ray imaging during laser welding [21]. Laser welding has a similar process to LPBF in terms of that a heat source is a laser and material is melted and solidified by the scanning of the laser continually. Chimura et al. revealed that keyhole depth increased as laser irradiation advanced and became constant before long. In welding, processing quality is known to be worse due to the molten droplet being scattered by the recoil pressure that occurred by evaporation of the metal, and it remains as a defect. The recoil pressure is the power that acts on the keyhole surface in opposite direction with the evaporation direction as a difference of the momentum when metal evaporated. Chimura et al. clarified that a keyhole was deepened with increasing recoil pressure.
About the two-dimensional melting of powder metal in LPBF, it was reported that the melt depth that is perpendicular to the scanning direction decreased [17]. It was supposed that this phenomenon was caused by the transition of the melting behavior from the melting form of local heating to the stable melting form. From this, about single-track analysis, it is supposed that the melt depth that is parallel to the scanning direction decreases based on a similar mechanism. However, the transition of melt depth itself has still been unexplained.
Recoil pressure was found to be proportional to the quantity of vaporization and inversely proportional to a vaporization area in a previous study on laser welding [21]. There is a difference between welding and LPBF, that is bulk materials are used in welding while powder metal is used in LPBF. In LPBF, the apparent density of powder bed is almost 60%. Therefore, it is supposed that the keyhole shape which is relevant to a vaporization area greatly changes when solidified bulk is formed and a void between the particles is integrated with a keyhole. Also, it is thought that as laser irradiation advances, powder metal around the irradiation point is heated and a sintered material is formed. After that, the quantity of vaporization decreases as thermal diffusion promotes and recoil pressure decreases.
Then, we conceived to acquire temperature distribution around the irradiation point using thermal imaging and observe the time change of keyhole using X-ray imaging. Based on these in-situ simultaneous observations, we can calculate recoil pressure quantitatively and elucidate relations with melt depth change.
The objective of this study is to clarify the transition mechanism of melt depth during single-track laser irradiation. During Ti powder was irradiated by laser beam, the high temperature width which is the area where the temperature of the powder rise and shape of the keyhole were measured by X-ray and thermal imaging. Besides, the relationship among high temperature width, keyhole shape, and the recoil pressure was discussed. Figure 1 shows a schematic of LPBF experiments conducted in this study using in-situ X-ray and thermal imaging. The following parameters were different from those in our previous paper [20]: powder metal, the material of the container, its transferring method, laser irradiation condition, the Xray tube voltage and current. Spherical pure Ti powder of average particle size under 45 µm (TILOP-45: OSAKA Titanium Technologies Co., Ltd.) which was produced by gas atomization method was used. Pure Ti was used because it is well used in AM and is easy to measure temperature due to clear melting point compared with Ti6Al4V. Table 1 shows the chemical composition of the powder described in the mill sheet provided by the supplier. The powder was placed into a rectangular sample container (130 mm long, 13 mm deep, 5 mm wide, 1 mm wall thickness) made of aluminum alloy A5052. The container was tapped 5 times to increase packing density. The powder bed was flattened using a ruler. The mass of the container with and without powder was measured and packing density was calculated from those differences. As a result of calculation from Eq. (1), packing density was estimated to be 58 %.

=
(1) Here M p is the mass of the powder, ρ Ti the density of titanium, Vs the volume of the sample holder [22]. The specimen was placed inside a vacuum chamber made of stainless steel.
The fused-silica window was located on the top part of the vacuum chamber as a window for laser transmission and measurement by a two-color pyrometer. The pressure in the vacuum chamber became less than 6.8×10 -2 Pa just after laser irradiation through evacuation by a turbo molecular pump 975 nm) were measured as R 1 and R 2 , respectively. The temperature T was calculated from the ratio of these intensities using Eq. (2) which was derived from Planck's equation [23].
Here C 2 = 1.438759 × 10 -2 m･K is the second radiation constant, λ n (n = 1, 2) is the wavelength of the emitted light, and R n is the radiation intensity at wavelength λ n . The size of the images was calculated to be 640 × 480 pixels with a pixel resolution of 13.3 µm. The exact size of the images was calculated from the width of sample holders, 3 mm. The black area of images expresses that the twocolor pyrometer cannot measure both R 1 , R 2 intensity mainly because this kind of area is out of the temperature measuring range of the two-color pyrometer.

X-ray image processing
The obtained X-ray images were processed for analysis using software (Fiji) [24]. All the data were not binarized at the same brightness because the brightness of the outline of the keyhole changes as the laser scan advanced. At first, X-ray images were reversed horizontally to match the scan direction to the right. Second, median filter and Fast Fourier Transform (FFT) were applied to the raw X-ray images to remove an impulse noise and frequency component noise, respectively. Third, smooth filter was applied to the processed images. Finally, the contrast of the processed images was adjusted. Image processing included the following steps: i) median filter, ii) FFT, iii) smooth filter, iv) B/C contrast adjustment.
i. median filter ii. FFT iii. smooth filter iv. B/C contrast adjustment

X-ray image quantification
After X-ray image processing, keyhole width L w and keyhole depth L d were measured as shown in

Definition of a surface of the powder bed
For all images (1750 pictures) from irradiation start, the average brightness of the atmosphere and the powder bed were acquired each. the interface between atmosphere and powder bed was defined based on those average. After that, the value that was almost the arithmetical mean of two of them was defined as the surface of the powder bed. As a result, the surface of the powder bed pointed at the low side of the brightness. Therefore, the brightness that was about 5 high from the arithmetical mean was defined as the surface of the powder bed. The decision of the surface of the powder bed was done by sight.
2. Measurement of keyhole width L w (i) X-ray image that the deepest part of the keyhole was easy to find as shown in Fig.5

Analysis of the spatter
Diameter r and velocity v of the molten droplets were measured using manual tracking of Fiji.

Thermal image processing and quantification
Thermal images were turned 180° to match the scan direction to the right. After thermal image processing, high temperature width L h was measured as shown in Fig.3 using the following steps.
1. In thermal images, the point where the temperature was measured for the first time was decided as an irradiation point. Furthermore, the image that temperature was observed for the first time was defined as the irradiation start time (0 s) and synchronized with the X-ray image.
2. Point "A" was moved from irradiation point to y+ axis and point "B" was moved from irradiation point to y-axis, the point that the R strength became lower than 10 was defined as the edge of the high temperature area.
3. Time change of distance from point "A" to "B" was measured as high temperature width L h . Figure 4 shows three examples of the obtained raw X-ray images. Figure 4(a) shows an X-ray image before laser irradiation start. The bright region on the upper side of the image is the atmosphere, and the dark region on the lower side of the image is the powder bed. Figure 4(b) shows the X-ray image during laser irradiation at t = 0.430 s. The slightly bright region compared with the powder and narrow region in the powder bed is the keyhole that was generated under the laser beam, which is often observed in the welding [21,25]. The slightly dark region compared with the powder bed which existed behind the keyhole (-x direction) is solidification bulk and molten droplet. The black and spherical object at the center of the X-ray image is a spatter. It was scattered above and fell again on the powder bed. Figure 4(c) shows the X-ray image during laser irradiation at t = 2.088 s. The black and spherical object on the right side of the solidification bulk is a molten droplet. The black region at the four corners of the X-ray image is the chamber that was reflected. The above fact was confirmed from thermal images later. These had already confirmed in our previous paper that used the same experimental equipment [20]. Figure 5 shows the melting behavior of the powder metal. Figure 5(a) shows the obtained raw Xray images for observation of spatter. As a specific example, molten droplets named S1~S5 were picked up and showed a moving direction with an arrow. Figure 5(b) shows processed images whose minimum and maximum brightness were 40 and 120, respectively to focus on keyhole shape in a powder bed. Furthermore, Figure 5(c) shows the measurement result of the keyhole depth L d and the keyhole width L w . The circle on the straight line shows the edge of the keyhole, the distance of the interval is measured as keyhole depth and keyhole width.

Time change of melting behavior
Molten droplets S1 and S2 generated just after laser irradiation start were scattered in the direction (over the powder bed) of the arrow which was shown in Fig5(a). As laser irradiation advanced, molten droplets S3~5 formed right under an irradiation point moved to reverse with the scanning direction and jumped into solidification bulk as shown at t = 1.680, 2.156, and 3.226 s.  Fig.5(b). Also, L d and L w increased at the same speed just after the irradiation start. Figure 7 shows the trace and moving velocity of molten droplets S1 ~ S5 using manual tracking in Fiji. Each plot in the trace expresses the place where a molten droplet existed for each frame (every 0.002 s) which corresponds to the frame rate of 500 fps of the CCD camera. The moving velocity was calculated by dividing the moving distance by 0.002 s (1 frame).

Analysis of the molten droplets (spatter)
The molten droplets formed just after irradiation start was scattered over the powder bed. As laser irradiation advanced and keyhole depth decreased, the molten droplets which were generated right under an irradiation point moved opposite to a scanning direction and jumped into solidification bulk.
The maximum moving velocities were 147, 214, 257, 268 and 271 mm/s for S1, S2, S3, S4 and S5, respectively. Although the moving direction was different from each other, the maximum velocity was almost at the same level. From this result, it was suggested that the same driving force acted at any time during laser irradiation. Figure 8 shows the variation of the temperature distribution with time with the corresponding crosssectional view of X-ray images during laser irradiation. All the thermal images showed that the temperature rose from the edge of the melt pool toward the irradiation point. Compared with Figs.8(c) and (d), S1~S5 which were scattered over the powder bed and jumped into backward solidification bulk would be molten droplets because their temperature is close to the melting point of titanium 1668℃ and their shape were spherical [22]. Figure 9 shows the time variation of the high temperature width L h , distance between A and B which was shown in Fig.8. Also, Fig.10 shows the thermal images which correspond to the three kinds of plots (black, red and light blue) shown in Fig.9. As shown in Figs.10(a) and (b), L h is overestimated by the molten droplet which existed on the measurement line that is parallel to the y-axis through an irradiation point as shown at t = 0.06 s in Fig.8(a). In Fig.10(c) outline of the temperature rose area could not be extracted because temperature around the irradiation point could not be measured.

Discussion
In this chapter, the time when L w , L d and L h became constant is defined using the coefficient of selfcorrelation. Besides, a model about the recoil pressure that pushed the molten droplets is constructed and this phenomenon is examined quantitively. Based on these, the transition mechanism of melt depth is clarified by considering the relationship among the recoil pressure and each value of L w , L d and L h .

Time variation of a keyhole shape and high temperature width
In terms of L w and L d in Fig.6, and L h in Fig.9, the time when became constant was examined quantitatively. The correlation coefficient was counted backward from 3.5 s when laser irradiation As a result, keyhole depth L d became constant at t = 1.812 s as shown in Fig.6(a). L w decreased temporarily and increased afterward at the peak which is shown in a circle in Fig.6(b). At the time of peak outbreak, following phenomena that L w decreased because the molten droplet formed right under the irradiation point and L w increased because the molten droplet jumped into backward solidification bulk as shown in X-ray images (at t = 2.156, 3.226 s) in Fig.5 was confirmed. After the consecutive datum with values less than 1.8 mm were omitted in the peaks shown in circles, the coefficient of selfcorrelation of L w was calculated. As a result, keyhole width L w became constant at t = 1.298 s as shown in Fig.6(b). The high temperature width L h became constant at t = 1.66 s as shown in Fig.9. Here, only using the data which were measured normally, the coefficient of self-correlation was calculated.
When the above results were summarized, the time when became constant was keyhole width L w , high temperature width L h, and keyhole depth L d in ascending order. It was suggested that the transition of L d was affected by L w and L h .

A dynamic evaluation of the molten droplet and investigation of the driving force
Then, the driving force that pushed a scattered molten droplet from a volume and scattering speed of a molten droplet was examined. Figure 11 shows that a simple model of the movement of a molten droplet. It is supposed that the molten droplet was moved by the driving force Then, the driving force F d was calculated through substitution of maximum velocity to v and substitution of the interval of record 0.002 s to t. Furthermore, Chimura et al. made a model of the recoil pressure that occurred by vaporization of metal [21]. From Eqs. (5) and (6), the recoil pressure was found to be inversely proportional to a vaporization area S and proportional to the quantity of vaporization rate ̇.
Here is the vaporization velocity of the atom, k the Boltzmann constant, b the boiling point, a the atomic mass, v the laser power consumed for vaporization, v the vaporization latent heat, the vaporization area. The numerator on the right side of Eq. (6) expresses the theoretical value F v of the volume expansion by the vaporization. Compared F d with the vaporization F v , it was able to be considered whether the driving force that pushed a molten droplet was due to the vaporization of powder metal or not. Here, ̇ was assumed to be constant and F v was calculated.
Laser power consumed for vaporization P v was calculated based on the assumption that 20 % of all the laser output was absorbed [26]. Besides, Fig.12 shows F d of S1~ S5 and the molten droplets which were observed except for S1~ S5.
As a result, vaporization F v became larger than all F d . Furthermore, the average of the early period of F d became larger than that of the latter half of F d . These results are examined in consideration of a temperature change in the next chapter.

Evaluation of melting behavior by thermal results
The results in Fig.12 were obtained without considering the time change of the quantity of vaporization rate ̇ accompanied with the temperature change around the irradiation point. It was revealed that L h increased gradually in Fig.9 as laser irradiation advanced. It was suggested that as L h increased, the temperature of the powder metal around the irradiation point rose and sintered material was formed. As a result, thermal diffusion was promoted and local heating was mitigated, and then laser power consumed for vaporization v decreased. Due to the occurrence of this phenomenon, the recoil pressure decreased as laser scan advanced. Thus, the theoretical value of vaporization F v shown in Fig.12 was the smaller value when temperature change around the irradiation point was considered, and a relative error with the experimental value of vaporization F v is shortened as shown in the schematic in Fig.13.
In the first place, it is difficult to measure the driving force to push a molten droplet during irradiation accurately, but in this study, it was calculated experimentally. The experimental value F d is not so far off by an order of magnitude compared to the theoretical value F v . The experimental values F d were calculated under the assumption that the scattering velocity was two-dimensional and not all the evaporated atoms contributed to the scattering. Therefore, the actual driving force F d is expected to be larger than calculated values F d . Therefore, from the results in this study and previous studies [10][11][12], it is reasonable to assume that the driving force to push a molten droplet was generated through the evaporation of the metal powder. If the F d can be measured quantitatively, the keyhole depth at that time can be roughly estimated. This can be used as a guideline to set the irradiation conditions with fewer defects for industrial use.

The relationship among keyhole shape, high temperature width and recoil pressure
The recoil pressure was calculated by dividing driving force F d by a vaporization area S calculated from keyhole shape. Then, the vaporization area was approximated as an elliptic cylinder. It was supposed that the elliptical semi-major axis is 2 and semi-minor axis is 4 , and then the height is L d . Figure 14 shows the relationship between the keyhole depth and the recoil pressure. The circle plot shows molten droplets which generated just after irradiation start, and the rhombus plot shows the molten droplets which generated after about 1.5 s. The tendency was showed that the keyhole depth increased with increasing the recoil pressure as shown in the previous study [21]. In Fig.14, the two plots at 60~65 Pa have small L d despite the large recoil pressure. This can be attributed to the large recoil pressure due to the heating at the laser tip even before the keyhole depth becomes large. The two plots at 60-65 Pa correspond to the molten droplets generated at 0.010 s and 0.032 s. At that time, the keyhole was shallow and the powder metal was heated by the laser tip. It has been revealed that the vaporization rate was largest when the powder was heated by the laser tip immediately after the start of laser irradiation, and the keyhole deepens later than the increase of recoil pressure [21].
Therefore, it can be understood that a large recoil pressure pushed the molten droplet immediately after the start of laser irradiation and the keyhole deepened later than the increasing recoil pressure in this study. Also, two circle plots with L d of 2.13 mm (23 Pa, 74 Pa) show another scattered molten droplet at the same time (t = 0.052 s). Normally, the same recoil pressure would push them. Therefore, it is conceivable that there are points with different recoil pressures even within the same keyhole depending on the multiple reflections of the laser in the keyhole.
Although the recoil pressure was calculated by the velocity of molten droplets, it is necessary to examine relations with the keyhole depth in the future by acquiring the quantity of vaporization by simulation. Furthermore, it turned out that heating form varied from local heating in the keyhole tip to heating in the surface of the keyhole wall by the multiple reflections of the laser beam as laser irradiation advances [21]. Because it was thought that the recoil pressure decreased when this phenomenon occurred, the quantity of vaporization is an important examination item and is the point that cannot be considered in this study. Figure 15 shows a schematic of the transition mechanism of melt depth by single-track laser irradiation after t = 0.13 s. The transition mechanism of melt depth by single-track laser irradiation was thought as follows. The increase of L d is seen early in irradiation, which is caused by the increase of the recoil pressure P T due to the decrease of L w and large quantity of evaporation ̇ (small L h ).

Transition mechanism of melt depth by single-track laser irradiation
Next, the decrease of L d is seen as laser scan advanced, which is caused by the decrease of the recoil pressure P T due to the increase of L w and decrease of quantity of evaporation ̇ (an increase of L h ).
At this point, the decrease in the quantity of evaporation can be attributed to the formation of the sintered material, which promotes thermal diffusion and reduces the laser power consumed for evaporation. In addition, the increase in L w considered to be due to the formation of the solidified bulk, which makes the voids between the powder metal apparent and the keyhole apparent. Finally, the transition to the constant L d is caused by stabilization of L w and L h followed by stabilization of P T .

Conclusion
By in-situ simultaneous observation methods combining X-ray and thermal imaging, the following knowledge about the melting behavior of the powder bed during the LPBF process was obtained.
1) Keyhole depth L d increased at first, gradually decreased as laser irradiation advanced, and then became constant finally. Keyhole width L w increased just after irradiation start and decreased thereafter. After that, L w increased again, became constant finally. High temperature width L h gradually increased as laser irradiation advanced. The time when became constant was keyhole width L w , high temperature width L h and keyhole depth L d in ascending order.
2) The transition mechanism of melt depth by single-track laser irradiation after t = 1.3 s was thought as follows. The increase of L d in the early period of the irradiation is due to the increase of the recoil pressure P T . This is because the decrease of L w (decrease of vaporization area S) and large quantity of vaporization (small L h ). Next, the decrease of L d is due to the decrease of the recoil pressure P T . This is because the increase of L w and decrease of quantity of vaporization (due to the increase of L h ). At last, the transition to the constant L d is caused by stabilization of L w and L h followed by stabilization of P T .

-Ethical Approval
Not applicable.
-Consent to Participate The authors declare that they all consent to participate this research.
-Consent to Publish The authors declare that they all consent to publish the manuscript.

-Authors Contributions
Not applicable. -Funding Not applicable.

-Competing Interests
Not applicable.
-Availability of data and materials All data generated or analysed during this study are included in this published article and its supplementary information files.    Molten droplets named S1~S5 were examples that show characteristic melting behavior. Molten droplets S1 and S2 which were generated just after laser irradiation start were scattered at the direction of the arrow which was shown in Fig5(a). As laser irradiation advanced, molten droplets S3~5 formed right under an irradiation point moved to reverse with the scanning direction.      Fig.11 Model of the spattering phenomenon for elucidating the driving force that pushed a molten droplet during the LPBF process. Fig.12 Comparison of the driving force F d that pushed a molten droplet and vaporization F v . F d is the driving force that pushed a molten droplet, F v is the theoretical value of vaporization. Fig.13 Comparison of the driving force F d that pushed a molten droplet and vaporization F v considering temperature change. The schematics that evaporation F v decreased due to the decrease of quantity of vaporization ̇ as laser scan advanced. It was thought that as L h increased, the temperature of the powder metal around the irradiation point rose and sintered material was produced. As a result, thermal diffusion was promoted and local heating was mitigated, and then the quantity of vaporization ̇ decreased. In addition, the result that the average of the early period of F d became larger than that of the latter half of F d was also considered because of the decrease of vaporization.
Fig.14 Relationship between the measured keyhole depth L d and recoil pressure P T . The recoil pressure was calculated by dividing driving force F d as shown in Fig.13 by a vaporization area S calculated from keyhole shape at that time. The vaporization area was approximated as an elliptic cylinder. It was supposed that the elliptical semi-major axis is 2 and semi-minor axis is 4 , and then the height is L d . The kind of plot corresponds to Fig.13. The circle plots are the value that divided F d of the early period of irradiation by a vaporization area S at that time. The rhombus plots are the value that divided F d of the irradiation latter half by a vaporization area S at that time. then increased again and became constant. The increase in L w considered to be due to the formation of the solidified bulk, which makes the voids between the powder metal apparent. It is assumed that the recoil pressure P T which is expressed as the ratio of these two parameters (̇ and S) increased after the start of laser irradiation, gradually decreased as laser irradiation advanced and then became constant. Due to the causes mentioned above, L d is thought to have caused the same changes as recoil pressure. Table 1 Chemical composition of Ti powder used in this study (mass %)