Mesofluidic Inline Separation for Produced Water Treatment (CRADA 537) October 2025 Carolyn A Burns Sarah Suffield Chinmayee Subban Michael J Minette Leonard F Pease Prepared for the U.S. Department of Energy under Contract DE-AC05-76RL01830 Choose an item. This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor Battelle Memorial Institute, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof, or Battelle Memorial Institute. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. PACIFIC NORTHWEST NATIONAL LABORATORY operated by BATTELLE for the UNITED STATES DEPARTMENT OF ENERGY under Contract DE-AC05-76RL01830 Printed in the United States of America Available to DOE and DOE contractors from the Office of Scientific and Technical Information, P.O. Box 62, Oak Ridge, TN 37831-0062; ph: (865) 576-8401 fax: (865) 576-5728 email: reports@adonis.osti.gov Available to the public from the National Technical Information Service 5301 Shawnee Rd., Alexandria, VA 22312 ph: (800) 553-NTIS (6847) email: orders@ntis.gov Online ordering: http://www.ntis.gov Choose an item. Mesofluidic Inline Separation for Produced Water Treatment (CRADA 537) October 2025 Carolyn A Burns Sarah Suffield Chinmayee Subban Michael J Minette Leonard F Pease Prepared for the U.S. Department of Energy under Contract DE-AC05-76RL01830 Pacific Northwest National Laboratory Richland, Washington 99354 Cooperative Research and Development Agreement (CRADA) Final Report Report Date: October 2025 In accordance with Requirements set forth in the terms of the CRADA, this document is the CRADA Final Report, including a list of Subject Inventions, to be provided to PNNL Information Release who will forward to the DOE Office of Scientific and Technical Information as part of the commitment to the public to demonstrate results of federally funded research. Parties to the Agreement: PNNL, APEX Petroleum Engineering CRADA number: PNNL 537 CRADA Title: Mesofluidic Inline Separation for Produced Water Treatment Responsible Technical Contact at DOE Lab: Leonard Pease Name and Email Address of POC at Company: Monica Chavez DOE Program Office: Fossil Energy and Carbon Management Joint Work Statement Funding Table showing DOE funding commitment: The Government’s estimated contribution, which is provided through PNNL’s contract with DOE, is $246,960, subject to available funding (see Table 1). Table 1. Joint Work Statement Funding Table (as planned) Year 1 Year 2 Total Funding Type Funds-in In-kind Funds-in In-kind Funds-in In-kind Apex PE, LLC $170,956 $76,202 $246,976 Dept. of Energy/PNNL $171,650 $75,310 $246,960 Totals $342,606 $151,512 $493,936 Executive Summary Mesofluidic inline separation developed by PNNL represents an opportunity to remove a key barrier in the treatment of produced water, namely suspended solids that clog downstream operations. The US alone produces over 800 billion gallons of produced water each year, most of which is reinjected underground (but not into the aquifers) as waste. The impact from treating and reusing even a fraction of this wastewater is immense as aquifers in the Midwest and elsewhere dry. The work described herein is essential to leveraging the opportunity to improve produced water quality (to allow for beneficial use) by combining mesofluidic inline separators with reverse osmosis systems. Providing water for agricultural and industrial uses in the American Southwest and Midwest by allowing the reuse of petroleum produced water may be critical to the long-term economics of the region. This is especially true as drought conditions are rapidly lowering aquifer levels in these regions. The incumbent technology for desalination is reverse osmosis (RO) due to its ability to treat a wide range of feedwaters and technological maturity. However, suspended solids cause RO (and other dissolved solids removal technologies) to become clogged and lose performance. A common strategy is to use prefilters upstream of RO systems, but these membrane-based filters easily clog and require regular maintenance. Unlike membrane filters, mesofluidic inline separators provide removal of suspended solids with much lower pressure drops than conventional filtration systems, permitting substantially higher flowrates. Although the amount of produced water from petroleum operations is vast, only a small fraction of it is reused or turned into potable water, for example, because of the lack of technologies to remove both dissolved and suspended solids at a significant throughput. This project shows that mesofluidic inline separators coupled with a commercial dissolved solids removal technology (RO) are positioned to do exactly that. Mesofluidic inline separators fit within commercial piping and are tunable for particle sizes of interest as described below. Indeed, mesofluidic inline separators have remarkably smaller footprints than competing filtration technologies and are easily transportable from jobsite to jobsite. This technology has no moving parts so that solids removal can be accomplished at much lower operating cost. Summary of Research Results Table 2. Summary of Milestones and Key Results Milestone/Deliverable Description/Key Results Date Complete 1. Complete APEX/PNNL contract Contract completed FY22 Q1 2. Design & build mesofluidic inline Separator physically realized FY23 Q4 separator 3. CFD model implemented Presentation of model results FY24 Q1 4. Test mesofluidic inline separator Separator tested; letter report FY24 Q2 5. Prepare manuscript Manuscript PNNL approved FY24 Q3 6. Prepare manufacturing licensing License opportunity marketed FY24 Q3 Highlights from milestones 2-4 are described in greater detail herein. Design and Build Mesofluidic Inline Separator PNNL explored first-of-a-kind ceramic mesofluidic separators to enhance produced water treatment. This work evaluated whether ceramic 3D printing may be used to create ceramic mesofluidic separators, which should be more mechanically robust than plastic (Pease, et al., 2022) but might be less expensive than metal (Burns, et al., 2021). A stated purpose of this effort for 3D printing was to evaluate the hypothesis that ceramic 3D printing costs less with the same quality. This hypothesis answers the question, “Can CoorsTek, Bosch, etc. print cost effectively?” Over the course of this work, PNNL engaged with three 3D printing companies (including CoorsTek and Bosch). The one that was most successful viz-a-viz our requests/designs (which have tight tolerances) has been Kyocera. Figure 1 shows test prints that vary the pin diameter and the number of supports. These results opened the opportunity to produce separators that separate particles of ~50 microns, a key size for flow streams that feed reverse osmosis systems, for example. Further work focused on the 16 micron separator as that will be able to remove an even larger fraction of particulate from produced water streams. Figure 1. Test prints that vary the pin diameter and the number of supports. Ceramic separators consisted of three types of parts: an entry piece, an exit piece, and one or more connective “pucks”. The current design requires at least 8 pucks. Kyocera showed success printing the entry and exit pieces in ceramic. However, the pucks for the 16-micron separator were challenging. Some appear suitable and others do not. Please note the inconsistent diameter of the pins and some of the pins are broken. Figure 2. Images of the ends of test prints. In net, a ceramic mesofluidic separator consisting of three types of parts: an entry piece, an exit piece, and multiple connective “pucks” was printed as promised. However, the 3D printed puck remains challenged. Figure 3 shows computed tomography of the inside of two of the pucks (labeled C and D) that had the best outside print of the most recent printing batch. The images present views of each puck from two directions (parallel and perpendicular to the pins). These images show that significant fractions of the pins are missing or out of place. If even a few of the pins are out of place the puck is defective and will not perform the desired separation. Taken together these printing errors indicate that ceramic printing is not yet able to produce the devices at the dimensions needed to separate particles at the 16-micron cutoff level. The technology continues to advance. Parts that were challenging to produce yesterday may potentially be produced with ease tomorrow. Figure 3. Images of two rods from two perpendicular perspectives. The design called for pins with a diameter of 300 micron and a gap of 225 microns. These computed tomography (CT) images are FFT transforms in which missing pins show up as dark patches. Figure 4. Images of pins. The bottom CT image is a contrast enhanced version of the top CT image. PNNL performed a head-to-head comparison of the costs of 3D printing from the three materials: ceramic, metal, and plastic. Table 3 shows the cost comparison. It shows that for the smaller sizes (e.g., a 16 micron separator), plastic printing is not really an option both because 300-micron diameter pins cannot yet be well printed and because the plastic is not as durable due to the loss of plasticizers. Although both ceramic and metal should maintain good durability, the cost difference between the two is significant. The ceramic printing costs remain nearly an order of magnitude higher than the costs of metal printing. However, as shown above the device cannot be printed reliably. Therefore, ceramic printing is not cost effective relative to metal. Remaining efforts as part of this effort will use metal or plastic. Table 3. Comparable Costs of Mesofluidic Devices Material Cost Durability Ceramic $18k-$46k Acceptable Metal ~$2.5k Acceptable Plastic Cannot print this small so far Short Note: 16-micron target. Success rates of 50% for ceramic 3D printing assumed. The relative prices are likely to change as each technology matures. To proceed with subsequent milestones, ceramic-resin composite separators were printed in-house. These were successfully printed with a 50-micron particle separation target. Figure 5. Image of the polymer mesofluidic device printed in white resin in 8 sections. Analytical-Computational Fluid Dynamics (CFD) Framework Please see Appendix A for a manuscript describing these accomplishments and key successes. Test of Mesofluidic Inline Separator Coupling with Reverse Osmosis Please see Appendix B for a manuscript describing these accomplishments and key successes. References Leonard F. Pease, Nathan R. Philips, Jason Serkowski, Timothy G. Veldman, Michael J. Minette, Carolyn A. Burns, Industrial Scale Mesofluidic Particle Separation, Chemical Engineering and Processing – Process Intensification 173 (2022) 108795. Carolyn A. Burns, Timothy G. Veldman, Jason Serkowski, Richard C. Daniel, Xiao-Ying Yu, Michael J. Minette, Leonard F. Pease, Mesofluidic Separation versus Dead-end Filtration, Separation and Purification Technology 254 (2021) 117256. Subject Inventions No new subject inventions were created as part of this Technology Commercialization Fund project and corresponding CRADA. Products Two manuscripts describing the modeling and coupled experiments have been prepared as shown in Appendix A and Appendix B. Appendix A: Analytical-Computational Fluid Dynamics (CFD) Framework Proceedings of the ASME 2025 International Mechanical Engineering Congress and Exhibition IMECE2025 November 16-20, 2025, Memphis, TN, USA IMECE2025-158876 MESOFLUIDIC SEPARATOR FLOW FIELDS AND SPLIT FRACTIONS Sarah R. Suffield, Carolyn A. Burns, Judith A. Bamberger, Michael J. Minette, and Leonard F. Pease Pacific Northwest National Laboratory Richland, Washington, USA ABSTRACT Here we evaluate the flow fields in mesofluidic separators NOMENCLATURE to determine the fluid split fraction between the two exits. Ap cross-sectional area of permeate Similar to microfluidic bump arrays, the flow fields around the Ax cross-sectional area of express lane (=sxsh) array of pins drive particles of sufficient size to one side of the a1-a3 coefficients device leaving one exit depleted in particles and the other exit b1-b4 coefficients enriched in particles. However, mesofluidic separators differ D hydraulic diameter of device (based on 2r if from microfluidic bump arrays at least because they can circular or sh if square), accommodate much larger particles and because they can Dp hydraulic diameter of the permeate operate at much higher flowrates. dp pin diameter The higher flowrates affect the flow fields within the device Dx hydraulic diameter of the express lane leading to vorticity and turbulence. In most flow situations, fD Darcy friction factor through the permeate turbulence and vorticities would smooth out particle fp friction factor through the pin array concentration profiles, canceling the influence of the bump fx Darcy friction factor of express lane arrays. However, experiments show that there is still a L length of the express lane adjacent to the pin array. separation under these conditions beyond that from gravity (a NL number of pins (or number of rows) in the direction surprising and unexpected result), and the onset of turbulence of flow within these devices is unclear (i.e., do the pins induce or reduce PD perimeter of device (2r if circular or 4sh if square), turbulence). Pp perimeter of permeate The larger particles require an exit large enough to permit Px perimeter of express lane these larger particles to leave without clogging the device. This Qf volumetric flowrates of the feed pathway provides a lower resistance pathway for these particles. Qp volumetric flowrates of the permeate How much of the flow proceeds through this pathway as opposed Qx volumetric flowrates of the express lane to the higher resistance pathway of the bump array remains r radius of the cross section unclear. A typical rule of thumb, suggesting that the area Ref Reynolds number of feed available to the exiting flow determines the exiting volumetric Rep Reynolds number of permeate flowrate, fails because the resistance to the flow governs; but, Rex Reynolds number of express lane semi-quantitative guidance on how to estimate these exiting S permeate flowrate split fraction flowrates has remained elusive. s segment length corresponding to the edge of the Here we use a combination of analytical and computational express lane models to evaluate the onset of turbulence with mesofluidic Scc center-to-center distance between pins in a row devices, determine the fraction of flow that exits through the sh length of the side of a square channel (sh and sx express lane, and estimate pressure versus flowrate curves. define the cross section of the express lane)sx Analytical expressions for the onset of turbulence and split length of the express lane fractions based on fluid resistance are determined. Further, we Vf average velocity of feed show that vortices form on one edge of the pins before the other Vp superficial velocity of the permeate due to the angle of the pins that breaks flow symmetry. Vp,max maximum fluid velocity through the permeate portion of the device Keywords: slurry, solids removal, deterministic lateral Vx velocity through the express lane displacement (DLD), bump arrays PNNL-SA-212979 1 © 2025 by ASME Greek Letters a pin, the velocity goes to zero due to a no-slip condition with  1 for aligned and 4 for staggered peak flow between the pins. As the flow leaves a row of pins it P pressure drop spreads out only to converge again as the next row of pins blocks  surface roughness a portion of the flow pathway. This sets up a periodic  angle defined from the circular center by the convergence as flow approaches a row of pins and divergence as isosceles triangle with two sides of length r and one flow leaves a row of pins. The onset of turbulence is known to side of length s be delayed (meaning it occurs at higher Reynolds numbers, e.g.,  dynamic viscosity of the slurry or fluid >1800) in convergent flow and to be accelerated (meaning it  the slurry or fluid kinematic viscosity occurs at lower Reynolds numbers, e.g., <1800) in divergent flow. The angle of the pins may also be influential but has not  slurry or fluid density been explored previously. Dincau, et al., have explored flow  angle of the pins with respect to the sidewall in conditions up to, but not including the Reynolds number, where degrees vortex shedding becomes important and found that the particle  constant size cutoff changes with flowrate [4]. Flowrates in the vortex Subscripts shedding regime and higher have not been reported from a f feed theoretical or modeling perspective. p permeate The second difference between industrially relevant x express lane mesofluidic devices and microfluidic bump arrays is that the former are tailored to fit within pipes or conduits in contrast to INTRODUCTION most microfluidic arrays that are designed so that edge effects Mesofluidic separators are like microfluidic bump arrays become unimportant (e.g., by including extra arrays pins with at least two key differences. Microfluidic bump arrays were between the particle flow pathways and the walls that serve no first reported by Huang, et al., [1] as a means of separating separation purpose but substantially increasing the cross section particles by size (see Fig. 1). Inglis, et al., [2] provided a of these devices—additions that would require much larger flows quantitative model to predict the performance of these devices in larger-scale industrial applications). These edges induce fluid specifically under laminar flow conditions (their formulas interactions between the walls of the pins and device walls and become increasingly inaccurate outside of the laminar range, the particles and fluid that transport between them. For example, which affect the structuring of the array of pins that drives larger particles require an exit large enough to permit these larger separation [3, 4]). The essential feature of these devices is an particles to leave without clogging the device. This pathway array of pins, the centers of which form a non-zero angle with provides a lower resistance pathway for these particles. How the sidewall. This angle breaks symmetry such that each pin in much of the flow proceeds through this pathway as opposed to at least a pair of downstream rows provides a distinct stagnation the higher resistance pathway of the bump array remains unclear. streamline. These stagnation streamlines form the boundaries of A typical rule of thumb suggesting that the area available to the flow channels. If the distance between the edge of pin and the exiting flow determines the exiting volumetric flowrate fails nearest stagnation streamline is smaller than the radius of a because the resistance to the flow govern, but semi-quantitative particle, that particle must move away from the first pin into the guidance on how to estimate these exiting flowrates has next adjacent flow channel. remained elusive. Mesofluidic separators challenge this model in two very specific ways. First, mesofluidic separators often operate at significantly higher flowrates than microfluidic systems [5-7]. The higher flowrates affect the flow fields within the device leading to vorticity and turbulence, which typically cancel stagnation streamline arguments. For this reason, authors FIGURE 1. IMAGE OF FLOW PROFILE IN A MESOFLUIDIC historically have limited their analyses of microfluidic bump INLINE SEPARATOR WITH RED INDICATING FASTER arrays to laminar conditions. In most flow situations, turbulence VELOCITIES AND BLUE INDICATING SLOW VELOCITIES. THE and vorticities would smooth out particle concentration profiles, EXPRESS LANE HAS THE HIGHEST FLOWRATES, BECAUSE IT canceling the influence of the bump arrays. Unexpectedly, HAS THE LEAST RESISTANCE EVEN THOUGH IT HAS A SMALL experiments show that there is still a separation under these CROSS-SECTIONAL AREA. NOTE THAT THERE IS AN conditions beyond that from gravity [5], and the onset of INTERNAL SPLIT FRACTION (ONLY WITHIN THE PIN ARRAY) turbulence within these devices is unclear (i.e., do the pins AND AN EXTERNAL SPLIT FRACTION (ACROSS THE WHOLE DEVICE) AS THE FLOW CAN REBALANCE BETWEEN THE END induce or reduce turbulence?). Therefore, a key question for OF THE BUMP ARRAY AND THE EXITS BASED, AT LEAST modeling efforts is to determine the flow regimes within PARTIALLY, ON THE PRESSURE DROPS. mesofluidic separators. This would be a relatively simple effort in the absence of the pins. In that case, the onset of turbulence In the remainder of this paper, we first model the fluid corresponds to a conduit Reynolds number of ~1800. However, mechanics of these mesofluidic separators. We then describe the the presence of pins adds wrinkles to this picture. At the edge of results for pressure drop and split fraction as a function of PNNL-SA-212979 2 © 2025 by ASME 𝜀 𝜀 flowrate using a combination of analytical and computational 1 𝑅𝑒𝑥 √𝑓𝐷 2.51 fluid dynamics. We conclude with insights into the onset of = −2 𝑙𝑜𝑔10 [ 𝐷 (1 − 𝐸𝑥𝑝 [− 𝐷 ]) + ] √𝑓𝐷 3.7 2.83∗26 𝑅𝑒𝑥 √𝑓𝐷 turbulence within these devices and the influence of asymmetry. (6) MODEL AND METHODS where fD is the Darcy friction factor in the permeate,  is the Here we present a model that uses both an analytical surface roughness, and D is the hydraulic diameter with framework with insight from computational fluid dynamics Reynolds number for the express lane (CFD) to understand the pressure drop, the split fractions, and the role of turbulence in these devices. We represent this system 𝑅𝑒𝑥 = 𝜌𝑉𝑥 𝐷𝑥 = 4𝑄𝑥 , (7) as though it has two resistances in parallel. One is the friction 𝜇 𝜈𝑃𝑥 through the pin array and one is the friction associated with the sidewalls. We treat flow through the pin array and the express where this Vx is the superficial velocity through the express lane lane as separate flow channels, recognizing that they are adjacent to the pins, Dx is the hydraulic diameter of the express physically in “fluid communication”. For the express lane, we lane,  is the slurry or fluid density,  is the dynamic viscosity of consider it to be an open channel with a hydraulic diameter the slurry or fluid, and  is the slurry or fluid kinematic viscosity. defined by the edges of the pins and the side walls. Because this The Darcy-Weisbach expression gives the pressure drop, P, definition of the express lane varies somewhat with position (the 𝑓𝑥 𝐿 pins are at a finite angle with respect to the wall), this model Δ𝑃 = 𝜌𝑄𝑥 2 , (8) 2𝐷𝑥 𝐴𝑥 2 permits exploration of the minimum, maximum, and median/average express lane position. where L is the length of the express lane adjacent to the pin array, For the express lane, we use the hydraulic diameter, Dx, and Qx is the volumetric flowrate of the express lane. which is For the pin array, we consider two expressions for the 4𝐴𝑥 pressure drop. The first one derives from expressions for flow 𝐷𝑥 = , (1) around tubes in a bank of tubes, similar to the pin array. Here 𝑃𝑥 the pin array has an angle with respect to the sidewall, which is where Px is the perimeter and Ax is the cross-sectional area. Our typically but not always small. Incropera, et al. [9], give cross sections are typically either rectangular or circular. For expressions for aligned and staggered arrangements of tubes, rectangular cross sections, neither of which are perfect for mesofluidic devices. For small or shallow angles, the aligned arrangement is most relevant, 𝐷𝑥 = 2𝑠ℎ 𝑠𝑥 , (2) though the development below incorporates a staggered 𝑠ℎ +𝑠𝑥 arrangement. In either case, the pressure drop, P, is then where sx is the length of the express lane and sh is the length of 𝜌𝑉𝑝,𝑚𝑎𝑥 2 the side of the rectangle completing the cross-sectional area Δ𝑃 = 𝑁𝐿 𝜒 ( ) 𝑓𝑝 , (9) 2 (=sxsh). For a circular cross section, the express lane has a flat edge and a curved edge. Then where NL is number of pins (or number of rows) in the direction of flow, and  is a constant, and friction factor in the permeate, 𝜃−𝑠𝑖𝑛 𝜃 𝐷𝑥 = 2𝑟 𝜃 , (3) fp, is given in Figs. 7.13-7.14 in Incropera, et al. [9] 𝜃+2𝑠𝑖𝑛( ) 2 𝑆𝑐𝑐 𝑉 aligned constrained by 𝑆𝑐𝑐 −𝑑𝑝 𝑝 𝑉𝑝,𝑚𝑎𝑥 = { 𝑆𝑐𝑐 , (10) 𝑠 𝜃 𝑉𝑝 staggered 2(𝑆𝑐𝑐 −𝑑𝑝 ) = 𝑟 𝑠𝑖𝑛 ( ), (4) 2 2 where r is the radius of the cross section,  is the angle defined where Vp is the superficial velocity of the permeate, Scc is the from the circular center by the isosceles triangle with two sides center-to-center distance between pins in a row, and dp is the pin of length r and one side of length s, and s is the segment length diameter. Then corresponding to the edge of the express lane. For the Reynolds number, Rex, if turbulent, roughness may then be used to 𝑁𝐿 𝜒𝑓𝑝 𝑆𝑐𝑐 2 determine the relationship between pressure drop and flow. The Δ𝑃 = ( ) 𝜌𝑄𝑝 2 , (11) 2𝛽𝐴𝑝 2 𝑆𝑐𝑐 −𝑑𝑝 Darcy friction factor, fx, expressions for the express lane for laminar and turbulent flow (see Afzal 2013 Eq. 41 [8]) are given, where Ap is the cross sectional area of the permeate,  is 1 for respectively, as aligned and 4 for staggered. Because the pin array typically has 64 only a slight angle with respect to the sidewall, the pins are often 𝑓𝑥 = , (5) nearly aligned and  does not differ substantially from ~1. In 𝑅𝑒𝑥 PNNL-SA-212979 3 © 2025 by ASME this formulation, the pressure drop increases quadratically with We may now combine the pressure drop expressions and flow, which was observed in Pease, et al. [5]. conservation of mass so that By conservation of mass, −1 𝑓 𝜒 𝑁𝐿 𝐷𝑥 𝐴𝑥 𝑆𝑐𝑐 𝜌𝑄𝑓 = 𝜌𝑄𝑥 + 𝜌𝑄𝑝 , (12) 𝑆 = [1 + √ 𝑝 ( )] . (14) 𝑓𝑥 𝛽 𝐿 𝐴𝑝 𝑆𝑐𝑐 −𝑑𝑝 where Qf, Qx, and Qp are the volumetric flowrates of the feed, where Ap is the cross-sectional area of the permeate. Essentially, express lane, and permeate, respectively. The inclusion of the flowrate split depends on one dimensionless group that must density serves as a reminder that, although constant density is range from zero to one. This solution is highly dependent on assumed to be approximately commensurate with measurements geometry, though flow remains a contributor via the friction to date, the density could vary among the three streams in ways factors. not captured below. An essential variable missing in this model is the friction factor through the pin array of the permeate, fp. Beale [10] (see his Eq. 11) provides a power law approximation with coefficients a1-a3 𝑎3 1 𝑏3 𝑓𝑝 = 𝑎1 (𝑎2 + ) (𝑏1 𝐸𝑥𝑝(−10−𝑏2 𝑅𝑒𝑝 ) − + 𝑏4 ) 𝑅𝑒𝑝 𝑅𝑒𝑝 (15) with a final factor included here that accounts for the offset in angle with coefficients b1-b4 and Rep is the Reynolds number of the permeate. This approximation does not directly address the influence of surface roughness, because surface roughness was not given explicitly by Beale [10] and Fig. 2a does not suggest a strong influence for surface roughness so long as there is some roughness. Figure 2 shows the friction factor as a function of a Reynolds number from the CFD model compared to data extracted from Figs. 7.13-7.14 in Incropera, et al. [9]. The figure shows some variation with respect to angle that is captured in the equations developed above and a modest contribution from friction. The CFD model results are close to those presented in Incropera, et al., [9]. The coefficients a1 and a3 were fit from Table 1 of Beale [10] and a2 was corrected based on the value of Fig. 7.13 of Incropera, et al., [9] at Rep=1.0.106 to find −4.605 𝑆 𝑎1 = 586.58 ( 𝑐𝑐 ) 𝑑𝑝 2 𝑆 𝑆 𝑎2 = −0.004205 ( 𝑐𝑐 ) + 0.01622 ( 𝑐𝑐 ) − 0.01235 𝑑𝑝 𝑑𝑝 2 𝑆 𝑆 𝑎3 = −0.2667 ( 𝑐𝑐 ) + 0.6613 ( 𝑐𝑐 ) + 0.604. (16) 𝑑𝑝 𝑑𝑝 FIGURE 2. THE FRICTION FACTOR, fp, THROUGH THE The coefficients b1-b4 were determined from the CFD as PERMEATE BASED ON PEAK FLOWRATE BETWEEN PINS VERSUS THE REYNOLDS NUMBER OF THE DEVICE BASED ON 𝑏1 = 0.004345𝜙 2 + 0.01162𝜙 THE HIGHEST FLOWRATE, Vp,max AS A FUNCTION OF THE ANGLE OF THE PINS WITH RESPECT TO THE SIDEWALL 𝑏2 = −0.005368𝜙 + 3.538 GENERATED FROM STAR-CCM+ RESULTS. 𝑏3 = −0.2933𝜙 2 + 3.501𝜙 𝑏4 = −0.005528𝜙 2 − 0.009157𝜙 + 1 (17) The permeate flowrate split fraction, S, may be defined as 𝑄𝑝 𝑄𝑓 −𝑄𝑥 𝑆≡ = . (13) 𝑄𝑓 𝑄𝑓 PNNL-SA-212979 4 © 2025 by ASME where  is the angle of the pins with respect to the sidewall in CFD has been used to develop a quantitative expression to degrees.1 Additional refinement of b2 may increase precision. improve analytical predictions. Specifically, we developed a The Reynolds number of the permeate is given as CFD model of an inline tube bundle, similar to textbook correlations for friction factor as a function of Reynolds number, 𝑅𝑒𝑝 = 𝜌𝑉𝑝 𝐷𝑝 = 4𝑄𝑝 (18) albeit including the angle of asymmetry. The commercial CFD 𝜇 𝜈𝑃𝑝 software STAR-CCM+ was used to develop the inline tube bundle model. The model included a single fluid region that and flowed through a 7x7 array of tubes (Fig. 3). Each tube had a diameter of 0.5 mm, and the orientation of the tube array was 4𝐴𝑝 𝐷𝑝 = , (19) varied from an inline configuration with a 0 degree tilt to a 4 𝑃𝑝 degree tilt orientation (see Fig. 3). We also conducted an X-ray computed tomography (XCT) where Pp is the perimeter and Ap is the cross-sectional area. Our scan to image the inside of the mesofluidic device. XCT image cross sections are typically either rectangular or circular. For scanning was conducted using a Nikon X-Tek/Metris XTH square cross sections, 320/225 kV instrument (Nikon Metrology, Belmont, CA, USA) 2𝑠ℎ (𝑠ℎ −𝑠𝑥 ) at 105 kV and 325 μA, with a 0.5-mm Cu filter reducing beam 𝐷𝑝 = , (20) hardening artifacts. During scans, each sample core was rotated 2𝑠ℎ −𝑠𝑥 continuously with brief stops for projection collection, totaling where Dp is the hydraulic diameter of the permeate, sx is the 3142 projections over a 360˚ rotation, capturing two frames per length of the express lane and sh is the length of the side of the projection with 500 ms exposure time for each frame. This rectangle completing the cross-sectional area (=sxsh). For a yielded a voxel size of 7.4 μm and a spatial resolution of ~15 μm. circular cross section, the express lane has a flat edge and a Images were reconstructed into 3D volumes with CT Pro 3D curved edge. Then for the permeate (version XT 2.2, Nikon Metrology, Inc). 2𝜋−𝜃 RESULTS AND DISCUSSION 𝐷𝑝 = 2𝑟 , (21) 2𝜋−𝜃+2𝑠𝑖𝑛( ) 𝜃 2 In this paper, we describe three key aspects of mesofluidic separator performance. First, we use the model described above constrained by Eq. 4. to predict the pressure drop versus flowrate. Second, we predict The Reynolds numbers are related via the split fraction between the permeate and the express lane. We note that the internal split fraction modeled here may differ 𝜌𝑉𝑓 𝐷 4𝑄𝑓 meaningfully from the split fraction that develops after the flows 𝑅𝑒𝑓 = = , (22) 𝜇 𝜈𝑃𝐷 exit the pin arrays due to system pressure drops (e.g., a lower pressure drop permeate flow may pull some express lane flow 𝑃 𝑅𝑒𝑥 = 𝑅𝑒𝐷 (1 − 𝑆) 𝐷 , (23) within the device into the permeate exit). Third, we evaluate the 𝑃𝑥 velocity profiles through the pins in consideration of the onset of 𝑃 turbulence. 𝑅𝑒𝑝 = 𝑅𝑒𝐷 𝑆 𝐷 , (24) An essential outcome of the model above is the relationship 𝑃𝑝 between the pressure drop through the device and the flowrate. where D is the hydraulic diameter of the device (2r if circular or Essential to that outcome is the development of expressions to sh if square), PD is the perimeter of the device (2r if circular or represent friction within the pin array that account for its tilt with 4sh if square), and Vf is the average velocity of the feed. respect to the sidewalls. Fig. 3 shows the resulting velocity plots A Reynolds-averaged Navier-Stokes (RANS) model with a for 0 and 4 degrees of tilt. Notably, after flow is established, the k–ε turbulence flow model was applied to the fluid region, which velocity at the bottom of a pin (see Fig. 3) is higher than the was assumed to be water. Flow across the tube bundle was velocity around the top of a pin. driven by a velocity inlet upstream of the bundle, and the Figure 4 shows the pressure drop versus the flowrate boundary downstream of the bundle was specified as an outlet. through the mesofluidic device using a fluid interconnect model Along each wall boundary, including the boundary around each that pairs a nearly aligned (small angle) tube model representing pin, orthogonal prismatic cell layers were applied to the mesh to the pin array with an open channel model representing the improve the accuracy of the flow solution near these surfaces. express lane portion of the flow channel. The agreement with experimental data is quite reasonable with best agreement 1 These correlations were generated from the following coefficients fit to CFD Angle a1 a2 a3 data for Scc/dp=1.6. 0 degrees 16.88 0.003576 0.7943 2 degrees 7.663 0.002294 0.6131 4 degrees 4.339 0.001288 0.4884 PNNL-SA-212979 5 © 2025 by ASME between modeling and experiment acknowledging that the pin The internal split fraction to the permeate is important because it diameters of as-built devices may be larger in diameter than as- influences the external split fraction and because the particles designed devices. The red data was taken early in a several week migrate relative to this split fraction. campaign and the blue data was take late in the campaign. Both Our modeling framework explored the split fraction as a data sets evaluate the flow of water through the device, though function of Reynolds number based on feed conditions and as a all the other tests during that time employed concentrated function of position down the separator channel. Figure 5 shows simulants with particles that may have accumulated on the that the split fraction is a function of the feed Reynolds number. surface. Indeed, the pins were discolored by the time of taking As the flowrate increases, more flow goes through the express of the second data set in blue suggesting that the pins were indeed lane than the permeate channel. The split fraction is controlled coated with a thin layer of simulant particles. The two model in part by the ratio of the friction factor in the permeate to the curves differ in that the upper dashed curve corresponds to a pin friction factor in the express lane. When this ratio increases there diameter that is 0.02 mm larger in radius than the lower solid is less resistance to flow through the express lane, so more flow curve, which is commensurate with the 4-30 micron (mode-95 goes through the express lane than the permeate. percentile) particles used in those experiments. (a) (b) FIGURE 3. VELOCITY VECTOR PLOTS FOR (a) 0 DEGREE AND (b) 4 DEGREE ANGLE OF THE PINS WITH RESPECT TO THEIR SIDEWALLS. VELOCITIES ABOVE AND BELOW A POST DIFFER WHEN SYMMETRY IS BROKEN. FIGURE 4. PRESSURE DROP VERSUS FLOWRATE. FIGURE 5. (TOP) FLUID SPLIT FRACTION (BLUE), S, AND FRICTION RATIO (RED), fp/fx, VERSUS FEED REYNOLDS A key metric of performance of mesofluidic separators is the NUMBER, Ref, BASED ON THE CHANNEL OPENING WITH fluid split fraction that goes through bump array or out the DIMENSIONS BASED ON THE DEVICE REPORTED IN BURNS, permeate exit. The remaining flow goes to the express lane. In ET AL. (2021). (BOTTOM) FLUID SPLIT FRACTION TO THE many cases, increasing the split fraction, or the product of the PERMEATE VERSUS EXPRESS LANE WIDTH. IN BOTH split fraction with the entering flowrate, are important to PANELS, THE DISJOINT IN THE CURVES OCCURS WHEN THE downstream applications. In the course of modeling, two split EXPRESS LANE TRANSITIONS BETWEEN LAMINAR AND TURBULENT FLOW (E.G., Rex=1800). fractions become important: the internal split fraction and the external split fraction. The external split fraction is the split Further, the internal split fraction varies with position down fraction as measured on the inlet and outlet. This split fraction the channel. Figure 6 shows that the pins change position, albeit is strongly influenced by inlet and outlet pressures. The internal gradually, down the channel. The express line narrows (flow split fraction is the split fraction as measured within the device. from right) and then widens as a column of pins ends. Figure 5 PNNL-SA-212979 6 © 2025 by ASME shows that the split fraction to the permeate decreases as the express lane widens. Remarkably, the variation in the internal split fraction can be meaningful. FIGURE 7. SCALED VELOCITY VERSUS SCALED POSITION FIGURE 6. CT IMAGE OF THE PINS SHOWING THAT THE FOR DIFFERENT FLOWRATES FIT TO LAMINAR AND TURBULENT MODELS. EXPRESS LANE PINS VARY AS A FUNCTION OF POSITION IN A MESOFLUIDIC DEVICE JOINED BY SCREWS. THE SPLIT FRACTION TO THE PERMEATE WILL INCREASE FROM LEFT TO RIGHT BUT THEN DROP AS THE NEW ROW OF PINS DIVIDES THE EXPRESS LANE. THIS IS THE FIRST IMAGE OF THE AS-BUILT INTERNALS OF A MESOFLUIDIC DEVICE. A key question is how turbulence affects the performance of mesofluidic separators. These separators function based on the interaction of fluid flow with pins. The pins are angled with respect to the sidewall of the channel. On the upstream side of every pin the fluid divides with some of the fluid going to one side of the pin and some of the fluid going to the other side of the pin. The point at which this division occurs is called the stagnation point and the streamline that anchors to this position is called the stagnation streamline. These streamlines form flow channels that govern the separation. Small particles stay within FIGURE 8. EXPONENTS FROM POWERLAW FITS a flow channel, whereas large particles consecutively bump from VERSUS REYNOLDS NUMBER OF FLOW IN THE GAP one flow channel into the next as they pass each row of pins. BETWEEN PINS. This description suffices (in the absence of edge effects) under laminar conditions. Others have shown that, as the Reynolds One means of determining the flow regime is to use number increases, vortices begin to form [4], but what happens computational fluid dynamics (CFD) to determine the velocity at higher Reynolds numbers has not previously been explored, profiles in the average direction of flow along a transverse that though separation at higher Reynolds numbers have been shown is perpendicular to the channel sidewall and then determine if at industrial scale flowrates. they adopt a parabolic profile, suggestive of a laminar flow The role of turbulence boils down to the question of which regime, or a fractional power law (e.g., a 1/7th power), suggestive flow regimes govern within the region defined by the pins in the of turbulent flow. Figure 7 shows these velocity profiles across mesofluidic separator. Determining the flow regime in the a range of flowrates scaled with wall away from the express lane entrance and exit regions is straightforward. However, the flow at 0 and the wall adjacent to the express lane at 1. Likewise, the regime within the pin region has been less obvious. The velocity velocity scales from zero where the pins are to 1 between the between the pins must be elevated because of the decreased pins. The express lane has a higher velocity because it has more cross-sectional area available for flow, suggesting an earlier open space and, therefore, less resistance. The figure also shows transition to turbulence. In contrast, the pins may serve to the laminar model (a parabolic profile) and a turbulent model laminarize the flow, and screens similar in local geometry to (here a 1/7th power law). The inset shows that the CFD evaluated these pin arrays are used to laminarize flow in wind tunnels, for velocities fall between these two model velocity profiles. example. Comparing “ideal” profiles to the data from the CFD run shows that at the lowest flowrate the velocity profile is indeed parabolic but at the higher velocities the velocity profile departs from the parabolic profile. The large peak on the right of the plot PNNL-SA-212979 7 © 2025 by ASME corresponds to the express lane. These profiles are Bump Arrays Separate Particles from Turbulent Flows?. In approximately periodic and reminiscent of those of Giese, et al. Proceedings of the ASME 2021 Fluids Engineering Division [12] for flow through porous media, although the specifics of the Summer Meeting (FEDSM 2021), August 10-12, 2021, geometry vary. Virtual, Online, 3, FEDSM2021-67696, V003T08A024. Although a transition between laminar and turbulent flow New York, New York: ASME. conditions at a distinct Reynolds number may arise under other https://doi.org/10.1115/FEDSM2021-67696. flow conditions, Figure 8 shows a more gradual transition for these flow conditions. We fit the velocity profiles between ~0.2 [4] Dincau, B.M., A. Aghilinejad, T. Hammersley, X. Chen, to ~0.6 with curves to determine the exponent. Each half of the J.-.H Kim. 2018. Deterministic lateral displacement (DLD) velocity profile was fit separately (i.e., the portion of the velocity in the high Reynolds number regime: high-throughput and profile on the permeate side, noted P, was fit separately from the dynamic separation characteristics, Microfluid Nanofluidics velocity profile on the express lane side, noted EL). The figure 22, 59. https://doi.org/10.1007/s10404-018-2078-9. below shows that the express lane halves have an exponent that decays as a function of the Reynolds number based on gap [5] Pease, L.F., N.R. Philips, J. Serkowski, T.G. Veldman, M.J. parameters (i.e., the width of the gap and the average velocity Minette, C.A. Burns. 2022. Industrial Scale Mesofluidic through the gap). The dashed line represents the exponential Particle Separation, Chemical Engineering Processing – expression for open pipes by Barenblatt and Goldenfeld [11] Process Intensification 173 (2022) 108795. (1995) under turbulent conditions. The halves of the velocity https://doi.org/10.1016/j.cep.2022.108795. profile near the permeate side appear to follow the laminar condition to higher gap Reynolds numbers before beginning to [6] Pease, L.F. T.G. Veldman, J. Serkowski, N.R. Phillips, R.C. decay. This suggests that the two halves of the velocity profile Daniel, M.J. Minette, X.Y. Yu, M.S. Fountain, C.A. Burns. behave differently with turbulence somewhat depressed on the 2020. Efficient Mesofluidic Separation of Large Particles in permeate side in contrast to the express lane side. These results Nuclear Slurries, Waste Management Symposium 2020 have implications for the way that turbulent diffusion should Proceedings, Paper 20408. affect the motion of particles through the separator. [7] Burns C.A., T.G. Veldman, J. Serkowski, R.C. Daniel, X.-Y. CONCLUSIONS Yu, M.J. Minette, L.F. Pease. 2021. Mesofluidic Separation In summary, we generated a model to evaluate the pressure versus Dead-end Filtration. Separation and Purification drop and split fraction versus flowrate for mesofluidic devices. Technology, 254, 117256. The model comprises an overall analytical framework informed https://doi.org/10.1016/j.seppur.2020.117256. by select CFD results. In this manner, the results are not limited to select instances dictated by CFD constraints but allow the [8] Afzal, N., A. Seena, A. Bushra. 2013. Turbulent flow in a model to capture and predict over a wide range of conditions, machine honed rough pipe for large Reynolds numbers: which benefits from the precision that CFD provides. general roughness scaling laws. J. Hydro-Environ. Res. 7, 81-90. https://doi.org/10.1016/j.jher.2011.08.002. ACKNOWLEDGEMENTS Pacific Northwest National Laboratory is a multiprogram [9] Incropera, F.P., D.P. DeWitt, T.L. Bergman, A.S. Lavine. national laboratory operated for the U.S. Department of Energy 2007. Fundamentals of Heat Transfer. Sixth Edition. John by Battelle Memorial Institute under Contract DE-AC05-76RLO Wiley & Sons, Hoboken, NJ. 1830. Funding for this work was provided by the U.S. Department of Energy under Department of Energy Contract [10] Beale, S.B. 2012. A Simple, Effective Viscosity Formulation DE-AC05-76RL01830. for Turbulent Flow and Heat Transfer in Compact Heat Exchangers. Heat Transfer Engineering, 33, 4-11. REFERENCES [1] Huang, L.R., E.C. Cox, R.H. Austin, J.C. Sturm. 2004. [11] Barenblatt, G.I. and N. Goldenfeld. 1995. Does fully Continuous particle separation through deterministic lateral developed turbulence exist? Reynolds number displacement. Science, 304, 987–990. independence versus asymptotic covariance. Phys. Fluids 7, https://doi.org/10.1126/science.1094567. 2078-2082. https://doi.org/10.1063/1.868685. [2] Inglis, D.W., J.A. Davis, R.H. Austin, J.C. Sturm. 2006. [12] Giese, M., K. Rottschafer, D. Vortmeyer. 1998. Measured Critical Particle Size for Fractionation by Deterministic and modeled superficial flow profiles in packed beds with Lateral Displacement Lab Chip, 6, 655–658. liquid flow. American Institute of Chemical Engineers https://doi.org/10.1039/B515371A. Journal, 44, 484-490. https://doi.org/10.1002/aic.690440225. [3] Pease L.F., J.E. Serkowski, T.G. Veldman, J. Williams, X.- Y. Yu, M.J. Minette, J.A. Bamberger, C.A. Burns. 2021. Can PNNL-SA-212979 8 © 2025 by ASME Appendix B: Test of Mesofluidic Inline Separator Coupling with Reverse Osmosis Submitted to the ASME 2026 Fluids Engineering Division Summer Meeting FEDSM2026 July 26-29, 2026, Bellevue, Washington FEDSM2026-XXXXXX PRODUCED WATER TREATMENT USING MESOFLUIDIC INLINE SEPARATION AND REVERSE OSMOSIS Carolyn A. Burns, Stony T. Akins, Jason Serkowski, Chinmayee Subban, Judith A. Bamberger, Michael J. Minette, Leonard F. Pease Pacific Northwest National Laboratory Richland, Washington, USA ABSTRACT INTRODUCTION Here we demonstrate the combination of mesofluidic inline Mesofluidic inline separators represent an opportunity to separators and reverse osmosis to remove dissolved and remove a key barrier in the treatment of produced water: nondissolved solids from produced water to drinking water suspended solids that clog downstream operations. The US concentrations. Mesofluidic inline separators have previously alone produces very large quantities of produced water each year, demonstrated selective particle separation by size in high bay most of which is reinjected as a waste product [1]. The impact environments, and reverse osmosis is a standard commercial from treating and reusing even a fraction of this water is approach to desalination. A unit operation upstream of the immense. Although the amount of produced water from reverse osmosis membrane is essential to remove non-dissolved petroleum operations is vast and some must be returned to solids and oil droplets to protect this membrane. However, rebalance geophysical stresses, only a small fraction of it is coupling these separators with reverse osmosis has not been reused or turned into potable water due, for example, to the lack explored previously. of cost competitive technologies to remove both dissolved and Here we describe the first design, construction, and bench suspended solids at significant throughput. testing of coupled mesofluidic inline separators with reverse The incumbent technology for desalination is reverse osmosis systems. We show that mesofluidic inline separators can osmosis (RO) due to its technological maturity and ability to treat increase the amount of solids that the system is subjected to a wide range of feedwaters. However, suspended solids may before clogging by approximately an order of magnitude. cause RO (and other dissolved solids removal technologies) to Further, the produced water that enters the mesofluidic inline become clogged and lose performance. A common strategy uses separator at ~5300 ppm emerges from the reverse osmosis dead-end filters upstream of RO systems, but membrane-based membrane at ~80-360 ppm, well within the drinking water filters, including the dead-end prefilters, commonly upstream of range. reverse osmosis membranes, easily clog and require regular These results comprise the first essential step in treating maintenance. produced water using this combination of technologies, because Unlike membrane filters, mesofluidic inline separators suspended solids can clog downstream unit operations including provide suspended solids removal with much lower pressure reverse osmosis systems. The combination of technologies drops than conventional filtration systems, permitting provides an important pathway for the treatment and reuse of substantially higher flow rates [2, 3]. Mesofluidic inline produced water (e.g., for surface discharge, reusable water, separators have remarkably smaller footprints than competing agricultural water, industrial use water, potable water, etc.). filtration technologies, are easily transportable between jobsites, Keywords: slurry, solids removal, deterministic lateral are tunable for particle sizes of interest, are fully passive, and displacement (DLD), bump arrays, water purification, produced have no moving parts so that solids removal can be accomplished water, brine, desalination, critical minerals, critical materials at much lower cost [4, 5]. Instead of bulky rate limiting filters to remove suspended solids, this technology fits simply within standard schedule pipes to operate at industrial flow rates and 1 meet safety standards and regulations [5]. The separator does scale is substantially similar to the separation precision achieved not require multiple operators but can be readily operated using at deployed scale. In this manner, the cost and time to cross the a single pump (perhaps even gravity flow with sufficient “valley of death” from early prototypes to commercial product is head/gravitational drop). Because these systems are now substantially reduced. produced by rapid advanced manufacturing techniques, Mesofluidic separators are not membranes. Physically, deployment can be very fast. mesofluidic inline separators partition particles by their size. The key to the technology1 is the interaction among fluid flow v fields, particles and posts.v Small particles follow the streamlines Numbered set by staggered posts in contrast to larger particles that cannot Up Post Arrays and so must consecutively flow to one side of the posts (Fig. 1 Industrial Pipe upper left). The post array is housed in a pipe with an open space along one side called an express lane (bottom open region in Fig. Express Lane 1 upperPucks right). Under flow, the post array forces large particles v Express Lane into this express lane. At the end of a post array (Fig. 1 lower (posts not shown) v v middle), the flow is simply partitioned with the express lane flow going one direction and the remaining flow depleted of large Numbered Up Post Industrial Pipe particles going another (i.e., to RO units). Arrays TABLE 1. SELECT PROOF-OF-PRINCIPLE PROTOTYPES. Express Lane Pucks v Scale Form Particle Slurry Flow (posts not shown) Express Lane factor – size rate material cutoff* FIGURE 1. (UPPER LEFT) SCHEMATIC OF PHYSICS OF Bench scale Block – 700 µm Glass beads 1-2 MESOFLUIDIC INLINE SEPARATOR SHOWING THAT SMALL polymer in water gpm PARTICLES SLALOM THROUGH THE POSTS WHILE LARGE Lab scale Block – 700 µm Glass beads Up to PARTICLES CONSECUTIVELY BUMP TO ONE SIDE DUE TO polymer in water ~4 POST STAGGERING [3]. (UPPER RIGHT) 3D PRINTED PUCK gpm WITH POSTS THAT DRIVE LARGE PARTICLES INTO THE OPEN Engineering Cylinder 40 µm Caustic ~0.5 “SMILEY” VOLUME TERMED THE EXPRESS LANE NEAR THE – metal simulants L/min BOTTOM [5]. (LOWER MIDDLE) ENGINEERING DRAWING OF “Half” scale Puck – 520 µm Abrasive 90-100 PUCKS WITHIN A PIPE HOUSING THAT DIVIDES THE FLOW polymer simulant gpm INTO A LARGE PARTICLE ENRICHED STREAM FROM THE *Based on formula of [8] were flow to have been laminar. Bench EXPRESS LANE AND A PARTICLE DEPLETED STREAM [4]. and lab scale devices had an equilateral rhombus arrangement of pins with a pin diameter of 1.2 mm, a pin gap of 1.8 mm and Mesofluidic inline separators are passive inserts that fit an offset of 0.3 mm. The engineering scale device had an within existing piping systems. These separators leverage equilateral rhombus arrangement of pins with a pin diameter of techniques learned over decades of microfluidic advances but 0.5 mm, a pin gap of 0.3 mm and an offset of 0.0101 mm. The have been scaled up to operate at industrially relevant conditions “half” scale device had a rhombus arrangement of pins with a [6]. Mesofluidic inline separators draw inspiration from a pin diameter of 1.2 mm, a flow wise pin gap of 1.815 mm, a microfluidic system that originally separated the constituents of transverse pin gap of 1.34 mm, and an offset of 0.3 mm. blood (e.g., red from white blood cells from platelets) and are now used to diagnose some cancers. These separators have been We have designed, constructed and tested several prototypes scaled up to accommodate broad particle size distributions, to of increasing complexity as shown in Table 1, each using rapid modularly fit within three-inch pipes (see Fig. 1), and to operate advanced manufacturing techniques. The flowrates and size at flowrates >100 gpm (as opposed to ~10 mL/h) [5]. Like separation targets were motivated by a nuclear waste processing microfluidics, numbering up is a key to rapid deployment [6]. application but suggest the capability of these separators for a Once the technology has been scaled up in a laboratory broad range of applications [4]. The bench scale prototype was environment, modular separating elements are assembled into constructed with post between parallel plates (i.e., a block form pucks, multiple modular pucks are assembled into a separation factor; see Fig. 1 with four vertical parallel plate separators stage, and multiple separation stages may be assembled into a stacked together above the express lane) and demonstrated separations cascade where multiple separation targets are particle separation at turbulent flowrates using a gravity drained required to satisfy product specifications (see Fig. 1). Multiple flow system with transient concentration pulses [3]. The lab cascades may number up to achieve a target throughput [7]. scale prototype was a substantially longer version of the bench Therefore, demonstrating separation precision at the laboratory scale (>3 ft; too large to fit on benchtop) with 13 channel output 1 U.S. Application No. 16/395,097 pending. 2 to evaluate particle distributions across the separator outlet. The added incrementally. Some tests proceed only through prefilter engineering scale prototype was constructed of ~5300 posts in a (as a control) and some through both the mesofluidic separator super nickel alloy similar to stainless steel with a circular and the prefilter. external form factor [2]. This prototype showed that mesofluidic inline separators are competitive versus metal dead-end filters in terms of particle size separated but at ~10 times more flow at the same pressure drop, suggesting operation orders of magnitude longer without clogging. The half scale prototype (Fig. 1) had a 3 inch external diameter with twenty pucks stacked together within a stainless steel tube to form a single separator module [5]. This prototype showed that polymer inserts performed well despite an abrasive simulant in an industrial high bay environment. Together, these studies validate the performance of mesofluidic inline separators relative to performance predictions and show that cutoff sizes tested (~40-600 m) span much of the range of suspended solids to be removed in produced water treatment. In this paper, we show that mesofluidic inline separators can be coupled with a commercial dissolved solids removal technology (RO) and quantify the increase in the dead-end filter lifetime before clogging. MATERIALS AND METHODS Unique produced water samples were obtained from two wells in the Permian Basin in West Texas (see Fig. 2). The particle size distributions from these two sites had a d50 of 16-31 microns for 33-28 #2H, and a d50 of 7-11 microns for 33-28 B#3H (see Fig. 3). The d90 is 30-81 microns for 33-28 #2H (overlapping at least partially the size range of the MoSci 63-90 FIGURE 2. IMAGES OF THE PRODUCED WATER SAMPLES micron particles used in these experiments to represent larger FROM APEX IN VIALS. THE TWO ON THE RIGHT ARE 33-28 #2H AND THE THREE ON THE LEFT ARE 33-28 B#3H. THE VIAL minimally setting particles found in produced water). These ON THE LEFT WAS PULLED TWO WEEKS EARLIER THAN THE particle sizes are consistent with the slow settling that was NEIGHBORING VIALS. THE BOTTOM IMAGE WAS observed. The particle size distribution emerging from wells COLLECTED APPROXIMATELY TWO HOURS AFTER PULLING varies over time, and these larger particles are likely typical of FROM THE CARBOY, SUGGESTING THAT OXIDATION proppant, cuttings, and larger particulate remaining from the PROCEEDS OVER TIME. initial fracturing process. The fully assembled system (see Fig. 4) can achieve pressures of 300 psi max. This safety limit restricts the salt concentration through the RO system to approximate 5000-7000 ppm, which lies within the TDS range of US produced waters (~1100-244000 ppm) [1]. The mesofluidic device used in these tests (see Fig. 5) was custom designed with dimensions as given in Table 2. During operation, a pump (see Table 3) was used to drive flow through the system at flows given in Table 4. Instant Ocean® Sea Salt or produced water was first driven through the mesofluidic separator as shown in Fig. 4. Flow that had proceeded through the pin array within the mesofluidic separator proceeded toward the dead-end filter (Hydronix 5 micron, SDC-25-1005). The remaining fluid from the mesofluidic separator that is enriched in particles was collected. Flow emerging from the dead-end filter was also collected for further analysis and was in some cases fed to the reverse osmosis module separately (Pro+Aqua, PRO-RO-1 with high salt membrane from Fresh Water Systems FIGURE 3. PARTICLE SIZE DISTRIBUTIONS. Filmtec SW30-4021 Sea Water 800 GPD RO Membrane Element). After steady-state flow was achieved, particles were 3 TABLE 2. MESOFLUIDIC INLINE SEPARATOR DIMENSIONS. Nominal functional range 45-175 microns Pin diameter 1 mm Center-to-center 1.53 mm Offset 8.2 microns Pins in first row 7 Width available to flow 20 mm Height 12.7 mm Length (w/o end pieces) 2.36 m Length (with end pieces) 2.71 m Conical entrance angle 5.0o (w.r.t. axis) TABLE 3. INSTRUMENTATION. Instrument Model Pump Dayton stainless steel centrifugal pump Model 5WXT1A Pump PPT05014GGG Pressure Digital pressure gauge (SSI Technologies, Inc. MediaGaugeTM MG-50-A-9V-R) pressure transducer and an analog pressure gauge (GRIFFCO) Flow Magnetic-inductive flow meter (Ifm electronic magnetic-inductive flow meter SM7001) Tank 15-gallon conical bottom Overhead RW16 basic IKA WERKE stirrer TABLE 4. PERFORMANCE. Added RO Feed Product Product FIGURE 4. IMAGE OF THE COMBINED SETUP INCLUDING A Mass Pressure Flowrate TDS Flow Split MESOFLUIDIC SEPARATOR, A PREFILTER, AND A REVERSE (psig) (gpm) (ppm) OSMOSIS MEMBRANE. THE MESOFLUIDIC SEPARATOR IS ON ~0-2 kg 290-299 0.52-0.53 82-357 41-50% THE RIGHT IN WHITE. THE REVERSE OSMOSIS UNIT Based on 4 L of produced water starting from a clean INCLUDING WHITE PREFILTER IS IN THE BLACK CART ON membrane based on DI water; produced water used was 33- THE LEFT. 28#2H. RESULTS AND DISCUSSION The purpose of this paper is to explore how mesofluidic inline separators can be coupled or cooperatively used with a commercial dissolved solids removal technology (RO) and quantify the increase in the dead-end filter lifetime before system clogging. Using the assembled system (see Fig. 4), we compared the performance of a prefilter with and without the mesofluidic separator installed upstream (see Fig. 4). While in some instances it may be possible to replace the prefilter with the mesofluidic separator, most reverse osmosis systems include a prefilter to ensure that a negligible quantity of particles interact with the reverse osmosis cartridge. The quantity of particles that prefilters can sustain is meaningful. For the cartridge we used FIGURE 5. CROSS SECTION OF A MESOFLUIDIC INLINE approximately a kilogram of particles in Instant Ocean® Sea Salt SEPARATOR (TERMED 45-175 AFTER THE NOMINAL (diluted to 5300 ppm to meet the pressure limit above) that could FUNCTIONAL RANGE IN MICRONS) WITH INLET ON LEFT be loaded before the cartridge lost flow. Because we used a AND BOTH OUTLETS ON RIGHT. LONGITUDINAL AXIS centrifugal pump, the flowrate may be a more sensitive indicator REDUCED TO FIT COLUMN WIDTH. of cartridge performance than pressure drop. Figure 6 compares 4 the flowrate sustained versus the cumulative mass of particles undesirable species from the produced water. Table 4 shows that added to the feed. The figure shows that whereas approximately the remaining total dissolved solids is within the potable water ~0.7 kg of particles can be fed to the system before the dead-end range, though the chemical composition of those solids remains filter began to clog in the absence of a mesofluidic separator to be determined to fully ascertain whether this water would (complete clogging ~1 kg); with an in-line mesofluidic separator classify as potable water. >7 kg could be added to the system. This represents an order of magnitude factor increase in solids fed. The improved performance is not free; the pressure drop does increase from ~6 psi to ~14 psi with the addition of the mesofluidic separator. FIGURE 7. VIALS SHOWING THE (LEFT) RAW PRODUCED WATER, (MIDDLE) WATER FROM THE MESOFLUIDIC SYSTEM AND PREFILTER, AND (RIGHT) TREATED WATER FROM RO MEMBRANE. Figure 8 also highlights that there is more than one location where species of interest could exit. For example, critical FIGURE 6. FLOWRATE AS A FUNCTION OF CUMULATIVE minerals could exit through either the reverse osmosis stream or MASS ADDED TO THE FEED WITH AND WITHOUT THE alternatively be enriched in the materials trapped by the prefilter MESOFLUIDIC SEPARATOR (NO RO). or among the particulate enriched in the mesofluidic separator. The ultimate target of this effort is the combination of a mesofluidic separator with prefilter and a reverse osmosis membrane. Operation proceeded using unique produced water samples from the Permian Basin in West Texas diluted (to ~5300 ppm down from ~13400 ppm) to prevent the pressure drop across the membrane from exceeding 300 psi. Particles (Mo-Sci LLC Mesofluidic 63-90 soda lime glass spheres with a size of 63-90 microns) were Produced Se arator added to the feed tank in twice the amount necessary to ensure Water that in the absence of the mesofluidic separator the prefilter would have clogged, and the flowrate for the reverse osmosis Particle- membrane would have been negligibly small (see Fig. 3). The enriched experiments showed that the performance with Instant Ocean is Treated Fluids not distinguished from that of produced water. The water Waters produced through the reverse osmosis membrane is not Dead-end Filtration distinguishable at the beginning, middle or end of the run with RO and Reverse Osmosis produced water (see Table 4). This result indicates that Waste Used mesofluidic separators can be useful in preserving the Filter functionality of reverse osmosis systems process produced water (see Fig. 7). FIGURE 8. REPRESENTATION OF THE MESOFLUIDIC- Figure 8 compiles images of characteristic fluids entering REVERSE OSMOSIS SYSTEM WITH IMAGES OF THE FLUIDS and exiting the combined system. As anticipated the particle rich AS THEY ENTER AND EXIT THE SYSTEM. THE PARTICLE RICH stream exiting from the mesofluidic separator is enriched in STREAM EXITING THE MESOFLUIDIC SEPARATOR particles that appear white as they jet out of the separator into the HIGHLIGHTS A PARTICLE STREAM IN INSTANT OCEAN beaker. Also of note is that there are three ways that flows can JETTING INTO A BEAKER. THE OTHER IMAGES ARE FOR exit this system. Here the focus has been on removing PRODUCED WATER. 5 CONCLUSION Can Bump Arrays Separate Particles from Turbulent Flows? In summary, we have demonstrated removal of suspended In Proceedings of the ASME 2021 Fluids Engineering solids from produced water using a mesofluidic inline separator. Division Summer Meeting (FEDSM 2021), August 10-12, This testing is a critical step toward obtaining industry 2021, Virtual, Online, 3, FEDSM2021-67696, acceptance for this revolutionary technology that promises to be V003T08A024. New York, New York: ASME. a key element in water treatment systems that turn produced https://doi:10.1115/FEDSM2021-67696. water from a waste into a benefit (e.g., for surface discharge, reusable water, agricultural water, industrial use water, potable [4] Pease, L.F., T.G. Veldman, J. Serkowski, N.R. Phillips, R.C. water, etc.). Daniel, M.J. Minette, X.Y. Yu, M.S. Fountain, C.A. Burns. 2020. Efficient Mesofluidic Separation of Large Particles in ACKNOWLEDGEMENTS Nuclear Slurries, Waste Management Symposium 2020 We express appreciation to the United States Department of Proceedings, Paper 20408. Energy (DOE) Fossil Energy and Carbon Management (FECM) Technology Commercialization Fund TCF-21-24900 for partial [5] Pease, L.F., N.R. Philips, J. Serkowski, T.G. Veldman, M.J. funding of this work. A portion of the research described in this Minette, C.A. Burns. 2022. Industrial Scale Mesofluidic report was conducted under the Laboratory Directed Research Particle Separation, Chemical Engineering Processing – and Development (LDRD) Program as a strategic investment at Process Intensification 173, 108795. the Pacific Northwest National Laboratory (PNNL), a https://doi.org/10.1016/j.cep.2022.108795 multiprogram national laboratory operated by Battelle Memorial Institute for the U.S. Department of Energy (DOE) under [6] Huang, L.R., E.C. Cox, R.H. Austin, J.C. Sturm. 2004. Contract No. DE-AC05-76RL01830. Continuous particle separation through deterministic lateral displacement. Science, 304, 987–990. REFERENCES https://doi.org/10.1126/science.1094567. [1] Scanlon, B., R. Reedy, P. Xu, M. Engle, J. Nicot, D. Yoxtheimer, Y. Qian, S. Ikonnikova. 2020. Can we [7] Pease, L.F., J.A. Bamberger, C.A. Burns, and M.J Minette. beneficially reuse produced water from oil and gas 2022. Concentrating Slurries Mesofluidically for Nuclear extraction in the U.S.? Science of The Total Environment, Waste Processing. Proceedings of the ASME 2022 Fluids 717, https://doi.org/10.1016/j.scitotenv.2020.137085 Engineering Division Summer Meeting, FEDSM2022- 87708 American Society of Mechanical Engineers, New [2] Burns C.A., T.G. Veldman, J. Serkowski, R.C. Daniel, X.-Y. York, NY. https://doi.org/10.1115/FEDSM2022-87708. Yu, M.J. Minette, L.F. Pease. 2021. Mesofluidic Separation versus Dead-end Filtration. Separation and Purification [8] Inglis, D.W., J.A. Davis, R.H. Austin, J.C. Sturm. 2006. Technology, 254, 117256. Critical Particle Size for Fractionation by Deterministic https://doi.org/10.1016/j.seppur.2020.117256. Lateral Displacement Lab Chip, 6, 655–658. https://doi.org/10.1039/B515371A [3] Pease L.F., J.E. Serkowski, T.G. Veldman, J. Williams, X.- Y. Yu, M.J. Minette, J.A. Bamberger, C.A. Burns. 2021. 6 Pacific Northwest National Laboratory 902 Battelle Boulevard P.O. Box 999 Richland, WA 99354 1-888-375-PNNL (7665) www.pnnl.gov