Chemical mechanical planarization (CMP) slurry quality control process and particle size distribution measuring systems

Abstract

A sensitive particle distribution probe uses special processing including a modified Twomey/Chahine iterative convergence technique and a specially constructed sample cell to obtain particle size distribution measurements from optically dense slurries, such as the slurries used in the semiconductor industry for chemical mechanical planarization. spectral transmission data is taken over the spectral range of 0.20-2.5 microns, utilizing specially constructed, chemically resistant sample cells of 50-2000 microns thickness, and miniature, fixed grating, linear detector array spectrometers. At wavelengths greater than one micron, the preferred design utilizes InGaAs linear detector arrays. An ultrasonic disrupter can be employed to breakup harmless soft agglomerates. In addition to direct particle size distribution measurement, the invention described here could be used to detect other fundamental causes of slurry degradation, such as foaming and jelling. The probe accomplishes continuous, real time sampling of undiluted slurry. A three-position chopper allows automated operation in an industrial environment without the need for frequent reference spectra, which would require taking the probe off-line. In other embodiments, the invention provides a quality control and/or particle size measuring system for cmp slurries using transmission data through an as-used cmp slurry flow. The process of the invention detects transmission through the flow, at select wavelengths, and determines changes in the logarithmic slope of transmission versus wavelength to detect acceptable or unacceptable cmp slurries. The process can further determine cmp slurry particle size through empirical extinction data stored in memory.

Description

RELATED APPLICATIONS

This application is a continuation-in-part of commonly-owned and U.S. application Ser. No. 09/069,682, filed on Apr. 29, 1998,

By first measuring the transmission of the sample cell filled only with the liquid portion of the slurry, then dividing that into the transmission expressed in Equation (1), one can isolate T_P(λ), which is the quantity of interest. Beer's Law is then solved for the particle volume extinction coefficient (βE(λ)), as shown in Equation (2), where L is the transmission path length or sample cell width. Equation (3) represents the formula for calculating the particle volume extinction coefficient in terms of the particle radius (r), the Mie extinction efficiency (Q_E), and the PSD (N(r)), where m is the particle's complex refractive index.
β_E(λ)=−ln(T_P(λ))/L  (2)
β_E(λ)=∫πr²Q_E(2πr/λ)N(r)dr  (3)

Equation (3) is inverted to solve for the particle size distribution. One class of inversion algorithms is the linear inversion, which provides a less preferred model for reasons that are explained below. The less preferred inversion method transforms the measurement equation into a linear system of equations by replacing the integral with a summation and by representing the collection of equations in the matrix form given by Equation (4). In this latter equation, elements of matrix Q consist of πr²Q_E. The Q matrix has m rows, one for each wavelength, and n columns, one for each radius; m must be greater than or equal to n. The N matrix is n by 1, and the elements consist of the particle size distribution. The β matrix is m by 1, and the elements consist of the measured spectral volume extinction coefficients.
Q·_RN_R=β_λ  (4)

Equation (4) can be formally inverted to solve for the particle size distribution, utilizing conventional inversion algorithms which constrain the solution to various conditions, such as smoothing (minimize the first or second derivative), or minimize the departure from a first guess, according to Twomey, Comparison of constrained linear inversion and an iterative nonlinear algorithm applied to the indirect estimation of particle size distributions, J. comp. Phys., Vol. 18, No. 2, pp. 188-200 (1975), which is hereby incorporated by reference to the same extent as though fully disclosed herein.

Constraints are required in all inversion algorithms because the existence of measurement error and quadrature error (replacing the integral with a sum) result in the fact that a family of partide size distributions will satisfy the measurement equation. For any inversion method, the uncertainty in the retrieved solution can be reduced by: (a) choosing a more sensitive measurement technique, (b) reducing the measurement error, (c) increasing the number of measurements, which reduces the effects of quadrature error.

Linear inversion techniques are computationally efficient, but they are a poor choice for the CMP slurry problem because the most popular constraint, i.e., that of smoothing, is a poor choice for slurry particle size distributions. These distributions are not necessarily smooth or continuous. Additionally, linear inversion algorithms can be unstable to an extent that produces physically unrealistic answers.

The CMP slurry measurement problem consists of detecting departures from the normal or specified particle size distribution, which makes a non-linear, iterative, inversion algorithm a natural choice and a more preferred model for use in practicing the invention. With the iterative approach, one can start with the normal particle size distribution as a first guess. The iterative calculations converge toward a final solution in an orderly fashion, where convergence is based upon a difference between the measured spectral extinction and that calculated from the last guess particles size distribution. Alternatively, one can start with a delta function as a first guess. Iteration is halted when this difference becomes less than some predetermined error bound. This preferred method of inverting equation (4) is based on previous work in the field of atmospheric remote sensing by Cerni, Aircraft-based remote sensing or tropospheric profiles for meoscale studies, Advances in Remote Sensing Retrievals, pp. 339-347 A. Deepak Publ., Hampton, Va. (1985); and Chahine, Inverse problems in radiative transfer: Determination of atmospheric parameters, J. Atmos. Sci., Vol. 27, pp 960-967 (1970) and Twomey (1975, referenced earlier), which are incorporated by reference herein to the same extent as though fully disclosed herein.

The algorithm given in Equations (5) and (6) is a preferred means of inverting the spectral transmission data to retrieve the particle size distribution. The superscripts I and I−1 refer to successive numbers of iterations. The subscripts P refer to different wavelengths, and indicate that all the measurements are utilized in adjusting the partide size distribution at a single r value. Additionally, one can improve the accuracy of the retrieval by adding conservation of mass (slurry percent solids by weight), and summing Equation (5) over all wavelengths.
N_P^(I)(r)=[1+(r_P^(I-1)−1)πr²Q_E(2πr/λ,m)]N_P^(I-1)(r)  (5)
r_P^(I-1)=β_E(λ)/[∫πr²Q_E(2πr/λ,m)N_P^(I-1)(r)dr]  (6)

EXAMPLE 1 VERIFICATION OF THE MODEL WITH EXPERIMENTAL RESULTS

Mie theory optical model results were verified with the use of an Acton SP-305 spectrometer system retrofitted with a sample cell according to FIG. 3. The sample cell was constructed to provide sapphire windows having a 40 mm diameter with the windows being held approximately 100 microns apart in a PVDF chemically resistant block. The detector module utilized one Si and one InGaAs photodiode to cover the broad 0.20-2.5 micron spectral range.

FIG. 5 shows a comparison between optical model predictions and experimental data for Solution Technology Klebosol 30N50 oxide slurry, which consists of SiO₂ particles immersed in a weak NH₄OH solution with a pH of 10.9. The product as tested contained 30% solids by weight, which is the basic product produced by Solution Technology. The product it is typically diluted to 18% solids by weight just prior to use in CMP wafer planarization. Thus, the tested product was even more optically dense than an actual CMP slurry using the product. As an input, the optical model utilized a particle size distribution provided by Solution Technology, shown in FIG. 6, which the manufacturer purports to have been measured with an electron microscope. The comparison shown in FIG. 5 demonstrates a remarkably good agreement between measured transmission data and optical model predictions.

Possible reasons for the observed small differences between the two curves (theoretical and actual results) of FIG. 5 include: (1) departure of the optical behavior of this unusually dense particulate suspension from that predicted by Mie theory, (2) departures of the sample particle size distribution from the typical particle size distribution provided by the slurry manufacturer, (3) errors in the particle size distribution measurements provided by the slurry manufacturer, due to the poor sample statistics provided by analysis of electron microscope imagery, (4) unexpected slurry liquid absorption bands, and (5) errors in the experimental spectral transmission measurement technique. The combined effects of these error sources are minor in this example. Klebosol 30N50 is described by the manufacturer as consisting of individual spheres, which are grown from seed in a saturated SiO₂ solution. As such, one would expect accurate predictions from Mie theory.

The experimental data shown in FIG. 5 was truncated at a transmission value of 0.030, below which the measured data indicated a leveling off and then an increase in transmission as wavelength decreased and optical depth increased. Such a result is unphysical, and indicates that the multiply scattered radiation, which is scattered in a near forward direction, has become comparable to or greater than the transmitted radiation. This result is expected to occur at some point with increasing optical depth and with a finite instrument field of view. The detector system utilized the SP-305 spectrometer, which is designed to have a nominal 1° field of view; and this scattering effect was predicted to be observed at a transmission value of approximately 0.050, i.e., an optical depth of 3.

EXAMPLE 2 VERIFICATION OF THE MODEL WITH EXPERIMENTAL RESULTS

FIG. 7 shows optical model predictions and experimental data for Cabot SC-1 oxide slurry, which consists of SiO₂ partides immersed a pH of 10.3. This sample was diluted to 12% solids by weight, which is the concentration at which it is used for CMP wafer planarization. The particle size distribution used as input to the optical model is plotted in FIG. 8, and represents a modified version of the Cabot SC-1 PSD measured by Bare et al., Monitoring slurry stability to reduce process variability, Micro. Vol. 15, No. 8, pp. 53-63 (1997) (the BH97 particle size distribution distribution probe. A modification to the BH97 particle size distribution consisted of multiplying each particle size distribution size bin by 0.56. The 0.56 factor was chosen to obtain good fit to the measured transmission data.

FIGS. 9 and 10 show how changes in this distribution size bin factor affect the transmission spectrum, and serve as another indication of the sensitivity of the spectral transmission measurement technique. Cabot SC-1 is a fumed silica product produced by combining reactant gases in a flame, and is known to consist of chains of tiny spheres fused together. Hence, the diameter of such a particle chain is not well defined. The extinction cannot be accurately modeled by Mie theory, and one should expect differences in the particle size distribution obtained by different measurement techniques. Given this uncertainty, a factor of 0.56 is reasonable.

In FIG. 9, the measured spectrum was truncated at a transmission value of 0.079, due to apparent errors introduced by multiple scattering. This higher value of transmission cut-off for SC-1 versus 30N50 (0.079 versus 0.030, above), is consistent with the larger particles present in the former slurry, which is known to produce more forward scattering. It is also consistent with more forward scattering produced by nonspherical versus spherical particles.

FIG. 11 depicts a schematic process diagram of process P1100 for use in operating the probe shown in FIG. 1. In step P1102, optically dense CMP slurry is diverted from the main slurry line to the sample cells 154 and 162. In step P1104, the flow of slurry is narrowed through the sample cells to provide an optical depth that permits meaningful spectral transmission data. Light is transmitted through the narrowed slurry along pathways 134 and 136 in step P1106. Pathways 140 and 142 deliver this light to the spectrophotometers 128 and 130 in step P1108. The spectrophotometers produce signals representative of the detected light and particles in the cells 154 and 162. These signals are transmitted to CPU 164 for processing according to the modified Twomey/Chahine technique according to equations 1-6.

At the conclusion of step P1108, step P1110 includes the detection of light transmitted along pathway 126 to spectrophotometers 128 and 130 due to the rotation of chopper blade 122 and the reflective action of mirror 202. The detector counts are transmitted to CPU 164 for registration of source lighting conditions without particle scattering from sample cells 154 and 162.

In step P1112, the detector background count is measured with chopper blade 122 positioned to place solid disk 204 in path 120 for blocking the transmission of light along either path 126 or 134. Spectrophotometers 128 and 130 again produce signals corresponding to detected light, and these signals are transmitted to CPU 164, which interprets the signals as background count information that can be subtracted from total counts received from light traveling along pathways 126 or 134.

In step P1114, CPU 164 uses stored detector signals from steps P1108, P1110, and P1112 to calculate, display and store a particle size distribution, as discussed above. Steps P1106-1114 are continuously repeated to perform real time measurements of the particle size distribution in the CMP slurry.

FIG. 12 shows one CMP slurry quality control system 1200 constructed according to the invention. CMP slurry 1202 from CMP slurry supply 1204 is transmitted through supply line 1206 to a sample cell 1208, e.g., sample cells 154, 162 of FIGS. 1 and 3. Sample cell 1208 provides for efficient and uniform CMP slurry flow 1210 through cell 1208 so that radiation 1212 may be transmitted therethrough, as discussed in FIG. 1 above in connection with beams 134, 136.

Source 1214 generates radiation 1212. By way of example, source 1214 can be a quartz tungsten halogen source, generating infrared and/or visible radiation 1212, or a deuterium source, generating ultraviolet radiation. Preferably, source 1214 is “broadband” so as to provide multiple wavelength bands which generate radiation 1212. However, multiple sources 1214a, 1214b . . . 1214n can be used, selectively, to generate desired radiation wavelengths 1212a, as required. For example, to generate ultraviolet light, source 1214b can represent a deuterium source; while to generate infrared or visible light wavelengths, source 1214a can represent a tungsten lamp. To switch between sources 1214, an arrangement such as shown in FIG. 1 can be used, or alternative techniques can be used to accomplish the same function, such as through mechanical actuation.

In the preferred embodiment, filters 1216 spectrally discriminate source radiation wavelengths 1212a emitted from source 1214 such that only selected wavelengths 1214b pass through filters 1216. Multiple filters 1216a, 1216b can be used to alternatively pass and select different wavebands to illuminate sample 1208. By way of example, filters 1216 are shown arranged on filter wheel 1218 which is rotated about axis 1219 by motor controller 1218a, selectively, to alternatively position filters 1216a, 1216b in the path of radiation 1212a. Filter wheel 1218, controller 1218a, and filters 1216 are known those skilled in the art of optics. In this manner, radiation 1212b of desired waveband can be selected by a user of system 1200. Filters 1216 are moved to block radiation 1212a as needed to select the appropriate wavelength band as emitted from source(s) 1214.

Although two filters 1215 are shown, those skilled in the art should appreciate that one or more filters can be used in system 1200 to achieve the objectives herein.

Radiation 1212c transmitted through sample cell 1208 corresponds to radiation also transmitted through CMP slurry flow 1210. A detector 1220 detects radiation 1212c and generates signals indicative of transmission of radiation 1212b through sample and flow 1208, 1210. These signals are interpreted by processor 1220, e.g., a computer, to determine a transmission value as a function of wavelength (or waveband). By way of example, if source 1214 generates radiation 1212a that is filtered by filter 1216a to 2.5 microns +/−0.2 micron, then detector 1220a can correspond to a near infrared detector, e.g., InGaAs, to detect transmission of radiation 1212c through sample and flow 1208, 1210. Transmission is determined by computer 1220 and associated with “2.5 microns.” At times, multiple detectors 1220a, 1220b . . . 1220n are required to detect all the wavelengths of interest from sources 1214a, 1214b . . . 1214n. Detectors 1220 can be inserted within system 1200, as needed, to measure appropriate wavelengths, or an appropriate optical technique such as illustrated in FIG. 1 can be used to achieve the same function.

Different slurry supplies 1204a, 1204b . . . 1204m can also be coupled to system 1200 in a manufacturing process; and each CMP slurry 1202a, 1202b . . . 1202m can then be coupled to sample cell 1208 as required through appropriate flow pathways 1206. Alternative sample cells 1208a, 1208b . . . 1208q can be used in system 1200, as needed, to acquire appropriate optical path lengths corresponding to enhanced detection of radiation 1212b through sample cell and flow 1208, 1210. As before, sample cells 1208 can be switched into system 1200 manually, or mechanically, or an optical configuration such as FIG. 1 can be used to achieve the same function (i.e., multiple samples are mounted within system 1200 and radiation of the appropriate wavelength is re-routed to the correct sample cell 1208 through different optical paths and beam splitters).

CMP slurry from flow 1210 leaves sample cell 1208 along slurry line 1222, which couples to semiconductor manufacturing process 1224. When system 1200 detects bad CMP slurry, as discussed herein (e.g., slurry with a particle distribution extending beyond a desirable range), then processor 1220 sends a warning signal to warning device 1226, e.g., a light, audible alarm or other device (e.g., a computer) coupled or proximate to manufacturing process 1224. In this manner, manufacturing process 1224 is informed, in real time, of CMP slurry quality control issues which can damage and destroy semiconductor surfaces used in integrated circuit devices.

Transmission values determined by system 1200 are preferably plotted with respect to wavelength, such as illustrated in FIG. 12A. Specifically, the natural log of transmission values (ln(t), axis 1240) is plotted against wavelength (λ, axis 1242), as shown. Accordingly, the slop of a line C which approximates ln(transmission(λ)) at time to may be determined, such as line C(t₀). At a later time t, line C may for example be plotted as C(t₀+t), indicating a change in the slope of ln(transmission(λ)). When the slope of line C changes by a sufficient amount, represented by angle β, determined empirically or by another measure, then the particle distribution sizes within the CMP slurry have changed and system 1200 sends a warning to manufacturing process 1224. FIG. 12A also illustrates wavelength measurement points λ₁, λ₂ used to determine the slope of line C, as known in the art. Each λ sample corresponds to a measurement point corresponding radiation passed through filter 1216, for example. Each waveband a is centered about a wavelength λ such that Δλ/λ is less than approximately 5%. For example, for λ=2.5 microns, a corresponds to 1bout 0.13 micron.

The function of source(s) 1214 and filter(s) 1216 can be replaced by laser diodes, if desired. Alternatively, filter(s) 1216 can be replaced by appropriate dispersive elements (e.g., gratings) located with detector(s) 1220, such as discussed in FIG. 1.

Processor 1220a preferably includes solid state memory to store one or more “reference transmission” data corresponding to a preferred transmission vs. wavelength curve, or ln(t) vs. λ data, for a known CMP slurry with acceptable particle size distribution. The reference transmission data further includes an acceptable variance of that data from optimal where CMP slurry is deemed “acceptable.” Accordingly, in this embodiment, system 1200 evaluates transmission data from flow 1210 in real time and compares that data to reference transmission data in memory 1220a, and generates a warning when the real time data exceeds the allowed variance, indicating an “unacceptable” CMP slurry. Memory 1220a can further include an array of curves or ln(t) vs. λ data corresponding to each CMP slurry 1202a, 1202b . . . 1202m, as appropriate, such that system 1200 can operate with multiple CMP slurries used in manufacturing process 1224. A user can select which reference transmission data to use at any one time through a user interface (e.g., a keyboard) at processor 1220.

Measuring particle size distributions within CMP slurry flow 1210 is also a feature of the invention. Typically, these distributions are centered about a particular particle size, e.g., 0.06 micron, as shown in FIG. 6. Other suitable center particle sizes in accord with the invention are between about 0.3 and 1.0 micron, though particle size distributions centered about a value between 0.1 and 0.3 micron, or 1.0 and 10 microns, are also envisioned and within the scope of the invention. Typically, the diameter of flow 1210 is approximately 100 microns for near-infrared wavelengths. Smaller flow diameters, i.e., down to 50 microns or smaller, are also envisioned, as are larger flow diameters up to approximately 2000 microns, all within the scope of the invention.

Mie theory can be used to determine particle size distributions in CMP slurry flow 1210. Alternatively, an empirical curve of extinction efficiency QE versus particle size diameter D is developed and stored in memory 1220a; and that empirical curve is compared to data obtained by system 1200 in real time. The particle size function preferably corresponds to πD/λ, where λ corresponds to the waveband of measurement. FIG. 12B illustrates exemplary Q_E versus πD/λ empirical data 1258 for different particle sizes, with QE on vertical axis 1260 and size parameter πD/λ on horizontal axis 1262. Bohren et al., Absorption and Scattering of Light by Small Particles, John Wiley & Sons, p. 319 (1983). To determine particle size D, Q_E is calculated directly as a function of transmission t, discussed above, and multiplied by λ/π.

FIG. 13 shows a process flow 1300 of the invention for detecting CMP slurry quality and/or particle size distribution. Process flow 1300 is representative for use of a system of the invention, such as illustrated in FIGS. 1 or 12. In process step 1302, the sample cell and CMP slurry flow is illuminated by radiation at a first waveband Δλ₁, e.g., 0.08 micron centered about 1.7 microns (λ₁). The system detector and processor then measure and determine a transmission value for λ₁, in process step 1304. In process step 1306, the sample cell and CMP slurry flow is illuminated by radiation at a second waveband Δλ₂, e.g., 0.03 micron centered about 0.6 micron (λ₂). The system detector and processor then measure and determine a transmission value for λ₂, in process step 1308. The systems and methods of the invention can detect further transmission values for other wavelengths and wavebands, as desired at step 1301, or calculate the slope of the transmission versus wavelength slope as set forth in step 1312. In step 1314, the slope measured in step 1312 is measured against a reference slope stored in system memory, or alternatively the current slope is compared to prior slope information, to evaluate change in the CMP slurry particle distribution. If the slope exceeds a predetermined amount from prior slope information, or from reference slope transmission data, then a warning is generated in step 1316. Otherwise, a next set of transmission data is taken in steps 1302-1308 to evaluate CMP slurry quality over time.

FIG. 14 depicts one CMP slurry particle measuring and quality control system 1400 constructed according to the invention. Two sample cells 1402a, 1402b (e.g., similar to sample cells 154, 162, FIG. 1) are used to extend the spectral range within which one obtains high accuracy transmission measurements to determine CMP slurry quality and/or particle size. Sources 1404a, 1404b generate radiation beam 1406 in different wavebands through beam splitter 1408. Other beam splitters and optics 1410 translate beam 1406 to appropriate grating spectrometers 1412a, 1412b, each with a mirror assemblies 1414a, 1414b used to isolate the desired waveband of interest.

The invention thus attains the objects set forth above, among those apparent from the preceding description. Since certain changes may be made in the above methods and systems without departing from the scope of the invention, it is intended that all matter contained in the above description or shown in the accompanying drawing be interpreted as illustrative and not in a limiting sense. It is also to be understood that the following claims are to cover all generic and specific features of the invention described herein, and all statements of the scope of the invention which, as a matter of language, might be said to fall there between.

Claims

1. A quality control process for detecting physical and/or chemical changes in a cmp slurry, comprising the steps of:

transmitting radiation through a flow of an undiluted optically dense slurry as used in a chemical mechanical planarization (cmp) process, the radiation having one or more wavelengths;

determining transmission of the transmitted radiation at each of the said one or more wavelengths; and

monitoring transmission, over time, to detect physical and/or chemical changes of the cmp slurry.

2. A process of claim 1, further comprising determining a slope of transmission as a function of the wavelengths.

3. A process of claim 1, further comprising the step of detecting changes in the particle size distribution of the cmp slurry.

4. A process of claim 3, further comprising the step of determining a change in the slope, over time, the change in slope indicating change in the particle size distribution.

5. A process of claim 3, further comprising determining a slope of a logarithmic of transmission.

6. A process of claim 1, further comprising the step of determining a slope of a logarithmic of transmission as a function of the wavelengths.

7. A process of claim 6, further comprising the step of determining a change in the logarithmic slope, over time, the change in the slope indicating change in a particle size distribution of the cmp slurry independent from a change in particle concentration.

8. A process of claim 1, further comprising the step of detecting changes in the particle size distribution of the cmp slurry wherein the particle size distribution corresponds to a value between about 0.03 and 1.0 micron.

9. A process of claim 1, further comprising the step of detecting changes in the particle size distribution of the cmp slurry wherein the particle size distribution corresponds to a value above about one micron.

10. A process of claim 1, wherein the step of transmitting the radiation comprises transmitting the radiation through the flow having a diameter of about 100 microns.

11. A process of claim 1, wherein the step of transmitting the radiation comprises transmitting the radiation through the flow having a diameter of between about 100-2000 microns.

12. A process of claim 1, wherein the step of transmitting the radiation comprises transmitting the radiation through a sample cell selected on the basis of desired accuracy.

13. A process of claim 12, further comprising selecting a sample cell defining a flow diameter of about 100 microns.

14. A process of claim 12, further comprising selecting a sample cell defining a flow diameter of between about 100-2000 microns.

15. A process of claim 1, wherein the step of determining transmission comprises determining transmission to an accuracy of at least about 1%.

16. A process of claim 1, wherein the step of transmitting comprises utilizing a grating to select the wavelengths of the radiation.

17. A process of claim 1, wherein the step of transmitting comprises using a laser.

18. A process of claim 1, wherein the step of transmitting comprises utilizing at least two filters to select the wavelengths.

19. A process of claim 1, further comprising generating a warning corresponding to the changes.

20. A process of claim 1, further comprising the steps of detecting changes in the particle size distribution of the cmp slurry and of comparing the transmission to a reference transmission indicative of a preferred particle size distribution within the flow.

21. A process of claim 20, further comprising the step of storing the reference transmission in memory.

22. A process of claim 1, further comprising the steps of:

(a) detecting changes in the particle size distribution of the cmp slurry;

(b) storing a plurality of reference transmissions, each reference transmission corresponding to a particular cmp slurry flow and particle distribution;and

(c) selecting one reference transmission and comparing the transmission to the selected reference transmission.

23. A process of claim 1, further comprising utilizing Mie theory to calculate particle sizes within the cmp slurry.

24. A process of claim 1, further comprising comparing transmission information with an empirical curve of extinction efficiency versus particle size diameter to determine particle sizes within the cmp slurry.

25. A process of claim 24, wherein the particle size diameter comprises a function of (pi) D/ lambda, where D is the particle size diameter and lambda corresponds to wavelength associated with the transmission.

26. A system for evaluating chemical mechanical planarization (cmp) slurry quality in a process, comprising:

a light source generating a beam of electromagnetic radiation for transmission through a flow of an undiluted optically dense slurry as used in a cmp process;

a spectral discriminator for isolating at least two wavelength bands of the radiation prior to transmission of the radiation through the flow;

a detector for detecting radiation transmitted through the flow; and

a processor for evaluating transmission of the wavelength bands through the flow to determine physical and/or chemical changes of the cmp slurry.

27. A system of claim 26, wherein the discriminator comprises two wavelength bandpass filters.

28. A system of claim 26, wherein the discriminator comprises a filter wheel.

29. A system of claim 26, wherein the discriminator is selected from the group consisting essentially of a laser and a grating.

30. A system of claim 26, wherein the processor comprises a computer.

31. A system of claim 26, further comprising memory, coupled to the processor, for storing one or more reference transmissions, each reference transmission corresponding to a particular cmp slurry flow and particle size distribution, the processor selecting one reference transmission and comparing the transmission to the selected reference transmission to detect changes in the particle size distribution.

32. A system of claim 26, further comprising memory, coupled to the processor, for storing data indicative of extinction efficiency as a function of particle size diameter, the processor comparing the transmission to the data to determine particle sizes within the cmp slurry.

33. A system of claim 26, wherein the processor comprises processing means to calculate a logarithm of transmission at each wavelength band and to determine a change in slope of logarithmic transmission versus wavelength band to detect changes in particle size distribution of the cmp slurry independently from changes in particle concentration.