Effects of gas pressure and catheter length on the breakup of discontinuous NiTi droplets in electrode induction melting gas atomization

NiTi powders used for selective laser melting have here been fabricated by the breakup of discontinuous droplets in electrode induction melting gas atomization (EIGA). The morphology, particle size distribution, and hollow ratio of the powder were characterized by scanning electron microscopy (SEM), laser particle size analyzer, and computed tomography (CT), respectively. The effects of gas pressure and catheter length on the particle size distribution and powder morphology were then studied. Furthermore, the effects of the classifier wheel speed on the particle size distribution and yield of the 15–53 μm powder in the classification process were also analyzed. The results showed that the average particle size (D50) of the NiTi powder first decreased and, thereafter, increased as the atomization gas pressure increased. This was also the situation with catheter length. Also, the yield of the 15–53 μm powder increased with an increase in the classifier wheel speed. The optimum parameters were a gas atomization pressure of 5 MPa, a tension length of 28 mm, and a classifier wheel speed of 660 r min−1. For this optimized condition, the D50 value and the yield of the NiTi powder were 57.54 μm and 46.4%. In addition, the flowability, hollow ratio, and oxygen content were 15.8 s/50 g, 0.31%, and 450 ppm, respectively.

Nowadays, additive manufacturing (AM) methods, including selective laser melting (SLM) [11], electron beam melting (EBM) [12], wire arc additive manufacturing (WAAM) [13], and laser-based directed energy deposition (LDED) [14] have gained significant attention in the manufacturing of NiTi parts [15][16][17][18][19][20][21][22][23].Since the AM method can reduce the three-dimensional manufacturing process to a two-dimensional one by layer-bylayer stacking, it can be used to fabricate NiTi with complex and precise structures.It should be noted that most AM NiTi parts use NiTi powder as a raw material.As a feedstock for SLM, the characteristics of the NiTi powder play an essential role in the preparation procedure and performance of the shaped products [24][25][26].The preparation methods for spherical NiTi powder for SLM mainly include electrode induction melting gas atomization (EIGA) [16], plasma spheroidization (PS) [27], and a plasma rotating electrode process (PREP) [28].Shi et al [27] obtained a spherical Ni-Ti composite powder, which was synthesized by blending elemental Ni and Ti powders.This mixture was, thereafter, spheroidized by the radio frequency plasma spheroidization technique.Chen et al [28] prepared a NiTi powder by a PREP, with the particle sizes ranging from 40 μm to 180 μm.Compared with the PS method and the PREP method, the EIGA method has the advantages of low cost, high yield and high production efficiency.In contrast to the vacuum induction melting gas atomization (VIGA) method, the EIGA method does not require a crucible, and the prepared powder has a higher purity.
To the best of the present authors knowledge, only powder with a particle size ranging from 15 μm to 53 μm can be used as a feedstock for SLM.When prepared by the EIGA method, the wide particle size distribution of the raw metal powder must be separated, which implies low productivity.With the wide use of an argon recovery device, the production cost of the gas atomized NiTi powder is mainly the cost of the NiTi alloy bar.Thus, it is of vital importance to reduce the production cost by increasing the yield.Many previous studies have focused on the SLM process and the use of NiTi powder [11,[15][16][17][18][19][20][21].Few studies have been conducted on the preparation process of the NiTi powder using EIGA, with the purpose of increasing its yield.
The atomization pressure and nozzle structure are important factors that affect the preparation of NiTi powder by the EIGA process.These factors directly affect the yield and quality of the fine powder, with particle sizes in the range of 15-53 μm used in SLM process.Moreover, the catheter length in the nozzle is a key factor that affects the size of the negative turbulent area and the structure of the flow field [29].It has also significant effects on the particle size, morphology, and hollow ratio of the NiTi powder prepared by EIGA.
Therefore, the purpose of the present study has been to study the effects of gas pressure and catheter length on particle size distribution and morphology.Furthermore, the effects of the classifier wheel speed on the particle size distribution and yield of 15-53 μm powder in the classification process have also been analyzed.Also, the properties of the 15-53 μm NiTi powder were finally investigated.The results of this study have provided a technical reference for the preparation of 15-53 μm NiTi powder for SLM.

Experimental materials and method
In the study, the raw material was a pre-alloyed NiTi rod with a diameter of 50 mm and a length of 750 mm (manufactured by Xi 'an Baoji Metal Materials Co., Ltd., China).The chemical composition of the NiTi rod is presented in table 1.The end of the NiTi50 rod was machined to an angle of 45°.Also, high purity argon (purity 99.999%) was used as the atomization gas, which was supplied by Sichuan Pangang Messer Gas Products Co., Ltd., China.
A laser image-particle size analyzer (Bettersize3000plus, China) was used to analyze the particle size distribution and sphericity of the powder.The oxygen nitrogen hydrogen analyzer (TC600, USA) was adopted to measure the contents of oxygen in the powder.Furthermore, the additive manufacturing mass analyzer (Explorer4, USA) was used to analyze the powder morphology.The flowability, tap density, and bulk density were measured by the powder comprehensive characteristic tester (BT-1000, China).Also, scanning electron microscopy (SEM), combined with energy dispersive spectroscopy (Sigma500, Germany) was used for the analysis of the element distributions on the surface of the NiTi powder.Finally, the hollow ratio was observed by computed tomography (YXLON FF35, Germany).

Results and discussion
The essence of gas atomization (GA) is the interaction between the high-velocity gas and the liquid, especially the liquid behavior in the supersonic gas flow [30].
In the VIGA process, the metal melt flows through the catheter in contact with the catheter wall.On the other hand, in the EIGA process, a discontinuous droplet passes through the center of the catheter (i.e., not in contact with the catheter wall).There are some differences in the breakup mechanism of these two processes.
In the process of EIGA, a molten droplet left the tip of the NiTi rod and firstly stretched to into longer stripe in the middle of catheter (figures 1 and 2).This stretching is caused by pressure difference between the melting chamber and the atomizing chamber.The process of gas atomization could be divided into two stages: primary break-up, secondary break-up.During the primary break-up stage, the molten metal was broken up into unstable liquid bodies (i.e., liquid film), which in turn broke down into drops (figure 1) [31].During the secondary break-up, the larger droplets, or liquid film were further broken up into smaller droplets in the supersonic expansion zone of the airflow (figures 1 and 2) [32].The gas pressure and catheter length were vital of importance for the breakup of the discontinuous droplets in EIGA.shows the D50 value powder and yield of 0-53 μm powder prepared at different gas pressures.With an increase in gas pressure, the D50 value of the powder did first gradually decrease, and then increase.This corresponded to the increase and, thereafter, decrease of the 0-53 μm powder yield.

Effect of gas pressure on the particle size distribution and morphology
The powder particle size distribution was very wide at a gas pressure of 3 MPa (figure 3(a)).Also, the D50 value of the powder was 91.56 μm, and the yield of the 0-53 μm powder was 20.39% (figure 3(b)).At a gas pressure of 5 MPa, the range of the powder particle size distribution was the narrowest (figure 3(a)), with the smallest D50 value of 57.54 μm and the highest 0-53 μm yield of 51.65% (figure 3(b)).With a further increase in the pressure, the D50 value of the powder increased and the yield of the fine powder decreased.At a gas pressure of 6 MPa, the D50 value was 62.06 μm and the 0-53 μm yield was reduced to 46.18% (figure 3(b)).
The gas atomization of liquid metals is dependent on the energy transfer from a high-velocity gas medium to a relatively low-velocity molten metal stream [31].Also, the gas-to-metal ratio (GMR) has an important effect on the particle size and ultimately affects the yield of the fine powder [33].The relationship between GMR and the  D50 value of the powder can be determined by equation (1) [34]: where k 1 is a constant to be determined by the states of the melt stream and the nozzle, D l is the diameter of the melt stream flow (m), and η m and η g are the kinematic viscosities of the molten metal and gas (m 2 •s −1 ), respectively.Also, M and G are the flow rates of the molten metal and gas (kg•m −3 ), respectively.W is the Weber number, and the magnitude of this value indicates the importance of surface tension [33].
According to the aerodynamic principle, the relationships between the atomization pressure and the flow rate and velocity of the nozzle outlet are determined by equations (2) and (3), respectively [35].
where S is the cross-sectional area of the nozzle outlet, P is the pressure in the gas outflow nozzle gap (which is approximately the pressure of the gas flowing into the nozzle), ρ is the gas density, K is the adiabatic coefficient of the gas, T 2 is the gas temperature before the compressed gas enters the nozzle, and P 1 is the pressure in the atomization chamber.In the present study, ρ = 1.784 kg cm −3 , K −1 = 1.66,T 2 = 288.15K, and P 1 = 1 × 10 5 Pa.The cross-sectional area of the nozzle outlet was here fixed without changing the nozzle clearance.As the atomization pressure increased, the velocity and flow rate of the gas also increased.The increasing kinetic energy of the gas caused the droplet to breakup into finer droplets.Finally, the finer droplets formed powder particles after spheroidization and solidification.As described in equation (1), higher GMR values are beneficial for finer powder with narrower particle size distributions (figure 3(a)).However, when the gas pressure was 6 MPa, the recirculation zone became larger and the collisions among the powder particles became more frequent, which resulted in aggregations and more satellites (figure 4(d)) [29].It also resulted in an increase in the D50 value of the powder and a decrease in the yield of the 0-53 μm powder (figure 3(b)).
Figure 4 shows the SEM morphology of the NiTi powder that was prepared for different gas pressures.Most powder particles showed good sphericity and a small proportion of satellites.Satellites were formed when small solid particles welded on a large semi-solid particle, whereas aggregations were formed when many small semisolid particles welded together [33].
For the pressure of 3 MPa, there were still a few melt sticks that did not break completely and coarse particles (figure 4(a)).Compared with the pressure of 5 MPa, the powder particles that were prepared at 4 MPa had some irregular shape and a lower sphericity.When the pressure increased from 3 MPa to 5 Mpa, it could be clearly seen that the particle size of the powder became smaller, which confirmed the results of the particle size distribution in figure 3. Nevertheless, satellites and aggregations appeared at a gas pressure of 6 MPa, which was due to a significantly increasing number of collisions (figure 4(d)).Therefore, the NiTi powder that was prepared at 5 MPa, with higher sphericity and fewer defects, has here been found to be more suitable for SLM.With an increase in catheter length, the D50 value did first gradually decrease, and then increase.When the catheter length increased from 26 mm to 28 mm, the cumulative distribution curve of the powder particle size tended to move to the left.However, when the catheter length was 29 mm, the curve drastically shifted to the right.Also, when the catheter length was 26 mm, the powder particle size distribution range was very wide (figure 5(a)).The D50 value of the powder was 97.34 μm, and the yield of the 0-53 μm powder was 22.30%.Furthermore, when the catheter length was 28 mm, the D50 value of the powder was the smallest (57.54 μm), and the 0-53 μm powder yield was the highest (51.65%).With a further increase in the catheter length, the average particle size of the powder increased, and the yield of the fine powder decreased.When the catheter length was 29 mm, the D50 value of the powder was 62.06 μm, and the 0-53 μm powder yield was reduced to 35.71% (figure 5(b)).
The pressure at the tip of the catheter was of large importance for the smooth outflow of molten metal during the atomization process.As the catheter length of the catheter increased, the pressure at the tip of the catheter changed from a positive pressure to a negative pressure [36].
When the catheter length was 26 mm, the pressure at the tip of the catheter was positive, and the metal melt had difficulties flowing smoothly into the atomization chamber.In addition, a metal melt reverse spray occurred, which caused a blockage in the catheter.When the catheter length was 29 mm, the distance of the metal melt channel became longer, and the heat loss of the metal melt in the process of flowing through the metal melt channel was increased.This reduced the superheat of the metal melt and the powder yield became lower.Furthermore, a part of the gas energy was lost on the catheter, and the airflow became too dispersed.Also, the atomization ability was weakened.
When the catheter length was 27 mm or 28 mm, the pressure at the outlet position was small and a negative pressure was formed.In addition, the metal melt could enter the atomization chamber smoothly.When the catheter length was 28 mm, the metal melt accelerated into the atomization chamber, which shortened the flow time and ensured a superheat of the metal melt.Figure 6 shows the SEM morphology of the NiTi powder for different catheter lengths.Satellite particles were found in the powder that was prepared with different catheter lengths during atomization.When the catheter length was 26 mm, the droplet was not stretched and did not become directly atomized.This resulted in metal droplets that were not completely broken up.Instead, they were spheroidized to form large particles.
For a catheter length of 29 mm, a lower superheat of the droplets led to longer spheroidization times.Solidification of large ligaments occurred before the spheroidization was completed, and this led to powder particles with a ligament shape (figure 6(d)).The powder particles shown in figure 5(b) were larger than those shown in figure 6(c).When the catheter length was 28 mm, the negative pressure below the catheter became larger, which contributed to a stretching of the metal droplets to longer stripes and the powder was completely broken.Therefore, the optimal catheter length has here been chosen as 28 mm.

Effect of classifier wheel speed on the particle size distribution and yield
The powder with different particle sizes was sieved into a 0-53 μm particle size range by an ultrasonic shaker (figures 7 and 8(a)).As can be seen in figure 8(a), satellites and aggregated powder appeared, which resulted in poor fluidity.The selective laser melting required powder with a particle size of 15-53 μm and good flowability.Therefore, the 0-53 μm powder was classified into particle sizes of 15-53 μm by a turbo air classifier.Also, the oxygen content of the turbo air classifier was controlled at a value below 50 ppm to reduce the oxidation of the powder.
In accordance with the principle of classification, the particles in the classifier wheel are exposed to two forces: one is the centrifugal force, F u , caused by the rotary inertia, the other is the drag force of the airflow, F f [37].The coarse particles will move along the trajectory and strike the classifier wheel blades.They will, thereafter, fall into the coarse powder collection when F u is larger than F f .Also, the fine particles will move into the center of the classifier wheel by the airflow, and enter the fine powder collection when F u is smaller than F f .
According to the systematic analyses of calculation formulas, which were obtained under different assumptions in a study of the cut size of the turbo air classifier, Gao et al [38] obtained a simplified equation (equation ( 4)): where d T is the cut size, Q is the airflow rate, and n is the classifier wheel speed.Also, ε, δ, and k are the parameters determined by the characteristics of the material, temperature, humidity, pressure of air, and the configuration of the classifier.As can be understood from equation (4), the cut size decreases when the classifier wheel speed increases.
As can be seen in table 2, with an increase in the classifier wheel speed, the D10 value of the 15-53 μm powder decreased gradually.When the classifier wheel speed increased from 540 r min −1 to 720 r min −1 (i.e., 540 r min −1 , 600 r min −1 , 660 r min −1 , and 720 r min −1 ) the D10 value of the 15-53 μm powder decreased by 25.26 μm, 23.08 μm, 21.99 μm, and 20.28 μm, respectively.Furthermore, the D90 value of the 0-15 μm powder decreased with an increasing classifier wheel speed.This led to an increased yield of the 15-53 μm powder.However, the small D10 value of the 15-53 μm powder, corresponded to a decreased fluidity.Furthermore, for classifier wheel speeds of 660 r min −1 and 720 r min −1 , the yields of the 15-53 μm NiTi powder were 46.4% and 47.5%, respectively.Also, the flowability was 15.8 s/50 g and 19.1 s/50 g, respectively.By taking the flowability and yield of the 15-53 μm powder into account, the optimum classifier wheel speed was 660 r min −1 .
For a classifier wheel speed of 660 r min −1 , figure 7 shows the particle size distribution of the NiTi powder before and after classification.The powder showed a Gaussian distribution with a low standard deviation, both before and after classification.This proved that the classification process did not affect the normal distribution of the powder.Furthermore, there was an overlap of the size differentiate distribution curves of the 0-15 μm and 15-53 μm powders, which was caused by Van-der-Waals attractive forces between the particles during the process of classification [29] (figure 7).The particle size distribution range of 15-53 μm was significantly narrower than that of the range of 0-53 μm.
SEM images of NiTi powder with particle sizes of the 15-53 μm and 0-15 μm powders are shown in figures 8(b) and (c), respectively, for a classifier wheel speed of 660 r min −1 .It was obvious that these powder particles were effectively separated during the classification process.The 15-53 μm powder was detached and the particles did not adhere to each other.Furthermore, the satellite ball particles on the surface of the powder did almost disappear.Also, the surface of 15-53 μm NiTi powder was smooth with a high sphericity, thereby ensuring high flowability.A small number of particles larger than 22 μm can be seen in figure 8(c), which corresponds to the data in table 2 and figure 7.

Properties of the 15-53 μm NiTi powder
The chemical composition of the NiTi powder that was been produced by EIGA, is presented in table 3. The actual composition of the NiTi powder was equivalent to the nominal composition, except that the oxygen  content had increased by 200 ppm in the production process.The element contents met the requirements of SLM for powder compositions.The element distributions on the powder particle surfaces were analyzed by surface scanning using energy dispersive spectroscopy (EDS).As can be seen in figure 9, the main elements Ni and Ti were adequately distributed on the surfaces of the spherical particles, and there was no obvious enrichment or dilution.
The hollow powder of the 15-53 μm NiTi powder is shown in figure 10.It is obvious that the gray-white powder with black center is a hollow powder.Based on the CT technology, the computer image processing and statistics showed that the amount of hollow powder only accounted for 0.31% of the total powder.Formation mechanism of the hollow powder is illustrated in figure 11.The metal droplet was compressed by gas to become a thinner ligament.The molten metal became unstable and was easily broken up into smaller droplets or liquid film when the drag force overcame the surface tension.Thereafter, the liquid film began to touch the gas and produced some waves, and these waves moved rapidly together with ligament deformations and rotations.This involved a bubble-formed gas in the ligament, and the ligament broke up into droplets.Some bubbles were retained in the droplets, which led to a hollow powder after solidification [39].
The content of the hollow powder affected the mechanical properties of the additively manufactured parts, which increased the possibility of cracking and fatigue cracking.This was detrimental to the tensile strength of the additively manufactured parts at room temperature and at high temperatures.
Table 4 summarizes the physical properties of the 15-53 μm NiTi powder, including flowability, sphericity, bulk density, tap density, and hollow ratio.According to table 4, the flowability, sphericity, hollow ratio, bulk density, and tap density of the powder were 15.8 s/50 g, 91.2%, 0.31%, 3.91 g cm −3 , and 4.24 g cm −3 , respectively.As can be seen in figure 12, 15-53μm NiTi powder was used to form a NiTi porous structure.By using the EOS M290 machine, SLM part was, thereafter, successfully manufactured with a scanning speed of 1100 mm s −1 , a laser power of 250 W, a hatch distance of 60 μm, and a layer thickness of 30 μm.The relative density of the NiTi porous structure was 98.58%.

Conclusions
The effects of gas pressure and catheter length on the particle size distribution and powder morphology have been comprehensively studied in the present work.This was also the situation with the effect of classifier wheel   (1) The D50 value of the NiTi powder did first decrease, and then increase, when the atomization gas pressure increased.This was also the result of the catheter length.For a gas atomization pressure of 5 MPa and a tension length of 28 mm, the D50 value of the NiTi powder was 57.54 μm and the yield of the 0-53 μm NiTi powder became 51.65%.
(2) Furthermore, the yield of the 15-53 μm NiTi powder increased with an increase in the classifier wheel speed.However, both its D10 value and flowability decreased.When the classifier wheel speeds were 660 r min −1 and 720 r min −1 , the yields of the 15-53 μm NiTi powder became 46.4% and 47.5%, respectively.However, the flowabilities were 15.8 s/50 g and 19.1 s/50 g, respectively.By taking both the flowability and yield of the 15-53 μm powder into account, the optimum classifier wheel speed was 660 r min −1 . (

Figure 3 (
a) shows the effects of gas pressure on the particle size distribution of the NiTi powder and figure 3(b)

Figure 1 .
Figure 1.Principle of atomization and catheter length.

Figure 3 .
Figure 3. (a) Particle size distributions of the NiTi powder prepared at different gas pressures; (b) The D50 value powder and yield of 0-53 μm powder prepared at different gas pressures.

3. 2 .
Effect of catheter length on the particle size distribution and powder morphology Figure 5(a) shows the effect of the catheter length on the particle size distribution of the NiTi powder and figure 5(b) shows effect of the catheter length on the D50 value powder and yield of 0-53 μm powder prepared.

Figure 5 .
Figure 5. (a) Particle size distributions of the NiTi powder for different catheter lengths; (b) The D50 value powder and yield of 0-53 μm powder prepared for different catheter lengths.

Figure 7 .
Figure 7. Particle size distributions of the powder before and after classification. d

Figure 9 .
Figure 9. SEM images of the surface of the NiTi powder.

Figure 10 .
Figure 10.CT image of hollow NiTi powder in the range of 15-53 μm.

Figure 11 .
Figure 11.Formation mechanism of the hollow powder.

Table 2 .
Effect of classifier wheel speed on the particle size distribution and yield.