A robust organic hydrogen sensor for distributed monitoring applications
Main
The unique properties of diatomic hydrogen (H2), such as its rapid diffusion rate and high energy density, make it suitable for various applications in different industries, including energy, transportation, petroleum refining, defence, space, food and pharmaceuticals1,2,3,4. However, its combustible nature, low flammability point (approximately 4 vol% H2 in the air) and the inability of people to detect it using their senses create safety concerns for hydrogen production, storage and usage5. There is, thus, a growing demand for sensitive and selective hydrogen sensors that can provide an early warning of potential leaks. To meet the large volume demands of the emerging hydrogen economies, such sensors must be cost-effective to manufacture, consume little power, operate at ambient conditions and have a minimal environmental impact. Among the different types of hydrogen sensors available, which include resistive, capacitive, optical, acoustic and magnetic systems6,7,8,9,10,11,12,13,14, resistive-based technologies are arguably the most promising because they are simpler and more scalable to manufacture and can be integrated with incumbent electronics15,16,17,18,19,20.
Various materials have been explored as the active element of such resistive hydrogen sensors to address the trade-offs between sensor performance (including sensitivity, rise time and fall time), shape, size, circuit complexity, power consumption and cost. These materials include noble metals such as palladium (Pd)21 and oxide materials such as zinc oxide (ZnO), molybdenum trioxide (MoO3), titanium dioxide (TiO2), niobium pentoxide (Nb2O5), tin(IV) oxide (SnO2), tungsten oxide (W2O3) and indium(III) oxide (In2O3)6,22,23,24,25,26,27. However, the typically low responsivity, high-temperature operation (requiring a power-hungry heating element) and long response time have limited their practical applications.
Various metal composites28,29,30,31,32,33 and organic materials, such as polypyrrole nanofibres and polyaniline34,35,36, have also been investigated for H2 detection. Polypyrrole nanofibre and polyaniline-based hydrogen sensors can exhibit a decrease in resistance with increasing hydrogen concentration due to the interaction with charged amine nitrogen sites on the polymers36. Pentacene-based organic field-effect transistors have also been employed to detect hydrogen by a different mechanism in which the contact resistance of the device at room temperature increases with the expansion of the palladium electrodes caused by the presence of hydrogen37,38. However, the suboptimal performance of these organic sensors in terms of sensing parameters, including low responsivity (≤0.9) and slow response time (up to 6 s), has impeded their deployment in practical applications.
In this article, we describe a hydrogen sensor that uses an organic semiconductor as the active layer. Our approach is based on the p-doping of organic semiconductors by oxygen species39,40,41. The current changes as a result of the interaction between hydrogen and the physisorbed oxygen species. The sensor offers high responsivity, fast response, low power consumption, low limit of detection (LoD) and good stability in harsh testing conditions. We also show that it can outperform a state-of-the-art commercial device in different practical sensing demonstrations.
Sensor fabrication and performance
Our sensor has interdigitated platinum electrodes patterned lithographically (Fig. 1a, right and Extended Data Fig. 1) and a thin layer of the organic polymeric semiconductor poly[3,6-bis(5-thiophen-2-yl)-2,5-bis(2-octyldodecyl)pyrrolo[3,4-c]pyrrole-1,4(2H,5H)-dione-2,2′-diyl-alt-thieno[3,2-b]thiophen-2,5-diyl] (DPP-DTT) deposited on top by spin-coating or printing (Fig. 1a, left). The polymer forms a conformal and smooth layer with a root mean square surface roughness of ~0.6 nm (Extended Data Fig. 1). Before the polymer layer was deposited by photoelectron spectroscopy, the work function of the platinum electrode was measured to be 5.3 eV (Extended Data Fig. 1), which matches well with the highest occupied molecular orbital energy level of DPP-DTT of −5.2 eV, making it a good hole-injecting contact. Completed devices were stored in dry air inside a dry box for 10 h before testing. The device was then transferred into a test chamber and exposed to a stream of synthetic air mixed with varying concentrations of H2 following the experimental procedures described in Methods (Extended Data Fig. 2).

a, Schematic representation of the DPP-DTT-based H2-sensing device composed (bottom to top) of glass, Pt and DPP-DTT. Also shown is the structure of DPP-DTT and a schematic representation of the energy levels of Pt and DPP-DTT with respect to the vacuum level. b, Current (I) versus voltage (V) curves of the sensing device were measured at various H2 concentrations. c, H2-concentration-dependent device current measured at a constant voltage of 1.0 V. The error bars indicate uncertainties in the device current mainly stemming from uncertainties in the thickness of the DPP-DTT thin films and represent the standard deviation from four individual devices. d, Change in responsivity (ΔR/R0) for various H2 concentrations and for different biasing voltages applied to the device. e, Measurement of the fast H2-dependent switching response for an applied bias of 0.5 V in ambient conditions (20 °C and 60% RH). Synthetic air was used as the carrier gas in the sensing measurements, as outlined in Methods. HOMO, highest occupied molecular orbital; LUMO, lowest unoccupied molecular orbital. It and I0 represent the measured device current over time and at t = 0 s, respectively.
Source data
Figure 1b shows the current versus voltage characteristics of the optimized device (Supplementary Fig. 1) recorded before and after exposure to hydrogen for a few minutes at room temperature. Under 100% synthetic air (0 ppm H2), a linear current versus voltage (I–V) characteristic was recorded, indicating the presence of Ohmic conduction under equilibrium conditions. Upon inserting various amounts of H2 from 100 to 1,000 ppm, the sensor current decreased logarithmically by 10,000-fold from microampere to sub-nanoampere levels (Fig. 1c) at room temperature and under 1.0 V bias, demonstrating its high sensitivity to H2. Such a drastic current decrease can be attributed only to carrier depletion induced upon exposure to H2. The current reduction in the low-voltage regime (0–0.1 V) was followed by a second conduction regime (>0.1 V), manifested as a sublinear I–V dependence. Qualitatively similar I–V characteristics have previously been attributed to Schottky barriers at both contacts in two-terminal semiconductor diodes42. We hypothesize that ambient oxygen p-dopes the DPP-DTT and shifts the Fermi energy so that the Pt/DPP-DTT becomes more Ohmic while the device conductivity increases (Extended Data Fig. 3). Exposure to H2 gas de-dopes the organic semiconductor and reduces its work function, giving rise to a higher Schottky barrier at each Pt–organic semiconductor interface to form a back-to-back diode equivalent circuit. Modulating the Schottky barriers by p-doping and subsequent de-doping leads to the strong current modulation observed in the I–V characteristics (Fig. 1b and Extended Data Fig. 4a) due to its exponential dependence on the barrier height.
The H2 concentration-dependent responsivity, defined as ΔR/R0, where R is the electrical resistance at the measured H2 concentration, and R0 is the resistance before H2 exposure, was calculated from the I–V characteristics for an applied bias between 0.1 and 1.0 V (Fig. 1d, Extended Data Fig. 4b and Supplementary Fig. 2). We found that the responsivity at higher hydrogen concentrations (1,000 ppm) saturated at around 1.0 V. Accordingly, 1.0 V was chosen for further temporal hydrogen response measurements at higher hydrogen concentrations. A bias-dependent ΔR/R0 of more than six orders of magnitude was observed upon increasing the H2 concentration from 100 to 1,000 ppm. The signal evolution at constant H2 concentrations recorded at 0.5 V, shown in Extended Data Fig. 4c, confirms the robust sensing behaviour of the device. The extracted LoD of the DPP-DTT sensor was approximately 192 ppb (Extended Data Fig. 4b).
The dynamic response of the sensor was studied next by subjecting the DPP-DTT device to pulses of synthetic air containing different H2 concentrations (Extended Data Fig. 4d) inside the gas-sensing chamber, which was maintained at 20 °C with 15% relative humidity (RH). Exposure of the sensor to 1% H2 (10,000 ppm) yielded a notably fast signal response with a rise time of ~0.84 s (Fig. 1e). The rapid response to hydrogen was also demonstrated by leaking hydrogen gas (99.99%) close to the organic sensor (Supplementary Video 1). The longer-term stability of the sensing process was confirmed by recording the response of the sensor over 1,000 sensing cycles at 0.1% H2 (1,000 ppm) (Extended Data Fig. 4e). A statistical analysis of the rise and fall times yields 0.83 (±0.08) s and 6.52 (±0.21) s, respectively (Extended Data Fig. 4f). These results demonstrate the excellent sensing capabilities of the organic device, which surpass most hydrogen sensor performance metrics as defined by the US Department of Energy and those published in the literature (Supplementary Table 1)43,44,45.
Mechanism of hydrogen detection
The proposed macroscopic and atomic-level mechanism for hydrogen sensing in our device is presented in Fig. 2a (left and right, respectively). In step (i), the organic sensor is introduced into an oxygen-containing environment (for example, ambient air) and is p-doped by oxygen molecules. This leads to an increase in the device current, and the biasing leads to the catalytic formation of oxygen adatoms on the surface of the Pt electrodes. In step (ii), hydrogen is introduced and immediately permeates through the organic semiconductor. In step (iii), the hydrogen molecules are split at the electrodes into adatoms and subsequently react with oxygen adatoms to form water. This reduces the oxygen-induced doping and is accompanied by a fast reduction in the device current, so that hydrogen is detected. The water molecules formed leave the device through evaporation, and the current remains low as long as hydrogen is present. In step (iv), the device remains exposed to ambient air although hydrogen exposure has ceased. Oxygen again p-dopes the organic semiconductor, leading to a rapid increase in the current (signal recovery) and the cessation of hydrogen sensing.

a, Schematic representation of the H2-sensing device and the hydrogen-sensing mechanism through oxygen doping and de-doping processes. b, DPP-DTT polymer chain with a physisorbed O2 molecule. c, Electronic DOS for the pristine DPP-DTT polymer (grey-shaded region) and the O2-physisorbed configuration (red line). The energy of the highest occupied state was set to zero. The O2-induced peaks above zero (above the valence band maximum) confirm that physisorbed O2 molecule dopants can act as acceptors for the DPP-DTT polymer.
Source data
To verify the proposed sensing mechanism, we undertook a series of density functional theory (DFT) calculations (Extended Data Fig. 5). The calculations targeted several processes and corresponding molecular structures involving oxygen- and hydrogen-related species in the bulk of the organic semiconductors of interest, as well as at the surface of the Pt electrode. We first examined the stability of oxygen molecules next to polymer chains of DPP-DTT and two other organic semiconductors, namely poly(indacenodithiophene-co-benzothiadiazole) (C16-IDTBT) and poly(3-hexyl)thiophene (P3HT) (Extended Data Fig. 5a,b). In all cases, we found that it is energetically favourable for an oxygen molecule to dissociate and form sulfonyl groups on the polymers (other O-chemisorbed structures are also stable, but their energies are higher than that of the sulfonyl configuration). However, as shown in a previous DFT study41, the activation energy for the process of a physisorbed O2 molecule next to a P3HT polymer chain reacting together to create a sulfonyl group on the thiophene is very high (approximately 2.0 eV). Hence, this process is largely suppressed in our device, and any oxygen molecules that have diffused into the bulk of P3HT (or DPP-DTT or C16-IDTBT, which have similar S-based heterocyclic groups) will probably remain as O2 in physisorbed configurations (Fig. 2b). Interestingly, density of states (DOS) calculations of the corresponding structures revealed that physisorbed O2 molecules create unfilled energy states inside the bandgaps and close to the respective valence bands of DPP-DTT, P3HT and C16-IDTBT. Figure 2c shows the DOS results for DPP-DTT. The results for P3HT and C16-IDTBT are presented in Extended Data Fig. 5a,b. The existence of these acceptor states is consistent with the effect of p-doping of oxygen on the polymers of interest (as well as other typical organic semiconductors)46,47.
Having identified the physisorbed oxygen molecules as the source of p-doping, we next aimed to understand the processes that enable the de-doping of the organic semiconductors in the presence of H2. The first mechanism considered was that hydrogen molecules react directly with physisorbed O2 molecules. We probed such a scenario with DFT-nudged elastic band (DFT-NEB) calculations and found that the transformation of physisorbed H2 and O2 into an H2O molecule and an S-O group on a P3HT polymer chain is an exothermic process with a prohibitively high energetic barrier of 2.50 eV. Another possibility was that hydrogen molecules reacted with O species chemisorbed on the polymer. However, as we mentioned earlier, the formation of such species is also suppressed due to their very high activation energies.
As the barriers to reactions involving O2 in these organic polymers are high, an alternative scenario focuses on the role of the Pt electrodes. It is known that Pt is efficient in dissociating hydrogen molecules and creating atomic H adatoms on its surface48. Accordingly, we performed a series of DFT-NEB calculations to investigate reactions of such H adatoms with O species on a Pt surface. The corresponding results are presented in Extended Data Fig. 5c–f. From these results, we deduce that, first, oxygen molecules are chemisorbed on the Pt surface, and then, in a multistep process, H adatoms react with the O adatoms to create H2O molecules. The latter have relatively small binding energies to the surface (0.2 eV), so some of these molecules will desorb. This process of eliminating p-type dopant oxygen species from the organic semiconductor in the presence of hydrogen leads to a reduction in current, as observed in our devices. To verify the role of the Pt electrode, a DPP-DTT device was prepared with gold (Au) electrodes. The experimental results (Supplementary Fig. 3) indicate that the DPP-DTT device did not show any significant response in the presence of hydrogen with Au electrodes, suggesting that the catalytic effects of Pt are crucial to the working mechanism of the device.
Finally, to exclude the possibility of electrochemical effects, we used a mixed ionic–electronic conductor, poly(3,4-ethylenedioxythiophene) polystyrene sulfonate, which exhibits high ionic conductivity like that of organic electrolytic materials, as the sensing layer (Supplementary Fig. 4). The high ionic conductivity overwhelmed the effect of oxygen p-doping, so that the device did not respond to hydrogen, which supports our conclusion regarding the resistive nature of our organic sensors.
Stress testing the organic hydrogen sensor
To investigate whether the p-doping and de-doping processes by O2 and H2, respectively, could be observed in other organic semiconductors, we studied several other hole-transporting organic semiconductors: C16-IDTBT, P3HT, 6,13-bis(triisopropylsilylethynyl)pentacene (TIPS-pentacene), poly[bis(4-phenyl)(2,4,6-trimethylphenyl)amine] (PTAA) and one electron-transporting material, namely 2,2′-((2Z,2’Z)-((12,13-bis(2-ethylhexyl)-3,9-diundecyl-12,13-dihydro-[1,2,5]thiadiazolo[3,4-e]thieno-[2″,3″:4′,5′]thieno[2′,3′:4,5]pyrrolo[3,2-g]thieno-[2′,3′:4,5]thieno[3,2-b]indole-2,10-diyl)bis(methanylylidene))-bis(5,6-difluoro-3-oxo-2,3-dihydro-1H-indene-2,1-diylidene))dimalononitrile (Y6). The detailed analysis of the response to hydrogen exposure for all sensors developed is presented in Fig. 3a and Extended Data Fig. 6a–d. The DPP-DTT sensor exhibited the highest responsivity of over three orders of magnitude, while the P3HT device showed the lowest. The Y6-based sensor had a low current due to the poor electron injection caused by the large mismatch of the Pt work function with the lowest unoccupied molecular orbital and highest occupied molecular orbital levels of Y6. The sensor showed no evidence of oxygen-induced p-doping and no response to H2 (Extended Data Fig. 6e). These findings also provide further support to the proposed sensing mechanism that relies on O2-induced p-doping and the subsequent de-doping upon exposure to hydrogen. Due to its superior performance among the range of active-layer materials tested, the DPP-DTT sensor was chosen for further analysis.

a, H2 concentration-dependent device responsivity of the organic materials DPP-DTT, C16-IDTBT, P3HT, TIPS-pentacene and PTAA. The error bars define the standard deviation of the mean responsivity by averaging over three different devices for each material. b, Temperature-dependent responsivity of the DPP-DTT H2 sensor. The error bars indicate the standard deviation in the responsivity from the mean from four different sets of data taken from a single device. c, Real-time current as a function of H2 concentration from 0 to 1,000 ppm at 120 °C for an applied bias of 1.0 V. d, Change in the responsivity of the DPP-DTT sensor in the presence of individual gases measured at a constant bias of 0.5 V and a concentration of 1% (10,000 ppm). e, Hydrogen concentration-dependent current of the DPP-DTT device, measured under various humidity conditions (15%, 50% and 80% RH). The error bars denote the standard deviation in the device current from the mean from four different sets of data taken from a single device. f, Real-time current response of the DPP-DTT device measured at various applied bias conditions (0.2 to 1.0 V in 0.2 V increments), as a function of switching the H2 concentration from 0 to 1,000 ppm at an 80% RH. Except for d, synthetic air was the carrier gas in all experiments, as outlined in Methods. EtOH, ethanol; MeOH, methanol; Me2CO, acetone; Tol, toluene.
Source data
The performance of the DPP-DTT sensors was evaluated under different environmental conditions, such as elevated temperatures, different interfering volatile gases and varying RH (Fig. 3b–f). A small increase was observed in the responsivity up to 40 °C for H2 concentrations up to 800 ppm (Fig. 3b). However, the responsivity of the sensor decreased on increasing the temperature from 60 to 120 °C. We attribute this to the lower adsorption rate for oxygen49 and H2 at higher temperatures50. Figure 3c shows the real-time response of the sensor to H2 concentrations from 0 to 1,000 ppm at 120 °C, which demonstrates its robust sensing capabilities even under such harsh conditions (full sets of the sensing characteristics of the DPP-DTT sensor measured at various temperatures from 40 to 120 °C are shown in Extended Data Fig. 7a–e). We also studied the long-term thermal stability of the DPP-DTT devices (Supplementary Fig. 5) and found that their response to hydrogen, in terms of current change, remained largely independent of thermal stress. Thus, the organic sensor does not require extended stabilization periods before it can be used for quantitative H2 concentration measurements.
The selectivity of the DPP-DTT sensor to H2 was also examined in the presence of various volatile organics (interfering species), including ethanol, methanol, acetone and toluene, which exhibit flammability and reducing properties like those of hydrogen and are commonly found in household environments51 (Fig. 3d and Extended Data Fig. 8a). The change in absolute responsivity for H2 was approximately four orders of magnitude higher than that for the other compounds, even in a lower responsivity region (0.5 V), indicating the possibility of a further reduction in the power consumption of the organic hydrogen sensor. In further experiments, the DPP-DTT organic device did not exhibit any response to non-volatile reducing methane gas (0.1% in air) when measured at 20 °C and 15% RH (Supplementary Fig. 6). The current response of the DPP-DTT sensor was also examined in the presence of the oxidizing-gas nitrogen dioxide. The oxidation of the DPP-DTT layer produced holes that increased the device current, in contrast to the effect observed with H2 (Supplementary Fig. 7). However, the device did not return to its initial state due to persistent conductivity and exhibited a very slow recovery, probably due to the irreversible oxidation of DPP-DTT. Importantly, the sensor was able to detect the presence of H2 in a complex gas mixture containing 1.0% H2, 1.0% O2, 1.0% N2 and 1.0% CO2 in Ar (Extended Data Fig. 9a). Before H2 exposure, the sensor showed a slight increase in the current when exposed to a stream of Ar gas containing 10% CO2 and O2 (Extended Data Fig. 9b,c); however, these small changes occurred at different timescales and did not interfere with the ability of the sensor to detect the presence of H2 rapidly in real time (Extended Data Fig. 9d).
The organic sensor showed no response when balloons filled with Ar, N2 and dry air were popped under the same experimental conditions to simulate sudden changes in the stoichiometry of the ambient atmosphere (Extended Data Fig. 10d), highlighting the specificity of the sensor to H2. The performance of the hydrogen sensor was also examined under atmospheric (21%) and high (99.9%) oxygen conditions for comparison. The corresponding I–V characteristics were measured at hydrogen concentrations of 0 and 1,000 ppm (Supplementary Fig. 8). We found that under 99.9% O2 conditions, the device current increased for both 0 and 1,000 ppm hydrogen concentrations. However, the device current remained largely unchanged despite the fourfold increase in O2 concentration, indicating that the responsivity of the device in various O2 environments remained unaffected. The only condition where the sensor failed to recover was when the atmosphere lacked O2. Extended Data Fig. 9e,f shows that the sensor response to H2 when inert gases were used as the carrier (Ar or N2) was initially strong, but its signal failed to recover due to the absence of O2, which is responsible for the subsequent p-doping.
To determine the impact of humidity on the H2-sensing capabilities of the device, we conducted measurements across a range of RH from 15% to 80% for various H2 concentrations ranging from 0 to 1,000 ppm (Fig. 3e and Extended Data Fig. 8b–d). The hydronium ions in water doped the organic semiconductor, such that the device current increased with humidity, as reflected in the I–V characteristics. In the presence of hydrogen, the organic semiconductor was de-doped, causing a decrease in the device current. However, hydrogen adatoms also react with hydroxyl ions, so that the device current increased and the hydrogen response reduced as the humidity was increased. The device exhibited robust dynamic operation with more than two orders of magnitude of current change upon repeated exposure to 0 and 1,000 ppm of H2 at 80% RH for various bias conditions ranging from 0.2 to 1.0 V in 0.2 V increments (Fig. 3f). In long-term stability testing of the device under high humidity conditions (80% RH), the device was found to be stable (Supplementary Fig. 9). Finally, the long-term stability of the DPP-DTT sensors was studied by storing them in ambient air and humidity (20–24 °C and 50–60% RH) while sporadically testing their response to 1,000 ppm of H2 under controlled conditions of 20 °C and 15% RH at 1 V. These data are plotted in Extended Data Fig. 10e. The sensor remained functional for nearly 2 years (646 days), with its response plateauing after an initial ‘burn-in’ period of ~60 days.
Our experiments demonstrate that the organic hydrogen sensor meets most of the target specifications for hydrogen safety applications. These include signal responsivity (0.25 relative resistance at 1% H2), LoD (<0.1%), response speed (t90 < 1 s at 1% H2), recovery speed (t10 < 60 s) and cost (potentially well under US$40 per unit), as defined by the US Department of Energy and in the literature (Supplementary Table 1)43,44,45. t90 is defined as the time required for the device current or resistance to change from 10% to 90% in the presence of hydrogen, while t10 is the time taken for the device current or resistance to change from 90% to 10% in the absence of hydrogen. This led us to investigate the practical utility of this technology in realistic sensing applications, which we will discuss next.
Quantitative hydrogen sensing in practical scenarios
Before deploying the sensor in the quantitative detection of H2 in practical scenarios, the device was calibrated with hydrogen concentrations ranging from background levels (500 ppb) to 1,000 ppm in ambient conditions (~20 °C and ~50% RH) using the same measuring electronics used in the demonstrations (Supplementary Fig. 10). We then studied the capability of the organic sensor to detect H2 under three sensing scenarios.
In the first scenario, the sensor acted as a dynamic leak detector (Fig. 4a) tasked to monitor the H2 concentration leaking from a pipe connected to a cylinder containing 5% H2 in synthetic air under ambient conditions. The leak was controlled by a mass-flow controller. Throughout the experiment, we maintained a flow rate of 500 standard cubic centimetres per minute. The sensing unit was slowly moved by hand. A rapid reduction of the current was observed as it was moved through the leak area, corresponding to a H2 concentration of 610 ppm (Fig. 4d).

a, Schematic illustration of the measurement set-up for the H2 leak detection system. An organic sensing unit was moved slowly towards the hydrogen-exposed region to detect the hydrogen concentration leaking from a pipe carrying hydrogen gas. The data were wirelessly transmitted to a smartphone. b, Schematic illustration of the experimental set-up of an early warning system. A wireless organic DPP-DTT sensor was used to detect a simulated H2 leak. A balloon containing 100% H2 placed at various distances below the sensor was popped in a controlled, closed-room environment to simulate the leak. c, Schematic of the experimental set-up for wirelessly monitoring H2 concentration inside a closed room, which had two sensing nodes on opposite walls. Approximately 15 l of 100% H2 were released onto the floor and allowed to diffuse throughout the room. d, Temporal hydrogen response of the DPP-DTT sensor in terms of normalized voltage when detecting a hydrogen leak, as shown in a. e, Real-time response of the organic DPP-DTT sensor to repeated H2 leaks simulated by sequentially bursting three balloons 100 cm from the sensor. f, Comparison of the responsivity of the commercial and organic sensors calculated from the balloon-popping experiments. g, Time-dependent normalized voltage variations at nodes 1 and 2 when detecting hydrogen inside the room, as shown in c. h,i, Comparison of responsivity versus power consumption (h) and rise time versus the LoD (i) for various organic, oxide, Pd and transition-metal dichalcogenide based resistive hydrogen-sensing devices reported in the literature in the last two decades (Supplementary Table 2). sccm, standard cubic centimetres per minute; TMD, transition-metal dichalcogenide. Schematic in b was created with BioRender.com.
Source data
In the second scenario, we used the sensor as a warning system for H2 leaks inside a closed room (volume ~18 m3, at room temperature and 55% RH) and compared its response with a commercial hydrogen sensor (details in Methods). H2-filled balloons were placed 100 cm below the sensor unit and popped with a needle, as shown in Fig. 4b. The response of the organic sensor was recorded wirelessly in real time. Figure 4e shows the response signal from the organic sensor from the three sequentially burst balloons. The hydrogen concentrations measured by the organic sensor were 300 ppm (balloon 1), 280 ppm (balloon 2) and 275 ppm (balloon 3). The organic sensor exhibited a fast response of ~3 s, which was maintained even after several consecutive balloon bursts (inset in Fig. 4e). Extended Data Fig. 10c shows the real-time signals from both the commercial and organic sensors at distances of 100, 150 and 200 cm from the bursting balloon. In each case, the organic sensor exhibited a faster and larger response than the commercial sensor. Moreover, the responsivity of the organic sensor was almost double that of the commercial sensor (Fig. 4f). The results indicate that, as a hydrogen-sensing unit (electronics together with sensor), the wireless organic sensor demonstrates better device performance in terms of responsivity compared to the wired commercial apparatus.
In the third scenario, two sensing nodes (nodes 1 and 2) were deployed at different positions inside a closed room (volume ~16 m3) with a ventilation system (Fig. 4c). The sensors were positioned on opposite walls, approximately 2.5 m apart. Supplementary Fig. 11 is a photograph showing the dimensions of the room. Approximately 15 l of 100% hydrogen were released from several popped balloons placed on the floor of the room. The distances from the hydrogen source to nodes 1 and 2 were approximately 3.0 and 2.2 m, respectively. The released H2 gas diffused throughout the room, and its concentration was monitored wirelessly in real time by the two nodes (Fig. 4g). The hydrogen concentrations measured at nodes 1 and 2 were approximately 605 and 740 ppm, respectively. This simple yet powerful demonstration shows the potential for these organic detectors to be used in wireless sensor networks.
The organic sensor was even used in a quantitative analysis of H2 concentration under dynamic conditions. Specifically, the wireless sensor was mounted on a drone and quantitatively detected the hydrogen concentration while flying through an area where an H2 leak had just occurred (Supplementary Fig. 12). The measured hydrogen concentration was approximately 320 ppm. Similarly, its high-fidelity dynamic sensing capability was also observed in a water-splitting experiment. Several cycles of H2 generation were monitored with an average rise time of 14 (±4) s (Extended Data Fig. 10a,b and Supplementary Video 2). The extracted hydrogen concentration in the exhaust gas (Extended Data Fig. 10b) was ~335 ppm, as determined from the calibration curve in Supplementary Fig. 10 using a measured RH of ~52% at 20 °C (Supplementary Fig. 13). The dynamic response of the organic sensor was also compared against that of the commercial sensor by monitoring the evolution of H2 concentration during a chemical reaction between Mg and acetic acid inside a round-bottom flask (Supplementary Fig. 15). The organic sensor showed faster response and recovery times than the commercial device.
Finally, to investigate the potential for cost-effective fabrication of these devices, we used a screen-printing process to deposit DPP-DTT onto interdigitated Pt electrodes and measured the performance of the devices (Supplementary Fig. 14a,b). The responsivity of the screen-printed devices was comparable to devices prepared by spin-coating, thus demonstrating that this technology is amenable to cheap, high-throughput manufacturing. Compared to other sensor technologies (Fig. 4h,i and Supplementary Table 2), our organic hydrogen sensor has the potential to be simpler to manufacture with superior performance (higher responsivity, lower power consumption and so on) and thus offers a promising technical solution for the safe and widespread use of hydrogen technologies.
Conclusions
We have described a reversible hydrogen-induced de-doping mechanism in organic semiconductors, which we used to develop a robust hydrogen sensor. The organic sensor exhibits a high responsivity (1.19 × 104), fast response time (<1 s), low power consumption (≤2 μW; with no requirement for an external heating element), low LoD (~192 ppb), stability in harsh testing conditions and high selectivity for hydrogen even in complex gas environments. We also showed that our organic hydrogen sensor can outperform a commercial device in different practical sensing demonstrations. With its potential for scalable and cost-effective fabrication due to the attractive properties of organic semiconductors, our hydrogen sensor could be of widespread use in the emerging hydrogen economy.
Methods
Interdigitated electrodes
Titanium (50 nm)/platinum (150 nm) interdigitated electrodes with 5-μm gaps (IDE2) patterned on a glass substrate were purchased from Micrux and used as electrodes. All IDE2 samples were cleaned by ultrasonication in water, isopropanol and acetone for 5 min each before depositing the organic semiconductor.
Active materials
This work used the following organic semiconducting materials: DPP-DTT, C16-IDTBT, P3HT, TIPS-pentacene and Y6. DPP-DTT, P3HT and TIPS-pentacene were purchased from Ossila, and PTAA and Y6 were purchased from 1-Material. All commercially supplied materials were used as received without further purification. C16-IDTBT (Mn 65 kg mol−1, dispersity of a polymer (Ð) 2.4) was synthesized in our laboratory.
Material preparation
DPP-DTT (5 mg ml−1, chlorobenzene), C16-IDTBT (6 mg ml−1, chlorobenzene), P3HT (8 mg ml−1, chlorobenzene), PTAA (8 mg ml−1, chlorobenzene), TIPS-pentacene (5 mg ml−1, toluene) and Y6 (10 mg ml−1, chlorobenzene) were dissolved overnight on a hotplate at 70 °C while being stirred at 1,000 rpm. The DPP-DTT solution was spin-coated at 4,000 rpm for 60 s on the interdigitated Pt electrodes, followed by annealing at 100 °C for 90 min. C16-IDTBT, P3HT, PTAA and Y6 were spin-coated at 2,000 rpm for 60 s on the interdigitated Pt electrodes, followed by annealing at 100 °C for 90 min. The TIPS-pentacene solution (10 μl) was drop cast onto the interdigitated Pt electrodes and annealed at 50 °C for 5 min. The thicknesses of the different organic semiconducting layers, measured with a stylus surface profilometer (Bruker), were 21 (±5), 32 (±6), 42 (±9), 49 (±8), 32 (±5) and 463 (±42) nm for DPP-DTT, C16-IDTBT, P3HT, PTAA, Y6 and TIPS-pentacene, respectively.
Characterization
All hydrogen sensors were electrically characterized inside a Linkam HFS600E-PB4 chamber. The equilibrium of the gas flow was achieved with two mass-flow controllers (ALICAT). The temperature of the sensing device inside the chamber was varied to study the temperature-dependent H2 sensing. All the electrical characterization measurements were obtained with Agilent B1500A instruments. Custom MATLAB programs were used to control the mass-flow controllers when mixing and diluting the sensing gas according to experimental requirements. This program was also used to provide hydrogen in pulsed form on top of the sensing surface, and this process was employed to evaluate the response of the device. The response from the other volatile gases (not H2) was measured using the DPP-DTT device inside a custom-designed chamber. For the response measurements, a particular amount of ethanol, methanol, acetone or toluene was soaked in tissue paper and placed in the chamber to achieve a 1% environment of each gas. A DHT11 humidity sensor, which has a measuring accuracy of ±5%, was used to calibrate the humidity inside the gas-sensing chamber (Linkam) as well as the ambient humidity level for all the humidity-dependent H2-sensing measurements. A commercially available Pt Clark-type electrode-based H2-detecting system (Unisense, approximately US$1,200 per electrode) was also used to detect H2. We compared its response with that of our organic hydrogen-sensing device. The surface morphology of the DPP-DTT-coated interdigitated Pt electrode was studied using an atomic force microscope (NT-MDT) in tapping mode. The work function of the Pt electrodes was measured by photoelectron spectroscopy in air using RIKEN KEIKI equipment. Gas cylinders were purchased from either Abdullah Hashim Industrial Gases & Equipment or Air Liquide. The responsivity of the sensor to hydrogen gas was calculated as (R − R0)/R0, where R0 is the device current (or resistance) before hydrogen exposure and R is the device current after exposure to air containing 1,000 ppm H2. The LoD of the DPP-DTT device was extracted from a linear fit of the H2 concentration-dependent responsivity curve using 3.3 × (standard deviation of the intercept/slope of the linear curve), where the standard deviation of the intercept = (number of data points used in the measurement times the standard error of the intercept). The power consumption of the sensor was calculated as P = V × I, where V is the bias applied to the device and I is the measured sensing current.
DFT methodology
DFT studies were carried out with the Vienna Ab Initio Simulation Package52 using a plane-wave basis (with an energy cutoff of 400 eV), projector augmented waves53, the generalized gradient approximation for Perdew–Burke–Ernzerhof54 exchange-correlation functional and the DFT-D3 method for the van der Waals interactions55.
To simulate the effect of oxygen and hydrogen reactions with the polymers of interest, we modelled them with a single periodic chain. The sampling of the reciprocal space in calculations of the electronic DOS used the tetrahedron method with 18 k-points in the direction of the periodic chain56. The structures were rendered with the software VESTA57. The activation energies of pertinent processes were obtained with the nudged elastic method58, which models the so-called minimum energy pathway as a ‘chain’ of subsequent configurations (termed images) along the reaction coordinate.
Proof-of-concept demonstrations
We have demonstrated hydrogen sensing in three example use cases: hydrogen-generating chemical reactions, water electrolysis and the bursting of a H2-filled balloon. We used prototype electronics59 with the organic sensor to send sensing data wirelessly to an Android mobile phone, which analysed the data in real time. Photographs of the circuit layout and the printed circuit board used for the measuring electronics are shown in Supplementary Fig. 16. In these demonstrations, a commercial hydrogen sensor from UNISENSE was used for comparison. Its low power consumption (~2 μW) makes the organic sensor suitable for safe hydrogen detection applications as this reduces the risk of sparking. The organic sensor can detect hydrogen concentrations well below the flammable range (>40,000 ppm), enabling pre-emptive security alerts before hydrogen reaches hazardous levels.
To demonstrate the detection of hydrogen from chemical reactions (Supplementary Fig. 15), a three-necked 500-ml round-bottom flask equipped with a condenser and two septa was charged with a mixture of magnesium shavings (20 g) and water (50 ml). The top of the condenser was fitted with the DPP-DTT organic sensor and the commercial hydrogen sensor, and a continuous flow of air was introduced through one of the septa. Once the responses of the sensors had become stable, 1 M aq. acetic acid (0.1 ml) was added through the other septum. Data from both sensors were simultaneously recorded as the response to the generated hydrogen. After hydrogen evolution ceased, another portion of the aqueous acetic acid was added, and this process was repeated several times.
For the water hydrolysis experiments, we designed a small reaction chamber made of glass. We used steel springs as electrodes inside the chamber. Further, 2.5 M aq NaOH was used as the electrolyte. The spring electrodes were partially covered with small cylindrical plastic pipes, which collected H2 and O2 from the cathode and anode electrodes, respectively. The reaction chamber was connected to a 9.0 V battery (Duracell). Data from the organic sensor were recorded remotely by the bespoke electronics.
The balloon-bursting experiments were performed at room temperature in ambient air. The 100% hydrogen-filled balloons were placed at varying distances (100–200 cm) from the organic and commercial hydrogen sensors at stationary positions. The balloons were then burst using a sharp needle, and the response from the two sensors was recorded simultaneously in real time.
Responses