Mesoporous Polymer-Derived Ceramic Membranes for Water Purification via a Self-Sacrificed Template.

Membrane separation has been widely used in water purification, and mesoporous ceramic membranes show a high potential in the future because of their high stability and resistance to harsh environments. In the current study, a novel polymer-derived ceramic silicon oxycarbide (SiOC) membrane was developed via a preceramic reactive self-sacrificed method and was further applied in a homemade dead-end system for water purification. A cyclosiloxane hybrid polymer was selected as the precursor and polydimethylsiloxane (PDMS) was used as the sacrificial template. Membrane pores were formed because of template removal during the sintering process, creating channels for water transportation. The pore size and porosity could be readily adjusted by changing the amounts and types of PDMS used in the fabrication process. The as-prepared SiOC membrane showed a high water permeability (140 LMH@2.5 bar) and high removal rate of rhodamine B (RhB), demonstrating its potential applications in water treatment. This work would provide an easy and scalable method to prepare ceramic membranes with a controlled pore size, which could be used for different water treatment applications.

Introduction

As a consequence of worsening environments due to growing population, climate change, and industrialization, lack of adequate water has become a global problem which needs to be solved soon.[1,2] Safe water resources have been challenged by wastewater discharged from municipal, agricultural, and industry uses.[1] To address this challenge, there is a great need for developing new technologies to treat seawater or wastewater for human uses, including drinking, house usage, and industry usage, beyond what can be obtained from the hydrologic water cycle.[3,4] Desalination of seawater and reuse of wastewater have been regarded as feasible methods to solve water scarcity and support population growth and sustainable development. Particularly in Middle Eastern and North African countries where water is a vital issue, the governments have been trying every single method to secure freshwater. The linchpin for desalination and wastewater reclamation is membrane technology, which shows superior performance than thermal approaches.[5]

Compared with the commonly employed polymeric membranes, which would suffer from biofouling and biodegradation, resulting in the failure of their performance,[6−8] ceramic membranes possess excellent chemical resistance, which could be easily cleaned for sustainable use. Another advantage of ceramic membranes is that they could be used under harsh conditions such as high temperature, high pressure, and highly contaminated feed solution.[9−11] Even though the ceramic membranes are highly stable, durable, and could be used for a long term, the membrane fabrication process is not so cost effective. The traditional ceramic membranes are processed through a powder sintering method at a high temperature and it is very difficult to control the pore size.[9] Thus, the high cost and difficulty to scale up become the main problems for the industrialization of ceramic membranes. Other materials such as metal–organic frameworks (MOFs),[12] covalent organic frameworks (COFs),[13] and graphene or graphene oxide[14,15] have also been investigated, but the high cost and low scalability hindered their applications.[3]

On the other hand, silicon oxycarbide (SiOC) has been developed as a new kind of ceramic for multiple applications.[16−18] It could be prepared from the preceramic polymers with much lower sintering temperatures, and the pore structures could be tuned easily.[19−21] SiOC ceramics also show good thermal stability, chemical resistance, and good mechanical properties, allowing them to sustain under harsh environmental conditions.[22,23] Recently, Dong et al. fabricated a microporous SiOC ceramic membrane with a diameter of 0.83 μm and demonstrated its possible application in oil–water separation.[24] However, the rejection of other organics has not been well-established with the novel membrane. Meanwhile, tuning the pore size in the mesoporous range for specific water treatment applications remains a challenge. Therefore, in the current study, a new kind of mesoporous SiOC ceramic membrane was prepared via a facile preceramic reactive self-sacrificed method. This method utilized the property of preceramic polymers, which could be processed like a polymer, to make the scalability possible. The pore size of polymer-derived SiOC ceramic membranes could be readily adjusted by changing the types and amounts of polydimethylsiloxane (PDMS). Furthermore, the SiOC ceramic membrane was applied in a membrane filtration system to investigate the potential applications in water treatment. Compared with other polymer-derived ceramics, including SiC, Si3N4, and SiBC, the preceramic used for SiOC fabrication is much more cost-effective. The possibility of tuning the pore size in the mesoporous range could provide a new route to fabricate inorganic membranes for different water treatment applications and guide in the design of a new generation of inorganic membranes.

Results and Discussion

The preparation of the SiOC membrane via a reactive self-sacrificed method is shown in Figure . Typically, the cyclosiloxane hybrid polymer was chosen as the matrix polymer precursor because of its high ceramic yield (83%@1000 °C) (Figure A). The high ceramic yield could decrease the size shrinkage during the sintering process and keep the “skeleton” of SiOC ceramic intact after sintering. The cyclosiloxane hybrid polymer could be formed via polymerization of 2,4,6,8-tetramethylcyclotetrasiloxane (D4H) and 2,4,6,8-trtramethyl-2,4,6,8-tetravinylcyclotetrasiloxane (D4V). PDMS was used as the self-sacrificed template because it has a similar chemical nature to that of the cyclosiloxane hybrid polymer, allowing homogeneous dispersion.[25] The details for the experiment could be found in the Experimental Section. PDMS tends to phase-separate during the curing procedure, forming nanosized PDMS regions in the cyclosiloxane hybrid polymer (Figure S1). As can be seen from Figure A, PDMS could be totally burnt off at temperatures above 600 °C. When the cylosiloxane hybrid polymer was sintered at 1000 °C, the PDMS regions would disappear, leaving nanopores inside the SiOC ceramics. The design of the SiOC membrane is at the molecular level, and the phase separation to create pathways for water is nanosized.

Figure 1

Preparation of the SiOC membrane via a reactive self-sacrificed process.

Figure 2

(A) TGA curves of the cyclosiloxane hybrid polymer and three kinds of PDMS in nitrogen, (B) TGA curves of SiOC ceramic in nitrogen and air, (C) XRD pattern of SiOC ceramic, and (D) Raman spectra of SiOC ceramic.

The thermal stability of the obtained SiOC membrane could be demonstrated by thermogravimetric analysis (TGA) (Figure B). The SiOC membrane showed no significant weight change in both nitrogen and air until 1000 °C, indicating its high thermal stability. It showed a typical amorphous X-ray diffraction (XRD) pattern (Figure C), and it would stay amorphous for application temperatures below 1000 °C. The Raman spectra (Figure D) show two characteristic features of disordered graphitic forms of carbon (D band: 1346 cm–1 and G band: 1542 cm–1), demonstrating the presence of free carbon in the prepared SiOC membrane.

The effect of the functional groups on PDMS was investigated by using vinyl-terminated (Mw = 25,000 g mol–1) and trimethyl-terminated PDMS (Mw = 63,000 g mol–1 and Mw = 139,000 g mol–1). As shown in Figure S2, the SiOC membranes prepared from vinyl-terminated PDMS showed a sphere-like morphology and the “spheres” seemed to be connected with each other. This might be due to the hydrosilylation reaction between the cyclosiloxane hybrid polymer and vinyl groups from PDMS, which disturbed the phase separation process, leading to the formation of “closed pores” during the following pyrolysis process.

The SiOC membranes prepared from trimethyl-terminated PDMS showed a morphology of uniformly dispersed spheres. The formation of a microsphere morphology should be attributed to the difference in the surface energy of PDMS and the cyclosiloxane hybrid polymer. The diameter of the spheres does not change much in SiOC membranes prepared from PDMS with a molecular weight of 63,000 g mol−1 because they can be easily dispersed in the cyclosiloxane hybrid polymer due to the low viscosity and a relatively short chain length. The diameter of the spheres could be easily tuned by changing the loading content of trimethyl-terminated PDMS with a molecular weight of 139,000 g mol−1, ranging from several μm to tens of μm (Figures and S3). The SiOC membrane prepared from 20 wt % PDMS (Mw = 139,000 g mol−1) showed larger spheres (Figure C), which could be ascribed to fewer PDMS chains at such loading, leading to larger sphere diameters compared with that made from 20 wt % PDMS (Mw = 63,000 g mol−1) (Figure A) and 30 wt % PDMS (Mw = 139,000 g mol−1) (Figure D) during the sphere formation process. It should be noted that if the loading content of trimethyl-terminated PDMS is too low (such as 10 wt %), the cyclosiloxane hybrid polymer tended to form spheres with wider diameter distribution (Figure S3) and such nonuniformity is not favorable in membrane separation because of the poor selectivity.

Figure 3

SEI (Secondary Electron Image) images of fracture surfaces of SiOC membranes prepared with (A) 20 wt % PDMS (Mw = 63,000 g mol–1), (B) 30 wt % PDMS (Mw = 63,000 g mol–1), (C) 20 wt % PDMS (Mw = 139,000 g mol–1), and (D) 30 wt % PDMS (Mw = 139,000 g mol–1).

Table shows the parameters of the as-prepared SiOC membranes obtained from the N2 abosorption-desorption test (Figure S4) and water permeability test. As the content of PDMS loading increased, the linear shrinkage and porosity increased, while the apparent density decreased. This might be due to the weight loss from PDMS during the sintering process. It could also be concluded that for SiOC membranes prepared with trimethyl-terminated PDMS the higher the molecular weight of PDMS used, the higher the apparent density of the corresponding SiOC membrane is. However, the SiOC membranes prepared with vinyl-terminated PDMS showed a higher density compared to other SiOC membranes prepared with the same PDMS content. The average pore size is within 100 nm, demonstrating that the as-prepared SiOC membranes are mesoporous.

Table 1

Parameters of SiOC Membranes

samples		diameter (mm)	linear shrinkage (%)	apparent density (g cm–3)	surface area (m2 g–1)	average pore size (nm)	porositya (%)
vinyl terminated PDMS	Mw = 25,000 g mol–1, 20 wt %	23.17	22.77	1.90	-	-	-
	Mw = 25,000 g mol–1, 30 wt %	23.14	23.23	1.62	-	-	-
trimethyl terminated PDMS	Mw = 63,000 g mol–1, 20 wt %	22.69	24.38	1.38	8.29	9.91	35
	Mw = 63,000 g mol–1, 30 wt %	22.42	25.28	1.28	11.75	6.19	45
	Mw = 139,000 g mol–1, 20 wt %	22.64	24.54	1.49	25.43	7.37	27
	Mw = 139,000 g mol–1, 30 wt %	22.23	25.90	1.20	28.58	7.14	41

Porosity was obtained from the water permeability test.

The hydrosilylation reaction between the cyclosiloxane hybrid polymer and vinyl-terminated PDMS would turn part of the PDMS into ceramic during sintering instead of 100% volatile molecules would be burnt off, leading to a higher apparent density. That would be one critical shortcoming for SiOC membranes prepared with vinyl-terminated PDMS. Because part of the vinyl-terminated PDMS could be turned into SiOC ceramic during the sintering process, the pathways for gas release might be blocked and the remaining volatile siloxane part would be trapped inside the SiOC membrane. In this case, during the sintering process, high internal pressure would lead to cracks within the SiOC membranes (Figure S5), which is unfavorable in the separation process.

The SiOC membrane prepared with PDMS of higher molecular weight showed a higher surface area. If the separated PDMS chain is not long enough (Mw = 63,000 g mol–1), PDMS would form closed pores, leading to a lower surface area but higher porosity. If the separated PDMS chain is sufficiently long (Mw = 13,900 g mol–1), the pores would connect with each other, forming a water-continuous pathway (higher surface area but lower porosity). Meanwhile, the SiOC membrane possessed a hydrophilic surface. Water drops could be quickly absorbed by the SiOC membrane, demonstrating a hydrophilic nature and pathways for water to transport inside the membrane.

The size exclusion effect is critical in the performance of the separation/desalination/wastewater treatment process. For microfiltration, ultrafiltration, and nanofiltration, the selectivity is mainly controlled by the size of pores. In this work, the pore size of the SiOC membrane is from 6.19 to 9.91 nm and it can be readily adjusted by changing the type and content of PDMS used. Because the pore size can be tuned as we wish, we can further design SiOC membranes with a desired pore size for different water treatments.

The water permeability and rejection performance were tested in a homemade dead-end membrane filtration system (Figure S6). All tests were repeated three times. As shown in Figure a, the water flux increased linearly with the increase of transmembrane pressure for all membranes. The water permeability could be related to the pore size distribution and the porosity of the membranes. When the PDMS (trimethyl-terminated, Mw = 63,000 g mol–1) content increased from 20 to 30%, the particle size of the resultant SiOC membrane did not show any significant change (Figure A,B) but the porosity increased from 35 to 45%. The high porosity should be caused by the pore-forming process during the sintering and results in a higher flux. Interestingly, for the samples prepared with trimethyl-terminated PDMS of Mw = 139,000 g mol–1, when the concentration of PDMS increased from 20 to 30%, the porosity of the SiOC membrane increased from 27 to 41%, with a negligible effect on the water permeability. The higher porosity did not contribute to a higher flux, which could be caused by the higher resistance due to the smaller pore size (7.14 nm) and more pores with closed ends, These closed pores would contribute to its porosity but does not contribute to water permeability.

Figure 4

(A) Water permeability of different kinds of SiOC membranes and (B) rejection of different feed solutions via the SiOC membrane. The membrane used for the rejection test is prepared from trimethyl-terminated PDMS, Mw = 63,000 g mol–1.

The rejection performance of the SiOC membrane (prepared from trimethyl-terminated PDMS, Mw = 63,000 g mol–1) was investigated and the result is shown in Figure B. Surprisingly, the membrane showed a high removal rate (95%) on rhodamine B (RhB) solution. Because the molecular size of RhB is much smaller compared with the pore size, the normal size exclusion mechanism might not be suitable to explain the unusual phenomena. Therefore, membrane adsorption combined with separation was proposed. The van der Waals forces exist between the aromatic structure of RhB and the sp2 free carbon phase in the SiOC network and the electrostatic ionic interaction is present between the carbon/silicon–oxygen complex and the positively charged surface of RhB.[26,27] After adsorption of RhB on the SiOC membrane, the pore size would be decreased further because of the attachment of the organics, resulting in a higher removal rate. These results showed that the SiOC membrane prepared via this method could be used for water treatment. The possibility of tuning the pore size and porosity could make it a good solution to produce filtration membranes for different applications.

Experimental Section

Chemicals

Vinyl-terminated PDMS (Mw = 25,000 g mol−1) and trimethyl-terminated PDMS (Mw = 63,000 and 139,000 g mol−1) were purchased from Alfa Aesar (China). 1,3-divinyl-1,1,3,3-tetramethyldisiloxane platinum (0) (Karstedt’s catalyst), 2,4,6,8-tetramethylcyclotetrasiloxane (D4H), and 2,4,6,8-tetramethyl-2,4,6,8-tetravinylcyclotetrasiloxane (D4V) were purchased from Aladdin (China).

Preparation of Porous SiOC Ceramic Membranes

An appropriate amount of PDMS and Karstedt’s catalyst (0.2 wt %) was first dissolved in D4V by magnetic stirring. D4H (the mass ratio of D4H/D4V is 1:1) was added to the solution and the solution was stirred at room temperature for 12 h. The resultant viscous samples were transferred to Teflon molds and degassed in a desiccator before they were cured in an oven (80 °C for 4 h and 150 °C for 4 h). The samples were pyrolyzed in flowing nitrogen at 1000 °C for 2 h with a ramp rate of 5 °C min–1. The SiOC membranes were carefully polished, washed with DI water and ethanol, and dried before the tests.

Characterizations

TGA curves were recorded on a Diamond TG/DTA instrument (PerkinElmer, original from America) from room temperature to 1000 °C at a heating rate of 10 °C min–1. The XRD pattern was obtained using D8 ADVANCE (Bruker, German) from 15 to 80° with a step size of 0.1°. SEI images were obtained using FEI Quanta FEG 250 (FEI, America). The fracture surfaces were sputter-coated with platinum before the test. Transmission electron microscopy (TEM) images were obtained using JEM2100 (JEOL, Japan). First, the precursor was sliced using a Gatan691 ion beam thinner (Gatan, America). Then, the precursor slice was dropped on a copper grid and dried before being loaded into the transmission electron microscope. The Raman spectra were collected on a Renishaw inVia Reflex (Ar+ laser excitation, Renishaw, the United Kingdom). Brunner–Emmet–Teller data were obtained using ASAP2020M (Micromeritics, America) through nitrogen absorption and desorption.

Water Permeability and Rejection Test

The water permeability and rejection performance were tested in a homemade dead-end membrane filtration system (Figure S6). A high-pressure metering pump (5979 Optos Pump 2HM, Eldex) was used to feed the solution into the system, and the membrane was secured in the membrane cell module (SS316 stainless steel). The effective membrane area was calculated as 2.54 × 10–4 m2. To investigate the pure water permeability of different kinds of SiOC membranes, a varied range of transmembrane pressures was applied from 0 to 2.5 bar. The filtration test was conducted in 60 min. The mass of the permeate was recorded using a balance, which was connected to a computer. The water flux was calculated according to the equationwhere J means the permeate flux (LMH, L m–2 h–1), V is the volume of the permeate (L), A represents the effective membrane surface area (m2), and t is the filtration time (h).

The overall membrane porosity ε (%) was calculated based on a wet–dry weight method through the following equationwhere Wwet and Wdry represent the weights of wet and dry membranes, respectively, and ρH2O and Vmem are the density of DI water and the volume of the SiOC membrane, respectively.

In the rejection test, the pure water in the feed tank was replaced by 1 M NaCl, 1 M Na2SO4, and 100 ppm rhodamine B (RhB). The rejection was calculated as followswhere Cp and Cf represent the concentration of salt or RhB in the permeate site and feed site, respectively.

The concentration of salt was examined by the conductivity change via a conductivity meter (Myron L), and the RhB concentration was detected using a UV-1800 spectrophotometer (Shimadzu, Japan).