线锯切割硅片的工艺技术之英文版

Wafer Slicing and Wire Saw Manufacturing TechnologyI. Kao PIand V. Prasad, J. Li, M. BhagavatDepartment of Mechanical Engineering, SUNY Stony Brook, NY 11794-2300Abstract Wire saw, with its ability to cut very thin wafers from large diameter crystalline ingots of semiconductor materials,has emerged as a leading technology for wafer production in semiconductor and photovoltaic industry. Nevertheless, the wiresaw cutting process remains lacking a theoretical methodology and is not properly understood. The modern times compulsionof more accurate and efficient manufacturing has made it imperative to understand this abrasive slurry cutting process and tooptimize it. As a sequel to this understanding, control tools can be developed to monitor the optimum process. In this paper,comparison are made between the wire saw and the inner diameter ID saw which has been used to cut semiconductor wafers.This comparison brings out superiority of wire saw over ID saw in most respects.Introduction Wire saw, operating on the Free Abrasive Machining FAM technique, is an emerging technologyfor large diameter thin crystal wafer production in semiconductor and photovoltaic PV industry. Its advantagesspan from producing very thin wafers with small kerf loss to high yield and productivity. Since medieval timeswire saws have been known to be used in cutting hard materials like granite slabs and other variety of stones.However, requirements on wire saw cutting in terms of wafer thickness and quality are very different in electronicsand photovoltaic applications than the traditional use of the process.An evaluation of wire saw cutting process has shown that it is a poorly understood phenomenon and no modelexists for simulation design and control. Even so, in crystal cutting applications, it has shown potential to producebetter surface finish and thinner wafers with much higher yield than ID saws. This, coupled with total absence ofany commercial US technology in this field makes this process worthy of detailed study.The primary objective of this initial study of the long-term research project is to evaluate the current wire sawslicing technology and to compare it with the ID saw technology most commonly used for wafer slicing. Such acomparison is expected to shed light on the superiority of wire saw over other conventional crystal sawingmethods, thus in turn bringing out the importance of concentrating efforts on the wire saw.Wire saw manufacturing processes Figure 1 shows the schematic of a wire saw. In the wire saw, a single strandof thin wire 175 μ m in diameter moves from a feed reel to a take-up reel. In between, the wire goes through theentrance side of storage system called the “ carriage ” and into the rectangular arrangement of fixed shafts withreplaceable wire guides. The wire wraps around the wire guides which have hundreds of grooves. This creates amultiple net of parallel wires known as web through which the ingot crystal is fed together with abrasive slurry toproduce a cut.wire storage carriagetachometertension unitlevel windTake-up reelFeed reelWebSpecimen Feed e.g. crystal Ingot slurry abrasiveShafts wire guidesAbrasive slurry SiC and diamond are mostcommonly used abrasives slurry also acts as coolantWire material stainless steel is generallyused typical diameter 150-300 μ mSpecimen can be fed from 4 sides ofof the web simultaneouslyFigure 1. Conceptual wire saw slicing setupThe current wire saw can cut up to four ingots simultaneously from four sides of the web, thus showing a vastimprovement in productivity over the ID saw. On account of the thin wire, the kerf loss in wire saw cuttingprocess is minimized. Additionally, the future use of large diameter crystals favors wire sawing because ingotdiameter capacity of wire saw is limited by only shaft spacing of wire guides of the web and the travel.Comparison between wire saw and ID saw Table 1 summarizes the principal differences between the wire sawand the ID saw.PROPERTY WIRE SAW ID SAWMethod of cutting Lapping GrindingTypical cut surface features Wire marks Chipping and fractureDepth of damage Uniform 10 to 15 μ m Variable 20 to 30μ mProductivity 300 to 2000 sq.in./hr 200 to 400 sq.in/hrNumber of wafers cut per run As high as 3200 wafers A single wafer per cutKerf loss per cut Typically 200 to 300 μ m From 300 to 500μ mMinimum thickness of cut wafer As low as 200μ m 300 μ mYield of 0.025 inch thick wafers 31 wafers per inch 27 wafers per inchMaximum diameter of cut crystal 300mm-diameter crystals Up to 200mm-diameterTable 1. Comparison between wire saw and ID sawThe FAM process of the wire saw produces considerably less depth of damage and more uniform surfaces thanthat in the grinding process of the ID saw. This results in low residual stresses in wire saw cut crystal wafers andmakes the wire saw more amenable to cutting of very thin wafers because the constructive interference of stressfields, when two cuts are made very close to each other, becomes less significant. Higher residual stresses are notdesirable since they lead to breakage of thin wafers in both the cutting process and post-process handling.In the PV applications where the post processing is not required, it is of cardinal importance to have leastdamaged wafer surfaces. The lesser surface damage, however, may not be as important to semiconductor industrywhere the sliced substrate is lapped, edged and polished well beyond the damage region. Nevertheless, smallerresidual stress in the substrate will make it easier to handle the cut wafers without breaking and to produce morehomogeneous wafers for the fabrication processes. The other field where a low damage on the cut surface is ofimportance is in slicing of optical material e.g., KTaO 3 crystals. Here residual damage to the material adjacentto the cut produces objectionable levels of birefringence in materials with high strain-optic coefficients.Also the well-known fact of getting a smooth surface by burnishing, wherein the working forces are much lessthan in polishing, makes wire saw more capable of giving good surface finish than the ID saw. The control of wiresaw in terms of wire stiffness and vibration and slurry concentration management can be facilitated by properlymodeling the manufacturing processes. Comparing to grinding and other cutting processes, the wire saw has muchless brute force during the cutting process and thus has a potential of more energetically efficient cutting. This alsocan be advantageous in the processes wherein the temperature rise of the crystal at the local contact areas is notpermissible from the metallurgical point of view due to possible change in its microstructure.Currently, the cost of consumables abrasives and slurry base for the wire saw operation is still higher thanthat of the ID saw. The cost and ease of process control of the ID saw gives it a slight edge over the current wiresaw. But this can mostly be attributed to total lack of methodical approach in research of wire saw manufacturingprocesses. Development of systematic modeling and control strategies can improve the wire saw manufacturingprocess and make it more cost-effective.Conclusion The objectives of the ongoing research in the modeling and control of wire saw are to providesystematic methodology to analyze and improve the wire saw manufacturing processes and to make it more cost-effective. Optical measurement techniques will also be introduced for in-situ and post-process measurements.Moreover, a US-based technology in wire saw manufacturing process and equipment technology will be developed.Acknowledgment The project has been supported by the NSF Grant DMI-9634889 1996--1999. We also like tothank the GT Equipment Technology, Inc., our industrial partner, for their supports in this project.References[1] R. K. Sahoo, V. Prasad, I. Kao, “ An Integrated Approach for Analysis and Design of Wafer Slicing by a Wire Saw.“ Toappear in ASME Transactions, Jl of Electronic Packaging.[2] Ray Wells, “Wire Saw Slicing of large Diameter Crystals“ Solid State Technology. September 1987. pp63-65[3] M. C. Huffstutler et al, “Wire Saw Cutting Technique for Optical Crystals“ Ceramics Bulletin 1966. Pp1098