"Fingerprint?" - A human fingerprint identifies a specific individual, but it doesn't provide detailed information (their height, hair color
). Factories have a similar need for fingerprints: Their day-to-day goal is to make sure each item made matches the standard product design. They could check this match by measuring a few dozen characteristics of each item. However, this would add significant delay and cost, and might require an expensive staff of "measurers." It would be better if one automatic measurement provided a comprehensive "fingerprint" that could simply be matched to that of the desired product design.
The above figure represents such a product fingerprint. It displays data collected automatically during the growth of a semiconductor device in our MBE system (see the previous image of a "resonant cavity photodetector"). A full explanation of the data is quite complex (it's given below for those of you who just HAVE to know). Here, the key points are: 1) as the device grew, data were filled in, scan line-by-line, from the bottom of image; 2) data from the growing device overwrote data from the standard, desired, reference device. At the point this image was taken, data for "today's" device had been collected up to the white line. Above the white line is the remaining data from the reference "golden" device. Without knowing anything else, you can see that the very complex data match well at the line. Thus, today's device is the same as the reference device. That's all the operator needs to know. Everything's fine, let the machines roll! If the data cease to match, it's time to ruin the engineer's evening off by calling him/her in to fix things. The operator does not have to understand the details. The operator does not have to spend a lot of time interpreting the measurements (and will thus probably be stuck with operating a whole bunch of machines simultaneously!).
OK, so you just have to know about the details: The image consists of 500 scan lines taken at 500 times during the growth of the device. At each time, the optical reflectivity of the growing layers was measured at 333 different wavelengths. If the device reflected strongly at a certain wavelength, this became a white pixel. If it reflected weakly at another wavelength, this became a dark pixel
This sequence of wavelength pixels became a scan line. The next set of reflectance measurements became the next scan line up, and so on. We used wavelengths at which the layers were partially transparent. Hence reflections occurred at both the growing crystal surface AND at the now buried boundaries between deeper layers. At each wavelength, the sum of these reflections could add or subtract from one another, leading to the complex oscillations you see. This particular device involved a series of identical layers (a "superlattice") that, as they grew, became more and more effective at reflecting light a certain specific wavelength. That's why the oscillations grow stronger but narrower as the device is completed at the top of the figure. Because data are taken at 500 times, at 333 wavelengths and they measure effects throughout the structure, they provide a very complete and precise "fingerprint" of the device's properties. Sound terribly complex? Yes, in some ways, but it was all done with the laptop PC I'm now using, some glass fibers, and a miniature solid state spectrometer built around a single optical integrated circuit (originally designed for fax machines). The entire measurement system cost less than the PC, yet it allowed us to grow the devices with a reproducibility of few tenths of a percent! (see the next image of that system)