Electrical Characterization of a Unimorph Vibration Energy Harvester with Al/AlN/Al Structure Realized by Magnetron Sputtering

Abstract

In this work, the realization of a unimorph vibration energy harvester with an Al/AlN/Al structure by magnetron sputtering is proposed. Starting from an Al substrate, the device with an Al/AlN/Al structure was obtained by using a magnetron sputtering in two different operative conditions. The realized energy harvester was investigated in the unimorph bender set-up. The electrical characterization was performed by estimation of the AlN d₃₁ piezoelectric coefficient and measurements of the output power. The estimated absolute value of d₃₁ was 0.48 pC/N and the maximum output power was about 17 μW with 9.81 m/s² (rms value) excitation acceleration.

1. Introduction

Vibration based energy harvesting is an important field of research for the development of self-powered wireless sensor network nodes and portable devices. The development of self-powered solutions without the need for batteries is of interest for many applications [1].

For achieving this result, ultra-low power electronics is important in order to reduce the power requirement as much as possible, and energy harvesting from the environment is a way to power wireless sensors and portable devices. Regarding the power range of interest for applications, ultra-low power electronics has an input power that is in the range from microwatt to milliwatt [2,3]. Regarding the energy-harvesting technique, there are different types of vibration-based energy harvesters. There are three main methods: electromagnetic, electrostatic, and piezoelectric [1,4]. In particular, piezoelectric energy harvesting is an easy solution for the conversion of mechanical vibration energy into electrical energy through the piezoelectric effect [1,4,5,6]. Vibration piezoelectric energy harvesters with an obtained output power that is within the input power range of ultra-low power electronics are reported in the literature [7]. Different piezoelectric materials have been studied for energy-harvesting applications, and among them, lead zirconate titanate (PZT) and polyvinylidene fluoride (PVDF) are well known solutions [8]. Another solution suitable for energy-harvesting applications is aluminum nitride (AlN), which has interesting properties such as a favorable figure of merit and low toxicity (it is a lead-free material) [7,9].

An important aspect for applications is the growth of AlN films oriented along the c-axis. Different methods have been investigated for the realization of high-quality AlN films, and among them, the sputtering technique is a well-known method, and aluminum (Al) is among the different studied substrates [7,9,10,11].

Power output values for AlN-based devices are reported in the literature in the range of tens/hundreds of microwatts, even if a comparison of devices is not immediate, as many factors such as design and operating conditions affect the harvester performances [7,9,12].

A case of interest is the Al/AlN/Al structure [11]. After the realization of AlN/Al structure, the second electrode in Al can be obtained by different techniques such as thermal evaporation [11], ion-beam assisted deposition [13], and magnetron sputtering [14].

A preliminary study of AlN film deposited on Al substrate by pulsed-reactive magnetron sputtering has already been presented in a previous paper [15]. Thick films with thickness of 20–30 µm were obtained, and the AlN thick films characterization was performed. The second electrode was realized by silver conductive paint, obtaining the Ag/AlN/Al structure.

In this work, the realization of a unimorph energy harvester with an Al/AlN/Al structure is reported. This structure has Al for both electrodes and was obtained as follows.

We started from an Al substrate and an Al target, and the AlN deposition on Al was made by operating the magnetron sputtering as already reported in a previous paper [15]. Then, as part of the activity of this work, the Al deposition on AlN was performed, and the used operative conditions of the magnetron sputtering for this Al deposition are reported.

The electrical characterization of the device in the unimorph bender set-up was performed. A well-known electromechanical model of the cantilever beam configuration for energy-harvesting applications uses the Euler–Bernoulli beam theory [16,17]. A simple solution, even if it is an approximate solution, is given by the use of only the first natural frequency of the model, for the case of excitation around a natural frequency of the bender [16]. In this paper, the analytical solution in the simple form of the first natural frequency of the model is used for the fitting of the experimental data in order to obtain the estimation of the d₃₁ AlN transverse piezoelectric coefficient. The experimental set-up, the analytical solution of the model, and the corresponding data interpolation function are described in Section 2. The value of the AlN piezoelectric parameter d₃₁ was estimated and the output power was measured. These results are presented in Section 3 and are discussed in Section 4. Finally, the main conclusions are reported.

2. Materials and Methods

2.1. Magnetron Sputtering System

A magnetron sputtering system was used, with a stainless steel chamber that is almost parallelepiped in shape (about 457 × 457 × 612 mm³). It is equipped with a large door (527 × 680 mm²) that allows for an easy insertion or extraction of substrates and samples. Figure 1 shows the inner view of the chamber, where a single cathode with a circular/planar (six inches diameter) aluminum 99.99% target is present. Over the target, there is a shutter, rotatable from the outside of the chamber, and over it, there is a rotatable structure for positioning the substrates over the target.

For the refrigeration of the cathode, a water-cooled closed circuit was used. A safety system with interlocks against mistakes in the system operation is present. Argon and nitrogen gases were used in the magnetron sputtering operation.

The substrate was placed at about 6 cm above the target.

A pulsed-power (TruPlasma DC 4001, Huettinger, Zielonka, Poland; 1 kW maximum output power) operating with a repetition frequency of 50 kHz (with a duty cycle of 90%) was used.

The base vacuum pressure of the sputtering chamber was always less than 0.1 mPa. A preliminary pre-sputtering in Argon was performed before each deposition.

The realization of the Al/AlN/Al structure was carried out in two steps. The first step was to use Al as substrate (commercial AISI 1050A aluminum) and, after smoothing its surface, to obtain the growth of the AlN film on the Al substrate by pulsed-magnetron reactive sputtering. This part was studied in a preliminary research activity [15]. The second step was to mask the AlN/Al structure, and, then, by pulsed-magnetron sputtering, the Al deposition was performed.

2.2. Unimorph Vibration Energy Harvesting

The Al/AlN/Al structure has the substrate in Al (with parallelepiped geometry and a measured thickness value of 1.87 mm), an AlN thick film as the piezoelectric material, and an Al thin film as the second electrode. This structure can be modelled as a parallel plate capacitor, with one plate (deposited Al) having a smaller area than the other (Al substrate).

The sample made with this structure was used as a unimorph-type energy harvester. In Figure 2, a sketch of the unimorph type cantilever is reported: indicated along the length of the beam are the distances from the clamp of the initial and final position (x₁ and x₂ respectively) of the Al film (top electrode), and the length L of the cantilever beam.

The sample was mounted on an electrodynamic shaker (model 2007E of the Modal Shop Inc., Cincinnati, OH, USA) and clamped as a cantilever beam. The shaker was driven by a signal voltage generator and a power amplifier. The frequency was varied, and the excitation acceleration was measured by an accelerometer placed on the clamp of the cantilever beam. The accelerometer output was connected to a signal conditioner and finally to an oscilloscope input channel. On another input channel of the oscilloscope the sample output voltage was visualized.

The unimorph harvester model with harmonic excitation proposed in the literature was used for the analysis of the frequency response of the voltage output [16]. The cantilever beam was assumed with no base rotation, and only the first natural frequency was considered.

2.2.1. Open Circuit Voltage

The open circuit condition is obtained when $R_L\gg {\displaystyle \frac{1}{\omega C_p}}$, where R_L is the load resistance, C_P is the capacitance of the sample, and ω is the angular frequency [16]. In the case of open circuit, the model shows that the root mean square (rms) value of the output voltage (V₀) is as follows [15]:

$$V_0=E{\displaystyle \frac{0.783\sqrt{m L} \frac{\left|{\chi }_1\right|}{C_p}}{\sqrt{{\left(\frac{{\chi }_1^2}{C_p}+{\omega }_1^2-{\omega }^2\right)}^2+{\left(2{\zeta }_1{\omega }_1\omega \right)}^2}}},$$

(1)

where E is the rms value of the excitation acceleration, ω₁ is the first natural angular frequency, ζ₁ is the damping ratio for the first mode, m is the mass per unit length of the beam, L is the length of the beam, and χ₁ is the coupling term for the first mode [15,16].

According to (1), a data fit was carried out by the following interpolation function:

$$V_0=m_1+{\displaystyle \frac{m_2}{\sqrt{{\left({\omega }^2-m_3^2\right)}^2+{\left(\omega m_4\right)}^2}}},$$

(2)

with m₁ in V, m₂ in V/s², m₃ in s⁻¹, and m₄ in s⁻¹.

With respect to Equation (1), in Equation (2), the m₁ term was added in the data fit to account for voltage offset in the measured signal. Therefore, from the experimental data V₀ and E, and the data fit given by Equation (2), the m₂ value was obtained. Comparing Equation (2) to Equation (1), it results in the following:

$$m_2=0.783 E \sqrt{m L} {\displaystyle \frac{\left|{\chi }_1\right|}{C_p}}.$$

(3)

Moreover, for the unimorph harvester, the following relation for χ₁ was used [15]:

$${\chi }_1=-d_{31}{Y}_p{b}_p{\displaystyle \frac{h_s}{2}}{\left.{\displaystyle \frac{d{\mathsf{\Phi }}_1\left(x\right)}{dx}}\right|}_{x=x_1}^{x=x_2},$$

(4)

where Y_p is the Young’s modulus of the piezoelectric material, d₃₁ is the piezoelectric constant, b_p is the width of the piezoelectric material, and h_s is the thickness of the Al substrate. The function Φ₁(x) in (4) is defined as follows [17]:

$${\mathsf{\Phi }}_1(x)={\displaystyle \frac{1}{\sqrt{mL}}}\left\{cosh\left({\displaystyle \frac{{\lambda }_1}{L}}x\right)-cos\left({\displaystyle \frac{{\lambda }_1}{L}}x\right)-{\sigma }_1\left[sinh\left({\displaystyle \frac{{\lambda }_1}{L}}x\right)-sin\left({\displaystyle \frac{{\lambda }_1}{L}}x\right)\right]\right\},$$

(5)

with λ₁ = 1.8751, and σ₁ = 0.7341 [17].

With the realized sample, the following values were used. The measured values were x₁ = 5 mm, x₂ = 72 mm b_p = 18.5 mm, L = 95 mm, h_s = 1.87 mm, and b_s (width of the Al substrate) = 30 mm. Moreover, the aluminum mass density value was taken, ρ_Al = 2700 kg/m³, and AlN Young’s modulus value was taken, Y_p = 345 GPa [18]. The mass per unit length of the beam was calculated: m = ρ_Al b_s h_s = 151.5 × 10⁻³ kg/m.

Finally, all the experimental data were obtained with E = 9.81 m/s².

From Equations (3)–(5), and using the values indicated above, the absolute value of the piezoelectric coefficient results in the following:

$$\left|d_{31}\right|={\displaystyle \frac{C_p{m}_2}{k}}$$

(6)

with $k \approx 1.13\times {10}^9 \mathrm{N}/{\mathrm{s}}^2$, C_p in F, and d₃₁ in C/N.

The capacitance C_p was measured at the first natural frequency by an LCR meter (LCR-6300, Good Will Instrument Co., Taipei, Taiwan). Moreover, as already reported, from the data fit given by Equation (2), the m₂ value was obtained.

By using Equation (6), the estimate of the absolute value of d₃₁ was obtained.

2.2.2. Output Power

In Figure 3, the energy harvester is represented as an equivalent electrical circuit consisting of an AC current source j(t) in parallel with a capacitor (sample capacitance C_P), and the electrical load is represented as a resistor with resistance R_L [12,16].

The maximum output power is obtained with a resistance value [12]:

$$R_L={\displaystyle \frac{1}{{\omega C}_p}}$$

(7)

The output power:

$$P_L={\displaystyle \frac{V_L^2}{R_L}}$$

(8)

where V_L is the rms value of the output voltage measured on the load resistance. The voltage was measured by using an oscilloscope (Yokogawa DL9140, Tokyo, Japan) with a voltage probe.

3. Results

3.1. Sample Realization with Al/AlN/Al Structure

The sample realization was made in two steps: the AlN/Al structure was the first step and then, after masking, the Al deposition.

As a first step, the AlN/Al structure was made following the operative conditions of one of the cases analyzed in the preliminary study [15]. The AlN thick film was made by pulsed-magnetron reactive sputtering, with the operative pressure set at 0.3 Pa. Argon and nitrogen flow rates were set by a mass flow controller for each gas, and the total value (i.e., the sum of nitrogen and argon flow rates) was kept equal to 25 sccm. The nitrogen concentration, i.e., the ratio between the nitrogen flow rate and the 25 sccm total flow rate, was set at 0.4. The output power value of the pulsed-DC power supply used was set equal to 150 W. The process lasted 50 h. As reported in [15], with these operative conditions, the resulting film thickness was h_AlN = 28.7 μm. This value was assumed in the present work as the same operating conditions were used.

As a second step, new with respect to the preliminary analysis, the obtained AlN/Al structure was masked in air by using Kapton tape in such a way that an area of 67 × 18.5 mm² of the AlN upper surface was left uncovered for the Al deposition. The Al deposition was made by using the magnetron sputtering system operated only with argon gas with flow rate of 25 sccm and pressure set at 2 Pa. The output power value (of the pulsed power supply) was set at 100 W. The process lasted 3 h. In this way, a parallel plate capacitor structure was obtained.

In

Table 1. Magnetron sputtering main operative conditions used for AlN deposition on Al substrate and Al deposition on AlN/Al structure.

Deposition	Power (W)	Pressure (Pa)	Operating Time (h)	Ar Flow Rate (sccm)	N2 Flow Rate (sccm)
AlN on Al substrate	150	0.3	50	15	10
Al on AlN/Al structure	100	2	3	25	0

, the magnetron sputtering main operative conditions for AlN deposition on Al substrate and Al deposition on AlN/Al structure are reported.

A copper tape with conductive adhesive (3M 1181 tape) was used to extend the electrode realized by Al deposition. Moreover, the copper tape was also placed on the Al substrate to obtain an easy connection. Kapton tape (Tesa 51408, Tesa, Norderstedt, Germany) was used for electrical insulation. The realized sample is shown in Figure 4.

3.2. Vibration Unimorph Energy Harvester

The sample with Al/AlN/Al structure was clamped near one end to make a cantilever configuration. The sample was mounted on top of the electrodynamic shaker, which was driven by a signal voltage generator and a power amplifier, and the accelerometer was placed on the clamp. The sample output voltage and the electrical signal related to the excitation acceleration were visualized by an oscilloscope. In Figure 5, the experimental set-up is shown.

The electrical load, consisting of the voltage probe connected to the oscilloscope, was R_L = 10 MΩ. The output voltage was measured varying ω in the range of 450–1600 rad/s. The maximum (rms value) of the output voltage was measured at f = 159.4 Hz (the same rms value was measured at f = 159.3 Hz and f = 159.5 Hz). The capacitance value, measured at the frequencies 159.3 Hz, 159.4 Hz, and 159.5 Hz with an LCR meter, was C_p = 3.88 nF. With these values of ω, R_L, and C_p, the open circuit condition $R_L\gg {\displaystyle \frac{1}{\omega C_p}}$ was satisfied.

In Figure 6 the experimental values (full circles) of the open circuit output voltage V₀ (rms value) versus angular frequency are reported. As already mentioned, the experimental data were obtained with the rms value of the excitation acceleration equal to 9.81 m/s². The open circuit output voltage, starting from low frequency, at first increases with frequency as the mechanical structure approaches the resonance vibration frequency. After the mechanical resonance, the output voltage decreases.

By using Equation (2), the numerical fit of the open circuit output voltage experimental values was carried out (continuous line in Figure 6). The values of m₁, m₂, m₃, and m₄ parameters resulting from the data interpolation are reported in

Table 2. Values of the parameters of the interpolation function fitted from the experimental data shown in Figure 6.

m1 [V]	m2 [V/s2]	m3 [s−1]	m4 [s−1]
0.0265	1.3966 × 105	1006.6	46.729

. The analysis of these values is reported in the Discussion.

Then, the frequency was set at f = 159.4 Hz, and twelve different values of load resistance R_L were used in the range 100 kΩ–10 MΩ. For each value of load resistance, the corresponding output voltage V_L was measured. The output power was calculated by using Equation (8), and the obtained results are reported in Figure 7. Among the results, the maximum value was P_L ≈ 17 μW with R_L = 259 kΩ.

4. Discussion

The capacitance of the device with an Al/AlN/Al structure was C_p = 3.88 nF measured at 159.4 Hz. For the discussion of this C_p value, the data reported in [15] relative to a deposition of AlN on Al substrate obtained in the past by magnetron sputtering, with the same operative conditions set for this research work, i.e., h_AlN = 28.7 μm, and a value of AlN relative permittivity ε_33r ≈ 9, are used for the evaluation. The model of parallel plate capacitor is considered. From the relation:

C_P = ε₀ ε_33r A/h_AlN

(9)

with ε₀ ≈ 8.854 × 10⁻¹² F/m, A = 67 × 18.5 mm² (the area of the deposited aluminum in this work), and, as above mentioned, with h_AlN = 28.7 μm, ε_33r ≈ 9, a capacitance value of about 3.4 nF results, which differs less than 15% from the measured value.

The values of the parameters of the interpolation function fitted from the experimental data shown in Figure 6 and reported in Table 2 are discussed.

The m₁ term, as already introduced, accounts for the voltage offset in the measured signal.

The m₂ term gives the estimation of the d₃₁ AlN piezoelectric coefficient. From the value m₂ = 1.3966 × 10⁵ V/s² and by using Equation (6), an absolute value of d₃₁ of 0.48 pC/m resulted, which is very close to the value of 0.52 pC/N obtained with the Ag/AlN/Al structure reported in a previous paper [15]. This value, even if it is not so high, is comparable with the experimental data obtained on AlN films indicated in the literature [19].

For the discussion of the m₃ value, as a preliminary analysis, from Equation (3), it results with $\left|{\chi }_1\right|={\displaystyle \frac{{C_{p}m}_2}{0.783 E \sqrt{m L}}}$ and therefore it can be calculated by using the values reported above, i.e., C_p = 3.88 nF, m₂ = 1.3966 × 10⁵ V/s², E = 9.81 m/s², m = 151.5 × 10⁻³ kg/m, and L = 95 mm. The term ${\displaystyle \frac{{\chi }_1^2}{C_p}}$ in Equation (1) can also be calculated: ${\displaystyle \frac{{\chi }_1^2}{C_p}}\approx$ 89.1 s⁻². By comparing between Equations (1) and (2), it results the relation $m_3^2={\displaystyle \frac{{\chi }_1^2}{C_p}}+{\omega }_1^2$, and now ω₁ can be obtained. From the values, it results with ω₁ ≈ m₃ = 1066.6 rad/s. From the theory, the value of the first natural angular frequency is given by the following relation [16]:

$${\omega }_1={\lambda }_1^2 \sqrt{{\displaystyle \frac{Y I}{m L^4}}}$$

(10)

where the product YI is the flexural stiffness of the beam. For the unimorph cantilever beam of this paper, in first approximation, the flexural stiffness is due to the Al substrate having a rectangular cross-section (thickness value h_s, width value b_s). Therefore, Y is the aluminum Young’s modulus, and I, the area moment of inertia of a rectangular cross-section, is equal to ${\displaystyle \frac{b_s{h}_s^3}{12}}$. The Al Young’s modulus value was taken as Y = 70 GPa [20]. From Equation (9), it results in a value of the first natural angular frequency, ω₁ ≈ 1071 rad/s. This value is in good agreement with the value of 1006.6 rad/s obtained for the m₃ parameter.

Moreover, comparing Equations (1) and (2) yields $m_4=2{\zeta }_1{\omega }_1$. From this relation, the damping ratio for the first mode results in ${\zeta }_1={\displaystyle \frac{m_4}{2{\omega }_1}}$. With m₃ ≈ ω₁, it results in ${\zeta }_1\approx {\displaystyle \frac{m_4}{2m_3}}\approx 0.023$.

The harvester output power was measured by using, as resistive loads, commercial discrete load resistance values and the voltage probe connected to the oscilloscope. The output power presents its maximum, among the analyzed cases (Figure 7), for R_L = 259 kΩ. This is in very good agreement with Equation (7), from which, with f = 159.4 Hz and C_p = 3.88 nF, results in R_L = 257 kΩ.

It can be observed that the measured output power value of 17 μW is within the input power range of ultra-low power electronics.

5. Conclusions

In this paper, an electrical characterization of an unimorph energy harvester with Al/AlN/Al structure realized by magnetron sputtering is presented. This structure was obtained by operating a magnetron sputtering in two different operative conditions. An output power P_L ≈ 17 μW with R_L = 259 kΩ at 9.81 m/s² (rms value) of excitation acceleration was measured, which is a value of interest for ultra-low power electronics. For the estimation of the AlN d₃₁ piezoelectric coefficient, a simple solution was adopted. The Euler–Bernoulli beam theory was used for the unimorph harvester model with harmonic excitation, and only the first natural frequency was considered. The output voltage analytical relation in the open circuit condition was written in the form of an interpolation function with four parameters for the data fit. From one of these parameters, the absolute value of the AlN d₃₁ piezoelectric coefficient was estimated to be 0.48 pC/N.