Field of the Invention
This invention relates to a surface analysis system and, more particularly, relates to analysis of a surface to determine the presence of particle contamination and/or defects on or below the surface of a material.
Background of the Invention
The use of optical detecting devices to determine the presence and/or size of particles is well known, and is shown, and such detecting devices have hertofore included lasers to provide the illuminating beam (see, for example, U.S. Pat. Nos. 3,406,289 and 4,571,079).
An optical surface scanning device, commonly referred to as a wafer scanner, has been heretofore utilized in semiconductor processing to identify surface adhering particulates. Semiconductor wafers are normally run through various manufacturing process steps (often as "virgin" wafer witness blanks) to determine the amount of contamination deposited on the wafer and assumed to be generated by the process. The process may be, for example, a photoresist coating process, a silicon oxide forming furnace operation, a photolithographic circuit masking process, or any of an array of others, but the wafer scanner has proved to be a useful tool to monitor any of them. Although independent techniques are also utilized to measure the microcontamination in various process fluids, the wafer scanner has nevertheless been utilized to provide monitoring where it directly effects the product, i.e., on the wafer itself.
As with any measuring tool, it is a fundamental requirement of apparatus measuring microcontamination to possess repeatability and accuracy. Repeatability or reproducibility must be inherent to achieve accuracy but places no guarantee of accuracy. Resolution implies an ability to separate measurement values of similar magnitude but again does not assure accuracy. Indeed, as with many types of instrumentation, the determination of performance with respect to resolution, repeatability, and accuracy can often be very difficult to assess and verify.
This proves to be the case with wafer scanners which appear to often generate an intolerable disparity of results when particles of known size were deposited on test wafer surfaces and analyzed. Considering that the contaminants are located in a thin plane (so that a shallow depth-of-field is required), the background light level is extremely low, and repetitive measurements are possible, it would appear that surface microcontamination measurements should be more easily accomplished than are aerosol measurements. However, aerosol instruments now known have far superior resolution than do now known wafer scanners.
A wafer scanner is, in reality, a fairly straightforward device, and three fundamental types of wafer scanners are generally known. In one, the wafer moves along a belt or other transport mechanism (usually stepper driven) and a laser beam is focused onto a small spot on the surface. The spot is dithered laterally by an oscillating mirror (or a rotating polygon), and the combination of the two motions generates a raster scan suitable for direct cathode ray tube display of coordinates illuminated on the wafer. The light reflected from the wafer is trapped or monitored for extinction, while the diffusely scattered light is analyzed for its bulk scattering (DC or low frequency changes in diffuse scattering) and point scattering events from particles or perhaps digs and scratches (AC or high frequency changes in diffuse scattering).
A second type of wafer scanner now generally known differs from the first above described type in that the wafer rotates on a chuck fixed to a transport stage, and light is incident and collected at fixed angles. The conversion of scanning polar coordinates to display cartesian coordinates can be accomplished easily in software or by electrical resolvers.
The third type of wafer scanner now known is the simplest optically. The wafer is totally illuminated by white light generated from a high intensity source such as an arc lamp. At an oblique angle, a vidicon (usually solid state) is positioned to view the wafer and particles appear directly on the vidicon output. This system is, however, the most complex in terms of analysis requiring post analysis of video data.
In all of the known wafer scanner systems, particles or defects are analyzed by measuring the amount of light scattered and compared to a calibration, or response, curve in a manner similar to most common aerosol or hydrosol counters.
Polystyrene latex microspheres (PSL) are the most desirable particle calibration material, although attempts have been made to generate defect patterns via microlithography to effect "calibration wafers". Particles cannot, however, now be simulated using pattern defects in any certain way, and the users and manufacturers of wafer scanners have found difficulty in using PSL particles as a preferred calibration material for a variety of reasons, including: PSL particles cannot be applied as a liquid suspension because no solvent is clean enough to evaporate without leaving a residue that could be confused with the PSL particles; application of dry PSL particles using standard aerosol nebulizing techniques require some other means of verifying that the particles would attach and "stick"; response sometimes appears to be a function of position; and the results generally appear nonmonotonic and thus ambiguous (i.e., smaller particles gave greater signals than larger ones). The last situation is one familiar to most researchers who have studied aerosol counters and most aerosol viewing geometries have to be tailored to achieve a monotonic response.
It is also the primary reason existing instruments have not adopted PSL as the calibration standard. Existing technology does not produce a monotonic calibration with uniform spherical particles.
As can be appreciated from the foregoing, an improved system and method could therefore be advantageously utilized for effecting analysis of the surface of an element to determine particle contamination and/or defects on or below the surface of a material.
Summary of the Invention
This invention provides an improved system and method for analyzing the surface of a material. Beams having different polarizations are directed at the surface and light scattered therefrom is collected and detected to produce electrical signal outputs representative of sensed contamination or defects. In the particularly described embodiment, a "P" polarized laser beam and an "S" polarized laser beam are generated and separatedly directed to the surface at a common point, and light scattered from both is commonly collected and then separated to provide electrical signals representative thereof. The electrical signals are then processed to provide an output indicative of contamination and/or defects.
It is therefore an object of this invention to provide an improved system and method for surface analysis.
It is still another object of this invention to provide an improved system and method for surface analysis utilizing beams having different polarizations.
It is still another object of this invention to provide an improved system and method for surface analysis utilizing polarized beams one of which is a "P" polarized laser beam and the other of which is an "S" polarized laser beam.
It is still another object of this invention to provide an improved system and method for surface analysis for sensing particle contamination and/or defects on or below the surface of a material.
With these and other objects in view, which will become apparent to one skilled in the art as the description proceeds, this invention resides in the novel construction, combination, arrangement of parts, and method substantially as hereinafter described, and more particularly defined by the appended claims, it being understood that changes in the precise embodiment of the herein disclosed invention are meant to be included as come within the scope of the claims.
Brief Description of the Drawings
The accompanying drawings illustrate a complete embodiment of the invention according to the best mode so far devised for the practical application of the principles thereof, and in which:
FIG. 1A is a simplified side view illustrating generation of standing waves in a region of overlapping wave fronts on a silicon wafer;
FIG. 1B is an expanded view of the overlapping wave fronts as shown in FIG. 1A;
FIG. 2 is a graph illustrating reflectivity of silicon as a function of angle and polarization;
FIG. 3 is a simplified side view illustrating a particle raised several microns above a reflecting surface;
FIG. 4 is a graph illustrating the scattering cross-sections for PSL on silicon at 0.degree. angle of inclination;
FIG. 5 is a graph illustrating the scattering cross-sections of PSL on silicon at 60.degree. angle of inclination;
FIG. 6 is a schematic illustrating the system of this invention; and
FIG. 7 is a schematic illustrating the processing unit shown in FIG. 6.
Description of the Invention
To understand the advantages of this invention, it is necessary to have a physical understanding of the light scattering process using MIE theory to calculate scattering responses.
For a particle suspended in a homogeneous medium, a single incident plane wave generates a single outgoing spherical wave centered at the particle. However, a particle resting on a surface, while totally illuminated by an incoming (incident) wave, is also partially illuminated by an outgoing (reflected) wave. In addition, portions of the scattered waves propagating toward the surface are partially reflected by the surface and then partially rescattered by the particle (i.e., multiply scattered).
All of the scattered wave components have permanent phase relationship and interfere in a complex manner. Of additional concern is that an interference zone is set up in the overlapping common volume, as shown in FIGS. 1A and 1B. The fringe contrast in this interference zone is a function of the amplitude of the reflected wave, while the fringe spacings depend on the angles of illumination (incidence angle .theta..sub.i) and state of polarization of the incident wave. The fringe spacing is given by: ##EQU1##
The fringe zone represents a standing wave pattern which is phase-locked to the surface. For "S" polarization (electric field vector (E) perpendicular to the plane of incidence), there is a 180.degree. phase shift, upon reflection, at all incidence angles for most surface materials, and thus the surface itself is generally in the first dark zone.
The contrast of the fringes (or the standing wave ratio (SWR)) is a function of the amplitude of the reflected wave which can be computed from material optical data. For silicon, the contrast is about 60% (SWR=2.5) for normal incidence and increases to 100% at grazing incidence.
For "P" polarization (E in the plane of incidence), it is necessary to resolve E.sub.p into its two orthogonal components: E.sub.p .parallel. parallel to the surface and the other E.sub.p .perp. perpendicular to the surface. E.sub.p .parallel. undergoes a 180.degree. phase shift, while the E.sub.p .perp. has 0.degree. phase shift until angles of incidence exceed Brewster's angle, whereupon a 180.degree. phase shift again occurs.
Thus, in general, for "P" polarization, the fringe zone is also phase locked on a dark zone for 0.degree. incidence angles gradually shifting in phase and lowering in contrast as incidence angles increase. Amplitudes for the reflected "P" and "S" polarization components for silicon are shown in FIG. 2. Silicon, having a complex reflective index, does not have a zero reflectance at Brewster's angle. For silicon, the "P" fringe contrast decreases from 60% at normal incidence to nearly 0% at Brewster's angle where the standing waves essentially disappear.
The existence of standing waves phase-locked to a surface generates some interesting possibilities for scattering responses. One immediate result is that for normal incident waves, a very small (<<.lambda.) particle attached to the surface would scatter very weakly as compared to a position elevated above the surface to a bright zone (in fact, nearly 4x less). At oblique incidence, a "P" wave would also scatter more than an "S" wave for such very small particles. Collecting geometries are, however, biased in favor of "S" collection because "S" scattering is greatest normal to the surface while "P" is greatest parallel to the surface. As particles increase in size, they extend into more than one fringe generating a complicated response.
Thus the amplitude of any reflected wave illuminating the particle is a function of the state of polarization of the incoming wave and the optical properties of the surface and must be accurately known to determine scattering response. However, for all but the smallest of particles, and at all but the largest of incidence angles, this reflected wave is itself attenuated by the particle's presence; the particle partially shadows the reflected beam.
Obviously, the problem would be simplified if the particle could be suspended several diameters above the reflective boundry to move the near-field away from the boundary. It must remembered, however, that the near-field varies as I/D.sup.-2, and the offset need only be a few microns for the shadowing effect to be lost.
To this end, several .lambda./2 layers of SiO.sub.2 can be grown on a silicon wafer to provide a situation such as shown in FIG. 3. For submicron particles, the effect of shadowing by the particle could then, for all practical purposes, be neglected. Tests have been conducted with monodisperse PSL particles elevated by such a standoff-layer where good agreement between theoretical (using MIE theory) and empirical values were found.
If an odd multiple of .lambda./4 spacing is utilized, the results would obviously change as the particle transitions from a dark to a bright zone. Aside from the fact that the standoff-layer simplifies the theoretical treatment, it also hints at possible significant changes in response for patterned versus "virgin", or bare, wafers.
However, the case of greatest interest is that of FIG. 1 corresponding to a bare wafer and the strengths of the reflected wave must be reasonably estimated if a useful representation of the response is to be obtained. The reason for this concern is apparent when it is realized that the incident beam results in a direct "collectable" backscattered contribution while the reflected wave generates a direct "collectable" forward scattered component. The forward scatter is substantially greater than the backscatter for all but Rayleigh sized particles, in which case they are equal. It is therefore apparent that the reflected wave's amplitude must be determined quite accurately. In addition, phase differences between all components must be determined to compute scattered intensities.
When a particle is very small, its effective scattering cross section is much less than its geometric cross section, and thus the effect that the particle has on the reflected wave can be neglected. However, as the particle approaches the wavelength in size, its scattering cross section becomes greater than its geometric cross section (Q.sub.sc >1, where ##EQU2## and the presence of the particle greatly effects the reflected wave. At 0.degree. incidence, the reflected wave is not effectively illuminating particles with Q.sub.sc >1 corresponding to sizes larger than .lambda.; its existence being precluded by the particle itself. Thus, for such sizes at incidence angles near 0.degree., the scattering that results is closely approximated by backscatter only. Only when the particle diminishes in size to where Q.sub.sc <1, or for larger incidence angles, can a reflected component begin to exist and generate measurable forward scatter.
For computation purposes, the reflected component generated (that portion potentially shadowed) was attenuated by reducing its amplitude in proportion to the particle's scattering efficiency factor. At this point, it could be seen that a compensating factor emerged which reduced dynamic range--as size increased to where strong forward scatter was dominant, it was the least likely to be generated. Theoretical values for bare silicon at 0.degree. incidence angles and for .+-.5.degree. collecting angles are shown in FIG. 4. Computations for sizes larger than 1 .mu.m are subject to increasing error, and while the mean trends are useful, the response oscillatory signatures are not exactly correct in phase and amplitude. However, the computed response in the submicron range is indistinguishable from actual measurements. For sizes larger than 1 .mu.m, the means response is essentially a D.sup.2 function. Clearly, this geometry generates an ambiguous response over much of the size range.
For polarization effects, the case of 60.degree. incidence angle can be examined. Here, it is a situation where the particle cannot alternate much of the reflected illumination since a large portion of it is generated by reflection well outside the region influenced by the particle. Again, for very small particles, it is illuminated by an unperturbed reflected wave, but as size increases, there is again some small attenuation of the reflected component which is again estimated from Q.sub.sc. The results for both "S" and "P" polarization at 60.degree. incidence angles and .+-.10.degree. collecting angles are shown in FIG. 5, along with experimental values for PSL. The agreements are seen to be quite good in the important submicron range of sizes.
The results of FIG. 5 might at first appear to be as difficult to utilize as those of FIG. 4. However, if scattering from both polarization measurements are independently analyzed there are two calibration relationships from which to choose, and also there is a polarization ratio available. For example, for the case of FIG. 5, it is possible to develop a monotonic calibration which is effected by using the "S" polarization response for sizes up to about 0.4 .mu.m and then switching to "P" polarization responses for sizes larger than 0.4.mu.m. This type of "polarization diversity" is utilized in the schematic of the invention shown in FIG. 6.
As shown in FIG. 6, "P" polarized He-Ne laser 14 is utilized, as is an "S" polalrized He-Ne laser 16. While not specifically shown, the lasers could also operate at different frequencies as well as having different polarizations. As shown, the "P" polarized beam 18 from laser 14 is reflected by mirrors 20 and 22, coupled through cylindrical lens 24, reflected by mirror 26, coupled through cylindrical lens 28, and reflected by mirror 30, which directs the beam to a point (monitoring region) on the surface of wafer-disk 32.
As also shown in FIG. 6, wafer-disk 32 is supported on vacuum chuck 34 so that the surface is at the monitoring region and the chuck is rotated, in conventional manner, by motor 36.
The "S" polarized beam 38 from laser 16 is refleted by mirrors 40 and 42, coupled through cylindrical lens 44, reflected by mirror 46, coupled through cylindrical lens 48, and reflected by mirror 50, which mirror directs are "S" beam to the surface of wafer-disk 32 so that the two beams impinge the surface at a common point at the monitoring region.
Light scattered due to particle contamination and/or surface defects is collected through lens arrangement 54, which includes a six element diffraction limited dark-field objective lens set 56, 57, 58, 59, 60 and 61, with the output from the lenses being directed to and through polarizing beam splitter 64. Polarizing beam splitter 64 separated or divides the scattered light into two components, one of which is representative of light scattered from the "P" beam and the other of which is representative of light scattered from the "S" beam. The component containing the light scattered from the "P" beam is directed through the beam splitter to detector 66 where scattered light is detected and representative electrical signals generated, which signals are then coupled through amplifiers 68 and 70 to provide a "P" signal output. The "S" beam component is directed from the beam splitter to detector 72 where the light sensed is utilized to form electrical signals representative thereof, which electrical signals are than coupled through amplifier 74 to provide an "S" signal output.
If the lasers have different frequencies, the polarizing beamsplitter 64 can be replaced with a dichroic (color-separating) beamsplitter or otherwise split and filtered to separate the two scattering signals. The "S" and "P" beams will thus be of different color (frequency).
As also indicated in FIG. 6, the outputs from amplifiers 70 and 74 are coupled to signal processor unit 76 which provides an output indicative of particle contamination and/or defects on and below the surface of the material. This output can be a monotonic output indicative of particle contamination sensed. For such an output, processing unit 76 may include switching means to accept process "S" polarization signals for particle sizes up to about 0.4 .mu.m and "P" polarization signals for particle sizes over 0.4 .mu.m.
Processing unit 76 is shown in greater detail in FIG. 7. As shown, unit 76 includes pulse height analyzers 78 and 80 connected to receive the outputs from amplifiers 70 and 74, with outputs from analyzers 78 and 80 being coupled to logic array or microprocessor 82. The analysis of the "S" and "P" signals is accomplished with the two independent pulse height analyzers 78 and 80. The measured scattering amplitudes generated by PHA's 78 and 80 are conveniently compared using unit 80 which can include preprogrammed logic arrays or via a "lookup" table using microprocessor based software to interpret the measured values and make decisions as to size or distinguish particles from surface defects.
A second use of polarization analysis is for surface defects. Surface defects in general, do not exhibit polarization sensitivity nearly to the extent that small particles do. All of the scattering occurs at one plane without the complicating effects of illumination by multiple beams. The primary effect is that of diffuse scattering by a small localized area. The scattering for "S" polarization versus "P" polarization is generally a constant for a particular viewing solid angle and does not reverse response magnitude with changing size as particles do. Particles can thus be distinguished from surface defects by polarization ratio measurements except at the size where polarization ratio or particles and defects are equal.
For other collecting angles, the response functions will differ, but in general, the "S" polarization will provide higher scattering response than "P" for smaller submicron sizes with the S/P ratio reversing as size increases. It is also possible to use a single laser which provides both controlled "S" and "P" outputs or circularly or elliptically polarized outputs.
As can be appreciated from the foregoing, this invention provides an improved surface analysis system and method that is well suited for determining particle contamination and/or defects on or below the surface of a material.