Experimental investigation on the height deviation of bumps printed by solder jet technology

Journal of Materials Processing Technology 243 (2017) 291–298

Contents lists available at ScienceDirect

Journal of Materials Processing Technology journal homepage: www.elsevier.com/locate/jmatprotec

Experimental investigation on the height deviation of bumps printed by solder jet technology Wei Xiong a , Lehua Qi a,b,∗ , Jun Luo a,b,∗ , Daicong Zhang a , Junhao Liang c , Hao Yi a School of Mechanical Engineering, Northwestern Polytechnical University, Xi an 710072, China Education Ministry Key Laboratory of Modern Design and Integrated Manufacturing Technology, Northwestern Polytechnical University, Xi’an 710072, China c School of Material, Northwestern Polytechnical University, Xi an 710072, China a

b

a r t i c l e

i n f o

Article history: Received 18 September 2016 Received in revised form 28 December 2016 Accepted 31 December 2016 Available online 31 December 2016 Keyword: Solder bumps Height deviation Oscillation Solidification

a b s t r a c t Solder jet technology is considered as a flexible and low cost method to print bumps directly for flip-chip packaging. However, the height deviation of solder bumps, which decides the quality and reliability of the packaging, is significantly influenced by the fluid dynamics and solidification behaviors during solder droplet impact. Here, an experimental investigation was first conducted to understand the influence of impact parameters on the height deviation of solder bumps. The results showed that the underdamped oscillation of droplets before complete solidification was the main reason for the bump height deviation, because this oscillation resulted in different bump shapes under different solidification rates. A clear threshold, dividing the regions of the large and small deviation of the solder bump height, was found by estimating the droplet total solidification time scale. “Deposition and reheating” process was presented and proved to be an effective method to reduce the height deviation. A solder bump array with the height of 223 ± 2 ␮m, was printed through this process. The dimensionless height deviation (h/h) of printed bumps was remarkably less than 1%. The present work provided a method of printing uniform height solder bumps. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Packaging electronic devices has become a considerable manufacture challenge due to the integration and multifunction of electronic circuits. Luo et al. (2012), Liu and Orme (2001) considered the solder jet technology as a flexibility and low cost method to print bumps directly for flip-chip packaging. During the process, the micro solder droplets are ejected from a nozzle and deposited onto a moving substrate directly to form solder bumps, then a chip is placed in contact with the pre-deposited solder bumps (also called solidified droplets) on the substrate. The common used solder bump materials include the PbSn solder and the lead-free solder (such as SnAg and SnAgCu). The study on micro molten droplet spaying and deposition was first originated by Orme et al. (1993). Later, various uniform droplet spraying apparatuses were developed. Luo et al. (2012) developed a pneumatic droplet generator

∗ Corresponding authors at: School of Mechanical Engineering, Northwestern Polytechnical University, Xi’an 710072, China. E-mail addresses: [email protected] (L. Qi), [email protected] (J. Luo). http://dx.doi.org/10.1016/j.jmatprotec.2016.12.031 0924-0136/© 2017 Elsevier B.V. All rights reserved.

and proposed a 2D axisymmetric model to understand the mechanism of droplet ejection. Amirzadeh et al. (2013) and Luo et al. (2016) produced droplets whose diameters were smaller than the nozzle diameters. He et al. (2014), Lee et al. (2008), and Fan et al. (2008) also developed a variety of generators to produce micro solder droplets. The basic theories and method to eject uniform droplets has been widely investigated, and how to print solder bumps with a uniform height became a crucial problem for the application of this technology to microelectronic packaging. The height deviation of solder bumps has a significant effect on microelectronic packaging quality and reliability. In order to provide an efficient electronic contact, it is essential to ensure the solder bumps with a uniform height. However, it is difficult to print uniform height solder bumps directly, because the fluid dynamics and solidification behaviors have a significant effect on the bump formation and many factors can caused the height deviation. Wu and Hwang (2015) recorded the impacting process of a single solder droplet using high-speed camera, and found several surface ripples on the solder bumps. Haferl et al. (2001) provided a numerical method to study fluid flow phenomenon of solder droplets impacting onto a substrate, and indicated that the increases of impacting velocity caused the final bump shapes changed from a “ball” shape

292

W. Xiong et al. / Journal of Materials Processing Technology 243 (2017) 291–298 Table 1 Physical properties of Sn 63 wt% Pb.

Fig. 1. Schematic diagram of micro-droplet deposition experimental system.

to a “Christmas tree” shape. Waldvogel and Poulikakos (1997) presented a theoretical model to elucidate the solidification behavior of molten droplets on a substrate. Their results revealed that increasing substrate temperature prolonged the total solidification time, and in turns, affected the final bump shapes. Naidich (1981) performed wetting experiments using metal droplets on different material metal surface under different substrate temperatures, which showed that an increasing substrate temperature usually led to an increased wettability. However, the influence parameters on the bump height deviation and the method of printing uniform height solder bumps were not further discussed. This research aimed to understand the influence of impact parameters on the printed bump height deviation. In the present work, a series of deposition experiments of solder droplets onto a smooth metallic substrate were carried out using a selfdeveloped micro-droplet deposition experimental system. The dynamic formation of solder bumps was analyzed by calculating some dimensionless parameters theoretically and observing the solder droplet impacting process experimentally. Because the solidification behavior cannot be visualized, the droplet complete solidification time was estimated to characterize the solidification rate quantitatively, and then the effect of solidification on the bump height was obtained. Based on the above research, a method of reducing the bump height deviation was presented and proved by printing a solder bump array. A method of printing uniform height solder bumps was provided in this work. 2. Experimental approach The schematic diagram of micro-droplet deposition experimental system is shown in Fig. 1. It mainly consisted of a self-developed drop-on-demand (DOD) generator, a droplet deposition system, a low-oxygen environment control system, and a high speed image recording system. The self-developed DOD generator was used to produce uniform droplets. It included a crucible, a micro nozzle, a piezoelectric ceramic, a vibrant bar, a heater, a temperature controller (Shimax, Japan) and cooling water. The droplet ejection process can be summarized as follows: The metal material was melted in the crucible. A micro displacement was induced by the piezoelectric ceramic and

Density

d

8474.4

(Kg/m3)

Surface tension Dynamic viscosity Latent heat of fusion Specific heat capacity Thermal conductivity Melting point of metal

d d Ld Cd kd Tf

0.494 0.0013 47560 186.2 48 456

(N/m) (N s/m2) (J/kg) (J/(kg K)) (W/(m K)) (K)

transmitted to the nozzle by the vibrant bar. A velocity transient was caused by the micro displacement and resulted in ejecting a solder droplet from the nozzle. The droplet deposition system was used to deposit droplets at precise locations. It included a Program Multiple Axes Controller (PMAC) (Delta Tau, America), an X-Y-Z motion platform, a substrate with a heater inside, and a temperature controller (Shimax, Japan). A copper substrate was chosen as the deposition surface. In order to remove the oxides and organic contaminants from the surface, the copper substrate was mechanically polished and cleaned before the experiment. A low-oxygen environment control system (Mikrouna, China) was used to maintain a low oxygen content below 20 ppm, so that the formation of oxides was hindered during the droplet generation, flight, and deposition process. The system was filled with inert gas (Nitrogen, 99.99 percent purity). Both the droplet generator and deposition system were located inside. The high speed image recording system was used to record images of impinging droplets during deposition. It included a highspeed CCD camera (MotionBLITZ cube1, Germany), a microscope (Computar, Japan) and a 100 W LED light. The high speed CCD camera was able to record 1000 images per second. Since the properties of Sn 63 wt% Pb alloy is close to those of the lead-free solder alloys, the lower cost Sn 63 wt% Pb alloy was used in the experiment to represent the solder materials. Its main properties were shown in Table 1. Before melted, the Sn 63 wt% Pb alloy was ground to remove the oxides from its surface. In the experiments, a sequence of pulse waveforms was applied to the piezoelectric ceramic, and each individual pulse created one droplet. Droplets were ejected at the frequency of 1 Hz and deposited onto the moving substrate. The droplet and substrate temperatures could be modified by temperature controller #1 and temperature controller #2, respectively. In addition, a tool microscope (Nikon MM400, Japan) was used to observe the final bump shape and measure the height of its contours. 3. Result and discussions 3.1. The dynamic formation process of solder bumps A spherical liquid-metal droplet was ejected from the nozzle, and then impacted on the substrate and solidified a bump. In order to analyze the reason for the bump height deviation, it is essential to achieve a comprehensive understanding of the dynamic deposition process. According to Schiaffino and Sonin (1997), the dynamic and thermal behaviors of droplets during impact can be characterized by some dimensionless parameters, which include Weber number (We), Ohnesorge number (Oh), Stefan number (Ste), Prandtl number (Pr), and melt superheat parameter (ˇ). Those dimensionless parameters are shown in Eqs. (1)–(5). We = Oh =

d Ud2 Dd d d



d d Dd

(1) (2)

W. Xiong et al. / Journal of Materials Processing Technology 243 (2017) 291–298

293

Fig. 2. Images of solder droplet ejecting from a nozzle, with the initial velocity of 0.54m/s. The images were captured by the high-speed CCD with time interval of 1 ms.

Ste = Pr =

Cd (Tm − Ts ) Ld d Cd kd

ˇ=

Td −Tf Tf −Tt

(3)

(4) , (5)where Ud is the impacting velocity. Dd denotes the

droplet diameter before impacting. Td is the initial droplet temperature, Ts is the substrate temperature. Other parameters are shown in Table 1. The We number depends on the droplet diameter Dd and impacting velocity Ud . Fig. 2 shows the images of droplet generation process. A 411 ␮m-diameter droplet ejected from the nozzle at a velocity of 0.54 m/s. Luo et al. (2012) and Qi et al. (2011) purposed the dynamic droplet model to predicted the droplet impacting velocity Ud , which is shown in Eqs. (6)–(8). dUd 1 1 = Dd3 d g − 2Cd Dd2 g Ud |Ud | Dd3 d 6 6 dt Cd =

24 6 + 0.4 + √ Re (1 + Re)

(6) (7)

Fig. 3. Prediction of droplet velocity Ud and Weber number We as a function of deposition distance S.

 S=

t

ud dt

(8)

0

where g is the gravitational acceleration vector. Cd is the drag coefficient, expressed by Eq. (7), Re = Ud Dd g /g is droplet Reynolds number. g and g are the density and dynamic viscosity of the protective gas, which is equal to 1.1452 km/m3 and 2.125 × 10−5 N s/m2 , respectively. S is the distance between the nozzle and the substrate surface. In the experiment, S was set between 1–40 mm to assure a high location accuracy. Fig. 3 shows the prediction of Ud and We number as a function of deposition distance S. We = 2.2–7.3 > 1, certainly not a low value. In addition, the Oh number in the experiment was about 0.000991. With this combination of We and Oh number, Schiaffino and Sonin (1997) showed that the deposition behavior pertained to the region where the droplet impingement was driven by the

Fig. 4. Images of droplet impinging on a copper substrate. The droplet kept alternate spreading and recoiling in order to balance the kinetic energy and the surface energy.

294

W. Xiong et al. / Journal of Materials Processing Technology 243 (2017) 291–298

Fig. 5. The time variation of dimensionless height ␰(t) during deposition. The oscillation of ␰(t) was occurred and damped by viscosity as well as solidification.

Fig. 7. Variation of total solidification time ␶sol with initial droplet temperature Td and substrate temperature Ts. ␶sol increases monotonically with the increasing of Td and Ts.

tinuously due to the initial kinetic energy lost by viscosity and solidification. Thus, the underdamped oscillation of droplets was ineluctable due to the combination of We and Oh number in the experiment. And this oscillation was the main reason for forming different bump shapes. The different bump shapes under different solidification rates would be further discussed in the following.

3.2. The effect of solidification on the final bump shapes The solder droplet spread with solidification behavior. A radio of spread and solidification time was expressed by Schiaffino and Sonin (1997), as shown in Eq. (10). Fig. 6. SME image of solder bump. Several small ripples was formed in the middle of the bump surface.

spr OhSte = sol Pr

dynamic pressure of impact and resisted primarily by inertia. An underdamped oscillation would occur after rapid spreading and be damped out by viscous effects. As can be seen in Fig. 4, a droplet impacted on the substrate at a velocity of 1m/s. the We and Oh number were about 7.05 and 0.000991, respectively. The droplet kept alternate spreading and recoiling in order to balance the kinetic energy and the surface energy (Fig. 4(b)–(f)). And the oscillation would stop when the kinetic energy was fully dissipated (Fig. 4 (f)–(g)). Dykhuizen (2016) used a dimensionless height (t) = h(t)/Dd to characterize the droplet height variation during deposition, and the variation of (t) with time during deposition can be described as Eq. (9):

where  spr is the time scale required for complete spread, and  sol is the total solidification time. In the experiment, the initial droplet temperature Td and substrate temperature Ts were 643 K and 313 K, respectively. According to Eqs. (3) and (4), the Ste and Pr number were 0.5599 and 0.005, respectively.  spr / sol = 0.11 indicated the solder droplet solidification was 9.09 times longer than its initial spreading. However, based on Schiaffino and Sonin (1997), the local solidification at the contact line always occurred and hampered the droplet spreading, and after the first oscillation period the solidification effects become visible, which played an important role in formation of the final bump shape. Fig. 6 shows a solder bump under scanning electron microscopy (SEM), as a typical example to express the solidification effects. A series of ripples in the middle of the surface were the direct result of the combined effect of droplet oscillation and solidification behaviors. While the top surface was smooth, elucidating that the oscillation was finished before the droplet solidified completely. Schiaffino and Sonin (1997) ignored the contact thermal resistance between the droplet and the substrate, and estimated the droplet total solidification time by Eqs. (11).

t

 (t) = 1 −

t0

v (t) dt Dd

(9)

where v(t), as a function of the deposition time t, is the instantaneous velocity of the droplet highest point. t0 is the time point when the droplet first contacted the substrate. Fig. 5 shows the time variation of dimensionless height (t) during deposition. The (t) oscillated and subsequently tended to a static value ((t)∞ = 0.75) after impacting the substrate. The oscillation of (t) was caused by the repeated changing of u(t), and the amplitude decreased con-

sol =

Dd2 4˛

 1 Ste

(10)

+ˇ

 (11)

W. Xiong et al. / Journal of Materials Processing Technology 243 (2017) 291–298

295

Fig. 8. Final bump shapes under different solidification rates, which can be divided into three types: “bulb”, “hat”, and “hemisphere”.

where ˛=kd /d Cd is the thermal diffusivity of solder droplet. Combining Eqs. (3) and (5),  sol can also be calculated by Eq. (12).

sol

 C D 2 = d d d 4kd

 

Ld

Cd Tf − Ts

+

Td − Tf Tf − Ts

(12)

Hence,  sol was determined by Td and Ts . In the experiment, Dd was about 411 ␮m. The variation of  sol with Td and Ts is shown in Fig. 7. Increasing the initial droplet temperature and substrate temperature prolongs the total solidification time  sol , and in turn, affects the final bump shape. Fig. 8 shows the final bump shapes under different solidification rates, which can be divided into three types: “bulb”, “hat”, and “hemisphere”. Fig. 8(a) shows the geometric contour of the bump that takes the “bulb” shape, in which the contact diameter is less than the maximum diameter of the upper part. The reason for forming this shape is that the rapid

local solidification ( sol = 4.2 ms) at the contact line hampered the spreading of the droplet, and resulting in the less contact diameter. Fig. 8(b) shows the geometric contour of the bump that takes the “hat” shape. A convex edge was formed at the contact line, which was caused by the longer solidification time ( sol = 4.9 ms) and the droplet spreading at the contact line. Fig. 8(c)–(f) show the geometric profile of bumps that takes the “hemisphere” shape. The reason is that the liquid metal flowing away from the solidified metal during the solidification process due to the slower solidification rate ( sol = 5.7 ms, 6.6 ms, 10.1 ms, and 27.6 ms, respectively). The convex edge was covered, and the final shape was formed like “hemisphere” due to the surface tension. Fig. 9 shows the variation between the bump height h and the dimensionless bump height h/Dd with the total solidification time  sol . A clear threshold which divides the regions of the large and small deviation of the solder bump height can be observed in the

296

W. Xiong et al. / Journal of Materials Processing Technology 243 (2017) 291–298

(13) and (14) on the basis of the statistical result from about 150 droplets. 1 hi n n

h=

(13)

i=1

=

1 hi − h n n

12 (14)

i=1

Fig. 9. The relationship between the bump height h and total solidification time ␶sol. A clear threshold in the range of 6.0 ms–8.9 ms, dividing the regions of the large and small deviation of the solder bump height, can be observed.

Table 2 Process parameters of the single-factor experiment. Num

Initial droplet temperature Td (K)

Substrate temperature Ts (K)

#1 #2

543, 583, 623, 663, 703, 743 623

353 353, 383, 413, 443, 473, 503

range of 6.0 ms to 8.9 ms h decreased sharply when  sol was less than 6.0 ms, and the final bump shape varied from “bulb” to “hat” and “hemisphere” with the increasing of  sol . It indicated the fact that the solidification had a marked effect on the bump height under this condition. While h remained stable and all the bumps took the “hemisphere” shape when  sol was longer than 8.9 ms, h/Dd also stable at 0.67 at this condition, which showed that the effect of solidification was not obvious.

3.3. The method to reduce the height deviation of solder bumps In order to further investigate the influence of impact parameters on the bump height deviation, a series of experiments was performed under the conditions listed in Table 2. The average h and standard deviation  of bump height were calculated by Eqs.

Fig. 10 shows the variation of h and  with initial droplet temperature Td and substrate temperature Ts . The results illustrated the fact that the bump height deviation was reduced by increasing the Td and Ts . An example of solder bumps under the lower and higher temperature is shown in Fig. 11. A variety of bump shapes were formed at the lower temperature (Td = 583 K, Ts = 353 K), as shown in Fig. 11(a). The height deviation reached ± 22.5 ␮m. While all the bump shapes were “hemisphere” at the higher temperature (Td = 703 K, Ts = 353 K), as shown in Fig. 11(b). The height deviation reached ± 5 ␮m. The reason is that the solidification rate of each droplet was not identical during deposition, because the substrate roughness was not uniform and other random factors. The lower temperature resulted in a marked influence of solidification on bump height. The solidification error among the bumps led to a variety of contour shapes of the bumps, and further led to the larger height deviation. Meanwhile, the higher temperature resulted in a little influence of solidification on bump height. The solidification error among bumps did not lead to an obvious change of the height, although it still existed. While, h and  increased suddenly when Td reached 747 K or Ts reached 503 K. The reason for this phenomenon is that the solder droplet rebounded away from the substrate. Because of the higher Td and Ts , the local solidification at the contact line was not formed. The droplet was easy to recoil or bounce and then solidified into a nearly sphere, resulting in the poor deposition accuracy and the large bump height deviation. Based on the above discussion, the printed bump height deviation can be suppressed by prolonging the droplet total solidification time. However, the initial droplet temperature and the substrate temperature have to be limited to avoid the droplet rebounding during deposition. To solve this problem, the “deposition and reheating” process was presented in our test. The solder droplets were first deposited at the lower initial droplet and substrate temperatures to avoid rebounding behavior, and then the printed

Fig. 10. Variation of average h and standard deviation ␴ with the initial droplet temperature Td and substrate temperature Ts h and ␴ decreased with the increasing of Td and T, but hand ␴ also increased suddenly when Td reached 743 K or Ts reached 503 K.

W. Xiong et al. / Journal of Materials Processing Technology 243 (2017) 291–298

297

Fig. 11. The final bump shapes under different conditions: (a) at the lower temperature, where the final shapes included “hat” and “hemisphere”, the height deviation reached ± 22. 5 ␮m; (b) at the higher temperature, where all of the final shapes were “hemisphere”, and the height deviation reached ± 5 ␮m.

Fig. 12. The solder dump array, with a high deposition accuracy and a uniform height, was printed through “deposition and reheating” process.

bumps were reheated to suppress the height deviation. Fig. 12 shows the solder bump array printed through “deposition and reheating” process. The solder droplets were first ejected at the

initial droplet temperature of 623 K since the most stable droplet generation was achieved at this temperature. And those droplets were deposited on the substrate, with temperature ranging from

298

W. Xiong et al. / Journal of Materials Processing Technology 243 (2017) 291–298

313 K (room temperature) to 443 K to avoid the droplet rebounding behavior. Then printed bumps were heated up to a temperature slightly higher than its melting point. After the metal bumps were melted, the liquid state was maintained for ∼3 min to ensure the totally deformation of metal droplets. All these processes were carried out in the inert gas environment to avoid oxidization of liquid metal. The height deviation reached 223 ± 2 ␮m and the dimensionless height deviation (h/h) was remarkably less than 1%. These phenomena clearly proved that “deposition and reheating” is an effective method for printed the uniform height solder bumps. 4. Conclusions (1) According to theoretical analysis and experiment, the underdamped oscillation of solder droplets was ineluctable during deposition because of the high We number and low Oh number in our experiment. And this oscillation before complete solidification was the main reason for forming different bump shapes. (2) Three types of final bump shapes were formed under different solidification rates. A clear threshold, dividing the regions of the large and small deviation of the solder bump height, was found. The solidification had a marked influence on the final bump height when the total solidification time was less than 6.0 ms, while the effect was weak and the dimensionless bump height h/Dd was stable at 0.67 when the total solidification time was longer than 8.9 ms. (3) “Deposition and reheating” process is an effective method of printing the uniform height solder bumps. A solder bump array with the height of 223 ± 2 ␮m was printed through this process, the dimensionless height deviation (h/h) was remarkably less than 1%. Acknowledgments This work was supported by the National Natural Science Foundation of China (No. 51675436), the Natural Science Basic Research

Plan in Shaanxi Province of China (No. 2015JQ5123), the Seed Foundation of Innovation and Creation for Graduate Students in NPU (Z2016015), and NPU Graduate Research Fellowship. References Amirzadeh, A., Raessi, M., Chandra, S., 2013. Producing molten metal droplets smaller than the nozzle diameter using a pneumatic drop-on-demand generator. Exp. Therm. Fluid Sci. 47, 26–33. Dykhuizen, R.C., 2016. Review of impact and solidification of molten thermal spray droplets. J. Therm. Spray Technol. 3 (4), 351–361. Fan, K.C., Chen, J.Y., Wang, C.H., Pan, W.C., 2008. Development of a drop-on-demand droplet generator for one-drop-fill technology. Sensor Actuat. A: Phys. 147 (2), 649–655. Haferl, S., Butty, V., Poulikakos, D., Boomsma, K., Megaridis, C.M., Nayagam, V., 2001. Freezing dynamics of molten solder droplets impacting onto flat substrates in reduced gravity. Int. J. Heat Mass Transfer 44 (18), 3513–3528. He, L., Wei, X., Zhang, H., Chen, P., Luo, H., Ma, M., Liu, S., Shu, X.-Y., 2014. Drop-on-demand jetting of piezoelectric diaphragm-piston for fabricating precision solder bumps. The 15th IEEE International Conference on Electronic Packaging Technology (ICEPT), 10–13. Lee, T.M., Kang, T.G., Yang, J.S., Jo, J., Kim, K.Y., Choi, B.O., Kim, D.S., 2008. Drop-on-demand solder droplet jetting system for fabricating microstructure. IEE Trans. Electron. Packag. Manuf. 31, 201–210. Liu, Q., Orme, M., 2001. High precision solder droplet printing technology and the state-of-the-art. J. Mater. Process. Technol. 115, 271–283. Luo, J., Qi, L.-H., Zhong, S.-Y., Zhou, J.-M., Li, H.-J., 2012. Printing solder droplets for micro devices packages using pneumatic drop-on-demand (DOD) technique. J. Mater. Process. Technol. 212, 2066–2073. Luo, J., Qi, L.-H., Tao, Y., Ma, Q., Visser, C.W., 2016. Impact-driven ejection of micro metal droplets on-demand. Int. J. Mach. Tools Manuf. 106, 67–74. Naidich, J.V., 1981. The wettability of solids by liquid metals. Progress Surf. Membr. Sci. 14, 353–484. Orme, M., Willis, K., Nguyen, T.V., 1993. Droplet patterns from capillary stream breakup. Phys. Fluids A Fluid Dyn. 5 (1), 80–90. Qi, L.H., Luo, J., Zhou, J.-M., Hou, X.-H., Li, H.-J., 2011. Predication and measurement of deflected trajectory and temperature history of uniform metal droplets in microstructures fabrication. Int. J. Adv. Manuf. Technol. 55 (9–12), 997–1006. Schiaffino, S., Sonin, A.A., 1997. Molten droplet deposition and solidification at low Weber numbers. Phys. Fluids 9 (11), 3172–3187. Waldvogel, J.M., Poulikakos, D., 1997. Solidification phenomena in picoliter size solder droplet deposition on a composite substrate. Int. J. Heat Mass Transfer 40 (2), 295–309. Wu, C.-H., Hwang, W.S., 2015. Study of solder jet bumping process using high-speed digital camera. Mater. Sci. Semicond. Process. 31, 38–43.