Plasma-enhanced chemical vapor deposition of amorphous carbon molecular sieve membranes for gas separation

Abstract

Amorphous carbon membranes were successfully synthesized onto a SiO₂–ZrO₂/α-Al₂O₃ nanoporous substrate via plasma-enhanced chemical vapor deposition (PECVD) at room temperature. PECVD-derived amorphous carbon membranes exhibited molecular sieving properties, showing ideal selectivities of 23 and 1750 for He/N₂ and He/SF₆, respectively, at 25 °C. The membrane maintained high selectivity even at high temperatures as high as 200 °C, indicating considerable stability of the plasma-deposited amorphous carbon layer.

---

Membrane-based gas separation is rapidly growing in a number of applications such as hydrogen purification, carbon dioxide capture, oxygen enrichment, and hydrocarbon separation, because of their low energy requirement and low operating cost. In this context, microporous inorganic membranes such as amorphous carbon membranes have recently received considerable attention owing to their unique characteristics such as an advanced thermal stability, anti-corrosive property, solvent resistance, and super-hydrophobicity.^1,2 Amorphous carbon membranes are comprised of micropores with a range of several angstroms, which can effectively reject the permeation of relatively large molecules, thereby exhibiting excellent permselectivity.¹ For instance, Yoshimune et al. reported flexible carbon hollow fiber membranes derived from sulfonated poly(phenylene oxide) with an ideal H₂/N₂ selectivity of 214 and a H₂ permeance of 3.67 × 10⁻⁸ mol m⁻² s⁻¹ Pa⁻¹ at 30 °C.³ The conventional synthesis of amorphous carbon membranes is via the pyrolysis and carbonization of polymeric precursors.^1,2 In this way, to obtain acceptable permeation performance, the synthetic process requires high operating temperatures that are usually well above 500 °C.

Another approach that might be useful for the synthesis of amorphous carbon membranes is the use of plasma-enhanced chemical vapor deposition (PECVD), which is a commonly used technique for the deposition of thin films. There is growing interest in utilizing PECVD to fabricate inorganic membranes that have the potential for application in molecular separation, since PECVD offers a low processing temperature (as low as room temperature) unlike the conventional methods (typically 500–600 °C).^4–8 For instance, Roualdes et al. prepared polysiloxane membranes using organosilicon precursors such as hexamethyldisiloxane (HMDSO) onto cellulose ester substrates and reported that plasma-deposited membranes had higher selectivity than conventional polysiloxane membranes.⁴ Kafrouni et al. used hexamethyldisilazane and ammonia as precursors and prepared silicon carbonitride membranes with a He/N₂ selectivity of 50 at 150 °C.⁵ Moreover, in our previous study, we demonstrated a low-temperature fabrication of organosilica membranes with excellent molecular sieving properties via a 2-step PECVD involving HMDSO/Ar-PECVD followed by HMDSO/O₂-PECVD.^6,7 The obtained membrane showed remarkable selectivities for He/N₂ and He/SF₆ of 7800 and 27 000, respectively, at 25 °C.⁷ We also found that such PECVD-derived organosilica membranes possess superior thermal stability at temperature as high as 500 °C, although the membranes were fabricated at room temperature.⁸ These studies have shown that PECVD is significantly attractive and promising as a method to fabricate inorganic membranes for molecular separation applications. Thus far, however, this approach has been limited to the fabrication of membranes consisting of silicon-based materials.

PECVD has also been applied to the fabrication of amorphous carbon thin films that are used for protective coatings such as gas diffusion barriers.^9–11 Recently, Karan et al. prepared innovative ultrathin diamond-like-carbon (DLC) membranes for organic solvent nanofiltration using PECVD at low temperatures.¹² They reported that the estimated pore size of a PECVD-derived amorphous carbon layer was approximately 1 nm, and revealed that the membranes showed an ultrafast permeation of organic solvents with a sharp molecular weight cut-off.⁹ Carbon thin films such as DLC and amorphous carbons have traditionally been recognized as a diffusion barrier material, but the results reported by Karan et al. demonstrated the great potential of PECVD-derived amorphous carbons as a material for separation membranes. This also posed an interesting question as to whether PECVD-derived amorphous carbons were applicable as a membrane material for other molecular separation applications, such as gas separation, which typically requires a much smaller pore size than that of Karan's. Herein, we report the first successful synthesis and molecular sieving characteristics of amorphous carbon membranes prepared via PECVD at room temperature.

Since gaseous atoms and molecules range in size from 0.26 to 0.55 nm, the size of the membrane micropores must be tuned by several angstroms in order to achieve a desirable gas separation performance. Our approach to fabricating such a microporous structure via PECVD is described in Fig. 1. A porous α-Al₂O₃ support was first coated with a SiO₂–ZrO₂ sol to form an intermediate layer, which had an average pore size of 1–2 nm, and then, an amorphous carbon layer was plasma-deposited onto the intermediate layer. We expected the presence of the nanoporous intermediate layer to prevent the penetration of deposits into the support and to reduce the thickness of the amorphous carbon layer in order to close all the pores in the support.

Fig. 1 Schematic of the fabrication of an amorphous carbon membrane via PECVD onto a porous support.

PECVD was conducted in a flow-type plasma reactor equipped with an RF coil operated at 13.56 MHz, as shown in Fig. S1 (ESI-1†). Propylene was used as a carbon source, and was fed to the reactor at a flow rate of 10 sccm. The pressure in the reactor was kept constant at approximately 150 Pa during the plasma-deposition. The gas permeation properties of the resultant membranes were measured in situ at room temperature via a constant-volume variable-pressure technique.

Scanning electron microscopy (SEM) showed that the plasma-deposited layer was continuous and formed a densely packed grain structure (Fig. 2(a)). The deposited layer covered the entire surface of the SiO₂–ZrO₂ intermediate layer with no obvious cracks or defects. The thickness of the plasma-deposited layer was approximately 3 μm after 20 min of deposition (Fig. 2(b)). The FT-IR spectra obtained from the PECVD-derived film showed absorption peaks at 2930 and 2960 cm⁻¹, which are characteristic of the asymmetric stretching of sp³ methylene (–CH₂–) and sp³ methyl (–CH₃),¹³ respectively, indicating a relatively high hydrogen content (Fig. 2(c)). The absorbance area of methylene and methyl groups increased linearly with the time of deposition, indicating the linear growth trend of the plasma-deposited layer (Fig. S2, ESI-2†). The carbon content estimated by X-ray photoelectron spectroscopy (XPS) was 89.3 at% (Fig. S3, ESI-2†). High resolution XPS spectrum of C_1s showed that some of the carbon atoms had been covalently bonded to the oxygen atoms (Fig. S4, ESI-2†). The Raman spectra of the film showed D and G broad bands (Fig. 2(d)), which is characteristic of amorphous carbon.¹⁴ These results indicate that the plasma-deposited layer is mainly composed of hydrogenated amorphous carbon.

Fig. 2 SEM images of (a) the surface and (b) the cross-section of a PECVD-derived membrane deposited onto a SiO₂–ZrO₂/α-Al₂O₃ substrate, and (c) FTIR and (d) Raman spectra of PECVD-derived film deposited onto a silicon wafer (20 min of deposition time).

Fig. 3 shows the single gas permeance of He (kinetic diameter, d_i = 0.26 nm), N₂ (0.364 nm), and SF₆ (0.55 nm) at 25 °C for a PECVD-derived amorphous carbon membrane as a function of plasma-deposition time. Note that the permeance with a deposition time of 0 indicates the permeance of the SiO₂–ZrO₂/α-Al₂O₃ substrate before the plasma-deposition. The amorphous carbon membrane was stable in the gas permeation measurement, withstanding repeated cycles of pressurization and evacuation. The amorphous carbon membrane showed higher selectivity and lower permeance as deposition time increased. With 5 min of deposition time, the membrane showed a He permeance of 9.2 × 10⁻⁹ mol m⁻² s⁻¹ Pa⁻¹, which was approximately one-thousandth that of the SiO₂–ZrO₂/α-Al₂O₃ substrate, indicating that a plasma-deposited layer creates major resistance for gas permeation. Meanwhile, the permeance ratios of He/N₂ and He/SF₆ were 19 and 140, respectively, which exceeded the ideal selectivity of Knudsen diffusion. The permeance of He was slightly decreased in the subsequent deposition, while the permeance ratios were increased significantly. For instance, the permeance ratios for He/N₂ and He/SF₆ were increased to 23 and 1750, respectively, after 20 min of deposition.

Fig. 3 Single gas permeance for He, N₂ and SF₆, and permeance ratio for He/N₂ and He/SF₆ at 25 °C for a PECVD-derived amorphous carbon membrane as a function of plasma-deposition time.

In Fig. 4(a), the single gas permeance at 25 °C for the amorphous carbon membrane at deposition times of 0, 5, 10, and 20 min are plotted as a function of the kinetic diameter. As mentioned previously, for all types of gases, permeance decreased as deposition time increased. Fig. 4(b) shows the kinetic diameter dependence of the relative permeance normalized by He permeance together with the ideal Knudsen-based permeance that would be expected from the permeance of He. Overall, after plasma-deposition, the relative permeance decreased as the kinetic diameter of gases increased, and was significantly lower than that expected from the Knudsen diffusion mechanism, indicating the molecular sieving characteristics of the plasma-deposited amorphous carbon layer. In comparing the relative permeance of He over N₂ and SF₆, the normalized permeance of N₂ reached a limit within 10 min of deposition time. In contrast, the normalized permeance of SF₆ continued to decrease as the plasma-deposition time proceeded. In other words, selectivity for He/N₂ reached a constant value, while that for He/SF₆ kept increasing throughout the process of plasma-deposition.

Fig. 4 (a) Single gas permeance and (b) normalized permeance at 25 °C for a PECVD-derived amorphous carbon membrane as a function of the kinetic diameter of gases. The dash-line indicates the ideal Knudsen-based permeance normalized by the permeance of He.

According to the gas-translation model,^15–18 gas permeation characteristics in microporous membranes, where the effect of adsorption can be negligible, are dependent on the ratio of the molecular diameter of permeation gases to the average pore size of the membrane. Hence, once a microporous molecular sieving layer is uniformly formed on a substrate, the selectivity for any given gas pair can be uniquely determined. Meanwhile, the gas selective properties of molecular sieve membranes are often limited by grain boundary defects. The presence of grain boundary defects allows undesired permeation through the membrane, particularly for the molecules that are too large to permeate through the microporous molecular sieving layer. Therefore, since the permeance ratio for He/N₂ reached a constant value, it is reasonable to suppose that the permeation of He and N₂ occurred mainly through the plasma-deposited layer. In contrast, it can be presumed that SF₆ mainly permeated through the grain boundary defects. The delayed decrease in the relative SF₆ permeance indicates the blockage of grain boundary defects along with the growth of the plasma-deposited layer.

Regarding the permeation of small-sized gases, the permeance ratio of He/H₂ showed little change during the plasma-deposition, and was about the same as the Knudsen ratio of 0.71. This suggests that the size of the permeation pathway in the plasma-deposited layer was much larger than the size of either He or H₂ (0.289 nm) so that He and H₂ could permeate in a Knudsen-like manner. It should also be noted that the permeance of CO₂ (0.33 nm) after plasma-deposition was similar to that of He, although the kinetic diameter and the molecular weight of CO₂ are larger than those of He. This can be ascribed to the permeation of CO₂ being affected by an attractive interaction with the plasma-deposited layer.

To further understand the gas permeation behavior through PECVD-derived amorphous carbon membranes, the temperature dependence of gas permeance was studied at temperatures of 200, 150, 75, and 30 °C, in a decreasing order. As described in Fig. 5, although plasma-deposition was conducted at room temperature, the membrane maintained high selectivity at high temperature, indicating considerable stability of the plasma-deposited amorphous carbon layer. The permeance ratios for He/N₂ and He/SF₆ at 200 °C were 16 and 1560, respectively, which showed excellent molecular sieving characteristics.

Fig. 5 Temperature dependence of single gas permeance for a PECVD-derived amorphous carbon membrane. The solid curves are fitted with the modified gas-translation model (ESI-3†).¹⁹

The permeances of He, H₂, CO₂, and N₂ were all increased with an increase in temperature, indicating that an activated diffusion was significant in the permeation through the membrane. An activated diffusion occurs when the permeating molecules are restricted in their movement such as when the dimensions of the permeating molecules are only slightly smaller than the pore diameter. The activation energies of permeation for He, H₂, CO₂, and N₂, as determined from the modified gas-translation model,¹⁸ were 12.6, 11.9, 5.2, and 15.8 kJ mol⁻¹, respectively. Despite the fact that the permeance ratio for He/H₂ was near that of Knudsen selectivity (in which the activation energy was assumed to be 0 in theory), activation energies observed for these gases were found to be high. Since the plasma-deposited layer contained a linear aliphatic structure, we speculated that the activation energies for these gases were associated with the movement of polymer-like segments. Regarding the permeation of CO₂, the activation energy for permeation was smaller than those for other gases. The apparent activation energy is the sum of the heat of adsorption and the activation energy of the rate-limiting kinetic step. This result indicates that gas permeation that is based on surface diffusion makes a strong contribution to the overall permeance of CO₂. Conversely, the permeance of SF₆ showed a temperature dependence similar to Knudsen diffusion, which decreased slightly with an increase in temperature. The activation energy of permeation for SF₆ was 3.1 kJ mol⁻¹. Knudsen diffusion occurs in somewhat larger pores, compared with activated diffusion.

Since the kinetic diameters of the evaluated gases follow the order of He (0.26 nm) < H₂ (0.289) < CO₂ (0.33) < N₂ (0.364) < SF₆ (0.55), this result makes it reasonable to expect that the effective pore size of the plasma-deposited amorphous carbon layer was larger than the size of N₂, but no more than the size of SF₆. Thus, the permeation of He, H₂, CO₂, and N₂ occurs mainly through the amorphous carbon matrix, whereas the permeation of SF₆ is more likely to occur through the grain boundaries. According to the normalized Knudsen-based permeance (NKP) analysis,¹⁹ the effective pore size of the plasma-deposited amorphous carbon membrane was estimated to be 0.53 nm (Fig. S5, ESI-4†). This clearly shows that we are able to develop amorphous carbon membranes with micropores that are applicable to gas separation. This finding may provide new insight into gas permeation through PECVD-derived amorphous carbon thin films, which was used to be considered as diffusion barrier materials.

Conclusions

In conclusion, a novel amorphous carbon membrane was deposited onto a nanoporous SiO₂–ZrO₂/α-Al₂O₃ substrate using plasma-enhanced chemical vapor deposition (PECVD) at room temperature. The resultant membrane showed molecular sieving characteristics with ideal selectivities of 23 and 1750 for He/N₂ and He/SF₆, respectively, at 25 °C. The temperature dependency of permeances reveals that the permeation of the small-sized gases (He, H₂, CO₂, and N₂) obeyed the activated diffusion mechanism, while that of the larger-sized gas (SF₆) was governed by Knudsen diffusion, which suggests that the effective pore size ranged between the size of N₂ and SF₆. The ideal selectivity reached 16 and 1560 for He/N₂ and He/SF₆, respectively, at 200 °C. This work demonstrated a novel approach to the fabrication of an amorphous carbon membrane for gas separation without high-temperature processing.

Acknowledgements

This research was supported by Core Research for Evolutional Science and Technology (CREST) from the Japan Science and Technology Agency (JST), and Grants-in-Aid for Young Scientists (B) (KAKENHI) Grant No. 25820382 from the Japan Society for the Promotion of Science (JSPS).

Notes and references

1. A. Ismail and L. David, J. Membr. Sci., 2001, 193, 1.
2. S. Saufi and A. Ismail, Carbon, 2004, 42, 241.
3. M. Yoshimune, K. Mizoguchi and K. Haraya, J. Membr. Sci., 2013, 425–426, 149.
4. S. Roualdes, J. Sanchez and J. Durand, J. Membr. Sci., 2002, 198, 299.
5. W. Kafrouni, R. Rouessac, A. Julbe and J. Durand, J. Membr. Sci., 2009, 329, 130.
6. T. Tsuru, H. Shigemoto, M. Kanezashi and T. Yoshioka, Chem. Commun., 2011, 47, 8070.
7. H. Nagasawa, H. Shigemoto, M. Kanezashi, T. Yoshioka and T. Tsuru, J. Membr. Sci., 2013, 441, 45.
8. H. Nagasawa, T. Minamizawa, M. Kanezashi, T. Yoshioka and T. Tsuru, Sep. Purif. Technol., 2014, 121, 13.
9. S. Vasqiez-Borucki, W. Jacob and C. A. Achete, Diamond Relat. Mater., 2000, 9, 1971.
10. N. Boutroy, Y. Parnel, J. M. Rius, F. Auger, H. J. von Bardeleben, J. L. Cantin, F. Abel, A. Zeinert, C. Casiraghi, A. C. Ferrari and J. Robertson, Diamond Relat. Mater., 2006, 15, 921.
11. O. Polonskyi, O. Kylian, M. Petr, A. Choukourov, J. Hanus and H. Biederman, Thin Solid Films, 2013, 540, 65.
12. S. Karan, S. Samitsu, X. Peng, K. Kurashima and I. Ichinose, Science, 2012, 27, 444.
13. M. Nagatsu, T. Sano, N. Takada, N. Toyoda, M. Tanga and H. Sugai, Diamond Relat. Mater., 2002, 11, 976.
14. A. C. Ferrari and J. Robertson, Phys. Rev. B: Condens. Matter Mater. Phys., 2001, 64, 075414.
15. P. Chu and L. Li, Mater. Chem. Phys., 2006, 96, 253.
16. J. Xiao and J. Wei, Chem. Eng. Sci., 1992, 47, 1123.
17. A. Shelekhin, A. Dixon and Y. Ma, AIChE J., 1995, 41, 58.
18. H. R. Lee, M. Kanezashi, Y. Shimomura, T. Yoshioka and T. Tsuru, AIChE J., 2011, 57, 2755.
19. T. Yoshioka, M. Kanezashi and T. Tsuru, AIChE J., 2013, 59, 2179.

---

Footnote

† Electronic supplementary information (ESI) available: Experimental details, FTIR and XPS results of prepared membranes, and additional results for the gas permeation measurement. See DOI: 10.1039/c6ra09381g

---