and Orientation-Dependent Photovoltaic Properties of ZnO Nanorods

Oct 16, 2007 - investigated by a lock-in amplifier with dc bias and Kelvin probe (KP) based measurements. The kinetic features of SPV responses are ...
0 downloads 0 Views 437KB Size
Download PDF
17136

J. Phys. Chem. C 2007, 111, 17136-17145

Size- and Orientation-Dependent Photovoltaic Properties of ZnO Nanorods Qidong Zhao, Tengfeng Xie, Linlin Peng, Yanhong Lin, Ping Wang, Liang Peng, and Dejun Wang* College of Chemistry, Jilin UniVersity, Changchun 130012, People’s Republic of China ReceiVed: July 10, 2007; In Final Form: September 1, 2007

ZnO nanorod arrays on an ITO substrate and nanorod powder have been prepared via a chemical method in aqueous solution at low temperature. Two dimensions of composite nanorods in the arrays were obtained by controlling the reaction time. SEM, XRD, UV-vis transmission, and PL measurements have been utilized to characterize the samples. The surface photovoltage (SPV) spectra of the three samples have been comparatively investigated by a lock-in amplifier with dc bias and Kelvin probe (KP) based measurements. The kinetic features of SPV responses are interpreted in terms of ac SPV phase spectra and SPV transients on a KP. We demonstrate that the photovoltaic properties of ZnO nanorods not only depend on the rod size, but also rely on the crystallographic orientation. The mechanisms therein have been discussed in detail. Our results could lead to better understanding of the photovoltaic properties in ZnO nanostructures.

1. Introduction In recent years, considerable attention has been paid to nanostructured ZnO material, which is mostly inspired by its diverse functional properties in mechanics, optics, electronics, magnetics, and catalysis, as well as the accessibility to facile fabrication of nanocrystals with various shapes (such as ball, belt, rod/wire, ring, tube, box, comb, and flower, etc.) and controlled dimensions (from quantum dots to three-dimensional networks).1-5 Bulk ZnO material has a direct wide band gap (Eg ≈ 3.37 eV, 368 nm) and a large exciton binding energy (∼60 meV) at room temperature.1 For ZnO systems with diverse nanostructures, influences of shape and dimension on surfacerelated properties are significant. For example, one-dimensional ZnO nanostructures with controlled tip shapes show modulated visible photoluminescence,6-8 as well as tuned current rectification characteristics.9 ZnO nanorod arrays with similar grain sizes but different morphologies are discrepant on field emission.10 These modulated properties have been much attributed to the varying oxygen vacancy distribution at the grain surface determined by different shapes. New features coming with a reduced grain size, such as a widened band gap11 and enhanced exciton-related luminescence,12 are mainly due to the quantum size effect or confinement effect, and an anomalous blue shift in the emission spectra of ZnO nanorods with reduced sizes that are beyond the quantum confinement regime has been recently observed, which is ascribed to an enhanced surface effect due to a larger surface-to-volume ratio.13 Potentially improved performance of ZnO in sensors, transducers, catalysts, and solar cells can be expected with properly designed nanostructures.3 Therefore, it is of practical interest to study the correlation between various nanostructures and the novel properties of ZnO materials for their future applications in functional devices. One-dimensional nanostructures, i.e., nanorods/nanowires/ nanobelts, are the smallest dimension structures that can be used for efficient transport of electrons and optical excitations. They * To whom correspondence should be addressed. Phone: +86-43185168093. E-mail: wangdj@jlu. edu.cn.

exhibit interesting electronic and optical properties associated inherently with their low dimensionality and reduced size (or quantum confinement effect) in two dimensions relative to those of bulk materials.14,15 They are expected to play an important role as promising building blocks of both interconnects and functional units in fabricating novel electronic, optoelectronic, electrochemical, and electromechanical devices with nanoscale dimensions.16,17 The preparation of ZnO materials with onedimensional structure has been advanced maturely.1,18 The sizedependent properties of ZnO nanowires/nanorods such as electron-phonon coupling,19 surface luminescence,20 Young’s modulus,21 and gas sensitivity22 have been observed recently. Furthermore, the crystallographic orientation or anisotropic crystal structure of nanorods has also been proved to have appreciable influence on many aspects of their properties. By tuning the face orientation of ZnO crystals, optimized photocatalytic activity has been realized.23 The photoelectrochemical performance of sensitized ZnO has been found dependent on the crystalline orientation in electrodeposited thin films.24 Thus, for a certain width and aspect ratio of nanorods, besides the size effect, it is crucial to distinguish the anisotropic effects on the properties of interest for improved function and integration of nanoscale devices. Semiconductor nanowire arrays have been previously predicted to be potential substrates for improved photocatalysis and photovoltaics.25 The realization of a purpose-built array of anisotropic metal oxide material with low cost provides an opportunity to study the relation between the structure and properties of well-oriented ZnO nanorods.26 To date, nanorod arrays of ZnO have been found applicable in various photoelectric systems, such as nanolasers,27 Schottky diodes,28 homojunctions,29 p-n heterojunction devices,30 field-effect transistors,31 photonic crystals,32 solar cells,33 biosensors,34 photocatalysis,35 hybrid plasma display panels,36 and current nanogenerators.37 The comprehensive understanding of the mechanisms of photoelectric response and charge transport in the array structures with various sizes and morphologies is demanding for the design and optimization of the devices, and it is essential to explore the mechanisms such as how the surface

10.1021/jp075368y CCC: $37.00 © 2007 American Chemical Society Published on Web 10/16/2007

Photovoltaic Properties of ZnO Nanorods

J. Phys. Chem. C, Vol. 111, No. 45, 2007 17137 the size and the crystallographic orientation of the nanorods have distinct effects on the spectrum-dependent photovoltaic properties. 2. Experimental Section

Figure 1. SEM images of ZnO nanorods: (a) sample A; (b) sample B; (c) sample C. The scale bars in the three images are 1 µm, 200 nm, and 500 nm, respectively.

or interface potential in the material will change under different wavelengths of light excitation for better utilization of its photoelectric properties. The surface photovoltage (SPV) method is a well-established technique for the characterization of semiconductors, which provides both optical and transport properties of different regions in the material under study, with high sensitivity to defect states in the sample at its surface, bulk, or any buried interface.38 The detected SPV response depends on the light absorption, charge separation, and charge transport properties of the system under investigation. A Kelvin probe (KP), as the first elaborate equipment for SPV detection, measures the surface work function and follows the change in the surface potential under steady illumination, which is sensitive to surface charge changing due to either a fast or a slow process. The other is a lock-in amplifier combined with a light chopper, which tracks ac voltage from a fixed capacitor structure, and it has superior temporal resolution compared to that of the KP via the optimized light-modulating frequency. In addition, dc bias could be applied to the capacitor structure to obtain an enhanced SPV response and more useful information on the sample under investigation. The measured ac SPV response amplitude reflects the mean potential change between the electrodes during a light-chopping period. This method has been shown effective in investigations on azo pigments,39 quantum structures,40 ZnO nanocrystals,41 Fe2O3 nanocrystals,42 and TiO2 powder.43 The phase value of the ac photovoltage signal correlated to the light modulation benefits analysis of the kinetics of the SPV generating process. SPV phase spectroscopy has been successfully applied for characterizing photogenerated charges in wafers and quantum well structures,44 and more recently, the two detecting techniques have both been utilized to study the photogenerated charges in porphyrin/TiO2-C6045 and porphyrin/Au46,47 systems. In a recent study, we prepared a ZnO nanorod array with a large grain size on an ITO substrate chemically and characterized the transfer behavior of photogenerated charges at the surface of nanorods and the interface of ZnO/ITO.48 In this study, we further investigate the photovoltaic properties of ZnO nanorods with different sizes or crystallographic orientations with respect to the probing electrode, including the SPV developing kinetics. The lock-in amplifier with a dc bias and KP-based SPV measurements were both carried out. We demonstrate that both

The preparation of samples was in light of refs 18 and 26. A piece of cleaned ITO glass (∼2 cm2) was wetted on the single conductive side with a droplet of 0.005 M zinc acetate dihydrate, Zn(Ac)2‚2H2O (from Aldrich), in ethanol, rinsed with clean ethanol after 10 s, and then blown dry with air. The above procedure was repeated three times. Then the ITO was heated to 350 °C in air for 15 min. After cooling, the ITO with ZnO seeds was put into a fresh aqueous solution of 0.1 M equimolar zinc acetate and methenamine, C6H12N4 (from Aldrich), in an autoclave, which was then placed in a regular laboratory oven and heated at 95 °C for 5 h, yielding the first sheet of ZnO nanorod arrays on ITO (sample A). Sample B is obtained by repeating the above procedure, but the heating duration at 95 °C is 40 min in another autoclave. Sample C is the white precipitate collected from the bottom of the same autoclave containing sample A. Subsequently, all three ZnO samples were thoroughly washed with ethanol and deionized water (with sample C centrifuged) and then dried in air at 50 °C separately. The morphology and dimension of the nanorods were observed on a scanning electron microscope (Shimadzu SSX550). UV-vis spectra were obtained on a CARY100 spectrophotometer. The PL spectrum of sample A was taken on a UVvis spectrophotometer (Edinburgh Analytical Instruments FS920). The XRD patterns of blank ITO and the samples were obtained on an XRD-6000, Shimadzu. The lock-in-based SPV measurement system is constituted of a source of monochromatic light, a lock-in amplifier (SR830DSP) with a light chopper (SR540), a photovoltaic cell, and a computer. A 500 W xenon lamp and a double-prism monochromator provide monochromatic light. A low chopping frequency of ∼23 Hz is used. The construct of the photovoltaic cell is a sandwich-like structure (see Figure 4). A spacer of mica (∼10 µm thick) was inserted between the sample and the counter electrode when sample A and sample B were measured, while for sample C, the powder sheet was directly sandwiched between two blank ITO electrodes. The effective overlapping area of the two electrodes in the test is ∼1 cm2 for all three samples. The potential of the irradiated electrode with respect to the back electrode denotes the signs of the applied dc bias between the electrodes. The back electrode was connected to the common grounding point of the lock-in amplifier. The correlated phase spectra were taken synchronously with SPV spectra by the computer, and the phase retardation value of zero corresponds to the light excitation starting time. Calibration of the system was done as in ref 44 to eliminate the possible phase shift that is not correlated to the SPV response, so that any phase retardation is only with respect to the phase of the light modulation and reflects the kinetics of the SPV response. For the ideal case of infinitely fast SPV formation processes, the SPV transient follows the excitation light sequence, which results in zero SPV phase retardation. At this time, a negative ac SPV response has a phase value of 180°, while a positive ac SPV response has a phase value of 0°, but in the practical case, the SPV transient will fall behind the excitation light sequence. For a practical negative SPV response, the slower monotonous onset of SPV response in a chopping period corresponds to the larger phase retardation with respect to 180°, i.e., smaller than 180° in value by less than 180° generally, and for a practical

17138 J. Phys. Chem. C, Vol. 111, No. 45, 2007

Zhao et al.

Figure 2. X-ray diffractogram of the samples: (a) blank ITO glass; (b) sample A; (c) sample B; (d) sample C. The main peaks belonging to ZnO are labeled.

positive SPV response, the slower monotonous onset of SPV response in a chopping period corresponds to the larger phase retardation with respect to 0°, i.e., generally smaller than 0° in value by less than 180°. KP-based SPV measurements were carried out on a commercial KP system (KP Technology Ltd., Scotland, U.K.). The width of the gold reference probe is 1.8 mm. The work function of the probe is employed as 5.1 eV after the manufacturer. The SPV was measured by tracking the contact potential difference (CPD) between the sample and the probe. The SPV spectra were obtained by scanning the monochromatic light through the visible and UV range with a rate of ∼30 nm/min. All the SPV measurements were operated under ambient conditions, and the raw data were not treated further. Constant light intensity at each wavelength was not used in the KP and ac SPV measurements, and the monochromic light intensity depended on the xenon lamp spectral energy distribution. The largest intensity of the incident monochromatic light is less than 80 µW/cm2. 3. Results and Discussion From SEM images (Figure 1), we know that ZnO nanorods ∼100 nm in width and ∼1.5 µm in length have grown vertically on the ITO substrate for sample A. Sample B is much smaller in width (about 60 nm). Sample C is a powder aggregate of disordered nanorods with a grain size similar to that of sample A, but with the rod axis much more parallel to the substrate. As indicated by the XRD patterns in Figure 2, vertically wellaligned ZnO nanorods with the c-axis perpendicular to the ITO substrate for samples A and B were obtained, and for sample C, the nanorods in the powder have adopted an orientation with the c-axis parallel to the substrate more than that in sample A or B, which is deduced from the predominant diffraction peaks corresponding to the (100) and (101) planes. This result is consistent with the observed SEM image of sample C. The UV-

vis transmission spectra in Figure 3a imply that, for sample A, only incident light above 370 nm can reach throughout the rods, whereas for sample B, the incident light above 330 nm can reach throughout the nanorods whether it is incident from the surface or interface. The PL pattern of sample A in Figure 3b shows the obvious visible photoluminescence and UV emission, which gives the evidence that the as-prepared nanorods are rich in surface defects such as oxygen vacancies.49 The lock-in-based ac SPV measurement was first carried out to explore the photovoltaic properties of the samples. The photovoltaic properties of sample A were examined, and the testing configuration in Figure 4a was applied first. The light was incident from the ITO substrate that the ZnO nanorods had grown on, and the results are shown in Figure 5. In Figure 5a, we have a weak SPV response when sample A is illuminated from the ZnO/ITO interface under zero bias, which features a sharp peak located at ∼375 nm. However, below 350 nm no SPV response for the band-to-band transition can be observed. The SPV peak position is ∼63 meV below the band gap (3.37 eV), and the response is ascribed to the dissociation of excitons via surface states or interface states. The explanation for the interesting phenomenon is as follows. Bulk ZnO exhibits a large absorption coefficient higher than 155 000 cm-1 at 350 nm and shorter wavelength, which corresponds to a small lightpenetrating depth of less than 65 nm, but a relatively small absorption coefficient of less than 7000 cm-1 exists at 383 nm and longer wavelength, which corresponds to a large penetrating depth of more than 1400 nm. In the spectral range from 383 to 350 nm, the absorption coefficient increases sharply.50 When it is incident from the ITO substrate, as the length of the ZnO nanorods in sample A is larger than the penetration depth of incident photons with wavelength below 350 nm, the light in that spectral range cannot reach the surface space charge layer at the free end of the ZnO nanorods, so the band-to-band transition-related SPV response cannot be yielded effectively

Photovoltaic Properties of ZnO Nanorods

J. Phys. Chem. C, Vol. 111, No. 45, 2007 17139

Figure 3. (a) UV-vis transmittance of the ITO substrate (curve A) and the ZnO nanorod array with the ITO substrate, sample A (curve B) and sample B (curve C). (b) PL emission of sample A excited at 325 nm.

Figure 4. Schematic of ac SPV measurement configurations for samples A and B: (a) The ZnO nanorod array is illuminated from the ITO substrate. (b) The array is illuminated from the ZnO nanorod surface.

due to the lack of the force for separating electrons and holes at the interface of the ZnO/ITO substrate. Also the ac SPV amplitude was enhanced only with increasing positive bias, because the macroscopic built-in field in the nanorods points from the substrate to the free ends of the nanorods, which is consistent with the direction of a positive electric field by external bias. In addition, the peak broadens toward the sub band gap spectral range with increasing positive bias, which implies that the SPV response coming from dissociated excitons could be affected by interface states.48 Under increasing negative bias to -1.0 V, the weak SPV response quenched and did not emerge again with opposite polarity. In Figure 5b, the corresponding phase spectra show the statistic kinetic characteristics for each ac SPV spectrum with a distinguishable response. For all the curves (A-E), the phase values are above zero, which means the photogenerated electrons generally transfer to the electrode from which light is incident. The pattern of curve A that phase retardation is gradually reduced with respect to 180° from 390 to 370 nm in the phase spectra is similar to the phase retardation regularity of the p-type Si wafer from sub band gap to super band gap energy reported in ref 44. Because of the nonlinear recombination of excess carriers in the material, an increasing absorption coefficient results in an increase of the excess carrier concentration at a given moment and a decrease of the momentary excess carrier lifetime. Then a shorter time is needed for excess carrier concentration to approach the balance value in a given lightexciting period. Consequently, increasing the absorption coefficient from 390 to 370 nm will reduce the phase retardation of the variations in ac SPV. The SPV intensity is a monotonous function of excess carrier concentration. In addition, the larger

absorption coefficient causes the light-excited region in ZnO nanorods to shrink toward the ITO substrate, so the distance that the charge has to pass is reduced, and the photocarriers could transfer to the ITO substrate in a shorter time, resulting in decreasing phase retardation. However, the pattern of curve B to curve E takes a reversed regularity from longer wavelength to 380 nm; i.e., increased phase retardation with respect to 180° is observed. The abnormal phenomenon implies some complex factors that control the separation and transport of photogenerated charges. A larger retardation means the ac SPV amplitude in one excitation period will take a longer time to reach the balance or saturated value (if possible), statistically. For curves B-D, the increased phase retardation takes place with enhanced SPV response under increased positive bias, within the spectral range of 405-380 nm. That means the SPV response is harder to saturate and enhances with illumination time within a chopping period when the positive bias is increased. When illuminated with the sub band gap light, nearly the whole part of each nanorod could be excited, which is evident from the UV-vis transmission spectrum. A certain amount of excitons could also be generated near the band gap and then dissociated, with the holes in excitons trapped by donor surface states, such as O2- adsorbed to the oxygen vacancies at the surface.51 The probability that electrons are depopulated from surface states as well as the dissociation of excitons increases under increasing positive bias. Then a longer time than that without bias is needed for photogenerated electrons to accumulate toward the interface and approach a saturated amount under positive bias. Furthermore, the positive bias could also act on the excited surface states in the spectral region between 390 and 405 nm. The external electric field could attract negative charge from ZnO to the ITO substrate, leading to enhanced SPV response nonlinearly with bias amplitude. Under the bias, the photogenerated electrons could overcome the trapping of surface states and move remotely to the substrate. Thus, the charge accumulates at the ITO substrate with time. Nevertheless, under a positive bias of 1.0 V, the SPV has begun to approach a saturated value in an illumination period. As observed in the corresponding phase spectrum of curve E in Figure 5b, the phase retardation starts to decrease with respect to 180° in the spectral range from 398 to 380 nm. The common characteristic of increased phase retardation from longer wavelength to 380 nm for curves B-E should be rationally assigned to an increasing amount of dissociated photogenerated excitons, while for the range from 380 to 360 nm, the normal regularity of gradually reduced phase retardation from sub band gap to super band gap energy is most likely contributed by the increasing absorption

17140 J. Phys. Chem. C, Vol. 111, No. 45, 2007

Zhao et al.

Figure 5. ac SPV spectra (a) and corresponding phase spectra (b) with the interface of substrate ITO/ZnO nanorods excited first for sample A. The cell configuration in Figure 4a was used. Inset of (a): SPV spectra under biases of -1.0, 0, 0.25, and 0.5 V. Both curve A in (a) and curve A in (b) were obtained under zero external bias.

Figure 6. ac SPV spectra (a) and corresponding phase spectra (b) of sample A with the ZnO nanorod surface excited first. The sample vs electrode configuration in Figure 4b was used. In (a), curves A, B, and C denote SPV with the ZnO surface excited first with biases of 0, -1.0, and +1.0 V, respectively. Curve D: ZnO/ITO interface excited first without bias for comparison. In (b), curves A and B denote the phase of SPV with the ZnO surface excited first with biases of 0 and -1.0 V, respectively. Curve C: ZnO/ITO interface excited first without bias for comparison.

coefficient near the band gap energy in the spectral region.44 It is indicated by the quenched SPV response for 300-350 nm that the interface of ZnO/ITO substrate is not capable of generating excess charge carriers effectively. The extra electric field induced by negative bias is opposite the macro built-in field in nanorods. Under negative bias the interface defect states are primarily filled with electrons, so that they are of limited electrical activity in accumulating charge by trapping photogenerated electrons, quenching the SPV response. Furthermore, the quenched SPV under negative bias does not emerge with opposite sign, which indicates that the negative bias is not strong enough to reverse the built-in field in sample A. When sample A was excited from the surface of ZnO nanorods with another blank ITO as the counter electrode, the configuration in Figure 4b was applied. The typical results are shown in Figure 6. In Figure 6a, we have both band-to-band transition (300-360 nm) and the exciton-related SPV response (370-390 nm) under zero bias by comparing curves A and D, and the SPV response corresponding to the exciton-related transition was enhanced a little in amplitude only with increasing negative bias. However, under increasing negative bias to -1.0 V, the SPV response for band-to-band transition weakened and disappeared, and under increasing positive bias, both SPV response bands weakened and quenched. The difference in the SPV response characteristics caused by the two opposite illuminating directions originates from the different excited regions in the nanorod array in the spectral region 300-350 nm. When the top surface of the nanorods is excited first, the SPV response is contributed by the excess carriers generated

in the space charge layers underneath both the top surface and the lateral surface. Under negative bias, for the exciton-related SPV response band, the situation is similar to that of the light excitation from the interface, and the SPV response comes from photogenerated excitons dissociated via trapping by surface states and interface states, but for SPV response related to bandto-band transition within 300-350 nm, which relies on the charge separation by the space charge layer beneath the top surface, the negative bias will induce positive charge onto the surface of the nanorods, neutralizing the negative charge carried by oxygen and disabling the charge separation therein. The surface states are depopulated of electrons, and their capability of trapping photogenerated holes is weakened. The positively charged surface states induced by negative bias cannot trap but repel the photogenerated holes, and the external electric field applied is not strong enough to pull the photogenerated holes to the top surface efficiently to yield a positive SPV response. Therefore, the SPV response within 300-350 nm disappears under a negative bias of -1.0 V, although the external electric field induced by negative bias is in the same direction as the built-in field in the nanorod. Meanwhile the SPV response related to excitons from 370 to 400 nm is enhanced, sharing the same mechanism as that with light incident from the ITO substrate. However, the SPV response quenched for both of the bands under positive bias. Consider that at this time the positive bias forms an external electric field opposite the macro built-in field within the volume of the nanorods. This will induce a negative charge to the free end surface from the bulk of the nanorods, causing a weakened macroscopic built-in electric field

Photovoltaic Properties of ZnO Nanorods

J. Phys. Chem. C, Vol. 111, No. 45, 2007 17141

Figure 7. ac SPV spectra and corresponding phase spectra (inset) of sample B with the substrate ITO/ZnO nanorod interface excited first under positive biases (a) and under negative biases (b). The sample vs electrode configuration in Figure 4a was used.

along the nanorods. As a result, the photogenerated electrons and holes cannot be effectively separated by the weakened builtin field and then transfer for both of the transition bands, leading to quenched SPV. In Figure 6b, curve B represents the phase spectrum corresponding to the ac SPV spectrum under a bias of -1.0 V. The abnormal phase retardation relative to 0° with a phase value of about -120° at 375 nm, which is much larger than that without bias, apparently presents the complexity of photogenerated charge-transfer kinetics. The detailed process of charge transfer therein cannot be directly specified. To study the size effect on the SPV properties of ZnO nanorod arrays, sample B, which has a smaller size both in length and width than sample A, was investigated in the same way as sample A. The typical results are shown in Figure 7. When excited from the interface of ZnO/ITO under zero bias, both exciton-related and band-to-band transition-related SPV response can be observed at this time. This feature could be well understood in terms of the absorption characteristics of the system. It is evident from the UV-vis transmission spectrum that the incident light above 320 nm can penetrate the nanorod array for sample B. The length of the nanorods is estimated to be shorter than 70 nm according to the absorption coefficient of ZnO. When illuminated from the interface of ITO/ZnO, the whole film can be excited, including the surface space charge layer in the free end of the nanorods. Therefore, both excitonrelated and band-to-band transition-related SPV response can be observed at this time under zero bias, which is distinguished from sample A obviously. At this time, both the macro built-in field in the volume of the nanorods and the space charge layer underneath the surface, as well as charge trapping by the surface states and interface states, contribute to the SPV development. Under positive bias, which is in the same direction as the builtin field, the excess carriers are thus separated more efficiently with increasing bias intensity, yielding enhanced SPV response for both bands (see Figure 7a). As shown in Figure 7b, when negative bias was applied in the experiment, the band-to-band transition-related SPV response weakened and quenched and even took a reversal when an increasing negative bias was applied for the exciton-related SPV response band. The reversed polarity of the SPV response under negative bias could be clearly judged by the corresponding phase spectra. This phenomenon is also distinguished from that of sample A. This result implies that the interface states could be fully filled with electrons under a strong enough negative bias; thus, excess holes could be trapped at these positions. Negative bias is opposite the direction of the built-in field in the surface space charge layers of the ZnO rods. The depletion layer at the surface is then neutralized

by the additional negative charge induced there. Consequently, the electrons cannot be separated effectively from holes for band-band transition. Instead, photogenerated holes tend to recombine with the electrons in the nanorods, and no SPV response in the spectral range of 300-350 nm can be observed under the negative bias in the experiment. However, for excitonrelated SPV response, excitons are generated in the entire bulk of the nanorods by illumination. The macro built-in field in the volume of the nanorods is reversed by the negative bias, and the electrons and holes are separated toward the direction opposite to that without bias under the coaction of the two electric fields, resulting in reversed SPV response peaked at 371 nm. We can deduce that the built-in field in sample B is weaker than that in sample A, as judged from the reversed SPV under negative bias for sample B. For the kinetics of the SPV response, the phase retardation approaches 90° at 375 nm even when no bias is applied, indicating a much slower transfer rate of excess carriers in sample B. The slow process comes from the increased density of surface defects due to the reduced reaction time and smaller grain size. Surface trapping of photogenerated charges by defects is expected to be more dramatic in sample B than sample A, which hinders the process of excess electron transport along the nanorods. The above comparison between sample A and sample B indicates the size effect of nanorods on the photovoltaic properties of ZnO nanorod arrays grown on the ITO substrate. We have observed that the orientation of the nanorods could also have a distinct influence on the SPV feature. The ZnO nanorod powder of sample C was bladed carefully onto the ITO substrate for characterization. The SEM image and the XRD pattern of sample C indicate that the size of the nanorods is similar to that in sample A, but they adopt more orientation with the c-axis parallel to the substrate in contrast to that in sample A. In SPV measurement, the powder layer was thick enough (more than 0.5 mm) so that no light could penetrate through. The powder was directly sandwiched between two ITO electrodes without extra treatment, which were connected to the lock-in amplifier. The ac SPV results of sample C are shown in Figure 8. In Figure 8a, we see that, under zero bias, the sample gives a rather weak ac SPV response, with photogenerated electrons transferring to the top electrode from which the light is incident (see the corresponding phase value), but when positive bias or negative bias is applied between the two electrodes, the SPV response is enhanced, and the spectral range with a significant SPV response expands to 550 nm under 1.0 V bias, which is another feature different from that of sample A. The expanded spectral range in sub band gap energy indicates

17142 J. Phys. Chem. C, Vol. 111, No. 45, 2007

Zhao et al.

Figure 8. (a) ac SPV spectra of sample C under different dc biases: 0, 0.25, 0.5, 1.0, -0.25, -0.5, and -1.0 V. (b) Phase spectra corresponding to the ac SPV spectra.

that more excess carriers related to surface state transition are involved in the SPV response, which is consistent with the PL result that the nanorod surface is abundant in defects. The correlated surface states could preferentially trap electrons or holes under different biases, and judging from the phase spectra, the photogenerated electrons transfer to the light-incident ITO from ZnO under positive bias, while the photogenerated holes transfer to the light-incident ITO from ZnO under negative bias. The nearly identical SPV response patterns in the whole spectral range under symmetrical external biases imply that the macro built-in field is very weak in a single nanorod, which is liable to be modified by an external bias. This characteristic of the nanorods can be well understood by considering a single nanorod. For an isolated nanorod, unlike the vertically aligned one on the ITO substrate, no preferential macro built-in field exists in its whole volume, because of the radical distribution of the built-in field from the center toward the rod surface. Then the excitons localized in one nanorod are less probably dissociated by the macro built-in field rapidly, but the electrons and holes can be separated from each other via preferential trapping by surface states. The excess electrons and holes can diffuse further from one nanorod to the adjacent one in contact and be further separated, leading to detectable SPV response. The possibility that part of the excess charge carriers diffuse from some nanorods to the contacting top ITO electrode also exists, which competes with the charge diffusion toward the powder bulk. As a result, without bias, the ac SPV response for the nanorod powder is weakened. In fact, the latter mechanism dominates. On the whole, the photogenerated holes are trapped by the ZnO nanorods, while some electrons diffuse to the top electrode, and a negative SPV response is generated, as judged from the corresponding phase value. However, under increasing external bias, more photogenerated charges are able to transfer over the barriers at the boundary between two rods in contact, and this mechanism dominates then, which results in increasing SPV response with slowly developing kinetics. By comparison of sample C with sample A under negative bias excited from the interface, we can infer that the macro built-in field contributed by lateral surface states is more easily weakened by electron filling, but the built-in electric field along the c-axis in a vertically aligned nanorod on the ITO substrate is harder to reverse, which is owed to the stronger built-in electric field in the rods along the axis toward the substrate and the fact that the negative bias in the experiment was not strong enough to induce a reversed electric field along the axis. The peak pattern at 462 nm is due to the inhomogeneous spectral

energy distribution of Xe arc light. As for the phase spectra of sample C, the smaller retardation of sub band gap response than super band gap response is due to the localized nature of the surface states. The excess charge carriers related to surface states are more localized in a single nanorod than those by band-toband transition and can hop with less probability across the boundaries between nanorods under an external bias, leading to a shorter time needed to approach saturated SPV response in an illumination period. The KP-based SPV results also support the above discussions. In the experiment, steady monochromatic light was incident directly onto the top surface of the sample. In Figure 9a, we have the SPV spectra of the three samples. The following points of interest deserve our attention. First, the SPV onset of curve C is dramatically red shifted compared to that of curves A and B in the sub band gap region. According to the rule of the SPV generating process, for sub band gap SPV, the sample absorbs photons of certain energy and then the electron transition from surface states to the conduction band or from the valence band to surface states takes place, yielding SPV response as long as electrons are separated from the holes. On the basis of the positive sign of SPV, we suggest the former mechanism is most likely responsible for the SPV observed and the surface is positively charged upon illumination. The photogenerated electrons in the conduction band of ZnO nanorods drift or diffuse, leaving holes localized at the surface states after transition. For super band gap SPV, upon excitation, electrons are excited from the valence band to the conduction band and then separated from the free holes by diffusion or drift. The holes tend to be trapped at the donor surface states and separated from the photogenerated electrons, because of the existence of a large density of surface defects in the nanorods, which are mostly adsorbed with O2-. Consequently, a positive SPV response is observed for both sub band gap and super band gap illumination. The widened spectral region of the more marked SPV response of sample C could be attributed to the dominating orientation adopted by the nanorods in the powder, which makes the surface states on the lateral face of the nanorods much exposed to the probe. Thus, many more holes trapped at the surface states are detected in the sub band gap spectral region, which is in good agreement with lockin-based ac SPV results under bias, and the existence of dense surface states with a wide energy distribution at the side surface of the nanorods is further confirmed. However, the KP-based SPV spectrum of sample A once again did not reflect the effect of abundant surface states in the nanorods, although the nanorods

Photovoltaic Properties of ZnO Nanorods

J. Phys. Chem. C, Vol. 111, No. 45, 2007 17143

Figure 9. (a) SPV spectra of sample A (curve A), sample B (curve B), and sample C (curve C) with illumination on top taken on the KP. Inset: Schematic setup of the KP-based SPV measurement. (b) CPD changes with 350 nm light switched on or off: curve A, sample A; curve B, sample B; curve C, sample C. Curve B has been shifted down by 100 mV and curve C by 300 mV for clarity.

themselves in samples A and C are essentially the same in almost every aspect except the orientation and the interfacial contact between the nanorods and ITO substrate. Obviously it is the orientation of the nanorods that precludes the probing of the SPV directly related to the surface defects on the side surface of the nanorods in KP-based SPV detection as well as in ac SPV. This result also demonstrates that the surface states at the polar plane (001) of ZnO nanorods are in smaller density and energy distribution. Second, the SPV strength of sample B is much stronger than that of sample A in SPV results from both the lock-in amplifier and KP, which also indicates that the surface states play a key role in separating photogenerated electrons and holes to yield SPV response. Because of the much smaller grain size in sample B than that in sample A, the surface states in sample B are much more abundant, which could trap more holes in a given time. The SPV intensity of sample C is the largest among all three samples, which could be ascribed to the much thicker nanorod layer than that of sample A or B. Third, the SPV onset of sample B has a small blue shift with respect to that of sample A, which can also be discriminated by comparison of sample A with sample B in ac SPV spectra. The SPV onset of sample A is at ∼391 nm and that of sample B at ∼387 nm. The difference is assigned to the reduced grain size of the nanorods in sample B, which have a smaller average width than those in sample A. The onset of SPV response originates from the dissociated excitons upon excitation, with holes then trapped at the surface, for both samples A and B. A slightly higher energy is needed for photogenerating excitons in the nanorods of sample B. We also monitored the SPV transient by tracking the surface potential with 350 nm light switched on or off. The kinetics of SPV response by KP provides further understanding of the effect of surface states. As shown in Figure 9b, when excited by light of 350 nm, samples A and C exhibit similar SPV onset kinetics. About 1 min is needed for samples A and C to reach the equilibrium SPV value. Sample B exhibits a slower SPV onset of about 100 s. However, for the relaxation process with the light switched off, sample B exhibits a 2 times longer decay time than samples A and C, while the relaxation kinetics for samples A and C are similar. The considerably long relaxation of sample B implies much more abundant defects exist at the surface or a higher density of surface states than those of samples A and C, which is reasonably assigned to the reduced size and higher ratio of surface to volume of the nanorods. It will take a longer time to release the trapped photogenerated holes by the surface states with a larger density. The above result is

consistent with the kinetics obtained in ac SPV results without bias. In comparison of the phase value at 375 nm of sample A (Figure 6b, curve A) with that of sample B (Figure 7b inset, curve A) excited from the interface without bias, a larger retardation of sample B than sample A is distinct. The surface states with a larger density might trap more photogenerated holes and release them more slowly in sample B, which causes larger retardation. Nevertheless, similar trapping densities by the surface states should exist for both samples A and C as they are obtained simultaneously in one autoclave and of similar size. It should be emphasized that the relaxation times for the two samples are of similar scale, which indicates that the release kinetics of photogenerated holes previously trapped by surface states controls the SPV relaxation process, less dependent on the crystalline orientation of the nanorods. The KP detection for sample C seems more sensitive than the lock-in method without bias, partially because only one mechanism dominates in the KP configuration, which is without the contact between the top electrode and the sample. The excess charge carriers with two polarities are further separated across grain boundaries between nanorods via diffusion, and the KP detects the accumulated change in the surface potential under steady illumination. However, with the aid of an external bias, the lockin-based method can resolve the sub-band-gap-related SPV as effective as the KP. On the basis of the above experimental results and discussion, we conclude that the size of the ZnO nanorods can modulate the transmission property of the array, and thus the SPV response property, especially in the super band gap spectral region with a large absorption coefficient. For ZnO films of vertically aligned nanorod arrays, the length of the rod modulates the absorption and penetration depth of the incident light, determining the position where the electron-hole pairs are generated and the possibility that they are separated into charge carriers. It is revealed that, beyond the dimension scale where the quantum size effect on the SPV property takes place,41 the light absorption modulating effect of the grain size on the photogenerated charges can be dramatic. This feature should be notable for many nanostructured semiconductors when the light-penetrating depth is on the nanoscale. In addition, the coupling of the grain size and shape could further complicate the photoelectric behavior of the nanomaterials in the aspects of quantum confinement and surface states. Finally, we summarize the orientation effect of nanorods on the SPV developing process in terms of surface states. The surface states of ZnO originate from the surface electronic structures related to oxygen vacancies formed during the crystal

17144 J. Phys. Chem. C, Vol. 111, No. 45, 2007 growth process and the modification by foreign species. In light of the former analysis, sub band gap SPV response between 400 and 500 nm can be observed only for the powder sample without bias and the nanorods in the powder adopt more orientation with the side surface exposed directly to the probe, which give evidence that the surface states are mostly distributed at side facets of the rods, the nonpolar lattice planes. The localized character of the surface states is exhibited clearly in comparison of the SPV spectra of samples A and C. The surface states on the (002) surface at the nanorod free end are much sparser than those on the lateral surface, so that much less SPV related to sub band gap transition can occur on the top surface of the well-aligned nanorods, as observed for sample A or B. Therefore, the orientation of nanorods related to the light incidence and probing electrode controls the probability that surface states are excited and detected. We can also expect that, by illumination of sub band gap energy on the ZnO nanorods (400-500 nm), the surface potential of the lateral surface can be positively altered while that of the ends of the nanorods should remain around the original level. Furthermore, the separation and transport of photoinduced charges for the nanorods with or without a growth substrate are also relevant to the surface states. For ZnO nanorod powder in the air, as no pronounced macro built-in field with a certain orientation exists in a single rod, the photoinduced charges are less probably separated by the built-in field, but the electrons and holes can be separated from each other via preferential trapping of holes by surface states at the side surface of the nanorod, and the electrons could further diffuse to an adjacent nanorod in contact. When an external bias is applied, which is roughly perpendicular to the c-axis, photogenerated charges are much trapped at modified surface states at the lateral surface by the electron injection or extraction effect of external bias. External bias promotes more excitons to dissociate into excess carriers via surface states and accelerate the transfer rate of the excess carriers across grain boundaries, yielding enhanced SPV. While in the air for the ZnO nanorods in the array grown on the ITO substrate, all the surface but the junction end of ZnO/ ITO is exposed to oxygen. The oxygen molecules adsorb at the oxygen vacancies at the ZnO surface, extract negative charge from the bulk, and form O2- at the surface, which yields a significant macro electric field along the rod axis as well as weak electric fields in the lateral surface layer. The macro builtin field along the rod axis is vertical to the substrate. It can accelerate the charge transport along the rod and help separate photogenerated charges effectively. Although the lateral builtin field around one rod is neutralized on the whole, the surface states localized at the side surface of the nanorod play a key role in helping separate the electron-hole pairs by selectively trapping the holes and promote the dissociation of excitons. When the nanorod array is under the bias that behaves the same as the built-in field, the excess carriers from excitons transfer along the axis to the ends of the rods and are less trapped by surface states. Nevertheless, electrons could be trapped by the unpopulated interface states at ZnO/ITO, yielding a detectable SPV response. Therefore, it is feasible that the spectrum-dependent SPV response of the ZnO nanorod array can be modulated by varying the size of the composite nanorods and tailoring their orientation on the substrate via controlled synthesis. Besides the size effect, the influence of the growth substrate as well as the grain crystallographic orientation on the photogenerated charge separation in nanocrystals should also be critically taken into account to explain the SPV properties. The orientation-depend-

Zhao et al. ent SPV response also demonstrates that SPV spectroscopy is much more sensitive to the surface part of the sample which is just facing against the probing electrode, which is distinguished from PL spectroscopy. 4. Conclusions In this work, we have qualitatively investigated the effects of the size and the crystallographic orientation of ZnO nanorods on the spectrum-dependent photovoltaic properties. The detected SPV response depends on the light absorption, charge separation, and transport properties in the sample. The size of the nanorods modulates the absorption and penetration depth of the incident light, determining the position where the electron-hole pairs are generated and the possibility that they are separated into charge carriers. The crystallographic orientation of ZnO nanorods with respect to the probing electrode determines whether the photogenerated electrons related to surface states can be effectively detected because of the localized character of the surface states. The surface states are confirmed to be mostly localized at the lateral nonpolar surface of the nanorods rather than at the (001) surface, which is of larger density for reduced nanorod size, and slower SPV relaxation kinetics was observed for the nanorods with a smaller grain size. The growth substrate facilitates the macro built-in field to form along the nanorod axis and effects the photogenerated charge transport. Therefore, for nanostructured ZnO materials, the photovoltaic properties could be modulated on the spectrum by the grain size, surface state distribution with respect to the probing electrode via controlled crystallographic orientation, and the charge transport pathway related to the substrate as well as the specific nanostructure. Our results are also helpful in understanding the photovoltaic properties in other types of ZnO nanostructures besides nanorods and benefit the design and fabrication of photoactive components of smart systems based on nanostructured materials. Acknowledgment. For financial support, we are grateful to the National Natural Science Foundation of China (Grant Nos. 20473033 and 20673049) and Science and Technology Developing Funding of Jilin Province of China (Grant No. 20040503). This work is also supported by the National Basic Research Program of China (Program 973) (Grant No. 2007CB613303). References and Notes (1) Wang, Z. L. J. Phys.: Condens. Matter 2004, 16, R829. (2) Fan, Z.; Lu, J. G. J. Nanosci. Nanotechnol. 2005, 5, 1561. (3) O ¨ zgu¨r, U ¨ .; Alivov, Ya. I.; Liu, C.; Teke, A.; Reshchikov, M. A.; Do’’an, S.; Avrutin, V.; Cho, S.-J.; Morkoc¸ , H. J. Appl. Phys. 2005, 98, 041301. (4) Fonoberov, V. A.; Balandin, A. A. J. Nanoelectron. Optoelectron. 2006, 1, 19. (5) Heo, Y. W.; Norton, D. P.; Tien, L. C.; Kwon, Y.; Kang, B. S.; Ren, F.; Pearton, S. J.; LaRoche, J. R. Mater. Sci. Eng., R 2004, 47, 1. (6) Meng, X. Q.; Zhao, D. X.; Zhang, J. Y.; Shen, D. Z.; Lu, Y. M.; Liu, Y. C.; Fan, X. W. Chem. Phys. Lett. 2005, 407, 91. (7) Rosa, E. D.; Guzman, S. S.; Jayan, R. B.; Torres, A.; Salas, P.; Elizondo, N.; Yacaman, M. J. J. Phys. Chem. C 2007, 111, 8489. (8) Kong, B. H.; Kim, D. C.; Cho, H. K. Physica B 2006, 376-377, 726. (9) Pan, N.; Wang, X.; Zhang, K.; Hu, H.; Xu, B.; Li, F.; Hou, J. G. Nanotechnology 2005, 16, 1069. (10) Zhao, Q.; Zhang, H. Z.; Zhu, Y. W.; Feng, S. Q.; Sun, X. C.; Xu, J.; Yu, D. P. Appl. Phys. Lett. 2005, 86, 203115. (11) Marotti, R. E.; Giorgi, P.; Machado, G.; Dalchiele, E. A. Sol. Energy Mater. Sol. Cells 2006, 90, 2356. (12) Fonoberov, V. A.; Alim, K. A.; Balandin, A. A.; Xiu, F.; Liu, J. Phys. ReV. B 2006, 73, 165317. (13) Chen, C. W.; Chen, K. H.; Shen, C. H.; Ganguly, A.; Chen, L. C.; Wu, J. J.; Wen, H. I; Pong, W. F. Appl. Phys. Lett. 2006, 88, 241905.

Photovoltaic Properties of ZnO Nanorods (14) Xia, Y. N.; Yang, P. D.; Sun, Y. G.; Wu, Y. Y.; Mayers, B.; Gates, B.; Yin, Y. D.; Kim, F.; Yan, H. Q. AdV. Mater. 2003, 15, 353. (15) Kuchibhatla, S. V.; Karakoti, A. S.; Bera, D.; Seal, S. Prog. Mater. Sci. 2007, 52, 699. (16) Sirbuly, D. J.; Law, M.; Yan, H. Q.; Yang, P. D. J. Phys. Chem. B 2005, 109, 15190. (17) Lu, J. G.; Chang, P. C.; Fan, Z. Y. Mater. Sci. Eng., R 2006, 52, 49. (18) Greene, L. E.; Yuhas, B. D.; Law, M.; Zitoun, D.; Yang, P. D. Inorg. Chem. 2006, 45, 7535. (19) Wang, R. P.; Xu, G.; Jin, P. Phys. ReV. B 2004, 69, 113303. (20) Shalish, I.; Temkin, H.; Narayanamurti, V. Phys. ReV. B 2004, 69, 24540. (21) Chen, C. Q.; Shi, Y.; Zhang, Y. S.; Zhu, J.; Yan, Y. J. Phys. ReV. Lett. 2006, 96, 075505. (22) Liao, L.; Lu, H. B.; Li, J. C.; He, H.; Wang, D. F.; Fu, D. J.; Liu, C.; Zhang, W. F. J. Phys. Chem. C 2007, 111, 1900. (23) Jang, E. S.; Won, J. H.; Hwang, S. J.; Choy, J. H. AdV. Mater. 2006, 18, 3309. (24) Nonomura, K.; Komatsu, D.; Yoshida, T.; Minoura, H.; Schlettwein, D. Phys. Chem. Chem. Phys. 2007, 9, 1843. (25) Wu, Y. Y.; Yan, H. Q.; Yang, P. D. Top. Catal. 2002, 19, 197. (26) Vayssieres, L.; Keis, K.; Lindquist, S. E.; Hagfeldt, A. J. Phys. Chem. B 2001, 105, 3350. (27) Huang, M. H.; Mao, S.; Feick, H.; Yan, H. Q.; Wu, Y. Y.; Kind, H.; Weber, E.; Russo, R.; Yang, P. D. Science 2001, 292, 1897. (28) Park, W. I.; Yi, G.-C.; Kim, J.-W.; Park, S.-M. Appl. Phys. Lett. 2003, 82, 4358. (29) Lin, C. C.; Chen, H. P.; Chen, S. Y. Chem. Phys. Lett. 2005, 404, 30. (30) Park, W. I.; Yi, G.-C. AdV. Mater. 2004, 16, 87. (31) Ng, H. T.; Han, J.; Yamada, T.; Nguyen, P.; Chen, Y. P.; Meyyappan, M. Nano Lett. 2004, 4, 1247. (32) Wang, X. D.; Neff, C.; Graugnard, E.; Ding, Y.; King, J. S.; Pranger, L. A.; Tannenbaum, R.; Wang, Z. L.; Summers, C. J. AdV. Mater. 2005, 17, 2103. (33) Law, M.; Greene, L. E.; Johnson, J. C.; Saykally, R.; Yang, P. D. Nat. Mater. 2005, 4, 455.

J. Phys. Chem. C, Vol. 111, No. 45, 2007 17145 (34) Liu, T. Y.; Liao, H. C.; Lin, C. C.; Hu, S. H.; Chen, S. Y. Langmuir 2006, 22, 5804. (35) Xu, F.; Yuan, Z. Y.; Du, G. H.; Ren, T. Z.; Bouvy, C.; Halasa, M.; Su, B. L. Nanotechnology 2006, 17, 588. (36) Yoon, S. H.; Yang, H.; Kim, Y. S. Nanotechnology 2007, 18, 205608. (37) Wang, X. D.; Song, J. H.; Liu, J.; Wang, Z. L. Science 2007, 316, 102. (38) Kronik, L.; Shapira, Y. Surf. Sci. Rep. 1999, 37, 1. (39) Xie, T. F.; Wang, D. J.; Zhu, L. J.; Wang, C.; Li, T. J.; Zhou, X. Q.; Wang, M. J. Phys. Chem. B 2000, 104, 8177. (40) Arora, B. M.; Ghosh, S.; Datta, S.; Kumar, S. Mater. Sci. Semicond. Process. 2001, 4, 489. (41) Lin, Y. H.; Wang, D. J.; Zhao, Q. D.; Yang, M.; Zhang, Q. L. J. Phys. Chem. B 2004, 108 3202. (42) Jing, L. Q.; Xin, B. F.; Yuan, F. L.; Xue, L. P.; Wang, B. Q.; Fu, H. G. J. Phys. Chem. B 2006, 110, 17860. (43) Cheng, K.; He, Y. P.; Miao, Y. M.; Zou, B. S.; Wang, Y. G.; Wang, T. H. J. Phys. Chem. B 2006, 110, 7259. (44) Donchev, V.; Kirilov, K.; Ivanov, Ts.; Germanova, K. Mater. Sci. Eng., B 2006, 129, 186. (45) Fungo, F.; Milanesio, M. E.; Durantini, E. N.; Otero, L.; Dittrich, Th. J. Mater. Chem. 2007, 17, 2107. (46) Zidon, Y.; Shapira, Y.; Dittrich, Th. Appl. Phys. Lett. 2007, 90, 142103. (47) Zidon, Y.; Shapira, Y.; Dittrich, Th.; Otero, L. Phys. ReV. B 2007, 75, 195327. (48) Zhao, Q. D.; Wang, D. J.; Peng, L. L.; Lin, Y. H.; Yang, M.; Xie, T. F. Chem. Phys. Lett. 2007, 434, 96. (49) Tam, K. H.; Cheung, C. K.; Leung, Y. H.; Djurisˇic´, A. B.; Ling, C. C.; Beling, C. D.; Fung, S.; Kwok, W. M.; Chan, W. K.; Phillips, D. L.; Ding, L.; Ge, W. K. J. Phys. Chem. B 2006, 110, 20865. (50) Muth, J. F.; Kolbas, R. M.; Sharma, A. K.; Oktyabrsky, S.; Narayan, J. J. Appl. Phys. 1999, 85, 7884. (51) Yan, Y. F.; Al-Jassim, M. M.; Wei, S. H. Phys. ReV. B 2005, 72, 161307.