John C. Stover

Calculation Of Surface Statistics From Light Scatter

The operation of a differential scatterometer developed at Montana State University is briefly described. The scatterometer takes and stores data under computer control. Analysis routines allow calculation of the surface power spectral density (PSD) function for the cases of one-dimensional surfaces [Z(x) - diamond-turned surfaces, for example] and isotropic two-dimensional surfaces [ Z(x,y) - polished surfaces, for example] . In addition, the zero and second moments of the PSD may be taken to provide bandwidth-limited values of the root mean square roughness (cr) and the root mean square slope (m). Results from several samples are used to check the vector perturbation theory [E. L. Church and J. M. Zavada, Appl. Opt. 14, 1788 (1975)] used by the computer to relate the scatter distribution function to the PSD. These experiments take advantage of the fact that the surface - and hence its PSD - remain a constant function during the measurements. Variations in the incident angle and polarization are introduced, and the resulting PSDs are calculated and compared. In another experiment, the min/max scatter angles (or, conversely, the min/max PSD spatial frequencies) are matched to those of a total integrated scatter (TIS) system. Integration over the light scatter data and the PSD allows direct comparison to the TIS and effective rms roughness obtained by the TIS system.

Modeling scatter from silicon wafer features based on discrete sources method

The discrete sources method (DSM) is extended to investi- gate complete mathematical models for light scattering from various kinds of silicon wafers features. Computer simulations demonstrate the usefulness of the DSM for comparing scatter signals associated with different surface features. Wafer features such as contaminating par- ticles, subsurface defects and pits are investigated.

Accurate sizing of deposited PSL spheres from light scatter measurements

This paper reviews a light scatter technique used to size particle depositions of polystyrene latex spheres on silicon wafers. The technique has proved to provide accurate (approximately 1% uncertainty) sizing of PSL sphere depositions. Measurements were made of NIST Standard Reference Materials as a means of checking the technique and an uncertainty analysis was performed using the techniques prescribed by NIST. The technique has the advantage that measurements are made of PSL spheres depositions on a substrate, which is the way in which they are used to calibrate defect scanners. In addition to presenting the details of the sizing technique the paper also discusses implications for scanner calibration accuracy.

Measurement and analysis of scatter from silicon wafers

The measurement of wafer surface roughness has become of increasing interest in the semiconductor industry in the last year. This interest is driven by the need to reduce background `haze associated with laser scanning particle counters, by the ever decreasing linewidth requirements and by a recent report that gate oxide breakdown voltage decreases as roughness increases. Scatter measurement offers the potential of being a fast, non-contact method of monitoring roughness; however, the ability to accurately calculate wafer roughness via scatter depends on various wafer surface characteristics. The paper presents data taken on a number of wafers and demonstrates that bare silicon scatters almost exclusively from surface topography and is isotropic over an appropriate spatial bandwidth, thus facilitating scatter measurement. Because silicon proved to be an excellent source of topographic scatter, an experiment was undertaken to compare the one dimensional roughness measurements made with an optical profilometer to the two dimensional (area) measurements made via scatter, and these results are also reported.

Design review of a multiwavelength, three-dimensional scatterometer

A design review of a five wavelength, laser-based, full reflection hemispherical BRDF instrument is given. A source box design which provides full polarization control of incident radiation is described. The receiver design uses a periscope mirror which allows near (< 1 .5°) retroreflection measurements with full received polarization control. Replacement of the periscope mirror with a beam splitter pellicle allows 0° retroreflection measurements. The goniometer design shown allows the measurement of full reflection hemisphere scatter with transmission measurements possible. A new software control approach which allows a user to define scans in scatter space without being concerned with mechanical axes motion is described.

Measurement Of Low Angle Scatter

Abstract. A low angle scatter instrument (LASI) has been designed and built at Montana State University under contract to the Los Alamos National Labora-tory. The instrument is capable of measuring scatter at 0.1 ° from the specular beam on most samples and at 0.01 on high scatter samples. Samples may be either reflective or transmissive components. LASI consists of a small aperture detector (100 Am nominal) that is scanned through a focused beam under computer control. The system measures light intensity over approximately nine orders of magnitude. Because the instrument produces some scatter, scans are made with and without the sample present. Following a normalization process, the two data sets are compared to determine the scatter due to the sample alone. Extensive software is used to control the data-taking process, analyze the data, and present the results in graphical form. In this paper a review of the instrument is presented, rather than an extensive survey of scattering data. The basic design problems involved in making these types of instruments are discussed, and procedures for taking measurements are given. We conclude that the major constraint on this type of measurement is scatter from the instrument itself. The difficulties involved in obtaining optics with small amounts of low angle scatter emphasize the need for this type of instrumentation.

Measurement and analysis of scatter from rough surfaces

The technique of characterizing the surface roughness of smooth, clean, front surface reflectors from BRDF measurements is in common use and has been well documented in the literature. It relies on using the Rayleigh-Rice vector perturbation theory to obtain the surface power spectral density function (or PSD) from which surface statistics may be found. These calculations are restricted to smooth surfaces only, as defined by the Rayleigh criterion. A number of potential industrial applications that would benefit from the fast, non-contact aspects of scatter/roughness characterization have not been implemented because the smooth surface restriction is clearly violated. This paper offers a brief arm-waving explanation of why BRDF cannot be used to obtain the PSD for rough surfaces, explores increasing the wavelength and/or the incident angle as measurement techniques to extend the PSD (from BRDF) calculation to rougher surfaces, and presents data showing that TIS measurements (which do not require that the PSD be calculation) can be used to find the rms roughness of surfaces are less severely limited by the impact of the smooth surface criterion.

Design review of an instrument to map low-level hydrocarbon contamination

An instrument capable of measuring and mapping hydrocarbon contamination densities as low as 3 milligrams/square foot has been developed for use in pre-bonding inspection of spacecraft components. The instrument makes use of the 3.4 micrometer absorption band present in hydrocarbons. Measurements above, one, and below the band are made to compensate for effects due to variations in substrate roughness. The instrument, dubbed SurfMap- IRTM, has been tested on substrates of aluminum and steel.

Wavelength scaling investigation of several materials

A logical approach to investigating materials that scatter non-topographically is to coat them with a thin layer that scatters only topographically (i.e., a layer that wavelength scales) so that the two scatter sources can be separated. This technique has been used to study scatter from beryllium mirrors using layers of aluminum and gold. In the course of these investigations it was learned that many aluminum surfaces do not wavelength scale (no information was available on gold). This was discovered fairly late in the study and caused a lot of extra expense and frustration. This paper presents wavelength scaling data from the uv to the mid-IR for several materials. The intention is to provide guidance for future studies of this nature.

An Investigation Of Anomalous Scatter From Beryllium Mirrors

Scatter from Beryllium mirrors often seems to be higher than would be expected based on mirror surface roughness data. This paper verifies the effect and shows that it is more severe in the IR than in the visible. The implication is that the excess scatter is caused by mirror defects that are non-topographic in nature. The conclusion is that profile, or roughness specifications should not be used when low scatter is the requirement of concern.