Report

Piezoelectric Nanogenerators Based on Zinc Oxide Nanowire Arrays

See allHide authors and affiliations

Science  14 Apr 2006:
Vol. 312, Issue 5771, pp. 242-246
DOI: 10.1126/science.1124005

Abstract

We have converted nanoscale mechanical energy into electrical energy by means of piezoelectric zinc oxide nanowire (NW) arrays. The aligned NWs are deflected with a conductive atomic force microscope tip in contact mode. The coupling of piezoelectric and semiconducting properties in zinc oxide creates a strain field and charge separation across the NW as a result of its bending. The rectifying characteristic of the Schottky barrier formed between the metal tip and the NW leads to electrical current generation. The efficiency of the NW-based piezoelectric power generator is estimated to be 17 to 30%. This approach has the potential of converting mechanical, vibrational, and/or hydraulic energy into electricity for powering nanodevices.

Wireless devices may allow in situ, real-time biomedical monitoring and detection, but such devices still require a power source. Ideally, such devices should be self-powered and not dependent on a battery. The body provides numerous potential power sources: mechanical energy (such as body movement, muscle stretching, blood vessel contraction), vibrational energy (acoustic waves), chemical energy (glucose), and hydraulic energy (body fluid and blood flow), but the challenge is their efficient conversion into electrical energy. If accomplished on the nanoscale, such power sources could greatly reduce the size of integrated nanosystems for optoelectronics (1), biosensors (2), resonators (3), and more.

We demonstrate an approach to converting mechanical energy into electric power with the use of aligned zinc oxide (ZnO) nanowires (NWs). The mechanism of the power generator relies on the coupling of piezoelectric and semiconducting properties of ZnO as well as the formation of a Schottky barrier between the metal and ZnO contacts. The nanogenerator has the potential of harvesting energy from the environment for self-powered nanotechnology.

Among the known one-dimensional (1D) nanomaterials, ZnO has three key advantages. First, it exhibits both semiconducting and piezoelectric (PZ) properties that can form the basis for electromechanically coupled sensors and transducers. Second, ZnO is relatively biosafe and biocompatible (4), and it can be used for biomedical applications with little toxicity. Finally, ZnO exhibits the most diverse and abundant configurations of nanostructures known so far, such as NWs (5), nanobelts (NBs) (6), nanosprings (7), nanorings (8), nanobows (9), and nanohelices (10). Although numerous studies have demonstrated novel nanodevices and applications based on NWs and NBs, little work has been done to address the power needs of these nanosystems.

Our study is based on aligned ZnO nanowires grown on c plane–oriented α-Al2O3 substrate, using Au particles as a catalyst, by the vapor-liquid-solid (VLS) process (11, 12). An epitaxial relation between ZnO and α-Al2O3 allows a thin, continuous layer of ZnO to form at the substrate, which serves as a large electrode connecting the NWs with a metal electrode for transport measurement (Fig. 1A). The NW grows along the [0001] direction and has side surfaces of Embedded Image (Fig. 1B). Most of the Au particles at the tips of the NWs either evaporate during the growth or fall off during scanning by the atomic force microscope (AFM) tip (fig. S1). For most of the NWs, the growth front is free of Au particles or has a small hemispherical Au particle that covers only a fraction of the top (inset, Fig. 1B). For the purpose of our measurements, we have grown NW arrays that have relatively less density and shorter length (0.2 to 0.5 μm), so that the AFM tip can exclusively reach one NW without touching another.

Fig. 1.

Experimental design for converting nanoscale mechanical energy into electrical energy by a vertical piezoelectric (PZ) ZnO NW. (A) Scanning electron microscopy images of aligned ZnO NWs grown on α-Al2O3 substrate. (B) Transmission electron microscopy images of ZnO NWs, showing the typical structure of the NW without an Au particle or with a small Au particle at the top. Each NW is a single crystal and has uniform shape. Inset at center: an electron diffraction pattern from a NW. Most of the NWs had no Au particle at the top. Inset at right: image of a NW with an Au particle. (C) Experimental setup and procedures for generating electricity by deforming a PZ NW with a conductive AFM tip. The base of the NW is grounded and an external load of RL is applied, which is much larger than the resistance RI of the NW. The AFM scans across the NW arrays in contact mode.

The measurements were performed by AFM using a Si tip coated with Pt film, which has a cone angle of 70°. The rectangular cantilever had a calibrated normal spring constant of 0.76 N/m (Fig. 1C). In the AFM contact mode, a constant normal force of 5 nN was maintained between the tip and sample surface. The tip scanned over the top of the ZnO NW, and the tip's height was adjusted according to the surface morphology and local contacting force. The thermal vibration of the NWs at room temperature was negligible. For the electric contact at the bottom of the nanowires, silver paste was applied to connect the (large) ZnO film on the substrate surface with the measurement circuit. The output voltage across an outside load of resistance RL = 500 megohms was continuously monitored as the tip scanned over the nanowires (note the defined polarity of the voltage signal). No external voltage was applied in any stage of the experiment.

Experimentally, both the topography (feedback signal from the scanner) (Fig. 2A) and the corresponding output voltage (VL) images across the load (Fig. 2B) were recorded simultaneously when the AFM tip was scanned over the aligned NW arrays. In contact mode, as the tip scanned over the vertically aligned NWs, the NWs were bent consecutively. The bending distance was directly recorded in the topography image, from which the maximum bending deflection distance and the elastic modulus of the NW as well as the density of NWs that have been scanned by the tip were directly derived (13).

Fig. 2.

Electromechanically coupled discharging process of aligned piezoelectric ZnO NWs observed in contact mode. (A and B) Topography image (A) and corresponding output voltage image (B) of the NW arrays. The discharging process was so rapid that each discharge event was characterized by only a couple of data points, which made it difficult to display the data by rainbow color (see fig. S2). (C) A series of line profiles of the voltage output signal when the AFM tip scanned across a vertical NW at a time interval of 1 min. The data show the registration of the electric signal with the location of the NW and the reproducibility of the event over an extended period of time. Colors represent the outputs required for a series of scans. In (A) to (C), the scanning speed of the tip was 12.081 μm/s, and the time spent to acquire and output one scan point was 2 ms. (D) Line profiles from the topography (red) and output voltage (blue) images across a NW. The peak of the voltage output corresponds approximately to the maximum deflection of the NW, indicating that the discharge occurs when the tip is in contact with the compressed side of the NW. When the tip touches the NW, PZ charges start to accumulate but no discharge occurs. Discharge occurs when the deflection reaches nearly the maximum ym. Note that the lateral deflection y includes the shape and contact geometry of the tip, which must be subtracted to derive the true deflection of the NW. (E) Line profile of the voltage output signal when the AFM tip scans across a vertical NW at 12.394 μm/s. The time spent to acquire and output one scan point was 0.05 ms, which was achieved at the maximum scan frequency of the AFM. The inset is an equivalent circuit of the measurement to be used for simulating the discharging process. (F) The resonance vibration of a NW after being released by the AFM tip, showing that the stored elastic energy is transferred mainly into vibrational energy after creating the PZ discharge event.

In the VL image, many sharp output peaks (like discharge peaks) were observed. These peaks, typically about 4 to 50 times the noise level, are rather sharp and narrow, and sometimes one or two pixels represent one voltage peak because of the limited scanning speed of the AFM tip, so that the color distribution in the plot is not easy to display (fig. S2). By reducing the scan range and increasing the scan frequency, more complete profiles of the discharge peaks were captured (fig. S3). Most of the voltage peaks are ∼6 to 9 mV in height. The density of NWs contacted by the tip is counted from Fig. 2A to be ∼20/μm2, and the average density of NWs whose voltage output events had been captured by the tip in Fig. 2B is ∼8/μm2, thus ∼40% of the NWs were contacted.

The location of the voltage peak is directly registered at the site of the NW. A time series of the voltage output line profiles across one NW acquired at a time interval of 1 min is shown in Fig. 2C. Because the dwell time for each data point (or pixel) is 2 ms, which is longer than the average lifetime of the voltage peak of ∼0.6 ms (Fig. 2D), the peak at which VL reached the maximum was possibly missed by the “slow” scanning tip, so that VL shows a chopped top (arrows in Fig. 2C). A sharp peak can be identified continuously at the location of the NW and in the NW output voltage in each scan of the tip. When the tip started to deflect the NW, no voltage output was observed (Fig. 2D); VL was detected when the deflection of the NW approached its maximum. When the NW was released by the AFM tip, VL dropped to zero, indicating that the output of piezoelectricity was detected toward the end of the AFM scan over the NW.

The shape of the discharge peak can be further improved and analyzed by maximizing the tip scanning frequency and reducing the scanning range. Shown in Fig. 2E is a line profile of VL when the tip was scanned over a single NW at a scanning velocity of 12.394 μm/s. The full width at half maximum (FWHM) of the voltage peak was estimated to be τ ∼ 0.6 ms. The damping behavior of the voltage peak can be qualitatively described with an equivalent circuit (14) (Fig. 2E). The NW is approximated as a resistor RI and a capacitor C (including the contribution from the system). The lifetime of the output voltage VL across the load RL (note the polarity of the voltage) is τ = (RL + RI)C (15). For the experiment we have designed, the resistance of the NW RI is negligible relative to RL (16, 17). Thus, the equivalent capacitance of the NW and the system is C ≈ τ/RL ≈ 1.2 pF.

The observed sharp voltage output for ZnO NWs was not observed for metal film (fig. S4), aligned carbon nanotubes (fig. S5), or aligned WO3 NWs (fig. S6) under identical or similar experimental conditions. These data rule out the possibility of friction or contact potential as the source of the observed VL response.

The efficiency of the electric power generated by this process can be calculated as follows. The output electrical energy from one NW in one piezoelectric discharge (PZD) event is Embedded Image (15), where V0 is the peak voltage of the discharge output. For simplicity, we approximately take the NW as a 2D object for easy analytical calculation. The elastic deformation energy (WELD) created by the AFM tip for displacing the NW is WELD = 3YIy2m/2L3 (15), where Y is the elastic modulus, I is the moment of inertia, L is the length of the NW, and ym is the maximum deflection of the NW. Dissipation of WELD mainly occurs in three ways: (i) mechanical resonance/vibration after releasing the NW (Fig. 2F); (ii) PZ discharge (ΔWPZD) for each cycle of the vibration; and (iii) friction/viscosity from the environmental medium. The mechanical resonance of the NW continues for many cycles, but it is eventually damped by the viscosity of the medium. Each cycle of the vibration generates ΔWPZD, but the AFM tip in the present experimental design catches only the energy generated in the first cycle of vibration. Taking into account the energy dissipated by the NW in the first cycle of vibration ΔWELD (15), the efficiency of converting mechanical energy to electrical energy is ΔWPZDWELD. Therefore, an efficiency of 17 to 30% has been received for a cycle of the resonance (table S1). The large efficiency is likely due to the extremely large deformation that can be borne by the nanowire (18).

The physical principle for creating the PZ discharge energy arises from how the piezoelectric and semiconducting properties of ZnO are coupled. For a vertical, straight ZnO NW (Fig. 3A), the deflection of the NW by an AFM tip creates a strain field, with the outer surface being stretched (positive strain ϵ) and the inner surface compressed (negative ϵ) (Fig. 3B). An electric field Ez along the NW (z direction) is then created inside the NW volume through the PZ effect, Ez = ϵz/d, where d is the PZ coefficient (19) along the NW direction that is normally the positive c axis of ZnO, with the Zn atomic layer being the front-terminating layer (20, 21). The PZ field direction is closely parallel to the z axis (NW direction) at the outer surface and antiparallel to the z axis at the inner surface (Fig. 3C). Under the first-order approximation, across the width of the NW at the top end, the electric potential distribution from the compressed to the stretched side surface is approximately between Embedded Image to Embedded Image [with Embedded Image (15), where T is the thickness of the NW). The electrode at the base of the NW is grounded. Note that Embedded Image and Embedded Image are the voltages produced by the PZ effect, which are each typically larger than a few tens of volts (22). The potential is created by the relative displacement of the Zn2+ cations with respect to the O2– anions, a result of the PZ effect in the wurtzite crystal structure; thus, these ionic charges cannot freely move and cannot recombine without releasing the strain (Fig. 3D). The potential difference is maintained as long as the deformation is in place and no foreign free charges (such as from the metal contacts) are injected.

Fig. 3.

Transport is governed by a metal-semiconductor Schottky barrier for the PZ ZnO NW (see movies S1 and S2). (A) Schematic definition of a NW and the coordination system. (B) Longitudinal strain ϵz distribution in the NW after being deflected by an AFM tip from the side. The data were simulated by FEMLAB for a ZnO NW of length 1 μmand an aspect ratio of 10. (C) The corresponding longitudinal PZ-induced electric field Ez distribution in the NW. (D) Potential distribution in the NW as a result of the PZ effect. (E and F) Contacts between the AFM tip and the semiconductor ZnO NW [boxed area in (D)] at two reversed local contact potentials (positive and negative), showing reverse- and forward-biased Schottky rectifying behavior, respectively (see text). This oppositely biased Schottky barrier across the NW preserves the PZ charges and later produces the discharge output. The inset shows a typical current-voltage (I-V) relation characteristic of a metal-semiconductor (n-type) Schottky barrier. The process in (E) is to separate and maintain the charges as well as build up the potential. The process in (F) is to discharge the potential and generates electric current. (G and H) Contact of the metaltip with a ZnO NW with a small Au particle at the top. The PZ potential is built up in the displacing process (G), and later the charges are released through the compressed side of the NW (H). (I) Contact of the metal tip with a ZnO NW with a large Au particle at the top. The charges are gradually “leaked” out through the compressed side of the NW as soon as the deformation occurs; thus, no accumulated potential will be created.

The contacts at the top and the base of the NW are nonsymmetric; the bottom contact is to the ZnO film in contact with Ag paste, so the effective contact is between ZnO and Ag. The electron affinity (Ea) of ZnO is 4.5 eV (23) and the work function (ϕ) of Ag is 4.2 eV; there is no barrier at the interface, so the ZnO-Ag contact is ohmic. At the tip of the NW, Pt has ϕ = 6.1 eV, and the Pt-ZnO contact is a Schottky barrier (24, 25) and dominates the entire transport process. Because the compressed side of the semiconductor ZnO NW has negative potential Embedded Image and the stretched side has positive potential (Embedded Image), two distinct transport processes will occur across the Schottky barrier.

We now consider the case of a ZnO NW without an Au particle at the top. In the first step, the AFM conductive tip that induces the deformation is in contact with the stretched surface of positive potential Embedded Image (Fig. 3, D and E). The Pt metal tip has a potential of nearly zero, Vm = 0, so the metal tip–ZnO interface is negatively biased for Embedded Image. Because the assynthesized ZnO NWs behave as n-type semiconductors, the Pt metal–ZnO semiconductor (M-S) interface in this case is a reverse-biased Schottky diode (Fig. 3E), and little current flows across the interface. In the second step, when the AFM tip is in contact with the compressed side of the NW (Fig. 3F), the metal tip–ZnO interface is positively biased for Embedded Image. The M-S interface in this case is a positively biased Schottky diode, and it produces a sudden increase in the output electric current. The current is the result of ΔV-driven flow of electrons from the semiconductor ZnO NW to the metal tip. The flow of the free electrons from the loop through the NW to the tip will neutralize the ionic charges distributed in the volume of the NW and thus will reduce the magnitudes of the potential Embedded Image and Embedded Image. Thus, VL starts to drop and reaches zero after all of the ionic charges in the NW are fully neutralized. This mechanism explains why the discharge curve in Fig. 2E is nearly symmetric. According to the model, the discharge occurs when the NW is bent nearly to its maximum deflection, so VL should have a small offset in reference to the corresponding topography peak along the direction of tip scan, as is observed experimentally in Fig. 2D.

We next consider the case of a ZnO NW with a small Au particle at the top as a result of VLS growth (Fig. 3, G and H). In the first step of the AFM tip displacing the NW, the tip is directly in contact with the ZnO NW but not the Au particle (Fig. 3G) because of its small size and hemispherical shape (Fig. 1B and fig. S1). A process similar to that described in Fig. 3E occurs, and no output voltage will be observed. When the tip is in contact with the Au particle, the tip is integrated with the particle as one metal contact, and at the local interface between the Au particle and the negatively charged, compressed side of the ZnO NW (Embedded Image), a forward-biased Schottky diode is formed; thus, the current flows from the tip through the Au particle to the interface region with negative PZ voltage (Embedded Image) (Fig. 3H). A discharge process similar to that shown in Fig. 3F occurs, and a sharp voltage output is produced.

In the case of a NW with a large Au particle that fully covers its top (Fig. 3I), the metal tip is directly in touch with the Au tip at the beginning of the forced displacement. Because of the conducting channel at the compressed side of the NW, the polar charges produced by the displaced Zn2+ and O2– ions caused by the PZ effect are immediately neutralized by external free charges as soon as they are produced by the deformation. Therefore, no accumulative potential profile is formed at the M-S interface, and no measurable output voltage is detected in the experimental setup we have designed (Fig. 1C).

We also measured the PZ voltage output for the same samples of aligned ZnO NWs in the AFM tapping mode. In this case, the deformation occurred longitudinally and there was no side displacement. The NW was vertically compressed, so the voltage created at the top of the NW was negative Embedded Image so long as the base electrode was grounded (Fig. 4A). There was no voltage drop across the width of the NW. Thus, across the metal tip–ZnO interface, a positively biased Schottky diode formed and the electrons could freely flow across the interface. Electrons flowed as the deformation occurred, and there was no accumulation of net charge in the NW volume. This type of slowly “leaked” current produces no detectable signal. Therefore, for the ZnO NWs whose AFM image in tapping mode is shown in Fig. 4B, no output voltage VL was detected beyond the noise level (Fig. 4C).

Fig. 4.

Electromechanically coupled discharging process of aligned piezoelectric ZnO NWs observed in tapping mode. (A) Experimental setup. (B and C) Topography image (B) and corresponding output voltage image (C) of the NWs. The tapping force was 5 nN, tapping frequency 68 KHz, and tapping speed 6 μm/s. The voltage output contains no information but noise, proving the physical mechanism demonstrated in Fig. 3.

We now estimate the possibility of powering nanodevices with the NW-based power generator. From table S1, the PZ energy output by one NW in one discharge event is ∼0.05 fJ, and the output voltage on the load is ∼8 mV. For a NW of typical resonance frequency ∼10 MHz, the output power of the NW would be ∼0.5 pW. If the density of NWs per unit area on the substrate is 20/μm2, the output power density is ∼10 pW/μm2. By choosing a NW array of size 10 μm × 10 μm, the power generated may be enough to drive a single NW/NB/nanotube-based device (2628). If we can find a way to induce the resonance of a NW array and output the PZ-converted power generated in each cycle of the vibration, a significantly strong power source may be possible for self-powering nanodevices. Furthermore, if the energy produced by acoustic waves, ultrasonic waves, or hydraulic pressure/force could be harvested, electricity could be generated by means of ZnO NW arrays grown on solid substrates or even on flexible polymer films (29). The principle and the nanogenerator demonstrated could be the basis for new self-powering nanotechnology that harvests electricity from the environment for applications such as implantable biomedical devices, wireless sensors, and portable electronics (29).

Supporting Online Material

www.sciencemag.org/cgi/content/full/312/5771/242/DC1

SOM Text

Figs. S1 to S6

Table S1

Movies S1 and S2

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article