Unlocking Ultra‐High Performance in Immersed Solar Water Splitting with Optimised Energetics

This research introduces a pioneering approach to solar water splitting technology, utilizing an innovative, highly efficient immersed system. The system incorporates a flexible array of electrochemical and photoelectrochemical cells, powered by high‐performance III‐V triple‐junction cells. Remarkably, this method significantly boosts the solar‐to‐hydrogen (STH) conversion efficiency, reaching a record 20.7% under 1 sun illumination, employing earth‐abundant catalysts operating at ambient temperature. These findings highlight extensive scope for further optimization, including minimizing optical transmission losses, mitigating shading effects, and reducing the overpotential of the electrochemical cells, thereby augmenting the STH efficiency to an estimated 28%. Through a comprehensive techno‐economic analysis, a levelized cost of hydrogen (LCOH) of 8.3 USD kg−1 is estimated, forecasting the potential for a reduction to a competitive 1.8 USD kg−1 with improved efficiency, increased capacity factors, and decreased production costs. A sensitivity analysis emphasizes the significant influence of factors such as III‐V cell cost, electrolyzer membrane cost and capacity factor on the LCOH. Overall, this study signifies crucial progress toward a highly efficient and economically viable solar water splitting solution, promising a sustainable route for hydrogen production.


Introduction
3][4] By stacking photoabsorbers with increasingly wider band gaps on top of each other, far more photons can be harvested, particularly at longer wavelengths.This arrangement also reduces thermalization losses.The photovoltage of a tandem cell is equal to the sum of its subcell photovoltages, hence the photovoltage increases as more subcells are added.However, since subcells are current-matched, the photocurrent density decreases as the solar spectrum is divided into increasingly smaller segments.
A typical electrochemical cell operating at room temperature requires around 1.6 V for water splitting, for which a double-junction cell is sufficient.Additional subcells simply supply excess photovoltage while lowering the photocurrent density.As such, double-junction solar water splitting devices have received much attention.[7] Thus far however, the STH efficiency has been limited by the bottom subcell.A record STH efficiency of 19.3% was achieved by a monolithic InGaP/InGaAs photoelectrochemical (PEC) device, equating to around 85% of its maximum theoretical STH efficiency. [1]While undeniably impressive, the maximum theoretical STH efficiency of this device is substantially less than 30% because it is difficult to tune the band gap of InGaAs to less than 1.2 eV due to increasing lattice mismatch with the GaAs substrate. [2,8]Si is inexpensive and has a band gap of 1.12 eV, [9] making it an attractive narrow-gap material, however, high-quality growth of III-V materials on Si remains challenging. [10,11]Alternatively, perovskite/Si systems are promising, with one such photovoltaic-electrochemical (PV-EC) system recently attaining an STH efficiency of 20%, [12] although the instability of perovskites continues to be problematic. [13][16] Triple-junction cells are of particular interest as they are a mature technology and remain among the most efficient multi-junction devices developed to date. [17]Because of this, their application to solar water splitting has gained attention in recent years, primarily in the form of PV-EC water splitting.
Since triple-junction cells still provide a much greater photovoltage than is necessary for water splitting, research has focused on developing systems that utilize the excess photovoltage.PV-EC water splitting has proven to be an ideal starting point as it offers the flexibility to connect multiple cells and electrolysers in series.By adjusting the ratio of cells to electrolysers, the combined photovoltage from the cells can be tailored to match the combined potential required by the electrolysers, thereby utilising the excess photovoltage.20] The success of multi-junction cells is largely due to the versatility of high-performance III-V materials.However, the associated material and manufacturing costs remain very high.In fact, it is likely that the widespread commercialisation of III-V-based solar water splitting devices will not be viable without the use of concentrator lenses, which substantially reduce material usage. [21,22]uch devices must therefore be able to operate stably under concentrated illumination, for which thermal integration is critical.The role of thermal integration is twofold; without it, high STH efficiencies cannot be achieved at high current densities [23] and the degradation of device components such as photoabsorbers, catalysts and membranes is greatly accelerated. [24,25]Immersed solar water splitting devices have a distinct advantage in this regard, since the electrolyte itself can be used as a heat transfer fluid.Immersed triple-junction PEC devices have been reported by Hannappel et al. [4] and Chorkendorff et al., [26] however these devices are monolithic and therefore cannot be connected in series, making it very difficult to utilise the excess photovoltage.This work introduces a novel approach that builds upon the concept pioneered by series-connected PV-EC systems.However, instead of using a bank of PV cells to power a bank of electrolysers, series-connected (photo)-electrochemical cells are employed in which all light-harvesting components are immersed.The light-harvesting components in question are fully decoupled triple-junction photoanodes fabricated by combining triplejunction cells with earth-abundant cocatalyst foils. [27]Each photoanode is more than capable of powering an electrochemical cell.By connecting multiple photoanodes and electrochemical cells in series, the surplus photovoltage from the photoanodes can be harnessed to power an additional electrochemical cell.The adjustable ratio of photoanodes to electrochemical cells allows the device to operate as close as possible to its maximum power point.With a ratio of three photoanodes to four electrochemical cells, the STH efficiency was raised from 16.2% to 20.7% under 1 sun, demonstrating the ability to utilise the excess power generated by immersed multi-junction devices.Additionally, theoretical calculations have been carried out to predict potential efficiency enhancements, while a detailed techno-economic analysis has been performed to assess the cost-effectiveness of the system.This innovative approach for constructing highly efficient solar water splitting systems using multi-junction photoelectrodes could potentially expedite the commercial application of water splitting technology.

Structure and Composition of Earth-Abundant Cocatalysts
Earth-abundant NiMo x and NiFe(OH) x were selected as cocatalysts for the hydrogen and oxygen evolution reactions (HER and OER), respectively.NiMo x was deposited hydrothermally on Ni foam, while NiFe(OH) x was cathodically electrodeposited on Ni foil to form a cocatalyst foil. [27]The morphology, structure, and chemical composition of the earth-abundant cocatalysts were analyzed using scanning electron microscopy (SEM), high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), energy-dispersive X-ray spectroscopy (EDX), X-ray diffraction (XRD), and X-ray photoelectron spectroscopy (XPS).Figure 1a,b shows that NiMo x forms a porous nanoflake array on top of a much denser bottom layer.Both layers are ≈1 μm thick.Additionally, small bright spots are visible on the nanoflakes in Figure 1c, which are likely metallic Ni crystallites.These observations are consistent with previous samples of NiMo x synthesized on Ni foil. [27]n contrast, NiFe(OH) x /Ni foil was synthesized via a slightly modified method from that reported previously; [27] a chloridebased precursor solution was used instead of nitrate, while the cathodic electrodeposition was performed at a more negative potential.NiFe(OH) x was previously observed to form a very thin porous layer 50-100 nm thick, composed of small needle-shaped structures.However, the new deposition method produces a vastly different morphology.The NiFe(OH) x is textured but not overly porous, as shown by the SEM images in Figure 1d-f.It can be observed that the NiFe(OH) x is quite granular, consisting of particles 100-200 nm across.Although the particles are fairly compacted, there are cracks in the material that help to expand the surface area.The STEM-EDX maps in Figure S1 (Supporting Information) show that the particles form a layer at least 2 μm thick, with a reasonably uniform distribution of Ni, Fe, and O, while the XRD pattern of NiFe(OH) x in Figure S2 (Supporting Information) displays only the same three peaks as that of the underlying Ni foil, indicating that the NiFe(OH) x is amorphous.
The XPS spectra of NiFe(OH) x in Figure S3 (Supporting Information) show that Ni 2+ and Fe 3+ are the dominant species, with Ni 3+ also present in sizeable quantities. [28]The Fe 2p spectrum in Figure S3b (Supporting Information) indicates that Fe is present in at least two phases, namely Fe 2 O 3 and FeOOH, while the O 1s spectrum in Figure S3c (Supporting Information) suggests a prevalence of oxyhydroxides, [29] which are catalytically active toward the OER. [30]Interestingly, there is no longer an abundance of hydroxide surface species, as was previously observed. [27]All assigned peaks are listed in Table S1 (Supporting Information).

Electrochemical Properties of Earth-Abundant Cocatalysts
The electrochemical properties of NiMo x /Ni foam and NiFe(OH) x /Ni foil were analyzed by measuring their I-V and J-V characteristics, respectively, in 1 m NaOH electrolyte in a three-electrode electrochemical cell, as shown in Figure 2. As previously observed, [27] both cocatalysts undergo redox transformations.In the case of NiMo x /Ni foam, the transformation at 0.1 V versus RHE is semi-reversible and can be attributed to the reduction of oxidized Ni and Mo species. [31]NiFe(OH) x /Ni foil, on the other hand, contains a large quantity of Ni 2+ , as confirmed by XPS, hence the cyclic voltammogram (CV) in Figure 2b  exhibits a redox couple at ≈1.45 V versus RHE corresponding to the reversible oxidation of Ni 2+ to Ni 3+ . [32]ecause of the redox transformations experienced by the cocatalysts, the reverse CV scans are used to determine the kinetic overpotentials.NiMo x /Ni foam (≈1.5 cm 2 ) has a HER overpotential of just 48 mV at 15 mA, in good agreement with that reported previously. [27]NiFe(OH) x /Ni foil has an OER overpotential of 302 mV at 10 mA cm −2 .This is a slight improvement on that reported previously, [27] possibly because there are far fewer hydroxide surface species.Electrochemical impedance spectroscopy (EIS) shows that the charge transfer resistance at the NiFe(OH) x surface is very low; the Nyquist plots in Figure 2c display small arcs, with arc diameter corresponding to charge transfer resistance.At 1.6 V versus RHE, the charge transfer resistance is less than 1 Ω.However, this is still slightly more resistive than previously observed, likely due to the increased thickness and reduced porosity of the NiFe(OH) x .While stability measurements were not performed, it has been shown in the preceding literature that both cocatalysts remain stable indefinitely during electrolysis. [27]

Device Performance and Characterization
Triple-junction cells are some of the most important lightharvesting devices developed to date.Although highly efficient double-junction cells do exist, [8,14] the addition of a third subcell greatly extends absorption into the IR.Triple-junction cells are also far more mature than many other multi-junction devices, being the first to achieve a PV efficiency of over 40% under concentrated illumination in 2007. [17]This has largely been facilitated by Ge (E g = 0.67 eV), which not only has a narrow band gap ideal for harvesting longer wavelengths, but also has a lattice constant fortuitously close to that of GaAs and other III-V alloys, [9] enabling the epitaxial growth of III-V subcells on Ge substrates.Additionally, the growth and optimization of triple-junction cells remains far less complex than for devices with four or more subcells.In-GaP/(In)GaAs/Ge cells are now commercially available and regularly attain PV efficiencies exceeding 30% under 1 sun. [17,33,34]ore recently, inverted metamorphic architectures have enabled InGaP/GaAs/InGaAs cells grown on GaAs substrates to achieve record PV efficiencies of 37.9% under 1 sun and 44.4% under concentrated illumination. [14,35]In this work, InGaP/InGaAs/Ge cells are used, for which a schematic and PV response are shown in Figure 3a,b, respectively.

Triple-Junction Photoanode
A triple-junction photoanode was constructed by combining a triple-junction cell with NiFe(OH) x /Ni foil, adopting a decoupled buried-junction structure.Ag paint and Ag bars were used to make electrical connections while glass, Teflon tape, and epoxy were used to encapsulate the device.The triple-junction cell provides a photovoltage that is easily sufficient for water splitting.The long-term stability of the photoanode was therefore tested, as shown in Figure 3c.A two-electrode electrochemical cell was used, with a Pt plate cathode.The photoanode maintains a nearconstant photocurrent density of 13 mA cm −2 unassisted for almost 9 days, while its J-V characteristics in Figure 3d remain virtually unchanged, demonstrating excellent stability.There is a slight drop in the photocurrent which can be at least partially attributed to a gradual decrease in the lamp intensity.The lamp intensity was reset after around 3 days, restoring the photocurrent to its original value.This feature is visible in Figure 3c.

Triple-Junction Solar Water Splitting Device
Having demonstrated the excellent performance of a single photoanode, a complete triple-junction solar water splitting device was then constructed.The triple-junction device, illustrated in Figure 4a and pictured in Figure S4 (Supporting Information), features four electrochemical cells, three of which combine triplejunction cells with NiFe(OH) x /Ni foil to create photoanodes.As before, Ag paint and Ag bars were used to make electrical connections while glass, Teflon tape, and epoxy were used to encapsulate the device.Connections were made such that components could be measured individually, as well as in series.NiMo x /Ni foam cathodes were simply clipped to the Ag bars (not pictured).
The cathodes (≈1.5 cm 2 ) were substantially larger than the photoanodes (0.25 cm 2 ) so as not to be a limiting factor in the device performance.
The J-V characteristics of the triple-junction device and its components are shown in Figure 4b,c.The PV response of a single photoanode was measured first, yielding a photocurrent density of 13 mA cm −2 , a V OC of 2.35 V and a V MP of 2.08 V, as shown in Figure 4b.All three quantities are slightly lower than those in Figure 3b because the encapsulating glass has an optical transmittance of 91%, while the active area has been reduced from 1 to 0.25 cm 2 , resulting in significant shading which lowers the photovoltage by ≈6% under 1 sun. [36]The electrical connections are also likely to introduce some resistance.The J-V characteristics of a single electrochemical cell were then measured, revealing that an applied potential of 1.55 V is required to operate at 13 mA cm −2 , normalized with respect to the (photo)anode.A single photoanode can therefore easily drive a single electrochemical cell, however a further 0.6 V goes unused, wasting power and limiting the STH efficiency.
To utilize this excess photovoltage, photoanodes and electrochemical cells can be connected in series until the combined excess photovoltage from the photoanodes is sufficient to drive an additional electrochemical cell.The photovoltage of the photoanodes increases linearly when connected in series, as does the potential required to drive the electrochemical cells.Figure 4b shows that the excess photovoltage from two photoanodes is capable of driving a third electrochemical cell, albeit at less than half the maximum current density.However, three photoanodes are ideal for driving four electrochemical cells; with an excess of 0.6 V per photoanode, a combined excess of up to 1.8 V is sufficient for driving the fourth electrochemical cell.The point of intersection of the two relevant J-V curves is very close to the maximum power point of the three series-connected photoanodes, indicating minimal power wastage.
Different ratios of photoanodes to electrochemical cells were then measured together as a complete system, the J-V characteristics of which are shown in Figure 4c.As before, each photoanode generates a photocurrent density of 13 mA cm −2 .With one photoanode driving one electrochemical cell (a 1:1 configuration), the triple-junction device can easily operate without an external bias.However, there is once again an excess photovoltage of 0.6 V.With two photoanodes driving three electrochemical cells (a 2:3 configuration), the triple-junction device operates poorly at just half the maximum photocurrent density.However, it is confirmed that a 3:4 configuration is ideal, operating very close to its maximum power point with minimal excess photovoltage.These results agree very well with those in Figure 4b.
Ordinarily, the STH efficiency can be calculated using Equation 1, where J is the unbiased photocurrent density,  F is the Faradaic efficiency and P in is the illumination power per unit area (100 mW cm −2 for 1 sun).
However, with different ratios of light-harvesting components N PV to electrochemical cells N EC , Equation 1 must be adjusted accordingly, provided that J is the photocurrent density through each individual light-harvesting component.
This slightly altered approach can also be applied to devices that consist of a single electrochemical cell driven by multiple light-harvesting components.Figure 4c shows that the 1:1 configuration generates a photocurrent density of 13.15 mA cm −2 at 0 V, equating to an STH efficiency of 16.2%.The 2:3 configuration only manages a photocurrent density of 6.86 mA cm −2 at 0 V, but with a higher ratio of electrochemical cells to photoanodes, this still equates to a decent STH efficiency of 12.7%.The 3:4 configuration performs exceptionally well, generating a photocurrent density of 12.63 mA cm −2 at 0 V.This equates to an outstanding STH efficiency of 20.7%.The short-term stability of the triple-junction device was also tested, as shown in Figure 4d.With such small compartments, evaporation of the electrolyte is an issue, hence the numerous sharp drops in the measured photocurrent.Nevertheless, the device maintained an STH efficiency of over 20% for 40 h.The impact of evaporation can be minimized in the future by enclosing the compartments, increasing the depth of the compartments and, in more practical systems, collecting the water vapor and recirculating it back into the electrolyte.

Computational Analysis
While the STH efficiency achieved is well below that of record PV-EC systems, [19] it is crucial to emphasize that this system represents a notable milestone for immersed devices tested under 1 sun conditions, as summarised by the comparison with previously reported systems in Figure 4e.Furthermore, there exists significant potential for further improvements to bridge the gap and enhance the overall efficiency.
To assess the potential enhancements in STH efficiency, theoretical modeling based on previously established methods [37,38] was employed to evaluate the current system.The triple-junction cell was modeled under 1 sun AM1.5G illumination using the transcendental solar cell equation, with realistic losses introduced via loss parameters fitted to experimental data.Catalyst performance was modeled using Nernst and Butler-Volmer equations, with additional resistive losses fitted to experimental data.Figure 5a shows the sequential improvements in the STH efficiency of the 1:1, 2:3, and 3:4 configurations.
First, by minimizing optical transmission losses, the photocurrent density can be increased by up to 2 mA cm −2 .This would immediately raise the STH efficiency to over 15% and 24% for the 2:3 and 3:4 configurations, respectively.Achieving this could be facilitated by using quartz glass and/or by depositing an AR coating.
Second, the shading effect can be effectively addressed.It is conceivable to reduce shading from 75% to less than 40% through device design and fabrication improvements and/or the use of larger triple-junction cells.The resulting reduction in photovoltage would be no more than 2% under 1 sun, [36] saving 100 mV per photoanode.While this would only result in minor improvements for the 3:4 configuration, it would significantly enhance the STH efficiency of the 2:3 configuration to over 21%.The improved photovoltage shifts the operating point of the system closer to the maximum power point of the triple-junction cell, thereby considerably increasing the operating current density and subsequently the STH efficiency.Although not modelled in this study, the photovoltage could theoretically be increased further with minimal impact on the photocurrent by using In-GaP/GaAs/InGaAs cells, since the subcell band gaps are slightly wider than those of InGaP/(In)GaAs/Ge cells, particularly for the bottom subcell. [35,39]However, such cells are still under development.
Third, the overpotential of the NiFe(OH) x cocatalyst is ≈300 mV at 10 mA cm −2 , while similar materials have been reported with overpotentials of just 200-250 mV. [30,40,41]It should be noted that catalysts with much lower overpotentials, such as those with finely controlled nanostructures [41] or added metal bases, [40] tend to demand more complex synthesis methods.As such, ease of synthesis will be an important consideration when scaling up in the future.Nevertheless, this presents an opportunity to save a further 50 mV or more per electrochemical cell.With the combination of these improvements, it becomes feasible for the 2:3 configuration to operate close to its maximum power point under 1 sun, approaching an STH efficiency of 28%.
Moreover, operating under concentrated illumination offers additional advantages.Photovoltage increases with illumination intensity, [42] while higher temperatures reduce kinetic overpotentials. [43]Since the electrolyte can be employed as a heat transfer fluid, immersed devices are well suited to operate under concentrated illumination, ensuring that all components are maintained at optimal temperatures.This will be particularly important when incorporating membranes, which are typically ion-conducting polymers that are far more susceptible to heat-induced degradation than inorganic materials and components. [44]Membranes will almost certainly be necessary in future designs for H 2 and O 2 gas separation.With regards to scale, there is little evidence to suggest that a solar water splitting system of this nature cannot be scaled up, given that individual components (photoabsorbers, lenses, catalysts, membranes, etc.) have reached commercial maturity.This will not be trivial however, as there are likely to be new engineering challenges associ-ated with current extraction, gas collection and thermal management.

Techno-Economic Analysis
To assess the commercial potential of this type of system, the levelized cost of hydrogen (LCOH) is calculated.The LCOH represents the average net price per unit of hydrogen generated that would be necessary to cover the expenses of installing, operating, and maintaining a hydrogen plant.The LCOH focuses only on hydrogen production (often referred to as the price "at the gate") and does not include storage or transportation costs.For this analysis, a plant cost model based on data from literature and suppliers is employed, along with discussion with industry experts.For comparison, the system is designed to produce 10 tons of H 2 per day over a 20-year lifetime (with construction occurring during year 0), assuming hydrogen production exclusively during peak sun hours without grid back-up.The parameters used for the analysis are summarised in Table 1 and detailed in the Supporting Information.All costs are based on the 2022 US financial year unless otherwise specified.This analysis does not consider any hydrogen, carbon, or oxygen credits.
With the parameters detailed below and in the Supporting Information, an LCOH of 8.3 USD kg −1 is calculated.This cost estimation is based on a PEC:EC ratio of 2:3, where PEC cells are defined as a light-harvesting subset of EC cells.Considering the large number of modules involved at the plant scale, this ratio could be more finely tuned to maximize the STH efficiency while minimizing the cost.Figure 5b details the cost breakdown for both the capital and operating expenditures (CAPEX and OPEX).As expected, the III-V element cost dominates the CAPEX with a 41% share, using a price of 468 USD 2022 per m 2 (393 USD 2017 per m 2 ) for the PV panels, based on the estimation by Grimm et al. [45] The installation and contingency costs follow with equal 14% shares.The balance-of-plant (BOP) costs cover the electrolyte and gas processing (3% and 7% of the total CAPEX, respectively) and the mounting structure (2%).Owing to the use of earth-abundant materials, the electrolyzer stack accounts for just 10% of the CAPEX, with 8% pertaining to the membrane only.The OPEX amounts yearly to only a very small fraction of the CAPEX (3%) and is dominated by administration and insurance costs (35%), staffing costs (17%) and replacement costs for the electrolyzer stack (22%).The maintenance for the PEC system (electronic components, wiring, compressor) and the piping each amount to 12% of the annual OPEX.
To determine the impact of individual parameters on the LCOH, a sensitivity analysis is conducted, presented in Figure 5c.Considering the significant impact of the III-V cell cost on the CAPEX, its value is varied from 5 down to 0.78 USD W −1 as used by Khan et al., [46] in agreement with NREL predictions. [47,48]his reduces the LCOH from 19.9 to 6.6 USD kg −1 .Another well-known significant expenditure for green hydrogen production systems is the electrolyzer membrane.A base value of 59.5 USD per m 2 is used for the membrane. [45]Nevertheless, several membrane-less electrolysis systems are under development, a number of which have already been commercialized.Eliminating the need for a membrane would lower the LCOH from 8.3 to 7.2 USD kg −1 .It is worth noting that the analysis only considers 6.1 peak sun hours (location: Alice Springs, based on typical meteorological year data from the Australian Bureau of Meteorology, available at www.marcoernst.net).Increasing the number of peak sun hours (hence the capacity factor of the system) from 6.1 to 9.1 h, by changing the location [46] or adding a tracking system, would reduce the LCOH from 8.3 to 5.7 USD kg −1 .Finally, increasing the STH efficiency of the system beyond 27.3% would also have a sizeable impact on the LCOH.This could be achieved by improvements in cell structure or manufacturing techniques, or by the use of concentrators (for which Khan et al. make a strong case). [46]Ramping up the STH efficiency to 45% would decrease the LCOH from 8.3 to 5.2 USD kg −1 .Using concentrators would also increase the number of peak sun hours as the direct normal irradiance is greater than the global horizontal irradiance for the location selected (Alice Springs), which would further increase hydrogen production and hence lower the LCOH.
The sensitivity analysis is based on current short-term target production prices for components such as III-V solar cells and membranes.As production scales up, these costs will decrease further, following the learning curve of other technologies.For instance, Khan et al. predict a 0.12 USD W −1 long-term price for III-V PV systems [46] and Grimm et al. a 148 USD kW −1 long-term price for electrolyzer stacks. [45]These combined cost reductions would result in an LCOH of 3.8 USD kg −1 .When also considering the increase in capacity factor and STH efficiency mentioned above, the long-term price drops to 1.8 USD kg −1 .This would make III-V-based systems competitive with other hydrogen production technologies.It may be possible to lower the price even further with the introduction of hydrogen tax credits, such as those proposed under the Inflation Reduction Act of 2022.
Before concluding the sensitivity analysis, it is important to note the significance of the financial parameters in this study.Discussions with industry representatives led to the selection of an 8% discount rate for the analysis, while recent literature varies between 5% [49] and 12%. [12,45]Varying the discount rate over the 3-15% range results in a variation in the LCOH from 6.4 to 11.6 USD kg −1 .The administration and insurance costs are also largely overlooked in the literature for green hydrogen production systems but have been estimated to be up to 5% for other hydrogen production technologies. [50]Varying their share from 0% to 5% of the CAPEX raises the LCOH from 7.6 to 11.2 USD kg −1 , which highlights once more the importance of minimizing the CAPEX to reach a competitive LCOH.Due to the size of the parameter space, the lack of boundary consistency or even specifications in the literature (BOP, storage capacity, compression, administration, insurance, etc.) and the recent worldwide inflation surge, it is difficult to compare this work with others in the literature.Nevertheless, our results align with other estimations in the literature, although still slightly higher than for Si-based hydrogen production systems. [12,45,46]

Conclusion and Outlook
In conclusion, this research work demonstrates an advancement in solar water splitting technology through the successful construction of a highly efficient immersed solar water splitting device.This device, equipped with a ratio-adjustable array of electrochemical and photoelectrochemical cells, shows remarkable potential in enhancing the STH efficiency, reaching a record 20.7% under 1 sun illumination.The use of earth-abundant catalysts and the capacity to be thermally integrated permitting operation under concentrated illumination further underscore its promising practical applicability.Additionally, this work also presents a comprehensive techno-economic analysis, providing valuable insights into the factors influencing the LCOH.The results of the sensitivity analysis emphasize the substantial impact of variables such as III-V cell cost, electrolyzer membrane cost and capacity factor on the LCOH, pointing to potential strategies for further cost reduction.Moving forward, this work serves as a significant milestone toward achieving highly efficient and economically viable solar water splitting.As this system is progressively optimized and new avenues of improvement are explored, it is believed that this technology holds great promise as a sustainable solution for hydrogen production.Materials Characterization-Structure and Morphology: The structure, morphology, and elemental composition of the cocatalysts were studied using scanning electron microscopy (SEM, FEI Helios NanoLab 600), scanning transmission electron microscopy (STEM, JEOL JEM-ARM200F) and energy-dispersive X-ray spectroscopy (EDX).Lamellas were prepared via a standard focused ion beam-SEM (FIB-SEM) procedure.Cu was deposited on the samples during lamella preparation for protection.The crystallinity of the cocatalysts was analyzed using X-ray diffraction (XRD, PANalytical X'Pert PRO MRD), performed using a Cu K X-ray source.

Experimental Section
Materials Characterization-Electronic Properties: Electronic properties were examined using X-ray photoelectron spectroscopy (XPS, Thermo ESCALAB250Xi), performed using an Al K X-ray source.The spectrometer was calibrated against Cu, Ag, and Au references.
Device Performance and Characterization-(Photo)electrochemical Measurements: (Photo)electrochemical measurements were performed in a quartz cube cell containing 1 m NaOH electrolyte.Standard two-electrode and three-electrode systems were used, with a working electrode, a Ag/AgCl reference electrode and a Pt coil counter electrode connected to a potentiostat (CHI660E).A 100 W Xe arc lamp fitted with an AM1.5G filter (Abet Technologies 11 002 SunLite) was used to provide simulated 1 sun illumination.Electrochemical impedance spectroscopy (EIS) was carried out from 1 Hz to 100 kHz.

Figure 2 .
Figure 2. a) I-V characteristics of NiMo x /Ni foam (≈1.5 cm 2 ) without iR-compensation.b) iR-compensated J-V characteristics and c) Nyquist plots of NiFe(OH) x /Ni foil.All measurements were performed in 1 m NaOH electrolyte.

Figure 3 .
Figure 3. a) Schematic and b) PV response of the triple-junction photoanode, measured prior to encapsulation.c) Long-term stability of the triplejunction photoanode with a Pt plate cathode, measured chronoamperometrically at 0 V. d) Two-electrode J-V characteristics of the triple-junction photoanode before and after the stability test.All measurements were performed under 1 sun illumination.Dark currents are shown as dotted lines.

Figure 4 .
Figure 4. a) Schematic of the triple-junction solar water splitting device.b) PV responses and two-electrode J-V characteristics of series-connected photoanodes and electrochemical cells, respectively, with maximum power points marked.c) Two-electrode J-V characteristics of the triple-junction device with different ratios of photoanodes to electrochemical cells.d) Short-term stability of the triple-junction device in the 3:4 configuration.e) Comparison of this work (circled) with previously reported solar water splitting systems.References are listed in Table S2 (Supporting Information).All measurements were performed in 1 m NaOH electrolyte under 1 sun illumination.Dark currents are shown as dotted lines.

Figure 5 .
Figure 5. a) Modeled STH efficiency of each configuration with sequential improvements.b) CAPEX and OPEX cost repartitions for the proposed system.c) Sensitivity analysis of the proposed system.

Fabrication-
Fabrication of Earth-Abundant Cocatalysts: NiMo x /Ni foam was synthesized by first sonicating Ni foam in 3 m HCl, acetone, and DI water.An aqueous solution containing 0.12 m Ni(NO 3 ) 2 •6H 2 O and 0.03 m (NH 4 ) 6 Mo 7 O 24 •4H 2 O was prepared and transferred to a 50 mL Teflon/stainless steel reactor.The Ni foam was immersed in the solution.The reactor was sealed and a hydrothermal reaction was carried out in an oven at 150 °C for 15 h.The obtained sample was rinsed with DI water and ethanol and then dried at 70 °C for 5 h, followed by calcination at 500 °C for 3 h under a constant flow of H 2 /Ar (5/50 mL per minute).NiFe(OH) x /Ni foil was synthesized by first sonicating 100 μm thick Ni foil in acetone, ethanol, and 0.25 m HCl for 10 min.Kapton tape was used to protect one side of the Ni foil during deposition.NiFe(OH) x was cathodically electrodeposited on the Ni foil from an aqueous solution containing 20 mm NiCl 2 , 20 mm FeCl 3 , and 0.1 m NaCl.Electrodeposition was carried out in a three-electrode cell via a single cyclic voltammetric scan from −0.5 to −2.0 V versus Ag/AgCl and back at a scan rate of 10 mV s −1 , with a Pt coil counter electrode.Fabrication-Construction of the Triple-Junction Solar Water Splitting Device: InGaP/InGaAs/Ge cells were fabricated by MicroLink Devices, Inc., USA.Three cells were used to make photoanodes, with NiFe(OH) x /Ni foil contacted directly to the back side of each cell.Ag paint and Ag bars were used to make electrical connections while glass slides, Teflon tape and epoxy (Loctite EA 9462) were used to construct and encapsulate the device.NiMo x /Ni foam cathodes were clipped to Ag bars.An individual photoanode was also constructed using a different epoxy (Loctite EA E-60NC).

Table 1 .
Parameters used for the techno-economic analysis.