LETTERS Nanowire dye-sensitized solar cells MATT LAW, 2* LORI E. GREENE1 2*. JUSTIN C. JOHNSON1 RICHARD SAYKALLY AND PEIDONG YANG1, 2, Department of Chemistry, University of Califomia, Berkeley, Califomia 94720, USA Materials Science Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA These authors contributed equally to this work. Published online: 15 May 2005; doi: 10.1038/nmat1387 xcitonic solar cells'-including organic, hybrid organic- random walk through the film. Drift transport, a vital mechanism in inorganic and dye-sensitized cells(DSCs)-are promising most photovoltaic cells, is prevented in DSCs by ions in theelectrolyte devices for inexpensive, large-scale solar energy conversion. that screen macroscopic electric fields and couple strongly with the The dSC is currently the most efficient and stable' excitonic moving electrons, effectively rendering them neutral carriers(that photocell. Central to this device is a thick nanoparticle film is, there is ambipolar diffusion). Under full sunlight, an average that provides a large surface area for the adsorption of light- injected electron may experience a million trapping events before harvesting molecules. However, nanoparticle DSCs rely on either percolating to the collecting electrode or recombining with an trap-limited diffusion for electron transport, a slow mechanism oxidizing species, predominantly Ii in the electrolyte. Transit times that can limit device efficiency, especially at longer wavelengths. for electron escape from the film average in the milliseconds . Ye Here we introduce a version of the dye-sensitized cell in which despite the extremely slow nature of such trap-mediated transpo the traditional nanoparticle film is replaced by a dense array of (characterized by an electron diffusivity, Dn s s-, several oriented, crystalline ZnO nanowires. The nanowire anode is orders of magnitude smaller than in TiO2 and ZnO single crystals. 7) synthesized by mild aqueous chemistry and features a surface area electron collection remains favoured over recombination because up to one-fifth as large as a nanoparticle cell. The direct electrical of the even slower multi-electron kinetics of I; reduction on oxide pathways provided by the nanowires ensure the rapid collection of surfaces. Electron diffusion lengths of 7-30 um have been reported carriers generated throughout the device, and a full Sun efficiency for cells operating at light intensities up to 0. 1 Sun 0. 8. This is of 1.5% is demonstrated, limited primarily by the surface area of strong evidence that electron collection is highly efficient for the 10- the nanowire array um-thick nanoparticle films normally used in devices. The anodes of dye-sensitized solar cells are typically constructed Insight into the factors that limit DSC performance is gained by sing thick films(-10 um)of TiO, or, less often, SnO, or Zno comparing theoretical cell efficiencies with those of current state-of- nanoparticles- that are deposited as a paste and sintered to produce the-art cells. The power conversion efficiency n of a solar cell is given electrical continuity. The nanoparticle film provides a large internal as n=(FF XIxI x Voc)/Pin, where FFis the fill factor, I I is the absolute surface area(characterized by a roughness factor, defined as the total value of the current density at short circuit, Vo is the photovoltage at film area per unit substrate area, of -1,000)for the anchoring of open circuit and Pin is the incident light power density. In principle, sufficient chromophore(usually a ruthenium-based dye) to yield the maximum s of a DSC is determined by how well the absorption high light absorption in the 400-800 nm region, where much of the window of its dye overlaps the solar spectrum. Record cells achieve solar flux is incident. During operation, photons intercepted by the current densities(and overall efficiencies)that are between 55 and dye monolayer createexcitons that arerapidly split at the nanoparticle 75%of their theoretical maxima at full sunlight, depending on the surface, with electrons injected into the nanoparticle film and holes exact dye used. Much of the shortfall is due to the poor absorption aving the opposite side of the device by means of redox species of low-energy photons by available dyes. Considerable efforts have (traditionally the 1-7/I] couple)in a liquid or solid-state'electrolyte. been made to develop dyes and dye mixtures that absorb better at The classic report of a 10% efficient TiO, DSC initiated a decade of long wavelengths,2, so far with little success. Another option for research into the electrical transport physics of nanoparticle anodes, improving the absorption of red and near-infrared light-thickening which were shown to collect injected electrons with high efficiency the nanoparticle film to increase its optical densityis unsuccessful despite their disordered, polycrystalline topology. because the film thickness comes to exceed the electron diffusion The nature of electron transport in oxide nanoparticle films is length through the nanoparticle network. fairly well understood. Time-resolved photocurrentand photovoltage One promising solution to this impasse is to increase the electron measurements%, 0 and modelling studies.2 indicate that electron diffusion length in the anode by replacing the nanoparticle film with transport in wet, illuminated nanoparticle networks proceeds by a an array of oriented single-crystalline nanowires. Electron transport trap-limited diffusion process, in which photogenerated electrons in crystalline wires is expected to be several orders of magnitude repeatedly interact with a distribution of traps as they undertake a faster than percolation through a random polycrystalline network. naturematerialsIVol4iJunE2005iwww.naturecom/naturematerials @2005 Nature Publishing Group
LETTERS nature materials | VOL 4 | JUNE 2005 | www.nature.com/naturematerials 455 Nanowire dye-sensitized solar cells MATT LAW1,2*, LORI E. GREENE1,2*, JUSTIN C. JOHNSON1 , RICHARD SAYKALLY1 AND PEIDONG YANG1,2† 1 Department of Chemistry, University of California, Berkeley, California 94720, USA 2 Materials Science Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA *These authors contributed equally to this work. † e-mail: p_yang@berkeley.edu Published online: 15 May 2005; doi:10.1038/nmat1387 E xcitonic solar cells1 —including organic, hybrid organic– inorganic and dye-sensitized cells (DSCs)—are promising devices for inexpensive, large-scale solar energy conversion. Th e DSC is currently the most effi cient2 and stable3 excitonic photocell. Central to this device is a thick nanoparticle fi lm that provides a large surface area for the adsorption of lightharvesting molecules. However, nanoparticle DSCs rely on trap-limited diff usion for electron transport, a slow mechanism that can limit device effi ciency, especially at longer wavelengths. Here we introduce a version of the dye-sensitized cell in which the traditional nanoparticle fi lm is replaced by a dense array of oriented, crystalline ZnO nanowires. Th e nanowire anode is synthesized by mild aqueous chemistry and features a surface area up to one-fi ft h as large as a nanoparticle cell. Th e direct electrical pathways provided by the nanowires ensure the rapid collection of carriers generated throughout the device, and a full Sun effi ciency of 1.5% is demonstrated, limited primarily by the surface area of the nanowire array. The anodes of dye-sensitized solar cells are typically constructed using thick fi lms (~10 μm) of TiO2 or, less often, SnO2 or ZnO nanoparticles4–6 that are deposited as a paste and sintered to produce electrical continuity. The nanoparticle fi lm provides a large internal surface area (characterized by a roughness factor, defi ned as the total fi lm area per unit substrate area, of ~1,000) for the anchoring of suffi cient chromophore (usually a ruthenium-based dye) to yield high light absorption in the 400–800 nm region, where much of the solar fl ux is incident. During operation, photons intercepted by the dye monolayer create excitons that are rapidly split at the nanoparticle surface, with electrons injected into the nanoparticle fi lm and holes leaving the opposite side of the device by means of redox species (traditionally the I– /I3 – couple) in a liquid or solid-state7 electrolyte. The classic report8 of a 10% effi cient TiO2 DSC initiated a decade of research into the electrical transport physics of nanoparticle anodes, which were shown to collect injected electrons with high effi ciency despite their disordered, polycrystalline topology. The nature of electron transport in oxide nanoparticle fi lms is fairly well understood. Time-resolved photocurrent and photovoltage measurements9,10 and modelling studies11,12 indicate that electron transport in wet, illuminated nanoparticle networks proceeds by a trap-limited diffusion process, in which photogenerated electrons repeatedly interact with a distribution of traps as they undertake a random walk through the fi lm. Drift transport, a vital mechanism in most photovoltaic cells, is prevented in DSCs by ions in the electrolyte that screen macroscopic electric fi elds and couple strongly with the moving electrons, effectively rendering them neutral carriers (that is, there is ambipolar diffusion)13. Under full sunlight, an average injected electron may experience a million trapping events before either percolating to the collecting electrode or recombining with an oxidizing species, predominantly I3 – in the electrolyte14. Transit times for electron escape from the fi lm average in the milliseconds15. Yet despite the extremely slow nature of such trap-mediated transport (characterized by an electron diffusivity, Dn ≤ 10–4 cm2 s–1, several orders of magnitude smaller than in TiO2 and ZnO single crystals16,17), electron collection remains favoured over recombination because of the even slower multi-electron kinetics of I3 – reduction on oxide surfaces. Electron diffusion lengths of 7–30 μm have been reported for cells operating at light intensities up to 0.1 Sun9,10,18. This is strong evidence that electron collection is highly effi cient for the 10- μm-thick nanoparticle fi lms normally used in devices. Insight into the factors that limit DSC performance is gained by comparing theoretical cell effi ciencies with those of current state-ofthe-art cells. The power conversion effi ciency η of a solar cell is given as η = (FF × |Jsc| × Voc)/Pin, where FF is the fi ll factor, |Jsc| is the absolute value of the current density at short circuit, Voc is the photovoltage at open circuit and Pin is the incident light power density. In principle, the maximum Jsc of a DSC is determined by how well the absorption window of its dye overlaps the solar spectrum. Record cells achieve current densities (and overall effi ciencies) that are between 55 and 75% of their theoretical maxima at full sunlight, depending on the exact dye used19. Much of the shortfall is due to the poor absorption of low-energy photons by available dyes. Considerable efforts have been made to develop dyes and dye mixtures that absorb better at long wavelengths20,21, so far with little success. Another option for improving the absorption of red and near-infrared light—thickening the nanoparticle fi lm to increase its optical density—is unsuccessful because the fi lm thickness comes to exceed the electron diffusion length through the nanoparticle network. One promising solution to this impasse is to increase the electron diffusion length in the anode by replacing the nanoparticle fi lm with an array of oriented single-crystalline nanowires. Electron transport in crystalline wires is expected to be several orders of magnitude faster than percolation through a random polycrystalline network. nmat1387-print.indd 455 mat1387-print.indd 455 10/5/05 3:46:18 pm 0/5/05 3:46:18 pm ©2005 NaturePublishingGroup © 2005 Nature Publishing Group
ETTERS Platinized nanowire array electrolyte electrode d Length (um) Figure 1 The nanowire dye-sensitized cell, based on a Zno wire array. a, Schematic diagram of the cell. Light is incident through the bottom electrode. b, Typical scanning electron microscopy cross-section of a cleaved nanowire array on FTO. The wires are in direct contact with the substrate, with no intervening particle layer Scale bar, 5 um. C, Magnified view of the oriented wires. In this array, wire length and diameter vary from 16 to 17 um and 130 to 200 nm, respectively. Scale bar, 500 nm. d, without (triangles) PEl added to the growth bath. Lines are least-squares fits to the data, and error bars represent one standard deviation. Using a sufficiently dense array of long, thin nanowires as a dye aspect ratio of our nanowires above 125 by using polyethylenimine scaffold, it should be possible to increase the DSC dye loading(and (PEI), a cationic polyelectrolyte, to hinder only the lateral growth so its absorption of red light) while simultaneously maintaining of the nanowires in solution, while maintaining a relatively high very efficient carrier collection. Moreover, the rapid transport nanowire density(Fig. 1b-d and Methods). The striking effect of provided by a nanowire anode would be particularly favourable for this molecule is seen by plotting nanowire length against diameter at ell designs that use non-standard electrolytes, such as polymer gels different growth times with and without PEI(Fig. le). The longest or solid inorganic phases, in which recombination rates are high arrays presented here(20-25 um) have one-fifth the active surface compared with the liquid electrolyte cell. Here we present the first area of a nanoparticle anode. rdered nanowire DSC(Fig. la)and illustrate how this topology The wire films are good electrical conductors along the direction could improve the understanding and performance of DSCs and of the wire axes. Two-point electrical measurements of dry arrays other types of excitonic solar cells. on FTO substrates gave linear current-voltage (I-v traces(se A high-performance nanowire photoanode must foremost Supplementary Information, Fig S1), indicating barrier-fre have a large surface area for dye adsorption, comparable to that of contacts between nanowire and substrate. Individual nanowires a nanoparticle film. We made ZnO nanowire arrays of high surface were extracted from the arrays, fashioned into field-effect transist area in aqueous solution using a seeded growth processthat was using standard electron-beam lithography procedures, and analysed modified to yield long wires. Briefly, a 10-15-nm-thick film of Zno to determine their resistivity, carrier concentration and mobility uantum dots was deposited onto F: SnO2 conductive glass(FTO)(Fig. S2). Measured resistivity values ranged from 0.3 to 2.0 Q2 cm, ubstrates by dip coating, and wires were grown from these nuclei with an electron concentration of 1-5 X 108 cm-and mobility through the thermal decomposition of a zinc complex. This two-step of 1-5 cm2V-s-l. Using the Einstein relation, D= kB Tu/e,w process is a simple, low-temperature and environmentally benign estimate an electron diffusivity D=0.05-0.5 cm2s- for single route to forming dense arrays(up to 35 billion wires per square dry nanowires. This value is several hundred times larger than tl centimetre)on arbitrary substrates of any size Solution-grown Zno highest reported diffusivity for TiO, or ZnO nanoparticle films in nanowire arrays reported previously have been limited to aspect operating cells 524. Moreover, the conductivity of the wire arrays ratios of less than 20, too small for efficient DSCs. We boosted the increased by 5-20% when they were bathed in the standard DSC naturematerialsVol.4IJuNe2005iwww.nature.com/naturematerials @2005 Nature Publishing Group
LETTERS 456 nature materials | VOL 4 | JUNE 2005 | www.nature.com/naturematerials Using a suffi ciently dense array of long, thin nanowires as a dye scaffold, it should be possible to increase the DSC dye loading (and so its absorption of red light) while simultaneously maintaining very effi cient carrier collection. Moreover, the rapid transport provided by a nanowire anode would be particularly favourable for cell designs that use non-standard electrolytes, such as polymer gels or solid inorganic phases, in which recombination rates are high compared with the liquid electrolyte cell22. Here we present the fi rst ordered nanowire DSC (Fig. 1a) and illustrate how this topology could improve the understanding and performance of DSCs and other types of excitonic solar cells. A high-performance nanowire photoanode must foremost have a large surface area for dye adsorption, comparable to that of a nanoparticle fi lm. We made ZnO nanowire arrays of high surface area in aqueous solution using a seeded growth process23 that was modifi ed to yield long wires. Briefl y, a 10–15-nm-thick fi lm of ZnO quantum dots was deposited onto F:SnO2 conductive glass (FTO) substrates by dip coating, and wires were grown from these nuclei through the thermal decomposition of a zinc complex. This two-step process is a simple, low-temperature and environmentally benign route to forming dense arrays (up to 35 billion wires per square centimetre) on arbitrary substrates of any size. Solution-grown ZnO nanowire arrays reported previously have been limited to aspect ratios of less than 20, too small for effi cient DSCs. We boosted the aspect ratio of our nanowires above 125 by using polyethylenimine (PEI), a cationic polyelectrolyte, to hinder only the lateral growth of the nanowires in solution, while maintaining a relatively high nanowire density (Fig. 1b–d and Methods). The striking effect of this molecule is seen by plotting nanowire length against diameter at different growth times with and without PEI (Fig. 1e). The longest arrays presented here (20–25 μm) have one-fi fth the active surface area of a nanoparticle anode. The wire fi lms are good electrical conductors along the direction of the wire axes. Two-point electrical measurements of dry arrays on FTO substrates gave linear current–voltage (I–V) traces (see Supplementary Information, Fig. S1), indicating barrier-free contacts between nanowire and substrate. Individual nanowires were extracted from the arrays, fashioned into fi eld-effect transistors using standard electron-beam lithography procedures, and analysed to determine their resistivity, carrier concentration and mobility (Fig. S2). Measured resistivity values ranged from 0.3 to 2.0 Ω cm, with an electron concentration of 1–5 × 1018 cm–3 and mobility of 1– 5 cm2 V–1 s–1. Using the Einstein relation, D = kBTμ/e, we estimate an electron diffusivity Dn = 0.05–0.5 cm2 s–1 for single dry nanowires. This value is several hundred times larger than the highest reported diffusivity for TiO2 or ZnO nanoparticle fi lms in operating cells15,24. Moreover, the conductivity of the wire arrays increased by 5–20% when they were bathed in the standard DSC 0 5 10 15 20 25 30 35 Length (μm) 400 350 300 250 200 150 100 50 0 Daimeter (nm) Platinized electrode Dye-coated nanowire array in electrolyte Transparent electrode e– a b c d e Figure 1 The nanowire dye-sensitized cell, based on a ZnO wire array. a, Schematic diagram of the cell. Light is incident through the bottom electrode. b, Typical scanning electron microscopy cross-section of a cleaved nanowire array on FTO. The wires are in direct contact with the substrate, with no intervening particle layer. Scale bar, 5 μm. c, Magnifi ed view of the oriented wires. In this array, wire length and diameter vary from 16 to 17 μm and 130 to 200 nm, respectively. Scale bar, 500 nm. d, Typical top view of a single nanowire, showing its faceting, surface texture and a slight taper to its tip. Scale bar, 50 nm. e, Wire length against diameter with (circles) and without (triangles) PEI added to the growth bath. Lines are least-squares fi ts to the data, and error bars represent one standard deviation. nmat1387-print.indd 456 mat1387-print.indd 456 10/5/05 3:46:22 pm 0/5/05 3:46:22 pm ©2005 NaturePublishingGroup © 2005 Nature Publishing Group
LETTERS Dye loading(moles x 10- per cm2 of substrate 12 A>》》 02cm2 10 百23 22 ur 8 ●Zn0 wIres ■T2 particles,smal ▲zno0 partcles,arge 0.75 4.4um ◇ Zno partcles, small Figure 3 Comparative performance of nanowire and nanoparticle cells. Short 0.50 circuit current density versus roughness factor for cells based on Zno wires, small TO, particles, and large and small Zno particles. The TiO, films show a higher maximum current than either of the Zno films and a larger initial slope than the small ZnO particles, consistent with better transport through TO, particle network The large Zno partice cells attain a smaller maximum current than the small 0.35 particles because the film thickness(and therefore the electron escape length) becomes larger than the electron diffusion length at a much lower roughness factor. ght intensity (mW cm-3 The wire data fall on the TiO, line and significantly exceed the current output 1.6 both types of Zno particle cells above a roughness factor of 100 A slight sag of the wire data off the TiO, line at high roughness factor may be a sign of excessive scattering within the opaque wire films. Cell thickness is directly proportional to oughness factor and is labelled for each cell type at a roughness factor of 200 Error bars are provided on only two points to maximize figure clarity, and they are an estimate of the maximum range of the values. The error bars for cells wi roughness factors below 250 are smaller than the size of the data points. Data points were made by measuring the roughness factor of dye-sensitized films through ultraviolet-visible spectroscopy of the desorbed dye in basic H,0 and then a re-sensitizing the films for fabrication into cells. See also Fig. S6. Fig S7 sh comparative trends in n, Vo and fill factor Cell size: 0.8 cm2. electrolyte( Fig. S3). Thus, facile transport through the nanowire Light intensity(mW c2 array is retained in device-like environments and should result in faster carrier extraction in the nanowire cell Solar cells were constructed with wire arrays of various surface Figure 2 Device performance under AM 1.5G illumination. a, Traces of current areas and tested in simulated sunlight. At a full Sun intensity of 100+3 mW cm-2, our highest-surface-area devices are characterized mall cell (active area: 0. 2 cm2?) shows a higher V and J than the large cell byl=53-5.85 mA cm2,v=0.61-0.71V,FF=0.36-0.38and (0.8 cm?). The ill factor and efficiency are 0.37 and 1.51% and 0.38 and 1.26%, efficiency n=1. 2-1.5%.(Fig 2a)The external quantum efficiency respectively. Inset, the extemal quantum efficiency against wavelength for the large of these cells peaks at 40-43% near the absorption maximum of th cell. b, Open-circuit voltage and fill factor against light intensity, and c, short-circuit dye and is limited chiefly by the relatively low dye loadings of the current density and efficiency against light intensity for cells with roughness factors nanowire films. m 75 to 200. Each of the four parameters is represented by data from two Figure 2b and c shows the effect of light intensity on the different devices in order to provide an estimate of the range of their variability. In performance characteristics of the wire cells. The open-circuit general, cells of high roughness factor have low V- and ill factor, but high J_ and voltage and short-circuit current depend logarithmically and efficiency(see Fig. S7). Note that each of the eight plots is taken from a different inearly on light flux, respectively. The fill factors are low compared with nanoparticle cells, do not vary with cell size(see Fig S4), and fall off with increasing light intensity owing to the development of a large photo-shunt of unknown origin. These poor fill factors halve the potential efficiency of our best nanowire cells, and they ar robust with respect to changes in the nanowire electrical properties (Fig. S5), electrolyte concentration and choice of substrate(Fto or naturematerialsIVol4iJuNe2005iwww.nature.com/naturematerial @2005 Nature Publishing Group
LETTERS nature materials | VOL 4 | JUNE 2005 | www.nature.com/naturematerials 457 electrolyte (Fig. S3). Thus, facile transport through the nanowire array is retained in device-like environments and should result in faster carrier extraction in the nanowire cell. Solar cells were constructed with wire arrays of various surface areas and tested in simulated sunlight. At a full Sun intensity of 100 ± 3 mW cm–2, our highest-surface-area devices are characterized by Jsc = 5.3–5.85 mA cm–2, Voc = 0.61–0.71 V, FF = 0.36–0.38 and effi ciency η = 1.2–1.5%. (Fig. 2a) The external quantum effi ciency of these cells peaks at 40–43% near the absorption maximum of the dye and is limited chiefl y by the relatively low dye loadings of the nanowire fi lms. Figure 2b and c shows the effect of light intensity on the performance characteristics of the wire cells. The open-circuit voltage and short-circuit current depend logarithmically and linearly on light fl ux, respectively. The fi ll factors are low compared with nanoparticle cells, do not vary with cell size (see Fig. S4), and fall off with increasing light intensity owing to the development of a large photo-shunt of unknown origin. These poor fi ll factors halve the potential effi ciency of our best nanowire cells, and they are robust with respect to changes in the nanowire electrical properties (Fig. S5), electrolyte concentration and choice of substrate (FTO or 0.2 cm2 0.8 cm2 515 nm 400 500 600 700 800 Wavelength (nm) 40 30 20 10 EQE (%) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Bias (V) 6 5 4 3 2 1 0 Current density (mA cm–2) 0.75 0.70 0.65 0.60 0.55 0.50 0.45 0.40 0.35 VOC and fill factor VOC FF 0 20 40 60 80 100 Light intensity (mW cm–2) 1.6 0 20 40 60 80 100 Light intensity (mW cm–2) 6 4 2 0 JSC (mA cm–2) JSC η η 1.4 1.2 1.0 0.8 0.6 0.4 0.2 Efficiency (%) a b c Figure 2 Device performance under AM 1.5G illumination. a, Traces of current density against voltage (J–V) for two cells with roughness factors of ~200. The small cell (active area: 0.2 cm2 ) shows a higher Voc and Jsc than the large cell (0.8 cm2 ). The fi ll factor and effi ciency are 0.37 and 1.51% and 0.38 and 1.26%, respectively. Inset, the external quantum effi ciency against wavelength for the large cell. b, Open-circuit voltage and fi ll factor against light intensity, and c, short-circuit current density and effi ciency against light intensity for cells with roughness factors from 75 to 200. Each of the four parameters is represented by data from two different devices in order to provide an estimate of the range of their variability. In general, cells of high roughness factor have low Voc and fi ll factor, but high Jsc and effi ciency (see Fig. S7). Note that each of the eight plots is taken from a different device. Active cell size: 0.8 cm2 . 0 200 400 600 800 1,000 1,200 Roughness factor 12 10 8 6 4 2 0 JSC (mA cm–2) 0 3 7 10 17 13 20 Dye loading (moles x 10–8 per cm2 of substrate) 2.2 um 18–24 μm 52 μm 4.4 μm ZnO wires ZnO partcles, large ZnO partcles, small TiO2 particles, small Figure 3 Comparative performance of nanowire and nanoparticle cells. Shortcircuit current density versus roughness factor for cells based on ZnO wires, small TiO2 particles, and large and small ZnO particles. The TiO2 fi lms show a higher maximum current than either of the ZnO fi lms and a larger initial slope than the small ZnO particles, consistent with better transport through TiO2 particle networks. The large ZnO particle cells attain a smaller maximum current than the small particles because the fi lm thickness (and therefore the electron escape length) becomes larger than the electron diffusion length at a much lower roughness factor. The wire data fall on the TiO2 line and signifi cantly exceed the current output from both types of ZnO particle cells above a roughness factor of ~100. A slight sag of the wire data off the TiO2 line at high roughness factor may be a sign of excessive scattering within the opaque wire fi lms. Cell thickness is directly proportional to roughness factor and is labelled for each cell type at a roughness factor of 200. Error bars are provided on only two points to maximize fi gure clarity, and they are an estimate of the maximum range of the values. The error bars for cells with roughness factors below 250 are smaller than the size of the data points. Data points were made by measuring the roughness factor of dye-sensitized fi lms through ultraviolet–visible spectroscopy of the desorbed dye in basic H2O and then re-sensitizing the fi lms for fabrication into cells. See also Fig. S6. Fig. S7 shows comparative trends in η, Voc and fi ll factor. Cell size: 0.8 cm2 . nmat1387-print.indd 457 mat1387-print.indd 457 10/5/05 3:46:24 pm 0/5/05 3:46:24 pm ©2005 NaturePublishingGroup © 2005 Nature Publishing Group
LETTERS 0.20 thick enough to support the sort of radial electric field (depletion layer)that is impossible in smaller TiO2 or ZnO nanoparticles with fewer carriers. This upward band bending at the nanowire surface should suppress recombination by corralling injected electrons within the wire cores. At the same time, an axial field along each 油订水 nanowire encourages carrier motion towards the external circuit. These macroscopic fields should act synergistically to increase 0.10 electron transport relative to nanoparticle cells, which lack such fields Ambipolar diffusion is consequently a less dominant mechanism in the nanowire devices A switch from particles to wires also affects the kinetics of Time delay (ps) charge transfer at the dye-semiconductor interface, as particle and wire films have dissimilar surfaces onto which the sensitizing dye adsorbs. Whereas Zno particles present an ensemble of surfaces having various bonding interactions with the dye, our wire arrays are dominated by a single crystal plane(the [100))that accounts for over 95% of their total area. We used femtosecond transient absorption spectroscopy to measure the rate of electron injection from photoexcited ruthenium dyes into nanowire and nanoparticle films. Dye-sensitized samples were excited with 400-nm, 510-nn or 570-nm pulses and the free carrier concentration of the oxide Figure 4 Transient mid-infrared absorption traces of dye-sensitized Zno nanowire was monitored with a mid-infrared probe(see Figs S8 and S9).The NW)and zno nanoparticle(NP) films pumped at 400 nm. The large difference in transient responses for wires and particles(Fig 4)were considerably injection amplitudes is due to the larger surface area of the particle film Injection in different. Injection in wires was characterized by bi-exponential wires is complete after-5 ps but continues for -100 ps in the particle case. A high- resolution trace (inset shows the ultrafast step (250 fs)and -3ps rise tme for a whereas the particle response was tri-exponential and significantly slower(time constants: 25 mA cm-2). In contrast, for other types of excitonic photocells, such as inorganic- the rapid saturation and subsequent decline of the current from cells hybrid devices, in which an oriented, continuous and built with 12-nm TiO2 particles, 30-nm Zno particles or 200-nm inorganic phase of the proper dimensions could greatly improve the Zno particles confirms that the transport efficiency of particle collection of both electrons and holes. films falls off above a certain film thickness, as we argued above. Crucially, the nanowire films show a nearly linear increase in s that maps almost directly onto the TiO2 data. Because transport in the METHODS hin TiO2 particle films is very efficient(with Js =7.8-8.7 mA cm-2 at a roughness factor of 250), this is strong evidence of an equally array of SYNTHESIS OF NANOWIRE ARRAYS addition, the nanowire cells generate considerably higher currents hlm of Zno quantum dots, 3-4 nm in diameter, by dip-coating in a concentrated ethanol solution. than either of the Zno particle cells over the accessible range of Nanowires were grown by immersing seeded roughness factors(55-75% higher at a roughness of 200). This is molecular weight, Aldrich) at 92"C for 25 hours. Because nanowire growth slowed after this period. direct confirmation of the superiority of the nanowire photoanode substrates were repeatedlly introduced to fresh solutio reaction time of up to 50 hours). The arrays were then rinsed with deionized water and baked in air at Better electron transport within the nanowire photoanode is a 00"C for 30 minutes to remove any residual organics and to optimize cell performance. product of both its higher crystallinity and an internal electric field SOLAR CELL FABRICATION AND CHARACTERIZATION Nanowire arrays were first sensitized in a solution(0.5 mmolF)of(Bu, N), Ru(dcbpyH)(NCS)(N71 rom the surrounding electrolyte and sweeping them towards the如知由抛 Dupont). The internal space of the cell ( roughly one-third of the thickness of the space-charge layer in the)b5AML须mM05M4由如3m学由 collecting electrode. The Debye-Huckel screening length of Zno was illed witha liqui semiconductor at the semiconductor-electrolyte junction) is about dificiency (eoe) values t 4 nm for a carrier concentration of 108 cm3, making our nanowires semon Lmp coupled to a monochromator, and calibrated with a silicon photodiode. naturematerialsVol.4IJuNe2005iwww.nature.com/naturematerials @2005 Nature Publishing Group
LETTERS 458 nature materials | VOL 4 | JUNE 2005 | www.nature.com/naturematerials indium tin oxide). The effi ciency of our devices is fairly fl at above a power density of ~5 mW cm–2. To assess the relative effi ciency with which carriers are extracted from the nanowire devices, we compare in Fig. 3 the shortcircuit current densities of the wire cells to those of TiO2 and ZnO nanoparticle cells as a function of the internal surface area (roughness factor). A hypothetical photoanode that maintained a near-unity carrier collection effi ciency independent of roughness factor would trace out a line in this plot that gradually tapered off at high surface areas to a large Jsc value (>25 mA cm–2). In contrast, the rapid saturation and subsequent decline of the current from cells built with 12-nm TiO2 particles, 30-nm ZnO particles or 200-nm ZnO particles confi rms that the transport effi ciency of particle fi lms falls off above a certain fi lm thickness, as we argued above. Crucially, the nanowire fi lms show a nearly linear increase in Jsc that maps almost directly onto the TiO2 data. Because transport in the thin TiO2 particle fi lms is very effi cient (with Jsc = 7.8–8.7 mA cm–2 at a roughness factor of 250), this is strong evidence of an equally high collection effi ciency for nanowire fi lms as thick as ~25 μm. In addition, the nanowire cells generate considerably higher currents than either of the ZnO particle cells over the accessible range of roughness factors (55–75% higher at a roughness of 200). This is direct confi rmation of the superiority of the nanowire photoanode as a charge collector. Better electron transport within the nanowire photoanode is a product of both its higher crystallinity and an internal electric fi eld that can assist carrier collection by separating injected electrons from the surrounding electrolyte and sweeping them towards the collecting electrode. The Debye–Hückel screening length of ZnO (roughly one-third of the thickness of the space-charge layer in the semiconductor at the semiconductor–electrolyte junction) is about 4 nm for a carrier concentration of 1018 cm-3, making our nanowires thick enough to support the sort of radial electric fi eld (depletion layer) that is impossible in smaller TiO2 or ZnO nanoparticles with fewer carriers. This upward band bending at the nanowire surface should suppress recombination by corralling injected electrons within the wire cores. At the same time, an axial fi eld along each nanowire encourages carrier motion towards the external circuit. These macroscopic fi elds should act synergistically to increase electron transport relative to nanoparticle cells, which lack such fi elds. Ambipolar diffusion is consequently a less dominant mechanism in the nanowire devices. A switch from particles to wires also affects the kinetics of charge transfer at the dye–semiconductor interface, as particle and wire fi lms have dissimilar surfaces onto which the sensitizing dye adsorbs. Whereas ZnO particles present an ensemble of surfaces having various bonding interactions with the dye, our wire arrays are dominated by a single crystal plane (the {100}) that accounts for over 95% of their total area. We used femtosecond transient absorption spectroscopy to measure the rate of electron injection from photoexcited ruthenium dyes into nanowire and nanoparticle fi lms. Dye-sensitized samples were excited with 400-nm, 510-nm or 570-nm pulses and the free carrier concentration of the oxide was monitored with a mid-infrared probe (see Figs S8 and S9). The transient responses for wires and particles (Fig. 4) were considerably different. Injection in wires was characterized by bi-exponential kinetics with time constants of less than 250 fs and around 3 ps, whereas the particle response was tri-exponential and signifi cantly slower (time constants: <250 fs, 20 ps, 200 ps). Our data on particle injection are in excellent agreement with published results25, validating our evidence for faster electron injection in nanowires. The nanowire dye-sensitized solar cell is an exciting variant of the most successful of the excitonic photovoltaic devices. As an ordered topology that increases the rate of electron transport, a nanowire electrode may provide a means to improve the quantum effi ciency of DSCs in the red region of the spectrum, where their performance is currently limited. Important differences in transport, internal electric fi eld distribution and light scattering should make comparative studies of wire and particle devices fruitful. Raising the effi ciency of the nanowire cell to a competitive level depends on achieving higher dye loadings through an increase in surface area. We are now extending our synthetic strategy to design nanowire electrodes with much larger areas available for dye adsorption. The advantages of the nanowire geometry are even more compelling for other types of excitonic photocells, such as inorganic–polymer hybrid devices26, in which an oriented, continuous and crystalline inorganic phase of the proper dimensions could greatly improve the collection of both electrons and holes. METHODS SYNTHESIS OF NANOWIRE ARRAYS Arrays of ZnO nanowires were synthesized on FTO substrates (TEC-7, 7 Ω per square), Hartford Glass Co.) that were fi rst cleaned thoroughly by acetone/ethanol sonication and then coated with a thin fi lm of ZnO quantum dots, 3–4 nm in diameter, by dip-coating in a concentrated ethanol solution. Nanowires were grown by immersing seeded substrates in aqueous solutions containing 25 mM zinc nitrate hydrate, 25 mM hexamethylenetetramine and 5–7 mM polyethylenimine (branched, low molecular weight, Aldrich) at 92 °C for 2.5 hours. Because nanowire growth slowed after this period, substrates were repeatedly introduced to fresh solution baths in order to obtain long wire arrays (total reaction times of up to 50 hours). The arrays were then rinsed with deionized water and baked in air at 400 °C for 30 minutes to remove any residual organics and to optimize cell performance. SOLAR CELL FABRICATION AND CHARACTERIZATION Nanowire arrays were fi rst sensitized in a solution (0.5 mmol l–1)of (Bu4N)2Ru(dcbpyH)2(NCS)2 (N719 dye) in dry ethanol for one hour and then sandwiched together and bonded with thermally platinized FTO counter electrodes separated by 40-μm-thick hot-melt spacers (Bynel, Dupont). The internal space of the cell was fi lled with a liquid electrolyte (0.5 M LiI, 50 mM I2, 0.5 M 4-tertbutylpyridine in 3-methoxypropionitrile (Fluka)) by capillary action. Cells were immediately tested under AM 1.5G simulated sunlight (300 W Model 91160, Oriel). Intensity measurements were made with a set of neutral density fi lters. External quantum effi ciency (EQE) values (uncorrected for transmission and refl ection losses) were obtained with a 150-W xenon lamp coupled to a monochromator, and calibrated with a silicon photodiode. 0 5 10 15 Time delay (ps) 0.04 0.03 0.02 0.01 0 Absorbance 0 20 40 60 80 100 Time delay (ps) 0.20 0.15 0.10 0.05 0 Absorbance NP NW Figure 4 Transient mid-infrared absorption traces of dye-sensitized ZnO nanowire (NW) and ZnO nanoparticle (NP) fi lms pumped at 400 nm. The large difference in injection amplitudes is due to the larger surface area of the particle fi lm. Injection in wires is complete after ~5 ps but continues for ~100 ps in the particle case. A highresolution trace (inset) shows the ultrafast step (<250 fs) and ~3 ps rise time for a nanowire sample. The slower time constant showed a weak dependence on pump wavelength (see Figs S8 and S9). Particles were synthesized6 and fi lms were prepared25 (using dye N719) as described elsewhere. Films were deposited on Al2O3 substrates. Spectra are offset by ~0.05 absorbance units for clarity. nmat1387-print.indd 458 mat1387-print.indd 458 10/5/05 3:46:24 pm 0/5/05 3:46:24 pm ©2005 NaturePublishingGroup © 2005 Nature Publishing Group
LETTERS Identical procedures were used to build and test DSCs based on TiO2 and ZnO partide films prepared 7. Kruger, L, Plass, R, Gratzel, M Cameron, P L Peter, L M. Charge transport and back reaction spectroscopy. I Plys. Chen. B 107, 7536-7539(2003). Th a 0.2M ously, sintered a second 8. ORean. B Gratzel, M. A low-cost, high-efficiency solar cell based on dye-sensitized colloidal TiO, films. Nature 353, 737-740(1991) and 9. Fisher, A C, Peter, L M, Ponomarev, E.A. walker, A B. wijayantha, K.G. U Intensity dependence of the back reaction and transport of electrons in dye-sensitized nanocrystalline TiO, ater All films were free of cracks. The small Zno particles 000) synthesized by heating 0.8 g zinc acetate dihydrate and 50 ml ethanol in an autosave at 125"Cfor 0, Oekermann, T, Zhang, D, Yoshida, T. Minoura, H Electron transport and back reaction 2 hours. The large particles were obtained as a commercial powder(200 mesh, 99.9999, Cerac) nanocrystalline TiO, films prepared by hydrothermal crystallization. J. Phys. Chen ELECTRICAL MEASUREMENS ng were dispersed from eth 11. Nelson, I. Continuous-time random-walk model of electron transport in nanocrystalline TiO, used to pattern and deposit contacts( 100 nm Ti) lnking the wires to prefabricated electroc 12. van de Lagemaat. ].& Frank, A ]. Nonthermalized electron transport in dye-sensitized nanocrystalline TiO, films transient photocurrent and random-walk modeling studies /. Phys. bbal back gate using a se r parameter analyser(4145B, Hewlett-Packard). samples ansport studies were made by encapsulating fired arrays (grown on FIO) in a matrix of spin-cast 13. Kopidakis, N, Schiff, E.A. Park, N -G, van de Lagemaat. L& Frank, A. J- Ambipolar diffusion of photocarriers in electrolyte-filled, nanoporous TO./. Phys. Chern. B. 104, 39 L Benkstein, K D, Kopidakis, N, van de Lagemaat, 1. Frank, A l Influence of the network geometry on electron transport in dye-sensitized titanium dioxide solar cells. I. Phys. Chem. rovided mechanical stability for the B.107,7759-7767(2003) Kopidakis, N Benkstein, MID-INFRARED TRANSIENT ABSORPTION MEASUREMENTS of photocarriers in dye-sensitized nanocrystalline TiO solar cells. J Phys. Chem. B 10 I kHz repetition 16. Kavan, L Gratzel, M Gilbert, S. E, Klemenz, C& Schell, H. L Electrochemical and rate. About 800 I of the beam was used to pump an optical parametric amplifier(TOPAS, Quanto photoelectrochemical investigation of single-crystal anatase. J. Am. Chem. Soc. 118,6716-6723 while 80 1 was retained and frequency-doubled in B-barium borate( BBO) for us beam. This beam was delayed by a motorized stage and directed to the sample. The signal and idler bean 17. Wagner, P& Helbig, R. Hall effect and anisotropy of the mobility of the electrons in zine axi . Phys. Cem. Sol. 35, 321 alses(1,000-3, 500 cm). The residual 810-nm beam and the residual signal a &N preparation methods and annealing temperatures nsitized solar cells. L Plys. Chem. B 106, 1000---10010(200 was directed to a separate delay stage and then to the sample. The pump beams were focused to a spot siae of 19. Frank, A I, Kopidakis, N&van de Lagemaat.).Electrons 102 solar cells pump energies of0.s2以 55-1179(2001) 20. Rene ligands: synthesis, characterization, and INDO/S analysis. Inorg Chem. 41, 367-378(2002) ump(reference). The sample signal was subtracted from the reference signal, and the result was divided narin dyes having thiophene moieties for highly efficient organic the reference to gne the differential transmittance, which was converted to effective absorbance. Th dye-sensitized solar cells. New I. Chem. 27, 783-785(2003). 22. Kron, G, Eserter, I, werner J. H. Ra orousTIO, sola Is-compuarison of electrolyte and solid-state devices. J. Phys. Chem. B 107, 3556-3564(2003). sample and directed through bandpass filte being focused onto a single-dement Hg Cdle detecto 23. Greene, Let al Low-temperature wafer scale production of Zno nanowire arrays. Angew. Chem. Int L Noack, V, weller, H. Eychmuller, A Electron transport in particulate ZnO electrodes: a simple both forward and reverse scans, checked for approach I Phys. Chem. B 106, 8514-8523(2002). each scan to minimize probing dye photoproducts. They were not mowed during the scan because small ZnO nanocrystalline thin films. 1. Phys. Chem. B107, 14414-14021(2003). inhomogeneities caused changes in the amplitude of the transient signal, obscuring the 26. Huynh, w. U, Dittmer, LL Alnvisatos, A P Hybrid nmanorod-polymer solar cells. Sciemce 295. Received 4 March 2005; accepted 31 March 2005; published 15 May 2005 27. Park, N.-G. et aL Morphological and photoelectrochemical characterization of core-shell nanoparticle films for dye-sensitized solar cells: Zn-O type shell on Snoz and TiO cores Langar I. Gregg, B. A. Excitonic solar cells. / Phys. Chen. B107, 2 Nazeeruddin, M K. et al Engineering of efticient panchromatic sensitizers for nanocrystalline ThO r Acknowledgements 3. Wang. P et al. A stable quasi-s0 solar cell with an amphiphilic ruthenium We thank M. Graetnel, A P Alivisatos, L Frechet. B. O'Regan, E. Kadnakowa, U. Bach, D Milliron and L Gur for discussions, T Lavarone and S. Hammzehpour for technical assistance and A P. Alivisatos for of the solar simulator. This work was supported by the US Department of Energy, Office of Basic 5. Tennakone, K. Kumara, GR.RA, Kottegoda, L. R.M.& Perera, V.E. S An efficient dye-sensitized Correspondence and requests for materals should be addressed to pY cell made from oxides of tin and zinc. SupplementaryiNformationaccompaniesthepaperonwww.nature.com/naturematerials. electrochemical solar cell based on nanostructured Zno electrodes. Sol Emergy Mater. Sol. Competing financial interests Cc73,51-58(2002) naturematerialsIVol4iJuNe2005iwww.nature.com/naturematerial @2005 Nature Publishing Group
LETTERS nature materials | VOL 4 | JUNE 2005 | www.nature.com/naturematerials 459 Identical procedures were used to build and test DSCs based on TiO2 and ZnO particle fi lms prepared by spin-coating or spreading pastes with a thin glass rod (doctor-blading). Films of TiO2 made from a commercial paste of 10–15-nm anatase crystals (Ti-Nanoxide T, Solaronix) were sintered at 450 °C for 30 minutes, treated with a 0.2 M aqueous TiCl4 solution for 12 hours as described previously2 , sintered a second time at 450 °C for 30 minutes and sensitized with dye for 24 hours. Pastes of small, spherical ZnO particles (30 ± 14 nm) and large, irregular ZnO particles (200 ± 75 nm) were formulated as described previously27, and sintered and sensitized similarly to the nanowire cells. Film thickness was varied by using different spacers for doctor-blading and/or by diluting the pastes with water. All fi lms were free of cracks. The small ZnO particles were synthesized by heating 0.8 g zinc acetate dihydrate and 50 ml ethanol in an autoclave at 125 °C for 2 hours. The large particles were obtained as a commercial powder (200 mesh, 99.999%, Cerac). ELECTRICAL MEASUREMENTS For the single wire studies, nanowires 8–10 μm long were dispersed from ethanol solution on oxidized silicon substrates (300 nm SiO2) and fi red in air at 400 °C for 30 minutes. Electron-beam lithography was used to pattern and deposit contacts (100 nm Ti) linking the wires to prefabricated electrode sets. Most devices showed ohmic I–V plots without annealing treatments. Measurements were made with a global back gate using a semiconductor parameter analyser (4145B, Hewlett-Packard). Samples for array transport studies were made by encapsulating fi red arrays (grown on FTO) in a matrix of spin-cast poly(methylmethacrylate) (PMMA), exposing the wire tips by ultraviolet development and dissolution of the top portion of the PMMA fi lm, and then depositing metal contacts by thermal evaporation. The insulating PMMA matrix prevented potential short circuits due to pinholes in the nanowire array and provided mechanical stability for the measurement. MID-INFRARED TRANSIENT ABSORPTION MEASUREMENTS Transient absorption measurements were made with a home-built Ti:sapphire oscillator (30 fs, 88 MHz) and commercial regenerative amplifi er (Spitfi re, Spectra-Physics) that operates at 810 nm and 1 kHz repetition rate. About 800 μJ of the beam was used to pump an optical parametric amplifi er (TOPAS, Quantronix), while 80 J was retained and frequency-doubled in β-barium borate (BBO) for use as the 405-nm pump beam. This beam was delayed by a motorized stage and directed to the sample. The signal and idler beams from the optical parametric amplifi er were combined in a AgGaS2 crystal to create tuneable mid-infrared pulses (1,000–3,500 cm–1). The residual 810-nm beam and the residual signal and idler beams were recombined in a BBO crystal to create sum-frequency generation at 510 nm and 575 nm. The 510-nm beam was directed to a separate delay stage and then to the sample. The pump beams were focused to a spot size of roughly 200–300 μm, with typical pump energies of 0.5–2 μJ. The pump beams were mechanically chopped at 500 Hz (synchronous with the laser), and separate boxcar integrators were triggered by the rejected and passed beams, allowing for independent detection channels of probe with pump (‘sample’) and without pump (‘reference’). The sample signal was subtracted from the reference signal, and the result was divided by the reference to give the differential transmittance, which was converted to effective absorbance. The probe beam, which was typically centred at 2,150 cm–1 with a bandwidth of 250 cm–1, was focused with a CaF2 lens to a size of roughly 100–200 μm. The probe beam was collected after transmission through the sample and directed through bandpass fi lters before being focused onto a single-element HgCdTe detector (IR Associates). An instrument response of 250–300 fs was determined by measuring the rise of free-electron absorption (less than 50 fs) in a thin silicon wafer after blue or green pump. Each transient plot is an average of points taken on both forward and reverse scans, checked for reproducibility. Each point consists of about 500 averaged laser shots. Samples were translated after each scan to minimize probing dye photoproducts. They were not moved during the scan because small inhomogeneities caused changes in the amplitude of the transient signal, obscuring the true kinetics. Received 4 March 2005; accepted 31 March 2005; published 15 May 2005. References 1. Gregg, B. A. Excitonic solar cells. J. Phys. Chem. B 107, 4688–4698 (2003). 2. Nazeeruddin, M. K. et al. Engineering of effi cient panchromatic sensitizers for nanocrystalline TiO2- based solar cells. J. Am. Chem. Soc. 123, 1613–1624 (2001). 3. Wang, P. et al. A stable quasi-solid-state dye-sensitized solar cell with an amphiphilic ruthenium sensitzer and polymer gel electrolyte. Nature Mater. 2, 402–407 (2003). 4. Rensmo, H. et al. High light-to-energy conversion effi ciencies for solar cells based on nanostructured ZnO electrodes. J. Phys. Chem. B 101, 2598–2601 (1997). 5. Tennakone, K., Kumara, G. R. R. A., Kottegoda, I. R. M. & Perera, V. P. S. An effi cient dye-sensitized photoelectrochemical solar cell made from oxides of tin and zinc. Chem. Commun. 15–16 (1999). 6. Keis, K., Magnusson, E., Lindström, H., Lindquist, S.-E. & Hagfeldt, A. A 5% effi cient photoelectrochemical solar cell based on nanostructured ZnO electrodes. Sol. Energy Mater. Sol. Cells 73, 51–58 (2002). 7. Krüger, J., Plass, R., Grätzel, M., Cameron, P. J. & Peter, L. M. Charge transport and back reaction in solid-state dye-sensitized solar cells: a study using intensity-modulated photovoltage and photocurrent spectroscopy. J. Phys. Chem. B 107, 7536–7539 (2003). 8. O’Regan, B. & Grätzel, M. A low-cost, high-effi ciency solar cell based on dye-sensitized colloidal TiO2 fi lms. Nature 353, 737–740 (1991). 9. Fisher, A. C., Peter, L. M., Ponomarev, E. A., Walker, A. B. & Wijayantha, K. G. U. Intensity dependence of the back reaction and transport of electrons in dye-sensitized nanocrystalline TiO2 solar cells. J. Phys. Chem. B 104, 949–958 (2000). 10. Oekermann, T., Zhang, D., Yoshida, T. & Minoura, H. Electron transport and back reaction in nanocrystalline TiO2 fi lms prepared by hydrothermal crystallization. J. Phys. Chem. B 108, 2227–2235 (2004). 11. Nelson, J. Continuous-time random-walk model of electron transport in nanocrystalline TiO2 electrodes. Phys. Rev. B 59, 15374–15380 (1999). 12. van de Lagemaat, J. & Frank, A. J. Nonthermalized electron transport in dye-sensitized nanocrystalline TiO2 fi lms: transient photocurrent and random-walk modeling studies. J. Phys. Chem. B 105, 11194–11205 (2001). 13. Kopidakis, N., Schiff, E. A., Park, N.-G., van de Lagemaat, J. & Frank, A. J. Ambipolar diffusion of photocarriers in electrolyte-fi lled, nanoporous TiO2. J. Phys. Chem. B. 104, 3930–3936 (2000). 14. Benkstein, K. D., Kopidakis, N., van de Lagemaat, J. & Frank, A. J. Infl uence of the percolation network geometry on electron transport in dye-sensitized titanium dioxide solar cells. J. Phys. Chem. B. 107, 7759–7767 (2003). 15. Kopidakis, N., Benkstein, K. D., van de Lagemaat, J. & Frank, A. J. Transport-limited recombination of photocarriers in dye-sensitized nanocrystalline TiO2 solar cells. J. Phys. Chem. B 107, 11307–11315 (2003). 16. Kavan, L., Grätzel, M., Gilbert, S. E., Klemenz, C. & Schell, H. J. Electrochemical and photoelectrochemical investigation of single-crystal anatase. J. Am. Chem. Soc. 118, 6716–6723 (1996). 17. Wagner, P. & Helbig, R. Hall effect and anisotropy of the mobility of the electrons in zinc oxide. J. Phys. Chem. Sol. 35, 327–335 (1974). 18. Nakade, S. et al. Dependence of TiO2 nanoparticle preparation methods and annealing temperatures on the effi ciency of dye-sensitized solar cells. J. Phys. Chem. B 106, 10004–10010 (2002). 19. Frank, A. J., Kopidakis, N. & van de Lagemaat, J. Electrons in nanostructured TiO2 solar cells: transport, recombination and photovoltaic properties. Coord. Chem. Rev. 248, 1165–1179 (2004). 20. Renouard, T. et al. Novel ruthenium sensitizers containing functionalized hybrid tetradentate ligands: synthesis, characterization, and INDO/S analysis. Inorg. Chem. 41, 367–378 (2002). 21. Hara, K. et al. Design of new coumarin dyes having thiophene moieties for highly effi cient organicdye-sensitized solar cells. New J. Chem. 27, 783–785 (2003). 22. Kron, G., Egerter, T., Werner, J. H. & Rau, U. Electronic transport in dye-sensitized nanoporous TiO2 solar cells—comparison of electrolyte and solid-state devices. J. Phys. Chem. B 107, 3556–3564 (2003). 23. Greene, L. et al. Low-temperature wafer scale production of ZnO nanowire arrays. Angew. Chem. Int. Edn Engl. 42, 3031–3034 (2003). 24. Noack, V., Weller, H. & Eychmüller, A. Electron transport in particulate ZnO electrodes: a simple approach. J. Phys. Chem. B. 106, 8514–8523 (2002). 25. Anderson, N. A., Ai, X. & Lian, T. Electron injection dynamics from Ru polypyridyl complexes to ZnO nanocrystalline thin fi lms. J. Phys. Chem. B 107, 14414–14421 (2003). 26. Huynh, W. U., Dittmer, J. J. & Alivisatos, A. P. Hybrid nanorod–polymer solar cells. Science 295, 2425–2427 (2002). 27. Park, N.-G. et al. Morphological and photoelectrochemical characterization of core–shell nanoparticle fi lms for dye-sensitized solar cells: Zn-O type shell on SnO2 and TiO2 cores. Langmuir 20, 4246–4253 (2004). Acknowledgements We thank M. Graetzel, A. P. Alivisatos, J. Frechet, B. O’Regan, E. Kadnikova, U. Bach, D. Milliron and I. Gur for discussions, T. Lavarone and S. Hamzehpour for technical assistance and A. P. Alivisatos for use of the solar simulator. This work was supported by the US Department of Energy, Offi ce of Basic Sciences. Correspondence and requests for materials should be addressed to P.Y. Supplementary Information accompanies the paper on www.nature.com/naturematerials. Competing fi nancial interests The authors declare that they have no competing fi nancial interests. nmat1387-print.indd 459 mat1387-print.indd 459 10/5/05 3:46:25 pm 0/5/05 3:46:25 pm ©2005 NaturePublishingGroup © 2005 Nature Publishing Group