The fine geometries required by semiconductor manufacturers were beyond the reach of traditional water jets and their nozzle technologies. Though small aperture nozzles delivered sufficiently fine beams of water, the nozzle aperture would increase with use causing unacceptable deviations from target geometries. In addition, traditional water jets rely on the impact forces of high-energy water beams to erode material. Manufacturers with expensive clean rooms have been concerned about these high pressures, since a relatively small leak at 40,000 psi can be devastating. Some water jets operate at lower pressures by employing an abrasive mixed with the water; however these can only provide cut widths down to 0.5mm.
The cut beams of abrasive water jets have traditionally been difficult to control. As dry abrasive is introduced into the pressurized water stream, a large amount of air is also introduced. This air destroys any hope of generating a consistent and dense coherent beam of water. The resulting spreading beam cannot produce the small cut widths or the 25 micron tolerance required in semiconductor singulation. This new wafer-jet method relies on abrasive to remove material and operates at relatively low pressure. In fact, there is so much abrasive within the cutting beam that "water jet" is almost a misnomer, since the technology's main attribute is very fine diameter cut beams.
The cut beam can be as small as 50 microns and does not suffer the variability problems of traditional water jets. Two novel solutions have been applied in this process: Wet abrasive mixing and long-life nozzles. To maintain 50 micron cut beams, the pressurized water stream must be completely devoid of air. As abrasive is introduced to the water-jet unit, it is first soaked with water at ambient pressure. The wet abrasive is then introduced to the mixing chamber and exposed to high-pressure water (Figure 2). The abrasive never contacts the pump or its seals. Once the abrasive/water mixture is pressurized, it moves through large diameter, high-pressure tubing to the nozzle manifold.
High-Speed Laminar Cut Beam The large tube diameter enables the system to move large volumes of pressurized slurry at very low speeds, preventing wear to the tubing, manifold, and joints. "Squeezing" the slurry through a small nozzle causes it to exit the nozzle in a very high-speed laminar cut beam. The size of the cut beam depends solely on the size of the nozzle aperture. The nozzle aperture does not widen during use because the exiting beam is kept laminar and straight by the lack of air in the pressurized stream and the length of the nozzle. By designing the nozzle length to match the abrasive size and the desired beam diameter, the pressurized slurry is "collimated" and must proceed through the nozzle in an orderly and predictable manner. Furthermore, the development of low cost, high-hardness nozzles, and the miniscule side impacts of the slurry minimize wear at the nozzle exit (Figure 3). Abrasive water-jet technology enables the use of QFN and photonic singulation. The cutting beam interacts with a substrate only along the vertical axis preventing the formation of shear forces. Therefore, devices are retained in their intended position and cut geometries remain consistent. Another benefit of this water and slurry-based method is the continual renewal of inexpensive abrasive (Al2O3 or garnet).
Abrasive Material Is Reusable The abrasive is never "dulled" by ductile or compliant materials. The process remains inexpensive and robust, even when singulating laminates of very dissimilar materials. Finally, a single nozzle acts as a point source for cutting, thus, enabling curvilinear cut paths for photonics. A first generation lab model of the microJet was employed to obtain the data in the table. Production data of a third generation unit should be available in Q4, 2002. Conclusion The laboratory results shown in the table justify a full production test of this new singulation technology. Both BGA and QFN sites have been identified, and the results are available on request. In addition, experiments are now underway to support tape-BGA and GaAs singulation. Acknowledgements The authors thank Shafi Islam and Advanced Interconnect Technologies, Batam Island, Indonesia, for invaluable support with this project.
|
Copyright © 2002