Effects of Sludge Particle Size and Density on Hanford Waste Processing

Transcription

1 Effects of Sludge Particle Size and Density on Hanford Waste Processing A.P. Poloski, B.E. Wells, L.A. Mahoney, R.C. Daniel, J.M. Tingey, S.K. Cooley Pacific Northwest National Laboratory P.O. Box 999, Richland, WA U.S.A. ABSTRACT The U.S. Department of Energy Office of River Protection s Waste Treatment and Immobilization Plant (WTP) will process and treat radioactive waste that is stored in tanks at the Hanford Site in southeastern Washington State. Piping and pumps have been selected to transport the high-level waste (HLW) slurries in the WTP. Pipeline critical-velocity calculations for these systems require the input of a bounding particle size and density. Various approaches based on statistical analyses have been used in the past to provide an estimate of this bounding size and density. In this paper, representative particle size and density distributions (PSDDs) of Hanford waste insoluble solids have been developed based on the new approach that relates measured particlesize distributions (PSDs) to solid-phase compounds. This work was achieved through extensive review of available Hanford waste PSDs and solid-phase compound data. Composite PSDs representing the waste in up to 19 Hanford waste tanks were developed, and the insoluble solidphase compounds for the 177 Hanford waste tanks, their relative fractions, crystal densities, and particle size and shape were developed. With such a large combination of particle sizes and particle densities, a Monte Carlo simulation approach was used to model the PSDDs. Further detail was added by including an agglomeration of these compounds where the agglomerate density was modeled with a fractal dimension relation. The Monte Carlo simulations were constrained to hold the following relationships: 1) the composite PSDs are reproduced, 2) the solid-phase compound mass fractions are reproduced, 3) the expected in situ bulk-solids density is qualitatively reproduced, and 4) a representative fraction of the sludge volume comprising agglomerates is qualitatively reproduced to typical Hanford waste values. Four PSDDs were developed and evaluated. These four PSDD scenarios correspond to permutations where the master PSD was sonicated or not-sonicated before being analyzed and whether agglomerates existed or not in the PSD samples. When critical pipeline velocity calculations are applied to these results, several percent of Hanford tank waste sludge are expected to exceed pipeline velocities of 4 ft/sec. Operation and waste processing at pipeline velocities in the >4 to 6 ft/sec range appear to be compatible with the Hanford sludge in 3-inch pipes. This manuscript has been authored by Battelle Memorial Institute, Pacific Northwest Division, under Contract No. DE-AC05-76RL01830 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes.

2 INTRODUCTION The U.S. Department of Energy (DOE) Office of River Protection s Waste Treatment and Immobilization Plant (WTP) is being designed and built to pretreat and then vitrify a large portion of the wastes in Hanford s 177 underground waste storage tanks. Because of the variability of the waste, WTP process piping must be capable of transporting a wide range of materials, including Newtonian and non-newtonian fluids, slurries with differing solids concentrations, and slurries with mechanical and chemical plugging potential. A minimum pipeline velocity is required to avoid mechanical plugging for slurries that exhibit settling behavior. This minimum or critical velocity is a function of the solid particle size and morphology, particle density, solids concentration, the carrier fluid, and the pipeline configuration and is defined as the minimum velocity demarcating flows in which the solids form a bed at the bottom of the pipe from fully suspended flows [1]. The analytical prediction method for the WTP critical velocities is provided in BNI [2]. The basis for these predictions was provided by the Oroskar and Turian [1] equation for the critical velocity, U c, and may be determined by ( ) ( ) D ρ ( ) LD gd S U = c gd S C v 1 C v (Eq.1) d μ L where g = gravitational acceleration (m/s 2 ) d = particle diameter (m) S = ratio of the solid density to liquid density, ρ S /ρ L C v = volume fraction of undissolved solid in the flow D = pipe diameter (m) ρ L = liquid density (kg/m 3 ) μ L = liquid dynamic viscosity (Pa s). To calculate the velocity at which a sliding bed will transition to a stationary bed, BNI [2] provides the Thomas equation [3,4] as U ( S 1) gμ L DρL T = 9 (Eq. 2) ρl μ L The undissolved solids and therefore their sizes affect the critical velocity calculation. In the absence of specific process stream data, and as the initial condition of the waste received at the WTP, the physical properties of the as-received waste are considered. Waste will be fed to the WTP by the Hanford waste feed delivery (WFD) system. High-level waste (HLW) feed includes insoluble solids consisting primarily of oxides and hydroxides of metals used to fabricate and reprocess nuclear fuels. These solid particles range in size and density from small, dense primary particles to large, low-density, diffuse flocs or soft

3 agglomerates and large, relatively dense, cemented aggregates and stable agglomerates (collectively termed hard agglomerates). Fig. 1 depicts these different particles. Fig. 1. Solid particles and soft and hard agglomerates (from [5]) The individual particle size in a waste stream may vary by five orders of magnitude and is affected by the constituents present. As stated in Jewett et al. [6], the smallest particles are many oxides and hydroxides, including ZrO 2 and FeOOH, whose diameters are in the 3 to 6 nm range. Other particles such as boehmite (AlOOH) and apatite are in the 0.1 to 1 μm size range. These submicron primary particles found in many HLW tanks form agglomerates typically 1 to 10 μm in size, but can reach 100 μm or more [7,8]. Some of the largest primary particles are gibbsite (Al(OH) 3 ) and uranium phosphate, which can exceed 20 μm in size. Thus, the individual primary particles vary in density as well as size, and it follows that the agglomerates formed therefore also vary in density. Many studies have been conducted to determine the particle size and constituents of the undissolved solids in the Hanford waste. However, the available information relating the particle size and the constituents, i.e., the density, is limited. This paper describes a new approach, described by Wells and co-workers [9], for relating measured particle-size distributions (PSDs) to solid-phase compounds and generating expected particle size and density distributions (PSDDs). COMPOSITE PSD RESULTS As part of the effort to make sure that Hanford tank waste feeds properly through the pipelines from the waste tanks to the WTP, it is necessary to understand the composite PSDs for the various waste types and states of agitation. Fig. 2 shows the probability and cumulative distributions for 1) minimal disturbance (i.e. unsonicated samples) with corresponding sonicated data and 2) sonicated data with corresponding minimal disturbance data. Because both composite PSDs derive from the same tank basis, it is appropriate to directly compare these two. The

4 minimal disturbance data shows a broad and even distribution of particle diameters(considering the roughness of the particle diameter source data) that ranges from around 0.5 to approximately 300 μm. The first 90% of the distribution encompasses diameters from 0.5 to 20 μm with the remainder forming a long tail from 20 to 300 μm. Fig. 2. Probability (a) and cumulative (b) composite volume based PSDs for all minimal disturbance data and all sonicated data. These distributions employ different tank bases. As expected, sonication shifts the PSD to lower particle diameters. The shift appears to be uniform across the entire range of sizes observed, although the PSD tail between 100 and 300 μm appears to be appreciably reduced. Quantiles of the PSDs provided in the figures are provided in Table I. After sonication, 90% of the material falls below 20 μm. The shift to lower particle sizes is not dramatic less than an order of magnitude. The small difference between the sonicated and minimal disturbance might seem at first to indicate that much of the material is either primary particles or hard agglomerates; however, it could also suggest that, were any large flocs present in the waste samples before analysis, the shear required to suspend the particles and obtain representative measurements was also sufficient to break those flocs. Table I. Cumulative Volume Based PSD Data from Fig. 2 IDENTITY OF SOLID PHASES IN SLUDGE The solid phases in the sludge portion of Hanford tank waste must be identified to better understand how to plan for an uninterrupted flow of the wastes. A hybrid approach was taken to identify and quantify the compounds present in the solid phase of Hanford tank sludges. The

5 solids predicted by the ESP 1 chemical thermodynamic model were taken as a first approximation. These predictions were then reviewed and revised by a panel of experts who compared the compounds to observations made on sludge solids, typically by microscopic analysis techniques. Compounds not predicted in ESP but observed by microscopic analysis were added according to the stoichiometric ratio of the analyte of interest. All aluminum hydroxide predicted by ESP in tanks containing Redox boiling waste was assigned as boehmite. Aluminum hydroxide/oxide predicted in other tanks is assigned as gibbsite. The end result is shown in Table II. Table II. Estimated Composition of Solid Phase in Hanford Sludges MODELING OF SOLID PHASE PARTICLE SIZE AND DENSITY DISTRIBUTION As described by Wells and co-workers [9], the sonicated PSD is assumed to represent the PSD for the individual primary particles and the hard agglomerates, while the minimal disturbance PSD is assumed to represent the PSD for the individual primary particles and both the soft and hard agglomerates. In lieu of data sets representing more complete sampling of Hanford waste, these PSDs are used as representations of the combined insoluble solid-phase inventory. The 1 ESP was supplied and developed by OLI Systems, Inc., Morris Plains, New Jersey.

6 solid-phase compounds for the insoluble Hanford waste, their relative volume fractions, crystal density, maximum primary particle size, and maximum agglomeration size have also been identified as described by Wells and co-workers [9]. To define the insoluble solid PSDD, the solid-phase compound data are modeled into the sonicated and minimal disturbance PSDs as will be described below. Quantifying the fractal dimension relating the agglomerate size to its density for the Hanford waste is subject to uncertainty as described by Wells and co-workers [9]. Thus, for each PSD modeled, a bounding case with the agglomerates set to the crystal density is considered. Additionally, cases in which the fractal dimension is adjusted such that the resulting volumeweighted average of the bulk solids (primary particles and agglomerates) approximates that estimated for in situ conditions in individual waste tanks (see Jewett et al. [6] for example) are evaluated. Four PSDD modeling approaches were considered: 1. Upper Bound Sonicated PSD Case Primary particles and hard agglomerates are assigned crystal density. 2. Lower Bound Sonicated PSD Case Primary particles are assigned crystal density. Density of hard agglomerates is assigned via fractal relation. 3. Upper Bound Minimal Disturbance PSD Case Primary particles, soft and hard agglomerates are assigned crystal density. 4. Lower Bound Minimal Disturbance PSD Case Primary particles are assigned crystal density. Density of soft and hard agglomerates is assigned via fractal relation. Those cases denoted as upper bounding are so-called given that all of the particulate is assigned a crystal density; higher density for a given particle is not achievable within a given solid-phase compound. The lower bound cases are denoted as such strictly in relation to the upper-bound cases. These cases are expected to be more representative of actual Hanford waste conditions given the observed pervasiveness of agglomerate particulates and their expected fractal dimensions [9]. The composite PSD is deconvoluted into PSDs for each species by solving a least-squares optimization problem. For each realization (i.e., set of parametric variations performed), the optimization problem is formulated by finding a set of weighting factors that minimizes the error between the target composition of species and the calculated composition of species. W j ( ε ) min (Eq. 3) where ε is an objective function defined by a least-squares error, and W j is a weighting factor for chemical species j. In the absence of particle-size cutoffs, the weighting factors would be equivalent to the volume fraction of each species. The objective function is defined as the square of the relative error between the mass composition of species supplied by the ESP modeling and the mass composition of species calculated in this deconvolution exercise.

7 2 M, j esp M j ε = (Eq. 4) j M j, esp where M j is the calculated mass fraction of chemical species j, and M j,esp is the mass fraction of chemical species j determined by ESP. Note that ESP results were used as the first approximation, but the final compounds and mass fraction thereof were determined as described in Section 3.2 of the Wells and co-workers report [9]. The mass faction of a particular species is determined in a way like that described by Wells and co-workers [9]. The PSD is described as a series of bins of various sizes. For each bin/species combination, the volume fraction of that bin that falls below the maximum observed diameter for the given species is defined. The volume fraction of a species present in a particular bin is proportional to the number of species present in that size range. The volume fraction is then converted to a mass by multiplying this value by the density of the species in that bin. The density can be defined as either crystalline or fractal. Crystals can exist from the minimal bin size to a defined input value based on the largest size observed by scanning electron microscopy (SEM) images. Flocs or agglomerates can exist from a minimal value that represents the agglomerate building block particle size to the upper bin limit. A fractal dimension is also used as an input value to scale the density from the crystalline density at the building block size to smaller values according to the fractal relation shown below. ρ ρ ij, floc j, crys where ρ ρ L L Di = d j F j, D 3 (Eq. 5) ρ ij,floc = the calculated density of the flocs for fraction of chemical species j for bin i ρ j,crys = the crystalline density of chemical species j ρ L = the liquid phase density D i, = the particle size in bin i d j, = the size of the floc building block particles for chemical species j F j,d = the fractal dimension for chemical species j. In this model, the threshold between crystals and flocs for a particular species, d j, was the primary building-block size as determined by analysis and the expert panel recommendations as provided by Wells and co-workers [9]. In Cases 1 and 3, this value is set to the maximum size of the distribution, and flocs or agglomerates are not considered. Case 1 is fit to the sonicated PSD, while Case 3 is fit to the minimal disturbance PSD. In Cases 2 and 4, the threshold crystal/agglomerate building block value, d j, was varied in multiple realizations as part of a Monte Carlo simulation. Again, Case 2 is fit to the sonicated PSD while Case 4 is fit to the minimal disturbance PSD. The total mass of each species can then be calculated through the following equation:

8 m N = bins j i= 1 where W Nspecies k = 1 j W f k ij f ik F j, D 3 ( ) ( ) Di V + i 1 f ij, floc f ij, floc ρ j, crys (Eq. 6) d j m j = the non-normalized mass of species j over the PSD histogram W j = a weighting variable for chemical species j that is adjusted to close the mass balance on the deconvolution problem N bins = the number of bins in the PSD histogram N species = the number of species considered in the PSD histogram f ij,floc = the fraction of flocs in bin i for chemical species j set to 0 or 1 based on the primary particle size d j, f ij = the volume fraction of material in a given bin i falling below the maximum observed diameter for a chemical species j V i = the volume fraction of the overall PSD for bin i. These values can be converted to a mass fraction through the following normalization. m j M j = N species m k= 1 k where M j is the normalized mass of species j over the PSD histogram. (Eq. 7) This optimization problem was implemented in an Excel spreadsheet. For a given realization, the weighting vector W j was solved to close the mass balance on the model problem. The result is a PSD where each size bin is deconvoluted by chemical species and agglomerates. At this point, densities are known for each chemical species in each particle-size bin. These values are then used as inputs to critical velocity equation calculations. Critical velocities are determined for each chemical species in each particle-size bin using the Oroskar and Turian equation and Thomas equations provided in WTP-GPG-M-0058, Rev. 0 [2]. In these calculations, the following assumed parameter values form the basis for the calculations: 2 cp carrier fluid viscosity 1.2 g/ml carrier fluid density 3 inch ID pipe This results in critical velocity values being assigned to the chemical species present in each bin in the PSD. Two results are then obtained for the following cases: Oroskar and Turian equation over entire particle-size range Thomas equation for fine particles, Oroskar and Turian [1] equation for coarse particles, interpolation of Thomas and Oroskar and Turian equations for transition region.

9 The viscous sublayer thickness was calculated at approximately 130 μm for the flow of water at 4 ft/sec in a 3-inch ID pipe. Thomas [3] states that his equation is valid for particle sizes smaller than 0.3 times the viscous sublayer thickness while conventional critical-velocity relations such as the Oroskar and Turian equation can be used for particle sizes above the boundary layer thickness. Therefore, the Thomas equation is used for particles smaller than 40 μm, and the Oroskar and Turian equation [1] is used for particles above 130 μm. A log-normal interpolation between the Thomas [3] and Oroskar and Turian [1] equations is used to determine the critical velocity in the transition region between 40 and 130 μm. Note that these ranges are only valid when water is the carrier fluid (e.g. after a process line flush). The viscous sublayer size is expected to become larger with increasing viscosity. Consequently, application of this sublayer thickness assumption maximizes the impact of the Oroskar and Turian equation [1] relative to the Thomas equation [3]. Given a threshold design velocity value, the fraction of the PSD having critical velocities above the threshold design value can be determined by a simple summation. This methodology is illustrated through an example problem which corresponds to Case 1 of this report. The input data are provided by Wells and co-workers [9]. Since a Monte Carlo approach was employed, many possible solutions to the equations presented above are calculated. Data in Fig. 3,4,5,6 and Table III represent a single instance or possibility of these possible solutions and are used exclusively as example data to illustrate the methodology used in the Monte Carlo approach. Representative results from the entire Monte Carlo simulation for each modeling case are presented after this example problem discussion. In this example, all bins are assigned crystal density values. This corresponds to the Case 1 scenario described by Wells and co-workers [9]. Given the PSD and solid composition data, the optimization problem shown by Eq. 1 to Eq. 5 was solved to deconvolute the PSD while maintaining the overall chemical composition. The resulting chemical species map is shown in Fig. 3. The volume fraction assigned to each species is proportional to the overall PSD over the particle-size range. While this proportionality may not match the reality of the sludge in the Hanford tank farm, not enough information exists to form the basis for a more realistic model. In this example, all of the species are assigned a crystalline density. The resulting density map is shown in Fig. 4. In cases where flocs or agglomerates are present (i.e. cases 2 and 4), a fractal relation is used to determine the floc density for various bin sizes.

10 Volume Fraction Particle Size (μm) Fig. 3. Example problem species map; chemical species shown on the colorbar axis Volume Fraction Particle Size (μm) Fig. 4. Example problem density mapping; density (g/ml) shown on the colorbar axis

11 At this point, particle sizes and densities are defined over the entire PSD. The Oroskar and Turian equation is used to calculate the critical velocities over all of the size bins and densities in the PSD. The resulting Oroskar and Turian critical velocity map is shown in Fig. 5. In a similar fashion, the combination of the Thomas and Oroskar and Turian equations are used to create the critical velocity map shown in Fig Volume Fraction Particle Size (μm) Fig. 5. Example problem Oroskar and Turian equation critical velocity mapping; critical velocity (ft/sec) shown on the colorbar axis Thomas Region Transition Region OT Region Volume Fraction P article S ize (μ m) Fig. 6. Example problem Thomas and Oroskar and Turian equation critical velocity mapping; critical velocity (ft/sec) shown on the colorbar axis 0

12 If a threshold design criteria is applied to the critical-velocity mappings shown in Fig. 4 and Fig. 5, the percent of the PSD above that threshold can be calculated. In the example problem, four design velocities are applied to Fig. 4 and Fig. 5. The bins that are above the specified threshold velocity are then summed up in each case. The results are summarized in Table III. Table III. Example Problem Results that Indicate Fraction of PSD with Critical Velocities Above Specified Threshold Design Values Design Velocity (ft/sec) Critical-Velocity Models Percent of distribution above Design Velocity using Oroskar and Turian equation (Fig. 4) 6.6% 2.1% 0.6% 0.1% Percent of distribution above Design Velocity using Thomas + Oroskar and Turian equations (Fig. 5) 1.6% 0.2% 0.1% 0.0% Critical-Velocity Criterion As described by Wells and co-workers [9], critical velocities are determined for each compound in each particle-size bin using the Oroskar and Turian [1] equation and the Thomas equation as provided in BNI [2]. The mass fraction of the solids in the flow is set to 0.154, and the liquid density and viscosity are assigned to be 1.2 g/ml and 2 mpa s [9], respectively. The pipe diameter is set to 3 inches. As noted in Wells et al. (2007), this approach does not reflect the actual concentrations of the constituents, and may thus indicate settling of particulate that comprises only a minimal fraction of the waste, and thus, if transported uniquely, would require lower velocities. Results are obtained for the following cases: Oroskar and Turian equation over entire particle-size range Thomas equation for fine particles, Oroskar and Turian equation for coarse particles, interpolation of Thomas and Oroskar and Turian equations for transition region. Critical-velocity criterion results for each of the PSDD approaches are provided in Table IV. Four threshold design velocities are evaluated for the volume percent of solid particulate that requires a critical velocity above these thresholds. No design margin is applied to the threshold design velocities. The current waste-feed delivery system has a minimum flow rate of 90 gpm, which translates to approximately 4 ft/sec in a 3-inch-diameter pipe. The threshold design velocity of 3.07 ft/sec represents the velocity required if the threshold of 4 ft/sec is assigned the typical 30% design margin [2]. Case 1 and 3 provide upper bounds for the data considered with the sonicated and minimal disturbance PSDs, respectively. The maximum volume exceeding the criterion occurs for Case 3 as expected [9]. The volume of particulate exceeding the criteria decreases exponentially with the threshold design velocity (Fig. 7).

13 Table IV. PSDD Critical Velocity Criterion Results

14 Fig. 7. Volume of solid particulate exceeding threshold design velocity criterion Examining the specific solid-phase compounds that exceed the threshold design velocity criterion allows for two considerations. First, there is an illustration of the relative impact of the compounds considered. Second, this information provides insight into the potential impacts of specific waste streams. The Case 1 representative PSDD percentage (relative to initial volume fraction) of each solid-phase compound that exceeds the 3.07 ft/s criterion is presented in Table V.

15 Table V. Percentage of Case 1 Representative PSDD Particulate Exceeding 3.07 ft/s Threshold Design Velocity by Solid-Phase Compound CONCLUSION Many slurry transfer models, including the Oroskar and Turian model, have been developed and applied to pipeline transfer of particulate, which tends to have larger particles than most Hanford waste. Thus, the applicability of the Oroskar and Turian model should be investigated for Hanford waste. Defining an acceptable threshold design velocity for pipeline transfer of Hanford waste will require a combination of an applicable critical-velocity equation and accurate characterization of the Hanford solid particulate. Therefore, we recommend additional experimental testing to confirm the critical-velocity correlations used for the WTP design. The recommended experimental testing should employ pipe runs representative of typical layouts that exist in the WTP, including critical components such as short elbows, miter bends, and vertical risers. Testing should also subject the pipe flow to the same inlet (feed) conditions that will exist during plant operations. Testing should be conducted to evaluate: Whether the design velocities predicted by the correlation for the range of anticipated PSDs are sufficient to maintain fully suspended slurry flow. What the critical velocity is for initiation of solid settling. Whether the specified flush flow rate and volume are sufficient to re-suspend and clear settled solids from the pipe runs.

16 Simulants for this testing initially should be designed based on the current state of knowledge from the PSDDs contained in this report. As described, however, the data contained in this report have limitations in which entire waste types are not characterized for PSDs. Additionally, the SEM images used in the current analysis to define the solid-phase compound sizes do not represent all of the Hanford waste types. Additional characterization would increase the confidence in the PSDD data by considering a wider range of Hanford waste. It is recommended that characterization testing be performed on composite samples representing different major Hanford waste types. Samples should be analyzed to determine the size and density of the solid particles. Seven composite samples prepared by the EFRT M12 task include the following waste types: 1C, 2C, CWP, CWR, R (boiling), TBP, and PFeCN. Additional archived samples, outside the scope of the M12 testing, may be available and consist of the following waste types: R (non-boiling), 224, and CWZr. Analysis of the listed samples would result in a waste-type profile that covers the major volume of Hanford sludge. Deliverables of the sample characterization should include the following: Sonicated and minimal disturbance PSDs Solid-phase compound identification and sample fraction Size range observed for particles of a specific compound Whether the particles appear to be crystalline or agglomerates Shape of the particles from a certain compound. REFERENCES 1. Oroskar, A.R. and R.M. Turian. (1980). The Critical Velocity in Pipeline Flow of Slurries. AIChE J. 26: Bechtel National Inc. (BNI) Minimum Flow Velocity for Slurry Lines WTP- GPG-M-0058 Rev. 0, BNI, Richland, Washington. 3. Thomas, A.D. (1976). The Role of Laminar/Turbulent Transition in Determining the Critical Deposit Velocity and the Operating Pressure Gradient for Long Distance Slurry Pipeline. Proceedings of the Sixth International Conference on the Hydraulic Transport of Solids in Pipes. Hydrotransport 6, Paper A2, pp Thomas, A.D. (1979). Predicting the Deposit Velocity for Horizontal Turbulent Pipe Flow of Slurries. International Journal of Multiphase Fluid Flow 55: Ilievski, D. and E.T. White. (1994.) Agglomeration During Precipitation: Agglomeration Mechanism Identification for Al(OH) 3 Crystals in Stirred Caustic Aluminate Solutions. Chemical Engineering Science 49(19): Jewett, J.R., S.D. Estey, L. Jensen, N.W. Kirch, D.A. Reynolds, and Y. Onishi Values of Particle Size, Particle Density, and Slurry Viscosity to Use in Waste Feed Delivery Transfer System Analysis. RPP Numatec Hanford Corporation, Richland, Washington. 7. Herting, D.L Results of Dilution Studies with Waste from Tank 241-AN-105. HNF- SD-WM-DTR-46 Rev. 0, Numatec Hanford Corporation, Richland, Washington. 8. Bunker, B.C., P.J. Bruinsma, G.L. Gruff, C.R. Hymas, X.S. Li, J.R. Phillips, D.R. Rector, P.A. Smith, L. Song, J.M. Tingey, and Y. Wang Colloidal Studies for Solid/Liquid Separation. TWRSPP , Pacific Northwest Laboratory, Richland, Washington.

17 9. Wells, B.E., M.A. Knight, E.C. Buck, R.C. Daniel, S.K. Cooley, L.A. Mahoney, P.A. Meyer, A.P. Poloski, J.M. Tingey, W.S. Callaway, III, G.A. Cooke, M.E. Johnson, M.G. Thien, D.J. Washenfelder, J.J. Davis, M.N. Hall, G. Smith, S.L. Thomson, and Y. Onishi Estimate of Hanford Waste Insoluble Solid Particle Size and Density Distribution. PNWD-3824, WTP-RPT-153 Rev. 0, Battelle Pacific Northwest Division, Richland, Washington.