Fabrication of High-Efficiency Piezoelectric Energy Harvesters of Epitaxial Pb(Zr,Ti)O₃ Thin Films by Laser Lift-off

Introduction

Recently, piezoelectric vibration energy harvesters (PVEHs) have attracted attention as promising power sources for battery-free wireless sensor nodes. Compared with other types of vibration energy harvesters, such as electrostatic or electromagnetic generators, PVEHs are advantageous because they are small, simple, and have high energy efficiency because piezoelectric materials have excellent electromechanical transfer characteristics.

Conventional MEMS PVEHs are composed of PZT-deposited silicon (Si) cantilevers. To maximize the output power of a PVEH, its resonance frequency should be adjusted to the peak frequencies of environmental vibrations, typically less than 200 Hz (Kim, Priya, and Kanno 2012). Thus, thinner cantilevers with large tip mass are preferable. However, thin Si cantilevers are brittle, making them break easily under large vibrations. One way to improve the toughness of PVEHs is to use metal cantilevers (Cao, Zhang, and Kuwano 2012; Kanda et al. 2009; Lin and Wu 2013; Tsujiura, Adachi, and Kanno 2012; Tsujiura et al. 2013). The fracture toughness of austenitic stainless steel (SS304) is more than 100 MPam^1/2, a hundred times larger than that of Si (Ericson, Johansson, and Schweitz 1988; Murase et al. 1993). Metal cantilevers can be reduced in thickness to less than a few tens of micrometers while maintaining sufficient toughness. These properties of metal cantilevers make it easy to lower their resonance frequency.

On the other hand, the most important research in MEMS PVEHs is improving power-generation efficiency. Piezoelectric materials with high piezoelectric coefficients and low relative dielectric constants are required to enhance power generation efficiency (Yeager and Trolier-McKinstry 2012). Conventional MEMS PVEHs are usually made of Pb(Zr,Ti)O₃ (PZT) thin films with compositions near the morphotropic phase boundary. PZT thin films on Si substrates exhibit large piezoelectric coefficients e_31,f (–8 to –10 C/m²), but their relative dielectric constant is ~1,000 (Tomioka et al. 2012; Trolier-Mckinstry and Muralt 2004). On the other hand, c-axis-oriented epitaxial PZT thin films have not only high piezoelectric coefficients but also low dielectric coefficients (Yeager and Trolier-McKinstry 2012). However, epitaxial PZT thin films are usually deposited only on epitaxial substrates such as MgO and SrTiO₃ or on Si substrates with multiple buffer layers (Isarakorn et al. 2010; Kanno, Fujii, and Takayama 1997; Yin et al. 2012; Yokoyama et al. 2005), making it difficult to microfabricate epitaxial PZT thin films including the substrate by conventional methods.

Morimoto et al. fabricated PVEHs of epitaxial PZT thin films, transferred from MgO substrates to stainless cantilevers, which had better power generation performance than PVEHs of polycrystalline PZT thin films (Morimoto et al. 2010). They fabricated the epitaxial PZT thin film on a MgO substrate; after bonding the PZT and stainless cantilever, they wet etched the substrate. However, this method often overetches the thick MgO substrate, degrading or damaging the PZT film. An alternative method is laser lift-off (LLO), a dry process (Tsakalakos and Sands 2000; Tsakalalos et al. 2003). In LLO, an excimer laser irradiates the PZT thin film from the backside of the MgO substrate, allowing the epitaxial PZT thin films to be removed from the MgO. Energy harvesters fabricated by LLO are reported by Do et al. (2012; 2013); Lee et al. (2010); Park et al. (2014). However, these energy harvesters were made of non-epitaxial PZT thin films transferred to flexible substrates.

In this study, we fabricated high-efficiency energy harvesters of epitaxial PZT thin films on stainless steel cantilevers. Using LLO, we transferred c-axis-oriented epitaxial PZT thin films to microfabricated stainless steel cantilevers, and then evaluated the power-generation efficiency of the harvesters.

Device fabrication

We fabricated PVEHs of epitaxial PZT thin films transferred to stainless steel (SS304) cantilevers. SS304 cantilevers (10×10 mm², 30 μm thickness) with built-in tip mass were prepared by two-step spray etching (Tsujiura, Adachi, and Kanno 2012; Tsujiura et al. 2013). Figure 1 shows the PVEH fabrication process. The c-axis-oriented epitaxial PZT thin films were grown on (001) single-crystal MgO substrates by radio-frequency (rf) magnetron sputtering (Figure 1(a)). The sputtering conditions of the epitaxial PZT thin films were reported in our previous study (Kanno, Fujii, and Takayama 1997). After PZT deposition, we deposited Cr layer as electrodes (Figure 1(b)).

Figure 1

Fabrication process of PVEHs of c-axis-oriented epitaxial PZT thin films: (a) Deposition of PZT thin film on single-crystal MgO by rf magnetron sputtering, (b) deposition of Cr electrode, (c) bonding with microfabricated SS304 cantilever, (d) excimer laser irradiation from the backside of MgO, (e) removal of MgO, (f) deposition of Pt electrode, (g) schematic of a PVEH of a c-axis-oriented epitaxial PZT thin film

The PZT thin films were transferred to the SS304 cantilevers using LLO. Prior to LLO, we bonded the PZT/MgO substrates to the SS304 cantilevers with epoxy (Huntsman, Araldite Rapid) so the Cr electrodes faced the SS304 cantilever (Figure 1(c)). Then, a KrF excimer laser (λ=248 nm) irradiated the backside of the MgO substrates (Figure 1(d)). Because the photon energy of KrF (5 eV) is lower than the band gap of MgO (8 eV) and higher than that of PZT (3.4 eV), the KrF beam transmits through MgO and is absorbed at the PZT/MgO interface (Do et al. 2012; 2013; Lee et al. 2010; Pandey et al. 2005; Park et al. 2014; Tsakalakos and Sands 2000; Tsakalaos et al. 2003). The absorbed photon energy separates the PZT thin film from the MgO substrate (Figure 1(e)). The laser pulses were scanned over the whole PZT thin films to eventually remove the MgO substrate from the PZT-bonded SS304 cantilevers.

Table 1 shows the LLO conditions. Figure 2 shows an optical image of the transferred epitaxial PZT thin film. The lattice-like surface pattern on the transferred PZT corresponds to the single-shot area of the laser pulses. Because the cantilevers were as thin as 30 μm, we fixed their backsides with electron wax to prevent the laser shots from deforming them. By optimizing the laser energy density to 25–30 mJ/mm², we transferred the c-axis-oriented PZT thin films to the SS304 cantilevers. After transferring the PZT thin films, we deposited Pt surface electrodes by rf magnetron sputtering, completing the PVEHs (Figure 1(f)).

Figure 2

Photograph of piezoelectric energy harvester after LLO

Piezoelectric properties and crystal structure of transferred PZT thin films

After transferring the epitaxial PZT thin films, we evaluated their electrical properties. We found their relative dielectric constant and dielectric loss to be 340 and 3.9%, respectively. These values are typical for c-axis-oriented epitaxial PZT thin films on Pt/MgO substrates (Kanno, Fujii, and Takayama 1997).

We also measured the piezoelectric properties of the transferred PZT thin films from the converse piezoelectric effect of the transferred PZT/SS304 unimorph cantilevers. Because the electrodes and epoxy layers were sufficiently thin, we ignored their effect. The piezoelectric coefficient e_31,f can be calculated by the following equation (Chun, Sato, and Kanno 2013).

where d₃₁ is the transverse piezoelectric constant, s₁₁, s₁₂ are components of elastic compliance, h is the thickness, E is Young’s modulus, ν is Poisson’s ratio, L is the beam length, and V is the applied voltage. The superscript p and subscript s represent the piezoelectric thin films and substrates, respectively. Figure 3 shows the tip displacement and piezoelectric coefficient e_31,f as functions of applied voltage. The tip displacement proportionally increased with applied voltage. The piezoelectric coefficient e_31,f was almost constant at −4.8 C/m², a typical value of epitaxial PZT thin films on Pt/MgO substrates (Morimoto et al. 2010).

Figure 3

Tip displacement and piezoelectric coefficient of the unimorph cantilever of the transferred PZT thin film on a stainless steel cantilever as a function of applied voltage

We evaluated the crystal structure of the transferred PZT thin films by X-ray diffraction (XRD). Figure 4 shows XRD patterns of the PZT thin films before and after transfer. The transferred PZT thin films exhibited strong c-axis orientation. These results indicate that LLO did not degrade the crystal structure of the film, allowing the PZT thin films to be transferred to a variety of materials without chemical etching.

Figure 4

XRD patterns of PZT thin films: (a) PZT thin film on MgO substrate (before LLO) and (b) transferred PZT thin film (after LLO)

Power-generation performance

Figure 5 shows a schematic of the system used to measure the power-generation performance of the PVEHs. We mounted the PVEHs on a vibration exciter and measured the output voltage using a frequency response analyzer (FRA). To measure the vibrational acceleration of the PVEHs, we attached an acceleration pickup to the base of the cantilever.

Figure 5

Schematic of measurement system for PVEHs

First, we evaluated the frequency response of the output voltage and determined the resonance frequency by sweeping the excitation frequency in an open-circuit state at an acceleration of 1.0 m/s². Simultaneously, we measured the cantilever displacement by a laser Doppler vibrometer. To avoid errors caused by the large deflection of the tip in this measurement, we measured the displacement at 5 mm from the fixed end.

Figure 6 shows the frequency response of the output voltage and the cantilever displacement at an acceleration of 1.0 m/s². The voltage and displacement curves had similar shapes, and their peaks both appeared at 143 Hz, where the maximum voltage reached 2.5 V_p-p. Using the half-power bandwidth method, we obtained the quality factor (Q) from the resonance curves of displacement and output voltage to be 157.

Figure 6

Output voltage in an open-circuit state and cantilever displacement at 5 mm from the fixed end as functions of frequency

Although we measured the resonant frequency of the raw SS304 cantilever to be 115 Hz at an acceleration of 1.0 m/s², the PVEH exhibited higher resonance frequency (143 Hz) after transferring the PZT thin films. We attribute this increase to an increase in stiffness of the cantilever from the PZT/epoxy layers.

We then evaluated how the effective electric power depended on the parallel-connected load resistance at the resonance frequency under an acceleration of 1.0 m/s². We calculated effective electric power from the following equation.

where P is the effective electric power, V_p–p is the peak-to-peak output voltage, and R is the load resistance. Figure 7 shows the output power and output voltage as functions of load resistance. The maximum output power of 1.8 μW was obtained at a load resistance of 90 kΩ.

Figure 7

Output power and output voltage as functions of load resistance. The oscillation frequency and acceleration were 143 Hz and 1.0 m/s², respectively

We evaluated the power-generation characteristics of the PVEHs using theoretical equations. The theoretical output power is expressed by the following equation (Renaud et al. 2008).

where m is the tip mass, a is the acceleration of the tip mass, ω₀ is the resonance frequency, K² is the generalized electromechanical coupling coefficient, Q is the quality factor, R is the load resistance, and C is the capacitance. Here, the optimal resistance R_opt to maximize P is expressed by the following equation.

As described above, we calculated Q to be 157 at an acceleration of 1.0 m/s². Then we calculated the fitting K², back-calculated from eq. (3) using the measured output power. The fitting K² was calculated to be 3.28×10⁻³. From eq. (3), we calculated the theoretical optimal load resistance and maximum effective output power at an acceleration of 1.0 m/s² to be 83 kΩ and 1.8 μW, respectively. These values agree well with the measured values, as shown in Figure 7. This result supports the validity of the experimentally measured optimal resistance and effective output power.

Figure 8 shows how the effective electric power depended on acceleration. As the acceleration increased, the output power increased to a maximum of 49 μW at an acceleration of 7.5 m/s². Although the output power theoretically increases with the square of acceleration, the experimental output power increased proportionally. In the previous study, we have observed that the Q of metal-based PVEHs rapidly decreased with the increase of acceleration (Tsujiura, Adachi, and Kanno 2012), and experimental output power was in good agreement with the calculated value if we took into account of the variation of Q. Therefore, proportional increase of output power of this study is also attributed to the decrease of Q with the acceleration.

Figure 8

Output power as a function of acceleration. The oscillation frequency and load resistance were 143 Hz and 90 kΩ, respectively

Table 2 compares the performance of our PVEH with those from other studies. The normalized power density (NPD) is defined as the effective output power divided by the volume of the PVEH (volume of the cantilever) and the square of gravitational acceleration (Janphuang et al. 2014). Note that the NPD values in Table 2 were calculated from the data of the references. The NPD of our PVEH exceeds those of PVEHs using non-epitaxial PZT thin films and is comparable to that of PVEHs using thinned bulk PZT. Transfer by LLO did not cause degradation, unlike wet etching, allowing us to use high-quality PZT thin films to fabricate PVEHs with high power-generation efficiency. Furthermore, thin SS304 cantilevers with built-in tip mass lowered the resonance frequency of the PVEH in spite of their small sizes. These results lead the way to high-efficiency microgenerators composed of epitaxial PZT thin films on flexible metal substrates.

Conclusions

In this study, we fabricated PVEHs of c-axis-oriented epitaxial PZT thin films. The epitaxial PZT thin films were deposited on MgO substrates by rf magnetron sputtering and were transferred to microfabricated SS304 cantilevers using LLO. The transferred epitaxial PZT thin films exhibited excellent piezoelectric properties, and we found no degradation caused by LLO transfer. At an acceleration of 1.0 m/s², the PVEHs exhibited a resonance frequency and generated output power of 143 Hz and 1.8 μW, respectively, and the output power reached a maximum of 49 μW at 7.5 m/s².

Compared with MEMS PVEHs composed of non-epitaxial films on Si cantilevers, our PVEH had higher volumetric power-generation efficiency and lower resonance frequency. From these results, we conclude that LLO is a promising means of transferring epitaxial PZT thin films to fabricate high-efficiency microgenerators.