Room Temperature Resistive Hydrogen Sensor for Early Safety Warning of Li-Ion Batteries

Abstract

Lithium-ion batteries (LIBs) have become one of the most competitive energy storage technologies. However, the “thermal runaway” of LIBs leads to serious safety issues. Early safety warning of LIBs is a prerequisite for the widely applications of power battery and large-scale energy storage systems. As reported, hydrogen (H₂) could be generated due to the reaction of lithium metal and polymers inside the battery. The generation of H₂ is some time earlier than the “thermal runaway”. Therefore, the rapid detection of trace hydrogen is the most effective method for early safety warning of LIBs. Resistive hydrogen sensors have attracted attention in recent years. In addition, they could be placed inside the LIB package for the initial hydrogen detection. Here, we overview the recent key advances of resistive room temperature (RT) H₂ sensors, and explore possible applications inside LIB. We explored the underlying sensing mechanisms for each type of H₂ sensor. Additionally, we highlight the approaches to develop the H₂ sensors in large scale. Finally, the present review presents a brief conclusion and perspectives about the resistive RT H₂ sensors for early safety warning of LIBs.

1. Introduction

In recent years, rechargeable lithium-ion batteries (LIBs) have become one of the most competitive energy storage technologies in the fields of portable electronic devices including cell phones and laptops, electric vehicles, and even large-scale energy storage (Figure 1a) due to their eco-friendliness and efficient storage of energy [1,2,3,4,5]. LIBs are usually sealed in stainless-steel, an aluminum or plastic package, consisting of Li-metal-oxide cathode and graphite anode, a polymer separator, and liquid electrolyte. The liquid electrolyte containing lithium salt and organic solvent is conductive, which connects the cathode and anode. Meanwhile, the polymer separator prevents short-circuiting of the cathode and anode. Li ions transport through the electrolyte causing the battery to recharge and re-discharge, as shown in Figure 1b [1].

However, “thermal runaway” of LIBs frequently occurs due to Li dendrite unavoidably growth on graphite anodes under fast or long cycling situations, which is usually accompanied by toxic and flammable gases generation causing serious safety incidents such as fire and explosion (Figure 1c) [6,7,9]. The new regulations of “Safety Requirements for Power Batteries for Electric Vehicles” and “Global Technical Regulations for Safety of Electric Vehicles” require timely warning at least five minutes before serious incidents. In megawatt power grid energy storage of LIBs, security challenges are more serious. Therefore, early safety warning of LIBs is a prerequisite for the widely applications of energy storage systems based on LIBs.

The present safety warning system of LIBs is based on a battery energy storage system, chiefly relying on the protection of battery manage system (BMS), smoke detection, and some special gases detection [10,11]. The current BMS protect a battery by detecting its external surface temperature, voltage, and state of charge. For the special gases (such as hydrocarbons and CO) and smoke detection, they come from the reduction and oxidation of electrolytes at high temperature after “thermal runaway” is occurring [12,13,14,15,16]. Therefore, a more timely and reliable detection method before “thermal runaway” occurs is urgently need.

Recently, some research studies concerning early warning of battery failure based on venting gas signal have been reported [7,8,17,18]. Especially, Jin and Cui et al. [7] reported that H₂ detection could be an efficient safety indicator during the Li dendrite growth period before “thermal runaway” due to the H₂ produced by the reaction of Li dendrite and common electrode polymer binders (i.e., polyvinylidene fluoride (PVDF)) (Figure 1d). When the gas accumulation inside a battery cell reaches a high level, the battery vent ruptures and the gases are leased. Furthermore, H₂ was first detected among released gases (H₂, CO, CO₂, HCl, HF, and SO₂) from the battery and much earlier (over 10 minutes) than the rapid temperature increasing and smoke and fire detection (Figure 1e,f)) [7]. Obviously, H₂ detection inside lithium-ion battery cell is the most effective measure for the early safety warning of LIBs. However, the research on H₂ detection inside battery cells is still in an early stage; literatures covering the relevant experimental research are very limited. Jin et al. [8] further investigated H₂ gas diffusion behavior and the H₂ detector installation of LIBs energy-storage cabin. The H₂ detectors in the cabin could warn about 116 s before thermal runaway, as they set a consistent warning value of 20 ppm for H₂ (Figure 1g).

Various H₂ gas sensors have been studied in the last few decades, such as FET [19,20,21], Schottky barrier [22,23,24,25], Capacitive [26], surface acoustic wave [27,28,29], and resistive [30,31,32]. Among them, resistive H₂ sensors are operated by the transduction of chemical reactions into resistance signals, offering some benefits of low detection limit (LOD) at the ppb level for flexible and simple devices, and could potential work at room temperature. Therefore, they are the most promising for the early safety warning of LIBs [33]. The operation of resistive H₂ sensors rely on the reactions between H₂ and sensing materials. Palladium (Pd)-based H₂ sensors are the current state-of-the-art resistive H₂ sensors, including Pd-based bimetals, and Pd-based composites [31,34,35,36,37]. In addition, homogeneous and heterogeneous designs based on other metals, metal oxide (MOXs), carbon nanotubes (CNTs), graphene, and transition dichalcogenides (TMDs) are studied for efficient H₂ sensing [38]. Although resistive hydrogen sensors have dramatically advanced in recent years, to the best of our knowledge, resistive H₂ sensors integrated inside of LIB cells has not been reported yet.

Due to the harsh application condition of H₂ sensor inside of LIB cell, special requirements are put forward. For examples, high temperature is dangerous for LIBs, so the internal H₂ sensor should operate without heater; for efficient and valid warning, the detection concentration of internal H₂ sensor is suggested to be higher than 20 ppm according to previous reports, with response time of less than 116 s [7,8]; owing to the detection under electrolyte vapors, the internal H₂ sensor needs great selectivity. Therefore, there remain huge challenges for the internal H₂ sensors applied for the early safety warning of LIBs, especially in terms of the accuracy, selectivity, long-term stability, and compatibility while operating at room temperature under harsh environment inside of LIB cell.

In this review, we present a comprehensive overview of the recent advances and underlying mechanisms in resistive H₂ sensors, which are possible to work inside of LIB cell, from three aspects: (1) operating at room temperature, (2) selectivity, and (3) internal gas sensors in LIB. We highlight the recent key advances and strategies for the realization of H₂ sensors in the applications of LIBs. H₂ sensors based on various materials are discussed, as well as the key H₂ sensing mechanisms. Then, we discuss the challenges and prospects of resistive H₂ sensors for the early safety warning of LIBs. We hope that this review could provide and stimulate valuable ideas for the evolution of H₂ sensors in LIBs systems, as well as and propel the development of LIBs.

2. Resistive H₂ Sensor Operating at Room Temperature

For LIBs, a high internal temperature probably induces additional violent exothermic chemical reactions for further heat generation, resulting in thermal runaway of LIBs. Hence, room temperature H₂ sensors applied inside of LIB is necessary. Resistive H₂ sensors operating at room temperature mainly rely on the catalytic adsorption of H₂ molecule on noble metal (for example Pd, Au, Pt). Among of them, Pd-based H₂ sensors are the current state-of-the-art resistive H₂ sensors. Beyond that, synergistic effect of heterojunctions, metallization effect of special metal oxide [39,40], and efficient H₂ sensing capacity of emerging materials (i.e., MXene, TMDs) at low temperatures have also been investigated to develop RT H₂ sensor.

2.1. Pd-Based H₂ Sensors

Basic H₂ Sensing Mechanism

Generally, the H₂ sensing mechanism of Pd-based H₂ sensors are catalytic adsorption of H₂ on Pd, which induces the dissociation of H₂ molecule into H atoms and the formation of PdH_x (Equations (1) and (2)) [31,41].

$${\mathrm{H}}_2\left(\mathrm{gas}\right)\to {\mathrm{H}}_2\left(\mathrm{ads}\right)$$

(1)

$$\mathrm{Pd}+\frac{\mathrm{x}}{2}{\mathrm{H}}_2\left(\mathrm{ads}\right)\to {\mathrm{PdH}}_{\mathrm{x}}$$

(2)

Numerous resistive Pd-based H₂ sensors have been developed, including pure Pd nanostructures and Pd-based composites, such as Pd-hetero metal, Pd-MOX, Pd-carbon material, and Pd-TMD. In this section, we discuss the sensing mechanisms, effect factors and key advances in recent several years of the resistive H₂ sensors based on Pd and Pd composites.

Pure Pd materials. According to Equations (1) and (2), PdH_x is formed when H₂ molecules adsorb on Pd surface. Conductivity of PdH_x is poorer than Pd, which generates H₂ response in resistance of Pd [31,42]. Plenty of resistive H₂ sensors based on pure Pd materials have been reported. With the significant development of nanoscience, various Pd nanostructures with further H₂ sensing mechanisms were developed to enhance the H₂ sensing capacities [43,44,45,46].

Kim et al. [43] and Jung et al. [44] investigated the nanograin effects and the α-to-β phase transition of Pd on H₂ sensing. When the Pd crystallites < 10 nm, abundant nanogaps were created between the nanograins (<2 nm). The nanogaps between Pd nanograins intrinsically stem the electrical conduction. Exposed to different concentration of H₂, different kinds of conducting pathways were formed in Pd, yielding switchable H₂ sensing behaviors (Figure 2a,b) [44]. Similar phenomena were demonstrated in Ref. [31] (Figure 2c–g): (i) when exposed to low concentrations of H₂ (2.5 to 100 ppm), α-PdH_x was formed expanding Pd nanograins, which closed the narrow nanogaps and further producing a new conduction pathway. Obvious negative resistance response was observed. (ii) Upon exposed to moderate concentrations of H₂ (250 to 2500 ppm), most of the nanogaps were closed, resulting in saturation of the negative resistance response. (iii) Upon exposed to high concentrations of H₂ (0.5 to 3%), all nanogaps were closed and β-PdH_x were formed, while surface electron scattering generated by β-PdH_x dominantly the electrical conductions, leading to significant positive resistance response. Consequently, ultrasmall grain size of Pd are critical for H₂ detection at low concentrations, α-to-β phase transition of Pd could detect a wide range of H₂ concentrations from sub-ppm to 4%. In addition, negative effect of oxygen on Pd-based H₂ sensing has been demonstrated in earlier literatures [47,48]. We will discuss the effect in a later section.

Pd-metal alloy. Recently, many researchers developed Pd-based alloys to improve the H₂ sensing properties, such as limit detection and response/recovery behavior [35,49,50,51]. Jung et al. [49] developed Pd/Pt and Pd/Au high surface-to-volume ratio nanopatterns (Figure 3a). Compared to Pd nanopattern, Pd/Pt nanopattern showed 45.5-fold higher response to 1% H₂ (Figure 3b,c), Pd/Au nanopattern showed about 73-fold and 4.6-fold enhancement in the response and recovery speed, respectively (Figure 3d–f). The dramatic improvements in response and recovery behaviors to H₂ are attributed to the ultrasmall size (<5 nm), ultrathin nanopattern (<15 nm), grain interfaces and the lower adsorption/dissociation energy of H₂ on Pd/Pt and Pd/Au surfaces (Figure 3g–j). Kim et al. [51] developed hollow Pd-Sn alloy nanotubes with high surface area of 223.0 m²/g for H₂ sensing and ultrafine grain size. Interestingly, the Pd-Sn alloy effectively prevented degradation of H₂ sensing performances caused by the α-β transition of Pd. As a result, the hollow Pd-Sn nanotubes exhibited outstanding sensing properties to a wide concentration range of H₂ (50 ppm to 3%), especially fast response/recovery rate to high concentration of H₂ (20 s/17 s to 2% H₂, Figure 4a–i). The noteworthy H₂ sensing mechanism was proposed: (i) The abundant interface gaps between the grains owing to larger atom size of Sn (0.141 nm) and smaller atom size of Pd (0.138 nm) dominated low concentration of H₂ (0.5 ppm to 0.02%) sensing behavior, due to the nanogap effect. (ii) For high concentration of H₂ (0.05% to 3%), the formation of β-PdH_x governed the H₂ sensing properties (Figure 4j). The H₂ sensing mechanism agreed with the previous reports [43,44].

Pd-MOXs. MOXs are the classical gas sensing semiconductor materials. However, their H₂ sensing properties are limited at room temperature. Previously, many researchers developed Pd-MOX composites to achieve good H₂ sensing at room temperature. Here, we overview some interesting representative approaches of H₂ sensors based on Pd-MOX composites, reported recently.

Zhang et al. [46] investigated the interconvertible effect on H₂ sensing of catalyst nanoparticles and semiconductor support in Pd-decorated PdO hollow shells. They prepared PdO hollow shells, which were subsequently treated by NaBH₄ to be partially reduced into Pd on the PdO surface. The Pd nanoparticles were discretely and physically inlaid on the surface of PdO with a ultrasmall size of ~ 2 nm (Figure 5a). The catalytic effect of Pd, as well as the Schottky-junction between Pd and PdO, enhanced the H₂ sensing performances even at 1 ppm (Figure 5b–d). Notably, inlaid Pd in PdO shells prevented the agglomeration of Pd nanoparticles, which generated long-term stability (Figure 5e).

In addition, Zhang et al. [36] designed an unique conduction model by inserting a high-conductive metallic core Au into less-conductive p-type PdO to boost the RT H₂ sensing performances (Figure 6). As a result, Pd decorating Au@PdO demonstrated an ~90 times larger in H₂ response than Pd decorating Pd@PdO. And the boosted H₂ response helped the Pd-Au@PdO sensor showed ultralow LOD of 0.1 ppm. Pd-ZnO nanoflowers has been demonstrated RT H₂ sensing in ppb level with experimental LOD of 300 ppb, due to the change in channel conductance of ZnO nanoflowers based on the incorporation of Pd [37].

Pd-carbon materials. Carbon nanotubes, graphene, and their derivatives are very promising in the room temperature gas sensor fields, due to their good electrical conductivity at RT. However, H₂ molecule is very difficult to adsorb on the surface of carbon materials, due to weak H₂ adsorption capacity. In regard of this case, lots of Pd-carbon material composites have been previously investigated for RT H₂ sensors. Even so, low LOD and quick recovery behavior remain big challenges. As is well known, Pd particles at ultrasmall size are very helpful for the RT H₂ sensing. However, they are very easy to aggregate to further cause degradation of overall H₂ response.

For that, we designed Pd-graphene system via DNA assistant for trace of H₂ sensing [33]. During the synthesis process of one-pot solution method, DNA suppressed the stacking of graphene layers due to π-π interaction between DNA and graphene, meanwhile in-situ anchored PdO₂ subnanoscale clusters on the exfoliated single layer graphene (Figure 7a,b). The design showed mimic wrinkled morphology and sensing mechanism of natural olfactory neuroepithelium, being named BONe (Figure 7a). The BONe boosted H₂ sensing performance (25 s/35 s at 5 ppm H₂ with experimental LOD of 50 ppb) at RT with yearlong durability (Figure 7c–e). High surface area of the BONe, good conductivity of graphene, as well as the subnanoscale of PdO₂, generated the ultra-sensitivity to ppb-level H₂ at RT; subnanoscale of PdO₂ anchored on graphene yielded yearlong stability. Notably, it was calculated that PdO₂ exhibited a d-band downshift of −2.58 eV, which was a much further downshift than that of PdO (−2.16 eV) and Pd (−0.179 eV), governing the complete and fast recovery behavior during H₂ detection (Figure 7f–h). A similar d-band theory was proposed previously, that tuning of d-band energy level balanced the adsorption and desorption capacities, and the lower the d-band level, the weaker the adsorption [52].

Pd-other materials. With the emerging of new materials (metal-organic frame MOF, Mxene, et al.), composites of Pd and kind of new materials were developed for H₂ sensing. The sensor based on flexible Ti₃C₂T_x MXene@Pd nanoclusters, reported in Ref. [53] delivered a response/recovery time of 32/161 s and sensitivity of 23% to 4% H₂ at RT (Figure 8a–c). The strong H₂ adsorption into lattice of Pd nanoclusters induced electrons doping in MXene, generating fast response behavior (Figure 8d). However, the limit of detection (0.5%) needs to be improved. The Pd-decorated sodium titanate nanoribbons (Pd-NTO NRs) developed by Zhang et al. [54] exhibited ultrafast response to 1% H₂ within 1.1 s at RT, and a wide detection range (0.8 ppm to 10% H₂). The excellent H₂ sensing capacity benefited from the laterally paralleled morphology and abundant oxygen vacancies on edge sites of nanoribbons, as well as monodispersed Pd nanoparticles in size of ~3.5 nm. Oxygen in air could block the reactive sites, leading to depress the H₂ sensitivity and retard the response/recovery rate [47,48]. For that, Kim and Penner et al. [55] designed patterned Pd nanowires covering ZIF-8 membrane (Pd NWs@ZIF-8) for H₂ sensing. Although the ZIF-8 membrane reduced the H₂ response slightly, 20-fold faster response/recovery rate (3.5% at 10/7 s to 1% H₂ versus 5.9% at 164/229 s of Pd nanowires) was achieved due to the molecular sieving and acceleration effects of ZIF-8 (Figure 8e) since the pore size of ZIF-8 is 0.34 nm, which is larger than the diffusion kinetic diameter of H₂ molecule (0.289 nm), but a little smaller than that of O₂ molecule (0.345 nm) (Figure 8e).

2.2. Other Noble Metal-Based H₂ Sensors

Similar to H₂ adsorption property of Pd, other noble metals also exhibit H₂ adsorption behavior [56]. Moreover, Au and Pt-based H₂ sensors has been demonstrated due to the formation of MH_x [57,58,59]. Guha et al. [57] reported Pt-functionalized rGO for excellent H₂ sensing at RT, 65 s/230 s against 5000 ppm H₂ with LOD of 200 ppm in air ambience. Interestingly, for this Pt-rGO composite, H₂ response was larger in N₂ environment than that in air ambience; however, sensor recovered faster in air than in N₂. In N₂ and air environments, H₂ first physisorbed on Pt-rGO surface and then dissociated to from Pt-H (Equation (3)). But in air, H₂ also reacted with the adsorbed oxygen on the Pt-rGO surface to form Pt-H and H₂O (Equations (4) and (5)).

$$\mathrm{Pt}+\frac{1}{2}{\mathrm{H}}_2\to \mathrm{Pt}-\mathrm{H}$$

(3)

$$\mathrm{Pt}+\frac{1}{2}{\mathrm{O}}_2\to \mathrm{Pt}-\mathrm{O}$$

(4)

$$\mathrm{Pt}-\mathrm{O}+\frac{3}{2}{\mathrm{H}}_2\to \mathrm{Pt}-\mathrm{H}+{\mathrm{H}}_2\mathrm{O}$$

(5)

The H₂O competed with H₂ molecules to adsorb on the Pt-rGO surface, reducing the H₂ response. On the contrary, the adsorbed H₂ and the dissociated H-atoms reacted with adsorbed oxygen to convert to water vapor while purging air, exhibiting fast recovery in air ambience. Similar effect of oxygen on Pd- and Pd@Pt-based H₂ sensing has been demonstrated in earlier literatures (Figure 9) [47,48].

2.3. MOXs-Based H₂ Sensors

Generally, most MOXs-based resistive gas sensors operated at high temperature (>150 °C). However, recent approaches were developed to realize RT H₂ sensing performances of MOXs, such as nanostructure design [60,61,62], composites of MOXs [63,64], and surface metallization [37]. Huang et al. [61] demonstrated the well-aligned MoO₃ nanoribbon arrays exhibited great H₂ response/recovery behavior at RT, with a response/recovery time of 3 s/16 s at 100 ppm H₂ which is much shorter than 59 s/151 s of randomly arranged MoO₃ nanoribbons (Figure 10a–e). The accelerated H₂ response/recovery rate was due to the fact that the high alignment of nanoribbons could increase the surface activity of MoO₃ and suppress the nanojunction effect. On the contrary, in the randomly arranged MoO₃, the interface diffusion of adsorbed oxygen caused serious nanojunction effect, resulting in much slower H₂ response/recovery rate.

Yun et al. [37] designed holey engineered 2D ZnO-nanosheets for supersensitive H₂ sensing (Figure 10f–j). The sensor exhibited 115% response to 100 ppm H₂ with short response/recovery time of 9 s/6 s at room temperature (Figure 10g). And the experimental LOD was 5 ppm (Figure 10f). Upon exposure to H₂, surface of the ZnO nanosheets became metallic Zn (Figure 10h,i). Metallization of Zn on the ZnO surface governed the gas sensing mechanism about high response and great selectivity to H₂ at room temperature (Figure 10h–j). Moreover, the synergetic effect of 2D nanosheets and interconnected holey/porous network of ZnO generated the excellent H₂ sensing performances by offering abundant active sites for H₂ molecules.

2.4. MoS₂-Based H₂ Sensors

In recent year, MoS₂ has shown great H₂ sensing potential due to its 2D van der Waals structures, H₂ adsorption capacity and enable RT operation [65,66]. Unfortunately, sluggish response/recovery limited the applications of MoS₂-based H₂ sensors. In regard of this case, some composites based on MoS₂ were investigated for H₂ sensing. Huang and Chen at al [38] designed hybrid interlinked MoS₂-ZnO nanotubes for RT H₂ sensing of 51.1% to 500 ppm with 14/19 s response/recovery and LOD of 10 ppm, due to the increased oxygen vacancies and surface-active sites. Hollow MoS₂/Pt-based chemiresistors, designed by Kim et al. [65], exhibited great H₂ sensing performance with fast response/recovery rate at RT (8.1/16 s for 1%, and 2.7/16 s for 4% H₂), due to the catalytic H₂ spillover of Pt, as well as sufficiently permeable pathways and maximized active sites for H₂ produced by the hollow MoS₂. Similarly, vertically aligned edge-oriented Pd/MoS₂ nanofilm was reported for great H₂ sensing properties at RT, response of 33.7% to 500 ppm with response/recovery of 16 s/38 s, and LOD of 50 ppm [66]. Even so, selectivity of RT H₂ sensors based on MoS₂ has not been investigated widely and deeply.

2.5. Other-Based H₂ Sensors

With the development of nanoscience and nanotechnology, new composites for RT H₂ sensors have been developed. Xu and Ou et al. [67] explored ultrathin nickel oxysulfide which exhibited a selective and fully reversible response to H₂ at RT for a wide range from 0.25% to 1%, due to the physisorption of H₂ on the surface. Dash et al. [68] developed RT H₂ sensors based on rGO-ZnFe₂O₄-Pd nanocomposite, showing high sensitivity and fast response/recovery rate (11.43% to 200 ppm with 18/29 s response/recovery) to a wide range of H₂ (50–1000 ppm), due to the synergistic effect of rGO, ZnFe₂O₄ and Pd nanoparticles.

Up to now, there are lots of efforts to develop room temperature resistive H₂ sensors. Unfortunately, no H₂ sensors integrated inside of LIB cells have been reported. Table 1 summarizes the recent and representative room temperature resistive H₂ sensors possibly integrated inside of LIB cells. Although new gas sensing materials are emerging all of the time, Pd and Pd-based materials have been the current state-of-the-art resistive H₂ sensing material operating at room temperature. In conclusion, Pd is still the most excellent catalyst for RT H₂ sensing with great selectivity due to the formation of PdH_x. But the LOD (ppm level) and response/recovery behavior (tens of seconds) of Pd-based H₂ sensors should be further improved. Sensors based on Pd-carbon materials should be state-of-the-art RT H₂ sensors displaying LOD of ppb-level (due to the catalytic H₂ adsorption/dissociation on Pd surface, as well as high surface area and RT good conductivity of carbon materials). However, they still suffer from slow response/recovery rate (tens of seconds) to ppb-H₂ at RT. According to previous investigations, atomically dispersed sub-nano clusters (perhaps single atom) of Pd are in favor of ppb-level H₂ detection at RT; optimal d-band energy level of Pd could yield complete and fast recovery; construction of conduction channel in Pd-based sensing material could generate fast response and recovery rate. Although some composites of MOXs exhibited good RT H₂ sensing property, they inherently showed cross-sensitivity to other gases (such as NO₂, CO and ethanol) and susceptibility to humidity. Despite some new structures and materials (ternary composites, MoS₂, and MXene) emerging for RT H₂ sensing, fast response in 1 s to ppb-H₂ with special selectivity is still a critical issue.

3. Selectivity

LIBs usually undergo the following changes with abuse: lithium precipitation at the negative electrode, decomposition of solid electrolyte interface (SEI) layer, melting of the separator, reaction of electrolyte with anode, decomposition of the positive electrode, decomposition, and vaporization of the electrolyte [8]. Multiple side reactions can occur simultaneously, producing gas mixture, including H₂, CO, water vapor, and electrolyte vapors. Therefore, selectivity of H₂ sensors inside of LIB is a critical issue for the early safety warning of LIBs.

Chemo-resistive H₂ sensors generally suffer from cross-sensitivity with other gases (for example, CO, hydrocarbon gases, water vapor, ethanol vapor, and NO₂). Synergistic effect of heterogeneous composite, adjustment of dangling bonds on the sensing material surface, and filter membrane are conventional means to improve the selective response to H₂. Pd has been extensively decorated on the major sensing materials to improve the H₂ selectivity, due to the high H₂ adsorption capacity [37]. Heterogeneous composites of MOX-MOX, MOX-carbon material, and MOX-MoS₂ were usually developed for great selective H₂ response due to the synergistic effect of heterojunctions. In our previous report [69], reducing the surface oxygen-containing functional groups via heat treatment under H₂ condition was verified to improve the H₂ selectivity by inhibiting the intercross response to oxidizing gas (NO₂) and organic gas (ethanol). As one of the obstinate problems for room temperature resistive gas sensors, humidity interference has been troubling researchers in decades. The humidity-resistant properties of Ag₂Te-based NO₂ sensor operating at room temperature was demonstrated in a recent report [70]. Adelung and Lupan et al. [71] developed ZnO-based H₂ sensor employing graphene oxide as molecular sieve (Figure 11a). The nanopores in the GO membrane, acting as a size-selective sieve, only allowed permeation of H₂ molecules among the tested gases (ethanol, methane, ammonia, acetone, methanol and H₂). In other reports [24,72], poly(methyl methacrylate) (PMMA) membrane layers have demonstrated the selective H₂ filtration effect, as shown in Figure 11b,c. ZIF-8 membrane has also been investigated to improve H₂ sensing selectivity [55]. Since the pore size of ZIF-8 is 0.34 nm, which is larger than the diffusion kinetic diameter of H₂ molecule (0.289 nm), but a little smaller than that of the other tested gases. For the internal H₂ sensor working under harsh conditions, filling of interface gases (electrolyte vapors, CO, et al.) in the filter membrane could be a better method to inhibit inter-cross selectivity.

4. Internal Gas Sensors in LIBs

Internal sensor in LIB is much more efficient and timelier for the early safety warning of LIBs. Internal sensors about temperature detection (Figure 12a) [73], and the flexible three-in-one microsensor of internal temperature, voltage and current detection (Figure 12b) [74] have been reported. Unfortunately, to the best of our knowledge, the internal gas sensors (including H₂ sensor) have not been reported yet. Internal gas sensors in LIB should be flexible, stable under serious condition filling with electrolyte vapors and poor air, while operating at room temperature. Flexible gas sensors, including H₂ sensors, have been widely investigated [35,45,53,75,76].

Lee et al. [24] demonstrated that a PMMA membrane coating on Pd-graphene sensing layer could prevent the NO₂, CO and CH₄ response completely (Figure 11b). Li et al. [77] developed Pd-CNTs/PDMS/POTS (PDMS, polydimethylsiloxane; POTS, 1H,1H,2H,2H-perfluorooctyltriethoxysilane) sensors for self-cleaning and humidity-insensitive H₂ sensors at RT. Superhydrophobicity of PDMS/POTS generated the waterproof and self-cleaning properties of the sensor, maintaining the H₂ sensing capacity under highly humidity conditions. Moreover, the self-healing of superhydrophobicity yield the long-term stability of the sensor (Figure 13a). In the work of flexible NO₂ sensor [78], semipermeable PDMS membrane was used to achieve the water-resistant property with an ultralow LOD of 8.3 ppb (Figure 13b). The research on internal H₂ sensor inside of LIB cell could be inspired by plenty of the flexible gas sensor research. A flexible thin film sensor with self-cleaning and H₂ permeable membrane could generate long-term stability inside of LIB cell.

Table 1 summarizes the recent and representative resistive H₂ sensors operating at room temperature based on inorganic materials.

Table 1. Recent Studies on Resistive H₂ sensors operating at RT.

Material & Morphology	H2 Response $\left|{\mathit{R}}_{{\mathit{H}}_2}-{\mathit{R}}_{\mathit{a}\mathit{i}\mathit{r}}\right|/{\mathit{R}}_{\mathit{a}\mathit{i}\mathit{r}}\times 100\mathit{\%}$	tresp/trec b	LOD c (ppm)	Measurement Range	Ref. d
Pd nanofiber yarn	1.37% at 0.1%	236 s/388 s at 0.1%	2	1 ppm to 4%	[43]
Pd nanofiber yarn	0.88% at 0.1%	76 s/384 s at 0.1%	1	1 ppm to 4%	[43]
Pd nanopattern	0.8% at 0.1%	230 s/680 s at 0.1%	2.5	2.5 ppm to 4%	[44]
Pd nanotube array	1000% at 0.1%	180 s/n.r. at 0.1%	100	100 ppm to 1%	[45]
Pd/Pt nanopattern	2% at 1%	7 s/35 s at 0.1%	10	10 ppm to 1%	[49]
Pd/Au nanopattern	n.r.	8 s/30 s at 0.1%	10	10 ppm to 1%	[49]
Pd/Mg film	3% at 1%	6 s/32 s at 1%	1	1 ppm to 4%	[50]
Pd–Sn Alloy Nanotube	1.63% at 200 ppm	20 s/18 s at 200 ppm	1	0.5 ppm to 3%	[51]
PdMo alloy nanosheet	18.7% at 1%	73 s/40 s at 1%	1	0.01% to 1%	[35]
Pd-PdO Hollow Shells	4.6% at 1 ppm	5 s/32 s at 1%	1	1 ppm to 1%	[46]
Pd-ZnO nanoflowers	45% at 10 ppm	137 s/165 s at 300 ppb	0.3	0.3 to70 ppm	[37]
Pd/ZnO	39.2% at 1000 ppm	68 s/n.r. at 1000 ppm	1000	0.1% to 2%	[75]
Pd-WO3 nanoparticle	1786.3 a at 100 ppm	41 s/n.r. at 100 ppm	1	1 to 100 ppm	[79]
MWCNT@Pd nanosheets	3.6% at 1%	74 s/35 s at 1%	5	0.02% to 1%	[80]
Pt/g- C3N4 film	51% at 10,000 ppm	39 s/5 s to 10,000 ppm	1%	1% to 10%	[81]
Pd-decorated crumpled rGO	14.8% at 2%	73 s/126 s at 2%	25	25 ppm to 2%	[82]
Pd/porous graphene	0.6% at 600 ppm	n.r.	600	600 ppm to 1.3%	[83]
Pd/graphene	66% at 2%	1.8 min/5.5 min at 2%	25	0.025% to 2%	[72]
Pd sub-nano clusters on graphene	15% at 5 ppm	25 s/35 s at 5 ppm	0.05	50 ppb to 5 ppm	[33]
Pd nanowires@ZIF-8 Core-shell	0.8% at 0.1%	8 s/30 s at 0.1% H2	600	600 ppm to 1%	[55]
Vertically aligned Pd/MoS2 nanofilm	33.7% at 500 ppm	16 s/38 s at 500 ppm	50	50 ppm to 1%	[66]
Pd-sodium titanate nanoribbons	12.0 a to 1%	1.1 s/13.5 s at 1%	0.8	0.8 ppm to 10%	[54]
Pd- sodium titanate nanoparticles	5.7 a to 1%	13.3 s/39 s at 1%	100	100 ppm to 5%	[54]
Ti3C2Tx@Pd nanoclusters	23% at 4%	32 s/161 s at 4%	0.5%	0.5% to 40%	[53]
rGO-ZnFe2O4–Pd	11.43% to 200 ppm	18 s/39 s at 200 ppm	50	50 to 1000 ppm	[68]
Pt/rGO	97% at 500 ppm	65 s/230 s at 5000 ppm	200	200 to 5000 ppm	[57]
Pt/3D graphene	6.1% at 1%	25 s/20 s at 1%	10	10 ppm to 1%	[59]
Pt-PdO nanowires	23% at 100 ppm	166 s/445 s at 0.1%	10	10 to 100 ppm	[84]
Au@PdO nanoparticle arrays	~180 a at 1%	n.r.	0.1	0.1 ppm to 1%	[36]
PdO-PdAu Ternary Hollow Shells	n.r.	2.2 s/23 s at 0.1%	30	30 ppm to 1%	[85]
Hollow MoS2/Pt	8.7% at 1%	8 s/16 s at 1%	500	500 ppm to 4%	[65]
2D holey ZnO	115% at 100 ppm	9 s/6 s at 100 ppm	5	5 to 100 ppm	[39]
WO3-TiO2 composite	5.62 a at 10,000 ppm	48 s/5 s at 10,000 ppm	1000	1000 to 10,000 ppm	[63]
MoS2-ZnO nanotubes	0.51 a at 500 ppm	14 s/19 s at 500 ppm	10	10 to 500 ppm	[38]
SnO2-coated β-Ga2O3 nanobelts	115% at 33 ppm	216 s/125 s at 33 ppm	33	33 to 1000 ppm	[64]
Nanocrystalline SnO2 thin film	48% at 3 ppm	135 s/46 s at 3 ppm	3	3 to 100 ppm	[60]
Well-aligned MoO3 nanoribbon arrays	~3% at 100 ppm	3 s/16 s at 100 ppm	100	100 to 500 ppm	[61]
ZnO microwires with GO membrane	3.42 a at 1000 ppm	114 s/30 s at 1000 ppm	10	10 to 1000 ppm	[71]
ordered mesoporous TiO2	298 a at 1000 ppm	85 s/198 s at 1000 ppm	100	100 to 1000 ppm	[62]
3D nickel oxysulfide micro-flowers	3.24% at 1%	20 min/33 min to 1%	0.25%	0.25% to 1%	[67]

n.r. indicates not reported. ^a Response is defined as $R_{H_2}/R_{air}$. ^b Response is defined as the time taken from baseline to 90% of the maximum response; recovery time is defined as the time taken from the maximum response to 10% of maximum response. ^c LOD is the limit of detection measured by experiment. ^d Ref. is the reference number.

Table 1. Recent Studies on Resistive H₂ sensors operating at RT.

Material & Morphology	H2 Response $\left|{\mathit{R}}_{{\mathit{H}}_2}-{\mathit{R}}_{\mathit{a}\mathit{i}\mathit{r}}\right|/{\mathit{R}}_{\mathit{a}\mathit{i}\mathit{r}}\times 100\mathit{\%}$	tresp/trec b	LOD c (ppm)	Measurement Range	Ref. d
Pd nanofiber yarn	1.37% at 0.1%	236 s/388 s at 0.1%	2	1 ppm to 4%	[43]
Pd nanofiber yarn	0.88% at 0.1%	76 s/384 s at 0.1%	1	1 ppm to 4%	[43]
Pd nanopattern	0.8% at 0.1%	230 s/680 s at 0.1%	2.5	2.5 ppm to 4%	[44]
Pd nanotube array	1000% at 0.1%	180 s/n.r. at 0.1%	100	100 ppm to 1%	[45]
Pd/Pt nanopattern	2% at 1%	7 s/35 s at 0.1%	10	10 ppm to 1%	[49]
Pd/Au nanopattern	n.r.	8 s/30 s at 0.1%	10	10 ppm to 1%	[49]
Pd/Mg film	3% at 1%	6 s/32 s at 1%	1	1 ppm to 4%	[50]
Pd–Sn Alloy Nanotube	1.63% at 200 ppm	20 s/18 s at 200 ppm	1	0.5 ppm to 3%	[51]
PdMo alloy nanosheet	18.7% at 1%	73 s/40 s at 1%	1	0.01% to 1%	[35]
Pd-PdO Hollow Shells	4.6% at 1 ppm	5 s/32 s at 1%	1	1 ppm to 1%	[46]
Pd-ZnO nanoflowers	45% at 10 ppm	137 s/165 s at 300 ppb	0.3	0.3 to70 ppm	[37]
Pd/ZnO	39.2% at 1000 ppm	68 s/n.r. at 1000 ppm	1000	0.1% to 2%	[75]
Pd-WO3 nanoparticle	1786.3 a at 100 ppm	41 s/n.r. at 100 ppm	1	1 to 100 ppm	[79]
MWCNT@Pd nanosheets	3.6% at 1%	74 s/35 s at 1%	5	0.02% to 1%	[80]
Pt/g- C3N4 film	51% at 10,000 ppm	39 s/5 s to 10,000 ppm	1%	1% to 10%	[81]
Pd-decorated crumpled rGO	14.8% at 2%	73 s/126 s at 2%	25	25 ppm to 2%	[82]
Pd/porous graphene	0.6% at 600 ppm	n.r.	600	600 ppm to 1.3%	[83]
Pd/graphene	66% at 2%	1.8 min/5.5 min at 2%	25	0.025% to 2%	[72]
Pd sub-nano clusters on graphene	15% at 5 ppm	25 s/35 s at 5 ppm	0.05	50 ppb to 5 ppm	[33]
Pd nanowires@ZIF-8 Core-shell	0.8% at 0.1%	8 s/30 s at 0.1% H2	600	600 ppm to 1%	[55]
Vertically aligned Pd/MoS2 nanofilm	33.7% at 500 ppm	16 s/38 s at 500 ppm	50	50 ppm to 1%	[66]
Pd-sodium titanate nanoribbons	12.0 a to 1%	1.1 s/13.5 s at 1%	0.8	0.8 ppm to 10%	[54]
Pd- sodium titanate nanoparticles	5.7 a to 1%	13.3 s/39 s at 1%	100	100 ppm to 5%	[54]
Ti3C2Tx@Pd nanoclusters	23% at 4%	32 s/161 s at 4%	0.5%	0.5% to 40%	[53]
rGO-ZnFe2O4–Pd	11.43% to 200 ppm	18 s/39 s at 200 ppm	50	50 to 1000 ppm	[68]
Pt/rGO	97% at 500 ppm	65 s/230 s at 5000 ppm	200	200 to 5000 ppm	[57]
Pt/3D graphene	6.1% at 1%	25 s/20 s at 1%	10	10 ppm to 1%	[59]
Pt-PdO nanowires	23% at 100 ppm	166 s/445 s at 0.1%	10	10 to 100 ppm	[84]
Au@PdO nanoparticle arrays	~180 a at 1%	n.r.	0.1	0.1 ppm to 1%	[36]
PdO-PdAu Ternary Hollow Shells	n.r.	2.2 s/23 s at 0.1%	30	30 ppm to 1%	[85]
Hollow MoS2/Pt	8.7% at 1%	8 s/16 s at 1%	500	500 ppm to 4%	[65]
2D holey ZnO	115% at 100 ppm	9 s/6 s at 100 ppm	5	5 to 100 ppm	[39]
WO3-TiO2 composite	5.62 a at 10,000 ppm	48 s/5 s at 10,000 ppm	1000	1000 to 10,000 ppm	[63]
MoS2-ZnO nanotubes	0.51 a at 500 ppm	14 s/19 s at 500 ppm	10	10 to 500 ppm	[38]
SnO2-coated β-Ga2O3 nanobelts	115% at 33 ppm	216 s/125 s at 33 ppm	33	33 to 1000 ppm	[64]
Nanocrystalline SnO2 thin film	48% at 3 ppm	135 s/46 s at 3 ppm	3	3 to 100 ppm	[60]
Well-aligned MoO3 nanoribbon arrays	~3% at 100 ppm	3 s/16 s at 100 ppm	100	100 to 500 ppm	[61]
ZnO microwires with GO membrane	3.42 a at 1000 ppm	114 s/30 s at 1000 ppm	10	10 to 1000 ppm	[71]
ordered mesoporous TiO2	298 a at 1000 ppm	85 s/198 s at 1000 ppm	100	100 to 1000 ppm	[62]
3D nickel oxysulfide micro-flowers	3.24% at 1%	20 min/33 min to 1%	0.25%	0.25% to 1%	[67]

n.r. indicates not reported. ^a Response is defined as $R_{H_2}/R_{air}$. ^b Response is defined as the time taken from baseline to 90% of the maximum response; recovery time is defined as the time taken from the maximum response to 10% of maximum response. ^c LOD is the limit of detection measured by experiment. ^d Ref. is the reference number.

5. Conclusions and Perspective

H₂ detection inside lithium-ion battery cells is the most effective measure for the early safety warning of LIBs. The special application condition proposed special requirements for H₂ sensors in terms of working at room temperature, response time, detection concentration, selectivity, stability, and even packaging.

Gas sensing material is the core of a gas sensor. In the past few decades, chemiresistive RT H₂ sensors based on various materials were investigated, including noble metals (Pd, Pt), metal oxides (ZnO, SnO₂, TiO₂, et al.), carbon materials (CNT, graphene, C₃N₄), MoS₂, MXene, and their composites. For each material, various approaches were studied to improve the H₂ sensing performances, such as morphology and structure design, catalyst addition, as well as establishment of heterojunctions. Obviously, Pd and Pd-based materials has been still the current state-of-the-art room temperature resistive H₂ sensing material. Although the LOD (ppm level) and response/recovery behavior (tens of seconds) of room temperature H₂ sensors has been developed, further exploration for the H₂ sensor experimentally applied for early safety warning of LIBs is required.

Thin H₂ sensor integrated inside of LIB cell could be much more efficient and timelier for the early safety warning of LIBs, which has not been reported yet. Great selectivity to H₂ is required for the H₂ sensor integrated in LIB cell, especially inhibiting the inter-cross sensitivity to electrolyte vapors. Sensor device covered by semipermeable membrane is promising and necessary for the internal gas sensor in LIB to prevent the device from electrolyte vapors and other gases released from the chemical reactions inside the battery while also improving long-term stability.

Research concerning resistive H₂ sensors integrated inside of LIB cells for early safety warning is still in an early stage. Notably, both the physical stability and the sensing stability of the internal gas sensor in a few years are crucial.

We overviewed the recently interesting advances of resistive H₂ sensors possibly used inside of LIB. Although there remain huge challenges and unexplored topics, we strongly believe that the resistive H₂ sensors will further grow in the future and unlock novel applications in LIBs. We hope that this review inspired the applications of H₂ sensors in early safety warning of LIBs.