Aspheric lens manufacture requires surface control

The size, weight, and both the number and cost of elements, are key parameters to optimize in an optical system. Cell-phone camera lenses, missile guidance systems, and extreme UV (EUV) projection optics are all designed around these and other parameters. New design solutions employ aspheric surfaces to simultaneously increase the lens performance while reducing the number and weight of optical elements. Deterministic surface-finishing techniques make flexible aspheric surface generation possible, and shorten the manufacturing time and cost of spherical lenses. Such techniques, however, have a major disadvantage: increased surface error in the mid-spatial frequency range from 0.5–1000mm^-1. Surface marks at these spatial frequencies degrade optical performance in both imaging and energy-delivery systems.

Manufacturing control of mid-spatial-frequency errors begins with metrology. The conventional technique uses a contact stylus that provides a line measurement of the surface. Mid-spatial-frequency error typically has directionality, which is called `lay'. To properly measure a surface with this property, the stylus must measure perpendicular to the lay direction. Deterministic surface finishing, however, typically produces spiral lay patterns that are hard to profile with a line (see Figure 1). Optical profilers therefore have a significant advantage over stylus tools in terms of speed, non-contact operation, and the ability to measure full-3D surface topography for mid-spatial-frequency information independent of the lay direction.

At Zygo, we have refined interference microscopy specifically for this kind of analysis. A new system, based on our NewView 6300, acquires data via scanning white-light interferometry coupled with Zygo's patented frequency domain analysis (FDA) method.¹ Most optical surfaces—especially spheres and aspheres—are curved surfaces: this means that long surface scans with very fine height resolution are necessary. These scans can be as long as 20μm, and yet require resolutions down to tens of picometers. Our FDA data acquisition provides this range of performance.

It is generally difficult to measure small surface features faithfully using imaging systems. Ours uses new fixed-focus magnification tubes optimized for low light scatter and minimal image distortion. The system achieves theoretical instrument transfer-function performance—that is, it transfers the image with optimum fidelity—and can be used to measure a wide range of optics, even those used at very-short EUV wavelengths.

The primary driver for this program was a need to measure free-form x-ray and EUV optical surfaces,² but the results are applicable to optical surfaces used for a broad range of applications. Mid-spatial frequencies scatter light at small angles, reducing image contrast. The effect scales as 1λ², making the problem 340 times worse at 13.4nm than at 248nm. Commercial optics manufacturers also want to control light scatter from mid-spatial frequency errors: by reducing errors to just below specification they minimize cost and maximize profits.

So, our instrument needed to have high repeatability, to be able to correct retrace errors, and to offer calibration of the instrument transfer function.

The instrument's short-term-repeatability goal was 50pm root mean square (RMS). The reference surface profile (reference) is defined by the average of many measurements and the repeatability is defined as the mean of the differences of a series of 10 measurements from the reference plus twice the standard deviation. Each measurement is comprised of, for example, 256 averages for a repeatability of 21pm RMS. This performance easily meets the EUV goal.

Retrace error exists in all interferometer systems in which a lateral mismatch occurs between the test and reference wavefronts: the mismatch causes the test and reference wavefronts to travel different paths in the interferometer. Our instrument compensates for these errors by first performing an extensive calibration on a known artifact. Once measured, any potential errors on production parts can be predicted and removed for up to 50 fringes of tilt, to less than 50pm.

Finally, we needed to know the instrument transfer function in order to measure the entire mid-spatial-frequency region. Full coverage of the 0.5–1000 mm^-1 region requires a sequence of measurements with several different magnification systems: specially-designed low-magnification objectives with curved reference surfaces cover the lowest of these frequencies,³ while 100× Mirau objectives resolve the higher spatial frequencies. The exact response of these objectives requires the measurement and knowledge of the instrument transfer function to correctly interpret the results.⁴

The fast-growing aspheric segment of optical manufacturing requires control of mid-spatial-frequency features to achieve the ultimate design performance. Zygo has developed an optical profiler that measures these features with picometer performance levels. This technology can be useful for commercial-quality optics. Future work in this area aims to broaden the spatial frequency range of measurement.