Solar water splitting provides a promising path for sustainable hydrogen production and solar energy storage. One of the greatest challenges towards large-scale utilization of this technology is reducing the hydrogen production cost. The conventional electrolyser architecture, where hydrogen and oxygen are co-produced in the same cell, gives rise to critical challenges in photoelectrochemical water splitting cells that directly convert solar energy and water to hydrogen. Here we overcome these challenges by separating the hydrogen and oxygen cells. The ion exchange in our cells is mediated by auxiliary electrodes, and the cells are connected to each other only by metal wires, enabling centralized hydrogen production. We demonstrate hydrogen generation in separate cells with solar-to-hydrogen conversion efficiency of 7.5%, which can readily surpass 10% using standard commercial components. A basic cost comparison shows that our approach is competitive with conventional photoelectrochemical systems, enabling safe and potentially affordable solar hydrogen production.
Water electrolysis (2H2O 2H2 + O2) combined with renewable power sources such as solar or wind provides a promising path for sustainable hydrogen production for fuel cell electric vehicles1 and power-to-gas storage of variable power sources2. One of the greatest challenges towards large-scale utilization of these clean energy technologies is reducing the hydrogen production cost. This may be achieved using photoelectrochemical (PEC) water splitting cells that directly convert water and sunlight to hydrogen and oxygen3. The current architecture of PEC water splitting cells4 resembles that of conventional water electrolysis cells5, comprising a sealed cell with two electrodes and a membrane that separates the O2 and H2 gas products. This architecture is well-suited for centralized hydrogen production in alkaline or polymer electrolyte membrane (PEM) electrolysers, but not for distributed solar hydrogen production in PEC cells. The low power density of the sunlight necessitates a large number of PEC cells, giving rise to critical challenges for gas separation, collection and transport to a centralized hydrogen storage and distribution facility safely and affordably. This work addresses these challenges by separating the hydrogen and oxygen production such that the hydrogen can be produced in a central generator far away from the solar field, where oxygen is produced. The ion exchange between the anode and cathode in the oxygen and hydrogen cells, respectively, is mediated by reversible solid-state redox relays that can be cycled many times with minimal efforts. Our separate cells approach works well for both electrolysis and PEC cells, as demonstrated in the following.
The conventional water electrolysis cell design is illustrated in Fig. 1a, featuring an alkaline electrolyser5. Both electrodes are dipped into one cell containing the aqueous electrolyte, with an ion exchange membrane or porous diaphragm separating the cell into anode and cathode compartments that generate O2 and H2, respectively. The two compartments can be separated using a salt bridge, as illustrated in Fig. 1b, but this is not a common practice because the series resistance of the salt bridge reduces the electrolysis efficiency. Recently, a new PEM electrolyser architecture was proposed, using a soluble molecular redox mediator (silicotungstic acid) that mediates the electron-coupled proton exchange between the oxygen and hydrogen evolution reactions (OER and HER, respectively)6. The redox mediator is reduced at the glassy carbon cathode of an electrolytic cell while water is oxidized at the Pt mesh anode, and then transferred to another cell with Pt catalyst that gives rise to spontaneous H2 evolution from the reduced mediator, as illustrated in Fig. 1c. Thus, the OER and HER are decoupled, and O2 and H2 are generated in separate cells. This provides potential advantages, as discussed in ref. 6, but it also has some critical drawbacks and limitations. Firstly, the reported efficiency of the electrochemical process in the electrolytic cell and the Faradaic efficiency for H2 evolution in the second cell were rather low, 63% and 68%, respectively, ending up in an overall efficiency of 43%. This is considerably lower than state-of-the-art PEM electrolysers that reach up to 68.3% based on the free energy metrics5, 7. Secondly, the redox mediator is acidic; therefore, precious Pt catalysts are required. And thirdly, given the dark colour of the redox mediator, the feasibility of this approach for PEC solar cells is questionable. To overcome these problems, we sought out a solid-state redox system that could mediate the ion exchange between the anode and cathode in alkaline aqueous solutions and be cycled multiple times. The material selection criteria of the desired redox mediator are discussed in the Supplementary Information. Accordingly, the NiOOH/Ni(OH)2 redox couple was selected for use as auxiliary electrodes (AEs), as illustrated in Fig. 1d. This idea was first disclosed by us in a patent application filed in 20158. A very recent report by another group demonstrates a similar concept with sequential H2 and O2 generation in two steps9, as opposed to continuous co-generation of H2 and O2 in separate cells presented herein.
NiOOH/Ni(OH)2 electrodes are widely employed in rechargeable alkaline batteries10, 11, where charged NiOOH electrodes reversibly convert to Ni(OH)2 on discharging: NiOOH + H2O + e−⇌ Ni(OH)2 + OH− (ref. 12). The NiOOH/Ni(OH)2 redox reaction in our AEs precedes the OER by more than 200 mV, as demonstrated in the cyclic voltammograms presented in Supplementary Fig. 1. This enables the AEs to serve as reversible redox relays that mediate the ion (OH−) exchange with the primary electrodes (anode and cathode) in separate cells, while the OER and HER take place only at the primary electrodes. Thus, O2 and H2 are produced in separate cells with no O2/H2 crossover. The operation principle of our two-cell water electrolysis system is illustrated in Fig. 1d and in more details in Supplementary Fig. 2. In the oxygen cell, the OER (4OH− O2 + 2H2O + 4e−) occurs at the anode, converting OH− ions supplied by the AE to O2 gas, while the AE transforms from NiOOH to Ni(OH)2. In the hydrogen cell, the HER (4H2O + 4e− 2H2 + 4OH−) occurs at the cathode, and the OH− ions generated by this reaction are consumed by the AE that transforms from Ni(OH)2 to NiOOH. Thus, one AE charges while the other one discharges. The electrons participating in the OER and HER reactions transfer from the cathode to the anode through a power source that drives the reactions, whereas the electrons participating in the AE redox reactions transfer from one AE to another through another metal wire. Summing all the reactions in the oxygen and hydrogen cells yields the overall water splitting reaction: 2H2O 2H2 + O2.
Owing to the remarkable cycling durability of the NiOOH/Ni(OH)2 AEs, which can be cycled thousands of times as demonstrated in rechargeable alkaline batteries13, electrolysis can be carried out repeatedly by cycling the AEs either by reversing the current polarity or swapping their places. To demonstrate this, Fig. 2 shows 40 electrolysis cycles in separate oxygen and hydrogen cells with Ni foil primary electrodes and commercial Ni(OH)2 AEs. Prior to this test, the AEs were activated through charge–discharge cycles during which the discharge capacity increased from 59% of the transferred charge (22.5 mAh) on the first cycle to 99.6% at the end of the activation process (see Supplementary Fig. 8). Following the activation process, a charged NiOOH electrode was placed near the anode in the oxygen cell, and a discharged Ni(OH)2 electrode was placed near the cathode in the hydrogen cell. The electrodes were connected by copper wires as illustrated in Fig. 1d, and electrolysis was carried out by forcing a constant current of 45 mA (5 mA cm−2) between the anode and cathode, while measuring the applied voltage, Vappl. Figure 2a shows Vappl as a function of time during 20 h of continuous operation in which the current polarity was reversed whenever the voltage reached a threshold limit of ±3 V. This limit was set to avoid oxygen evolution in the hydrogen cell and vice versa, as explained in the Supplementary Information.
In every cycle, Vappl increased gradually from 1.8 V at the beginning of the cycle to 2.3 V close to its end, ending with a sudden jump to the threshold limit (±3 V). The changes in Vappl followed the changes in the potential difference between the AEs in the oxygen and hydrogen cells, ΔUAE = UAE, O2 − UAE, H2, as shown in Supplementary Fig. 15(b). This resembles the discharge curve of a battery, indicating that the changes in Vappl track the state of charge of the AEs. The sudden jumps at the end of the cycles indicate complete discharge of the AEs, which must be recharged to continue the process. Here, recharging is achieved by reversing the current polarity and performing another electrolysis cycle with the oxygen and hydrogen cells being swapped, as illustrated in Supplementary Fig. 2. In Supplementary Fig. 12 we demonstrate cycling by swapping the AEs instead of reversing the current polarity as shown here. The two methods are equivalent and yield very similar results.
The first electrolysis cycle in Fig. 2 lasted 30 min before reaching the threshold voltage. Subsequently, the cycle duration decreased by 0.3%, on average, from one cycle to another, reaching 26 min after 40 cycles, as shown in Fig. 2b. This drift is due to incomplete charging of the AE in the hydrogen cell. However, the AE can be recharged back to its initial state whenever necessary as demonstrated in Supplementary Fig. 13. The cycle duration depends on the initial charge of the AEs and on the electrolysis current. It can be extended by increasing the charge and reducing the current. In Fig. 2 the AEs were charged to 22.5 mAh, a small fraction of their rated capacity (1,300 mAh, according to the vendor). Consequently, the cycle duration was short (30 min at 45 mA). In Supplementary Fig. 10 we show much longer cycles of >6 h achieved by charging the AEs to 448 mAh. During the whole test (125 h), gas bubbles evolved on the primary electrodes, but not on the AEs (see Supplementary Movie 1). Gas chromatography measurements confirmed stoichiometric O2 and H2 production in the oxygen and hydrogen cells, respectively, with Faradaic efficiency of ~100% and no O2/H2 crossover up to ~80% of the charged capacity of the AEs (see Supplementary Figs 19 and 21, respectively). Thus, the results presented here demonstrate that alkaline water electrolysis with continuous co-production of H2 and O2 gases in separate cells can be carried out using reversible solid-state AEs that can be cycled multiple times.
Next, we examine the electrolysis efficiency of our system and the polarization loss incurred by the AEs. At 100% Faradaic efficiency, the electrolysis efficiency equals the voltage efficiency, ηV = Vrev/Vappl, where Vrev = 1.23 V is the reversible voltage of water electrolysis (at 25 °C). Averaged over a 125 h electrolysis test with 20 cycles of >6 h (Supplementary Fig. 10), the applied voltage (Vappl) was 2.12 V, of which 0.12 V was the average voltage drop on the AEs, ΔUAE (see Supplementary Fig. 15(c)). Thus, the average electrolysis efficiency of our system was 58%. This is an excellent result for a simple electrolysis system with Ni foil electrodes operated under mild conditions (1 M NaOH, ambient temperature). The electrolysis efficiency can be readily enhanced by using rare-earth OER and HER catalysts such as RuO2 and Pt, respectively, as well as by using concentrated alkaline solution and operating the cell at elevated temperature, as commonly done in alkaline electrolysers5. These routes were not pursued in this study because we aim to demonstrate the effectiveness of our separate cells approach rather than break the efficiency record. Thus, more important than the overall electrolysis efficiency of the entire system is the polarization loss incurred by the AEs. This was, on average, 0.12 V, which is only 5.7% of the applied voltage. This is a reasonable toll, comparable to the Ohmic loss across the membrane in other PEC cell architectures14, 15, 16. ΔUAE increases with increasing current density, but the increase is slow, and at a current density of 40 mA cm−2 in the hydrogen cell and 4 mA cm−2 in the oxygen cell ΔUAE reaches only 0.16 V, see Supplementary Fig. 11. It is noteworthy that the polarization loss incurred by the redox reactions of the AEs scales logarithmically with the current, whereas the Ohmic loss across the membrane in conventional electrolysis scales linearly with the current.
Having demonstrated that the hydrogen cell can be separated from the oxygen cell in a simple alkaline water electrolysis system without degrading the efficiency with respect to alternative electrolysis architectures, we show next how this concept can be applied for the design of a hydrogen refuelling station with unsealed PEC solar cells or photoelectrochemical–photovoltaic (PEC–PV) tandem cells7, 17 connected by metal wires to a centralized H2 generator at the refuelling station, as illustrated in Fig. 3. The motivation for proposing this disruptive concept becomes clear when considering the requirements to separate and collect the H2 gas produced in a large number of conventional PEC cells distributed in the solar field and transport it to a centralized H2 storage and distribution facility, and the complex measures that must be taken to ensure safe operation in compliance with strict fire protection standards18. Considering as a case study a hydrogen refuelling station with a production rate of 400 kg H2 per day, an average insolation of 180 W m−2 and a solar-to-hydrogen (STH) conversion efficiency of 10%, the net area of the PEC solar cells should be >30,000 m2. Using 10 × 10 cm2 PEC cells, the largest PEC cells reported to date19, means an array of >3,000,000 cells. To collect the H2 from all these cells they must be sealed and fitted with H2 gas piping manifold. In addition to the immense piping construction, the cells must be fitted with membranes to prevent intermixing of H2 and O2, a highly flammable gas mixture, adding complex engineering and material challenges20. Furthermore, to comply with the strict fire protection standards of hydrogen technologies18, complex safety measures would have to be taken to ensure that no gas leaks occur within the PEC cells and at the joints along the H2 gas piping manifold, and that unexpected accidents are immediately extinguished. The required safety measures give rise to tremendous efforts and expenses, as discussed in the Supplementary Information. These challenges, in addition to efficiency and stability challenges, render solar H2 production using conventional PEC cells economically questionable21, 22. Our separate cells architecture overcomes these hurdles in an elegant way that is also very feasible, as demonstrated below.
Since the performance of current photoanodes for PEC water splitting is not sufficiently high for their integration with high-end PV cells7, 23, we demonstrate the utilization of the AEs in a PV-coupled electrolysis system that simulates a PEC–PV tandem cell. The system comprises of a Si PV module connected to oxygen and hydrogen cells with primary Ni foil and Pt-plated stainless-steel mesh electrodes, respectively, and activated NiOOH and Ni(OH)2 AEs. The system design follows the constraints that an analogue PEC–PV tandem cell would be subjected to; namely, the oxygen cell conformally maps the PV module, and the AEs in the oxygen and hydrogen cells are interchangeable. The inset in Fig. 4 shows a schematic illustration of our solar water splitting system. The geometry of the cells is depicted in Supplementary Fig. 22.
The current density–voltage (J–V) characteristics of the PV module, measured under simulated solar radiation (AM1.5G), and of the two-cell water electrolysis system are depicted in Fig. 4 by the red and blue curves, respectively. These J–V characteristics were measured separately, prior to connecting the PV module to the electrolysis system. The two curves cross at V = 1.7 V and J = 6.7 mA cm−2. However, the operation points of the coupled system were at slightly higher voltages and lower current densities, as shown by the green dots in Fig. 4. This small deviation is due to additional coupling losses such as wire and contact resistances as well as the voltage drift of the electrolysis system, as was discussed before. It is noteworthy that the short-circuit current density of this particular PV module, Jsc = 7.0 mA cm−2, is considerably lower than the vendor’s specifications (8.3 mA cm−2)24. The black dashed line curve in Fig. 4 shows the expected J–V curve of a PV module with Jsc = 8.3 mA cm−2.
The STH efficiency of the coupled system is calculated according to the following equation23:
where Isc is the short-circuit current of the PV-coupled electrolysis system (measured at standard AM1.5G insolation conditions), ηF is the Faradaic efficiency, 1.23 V is the reversible voltage (Vrev) of the water splitting reaction (at 25 °C), Amodule is the area of the PV module (6.03 cm2), and Pin is the power density of the incident light (100 mW cm−2). Taking ηF = 100% (see Supplementary Fig. 19) and Isc/Amodule = Jsc = 6.05 mA cm−2 (Fig. 4), the STH efficiency was 7.5%, averaged over 1 h of operation. Taking the photoactive area of the PV module (4.9 cm2) instead of the total area (6.03 cm2), which includes inactive area due to interconnects, yields an average STH efficiency of 9.1%. The ratio between the photoactive and inactive area can be readily increased in large-area PV modules; therefore, the higher STH value (9.1%) can be readily approached by sizing-up the components. Furthermore, considering the difference in Jsc between the particular PV module that was used in this test (7.0 mA cm−2) and the vendor’s specifications (8.3 mA cm−2), it is expected that the STH would increase by up to 18.6%, relative to the present result, by selecting a better PV module. This would bring the STH efficiency up to 8.9% based on the total area of the PV module, or 10.8% based on the net photoactive area. Further improvement would be possible by tailoring the J–V characteristics of the PV module to cross the J–V characteristics of the electrolysis system at its maximum power point, for example, using a d.c./d.c. power converter. Assuming a 90% d.c./d.c. power conversion efficiency, this would bring the STH efficiency up to 11.7%, based on the net photoactive area (see Supplementary Fig. 24).
Despite the use of a PV module of rather low performance that does not optimally match the electrolysis system, and primary electrodes that have not been optimized for maximum electrolysis performance, the efficiency of our system is comparable to state-of-the-art solar water splitting systems including both buried junction25, 26 and PV-coupled electrolysis configurations27, 28, 29 (see Table 1). Although higher STH efficiencies have been reported, they were obtained using high-efficiency non-commercial28 or expensive double-junction25, 26 or even triple-junction concentrator PV cells29, rare-earth catalysts26, 29, and without product separation26, 27, 28. More important than the STH efficiency, which depends greatly on the power conversion efficiency (PCE) of the PV cell used to drive the water splitting reaction, is the normalized STH/PCE ratio that is independent of the PV efficiency7. On the basis of this figure of merit, our system scores respectively high (see sixth column in Table 1). Besides efficiency, another important feature of solar water splitting systems that has great impact on their safety and cost is whether the hydrogen production is distributed or centralized. Examples of distributed systems are buried junction25, 26 and PEC–PV tandem cells7, 17; whereas centralized systems are present in PV-coupled electrolysis27, 28, 29 and our separate cells architecture. Centralized hydrogen production has important operational advantages in terms of H2 gas collection, process compactness, safety, and so on, whereas PEC–PV tandem cells offer potential advantages in efficiency7 and cost21, 30. Our separate cells approach yields the best of both worlds: it enables centralized hydrogen production, as illustrated in Fig. 3, yet it also works for PEC–PV tandem cells, as demonstrated in the following.