Piezoelectric thick film for power-efficient haptic actuator

Although human–machine interface (HMI) strongly depends on vision and hearing, it calls for adding the sense of touch.^1,2 Hence, surface haptic based on piezoelectric actuators is developed to generate tactile effect on surfaces.^3–5 Piezoelectric actuators can induce an ultrasonic wave at certain frequencies.^6–8 A squeezed-air film appears between the vibrating surface and a sliding finger, which in turn decreases the coefficient of friction.^9–11 This enables real-time control of the friction coefficient of someone's finger, thus emulating artificial textures.

Bulk piezo-ceramic haptic actuators stand for the nominal technological solution. They enable large deflections and are typically made of PZT ceramics with thickness in the mm-range,^12,13 which require large voltage, typically beyond 100 V, and which prevents using collective fabrication means, inducing fabrication costs difficult to overcome.

One alternative enabling acceptable deflection and technology compatible with mass production is based on piezoelectric thin films.^14,15 However, the main drawback of piezoelectric thin film actuators for haptic is large power consumption induced by their large capacitance. For example, Hua et al. recently demonstrated a transparent haptic device based on 0.5 μm-ITO/2 μm-PZT/0.5 μm-ITO metal–insulator–metal (MIM) stack on Corning Lotus™ NXT Glass (71 × 60 × 0.5 mm³).¹⁵ The haptic device showed an out-of-plane deflection of 1 μm at 10 V_pp at 22.5 kHz. The power consumption of the whole haptic device reached 3.6 W.

Efforts have nonetheless been devoted to developing power-efficient devices. Glinsek et al. demonstrated a transparent haptic device based on a PZT thin film and ITO interdigitated (IDE) electrodes on fused silica glass.¹⁶ The device exhibited a Lamb mode at 73 kHz and an out-of-plane peak-to-peak displacement of 3 μm at 150 V unipolar. The benefit of the IDE structure is a smaller capacitance than MIM (120 pF for each capacitor in this case), which therefore leads to much lower power consumption, namely, 75 mW per actuator (at 150 V unipolar = 92 V_rms). Note though that the footprint of this device is very much reduced (15 × 3 × 0.5 mm³). A simple linear increase in the width of the actuators, in order to make a fair comparison with the previous work, would multiply by 20 the value of the capacitance, reaching then 1.5 W per actuator. These examples not only show that consumption is substantial in various haptic devices from the literature but also that it is not straightforward to come up with a fair comparison between designs.

To obtain low-power haptic devices, our idea is to develop piezoelectric thick films actuated with a MIM structure. This enables decreasing the capacitance value (C) while keeping a similar out of plane deflection at a given voltage, as explained hereafter. The 1D-lateral force F induced by voltage U applied to a piezoelectric layer on a plate is¹⁷ 

where e_31,f is the effective piezoelectric coefficient, t_film is the film thickness, A_film is the cross-sectional area of the film, which equals film thickness times film width (i.e., A_film = t_film × w_film). Equation (1), therefore, simply reads F = e_31,f × U × w_film. This indicates that the induced force does not depend on PZT thickness. As will be shown in the finite element modeling (FEM) section, the increase in the structure's stiffness remains negligible if the film's thickness remains within 10% of the substrate's thickness. In addition, increasing the film's thickness proportionally decreases C, the capacitance of the actuator. As the power consumption is proportional to the capacitance value, this in turn linearly decreases the overall consumption of the haptic device. Indeed, power consumption P equals C × f × U_rms², with U_rms the applied voltage at frequency f.

In this work, we first investigated the impact of PZT film thickness on the performance of a simple haptic device using FEM. It reveals that the device deflection is hardly ever impacted by PZT thickness. A thicker film enables decreasing power consumption as discussed before. Based on this analysis, we fabricated two haptic transducers with identical dimensions but actuated, respectively, with thin (0.5 μm) and thick (10 μm) Pb(Zr_0.53Ti_0.47)O₃ (PZT) films. The thin and thick films were based, on the one hand, on a standard sol-gel approach and, on the other hand, on a composite slurry coupled with an infiltration sol-gel process. The fabricated haptic devices moved by three actuators each exhibited standing Lamb waves at 62.0 kHz (thick films) and 71.5 kHz (thin films). The capacitance of one thin film actuator is 160 nF whereas that of one thick film actuator is only 12 nF, both measured at 1 kHz. When 13 V_pp (peak to peak) were applied to each actuator, the thin and thick film haptic devices exhibited, respectively, a peak-to-peak deflection of 0.4 and 1 μm (limit of detection with a finger) and a consumption of 2.3 W (thin film) and 150 mW (thick film). This experiment, therefore, proves that increasing the thickness of PZT films in haptic devices decreases drastically the power consumption and not at the expense of reducing the haptic effect.

A 2D FEM was performed using COMSOL 5.5 to study the relationship between PZT thickness and deflection of a simple haptic plate. A scheme of the haptic device (34 × 10 × 0.65 mm³) is shown in Fig. 1(a). Three PZT capacitors (2 × 10 mm²) are placed symmetrically on the silicon substrate with spacing of 11.3 mm between each other. Electrodes, passivation, and barrier layers were not modeled because they are too thin to influence the device performance. A voltage of 40 V was applied to each PZT actuator. The piezoelectric coefficient e₃₁ and relative permittivity ε_r were set to −7 C m⁻² and 1000, respectively. The damping loss factor calculated as Δf/f_r, where f_r and Δf are resonant frequency and resonance width at full width at half maximum, respectively, is 0.0063 for 0.65 mm-thick platinized silicon. Figure 1(b) shows the modeled maximum deflection of the haptic device from Fig. 1(a) and each actuator's capacitance as a function of PZT thickness. A linear decrease in the capacitance is observed with PZT thickness, meaning that 1, 10, and 100 μm-thick films, respectively, exhibit capacitances of 180, 18, and 1.8 nF. As expected, the corresponding peak displacements vs PZT thickness of the haptic device spread in a much smaller range, varying from 1.6 to 1.8 μm only. The displacement is slightly improved when PZT increases because adding PZT has a softening effect, visible on the acoustic resonance. Indeed, the frequency of the latter decreases when PZT thickness increases, as displayed in Fig. 1(c). It makes sense because PZT is softer and heavier than Si. Figure 1(d) represents the mode shape of a 10 μm-PZT haptic device at its A₀-Lamb wave mode resonance (66.1 kHz) where the maximum deflection appears.

In this part, only the thick films fabrication process is detailed. The thin film process is very well established, for instance, in Ref. 16 and the details are given in the supplementary material. The comparison between thick and thin film haptic devices will though be performed in the main paper.

PZT thick films were fabricated by the deposition of a composite slurry (1 g/mL), which was synthesized by dispersing crystalline Nb-PZT [Pb(Zr_0.53Ti_0.47)_0.98Nb_0.02O₃] powder into PZT [Pb(Zr_0.53Ti_0.47)O₃] chemical solution deposition (CSD) solution. More details on processing and characterization are provided in the supplementary material. The slurry was deposited on platinized silicon via spin coating and was annealed at 700 °C to sinter the powders and crystallize the amorphous deposit. The deposition-heating-crystallization process was repeated up to seven times to obtain the desired thickness. The cross section scanning electron microscopy (SEM) micrograph of a four-layer-thick film reveals a porous microstructure [Fig. 2(a)] in which individual deposited layers can be observed. Its thickness is 6 μm.

To improve the density of the films, a so-called “x(1C + 4S)” procedure¹⁸ was deployed, in which one composite slurry deposition (1C) is followed by four CSD solution depositions (4S) to enable infiltration of the sol into the composite. x denotes the number of deposition-infiltration cycles. Details about this process can be found in the supplementary material (Fig. S2). The cross section of 4(1C + 4S) films displays a much denser microstructure [Fig. 2(b)] than the sample without infiltration [Fig. 2(a)]. 10 μm-thick films were achieved with a 7(1 C + 4S) procedure [Fig. 2(c)].

The x-ray diffraction (XRD) pattern of the 10 μm-thick film with sol-infiltration [7(1C + 4S)] is shown in Fig. 2(d). A pure perovskite phase is identified using the powder diffraction file (PDF) No. 01–070-4264.¹⁹ The presence and intensity of {100}, {110}, and {200} peaks indicate random orientation, which is a typical feature of nanoparticles-based thick films.²⁰ 

Figure 3(a) shows the polarization-electric field [P(E)] loops of 6 μm-thick PZT film deposited without infiltration. The remanent polarization (P_r) is 25 μC cm⁻¹, and the coercive field (E_c) is 125 kV cm⁻¹. E_c is rather large and linearly shifts as voltage increases. It suggests substantial leakage current as observed in the current density j(E) loops [Fig. 3(b)]. ε_r and tanδ are, respectively, 400 and 0.2, as shown in Fig. 3(c).

Figure 3(d) shows the P(E) loops of 10 μm-thick PZT films deposited by the 7(1 C + 4S) sol-infiltration process. P_r is 27 μC cm⁻¹ and E_c is 42 kV cm⁻¹, values that are standard for PZT films. E_c is divided by three after sol-infiltration. It suggests lower leakage current, which is confirmed in the j(E) loops of Fig. 3(e). At high field, current is indeed much smaller. The j(E) loops also show two sharp peaks in current that are linked with the switching of PZT ferroelectric domains, as expected. ε_r is 1000 and tanδ is 0.05 [Fig. 3(f)], values comparable to the values of our sol-gel deposited PZT thin films (ε_r = 900 and tanδ = 0.08, as shown in Fig. S5).

Two slider haptic devices with three actuating areas each were fabricated according to the modeling [cf. Fig. 1(a)], one based on 0.5 μm-thick sol gel films (for comparison) and one with 10 μm-thick sol-infiltrated PZT films. Both devices have the same dimensions, namely, 34 × 10 × 0.65 mm³.

The three PZT actuators were obtained by depositing and patterning, respectively, 100 nm-thick and 500 nm-thick Pt top electrodes of 2 × 8 mm² on PZT thin and thick films. In Fig. 4(b), the P(E) loop of the marked actuator confirms its ferroelectric nature with P_r = 15 μC cm⁻². The corresponding current curve does not display excessive leakage at high field [Fig. 4(c)]. In addition, the C(E) loop exhibits an expected value of capacitance around 12 nF per actuator [cf. Fig. 4(d)]. The losses are around 5% at maximum, which is a rather low value. The respective curves for the thin film actuators are displayed in Fig. 4(d) [P(E) loop] and Fig. 4(e) [C(E) loop]. The polarization is very similar to the one obtained for thick films. However, the capacitance is much larger, as expected, reaching 160 nF at maximum, with associated dielectric losses around 8%.

The observed resonant frequencies of the thin and thick film actuators are, respectively, 71.5 and 62.0 kHz, corresponding to the frequency at which the deflection is the largest. Note that these values are not exactly the same as the ones predicted by modeling (67 kHz for the thin films and 66 kHz for the thick films) because the actual final dimensions, and especially the lengths, was slightly different from initial design due to imperfect cut.

The out-of-plane displacement of the thick film haptic device was recorded by scanning across the area of the device, as plotted in Fig. 5(a). 120 V peak-to-peak and unipolar (which means 60 V DC + 60 V AC) were applied to the three actuators in series, which corresponds to 40 V applied to each actuator. A typical anti-symmetric Lamb wave A₀ was observed, which agrees well with the mode shape predicted by FEM. The maximum displacement reaches 3.0 μm peak-to-peak. We also observed that only 13.3 V is needed to achieve the minimum deflection that can be detected by a finger, namely 1 μm. A 2D-scan over the entire device surface [Fig. 5(b)] of the deflection at resonance has also been performed. The same Lamb stationary wave with four nodes equally spaced along the length was observed. Note also that the deflection along the y-axis stays unchanged as expected. The thin film haptic device could only withstand 20 V peak-to-peak and unipolar (which means 10 V DC + 10 V AC) for the three capacitors in series, which means 6.7 V per each actuator. The maximum deflection at the Lamb wave resonance at 71.5 kHz only reached 0.2 μm peak-to-peak, as shown in Fig. S7. The deflection is proportional to the applied voltage, so peak-to-peak deflection at 13.3 V should be 0.4 μm, which is 2.5 times less than the thick film device.

Note that increasing the film thickness could induce a change in piezoelectric properties due to stress relaxation or clamping effect, as described in Ref. 21. In our specific case, e_31,f values obtained by fitting the model to experimental deflection were −7 and −5 C m⁻² for thick and thin films, respectively. Therefore, there is a significant increase in piezoelectric properties in our thick films (+40%). The converse e_31,f coefficient of the PZT thick film was also experimentally measured through a cantilever structure.²² Its value is −7.6 C m⁻², which is in line with the value extracted from the haptic experiment. All the details are described in the supplementary material.

As exposed earlier, the purpose of developing PZT thick film-based haptic devices is to reduce the power consumption. The power consumption P to drive one capacitor can be estimated as P = CfU_rms², where C, U_rms, and f are capacitance, rms voltage, and resonant frequency, respectively. In order to perform a fair comparison, we applied the same voltage to the thin and the thick film haptic devices, namely, 40 V unipolar (20 V DC + 20 V AC, at 24 V_rms). The results are displayed in Table I and show that one thin film actuator consumes 750 mW, whereas one thick film actuator consumes 50 mW. Therefore, the haptic device with thick films consumes 15 times less than the one with thin films. As shown before, the thick film haptic device can withstand 120 V with the actuators in series. In this case, the consumption of all actuators is 450 mW. When 40 V (unipolar, peak-to-peak) are applied to the actuator at the resonance, the directly measured power consumption (U_rms × I_rms) is 500 mW, which is in line with the value (450 mW) obtained through Cf(U_rms)². More measurement details are described in the supplementary material.

Finally, we can also compare approximately these values with the ones reported in the literature in the introduction by normalizing the consumption with the actuators' width. Hence, if all actuators were 10 mm in width and if they were driven in order to reach 3–4 μm in deflection, Hua's¹⁵ and Glinsek's¹⁶ would, respectively, consume 3.5 W and 250 mW. In these conditions, our thick film haptic device consumes 450 mW, which is much less than Hua's because our capacitance is smaller. It is though twice as much as Glinsek's but our device enables working at lower voltage (40 V_pp per actuator vs 150 V_pp). Note also that we used a 0.65 mm-thick Si substrate as haptic plate, which is more difficult to deflect than the glass substrate used in Ref. 16 (0.5 mm thick). At constant voltage, modeling shows that the deflection of a haptic device with 0.5 mm-glass is four times larger than the same one with 0.65 mm-Si. Therefore, this indicates that our thick film actuators should only need 10 V_pp to actuate the same glass haptic plate with the same dimension to reach a deflection of 3 μm. Our solution clearly calls for even thicker films, as suggested by our modeling in Fig. 1(b).

As suggested by simple equations, finite element modeling demonstrated that using thick piezoelectric films for haptic devices enables decreasing power consumption. Dense 10 μm-thick PZT films were fabricated. They were based on a PZT composite slurry complemented with a sol-infiltration process. Thanks to their dense microstructure, these PZT thick films exhibit ferroelectric properties comparable with state-of-art sol-gel PZT thin films. Two haptic devices were fabricated, respectively, based on PZT thin and thick films. Both are functional and exhibit an A₀ Lamb wave resonant mode at 71.5 and 62.0 kHz, respectively. The power consumption per thick-film actuator is 50 mW for generating a 1 μm out-of-plane displacement, which represents a 15-fold reduction compared with our 500 nm-PZT thin film haptic device. Our PZT deposition technique enables reaching 10 μm in thickness. Modeling clearly shows that reaching 100 μm in PZT thickness would allow the same deflection with the same voltage, but this would also reduce consumption by one order of magnitude, down to a few mW per actuator.

See the supplementary material for further details about processing of solutions, powders, slurries, and films. Microstructural and electrical characterization of the films and haptic devices is also further described.

The authors acknowledge the Fonds National de la Recherche (FNR) of Luxembourg for supporting this work through the Project No. PRIDE17/12246511/PACE. S.D. and B.M. acknowledge the support of Slovenian Research Agency (core funding P2-0105).