A research team at the National Institute of Standards and Technology in Gaithersburg, Md., has determined that measurements of surface roughness taken using white-light interferometry can differ substantially from those taken using phase-shifting interferometry or stylus-based profiling. It has reported on the findings at recent conferences, and a paper describing the work is tentatively scheduled to appear in Applied Optics in October.The scientists have examined the measurements of surface roughness of various standards taken with five white-light interferometers produced by three manufacturers. They have found that the measurements of surfaces with an average roughness of 50 to 300 nm can differ by up to 80 percent from those taken with the other techniques, which generally deviated from each other by only about 6 nm or less. The error of white-light interferometry, which peaks at average surface roughnesses of 100 to 200 nm, appears to be unrelated to the specific instrument or to the shape or randomness of the surface profile.Part of the discrepancy, they suggest, may be the result of diffraction and could be managed using techniques that employ coherence and phase information to minimize the problems it can display in the measurement of step heights. More complicated sources of error may be involved, however, and the researchers are investigating theoretical models to explain the phenomenon.