A Monolithic Photovoltaic-Photoelectrochemical Device for Hydrogen Production via Water Splitting

See allHide authors and affiliations

Science  17 Apr 1998:
Vol. 280, Issue 5362, pp. 425-427
DOI: 10.1126/science.280.5362.425


Direct water electrolysis was achieved with a novel, integrated, monolithic photoelectrochemical-photovoltaic design. This photoelectrochemical cell, which is voltage biased with an integrated photovoltaic device, splits water directly upon illumination; light is the only energy input. The hydrogen production efficiency of this system, based on the short-circuit current and the lower heating value of hydrogen, is 12.4 percent.

The direct photoelectrolysis of water at the semiconductor electrolyte interface is not only of scientific interest but also could produce benefits such as cost and energy savings over coupled photovoltaic (PV)-electrolysis systems for H2 production (1). However, the lack of a stable, efficient light-absorption system coupled with suitable semiconductor-redox energetics has kept photoelectrochemists from reaching this goal.

Obstacles to direct photoelectrolysis of water are the lack of efficient light absorption (for reasonable solar efficiencies, the band gap must be less than 2.0 eV), corrosion of the semiconductor (thermodynamically, most useful semiconductors are photochemically unstable in water), and energetics (the difficulty of matching the semiconductor band-edge energies with the H2 and O2 evolution reactions) (2, 3). The most photochemically stable semiconductors in aqueous solution are oxides, but their band gaps are either too large for efficient light absorption or their semiconductor characteristics are poor (4). Semiconductors with better solid-state characteristics are typically thermodynamically unstable with respect to oxidation (2). However, p-type semiconductors generally offer some protection against photocorrosion, because under illumination the surface is cathodically protected. Heller (5, 6) has shown that p-type indium phosphide is stable in strong acid under illumination and H2 evolution but requires an external bias for water splitting. In earlier work, we identifiedp-type gallium indium phosphide (p-GaInP2) as perhaps a perfect semiconductor for water splitting. Although its band gap is in the ideal range (1.8 to 1.9 eV), the energetics of its band edges are not correct for water splitting, needing an additional 300-mV bias (7). Although it is possible to move its band edges to some extent, their energetic positions are still inadequate to effect water splitting (8).

A number of approaches have been tried to overcome some of the obstacles to the direct splitting of water. Cells with two simultaneously illuminated semiconductors have been used (9-11). Multiple-junction amorphous silicon devices have also been reported by Sakai et al. (12) and Lin et al. (13). Augustynskiet al. (14) have reported a water splitting system using a dye-sensitized nanocrystalline solar cell coupled to a tungsten trioxide electrode. However, each of their systems operated separately and they were not integrated into a monolithic device. Although all of these systems had some success, their efficiencies were rather low (<7%).

We now report a direct water electrolysis system based on a novel, integrated, monolithic photoelectrochemical (PEC)/PV device (Fig.1A). This device is patterned after the GaInP2/GaAs p/n,p/n tandem cell device grown at the National Renewable Energy Laboratory (15). The solid-state tandem cell consists of a GaAs bottom cell connected to a GaInP2top cell through a tunnel diode interconnect. The topp/n GaInP2 junction, with a band gap of 1.83 eV, is designed to absorb the visible portion of the solar spectrum; and the bottom p/n GaAs junction, with a band gap of 1.42 eV, absorbs the near-infrared portion of the spectrum transmitted through the top junction. Although single-gap electrodes have a solar conversion efficiency limit of 32%, tandem-junction devices have an efficiency limit of 42% (16). The maximum theoretical solar-to-electrical efficiency for the present combination of band gaps is about 34% (17). Our device differs from the standard solid-state tandem cell in that a PEC Schottky-type junction has replaced the topp/n junction. This device then is a PEC device that is voltage-biased with an integrated PV device. Under illumination, electrons flow toward the illuminated surface and holes flow toward the ohmic contact. Water is split directly upon illumination, light being the only energy input.

Figure 1

(A) Schematic of the monolithic PEC/PV device. (B) Idealized energy level diagram for the monolithic PEC/PV photoelectrolysis device.

The p-GaInP2/GaAs cells used in this study were grown by atmospheric-pressure organometallic vapor-phase epitaxy. Details of material growth and cell processing have been previously reported (18). The cell structure consists of a top layer of epitaxially grown p-Ga0.52In0.48P (abbreviated as p-GaInP2) 4.0 ± 0.5 μm thick, connected in series via a low-resistivity, grown-in tunnel junction (TJ) to a GaAs p/n bottom cell on a GaAs substrate. The photoelectrochemical characteristics of the present cell were compared with those of a similarly grownp-GaInP2 epilayer on a GaAs substrate. We have studied a number of similar devices with varying thicknesses of the top layer (19). The present configuration produced the best overall results and is reported here.

Standard chemical and electrochemical procedures were used (20), with the surface of the samples being coated with a platinum catalyst (21). Illumination for the photoelectrolysis was produced by a fiber optic illuminator with a 150-W tungsten-halogen lamp. To measure the light irradiance on the surface of the sample, a calibrated PV tandem cell was mounted into the electrode holder inside the cell in the manner similar to that used for the photoelectrodes. The measured light irradiance was 1.19 ± 0.05 W/cm2: about 11 suns (22).

For the PEC/PV configuration to work properly, the GaAs bottom cell must provide sufficient voltage to overcome any energetic mismatch between the band edges of the GaInP2 and the water redox reactions, and must also provide any additional voltage needed to overcome overvoltage losses from the H2 and O2evolution reactions. The total photovoltage output must include the thermodynamics for water splitting (1.23 V); polarization losses ηa and ηc for the anodic and cathodic processes, respectively; and the current-resistance potential drop in the bulk of the electrolyte, which can be significant when gas evolution occurs. An idealized energy-level diagram for the photolytic splitting of water with this device (Fig. 1B) illustrates the process, which includes two photons and one separated electron-hole pair. Light incident on the PEC/PV configuration first enters the wide band gapp-GaInP2 layer, in which the more energetic photons are absorbed, resulting in electron-hole pair excitation and producing photovoltage V ph1. Less energetic photons penetrate through the GaInP2 and are absorbed by a GaAs bottom p/n junction generating photovoltageV ph2. One set of holes and electrons are recombined at the tunnel junction. If the resultant photovoltageV ph= V ph1 +V ph2 is greater than that required for photoelectrolyis for this particular cell configuration, it will drive the water reduction reaction at the semiconductor electrode and water oxidation at the counterelectrode. Two photons are required to produce one electron in the external circuit, so four photons are required to produce one molecule of H2. This is an example of a D4 scheme (16).

This configuration can easily be designed to do oxidation as well as reduction at the semiconductor-electrolyte interface. For water-splitting reactions, however, H2 production at the surface of the semiconductor is preferred for two reasons: (i) the H2 evolution reaction has the lowest overvoltage losses, so the requirement for a catalyst is less, thus allowing an optimized counterelectrode to be used for the more complex O2evolution reaction; and (ii) under illumination, the semiconductor surface is cathodically protected.

Figure 2 shows photocurrent-voltage curves for platinum-catalyzed p-GaInP2(Pt) andp-GaInP2(Pt)/TJ/GaAs electrodes measured in a two-electrode configuration. The open circuit voltages in the dark were ∼–0.64 and –0.75 V for the p-GaInP2(Pt) andp-GaInP2(Pt)/TJ/GaAs electrodes, respectively. The dark reduction current was small (in microampere range) for both electrodes. Under illumination, the p-GaInP2(Pt) electrode started to generate hydrogen at a voltage 500 mV negative of 0 V bias, indicating that additional external voltage was needed for this semiconductor to split water. Under illumination, thep-GaInP2 (Pt)/TJ/GaAs electrode showed an open-circuit voltage of ∼0.55 V, indicating the extra photovoltage generated by the GaAs cell. Evolution of H2 started immediately, 400 mV positive of a short circuit. The photocurrent density reached a limiting value of 120 mA/cm2 at ∼0.15 V and remained almost constant with increasing bias. Copious amounts of gas bubbles could be observed at the semiconductor surface. Because these gas bubbles can grow to sufficient size to act as miniature lenses, causing pitting of the semiconductor electrode, 0.01 M of the surfactant Triton X-100 was added to the solution to promote the formation of smaller bubbles that left the sample surface more rapidly. The lower saturated photocurrent for thep-GaInP2 (Pt)/TJ/GaAs electrode as compared to the p-GaInP2 electrode is taken to indicate that the p/n GaAs bottom cell is the current-limiting junction.

Figure 2

(left). Current-voltage characteristics forp-GaInP2(Pt)/TJ/GaAs (curve 1) andp-GaInP2 (curve 2) electrodes in 3 M H2SO4 under white light illumination.

The photocurrent time profile for thep-GaInP2(Pt)/TJ/GaAs electrode at short-circuit condition is shown in Fig. 3. The current density decreased slightly for about 300 s but then stabilized at about 120 mA/cm2 and remained constant throughout the first part of the experiment. The inset in Fig. 3 shows the long-term current stability of the p-GaInP2(Pt)/TJ/GaAs electrode. After 20 hours, the initial current density of 120 mA/cm2decreased to 105 mA/cm2. Gas bubbles continued to be evolved from the entire surface. Visual inspection of the electrode surface after the experiment revealed some localized damage of the sample along the upper edge of the epoxy coating, where gas bubbles collect and are held by the “lip” of the epoxy. The remaining portion of the sample appeared undamaged. Because these experiments were done with the electrode in a vertical position, a more solar-like slanted position (not possible with our current configuration) would reduce or eliminate the accumulation of bubbles at this edge.

Figure 3

(right).Photocurrent time profile at short circuit for PEC/PV tandem cell in 3 M H2SO4 with 0.01 M Triton X-100 under tungsten-halogen white light illumination. Efficiency = 120 mA/cm2 × 1.23 V × 100/1190 mW/cm2 = 12.4%. CD, current density.

The products of photoelectrolysis were collected and analyzed with a mass spectrometer. Within our experimental error, the ratio was 2:1 H2:O2 as expected (23). The efficiency of H2 production was calculated with the following equation: efficiency = (power out)/(power in). The input power is the incident light intensity of 1190 mW/cm2. For the output power, assuming 100% photocurrent electrolysis efficiency, the H2 production photocurrent of 120 mA/cm2 is multiplied by 1.23 V, which is the ideal fuel cell limit at 25°C (the lower heating value of hydrogen). Calculated by this method, the H2 production efficiency of our system is 12.4% (24).

Gerischer (2) has calculated for a two-layer semiconductor system (a tandem cell) that the maximum efficiency of photoelectrochemical water splitting would be 24%. Bolton et al. (16) estimated 16% as a maximum realizable chemical conversion efficiency for a tandem system of this type (a D4 scheme in his nomenclature). Weber and Dignam (25,26) calculated that for different tandem cell configurations, the maximum water-splitting efficiency would fall between 10 and 18%. Their efficiency analysis took into account various solid-state and electrochemical losses. Our results are consistent with these calculations.

The key to making this system work appears to be the requirement that the bottom cell be the limiting electron provider. Because these tandem systems operate by requiring two photons (one per junction) to produce one electron in the external circuit, great care is taken in the solid-state systems to match the photon absorption characteristics so that equal numbers of photocarriers are generated in the top and bottom cells. However, when the photocurrent-matched configuration is placed in an aqueous environment, the system either produces no current in the external circuit or decomposes (27). A thicker, topp-type layer and the resultant mismatch of electron-hole formation in the two-junction region appear to be the key for proper operation.

  • * To whom correspondence should be addressed. E-mail: jturner{at}


View Abstract

Navigate This Article