Si‑Rich Alumina Supports Enhance Vertical SWCNT Growth and Nanofiltration Membrane Selectivity

Abstract

We examined how the thermal stability of alumina support layers, fabricated by radio‑frequency sputtering under different atmospheres, influences the growth of vertically aligned single‑walled carbon nanotube (VA‑SWCNT) arrays. Reactive sputtering with an O₂–Ar mixture produced a Si‑rich alumina alloy that remains stable at 850 °C, whereas non‑reactive argon‑only deposition yields a less stable film with pronounced defects after annealing. The Si‑rich layer promotes the growth of VA‑SWCNTs with a narrow diameter distribution (<2.2 nm), resulting in nanofiltration membranes that exhibit superior ion rejection owing to their smaller pore sizes.

Background

Single‑walled carbon nanotubes (SWCNTs) are promising for high‑strength composites, flexible electronics, and, notably, nanofiltration membranes. Their atomically smooth inner walls provide frictionless channels, enabling exceptionally high fluid flux while maintaining sharp size‑exclusion and electrostatic selectivity. Controlling the diameter distribution and density of SWCNTs is therefore critical for optimizing membrane performance.

Chemical vapor deposition (CVD) is the standard method for large‑scale SWCNT synthesis. Transition‑metal nanoparticles (Fe, Ni, Co) act as catalysts, and when densely packed, they self‑assemble into vertically aligned arrays (VA‑SWCNTs) that are ideal for membrane fabrication. However, CVD operates at 500–900 °C, where catalyst particles undergo rapid diffusion and ripening, reducing catalyst lifetime and increasing nanotube diameters.

Beyond catalyst chemistry, the catalyst–substrate interface governs thermal stability. Oxide supports such as SiO₂, Al₂O₃, MgO, and ZrO₂ have been employed to stabilize catalysts and improve SWCNT yield. Among them, alumina thin films have shown superior performance, likely due to their ability to prevent unwanted metal oxide formation and promote uniform catalyst dispersion.

Previous work has indicated that sputtered alumina outperforms films deposited by e‑beam evaporation or atomic‑layer deposition, suggesting that the film’s stoichiometry and impurities introduced during deposition play a role. This study investigates the effect of oxygen incorporation during sputtering on alumina thermal stability, VA‑SWCNT growth, and resulting nanofiltration membrane selectivity.

Methods

Alumina and Fe/Mo Catalyst Preparation

Alumina films were deposited on p‑type Si (100) wafers by 13.56 MHz RF sputtering from a 99.99 % Al₂O₃ target. For non‑reactive sputtering, the chamber was pumped to ~3 × 10⁻⁵ Torr, then filled with Ar to 5.8 mTorr and sputtered at 210 W (≈4.8 W cm⁻²), yielding ~0.6 nm min⁻¹ deposition up to 30 nm. Reactive sputtering added O₂ to raise the pressure to 6.2 mTorr and reduced the rate to ~0.5 nm min⁻¹.

Fe/Mo catalyst bilayers (0.5 nm Fe / 0.2 nm Mo) were deposited by e‑beam evaporation under <4 × 10⁻⁶ Torr. The wafers were diced into 1 × 1 cm² chips for annealing.

Annealing and CVD Growth

Annealing and growth were performed in a custom atmospheric‑pressure CVD setup with a quartz tube furnace. The temperature ramped to 850 °C at 50 °C min⁻¹ under He (515 SCCM) and H₂ (≤400 SCCM). After 12 min annealing, the H₂ flow was reduced to 15 SCCM, and growth was initiated with a mixture of C₂H₄ (100 SCCM), H₂ (15 SCCM), and He (515 SCCM). For annealing‑only runs, C₂H₄ was omitted. The growth protocol follows that detailed in our previous study [22].

Characterization

Surface morphologies were examined by atomic force microscopy (AFM) in tapping mode. Cross‑sectional TEM samples were prepared by Ar ion milling, and TEM (JEM‑ARM200F) coupled with EDX (QUANTAX 400) provided elemental maps. Raman spectra (632.8 nm HeNe laser, ×100 objective, <0.1 mW) assessed graphitic quality. Nanotube diameters were extracted from TEM images taken with a Philips CM300‑FEG TEM.

Membrane Fabrication and Nanofiltration

Low‑stress SiNₓ was LPCVD‑deposited to seal the VA‑SWCNT forest, forming a mechanically robust membrane. Argon ion milling removed catalyst residues and alumina on the growth side, followed by O₂ plasma etching to expose the tube ends. The resulting SiNₓ/CNT composite was confirmed defect‑free by SEM (JEOL 7401‑F) and used in pressure‑driven ion rejection tests. 2 mL of 1 mM KCl or 0.5 mM K₂SO₄ was pressurized at 0.69 bar through the membrane; permeate samples were analyzed by capillary electrophoresis (Agilent 3D CE). Rejection coefficients were calculated as 1 – (c_permeate / c_feed).

Results and Discussion

Alumina Thermal Stability

AFM images after annealing revealed stark contrast between the two deposition methods (Fig. 2). Non‑reactive sputtering produced ~180 pits μm⁻² with ~2 nm depth and 10–50 nm diameter, indicating significant defect formation. In contrast, reactive sputtering yielded a nearly defect‑free surface (RMS roughness 0.2 nm), with well‑defined sub‑2 nm Fe/Mo nanoparticles. These observations suggest that oxygen during sputtering enhances thermal stability by incorporating Si into the alumina matrix.

Alumina Composition

Cross‑sectional TEM revealed an interfacial silicate layer in both films, likely due to Si diffusion into the alumina during deposition. EDX analysis of the bulk layer (layer 2) showed a markedly higher Si/Al ratio (~10×) in the reactive film, confirming Si enrichment. Density‑functional theory predicts that Si–O bonds are stronger than Al–O, widening the thermal stability window and preventing volatilization of AlOₓ species at high temperature.

VA‑SWCNT Growth

TEM imaging confirmed SWCNT growth on both supports. The reactive film produced a mean diameter of 1.2 nm (σ = 0.4 nm), while the non‑reactive film yielded 1.4 nm (σ = 0.5 nm). Both distributions followed a log‑normal trend but were skewed toward smaller diameters. Raman spectra displayed a pronounced G‑band shoulder (~1570 cm⁻¹) for tubes grown on the stable support, indicating high structural quality (G/D ≈ 10). Growth from the unstable film terminated earlier, likely due to enhanced Fe/Mo subsurface diffusion caused by a defective support.

Ion Transport

Membranes fabricated with VA‑SWCNTs from the stable alumina exhibited 15–20 % higher KCl rejection and ~12 % higher K₂SO₄ rejection compared to those from the unstable film (Fig. 6). The improvement is attributed primarily to more efficient electrostatic exclusion in the narrower pores, as the hydrated radii of the ions are far smaller than the pore diameter. Size‑exclusion contributions appear negligible under the tested conditions.

Conclusions

1. Reactive O₂–Ar sputtering produces Si‑rich alumina layers with superior thermal stability at 850 °C, preventing defect formation during annealing.

2. The stable support yields VA‑SWCNT arrays with a narrower, sub‑2 nm diameter distribution.

3. Nanofiltration membranes built from these arrays achieve higher ion rejection, thanks to smaller pore sizes.

Future work may combine reactive sputtering with post‑treatments such as ambient annealing, O₂ plasma, or ion beam bombardment to further enhance support stability and membrane performance.