Elsevier

Desalination

Volume 316, 1 May 2013, Pages 53-66
Desalination

Preparation of polysulfone membranes via vapor-induced phase separation and simulation of direct-contact membrane distillation by measuring hydrophobic layer thickness

https://doi.org/10.1016/j.desal.2013.01.021Get rights and content

Abstract

Polysulfone (PSf) flat-sheet membranes with bi-continuous porous surfaces for direct-contact membrane distillation (DCMD) were successfully fabricated using a vapor-induced phase separation (VIPS) method. The present investigation revealed how the surface and cross-sectional morphology of PSf membranes and the desalination performance in DCMD were affected by exposure time, PSf content and relative humidity of air. In the VIPS process, an increase in exposure time led to a replacement of the bi-continuous top surface with a dense skin and to a large decrease in the permeate flux in DCMD. The best PSf membrane fabricated in this study had a mean pore radius of 0.32 μm, water contact angle of 106.4°, the liquid entry pressure of water of 300 kPa, and total porosity of 82.1%. During the DCMD process with a 35 g/L sodium chloride solution, the best membrane produced a permeate flux of 30.0 kg m 2 h 1 and a very low conductivity of distilled water at hot-feed and cold-distillate side temperatures of 73 °C and 25 °C, respectively. The thickness of the hydrophobic layer of the membranes was first measured using the weight method, and its value was used in a simulation of the DCMD process.

Highlights

► PSf hydrophobic membranes were prepared with VIPS. ► Bi-continuous top surface was good for the permeate flux in DCMD. ► There was hardly any wetting during 90 h desalination test. ► The thickness of the hydrophobic layer was first measured using the weight method.

Introduction

Membrane distillation (MD) is a new membrane separation process that has many industrial applications, mostly in water desalination [1], [2], [3], [4], [5], [6] and wastewater treatment [7], [8], [9], [10]. The permeate flux of MD is driven by the vapor pressure difference across the membrane that results from the temperature difference and solution composition gradients in the boundary layers adjacent to the membrane [11]. The MD process may be used as a substitute for conventional separation processes such as multistage vacuum evaporation, reverse osmosis, and distillation [12]. Compared to those processes, the advantages of MD are as follows: (1) lower required operating temperature and vapor space than conventional distillation, (2) lower operating pressure than RO, (3) 100% (theoretical) rejection of non-volatile solutes, (4) no limitation due to high osmotic pressure, and (5) lower energy consumption than multistage vacuum evaporation [13].

For an MD process, the porous membrane acts only as a barrier, but it plays a crucial role. El-Bourawi [11], Khayet [14], Susanto [15] and Alkhudhiri [16] have reviewed the characteristics of most commercially available membranes used in MD processes. MD membranes must be hydrophobic and porous and are mainly composed of polypropylene (PP), polyethylene (PE), polyvinylidene fluoride (PVDF) and polytetrafluoroethylene (PTFE). Polysulfone (PSf) is rarely used in MD membranes, except for polyethersulfone (PES) and PSf membranes modified by fluorinated surface-modifying macromolecules (SMMs) [17], [18], [19], [20] that increase the hydrophobicity of the membrane surface.

PSf is also a hydrophobic material, although its hydrophobicity is not as good as that of PVDFs. Porous PSf membranes can be used in MD processes once surface hydrophobicity is improved [20]. Compared to PVDF and PTFE, the most widely researched MD membrane materials, PSf exhibits superior rigidity that endows it with a characteristic propensity against contraction. Membrane contraction is problematic in MD industrialization [21]. PSf is a good membrane material for MD; the key is to improve the hydrophobicity of the membrane surface.

Superhydrophobic surfaces reported in the literature are mainly found in materials that exhibit both hydrophobicity and surface roughness; vapor-induced phase separation (VIPS) is used to construct rough surfaces. Xie [22], Peng [23] and our lab [24] have prepared porous, highly hydrophobic PVDF surfaces with micro- and nanoscale hierarchical roughness via VIPS. VIPS is a method wherein nonsolvent (usually water) vapor is introduced into a polymer solution to generate phase separation and a porous structure [25]. A recent research trend focuses on how to apply VIPS to the formation of a super-hydrophobic surface. The PVDF porous membranes prepared by our lab were successfully used in vacuum membrane distillation (VMD) and direct-contact membrane distillation (DCMD) processes [24], [26]. This paper reports research results for PSf membrane synthesis via VIPS.

Although good progress has been made in the preparation of MD membranes using NIPS in recent years, almost all the MD membranes prepared in the literature suffer from low membrane surface porosity. This is because the coagulant is usually water, a strong non-solvent that induces rapid phase inversion and leads to the formation of a dense skin with low surface porosity. Membrane experts believe that a porous surface will form when phase separation occurs at a region close to the doping solution critical point [27]. One way to induce this phenomenon is to use a casting solution with low polymer content. Another possible approach is to induce delayed demixing using a mixture of non-solvent and solvent as the coagulant. Bonyadi et al. [28] increased the outer surface porosity of PVDF fibers by introducing a solvent-dope solution two-phase flow in the air-gap region prior to entering the external coagulation bath. The PVDF concentration was lowered to a point closer to the critical point, which create delayed demixing during membrane formation. Because PVDF content is low just before demixing, this decreases the mechanical strength of the final membrane.

In this paper, we report the application of an efficient approach to increase the top surface porosity of PSf membranes by controlling the VIPS exposure time. Lee et al. [29] first found that the VIPS of a PSf/N-methylpyrrolidone (NMP) solution occurred via spinodal decomposition on the basis of time-resolved small-angle light scattering data, even though the membrane structure obtained was not bi-continuous. Such inconsistency between the phase-separation mechanism and membrane morphology was resolved by the work of Kuo et al. [30], who observed that, during VIPS, the nascent bi-continuous structure formed after spinodal decomposition coalesced to cellular pores when low-volatility solvent was used. The polymer-rich phase was slow to vitrify when the solvent concentration was kept high in the demixed solution, due to its low volatility. Tsai et al. [31] found that the bicontinuous structure developed into a cellular structure if the air gap was long enough to allow the coarsening process to proceed when PSf hollow fiber membranes were fabricated using a dry/wet spinning process. Su [32] determined the composition paths using FTIR microscopy and found that both spinodal decomposition and nucleation-and-growth can occur during VIPS, depending on the duration that the solution persisted in the meta-stable region. Any spinodal decomposition that occurred was dependent on this duration: the shorter the duration, the higher the probability that the solution phase-separated in this manner. The duration in the meta-stable region was not only related to the amount of water needed to change the solution from binodal to spinodal (i.e., the gap between them on the ternary phase diagram) but also depended on whether a skin that retarded water penetration into the cast film during membrane formation was formed. According to the above discoveries, a bi-continuous membrane structure without a skin (an ideal MD membrane) could be obtained via VIPS if a low-PSf solution was used.

Tsai et al. [33] found that the collapse of lacy to cell-like structures in PSf membranes can be effectively retarded with the use of 2-pyrrolidinone (2P), a poorer solvent for PSf. The nascent lacy structure and pore connectivity were maintained. The more pronounced effect of 2P on the formation of poly(methyl methacrylate) (PMMA) membranes arises due to greater hydrogen bonding between 2P and PMMA than between 2P and PSf.

Although the role of VIPS in the formation of PSf membranes was widely studied, as discussed in the preceding paragraph, no literature report has identified the importance of VIPS in the preparation of MD membranes. However, the role of VIPS in determining the morphology of membranes has been investigated. A few reports [34], [35] have noted that ambient humidity may influence membrane morphology. An investigation of the interplay between VIPS and membrane morphology is still needed to gain insight into the morphological control of membranes.

To date, most studies on VIPS have focused on phase-separation mechanisms and mass transformation during membrane formation. To the best of the authors' knowledge, no study has concentrated on the preparation of hydrophobic PSf membranes via VIPS, especially structural changes occurring in VIPS. This study therefore investigated the PSf membrane surface and bulk changes in VIPS, in which different membrane morphologies were produced by controlling exposure time, polymer content in the casting solution and relatively humidity (RH) of air. The DCMD performance of porous PSf membranes is related to membrane morphology in addition to the membrane preparation factors outlined above. The prepared membranes were characterized using mean pore radius(r), maximum pore radius (rmax), total porosity, scanning electron microscopy (SEM), the liquid entry pressure of water (LEPw) and contact angle (CA) measurements. Membrane performance was investigated in DCMD and the results were compared to PSf membrane data in the literature. The best membranes were chosen to undergo long-term DCMD tests.

To simulate mass and heat transfer in hydrophilic/hydrophobic membranes, we must know the thickness of the hydrophobic layer, which is difficult to determine, even using SEM, because of the difficulty of determining the boundary between the hydrophilic and hydrophobic layers. The thickness of the hydrophobic layer or the air gap in the membrane was calculated by weighing the membrane and applying this value to a mathematic model to predict the permeate flux in DCMD.

Section snippets

Materials

PSf (UDELTM 3500, specific gravity 1.24) was obtained from Solvay Shanghai Co. Ltd., and 1,2-propylene glycol (PG), used as an additive, was purchased from the Tianjin Fu Chen Chemical Reagent Factory. Isopropyl alcohol (IPA, GR grade, Merck) was used as a membrane-wetting liquid. Dimethylacetamide (DMAc) (analytical reagent) was purchased from the Tianjin Yongda Chemical Reagent Co. (China). Sodium chloride (NaCl) was purchased from the Beijing Chemical Works.

Membrane preparation

PSf membranes were prepared via a

Theory

The heat transfer that occurs in DCMD can be divided into four regions [1], [26], [37] (Fig. 2). The heat and mass transfer can be simulated according to the literature [37]. The thickness of the membrane δ could be measured. The thickness of the hydrophobic layer is the air gap thickness in the hydrophobic/hydrophilic membrane. The ratio of air gap thickness to the total membrane φ could be measured by weight method, and the thickness of the air gap in membrane δa could be calculated according

Results and discussion

Four cases with varying PSf content in the casting solution and RH of air were studied. The sample with a PSf content of 13 wt.% and an RH of 68% is referred to as membrane 13%PSf 68%RH. The 13%PSf 58%RH, 15%PSf 68%RH and 13%PSf 58%RH membranes are similarly named. For all the membranes fabricated in this study, when an aqueous solution of NaCl at a concentration of 35 g/L was used as feed solution and the conductivity of distilled water was in the range of 20 μs/cm and 40 μs/cm, no wetting was

Conclusions

A VIPS method to prepare flat-sheet PSf microfiltration membranes was explored. The surface and cross-sectional morphology of the PSf membranes and the performance of desalination in DCMD were affected by exposure time, polymer content and relative humidity (RH) of air. An increase in exposure time during the VIPS process led to the replacement of the bicontinuous top surface with a dense skin, and the permeate flux in DCMD decreased greatly. The water contact angle, thickness, porosity, mean

Nomenclature

    C

    membrane distillation coefficient (kg/m2 s Pa)

    B

    tortuosity of the membrane

    Ba

    permeance coefficient

    D

    diffusion coefficient of vapor in air (m2/s)

    h

    heat transfer coefficient (W/m2K)

    I0

    intercept defined in Eq. (2) (mol m 2 S 1 Pa 1)

    L

    equivalent length (m)

    LEP

    liquid entry pressure (Pa)

    LP

    effective pore length (m)

    M

    molecular weights of water and air (kg/mol)

    ΔP

    transmembrane pressure (Pa)

    Pf,m

    vapor pressure at the feed membrane surface (Pa)

    Pm

    average partial pressure in the membrane pore (Pa)

    Pp,m

    vapor pressure at

Acknowledgments

The authors thank the National Natural Science Foundation of China for financial support (21176008).

References (41)

Cited by (0)

View full text