Energy harvester using piezoelectric nanogenerator and electrostatic generator

During the last few decades, low-power microelectronics have been achieved due to technological advances, primarily increased circuitry and energy storage system efficiency.¹ Today, the electrical power between 10 and 100 μW can be adequate as the supplier for many microelectronics applications.² As a result, the interest in energy harvesting (EH), a viable supply source for wearable and mobile electronics, has increased in the last decade.¹ Among the various sources (e.g., solar, thermal, mechanical, and RF), mechanical energy has become an attractive choice since it is abundant in the environment (e.g., human motion, ocean waves, and machinery vibrations).^3–6 Consequently, there has been active effort to develop advanced vibration-based harvesters that convert ambient vibrations into electrical power.

Piezoelectric conversion is a common way to harvest mechanical energy.^1,3,6 To date, most of the piezo-EH devices utilize ceramic-based materials such as lead zirconate titanate (PZT).⁷ However, in such applications (e.g., human body) where the piezoelement may have to bend or stretch, ceramic-based materials rapidly deteriorate due to their brittle characteristic. This problem can be potentially solved by utilizing flexible and soft materials such as piezoelectric nanocomposites (PNCs).^8,9 Over the last decade, the interest in PNCs has increased due to their mechanical flexibility, lower cost, and biocompatibility compared to ceramics.^1,8,10,11 In 2011, Cheng et al.¹² developed a flexible PNC by fabricating a conductive polymer composite film incorporating ZnO nanorod arrays. Ghosh and Mandal¹³ reported a bio-PNC made with a fish scale exhibiting 5.0 pCN⁻¹. Zhang et al.¹⁴ fabricated and characterized a fabric-PNC consisting of barium titanate (BTO) and nanowire-polyvinyl chloride (PVC) wires. In a recent study, our group demonstrated PNCs with different viscoelastic properties based on polydimethylsiloxane (PDMS) matrix, barium titanate (BTO) nanoparticles, and carbon nanotube (CNT) conductive additives.¹⁵ Although PNCs are well-tailored for many applications, their performance depends on specific excitations (impacts, vibrations, etc.). Therefore, hybrid structures, incorporating different materials and/or EH processes, may also be necessary to harvest energy effectively in applications where the irregular and periodic motions are the cases.^16–18 

Electrets, called “electrical magnets,” contain permanent electrical polarization for many years through charged dielectric materials such as Teflon fluorinated ethylene propylene (FEP), Polytetrafluoroethylene (PTFE), and Parylene.^2,19 In the 1960s, West and Sessler's discovery of electret-based microphone significantly expanded the widespread usage of electrets.^19,20 Over the past two decades, due to the advances of high-voltage electret thin films, more researchers have integrated electrets into vibration-based EH structures.^21–24 Pondrom et al.²⁴ demonstrated that an electret-based generator could generate a power of 0.6 mW at 36 Hz. Zhang et al.²⁵ harvested a power of about 0.2 mW by using sandwiched fluoroethylene propylene films. Later, the same group proposed a harvester based on wavy fluorinated ethylene propylene electret films.²⁶ In 2018, Feng et al.²³ demonstrated high-performance gap-closing harvesters using electret-polarized dielectric oscillators. Gong et al.²² showed that a monocharged PTFE electret could be utilized for EH and obtained a power output of ∼35.6 μW in shoe applications. Later, Feng et al.²⁷ proposed an electret-dome device producing a maximum power density of ∼1 mW cm⁻³. In 2020, our group demonstrated a low-cost electret-based device generating power of ∼15 μW.²¹ Researchers demonstrated microelectromechanical systems (MEMS) scale electret-based devices.^28–31 Moreover, fluidic EH based on jumping or sliding drops on electrets were also reported.^32,33

The simultaneous use of multiple energy conversion mechanisms, such as piezoelectric, electrostatic, triboelectric, and electromagnetic, has been proposed to maximize the power output per unit area.^{16,17,34–38} For example, while Xu et al.³⁵ used piezoelectric and electromagnetic conversions, Wang et al.³⁹ suggested a piezoelectric–triboelectric nanogenerator. Khbeis et al.^40,41 demonstrated a hybrid approach that generates a power density of 1.17 μW mm⁻³ combining piezoelectric and electrostatic conversions. Lee et al.⁴² reported a power density of 16 μW cm⁻³ using a hybrid structure for wearable nanogenerators. By comparing stored energies of 0.45 μJ and 0.12 μJ for the hybrid and polyvinylidene difluoride (PVDF) alone, Lagomarsini et al.¹⁸ discovered that a piezoelectric–electrostatic generator could generate four times more energy. Moreover, Madinei et al.⁴³ introduced a MEMS scale hybrid EH device combining piezoelectric and electrostatic conversions.

In this study, we demonstrate a mechanical EH device combining a piezoelectric nanocomposite (PNC) with a metalized electret (Teflon FEP). Our device can harvest energy under cyclic stresses, thanks to its supple, self-recovery, and airbag-based structure. The viscoelastic properties of our PNC can be easily regulated by altering the ratio between the monomer and the cross-linker of the polymer matrix. This functionality will provide our structure with a tunable mechanical impedance, allowing us to develop EH devices for various applications, such as wearables, embedded in shoe soles. While PNCs are well suited for many applications, their conversion efficiency highly relies on mechanical excitation. Here, the electret-based electrostatic conversion can help to increase overall efficiency, where the excitations might be erratic (impacts, vibrations, etc.).

Figure 1(a) depicts our device's structure assembled with an Au-coated PNC, a metalized (Al) Teflon FEP electret, a Styrofoam polymer, and a Cu tape electrode. The PNC is composed of polydimethylsiloxane (PDMS) elastomer, BaTiO₃ (BTO) nanoparticles, and carbon nanotubes (CNTs). The porous PNC structure is preferred since it is less dense than its bulk counterpart, leading to increased impedance matching with soft surfaces such as the human body. In addition, our previous work¹⁵ indicated that softer PNCs with a small elastic module improve energy harvesting performance due to greater deformation for a given load. The Al electrode layer is shared with both the electret and the PNC. Figures 1(b)–1(d) demonstrate the functional mechanism of the device. The arrows illustrate the electric dipole orientations in the PNC. Figure 1(b) exhibits the initial condition without any applied force. Due to the poling method (Corona discharge) limitations, the maximum thickness of the PNC was restricted to 2.5 cm. Based on our previous experience with electret-based electrostatic EH,²¹ the thickness of the Styrofoam, which defines the air gap, was selected as 4 cm for optimal electrostatic EH performance. There is no electron flow in the initial condition, as the charges of the electrodes stay neutralized with the PNC and the electret resulting in zero potential difference between the electrodes. As shown in Fig. 1(c), when the force is applied on the top side of the device (pressing), the PNC is compressed between the Au and Al electrode layers. The piezoelectric response of the PNC initiates a potential difference between two electrodes, resulting in electron flow from the Au to the Al. Eventually, the Styrofoam is also compressed, and the Teflon FEP electret comes closer to the Cu electrode. At this stage, the Al electrode receives electrons from the Cu electrode due to the electrostatic induction. When the force is removed, the device moves toward its initial position (releasing), as exhibited in Fig. 1(d). Due to the self-recovery structure of the PNC, tensile stress occurred in the composite. This resulted in an electron flow from the Al electrode to the Au electrode. Moreover, the Styrofoam tends to recover to its original state, and the positive charges in the Cu electrode attract electrons from the Al electrode to achieve electrostatic equilibrium.

Piezoelectric nanogenerators (PNGs) generate electrical power employing stress-induced electric potential, caused by immobile piezoelectric charges accumulated at the opposite ends of a deformed PNC.⁴⁴ Open-circuit voltage, V_OCP generated by a PNG, can be described by

where ε, d₃₃, Y, h₀, and ε_r are the strain in the perpendicular direction, the piezoelectric coefficient, Young's modulus, the thickness, and the dielectric constant of the PNC material, respectively.³⁹ The relative displacement of the counter electrode to the base electrode in an electrostatic EH creates capacitance variability and charge generation.² With the help of Kirchhoff's law, the voltage across the load resistance V_LE in an electret-based system can be determined by

where V_S, Q, A, ε_r, ε₀, d₀, d, and t are the electret surface potential, instant charge, electret surface area, dielectric constant of the electret, dielectric constant of the air, initial air gap, displacement of the counter electrode, and thickness of the electret, respectively.⁴⁵ 

Figure 2(a) exhibits the fabrication of our PNC described as follows: (1) CNTs and BTO nanoparticles were first dispersed (with the 1:13 weight ratio) in ethyl alcohol (ACS reagent ≥99.5% from Sigma Aldrich). (2) The mixture was magnetically stirred for 5 h, followed by ultrasonication for 1 h. (3) It was subsequently mixed with the PDMS monomer (Sylgard 184 from Dow Chemical) using a revolutionary mixer (KK-400W, Mazerustar) for ∼300 s. (4) The mixture was placed in a vacuumed oven at 70 °C until the alcohol was evaporated entirely. (5) The NaCl salt was added to the mix (to obtain 25% porosity), followed by degassing for 1–2 h. (6) The PDMS curing agent was added to the mixture, which was then uniformly dispersed using the mixer for 3 min. The mass ratio between the PDMS monomer and the curing agent was adjusted as 20:1. (8) The uncured mixture was poured into a mold and degassed in a vacuum chamber. (9) It was then cured in the oven at 120 °C for 20 min. Subsequently, the solidified PNC was kept at room temperature for 24 h to ensure that it was completely cured. (10) It was then inserted into the DI water to dissolve NaCl. (11) The porous PNC sample with a diameter of 3 cm and a thickness of 0.25 cm was removed from the water, and the top surface was coated with Au using an ion sputter coater (SEC MCM-100). (12) Finally, microscopic images of the PNC were taken using a scanning electron microscope (SEM). More details about the PNC fabrication process can be found in our previous work.¹⁵ 

As the electret material, we employed the Teflon FEP (fluorinated ethylene propylene) due to its high breakdown voltage, low cost, and ability to retain surface potentials for a long time. It has a dielectric strength of 100–140 V μm⁻¹ and a standard surface charge density of 0.1–0.25 mC m⁻².^19,20 The metalized Teflon FEP has been used in various applications, including Hubble Space Telescope.⁴⁶ We charged a metalized Teflon FEP (thicknesses of the Al and the Teflon layers are 0.1 μm and 12.7 μm) from the Sheldahl by employing the corona discharge method, which is based on the projecting ions on the surface of an electret.

Figure 2(b) depicts our corona discharge setup, which consists of a probe, a metal grid (removable) located between the probe and the sample, a grounded electrode, and a metal chamber. To generate electric field and plasma, a high voltage DC power supplier (Spellman model CZE1000R) was connected to the probe. To control the electret surface potential V_s, another power supply (Keithley model 248) is connected to the grid. Moreover, our setup includes a heating system with a resistive heater, an AC voltage regulator to control the radiator power, a thermocouple, and a digital thermometer. The Teflon FEP electret was charged to a surface voltage of −1500 V using the corona setup at room temperature (22 °C) through applying −1500 V to the grid and −15 kV to the probe for 10 min.

To obtain a piezoelectric effect, nanocomposites need to be heated, causing the molecules to move more quickly, in conjunction with an external electric field's application, which allows the dipoles to rearrange.^8,11,15 Therefore, our as-fabricated PDMS/BTO/CNT sample was heated to the Curie point along with a strong electric field using the corona setup. To accomplish this, the corona chamber was heated to 150 °C, and then 25 kV (determined using 100 kV cm⁻¹ for 0.25 cm thickness of the as-fabricated PDMS/BTO/CNT) voltage was applied to the sample for 2 h. However, the metal grid was not employed (removed), and the generated ions were exposed directly to the Teflon FEP. The distance between the probe and the sample was measured as 5 cm.

Figure 3(a) displays the piezoelectric meter, PolyK model PKD3–2000, which measures d₃₃ from 1 to 2000 pCN⁻¹ at 110 Hz frequency under 0.25 N force, we utilized to measure the piezoelectric response of our PNC sample. We employed a Monroe Electronics model 279 ISOPROBE^® electrostatic voltmeter, shown in Fig. 3(b), with a potential, V_s (can be used to determine the charge density) of the as-charged Teflon FEP electret. It was well established that Teflon FEP has excellent charge storage capability.¹⁹ After our experiments, no substantial decay was observed in the electret surface potential. The piezoelectric coefficient (d₃₃) of the PDMS/BTO/CNT and the surface potential of the as-charged Teflon FEP were measured and recorded as 49 ± 3 pCN⁻¹ and −1390 ± 100 V, respectively, as shown in Fig. 3(c). The electric field distribution, which depends on geometries and voids, has a noticeable impact on the porous materials' piezoelectric coefficients.^47,48 Piezoelectricity in the PDMS/BTO/CNT, therefore, occurs as a result of the dimensional changes in the dipoles together with the charged voids. Due to the transduction method, porous piezoelectric materials may exhibit large d₃₃ coefficients but very small d₃₁.⁴⁹ Since the demonstrated EH device here operates under compressive (not stretching or bending) loading, it is d₃₃ and not d₃₁ that determines the performance.

Figure 3(d) displays our vibration-based characterization setup, which applies cyclic pressing, used to test our EH device's performance. It consists of a shaker (50 W_rms, approximately ±5 mm) and an amplifier (30 W_rms) from Smart Material EH, an oscilloscope (Tektronix TBS1064), a function generator (Agilent 33120A), and a fixture. In the same figure, our partially assembled EH device (Cu electrode layer is not shown) is also demonstrated. The Styrofoam layer has outer and inner diameters of 3 cm and 2.5 cm and a thickness of 0.4 cm. The assembled EH device has a total volume of ∼4.6 cm³ with a 3 cm diameter and a total height of ∼0.65 cm. When the shaker excites, the moving part (assembly of a rod and rubber stopper) applies cyclic impacts on the top of the EH device. The applied force was measured as ∼1 N using a miniature load cell mounted to the fixture (Futek LSB200 S-Beam Jr.) before testing the EH device. For different testing vibration frequencies, this force measurement was repeated, and the amplitude of the shaker was changed (if needed) to keep the force the same (∼1 N).

First, we measured open-circuit voltage and short-circuit current outputs, shown in Figs. 4(a)–4(d), under the same dynamic load (∼1 N) but different frequencies (5–20 Hz). Figure 4(a) displays the open-circuit voltage outputs of the piezoelectric nanogenerator, PNG, while Fig. 4(b) shows that of the electret-based electrostatic generator, EEG. The measured maximum peak-peak voltage outputs of the PNG and the EEG were ∼4.6 V and ∼500 V at 20 Hz. Very high voltage outputs are not surprising for electrostatic EH, as seen in the literature.^21,22 The release of the PNC occurs at a slower speed than its pressing due to its viscoelastic properties.¹⁵ This resulted in higher positive (pressing) and smaller negative (releasing) sides of the PNG voltage outputs. Due to viscoelastic characteristics (difference in the strain rate between the pressing and releasing) of the PNC, the peak-peak voltage outputs were affected by vibration frequency. As can be seen in the graphs, the EEG's peak-peak voltage values were not affected by the increment of frequency. The electrostatic converters' response depends on relative displacements of the electrodes toward each other. In our device, these displacements were not affected by changing the frequency. The numbers of peaks per second were increased for both the PNG and EEG with the increment of frequency. Figures 4(c) and 4(d) display short-circuit current outputs of the PNG and EEG under the same dynamic loading conditions with those of the open-circuit voltage measurements. The maximum peak-peak current value obtained from the PNG is 1.84 μA (at 20 Hz) and that from the EEG is 1.1 μA (at 10 Hz). It can be seen that the peak value of the current on the positive pulses is slightly higher than that on the negative pulses. The current inconsistency between pulses is possibly caused by charge loss.

Figures 5(a) and 5(b) demonstrate the load voltage and power at various resistances of the PNG and EEG, respectively. Note that the vibrations applied were at 20 Hz with a force of ∼1 N. By increasing load resistances (RLs), the voltage amplitudes rise and then appear to be saturated. The power outputs corresponding to load resistances were obtained by using the below equation:

where V_L and R_L are the measured voltage through the load and its resistance value of the load, respectively. From the graphs, the maximum power value of the PNG can be found as 16 μW at the matched load of 1 MΩ and that of the EEG as 25 μW at the matched load of 10 MΩ. Here, the volumetric power densities can be calculated as 3.5 μW cm⁻³ and 5.3 μW cm⁻³ for the PNG and EEG, respectively. This indicates that the EEG could generate higher power densities than the PNG. By using the power outputs, the load input (corresponding to 1.4 kPa for F = 1 N and A = 7 cm²) and the determined deformation energy density of the PNC (1.1 MJ cm⁻³ for 1-s loading) from our previous study,¹⁵ the energy conversion efficiencies were calculated as 0.24%, 0.36%, and 0.6% for the PNG, EEG, and the combined, respectively.

The PNG's mechanical impedance, consequently its piezoelectric response, can be tuned by changing the mixing ratios between the PDMS monomer and curing agent (such as 5:1, 10:1, and 20:1 by weight) during its fabrication. Our previous work¹⁵ demonstrated that the storage modulus (measured at 20 Hz) of the PNC was changed from ∼2.2 MPa to ∼0.7 MPa (consequently d₃₃ from 20 pCN⁻¹ to 65 pCN⁻¹) by just changing the ratio from 10:1 (harder) to 20:1 (softer).

Moreover, capacitor charging measurements were conducted to characterize the EH performance of our device, with vibrations at 20 Hz and force around 1 N. Figure 5(c) shows the diagram for used circuit configurations. The commercially available EGP10G and 1N4007 diodes were used in the PNG and the EEG circuits, respectively. In both circuits, a 47 μF capacitor and a 10 MΩ resistor were used. A multimeter (Fluke 189) and data logging software (FlukeView) were employed to measure voltages through the capacitors and plot the data. Figure 5(d) shows the capacitor charging by plotting the variations in voltage vs time. Also, an LED bulb powered by the stored energy was demonstrated in the same figure. By employing the below equation:

where C is the capacitance and V_c is the capacitor voltage, the stored energies U were calculated as 74.5 μJ and 25.9 μJ for the PNG and EEG, respectively. This provides us a total stored energy of ∼100.4 μJ in 2-min, corresponding to ∼50.2 μJ min⁻¹.

In conclusion, we have demonstrated an energy harvester, which comprised an in-house fabricated PNC (PDMS/BTO/CNT) and Teflon FEP as an electrostatic layer. The PNC layer not only served as the piezoelectric material but also assisted as the structural component of the device. In addition, the mechanical impedance can be modified by changing the PNC's monomer/cross-linker ratio, allowing us to fabricate devices for different mechanical energy conditions (impacts and vibrations). Measurements were conducted to characterize the device's performance under applied cyclic loading (pressing and releasing) with various frequencies (5–20 Hz). The total power density of ∼8.8 μW cm⁻³ was generated, and the stored energy of ∼50.2 μJ min⁻¹ was obtained by charging of the capacitors. It is encouraging to compare these results with findings, such as the total power output of 41 μW and stored energy of 0.45 μJ, in the literature of piezoelectric–electrostatic harvesters.^18,40–42 Such results also show that our harvester promises a solution for the 10–100 μW power requirement of microelectronics,² where low-frequency motions are available. Furthermore, potential future work on optimized system design and more efficient EH circuitry would also be beneficial.

The authors would like to acknowledge funding from the United States Office of Naval Research (ONR), Grant No. N000140610961. The authors also thank Valerie Rennoll and Ian Maclane for helpful discussions.