Jet Dispense for Electronic Packaging Applications Horatio Quinones / Erik Fiske / Alec Babiarz / Lian Fang Asymtek Headquarters 2673 Carlsbad CA, 92008, USA Abstract The scale of components used has consistently grown smaller and denser with the advances and improvements in technology of the electronics and optoelectronic industry,. This has allowed for more efficient use of resources (a silicon chip that once held hundreds of transistors now holds millions) and more convenient and consumer friendly products (palm sized computers compared to warehouse sized computers forty years ago). However, as advances lead to decreases in the scale of components, similar advances must be made in the production process in order to assemble these components. When applying the Surface Mount Adhesives (SMA), for instance, it is vital that the adhesives do not cover any part of the electrical leads, as this may decrease the quality of the solder connection and increase the electrical resistance. As components decrease in size, so does the gap between the leads. It is therefore necessary for the size of the adhesive dot to decrease to the same degree as the components decrease. Current jet technology can produce dots of surface mount adhesive of approximately 225E-6 m or about 3.5E-9 liters of volume. The underfilling of 3D-packages with small gaps, yet large geometries, that are required to accomplish thin packages, as well as the underfilling of small die present yet another challenge to the consistency and accuracy of dispensing processes. The challenge of jetting abrasive materials is addressed here by monitoring wearout evolution on the jet itself and corresponding effects on the jetted fluid characteristics. Underfill jetting accuracy from small to large die, is presented, this volumetric consistency exceeds today most process requirements. This paper proposes a solution to this dispense challenge by way of the non-contact Jetting Underfill for both, traditional capillary flow and Forced Flow Underfilling, commonly known as No-flow Underfilling. An alternative to the traditional seal pass dispensing is proposed whereby dispensing time is shorten significantly and acceptable and reliable die fillet is accomplished by way of jetting techniques. Display panels also require fluid dispensing; this paper will address the advantages in non-contact and high throughput that can be obtained with jetting of UV materials at frequencies as high as 200 Hz. A model is proposed to simulate jetting process analytically along with some experimental data. Key words: ball-needle, jetting, non-contact, high viscosity, underfill, pre-applied Introduction Fluid dispensing on electronics allows for a flexible and robust process in a manufacturing environment. Contrary to other methods including injection molding, plating, screening, the needle and jetting heads transiting provides a readily high degree of flexibility for dispensing fluid. In addition to these, the jetting technology offers noninvasive process; the actual nozzle does not need to have contact or small proximity to the surface upon which the fluid is delivered. The jet used in this research was the Asymtek DJ- 9K series non-contact DispenseJet (figure 1). This DispenseJet is normally closed, positive shut-off, air-actuated with a return mechanism driven by an elastic spring which uses a needle and seat to expel precise volumes of fluids. Air pressurized is regulated by a solenoid to retract the needle assembly from the seat. Upon this air pressure release, the needle is pressed by the spring and travels a predetermined distance at velocities high
enough to create a sizable compressive elastic wave that propagates to the fluid meniscus at the nozzle orifice and causes separation of the fluid in a form of a droplet by the momentum transfer that carries forces larger than that of the surface tension of the fluid in the nozzle outlet. Several parameters are responsible for the process: fluid pressure, stroke length, time of the needle at the open position (valve-ontime). Material properties and valve geometry are also instrumental in defining a robust manufacturing process: material viscosity, nozzle and seat geometries and thixotropic index, to name a few. Theory A typical signal trace for a jetting process is depicted in figure2. There on can observe the delay in the response of the solenoid after triggering. This delay time is somewhat more pronounced for the very first actuation or actuations taking place after some waiting time. The result may be of a somewhat smaller volume droplet. This time retardation i.e., time for the needle to start moving includes solenoid time response and pressure buildup until overcomes the spring force and system frictions. Trace (a) in figure 2 depicts the solenoid input signal, about 10E-3 seconds duration. c b a Figure 2. Jet traces: (a) Solenoid signal; (b) Needle displacement; (c) Air cavity pressure. Figure 1. Diagram of the cross section of Asymtek DJ-9K Jet. There are several parameters that can be adjusted in order to alter the size of the dots produced. The parameter adjustments affect the dots in two different ways, either by changing properties of the fluid, or by changing mechanical properties of the jet fluid properties was the nozzle heater temperature setting. Throughout this paper when reference to temperature is given, the temperature is of the heater setting and not the exact temperature of the fluid. The fluid temperature is always a few degrees less, due to thermal conduction losses in the system. The pressure profile, of trace (c) of figure 2 shows an initial sharp increase to an initial flattening that coincides with the needle startup displacement, trace (b) of figure 2. In this particular jet configuration, that time is about 2E-3 seconds, however the time to full open of the needle is about 3E-3 seconds. Given above system response, one can accomplish actuation frequencies of about 250 Hz, that is 4 milliseconds actuation time. Observation of decrease of displacement (trace (b) figure 2) one can conclude that the exhaust air or pressure decrease is occurring faster than the actual elongation of the spring that drives the needle. The flattening occurring towards the end of the pressure decrease can be attributed to that of the volume of air pressure remaining in the air
passages outside the air chamber and is irrelevant to the process for it does not affect the needle velocity. One can observe that the displacement, (trace c in figure 2), is zero which indicates that the actuation has ended by then. In this particular setting, the velocity of the needle is relatively constant during closing. From trace (b) one can infer velocities of about 1 to 3 meter per second. It should be noted that this velocity could be control by the spring constant. The role of the fluid on the velocity of the needle is of little significance. Jetting Volumetric Accuracy It has been reported [1] that volumes of a few nanoliters is possible for high viscosity fluid jetting. Variations of volume dispense could have several root causes. For those materials where the viscosity varies significantly with temperature and time, the challenge to deliver consistent volumes will ultimately depend on both, temperature management and periodic mass calibration [2]. Experimental data show that lower viscosity may result in volume increase as depicted in figure 3. There one can observe steady increase in volume jetted; temperature increase was caused by external source, i.e. table holding samples. Weight (mg) Heated Substrate 14 13 12 11 10 9 8 1 7 13 19 25 31 37 43 49 55 Unit Figure 3. Mass variation as function of temperature This temperature bias is a very common practice in traditional fluid dispensing as well as in jetting. Recall that low viscosity resulted from higher temperature results in faster fluid velocities and thereby possible faster manufacturing process. Figure 3 in addition to part-to-part variation, shows a consistent temperature increase bias due to gradual environment increase with time. Temperature variation of just a few degrees may cause large volumetric variations. Jetting requires implementation of dynamic temperature control. Ultimately the volumetric accuracy may just be as good as our temperature management. Figure 4 depicts jetting dispensing accuracy once a sound heat management process is in place. Mass (mg) 14 13 12 11 10 Jeting Test Active Air Cooling 9 8 0 10 20 30 40 50 60 70 Figure 4. Jetting mass variation under dynamic temperature control. The nature of the jetting process, i.e., discrete fluid dispensing in form of droplets lends itself for a method with high accuracy for a large range of volumes. The volumetric accuracy is based on that of the dot-to-dot variation. Typical dot size can be of just a fraction of a milligram. Ultimately, because the discrete dispensing, the accuracy for jetting small fluid volumes may be a function of our shot size. As mentioned above, for those materials where the viscosity may depend on time (polymerization or gelling), the implementation of periodic mass calibration may be a practical solution. Determination of shot size changes (mass per droplet), can be used to define number of droplets needed for a given desired mass of say dispended line, underfill, etc. Figure 5 shows the discretization effect when mass as small a 1-milligram is jetted.
1.3 1.2 1.1 1 0.9 0.8 Jetting Underfill Sampled Data 30C.174 mg/ dot 40C.219 mg/ dot 45C.282 mg/ dot 50C.382 mg/ dot in the fluid cavity, the fluid near the contact ball-needle-seat point of contact are very high indicating that some fluid compressibility could very well take place. Figure 6 shows the vector filed of fluid displacement during regular jetting operation. There one can see the fluid back motion away from the orifice as well as the displacement of the fluid toward the nozzle outlet. Figure 5. Discretization effect for jetting small volumes. Large masses can be jetted as well with high volumetric accuracy at it is shown in figure 6 below. Temperature was control within a couple of degrees throughout the entire jetting process. 50 45 40 35 30 25 20 Asymtek DJ-9K 0 10 20 30 40 5 Sampl es Figure 4. Large mass Jetting of abrasive underfill fluid. (Asymtek DJ-9K series) Incompressible fluid and non-slip theory were assumed in the numerical simulations. It was determined that the Reynolds number was very low and therefore turbulence conditions were not present (Re < 1). The ball-needle was accelerated by the spring and the transfer of momentum into the fluid causes the drop to form when surface tensions are overcome. The set of equations needed to define the problem consisted of the mass conservation equation. Calculations show that the fluid velocity at the orifice of the nozzle is more than an order of magnitude higher than that of the ball-needle itself. [1] For the pressure map 0 Figure 5. Fluid displacement mapping during regular jetting operation. Jetting Abrasive Materials It has always been a challenge to handle abrasive materials in various processes, jetting is no exception. The nature of most underfill materials used in microelectronic packaging is the presence of abrasive fillers such as silica. This hard material will cause wearout on those surfaces where it maintains contact. Figure 6 shows the wearout caused by continuous jetting actuations on the needle head surface. Figure 6. Needle head and shaft wearout caused by abrasive material. The functional failure will eventually be caused by the shaft wearout in a way of fluid leakage. The quality of the material jetted is not affected by this wearout. The enlargement
of the nozzle diameter due to the abrasiveness of the fluid tends to cause slight increase in the volume of the fluid droplet. Figure 7 shows the consistency of abrasive material jetting for several million actuations required for most manufacturing process. as wee. Volumes of less than 1 mg are required for small die. Figure 9 below show the jetting of less than 1mf for a die with less than 1 mm on the side and a small gap. Mal fillet is desired as well as high volumetric accuracy (<15% variation). DJ9K 15 mg Mass Flow Calibration Af ter 62 Mi l l i on Actuati ons 25 23 21 19 17 15 13 11 9 7 5 0 20 40 60 80 100 Sampl es Figure7. Volumetric consistency of abrasive fluid jetting. The ability of jetting fluids while the head is in motion, Jetting-On-the-Fly (JOF) allows for high fluid flow dispensing. This feature is of particular advantage when large fluid volumes are required as it is the case of large die, CSP and BGA underfilling. The jetting attribute of non-contact allows for fluid dispensing closer to the edge of the component as compare to that of regular needle dispensing. This results in smaller fillet and more homogeneous fillet around the component. Often, underfill jetting does not require seal pass. Figure 8 shows a large die underfilled with a jet, one can observe a very homogeneous fillet around, no seal pass was performed. Figure 8. Large die underfill by jetting without seal pass. (17x12 mm 2 ) The large range of jetting dispensing is of a great use when underfilling small components Figure 9. Small die jetting underfill. (<1 mm 2 ). Another characteristic of jetting underfill materials is the potential to form small fillet and thereby small wet-out-area (also referred as fluid tongue). This property is of most used when other components are next to the underfilled and are required to be fluid free, or simply uncontaminated. Proximity of the die is another instance where small fillets and tongues are required. Figure 10 depicts a large die with a very small wet-out-area, 250E-6 meters. The fillet in this instance is about 50E-5 meters. 250 µm Figure 10. Jetting underfill with small fillet, 50E-6 m, and small wet-out-area, 250E-6m. As described above, during fluid jetting the material is separated by momentum transfer and it does not require the mechanical pulling used in other dispensing methods including needle dispensing. This property allows for large gap dispensing, nozzle can placed be above component surface. The fluid stream of the droplet is in the order of magnitude of that of the nozzle orifice.
Pre-Applied Reinforcement Jetting All this advantages can be used to enable dispensing from large gaps and between small spaces. A good example of this jetting process is that of the pre-applied underfill or ball reinforcement underfill for bum technology. Asymtek has developed this process. Figure 11 depicts the ball reinforcement pre-applied underfill for a BGA substrate. The advantages of this process from reliability and rework ability are immense. Under ball metallurgy (UBM), may not need to be so robust, the mechanical role of the pre-applied underfill may allow weaker and cheaper interfaces. Figure 11. Pre-applied Bump reinforcement jetting for a BGA. It is possible to obtain different crown height by jetting. A long time problem of contamination of bumps by other dispensing methods is altogether solved by simply jetting small streams of fluid between the bumps. Findings Throughout this work several general trends were observed. Fluid cavity stream maps were determined by numerical analysis Volumetric accuracy of the jetting technology for both, non-abrasive and abrasive materials was very well established. Although wearout is observed in the jetting tooling for abrasive materials, such wearout evolution is relatively slow compared to other technologies. High flow rates for large components and large bump technologies can be accomplished with the jet. Fluid flow rates of more than 120 mg/s can be attained when JOF was used. Small fluid streams (100E-6 m), is a property of jetting that enable new applications such as pre-applied reinforcement. High dispensing gaps are possible when jetting (>2 mm). Volumetric accuracy and consistency was demonstrated for a large variety of materials and jetting settings. Traces of pressures, electrical signal and needle displacement define the lower limits for the frequency of droplet jetting. Zit is possible to obtain frequencies as high as 200 Hz for the Asymtek DJ-9K series jet. Controlling temperature during jetting resulted in high volumetric accuracy for materials where viscosity if highly dependent of temperature. Conclusions Numerical calculations for stream functions of the fluid inside the cavity agree with empirical observations very closely. Some materials may change properties as temperature varies and therefore could lead to variations on the volume and or mass jetted. The introduction of dynamic temperature control on jetting improved the volumetric for a large range of volumes accuracy to unprecedented level. A clear understanding of the needle displacement during jetting was possible by observing the traces of the cavity pressure, needle displacement and solenoid trigger signal. Jetting technology is an enable technology for pre-applied dispensing. Volumetric accuracy for a very large range is possible for jetting from the mere fact tat discrete droplet can be generated. Droplets from a few nanoliters to a few milliliters can be accomplished in jetting. Abrasive materials can be jetted with constancy for long times, several millions actuations. Line writing is improved, no dog-bone effect. Jetting virtually eliminates height sensing. No dripping occurs since a positive shut-off mechanism is in place. Dispensing through small spaces is possible by the nature of the droplet shape as ejected from the nozzle orifice.
Acknowledgement The authors would like to thank Mr. D. Fleming for work and data analysis he conducted for this study. Reference [1] H. Quinones et al, Jetting for Next Generation Packages, Pan Pacific Conference Proceeding February 2003. [2] Asymtek Mass Calibration, U.S. Patent number 5,906,682.