Cation-selective layered silicon oxide membranes for power generation

Inorganic two-dimensional membranes offer a new approach to modulating mass transport at the nanoscale. These membranes, which can harness the van der Waals gap as a nanochannel and address persistent challenges in organic membranes, are limited to a few material libraries, such as graphene, graphene oxide, molybdenum disulfide, and boron nitride. Here we report for the first time the development of cation-selective layered silicon oxide membranes, in which the nanochannels, specifically the van der Waals gap, can allow cation diffusion flux to generate an electromotive force for a long time. Considering the abundance and well-known properties of silicon oxide, this inorganic membrane can provide a promising route for membrane separation in a variety of applications.


Introduction
In early nanofluidic research on desalination [1], proton conduction [2], and salinity gradient energy conversion [3,4], silica nanochannels were considered representative ion transport channels among inorganic materials. Owing to their advantages of robustness, chemical stability, and high surface charge density, silica nanochannels can potentially replace conventional ion-exchange organic membranes with inadequate ion transport abilities [1][2][3]. Silica nanochannels are typically prepared by conventional lithographical approaches that generate channel sizes in the tens of nanometers range [5]. However, these lithography-based nanochannels necessitate extensive fabrication processes, making them expensive. It is challenging to fine-tune the properties of the channels at nano-or angstrom-scale precision on a large scale.
Meanwhile, two-dimensional (2D) membranes have been identified as potential candidates among inorganic materials, for enhancing the salinity gradient energy (blue energy) conversion between seawater and river water. Superior performance has been discovered in 2D nanofluidic membranes made of graphene, graphene oxide, molybdenum disulfide, boron nitride, clay minerals, and other atomically thin materials [6][7][8][9][10][11][12][13][14][15][16][17][18]. These efforts have provided evidence for the significance of layered structures that yield angstromscale channels inherent to the van der Waals gap. Layered membranes formed by stacking thin nanosheets have the advantages of high selectivity, easy fabrication, and large scalability. To this end, we recently developed highly cation-selective graphene oxide membranes with nanochannel sizes in the range 4−5 Å and demonstrated that electrical energy could be generated with high energy conversion efficiency [15]. Although silica membranes are excellent candidates because of their hydrophilic properties and the net charge provided by the oxidized sites, only few studies have been conducted with a layered form of silica to date.
There are several important features critical to the preparation of layered silica membranes that enable the sieving of ions. First, the synthesized membranes should be thermodynamically stable in an ambient and humid environment. Accordingly, the formation enthalpy for silica membranes is sufficiently low to prevent the collapse of the crystal structure. Second, the nanochannel size should be maintained as the strong hydrogen-bonded water molecules will likely permeate into the weak vdW-bonded interlayers. Third, the desired ion separation within the nanochannel may be critical to viable ion sieving, for example, in generating an electromotive force.
In this work, we report an approach for achieving layered Si oxide membranes using a topochemical etching of CaSi 2 crystals, where non-equilibrium steady-state cation diffusion occurs through the membranes for energy generation (figure 1). The CaSi 2 framework features [Si] − and [Ca] 2+ layers that are alternately stacked along the direction of the c-axis. Then, Ca 2+ is selectively etched from CaSi 2 , and the silicone backbone is further protonated or oxidized (Si 4+ ) to form layered Si oxide. With the layered Si oxide membranes, we demonstrate for the first time that the nanochannel of the membranes exhibits Na ion selectivity up to 85% and provides a stable voltage of 0.15 V for an extended period (10 h). The following sections describe the preparation method and performance of the layered Si oxide membranes for energy generation.

Fabrication of two-dimensional silicon oxide nanosheets
CaSi 2 (99.9%, Nanoshel), SnCl 2 (98%, Sigma Aldrich), HCl (1.25 M ethanol solution, Sigma Aldrich) and NaCl (>99%, Sigma Aldrich) were used in this study. SnCl 2 was used as an oxidant to deintercalate Ca 2+ ions of CaSi 2 . Then, 0.2 g of CaSi 2 powder and 0.6 g of SnCl 2 were added to 60 ml of ethanol solution and allowed to react for 48 h at 60 • C (CaSi 2 + SnCl 2 + CH 3 CH 2 OH → SiO x + Sn + CaCl 2 + reduzate) [19]. The obtained products were treated with ultrasonication for 30 min for exfoliation. Sn metal nanoparticles, a by-product of the reaction, were removed by reacting the solution with dil. HCl for 10 min.

Characterization
The morphologies of the layered silicon oxide were observed by scanning electron microscopy (SEM, JEOL Ltd, JSM-7001F). A Rigaku Ultima IV x-ray diffractometer with Cu-Kα (λ = 1.5418 Å) radiation (40 kV, 150 mA) was used to perform x-ray diffraction (XRD) analysis. Fourier transform infrared (FT-IR) spectroscopy was performed on a Vertex 70 (Bruker) apparatus. Transmission electron microscopy (TEM, JEOL Ltd, JEM-2100) observations were performed to investigate the microstructure of the layered silicon oxide.

Ion selectivity evaluation
A Ag/AgCl ink electrode (#011464, ALS Co., Ltd) was pasted to both ends of the layered silicon oxide membrane-based cell. The pasted electrodes were dried at 25 • C for a day. Commercial epoxy (S-208, DEVCON) was pasted on the layered silicon oxide-based membrane. The epoxy was dried at 25 • C for a day. A digital multi-meter (Agilent 34401 A) was used to perform the electrical measurements. All the measurements were performed at 25 • C.

Results and discussion
The structure of the CaSi 2 compound consists of corrugated Si planes and Ca 2+ planar monolayers (figure 2(a)) [19,20]. The layered silicon oxide was prepared by the deintercalation of Ca 2+ from CaSi 2 using oxidants. Ca 2+ is selectively etched from CaSi 2 without destroying the silicone backbone in a typical topochemical etching [21][22][23], which is then protonated or oxidized (Si − → Si 0 ) to form layered silicane or polysilane ( p3m1) [24]. However, the resulting Si nanosheets are metastable; therefore, they are likely to be transformed into Si (Fd3m) and then SiO 2 − x at ambient condition [19]. In contrast to using silicon anodes for batteries or electronic applications where oxidation is a significant problem in electrical contact [19,21,25], oxidation of Si is preferable for salinity gradient energy conversion. SnCl 2 was used as an oxidant in this work. As shown in figure 2(b), the color of the solution changes to brownish yellow during Ca 2+ deintercalation. This color change indicates the formation of oxidized Si nanosheets [20]. Figures 2(c) and (d) show the SEM images of the CaSi 2 particle that has no cleave structure. However, after Ca 2+ deintercalation of CaSi 2 , its cleavages are clearly visible, and as a result, remaining Sn metal particles are formed as a by-product between the Si layers (figures 2(e) and (f)). The oxidized Si nanosheets were exfoliated by ultrasonic treatment (supplementary information, figure S1) and then treated with HCl to remove the remaining Sn metal particles.
The surface morphology of the oxidized Si nanosheets was examined using SEM and TEM. The SEM images revealed the sheet-like structures of the stacked and folded oxidized Si nanosheets (figure 3(a)). Figure 3(b) depicts a TEM image of oxidized Si nanosheets with a sheet-like morphology and a lateral dimension greater than 500 nm. The TEM image also shows that the oxidized Si nanosheets have a high degree of transparency, indicating their ultrathin nature. The selected-area electron diffraction pattern in the inset shows that the oxidized silicon nanosheets are mostly amorphous, with no coherent scattering domains [26,27]. Figure 3(c) shows the atomic force microscopy image of the exfoliated oxidized Si nanosheets. The lateral dimension of the layer was a few microns, and the height profile shows that the sheet was about 2-6 nm thick.
Furthermore, to identify the structure of the oxidized silicon nanosheets, the XRD patterns of the nanosheets were compared to those of CaSi 2 and Ca 2+ deintercalated CaSi 2 ( figure 3(d)). After Ca 2+ and Sn were removed (blue, figure 3(d)), the diffraction peaks corresponding to CaSi 2 (navy, figure 3(d)) in oxidized silicon nanosheets almost disappeared, and the diffraction peaks of Sn metal particles (pink, figure 3(d)) were not detected. Oxidization of Si nanosheet was analyzed by the x-ray photoelectron spectroscopy (XPS) (supplementary information, figure S2). Si 2p XPS peak centered at 103.2 eV. The peak could be resolved into four components at 103.6 eV, 102.7 eV, 101.9 eV, and 100.4 eV, representing Si 4+ , Si 3+ , Si 2+ , and Si + , respectively [28,29]. The O/Si atomic ratio of the oxidized Si nanosheets from the XPS analysis is about 1.91, which is consistent with the value based on TEM and energy dispersive x-ray spectroscopy (EDS) analysis (supplementary information, figure S3). The type of functional groups in the oxidized Si nanosheets was confirmed using FT-IR spectroscopy, as shown in figure 3(e). The bands observed at 870, 1040, 1632, 2160, and 3429 cm −1 correspond to the vibration of υ (Si-H), υ(Si-O-Si), υ(Si-OH), υ(Si-H), and υ(-OH), respectively [26,30]. The broad band observed at 3429 cm −1 corresponds to the hydroxyl groups present on the oxidized Si nanosheets formed during the Ca 2+ deintercation step. Based on the FT-IR data, we estimated the surface charge density of oxidized Si nanosheets by measuring the Zeta potential at pH 7 (supplementary information, figure S4). The Zeta potential of the oxidized Si nanosheets was in the vicinity of −36 mV, which is sufficient to attract cations while repelling anions [8,31].
By vacuum-assisted filtration, the exfoliated oxidized Si nanosheets can be easily restacked to form a brown paper-like flexible membrane (figure 4(a)). Figure 4(b) shows the cross-sectional SEM image of the oxidized Si restacked membrane. The cross-sectional image of oxidized Si membrane shows the typical layered microstructure [17,18,32]. The channel size distribution of the oxidized Si membrane was analyzed by Brunauer-Emmett-Teller measurements. The N 2 adsorption-desorption isotherms of the oxidized Si membrane are type IV isotherms with H3 hysteresis loops, which indicates that plate-like particles are generated by the aggregation of oxidized Si nanosheets (supplementary information, figure S5) [33,34]. The channel size distribution was calculated from the nitrogen desorption curve using Barrett-Joyner-Halenda method and found to be 3.8 nm ( figure 4(c)). This implies that the oxidized Si membrane can act as an ion transport medium but not a rejecting barrier. Notably, the channel size of the oxidized Si membrane is suitable for salinity gradient energy conversion of seawater and river water by considering the electric double layer overlap size [35][36][37].
The selective ion transport was investigated by the confined Si oxide nanochannel. The cell is composed of nanochannel connecting two reservoirs: one with a high concentration and the other with a low concentration of NaCl solution ( figure 4(d)). The nanochannel (width = 1 cm, length = 3 mm, height = 5 µm) is physically confined to an epoxy so that its size does not expand in liquid solution [6,15]. A pair of Ag/AgCl electrodes were used to record the open circuit voltage. Figure 4(e) depicts the voltage output versus time for a NaCl electrolyte concentration gradient of 50 (0.5 M/0.01 M). The voltage output increases and saturates at 0.15 V, which equivalent to the sum of the voltage output induced by the selective Na + diffusion in the oxidized Si membrane (diffusion potential, V diff ) and the redox potential at the Ag/AgCl electrodes from unequal chloride concentration (V redox ) [3,8,17,38]. Therefore, the pure V diff induced by the oxidized Si membrane can be extract by subtracting the contribution of V redox from the voltage output. Accordingly, the cation selectivity of the oxidized Si membrane can be quantified by the V diff through the following equation: where, t + and t − are the cation and anion transference numbers respectively. F, R, T, a H , and a L are the Faraday constant, universal gas constant, absolute temperature, and the activity of high concentration electrolyte, and that of low concentration electrolyte, respectively. The cation selectivity is indicated by t + − t -. The cation selectivity of the oxidized Si membrane is calculated to be approximately 85%. However, for the commercial polyvinylidene fluoride (PVDF) filter membrane, the output voltage value decreased rapidly and eventually became zero, which indicates there is no cation selectivity. The oxidized Si membrane can provide stable voltage induced by the selective Na + diffusion for an extended period (10 h), which is distinct from the results of previous studies, where the duration is <1 h [8,39]. Further insights into the cation selectivity of the diffusion process can be obtained by analyzing the elemental distribution of the membrane after the diffusion of ions (supplementary information, figure S6). EDS elemental mapping at the cross-sectional channel reveals that the presence of Cl ions is not apparent, whereas that of Na ions is noticeable. This indicates that the oxidized Si membrane is cation-selective [31]. It also shows that oxidized Si membrane maintains the O/Si atomic ratio of about 1.97 after the diffusion test.

Conclusion
Our work demonstrates that: (a) the cation-selective membranes are made of layered Si oxide and (b) the membranes enable cation transport for an extended period. These results are because of the comparable nanochannel size to that of cation, the negative surface charge of Si oxide attracting the cations, and channel size stability that does not swell in water. We also demonstrated that the cell is readily applicable for generating output power, implying that the cell is feasible for application in the use of seawater in energy conversion. However, layered Si oxide membranes present several shortcomings, which should be addressed. For example, (a) the nanochannel size should be tunable to enhance the ion diffusivity and thus output power; (b) the surface charge of Si oxide should be optimized to increase the selectivity; and (c) potential large-scale cells should be explored in the context of cell-module-pack design. By addressing these critical issues, layered Si oxide can emerge as a promising inorganic membrane for chemical energy conversion.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.