..............................................................

A photovoltaic device structurebased on internal electron emissionEric W. McFarland* & Jing Tang†

* Department of Chemical Engineering, † Materials Department, University ofCalifornia, Santa Barbara, California 93106-5080, USA.............................................................................................................................................................................

There has been an active search for cost-effective photovoltaicdevices since the development of the first solar cells in the 1950s(refs 1–3). In conventional solid-state solar cells, electron–holepairs are created by light absorption in a semiconductor, withcharge separation and collection accomplished under the influ-ence of electric fields within the semiconductor. Here we report amultilayer photovoltaic device structure in which photon absorp-tion instead occurs in photoreceptors deposited on the surface ofan ultrathin metal–semiconductor junction Schottky diode.Photoexcited electrons are transferred to the metal and travelballistically to—and over—the Schottky barrier, so providing thephotocurrent output. Low-energy (,1 eV) electrons have sur-prisingly long ballistic path lengths in noble metals4,5, allowing alarge fraction of the electrons to be collected. Unlike conventionalcells, the semiconductor in this device serves only for majoritycharge transport and separation. Devices fabricated using afluorescein photoreceptor on an Au/TiO2/Ti multilayer structurehad typical open-circuit photovoltages of 600–800 mV and short-circuit photocurrents of 10–18 mA cm22 under 100 mW cm22

visible band illumination: the internal quantum efficiency (elec-trons measured per photon absorbed) was 10 per cent. Thisalternative approach to photovoltaic energy conversion mightprovide the basis for durable low-cost solar cells using a variety ofmaterials.

The device structure is a solid-state multilayer; it consists of aphotoreceptor layer deposited on a ,10–50-nm Au film, which caps200 nm of TiO2 on an ohmic metal back contact (Fig. 1). Au and

TiO2 are chosen and prepared so that a Schottky barrier, f, ofapproximately 0.9 V is formed at the metal–semiconductor inter-face, and the dye, merbromin (2,7-dibromo-5-(hydroxymercurio)-fluorescein), is selected such that the photoexcited donor level isenergetically above the barrier. The photon-to-electron conversionin this device occurs in four steps. First, light absorption occurs inthe surface-absorbed photoreceptors, giving rise to energetic elec-trons. Second, electrons from the photoreceptor excited state areinjected into conduction levels of the adjacent conductor, wherethey travel ballistically through the metal at an energy, 1e, above theFermi energy, Ef. Third, provided that 1e is greater than the Schottkybarrier height, f, and the carrier mean-free path is long compared tothe metal thickness, the electrons will traverse the metal and enterconduction levels of the semiconductor (internal electron emis-sion). The absorbed photon energy is preserved in the remainingexcess electron free energy when it is collected at the back ohmiccontact, giving rise to the photovoltage, V. The photo-oxidized dyeis reduced by transfer of thermalized electrons from states near Ef inthe adjacent metal. Like electrochemical dye-sensitized solar cells6,7,this structure physically separates the photon absorption processfrom charge separation and transport; but it does not have thedisadvantage of needing a reducing agent in an electrolyte forintermolecular charge transport.

In addition to the desired production of ballistic electrons in themetal, there are other processes available for energy dissipation bythe electron in the excited photoreceptor. The efficiency of thedevice will depend upon favouring specific energy transfer pathwaysfrom the major competing processes: (1) radiative intramolecularde-excitation with photon emission (luminescence), (2) non-radia-tive intramolecular or intermolecular de-excitation with directcoupling to phonons, (3) non-radiative intermolecular de-exci-tation with coupling to the metal conduction electrons, and (4)electron transfer of the energetic electron from the photoexcitedstate into the unoccupied conduction states of the metal. Bothpathways (3) and (4) may produce energetic electrons in metalconduction levels above the Fermi energy (hot electrons) that alsohave sufficient energy and momentum to traverse the Schottkybarrier; it is these electrons that give rise to the primary photo-current in the device. We believe that the dominant pathway is (4) inour device, but we cannot rule out contributions to the measuredphotocurrent from pathway (3). Pathway (2) as well as non-transmitted electrons from (3) and (4) constitute the principalcompeting ‘quenching’ processes whereby the electronic excitationenergy is dissipated as heat, and neither appears as light nor electronfree energy in the device current.

The current–voltage characteristics of a representative devicemeasured in the dark and under 100 mW cm22 visible band

Figure 1 Electron transfer in the operating photovoltaic device. Process A, photon

absorption and electron excitation from the chromophore ground state, S, to the excited

state, S*; B, energetic electron transfer from S* into and (ballistically) through the

conducting surface layer and over the potential energy barrier into the semiconductor;

C, electron conduction as a majority carrier within the semiconductor to the ohmic back

contact and through the load; and D, reduction of the oxidized chromophore, Sþ, by a

thermal electron from the conductor surface. Shown schematically are the relative

energies of the electron levels within the device structures, the Schottky barrier, f, the

Fermi energy, E f, and the semiconductor bandgap, Eg.

Figure 2 Current–voltage characteristics of the multilayer merbromin/Au/TiO2/Ti

photovoltaic device. The response is shown for dark conditions (curve A), and for

1,000 W m22 broadband visible illumination (B). For comparison, curves obtained from

the same device before adding merbromin are also shown for dark conditions (C), and

under the same illumination (D).

letters to nature

NATURE | VOL 421 | 6 FEBRUARY 2003 | www.nature.com/nature616 © 2003 Nature Publishing Group

illumination before and after applying merbromin are shown in Fig.2. After deposition of merbromin, the surface-activated device hadan open-circuit photovoltage, Voc, of 685 mV, and a short-circuitphotocurrent, jsc, of 18.0 mA cm22. The fill factor was determined tobe 0.63 from the current–voltage characteristics under illumination.Before deposition of merbromin on the Au surface, excitation ofdefect levels in the thermally oxidized TiO2 is responsible for a weakvisible photoresponse (Voc ¼ 325 mV, jsc ¼ 0.9 mA cm22).

The photovoltage, V, is the difference, under illumination,between the Fermi level of the semiconductor and the Fermi levelof the ultrathin metal film. For the surface-sensitized Schottkydiode, the expression for the maximum value under open circuitconditions, Voc is identical to a conventional Schottky solar cellwhere photon absorption occurs by bandgap excitation:

Voc ¼nkT

e

� �½lnðjg=ðT

2A** ÞÞ þ efB=kT� ð1Þ

where n is the ideality factor (,1), k is Boltzmann’s constant, e is thecharge on the electron, A ** is the Richardson constant, and T is thedevice temperature5. The photocurrent density produced by photonabsorption, jg, is balanced by the effective forward bias current fromVoc such that no net current is observed. The maximum photo-voltages and photocurrents can be achieved by choice and prep-aration of the device materials to control and match the Schottkybarrier height, f, the semiconductor conduction band positions,and the chromophore donor/acceptor levels.

The photocurrent produced from a monochromatic photon flux,F(1g), is determined largely by the photon capture efficiency and

the internal quantum efficiency, IQE, which is the number of hotelectrons injected into the semiconductor (and thus detected) perabsorbed photon:

jgð1gÞ ¼ Fð1gÞhgð1gÞIQEð1gÞ ð2Þ

IQEð1gÞ ¼

ð1g

0

hCMð1g;1eÞhMSð1eÞhS d1e ð3Þ

where hg(1g) is the efficiency for absorption of a photon of energy1g; hCM(1g,1e) is the probability that absorption will result in theinjection of an excited electron into the metal at an energy, 1 e, abovethe Fermi energy; hMS(1e) is the efficiency for charge transportacross the metal film and into the semiconductor; and hS is thecharge collection efficiency in the semiconductor. Experimentally,the incident photon-to-electron conversion efficiency, IPCEð1gÞ ¼jð1gÞ=Fð1gÞ; is typically determined by measuring the photocurrentfrom a monochromatic photon source. The IQE is then calculatedby correcting the IPCE for photon absorption in the dye asIQEð1gÞ ¼ IPCEð1gÞ=hgð1gÞ:

What makes the IQE for this device structure acceptable is thesurprisingly high value of hMS(1 e). At energies of ,1 eV above theFermi level, the ballistic mean free path for electrons in Au and othermetals with low-lying or filled d bands has been measured to beextremely long, ,20–150 nm (refs 4, 5). Thus, despite the compet-ing processes of quasi-elastic scattering by phonons and inelasticscattering from other electrons, provided the metal conducting filmis ultrathin, a significant fraction of the injected electrons will reachthe Schottky barrier.

Although the relationship between Voc and jg in our devices isidentical to that of a conventional semiconductor Schottky solarcell, equation (1), the bulk semiconductor is not utilized for photonabsorption; thus the bandgap and semiconductor thickness con-straints are largely removed. The significance of the Schottky barrierheight, however, is increased. The absorption process occurs in thephotoreceptor, which is selected to balance high solar-spectrumabsorbance with maximum photovoltage as well as chemical stab-ility. In this design, several different chromophores could be usedsimultaneously to provide more efficient conversion (for example,dyes, quantum structures); however, in an idealized cell using asingle dye, the optimal chromophore absorption maximum wouldbe the same as the ideal bandgap of a conventional Schottky diodesolar cell (,1.5 eV; refs 2, 8). Thus, the theoretical maximum poweroutput of the Schottky device utilizing internal electron emissioncan be no greater than an idealized conventional cell for the solarspectrum, ,25% (refs 2, 8).

The physical and electronic coupling of the chromophore to themetal conduction levels are crucial to the performance of the device.In aqueous solution, merbromin has an absorption maximum at511 nm wavelength; but when merbromin is attached to the Au film,the primary absorption peak appears to split into a blue-shiftedpeak and a stronger red-shifted peak, giving rise to a broadening andred-shift of the overall spectrum (Fig. 3a). The wavelength-depen-dent photoresponse of the device is shown in Fig. 3b, where we plotthe IPCE under short-circuit conditions, IPCE¼ jð1gÞ=Fð1gÞ;equation (2). The general features of the IPCE response spectrum,and the broad maximum, are approximately the same as the sur-face-bound dye absorption; this is consistent with the mechanism ofaction described in Fig. 1, with both the associated red- and blue-shifted dye states coupling to the metal such that hot electrons areproduced. The IQE is determined by correcting the IPCE for thefraction of incident photons absorbed in the dye (Fig. 3c). Althoughthe broad absorption edge of the TiO2 overlaps with the dyeadsorption between 400 nm and 450 nm, above 500 nm the TiO2

absorption is negligible and the IQE of approximately 10% reflectsthe efficiency of the internal emission process.

The main considerations in choosing the materials for the deviceare: (1) the relative energies of the donor/acceptor levels in the

Figure 3 The wavelength-dependent photoresponse and absorbance of the photovoltaic

device. a, The absorbance of merbromin adsorbed to the surface of the Au/TiO2/Ti device

(solid line). Shown for reference is the absorbance (in arbitrary units) of merbromin in

water (dotted line). b, The incident photon-to-electron conversion efficiency (IPCE) of the

dye-sensitized device under 0.4 mW cm22 illumination. c, The internal quantum

efficiency (IQE), determined by correcting the IPCE by the absorption efficiency of the

affixed dye, reflects the absorbed photon-to-electron conversion efficiency of the dye.

letters to nature

NATURE | VOL 421 | 6 FEBRUARY 2003 | www.nature.com/nature 617© 2003 Nature Publishing Group

photoreceptor, the conductor Fermi level, the barrier height, andthe position of the semiconductor conduction band edges, (2) theballistic mean free path of electrons, and (3) the physical andelectronic coupling of the chromophore to the conductor forhigh absorbance and high electron transfer efficiency. The devicesstudied here represent only one of several different configurations ofphotovoltaics taking advantage of ballistic electron transport andinternal electron emission in a photovoltaic device. For example,modifications of the structure to utilize hot hole injection (ratherthan hot electrons) in a p-type junction9 would allow use of holeconducting polymers instead of inorganic semiconductors.

The IPCE and overall energy efficiency are limited by low dyecoverage (8 £ 1014 molecules cm22) and the resulting low photonabsorption. Significant increases are expected with improved opti-cal design (reduced surface reflection), decreased metal thickness,increased dye loading, and an engineered surface morphology withsignificantly higher surface area structured such that multiple passesthrough a dye-covered surface are possible for each photon.Although the ultimate efficiency of an optimized device based onthe concept presented here is approximately the same as an idealconventional semiconductor cell, there appear to be practical andeconomic advantages in terms of the wide choice of inexpensive,durable, and readily synthesized device materials that may beutilized. A

MethodsDevice fabricationDevices were fabricated on titanium foil substrates (Alfa Aesar), which served as ohmicback contacts. A 250-nm layer of titanium (99.9999%) was evaporated under vacuumonto the foil following cleaning and polishing (using 10 mm grit). A 200-nm layer of TiO2

was grown on the substrate by thermal oxidation at 500 8C. The polycrystalline TiO2 ispredominately rutile phase, with oxygen vacancies giving rise to n-type doping. Au filmswere electrodeposited onto the TiO2 from a solution containing 0.2 M KCN and 0.1 MAuCN at pH 14. The TiO2 served as the working electrode, with a Pt wire counterelectrode. A 100-ms galvanostatic pulse at 2200 mA cm22 was used to nucleate Auuniformly on the surface, followed by a periodic galvanostatic pulse train of 5 ms atþ0.2 mA cm22 and 5 ms at 21.7 mA cm22 for 10 s to form a film ,10–50 nm thick.Photoactive merbromin (2,7-dibromo-5-(hydroxymercurio)fluorescein disodium salt,5 mM in water) was adsorbed onto the surface by immersion at room temperature for10–12 h, followed by rinsing in water.

CharacterizationCurrent –voltage (I–V) curves were measured using a voltage ramp rate of 0.05 V s21 in thedark and under illumination from a 250-W tungsten lamp (Oriel, 6129), with intensitymeasured using a radiometer (IL1700, International Light). The fill factor was calculatedat 1,000 Wm22 by dividing the maximum product of current and voltage from theilluminated I–V curve by the product of open-circuit voltage and short-circuit current atthe same illumination. The spectral response was determined using a 150-W Xe lamp andmonochromator (Oriel 7240). IPCE was calculated from the current density under short-circuit conditions and the photon flux as measured by the radiometer. The opticalabsorbance (and absorption efficiency, hg(1g)) of the dye on the device surface and dyephoton absorption was determined from the transmission and reflectance of a devicefabricated on a transparent substrate before and after application of the dye, using anintegrating sphere (LabSphere) and fibre-optic coupled monochromator (Ocean Optics).Free-solution dye absorbance was measured with an optical spectrometer (UV-1610,Shimadzu). Dye loading was determined by detaching the dye from the activated devicesurface in 1 mM NaOH solution, and determining the amount removed from thedifference in optical absorbance at 511 nm of the NaOH solution.

Received 17 July; accepted 18 November 2002; doi:10.1038/nature01316.

1. Chapin, D. M., Fuller, C. S. & Pearson, G. L. A new silicon p-n junction photocell for converting solar

radiation into electrical power. J. Appl. Phys. 25, 676–677 (1954).

2. Archer, M. D. & Hill, R. (eds) Clean Electricity from Photovoltaics (Series on Photoconversion of Solar

Energy, Vol. 1, Imperial College Press, London, 2001).

3. Goetzberger, A. & Hebling, C. Photovoltaic materials, past, present, future. Sol. Energy Mater. Sol. Cells

62, 1–19 (2000).

4. Seah, M. P. & Dench, W. A. Quantitative electron spectroscopy of surfaces: a standard data base for

electron inelastic mean free paths in solids. Surf. Interf. Anal. 1, 2–11 (1979).

5. Frese, K. W. & Chen, C. Theoretical models of hot carrier effects at metal-semiconductor electrodes.

J. Electrochem. Soc. 139, 3234–3249 (1992).

6. Gratzel, M. Photoelectrochemical cells. Nature 414, 338–344 (2001).

7. Green, M. A., Emery, K., King, D. L., Igari, S. & Warta, W. Solar cell efficiency tables (version 17). Prog.

Photovolt. Res. Applic. 9, 49–56 (2001).

8. Sze, S. M. Physics of Semiconductor Devices 2nd edn, Ch. 14 (Wiley & Sons, New York, 1981).

9. Nienhaus, H. et al. Electron-hole pair creation at Ag and Cu surfaces by adsorption of atomic

hydrogen and deuterium. Phys. Rev. Lett. 82, 446–448 (1999).

Acknowledgements We thank M. White, A. Tavakkoly, A. Kochhar, N. Shigeoka, G. Stucky and

W. Siripala for technical assistance and discussions. The project was supported by Adrena Inc.

Financial support for J.T. was provided by the NSF-MRSEC funded Materials Research

Laboratory (UCSB).

Competing interests statement The authors declare competing financial interests: details

accompany the paper on Nature’s website (ç http://www.nature.com/nature).

Correspondence and requests for materials should be addressed to E.W.M.

(e-mail: mcfar@engineering.ucsb.edu).

..............................................................

Hydrothermal recharge anddischarge across 50 km guided byseamounts on a young ridge flankA. T. Fisher*†, E. E. Davis‡, M. Hutnak*, V. Spiess§, L. Zuhlsdorff§,A. Cherkaoui*, L. Christiansenk, K. Edwards{, R. Macdonald‡,H. Villinger§, M. J. Mottl#, C. G. Wheatq & K. Becker{

* Earth Sciences Department, † Institute for Geophysics and Planetary Physics,University of California, Santa Cruz, California 95064, USA‡ Pacific Geoscience Center, Geological Survey of Canada, Sydney, BritishColumbia V8L 4B2, Canada§ Earth Sciences Department, University of Bremen, Bremen D-28359, GermanykDepartment of Earth and Planetary Sciences, Johns Hopkins University,Baltimore, Maryland 21218, USA{Rosenstiel School of Marine and Atmospheric Science, University of Miami,Miami, Florida 33149, USA# School of Earth and Ocean Science and Technology, University of Hawaii,Honolulu, Hawaii 96822, USAq Global Undersea Research Unit, University of Alaska, Fairbanks, Alaska 99775,USA.............................................................................................................................................................................

Hydrothermal circulation within the sea floor, through litho-sphere older than one million years (Myr), is responsible for 30%of the energy released from plate cooling, and for 70% of theglobal heat flow anomaly (the difference between observedthermal output and that predicted by conductive coolingmodels)1,2. Hydrothermal fluids remove significant amounts ofheat from the oceanic lithosphere for plates typically up to about65 Myr old3,4. But in view of the relatively impermeable sedi-ments that cover most ridge flanks5, it has been difficult toexplain how these fluids transport heat from the crust to theocean. Here we present results of swath mapping, heat flow,geochemistry and seismic surveys from the young eastern flankof the Juan de Fuca ridge, which show that isolated basementoutcrops penetrating through thick sediments guide hydrother-mal discharge and recharge between sites separated by more than50 km. Our analyses reveal distinct thermal patterns at the seafloor adjacent to recharging and discharging outcrops. We findthat such a circulation through basement outcrops can besustained in a setting of pressure differences and crustal proper-ties as reported in independent observations and modellingstudies.

Hydrothermal circulation on ridge flanks (crust older than1 Myr) advects lithospheric heat from much of the sea floor,contributing to enormous fluxes of fluid, energy and solutes1,6,7. Itis easy for fluid to enter and leave the crustal reservoir on mostyoung sea floor, where sediment cover is incomplete and permeablebasement rocks are widely exposed, but mechanisms by which fluidspenetrate through thick and more continuous sediments haveremained enigmatic5. The primary difficulty is that forces availableto drive hydrothermal circulation on ridge flanks are modest7–10,being limited mainly to the difference in fluid pressures below

letters to nature

NATURE | VOL 421 | 6 FEBRUARY 2003 | www.nature.com/nature618 © 2003 Nature Publishing Group