An improved method of infrared imaging of bulk defects in cadmium zinc telluride (CdZnTe) wafers has been developed. The method is intended primarily to be a means of identifying those portions of large CdZnTe wafers that are suitable to be "mined" for use in fabricating focal-plane arrays of photodetectors for x-ray and g-ray astronomy. Suitable portions are those that exhibit acceptably high degrees of uniformity of x-ray spectral response. The present method of infrared imaging is useful for identifying the suitable portions because, as described below, there is a correlation between (1) x-ray spectral responses and (2) infrared images of bulk defects that affect those responses.
Prior to the development of the present method, numerous investigators had used infrared-transmission imaging to document the distribution of bulk defects in CdZnTe. Incandescent lamps were used as the sources of radiation, and the infrared images were detected by silicon charge-coupled-device cameras operating at wavelengths just beyond the visible range. The present method is also one of infrared transmission imaging, but the wavelength range and the means of implementation are different.
Figure 1 schematically depicts the apparatus used in the present method. The source of radiation is a large-area black body at a temperature of 70 °C. The radiation detector is an infrared radiometer that operates in the wavelength range of 8 to 12 µm; it includes an HgCdTe photodetector cooled to 77 K by liquid nitrogen. A three-axis translation stage is used to manipulate a CdZnTe specimen wafer. Various lenses, including a microscope objective, are used to optimize images of defects.
During the development of the present method, experiments were performed to determine whether the infrared images produced by the apparatus described above could be used to identify the desired portions of CdZnTe wafers. In these experiments, CdZnTe specimen wafers of two different sizes (15 by 15 by 2 and 26.9 by 26.9 by 2 mm) were set up as planar photodetectors and exposed to a collimated beam of x rays from a 160-kV microfocus x-ray tube. The collimated beam was either 100, 250, or 500 µm wide. Each specimen was mounted on a computer-controlled, motorized translation stage and was translated in 100, 250, or 500 µm increments across the detector plane. At each increment of position, the CdZnTe detector output was processed into an x-ray-spectral response by a simple pulse-height-analysis system.
The bulk defects that can be seen in the infrared images include grain boundaries and twin boundaries decorated with tellurium inclusions, and pipelike voids. The results of the experiments show that there is a correlation between poor x-ray-spectral response and grain boundaries decorated with tellurium inclusions (see Figure 2).
It would be natural to ask why the infrared imaging method is preferable to generation of x-ray spectral images of wafers. The answer is simply that it would take a long time to scan a wafer [about 80 hours at 500-μm resolution for a 5-in. (127-mm)-diameter wafer] and most of that time would be wasted because of large defect densities encountered in practice. Instead, one could use the present infrared-imaging method to screen an entire wafer quickly to identify areas with acceptably low defect densities and dimensions large enough for fabricating photodetector arrays.
This work was done by Bradford Parker, J. Timothy Van Sant, Richard Mullinix, C. M. Stahle, A. M. Parsons, and J. Tueller of Goddard Space Flight Center, Bruno Munoz of Unisys Corp., S. D. Barthelmy of Universities Space Research Associates, and S. J. Snodgrass of Raytheon STX. For further information, access the Technical Support Package (TSP) free on-line at www.nasatech.com/tsp under the Physical Sciences category.