Batteries to Keep Solar‐Driven Water Splitting Running at Night: Performance of a Directly Coupled System

Direct solar‐powered hydrogen generation (so‐called “green” hydrogen) is promising as a renewable fuel that can be generated anywhere there is sunshine and water. Many attempts are made to integrate a water electrolyzer (EC) and solar cell at different levels (a so‐called artificial leaf) to take advantage of the reduced losses from the lack of wiring and optionally increased portability afforded by an integrated unit. However, in many cases, EC catalysts degrade as electrodes depolarize when shut down at night. Much less attention is paid to the need for a minimum current across the EC under insufficient illumination to prevent excessive cyclic degradation. Directly coupling a battery to keep an artificial leaf running at night can address this need and, in theory, also increase solar‐to‐hydrogen (STH) efficiency. A seven‐cell silicon heterojunction module, two bifunctional NiFeMo ECs in series, and a commercial Li‐ion NMC battery are selected to provide the same amount of solar output power despite different working voltages and tested in a series of simulated diurnal cycles. The increased average STH efficiency per cycle (11.4% vs. 10.5% without the battery) is analyzed and discussed with implications for future artificial leaf design and implementation.

DOI: 10.1002/solr.202100916 Direct solar-powered hydrogen generation (so-called "green" hydrogen) is promising as a renewable fuel that can be generated anywhere there is sunshine and water. Many attempts are made to integrate a water electrolyzer (EC) and solar cell at different levels (a so-called artificial leaf ) to take advantage of the reduced losses from the lack of wiring and optionally increased portability afforded by an integrated unit. However, in many cases, EC catalysts degrade as electrodes depolarize when shut down at night. Much less attention is paid to the need for a minimum current across the EC under insufficient illumination to prevent excessive cyclic degradation. Directly coupling a battery to keep an artificial leaf running at night can address this need and, in theory, also increase solar-tohydrogen (STH) efficiency. A seven-cell silicon heterojunction module, two bifunctional NiFeMo ECs in series, and a commercial Li-ion NMC battery are selected to provide the same amount of solar output power despite different working voltages and tested in a series of simulated diurnal cycles. The increased average STH efficiency per cycle (11.4% vs. 10.5% without the battery) is analyzed and discussed with implications for future artificial leaf design and implementation.
of the PV-generated energy over the whole diurnal cycle of operation results in a noticeable reduction of the power delivered to the electrolyzer for the same amount of solar energy collected. This facilitates reduction of the overpotential loss in the electrolyzer and consequently increases the overall STH efficiency despite the voltage loss inevitably associated with the battery chargeÀdischarge cycle.
In this work, we present an experimental proof of concept of our previously described theoretical framework [19] by coupling a solar electrolyzer with a working voltage matched to that of an appropriately sized battery. We compared two situations using exactly the same solar electrolyzer setup, under the same duration of illumination, with and without a battery attached. By working on both sides of the solar module maximum power point (MPP), we also exclude effects associated with different output power to the two systems working at different voltages by ensuring that the coupling factors of both systems' working points are almost identical. Combining a shingled seven-cell silicon heterojunction solar module with a pair of Earth-abundant NiFeMo catalyst [20] electrolyzers in series and, a commercial Li-ion battery, we show that such a system not only works as expected but can be done with relatively inexpensive, accessible, and abundant materials.

The Concept
As detailed by Astakhov et al., [19] if the current from solar illumination can be stored and used later during the night, a solar electrolyzer should see efficiency gains from working at a lower voltage working point. In Figure 1, a battery attached to an artificial leaf will store part of the charge collected in the day (Q B ), to be released to the electrolyzer at night for conversion into hydrogen (Q H2d ) at a lower current rate (I ECd < I Bl ), but over a longer time. By coupling a solar electrolyzer [20] in parallel to a battery (PEB) of sufficient size and voltage, the total current density in the electrolyzer can be reduced during illumination but a higher current is drawn from the PV module. Running the electrolyzer on the battery-stored energy in the night, all charge collected during illumination (Q B ) can be utilized to split water if the battery has no charge losses, and the discharge current, I Bd ,is sufficiently high to release all the charge before the next cycle. [19] The first condition is achieved by very high Faradaic (Coulombic) efficiencies in modern batteries exceeding 99% per cycle. [21,22] This is especially so for low state of charge utilization and relatively low C-rates (battery charge/discharge rates relative to 1 h for full charge) for charge and discharge below 1 C. The second condition is fulfilled when the product I Bd Â t d equals or exceeds Q B after accounting for storage and transmission losses. In the experiment, we control the discharge time to achieve exact charge balance to assess the change in STH efficiency accurately.
In Figure 2a, without the battery, the solar module (PV) only powers the electrolyzer (EC) under illumination (P ECl ) to generate hydrogen (P H2l ). With the battery attached, power diverted from the PV to the battery under illumination (P B ) will be used to power the electrolyzer (EC) in the dark (P ECd ) and result in two-time-resolved streams of hydrogen for the light (P H2l ) and Figure 1. Charge conservation and relative current levels within the PEB test. Ideally charge is conserved by the battery (Area Q B ¼ Area Q H2d ) during the "day" (t l ) and released over a longer time at "night" (t d ) for one diurnal cycle.  The experimental setup detailed earlier enables us to realize the dual modes of operation using a single setup without stopping the experiment to switch between PVÀEC and PEB modes of operation. This enables both respective currentÀvoltage profiles detailed in Figure 3a,b to be realized with one setup. In Figure 3a, showing a PVÀEC system, a hypothetical example EC (black curve) and PV module (red curve) together will result in a working point at the intersection of the two curves (WP PVÀEC ) under illumination. With the onset voltage of hydrogen evolution set at the left vertical axis, the area encompassed by the gray square at the working point represents the total voltagerelated kinetic loss for hydrogen generated at this working point.
Adding an appropriate hypothetical battery (Figure 3b, light green curve) to transform the PVÀEC system in Figure 3a to a PEB system should shift the working point of the PVÀEC system under illumination to a lower-voltage working point determined by where the new load curve (blue curve, EC-B) intersects the PV curve (WP PEB ). The PV current at this lower voltage is always higher than before due to the shape of the PV curve, even if it was working at the PV MPP before. The current, however, is now split between the battery and the EC. The current going into the EC is determined by the vertical intersection of the new working point and the initial EC curve with the battery charging current, making up the difference between these two points. This results in lower kinetic losses illustrated by the blue overlapping shaded regions, which sum up to a smaller area than before. Essentially, adding a battery is equivalent to working above the EC curve without adding more catalyst or increasing catalyst loading with the condition that all the current stored is utilized in the water-splitting reaction later.
In the directly coupled PEB device presented in Figure 3b, the voltage efficiency loss from the battery chargeÀdischarge overpotential (ΔV batt ) is directly offset by the voltage efficiency gain from the electrolyzer. The lower PEB working voltage for the same amount of energy/power input results in an increase in total current, and thus total charge converted to hydrogen, leading to a net gain in hydrogen produced.
The working point during illumination shifts from the black point before attachment (WP PVÀEC ) to the blue point after attachment of the battery (WP PEB ). Working currents with the battery attached are found from the vertical intersection of WP PEB and the EC curve and battery charge curve. In the dark, the working point for the electrolyzer shifts to the intersection between the battery discharge curve (WP PEB D ). Letting the left axis be E 0 , where E 0 is the reversible cell voltage of water splitting at standard conditions (in our case, E 0 ¼ 2 Â 1.23V for a pair of electrolyzers in series), the shaded boxes indicate the kinetic losses associated with each scenario, while the battery chargeÀdischarge overpotential is shown by ΔV batt . Note that the areas of both rectangles in plot b) are generally not indicative for the energy analysis because they represent powers acting for different periods of time.

Results and Discussion
The PV component of the system is represented by a seven-cell silicon heterojunction module (SHJ) minimodule with an opencircuit voltage of 5.05 V, short-circuit current of 114.9 mA, and 19.9% efficiency. The MPP is located at MPP voltage of 4.1 V and MPP current of 108 mA. CurrentÀvoltage characteristics of the PV module are presented by the blue curve ( Figure 4). The EC polarization curve is plotted in gray and mirrored about the x-axis for clarity with the red line (EC dark) representing the same curve operating on battery power in the dark.
Connecting only the EC during illumination (PVÀEC mode), an average current of 96 mA was measured, passing through the EC at 4.23 V (gray square, PVÀEC EC ), while the PV working  The PVÀEC test ran at a higher voltage than that of the PEB test, but the total current from the PEB test is higher. The effective working point of the PEB system exceeds the possible electrolyzer (EC) currentÀvoltage curve during illumination, with the total current split between the battery and the EC in an %45:55 ratio.
In the dark, for the PEB mode, the battery discharges through the EC, resulting in the working points of the battery (PEB Bd ) and electrolyzer (PEB ECd ) overlapping. The EC operates at a significantly lower voltage when powered by the battery due to the battery's inherent overpotential loss. Although there is a loss in energy from voltage overpotential loss from the battery, this is exactly compensated by the reduced overpotential loss in the EC cell, and total overpotential losses during light and dark periods are the same in the PEB system, as we described in detail in our previous work. [19] This can be understood simply by considering the initial PV input potential during illumination, and the final hydrogen generation potentials are unchanged and are both independent of the intermediate storage potential.
Using the same solar module and the same electrolyzer under the same conditions with and without a battery attached is meaningless without considering the working point and the corresponding PV coupling factor. Different working voltages will cause the PV power output to change. Coupling factor can be thought of as the proportion of maximum power output of the PV element that is utilized at a constant working voltage and will determine the total input power into the EC and/or battery during one diurnal cycle. Coupling factor is related to the use of the total proportion of the maximum power of the PV input. Beyond the direct effect on total input power supplied to the electrolyzer and/or battery in the particular illuminated period, coupling has no direct effect on the operation of the EC cell in the night, while being under battery power. Coupling factor can be omitted at night, as the PV input is zero. Thus, it is important to keep a constant coupling factor, C, between both tests defined by the following relation.
where P pv is the power output at the working point of the PV module and P mpp is the MPP power output. Any value below 1 means that the power generation potential of the PV device is not fully utilized.
To determine the impact on the addition of the battery to the STH efficiency, the amount of energy delivered by the PV module should be the same in both scenarios. This is done by ensuring that the illumination time and the coupling factor of both modes are as close to identical as experimentally possible.
In Figure 5, we can see that the coupling factors, C, of the PVÀEC test and the PEB tests were almost identical at 0.925 and 0.932 on average. This results in a PV power output that is almost constant across both tests during illumination despite the PVÀEC working at 4.4 V and the PEB working at %4 V. The splitting of current between battery and solar cell also results    www.advancedsciencenews.com www.solar-rrl.com in less-noisy power output at the EC cell likely due to lower bubblerelated loss, [23] due to a lower current density across the EC cell in both the light and dark parts of the test. Software charge counting was used to automatically determine the time required to discharge the battery back to its original state of charge to complete a single diurnal cycle with 100% charge utilization. Note that the variation of power for the PVÀEC system is from 0 to 400 mW, while the power fluctuates between 100 and 220 mW for the PEB system.

Comparing PVÀEC and PEB Systems' Energy Output
By plotting the energy input and output from the electrolyzer of both systems to the voltage, the difference of the two tests' run time can be effectively reconciled. In Figure 6, we plot the solar energy input as E pv and hydrogen energy output from the electrolyzer as E H2 PV. These represent the boundaries of maximum energy input and output of the system in one diurnal cycle.
Here we can see clearly that the voltage of the PEB system is always lower than that of the PVÀEC system, thus having a higher electrolyzer voltage efficiency [24][25][26] throughout 58.1% compared with 62-65.1% (voltage efficiency ¼ E 0 =V EC Â 100%). Furthermore, we can see that the voltage drop of the battery from charge to discharge is inconsequential as the electrolyzer runs for a longer time at a lower potential. Figure 6 also shows that the total amount of energy input into the electrolyzer for both cases is almost identical with a difference of 0.6%. The PEB has a slightly lower energy value (E PEB < E PV-EC ) due to the addition of the battery and additional wiring adding some losses to the system. However, looking at the amount of H 2 produced, there is a total increase in 8.4% (E H2 PEB > E H2 PV-EC ). Due to the lowered working PV voltage for the PEB system, which is more or less pinned to the charging voltage of the battery, the nearly identical energy supplied by the PV module results in a higher amount of charge delivered to the electrolyzer. Furthermore, the lower current density across the electrolyzer electrodes leads to more efficient hydrogen production. Attaching a battery is thus akin to having a larger-or higher-efficiency electrolyzer. In the PEB tests, 53.8% of hydrogen was produced under illumination and 46.2% was produced under battery power.
As the x-axis in Figure 6 starts at E 0 , the voltage between the y-axis and the two vertical lines in the PEB curve represents the overpotential losses of the EC for each mode of operation. As the battery and electrolyzer are in parallel, the operating voltages also reflect the chargeÀdischarge voltage of the battery and thus the difference between them also represents the battery overpotential loss. Note that the electrolyzer overpotential in the dark is lower than under illumination by the battery overpotential loss.

STH Efficiency
STH conversion efficiency (η STH ) is commonly used to quantify the performance of solar electrolyzers under standard conditions of AM1.5G irradiance (100 mW cm À2 ) and no applied bias. [27] It is commonly defined as the ratio of the amount of chemical energy produced and the amount of solar energy incident on the solar cell/module.
where η F is the Faradaic efficiency of the electrolyzer,E 0 is the reversible cell voltage of water splitting at standard conditions, I PV is the current of the solar cell/module and of area A PV , under incident solar power density P in . This can also be seen simply as a product of the efficiencies of both the electrolyzer and the PV module and PV coupling factor.
However, these two equations are not directly applicable to the PEB system as the total time the EC is run and the total time under illumination has been decoupled. Also, as the tests are run back to back with no break, it was also not possible to accurately split the streams of gas for direct comparison; thus, an assumption of perfect Faradaic efficiency of the catalyst (η F ¼ 100%) is used to simplify the analysis. This assumption helps to isolate gains and losses to the change in the system from specifically adding the battery. Nonideal catalyst Faradaic efficiency, as long as it is more or less constant between the two working modes, will not meaningfully affect the analysis, as they can be considered constant factors. Instead, comparing total chemical energy evolved and light energy   Figure 6. Solar energy input and solar-generated hydrogen curves plotted against voltage for one representative diurnal cycle. Overall solar energy collected by the PEB system is lower than the PVÀEC system but results in more H 2 produced. Lines plotted below the final amounts of energy collected (stars) trace out the energy collected over time to distinguish the energy produced during the dark (E H2d ) and light (E H2l ) working stages of the PEB test and their respective working voltages. Note that the x-axis begins from double the thermodynamic potential of water splitting at standard conditions (E 0 ). www.advancedsciencenews.com www.solar-rrl.com captured in both PVÀEC and PEB is more instructive.
where E H2 light and E H2 dark refer to the amount of hydrogen energy produced. E total is the total energy incident on the solar module of area A PV , with incident solar power density P in for the test duration under AM 1.5G illumination solar irradiation. Total illumination time of the respective systems is t l and total experimental time is t.
Under certain circumstances, simplified analysis gives acceptable results which can be used for a quick check of the system at the design stage to save time and effort. For the analysis of rectangular irradiance profiles, battery Coulombic/Faradaic efficiency of 100% are assumed. Equation (5) also can be simplified for the case of the rectangular irradiance profile to a simple sum of currents to the electrolyzer and battery during the illumination phase. [19] η Ã STH PEB ð%Þ ¼ where I B light is the current into the battery during illumination and I ECBl is the total current input to the battery and electrolyzer under illumination (or in lossless circuits, the PV current, I PV may be used). From Equation (4) and (5), we obtain the STH efficiencies of both systems, showing an absolute increase of 0.88% with the addition of the battery. We can see that η* STH PEB (%) and η STH PEB (%) differ by only 0.06%, and thus this proves that the method proposed in our earlier work [19] is sound and the simplified lossless analysis can be used as a good approximation of the total efficiency of a PEB system for the fast optimization or comparison of different PEB systems. Equation (6) is much more convenient as instantaneous current and power values during illumination can be used without running the whole system for a full diurnal cycle or can even be estimated at the design phase from the PV curve without running the experiment.

Energy and Loss Analysis
By plotting the maximum efficiencies at each device following the schematic in Figure 6, we can estimate the losses associated with each stage of energy conversion. In Figure 7, this is reflected in the maximum values at the top of each colored bar, with values obtained without the battery on the left (PVÀEC) and with the battery attached on the right (PEB). There is a significant but almost equal loss in energy efficiency from utilizing the PV module away from its MPP (19.9%) to their respective working points of 18.3% and 18.5% as per experimental design, represented by the coupling loss. Next is the red portion, which is the delta in efficiency from the PV output and EC input voltage and currents. This is generalized as being resistive losses. Because the power through the batteries is also an electrolyzer input, any losses from the batteries manifest as a larger "resistive loss." Note that in this analysis Faradaic/Coulombic losses from the battery are excluded due to charge counting, and voltage losses from the battery are compensated directly by the electrolyzer upon production of hydrogen. The low losses from battery storage are rather expected, given that the size of the battery (160 mAh) and total utilization of the battery during charge was %0.392 mAh or less than 1% maximum capacity at a charge rate of 47 mA (0.294 C-rate) and discharge rate of 33 mA (0.2 C-rate).
The blue portion of the graph details the power entering the electrolyzer, while the green portion is the final STH efficiency of both systems, with the ratio between them giving the efficiency of the electrolyzer and the loss given by the delta between the two thresholds. Note that in this analysis the total efficiency of the PEB is an average of the voltage efficiency of the light and dark times. From here it is clear that the addition of the battery causes η STH to increase via the increased efficiency of the electrolyzer. Using the battery to store charge, the positive relationship between voltage and efficiency of an electrolyzer is effectively decoupled.

Conclusion
In our work, a battery has been added in parallel to an artificial leaf (PVÀEC system) without power electronics for continued operation in the dark. The main function of the battery is to ensure electrolyzer current throughout the diurnal operation cycle to prevent depolarization of the electrolyzer when illumination is insufficient. In a broad sense, it is also a logical extension of the PVÀEC system to realize a sustainable combination of PV generation with short-and long-term storage. We demonstrate that a simple parallel connection of a battery system is sufficient for the self-sustained operation of the artificial leaf without power conversion or matching electronics. The battery voltage effectively sets the working voltage of the artificial leaf system when illuminated and the electrolyzer in the dark. The use of a commercial battery and Earth-abundant watersplitting catalysts shows that such a system need not be bespoke, can easily be implemented, and, possibly scaled quickly. Furthermore, the effect demonstrated is not specific to any PV, EC, or battery technology and can be observed as long as the principles of power and voltage matching, as described in the article, are followed. This result holds promise for simple, cost-effective, yet, highly efficient solar water-splitting systems.
As it has been predicted in our earlier work, [19] the battery can facilitate improvement of STH efficiency in addition to the stabilization function. This gain in STH efficiency is confirmed experimentally with high-efficiency components, appropriate voltage matching, and appropriate power matching of the components. The distribution of the PV current over the diurnal cycle with aid of the battery leads to the increase in STH efficiency from 10.5% to 11.4%. This is a result of reduced electrolyzer current density and related increase in electrolyzer voltage efficiency.
Under periodic irradiance, such as irradiance profiles closer to those found in the field, the artificial leaf system with storage (PEB system) has the potential to outperform an equivalent system without storage in terms of STH efficiency even though adding a battery brings an inevitable potential loss. The lowered electrolyzer current density and persistent electrolyzer operation of a PEB system also have the potential to extend electrolyzer lifetimes and lower the electrolyzer geometric factor when compared with a similar artificial leaf system. Though the gain in efficiency is moderate, it may be relevant when storage is already planned to be part of the system, as an alternative mode of operation. Future work may involve quantifying the gain in the electrolyzer lifetime from this mode of operation and possibly with more economically robust battery technology or simulate the use of grid-linked storage to make such operating modes relevant to mass market hydrogen production.

Experimental Section
A seven-cell shingled minimodule for the experiment was prepared using laser-cut silicon heterojunction solar cells. For the full details of the preparation, see the study by Chime et al. [28] The minimodule with area of 22.4 cm 2 was coupled with a pair of electrolyzers in series (1.5 cm 2 NiFeMo symmetric bifunctional electrodes on 1 cm-wide Ni foam strips in 1 M KOH, with onset voltage of 1.5 V each) and, a 160 mAh commercial NMC LiÀPo pouch cell battery with a voltage range of 3.7-4.2 V. The pair of electrolyzers was connected in series and treated as a single unit ("the EC") with an onset voltage of %3 V.