Design Principle and Loss Engineering for Photovoltaic-Electrolysis Cell System.

The effects of exchange current density, Tafel slope, system resistance, electrode area, light intensity, and solar cell efficiency were systematically decoupled at the converter-assisted photovoltaic-water electrolysis system. This allows key determinants of overall efficiency to be identified. On the basis of this model, 26.5% single-junction GaAs solar cell was combined with a membrane-electrode-assembled electrolysis cell (EC) using the dc/dc converting technology. As a result, we have achieved a solar-to-hydrogen conversion efficiency of 20.6% on a prototype scale and demonstrated light intensity tracking optimization to maintain high efficiency. We believe that this study will provide design principles for combining solar cells, ECs, and new catalysts and can be generalized to other solar conversion chemical devices while minimizing their power loss during the conversion of electrical energy into fuel.

Introduction

Splitting water into hydrogen is a prominent pathway for solar energy conversion and storage.[1] The recent development of converting excess electricity into hydrogen supports the advent of the hydrogen economy.[2] Additionally, the hydrogen generated from solar water-splitting reaction shows the possibility to meet the current hydrogen demand.[3] Three different systems have been suggested for using solar energy to convert water into hydrogen. These systems are the photoelectrochemical electrode (PEC), photocatalyst, and photovoltaic–electrolysis cell (PV–EC) systems. Although the state-of-the-art PEC and photocatalyst have 10 and 5% solar-to-hydrogen conversion (STH) efficiency, respectively, a much higher efficiency has been achieved with PV–EC systems.[4−15] The ease with which the PV–EC systems can be scaled up is regarded to be a great advantage for industrial applications. This superior efficiency can be realized by combining the following two superior pre-existing infrastructures: efficient photovoltaics (PVs) as the power generator and an electrolysis cell (EC) operated at low overpotential for the hydrogen production.[16−19] Moreover, the compact EC structure fabrication with low electrolyte resistance and finely controlled electrolyte flow for mass transfer limit preclusion are required to maximize the efficiency of the overall system.

The primary requirement for designing an efficient PV–EC system for commercialization is highly efficient and stable photovoltaic (PV) and electrolysis cell (EC) with low overpotential. In the case of the EC system, adoption of efficient water-splitting catalysts and proper cell configuration to minimize the solution resistance and prevent the mass transfer limit at the solution/electrode interface are needed. More importantly, an efficient combination of independent PV and EC systems is also a critical issue for improving the STH efficiency of the PV–EC system. For instance, to use the current close to the short circuit current of PVs, Grätzel group used a NiFe double hydroxide catalyst electrode that has a 16 times larger area than the light-irradiation area on perovskite PVs (15.7% efficiency) and achieved 12.3% STH efficiency, which is equivalent to 78.3% of the PV maximum power.[10] Nocera group highlighted the appropriate number of series-connected Si PVs (16% PV efficiency) on the PV–EC system by comparing the STH efficiency of systems based on three series-connected PVs (2.8%) and four series-connected PVs (10%).[20] Additionally, Fujii group varied the ratio of series-connected ECs and PVs (31.2% of its efficiency) to achieve 24.4% STH efficiency, by consuming 78.2% of the PV-driven maximum power for hydrogen conversion, which is far more efficient than the same number of PVs and ECs connected PV–EC system, with a 14.7% STH efficiency.[21] Recently, by optimizing the electrode area of the EC, solar light density, connection methodology, and temperature, Jaramillo group achieved the highest ever STH efficiency (∼30%).[22] Although the STH efficiency record has been continuously broken by these pioneering works, it seems that general guidelines to analyze the determinants of each subcompartment in the PV–EC system do not exist. Therefore, further interface engineering and optimization of each compartment can help design a record with high STH-efficient systems.

Here, we conduct a systematic analysis of the PV–EC system to investigate the efficiency determinants by studying models and performing experiments. The decoupled key parameters investigated in our study included the following: (i) existence of a dc/dc converter, (ii) catalyst overpotential, (iii) PV efficiency, (iv) Tafel slope, (v) electrolyte resistance, (vi) surface by modeling, (vii) converter efficiency, (viii) the number of series-connected PVs, and (ix) light intensity. To validate the design principle that we suggested by both modeling and experiments, we built converter-assisted PV–EC (PV-Conv-EC) systems with an independently defined PV and EC. The PV that we used was the record-high single-junction GaAs PV (26.5%) fabricated by LG Electronics Inc., which was recently marked in the National Renewable Energy Laboratory (NREL) solar cell efficiency chart. The superior efficiency of GaAs PVs is due to the rear junction structure consisting of a top base (n-GaAs) layer and a bottom emitter (p-Al0.3Ga0.7As) layer for minimizing carrier recombination. Additionally, fill factors can be maximized in this structure because the band offset caused by heterojunction between base and emitter layers can be further decreased by the n-graded layer.[23] The thickness of the n-graded layer was optimized as 80 nm, which is the perfectly optimized distance for minimizing valence band bending related to the high recombination rate. Because the fabrication procedure can be used for all GaAs wafers regardless of their size, the scalable and high-performing GaAs PVs can be used as the power supply for designing our PV–EC system. The EC system consisted of IrO, Pt/C nanocrystalline materials on carbon electrodes and membrane-electrode-assembled (MEA) configuration with low electrolyte resistance. However, the current analysis using these two catalysts can be generalized for use with other cheap and earth-abundant catalysts. Note that the following analyses were made by characterizing the parameters of catalysts such as Tafel slope and overpotentional. On the basis of our design principle, we finalized the PV–EC system composed of GaAs PVs, dc/dc converter, and MEA EC. The EC electrode area used in this work was 6 cm2, and 40–120 mW of power can be stored as hydrogen energy depending on the PV area (2–6 cm2). The maximum STH efficiency that we achieved by controlling the subcompartments was 20.6%, and 78% of the maximum PV-driven electricity was converted into hydrogen energy. We believe that the new guideline and novel analysis proposed in this work will open up new possibilities not only for analyzing the efficiency of a PV–EC system but also by providing insights into the way PV-Conv-EC systems can be implemented regardless of the PV efficiency, PV and EC connection variation, or catalysts.

Results and Discussion

Design Principle by Analysis in Terms of Subcompartment Variables of the PV-Conv-EC System

To decouple the determinants of the efficiency of a PV–EC system, the model system with an independent PV and EC governing the current density–voltage (j–V) equation can be expressed as followsEquation is the diode equation of PV materials consisting of a short circuit current density (jSC), dark current density (j0), charge of electron (q), series resistance (Rs), ideal diode factor (n), ideal gas constant (k), temperature (T), shunt resistance (Rsh), photogenerated voltage (VPV), and current density (jPV). Equation is the EC j–V relationship consisting of thermodynamic water electrolysis potential (V0, 1.23 V at 298 K), Tafel slope of the cathode and the anode (τcat and τano, respectively), exchange current density of the cathode and the anode (j0,cat and j0,ano), solution resistance (Rsol), applied voltage on EC (VEC), and current at EC (jEC). The EC electrode and light irradiance area to the PV are regarded as the same unless otherwise stated. Moreover, in eq , the total overpotential of the EC catalysts for reaching 10 mA cm–2 (η10mA) except for the solution resistance can be denoted as followsBecause the standard potential difference between water oxidation and reduction is 1.23 V, which is identical to the Gibbs free energy content to form hydrogen and oxygen under standard conditions, at least 1.23 V of open circuit voltage is necessary to split water. However, considering that conventional single-junction PVs such as Si (0.7 V) or GaAs (1.1 V) exhibits less open circuit voltage than 1.23 V, multiple PVs should be series-connected to supply enough potential.[20,23] Therefore, we designed the model system based on two series-connected PVs.

Before performing the detailed calculation, parameters affecting the efficiency of a PV–EC system can be predicted by a simplistic model. Because conventional PV–EC system is composed of a direct electrical connection between PV and EC, this situation can necessitate the current density and voltage of PV and EC to be identical as follows: jPV = jEC and VPV = VEC. Therefore, graphically, this necessity determines the operational state of the PV–EC system as the intersection point of individual PV and EC j–V curves (Figure a). For instance, the intersection point can be obtained at 1.59 V of voltage and 15 mA of current from two intersecting j–V curves, a series-connected PV with a 29% efficiency and an EC with 300 mV of η10mA, in Figure a. This point reflects the operating situation of the modeled PV–EC system. From this point, the amount of power that is stored as hydrogen (pH2) per square centimeter and wasted as loss due to overpotential (pkin) per square centimeter can be estimated. pH2 can be expressed as the intersection current multiplied by 1.23 V because chemical energy stored as hydrogen is identical to the Gibbs free energy change (2H2 + O2 → 2H2O). Thus, considering that pH2 is proportional to the intersecting current, the intersection point close to short circuit current of the PV indicates higher pH2. In this model, we thought that parasitic current triggered by charging the electrodes, catalyst materials degradation, or side reaction was negligible compared with water-splitting reaction. Thus, we assume that the faradaic efficiency (ηF) was 1, 18.5 mW cm–2 of pH2 was generated, and 5.3 mW cm–2 of pkin was wasted. However, the PV–EC operating power (pH2 + pkin, 23.8 mW cm–2) is always lower than the maximum power of PV (pPV,max, 29 mW cm–2), as shown in Figure a. The discrepancy between the pPV,max point and high pH2 point is due to the current density–voltage coupling. If it is possible to make the EC use pPV,max regardless of the intersection current by decoupling the current–voltage of the PV and EC, a higher pH2 can be achieved.

Figure 1

Design principle of the PV-Conv-EC system based on an independent PV, the EC performance, and the existence of a converter. (a) Hydrogen power per square centimeter (pH2) and kinetic loss per square centimeter (pkin) at a given current density–voltage (j–V) curve of the PV and EC. The intersection between the PV and EC j–V curve has a lower voltage and a higher current density than the pPV,max point. (b) pH2 and pkin after the dc/dc converter assistance on (a). The extra pH2 can be achieved by pPV,max utilization. (c) pH2 and pkin of the PV-Conv-EC system at each pPV,max and the overpotential of reaching the 10 mA cm–2 (η10mA) condition. Extra pH2 gain by the application of the converter is also displayed. (d) pH2 of PV–EC and the PV-Conv-EC system depending on the EC performance (Tafel slope and η10mA) at 30% PV efficiency.

To resolve the limitation of current density–voltage coupling between PV and EC, we propose a second model system that represents the PV–EC system with a (i) dc/dc converter. PV-linked converter application is a well-known method to help electrical power consumption devices to use the maximum electrical power of PVs.[24−26] Therefore, the converter application allows pPV,max to be consumed on the EC by converting PV voltage and current density at pPV,max point into voltage and current density following the EC j–V relationship, called the “maximum power point tracking” (MPPT). In this modeling, we assume that the converter efficiency is 100%, indicating that the power generated by the solar cell can be fully consumed by the EC without power loss. Comparison of the MPPT process with the previous model is displayed in Figure b, which is equivalent to the PV-Conv-EC system model assuming that no power loss occurs during the converter application. Compared with pH2 in Figure a (18.5 mW cm–2), the MPPT process on the PV–EC system shows higher pH2 gain (22 mW cm–2), indicating that 3.5 mW cm–2 of “extra pH2” can be additionally stored. Although 7 mW cm–2 of pkin still exists even after using the converter, the total loss of the PV-Conv-EC system is far lower than that of the converter-unassisted model, with 10.5 mW cm–2 of electrical power loss. Additionally, the converter application can always ensure maximum power utilization generated from the solar cell regardless of the light intensity, which will be further discussed later.

To analytically investigate the effect of the catalyst performance on the PV-Conv-EC system, the amount of pH2 was determined along with the (ii) catalyst overpotential parameter (η10mA) at (iii) various PV efficiencies. Figure c describes the pH2 gain by the PV-Conv-EC system with different subcompartment applications; in this work, two η10mA (300 and 600 mV) and three PVs with different efficiencies (10, 20, and 30%) are selected as independent EC catalysts and PV performance descriptors, respectively; especially, the 300 mV of η10mA can be almost achieved even with the transition-metal-based catalyst materials such as NiCeO water oxidation and NiMo water reduction catalysts with 280 and 40 mV overpotential reaching 10 mA cm–2, respectively.[27,28] Solar cells with 10, 20, and 30% efficiency were modeled based on the representative examples such as a series-connected organic solar cell, perovskite solar cell, and GaAs solar cell, respectively.[23,29,30] The j–V curves of PVs are shown in Figure S1. At each pH2 gain, the extra pH2 is denoted as the check patterns in Figure c to indicate the amount of additional pH2 gain after the converter application on the PV–EC system. Thus, the pH2 region except for the extra pH2 is identical to the pH2 gain by the PV–EC system without the converter application. Focusing that the pH2 with the PV–EC system without the converter is almost consistent despite catalysts with different η10mA utilization, the converter application is an indispensable factor to thoroughly receive benefits of low overpotential catalysts. Taking into account the importance of the converter application, we compare the pH2 gain of the PV-Conv-EC system at each independently varied PV and EC performance parameters. The highest pH2 gain (23 mW cm–2) is observed for the PV-Conv-EC system in combination with PV with the highest electrical conversion efficiency (30%) and the lowest catalyst η10mA (300 mV). With the same PV utilization, however, the pH2 is only 19.4 mW cm–2 when a catalyst with relatively higher η10mA (600 mV) was used. This indicates that 3.6 mW cm–2 of additional pH2 can be achieved by improving the performance of the catalysts. By contrast, in the case of 10% PV, only 1.3 mW cm–2 of additional pH2 gain is achieved when lower overpotential catalysts are used. Considering that the additional pH2 by catalyst utilization with low overpotential at 30% PV is approximately threefold higher than that with 10% PV, the dominance of catalyst overpotential becomes more firmly established with highly efficient PV utilization by the PV-Conv-EC system.

The pH2 variation can be further extended in terms of specific EC catalyst parameters such as the summation of (iv) the Tafel slope of the catalysts (τcat + τano) and η10mA based on eqs and 3, as shown in Figure d. Although pH2 gains by both PV–EC and PV-Conv-EC systems are highly dependent on η10mA, the pH2 profiles along the varied catalyst η10mA are different. For instance, pH2 at PV–EC without the converter was almost constant when η10mA is lower than 700 mV and started to decrease abruptly when η10mA exceeded 700 mV. By contrast, instead of an abrupt pH2 decrease, a gradual decrease in pH2 is observed at the PV-Conv-EC system. Furthermore, compared with the PV–EC system, pH2 of the PV-Conv-EC system is dependent more on τcat + τano at a low η10mA. For instance, under a fixed η10mA (300 mV), consistent pH2 was kept at 18.5 mW cm–2 with the PV–EC system without the converter application. By contrast, 22.6 mW cm–2 of pH2 can be increased to 23 mW cm–2 when a catalyst with 90 mV dec–1 of τcat + τano was used instead of catalyst with 200 mV dec–1 at the PV-Conv-EC system. We also found that the decrease in τcat + τano remarkably fosters pH2 to increase when catalysts with lower η10mA are used. This is because relatively small pH2 increase was observed with a decrease in the Tafel slope at a high overpotential such as pH2 increased from 16.2 to 16.3 mW cm–2 when the τcat + τano was decreased from 200 to 90 mV dec–1 at an η10mA of 1000 mV. Considering that pH2 increase was more than 0.4 mW cm–2 because of the decrease in the Tafel slope at 300 mV of η10mA, a relatively small amount of additional pH2 (0.1 mW cm–2) was obtained with the catalysts with 1000 mV of η10mA. This case study shows that the τcat + τano decrease can help the PV-Conv-EC system to take advantage of high pH2, particularly using catalysts with a low overpotential. From the results shown in Figure c,d, we can conclude that high pH2 (23 mW cm–2) gain is achieved with the catalysts with low τcat + τano (90 mV dec–1) and η10mA (300 mV) parameters and that are identical to 77% power of pPV,max (30 mW cm–2), far higher than 60% pPV,max to pH2 ratio with the conventional PV–EC system.[31]

Even if efficient catalysts with low overpotential and Tafel slopes are used on the EC, (v) solution resistance of the EC can be affected by the pH2 value based on the solution resistance term in eq . Additionally, in a practical situation, it is possible to control (vi) the EC to PV surface area ratio. This control can alter the current-related constant such as the ratio among jSC, j0, j0,cat, and j0,ano, triggering pH2 to be varied in return. Figure S2 shows the pH2 variation in terms of solution resistance and AEC/APV ratio to analyze their dominancy. In real situations, many variables related to the catalyst or solar cell will not exactly follow the relationship that we verified in Figure S2. However, Figure S2 tells us that an efficient and a low-cost EC catalyst should be used for the PV-Conv-EC system for effectively increasing pH2 instead of increasing the EC electrode surface area. The low resistance and high AEC/APV configuration guarantee a higher pH2 gain with the PV-Conv-EC system in terms of light-irradiation area on PVs. In this regard, MEA EC configuration perfectly matched this requirement because of the thin membrane that helps to minimize the electrolyte resistance and to control the EC electrode area by simply preparing an optimal area of supporting electrodes and membranes.

Scheme indicates the schematic of the PV-Conv-EC system composed of independent devices to meet all requirements suggested from Figures and S2. To be specific, series-connected GaAs PV modules, buck-type dc/dc converter, and MEA ECs are consecutively linked using wires. In addition to the low resistance and feasibility to control the AEC/APV ratio, the MEA configuration of the EC also warrants high purity of hydrogen and low resistance because of the existence of thin proton exchange membrane.

Scheme 1

Schematic of the System That Is Based on Two Series-Connected Single-Junction GaAs PVs Equipped with a dc/dc Converter and an MEA EC

The electrolyte was continuously circulated only at the anode compartment of the EC.

Improving Design Principle by Implementing the PV-Conv-EC System

For implementing the PV-Conv-EC system, we assembled the EC and verified its performance in terms of parameters derived from catalyst materials, system configuration, and product purity (Figure ). To be specific, IrO and Pt/C nanocrytallines were used as water oxidation and reduction catalysts (Figure S3). On the basis of this catalytic property of materials, the Tafel slope and overpotential of the catalytic material were analyzed by converting the j–V curve into an overpotential–log j curve (Figure a). We specify the log j range of 0.9–1.3 to purely analyze the water-splitting reaction of our catalysts without the charging effect. The slope of the overpotential–log j curve of each catalyst corresponds to the Tafel slope. Moreover, after combining catalyst electrodes into an EC, the performance of EC can be overlaid on the overpotential–log j slope after iR compensation process by resistance derived from Figure S4. The slope (98 mV dec–1) for increasing 1 order of current density at the EC was perfectly consistent with the summation of the Tafel slope of the cathode (32 mV dec–1) and anode (66 mV dec–1) electrodes. Furthermore, the overpotential for reaching 10 mA cm–2 (300 mV) was also identical to the summation of the cathode (30 mV) and anode (270 mV) overpotentials. Altogether, it is valid that no parasitic current is generated after implementing the EC with both catalyst materials. Seemingly, we can predict the final pH2 with the given catalyst performance thanks to the proposed design principle as suggested above.

Figure 2

Electrochemical analysis of the electrode material and the MEA EC. (a) Tafel slope of each cathode and anode and the MEA EC system. The sum of the Tafel slope and the overpotential of each electrode is similar to the MEA EC. (b) j–V curve of the MEA EC at different electrolyte resistances controlled by the distance between the cathode and anode. The inset displays the similarity of the j–V curve at each resistance after iR compensation.

To better achieve low resistance on the EC, MEA configuration was implemented with nanocrystalline catalyst decoration on its electrode. In MEA cell with 6 cm2 of electrode area, proper distance between electrodes should be used to minimize the electrolyte resistance and product separation. Because the electrolyte resistance is in proportion to the distance between the cathode and anode electrodes, distance between the electrodes was controlled by Nafion thickness. The electrolyte resistance decreased as the thin membrane was assembled (Figure S4b). Thus, the MEA EC with low resistance guarantees higher current density at the same applied voltage (Figure b). Considering that the j–V performance after solution resistance compensation is similar regardless of the electrolyte resistance, it is plausibly attributed to an identical catalytic performance even under different electrode distances. Additionally, ηF was measured at each MEA EC (Figure S5). Regardless of the distance between electrodes in the EC, H2 faradaic efficiency (ηF) was almost 100%. To be specific, the EC with 2.1 Ω cm2 of resistance showed 99.6% of ηF, indicating that even the MEA EC with the lowest electrolyte can guarantee pure hydrogen collection without hydrogen crossover into the anode compartment. The minimal amount of O2 (0.2% of produced H2 at 120 s) at the cathode compartment also supports this. Instead, O2 existed in the anode chamber (ηF = 89%) of the EC (Figure S6). Because an EC with the lowest resistance guarantee purified H2 storage and ηF, we applied the EC with the lowest resistance to PV-Conv-EC connection.

The electrical activity of the PV-Conv-EC system can also be illustrated as equivalent circuits (Figure a). Compared with conventional PV–EC system circuits, which are identical to the combination of independent PV and EC circuits, a converter between independent systems is the distinguished feature of the PV-Conv-EC system circuit.[32] The resistance and capacitor at each EC electrode represent power consumption due to the water-splitting reaction and the charging effect caused by the electrodes. The PV circuit contains shunt and series resistance. In the case of the converter circuit, the degree of switch (S1) on/off ratio, called the “duty ratio”, and the inductor help the EC to use the maximum power of the solar cell (pPV,max) regardless of the current–voltage coupling. When S1 is off, S2 is automatically turned on to operate its circuit. In our case, the buck-type converter is used, and this helps decrease the voltage to be applied to the EC compared with the voltage at the PV. Despite the decreased voltage at the EC, the current on EC is far higher than the photogenerated current in return. Additionally, based on the fact that the average voltage gain at the power-consuming part is identical to the photogenerated voltage, multiplying it with the duty ratio of the converter, theoretical input duty ratio is determined to be VEC/VPV, where the power generated by the PV at VPV and the power consumed by the EC at VEC are equivalent.[33] Altogether, to use the pPV,max at the EC compartment, the duty ratio for the MPPT process should be predicted as VEC,MPPT/VPV,max, where VPV,max and VEC,MPPT indicate the pPV,max voltage and the voltage that was applied to the EC to consume the pPV,max, respectively. In addition, the converter circuit has complementary metal–oxide–semiconductor (CMOS), it is unable to escape from power loss during the conversion process. Power losses in CMOS circuits include the conduction loss caused by the parasitic resistance and the switching loss attributed to the switching operation of the metal oxide semiconductor field effect transistors (MOSFETs).[33] Thus, even with the MPPT process, only ηconvpPV,max (ηconv < 1, ηconv = converter efficiency) of power is used by the EC to generate hydrogen.

Figure 3

solar-to-hydrogen energy conversion process of the PV-Conv-EC system at different PV illumination areas. (a) Diagram indicates circuit of the PV-Conv-EC system. Maximum power of PV (pPV,max) can be provided to the EC with the optimum duty input to the dc/dc converter, which is called MPPT. ηconvpPV,max is distributed to the pH2 and pkin. (b) Predicted STH efficiency based on power–voltage curve and I–V curve, assuming 100% of pPV,max consumption on the EC. The filled circle and hollow circle represent the pPV,max point and the point where the EC consumes pPV,max at each surface area of the PV. MPPT can be achieved by inputting the theoretical duty (DT), which is the ratio between the voltage of the EC that consumes pPV,max (VEC,MPPT) and the voltage at the pPV,max (VPV,max). (c) Current density and STH efficiency under chopped illumination on the PV-Conv-EC system under optimized duty (DO). The recovery of current density after each chopping cycle indicates the stability of the system. (d) Converter efficiency (ηconv) and (VEC/VPV) (1/D) at each input duty (D) under various APV conditions. The range of D was DO ± 0.02 to show the duty dependency near the MPPT region. (e) Converter efficiency (ηconv) and (VEC/VPV) (1/D) with a wide range of input duties (D) with APV = AEC = 6 cm2 configuration under 100 mW cm–2 light irradiance. The ηconv and (VEC/VPV) (1/D) values are almost consistent throughout the D range.

With the independent PV and EC performance parameter, pH2 at the implemented PV-Conv-EC system can be predicted (Figure b). The I–V curve of the series-connected GaAs PVs is shown in Figure S7. STH efficiency can be derived from pH2 from overlaid graphs by the following equationAlthough the original STH efficiency is expressed in terms of hydrogen evolution rates (n) and its free energy (236 000 J mol–1) divided by APV and light density (psol), the numerator of the equation can be converted into current measured on the EC compartment (IEC), 1.23 V, and ηF at the EC. We assume that the converter can transfer the full amount of the PV-generated power to the EC. Through calculation, STH efficiencies can be estimated to be 21.5% (17.52 mA cmPV–2), 20.9% (17.03 mA cmPV–2), and 20.6% (16.78 mA cmPV–2) when AEC/APV ratios were 3, 1.5, and 1, respectively. As calculated in Figure d,0.996 of ηF was used for STH prediction. Additionally, based on the overlaid power (P)–V curve in Figure b, the theoretical duty for the MPPT (DT) as we mentioned previously was expected to be 0.84, 0.85, and 0.87 at each configuration when AEC/APV ratios were 3, 1.5, and 1, respectively. Without the converter application, only 18.4% STH efficiency can be estimated because of the voltage–current coupling at all configurations.

By connecting the independent PV, converter, and EC, as shown in Scheme , we can generate and collect hydrogen (Figure S8). In this system, the STH efficiency can be converted from IEC measured by potentiostat as an ammeter (Figure c). In a practical situation, the optimized input duty for MPPT (DO) during the PV-Conv-EC operation was not identical to the DT predicted from Figure b. Therefore, an extra procedure for determining the DO for each configuration is necessary. The MPPT process in the experiment was conducted by deciding whether pPV,max and measured power (IPV × VPV) at the PV compartment were identical during the system operation. After the converter reached the DO, the current was measured and converted into STH efficiency, as derived from eq . The measured current densities per light-irradiated area (converted STH efficiency from measured current on EC) were 16.81 mA cmPV–2 (20.6%), 15.96 mA cmPV–2 (19.6%), and 15.56 mA cmPV–2 (19.1%) when AEC/APV was 3, 1.5, and 1, respectively. STH efficiencies were lower than the value driven from Figure b. The lower efficiency can be related to the converter efficiency because the converter actually used a small amount of PV power for its operation. Without the converter application, 18.4% STH efficiency can be measured (Figure S9).

For an in-depth understanding of (vii) ηconv and the discrepancy between DO and DT, we analyzed the converter efficiency with DO condition at the given AEC/APV configuration. At each DO, ηconv can be calculated by dividing the electrical power consumed on the EC by photogenerated PV power as followsBecause all used IEC, VEC at the EC and IPV and VPV at the PV can be measured by a potentiostat and voltmeter, ηconv at each configuration can be derived (Figure d). Interestingly, ηconv increased when higher AEC/APV EC configuration was used. For instance, approximately 95.9% ηconv can be achieved when AEC/APV was 3, which is far higher than 92.8% ηconv when AEC/APV was 1.

Note that the degree of consistency between DO and DT (DT/DO) in Figure b becomes larger with higher AEC/APV EC configuration. Moreover, the values between ηconv and DT/DO are almost identical, indicating a correlation between the two parameters. The correlation can be theoretically derived from the following equationThe transformation of the IPV/IEC value into DO is based on our observation of equivalency with two values (Table S1). The consistency between IPV/IEC and DO can be attributed to the relatively large inductance (1 mH) of our converter with an operation frequency of 20 kHz, which significantly reduces the ripple current of the inductor in the converter. Additionally, because VEC/VPV during the MPPT process was identical to DT, ηconv and DT/DO values agreement can also be derived (Figure d). Despite controlling the input duty (D) so as not to be identical to DO at each configuration, equivalence between ηconv and (VEC/VPV) (1/D) was still maintained, indicating the validity of eq .

We also acknowledged that ηconv becomes lower as DO becomes higher. The high DO value is identical to the technical situation where switch S1 in Figure a is mostly in the turn-on position. Hence, this indicates that a longer time in the turn-on position can severely reduce the conversion efficiency of our system. To prove that ηconv is dependent on the input duty ratio and not on the amount of PV-generated electrical power, we investigated ηconv in terms of deliberately controlled input duty ratio (D) at a fixed AEC/APV (=1) configuration and found that ηconv increased with lower D (Figure e). This further strengthens our claim that a small duty can actually increase the converter efficiency. The ηconv in terms of D at a fixed AEC/APV configuration followed a similar trend as the ηconv in terms of DO in Figure d, indicating that rather than the amount of PV-generated power, the duty ratio value plays a huge role in the ηconv of our system. Therefore, we can hypothesize that as the number of series-connected PV increases, lower optimized duty will be necessary for the MPPT process and also for the generation of high-converter efficiency.

To further understand the converter efficiency and its operation with various duties, we observed the operation of the PV-Conv-EC system by varying (viii) the number of series-connected PVs. The DO for MPPT was completely different from two series-connected PV systems. For instance, by series-connecting three GaAs PVs, much lower DO is required compared with a two series-connected PV configuration (Table ). Interestingly, under PVs with three series-connected configurations, dramatic improvement in the STH efficiency (19.5%) can be achieved when AEC/APV = 1, compared with 12.2% STH efficiency without the converter application (Figure S10c). The ηconv of three series-connected PVs that used the PV-Conv-EC system was higher than that of the two series-connected PV system because of the lower DO required for the MPPT process at AEC/APV = 1 configuration. Therefore, the number of series-connected PVs plays a significant role in determining the ηconv of the converter MPPT process. Moreover, by using boost-type converter, the system also allows MPPT of nonseries-connected PVs to operate water electrolysis, although VOC is much lower than 1.23 V as shown in Table . Despite the low ηconv with non-series-connected PV, hydrogen can be generated where conventional PV–EC system is not available (Figure S10d).

Table 1

PV, STH, Converter Efficiency, and Duty-Controlled Value at Different configurations of the PV-Conv-EC System Under 100 mW Cm–2 Light Irradiancea

configuration	PV efficiency (%)	STH efficiency (%)	DO	ηconv
APV/AEC = 1:1, 2 series-connected PVs	26.5	19.1 [18.4]	0.94	0.928
APV/AEC = 1:3, 2 series-connected PVs	26.5	20.6 [18.4]	0.87	0.959
APV/AEC = 1:1, 3 series-connected PVs	26.3	19.5 [12.1]	0.605	0.955
APV/AEC = 1:1, parallel-connected PVs (boost-type converter)	25.5	17.2 [0]	0.450	0.867

The square bracket in the STH efficiency indicates the STH efficiency without a converter application on the PV–EC system.

In an identical PV and EC area, the PV-Conv-EC system can also be used to various (ix) solar light power densities. By measuring the I–V curve of the PV at various light densities, the current and DT based on the VEC,MPPT/VPV,max ratio can be predicted as shown in Figure a. When the solar power densities are 70 and 30 mW cm–2, the estimated current of the EC at each light density divided solar cell area will be 11.97 (20.9% STH) and 5.09 mA cmPV–2 (20.8% STH). Similar to that shown in Figure b, relatively lower power generated from the PV would guarantee higher STH and lower DT for MPPT (Figure a). By measuring the IEC generated from different light irradiance powers, the STH efficiency can be performed and the duty can be optimized (Figure b). Interestingly, we found that MPPT can occur at each light irradiance power for achieving the highest STH efficiency by the use of pPV,max. Therefore, our system can always use the maximum power of the PV at any time even though it changes because of the light intensity. The STH efficiency at a small light irradiance power was provided at both high STH and converter efficiency. Current densities of 11.29 mA cmPV–2 (19.8% STH) and 4.92 mA cmPV–2 (20.1% STH) can be achieved, respectively, at 70 and 30 mW cm–2 of light irradiance to PVs. The stability of our system was also confirmed under the 30 mW cm–2 light condition (Figure c). These show the possibility that our system can truly achieve the maximum STH even with the varying intensity of solar light through the combination of an automatic perturbation and observation (P&O) algorithm. Moreover, if we can obtain spectra on each solar light intensity, it would be possible to firmly analyze the relationship between the STH efficiency and solar light intensity based on light absorption and conversion prediction of our GaAs PVs as the Deutsch group suggested and the modeling we did on Figure .[34] It will further improve the robustness of our design principle so that our system has more industrial applications.

Figure 4

solar-to-hydrogen energy conversion process of the PV-Conv-EC system at different solar power densities. (a) Prediction of STH and DT based on the I–V curve of PV and EC. (b) STH efficiencies under light-chopped illumination at DO. The STH efficiencies under each condition are converted from each current and solar power density. (c) Current density measured at the EC for 4000 s at 30 mW cm–2 solar power density. The current density was converted into STH with 1.23 V and light power density.

Conclusions

The advantage of this approach is that the decoupled factors in terms of PV and EC variables can be investigated to determine how they can affect the final efficiency of a PV–EC system. Our findings provide new insights into the selection of independent PV and EC compartments for achieving the desired efficiency of a solar-driven hydrogen evolution reaction. The individual compartment-dependency and the interface optimization presented here highlight the selection of an EC catalyst and the optimized configuration that derives the highly efficient photoelectrolysis hydrogen evolution, minimizing the loss. Through the optimization modeling and experiments, the importance of EC performance especially with highly efficient PVs, converter existence, and optimized number of series-connected PVs are indispensable for a high STH-efficiency PV-Conv-EC system. As a result, 20.6% STH efficiency and 78% PV electricity-to-hydrogen conversion efficiency can be achieved. Even if the rare metal electrocatalysts were used for this work, we definitely believe that state-of-the-art earth-abundant catalysts can be also used to our design principle because of their superior characteristics.[27] This PV-Conv-EC system design rule can help advance the commercialization of solar-driven hydrogen fuels for a future clean-energy society.

Methods

Model Study

From the analysis of Figure c, the fixed parameters of eqs and 3 were calculated as follows: τcat = 30 mV dec–1, τano = 60 mV dec–1, I0,cat = 1 mA, and Rsol = 1 Ω cm2. The τcat, τano, and I0,cat values correspond to the Tafel mechanism of hydrogen evolution, the one-proton-one-electron-involved water oxidation mechanism, and the exchange current of platinum for hydrogen evolution, respectively.[35,36] In Figure d, the variation in the τcat + τano and η10mA values were brought about by the anode performance under a fixed cathode performance condition (cathode exchange current = 1 mA and Tafel slope = 30 mV dec–1). The PV used in this figure was with 30% efficiency, as shown in Figure S1. The resistance was also fixed, Rsol = 1 Ω cm2, as in Figure c,d. In Figure S2, the anode performance was fixed as the anode exchange current density = 10–3.5 mA cm–2 and the Tafel slope = 60 mV dec–1. The cathode performance was fixed as cathode exchange current = 1 mA cm–2 and the Tafel slope = 30 mV dec–1. The PV used in this figure was with 30% efficiency, similar to the efficiency outlined in Figure S1. We assume that the temperature was maintained at room temperature in the modeled solar cell or PV-Conv-EC system.

Fabrication of the Solar Cell

The GaAs solar cell was fabricated by the following process. First, the device layer was deposited on the GaAs wafer by metalorganic chemical vapor deposition (MOCVD) in the following order: trimethylgallium (TMGa), arsine (AsH3), trimethylaluminum (TMAl), phosphine (PH3), and trimethylindium (TMIn). Next, by using the e-beam evaporator, the front and back of the gold electrode were deposited. The front layer was additionally electroplated for decreasing the resistance with the optimized grid structure. To control the surface area, the mesa-etching process was conducted. To be specific, the cell with a designated area (1 cm2 in this case) was covered with a photoresist and etched with solution. After the etching is completed, the photoresist was peeled off. After cutting the cell into the desired form, the cell was bonded on a printed circuit board (PCB) substrate, and the front electrode was wired with the substrate. Finally, for the antireflective coating, the ZnS and MgF2 layers were thermally evaporated and deposited on the solar cell with optimized thickness (∼50 and ∼100 nm for ZnS and MgF2, respectively).

Synthesis of Electrode

The metal oxide nanocrystalline was synthesized using the hot injection method. After washing the nanocrystalline with toluene and acetone, 20 mg of the synthesized nanocrystalline was annealed under 250 °C for 1 h for removing organic ligands. The annealed nanocrystalline was dispersed in 50 mL of 1 mg ml–1 K2IrCl6 (STREM, 99%) and heated at 60 °C for 6 h under continuous stirring. The powder was washed with water and ethanol. After washing, annealing was conducted under 250 °C for 1 h. After the annealing procedure, 1.2 mg of IrO powder was dispersed in 120 μL of ethanol and 9 μL of Nafion 117 solution (Aldrich, ∼5 wt % mixture of alcohol and water). A drop of ink was placed on 6 cm2 carbon fiber paper (FuelCellStore, Spectracarb 2050A-0850) and dried under ambient condition for 1 day. The cathode was prepared with 1.2 mg of 20 wt % Pt/C (Alfa) and dispersed in 120 μL of ethanol and 9 μL of Nafion 117 solution. A drop of ink was placed on 6 cm2 carbon fiber paper and dried under ambient air for 1 day.

Converter Designing

To implement the converter, the FR-4 PCB was designed on the breadboard with appropriate devices. The switches of the converter (S1 and S2) were both implemented by MOSFET (IRF 540). The 1 mH ferrite core inductor, SRR 1208-102KL, was used as the inductor. The appropriate duty cycle control was exercised with the digital signal processor TMS320C28346 from Texas Instruments. The switching frequency was 20 kHz.

Material Characterization

The IrO nanocrystalline was analyzed using a transmission electron microscope (JEOL, JEM-2100F). The Pt/C nanocrystalline was measured using a field-emission scanning electron microscope (Carl Zeiss, SIGMA).

MEA Cell Construction

The configuration of the cell is shown in Scheme . The Nafion membrane was washed under boiling 3 wt % H2O2 for 1 h, boiling 1 mol L–1 H2SO4 for 1 h, and boiling deionized water for 1 h before each assembly procedure. The interface between the graphite bipolar plate and the Nafion membrane was joined with a silicon (∼0.2 mm) gasket with both the cathode and anode sites. The cell was sealed with torque of a 75 kgf cm at each screw driver. The carbon fiber paper electrode area was fixed at 6 cm2 for each cathode and anode.

Electrochemistry

Electrochemical measurements were recorded using a potentiostat (CHI 760E, CH Instruments). For three electrochemical measurements, a Pt foil (3 × 3 × 0.01 cm, 99.997% purity, Alfa) was used as the counter electrode, and Ag/AgCl electrode (BASi, 3 M NaCl) was used as a reference electrode. In the case of 3 electrode system, 0.5 cm2 of catalyst electrode performance was measured. The working electrode was prepared by following the procedure similar to the previous drop-cast carbon fiber paper. The scan rate was 10 mV/s. The catalyst property was shown after polarization between the forward and reverse CV scan current and further iR compensation using V = Vapplied – iR.

The electrochemical potentials were converted into the RHE scale by the following equationFor the two-electrode electrochemical measurements, the counter and reference electrodes were connected to the cathode compartment of the MEA cell and the working electrode was connected to the anode compartment. The electrolyte was prepared with 0.1 mol L–1 HClO4 for both the three and two electrochemical measurements. The current detection during the photovoltaic-based electrolysis was measured by the same potentiostat connected to the ammeter. The voltage during photoelectrolysis was measured based on the converter voltage measurement.

Gas Measurement

Gas quantification was performed using gas chromatography (GC). The hydrogen evolution compartment of the MEA cell was connected and hermetically sealed in a glass flask. Before electrolysis, Ar gas was purged for 15 min to eliminate the gas content. After sealing of the gas, bulk electrolysis was started under 1.6 V until a charge was accumulated. For each given charge, 1 mL of gas was collected using a gas-tight syringe and injected into a gas chromatography instrument (NARL8502 model 4003, PerkinElmer). The oxygen evolution was analyzed using a fluorescence-based oxygen probe (NEOFOX-KIT-PROBE, Ocean Optics). A similar procedure for the detection of H2 evolution was performed: the MEA cell anode compartment was tightly linked with a glass flask containing an electrolyte. After purging with Ar for 15 min, bulk electrolysis was conducted under 1.6 V while the oxygen probe detected the oxygen in the glass flask.

Solar Cell IV Curve Measurement

For a single GaAs PV, the current–voltage performance was measured by applying external potential by an I–V test system (K3000, MsScience). The solar simulator (K3000, MsScience, class AAA) was used to irradiate the standard light condition on the GaAs PV and PV-conv-EC system under the AM 1.5 G (100 mW cm–2) condition or a specific solar power density condition. During the PV performance measurement or PV-Conv-EC system operation, the PVs are always cooled with a fan for maintaining consistent room temperature. The PV area was controlled by a combination of 1 cm2 of monolithic GaAs PV. The light intensity variation was measured using an optical power meter (Newport, model 1916-R).