What Will it Take to Get to 250,000 ppm Brine Concentration via Ultra-high Pressure Reverse Osmosis? And is it Worth it?

Arezou Anvari¹,ᵃ, Jishan Wu¹,ᵃ, Arian Edalat², Nikolay Voutchkov³, Ahmed Al-Ahmoudi³, Subir Bhattacharjee⁴, Eric M.V. Hoek¹,*

Abstract

The possibility of ultra-high pressure reverse osmosis (UHPRO) to replace thermal desalination promises reductions up to 50% in both energy consumption and capital cost for brine concentration.However, commercially available reverse osmosis (RO) membranes suffer significant  performance degradation under ultra-high pressure due to severe membrane compaction and embossing. High temperature operations such as in Middle East seawater desalination and brine mining further exacerbate both issues due to softening of the polysulfone support membrane. Assuming that pressure and temperature-related compaction issues can be resolved through innovations in membrane materials and module design, this study asks the questions, “What will  it take to get to 250,000 ppm via ultra-high pressure reverse osmosis? And is it worth it?” This work provides a comprehensive examination of the RO process at high pressure—specifically SWRO to HPRO to UHPRO. We elucidate the influence of thermophysical properties, hydraulic pressure, permeate flux, water permeability, feed spacer design, concentration polarization and salt  rejection and stage design on recovery and specific energy consumption (SEC). Our analysis of high rejection (HR) and high flux (HF) RO membranes reveals a dual-edged nature of HF membranes: while they boast superior permeability, enabling reduced trans-membrane hydraulic pressure, they require higher system level water flux, which induces significant concentration  polarization (CP). This in turn leads to CP-limited product water recovery, which is a novel finding. In addition, membrane compaction drives up the applied pressure and SEC, and also the capital cost of UHPRO systems. We further assess changes in feed spacer designs that could dramatically reduce CP and enhance product water recovery. This paper presents a holistic roadmap of how system design drives these parameters. The findings herein are poised to guide both technological refinements and best practices for RO membrane brine concentration, propelling the water treatment industry towards a more sustainable future.

1 Introduction

Over the last six decades, the emergence and evolution of reverse osmosis (RO) membrane technology has revolutionized water treatment, particularly desalination [1, 2]. Originally a conceptual framework confined to laboratories, RO technology has burgeoned into an incredibly versatile water treatment technique and its range of applications now extends to seawater and brackish groundwater desalination, various ultra-pure water and process separations in addition to industrial and municipal wastewater recycling [3]. Recent advancements in RO membrane materials [4] and process technologies [5] have further enhanced its capabilities. These improvements have driven the efficiency of seawater desalination to within 10 to 20% of the thermodynamic performance limits. Similarly, for brackish and low-salinity water treatment, these enhancements have achieved efficiency levels within 30 to 40% of the theoretical limits [6]. Despite these significant improvements in RO membrane materials and process technologies over several decades, the pursuit of innovations in brine concentration continues. Alternative separation concepts like osmotically-assisted reverse osmosis (OARO), membrane distillation (MD), forward osmosis (FO), capacitive de-ionization (CDI), microbial fuel cell-assisted electro-dialysis (MFCED), and humidification-dehumidification (HDH) are actively being pursued for industrial development [1]. These techniques retain considerable appeal due to their potential to deliver superior water quality and enhanced recovery rates for hyper-saline brines. They also demonstrate the possibility to desalinate waters that challenge the capabilities of RO, such as high-temperature waters and hyper-saline brines.

One major constraint of modern RO membrane technology is the limited availability of high-pressure RO membrane modules and pressure vessels. Typical seawater RO products are not designed to exceed 80 bar (1,200 psi) of operating pressure [7]. Notwithstanding potential recovery limitations imposed by mineral scaling, this pressure limitation prevents one from achieving economically viable product water recovery for feed waters with higher salinity than seawater. If an RO membrane product were developed to withstand continuous operating pressures needed to concentrate brine high enough that it is suitable for crystallization (>250,000 mg/L). Additionally, with a high-pressure tolerant RO membrane one could realize an FO/RO integrated process capable of seawater desalination at high recovery using simple salt-based draw solutions, i.e., no heat required to recover thermo-reversible polymers [8] or thermolytic salts [9]. Further, one could employ RO to further concentrate and valorize SWRO brine, which appears of great interest in the Middle East [10] to concentrate other hyper-saline brines such as oil and gas produced waters from hydraulic fracturing and brines targeted for direct lithium extraction [11].

Given the above review and simple analysis, one might ask the question, “Why cannot commercially available RO membrane technology withstand higher operating pressures and temperatures?” Most modern RO membranes are thin film composites (TFC). These membranes are constructed from polyamide (PA) coating film chemistries [12] polymerized in situ onto polysulfone (PSF) porous supports previously phase inverted onto polyester (PET) non-woven fabrics. Over 50% of performance (permeate production) loss of the composite RO membranes have been reported at high applied pressures [13]. This significant performance degradation is primarily ascribed to the compaction or densification of the mesoporous PSF support layer, a finding corroborated by several studies [13-15]. Pendergast et al. [14] examined hand-cast composite RO membrane compaction at pressures reaching 35 bars, documenting a reduction in the thickness of the PSF support layer between 20% to 30%, accompanied by up to 50% loss of water permeability. Davenport et al. [15] reported a 35% decrease in permeability and a 60% reduction in cross-sectional thickness of commercial seawater RO membranes when operated at 150 bars. In a more recent study, Wu et al. [13] explored commercial RO membrane compaction at pressures up to 207 bars, unveiling a thickness reduction in the PSF support layer between 38% 98 to 60%, resulting in over 50% water permeability loss [13]. The densification and collapse of skin layer pores, situated directly beneath the PA coating film, extends the effective diffusion path length through the composite PA-PSF membrane structure, thereby diminishing the permeability of both salt and water [13].

Having identified a significant performance gap in RO membrane technology, it is clear why interest in traditional thermal desalination processes persists, while interest in new alternative desalination processes and technologies continues to grow. This study evaluates SWRO, HPRO and UHPRO processes. The key areas we examine are:

1. The effect of the number of elements in series on the overall desalination performance.

2. The role of spacer designs, particularly focusing on porosity and thickness, in influencing the efficiency of desalination processes. membranes in desalination.

3. The comparative performance of high-rejection (HR) membranes versus high-flux (HF) long-term performance.

4. The potential advantages and drawbacks of compaction-resistant membranes in sustaining .

5. The energy consumption analysis on SWRO, HPRO and UHPRO processes at different recoveries and operating pressures.

2 Model development

The reverse osmosis (RO) membrane system is modeled by interpreting each spiral-wound element as a series of thin, parallel rectangular flow channels, with a feed spacer of a defined thickness and porosity. The curvature of the element is disregarded, as it was determined that assuming a rectangular geometry rather than a bent envelope doesn't notably affect the depiction of flow behavior within the feed channel [16]. Each flow channel is discretized into small incremental unit volumes. The local influent, effluent, residual pressures, flows, and concentrations are calculated from a forward difference solution to the differential mass and momentum balances presented. A local mass transfer coefficient, influenced by the local retentate flow rate, is calculated to determine the effects of solute concentration polarization on bulk, membrane, and permeate concentrations. Locally determined mass and momentum transport parameters are then integrated across the entire system to determine global system performance.

The schematic of the model is illustrated in Fig.1, where NF pretreated feed with 32 g/L NaCl is pumped into a series of SWRO elements with a recovery rate of 50%. The retentate from the first SWRO stage is then pumped into a second stage HPRO system, also at 50% recovery. Finally, the retentate is concentrated up to 250 g/L in the last UHPRO stage.

To construct the model, several key assumptions are made. The feed stream chemistry is considered to be pure NaCl brine, with the thermo-physical properties strictly applicable to the feed stream. The simulations do not consider any deleterious effects of fouling, scaling, membrane compaction, or chemical cleaning. Baseline water permeance (A) and solute permeability (B) values are assumed to be 1 LMH/bar and 0.06 LMH for seawater RO (SWRO), 0.8 LMH/bar and 0.08 LMH for high-pressure RO (HPRO), and 0.6 LMH/bar and 0.1 LMH for UHPRO, respectively. The lower A value for HPRO and UHPRO are attributed to membrane compaction due to the higher applied pressures. In other simulations, the A and B values of SWRO are simulated within HPRO and UHPRO systems. In scenarios where high-flux SWRO membranes are used, the A and B values are set at 2.4 LMH/bar and 0.0912 LMH, respectively.

 Fig. 1. Schematic of the full-scale model setup. NF pretreated feed with 32 g/L NaCl pumped into a series of SWRO elements with 3057.56 gpm flow rate at 50% recovery. The retentate of the first stage of SWRO is pumped into the second stage HPRO system with 50% recovery. And then with final UHPRO stage, the retentate is concentrated up to 250 g/L.

2.1 Governing equations

In the membrane system, the primary driver for water permeation is the difference between the feed side hydraulic pressure and the trans-membrane osmotic pressure. For solute permeation, the concentration gradient across the membrane is the dominant force. Based on the solution diffusion model [17],water flux, Jw (=vp), the local permeate velocity) and solute flux ,Js (=vpcp ),are described by

where A is the intrinsic water permeability, B is the intrinsic solute permeability, ∆p is the available hydraulic pressure, ∆π is the trans-membrane osmotic pressure. In addition, ∆cm(=cm-cp) is the trans-membrane concentration difference, cm and cp are solute concentrations at the membrane-solution interface at the feed and permeate sides, and rs (=1-cp/cm) is the intrinsic local salt rejection.

According to Mulder [18], one can re-arrange eq (2) to obtain cp=cmB/(vp+B), which indicates that local permeate concentration varies with solute concentration, solute permeability, and permeate velocity. The expression for permeate concentration can be additionally re-arranged to produce a useful expression for the local (real) rejection by an RO membrane,

The rejection is seen to depend on permeate velocity. The system's local separation performance is shaped by variables like hydraulic pressure, retentate crossflow velocity, and concentration. These variables are further affected by transport within the spiral wound element and membrane fouling phenomena. In our model, while water and solute permeability serve as fitting coefficients and may change over time due to membrane compaction [13], we assume that water permeability, solute permeability, and solution osmotic coefficient are held constant throughout the system. Factors such as solute rejection, permeate velocity, osmotic and hydraulic pressures, solute concentration, and retentate velocity differ from inlet to outlet, as detailed below.

The applied hydraulic pressure drives water permeation through RO membranes. Yet, as retentate flows over the membrane and feed spacer surfaces, hydraulic pressure losses occur from the system's inlet to its outlet [19, 20]. In our model, the Poiseuille equation is combined with the empirical friction factor of Schock and Miquel [16] to describe the available hydraulic pressure at each axial location within the feed side crossflow channel as

Here ux is the crossflow velocity through the spiral wound elements, fsp(= 6.23Re-⁰.³) is an empirical friction factor accounting for the presence of a feed spacer, p is the solution density, dH is the hydraulic diameter, and Re is the local Reynolds number.

We assume the solution's density and viscosity in each stage to be independent of temperature and pressure assuming the process to be isothermal and largely isobaric. Concentration dependence of these properties is considered (see Section 2.6). However, retentate velocity, Reynolds number, and spacer friction factor vary throughout the system. We also allow for local variations in hydraulic diameter to account for changes in channel cross-section due to fouling. Factors such as total feed flow rate, the number of parallel pressure vessels (modules), leaf count per element, leaf and spacer dimensions (including thickness and porosity), play a crucial role in describing the module's hydraulic pressure drop [16].

The local crossflow velocity is determined from a differential mass balance on water flowing through the spiral wound elements as

where u0 is the inlet flow velocity (at x = 0). In eq (5), inlet velocity for each stage is determined by dividing the influent volumetric flow rate by NMNLεsp WH*,where NM is the number of modules (pressure vessels containing spiral wound elements) into which the feed water is pumped in parallel, NL is the number of membrane leafs (crossflow channels) per spiral wound element, εsp is the effective porosity of the flow channel created by the feed spacer, W is leaf width, and H* is the effective local channel height. The local permeate volumetric flow rate, Qp, is  2.vp NM NL W. dx where dx is the length of membrane considered. The factor “2” accounts for water permeating through both top and bottom sides of a membrane leaf.