Supplementary materials for Tactile Feedback Display with Spatial and Temporal Resolutions

Transcription

1 Supplementary materials for Tactile Feedback Display with Spatial and Temporal Resolutions Siarhei Vishniakou,, Brian W. Lewis,, Xiaofan Niu, Alireza Kargar, Ke Sun, Michael Kalajian,, Namseok Park, Muchuan Yang, Yi Jing, Paul Brochu, Zhelin Sun, Chun Li,#, Truong Nguyen,, Qibing Pei, Deli Wang,,, * Department of Electrical and Computer Engineering, University of California - San Diego, CA California Institute of Telecommunication and Information Technology, University of California - San Diego, CA Material Science Program, University of California - San Diego, CA Department of Material Science and Engineering, University of California - Los Angeles, CA 90095

2 A. ZnO TFT and pressure sensor, and array fabrication, testing and properties Figure S1 shows an optical image of a ZnO TFT sensor array on a glass substrate. Gate, source, and drain are ITO. The dimensions of one device are 5µmx190µm. '()*$+(,-.$-/0()/1*,$230$/34)(/)$ $,0("4$ 789$+(,$ 789$ -3:0/*$ +(,-$ Fig S1. Optical image of a ZnO TFT sensor array.!"#$%&$ Fig S2. Drain current change in a ZnO TFT as a function of pressure at different gate voltages. The TFT has strong response in the ON state (Vg=+15V) and little response in the off state (Vg=-15V).

3 The ZnO TFT devices show response to pressure applied on top of the TFT. Figure S2 shows that the ON state (Vg = +15V) current linearly increases with applied pressure while there is very little current variation in the OFF state (Vg = -15V). We are currently working on identifying the exact mechanism that creates the pressure response. At this point, we believe that the change of drain current with applied pressure is due to the piezoelectric effect of the ZnO nanostructures (references 34,35 in man text). Our previous study indicates that the nanocrystallites in the sputtered ZnO thin film are aligned with the c-axis and normal to the substrate surface [1]. Under pressure, the aligned ZnO nanocrystals generate piezopotential across the ZnO thin film. The resulting change in the vertical electric field inside the channel modulates the drain current. We observed the increase of the drain current indicating an enhanced gate effect. Figure S3. Current vs time at a steady bias to TFT(with pressure applied at t=9-14s). Figure S3 shows drain current as a function of time while a constant bias is applied to a ZnO TFT. A constant pressure is applied during the interval of [9;15] seconds. Current baseline drifts under a constant bias regardless of applied pressure. We believe the current drift is caused by the charging of the surface states of ZnO nanoparticles in the thin film [2]. This current drift, although small (less than 6% for the first 20 s), prevents reliable data processing of a pressure signal. The surface state effects on the transport properties of a nanostructured FET have been previously carefully studied by our group in the case of single nanowire FETs [3,4]. To solve the problem of current drift, we used a pulsed measurement technique. The TFT is biased for a short time, current is measured, and then bias is removed to allow TFT to return to initial state. This cycle is repeated continuously to obtain drain current as a function of time. This technique limits the amount of time the surface states can charge and results in a steady current at a given bias, as has been done previously in the measurement of CNT transistors [5]. Figure S4 shows the current change with time with pulse frequency of 1Hz, 50% duty cycle and Vg-high = 5V. The baseline is flat during both parts of a measurement cycle. Higher pulse frequencies and shorter duty cycles increase the stability of the current and subsequently, pressure measurement.

4 Fig S4. Current vs time during a pulsed measurement (1 Hz, 50% duty cycle, V g high =+5V, V g low =0V). Using the pulsed measurement technique (Figure S4) and our in-house built single device measurement system (Figure S5), we were able to accurately measure drain current as a function of time as different weights are added on top of the device, as shown in Figure 1E. Next, we found the amount of current change at each pressure by averaging the current measurement in the middle 10 seconds of each weight application. The current at no pressure applied is the baseline current. At every pressure, we subtracted baseline and then divided by baseline to obtain a percentage change. The percentage change is plotted in Figure S6 top. The data follows a linear fit within 0.1%, with residuals shown in Figure S6 bottom. We also calculated the noise level by finding the standard deviation of drain current at the middle 10 s of each pressure application in Figure 1F. The noise level is lower than 1.6 na (see Figure S7). We can estimate the sensitivity of our pressure measurement system by requiring a signal-to-noise ratio of 3 (SNR=3). This extrapolates the sensitivity of our system to be around 15KPa, which is at the level of the human gentle touch [6]. Figure S5. Circuit design for single device testing.

5 Fig S6. (top) Same as figure 2F. (bottom) Differences between linear fit and actual measurements. All measurements are within 0.1% of the fit. Figure S7. The noise level of the pressure measurement in Figure S6 (top).

6 Figure S8. Layout of the 8 x 8 ZnO TFT sensor array. The highlight shows the active channel region with the dimension of 5 µm (L) x 190 µm (w). Figure S9. Block diagram of the design of the TFT array reader. Figures S8 and S9 show the layout of the 8x8 sensor array and the design block diagram of the array reader, respectively. The 8x8 array of ZnO TFTs is arranged in rowand-column fashion to limit the number of electrodes. The entire array is first scanned without any pressure application to acquire a baseline of current levels. The array is then scanned continuously with pressure applied. The devices that are pressed show

7 increased drain current. The difference between the current measurement with applied pressure and current baseline is plotted as a 2D or 3D map to visualize the input pressure signal (see Figure 3 in main text). B. Polymer actuator, array, and array driver Figure S10. An 8x8 polymer actuator array with schematics showing the blocking force and displacement measurements. Actuation of the dielectric elastomer polymer is achieved through high-voltage bias in the 3-4kV range. Figure S10 shows a diagram of the actuator displacement and force measurement. We used an Acuity AccuRange 200 TM laser displacement sensor. The laser displacement system shines a laser beam on the actuator top surface and the reflection spot separation between the original (OFF state) and diffracted (ON state) is measured. Silver paste was smeared on the specific dot to improve surface reflection. For the force measurement, a load cell (Transducer Techniques, GSO-10) was used. Blocking force was measured by placing the load cell shaft right on top of the actuator. The analog data from both sensors were then converted to real values. The input voltage to the actuator array circuit was measured using a high voltage probe and is considered to be accurate. However, due to a resistor network designed to minimize current spikes, limit the maximum current draw, and provide a stable load, actual voltages applied to the actuators were measured to be less than the voltage measured at the source. High-voltage measurements have an additional error due to the design of the high voltage probe.

8 Figure 11. Pulsed actuation at 1 Hz. Figure S11 shows the actuation at 1 Hz with 50% duty cycle, measured from a 1 layer IPN actuator under 3.7kV. Figure S12 shows the measured force produced by an actuator under 3.7kV bias. The actuators consist of one and two IPN stacking layers, and the results show that the output actuation force increases as more layers of IPN films were stacked together. However, the actuation displacement decreases with layer stacking, as shown in Figure S13. Figure S12. Measured actuation force from one or two IPN layered actuators.

9 Figure S13. Displacement from one or two IPN layered actuators. In order to drive the polymer actuators, a high voltage has to be applied with varying duty cycles and frequencies. Figure S15 shows the block diagram of the actuator driver circuit. The actuator arrays were tested as described in the method section in the manuscript. Figure S14. Block diagram of the actuator array driver.

10 C. Integration and tactile feedback display demonstration Figure S15. Pictures of the integrated tactile feedback display system. Figure S16 shows the setup of the integrated system, with the sensor array/reader and actuator array/driver connected through a PC running a Matlab program. Information on the touch sense is measured by the sensor array and the electronics that operate it. That information is digitized and sent to a PC by an Arduino microcontroller system. The PC processes the data in a Matlab environment to obtain a map of pressure information. After the PC processes the data to obtain the pressure data, it either stores it for later replay or converts it to a driving pattern and sends it to another Arduino microcontroller. This second Arduino microcontroller interprets the data it receives to replay the pressure data on the actuator array, thus reproducing the initial touch sense data. The full system allows the recording and reproduction of the pressure data related to the touch sense. One of the key data points necessary to demonstrate the tactile feedback display with touch is the relationship between the input pressure from the sensor side and output force on the actuator side. We were not able to directly measure the output force of the actuator due to limits of our measurement system. However, we propose a possible calibration to obtain a linear relationship between the input force and output displacement of the actuator. The calibration curve, which relates charging time of the actuators to the measured input pressure (green line) as well as the output displacement vs. input pressure relationship (blue line) are shown in Figure S17. We find this to be a reasonable compromise to make as the force required to displace the actuators will increase with larger displacements. Once the measurement between output

11 displacement and perceived pressure is made, it is a trivial matter to apply a calibration curve to obtain a 1:1 relationship between input pressure and output pressure. The red box indicated the detection limit of our current devices, which is about 15 KPa. The human touch pressure range was also shown. ;9.*2$479:)$ %,20"-3"45$6"."4$78$ 79+$:9++,24$0504,.$ ()*+#,$-.,$/.01$ Fig S16. Proposed relationship between the actuator output displacement vs input force on a sensor. This relationship can be attained by adjusting the charging time of the actuators as shown by the green curve. D. Video clip narratives Video clip A: Tactile FD_SI _System_Integration : This movie is a demonstration of our integrated touch pressure sensing and reproduction system. All of the processing is done in real time. The top left image shows our pressure-sensitive TFT array. You can see small allen wrenches with a nonconductive polymer coating pressing on the array, inducing pressure. The pressure is detected in the TFT array in form of a current increase on a specific TFT in the array. This information is measured by a circuit, described in the paper and supplemental materials, and then transmitted to a PC. The PC, running a Matlab program which processes the TFT data, obtains a 2D surface plot of pressure shown in the bottom half of the screen. Please note that stray points are caused by measurement noise and that column 4 and 8 do not respond to pressure due to fabrication faults. This splits the 2D surface plot when column 4 is pressed upon. This surface plot is processed to obtain a graph of local maxima points, shown in the bottom right. These points correspond to the TFT array positions of applied pressure. The local maxima points are then sent from the PC to the actuator array system, reproducing the inputted pressure. This video shows successful one and two point pressure sensing and reproduction. Video clip B: Tactile FD_SI_Spatial_Mapping : This movie illustrates spatial editing of touch pressure. All of the processing is done in real time. Top left image shows pressure applied on the TFT array. During the first 15 seconds original signal is directly reproduced on actuator array in top right. Next 15 seconds show transpose of the original input. For a dot with coordinates (x,y), we actuate dot with coordinates (y,x). Driving pattern is shown on bottom right. Other stray signals are present due to noise. Noise creates local maxima which get interpreted as

12 pressure points. Next 15 seconds show a spatial translation, with coordinates (x,y) shifting to (x+2, y-2). Note that the shift is cyclical, so coordinate 0 becomes 8. Finally, for the last 15 seconds the system returns to direct reproduction of input. Again, this is real-time processing, with the input constantly being applied to the sensor array, but reproduced differently depending on the current mode of the system. References: 1. Sun, K. et al. Crystalline ZnO Thin Film by Hydrothermal Growth, ChemComm, 47, (2011). 2. Bubel, S., Mechau, N., Hahn, H., & Schmechel, R. Trap states and space charge limited current in dispersion processed zinc oxide thin films, J. Appl. Phys. 108, (2010). 3. Dayeh, S. A. et al. Influence of Surface States on the Extraction of Transport Parameters from InAs Nanowire Field Effect Transistors, Appl. Phys. Lett. 90, (2007). 4. Dayeh, S.A., Soci, C., Yu, E.T., & Wang, D. Transport Properties of InAs Nanowire Field Effect Transistors: Effects of Surface States, J. Vac. Sci. Tech. B 25, 1432 (2007). 5. Estrada, D. Dutta, S. Liao, A. & Pop, E. Pulsed measurements Reduction of Hysteresis for Carbon Nanotube Mobility Measurements Using Pulsed Characterization, Nanotechnol. 21, (2010). doi: / /21/8/ Mascaro, S. A. & Asada, H. H. Photoplethysmograph fingernail sensors for measuring finger forces without haptic obstruction, IEEE Trans. Robot. & Autom. 17, (2001).