Synthesis of C-Plane Oriented Hexagonal Tungsten Oxide Membranes on Tubular Substrates and Their Acetic Acid/Water Separation Performances

Hexagonal tungsten oxide (h-WO3) membrane is a novel candidate for dehydration of acetic acid (CH3COOH)/water mixtures owing to its molecular sieving property and acidic resistance. Meanwhile, c-plane orientation is an important factor for h-WO3 membranes because the pores of h-WO3 run along its c-axis. However, so far, high c-plane orientation has not been successful on tubular substrates. Here, the effect of synthesis conditions of h-WO3 membranes on tubular substrates against c-plane orientation and CH3COOH/water separation performance are investigated. The h-WO3 membranes were prepared by hydrothermal synthesis from a precursor sol containing various amounts of sodium tungstate (Na2WO4) in the presence of tubular substrates with seeds embedded on their outside surface. The seeding method and the amount of Na2WO4 in the precursor sol significantly affected both crystal orientation and densification of the membrane. A precursor sol with appropriate amounts of Na2WO4 was essential to simultaneously satisfy high c-plane orientation and densification of the membrane while excess Na2WO4 drastically decreased the degree of c-plane orientation. A highly c-plane oriented h-WO3 membrane was successfully obtained under the optimized condition, which exhibited a maximum separation factor of 40.0 and a water permeance of 1.53 × 10−7 mol·m−2·s−1·Pa−1 in a 90:10 wt % CH3COOH/water mixture. The water permeance approximately doubled compared to the previous report, possibly owing to the significantly higher degree of c-plane orientation. Furthermore, it was found that its separation ability can be maintained while stored in 90:10 wt % CH3COOH/water mixture with pH < 0 for more than 500 h.


Introduction
Dehydration of organic solvents is one of the most important technologies in chemical processes. Currently, the primary technology for dehydration of organic solvents is distillation; however, distillation-based separations are energy consuming for azeotropic and close-boiling-point mixtures [1]. Therefore, an energy-efficient separation technology is desired to reduce the environmental load of these processes. Membrane-based separations were purchased from NGK Filtech Co., Ltd., Chigasaki, Japan The former tubes were cut into pieces with lengths of 10 mm using a cutting machine and were adopted as the tubular substrate for the membranes. The tubular substrates were completely dried at 343 K after removing contaminants on the surface by carrying out sonication in distilled water and ethanol (purity: 95%) for 10 min, respectively. Sodium tungstate dihydrate (Na 2 WO 4 ·2H 2 O, purity: >98%), CH 3 COOH (purity: >99%) and distilled water were used for membrane synthesis. CH 3 COOH and distilled water were also used in the PV experiments. All chemical reagents used for membrane synthesis and performance tests were purchased from Nacalai Tesque Inc., Kyoto, Japan and were used as-received. Distilled water was produced in our laboratory using a water distillation unit (RFD240NC, Toyo Seisakusho Kaisha, Ltd., Kashiwa, Japan).

Preparation of Seeds and Seeding on Tubular Substrates
In total, 36.75 g of CH 3 COOH was rapidly poured into an Na 2 WO 4 solution of 27.30 g containing 7.0 mmol of Na 2 WO 4 , under continuous agitation. Precipitates appeared immediately after CH 3 COOH addition, forming a suspension of seeds. Next, seeding was performed by dipping the tubular substrates into the aforementioned suspension for 5 min and subsequently dried at room temperature and 343 K for 10 min, respectively. This seeding process was repeated three times to ensure attachment of seeds on the surface of tubular substrates. In this study, two ways of seeding were tested. One is directly dipping the tubular substrates into the seed suspension as we have done in our previous study. Another is dipped into the seed suspension with both ends of the tubular substrates sealed using silicon stoppers to avoid attachment of seeds on the inner wall of the tubular substrates. The two seeding methods are illustrated in Figure 1.

Materials
Asymmetric porous -alumina tubes with an outer diameter of 12 mm, wall thickness of 1.5 mm, length of 500 mm, overall porosity of ~35% and outer surface pore size of 0.1 m were purchased from NGK Filtech Co., Ltd., Chigasaki, Japan The former tubes  were cut into pieces with lengths of 10 mm using a cutting machine and were adopted as the tubular substrate for the membranes. The tubular substrates were completely dried at 343 K after removing contaminants on the surface by carrying out sonication in distilled water and ethanol (purity: 95%) for 10 min, respectively. Sodium tungstate dihydrate (Na2WO4•2H2O, purity: >98%), CH3COOH (purity: >99%) and distilled water were used for membrane synthesis. CH3COOH and distilled water were also used in the PV experiments. All chemical reagents used for membrane synthesis and performance tests were purchased from Nacalai Tesque Inc., Kyoto, Japan and were used as-received. Distilled water was produced in our laboratory using a water distillation unit (RFD240NC, Toyo Seisakusho Kaisha, Ltd., Kashiwa, Japan).

Preparation of Seeds and Seeding on Tubular Substrates
In total, 36.75 g of CH3COOH was rapidly poured into an Na2WO4 solution of 27.30 g containing 7.0 mmol of Na2WO4, under continuous agitation. Precipitates appeared immediately after CH3COOH addition, forming a suspension of seeds. Next, seeding was performed by dipping the tubular substrates into the aforementioned suspension for 5 min and subsequently dried at room temperature and 343 K for 10 min, respectively. This seeding process was repeated three times to ensure attachment of seeds on the surface of tubular substrates. In this study, two ways of seeding were tested. One is directly dipping the tubular substrates into the seed suspension as we have done in our previous study. Another is dipped into the seed suspension with both ends of the tubular substrates sealed using silicon stoppers to avoid attachment of seeds on the inner wall of the tubular substrates. The two seeding methods are illustrated in Figure 1.

Membrane Synthesis
The membranes were synthesized on the outer surface of the tubular substrates. Membrane synthesis was carried out using precursor sols prepared by mixing CH3COOH with Na2WO4 solutions with a variety of concentrations as follows. Na2WO4 solutions were made by completely dissolving Na2WO4•2H2O in distilled water. To this solution, CH3COOH was added dropwise under continuous agitation until the pH reached 2 ± 0.2 and was continuously mixed for an additional 5 min. The resulting precursor sols were approximately 60 mL and the amount of Na2WO4 in the sols were 3.5, 5.0, 6.0 or 7.0 mmol, respectively. Then, the seeded tubular substrates prepared in Section 2.2 were fixed on a handmade polytetrafluoroethylene holder and placed in polytetrafluoroethylene vessels.

Membrane Synthesis
The membranes were synthesized on the outer surface of the tubular substrates. Membrane synthesis was carried out using precursor sols prepared by mixing CH 3 COOH with Na 2 WO 4 solutions with a variety of concentrations as follows. Na 2 WO 4 solutions were made by completely dissolving Na 2 WO 4 ·2H 2 O in distilled water. To this solution, CH 3 COOH was added dropwise under continuous agitation until the pH reached 2 ± 0.2 and was continuously mixed for an additional 5 min. The resulting precursor sols were approximately 60 mL and the amount of Na 2 WO 4 in the sols were 3.5, 5.0, 6.0 or 7.0 mmol, respectively. Then, the seeded tubular substrates prepared in Section 2.2 were fixed on a handmade polytetrafluoroethylene holder and placed in polytetrafluoroethylene vessels. The aforementioned precursor sol was slowly poured into the vessels. The polytetrafluoroethylene vessels were sealed in a stainless-steel autoclave (HU-100, SAN-AI Kagaku Co. Ltd., Nagoya, Japan) and transferred to a convection oven (MOV-450, AS ONE Corporation, Osaka, Japan) in order to perform hydrothermal synthesis at 453 K for 24 h. After the oven was cooled down to room temperature, the membranes were rinsed by sonicated in distilled water and ethanol several times to clean the membrane surface and to remove the precursor sol remaining inside the tubular substrate. Finally, the membranes were dried at 343 K overnight and were stored at room temperature.

Characterization of Seeds and Membranes
The crystalline phases of the seeds and membranes were characterized using X-ray diffraction (XRD, Ultima IV, Rigaku Corporation, Akishima, Japan) equipped with Cu-Kα radiation (λ = 1.54050 Å). The degrees of c-plane orientation of the membranes were evaluated using the peak intensities of (001) plane and (200) plane of h-WO 3 , which are the representative c-plane and a-plane, respectively. To quantitatively compare the degree of c-plane orientation, c-plane orientation index; R, defined by the following Equation (1) was adopted. R = (I m_001 /I m_200 )/(I p_001 /I p_200 ) Here, I m_001 and I m_200 are the peak intensities of (001) and (200) planes of h-WO 3 obtained from the membranes and I p_001 and I p_200 are the peak intensities of (001) and (200) plane registered in the standard powder diffractogram (ICDD #01-075-2187). R > 1 indicates that the membrane exhibits c-plane orientation and a larger R corresponds to a higher degree of c-plane orientation. Meanwhile, R = 1 or R < 1 indicate that the membranes exhibit random (equivalent to standard powder data) or a-plane orientation, respectively. The morphologies of the seed, surfaces and cross-sections of the membranes were observed using a scanning electron microscope (SEM, S-4800, Hitachi High-Tech Corporation, Tokyo, Japan).

Evaluation of Membrane Performance
The PV performances of the membranes were considered from the separation performance against 90:10 wt % CH 3 COOH/water mixtures. The membrane was soaked in the 90:10 wt % CH 3 COOH/water mixtures and the testing temperature and time were fixed to be 353 K and 10 h. The permeate collected within the first hour was discarded to make sure the performance stabilized and the permeate thereafter was collected for analysis. The change in feed concentration by discarding the permeate collected in the first hour was less than 0.005%. Separation factor; α, was calculated using following Equation (2): Here, X 0 and X 1 are the mass fractions of water in the feed and in the permeate and Y 0 and Y 1 are the mass fraction of CH 3 COOH in feed and in the permeate, respectively.
Additionally, the total flux and permeability of each component was evaluated. The total flux was the weight of the permeate per unit membrane area collected in one hour. The permeability of a particular component (P c ) was calculated using the following Equation (3): Here, m is the amount of the collected permeate (g), x c is the mass fraction of the target component (g/g), M c is the molecular weight of the target component and ∆p c is the partial pressure differences (Pa) of the target component across the membrane, respectively. The CH 3 COOH concentrations of the feed and the permeate were measured using a reflectometer (PAL-RI, Atago Co. Ltd., Tokyo, Japan).

Characterization of the Seed
The seed size against the pore size of the porous substrate is reported to be an important parameter in membrane synthesis. If the seed is too small compared to the substrate pores, the seed will penetrate into the substrate and will not remain on its surface [27]. In contrast, if it is too large, the seed layer will become inhomogeneous and defective [28]. The appropriate seed size is considered to be smaller than 1.5 m to obtain a homogeneous surface [27] but larger than the substrate pores. Thus, the seeds and the tubular substrates were characterized. The SEM image, particle size distribution, and XRD diffractogram of the seeds are shown in Figure 2. The morphology of the seed was isotropic irregular shape ( Figure 2a) with an average diameter of 253 nm. Approximately 90% of the seed size was between 100 and 500 nm, which is expected to be effective to use as seeds for substrate with a nominal outer surface pore size of 0.1 m (100 nm). The seed had an amorphous structure according to XRD as shown in Figure 2c.

Characterization of the Seed
The seed size against the pore size of the porous substrate is reported to be an important parameter in membrane synthesis. If the seed is too small compared to the substrate pores, the seed will penetrate into the substrate and will not remain on its surface [27]. In contrast, if it is too large, the seed layer will become inhomogeneous and defective [28]. The appropriate seed size is considered to be smaller than 1.5 m to obtain a homogeneous surface [27] but larger than the substrate pores. Thus, the seeds and the tubular substrates were characterized. The SEM image, particle size distribution, and XRD diffractogram of the seeds are shown in Figure 2. The morphology of the seed was isotropic irregular shape ( Figure 2a) with an average diameter of 253 nm. Approximately 90% of the seed size was between 100 and 500 nm, which is expected to be effective to use as seeds for substrate with a nominal outer surface pore size of 0.1 m (100 nm). The seed had an amorphous structure according to XRD as shown in Figure 2c. Cross-sectional and surface SEM images of the tubular substrates are shown in Figure  3. The tubular substrates consist of three different layers; fine-grained, intermediategrained and coarse-grained layers, as shown in Figure 3a. The fine-grained layer comprising the outer surface of the tubular substrates was approximately 100 m thick and con- sist of well packed alumina particles of several hundred nanometers as shown in Figure  3b. The spaces between the particles were mostly smaller than the seeds and no large defects could be found. This observation also supports that the relationships between seed size and substrate pores were adequate for effective seeding.  Cross-sectional and surface SEM images of the tubular substrates are shown in Figure 3. The tubular substrates consist of three different layers; fine-grained, intermediate-grained and coarse-grained layers, as shown in Figure 3a. The fine-grained layer comprising the outer surface of the tubular substrates was approximately 100 µm thick and consist of well packed alumina particles of several hundred nanometers as shown in Figure 3b. The spaces between the particles were mostly smaller than the seeds and no large defects could be found. This observation also supports that the relationships between seed size and substrate pores were adequate for effective seeding.

Characterization of the Seed
The seed size against the pore size of the porous substrate is reported to be an important parameter in membrane synthesis. If the seed is too small compared to the substrate pores, the seed will penetrate into the substrate and will not remain on its surface [27]. In contrast, if it is too large, the seed layer will become inhomogeneous and defective [28]. The appropriate seed size is considered to be smaller than 1.5 m to obtain a homogeneous surface [27] but larger than the substrate pores. Thus, the seeds and the tubular substrates were characterized. The SEM image, particle size distribution, and XRD diffractogram of the seeds are shown in Figure 2. The morphology of the seed was isotropic irregular shape ( Figure 2a) with an average diameter of 253 nm. Approximately 90% of the seed size was between 100 and 500 nm, which is expected to be effective to use as seeds for substrate with a nominal outer surface pore size of 0.1 m (100 nm). The seed had an amorphous structure according to XRD as shown in Figure 2c. Cross-sectional and surface SEM images of the tubular substrates are shown in Figure  3. The tubular substrates consist of three different layers; fine-grained, intermediategrained and coarse-grained layers, as shown in Figure 3a. The fine-grained layer comprising the outer surface of the tubular substrates was approximately 100 m thick and con- sist of well packed alumina particles of several hundred nanometers as shown in Figure  3b. The spaces between the particles were mostly smaller than the seeds and no large defects could be found. This observation also supports that the relationships between seed size and substrate pores were adequate for effective seeding.   Figure 4 shows the surface SEM images of the tubular substrate before and after seeding with and without the silicon stoppers at both ends of the tubular substrate. Seeds almost cover the surface after seeding in both cases; however, the attached seed layer seemed to be thicker when using silicon stoppers. For the sample seeded without silicon stoppers, some areas showed morphology similar to the substrate surface, implying that the seed layer was relatively thin. Meanwhile, for the sample with silicon stoppers, although some cracks were observed, the surface morphology completely changed possibly due to a thicker seed layer formation. Figure 5a-c show the XRD diffractogram of the tubular substrate before seeding and after seeding without and with silicon stoppers. After seeding, a broad peak was detected between 25 to 35 degrees in both cases. This is expected to be derived from the amorphous h-WO 3 seed layers, considering the XRD diffractogram of the seed shown in Figure 2c. The intensity of the broad peak was slightly stronger for the sample seeded in the presence of silicon stoppers, implying thicker seed layer formation as anticipated from the SEM image. This might be owing to the air remaining inside the tube when both ends are sealed with silicon stoppers, leading to suction and filtration of the seed solution into the substrate. Figure 4 shows the surface SEM images of the tubular substrate before and after seeding with and without the silicon stoppers at both ends of the tubular substrate. Seeds almost cover the surface after seeding in both cases; however, the attached seed layer seemed to be thicker when using silicon stoppers. For the sample seeded without silicon stoppers, some areas showed morphology similar to the substrate surface, implying that the seed layer was relatively thin. Meanwhile, for the sample with silicon stoppers, although some cracks were observed, the surface morphology completely changed possibly due to a thicker seed layer formation. Figure 5a to Figure 5c show the XRD diffractogram of the tubular substrate before seeding and after seeding without and with silicon stoppers. After seeding, a broad peak was detected between 25 to 35 degrees in both cases. This is expected to be derived from the amorphous h-WO3 seed layers, considering the XRD diffractogram of the seed shown in Figure 2c. The intensity of the broad peak was slightly stronger for the sample seeded in the presence of silicon stoppers, implying thicker seed layer formation as anticipated from the SEM image. This might be owing to the air remaining inside the tube when both ends are sealed with silicon stoppers, leading to suction and filtration of the seed solution into the substrate.    Figure 4 shows the surface SEM images of the tubular substrate before and after seeding with and without the silicon stoppers at both ends of the tubular substrate. Seeds almost cover the surface after seeding in both cases; however, the attached seed layer seemed to be thicker when using silicon stoppers. For the sample seeded without silicon stoppers, some areas showed morphology similar to the substrate surface, implying that the seed layer was relatively thin. Meanwhile, for the sample with silicon stoppers, although some cracks were observed, the surface morphology completely changed possibly due to a thicker seed layer formation. Figure 5a to Figure 5c show the XRD diffractogram of the tubular substrate before seeding and after seeding without and with silicon stoppers. After seeding, a broad peak was detected between 25 to 35 degrees in both cases. This is expected to be derived from the amorphous h-WO3 seed layers, considering the XRD diffractogram of the seed shown in Figure 2c. The intensity of the broad peak was slightly stronger for the sample seeded in the presence of silicon stoppers, implying thicker seed layer formation as anticipated from the SEM image. This might be owing to the air remaining inside the tube when both ends are sealed with silicon stoppers, leading to suction and filtration of the seed solution into the substrate.   The effect of seeding was further investigated by synthesizing the h-WO 3 membranes without and with seeding using precursor sols containing 3.5 mmol of Na 2 WO 4 . Figures 6  and 7 show the SEM images and the XRD diffractograms of the membranes synthesized at 453 K for 24 h without and with seeding. When using the tubular substrates without seeding, although rod-like crystals were deposited on the substrate surface, the thickness of membrane was uneven (Figure 6a-1) and the substrate surface was not completely covered ( Figure 6a-2). The substrate was clearly visible in the gaps between the crystals at high magnification as shown in Figure 6a-3, indicating that the densification of the membrane was insufficient. Moreover, the rod-like crystals seem to be mainly tilted parallel to the substrate surface, implying a rather a-plane orientation of crystals. This was further confirmed from its XRD diffractogram showing a highest peak from reflection of a-plane of h-WO 3 as shown in Figure 7a. This may be due to the less nucleation occurring on the substrate surface in the absence of seeds. Details will be discussed afterwards comparing the results with seeding.

Effect of Seeding on Tubular Substrates
The effect of seeding was further investigated by synthesizing the h-WO3 membranes without and with seeding using precursor sols containing 3.5 mmol of Na2WO4. Figures 6  and 7 show the SEM images and the XRD diffractograms of the membranes synthesized at 453 K for 24 h without and with seeding. When using the tubular substrates without seeding, although rod-like crystals were deposited on the substrate surface, the thickness of membrane was uneven (Figure 6a-1) and the substrate surface was not completely covered (Figure 6a-2). The substrate was clearly visible in the gaps between the crystals at high magnification as shown in Figure 6a-3, indicating that the densification of the membrane was insufficient. Moreover, the rod-like crystals seem to be mainly tilted parallel to the substrate surface, implying a rather a-plane orientation of crystals. This was further confirmed from its XRD diffractogram showing a highest peak from reflection of a-plane of h-WO3 as shown in Figure 7a. This may be due to the less nucleation occurring on the substrate surface in the absence of seeds. Details will be discussed afterwards comparing the results with seeding. Meanwhile, when using the tubular substrates with seeding, dense membranes with thickness of more than 10 m were formed and rod  -like crystals seemed to well cover the surface of the substrate regardless of the use of silicon stoppers, as shown in Figure 6b,c, absence or presence of silicon stoppers. Seeding without silicon stoppers was found to slightly improve the c-plane orientation of the membrane; however, the c-plane orientation index is still similar to the value of the standard powder diffraction, meaning that cplane orientation was still weak. In contrast, the c-plane orientation index increased approximately ten times when substrates were seeded with silicon stoppers, clearly evidencing the effectiveness of sealing both ends of the tubular substrate with the silicon stoppers. This improvement may be owing to the thicker seed layer formed on the substrate surface.  Meanwhile, when using the tubular substrates with seeding, dense membranes with thickness of more than 10 µm were formed and rod-like crystals seemed to well cover the surface of the substrate regardless of the use of silicon stoppers, as shown in Figure 6b,c, respectively. This indicates that the seeding was effective to increase the homogeneity of the membrane and promote their densification. In contrast, the appearances of the membrane surfaces were quite different when using substrates seeded without and with the silicon stoppers. The rod-like crystals stood more vertically to the substrate surface when membranes were synthesized using tubular substrates seeded in the presence of silicon stoppers. According the XRD diffractograms of the synthesized membranes, the reflection from (200) plane; a-plane of h-WO 3 drastically decreased while the reflection from (001) plane; c-plane of h-WO 3 drastically increased when employing tubular substrates seeded using silicon stoppers (Figure 7c). This means that the seeding method significantly affects the crystal orientation of the h-WO 3 membranes. Figure 8 shows the c-plane orientation index; R of the membranes synthesized using substrates without and with seeding in the absence or presence of silicon stoppers. Seeding without silicon stoppers was found to slightly improve the c-plane orientation of the membrane; however, the c-plane orientation index is still similar to the value of the standard powder diffraction, meaning that c-plane orientation was still weak. In contrast, the c-plane orientation index increased approximately ten times when substrates were seeded with silicon stoppers, clearly evidencing the effectiveness of sealing both ends of the tubular substrate with the silicon stoppers. This improvement may be owing to the thicker seed layer formed on the substrate surface. The mechanism of how the seeding and its amount effected the crystal orientation of the membrane may be explained as follows. When no seeds existed on the surface of the tubular substrate, the number of nuclei that nucleated on the substrate surface during the initial stage of hydrothermal synthesis should be small. Thus, the open space between each nuclei shall be large, leaving enough area for the nuclei to grow freely. This must have allowed crystal growth of h-WO3 crystals in random directions as shown in Figure  9a. When a thin seed layer exists, the nucleation should be slightly promoted and the number of nuclei crystallized on the substrate surface shall increase. Thus, the space between the nuclei shall become smaller and the growth of the h-WO3 pillars shall be disturbed by the neighboring crystals. Therefore, spatially limited growth must have partially took place, as shown in Figure 9b, resulting in a slight c-plane orientation. Furthermore, in the case of thicker seed layer, h-WO3 may be more likely to crystallize near the substrate surface during hydrothermal synthesis. The growth of the h-WO3 crystals comprising the membrane is almost limited in two dimensions due to the substrate, and neighboring crystals spatially limit the growth of rod-like crystals parallel to the substrate surface. Therefore, the crystals may have grown vertically to the substrate surface as shown in Figure 9c. In addition, because the seeds did not attach on the inner walls of the tubular substrate, the tungsten source in the precursor sol may have been selectively consumed at the outer surface of the tubular substrate, leading to increase in crystallization of neighboring crystals. This must have led to membrane formation with c-axes oriented vertical to the surface of the underlying substrate. Silicon stopper was applied hereafter to ensure efficient seeding, to prevent seeding on the inner wall of the tube, and to promote the cplane orientation of the final membrane. The mechanism of how the seeding and its amount effected the crystal orientation of the membrane may be explained as follows. When no seeds existed on the surface of the tubular substrate, the number of nuclei that nucleated on the substrate surface during the initial stage of hydrothermal synthesis should be small. Thus, the open space between each nuclei shall be large, leaving enough area for the nuclei to grow freely. This must have allowed crystal growth of h-WO 3 crystals in random directions as shown in Figure 9a. When a thin seed layer exists, the nucleation should be slightly promoted and the number of nuclei crystallized on the substrate surface shall increase. Thus, the space between the nuclei shall become smaller and the growth of the h-WO 3 pillars shall be disturbed by the neighboring crystals. Therefore, spatially limited growth must have partially took place, as shown in Figure 9b, resulting in a slight c-plane orientation. Furthermore, in the case of thicker seed layer, h-WO 3 may be more likely to crystallize near the substrate surface during hydrothermal synthesis. The growth of the h-WO 3 crystals comprising the membrane is almost limited in two dimensions due to the substrate, and neighboring crystals spatially limit the growth of rod-like crystals parallel to the substrate surface. Therefore, the crystals may have grown vertically to the substrate surface as shown in Figure 9c. In addition, because the seeds did not attach on the inner walls of the tubular substrate, the tungsten source in the precursor sol may have been selectively consumed at the outer surface of the tubular substrate, leading to increase in crystallization of neighboring crystals. This must have led to membrane formation with c-axes oriented vertical to the surface of the underlying substrate. Silicon stopper was applied hereafter to ensure efficient seeding, to prevent seeding on the inner wall of the tube, and to promote the c-plane orientation of the final membrane. Membranes 2021, 11, x FOR PEER REVIEW 10 of 19 Figure 9. Illustration of the anticipated mechanism of the different crystal orientation observed for h-WO3 membranes prepared under different seeding conditions: (a) without seeding; (b) with seeding in the absence of silicon stoppers; (c) with seeding in the presence of silicon stoppers. Figure 10 shows SEM images of the membranes synthesized using precursor sols containing 3.5, 5.0, 6.0 and 7.0 mmol of Na2WO4, respectively. Dense membranes seemed to form in all conditions. The thicknesses of the formed membranes were found to increase from around 10 m to approximately 15 m by increasing the amount of Na   2WO4 in the precursor sol from 3.5 to 7.0 mmol. The amount of Na2WO4 in the precursor sol was also found to affected the membrane structure. The membranes synthesized with 3.5 and 5.0 mmol of Na2WO4 were comprised of rod-like crystals oriented vertically to the substrate surface, as shown in Figure 10a Figure 10 shows SEM images of the membranes synthesized using precursor sols containing 3.5, 5.0, 6.0 and 7.0 mmol of Na 2 WO 4 , respectively. Dense membranes seemed to form in all conditions. The thicknesses of the formed membranes were found to increase from around 10 µm to approximately 15 µm by increasing the amount of Na 2 WO 4 in the precursor sol from 3.5 to 7.0 mmol. The amount of Na 2 WO 4 in the precursor sol was also found to affected the membrane structure. The membranes synthesized with 3.5 and 5.0 mmol of Na 2 WO 4 were comprised of rod-like crystals oriented vertically to the substrate surface, as shown in Figure 10a-2,b-2, while those synthesized with 6.0 and 7.0 mmol of Na 2 WO 4 contained many rod-like crystals facing parallel to the substrate surface as shown in Figure 10c-2,d-2. Figure 11 shows the XRD diffractograms of h-WO 3 membranes synthesized using precursor sols containing 3.5, 5.0, 6.0 and 7.0 mmol of Na 2 WO 4 , respectively. Only peaks related to h-WO 3 were detected and those from the substrate were not, indicating that the h-WO 3 membranes without large defects formed thicker than the penetration depth of the X-ray. The peaks derived from c-plane and a-plane of h-WO 3 are marked with hexagonal symbols and rectangular symbols, respectively. The membranes synthesize with 3.5 and 5.0 mmol of Na 2 WO 4 showed obviously higher intensity of c-plane than a-plane of h-WO 3 while those synthesize with 6.0 and 7.0 mmol of Na 2 WO 4 showed the opposite trend. This matched well with the change in morphology of the membranes observed from the SEM images shown in Figure 10; i.e., change in direction of rod-like crystals from vertical to parallel to the substrate surface. This implies that an excess amount of Na 2 WO 4 in the precursor sol has an adverse effect on the c-plane orientation of the membranes. Figure 10. Cross-sectional and surface SEM images of the membranes synthesized using precursor sols containing different amounts of Na2WO4: (a-1) low and (a-2) high magnification cross-sectional SEM images of the membrane synthesized using precursor sol containing 3.5 mmol of Na2WO4; (a-3) low and (a-4) high magnification surface SEM images of the membrane synthesized using precursor sol containing 3.5 mmol of Na2WO4; (b-1) low and (b-2) high magnification crosssectional SEM images of the membrane synthesized using precursor sol containing 5.0 mmol of Na2WO4; (b-3) low and (b-4) high magnification surface SEM images of the membrane synthesized using precursor sol containing 5.0 mmol of Na2WO4; (c-1) low and (c-2) high magnification cross-sectional SEM images of the membrane synthesized using precursor sol containing 6.0 mmol of Na2WO4; (c-3) low and (c-4) high magnification surface SEM images of the membrane synthesized using precursor sol containing 6.0 mmol of Na2WO4; (d-1) low and (d-2) high magnification cross-sectional SEM images of the membrane synthesized using precursor sol containing 7.0 mmol of Na2WO4; (d-3) low and (d-4) high magnification surface SEM images of the membrane synthesized using precursor sol containing 7.0 mmol of Na2WO4. Figure 11 shows the XRD diffractograms of h-WO3 membranes synthesized using precursor sols containing 3.5, 5.0, 6.0 and 7.0 mmol of Na2WO4, respectively. Only peaks related to h-WO3 were detected and those from the substrate were not, indicating that the h-WO3 membranes without large defects formed thicker than the penetration depth of the X-ray. The peaks derived from c-plane and a-plane of h-WO3 are marked with hexagonal symbols and rectangular symbols, respectively. The membranes synthesize with 3.5 and 5.0 mmol of Na2WO4 showed obviously higher intensity of c-plane than a-plane of h-WO3 while those synthesize with 6.0 and 7.0 mmol of Na2WO4 showed the opposite trend. This matched well with the change in morphology of the membranes observed from the SEM images shown in Figure 10; i.e., change in direction of rod-like crystals from vertical to parallel to the substrate surface. This implies that an excess amount of Na2WO4 in the precursor sol has an adverse effect on the c-plane orientation of the membranes. The effect of Na 2 WO 4 amount on the degree of c-plane orientation of the h-WO 3 membranes was quantitatively evaluated by the c-plane orientation index; R, as shown in Figure 12. The R of the membranes synthesized with both 3.5 and 5.0 mmol of Na 2 WO 4 showed 10 times higher value than that of the standard powder data, which were 10.4 and 11.3, respectively. This was also significantly higher than that of our previous report; which showed a maximum R of 2.0, meaning a membrane with high c-plane orientation has been realized on the tubular substrate. Meanwhile, when the amount of Na 2 WO 4 in the precursor sol was further increased to 6.0 and 7.0 mmol, R drastically decreased to a lower value than 1. Especially, when the amount of Na 2 WO 4 in the precursor sol was 7.0 mmol, R of the membrane became almost one-tenth of that of standard powder data. This further proved that the amount of Na 2 WO 4 in the precursor sol significantly affected the c-plane orientation of the synthesized h-WO 3 membranes. In total, 3.5 to 5.0 mmol of Na 2 WO 4 was found to be effective to prepare h-WO 3 membranes with high c-plane orientation. The effect of Na2WO4 amount on the degree of c-plane orientation of the h-WO3 membranes was quantitatively evaluated by the c-plane orientation index; R, as shown in Figure 12. The R of the membranes synthesized with both 3.5 and 5.0 mmol of Na2WO4 showed 10 times higher value than that of the standard powder data, which were 10.4 and 11.3, respectively. This was also significantly higher than that of our previous report; which showed a maximum R of 2.0, meaning a membrane with high c-plane orientation has been realized on the tubular substrate. Meanwhile, when the amount of Na2WO4 in the precursor sol was further increased to 6.0 and 7.0 mmol, R drastically decreased to a lower value than 1. Especially, when the amount of Na2WO4 in the precursor sol was 7.0 mmol, R of the membrane became almost one-tenth of that of standard powder data. This further proved that the amount of Na2WO4 in the precursor sol significantly affected the c-plane orientation of the synthesized h-WO3 membranes. In total, 3.5 to 5.0 mmol of Na2WO4 was found to be effective to prepare h-WO3 membranes with high c-plane orientation. The reason why the amount of Na 2 WO 4 in precursor sol affected the c-plane orientation of the membranes may be explained as follows. When the amount of Na 2 WO 4 in the precursor sol was below 5.0 mmol, nucleation and crystallization of h-WO 3 are expected to be mainly occurring near the seeds embedded on the tubular substrate surface. Numerous nuclei must have selectively grown vertically to the substrate surface due to the neighboring nuclei which limits the growth parallel to the substrate surface, as discussed in Figure 9 of Section 3.2. In contrast, when the amount of Na 2 WO 4 in the precursor sol exceeded 6.0 mmol, nucleation and crystallization of the h-WO 3 may have taken place primarily in the bulk precursor sol due to the high tungsten concentration. The h-WO 3 crystals nucleated in the bulk precursor sol shall attach on the substrate surface and become incorporated into the h-WO 3 membrane starting to grow on the surface of the tubular substrate as shown in Figure 13, thus decreasing the degree of c-plane orientation of the membrane. Furthermore, the number of nuclei generated in the bulk precursor sol is likely to increase as the precursor sol reaches closer to saturation and this must have led to attachment of more h-WO 3 crystals from the bulk precursor sol onto the substrate surface. A precursor sol containing a moderate amount of Na 2 WO 4 , which is large enough to grow and densify the membrane but low enough to prevent vast nucleation and crystallization of h-WO 3 in the bulk precursor sol, seems to be the key for realizing a h-WO 3 membrane satisfying both high densification and high c-plane orientation. The reason why the amount of Na2WO4 in precursor sol affected the c-plane orientation of the membranes may be explained as follows. When the amount of Na2WO4 in the precursor sol was below 5.0 mmol, nucleation and crystallization of h-WO3 are expected to be mainly occurring near the seeds embedded on the tubular substrate surface. Numerous nuclei must have selectively grown vertically to the substrate surface due to the neighboring nuclei which limits the growth parallel to the substrate surface, as discussed in Figure 9 of Section 3.2. In contrast, when the amount of Na2WO4 in the precursor sol exceeded 6.0 mmol, nucleation and crystallization of the h-WO3 may have taken place primarily in the bulk precursor sol due to the high tungsten concentration. The h-WO3 crystals nucleated in the bulk precursor sol shall attach on the substrate surface and become incorporated into the h-WO3 membrane starting to grow on the surface of the tubular substrate as shown in Figure 13, thus decreasing the degree of c-plane orientation of the membrane. Furthermore, the number of nuclei generated in the bulk precursor sol is likely to increase as the precursor sol reaches closer to saturation and this must have led to attachment of more h-WO3 crystals from the bulk precursor sol onto the substrate surface. A precursor sol containing a moderate amount of Na2WO4, which is large enough to grow and densify the membrane but low enough to prevent vast nucleation and crystallization of h-WO3 in the bulk precursor sol, seems to be the key for realizing a h-WO3 membrane satisfying both high densification and high c-plane orientation. Figure 13. Illustration of the anticipated effect of bulk precursor sol nucleation on crystal orientation of membrane. Figure 13. Illustration of the anticipated effect of bulk precursor sol nucleation on crystal orientation of membrane. Table 1 shows the PV performance of the h-WO 3 membrane obtained from precursor sols with various amounts of Na 2 WO 4 . The membrane performance drastically deteriorated when the amount of Na 2 WO 4 exceeded 6.0 mmol. The total flux of membranes synthesized with precursor sols with 3.5 and 5.0 mmol of Na 2 WO 4 approximately doubled compared to those prepared with 6.0 and 7.0 mmol of Na 2 WO 4 . The separation factor also drastically changed between 5.0 and 6.0 mmol of Na 2 WO 4 . This seemed to be closely related to the change in the degree of c-plane orientation of the membranes shown in Figure 12. The membranes showing high performance exhibited an obviously high degree of c-plane orientation. Therefore, the crystal orientation must have highly affected the membrane performance. To understand this reason, the permeance of water and CH 3 COOH were compared. The fluctuation in CH 3 COOH permeance was rather small regardless of the change in total flux while the change in water permeance drastically increased along with the increase in total flux. This means that the pathway for only water molecules increased. As we have revealed in our previous study, the separation mechanism of h-WO 3 membranes is based on molecular sieving [26]. Because the pores of h-WO 3 is running along the c-axis of h-WO 3 crystals and the water-permeable pores are smaller than CH 3 COOH molecules, the higher c-plane orientation must have resulted in more pathways for water molecules without increasing those for CH 3 COOH molecules. The growth of non-c-plane oriented h-WO 3 crystals attached and incorporated in the membrane must be prevented to realize high membrane performance.  Figure 14 shows the PV performance for separation of 90:10 wt % CH 3 COOH/water mixture of the highly c-plane oriented h-WO 3 membranes obtained in this study (prepared by precursor sols containing 3.5 and 5.0 mmol of Na 2 WO 4 ) compared to our previous study and membranes comprised of conventional materials. Compared to our previous report, the total flux was drastically enhanced in this study. Indeed, the total flux was increased by more than twice of that of the previous study. Moreover, the performances of the developed membranes seem to be higher than conventional polymeric membranes [29][30][31] and are becoming closer to those of conventional inorganic membranes [9,12,14,20,[32][33][34][35], including zeolite membranes [9,12,14,20,34,35], which are expected to be promising candidate membranes for separating CH 3 COOH/water mixtures.  [26]; the cross marks indicate performances of conventional polymeric membranes reported in references [29][30][31]; the triangle marks indicate performances of conventional inorganic membranes reported in references [32,33]; the diamond marks indicate the conventional zeolite membranes reported in reference [9,12,14,20,34,35].)

PV Performance
The stability of the h-WO3 membrane was evaluated by a long-term-separation test using the membrane synthesized with 3.5 mmol of Na2WO4. The PV performance of the membrane was repeatedly evaluated before and after storing in a 90:10 wt % CH3COOH/water mixture with a pH below 0 for a particular period at room temperature. The result is shown in Figure 15. Even after repeated testing for more than 500 h, the separation factor was maintained almost constant. Both water and CH3COOH permeance decreased by about 20% from the first test to the second test; however, the permeance was still high compared to our previous report [26] and maintained nearly constant thereafter for more than 300 h. This indicated that the h-WO3 membrane is durable against acidic solvents and that its performance can be maintained even during a long-term-separation test.  [26]; the cross marks indicate performances of conventional polymeric membranes reported in references [29][30][31]; the triangle marks indicate performances of conventional inorganic membranes reported in references [32,33]; the diamond marks indicate the conventional zeolite membranes reported in reference [9,12,14,20,34,35].) The stability of the h-WO 3 membrane was evaluated by a long-term-separation test using the membrane synthesized with 3.5 mmol of Na 2 WO 4 . The PV performance of the membrane was repeatedly evaluated before and after storing in a 90:10 wt % CH 3 COOH/water mixture with a pH below 0 for a particular period at room temperature. The result is shown in Figure 15. Even after repeated testing for more than 500 h, the separation factor was maintained almost constant. Both water and CH 3 COOH permeance decreased by about 20% from the first test to the second test; however, the permeance was still high compared to our previous report [26] and maintained nearly constant thereafter for more than 300 h. This indicated that the h-WO 3 membrane is durable against acidic solvents and that its performance can be maintained even during a long-term-separation test.
The stability of the h-WO3 membrane was evaluated by a long-term-separation test using the membrane synthesized with 3.5 mmol of Na2WO4. The PV performance of the membrane was repeatedly evaluated before and after storing in a 90:10 wt % CH3COOH/water mixture with a pH below 0 for a particular period at room temperature. The result is shown in Figure 15. Even after repeated testing for more than 500 h, the separation factor was maintained almost constant. Both water and CH3COOH permeance decreased by about 20% from the first test to the second test; however, the permeance was still high compared to our previous report [26] and maintained nearly constant thereafter for more than 300 h. This indicated that the h-WO3 membrane is durable against acidic solvents and that its performance can be maintained even during a long-term-separation test. Figure 15. Performance of highly c-plane oriented h-WO3 membrane during long-term-separation test. Figure 15. Performance of highly c-plane oriented h-WO 3 membrane during long-term-separation test.
So far, several zeolite membranes have been evaluated using long-term separation tests [10][11][12][13]15,17]. The initial decrease in permeability has been observed in almost all cases. Masuda et al. anticipated that this is due to the plugging by partial insertion of the CH 3 COOH molecule into the zeolite pores [13]. Thus, a similar phenomenon may have taken place in this study. However, compared with a conventional study tested in a similar testing condition of 90:10 wt % CH 3 COOH/water mixture at 348 K, the permeability of MOR-type zeolite membrane decreased to nearly half after 30 h of PV test [11] while the developed h-WO 3 membrane decreased by only 20% and stabilized. This implies that the plugging of h-WO 3 pores caused by the CH 3 COOH molecule was much less compared to the mordenite pores. From this result, there is a possibility that the h-WO 3 membrane has a better fouling resistance compared to the conventionally reported membrane.

Conclusions
In this work, the effect of synthesis conditions of h-WO 3 membranes with c-plane orientation on tubular substrates and their CH 3 COOH/water separation performance were investigated. The h-WO 3 membranes were prepared by hydrothermal synthesis using a precursor sol containing various amounts of Na 2 WO 4 in the presence of tubular substrates with seeds embedded on only their outside surface. The method to embed the amorphous h-WO 3 seeds selectively on the outer surface of the tubular substrates was found to play an important role against densification and c-plane orientation of the h-WO 3 membranes. In addition, the amount of Na 2 WO 4 in the precursor sol significantly affected crystal orientation of the membrane. A precursor sol with 3.5 and 5.0 mmol of Na 2 WO 4 was effective to simultaneously satisfy c-plane orientation and densification of the membrane while an excess amount of Na 2 WO 4 drastically decreased the degree of c-plane orientation. A highly c-plane oriented h-WO 3 membrane was successfully obtained under the optimized condition, which had a c-plane orientation index of approximately 10 and exhibited a maximum separation factor of 40.0 with a water permeance as high as 1.53 × 10 −7 mol·m −2 ·s −1 ·Pa −1 in a 90:10 wt % CH 3 COOH/water mixture. This corresponded to a greater than five-times higher c-plane orientation index and a greater than two-times higher water permeance compared to our previous study. Furthermore, it was found that the CH 3 COOH/water separation ability of the highly c-plane oriented h-WO 3 membrane can be maintained while being stored in 90:10 wt % CH 3 COOH/water mixture with pH < 0 for 500 h, proving its high acidic resistance and stability of performance in an acidic environment. The highly c-plane oriented h-WO 3 membrane seems to be a possible candidate for separation of water from acidic solutions.
Supplementary Materials: The following are available online at https://www.mdpi.com/2077-037 5/11/1/38/s1, Figure S1: Difference in c-plane orientation of h-WO 3 membranes prepared on (a) flat-type substrate and (b) tubular substrate under the same preparation condition in previous study.