Microenergy Harvesters Based on Fluorinated Ethylene Propylene Piezotubes

Energy harvesting from vibrations provides power to low‐energy‐consuming electronics for standalone and wearable devices as well as for wireless and remote sensing. In this contribution, compact tubular ferroelectret energy harvesters utilizing a single‐tube design are presented. Such single‐tube harvesters can be fabricated from commercially available fluorinated ethylene propylene (FEP) tubes with wall thicknesses of 25 and 50 μm, respectively, by mechanical deformation at elevated temperature. It is demonstrated that the generated power is highly dependent on parameters such as wall thickness, load resistance, and seismic mass. Utilizing a seismic mass of 80 g at resonance frequencies around 80 Hz and an input acceleration of 1 × g (9.81 m s−2 rms), output powers up to 300 μW can be reached for a transducer with 25 μm thick walls.


Introduction
Nowadays the popularity of mobile electronic devices is undoubtedly growing. Currently, rechargeable batteries are used to power such devices. This charging presents problems in remote locations where no power supplies are available. Therefore, self-powered technologies from ambient sources commonly associated with energy harvesting from heat, light, and/or mechanical vibrations have to be utilized. [1][2][3][4] Vibrational energy harvesting using piezoelectric transducer mechanisms has long been based on piezoceramics such as lead zirconate titanate (PZT). However, with increasing restrictions on the use of toxic lead, alternatives to lead-based materials have to be investigated. This has stimulated an extensive research on lead-free piezoelectric materials in the form of ceramics [5,6] as well as ferroelectric polymers such as polyvinylidene fluoride (PVDF) and its copolymers. [7][8][9][10][11] The latter materials are advantageous to ceramics due to their high flexibility, excellent processability, low dielectric constant, and low acoustic impedance combined with low manufacturing costs. A disadvantage compared to piezoceramics is their lower efficiency due to a small piezoelectric d 33 coefficient of 10-40 pC N À1 , which is more than an order of magnitude smaller than that of PZT ceramics.
In contrast, ferroelectrets or piezoelectrets [12,13] have been shown to exhibit significantly higher d 33 coefficients in the range of 100-3000 pC N À1 due to mechanically soft air voids embedded in a poled semirigid polymer framework. [14][15][16] Several harvesting applications of such material composites have been demonstrated profiting from their very large piezoelectric d 33 coefficients. [17][18][19][20][21][22][23][24][25][26] Among them are cellular polypropylene (PP) devices whose generated power was originally about 1 μW for a harvester of an active area of about 1 cm 2 and a seismic mass of a few grams, for an extrapolated acceleration of g ¼ 9.81 m s À2 . [17][18][19][20][21] Design optimization led in relatively short times to a significantly higher power output of more than 100 μW for somewhat larger masses. [20] However, the disadvantage of PP is that the charge and therewith the piezoelectricity is thermally stable only up to þ60 C. [13] Therefore, other voided polymers have been considered as well. [22][23][24][25][26][27][28] Best results so far have been obtained with ferroelectret harvesters made of laminated films of fluorinated ethylene propylene (FEP) combining good thermal stability with high output power of up to 100 μW for seismic masses on the order of 0.1 g. [28] Other promising substitutes to PP ferroelectrets with better thermal stability are polytetrafluoroethylene (PTFE) piezotubes [29] and FEP tube arrays, [30,31] where the latter exhibit promising d 33 coefficients but have not been explored for energy harvesting applications yet.
Consequently, single air-filled FEP tubes as miniaturized polymer ferroelectret harvesters were investigated. Therefore, single FEP tubes were deformed at elevated temperatures, subsequently metallized on both sides, and then polarized in high electric fields. For the determination of the power output they were exposed to mechanical vibrational stress in the thickness direction (  mode) at low frequencies (10 Hz to 1 kHz) utilizing a seismic mass. The power generated was determined as a function of the exciting frequency, seismic mass, and load resistance for two different FEP tubes with 25 and 50 μm thick walls, respectively. Experimental results were then compared with an existing analytical model and finally with the output powers obtained from harvesters with different geometrical structures.

Results and Discussion
Two commercial FEP tube types from ZEUS Ltd (USA) with equal diameters of 1 mm and wall thicknesses of 25 and 50 μm, respectively, were used for device fabrication. For the forming of a tube to a stadium-like cross section a 30 mm long piece of FEP tube was placed between two parallel metal plates and heated up to þ250 C for 10 min. The distance between the plates was reduced gradually during heating until stabilized by two metallic spacers of calibrated thickness of 0.4 mm. Figure 1 shows typical photomicrographs of such stadiumshaped cross sections of tubes with 25 and 50 μm thick walls, respectively. The shape was preserved at room temperature after removal of the plates. Electrical poling of the produced specimens was conducted by direct-contact charging in ambient air at room temperature. Therefore the tube was first metallized on the upper and lower flat areas with Al electrodes, as shown in Figure 1c (metallized area: 1 mm Â 20 mm) and then charged by applying a bias voltage up to AE6 kV from a high-voltage power supply HSN-35 (FUG GmbH) for a few seconds, sufficiently to fully charge the devices. [31] It should be noted that the voltage V pol at which the maximal residual polarization of the tubular structure is achieved can be approximately estimated as [32][33][34] : where E B corresponds to the threshold electric breakdown field strength in the air channel (Paschen's law), [35] while d wall and d air denote the thickness of the wall and air channel with corresponding relative dielectric permittivities of ε wall and ε air , respectively. It must also be taken into account that the breakdown strength E B in Equation (1) depends on d air . For the present estimation, the values for E B of 62 and 65 kV cm À1 were used for air gaps of 350 and 300 μm, respectively, while ε air ¼ 1 and ε wall ¼ 2.1. Under such conditions, Equation (1) delivers a value of 4.5 and 4.6 kV for the particular structures shown in Figure 1a,b, respectively. After poling, the dynamic piezoelectric d 33 responses of the obtained specimens were determined, as they represent one of the most important parameters of harvester devices. For measuring the dynamic d 33 coefficient, the tubular sample and a seismic mass m s placed on it were accelerated sinusoidally by a Bruel&Kjaer shaker. Thereby, the sample was loaded with two forces, namely the static force m s Â g and the dynamic force, m s Â a, where a is the dynamic acceleration. The dynamic acceleration a was measured with an accelerometer in combination with a charge amplifier. Simultaneously, the charge Q generated by the tubular specimen in short circuit was measured by a second charge amplifier of the same model. In this way, the charge sensitivity S of the ferroelectret harvester could be determined as 9.81 Â Q rms /a rms , where Q rms and a rms are the rms values of charge and dynamic acceleration. [22] In addition, the dynamic piezoelectric coefficient can be calculated as d 33 ¼ Q rms /m S Â a rms for a fixed frequency significantly lower than the resonant frequency. In this work, the dynamic d 33 coefficients were determined for a frequency of 20 Hz. More information about the measuring setup can be found in Section 4.
Exemplarily, the experimental results for simultaneously measured Q rms and a rms for a harvester fabricated from an FEP tube with 50 μm thick walls are shown in Figure 2a,b for various m s , while the calculated charge sensitivities S for the same seismic masses are shown in Figure 2c. One can learn from Figure 2c that the tubular transducer displays remarkable and flat frequency responses for different m s up to the resonance region located between 100 and 200 Hz. In addition, Figure 2d shows for the same harvester the dynamic d 33 coefficients at the fixed frequency of 20 Hz as a function of seismic mass for a whole set of utilized m s values and dynamic loads. It is remarkable that an increase in the dynamic mechanical load up to about 5 kPa does not weaken the piezoelectric response (see Figure 2d): the dynamic d 33 coefficient stays constant at about 80 pC N À1 for seismic masses up to 180 g, which corresponds under the gravity of earth to a static pressure of 88 kPa. At the same time, the present experiments showed that a thin-walled sample has a noticeably higher dynamic response, %290 pC N À1 , which, however, remains constant only up to m s ¼ 80g. For a higher load, it begins to decrease remarkably. Based on these results, it can be concluded that a decrease in the thickness of the wall would lead to a significant increase in the dynamic piezoelectric response under low dynamic loads, and to a significant degradation under elevated mechanical loads.
The same experimental setup was used to measure the output power generated by tubular harvesters excited at various frequencies and load resistances R L . The output power was obtained experimentally from the relation where I and Q R L rms represent the current and charge (rms) through the load resistor, and ω represents the angular frequency of the shaker. The value P N normalized to an acceleration of g ¼ 9.81m s À2 is then given by where a rms is the measured acceleration. Results of the output power P N generated by a thick-walled ferroelectret harvester for m s ¼ 60g and various load resistances R L are shown in Figure 3a as a function of the vibrational frequency. The figure indicates that the output power increases with increasing frequency below the resonance frequency. A maximum value of P N ¼ 20 μW is obtained at the resonance frequency of 140 Hz and an optimum load resistance of R opt ¼ 190 MΩ corresponding to a harvester capacity. C s ¼ 6 pF. This is especially well seen in Figure 3b, which shows the peak output power P max N as a function of the load resistance. As expected, the output power increases at low resistance R L proportional to R L and decreases proportionally to 1=R L for R L much larger than the optimal load resistance R opt . [23] In the present case, the total harvester capacitance C s is greater than the capacitance of the tubular transducer itself, which amounts to about 2 pF. This is understandable due to the relatively large parasitic capacitance between the seismic mass and the support plate compared to the actual transducer.
It was previously revealed that the normalized power output P N generated by a ferroelectret harvester in a load resistance R L in response to an input acceleration g at the circular frequency ω can be written as [17,19] where ω 0 is the resonance circular frequency of the harvester, C S is the total harvester capacitance, consisting of the sum of the capacitance of the tubular transducer and the parasitic capacitance of the measuring setup, and ζ ¼ Δω/2ω 0 is the damping ratio corresponding to half of the half-power bandwidth   www.advancedsciencenews.com www.aem-journal.com Δω/ω 0 . According to Equation (4), the generated power is highly dependent on the seismic mass used. In addition, the normalized peak power P max N generated in an optimal load resistance R opt ¼ ðC s ω 0 Þ À1 can be expressed as [17,19] where Y is Young's modulus of a tubular harvester in the direction of compression, A the transducer area loaded by the seismic mass m s , and t the total thickness of the device. According to Equation (5), P max N is proportional to m 3=2 s . The latter dependence was experimentally verified for ferroelectret energy harvesters based on the longitudinal piezoelectric effect. [17] In this work, the influence of the seismic mass on the output power of tubular FEP harvesters was also investigated. Figure 4a shows the frequency dependence of P N obtained for a thickwalled harvester at various seismic masses m S for optimal load resistances. Two kinds of effects can be easily identified with increasing m s . First of all, the normalized power output grows significantly for increasing seismic mass. As a result, the value of P max N is enhanced from about 3 μW for m S ¼ 20 g to about 100 μW for m S ¼ 160 g.
The second effect is the shift of the resonance maximum toward lower frequencies with increasing seismic mass. The experimentally obtained dependence of P max N on m s is shown in Figure 4b. A fit by the power-law function P max N ¼ c Â m b s , where c and b are the fitting parameters, results in c ¼ (3.73 AE 1.29) Â 10 À8 W g À1 and b ¼ 1.55 AE 0.07 is also shown in Figure 4b. The value for parameter b agrees well with the expected mass dependence from Equation (5) where P max N $ m 3=2 S . The output power P N of a thin-walled tubular harvester (wall thickness of 25 μm) as a function of vibrational frequency, load resistance, and seismic mass shows that the efficiency of harvesters is much higher, reaching P max N values of about 300 μW for a seismic mass of 80 g. Figure 5 compares the frequency responses of P N for thin-and thick-walled harvesters under the same experimental conditions utilizing a seismic mass of 80 g and a load resistance of 330 and 190 MΩ, respectively. It turned out that the peak output power of a thin-walled device is about one order of magnitude higher than that of the thickwalled specimen. Furthermore, the peak position of the generated power is shifted from about 135 Hz to a frequency of about 78 Hz for a thin-walled device, which is a clear advantage of such a harvester, as most of the energy from ambient mechanical vibrations is concentrated at very low frequencies. [36] In the next step, the experimental dependencies shown in Figure 5 were fitted to Equation (4) utilizing the directly accessible parameters such as R L , g, and m s and experimentally determined values, such as the d 33 constant (see Figure 2d) and the separately measured device capacitance C s . The fit of Equation (4) then was used to determine the two variables ζ and ω 0 , which largely depend on the mechanical properties, the geometry of the sample, and the seismic mass. Both fit parameters are listed together with the other fixed parameters in Table 1. In these calculations, experimental values for the dynamic d 33 , determined at 20 Hz, were used. C S was measured separately using a capacitance meter. The obtained fits are also shown in Figure 5 by dashed lines for both devices and agree very well with the experimental results.
The normalized power output of 100 μW, generated with a seismic mass of around 100 g, falls in the range of generated powers of previously reported ferroelectret energy harvesters, e.g., based on cross-linked PP using the d 33 effect. [18][19][20] The power output related to the seismic mass is, however, still lower than that generated by thinner-walled FEP layered harvesters. [28]    The solid lines represent the experimental results for the seismic mass of 80 g, while the dashed lines correspond to a fit using Equation (4) (see Table 1).
www.advancedsciencenews.com www.aem-journal.com However, it has to be mentioned that in the present case the total capacitance C S of the harvesters, including the parasitic capacitance, is in the range of 6 pF and therewith by a factor of at least 3 larger than the actual transducer capacitance. According to Equation (5), the generation of a far larger power can be expected if the parasitic capacitance is reduced. Another effective way to increase power output can be deduced from the comparison between thin-and thick-walled devices. Such a comparison clearly indicates that a further reduction of the wall thickness would drastically improve the power output. Before concluding, it should be noted that the long-term stability of the developed harvesters is an important issue and depends on the temperature, humidity, and utilized mechanical load. It is known for ferroelectrets in general that their resistance to fatigue and aging is mainly determined by two factors: their mechanical and charge storage stability, where the latter is limited by the allowed temperature range. In the present case, the ferroelectrets are FEP-based and they have been shown to provide a lifetime at ambient temperature of up to 50 years and more. [16,[37][38][39] Concerning the mechanical stability, it has been demonstrated in this study that both thin-and thick-walled harvesters have certain limits for the applicable mechanical load. Under high loads, the tubular structures can partially collapse, which leads to the degradation of the piezoelectric response as shown in Figure 2d. Further investigation of device stability factors, such as the frequency dependence of the mechanical fatigue, influence of humidity and temperature, as well as their not trivial interplay, need, however, more detailed research, which is beyond the scope of this work and will be published in an upcoming article.

Conclusions
In this work, compact ferroelectret energy harvesters of tubular design are introduced. These harvesters can be easily fabricated from commercially available FEP tubes with a wall thickness of 25 and 50 μm. With seismic masses of 20-180 g, the generation of power up to 300 μW at frequencies around 100 Hz is possible for an input acceleration of g (rms). It was experimentally verified that the power generated at the resonance frequency into the optimal load resistance is proportional to m 3=2 S . Energy harvesters based on the present design have several advantages: First, they are rather compact: without considering the volume of seismic mass used they have an active area of about 20 mm 2 and a height of 0.4 mm (see also Figure 1c). However, if necessary, the separate tubes can be fused or stacked together, forming an array with a much larger active area. [30,31] Second, harvesters with seismic masses ranging from grams to kilograms can be realized by adjusting the wall thickness, suggesting that devices with thicker walls can withstand large mechanical loads. [29] Third, the use of FEP as a base material ensures good temperature stability [23,28,[37][38][39] sufficient for most room temperature applications. The obtained experimental  Figure 6. Schematic of experimental setup used for energy harvesting evaluation.
www.advancedsciencenews.com www.aem-journal.com results are an excellent base for further optimizing the tube design, which can be used not only in vibration-based energy harvesters, but also in accelerometers.

Experimental Section
The experimental setup for characterizing the present ferroelectret energy harvesters is schematically shown in Figure 6. The setup included a seismic mass m S placed on top of the tubular transducer, both mounted on a platform driven by an electrodynamic shaker (B&K 4809), which was fed by an audio analyzer (dScope Series III, PrismSound) through a buffer amplifier (B&K 2706). The harvester-generated current through a load resistor R L was measured by a charge amplifier (B&K 2635) and rectified by the audio analyzer. In parallel, the acceleration was measured by an accelerometer (B&K 4393) mounted directly on the platform through a charge amplifier (B&K 2635) and the audio analyzer. Both signals were recorded at different frequencies for various load resistors and seismic masses and were used to calculate the frequency dependence of the output power. The same setup omitting the load resistor was used to measure the dynamic d 33 coefficients.