AU608887B2 - System for measuring the size distribution of particles dispersed in a fluid - Google Patents

Description

47 AU-A]j-21236/88 ,~il~h PCT WORLD INT E RG ZAT INTERNATIONAL APPLICATION PUBL HED UNDER TE PATENT COOPERATION TREATY (PCT) (51) International Patent Classification 4 (11) International Publication Number: WO 89/ 00286 G01N 15/02, 21/53 1 (43) International Publication Date: 12 January 1989 (12.01.89) (21) International Application Number: PCT/US88/02222 (81) Designated States: AT (European patent), AU, BE (European patent), CH (European patent), DE (Euro- (22) international Filing Date: 29 June 1988 (29.06.88) pean patent), FR (European patent), GB (European patent), IT (European patent), JP, LU (European patent). NL (European patent), SE (European patent).

(31) Priority Application Number: 068,139 (32) Priority Date: 29 June 1987 (29.06.87) Published With international swarch report.

(33) Priority Country: US Before the expiration of the time limit for amending the claims and to be republished in the event of the receipt of amendments.

(71) Applicant: COULTER ELECTRONICS OF NEW ENGLAND, INC. [US/US]; 29 Cottage Street, Amherst, MA 01002 (72) Inventor: BOTT, Steven, E. Shelburne Falls Road, Conway, MA 01341 (US).

(74) Agents: LAPPIN, Mark, G. et al,; Lahive Cockfield, I.x 2 3 MAR IM State Street, Boston, MA 02109 (US).

AUSTRALIAN

a1I 30 JAN1989 PATENT OFFICE (54) Title: SYSTEM FOR MEASURING THE SIZE DISTRIBUTION OF PARTICLES DISPERSEDIN A FLUID (57) Abstract An, apparatus and method is disclosed which provides a measure of the size distribution of particles dispersed in a fluid based upon an optimum combination of CLS measurements and DLS measurements. The apparatus includes a light soutre (18) providhig a light beam (20) which is focused by a lens (16) on the sample Scattered light is sensed by a light detector The measurement is characterized by relatively high resolution particle sizing. DLS data is optimally combined with CLS data, and either a size distribution or a histogram may represent distributions weighted by mass, volume, number, surface area, or other measures, i~ 1 -i S 89/00286 PCT/US88/02222 -1- SYSTEM FOR MEASURING THE SIZE DISTRIBUTION OF PARTICLES DISPERSED IN A FLUID RFRv rTF Xr' REPLATED APPLICATION The nt application is a continuation-in-part o Serial No. 817,048, now U.S. Patent No. 4,676,641, "System-foMeasuring the Size niistri h iti nn' f Partic1es ni spersed in a Fluid".

FIELD OF THE INVENTION This invention relates to light scattering instrumentation and more particularly to light scattering systemp-for measuring the size distribution of particles dispersed in a fluid.

BACKGROUND OF THE DISCLOSURE There are several prior art techniques for measuring the distribution of the particle size in a sample by using light scattering. Generally, to measure the sizes of individual particles, for example, in a flowing stream of a liquid or gas, the particle-containing sample is illuminated by a constant light source and the intensity of light scattered by the particle is detected. A particle scatters the light by an amount directly related to the particle size; in general, bigger particles scatter more light than smaller particles. The relation between the amount of scattering and particle size may be determined either from theoretical calculations or through calibration process. With knowledge of this relation, for a single particle at a time, the detected scattered light intensity provides a direct measure of the SUBSTITUTE

SHEET

WO 8900286PCT/US88/02222 -2particle size. The distribution of particle sizes in a sample may be determined by individually, passing each particle in the sample, or a suitable portion of the sample, through the scattered light detection apparatus and tabulating the sizes of the v.

4 rious particles. In practice, this method is generally restricted to particles greater than 0.5 microns.

Moreover, this method is relativwely s~ow since particles must be detected individually. This ttechnique is referred to in the prior art as optical particle counting (OPC).

A second prior art technique of particle sizing by light scattering is referred to as static -or "classical" light scattering (CLS) This method 1ig based upon illumination of a sample containing the particles-to-be-sized followed by the measurement of the intensity of scattered ligl ,t at several predetermined angles. Bec-,ause of intra-particle destructive interference, the intensity of light 2scattered from a particle depends on both the size and composition of the particle and the angle at which the measurement, is made. This method of particle sizing based~ on the angular dependence of~ the scattered intensity can be used to measure the 2size distribution of a group of particles, as opposed 2to the first method noted above which is restricted to individual particles.

To implement the CLS measurement method, a sample of particles dispersed in a f luid is 3illuminated along an input axis, and the intensity of scattered light is measured at several predetermined angles. The scattered light intensity at each angle may be measured simultaneously with a multitude of SUBSTITUTE SHEET i SWO 89/00286 PCT/US88/02222 -3detectors or consecutively, by moving a single detector around the sample to permit measurement of the intensity at each desired angle.

For large particles, for example, having diameters greater than 1 micron, the scattered light flux is concentrated in the forward direction relative to the input axis. Irrstruments for sizing large particles are referred to as laser diffraction devices. For sizing of smaller particles, for example, having diameters as low as 0.2 microns, the scattered light flux has significant magnitude both at lower and higher scattering angles relative to the input axis. The angular intensity measurements used on smaller particles are termed total integrated, or average, intensity measurements and may be displayed in a form known as Zimm plots.

A third prior art technique for particle sizing by light scattering is dynamic light scattering (DLS), also known as photon correlation spectroscopy (PCS) or quasi-elastic light scattering (QELS). See B. E. Dahneke, "Measurement of Suspended Particles by Quasi-Elastic Light Scattering," John Wiley Sons, Inc., New York, 1983. This technique is based on measuring the time-fluctuations of the intensity of light scattered from an illuminated sample containing a group of particles which are diffusing through a fluid, that is, randomly moving due to collisions with solvent molecules and other particles. For example, the particles may be 3 macromolecules dissolved in a liquid, where the macromolecules may be ionized by the loss of A small number of charged &toms.

SUBSTITUTE SHEET WO 89/00286 PCT/US88/02222 -4- In accordance with the DLS technique, scattered light intensity is measured as a function of time at a selected angle with respect to an illumination input axis. The light intensity detected at any instant at the detector is dependent on the interference between the light scattered from each illuminated particle in the scattering volume.

As the particles randomly diffuse through the solution, the interference of the light scattered from them.changes and the intensity at the detector therefore fluctuates. Since smaller particles diffuse faster, the fluctuations resulting from the motion of relatively small particles vary faster than those resulting from the motion of larger particles.

Thus, by measuring the time variation of the scattered light fluctuations at the detector, information representative of tne distribution of particle sizes is available. More particularly, the autocorrelation function of the measured intensity is related to the distribution of particle sizes in the fluid. Conventional DLS instruments such as the Model N4 photon correlation spectrometer manufactured by Coulter Electronics, Inc., Hialeah, Florida, provide autoonrrelztion signals for the detected intensity suitable for ,,easu:ing distributions including particle sizes as '.ow as 0.003 microns.

Accordingly, suci& devices have a size measuring range extending considerably below the above-noted individual particle and CLS methods.

Y 30 Particle sizing measurements by the known DLS techniques are generally made in the following manner. The particles-to-be-sized are suspended or dissolved in a fluid, forming a sample. The sample is illuminated by a laser beam directed along an SUBSTITUME -SHEET 'WO 89/00286 PCT/US88/0222 input axis. Although a laser is generally used to generate the beam, a non-coherent light source may alternatively be used.

The light scattered from the particles in the sample is detected by a photodetector, such as a photomultiplier, which is positioned at a predetermined angle. The particular angle miay be selected by the operator, but usually only one angle is measured at a time. The photodetector produces a signal which varies with time as the scattered light intensity incident on the photodetector varies. This time-varying signal is applied to an autc-orrelator analyzer, to compute the autocorrelation function of the photodetector signal. Typically, the autocorrelator computes the value of the autocorrelation function of the detected scattered light at as many as one hundred discrete time points. This autocorrelation function contains the information about the fluctuations in the detected scattered light, from which information about the distribution of particle sizes in the sample can be extracted. Thus, the autocorrelation function (acf) is the raw data of a DLS measurement. While most conventional DLS measurements are performed using this autocorrelation step, it is known that the acf of the intensity signal corresponds to the Fourier Transform of the power spectrum of that signal.

Accordingly, a spectrum analyzer may be used in place of the autocorrelator co generate a power spectrum signal including the same information representative of the particle size distribution as is resident in the autocorrelation function. The frequency domain information resident in the power spectrum signal can be used to determine the particle size distribution.

SUBSTITUTE

SHEET

1 WO 89/00286 PCT/US88/02222 -6- In the prior art, there are several techniques for extracting the particle size distribution from the acf. For use with these techniques, the relationship between the acf and the size distribution can be expressed as: g(t) K(x(r)) where g(t) is the acf (or a function closely related to the acf), x(r) is the sought distribution of particle sizes (x is a function of r, the particle radius), and K is a function (or operator, linear or non-linear) which relates particle size to the acf.

Thus, given the exact form of K, the autocorrelation function resulting from any distribution, of particle sizes would be known.

Since the acf, is what is actually measured in practice, the above relation must be inverted to yield the particle size distribution: x(r) K-l(g(t)) Accordingly, for the measured acf for a sample of particles, the size distribution for those par':icles can be extracted by applying the operator K 1 to the measured acf, The operator K-1 is the generalized inverse of the operator K. In practice however, the acf is "ill-conditioned" so that the inversion process is generally difficult and complex, although there are a number of known techniques for performing the inversion.

An example of the form K-1 for one particular commonly used prior art extraction technique is: (2tK Y. =)-tg In this example, X is a vector whose components are the proportions of the particles of each size, a is a SUMSTITUTE

SHEET

WO 89/00286 PCT/IJS88/02222 -7vector whose components are the values of the acf at different points in time, as computed by the autocorrelator, S is a matrix relating A to and U is a matrix which increases the conditioning of the inversion. Et is the transpose of the matrix I.

Alpha conttols the amount of conditioning imposed on the solution. The inverse operator K 1 in this case can be written K- 1 (KtE aH)-lt, where alpha is a smoothing parameter determined conventionally. The inversion is usually performed along with some non-negativity constraints imposed on the solution; these constraints are formally part of the inverse operator K 1 Other known methods for inversion are the histogram method, the singular value decomposition method, the delta function method, and the cubic spline method.

The size distribution, obtained from this extraction or "inversion" process can be expressed either as a continuous distribution as implied by where the distribution is defined for particles of any size, or as a discrete size histogram expressed by the vector x, where the distribution of particle sizes is defined at only a set number of particle sizes. The vector x is representative of (a group of numbers (x(rl), x(r 2 x(rn)) giving the relative proportion of scattered light intensity from particles of size r l respectively. The size distribution x(r) is referred to as a size histogram x herein below.

The size distribution x(r) and size histogram, x, are "intensity weighted" functions since these are representative of the relative proportion of particles as charact.rized by the subsTITUT

"^SEE

_i i i, i iii WO 89/00286 PCT/US88/0222 -8relative amount of scattering intensity from particles of each size. However, these intensity weighted functions are dependent on the angle at which the measurement of scattered light was made.

That is, the apparent proportion of particles of ch size, as evidenced by the scattered light intensity contribution of particles of different sizes, depends on the angle at which the measurement is made. Thus, size distributions made at different angles cannot be directly compared using the intensity weighted distribution x(r) or histogram x.

Accordingly, if the amount of light scattered per particle, as a function of the scattering angle, is known, either through theoretical calculations or by an empirical method, the intensity weighted size distributions x(r) and histogram x at each angle can be directly compared by first converting those functions to corresponding mass, volume, or number weighted size disltributions.

For example, the intensity weighted histogram A may be converted in accordance with; -V4 In this expression, v is the mass, volume or number weighted size histogram and Q is the conversion matrix between the intensity weighted histogram, x, and the mass, volume or number weighted histogram, X. Similarly, the size distribution x(r) may be converted into a corresponding mass, volume or number 3Q distribution function Since all of these converted histograms and distribution functions SUBSTITUTE SHEET

A

I

i\ WO 89/00286 PCT/US88/02222 With the latter method, the data collected at different angles are substantially independent, and therefore data collected at one angle complement those collected at other angles. For example, data collected at lower scattering angles are generally more sensitive to the presence of large particles in the sample while, conversely, Cata collected at large scattering angles are more sensitive to the presence of smaller particles. A sample containing both large and small particles can therefore be accurately sized by using the data from two or more angles, where relatively lower angle or angles provide information about the larger particles and relatively high angle or angles provide information about the smaller particles. In contrast, measurement at a single low angle would provide relatively little and possiblly obscured information about the smaller particles and hence the sizing resolution would be poor.

The prior art method of using several angles to enhance the sizing resolution involves simply making measurements at two or more angles and averaging the volume weighted histograms resulting from the measurements made at the two or more angles. Symbolically, the process of combining information obtained at several scattering angles by averaging results can be expressed by: x1 K-l(gl(t),e 1 A2 K-l(g 2 2 1 i xm K'l(gm(t),8) where the subscripts refer to measurements made at the scattering angles 01 through e m The inclusion of 0 as an argument of the operator K-1 indicates that the inversion process, that is, the SUBSTITUTE SHEET PCT/US88/02 2 22 WO 89/00286 -9provide the desired angle-independent information about the size distribution partioles, they are referred to generally below as y, o- n v(r), respectively.

I A volume weighted histogram and distribution function provide a measure of the proportion of the total volume of particles in a sample as a function of particle size. For example, 50% of the volume of a sample of particles might come from particles of size 0.1 micron and the remaining 50% from particles of size 0.3 microns. Similarly, the mass weighted histograms and distributions provide a measure of the mass of particles in a sample as a function of size and the number weighted histograms and distribution of the numbers of particles in a sample as a function of size. For particles of the same density, the mass and volume weighted histograms and functions are the same. Volume, mass and number weighted size histograms and distributions are generally more useful than the corresponding intensity weighted size histograms or distributions since the,former relate to quantities which can be directly measured by other means.

All of the prior art light scattering measurement techniques are characterized by low resolution and poor reproducibility, the princ.-pt1 drawbacks of such methods. With respect to DLS sizing measurements, efforts have been made to try to increase the resolution. The general methods used to increase resolution either attempt to improve the signal-to-noise ratio of the measurement by collecting intensity data over a long period or over a large number of short periods and then averaging the results, or by using intensity data collected at several angles.

SUBSTITUTE

SHEET

measurement conditions, where m is an integer equal to or greater than one, and n is an integer equal to or greater than zero, and the sum of m and n is S WO 89/00286 PCT/US88/02222 -11operator K- 1 depends on the scattering angles. Each of the m intensity weighted histograms, may be converted to an angle-independent volume weighted histograim, and then the m volume weighted histograms may be averaged to produce the "enhanced" resolution result, X: vX (1/m)[Xl v2 Vml However, this volume weighted distribution, X, is not necessarily the solution which is the best "it to all the data'. The size resolution obtainable for a single measurement at a single angle is quite low and the presence of particles of some sizes may not be detected at some angles. Thus, even when the intensity histograms are converted to volume histograms, the histograms obtained at different angles may give very different and apparently contradictory information.

Briefly, the invention disclosed in U.S.

Patent Applicati.on Sei No.l-rO1-44 is an apparatus and method which provides a measure of the size distribution of particles dispersed in a fluid based upon an optimum combination of CLS measurements and DLS measurements, providing a resultant measurement characterized by relatively high resolution pilrticle sizing. More particularly, in accordance with that invention, DLS data representative of the autocorrelation function, or power spectrum, of the detected intensity of scattered light at a plurality of angles about a sample, is optimally combined with CLS data representative of the average total detected intensity at those angles, to provide an angle-independent, high resolution size distribution Sv(r). The size distribution may be expressed in terms of the continuous function v(r) or the SUBSTITUTE

SHEET

to said m correlation signals for generating a composite correlation signal, ii. means for generating a /3 3 WO 89/00286 PCT/US88/02222' -12histogram y, and may represent distributions weighted by mass, volume, number, surface area, or other measures.

By way of example, in combining the DLS and CLS data, an angle independent volume weighted histogram may be determined from: X J-l(g 1 where are the determined autocorrelations of the detected light intensities at m scattering angles and where i(em) are the detected average intensities at the respective m scattering angles; as well as n additional angles m+n, where m is an integer equal to or greater than one and n is an integer equal to or greater than zero. In this form of the invention, the DLS measurements are made m angles and the CLS measurements are made at m+n angles, including the same angles at which DLS measurements are made. J-1 is a single operator which acts simultaneously on all of the autocorrelation functions and average intensity values to provide the "best fit" to all the data. This is in contrast to the m separate K 1 operators, one for each angle, set forth in equations above. In various forms of the invention, rather than different angles, acf and intensity measurements may be made at the same angle, but under difficult conditions, for example, temperature, hydrodynamic solution characteristics, or polarization angles, which establish independent intensity characteristics at the sensor.

SUBSTITUTE SHEET ft4 Ii WO 89/00286 PCT/US88/02222 -1- With the invention, the operator j 1 incorporates the information frcm the CL$ measurements as well as che independent tinformation from the DLS measurements, in a manner appropriately normalizing the autocorrelation functions measured at the different scattering conditions. For simplicity, the following descriptions will characterize the various measurements as being made at angles denoted Gi although it is -nly necessary that the measurements be made under conditions hich result in independent intensity characteristics.

The resultant distribution, y, based upon the J- 1 transformation of the autocorrelation functions and the classical scattered intensities, simultaneously in a single procedure, provides an increase in the sizing resolution of the determined particle distributions compared to the prior art techniques which are based upon either the autocorrelation functions £f the classical scattering intensities, but not both.

Briefly, according to the invention disclosed in U.S. Patent Application Serial No.

817,048, a system is provided for measuring the size distribution v(r) of particles dispersed in a fluid sample, where r is representative of particle size.

The system includes means for illuminating the sample with a light beam directed along an input axis.

Either a coherent or a non-coherent light source may be used.

A light detector detects the intensity of light from the light beam at m points angularly dispersed from said input axis at a plurality of angles em, where m is an integer equal to or greater than one). The detector generates m intensity SUBSTITUrT SHrE ;11 KI:: WO 89/00286 PCTUJS88/02222 -14signals, each of the intensity signals being representative of the detected intensity of the light from the light beam as a function of time at a corresponding one of the m points. In various forms, the invention may be embodied in a homodyne or a heterodyne configuration. In the homodyne form, only scattered light is detected at the m points during the intensity signal measurements, while in the heterodyne form, a portion of the beam is directly incident on the detector at the m points, so that the intensity signal corresponds to a beat signal resulting from both scattered and non-scattered portions of the light beam.

In one form, an autocorrelation processor generates m correlation signals each of the correlation signals being representative of the autocorrelation function of a corresponding one of the intensity signals. Each of the correlation signals equals an associated transformation Ji of the distribution where The transformations may be linear or non-linear. Since the autocorrelation functions for the intensity signals are the Fourier Transforms of the power spectra of those signals, the autocorrelation processor is, in one form of the invention, an autocorrelator which directly generates the m correlation signals as m time domain autocorrelation signals gi(t), where t is time and In other forms, the autocorrelation processor includes a spectrum analyzer which generates the m correlation signals as m frequency domain power spectrum signals Gi(f), where f is frequency and Since the power spectrum signal is the Fourier Transform of the autocorrelation signal, the power spectrum signas SUBSTITUTE

SHEET

WO 89/00286 PCT/US88/02222 Gi(f) may be used to provide the same information as the autocorrelation signals gi(t).

A light detector also detects the time average intensity of scattered light from the light beam at the m points as well as n additional points angularly displaced from the input axis, where n is an integer greater than or equal to zero. The latter detector generates average signals representative of the time average of the intensity of scattered light detected at the respective m+n points.

A size processor, responsive to the correlation signals and the average signals, generates a signal representative of the distribution The size processor generates a composite correlation signal representative of a weighted direct sum of the m correlation signals. The size processor determines a composite transformation operator J-3 which is related to the transformations Ji and the n average signals.

The size processor transforms the composite correlation signal in accordance with the determined composite transformation operator thereby providing a resultant signal which incorporatei? the size distribution information of both the CLS and DLS data and is representative of the siae distribution v(r).

In accordance with the invention, either the composite correlation signal or the composite transformation operator is substantially scaled to the average intensities of the scattered light at the respective ones of the m points. This scaling, or normalization, permits the DLS data represented by the composite correlation signal to be optimally combined with the CLS data represented by the average signals.

SUBSTITUTE

SHET

SUBSTITUTE SHEET or.

WO 89/00286 PCT/US18/02222 -16- In one form of the invention, the transformations Ji are linear transformations and the composite transformation operator J-1 is the generalized inverse of the operator corresponding to the direct sum of the operators for the associated transformations Ji, The inverse transformation operator J-1 may correspond to the inverse of the matrix corresponding to the direct sum of the associated transformations Ji. Alternatively, the operator J-1 may correspond to [ata al]-iat where a is the matrix corresponding to the direct sum of the matrices coresponding to the associated transformations Ji, at is the transpose of the matrix is a conditioning matrix, and alpha (cc) ~s a smoothing parameter. Further, all components of the vector representative of the distribution v(r) may be constrained to be greater than or equal to zero.

In another form, the associated transformations are non-linear, with the size distribution being characterized by where R is a characterization parameter vector having k components. In this form, the composite transformation operator J- 1 is the E solution algorithm for 8 m q I 1 (Jij[v 2 0, a 8 pS i=1 j=1 where i is an integer m, j is an integer I is an integer k pa is the ath component of e and where gi(tj) is the autocorrelation function of the intensity signal for the ith of Said angle at the jth time interval and SUBSTITUTE SHEET SUBSTITUTE SHEET I l l k I 1 WO 89/00286 PCT/US88/02222 -17- Jij is an operator related to the associated transformations. The model size distribution v(r,p) may for example be characterized in terms of parameters K and a, the mean particle size of the actual size distribution, and the standard deviation of the actual size distribution, respectively. The solution algorithm for the p% minimizes the squares of the residuals Jij[v(r,p)]-gi(tj) for the various points in time tj for the various acf's gi. More particularly, v(r,p) may have the form: v(rR) 1 e-(r-r) 2 /2a 2 2ir 2 where the Jij operator has the form: dr e-r(r, ei)tj o 0 where r(r, Oi) has the form: r(r, ei) [4 n sin 1 9i] 2 kBT 2 61r where n is the rv.ractive index of the sample, X, is the wavelength of the light illuminating the sample, k B is Boltzmann's constant, n is the viscosity of the sample and T is absolute temperature.

In another form, the composite correlation signal operator controls the weighted direct .sum of the m correlation signals to be unity normalized and the composite transformation operator is substantially scaled to the average intensities of light scattered from the light beam at the respective ones of the m points. In yet another form, the composite correlation signal generator controls the weighted direct sum of the m correlation signals to SUBSTITUTE

SHEET

SUBSTITUTE

SHEET

C

PCT/US88/02222- WO 89/00286 -18be substantially scaled to the average intensities of light scattered from the light beam at the respective ones of the m points.

In other forms of that invention, the general method of using information exacted from measurements at two or more scattering angles can be applied to determine size and shape information about particles which are rod-like, ellipsoids or other forms, including "Gaussian coils".

In other forms of that invention, instead of making CLS and DLS measurements at two or more Scattering angles, such measurements .ay be made at one angle under different sets of conditions, for example, different temperatures or hydrodynamic solution characteristics or polarization of the light beam, providig complementary information which is processed to yield enhanced particle characteristic resolution for a wide variety of dynamic systems.

A critical aspect of the invention disclosed ,q 6 in U.S. Patent Application Serial N .18-1-O is the recognition that by combining CLS (classical light scattering) data and DLS (dynamic light scattering) data from multiple scattering angles: the resolution and repeatability of particle sizing may bz improved over prior art CLS or single angle DLS measurements.

To combine CLS data and DLS data is difficult and the determination of conditions under which the CLS and DLS data can be combined in a single, simultaneous analysis is a complicated process. In particUlar, one major problem in combining these two types of light scattering measurements is in properly normalizing the DLS data taken at different angles. Briefly, "normalization" refers to adjusting the amplitudes of the sU ITUTE S/ SUBSTITUTE

SHEET

i _1 I i SUBSTITUTE

SHEET

SWO 89/00286 PCT/US88/02222 -19autocorrelation functions (acfs) at different angles with respect to the CLS data so that data from all the DLS angles can be analyzed within one comprehensive model.

The normalization of the DLS data to the CLS data is important in view of the following. A CLS measurement at a single angle is the average value of the light scattered from particles of all sizes in the sample being measured, weighted by the intensity of light scattered by particles of each size. The amount of light scattered at a particular angle from a particle of a particular size depends on both the size of the particle and the scattering angle. To make particle size measurements using CLS data alone (that is, average intensity measurements at a multitude of angles), measurements at the selected angles are made and then an analysis procedure is used to find a particle size distribution corresponding to the measured pattern of angular scattering intensity. The selected distribution must be such that the light scattered from every size of particle in the distribution, weighted by the angle dependent intensity of light scattered by particles of each size, must be close, at each angle, to the measured value of the average scattered intensity at that angle.

CLS data is thus a subset of DLS data since the magnitude of the CL5 intensity at a particular angle is simply the value of the zero time point of the homodyne DLS autocorrelation function (acf) at that angle. The relation between CLS data and DLS data may be appreciated by considering a plot of the CLS data in a cartesian coordinate system in which the x axis represents scattering angles (e.g.

SUBSTITUTE

SHEET

L JL=LLJa a o eacn size, a is a SUBSTITUTE

SHEET

WO 89/00286 PCT/US88/02222 from 0 to 180 degrees), and the y axis represents scattering intensity. In such a coordinate system, a given distribution of particles would be characterized as a curve in the X-Y plane. The height of the curve at any point would be the scattering intensity at the angle corresponding to that point. This curve, or at least a number of points along this curve, correspond to a CLS measurement.

The z axis represents the delay time of an acf of the scattered light intensity. Generally, acf's have the form of decaying exponentials (or sums of decaying exponentials). A number of discrete poi'nts along the CLS curve (in the xy plane) correspond to the angles at which DLS measurements may be made. The DLS data for each point (or angle) defines a curve in a plane parallel to the Y-Z plane. These acfs (at one or more angles) are the DLS data. The shape of the acf at each angle in i general would be slightly different unless all the particles were the same size. Thus, the matrix of acf data is used in a measuremnt of the type defined in U.S. Patent Applicai"on-Seral No.r 7-4 In the prior art, only the curve in the X-Y plane alone or a single one of the acfs wotlId be analyzed to obtain a (low resolution) particle size distribution The system of U.S. Patent Applicaion -Seria- No.'8-1-7,04 provides particle distribution measurements utilizing the two dimensional "measurement" surface on which all the values of the acfs at all delay times and all scattering anglres lie. The intersection of this surface with the X-Y plane is the CLS data curve.

SUBSTITUTE SHEET i i I I wO 89/00286 PCT/US88/02222 -21- To analyze the matrix of data shown, a model is made describing how the measurement surface, or more precisely, the discrete points on the measurement surface at whi-h actual measurements are made, varies when the parameters histogram bin heights) of the particle size distribution vary.

Then, when a particular measurement is made, the particle size distribution which leads to a measurement surface closest to the measured measurement surface is the best estimate of the true particle distribution. The best estimate is found by a curve fitting algorithm, such as a non-negative least square (NNLS) algorithm.

The importance of normalization to this process is that in order to accomplish the curve fitting, the measured acfs (which are ordinarily measured without regard to their absolute height) need to be normalized so that their amplitudes, i.e.

zero time values, are exactly the amplitudes of the CLS data curve at the angles at which the acfs are measured. If this normalization is not undertaken, the measured measurement surface generally differs from the model measurement surface since in the model measurement surfaces all acfs at time zero must lie on the CLS data curve. Thus, the measured data are normalized to be appropriate to this model.

In order to perform the DLS data normalization, CLS data the average intensity of scattered light) must be measured at every DLS angle to obtain the normalization constants. For this reason, U.S. Patent Appl ca-on-Se -ial- 4,616,641 No.k28-3-748 defines an analysis in which the DLS data is measured at m angles (m 0) and the CLS data is measured at those m angles and at n additional SUbSTuE SHEET IirU SUBSTITUTE SHEET WO 89/00286 PCT/US88/02122 -22angles, i.e. CLS data needs to be measured at all the DLS angles and optionally at some additional angles.

However, it is difficult to accurately obtain the CLS data used for the DLS normalization.

More importantly, small errors in the normalization constants lead to serious errors in the extracted particle size distribution. Therefore, it is desireable to find an improved model by which to analyse the data.

Because of the mathematical form of acfs, the amplitude information is mathematically separable from the rest of the informatior contained in the acf, namely the information contained in its shape, By performing a mathematical transformation in one of several ways, the DLS amplitudes can be removed from the model, thereby allowing the curve fitting to be accomplished without knowledge of the amplitudes of 'he acfs, i.e. without a normalization in the manner defined in U.S. Patent ANp-ioat-oon-Seria- No, t aOh 2 It is an object of the present invention to provide an improved apparatus old method for measuring the distribution of particle sizes dispersed in a fluid, Another object is to provide an improved particle size distribution measuring apparatus and method characterised by relatively high resolution.

SUMMARY OF THE INVENTION Briefly, the present invention is an improved apparatus and method which provides a ieasure of the size distribution of particles dispersed in a fluid based on an optimum combination SUBSTITUTE

SHEET

indicates that the inversion process, that is, the ISUBSTITUTE SHEET 1 'WO 89/00286 PCT/US88/0222 -23of CLS measurements and DLS measurements, without requiring normalization of the DLS acf's to the CLS values at the appropriate angles.

In one form of the invention, the acf's may be determined in the same manner as described in U.S.

Patent App.liation 4-Ser-l No. 17,- 4 All the shape information in the acf's then is extracted by calculating the moments of the unknown particle size distributions based on the acf's at each DLS angle.

Each acf provides a differently weighted estimate of the moments of the particle size distribution. Ways of extracting the moments from an acf are well known and detailed in the scientific literature. The zeroth moment of the particle size distribution is directly related to the amplitude of the acf. The higher order moments and the inverse moments contain the information about the shape of the acf. By determining a sufficient number of moments from an acf, substantially all the shape information in that acZ can be utilized.

After extracting the moments, a curve fitting algorithm is used to determine the particle size distribution mostly closely matchirng the CL

S

data curve (if this has been measured) and all the extracted nmments of all the acfs, except the zeroth moments which contain the amplitude information (i.e, normalization). Thus, with this form of the invention, all the aOf information and all the CLS data is used in the analysis except for information concerning the amplitude of the acf's. Since no acf amplitude information is used, no normalization of the acfs to CLS data need be done. Further, CLS data need not be measured at every DLS angle, siUTE SUBSTITUTZ

SHEE

L_1 i i o=veraL angles.

SUBSTITUTE

SHEET

i1 WO 89/00286 PCT/US38/02222 -24- In another form the invention not requiring normalization of the DLS data to CLS data, the amplitude information normally present in an acf is explicitly removed by normaliztng the amplitude of each acf to unity. Such a normalization is trivial to perform and, more importantly, obviates the necessity of making CLS measurements at each DLS angle and introducing significant error into the DLS measurements by normalizing them with the CLS measurements. The unity normalized DLS measurements are related by a special transformation ("DLS kernel") to the particle size distribution. The transformation differs from the DLS kernel which relates a particle size distribution to acfs normalized by CLS data (as described in patent a--p-pLa-1.ton-Sera.l No. &l-1- 7 04) The unity normalization in this form of the present invention is, however, such that CLS measurements need not be made at every DLS angle.

In a third form of the present invention not requiring normalization of the DLS data to the CLS data, the data analysis may be performed without using information about the acf amplitudes by allowing the amplitudes to be "floating" parameters in the curve fitting process. Allowing the amplitudes to float means that instead of normalizing the DLS kernel or acfs with the CLS data, the amplitudes of all the acfs are treated as unknown, adjustable parameters in the curve fitting. This process effectively allows the model measurement surface wider latitude in matching the measured acfs, by permitting frea adjustment of the acf amplitudes to more closely approach the measured measurer et surface. In patent applicao-.-s.a-- No. 84-04 U4 S SUBSTITUTE SHEET SUBSTITUTE

SHEET

SWO 89/00286 PCT/US88/02222 the amplitudes of the acts were normalized by CLS measurements. In that case, any small error in measuring a CLS data point forces ,~:rei'ponding errors in the entire acf, which for example, may comprise 100 data points which the CLS data point was being used to flormalize. In this third form of the present invention, the amplitude of the acf is determined by the acf itself (through the curve fitting process), preventing the situation where a single measurement controls the errors in a large number of other measurements. Again, CLS data need not be measured at every DLS angle and no normali;ation at all would be performed.

More particularly, the present invention is a system for measuring the size distribution v(r) of particles dispersed in a fluid sample, where r is representative of particle size. The system comprises a light source for illuminating the sample with a light beam directed along an input axis. A detector, or set of detectors, detects the intensity of light scattered from the light beam by the sample for i measurements, each made under an associated one of a set of m measurement conditions, and also for n measurements, each made under an associated one of a set of n measurement conditions, where m is an integer equal to or greater than one, and n is an integer equal or greater than zero, and the sum of m and n is equal to or greater than two, and at least two of the m+n measurement conditions are different.

The detector is adapted to generate m intensity signals, each of the intensity signals being representative of the detected intensity of the scattered light as a function of time under a corresponding one of the measurement conditions of SUBSTITUTE

SHEET

WO 89/00286 PCT/US88/02222 -26said set of m measurement conditions. The detector also generates n average signals, each of the n average signals being representative of the average intensity of the scattered light detected under a corresponding one of the measurement conditions of the set of n measurement conditions.

An autocorrelator generates m correlation signals. Each of the correlation signals is representative of the autocorrelation function of a corresponding one of the intensity signals and is equal to an associated transformation of the distribution v(r).

A size processor is responsive to the correlation signals and the average signals. The size processor includes a preprocessor responsive to the m correlation signals for generating a composite correlation signal. The size processor also generates a composite scattered light signal from the composite correlation signal and the n average signals. The size processor also determines a composite transformation operator J- related to the associated transfetmations and the n average signals. The composile scattered light signal is transformed in accordance with the determined composite transformation operator J-l to generate a size distribution signal representative of the distribution v(r).

In the system, the m+n measurement conditions are controlled whereby the intensity signals are mutually independent and the average signals are mutually independent.

In one form of the invention, the preprocessor is responsive to the m Correlation signals for generating m moment signals, where the m SUBSTITUTE

SHEET

I_ i i -i SUsTIU SHET -i: 'WO 89/00286 PCT/US88/02222 -27moment signals are representative of at least one of the moments of each of the m correlation signals other than the zjroth moments of the m correlation signals. The preprocessor further generates a weighted direct sum signal, where the weighted sum signal is representative of the direct sum of the m moment signals and corresponds 'o the composite correlation signal.

In another form of the invention, the preprocessor is responsive to the m correlation signals for generating m ur.ity normalized correlation signals, where each of t"e unity normalized correlation signals is representative of the unity normalized form of the corresponding one of the m correlation signals. The preprocessor generates a direct sum signal, which is representative of the direct sum of the unity normalized correlation signals, and corresponds to the composite correlation signal.

In still another form of the invention, the preprocessor generates a direct sum signal, where that sum signal is representative of the direct sum of the m correlation signals and corresponds to the composite correlation signal.

In each of the above forms of the invention, the composite scattered light signal may be the direct sum of the composite correlation signal and the average signals.

In these forms of the invention, the "direct sum" may be defined as set forth in Mostow, G.D., Sampson Linear Algebra, McGraw Hill Book Company, New York 1969, page 165 (particularly, a vector space U is the direct sum of the subspaces if every vector x in U can be expressed i: SUBSTITUTE SHEET SUBSTITUTE

SHEET

WO 89/00286 PCT/US88/02222' -28one and only one way as a.sum where Xj is in Further, a "composite signal" of a group of signals is defined as any transformation of the elemental signals which preserves all or the bulk of the information contained by the union of the elemental signals. A "direct sum" is a simple form of a composite signal as so defined above. Moreover, the moment signals, however they might be combined, are also a composite signal derived from the original acf's.

A composite transformation operator is defined as a transformation operator comprising the mathematical relationships between a particle size distribution and each datum in a composite signal.

BRIEF DESCRIPTION OF THE DRAWINGS The foregoing and other objects of this invention, the various features thereof, as well as the invention itself, may be more fully understood from the following description, when read together with the accompanying drawings, in which: Fig. 1 shows in schematic form, an exemplary system embodying of the present invention; Fig. 2 shows, in block diagram form, an embodiment of the detector and signal processor of the system of Fig. 1; Fig. 3 shows, in block diagram form, another embodiment of the detector ard signal processor of the system of Fig. 1; Fig. 4 shows a flow chart illustrating the operation of the system of Figs. 1 and 2; and Figs. 5-7 show detailed flow charts for three implementations of the data processing step of the flow chart of Fig. 4.

SUBSTITUTE SHEET signals.

SUBSTITUTE

SHEET

'WO 89/00286 PCT/US88/02222 -29- Figs. 8 and 9 illustrate the mass weighted size distribution determined by an embodiment of the present invention for an exemplary sample of polystyrene latex spheres dispersed in water.

Fig. 10 shows a flow chart illustrating the operation of an embodiment of the present invention; Fig. 10A shows, in schematic form, an exemplary sensor format for the present invention; Figs. 11-13 show detailed flow charts for three implementations of the data processing step of the flow chart of Fig. DESCRIPTION OF THE PREFERRED EMBODIMENTS Fig. 1 shows an exemplary system 10 in accordance with the invention. The system provides an output signal representativ2 of the size distribution of particles v(r) dispersed in a fluid sample 12, where r is representative of particle size. The sample may, for example, include discrete particles or macromolecules suspended in a liquid, or may include ionized macromolecules in a solvent, or may include discrete particles in an aerosol, or any configuration wherein the particles-to-be-sized are dispersed in a fluid, and are subject to Brownian motion in that fluid.

In the embodiment of Fig. 1, the sample 12 is positioned within a bath chamber 14 filled with a temperature-controlled index of refraction matching medium. The system 10 includes a light source 18 which provides a collimated light beam 20 directed along an input axis 24 and focussed by lens 16 onto the sample 12. In the presently described embodiment, the light source 18 is a laser, although in other forms of the invention, a non-coherent light SUBSTITUTE

SHEET

SUBSTITUTE

SHEET

I-

WO 89/00286 PCT/US88/02222 source may be used. The present embodiment is a homodyne configuration in which substantially no non-scattered light is permitted to reach detector 28. In a heterodyne form of the invention, portion of the light beam from source 18 may be coupled directly to the sensor 30, for example, by a fiber optic link indicated by the broken line 22 in Fig. 1.

A light detector 28 includes a sensor having a sensing axis 32 which is positionable at a plurality of points equidistant from and dispersed angularly about; the sample 12 along an arc 36. As shown, the input axis 32 is displaced by an angle 8 with respect to axis 24. In various forms of the invention, a detector may be successively positionable along arc 36, or alternatively, a plurality of light detectors might be fixedly positioned at discrete points along the arc 36. The detector ?8 provides output signals along lines 28a to a signal processor 40 In the present embodiment, aperture defining devices 37 and 38 are positioned With respect to sensor 30 and the axis 32 in a manner restricting the light detected at sensor 30 to be within a predetermined coherence area.

In addition, the present embodiment includes a pair of polarizers 39a and 39b positioned along the axis 24 before the sample and along the axis 32, respectively. The filters 39a and 39b permit passage only of portions of the light beam and scattered lii;ht, respectively, having predetermined polarization. By selectively controlling the polarization angle of these filters, substantially independent intensity signals may be generated for a single angle 9 for each orthogonal polarization angle. For example, in one form, the filter 39a SUBSTITUTE

SHEET

II

weighted direct sum of the m correlation signals to SUBSTITUTE

SHEET

WO 89/00286 PCT/US88/02222 -31passes right circularly polarized light and the filter 39b may be selectively adapted to pass right circularly polarized light or left circularly polarized light. Alternatively, the filter 39a passes vertically polarized light and the filter 39b may be selectively adapted to pass light characterized by one of two different orthogonal polarizations.

Fig. 2 shows one form for the detector 28 and the signal processor 40 in which the detector 28 includes a photomultiplier 44 and associated pulse discriminator 46 and an integrator 48. In this form, the signal processor 40 includes an autocorrelator a signal analyzer 52 and a controller 54.

Fig. 3 shows an alternate form in which the detector 28 includes an interferometer 60 and a photomultiplier 62, which provide a signal representative of the power spectrum of the light intensity at the detector 30, and a photomultiplier 63 and an integretor 64, which provide a signal representative of the time average of the light intensity at detector 30. The signal proceZsor includes a signal analyzer 66 and a controller 68.

In Figs. 2 and 3, the signal analyzers provide the output signal representative of the size distribution v(r).

In operation, briefly, the light source 18 illuminates the sample 12 along the input axis 24 and the sensor 30 of detector 28 detects the intensity of the light scattered by the sample at a plurality of points along arc 36. The intensity measured at the detector 28 is an interference pattern resulting from the in phase contributions of the light scattered from each molecule in the scattering volume of sample S-13 S

E~

SUBSTITUTE

SHEET

WO 89/00286 PCT/US88/02222 -32- 12. In liquid or gaseous samples, the molecules are in motion and the interference pattern at the detector 28 is modulated by the motions of the scattering particles. In the absence of external fields, the motions of the particles are random, and are due just to thermal fluctuations. The fluctuations in scattered electric field at the detector caused by these random motions comprise a stationary random process. The second momeint of the process is defined by 1 T/2 1 g(t) lim If (1) T- -T/2 where e(r) is the electric field of the scattered light at the detector 28 at time T. g(t) is the autocorrelation function (hereinafter abbreviated "acf") of the process. The acf is a measure of the correlation between the configuration of the scattering molecules at a given time compared to that at any later time, e.g. at very short delay times, the configuration of particles as well as the scattered electric field measured at the detector 28 closely resembles the original configuration; as time passes, that resemblance diminishes. Because the 2 degree of correlation depends on the speed with which and the mechanism by which the configurations change, the acf provides characterization of the dynamics of the particles in the sample. In the configuration of Fig. 2, the acf of the scattered light is determined Sby a digital correlator 50. Alternatively, in the configuration of Fig. 3, the photomultiplierdiscriminator-autocorrelator elements of Fig. 2 are replaced by the interferometer 60 and photomultiplier 62. In the latter configuration, the power spectrum, SUBSTITUTE SHEET i -j L i~YII--L-- LIII WO 89/00286 PCT/US88/02222 -33which is the Fourier Transform of the acf is measured. In both the configuration of Fig. 2 and that of Fig. 3, the information obtained is equivalent and both the acf and power spectrum signals are referred to herein as the DLS data. In practice, the choice between the two forms is determined by the rapidity of fluctuations in the light scattered by the sample 12. Preferably, fluctuations decaying on a time scale slower than 0.01 microseconds are measured by autocorrelation and those faster than 0.01 microseconds are measured by interferometry.

The photomultiplier 44 and integrator 48 of Fig. 2 and the photomultiplier 63 and integrator 64 of Fig. 3 provide signals representative of the average detected intensity at the detector 30. Those signals are referred to herein as the CLS data.

The elements 18, 28, 44, 46, 48 and 50 of the system 10 of Figs. 1 and 2 may be implemented in part by commercially available devices, such as the Model LSA2+ photon correlation spectrometer and Model 1096 Correlator, manufactured by Langley Ford Instruments, division of Coulter Electronics of New England, Inc., Amherst, Massachusetts.

Alternatively, these elements may be implemented by the Coulter Model N4 photon correlation spectrometer, or by the System 4700 spectrometer manufactured by Malvern Instruments, Inc., Framingham, Massachusetts or Series MMI000 spectrometer manufactured by Amtec, Villeneure-Loubet, France, together .ith a Coulter Model 1096 correlator.

Fig. 4 shows a flow chart illustrating the general operation of the system of Figs. 1 and 2, including a data collection phase, denoted A, and a SUBSTITUTE

SHEET

kI L i

J

-14 LSUBSTITUTE SHEET WO 89/00286 PCT/US88/02222' -34data processing or analysis phase, denoted B. In accordance with the invention, as shown in Figs. 1 and 2, the data collection phase A is performed with elements 18 and 28 in two modes. In the first mode, DLS measurements are made an m different angles and the autocorrelator 50 provides m autocorrelation functions, denoted acf 1 acf m in Fig. 4. In the second mode, the CLS measurements are made at the same m angles, where m is an integer greater than or equal to one, and at n additional angles, where n is an integer greater or equal to zero and the integrator 48 provides m+n integrated intensity values, denoted int int m+n in Fig. 4. These acf and integrated intensity measurements may be made at the same time or sequentially since the resultant data for each measurement is substantially independent. In various forms of the invention, rather than different angles, successive pairs of acf and average intensity measurements can be made at the same angle, but under different conditions, for example, temperature, hydrodynamic solution characteristics, or polarization angles, which establish independent intensity characteristics at the sensor 30. For example, the measurement conditions may be controlled by selectively controlling the polarization of light incident on the sample (the "input light") and the polarization of light scattered from the sample and incident on the detector (the "output light"). The polarization of the input and output light may be controlled to be at the same polarization horizontal/horizontal, vertical/vertical, right circular/right circular, or left circular/left circular) or at orthogonal polarizations horizontal/vertical, suBSTITUTE

SHEET

i: I L IL

S

O B STI SIET WO 89/00286 PCT/US88/0222 vertical/horizontal, right circular/left circular, left circular/right circular. Also, successive pairs of acf and average intensity measurements can be made at various combinations of angles and these conditions.

For any of the combinations of angles and conditions at which the CLS and DLS data is measured, Figs. 5, 6 and 7 illustrate three different signal processing methods which can be implemented by the signal analyzer 52 under the control of controller 54. Each method includes a Normalization Phase, denoted A, and Model Phase, denoted B, and Analysis Phase, denoted C.

In each of the forms of Figs. 5-7, in the Normalization Phase, the m acf's are -irst normalized with the angular intensity (CLS) data so that those autocorrelation signals are scaled to the values of the average detected intensities at the m points of detection, forming m normalized acf's. A composite acf is then formed from the direct sum of the m normalized acf's.

In the Model Phase, an analytic, or discrete, model is constructed for use in the Analysis Phase so that the inversion of the equation v J l g l t gm(ft) i (i01 i(Em) i(em+l), .i may be performed. Each of Figs. 5. 6 and 7 indicates a different exemplary and known form for constructing the model in a conrentional manner for ill-conditioned f'l:.tions.

In Fig. 5, a model size distribution is generated as a histogram having all non-negative components and a kernel J is formed for the SUBSTITUTE

SHEET

SUBSTITUTE

SHE':"

WO 89/00286 PCTiUS88/02222' -36transform. To find the size diotribution v(r) using this kernel, a prograutaid digitaX computer using Non-Negative Least Squares (NNLS) techniques performs a "best fit", see C. Lawson and R. Hanson, "Solving Least Square Problems", Prentice Hall, Inc., Englewood Cliffs, 1974.

In Fig. 6, a model size distiibution is generated as a smoothed histogram with all non-negative components. A kernel J and a smoothing martrix H are formed and a aharacteristic alpha (m) is generated. Then matrix operations are performed on J, H and alpha to reduce the prublem to determining the residual vector having the minimum norm subject to linear constraints. To find the size distribution v(r) using J, H and alpha, a programmed digital computer using Least Distance Programming techniques, minimizes the norm of the residuals, [(ivj-gi(tj), subject to linear constraints, and matrix operations are then performed to determine see C. Lawson and R. Hanson, "Solving Least Square Problems", Prentice Hall, Inc., Englewood Cliffs, New Jersey, 1974.

In Fig. 7, the model size distribution is determined in terms of a parameterized distribution D(pl, p2, where the parameters pl, p2...pk determine the size distribution. Then an analytical form is determined for derivatives of V with respect to the A programmed digital computer uses Non-Linear Least Squares fitting techniques to determine the best values of the Pi characterizing v(r) using the program GRADLS; see P. R. Bevington, "Data R Auction and Error Analysis for the Physical Sciences", McGraw Hill Book Co., New York, 1969.

SUBSTITUTE SHEET SUBSTITUTE

SHEE

'WO 89/00286 PCT/US88/02222 -37- In all of the forms of the invention shown in Figs. 5-7, the acf's are normalized and then a composite acf is formed. However, in alternate forms of the invention, the acf's may be unity normalized and then used to form the composite acf Then the kernel J may be normalized and the normalized transformation operator may be used to generate the size distribution. Moreover, the composite acf's in Figs. 5 and 6 are direct sums of the normalized acf's (and the CLS intensities), where linear transforms are used. In other forms oP the inventions, as in Fig. 7, non-linear transforms are used.

Additionally, as described above, the scattered light for the sample is measured alone, forming the basis for a homodyne analysis.

Alternatively, the measurement may be made on the scattered light as augmented by non-scattered light from the source, so that a heat signal is in effect produced as the basis for a heterodyne analysis. In those forms, the signals g(t) may be the true autocorrelation of the detected intensity, or may be merely "closely related", for example, where a background level is subtracted out and the square root of the resultant signal is taken. The latter is particularly appropriate for gaussian light in a homodyne configuration.

In the preferred embodi:;,ent of Figs. 1 and 2, the light source 18, detector 28 (including elements 44, 46 and 48) and the autocorrelator 50 are provided by a Langley Ford Model LSA2+ photon correlation spectrometer and a Langley Ford Model 1096 correlator. The signal analyzer 52 and controller 54 are in the form of a Model Universe 137/T digital computer manufactured by Charles River SUBSTITUTE

SHEET

i ii L- 111: surrace. In patent :rplication -Seria No.1 84744-, SUBSTITUTE

SHEET

WO 89/00286 PCT/US88/02222 -38- Data Systems, Natick, Massachusetts, having a UNOS operating system with a Fortran 77 Compiler, as produced by Absoft Corpuoat:L n, Royal Oak, Michigan.

With this configuration, acf and CLS data is collected at two angles, for example, 900 and As the data is collected, it is automatically stored in the internal memory of the 1096 correlator. After the data at two angles has been collected and stored, the data is transferred to the 137/T computer through a serial (RS-232) port. The 1096 correlator is set to send the data in its 'five channels per line' fo-mat. The data, received through the RS-232 port is directly stored to a disk file on the 137/T computer.

The format of the data from the 1096 correlator, as stored in the 137/T computer disk file, is changed into the correct format for the signal processing program, cont2ng.fm. This format conversion is accomplished by processing the two data sets (one for each angle) using the 137/T computer programmed in accordance with the program condense.fm. Under control of the program, condense.fm, the 137/T computer reads the acf data along with the sample times used in the measurements (also contained in the data sets from the 1096 correlator) and condenses the 256 channels of data for each angle into two sets of 60 channels of data, by combining several channels of the original data into one channel of condensed data. The processed data is written out to a new disk file on the 137/T computer. This process of condensing the data is done only to speed the subsequent data processing, it has no material ailfect on the sizing results obtained. Under the control of the condense.fm SUBSTITUTE SHEET SUBSTITUTE SHEET SWO 89/00286 PCT/US88/02222 -39program, the 137/T computer then writes out the acf time points for the first angle, corresponding to the 60 condensed acf points at which the data is measureu. The data is written out in the FORTRAN format 5e15.6. Following the acf time points for the first angle are written out the 60 acf time points for the second angle in the same format. This is followed by the 60 condensej acf points for the first an. e and cnen the 60 condensed acf points for the second angle. The acf points are written out in the FORTRAN format 4e17.11.

At this point, a new disk file contains reformatted and condensed data equivalent to the original data collected by the 1096 correlator. This new data file, which contains only time points and acf points for the two angles, is prefixed by a header which informs the signal processing program cont2ang.fm how the data should be handled and gives the program some other information, such as the temE:-rature and viscosities of the sample, the scattering angles, and the like. The contents of the header are described in CONTIN Users Manual, Postfaeh 12 2209,D-6900 Heidelberg, BRD. A sample header is prefixed to the file using the 137/T computer editor, ted.

Once the header is prefixed to the data file containing the condensed two angle data, cont2ang.fm is called to process the data and provide the size distribution of the particles contained in the sample. Figs. 8 and 9 shows the mass weighted size distribution, v(r) generated with this configuration where the sample comprised a mixture of 450 A and 850 A (radius) polystyrene latex spheres (from Seragen Diagnostics) dispersed in water. The measurements SUBSTITUTE

SHEET

SUBSTITUTE

SHEET

WO 80/00286 PCT/US88/02222 were made at scattering afigles of 1440 (300 second measurement) and 63.20 (900 second measurement) at a temperature of 20°C. Fig. 8 shows a portion of the data output from the cont2ang.fm program for this example, and Fig. 9 shows a graph representative of this data.

The larger sized particles in this sample are less than twice as large as the smaller ones; for light scattering measurements, such relatively closely space peaks are extremely difficult to resolve. The enhanced resolution resulting from combining DLS and CLS data allows clear separation of the two peaks, as shown in Fig. 9.

Figs. 1-3 also represent embodiments of the invention in which it is not necessary to normalize the D1S data to the CLS data as is required in the embodiments described in conjunction with Figs. 5-7.

In the "non-normalizing" embodiments, the signal analyzer 52 and controller 54 of Fig. 2 and the signal analyzer 66 and controller 68 of Fig. 3 differ from the corresponding elements in the "normalizing" forms of the invention described above. In the preferred form, those elements may also have the form of a Model Univetae 137/T digital computer manufactured by Charles River Lata Systems, Natick, Massachusetts, programmed as described in more detail below.

Fig. 10 shows a flow chart illustrating the general operation of the system of Figs. 1 and 2 in one "non-normalizing" form of the invention, including a data collection phase, denoted A, and a data processing or analysis phase, denoted B. In accordance with this non-normalization form of the invention, as shown in Figs. 1 and 2, the data SUBSTITUTE

SHEET

~I 1--

L,

I SUBSTITUTE SHEET 'WO 89/00286 PCT/US88/02222 -41collection phase A is performed with elements 18 and 28 in two modes. In the first mode, DLS measurements are made an m different measurement conditions, where m is an integer greater than or equal to one, and the autocorrelator 50 provides m autocorrelation functions, denoted acf 1 acf m in Fig. 10. In the second mode, the CLS measurements are made at n different measurement conditions, where n is an integer greater than or equal to zero, and m+n is greater than or equal to two, and the integrator 48 provides n integrated intensity values, denoted int int m+n in Fig. 4. At least two of the m+n measurement conditions are different. These acf and integrated intensity measurements may be made at the same time or sequentially, unless the particle size distribution itself changes substantially with time, since the resultant data for each measurenent is substantially independent. Preferably, each of the m measurement conditions differ from each of the n measurement conditions, but there may be CLS and DLS measurements made at the same conditions. In minimum systems, by way of example, for m=1 and n=l, a single DLS measurement and a single CLS measurement at different measurement conditions may be performed, or for m=2 and n=0, two DLS measurements, at different measurement conditions, may be performed. In the latter case, the DLS measurements in effect inherently include some CLS data.

In various forms of the invention, the acf and average intensity measurements are made under different measurement conditions, for example, angles, temperature, hydrodynamic solution characteristics, or polarization angles, i.e. any conditions which establish independent intensity SUBSTITUTE

SHEET

i :1= WO 89/00286 PCT/US88/02222- -42characteristics at the sensor 30. Also, the various acf and average intensity measurements can be made at various combinations of angles and these conditions.

In the embodiment illustrated in Fig. 1, the measurements are made by a single sensor which is controllably positioned to the desired angular positions, as measured with respect to the input axis 24. Polarization controllers in the beam path, and/or sample temperature controllers might also be used to establish independent measurement conditions. However, in the present invention, it is particularly advantageous to utilize separate sensors for CLS measurements and for DLS measurements, since the known criteria for optimal collection optics are diametrically opposed for the respective CLS and OLS measurements. CLS measurements have a highest signal to noise ratio when, at a given angle, scattered light is collected from a relatively large portion of the sample. Conversely, for DLS measurements, high signal to noise ratio is obtained when the collection optics view a relatively small portion of the sample.

It has been found that two separate arcuate photodiode detector arrays (a forward direction array for low, e.g. 6-60 degrees, angle with respect to the beam input axis), and a side direction array for the high, e.g. 60-120 degrees, angle or 90 degrees plus or minus relatively small angles) are well suited for the CLS measurements, and a separate arcuate array 45, 90 and 150 degrees) of fiber optic coupled photomultiplier detectors are well suited for the DLS measurements. This configuration is illustrated schematically in Fig. 10A, where, for simplicity, only the light source, sample, CLP sensor arrays, and DLS sensor array are shown. In Fig. 10A, elements SUBSTITUTE' 8HL'Ur in ouner rorms or mne invention, a non-conerent light SUBSTITUTE

SHEET

'WO 89/00286 PCT/US88/02222 -43corresponding to elements in Fig. 1 are identified with the same reference designations. In the configuration of Fig. 10A, the sample 12 is held in a rectangular ctoss-section cuvette having opposed pairs of parallel planar lateral sidewalls. The CLS low angle (near zero degrees) sensors and CLS high angle (near 90 degrees) sensors are in arcuate arrays lying in the X-Y plane and the X-Z plane respectively. The DLS photodiode array also lies in the X-Z plane.

For any of the combinations of angles and/or conditions at which the CLS and DLS data are measured, Figs. 11, 12 and 13 illustrate three different signal processing methods which can be implemented by the signal analyzer 52 under the control of controller 54. Each method includes a Preprocessing Phase, and Model Phase, and Analysis Phase. In each of Figs. 11-12, the Analysis Phase is the same non-negative least squares (NNLS) curve fitting process as described in conjunction with Fig..

although alternative curve-fitting techniques may be used. In Fig. 13, the Analysis Phase is similar to that shown in Fig. 7.

In the form of Fig. 11, in the Preprocessing Phase, each of the m acf's are first processed to extract their moments (except the zeroth moment), and then the resultant m moment signals are combined with the n int signals to form a weighted direct sum. The latter composite scattered light signal is then transferred to the Analysis Phase.

The size distribution Model Phase is substantially the same as that described in connection with Fig. 5, except that the model is SUBSTITUTE

SHEET

angle. For example, in one form, the filter 39a SUBSTITUT S4EET m_ me gmg is_ WO 89/00286 PCT/US88/0222 -44constructed for use in the Analysis Phase so that the inversion of the equation =J-l(ml, ,m m i(em+1), may be performed, when mi represents the moments (except the zeroth moment) for the ith acf.

Alternatively, the forms for constructing the model shown in and described in conjunction with Figs. 6 and 7 may be used.

In Fig. 11, a model size distribution is generated as a histogram havinq all non-negative components and a kernel J is formed for the transform. To find the size distribution v(r) using thi's kernel, a programmed digital computer using Non-Negative Least Squares (NNLS) techniques performs a "best fit." More particularly, to extract the inverse moments from the acf's, the homodyne scattered light acf, must first be processed to yield the heterodyne acf, g(t): baseline))/g(0) where the baseline can be measured conventionally (see, e.g. B. Chu, "Laser Light Scattering", Academic Press, 1976). Then, the natural log of g(t) is curve fitted to a power series in the delay time, t: ln(g(t))=K 0 /0!+Klt/l!+K 2 t 2 The (Ki, are related to the inverse moments cf the particle size distribution, {ki, by the equation k-i=CiKi where C=31rt/(kBTq 2 r is the viscosity of the fluid suspending the particles, kB is Boltzmann's constant and T is the absolute temperature. q is the SUBSTITUTE SHEET from each molecule in the scattering volume of sample WO 89/00286 PCT/US88/02222 magnitude of the scattering vector, defined by q=4r n sin(0/2)/. n is the refractive index of the fluid suspending the particles, 0 is the scattering angle and X is the wavelength of the incident light. The inverse moments are the average values of the inverse central diameters of the particle size distribution, e.g. the 1st inverse moment is the average value of the inverse diameter, the 2nd inverse moment is <<l/d 2 The ordinary moments are <<d 2 The process by which this inverse moment extraction is accomplished is described in detail in D.E. Koppel, J. of Che. Phys., vol. 57, p. 4814, 1972.

If a sufficient number of inverse moments are taken, the particle size distribution is completely specified. In the preferred embodiment, the 1st and 2nd order inverse moments are sufficient, taken together with the 1st order moment defined by:

A

kl where 0 A equals the number of time points in the acf.

The zeroth order moment, which contains the normalization factor, is omitted. Positive subscripts on k are indicative of moments and negative subscripts are indicative of inverse moments.

The particle size distribution is related to the moments extracted from the acfs. As a result, the matrix equation v=aq, relates the product of the kernel matrix, X, by the size distribution vector, v, to the measurement vector, g, where the components of the vector, v, are numbers These components of v represent the relative amount of particles of size indexed by i, that is, the desired size distribution. The components of the measurement SUBSTITUTE

SHEET

b In the latter configuration, the power spectrum, SUBSTITUTE SHEET WO 89/00286 PCT/US88/02222 -46vector, g, are the moments and inverse moments extracted from the acf's and the CLS data. The components of the matrix, K may ib defined in the following manner. First, the matrix may be expressed as equation as a group of separate equations:

B

X vi*ci(9)*di=kj(e) i=1 where B equals the number of bins in the size histogram, N equals the number of inverse moments, M equals the number of ordinary moments, ci(0) is the intensity conversion factor for the ith size bin in the distribution and for the scattering angle, 8, and di is the particle diameter for particles in the ith size bin in the distribution. The moments kj are labelled with the scattering angle, 0, to indicate that the different moments are extracted for each DLS scattering angle used in the analysis. The equation above defines the DLS-related components of the kernel matrix, t, giving: Kji()=ci(9)*dJ.

The matrix K is augmented, by direct sum, with the CLS portion of the kernel, in the same manner as set forth in U.S. Patent Application Serial No. 81' 7 ,48.

The matrix equation, KyvS, with components defined as above is then solved with the non-negative least squares (NNLS) curve fitt:ing algorithm, as described in conjunction with Fig. In the form of Fig. 12, the m acf's are unity nomalized normalized so that g(0)=1 for SUBSTITUTE

SHEET

I -i ~i i including a data collection phase, denoted A, and a SUBSTITUTE

SHEET

WO 89/GP286 PCT/US88/02222 -47each acf) aid then a "modified" kernel is used in the form described by the equations:

B

I vici i=O e-r(k) =g(tj)

B

vici(k) i=1 where: and number of time points in each acf (A) where k indexes the scattering angle, i is the size S(histogram bin number), j is time (for the acf's), x is the weight distribution and C(k) is the conversion of weight to intensity at the kth angle. Ci( k corresponds to Ci(e), as used above, where e is the kth angle). Finally, the curve fit is determined by 2 minimizing the (Vi} in the expression 2O m A k1 I k=l j=1l B (k) vici k [er -g(tj)] 2 i=l This process allows the information in the acf to be directly utilized, rather than first extracting the moments, as in the process of Fig. 11. In addition, all information in the acf's is utilized rather than just that in the nth(n0O) order moments. The Model Phase and Analysis Phase are substantially the same as those in Fig. 11.

SUBSTITUTE SHEET m, ,l Iiv a J.

polarizations horizontal/vertical, SUBSTITUTE

SHEET

WO 89/00286 PCT/US88/02222 -48- In the form of Fig. 13, generalized curve fitting techniques are applied to the acf's which have not been normalized at all. Generally, B k) (k) A(ek) I viti e (tj) i=1 where: and number of time points in each acf (A) where A(ek) is the amplitude of the non normalized kth acf. V is obtained by curve fitting the equations m A B r(k) 2 A(Ok) I vic er i tj-g (tj)]=0 vi k=l j=l i=l where Im A B k) 2 a8 I vic e tj-g k=1 j=1 i=1 where where v i is constrained to be non-negative for all of the histogram bins and A(9k) is constrained to be non-negative for all of the DLS measurement angles.

The invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the

Q\M

,OosrUT^ W -L Q A.VLV111=U YuL une SUBSTITUTE

SHEET

WO 89/00286 PCT/US88/G2222 -49appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

SUBSTITUTE

SHEET

Claims (37)

1. A system for measuring the size distribution v(r) of particles dispersed in a fluid sample, where r is representative of particle size, comprising: A. illumination means for illuminating said sample with a light beam directed along an input axis, B. detector means for detecting the intensity of light scattered from said light beam by said sample for m measurements made under an associated one of a set of m measurement conditions, and for n measurements made under an associated one of a set of n measurement conditions, where m is an integer equal to or greater than one, and n is an integer eqt,;. to or greater than zero, and the sum of m and n is equal to or greater than two, and at least two of said m+n measurement conditions are different, said detector means further including: i. means for generating m intensity signals, each of said intensity signals being representative of the detected intensity of said scattered light as a function of time under a corresponding one of the measurement conditions of said set of m measurement conditions, SUBSTITUTE SHEET has no material affect on the sizing results obtained. Under the control of the condense.fm SUBSTITUTE SHEET WO 89/00286 PCT/US88/02222 -51- ii. means for generating n average signals, each of said n average signals being representative of the average intensity of said scattered light detected under a corresponding one of the measurement conditions of said set of n measurement conditions, C. autocorrelation means for generating m correlation signals, each of said correlation signals being representative of the autocorrelation function of a corresponding one of said intensity signals and being equal to an associated transformation of said distribution v(r), D. size processing means responsive to said correlation signals and said average signals, including: i. preprocessor means responsive to said m correlation signals for generating a composite correlation signal, ii. means for generating a composite scattered light signal from said composite correlation signal and said n average signals -t iii. means for determining a composite transformation operator J" 1 related to said associated transformations and said n average signals, SUBSTITUTE SHEET -1 Diagnostics) dispersed in water. The measurements SUBSTITUTE SHEET T WO 89/00236 PCT/US88/02222 -52- iv. means for transforming said composite scattered light signal in accordance with said determined composite transformation operator J- 1 to generate a size distribution signal representative of said distribution and E. means for controlling said m+n measurement conditions whereby said intensity signals are mutually independent and said average signals ai- mutually independent.

2. A system according to. clairn 1 wherein said preprocessor means includes i. means responsive to said m correlation signals for generating m morent signals, said m moment signals being representative of at least one of the moments of each of said In correlation signals other than the zeroth moments of said m correlaticon signals, ii. means for generating a weighted direct sum signal, said weighted sum signal being representative of the direct sum of said m moment signals and corresponding to said composite correlation signal.

3. A system according to claim 2 whorein said means for generating said composite scatter 7-d light signal includes means for generating the direct sum of said composite correlation signal and said n average SUBSTITUTE: SHEEI) accordance with this non-normalization form of the invention, as shown in Figs. 1 and 2, the data SUBSTITUTE SHEET 'WO 89/00286 PCT/US88/02222 -53- signals, said direct sum rresponding to said composite scattered light signal.

4. A system according to claim 1 wherein said preprocessor means includes: i. means responsive to said m correlation signals for generating m unity normalized correlation signals, each of said unity normalized correlation signals being representative of the unity normalized form of the corresponding one of said m correlation signals, ii. means for generating a direct sum signal, said sum signal being representative of the direct sum of said unity normalized correlation signals, and corresponding to said composite correlation signal.

5. A system according to claim 4 wherein said means for generating said composite scattered light signal includes means for generating the direct sum af said composite correlation signal and said n average signals, said direct sum corresponding to said composite scattered light signal.

6. A system according to cltim wherein said preprocessor includes: i. means for generating a direct sum signal, said sum signal being representative of the direct sum of said m correlation signals and corresponding to saiV csoposite correlation signal. SUBSTITUTEL SHLT, conditions which establish independent intensity SUBSTITUTE SHEET WO 89/00286 PCT/US88/02222 -54-

7. A system according to claim 6 wherein said means for generating said composite scattered light signal includes means for generating the direct sum of said composite correlation signal and said n average signals, said direct sum corresponding to said composite scattered light signal.

8. A system according to claim i wherein said illumination means includes means for directly illuminating said m points with portion of said light beam.

9. A system according to claim 1 wherein said illumination means includes means for preventing any of said beam from directly illuminating any of said m points. A system according to claim 1 wherein said autocorrelation means includes an autocorrelator means for generating said m correlation signals as time domain autocorrelation signals gi(t) where t is time and m, each of said m autocorrelation signals corresponding to the autocorrelation of a corresponding one of said m intensity signals.

11. A system according to claim 1 wherein said autocorrelation means includes a power spectrum means for generating said m correlation signal as frequency domain power spectrum signals Gi(f) where f is frequency and m, each of said m power spectrum signals corresponding to the power spectrum of a corresponding one of said m intensity signals. SUBSTITUTE SHEET .;i WO 89/00286 PCT/US88/02222

12. A system according to claim 1 wherein said associated transformations are linear transformations and wherein said composite transformation operator J-1 is the generalized inverse of the operator corresponding to the direct sum of the operators for said associated transformations.

13. A system according to claim 12 wherein saId inverse transformation operator J- 1 corresponds to the inverse of the matrix corresponding to the direct sum of the matrices of said associated transformations.

14. A systsm according to claim 12 wherein said inverse transformation operator J- 1 corresponds to C a c41]- 1 t where 3 is the matrix corresponding the direct sum of the matrices corresponding to said associated transformations, at is the transpose of the matrix g, H is a conditioning matrix, and alpha is a smoothing parameter. A system according to claim 12 wherein said inverse transformation operator J-1 corresponds to [cta ac]- 1 at where i is the matrix corresponding the direct sum of the matrices corresponding to said associated transformations, 'it is the transpose of the matrix H is a conditioning matrix, and alpha is a smoothing parameter and where all components of the vector representative of said distribution function v(r) are constrained to be greater than or equal to zero. SUBSTITUTE SHEET L~urLUI11lc wiTrn rig. n, except mnart rne moue. ib SUBSTITUTE SHEET i -NORM PCT/US 9 3/0222 WO 89/00286 -56-

16. A system according to claim 1 wherein said associated transformations are non-linear and wherein said size distribution is characterized by v(r,E), where Ir is a characterization parameter vector having K components, and wherein said composite transformation operator J-1 is the E solution algorithm foV m q a 1 X' apa i=1 j=1 Jij(v (r,pfl-ig(tj) 2 0, 91 where i is an integer m, j is an integer k is an integer k P2 is the ath component of E and where gi(tj) is the autocorrelation function of the intensity signal for the i th of said angle at the jth time interval and Jij is an operator related transformations.

17. A system according to n 1 1.

18. A system according to n=0.

19. A system according to n 1 1. A system according to n=O.

21. A system according to n 1. to the associated claim 1 wherein m 1 and claim 1 wherein m 2 and claim 2 wherein m L 1 and claim 2 wherein m A 2 and claim 4 wherein m 1 and SUBSTITUTE SHEET SUBSTITUTE SiHEET X WO 89/00286 PCT/US88/02222 U. -57-

22. A n=0.

23. A n 1. system according to claim 4 wherein m 2 and system according to claim 6 wherein m 1 and

24. A system according to claim 6 wherein m 2 2 and n=0. A system according to claim 1 wherein said measurement condition controlling means comprises means for establishing the position of said detector means whereby one or more of said m intensity signals is representative of the intensity of light from said light beam at points angularly displaced from said input axis, and one or more of said n average signals is representative of the average intensity of said scattered light at points angularly displaced from said input axis different from said points for which said intensity signals are generated,

26. A system according to claim 1 wherein said measurement condition controlling means comprises means operative during at least one of said m+n measurements for selectively controlling the polarization of light from said bearr incident on said sample to have a first predetermined polarization, and for selectively controlling the polarization of light scattered from said sample incident on said detector to have a said first polarization. SUBSrITUTE SHEET S C- ur sU W ll indexed by i, that is, the desired size distribution. The components of the measurement SUBSTITUTE SHEET WO 89/00286 PCT/US88/02222 -58-

27. A system according to claim 1 wherein said measurement condition controlling means comprises means operative during at least one of said m+n measurements for selectively controlling the polarization of light from said beam incident on said sample to have a second predetermined polarization, and for selectively controlling the p,'arization of light scattered from said sample incidant sn said detector to have a said second polarization, and wherein said second predetermined polarization is orthogonal to said first predetermined polarization.

28. A system according to claim 26 wherein said first predetermined polarization is circular.

29. A system according to claim 27 wherein said first and second predetermined polarizations are circular. A system according to claim 26 wherein first predetermined polarization is linear.

31. A system according to claim 27 wherein first and second predetermined polarizations are linear.

32. A system according to claim 1 wherein said measurement condition controlling means comprises means operative during at least one of said m+n measurement for selectively controlling the temperature of said sample to m at a first predetermined temperature and operative during at least one other of said m+n measurements for selectively controlling the temperature of said sample to be at a second predetermined temperature, SU'ST TUTE SHEET i- SUBSTITUTE SHEET WO 89/00286 PCT/US88/02222 -59-

33. A system according to claim 25 wherein said detector means includes first sensor means for generating said m intensity signals and a second sensor means for generating said n average signals.

34. A system according to claim 33 wherein said second sensor means includes a first curvilinear array of spaced apart photosensors and a second curvilinear array of spaced apart photosensors, said first array extending along an arc disposed about said sample, wherein the photosensors in said first array are offset from said sample by relatively small angles measured from said input axis, and said second array extending along an arc d:sposed about said sample wherein the photosensors in said second array are offset from said sample by ninety degrees plus relatively small angles measured from said input axis. A system according to claim 34 wherein said first and second arrays lie in orthogonal planes.

36. A system according to claim 34 including means for supporting a container for said sample, said container having opposed parallel planar lateral surfaces, said container having a rectangular cross-section bounded by said lateral surfaces, and wherein said supporting means includ means for ma %ining one pair of lateral su~ieces pe zndicular to said input axis, and the other pair of lateral surfaces parallel to said input axis. SIBSUSITUTZ SHEET SUBSTITUTE SHEET WO 89/00286 PCT/US88/02222

37. A system according to claim 35 including means for supporting a container for said sample, said container having opposed parallel planar lateral surfaces, said container having a rectangular cross-section bounded by said lateral surfaces, and wherein said supporting means includes means for maintaining one pair of lateral surfaces perpendicular to said input axis, and the other pair of lateral surfaces parallel to said input axis.

38. A method for measuring the size distribution v(r) of particles dispersed in a fluid sample, where r is representative of particle size, comprising the steps of: A. illuminating said sample with a light beam directed along an input axis, B. detecting the intensity of light scattered from said light beanm by said sample for m measurements made under an associated one of a set of m measurement conditions, and for n measurements made under an associated one of a set of n measurement conditions, where m is an integer equal to or greater than one, and n is an integer equal to or greater than zero, and the sum of m and n is equal to or greater than two, and at least two of said m+n measurement conditions are different, including the substeps of: i. generating m intensity signals, each of said intensity signals being depresentative of the SUBSITUTE SHEET oy=%a.Lu LUL riu wi;nout aeparting rrom une spirit or essential characteristics thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the WO 89/00286 PCT/US88/02222 -61- detected intensity of said scattered light as a function of time under a corresponding one of the measurement conditions of said set of m measurement conditions, ii. generating n average signals, each of said n average signals being representative of the average intensity of said scattered light detected under a corresponding one of the measurement conditions of said set of n measurement conditions, C. generating m correlation signals, each of said correlation signals being representative of the autocorrelation function of a corresponding one of said intensity signals and being equal to an associate transformation of said distribution v(r), D. in response to said correlation signals and said average signals, including the substeps of: i. preprocessing said m correlation signals for generating a composite correlation signal, ii. generating a composite scattered light signal from said composite correlation signal and said n average signals SUBSTITUTE SHEET SUBSTITUTE SHEET WO 89/00286 PCT/US88/02222 -62- iii. determining a composite transformation operator J- 1 related to said associated transformations and said n average signals, iv. transforming said composite scattered light signal in accordance with said determined composite transformation operator J-1 to generate a size distribution signal representative of said distribution and E. controlling said m+n measurement conditions whereby said intensity signals are mutually independent and said average signals are mutually independent.

39. A method according to claim 38 wherein said preprocessing means includes the substeps of: i. in response to said m correlation signals, generating m moment signals, said m moment signals being representative of at least one of the moments of each of said m correlation signals other than the zeroth moments of said m correlation signals, ii. generating a weighted direct sum signal, said weighted sum signal being representative of the direct sum of said m moment signals and corresponding to said composite correlation signal. SUBSTITUTE SHEET Li, i ij WO 89/00286 PCT/US88/02222 -63- A method according to claim 38 wherein said step of generating said composite scatterei light signal includes the step of generating the direct sum of said composite correlation signal and said n average signals, said direct sum corresponding to said composite scatteoed light signal.

41. A method according to claim 38 wherein said preprocessing steps includes the substeps of: i. in response to said m correlation signals, generating m unity normalized correlation signals, each of said unity normalized correlation signals being representative of the unity normalized form of the corresponding one of said m correlation signals, ii. generating a direct sum signal, said sum signal being representative of the direct sum of said unity normalized correlation signals, and corresponding to said composite correlation signal.

42. A method according to claim 41 wherein said step of generating said composite scattered light signal includes the step of generating the direct sum of said composih correlation signal and said n average signals, said direct sum corresponding to said composite scattered light signal. SUiSiTUr IIIl O68L9stC I a8 IIII1111 4 in I} XM11} -4 dO N: -VI-4O id F Z WO 89/00286 PCT/US88/02222 -64-

43. A method according to claim 38 wherein said preprocessing step includes the step of: i. generating a direct sum signal, said sum signal being representative of the direct sum of said m correlation signals arid corresponding to said composite correlation signal.

44. A method according to claim 43 wherein said step of generating said composite scattered light signal includes the step of generating the direct sum of said composite correlation signal and said n average signals, said direct sum corresponding to said composite scattered light signal. SUBSTITUTE SHEET