Search

Silica scaling of RO membranes was evaluated via real-time direct surface imaging demonstrating a capability for detecting the onset of silica scale formation and its evolution. Silica scaling was detected significantly earlier than by traditional flux decline measurements. The observed rate of silica particle nucleation followed classical nucleation theory while the growth of individual silica particles, at the early stages of silica scaling, was governed by diffusional growth. SEM and optical images of the membrane surface suggest that silica scaling occurs through the formation of both primary silica particles and their agglomerates (~1–30 µm), as well as a gel-like silica film embedded with silica particles both of which contribute to permeate flux decline. At low silica saturation index at the membrane surface (SIm≤1.93) silica gel film formation resulted in a smoother and less porous film than at higher silica saturation (SIm≥2.72). At the higher silica saturation levels (SIm=2.72–3.50), silica scaling resulted in larger observed particles as well as rapid permeate flux decline. The silica scale layer thickness was in the range of ~0.1–3.5 µm, with surface roughness being higher by a factor of 2.6–8.3 relative to the native membrane. Results of the present study suggest that there is merit in exploring the application of the present approach for early detection and monitoring of silica scaling in RO plants in support of strategies for silica scale mitigation.

Hydraulic permeability and molecular weight cutoff (MWCO) performance tuning of polysulfone (PSf) ultrafiltration (UF) membrane was attained via surface modification with a tethered layer of poly(acrylic acid) (PAA). Surface nano-structured (SNS) PAA layer was synthesized onto base PSf UF membranes by atmospheric pressure plasma (APP) surface activation followed by graft polymerization (GP) of acrylic acid (AA). Through a systematic study, it was shown that effective SNS-PAA-PSf UF membrane performance tuning was feasible by adjustments of APP and graft polymerization conditions. It was shown, for the first time, that SNS-PAA-PSf membranes can be synthesized with a range of hydraulic permeability (spanning a factor of 1.1–2.6 in magnitude) for a given MWCO, or a range of MWCO (spanning a factor of 1.5–2.3 in magnitude) for a given hydraulic permeability, thereby overcoming the hydraulic permeability-MWCO tradeoff.

Ultrafiltration as a pretreatment for RO feedwater with enhanced UF backwash, which combines continuous with pulse backwash, was investigated in a novel UF-RO process integration. Direct supply of RO concentrate to the UF module served for UF backwash which was further enhanced with pulse backwash generated using bladder-type hydraulic accumulators. Model analysis of the hydraulic accumulator operation, which was validated via a series of field experiments, demonstrated a capability for accumulator charging directly from the RO concentrate stream within a period of 30–40 s. Moreover, pulse backwash over a short period (~5 s) which was added to the continuous UF backwash (directly from the RO brine stream), enabled peak UF backwash flux up to a factor of 4.2–4.6 higher than the normal filtration flux. The above mode of UF operation with multiple consecutive backwash pulses was more effective than with a single pulse, while inline coagulation further increased the UF performance. Relatively long-term field operation (over eight days where) of the UF-RO system with self-adaptive triggering of UF backwash, whereby the number of consecutive pulses increased when a higher membrane fouling resistance was encountered, was highly effective in enabling stable UF operation over a wider range of water quality conditions and without the need for chemical cleaning. These encouraging results suggest that direct UF-RO integration with enhanced pulse UF backwash is an effective approach for dead-end UF filtration without sacrificing water productivity.

Reverse osmosis (RO) polyamide (PA) membrane integrity loss, performance degradation, and alteration of surface properties due to exposure to chlorine were evaluated experimentally via a fluorescent marker (uranine) based method, water permeability and salt flux measurements, and surface characterization via XPS, AFM and contact angle measurements. Membrane exposure to chlorine (4–200 mg/L NaOCl solutions) revealed that although membrane surface roughness increased with chlorine exposure intensity (ppm-h), surface hydrophilicity increased as inferred by the decline (up to 5.2–9.6%) of the surface energy of hydration. Comparative analysis of marker transport for membranes that have undergone chlorine exposure of 125–2000 ppm-h demonstrated that the membrane marker permeability coefficient (B) increased by up to a factor of ~5 relative to the intact membrane. While the severity of membrane integrity loss/performance degradation correlated to a reasonable degree with ppm-h of chlorine exposure, the severity of membrane integrity loss, at the same chlorine ppm-h, was greater for higher exposure concentration. Membrane integrity loss, over the same exposure levels, was quantified by an equivalent cylindrical breach that was in the range of about 14–40 µm, increasing in size with the intensity of chlorine exposure.

Polyamide (PA) thin film composite (TFC) BWRO and SWRO membranes were surface nano-structured (SNS) with tethered hydrophilic poly(acrylic acid) (PAA) layers. BWRO and SWRO membranes having water (Lp) and salt (B) permeability coefficients in the ranges of 2.3–3.4 L m−2 h−1·bar−1 and 0.15–0.54 L m−2 h−1, respectively, were utilized as base membranes onto which tethered PAA layers were synthesized via acrylic acid (AA) graft polymerization (GP), post-membrane surface activation with atmospheric pressure plasma (APP). Both APP surface activation and GP conditions impacted SNS-PAA-PA membrane performance evaluated in terms of the membrane Lp and B parameters. Surface structuring with a tethered PAA layer enabled tuning of membrane performance (in terms of Lp and B) to achieve water/salt selectivity (evaluated as Lp:B ratio) that was significantly higher (by up to 56%) relative to the base PA membranes. Depending on APP surface activation and AA graft polymerization conditions, membranes could be tuned to have essentially the same salt rejection over a range of permeability or a given permeability over a range of salt rejection. It was also shown that a membrane performance could be achieved that overcomes the permselectivity tradeoff.

The use of 2D finite element CFD model and the common film model (FM) for analysis of FO performance data was examined to assess the merits of conclusions derived from the above approaches regarding characterization of membrane and support layer resistances to water and salt permeation. The FM model leads to estimation of porous layer resistivity (K) that implies tortuosity/porosity ratio that is significantly higher than is physically attainable for typical membrane porous supports. Comparative FO data analysis with the 2D CFD model and FM approximation reveals that the latter approach overestimates (by an order of magnitude and greater) the significance of the permeation resistance of the FO membrane support layer. It is also shown that membrane (water and solute) permeabilities extracted from FO operation in the RO mode (FO/RO) are significantly higher than obtained from direct analysis of FO performance data via detailed 2D CFD modeling. The present analysis suggests that improvements in FO water permeability are more likely to emerge from developing higher permeability membrane skin and improved channel hydrodynamics (to decrease external concentration polarization) rather than from more permeable and thinner membrane support layer as has been asserted in various studies.

Forward osmosis (FO) desalination was investigated via 2-D numerical model of the fully coupled hydrodynamics and mass transfer equations. The model was formulated for a detailed composite channel structure (feed and draw channels, membrane skin layer and porous support) being capable of describing co-current or counter current cross operation where the membrane skin faces the salt feed solution (SFF) or where the membrane skin faces the draw solution (SFD). Simulations based on existing experimental FO data confirmed that FO operation in a counter-current/SFD mode provides slight improvement with respect to water flux, and reduced cross migration of feed and draw solutes relative to the co-current mode of operation. Analysis of existing FO data also revealed the dependence of the intrinsic membrane water permeability and solute transport coefficients on draw solute concentration. Simulation results indicated significant cross membrane migration of feed and draw solutes for long (~1 m) relative to short (~10 cm) FO channels. Moreover, up to an order of magnitude decline of draw solute concentration difference (along the membrane) can be encountered at the draw channel exit region. Simulation results suggest that accurate assessment of FO performance in long channels is critical for full-scale plant design in order to minimize salt leakage, optimize recovery, and setting accurate inlet/outlet conditions to enable simulations of membrane elements in series.

Membrane mineral scaling in semi-batch RO (SBRO) and steady-state RO with partial recycle (SSRO-PR) were investigated experimentally in a spiral-wound pilot system with gypsum as a model scalant. Scaling in SSRO-PR operation, at a given level of solution supersaturation, was compared to SBRO operation at the equivalent time-averaged solution supersaturation, and also for operation at equivalent water recovery. Real-time tracking of mineral scaling revealed that SBRO was more prone to mineral scaling, in terms of both crystals formation rate and scaled membrane surface area, relative to SSRO-PR. The growth rates of crystals in both SBRO and SSRO-PR followed diffusional growth and were similar at the same level of solution supersaturation. However, the crystallization induction time was shorter and rate of crystal formation higher in SBRO relative to SSRO-PR. This behavior was consistent with the higher level of gypsum supersaturation reached in the SBRO filtration period when compared to SSRO-PR at either one of the above operational scenarios. For a given recovery, improved scale mitigation in SBRO via extended flushing to reset the “crystallization induction clock” may be feasible, but the tradeoff is an extended filtration period with a higher level of solution supersaturation.

Semi-batch reverse osmosis (SBRO) desalination was investigated via process modeling and experimentally using a spiral-wound SBRO pilot system. The salinity increase during SBRO filtration necessitates periodic flushing of the excess salinity buildup in the system. Accordingly, concentrate flushing efficacy and its impact on SBRO performance were investigated for multi-cycles of filtration and concentrate flushing. Progressive cycle-to-cycle concentrate salinity rise, due to incomplete (concentrate) flushing, was apparent leading to corresponding SEC increase, but a stable cycle-to-cycle operation was ultimately achieved. Flushing duration longer than the convective residence time improved concentrate flushing. However, for a given recovery, longer flushing required longer filtration. The analysis confirmed that the SEC for an ideal SBRO operation (i.e., complete flushing of excess concentrate salinity) is lower relative to SPRO. However, ideal operation is likely to be a challenge with practical SBRO systems as it would necessitate complete concentrate flushing over the course of one convective residence time. Indeed, experimental evaluation of SBRO demonstrated higher SEC relative to single-pass RO (SPRO). Clearly, experimental quantification of concentrate flushing efficacy is critical for establishing SBRO effectiveness. Irrespective of SEC considerations, both SBRO and SPRO with partial recycle offer the flexibility of desalting operation over a wide recovery range.

The biofouling propensity and cleaning effectiveness were evaluated for a new generation of polyamide thin-film composite (PA-TFC) membranes that are surface nano-structured (SNS) with terminally anchored hydrophilic polymer chains. The SNS-PA-TFC membranes were formed via surface graft polymerization onto the surface of polyamide base membranes previously activated via surface treatment with atmospheric pressure plasma. Poly(methacrylic acid) (PMMA) and polyacrylamide (PAAm) SNS-PA-TFC membranes were synthesized and evaluated with respect to their biofouling resistance using secondary wastewater from a municipal wastewater treatment (MWT) plant. Biofouling resistance and cleaning effectiveness were quantified via flux decline measurements in addition to imaging of the biofouled membranes before and after DI water and chemical (Na 2 ·EDTA) cleaning. Increased biofouling resistance was highest for the PMAA-SNS-PA-TFC membrane being a factor of 4.2 greater than for the commercial membrane of about the same salt rejection used in the MWT plant. Moreover, permeability recovery for the PMAA-SNS-PA-TFC membrane was higher by factors of up to ~1.2 relative to the reference commercial membrane upon cleaning with DI water and more aggressive chemical cleaning at high pressure. The present evaluation of the SNS-PA-TFC membranes suggests that biofouling resistance and cleaning effectiveness of RO membranes can be enhanced via hydrophilic brush layers.

The influence of a pre-existing biofilm on RO membrane mineral scaling was evaluated using gypsum as a model scalant. The biofilm was established using microfiltered secondary treated wastewater effluent on-site in a wastewater treatment facility. Mineral scaling was then monitored via direct visual observation of crystal growth on the membrane surface in a transparent plate-and-frame RO cell. SEM imaging of the membrane surface and sectioned biofilm revealed gypsum crystals within the biofilm matrix, including the protrusion of crystal rods from within the biofilm. Both mineral scale surface coverage and crystal number density were greater in the presence of the biofilm. Moreover, individual mineral crystal growth rate within the biofilm was significantly higher relative to growth in the absence of a biofilm. Analyses of crystal growth and nucleation rates, within the biofilm and in its absence, suggest concentration polarization enhancement within the biofilm. The present study suggests that mineral scaling may not be restricted to tail elements in RO plants, but could also occur in lead elements where biofilms can enhance concentration polarization. Therefore, understanding of the coupling of biofouling and mineral scaling may be of particular significance to the operation and design of RO plants in water reuse applications.

The mineral scaling propensity of aromatic polyamide membrane surfaces was evaluated for surfaces ranging in roughness from about 1 to 94 nm, surface feature heights ranging from a few nanometers to as high as 400–450 nm and average peak separation of ∼140 nm to 2 μm. The polyamide surfaces were interfacially polymerized onto quartz crystal microbalance (QCM) sensors and silicon wafer substrates which were previously structured with a polyelectrolyte film formed by a layer-by-layer self-assembly method. Mineral scaling was quantified via QCM measurements of the time evolution of surface crystallization along with SEM and AFM surface characterization. Crystal surface mass density and rate of surface crystallization increased with increasing surface roughness. However, surface scaling propensity was also significantly impacted by the details of surface topography revealing fewer but larger crystals on rough surfaces that possessed large features with large separation distances. It is hypothesized that the separation distance between features and feature heights (dictating the geometry of surface “valleys”), for a given surface roughness, could affect the availability of surface nucleation sites as well as constrain growth of crystals within surface valleys.

An experimental method was developed for quantifying mineral scale nucleation and growth on RO membranes enabling comparison of the effectiveness of antiscalants in retarding mineral scale formation. Mineral scale growth and inhibition was evaluated, for calcium sulfate dihydrate (gypsum) as a model scalant, from direct visual observations of crystal growth on the membrane surface in a plate-and-frame RO cell. Comparison of membrane scaling with and without the use of antiscalants was carried out based on RO scaling tests conducted under identical operating conditions with respect to the initial flux and crossflow velocity, at identical initial concentration fields within the RO membrane channel and at the membrane surface. The range of gypsum saturation index (SIg) at the membrane surface was 1.3–2.3 enabling determination of gypsum scaling for the above SIg range from a single RO scaling test, given direct observation of crystal nucleation and growth on the membrane surface. The present work demonstrates that it is possible to distinguish antiscalants modes of scale inhibition and corresponding scale suppression effectiveness, over a reasonable range of solution supersaturation. The current approach introduces antiscalant crystal nucleation and growth retardation factors that can aid in the development, selection and dose optimization of antiscalants for use in RO desalination.

Surface nano-structuring of polyamide desalination membrane with a hydrophilic poly(methacrylic acid) was shown to reduce the membrane mineral scaling propensity as demonstrated with calcium sulfate dihydrate (gypsum). A two-step approach was employed, whereby the active polyamide (PA) layer of a thin-film composite (TFC) synthesized membrane was activated with impinging atmospheric plasma, followed by a solution free-radical graft polymerization (FRGP) of a water soluble methacrylic acid (MAA) monomer, at 60 °C and initial monomer concentration of 5–20% (v/v), onto the surface of the PA-TFC membrane. The approach of creating a layer of end-grafted poly(methacrylic acid) (PMAA) surface chains was first developed and evaluated using a surrogate polyamide membrane layer interfacially polymerized onto a thin poly(ethyleneimine) (PEI) film coated onto a silicon wafer. The resulting PMAA–PA-PEI–Si surrogate membrane surface was hydrophilic with a water contact angle range of 10–17°. Structuring of the PA-TFC membrane at equivalent FRGP reaction conditions resulted in membranes of higher permeability (by a factor of 1.3–2.26) relative to a commercial RO membrane of a similar surface roughness (∼70 nm) and salt rejection. Flux tests of membrane mineral scaling demonstrated that membrane mineral scaling propensity can be measurably reduced, relative to commercial membrane of the same salt rejection, while yielding equivalent or higher water permeability. The onset time for gypsum scaling for the optimal membrane surface (prepared at 10% (v/v) initial MAA concentration) was retarded by a factor of 2–5 relative to the commercial RO membrane. Current work is ongoing to further optimize the surface structure in order to increasing scaling resistance and assess the impact of surface structuring on nucleation of mineral salt crystals.

An analysis of the impact of increasing RO membrane permeability on the reduction in water desalination cost was carried out for RO desalting operated up to the limit imposed by the thermodynamic restriction. The premise of the present assessment is that, the current generation of high permeability RO membranes makes it feasible to carry out RO desalination up to the thermodynamic restriction limit. In this limit, the ratio of membrane to energy cost can be expressed as a function of the water recovery level and a dimensionless cost parameter that accounts for the purchase cost of electrical energy and membrane area, as well as feed water salinity, salt rejection requirement and membrane permeability. The present analysis suggests that, given the present day electrical energy and membrane prices, the benefit of developing membranes of even greater permeability is primarily at low water recoveries for inland brackish water desalting. At low water recoveries, however, there is typically an added cost associated with brine management for inland water desalting. In RO desalting of seawater, on the other hand, the specific cost (i.e., per permeate produced) of energy is much higher relative to the membrane cost, and there is lower economic incentive for developing higher permeability membranes if the objective is to lower the cost of seawater desalination. The analysis suggests that further significant improvement in RO membrane permeability is less likely to be the primary driver for a major reduction in the cost of seawater desalination. However, it is expected that significant reduction in RO water production cost can arise from a variety of other process improvements including, but not limited to, development of improved fouling-resistant membranes, more effective and lower feed pretreatment and brine management costs, optimization of process configuration and control schemes, as well as utilization of low cost renewable energy sources.

The energy optimization of two-pass membrane desalination, at the limit imposed by the thermodynamic restriction, was investigated and compared with a single-pass membrane desalting operation at the equivalent targeted overall salt rejection and permeate product recovery. The analysis considered the effect of pump and energy recovery efficiencies, membrane rejection, and retentate recycling from the second to the first-pass. The optimization results suggest that when the desired overall salt rejection can be achieved via a single-pass, then this process configuration will result in a lower specific energy consumption (SEC) relative to the two-pass membrane desalting process. However, if the desired overall rejection (or specific ion rejection, e.g., boron) cannot be achieved with the highest available rejection membrane in a single-pass, then a two-pass configuration is a feasible solution. In the latter case, the lowest energy consumption will be attained when a membrane of the highest available salt rejection is used in the first-pass. In cases in which desalting is accomplished at recoveries below the critical water recovery (i.e., the optimal recovery for a single stage), an operational sub-domain may exist in which the SEC for a two-pass process can be lower than for a single-pass counterpart, but only if the single-pass process is not operated at its globally optimal state. Retentate recycling from the second-pass to the first-pass feed can reduce the SEC for the two-pass process. However, optimization solution of the two-pass process is always a single-pass process as the optimal desalination process.

A 3-D (3-dimensional) numerical modeling approach to analyze the coupling of mineral scale, concentration polarization and permeate flux was developed and demonstrated for a gypsum membrane scaling test in a plate-and-frame RO membrane module geometry. The impact of concentration polarization on mineral gypsum scaling, and in turn the impact of mineral scale on the concentration polarization field, were explored via 3-D finite-element numerical solutions of the coupled fluid hydrodynamics and solute mass transfer equations, along with detailed experimental data on the extent and location of mineral scale. Numerical simulations of the concentration field for a scale-free membrane revealed that the regions of highest supersaturation with respect to calcium sulfate corresponded to regions of highest gypsum scale density, as observed through real-time imaging of the membrane surface in the gypsum scaling test. 3-D simulations of the concentration field in the presence of mineral scale revealed that the concentration polarization modulus and permeate flux were largely unaffected in contiguous scale-free regions of the membrane. However, near and just downstream of individual crystal formations, the local concentration polarization modulus decreased (by ∼5%) and the permeate flux increased (by ∼2%) relative to the same positions in the absence of scale. The model-calculated and experimental permeate flux decline agreed closely with an average absolute error of 1.8%. The present study suggests that flux decline due to mineral scaling can be reasonably described by the surface blockage mechanism.

The feasibility of operating brackish water reverse osmosis (BWRO) desalination in a feed flow reversal (FFR) mode so as to mitigate mineral scaling was evaluated with feed solutions containing calcium, sodium, sulfate and chloride ions. The premise of the FFR approach is to enable dissolution of surface mineral crystals, that form first at the BWRO plant tail elements, by periodic exposure to the undersaturated feed solution. Automatic triggering of FFR was accomplished by integrating an Ex-situ scale observation detector (EXSOD) with the BWRO plant controller. The EXSOD system consisted of a small high pressure plate-and-frame RO cell, with its membrane surface digitally imaged in real-time, to monitor the onset and development of membrane mineral scaling. During normal feed flow (NFF) operation, the EXSOD monitored a concentrate side-stream from the BWRO plant tail element, signaling the BWRO plant controller to initiate FFR when the specified scaling threshold was reached. Once the plant feed flow was reversed, the EXSOD feed stream was switched from the BWRO plant concentrate to the plant's permeate in order to dissolve mineral crystals formed on the EXSOD membrane during NFF operation. Successful operation of automated FFR, whereby scale formation was prevented, required adjustment of the EXSOD operating conditions such that the saturation index in the EXSOD RO cell was at a level that would not result in scale detection significantly before or after scaling would occur in the RO plant. Although the onset of mineral scaling was detected at the level of appearance of the first observable mineral crystal, the results suggest that it would be more practical to trigger FFR based on a threshold that considers both the mineral crystal number density and fractional surface scale coverage. Current work is continuing to demonstrate the broad applicability of FFR mode of RO mineral scale-free operation (with reduced or no antiscalant addition) over a wide range of solution compositions and operating conditions.

A neural network-based modeling approach with back-propagation and support vector regression algorithms was investigated as a mean of developing data-driven models for forecasting reverse osmosis (RO) plant performance and for potential use for operational diagnostics. The concept of plant “short-term memory” time-interval was introduced to capture the time-variability of plant performance since both a state of the plant model and standard time-series analyses for both flux decline and salt passage did not result in realistic predictive horizons for practical purposes. Past information of normalized permeate flux and salt passage were introduced as unique input variables along with process operating parameters to capture short-term plant performance variability. Sequential models, where the time-variation within each forecasting time-interval was also taken as input information, and marching forecasting models, where target values were predicted at fixed future times from past plant information, were developed. Models were trained, with normalized permeate flux and salt passage, for various model architectures, memory time-intervals and forecasting times using both back-propagation and support vector regression approaches. State of the plant models (without forecasting) were able to describe the relatively small permeate flux variations but were unable to capture salt passage trends (for any present time condition) since unsteady state phenomena could not be properly described without plant memory information. Forecasting of plant performance, with both sequential and marching models, yielded good predictive accuracy for short-term memory time-intervals in the range of 8–24 h for permeate flux and salt passage for forecasting times up to 24 h. Current work is ongoing to extend the approach for longer time scales and to incorporate data-driven forecasting models of RO plant into control strategies and process diagnostics.

Organic solute permeation, sorption, and rejection by reverse osmosis membranes, from aqueous solutions, were studied experimentally and via artificial neural networks (ANN)-based quantitative structure–property relations (QSPR), for a set of fifty organic compounds for polyamide and cellulose acetate membranes. Membrane solute sorption and passage for dead-end filtration model experiments were quantified based on radioactivity measurements for radiolabeled compounds in the feed, permeate and the membrane, while solute rejection was determined from a mass balance on the permeated solution volume. Artificial neural networks-based quantitative structure–property relations models were developed for the organic passage ( P ), sorbed ( M ) and rejected ( R ) fractions using the most relevant set of molecular descriptors selected from a pool of 45 molecular descriptors by means of a correlation-based feature selection method and self-organizing maps (SOM). The analysis included pre-screening with principal components analysis and SOM of the chemical domain for the study chemicals as defined by chemical descriptors to identify the applicability domain and chemical similarities. The QSPR models predicted the P and M mass fractions within the range of the standard deviations of measurements for the experimental data set of fifty compounds. Mass balance closure (requiring that M , P and R sum to unity) was satisfactory for the experimental data set of fifty compounds and for an external set of 144 test chemicals, which were not included in the model development. Somewhat higher prediction errors were encountered for a few chemicals that were not well represented within the present chemical domain. The quality of the QSPR/NN models developed suggests that there is merit in extending both the present compound database and the present approach to develop a comprehensive tool for assessing organic solute behavior in RO water treatment processes.

The development of calcium sulfate dihydrate (gypsum) mineral scale in reverse osmosis (RO) membrane desalting was investigated by direct real-time observation of crystal growth. Gypsum scaling studies were conducted in a specially modified plate-and-frame reverse osmosis cell fitted with an optical window, with dark-field membrane lighting arrangement within the membrane cell to enhance crystal boundaries and allow recording of digital surface images magnified through an optical microscope. The evolution of the surface number density (SND) of gypsum crystals resembled a sigmoidal population growth process with an increasing rate of crystal formation at higher solution supersaturation (with respect to gypsum) at the membrane surface. The rate of formation of new crystals declined as the surface scale coverage increased, with surface scale progressing by diffusion-controlled growth of existing crystals as the gypsum crystal SND approached its plateau. The apparent crystallization induction time (i.e., observation time for the appearance of surface crystals) decreased with increased supersaturation, with single crystals growth reasonably characterized by a diffusion-growth model and a single crystal fluid mass transfer coefficient that was generalized by a Sherwood number type correlation. The present approach provides a convenient method of characterizing the kinetics of gypsum scale formation via quantification of single crystal mass transfer coefficient and evolution of the crystal SND. Such information can guide the development of comprehensive RO membrane scale formation models and characterization of mineral scaling propensities of RO membranes and source water where mineral salt scaling is of concern.

Concentration polarization in a rectangular reverse osmosis (RCRO) and a laboratory plate-and-frame RO (PAFRO) membrane channels were investigated using a finite-element numerical model consisting of the fully coupled governing equations for fluid dynamics and mass transfer. The model considers both the concentration dependence of model parameters and axial variation of permeate flux. In both the RCRO and PAFRO channel geometries the flux varies axially along the membrane commensurate with the coupled development of the concentration boundary layer. For the case of the RCRO channel, the effect of permeate flux was quantified via the computed feed-side mass transfer coefficient (commonly used in estimations of concentration polarization) in terms of a dimensionless correlation for the average Sherwood number, S h ¯ = A 1 ( 1 + A 2 R e A 3 S c A 4 ( H / L ) A 5 R e ¯ p ) [ R e S c ( H / L ) ] 1 / 3 , that reduces to the form of the classical Leveque solution in the limit of vanishing permeate flux. Simulations of desalting in a PAFRO channel demonstrated the importance of accounting for local permeate flux variability and coupling of the hydrodynamics and mass transfer model equations in order to predict local concentration variations and in particular local maxima of concentration polarization, especially in regions of flow recirculation where scaling by sparingly water soluble mineral salts could be first triggered should their concentrations exceed saturation.

A two-stage reverse osmosis (RO) process for high water recovery (up to 95%) desalination of Colorado River water was evaluated and demonstrated at the pilot scale. The two-stage process involved intermediate chemical demineralization (ICD) of the concentrate stream from a primary RO (PRO) process followed by secondary RO (SRO) desalting. Using alkaline-induced precipitation in a pilot-scale solids contact reactor (SCR), the membrane scaling propensity of the PRO desalting step was reduced, allowing for further recovery of product water via SRO of the demineralized PRO concentrate stream. When operating the SCR effluent above pH 10, the removal of Ca2+, Ba2+, Sr2+, and silica – the principal scale-forming precursors – were attained via ICD at levels upwards of 94%, 97%, 88%, and 67%, respectively. Despite significant process variations, 95% overall water recovery desalination was demonstrated to be feasible in the pilot scale via the PRO–ICD–SRO approach, with SRO desalting operated at specific permeate flux varying by ≤7% and at stable normalized salt passage of 2.9 ± 0.7% over extended periods of time. This study suggests that long-term implementation of the PRO–ICD–SRO process to achieve overall water recovery ≥95% would require robust on-line pH control strategy to attain consistent SCR effluent pH, along with appropriate selection and use of make-up antiscalants for the secondary RO desalting step, primarily to control membrane scaling by silicates and gypsum.

A novel RO ex situ scale observation detector (EXSOD) was developed for direct visual real-time monitoring of mineral surface scaling on RO membranes. This system consists of a plate-and-frame RO cell with an optical window capable of withstanding pressures in excess of 2.4 MPa. Membrane surface imaging under RO operation is accomplished by high resolution digital photography using an optical microscope with lighting arrangement to enhance the boundaries of semi-transparent crystals. The utility of the direct membrane surface observation cell was demonstrated in a series of experiments conducted with a commercial brackish water RO membrane and with gypsum selected as the model scalant. The present study demonstrated, for the first time, a capability for early detection of mineral salt scaling at very early stages of formation, in advance of any measurable flux decline. The approach also illustrates that surface crystal development on RO membranes can be quantified with respect to both single crystal growth and the evolution of crystal number surface density. For the growth of gypsum, crystal surface number density evolved according to a sigmoidal growth curve, with crystal growth diminishing as gypsum rosette needles began to overlap. Dissolution of surface crystals occurred once solution at the membrane surface was undersaturated with respect to gypsum. It is suggested that the EXSOD system can be used for on-line monitoring of scaling and fouling in RO plants, evaluation of the impact of colloidal and biofouling as well as the effectiveness of antiscalants and membrane cleaning strategies.

The integration of membrane RO desalting and accelerated precipitation softening (APS) was evaluated for achieving high product water recovery (>95%) from desalting of mildly brackish surface water. A systematic approach that included laboratory RO demonstrations, membrane scaling diagnostics, and mineral solubility analysis was undertaken to develop an effective physical/chemical APS strategy for reducing the concentration of scale-forming ions in the primary RO (PRO) concentrate. APS demineralization, as an inter-stage process, between the PRO and secondary RO (SRO) desalting, involved alkaline pH adjustment and calcite crystal seeding of the PRO concentrate, followed by microfiltration and pH reduction by acid dosing to avoid calcite scaling in the SRO stage. Colorado River water desalting case study demonstrated that, at 90% PRO recovery, the PRO concentrate stream contained barite, calcite, gypsum, and silica above their solubility limits by factors of 122, 47, 1.1, and 0.8, respectively. Further desalting was not feasible with traditional scale-control strategies (e.g., pH reduction and antiscalant addition). However, APS demineralization enabled significant concentration reduction of calcium (>90%), barium (>95%), and strontium (∼78%) and moderate reduction (10–20%) of magnesium and silica. APS kinetics were favorable even in the presence of antiscalant carryover from the PRO stage. The study demonstrated that high recovery desalination of up to 98% is feasible with the PRO-APS-SRO sequence, which reduces the limitations imposed by membrane scaling.

The utility of a laboratory diagnostic approach was investigated for a systematic evaluation of the impact of operating conditions, membrane type, and antiscalant effectiveness on membrane mineral salt scale formation. The study focused on low-pressure reverse osmosis (RO) membrane desalting of brackish water with high potential for scaling by calcium sulfate dihydrate (gypsum). Gypsum scale formation, in the surface crystallization regime, was studied in a plate-and-frame RO system via flux decline measurements and membrane surface imaging. Membrane scale coverage increased with increasing level of surface gypsum supersaturation. However, membrane scaling propensity, as quantified by both the surface scale coverage and flux decline, indicated that there were measurable differences among the four commercial membranes studied, particularly in response to antiscalant treatment. Axial development of scale coverage indicated higher scale coverage as the membrane channel exit was approached, consistent with increased level of gypsum supersaturation along the membrane. As a result, antiscalant effectiveness, at a given dosage, was also location dependent. The present study illustrates a practical approach of combining flux decline and surface scale coverage measurements as a means of quantifying the impact of scale formation and its dependence on operating conditions and membrane type. Clearly, further studies are needed to elucidate the impact of membrane surface roughness and charge as well as module geometry on surface mineral salt scale formation and its mitigation.

A numerical model of non-isothermal pervaporation was developed to investigate the development of the velocity, concentration and temperature fields in rectangular membrane module geometry. The model consists of the coupled Navier-Stokes equations to describe the flow field, the energy equation for the temperature field, and the species convection-diffusion equations for the concentration fields of the solution species. The coupled nonlinear transport equations were solved simultaneously for the velocity, temperature and concentration fields via a finite element approach. Simulation test cases for trichloroethylene/water, ethanol/water and iso-propanol/water pervaporation, under laminar flow conditions, revealed temperature drop axially along the module and orthogonal to the membrane surface. The nonlinear character of the concentration and temperature boundary-layers are most significant near the membrane surface. Estimation of the mass transfer coefficient assuming isothermal assumption conditions can significantly deviate from the non-isothermal predictions. For laminar conditions, predictions of the feed-side mass transfer coefficient converged to predictions from the classical Leveque solution as the feed temperature approached the permeate temperature.

The axial development of calcium sulfate dihydrate (gypsum) scaling on selected reverse osmosis (RO) membrane surfaces was investigated experimentally in a plate-and-frame RO system. Scaling experiments with model solutions demonstrated progressive axial development of surface gypsum crystals along the membrane surface. The impact of surface crystallization was characterized via flux decline measurements and by both optical and high-resolution scanning electron microscopy (SEM) surface imaging. Surface coverage by gypsum scale (percent area covered and gypsum surface mass density) increased along the membranes (towards the membrane exit), consistent with the increase in concentration polarization. At the membrane channel entry, surface gypsum crystals were at their initial growth stage formed as primarily needle-like structures. With increasing axial distance from the entry, surface crystallization resulted in crystal structures that transitioned from the needle and plate-like structures, in the submicron size range, to partial rosettes to complete rosette structures in the mm size range. Significant differences were observed in the extent of surface scale coverage and surface crystal size among the membrane studied. Antiscalant addition to the feed solution led to progressive decline in the percent of area covered by scale with increasing antiscalant dosage and corresponding decrease in flux decline. Antiscalant addition resulted in perturbed surface crystal morphology of flattened and/or fused rosette crystal arms and to complete elimination of surface crystals at a sufficiently high antiscalant dosage. The present study demonstrates that gypsum scale development is affected by the formation of surface crystals on the membrane surface, thereby suggesting that there is merit to expanded research on the direct impact of surface topology and chemistry on surface crystallization of mineral salts.

Pervaporation removal of methyl tert-butyl ether (MTBE) from dilute aqueous solutions was demonstrated using ceramic-supported polymer (CSP) membranes. These membranes were created by the free radical graft polymerization of vinyl acetate onto a vinyl trimethoxysilane-activated porous silica substrate having a native average pore size of 500 A. The resulting membranes consisted of poly(vinyl acetate) chains terminally and covalently anchored to the membrane surface. These pervaporation membranes performed with separation factors ranging from about 68 to 577 at total permeate fluxes of 0.31–0.70 kg/m2 h over feed concentrations of 74–7400 ppm (0.01–1%, v/v). MTBE flux was controlled by the feed-side concentration boundary layer, while water permeation was controlled by the membrane resistance. The primary effect of the grafted PVAc in these CSP membranes was to increase separation by rejecting water, as opposed to increasing the flux of MTBE.

The topology of porous silica membranes (500 A native pore size) that were surface modified by silylation with vinyl trimethoxysilane (VTMS) and graft polymerization with poly(vinyl acetate) (PVAc) was investigated by atomic force microscopy (AFM). The polymer-modified membrane surfaces were populated by terminally anchored polymer chains with the depth of modification of about 200 nm as assessed from scanning electron microscopy (SEM) images. Surface roughness increased upon silylation but decreased as the result of surface graft polymerization. AFM imaging of surrogate silicon wafers, silylated or grafted in a manner similar to the membranes, revealed that silylation increased the root-mean-square (RMS) surface roughness by about 1.0 nm, while poly(vinyl acetate) modification reduced the RMS surface roughness by approximately 0.4 nm. The distributions of surface heights for the modified surfaces were asymmetric with the tails of the distribution representing the sparse larger features. Surface roughness analysis revealed that the tails make a significant contribution to the measured RMS surface roughness. Although the RMS surface roughness values of the polymer-modified surfaces were essentially similar (0.81–0.85 nm), skewness of the distribution decreased while the width of the Gaussian portion of the distribution increased with decreasing initial monomer concentration.

Asymmetric tubular alumina-supported poly(vinyl acetate) (PVAc) and poly(vinyl pyrrolidone) (PVP) membranes were created and characterized by pervaporation separation of binary mixtures of methanol and methyl tert-butyl ether (MTBE). The active separation layer was created by free-radical graft polymerization of PVAc and PVP onto a vinylsilane-modified alumina substrate with an average native pore diameter of 50 A. The separation layer consisted of a surface phase of terminally anchored polymer chains with estimated radius of gyration about a factor of 4.5–6.8 larger than the membrane pore radius. The methanol separation factors for the PVP and PVAc-grafted pervaporation membranes reached values of 26 and 100, respectively, at the lower range of methanol concentrations (1–5% v/v) tested in the study. The separation impact of the grafted polymer chains was apparent given that the native (unmodified) and vinylsilylated alumina membranes lacked selectivity for the MTBE/methanol system. Total permeate flux attained with the PVAc and PVP-based membranes ranged from 0.055 to 1.26 and 0.55 to 6.19 kg/m2 h, respectively, over the respective feed methanol concentration ranges of 1–90 and 5–90% (v/v). A tradeoff between separation and permeate flux was apparent from the decrease in permeation flux with increasing separation factor.

Fouling-resistant ultrafiltration ceramic-supported polymer (CSP) ultrafiltration membrane was developed for the treatment of oil-in-water (o/w) microemulsions. The CSP zirconia-based membrane was prepared via free-radical graft polymerization of vinylpyrrolidone onto the membrane surface. Pore reduction of about 25–28% was encountered upon modification as revealed by hydraulic permeability measurements. Oil rejection was higher for grafted membranes with smaller resultant pore size. Atomic force microscopy imaging of polymer-modified surfaces showed complete surface coverage by the polymer chains. The native zirconia membrane was irreversibly fouled after treating the o/w microemulsion for a brief period, while the CSP membranes maintained its pre-filtration hydraulic permeability even after many filtration runs. The present surface modification was found to be effective in preventing irreversible membrane fouling despite significant membrane surface roughness. Relative to the native membrane, oil rejection of the CSP membrane increased more than two-fold for the range of studied oil droplet size (18–66 nm), while surfactant rejection remained low for both the native and modified membranes.

The hydrodynamic response of a graft-polymerized membrane was demonstrated for a microporous silica-poly(vinylpyrrolidone) (silica-PVP) membrane. The membrane pores were modified by graft polymerizing vinyl pyrrolidone onto the membrane pore surface, resulting in a polymer surface layer of covalently tethered polymer chains. The hydraulic permeability of the modified membrane increased with increasing transmembrane pressure owing to flow-induced deformation of the grafted polymer chains. The dynamics of the modified pores was investigated by membrane hydraulic permeability studies along with a two-region hydrodynamic pore flow model. The thickness of the grafted polymer layer decreased with increasing pore-wall shear rate by up to about 47%, relative to the thickness at the zero shear limit, depending on the surface density and length of the grafted chains. Although the effective pore size of the polymer-grafted membrane was reduced by 5–36% (at the zero shear rate limit), about 18–59% of the pore size loss was regained at high pore-wall shear rates. Increasing the degree of shear-induced permeability change is feasible by increasing the ratio of the polymer chain length/pore size ratio as well as the surface density of the grafted polymer phase. The present results suggest that hydrodynamic pore size control could provide an additional useful degree of freedom in operating polymer-modified filtration membranes.

A novel asymmetric ceramic-supported polymer (CSP) pervaporation membrane was developed using free-radical graft polymerization of polyvinyl acetate (PVAc) onto a porous tubular silica substrate. The resulting membrane was characterized by pervaporation removal of trichloroethylene (TCE) and chloroform from dilute aqueous solutions. PVAc was chosen since it has a high affinity for TCE and chloroform and a low affinity for water, thus making the membrane permselective toward these solutes. This study has shown, for the first time, that pervaporation is possible even with a large substrate starting pore size (∼500 A) and where the active separation phase is a macromolecular layer of terminally anchored chains. Performance of the CSP membrane was assessed at various hydrodynamic conditions and feed concentrations. Resistance of the PVAc/CSP membrane was found to be negligible compared with the concentration boundary layer resistance. The enrichment factor, defined as the ratio of total solute concentration in the permeate to that in the feed, increased with polymer graft yield. For a given graft yield, the enrichment factor approached an asymptotic plateau value, as the tube-side Reynolds number was increased due to increased water flux in response to the rise in transmembrane pressure. Preliminary analysis suggests that the CSP pervaporation membrane performance could be substantially increased by optimizing reaction conditions to produce a higher polymer graft density.

A hybrid ceramic-polymeric membrane was developed by the growth of covalently-bonded polyvinylpyrrolidone (PVP) chains from the surface of a porous silica support membrane via a graft polymerization process. These ceramic-supported polymeric (CSP) membranes can be produced with variable surface density and length of the terminally anchored polymer chains. Hydraulic permeability measurements demonstrated the effect of the grafted chain density and length on the water permeability of the CSP membranes. The potential application of these CSP membranes for the treatment of oil-in-water emulsions, was evaluated in cross-flow filtration experiments. This preliminary evaluation indicated improved performance of a CSP membrane, relative to an unmodified support, since the CSP membrane produced a lower permeate concentration at an equivalent permeability. The improved selectivity and reduced fouling tendency is attributed to the increased hydrophilicity of the membrane surface provided by the grafted PVP chains.

The surface of 4100 A-pore-size silica membranes was modified with a covalently bonded polyvinyl-pyrrolidone brush layer by a novel graft polymerization process. Hydraulic permeability measurements performed with six different solvents and both unmodified and modified membranes suggest that the permeability of the modified membrane is determined by the configuration of the terminally anchored polymer chains. In the modified ceramic-supported polymeric (CSP) membrane, the swelling of the polymer brush layer increases as the solvent power increases, resulting in a decrease in the pore radius and subsequently, the permeability.

(a) The range of membrane permeabilities listed in the second paragraph of Section 1 (line 2): “. . .range of Lp =0.5–0.8×10−10 m3/m2 sPa. . .” should be replaced by “. . .range of Lp ≈0.39–2.2×10−11 m3/m2 sPa. . .” (b) The range of brackish water salinity values listed in the second paragraph of Section 1 (line 4) should be corrected as below: “. . .salinity in the range of 1000–2000mg/L. . .” should be replaced by “. . .high salinity in the range of ∼7500–15,000mg/L. . .” (c) The ranges of values for the membrane permeability, price and life time in the second paragraph in Section 4 (lines 5–7): “. . . (i.e., Lp =0.5–0.8×10−10 m3/m2 sPa) is taken to be in the range of 20–40$/m2 [10] (thusmA =0.76–1.52×10−3 $/m2 h assuming membrane life of 3 years). . .” should be replaced by “. . . (i.e., Lp ≈0.39–2.2×10−11 m3/m2 sPa) is taken to be in the range of 10–20$/m2 [10] (thusmA =0.63–1.3×10−7 $/m2 s, assuming membrane life of 5 years). . .” (d) In Section 4, third paragraph, line 8, the segment: “. . .with the Dow FilmTec SW30XLE-400i, Lp =0.78×10−10 m3/m2 sPa. . .” should be replaced by “. . .with the Dow FilmTec SW30XLE-400i, Lp =0.39×10−11 m3/m2 sPa. . .” (e) The value of RMEC in the third paragraph of section 4, line 7: “As an example, for RMEC =0.01. . .” should be replaced by “As an example, for RMEC =0.04. . .” (f) In Section 4, third paragraph, line 9, the text: “. . ., the ratio of the specific membrane cost (SMCtr) to the specific energy consumption (SECtr) is in the. . .” should be replaced by “. . ., the ratio of the specificmembrane cost (SMCtr) to the specific energy consumption (SECtr), forwater recovery rangeof∼30–50%, is in the. . .” (g) In Section 4, fourth paragraph, lines 1 and 2, the value of brackish water TDS and MER: “For desalination of mildly brackish water of ∼1000mg/L TDS, RMEC =1 and the estimated MER decreases from 6 to 0.04. . .” should be replaced by “For desalination of brackish water of . . .3500mg/L TDS, RMEC =1 and the estimated MER decreases from 6 to 0.07. . .”