Nanoporous metal–polymer composite membranes for organics separations and catalysis

Abstract

Metallic thin-film composite membranes are produced by sputtering metal films onto commercial polymer membranes. The separations capability of the membrane substrate is enhanced with the addition of a 10 nm Ta film. The addition of a tantalum layer decreases the molecular weight cutoff of the membrane from 70 kDa dextran (19 nm) to below 5 kDa (6 nm). Water flux drops from 168 LMH/bar (LMH: liters/meters²/hour) (polymer support) to 8.8 LMH/bar (Ta composite). A nanoporous layer is also added to the surface through Mg/Pd film deposition and dealloying. The resulting nanoporous Pd is a promising catalyst with a ligament size of 4.1 ± 0.9 nm. The composite membrane’s ability to treat water contaminated with chlorinated organic compounds (COCs) is determined. When pressurized with hydrogen gas, the nanoporous Pd composite removes over 70% of PCB-1, a model COC, with one pass. These nanostructured films can be incorporated onto membrane supports enabling diverse reactions and separations.

Introduction

The development of advanced materials for improved separations and reactions is important for environmental and industrial applications. As the industry seeks to reduce energy consumption and emissions, improved separations technologies can be a major tool to facilitate these goals [1,2,3]. Improved membrane technologies are already being applied to address these challenges, and further advances in membrane technology will improve that outlook. Most membranes used in water treatment are polymer based and may be a composite of several polymer layers to improve stability while maintaining optimal separations properties [1,4,5].

Thin-film composite (TFC) membranes are a common membrane type in which a relatively thin polymer layer is formed on the top of a preexisting porous ultrafiltration membrane support. This support layer typically has a much larger pore size and serves to stabilize the thin selective layer which is formed on the support’s surface [5,6]. These membranes are widely used due to their relatively low production expense and strong performance properties [5,7]. Their high mechanical flexibility allows for production via roll-to-roll processes and incorporation of the membranes into spiral wound modules which offer process advantages such as high packing densities [2,7].

This research focuses on an alternative type of TFC membrane (synthesized without interfacial polymerization), one with a metallic thin-film top layer. The addition of a metallic layer to the surface of a membrane allows modification of membrane flux using only a plasma treatment and deposition of a thin film of metal, in this case tantalum. Producing a top layer in this way allows for the production of very thin films which can drastically alter pore size of the membrane while maintaining appropriate solvent flux. This capability is of interest for separations and shows special suitability for organic solvent-based separations applications since the metallic film top layer is highly resistant to solvents compared to many polymers. Solvent-based separations are a field of increasing importance and are applicable in high-value separations such as those in the pharmaceutical industry [8]. This type of composite structure allows for the flexible nature of the polymer backing to be retained, while the metallic film layer reduces pore size and adds a surface layer which remains stable across a broad range of operating conditions.

In addition to modifying separations properties of the membrane, the addition of metallic films may be used to introduce catalytic capabilities into the membrane. The deposition of an alloy film allows subsequent dealloying to produce a nanoporous metallic film anchored to the top of membrane. This nanoporous layer possesses a sponge-like structure with high porosity and small feature size. When these films are supported by an ultrafiltration membrane substrate, the resulting composite shows potential as a catalyst system. A wide variety of metals may be used to produce porous films for these composites, in this case, a Pd catalyst was produced for specific water treatment applications.

Membranes produced entirely from metal encompass a field all of their own with applications in wastewater treatment, and dairy and wine industries [9,10]. These membranes are often produced through metal powder sintering and typically have the advantages of resistance to damage from solvents, extreme pH, and temperature [9]. However, high costs and large pore size can limit the application of these membranes using existing technology. A variety of studies are working to advance the state of the art in the field and explore alternate fabrication strategies [11,12]. One of the primary goals of this work on developing composites using metal thin films on polymer membrane substrates is to gain some of the benefits of both metal and polymer membranes and combine them into a single platform.

Sputtering and nanoporous metals

Physical vapor deposition, in general, and magnetron sputtering specifically, are a commercially mature and versatile methods for the production of thin films [13]. One of the primary virtues of sputtering is its flexibility. Most metals can be deposited via sputtering and there are few requirements for the substrate material onto which the thin film is deposited, beyond vacuum stability. Large-scale application of sputtering processes is widespread in the industry for the production of window films and conductive coatings for flexible circuits [14,15]. Using large planar magnetron sources, substrates with a width of 36” have been successfully coated with metal oxides in a roll-to-roll process [15,16]. In this work, commercially available ultrafiltration (UF) polysulfone (PSf) membranes have been used as substrates for thin-film deposition. These same UF PSf membranes are often used to produce traditional polymer TFC membranes for water treatment.

Work has been done previously by ourselves and other groups [17,18,19] in characterizing the behavior of metal thin-film-polymer composites of this general type. Films deposited using sputtering techniques often possess a small amount of porosity due to the various growth modes of thin films from the shower of incident atoms. In addition to their inherent porosity, thin films deposited on a porous substrate are affected by the structure of the substrate. The thin film may be templated by the substrate. Essentially, the porosity of the substrate is propagated through the deposited film. These metallic thin-film composite (MTFC) membranes are most interesting for separations as the plasma cleaning and thin-film layer modify flux and rejection properties significantly.

In addition to this templated porosity, additional porosity may be generated in a metallic alloy film through dealloying. This process involves the selective dissolution of one component of an alloy, typically a less noble metal, to generate a nanoscale pore structure. Under optimized conditions, as atoms of the sacrificial component of the alloy are removed, atoms of the other, more noble, component surface diffuse to form a porous network of interconnected wires or ligaments [20,21,22]. This technique has been applied to a variety of metallic systems [23,24,25,26] for applications as disparate as the biomedical field [27], energy applications [25,28], sensing [29,30], and catalysis [28,31,32,33]. Dealloying is applied here by carefully choosing a precursor alloy film to deposit on top of the PSf UF, in this case, a Mg/Pd alloy. After this Mg/Pd alloy film is deposited, the film is then submerged in water etching Mg away and creating nanoporous Pd (npPd) on the top of the membrane. This membrane is referred to as a nanoporous metallic thin-film composite (npMTFC) membrane.

Catalysis application

The npMTFC membranes in this case are produced with a nanoporous palladium layer forming the top surface. This is done to treat specific types of polluted waters by catalyzing a hydrodechlorination reaction with H₂. Chlorinated organic compounds (COCs) are a difficult class of pollutants for environmental remediation using traditional methods as they are generally very difficult to break down [34]. Significant research has been focused on alternative treatments methods. Pd-based catalysts [35,36,37], sometimes incorporating Fe to produce hydrogen through corrosion reactions with surrounding water [38,39,40], have been effective in initiating dechlorination of the compounds, removing the compounds by converting them in situ. The Pd serves as a catalyst which allows the H₂ to dissociate, forming H radicals which will in turn dechlorinate the pollutant. In this research, a 2-chlorobiphenyl (PCB-1) was chosen as a model COC. Trichloroethylene (TCE) was chosen as another target COC to show a broader application of the membranes, and its dechlorination follows a similar pathway [41]. Previous work [17] demonstrates that npPd shows promise to act as a similar type of nanostructured catalyst for dechlorination of chlorinated organic pollutants when provided with hydrogen for reaction.

In this work, commercial UF PSf membranes will be modified with plasma treatment and a thin tantalum layer to produce an MTFC membrane. These composite membranes will be shown to exhibit improved separations capabilities compared to the original membranes. Furthermore, these MTFCs will have an additional Mg/Pd alloy layer added on top, which is made nanoporous through dealloying in water. The resulting npMTFC membranes are seen to retain the improved separations properties of the MTFC and also to function as catalysts for the dechlorination of COCs such as PCB-1. Fabrication of these membranes is explored in depth and their structures are characterized with imaging techniques, as well as traditional membrane characterization techniques, such as flow cell testing and molecular weight cut-off (MWCO) determination. Finally, the catalytic application of the npMTFC with npPd is investigated in permeation catalysis experiments.

This work represents significant advancements over our previous work synthesizing and testing composites of this type [17]. In the previous study, no significant separations were performed using the membrane due to acid damage caused by dealloying. In this work, Mg/Pd precursor films are used as precursor materials for nanoporous Pd synthesis allowing water-based dealloying. This less-aggressive etchant crucially preserves the pore structure in the underlying membrane allowing the system to function for separations, as well as for catalysis work. Avoiding the use of acid in production allows for less costly and more environmentally friendly fabrication.

Results and Discussion

Fabrication of the composite membranes proceeded in two stages. First, a plasma treatment was applied, and a thin film of Ta was deposited on the UF PSf membrane substrate to generate the MTFC membrane. Next, if the membrane was to be used for catalysis experiments, an alloy film of Mg/Pd was deposited on the top of the Ta layer and made porous through dealloying. This process resulted in the final npMTFC membrane. After the fabrication process was optimized, the composite’s structures were characterized through electron microscopy techniques. Connections were then made between the membrane’s structures as imaged and their properties for liquid separations and catalysis.

Composite structure

Before the MTFC or npMTFC membranes were produced for this study, the substrate preparation and film deposition parameters required optimization. Under incorrect deposition conditions, the films will not form the optimal structure and will often delaminate due to insufficient bonding between film and substrate, or excessive intrinsic stress in the film. To promote bonding between the deposited film and substrate, an RF plasma cleaning is typically performed before film deposition begins. Plasma cleaning promotes film adhesion by increasing substrate surface energy and removing adsorbed gasses [42].

To better understand how the plasma cleaning step changed the PSf surface, both XPS and Zeta potential measurements were performed on the pristine and cleaned surfaces, and these results are shown in Figs. S1 and S2. X-ray photoelectron spectroscopy (XPS) results confirm that the chemical bonding on the PSf surface is altered by the plasma treatment. The number of carbon–carbon bonds is reduced with plasma treatment, and sulfide forms as some of the sulfone and sulfonate groups detected in the as-received membrane are removed or converted. Zeta potential measurements indicate there is an increase in the negative charge on the membrane’s surface after plasma cleaning. These results are further supported by contact angle measurements, which show a contact angle of 72.6 ± 2.4° for the unmodified UF PSf, and an angle of 20.7 ± 3.5° for the membrane after plasma treatment. This is in line with the results of other researchers [43,44,45], who have noted that Ar plasma treatment alters O/C ratio at the membrane surface and increases surface energy. These effects serve to improve the bonding of the deposited film to the substrate material, resulting in an enhanced composite.

Optimization discussion

In addition to modifying pore size and flow properties, the Ta layer serves as an interlayer material for subsequent Mg/Pd film depositions. Ta not only possesses good ductility and corrosion resistance but also bonds strongly to carbon and oxygen making it ideal as an interlayer material. Pd especially, being a noble metal, only bonds weakly to the PSf and was found to delaminate without the Ta mediating layer. For this reason, all npMTFC membranes were produced with a 10 nm Ta interlayer to aid in membrane stability.

To produce composites with ideal structures, it was desirable to better understand the way the film grew on the polymer membrane under different deposition conditions. This allowed optimization of film conformation to the underlying pore structure and reduction of intrinsic film stresses improving composite stability. To develop this understanding, the parameter space for film deposition was explored by varying argon working pressures (2.5, 5, and 10 mtorr). Depositions with an applied substrate bias of 10 W RF were also tested as were depositions with no applied substrate bias. The resulting film structures were characterized via SEM and are shown in Fig. S3. Application of an RF bias to the substrate during deposition will generally result in a denser film, with more compressive residual stresses. In these experiments, substrate biasing was found to increase the adherence of the film to underlying structures, promoting the propagation of the substrate’s pores through the film. Higher working pressures caused the film to span the substrate’s pores more easily. However, under optical analysis, films deposited at higher pressures were found to have significant cracking, especially the 10 mtorr depositions. Taking all these factors into consideration, 100 nm Mg/Pd films deposited using light (10 W RF) substrate bias at 2.5 mtorr were found to give optimal film structure and adhesion during dealloying.

MTFC membrane fabrication

Once the cleaning and deposition studies had been performed, the ideal deposition parameters were used to generate MTFC membranes. Comparing the MTFC with only a 10 nm Ta layer [Fig. 1(b)] to the UF PSf substrate [Fig. 1(a)] shows the extent to which the underlying UF PSf membrane structure has been modified by the addition of the thin metallic film. There is no clear modification of pore size with the plasma treatment and addition of the Ta layer from SEM imaging. This is deceptive as later water flux studies will show and may largely be an effect of the improved conductivity of the material improving imaging conditions in the SEM. The mechanism of pore size reduction may not take place on the very surface of the membrane. In the ORION, sputtering sources are oriented at an angle of about 70° to the substrate surface. Looking at the geometry of the pore, it is possible that the Ta is deposited up to a depth two to three times the diameter of the pore (12.5 ± 3.6 nm). This Ta deposited inside the pore mouth is the most likely mechanism for the pore size reduction which will be discussed further in Flux characteristics section. Since this reduction is seen inside the pore mouth, it may be difficult to detect with surface imaging techniques. These surface images confirm that the Ta film is well adhered to the substrate material with good coverage and no cracking or delamination.

Figure 1:

(a) SEM image of unmodified UF PSf membrane surface. (b) SEM image of MTFC surface (UF PSf after plasma cleaning and deposition of thin Ta layer).

Precursor film deposition

The propagation of membrane pores through the as-deposited film is further illustrated by the images shown in Figs. 2(a) and 2(b) of the precursor Mg/Pd films deposited on the top of the MTFC. This structure is a UF PSf membrane with Ta layer and Mg/Pd film that has not yet been dealloyed to generate additional porosity for the npMTFC. The cross-section made via the focused ion beam (FIB) revealed a pore in the deposited films [Fig. 2(b)], which reinforced the conclusion of a porous film based on the plan-view images. This templating of as-deposited films supports the potential of the thin-film deposition process to modify separations properties of the precursor membranes. Pore size here is clearly altered by the incorporation of additional film layers. Subsequent permeation experiments further explore this phenomenon.

Figure 2:

(a) Plan-view image of precursor structure to npMTFC (UF PSf with Ta layer and Mg/Pd film). (b) FIB cross-section of the as-deposited Mg/Pd film on the top of the UF PSf membrane. From bottom to top: first is the highly porous UF PSf membrane with large pores, above that is a thin (10 nm) Ta interlayer, then a 100 nm Mg/Pd alloy film. A pore in the film is marked by a red arrow and begins in the UF PSf substrate material and propagates through the film layer. (c) Top-down image of npMTFC structure. A bimodal structure is seen, with some larger pores likely generated by the underlying UF PSf membrane porosity and then small pores that result from dealloying. (d) Cross-sectional image of npMTFC membrane after FIB milling. The bottom component is the UF PSf membrane. Above this is the Ta interlayer. Further above is the npPd layer.

npMTFC membrane structure

Once the as-deposited structure of the MTFC films was characterized, the films were dealloyed. This process consisted of submerging the precursor alloy films (75/25 at.% Mg/Pd) into water and agitating them for 1 h using a shaker table. Since magnesium easily corrodes in water, the Mg component of the film was preferentially etched away leaving a Pd-rich structure. Afterward, it was found that about 5 at.% Mg remained in the film, producing a thin film of nearly pure Pd, and generating the npMTFC membrane from its MTFC precursor. Film thickness decreased by 40% from 100 to 60 nm. Both plane view [Fig. 2(c)] and FIB cross-section imaging [Fig. 2(d)] were performed on the npMTFC to further analyze its structure.

Based on film thickness reduction and composition, the porosity of the nanoporous film is estimated to be 42%. This estimate is made based on composition and thickness measurements and assumes no significant Pd is lost during dealloying. This assumption is supported by the high corrosion resistance of Pd, as well as other studies on corrosion of Pd alloys [46,47], which show no or very low corrosion rates of Pd over long time spans. This level of porosity means the nanoporous layer should have minimum resistance to water flow. The combination of these traits is what makes the npMTFC membrane an exciting candidate for catalytically active membrane production.

The nanoporous structure shows a well-formed network of Pd ligaments with a characteristic ligament size 4.1 ± 0.9 nm. The film is seen to have shrunk significantly in thickness but remains well adhered to the membrane substrate and does not show evidence of widespread cracking or other defects. Additionally, there is a clear bimodal structure to the nanoporous structure. While the majority of pores are of the same order as the ligaments, some pores are larger, about 20 nm in diameter from surface imaging. These pores are likely those seen in the as-deposited film structures, originating from the porous polymer support and continuing through the film’s structure.

Further analysis of film structure and composition was performed using transmission electron microscope (TEM) in STEM mode, with accompanying energy-dispersive x-ray spectroscopy (EDS) mapping of the same area. Figure 3 shows a representative result of this analysis. The multi-layer structure of the composite is shown especially clearly in the EDS map. Quantification from this lamella is given as Table S1 and confirms the low Mg content remaining in the Pd film. The fine ligament size of the Pd structure is clearer in the STEM images. The conformation of the deposited film to the polymer membrane’s roughness is also well illustrated here. There is a dip in the membrane structure in Fig. 3, but the Ta clearly penetrates into the feature, as does the npPd structure.

Figure 3:

STEM imaging (a) and EDS mapping (b) of npMTFC composite membrane lamella. HAADF-STEM imaging provides a strong Z contrast, so the dark areas to the left of the frame are the low atomic number polymer regions. Brighter areas towards the right consist of either the denser Ta film or the sponge-like nanoporous Pd film.

Based on these imaging techniques, certain hypotheses about the membrane’s flow properties can be made. Due to its high porosity and the presence of large (20 nm) pores in the film, the npPd layer of the npMTFC should not be a major impediment to flow. This means the underlying MTFC structure, that is the thin Ta film on the UF PSf support, will likely determine flow behavior. Permeation testing and rejection studies will give further information on the influence of the film to the membrane’s pore size.

Flux characteristics

In addition to imaging techniques, the structure of the membrane and film were further probed using traditional membrane testing techniques. To investigate the effects of each processing step on the permeation behavior of the membrane, dead-end cell flux measurements were taken at each stage. The MTFC membrane was tested for stability in solvent permeation settings with isopropanol (IPA) flux experiments. Finally, pore size of the membranes was probed using rejection studies with a series of marker molecules.

Pure water flux

Permeation testing of the pristine PSf UF membrane was performed (168 ± 30 LMH/bar), then the PSf UF after RF plasma cleaning and the deposition of the thin tantalum interlayer (MTFC) was tested (8.8 ± 5.3 LMH/bar), and finally, the permeance of the composite membrane with npPd layer (npMTFC) was determined (9.6 ± 0.6 LMH/bar). Details of these flux studies are shown in Fig. S4. Permeance for the MTFC with undealloyed Mg/Pd top layer was not determined, because the alloy is not stable in water and would dealloy during flux determination. As illustrated by the large standard deviation associated with the permeance value, a larger distribution of fluxes is seen for the MTFC membranes than other types. For the UF PSf substrate, a standard deviation of 18% of the measured permeance is determined for the samples, while for the MTFC, the deviation is 60% of the mean permeance value. There are two likely causes of this variation, changes in the structure of the deposited Ta film layer and variation in the UF PSf substrate. Only a fraction of the permeance variation measured in the MTFC membranes can be attributed to the substrate material as indicated by the lower standard deviations in permeance testing of the UF PSf. Therefore, some variability in the structure of the Ta film deposited on the surface of the polymer membrane is the likely explanation.

A large drop in permeance is seen after the pristine UF membrane is modified via plasma cleaning and with the addition of the 10 nm Ta layer. No further flux reduction is measured after the npPd layer is produced on top of the Ta. Therefore, it is clear that the Ta film addition is the primary flow inhibitor. To determine a physical cause for this permeance reduction, flow is modeled with Hagen–Poisseuille’s equation [Eq. (1)], where $J_v$ is the membrane flux, $N$ is the pore number, $P$ is the transmembrane pressure, $\mu$ is the viscosity, and $L$ is the membrane thickness.

$$d_{\text{c}}=2{\left(\frac{8J_v\mu L}{\pi N\text{Δ}P}\right)}^{\frac{1}{4}}.$$ (1)

By depositing a Ta film on the top of the UF PSf membrane, only a few of these parameters may be altered. Membrane thickness is essentially unchanged since the film is so thin, compared to the UF PSf layer which is microns thick. Membrane pressure is kept constant between runs, as is viscosity. Therefore, it is assumed means only pore number and effective pore diameter may account for the flux reduction. A reduction in pore size makes more sense intuitively, but further experiments determining membrane separations capabilities were performed to probe effective pore diameter of the composite membranes.

To quantify this effect, pore size of the composites was estimated as shown in Table 1. The average pore size of the UF PSf substrate was averaged from 100 pore measurements taken directly from SEM images. By taking ratios of pore diameter and flux between the composite and the measured UF PSf substrate, pore diameter may be calculated as in Eq. (2) [17,48], where $J$ is the membrane flux, and $d$ is the effective pore diameter. $J_0$and $d_0$ represent measured flux and pore diameter values for the UF PSf. The pore sizes of the MTFC and npMTFC composites were estimated based on flux values measured for those membranes. This relation assumes negligible change in pore number and membrane thickness between composites. Reasonable constraints considering the templating seen via SEM imaging and the low thickness of the deposited films.

TABLE 1:

Summary of water flux behavior of composite membranes and associated pore size.

Membrane	Permeance (LMH/bar)	Estimated pore diameter (nm)
UF PSf	168 ± 30	12.5 ± 3.6
MTFC	8.8 ± 5.3	6.0 ± 0.9
npMTFC	9.6 ± 0.6	6.1 ± 0.1

UF PSf pore diameter is measured from SEM images; MTFC and npMTFC diameters are estimated based on flux measurements.

$$\frac{J}{J_0}={\left(\frac{d}{d_0}\right)}^4.$$ (2)

By propagating the variations in permeance through Eq. (2), it can be seen that the distribution of permeance seen in the MTFC membranes corresponds to a variation in pore diameter of 0.9 nm as shown in Table 1. If variation in Ta film structure is the origin of the permeance distribution, then only a small change in that thickness could account for the variance measured in these experiments. The exact relationship between deposited film thickness and pore diameter of the composite requires further study. However, it is possible that a variation of only 0.5 nm in film thickness around the pore mouth could account such a change in pore diameter of the resulting composite structure. Variations in film roughness or growth patterns could similarly affect permeance of the resulting composites. Current research in flexible electronics and plasmonics [49,50,51] can serve as a guide for physical vapor deposition synthesis techniques of continuous metal films of low thicknesses. These MTFC membranes were further tested for stability in a cross-flow cell with results summarized in Fig. S5. Similar DI permeabilities were measured, and the Ta films remained well adhered to the UF PSf substrate despite the shear flow for over 24 h of testing, indicating strong bonding and stability of the composite.

In an effort to demonstrate the flexibility of this technique across a variety of substrates, MTFC membranes were also produced using an alternative membrane substrate, a UF polyethersulfone (PES) membrane. The base membrane was tested, as was an MTFC with UF PES substrate and a 10 nm Ta layer and an MTFC with UF PES substrate and 20 nm Ta layer. Results are summarized in Fig. S6. The flux behavior fit with expected trends, permeance of the base UF PES was measured to be 136.7 LMH/bar, permeance dropped to 9.7 LMH/bar with the addition of 10 nm Ta, and further dropped to only 1.5 LMH/bar with 20 nm Ta. Composite production easily translated to a completely new membrane material in this test with no needed process modification and formed a film capable of modifying composite permeance as expected. The 20 nm deposited Ta layer seems to have decreased pore size of the membrane more than the 10 nm based on flux measurements, which could allow tailoring of composite pore size to specific separations applications.

Solvent flux testing

To support the application of these membranes to solvent-based separations, permeation studies were performed, alternating feed solutions of DI water and IPA in dead-end cells. Some membranes may swell or deteriorate in solvents changing pore size or even degrading entirely. If the membrane is unaltered by the presence of a solvent, then in the absence of charge effects membrane flux will change by a factor equal to the ratio of the two permeates’ viscosities as seen in Eq. (1). This is borne out in flux studies shown in Fig. 4. The average pure water flux after steady state was reached was 43 LMH for the MTFC at 4 bar, while average IPA flux at the same pressure reduced to 21 LMH. Since the viscosity of IPA is 2.2 times that of water at 25 °C, this behavior fits the expected relation indicating the stability of the MTFC under use in IPA with no significant change in pore size.

Figure 4:

Permeance of MTFC composite membrane cycled between IPA and DI water under 4 bar pressurization. Viscosity corrected flux is shown here, indicating a good fit with expected behavior for a membrane unmodified by the presence of the organic solvent.

Rejection study

To more fully understand the modification of pore structure with the addition of metallic films, rejection studies were performed. By passing a solution containing a molecule of known size through the membrane and measuring if that molecule permeates the membrane or is rejected by the membrane, an idea of effective pore size may be determined. The lowest molecular weight at which a rejection of 90% is seen is given as the MWCO. This MWCO is used to judge a given membrane’s suitability for the application to a specific type of separation. Membrane rejection was calculated from permeate and retentate concentrations according to Eq. (3) [52], where $c_p$ is the concentration of solute in the permeate, $c_{\text{r}}$ is the concentration of solute in the retentate solution, and $R_i$ is the observed solute rejection for a given solute.

$$R_i=1-\frac{C_{\text{p}}}{C_{\text{r}}}.$$ (3)

Rejection studies have been performed on the pristine PSf UF membrane, the MTFC, and the npMTFC. Dextran solutions tagged with dyes were used as marker molecules, as well as sucrose for a low-end size test. Rejection values were calculated using Eq. (3) based on permeate and feed concentrations. [52] Results are summarized in Fig. 5 and given in detail as Table S2. In this study, PSf UF was found to have an MWCO of about 70 kDa. The MWCO is seen to be significantly decrease after the UF PSf is plasma cleaned and coated with a thin Ta layer to produce the MTFC. After this treatment, the MWCO is between 5 kDa which is almost completely rejected, and 342 Da sucrose which permeates the membrane. The rejection properties of the membrane are largely unchanged through the addition of the Mg/Pd alloy layer and subsequent dealloying steps. This is consistent with the hypothesis that the Ta film layer is the most important modifier of pore size for the composite. Effective diameter of the dextran compounds was determined from molecular weight using Eq. (4) [53]. Here, $r_{\text{p}}$is the hydrodynamic radius, and MW is the molecular weight of the dextran. Diameter of sucrose is based on hydrodynamic radii found in literature [54].

Figure 5:

Rejection study of composite membranes and different molecular weights of dye tagged dextrans and sucrose. Headspace of dead-end cell was pressurized at 4 bar with N₂. Diameters of given molecules are given in the key.

$$r_{\text{p}}=0.488{\text{MW}}^{0.437}.$$ (4)

Change in MWCO indicates a change in the pore size of the membrane. This is to be expected as the metal atoms that are deposited onto the membrane surface come in at an angle of approximately 45°. Some thickness of the film will be deposited inside the pore mouth, effectively reducing their size. This apparent reduction in pore size persists even after the addition of the Mg/Pd film and subsequent dealloying process. Rejection at this size scale serves to show the membrane has few imperfections, such as cracks in the film, as the water would flow preferentially through these relatively large channels reducing the rejection performance.

The minimal change in flux between MTFC and npMTFC composites suggests that the npPd layer does not significantly alter resistance to water flow in the composite. This matches expectations due to the presence of large pores in the bimodal pore structure. The npMTFC rejections of dextran molecules correspond closely to those of the MTFC, which aligns well with the hypothesis that the underlying Ta-coated UF PSf structure is the selective layer. However, sucrose rejection is higher for the npMTFC (40%) than for the MTFC (20%). Therefore, the nanoporous structure must have some influence on the flow of water through the membrane. It is likely that a fraction of the permeate passes through the smaller pores in the nanoporous structure which may reject small molecules while most of the flow likely passes through the larger pores.

Catalysis performance

Beyond separations applications, these composite membranes have the capability to catalyze reactions in the permeate as it passes through the membrane. The many small, well-distributed pores of the underlying UF PSf ensure good contact between the permeate and the catalyst structure anchored on top. Furthermore, the npPd film has a high surface area with many curved surfaces, which indicate it may function extremely well as a catalyst. The Pd here will be used for dechlorination of COCs in the presence of hydrogen. PCB-1 is used here as a model COC with some additional experiments performed using TCE to demonstrate a broader applicability of the catalyst to COCs in general. COCs are pollutants found in waste water streams and water supplies across the world which pose a health risk to nearby populations. These compounds are difficult and costly to remove using traditional means [34]. In the presence of H₂, Pd will act as a catalyst to remove the chlorine group from the compound reducing its toxicity and easing clean-up [35,38]. In this case, PCB-1 will be dechlorinated to form biphenyl.

Batch reactor dechlorination studies

Performance of the npPd was tested in a batch reactor configuration to study the performance of the Pd structure without the polymer backing. The npPd films were delaminated from the membrane by adding a pure Mg sacrificial layer between UF PSf membrane and Mg/Pd alloy film. The films delaminated during dealloying and were placed into a sealed dead-end cell for batch testing. The headspace was pressurized with pure H₂. Under hydrogen pressurization at 4 bar, the concentration of hydrogen in solution was 6.2 ppm according to Henry’s law [55]. The results of this study are shown in Fig. 6 for different catalyst loadings. As expected, PCB-1 concentration decreases with time and biphenyl is produced. For 100% H₂ utilization, 1 mole H₂ may dechlorinate 2 moles PCB-1. Therefore, a 5 ppm solution of PCB-1 solution requires less than 0.1 ppm of dissolved H₂ for complete dechlorination in a given volume. These batch experiments are performed at much higher H₂ concentrations, with additional H₂ continually supplied as dissolved H₂ is consumed.

Figure 6:

(a) Batch dechlorination study under 4 bar H₂ pressure with various loadings. (b) Detailed results of high loading batch test. Nanoporous Pd films were delaminated and suspended in PCB-1 solution in blanked dead-end cell. Cell was opened for sample retrieval at relevant time intervals.

This reaction may be approximated with a pseudo-first order reaction equation [38] as in Eq. (5), where $C$ is the concentration of PCB-1, $t$ is the time in minutes, and $k_{\text{obs}}$ is the measured reaction rate/h. Fitting the linear portion of the batch testing results with different loading and normalizing the rate with respect to catalyst loading gives an average $k_{\text{mass-}}=1.3\pm 0.3/\text{h}/\text{mg}$.

$$\frac{\text{d}C}{\text{d}t}=-k_{\text{obs}}C.$$ (5)

Permeation mode dechlorination of PCB-1

To take advantage of the favorable mass transfer of species to the catalyst surface that comes with the convective flow of the permeate through the membrane supported catalyst, the composite membrane was tested for dechlorination in dead-end permeation format. Headspace of the dead-end cell was pressurized with nitrogen for a control run, and then hydrogen gas. These raw results are shown in Fig. S7. As expected, no biphenyl was produced with nitrogen pressurization, since there was no dissolved hydrogen present. In Fig. 7, hydrogen pressurization results are given, with PCB-1 dechlorination plotted against dissolved H₂ concentration. When the headspace was pressurized with hydrogen, 50% of PCB-1 was removed from the permeate at 1.4 bar H₂ pressurization. 72% of PCB-1 was removed from the permeate at both 4 and 8 bar H₂ pressurization, with 60 and 65% of that concentration being detected as biphenyl, respectively.

Figure 7:

Permeate concentration of PCB-1 and biphenyl plotted against dissolved H₂ concentration in solution. Includes results from both pure H₂ headspace pressurization (solid symbol)) and 5/95% H₂/Ar mixture (open symbol). For zero hydrogen, there was no conversion.

Since this reaction may be modeled as a pseudo-first order reaction, it is expected that more PCB-1 will be dechlorinated, and more biphenyl will be produced, at higher retention times and with greater dissolved H₂ concentration. This expected relation is complicated due to the relationship between retention time and dissolved H₂ concentration. Increased H₂ pressure will increase flux, which decreases retention time and expected dechlorination. In opposition to this effect is Henry’s law which states increased H₂ pressure will increase dissolved H₂ concentration. Essentially, as retention time decreases the dissolved H₂ concentration increases. Of these competing effects, it appears that dissolved H₂ concentration is most important in this pressure/retention time regime as the data show biphenyl production increases with increasing H₂ concentration.

To further investigate the relation between H₂ concentration, retention time, and dechlorination performance, a series of experiments with a gas mixture below the flammability limit of hydrogen concentration was also performed. In these experiments, a 5/95 H₂/Ar gas mixture was used to pressurize the headspace. Pressures of 17.2 and 27.6 bar were used, corresponding to H₂ concentrations of 1.3 and 2.1 ppm, respectively, in solution. This was determined according to the H₂ partial pressure and Henry’s law; the raw data are reported in Fig. S8. Data for experiments with both pure H₂ and 5/95 H₂/Ar mixtures are combined in Fig. 7. Two runs were performed at equivalent H₂ partial pressures and therefore dissolved H₂ concentrations, but at different headspace pressures due to the gas composition. One run was performed at 1.4 bar of pure H₂ and the other at 27 bar of 5/95 gas mixture. This allowed an independent check of the influence of retention time on the reaction rate. The dechlorination results for the two tests were very similar, despite a 10-fold change in retention time. This indicates that the most important variable, at least in this pressure/retention time regime, is H₂ concentration rather than retention time. These gas mixture experiments also serve to demonstrate the viability of this catalysis for application as a 5% mixture of H₂ gas is easier and safer to employ since it is below the flammability limit for the gas.

Permeation mode dechlorination of TCE

To ensure the general application of these Pd-based npMTFC membranes to COC dechlorination, another model COC was chosen for dechlorination, TCE. TCE dechlorination was measured using chloride probe analysis. For each mole of TCE which is completely dechlorinated, three moles of chloride are produced in solution. Some TCE molecules may only be partially dechlorinated, with the removal of just one or two chlorine groups. 100% dechlorination then, would correspond to a chloride molar concentration equal to three times the starting TCE concentration. In permeation mode testing with hydrogen pressurization, chloride concentration increased drastically compared to the control run performed with N₂ gas in the headspace which showed no chloride generation. Using this balance, per cent TCE dechlorinated in the permeate may be determined from chloride measurements, results are shown in Fig. S9. At 4 bar H₂ pressure, 58% of TCE was dechlorinated in the permeate, while at 8 bar, only 45% was dechlorinated in the permeate. Residence time for the 8 bar run was 2.8 ms while retention time was 4.8 ms for the 4 bar run. The same trade-off in dissolved H₂ concentration is found here as for the PCB-1 dechlorination runs, with a dissolved H₂ concentration of 6.2 ppm at 4 bar and 12.3 ppm at 8 bar. Significantly, more dechlorination was observed in the retentate for TCE than for PCB-1 tests, indicating a greater portion of the dechlorination took place before the solution permeated the membrane structure.

Overall, these composite membranes with incorporated nanoporous metallic structures have shown themselves to be effective catalysts for PCB-1 and TCE dechlorination. The membranes have functioned well in both batch and permeation arrangements, with both pure hydrogen gas and with a hydrogen/argon gas mixture. With the short residence time that accompanies their low thickness, the npPd films have still dechlorinated significant amounts of COCs in solution, reinforcing the nature of the structure as an effective catalyst template.

Conclusions

This research represents a progression and expansion of previous research in incorporating nanoporous metal films into reactive membranes [17]. A new Mg/Pd alloy system has been developed to produce the desired catalyst structure. This is a significant improvement over previous research, where Fe/Pd films were dealloyed using sulfuric acid. Dealloying with water is more cost-effective, safer, and more environmental friendly than processes requiring concentrated acid. A purer npPd catalyst is also produced as a result of this new technique compared to previous methods. The other primary improvement of the gentler dealloying method is that the underlying pore structure of the polymer membrane is undamaged. This preserves the improved separations properties of the underlying membrane and Ta layer, allowing flux modifications for separation integration or to control residence time.

Furthermore, this research has shown that a plasma treatment and the addition of a thin metallic layer of only 10 nm thickness will change the membrane’s separation properties significantly. The MWCO of the UF PSf precursor membrane was about 70 kDa, while after the addition of the thin films, the MWCO of the composite is reduced to below 5 kDa. These composite membranes were also stable under solvent permeation as indicated by experiments with IPA. Cross-flow testing further reinforced the stability of these composites under different flow regimes. The deposition of a thin metal film allowed for a significant reduction of substrate membrane pore size and enhancement of separations capabilities. This technique could be used with a variety of membrane substrates with little alteration, allowing tuning of pore size and separations capabilities compatible with separations in harsh environments. With further development, these types of membranes could be applied to separations in pharmaceutical fields where solvent separations are particularly impactful. The flexible nature of the MTFC composite allows for these metal films to be applied in preexisting systems designed for polymer membranes, which are already in use commercially.

The npMTFC membrane produced with a npPd film top layer was characterized with SEM and TEM and shown to be well adhered to the membrane with a bimodal pore-ligament structure suitable to function as a catalyst. These nanostructured films were found to catalyze dechlorination of PCB-1, effectively removing 72% of PCB-1 from solution in permeation mode with a residence time of only 2 ms. Due to the very thin nature of these films and their porosity, very little Pd is required to manufacture this membrane. The amount of Pd used to produce 1000 cm² of this membrane is less than 30 mg. Using a rough cost of $70 g⁻¹ for Pd, this works out to less than $2.10 in added materials cost. There are additional fabrication costs associated with this synthesis technique but producing catalytic membranes with nanostructured films is a technique to maximize the effectiveness of an expensive material. In this research, the nanoporous top layer used consisted of Pd, but nanoporous metals can be made of a variety of metals including almost all noble metals easily. This allows the application of these structures to a range of reactions and processes.

Methods

Membrane fabrication

Ultrafiltration membranes were used as substrate materials for thin-film production. UF PSf membranes from Solecta were chosen due to their high porosity and solvent resistance. An additional series of dead-end cell flux tests included in the supplement used UF polyethersulfone (PES) membranes from Solecta as a basis for composite production. An ORION magnetron sputtering system (AJA International) was used for metal depositions with an argon working environment. A series of experiments were performed over different working pressure ranges and substrate biasing conditions to optimize as-deposited film structure and adhesion to the membrane substrate. An optimal working pressure (2.5 mtorr), substrate bias (10 W RF), and interlayer thickness (10 nm) were determined. Interlayers are commonly used in magnetron sputtering to help the thicker film adhere strongly and uniformly to the substrate material. Such is the case with this work, but additionally the thin film modifies pore size of the membrane as is discussed in the Results section. Tantalum was chosen as an interlayer material due to its ductility and high resistance to corrosion. After deposition of the Ta interlayer, a thicker layer of Mg/ Pd alloy was co-deposited onto the membrane. This involved simultaneous use of a Mg target (AJA Int. Purity: 99.95+%) and a Pd target (AJA Int. Purity: 99.98%) while a bias was applied to the substrate at 10 W RF. A precursor film thickness of 100 nm was used for these experiments with a composition of 75 at.% Mg and 25 at.% Pd as determined by subsequent EDS measurements. Deposition rate measurements were performed for Ta and for Mg/Pd co-sputtering by deposition of films of over 100 nm thickness using a defined deposition time period at a set target bias onto single crystal silicon wafers. Film thickness was measured via SEM cross-sectional imaging and normalized with time to give a deposition rate. Deposition time was then scaled to produce films of desired thickness. Film composition was controlled by regulating the power applied to the relevant target materials. This varied over the lifetime of the targets but was nominally kept to 140 W bias for Mg and 30 W bias for Pd. Total time of exposure to plasma for the membrane was consistently less than 10 min for all depositions. Temperature of the substrate was monitored via thermocouple and no increase was detected during deposition.

After deposition of the alloy film with correct composition, the Mg was selectively etched from the film by submerging the film in DI water. Films were submerged for 1 h and shaken to dealloy the entire film thickness. Film thickness decreased to about 60 nm after dealloying. This shows a significant amount of film contraction, which is common in many dealloying systems. Almost all Mg is removed in this process with the final film composition containing <5 at.% Mg as determined by EDS in a transmission electron microscope (TEM).

Materials characterization procedures

Film structures were characterized using electron microscopy with FIB cross-sectioning preparation. An FEI Helios Nanolab 660 was primarily used for imaging. EDS composition measurements were also taken using this instrument at an e-beam accelerating potential of 10 kV. Some additional images, as well as EDS measurements, were taken using an FEI Quanta 250. Supplementary SEM images for film optimization were taken with a Hitachi S-4300 SEM. A lamella of the npMTFC membrane was prepared in the FEI Helios FIB–SEM using the J-cut technique for further imaging and analysis in TEM [56]. EDS was performed using an Oxford EDX detector. An FEI Talos F200X G2 TEM was used for imaging of this sample with Super-X EDS for compositional analysis. Surface chemical analysis was performed via XPS using a Thermo K-alpha XPS system.

Membrane performance studies

The properties of the composite membranes were tested in a dead-end permeation flow cell (GE Osmonics Sepa ST). This allowed testing of the flat sheet membrane samples, with a sample area of 13.2 cm². These flow cells allow for the membrane sample to be placed in the bottom, supported by a sintered stainless steel disc. Water or feed solution is loaded above the membrane and the headspace is pressurized with a chosen gas. As the feed solution permeates the membrane, the permeate is collected and gravimetrically tracked for flux measurement. An initial compaction step during permeation testing was not included in results, steady state membrane fluxes have been reported here. Permeates were also tested for other purposes, such as membrane separations or chemical degradation testing. One stability study for the MTFC membranes was performed using a Sterlitech cross-flow cell which required a rectangular membrane with an active filtration area of 20.6 cm².

To understand the separation performance of these composite membranes, and how the performance changed through successive steps of fabrication, a series of rejection tests were performed. Dextran of various molecular weights tagged with markers was used for MWCO determination. Dextran tagged with blue dye was used for 5 kDa (Sigma Lot BCBS3284, St. Louis, MO) and 10 kDa (Sigma BCBT6139) weights, FITC and TRITC labeled dextrans were used for 40 kDa (Chondrex 180209, Redmond, Washington) and 70 kDa (Chondrex 180229) weights, respectively. Sucrose (Fisher Lot 136421, Waltham, Massachusetts) was used for a lower end test at 342 Da. The dye solution concentration was determined using a plate reader (Biotek Synergy H1 Hybrid Reader, Winooski, Vermont), while sucrose concentrations were determined using a total organic carbon analyzer (TOC-5000A Shimadzu, Kyoto, Japan) against standard curves.

Dechlorination performance of the npMTFC’s was tested using PCB-1 (Ultra Scientific RPC-006, Santa Clara, California) as a model-chlorinated organic compound. PCB-1 solutions (5 ppm) used for dechlorination were made in 50/ 50 water to ethanol mixtures as solvents to improve PCB-1 solubility. Experiments were performed with N₂, H₂, and Ar/H₂ mixtures of 95%/5% composition as headspace gasses in the permeation cell. Samples were taken and extracted in hexanes for further analysis of PCB-1 and biphenyl concentration. Extracts were tested via a CP-3800 gas chromatograph (GC) with a Varian Saturn 2200 mass spectrometer (MS) and an Agilent DB-5 ms column (Santa Clara, California). Samples were tested against a standard curve with a lower limit of 0.1 ppm with biphenyl-d₁₀ used as an internal standard. TCE dechlorination was also tested using an 80/20 water to ethanol solution at a concentration of 30 ppm TCE (Aldrich HA00250AA). TCE dechlorination was determined by the measurement of chloride ion concentration in solution. For each mole of TCE completely dechlorinated, 3 moles of chloride was formed. Chloride concentration was measured using a chloride ion selective electrode (Thermo Fisher Scientific 9617BNWP, Waltham, Massachusetts) against a prepared standard curve.