edfas.org 37 ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 4 areas, each structure has different distance to the incident surface anddifferent reflectance as shown in Fig. 3. This all has an effect on the reflected intensity to the point where some areas will have no measurable reflectivity. The black areas in Fig. 4will be very difficult tomeasure as there is little to no reflectance. Thismay requiremoving the point where the thickness is measured to get a valid measurement. MEASURING THE REMAINING SILICON THICKNESS The RST can be measured using either independent measurement systems or a measurement system that is integrated into the sample preparation system. There are advantages and disadvantages to both. A typical independent measurement system requires the operator to select the measurement point, adjust the optical focus and point position to get the best, most reliable reading. An experienced operator can get reliable measurements, but the measurement position will be loosely controlled, and a lot of operator time will be consumed. An automated measurement system, either external to, or incorporated into the polishing machine, can precisely control the measurement position, and will run without operator involvement. The problem is the reliability of the measurements in difficult circumstances. Measuring the thickness after grinding and coarse polishing steps can result in invalid or missing thickness measurements. Multiple measurements at slightly different positions and complex analysis software helps but is not foolproof. In any case, the thickness measurements are used to correct themechanicallymeasured profile for the generation of the tool path. EDGE EFFECTS FROM SURFACE SLOPE AND TOOL TRAVEL A lapping or polishing tool has a flat face and the tool can seldom travel completely off the die surface. There are usually surface mount components, capacitors, and filters which limit the travel of the tool beyond the die edge. Unless the tool inside edge reaches the die outside edge, the profile near the die edges will not be controlled. In addition to this, the higher the surface slope, the less tool surface is in contact with the die surface. This results in lower material removal from areas with higher slope. Unless there is a compensating algorithm in the tool path calculation, the highest slope areas of the die surface will end up the thickest after processing. This can easily result in edge areas much thicker than the die center. No tool path algorithm can compensate for a tool that does not travel beyond thedie edge. If there is less than the tool diameter clearance from the die edge to obstructions that restrict tool movement, the die edges will be thicker than the rest of the die. This edge distortion area will be equal to, or greater than thedistance the tool center travels beyond the die edge less the tool radius fromthe die edge. That is, if the tool center only moves to the die edge, the edge distortion will extend, at least, the tool radius from the die edges. The best case would be enough clearance and a tool pattern that moves the inside of the tool edge to the die edges. If this is not possible, which it never is, there will be edge thickness distortions where the edges of the die are thicker than desired. MULTIPLE THICKNESS CONTROL STEPS It may not be possible to reliably measure RST after the bulk removal of silicon by grinding. A grinding tool is very coarse and creates a rough surface with structural damage extending 50 to 100 microns below the surface. A quick very coarse lapping step can be added to the process, smoothing the surface and eliminating some of the damage. After this, some level of reliable RSTmeasurements can be made. Fig. 4 The very dark areas contain the highest density of active structures. The brightest areas are metallization. The picture is taken with visible light.
RkJQdWJsaXNoZXIy MTMyMzg5NA==