Research Update: Liquid gated membrane filtration performance with inorganic particle suspensions

Membrane technology is ubiquitous and widely used in water treatment, chemical and petrochemical industries, food and beverage processing, biopharma and bioprocessing, and in textile, paper, and pulp industries.¹ Membranes are utilized not only for making industrial processes more economical but also to achieve high levels of purity, selectivity, and throughput.² While membrane-based processing offers many benefits, it also suffers from fouling that leads to decline in the permeate flux and impacts solute rejection,^3,4 and ultimately increasing energy demands.

For water and water-related processes, most of the energy used is in the form of electricity and is estimated to be ∼13% of U.S. electricity generation (3.6 trillion kWh/yr or 12.3 Quads).⁵ In addition to energy, water processing-related CO₂ emissions are ∼290 × 10⁶ metric tons (approximately 5% of all U.S. CO₂ emissions) and equivalent to greenhouse gas (GHG) emissions of estimated 53 × 10⁶ cars. Out of this incredibly high number, municipal wastewater treatment is responsible for ∼18% of the total water-related carbon emissions.⁶ Electric Power Research Institute (EPRI) projects in excess of 20% increase in electricity demand just for wastewater treatment alone from recent estimates, reaching 90 × 10⁹ kWh/yr (0.3 Quads) by 2050.⁷ Hence, there is a need for innovative energy-efficient water treatment processes and technologies to reduce water-related energy usage and its environmental impact.

Recent advances in horizontal drilling and hydraulic fracturing technologies have led to a boom in domestic shale oil and natural gas wells. The growing prevalence of fracturing wells has created stress on surface and ground water supplies, with upwards of 6 × 10⁶ gallons of water used per well for drilling and fracturing.⁸ Although several options for produced water management and treatment exist, including thermal distillation, reverse osmosis membrane systems, chemical precipitation, and electro-coagulation, filtration to remove suspended solids and reuse of the fluid is the preferred practice, representing a prime opportunity for energy efficient water treatment technologies.

In the field of water treatment, fouling, due to deposition and adhesion of organic and inorganic particles and colloids onto membrane surfaces and internal pore structures, is the greatest challenge associated with the operation of microfiltration (MF) and ultrafiltration (UF) membrane processes.⁴ As a differential pressure-dependent process, bulk flow convection across a semi-permeable membrane can require significant energy input to maintain desired levels of throughput. Over time, fouling of MF/UF membranes during operation increases hydraulic resistance to flow, reducing filtration flux and inducing detrimental effects on the efficiency and economics of processing.⁹ 

Researchers have proposed and explored several membrane modification methods to reduce and minimize fouling.^3,10 These methods and techniques often utilize surfactants, coatings, chemical grafting, polymer blends, and plasma treatments and have been comprehensively reviewed by Miller et al.¹¹ Recently, Hou et al.¹² introduced a novel mechanism of liquid trapped inside porous membranes to act as a reconfigurable gate that can tailor the transmembrane pressure (TMP),¹³ allow phase separation, and minimize fouling buildup suitable for particle filtration (PF), MF, and UF processes. This mechanism, referred to as liquid gated membranes (LGMs), addresses some of the limitations of existing membrane surface modification technologies for minimizing fouling and increasing the efficiency of membrane processes.¹¹ The LGM mechanism allows capillary-stabilized trapping of an immiscible liquid within the pores of a conventional membrane providing a slippery, non-adhesive lining on the pore surfaces. In this configuration, the interfacial tension between the feed and the membrane surface can be minimized, directly reducing the pressure required to promote transport and operate at high throughput levels suitable for filtration applications. This property can also be tuned to enable separations of complex mixtures of immiscible phases and is compatible across different filter media. Furthermore, the presence of this gating liquid acts as a barrier against adsorption of foulants to the membrane by limiting molecular interactions of the foulants with the membrane surface.

Recently, Bazyar and Javadpour et al.¹⁴ investigated the LGM mechanism by performing gas-liquid porometry using different gating liquids on polyvinylidene difluoride (PVDF) MF membranes and demonstrated the importance of interfacial tension and viscosity of the gating liquid for LGMs. They also evaluated the thickness of the gating liquid lining the pores when the pores are in an open state and re-infusion times of different gating liquids. In order to further understand the behavior of gating liquids during transport of other immiscible fluids, Bazyar and Lv et al.¹⁵ also conducted liquid-liquid displacement porometry on LGMs. Their investigations confirmed the presence of gating liquid along the open pores in the form of a liquid lining^12,14 in addition to the presence of gating liquid in the pores that are closed during transport. Under the tested conditions, their results showed that in comparison to a solvent pre-wetted membrane having all of its pores open, nearly 57% of LGM pores were open for transport. Sheng et al.¹⁶ applied the LGM mechanism to elastomeric sheets (silicone rubber and polydimethylsiloxane) with through holes that were deformed under mechanical stretching and introduced an approach that may benefit larger pore size MF and PF membrane technology applications. In terms of LGM performance in the presence of potential fouling, Overton et al.¹⁷ recently demonstrated up to 60% passive flux recovery of LGMs in the presence of whey protein fouling for polytetrafluoroethylene (PTFE) MF membranes. Potential benefits of additional liquid-liquid and liquid-gas interface-based interactions have been explored in diverse application areas, including a convenient method to control microscale fluid flow, anti-thrombosis medical devices, and thermoresponsive mobile interfaces to name a few.^13,18–21

In this research update, emphasis has been placed on the filtration performance of LGM technology for separation of inorganic microparticles from an aqueous feed. A hydrophilic bentonite suspension was selected to mimic filtration of produced wastewater relevant for drilling applications in the oil and gas industry.^22,23 We compare baseline (BL) performance between conventional polymeric MF membranes and LGMs, as well as cleaning efficiency with repeated cycling. Results show that despite eventual fouling, LGMs operate with lower TMP, higher throughput, and longer time to foul compared to conventional membranes. Longer time to foul is due to reduced irreversible fouling, which is measured as pressure recovery and mass accumulation, after backwashing.

Unlaminated PTFE membrane filters with a nominal 5 µm pore size and 150-250 µm thickness (Sterlitech, PTU504750) were punched into 25 mm diameter discs (Figs. S1 and S2). For each sample, two baseline measurements of de-ionized (DI) water flow were recorded to evaluate initial transmembrane pressures of the pristine membranes. After each baseline run, the membranes were dried in a 70 °C oven for 1 h prior to subsequent use. The pristine membranes were converted into liquid gated membranes by infiltration with perfluoropolyether (Dupont Krytox 103) at a dose of approximately 13 μl/cm² (corresponding to the expected membrane void volume), as previously demonstrated.¹² Interfacial energy measurements were performed to confirm stable wetting configurations (see the supplementary material) for the chosen feed. Subsequently, the infiltrated membranes were placed under vacuum (635 mm Hg) for 1 h to promote infusion of the perfluoropolyether throughout the membrane pores. The feed flow rate was maintained at 2 ml/min, corresponding to a flux of 245 l/m² hr (LMH), with a peristaltic pump (Cole-Parmer Masterflex L/S with Model 77800-60 head). In the presence of nanoclay based feed, once TMP had increased to 120 kPa, backwashing was initiated at 2 ml/min for 10 min (Fig. S3).

For all experiments, pressure was measured with a digital pressure transmitter (Dwyer #628CR-10-GH-P1-E1-S1) interfaced with an external multichannel Data Acquisition (DAQ) card (National Instruments NI-9207). Data were collected via a custom LabView program. All reported data have an N = 3 or N = 5 sampling size and error bars are ±1 standard deviation.

For inorganic fouling studies, a bentonite suspension was chosen to simulate particulate matter in feed waters to represent a type of process water commonly treated with MF membranes. The suspension was composed of 500 mg/l hydrophilic nanoclay (Sigma Aldrich, CAS# 1302-78-9, Fig. S4) in de-ionized water. The feed suspensions were stirred continuously and monitored throughout the experiments to ensure that no settling occurred.

The filtration efficiency of nanoclay particles was measured using Beckman Coulter Multisizer 3. A size distribution and corresponding number of average particles were measured for the feed solution and compared to the particle concentration in the permeate solutions. For these membranes, the total number of particles larger than 5 μm in the permeate volume was compared to the initial concentration present in the feed to calculate the filtration efficiency of the membranes for this type of feed. Optical microscopy and Zeiss SEM were used for imaging the nanoclay particles.

Thermogravimetric analysis (TGA) was performed in air on non-fouled and fouled membranes with a TA Instruments Discovery TGA. Temperature was ramped at 20 °C/min up to 300 °C with a 60 min isothermal stabilization to determine the gating liquid content, followed by a 20 °C/min ramp up to 650 °C to determine the remaining inorganic content.

Pressure data were utilized for determining the mode of fouling and onset of fouling based on constituent fouling models as presented earlier.^24–26 Goodness of fit (GOF) analysis based on least square regression was conducted to determine the modeling fits. For the sake of simplicity, R² results have been tabulated for the constituent fouling models (standard blocking, cake formation, intermediate blocking, and complete blocking). A color map indicating the degree of GOF was used for illustration purposes.

Transmembrane pressure (TMP) was measured during constant flux dead-end filtration of an inorganic particle suspension. As previously observed with flow of a pure liquid through the membrane, the initial breakthrough pressure for the LGM, as indicated by the plateauing of the pressure curve at t ≈ 2 min coinciding with permeate flow, was approximately 16% lower than what was necessary to achieve permeate flow for the untreated membrane (Fig. 1). Flow was terminated once a maximum pressure of 120 kPa was reached. Under these conditions, the LGM sustained approximately a threefold longer filtration period before a backwashing procedure was evoked. Similar to the results presented earlier,^12,17 the presence of the conformal gating liquid coating on the membrane provides a low adhesion mobile interface at the membrane surface and appears to significantly delay the onset of fouling and reduces the fouling buildup rate as evident from Fig. 1. Average TMP for baseline (BL) runs with DI water without any foulant (not marked on Fig. 1) for the LGMs and the control membranes were at 29 ± 4 kPa and 35 ± 4 kPa, respectively.

Membrane fouling increases the pressure required to generate the desired volume of product water, requiring expensive chemical cleaning for fouling removal²⁷ that contributes to additional capital expenditure and operational expenses. Particulate fouling increases the resistance to flow across the membrane commonly causing pore blocking and eventually leading to cake layer formation over the course of a filtration run that can typically be partially removed by periodic backwashing.

Backwashing is used to maintain high flux in membrane systems between chemical cleanings and is used to minimize the frequency of resource-intensive chemical treatments, which generate waste, deteriorate the structural integrity of membranes, and often require significant downtime and infrastructure investment.²⁸ Backwashing is accomplished by flowing permeate water through the membrane from the permeate side to the feed side to remove foulants from the membrane surface.

Following nanoclay fouling to the extent where backwashing is required (after reaching 120 kPa in these experiments), we have compared the propensity for fouling removal of LGMs and control membranes (Fig. 2). To evaluate the degree of reversible fouling, the difference in pressure required for liquid (de-ionized water) flow before and after a single 20 ml backwashing cycle was compared. The results show LGM returning nearly to the initial BL TMP post-backwashing, suggesting highly reversible fouling behavior, whereas the non-gated control membrane demonstrates very little improvement despite backwashing, indicating the presence of irreversible fouling.^3,10,11,28 Initial BL data for the non-gated control membranes are marked in Fig. 2 for comparison.

We expected, therefore, that LGMs would show potential in reducing the frequency and duration of backwashing on fouled membranes. The latter will provide multiple advantages to the filtration cycle, including to minimize the use of cleaning chemicals and to reduce pumping of backwash water, thus decreasing the loss of product water. To demonstrate this property, five back-to-back filtration and backwashing cycles were conducted. Figure 3(a) shows the overall increase in the throughput of LGM membranes compared to the control membranes, illustrating threefold higher throughput. This is attributed to the reduced irreversible fouling buildup observed in LGMs.

Figure 3(b) shows the time required for the TMP buildup to reach 120 kPa for both the non-gated control and the LGM samples over multiple backwashing cycles using 20 ml of backwash water per cycle. Although LGMs maintain a longer total operational cycle time (4100s vs 1500s) throughout several experimental cycles, the duration of the individual filtration period becomes shorter for both membranes as more feed volume is processed (see also Fig. S5), suggesting that backwashing alone under these process parameters is not sufficient for removal of all the accumulated inorganic fouling alluding to irreversible fouling buildup.

Figure 3(c) demonstrates the potential of LGMs to reduce the amount of backwashing by comparing normalized post-wash TMPs for control and LGM samples after 10 ml and 20 ml washes per cycle. LGM TMPs with 10 ml backwashing are lower than the control membrane TMPs with 20 ml backwashing for all 5 cycles. Different modes of fouling and/or reduced irreversible fouling can be attributed to the improved LGM performance with potential to further minimize backwashing water consumption.

Visual inspection and microscopy of the membrane surfaces after testing provide little insight into the magnitude of nanoclay fouling (Fig. S6). Visually, the nanoclay does not provide sufficient optical contrast to accurately quantify. Microscopically, the intrinsic roughness of the membrane creates difficulty in ascertaining the location of fouling particles. Ultimately, these methods of characterization are truly only relevant for surface fouling. Considering the relatively large thickness of these membranes (150-250 μm), the majority of the surface area where fouling can occur is located within the internal bulk structures of the membranes. To evaluate the propensity to fouling overall, thermogravimetric analysis (TGA) of the membranes was performed. Figure 3(d) reports mass accumulation on the membranes, as determined by TGA, after multiple filtration runs indicating irreversible fouling buildup causing pore clogging or blocking. A significant result is that not only are the LGMs processing threefold the feed volume of the control membranes but also their mass accumulation is 40% less than the control membranes. Probing the morphology of the control and liquid gated membrane reveals a stark disparity in roughness between the surfaces (Fig. S6). The underlying PTFE features many exposed asperities which are masked by the gating liquid, presenting a smoother surface. It has been reported that interfacial interactions between spherical particles and substrates are enhanced by surface roughness.²⁹ The ability of LGMs to retain less solids within the membrane matrix can be beneficial in product recovery applications, where a reduction in material lost to the membrane can increase overall process yields. Membranes modified to increase processing capacity can result in scaled down treatment plants with less capital investment for operations.

In order to quantify the mode of fouling and onset of fouling, constituent fouling model analysis was conducted for cycles 1, 3, and 5 from filtration experiments summarized in Figs. 3(a) and 3(b). Table I shows the fouling performance of the non-gated PTFE control membranes compared to the PTFE LGMs. Darker green depicts the highest R² values and best GOF, whereas darker red shows the poorest GOF. The onset of fouling is clearly delayed in the case of LGMs for cycle 1 with threefold longer duration for TMP to build up to 120 kPa due to fouling. Similar modes of fouling for both control membranes and LGMs were observed for cycle 1. The GOF analysis suggests a combination of complete and intermediate pore blocking as dominant fouling modes for the control membranes and complete blocking as the major fouling mode for LGMs. Mechanistically, complete blocking assumes that the pore openings become sealed and flow through them is prevented, thus reducing the overall number of pores available to the incoming feed. Similarly, intermediate blocking also assumes sealing of the pore entrances by a fraction of particles and a subsequent deposition of particles onto the top surface. The fouling model fits for cycle 1 are plotted in Figs. S7 (control membranes) and S8 (LGMs). The state of fouling by the end of cycle 1 appears mostly of reversible nature in the case of LGMs as it returns to baseline pressure post-backwashing, as shown in Fig. 2. Pressure recovery behavior in experiments and fouling data fits for cycle 1 infer that initially surface fouling (i.e., complete and intermediate blocking) raises the transmembrane pressure and internal membrane fouling is insignificant with LGMs, likely due to liquid-lining of pores. By cycle 3, control membranes show cake formation, in which particles build up at the membrane surface in a permeable cake of increasing thickness and hydraulic resistance, as the dominant fouling mode causing an immediate pressure buildup. Conversely, LGMs depict intermediate and standard blocking as the primary fouling modes (Table S1). Standard blocking occurs when there is accumulation on the pore walls, leading to pore constriction and reduced permeability. LGMs also show an extended time to foul (nearly 4 times longer) and a delayed onset of fouling for cycle 3. However, by cycle 5, both LGMs and control membranes show cake formation as the primary fouling mode that corresponds to increased irreversible fouling. LGMs continue to show the same trend of longer time to foul albeit at a reduced rate (Table S1). It should be noted that depending on the concentration of foulants in the feed, cake formation is an inherent limitation of dead-end configuration.

The filtration efficiency (the amount of particles removed as a percentage of particles present before filtration) of LGMs and non-gated control membranes is shown in Fig. 4. LGMs not only have higher throughputs but also their filtration efficiency is comparable to the control membranes for the tested nanoclay concentration. It is possible that different modes of fouling may result in different filtration efficiencies for LGMs. Our data show that filtration efficiencies for cycle 1 and cycle 5 are similar as both types of membranes have similar fouling modes (Tables I and S1). An important consideration is once cake formation becomes the dominant fouling mode in combination with irreversible fouling as in cycle 5 (Table S1), the cake layer can act as an additional depth filter, making it difficult to distinguish between filtration performance due to the membrane as compared to filtration due to the cake layer formation. The feed used in these studies is composed of a polydisperse suspension of inorganic particles (Fig. S4). The fairly wide distribution of particle size is representative of what can be expected in many industrial applications. The size of the particles with respect to the membrane pore sizes and the relative distribution of these particles in the suspension will greatly impact which type of fouling can occur. In the case where the feed consists of mostly larger particles, surface fouling (i.e., complete, intermediate, and cake formation) leads to pressure buildup. But when many small particles are also present in the feed, not only is internal fouling (i.e., standard blocking) possible but also the dynamics of fouling and recovery becomes more complex.^30–32 Besides physical dimensions, particle-particle and particle-membrane interactions resulting from surface energy,³³ electrostatic charge, and hydrodynamic forces can dictate fouling susceptibility.^34,35 Further research is needed on the impact of liquid gating mechanism in minimizing fouling with respect to the abovementioned interactions.

Loss of gating liquid due to leaching in the permeate was also quantified by measuring the amount of gating liquid remaining retained in the membrane under various scenarios. Figure 5 shows the comparison of gating liquid initially priming the membrane prior to any filtration and following flow of de-ionized (DI) water and filtration with nanoclay feed with backwashing. The presence of nanoclay resulted in a 15% decrease in the gating liquid volume over 5 cycles, highlighting the need for replenishment of gating liquid for processes requiring longer batch runs or continuous use. The high surface energy and surface area of the hydrophilic nanoclay particles present in the feed likely promote accelerated loss of the gating liquid, suggesting an opportunity for further optimization of gating liquid chemistry for this application.

In this work, the performance of liquid gated membranes in the presence of a model inorganic foulant—nanoclay—has been investigated. These results confirm that membranes with liquid gating reduce TMP as proposed earlier.¹² We also show that LGMs increase throughput and reduce fouling buildup. In this particular study evaluating the microfiltration of a polydisperse nanoclay suspension, LGMs exhibited a 10%-15% reduction in critical TMP; they take nearly threefold longer time to foul; and they produce over twofold times higher throughput; LGMs also have the ability to return to baseline pressures after backwashing demonstrating extended reversible fouling; they can also reduce the amount of backwashing water; and they have less than half the mass accumulation (irreversible fouling) compared to non-gated membranes. Filtration performance was dominated by fouling for the selected feed concentration as both the LGMs and the control membranes eventually suffered from different fouling modes. However, the onset of fouling and impact of fouling were significantly delayed with LGMs. The demonstrated performance of LGMs serves as an early stage validation of this technology. Future work includes application-specific development and testing of the LGM mechanism demonstrating feasibility for high-impact applications such as water treatment, bioprocessing, and other industrial filtration processes.

See supplementary material for further analysis, materials characterization, setup details, and supporting results.

This work was supported in part by the Advanced Research Projects Agency-Energy (ARPA-E), U.S. Department of Energy, under Award No. DE-AR0000326. This work was performed in part at the Center for Nanoscale Systems (CNS), a member of the National Nanotechnology Coordinated Infrastructure Network (NNCI), which is supported by the National Science Foundation under NSF Award No. 1541959. CNS is part of Harvard University.