Editorial Board

Hydrogen gas is considered as one of the potential energy sources for various industrial fields because of its abundance in nature and its eco-friendly byproduct (i.e., water) [1]. However, the gas is inherently flammable, colorless, and odorless. Therefore, its leakage should be detected with maximum rapidity [2]. From this perspective, hydrogen gas sensors are required urgently for preventing severe accidents.

The United States Department of Energy (DOE) has set requirements for H₂ sensors, e.g., concentration range (0.1–10%), operating temperature (-30–80 °C), and response time (< 1s) [3]. As a result, substantial efforts have been undertaken to develop highly sensitive and selective H₂ sensors. However, it is challenging to satisfy the above-mentioned requirements.

Among the various types of H₂ sensors, chemiresistive sensors (which are operated by the variations in electrical signals from chemical reactions) have been considered as a potential candidate owing to their rapid response, inexpensive fabrication, and portability. Palladium is conventionally used as a highly effective material for sensing hydrogen gas because of its high selectivity and room-temperature operation [4]. However, Pd-based sensors can be degraded straightforwardly by ambient air and severe volume expansion of Pd. Hence, these sensors are unsuitable for practical use. However, semiconducting metal oxide (SMO)-based chemiresistive sensors have shown high reliability and durability even in ambient air [5]. Nonetheless, SMO-based sensors generally react with numerous gas species, thereby displaying low selectivity toward H₂ gas [6,7].

Here, we propose heterogeneous double-layered H₂ sensors for the first time. These consist of thin-layered SnO₂ nanosheets (NSs) overcoated with a d-space-controlled nanocellulose/graphene oxide (GONC) molecular sieving membrane (SnO₂ NSs@GONC). The d-space of GO was expanded by a nanocellulose fibril buffer layer between GO. This enabled the rapid diffusion of H₂ into the SnO₂ sensing layer. Meanwhile, the GONC molecular sieving layer effectively blocked interfering gas species (e.g., H₂S, CO, acetone, and ethanol). We observed that when a GONC membrane was attached on the SnO₂ sensing layer, GONC selectively filtered hydrogen gas and induced super-selectivity toward it even when the SMO-based sensor was used.

A key strategy for fabricating SnO₂ nanosheets (NSs) overcoated with a d-space-controlled nanocellulose/graphene oxide (GONC) molecular sieving membrane (hereafter, SnO₂ NSs@GONC) is shown in Fig. 1. First, we employed the GO sacrificial templating route for synthesizing porous SnO₂ NSs as a chemiresistive sensing layer [10]. Owing to the strong chemical interaction between numerous functional groups on the GO NSs and Sn precursors, the Sn precursor molecules were decorated uniformly on the GO surface (Fig. 1) [11]. Then, the Sn precursor-loaded GO NSs were calcined at 400 °C for 1 h to form porous SnO₂ NSs.

Fig. 1. 
Schematic illustration of synthesis of SnO₂ NSs@GONC by sacrificial GO templating route and membrane assembly process. Each process is explained in detail in the Experimental Section.

Owing to the thermal decomposition of GO and oxidation of the Sn precursor, the porous SnO₂ NSs were formed after the calcination process. The nanocellulose and GO NSs were employed as a gas filtering membrane. Note that the nanocellulose was used for a d-spacing controller of GO NSs. By applying the simple solution mixing process with nanocellulose and GO NS aqueous solution, nanocellulose fibers were induced to self-assemble on the GO NSs (GONC). Then, the as-prepared SnO₂ NSs were drop-coated on the sensing substrate, and a GONC membrane was decorated on the SnO₂ NSs using the solution drop-coating method. Nanocellulose can play a critical role as a filler layer between GO NSs and thereby, expand the d-space channel pores of GO NSs. As a result, target hydrogen gas molecules may diffuse rapidly into GONC through expanded channel pores of GONC, and interfering gas analytes can be blocked effectively by the surface pores of GONC.

The microstructures of the porous SnO₂ NSs were investigated by a transmission electron microscope (TEM). As shown in Fig. 2(a), Sn precursors were decorated uniformly on the GO NSs (hereafter, Sn-GO NSs), thereby exhibiting ultrathin two-dimensional (2D) morphology. After the calcination step, the TEM image of SnO₂ NSs exhibited morphological characteristics similar to those of Sn-GO NSs. This implies that the GO NSs effectively function as a sacrificial template for 2D morphological SnO₂ NSs during the calcination process (Fig. 2(b)). Furthermore, the sub-10 nm grain sizes of SnO₂ induce a high porosity. This is shown in the magnified TEM image of SnO₂ NSs (Fig. 2(c)). Moreover, energy dispersive spectroscopy (EDS) mapping analysis verified that Sn and O were distributed homogeneously on the SnO₂ NSs (Fig. 2(d)). Atomic force microscope (AFM) analysis was carried out to further investigate the thickness of the SnO₂ NSs (Fig. 2(e)). The AFM result clearly reveals that the SnO₂ NSs exhibited sub-10 nm thickness with a grain size distribution of 4–6 nm (Fig. 2(f), (g)).

Fig. 2. 
TEM images of (a) Sn precursor-assembled GO (Sn-GO NSs), (b, c) porous SnO₂ NSs after heat treatment at 400 °C, (d) energy dispersive spectroscopy (EDS) mapping of SnO₂ NSs, (e, f) atomic force microscopy (AFM) analysis of SnO₂ NSs and thickness distribution, (g) grain size distribution of SnO₂ NSs.

To investigate the channel pores of the d-space-controlled GONC membrane, we carried out high-resolution X-ray diffraction (XRD) analysis with GO and GONC samples. We can accurately estimate the d-space values of GO and GONC using Bragg's law on XRD peaks of GO and GONC samples, respectively. As shown in Fig. 3(a), GONC showed a large d-space value (1.01 nm) compared with GO (0.844 nm). This implies that one-dimensional (1D) nanocellulose between GO NSs plays an effective role as a filler layer for increasing the d-space of GO NSs (schematic image in Fig. 3(a). Based on the cross-sectional scanning electron microscope (SEM) result, we verified that the 34 μm-thick GONC membrane was decorated uniformly on the 21–22 μm-thick SnO₂ NSs (Fig. 3(b)). Furthermore, nanocellulose as a filler between GO membranes was observed by magnified SEM analysis (red box image in Fig. 3(b)).

Fig. 3. 
(a) X-ray diffraction (XRD) patterns of GO and GONC with corresponding d-space values, and schematic images of GO and GONC, (b) cross-sectional SEM images of SnO₂ NSs@GONC, and magnified image of GONC surface.

The sensing measurement test was carried out using the three samples (i.e., SnO₂ NSs, SnO₂ NSs@GO, and SnO₂ NSs@ GONC) to investigate the membrane effect on gas sensing characteristics. Note that we conducted sensing tests with various gas species including H₂, H₂S, CO, CH₃COCH₃, and C₂H₅OH. To fabricate the sensing devices, the porous SnO₂ NSs were drop-coated on an aluminum sensing substrate having an interdigitated Au electrode (width = 25 μm; gap size = 150 μm) [12]. Then, the GO or GONC molecular sieving membrane was overcoated on the SnO₂ NSs sensing layer. As shown in Fig. 4(a), pristine SnO₂ NSs and SnO₂ NSs@GONC sensors showed significant H₂ responses (R_air / R_gas > 10 @ 0.8 % H₂), whereas SnO₂ NSs@GO exhibited relatively low H₂ responses (R_air / R_gas < 5 @ 0.8 % H₂). Moreover, SnO₂ NSs@GO showed low H₂ responses over a wide range of hydrogen-gas concentration compared with SnO₂ NSs and SnO₂ NSs@GONC (Fig. 4(b)). These sensing behaviors indicate that the pristine GO membrane having a small d-space (0.844 nm) could disrupt the gas diffusion into the SnO₂ sensing layer. Meanwhile, the GONC membrane showed high H₂ permeability considering the H₂ response of SnO₂ NSs@GONC. Because the d-space pores of GONC were expanded by the nanocellulose filler layer, the GONC membrane could show significantly higher H₂ permeability. To investigate the selectivity properties of each sensor, we compared the gas responses to H₂S, CO, CH₃COCH₃, and C₂H₅OH. The comparison results revealed that SnO₂ NSs showed low selective sensing properties owing to the high responses to various gases (R_air / R_gas > 5 for H₂S, CH₃COCH₃, and C₂H₅OH) (Fig. 4(c)). However, it is noteworthy that SnO₂ NSs@GO and SnO₂ NSs@GONC showed negligible responses to H₂S, CH₃COCH₃, CO, and C₂H₅OH (Fig. 4(c)). This implies that GO and GONC membranes effectively blocked the interfering gas species (H₂S, CO, CH₃COCH₃, C₂H₅OH) owing to their large kinetic diameter (0.360 nm, 0.376 nm, 0.460 nm, and 0.450 nm, respectively). In general, the surface pores of the GO-based membrane had sub-0.3 nm sizes and thereby, blocked the larger gas molecules [13]. However, H₂ molecules having a kinetic diameter of 0.289 nm could pass through the GO surface, and the expanded d-space of GONC could accelerate the diffusion of H₂ molecules to the SnO₂ sensing layer.

Fig. 4. 
(a) Hydrogen gas sensing properties of SnO₂ NS-, SnO₂ NSs@GO-, and SnO₂ NSs@GONC-based sensors, (b) graph of hydrogen gas response values with respect to hydrogen gas concentration, for each sensor, (c) sensing response kinetics of each sensor to H₂S, CO, acetone, and ethanol, (d) sensing selectivity plots of SnO₂ NS-, SnO₂ NSs@GO-, and SnO₂ NSs@GONC-based sensors

Based on these principles, the SnO₂ NSs@GONC sensor showed a higher H₂ selectivity than the SnO₂ NSs and SnO₂ NSs@GO sensors (Fig. 4(d)). Note that we used S_H2 / S_H2S as an H₂ selectivity factor because the response of the SnO₂ NS-based sensor to H₂S was the dominant response to interfering analytes.

To summarize, we successfully developed the d-space-controlled (0.844–1.01 nm) GONC membrane for exceptional H₂ sensing selectivity. Because of the self-assembly of nanocellulose on the GO NSs, the GONC membrane had expanded d-space pores compared with the conventional GO membrane. Furthermore, we combined the GONC membrane with the SnO₂ NSs-based sensing layer by a simple drop-casting method. Because of the synergistic effect between the GONC membrane (molecular sieving layer) and porous SnO₂ NSs (sensing layer), the SnO₂ NSs@GONC sensor showed a dramatically enhanced selectivity to hydrogen gas. The miniscule size of the surface pores of GONC can effectively block the interfering gas species, whereas the expanded d-space of GONC enables the acceleration of H₂ diffusion to the SnO₂ sensing layer. Based on this improvement, d-space-controlled GO-based membrane combined with sensing layer would facilitate the control of the gas selectivity of chemiresistive gas sensors.