To help wafer manufacturers perfect their polishing processes and monitor product quality, Hologenix (Huntington Beach, CA) sells the PBS-1000 Subsurface Defect Measurement System. While Hologenix is a sales organization, the system was designed at VTI Inc (Dayton, OH, (937) 427-1104), which chose a PCI-bus multifunction card from United Electronic Industries, Inc as a key component.

Figure 1. An industrial PC on the bottom rack equipped with an analog and digital I/O card from United Electronic Industries controls the various subsystems that implement the PBS-1000 subsurface defect measurement system.

To perform its nondestructive subsurface measurements, the system aims a laser probe beam at the wafer at a large angle of incidence, and it measures backscatter from internal defects. The bulk scatter from a perfect single crystal material is extremely low or nonexistent. In contrast, defects such as slips and dislocations form a discontinuity in the crystal structure with an associated change in the refractive index around the defect. This discontinuity causes the light to scatter. Because the bulk scatter is very low, any scattered light must be associated with some type of defect

Figure 2. In operation, the PBS-1000 shines a laser probe on the test wafer, while two photodiodes detect backscatter from subsurface defects.

Working off this principle, the PBS-1000 shines a 0.30-mm diameter spot from a 25 mW helium neon laser at 632.8 nm on the test surface (Fig 2). At this wavelength, the system can read defects 3 microns deep in silicon and 0.4 microns in gallium arsenide. The system steps across the wafer, which can be as large as 300 mm in diameter, with a typical test time of 20-30 minutes for a full wafer map. If the operator sees a region of interest, he can rescan with higher resolution to create a defect map with step sizes down to 10 microns.

Because the system shines polarized light on the surface, the backscatter is also polarized, and the strength of each polarization orientation (P or S) differs with the type of defect and its depth. The PBS-1000 uses a pair of photomultiplier detectors to read light of both polarizations simultaneously. The system is constantly stepping across the wafer to create a map of the entire surface, and thus the data-acq board-a PowerDAQ II PCI-bus card with a 300-kHz 16-bit A/D-must sample both tubes at exactly the same time. For this purpose it supplies dual sample/holds on the card itself, not requiring an external panel with those amplifiers. The operator can also adjust the detectors' sensitivity thanks to the board's dual analog outputs. Finally, the developer needed the onboard counter/timers to count pulses to keep track of the laser beam's location over the wafer.

Figure 3a. An area map of a wafer indicates obvious polish damage near the center. The red spots are surface pits as small as 20 nm in diameter.

The result appears as a defect map on the system's 17" color monitor. Fig 3a shows a mapping of a 150-mm silicon wafer after cleaning. Data for this image was taken every 0.25 mm in both the X and Y directions. The red areas in the middle show some of the damage that resulted from polishing.

Picture4

To take a closer look at the damage, the operator moved to high-resolution mode to examine a 20 mm x 20 mm area (Fig 3b). Here the system takes data every 0.05 mm in both directions to reveal greater detail. The important parts of this image are the small red spots, which are small pits, 20-40 nm in diameter, that result from subsurface damage in the wafer.

When VTI's president Fred Orazio, Jr, decided to upgrade the hardware in the system to make it smaller and more cost effective, he considered data-acq cards from a number of vendors. Those that had all the functionality he needed consisted of a mothercard and daughtercard combo, but he was looking to keep the hardware as clean as possible by minimizing the number of boards. When you consider that he wanted a 16-bit A/D with simultaneous sampling, analog outputs, digital I/O and counter/timers all on one card, "the universe of possible cards wasn't terribly large," says Fred, who has been quite pleased with the performance of the UEI board.

The only minor glitch that developed concerned support software. The PowerDAQ II 32-bit driver is well developed, notes Fred, but the 16-bit drivers that he needed because of the previous choice of operating environment are understandably not quite so sophisticated. Nonetheless, Fred was trying to upgrade the hardware in the PBS-1000 in a short period of time and thus didn't want to do any modifications to the system's existing software, which is written in C++. He did run across a few minor problems when trying to get the 16-bit driver to do exactly what he wanted, but the UEI support team helped to resolve the problems immediately.