With the surge in the use of fossil fuels after the industrial revolution, the concentration of carbon dioxide in the atmosphere has increased from approximately 0.028 % before the industrial revolution to approximately 0.04 % [1], and has continued to increase by 20 % every five years. Various efforts have been devoted to reduce carbon emissions worldwide to mitigate their numerous detrimental side effects, such as abnormal climate and greenhouse effect. Historically, internal combustion engine vehicles have greatly contributed to improving transportation efficiency. However, these vehicles account for approximately 9 % of total carbon emissions and are among the major causes of negative environmental effects; therefore, an alternative strategy for transportation is urgently needed. Electric vehicle technology is rapidly advancing, and its implementation has expanded from conventional cars to large vehicles, such as buses and trucks, with the development of electric energy storage devices. However, large vehicles used for transportation with high daily mileage and heavy cargo weight exhibit a decreased energy efficiency because multiple batteries are installed; therefore, electric vehicle technology is not suitable for such systems.

Hydrogen has attracted considerable attention as a clean renewable energy source because it is widely available and can be used for eco-friendly energy generation. Hydrogen energy does not emit greenhouse gases such as carbon dioxide, which are generated during the combustion of hydrocarbons that are currently used as an energy source. Moreover, because hydrogen has a high energy density of 120– 142 kJ/g (3–4-fold higher than that of conventional fossil fuels) and can be stored and transported for long periods in large amounts, it is considered as a reasonable energy source for large mobilities, such as aviation systems, ships, and trains [2]. The European Union has initiated mass production of hydrogen fuel cells through proactive strategies to secure hydrogen energy technology. In Germany and France, hydrogen trains began operating in 2018 and 2022, respectively. In the United States, local governments and private companies are expanding their hydrogen infrastructure by using hydrogen-powered trucks and forklifts. In China, since 2019, the government has established and supported strategies for using electric vehicles as passenger cars and hydrogen vehicles as commercial vehicles, which is currently distributing 70 % of buses and 30 % of trucks that run on hydrogen [3,4].

For more than 100 years, hydrogen has been used in ammonia manufacturing, the petroleum industry, and glass manufacturing; therefore, safety guidelines related to hydrogen use are wellestablished at the industrial level. Nevertheless, there is a potential risk of explosions because hydrogen has a very low ignition energy (0.02 mJ), wide explosive range (4–75 vol%), high diffusion rate, and good penetrability because of its small molecular size [5]. The hydrogen storage tank explosion in Gangwon-do, South Korea, and hydrogen charging station explosion in Oslo, Norway in 2019 indicates the presence of a loophole in existing hydrogen safety guidelines. Therefore, as hydrogen energy-related infrastructure is gradually being constructed in downtown areas to achieve a hydrogen economy, it is necessary to establish new safety management systems suitable for the new infrastructure to ensure continuous development and eliminate safety concerns. Particularly, it is crucial to detect hydrogen leaks during the process of extracting, transporting, and utilizing hydrogen gas within a rapidly growing infrastructure. Thus, unlike the static hydrogen safety infrastructure applied in existing industrial plants, a hydrogen sensor with high target-gas selectivity, light weight, and a short response time is needed to ensure a dynamic, safe system for hydrogen-mobilitybased infrastructure applications.

Among the various hydrogen sensors currently under development, this review article describes the operating principles and current development status of a sensor suitable for hydrogen mobility with high selectivity for hydrogen and miniaturization.

2.1. Catalytic combustion hydrogen sensors

Catalytic combustion hydrogen sensors comprise sensing elements and catalytic metals such as Pd, Pt, and Ru. Hydrogen is spontaneously oxidized at a temperature above the ignition point (585 °C) when the environment does not contain a catalyst or ignition source [6]; however, the ignition point decreases to 300–500 °C in the presence of a catalytic metal such as Pt [1]. When the temperature of the sensing element increases during an exothermic reaction between hydrogen and oxygen on the surface of the catalytic metal, the resistance value of the sensing element changes, and the hydrogen concentration is measured in terms of the change in the resistance value. As another mechanism, the Seebeck effect is utilized to detect the hydrogen concentration by measuring the electromotive force generated upon a temperature change during hydrogen combustion. The relationship between the temperature change and electromotive force as well as the chemical equations for the hydrogen reaction over a Pt catalyst are as follows:

where U is the electromotive force, α is the Seebeck coefficient, and ΔT is the temperature change. Catalytic combustion hydrogen sensors were initially structured as the bead type, but were recently changed to a planar structure with the development of microelectromechanical systems (MEMS) technology. Catalytic combustion hydrogen sensors have limited applicability in portable devices because of their high operating temperatures and high power consumption. Therefore, numerous studies have been conducted to develop hydrogen sensors that can operate in room temperature/low-power environments [7–11]. Henriquez et al. [12] developed a microrod-like Pt-nanostructurebased resistive temperature detector on suspended SiO₂ prepared by MEMS process, as shown in Figs. 1(a)–(c). This sensor reduces the power consumption to 4 mW and operating temperature to 72.3 °C. Additionally, various studies have been conducted to improve sensitivity of the sensors by increasing the surface area of the catalysts. These studies used nanoparticles (NPs) of noble metals, such as Pd, Pt, and Au, because of their large surface area and high catalytic activity [13–16]. Previously, materials such as graphene, MnO₂, SnO₂ were used to support NPs; however, these materials have the disadvantage of high power consumption based on their high heat capacities [9,17–19]. To resolve this problem, studies are underway to combine functional ligands between NPs for stabilization [20]. When the catalyst NPs were stabilized by the ligand positioned between the NPs, the sensitivity to hydrogen increased because of the increased catalyst surface area. By applying this method, Pranti et al. [21] developed high-selectivity hydrogen sensor using Pt NPs linked by a 4,4′′- diamino-p-terphenyl ligand [Figs. 1(d)–(f)].

Figure 1. (a) Structural diagram of catalyst combustion hydrogen sensor containing low power microheater. (b) Microscopic image of catalyst combustion hydrogen sensor and Pt nanostructure. (c) Response of hydrogen sensor depending on hydrogen concentration. Adapted with permission from [12], Copyright 2021, American Chemical Society. (d) Structure of nanoparticle combined with functional ligand (DATER). (e) Change in sensor output value according to ligand-stabilized catalyst. (f) Response of hydrogen sensor to toxic gases at an identical concentration (H₂, C₂H₆, and CH₄). Adapted with permission from [21], Copyright 2019, IEEE.

2.2. Transistor hydrogen sensors

Field-effect transistor (FET) hydrogen sensors have been widely investigated as portable sensors because of their excellent reliability, reproducibility, small size, and low power consumption [22–24]. These sensors operate based on the principle that when hydrogen molecules are adsorbed onto the surface of a catalytic metal, they dissociate into two hydrogen atoms. The dissociated hydrogen atoms diffuse into the catalytic metal/insulator interface and are polarized to affect the charge-carrier concentration [Fig. 2(a)], thereby controlling the conductivity of the active layer [25]. The catalytic reaction of hydrogen requires a high active energy; thus, general hydrogen gas sensors are operated at elevated temperatures (≥300 °C) for advanced monitoring. Therefore, time for stabilization and additional heating power are required, which incur additional power consumption and are considered as weaknesses of FET-type hydrogen sensors for portable monitoring systems. Therefore, FET sensors that are operable at room temperature are actively being examined. Recently, Choi et al. [26] induced a catalytic reaction using ultraviolet light rather than using thermal energy. By combining the ZnO-NP/Pd catalyst layer with an AlGaN/GaN-heterostructure-based FET hydrogen sensor [Fig. 2(b)], the device could detect hydrogen gas at room temperature. Moreover, the dual-gate structure of the FET, which can stably monitor the target hydrogen gas over a wide temperature range, was studied. As shown in Figs. 2(c) and 2(d), the reference gate (e.g., Ti) does not react with hydrogen, which is in contrast to the sensing gate (e.g., Pt) [27–29]. Therefore, the operation was stable because although the two different electrode materials had different chemical reactivities with hydrogen, they showed similar thermal sensitivity behaviors in the operating range (20–80 °C) [27]. Furthermore, Sharma et al. developed a hydrogen sensor operable even at a high temperature of ~250 °C by using graphene-Pd-Ag and Pt as the sensing and reference gates, respectively [Figs. 2(e) and 2(f)] [30]. In addition to developing a sensing material or device structure, the detection sensitivity of hydrogen sensors can be improved by using novel electrical analysis methods [31–33]. Shin et al. [34] reported an improvement in the low-frequency noise and signal-to-noise ratio for FET hydrogen sensors by using body-tosource junction bias control.

Figure 2. (a) Operating principle of transistor hydrogen sensor [1], (b) structure of ultraviolet-assisted AlGaN/GaN field-effect transistor (FET) hydrogen sensor [26], (c) difference in output voltage depending on the temperature of dual-gate FET hydrogen sensor with Ti reference gate and Pt sensing gate, (d) difference in the output voltage of two gates depending on the hydrogen concentration. Adapted with permission from [27], Copyright 2009, IEEE. (e) Structure of FET hydrogen sensor with dual gate electrodes and (f) preparation method. Adapted with permission from [30], Copyright 2018, Springer Nature.

2.3. Schottky diode hydrogen sensors

The basic principle of Schottky diode hydrogen sensors is similar to that of FET hydrogen sensors. When hydrogen molecules react with catalytic metals such as Pt or Pd and decompose into two hydrogen atoms, the hydrogen atoms diffuse between the catalytic metals and are adsorbed in the form of a dipole layer at the interface of the metal semiconductor or metal oxide. The dipole layer forms an electric field near the semiconductor surface in the direction opposite to the built-in field, thereby lowering the Schottky barrier and increasing the current flow at the same applied voltage [35]. Si is mainly used as a semiconductor material in Schottky diodes, although various compound semiconductor materials have also been used, including SiC, GaAs, AlGaAs, InP, GaN, and AlGaN. Among these materials, Schottky diode hydrogen sensors based on SiC and GaN/AlGaN have been extensively investigated. Si has previously been used as a semiconductor material but has a narrow bandgap of 1.1 eV, making it difficult to increase the operating temperature of devices with an Si substrate to ≥250 °C. Furthermore, Si sensors containing a p-n junction cannot be used at ≥150 °C because junction leakage and metallization degradation readily occur at high temperatures. However, SiC has been widely studied in devices operating at high temperatures because it has a wide bandgap of 3.3 eV, which is three-fold wider than that of Si, higher thermal conductivity (4.9 W/cm·K), and higher breakdown voltage, as well as contains fewer crystal dislocation defects than other semiconductor materials with a wide bandgap [36]. Qi et al. [37] evaluated a SiC Schottky diode hydrogen sensor with a trench-insulator structure; a Pd/Ta₂O₅/SiC Schottky diode hydrogen sensor operating at 300 °C with a SiC substrate was also reported [shown in Figs. 3(a)–(c)] [38]. Chen et al. [39] developed a Schottky diode hydrogen sensor based on a Pt thin film/NP hybrid structure and GaN/AlGaN. This sensor had short reaction/recovery times of 18/12 s and a high sensing response value of 2.35 × 10⁷ when it was exposed to 1% hydrogen at 300 K, as shown in Figs. 3(d) and 3(e).

Figure 3. (a) Schematic, (b) reaction of Pd/Ta2O5/SiC hydrogen sensor with 120–5,000 ppm hydrogen at 300 °C, (c) preparation of the Pd/Ta₂O₅/SiC hydrogen sensor. Adapted with permission from [38], Copyrights 2021, Sensors. (d) Schematic and (e) reaction of AlGaN/GaN Schottky diode hydrogen sensor with 1,000 ppm and 1 % hydrogen. Adapted with permission from [39], Copyright 2021, Elsevier.

2.4. Optical hydrogen sensors

The optical properties of metals such as Pd, Y, and Mg are altered upon phase transition to metal hydrides under a hydrogen atmosphere [40]. Hydrogen gas can be detected based on their optical property changes such as by measuring the transmitted/reflected light or difference in the wavelength of the surface plasmon absorption band of the metal. Optical hydrogen sensors can be operated remotely without electrodes; therefore, they have simpler structures than those of other sensors. In addition, they can be easily integrated into other devices because they are not subject to electromagnetic interference. Pd has been widely investigated as a material for optical hydrogen sensors because it can dissociate hydrogen at room temperature and has a wide sensing range of 10¹–10⁴ Pa [41]. However, disadvantages of Pd include that the reaction is time-consuming and is nonlinear because of the hysteresis phenomenon during the phase change from α-PdH_x to β−PdH_x. These phase changes may be suppressed by alloying Pd with other metals, such as Hf [41], Au [42,43], Ni [44], and Cu [45]. Nugroho et al. [43] deposited a Pd₇₀Au₃₀ alloy on a glass substrate and coated the top with Polytetrafluoroethylene/Poly (methyl methacrylate), as shown in Fig. 4(a), to develop a plasmonic hydrogen sensor with high hydrogen selectivity and a reaction time of <1 s [Fig. 4(b)]. In addition, Bannenberg et al. [46] reported a Ta_1−γPd_γ thin film prepared by alloying Ta and Pd, which did not undergo hysteresis at either room temperature or high temperatures while reacting in the range of 10⁷ Pa.

Figure 4. (a) Structure of PdAu @polytetrafluoroethylene (PTFE)@poly(methyl methacrylate) (PMMA) hydrogen sensor and (b) comparison of reaction rates between PdAu @PTFE@PMMA hydrogen sensor and PdAu@PTFE hydrogen sensor immediately after preparation and after 4 months. Adapted with permission from [43], Copyright 2019, Springer Nature. (c) Structure of optical hydrogen sensor using isoreticular metal organic frameworks-20 (IRMOF-20), (d) response/reaction graph at 8 % hydrogen concentration, and (e) changes in location and strength of surface plasmon absorbance peak at different hydrogen concentrations. Adapted with permission from [49], Copyright 2019, American Chemical Society.

From another perspective, conventional optical fibers have a wide detection range; however, their sensitivity is low in most cases. Plasmon-active fibers have been studied as alternatives for improving sensitivity of hydrogen sensors [47,48]. Additionally, alternative materials to the costly noble metals used in optical hydrogen sensors have been explored. Transition metal oxides such as TiO₂, NiO_x, and WO₃ have recently been used. Furthermore, hydrogen sensors using metal-organic frameworks, which are widely used as hydrogen storage materials, have been developed. Miliutina et al. [49] developed an optical hydrogen sensor using a plasmon-based metal-organic framework structure. The gold-coated metal-organic framework structure on the optical fiber dramatically altered the plasmon band wavelength position and increased the intensity; this sensor showed excellent selectivity and a high response rate [Figs. 4(c)–(e)].

2.5. Chemiresistive hydrogen sensors

Chemiresistive hydrogen sensors measure hydrogen concentrations based on the principle that the resistance or conductivity changes following surface reactions or the adsorption of gas molecules onto the detection material. These sensors are suitable as portable sensors because of their high sensing performance, low manufacturing cost, and ease of miniaturization [50–53]. Thus, chemiresistive hydrogen sensors have been rapidly developed in recent years. Pd-based resistance sensors have a simple structure, high selectivity, and operability at room temperature. The most widely examined chemiresistive sensors are those based on semiconducting metal oxides; however, these sensors cannot be applied in various fields because of their high operating temperatures (200–400 °C), high power consumption, and rigidity. Thus, low-power, flexible hydrogen sensors were recently developed using graphene [54], carbon nanotubes [55], and organic/ inorganic 2D hybrid materials [56,57]. Among these sensors, carbon crystal structures can detect hydrogen quickly based on their high hydrogen sensitivity but readily react with other gases because of their large surface areas [58]. To overcome this limitation, studies have been conducted to evaluate the effects of doping noble metals such as Pd [54,55]. Flexible hydrogen sensors were recently manufactured using carbon nanofibers doped with Ni-Pt through electrospinning and chemical reduction [shown in Figs. 5(a) and 5(b)] [58]. Furthermore, two-dimensional transition metal dichalcogenide materials such as MoS₂ have a large surface area [59], high hydrogen affinity [60], and tunable electrical/chemical properties [61,62], and have thus attracted attention as novel hydrogen sensor materials.

Figure 5. (a) Flexible hydrogen sensor using carbon nanofiber (CNF)@Ni-Pt, (b) reactions in 1 % hydrogen atmosphere in flat and bend states. Adapted with permission from [58], Copyright 2021, American Chemical Society. (c) Structure of hollow structure of Pt-decorated MoS₂ (h-MoS₂/Pt) hydrogen sensor, and (d) response of hydrogen sensor at 0.4–0.8 % hydrogen concentrations. Adapted with permission from [63], Copyright 2021, American Chemical Society.

By exploiting these characteristics, researchers developed a Pt NPdecorated ultrathin MoS₂ hollow sphere hydrogen sensor using the spillover effect of Pt nanoparticles, as shown in Fig. 5(c). The sensor shows response times of 8.1 and 2.7 s for 1 and 4 % hydrogen, respectively, at room temperature [Fig. 5(d)] [63].

Transportation using hydrogen energy is considered to be sustainable because of its eco-friendliness and high energy efficiency. International think tanks, such as Deloitte, have proposed that the main application targets of hydrogen mobility should be commercial vehicles, aviation systems, ships, and micro-mobility in addition to hydrogenelectric vehicles. Therefore, the global market of this field is expected to grow at a rate of 29.7 % per year. However, technologies that maintain the safety of the new hydrogen infrastructure must be supported to enable continuous growth of the hydrogen economy. As described in this review, although hydrogen sensors with various underlying mechanisms have been developed (Table I), it remains challenging to selectively detect hydrogen alone in an actual environment wherein multiple gases are mixed. Therefore, to develop hydrogen sensors that can quickly and reliably detect hydrogen in various environments, different types of hydrogen sensors should be combined and integrated. In addition, composite materials must be developed to reduce the content of the costly noble metals currently used in hydrogen sensors, such as Pd and Pt. Furthermore, modularization technology based on software is required to improve the detection accuracy by correcting the errors and noise of hydrogen sensors.

Table 1 . Summary of sensing performances of hydrogen sensors by operation type.

Type of Hydrogen Sensor	Response/Recovery time (Concentration or Pressure of H2)	Operation Temperature	Sensitivity @ [H2]	Sensing Materials
Catalytic Combustion [12]	11 s/10 sa (1%)	72.3 °C	0.46%c @ 1%	Pt
Transistor [26, 30]	8 s/11 sa (4%)	RT	25%c @ 4%	AlGaN/GaN with ZnO NPs/Pd film
	68 s/57 sa~16 s/14 sa (1000ppm)	RT ~ 245 °C	8 μAd @ 2000 ppm	Graphene decorated Pd-Ag alloy NPs
Schottky Diode [38, 39]	7.1 s/18 sa (120 ~ 5000ppm)	300 °C	1000e @ 5000 ppm	Pd/Ta2O5/SiC
	18 s/12 sa (1%)	RT	2.35×107 f @ 1%	AlGaN/GaN with Pd NPs/Pt film
Optical [43, 47]	10 s/5 sb (8%)	RT	40 nmg @ 8%	IRMOF-20
	1 s/5 sa (1mbar)	RT	n.r. @ 1mbar (4 nmg @ 4%)	Pd70Au30@PTFE/PMMA
Chemiresistive [63, 64]	8.1 s/16 sa (1%)	RT	8.7%c @ 1%	Pt decorated hollow- MoS2
	1.81 min/5.52 mina (2%)	RT	66.67%c @ 2%	PMMA/Pd NP/ Single Layer Graphene

^aResponse/recovery times are defined as the time required to reach 90 and 10 % of saturation resistance or current, respectively.

^bResponse/recovery times are defined as the time required to reach 95 and 5 % saturation resistance or current, respectively.

^cSensitivity is defined as (R_H2 - R_air)/R_air× 100 (%) or (I_H2 - I_air)/I_air × 100 (%).

^dSensitivity is defined as variation in the drain current.

^eValue indicates that sensitivity is defined as R_Ar/R_H2.

^fValue indicates that sensitivity is defined as (I_H2 - I_air)/I_air.

^gValue indicates that sensitivity is defined as Δλ_peak.

^hn.r indicates not reported, and RT is room temperature.

Abbreviations: IRMOF-20, isoreticular metal organic frameworks-20; NP, nanoparticle; PTFE, polytetrafluoroethylene; PMMA, poly (methyl methacrylate).

This work was financially supported by a National Research Foundation of Korea (NRF) grant funded by the Korean government (Grant no. 2022R1F1A1072339) and a Pusan National University Research Grant, 2021.