Dual-Function Tactile Sensor with Linear Pressure and Temperature Perception at Low Degree of Coupling

Multiinformation tactile perception is highly important for detecting the change of external environment and realizing the compliant control of the robot. However, the integration of dual‐mode or multimode sensors with low coupling remains a challenge. Herein, a dual‐function tactile sensor with linear pressure and temperature perception is reported. The pressure detection part contains the pyramid conductive film and flexible hemispherical electrode, which works as the piezoresistive material. A spiral nickel film works as a temperature‐sensitive material for sensing the temperature of objects during approaching and while contacting. As a result, the dual‐function sensor can detect the pressure in 0–100 kPa with a sensitivity of 11.5 kPa−1, while the temperature is in 25–65 °C with a sensitivity of 0.0032 °C −1. To be state of the art, the tactile sensor array is used for an active safety control of robotic arms, which can realize the perception and timely avoidance of sudden changes in external temperature and object collisions.

sensor and the external object or the environment. [38,39] For discrete sensors, tactile sensors mainly include piezoresistive, [40] capacitive, [41] piezoelectric, [42] triboelectric, [43] pyroelectric, [44] transistors, [45] and semiconductors. [46] Usually, the temperature sensor has two types, one is thermal resistance type [47,48] and the other is thermoelectric type. [49,50] Significantly, the properties of temperature-sensitive materials are often affected by the pressure. In return, the properties of pressure-sensitive materials are often affected by the temperature too. [40,51,52] As a result, there is an inevitable coupling effect between pressure and temperature sensing. Many attempts have been done to design the decoupled temperature-pressure sensor; however, the coupling effect seems not to be eliminated completely even when using the complicated signal process. [53,54] One of reasons is lack of good sensitivity [55] and linear response [56] in a large scope both for pressure and for temperature in the flexible sensor.
Herein, we present a dual-function tactile sensor with linear pressure and temperature perception at low degree of coupling. In brief, the tactile sensor is integrated by vertically stacking two discrete sensors together. It contains five layers, a temperature-sensitive layer, a pyramid conductive film, a sealing layer, a flexible hemispherical electrode, and a flexible substrate. The interlocking structure between the hemisphere and the pyramid enables the tactile sensor to have a good linearity in the range of 0-100 kPa with a sensitivity of 11.5 kPa À1 . The temperature-sensitive layer is pasted on the upside of the tactile sensor, which also has perfect linearity in the range from 25 to 65°C with a sensitivity of 0.0032°C À1 . Significantly, the proposed tactile sensor is feasible and scalable for e-skins with superior stability, mechanical flexibility, and deformability, which can withstand large compression, bending, and torsion in the applications. Finally, the tactile sensor array is used for an active safety control of robotic arms, which can realize the perception and timely avoidance of sudden changes in external temperature and object collisions. Figure 1a illustrates a 4 Â 4 sensor array for robotic perception with the dual-function tactile sensor. The enlarged sensor unit is illustrated in Figure 1b, which contains five layers, a spiral nickel film as the temperature-sensitive layer, a carbon black/PDMS composite layer on the PDMS pyramid as the conductive film, a PDMS braced frame, a flexible hemispherical electrode, and a PI substrate. Among them, the pyramid conductive film and hemispherical electrode form a conductive path for pressure detection. The dimension of the sensor unit is defined within 10 mm Â 10 mm Â 1 mm. For the designed structure, the pyramid begins to deform ( Figure 1c) under a small normal pressure, while the pyramid structure and the hemispherical electrode are embedded into each other ( Figure 1d) under a large normal pressure (i.e., 5 kPa). Meanwhile, the pyramid substrate is also deformed, resulting in an increase in the contact area. With the further increase of the contact pressure (i.e., 20 kPa), the central pyramid is completely embedded in the hemispherical electrode, and the contact area reaches a saturation (Figure 1e). The outer pyramid enters the contact area, leading to the output current increasing continuously. After determined deformation behavior of the interlocking structures, a two-layer conductive piezoresistive material is fabricated by spraying of carbon black/PDMS composite. Figure 1f shows a pyramidal conductive structure, in which the outer light grey illustrates a low-conductive composite, while the inner dark grey represents a high-conductive composite. In the structure, the current lines are supposed to vertically pass through the outer low-conductive piezoresistive film into the inner high-conductivity film first and then shrink at the interfaces (Figure 1f ). The above conductive film is taken as one electrode in the tactile sensor. The bulk resistance of the pyramidal conductive structure is approximately inversely proportional to the contact area. The relationship between total shrinkage resistance of contact interface and contact pressure can be calculated by Equation (1).

Results and Discussion
where ρ 1 is the resistivity of pyramid structure, ρ 2 is the resistivity of hemispherical electrode, F is the contact force, l is the root mean square value when bump height is normally distributed, and E* is the constant related to Young's modulus and Poisson's ratio. Figure 1g shows an enlarged scanning electron microscope (SEM) image of the pyramid PDMS, in which a clear boundary between pyramid PDMS and piezoresistive films is identified by the red dash line. Nevertheless, there is no clear boundary between two layers of piezoresistive composites because they are fabricated in one cue process as well as using the same carbon black and PDMS with only different concentrations. The equivalent circuit model is illustrated in Figure 1h. The total resistance is sum up of the resistance of flexible hemispherical electrode, resistance of pyramidal structure, and the contact resistance between the two parts ( Figure 1h). Figure 1i shows the SEM image of the contact condition between pyramid and hemispherical electrode, in which the wrinkled structure on the hemispherical surface is clearly observed. Meanwhile, the deformation of the pyramid tip can be identified upon pressure. The nanoscaled wrinkled structure will improve the contact area between it and the pyramid tip, which is the core for high sensitivity in the low-pressure scope. Also, the wrinkle has similar strength to the pyramid tip due to the same PDMS material, leading to the long-term stability under the cycle loading/unloading processes. The coplanar electrode structures are commonly used in linear pressure sensors based on the piezoresistive principle ( Figure 1j). The contact resistance of each contact area of the microstructure piezoresistive film is connected in series. Therefore, the total resistance of the sensor is dominated by the contact resistance at this state. As a result, the coplanar electrode can significantly improve the sensitivity of the sensor. Figure 1k shows the enlarged image of one hemisphere in an uncompressed state.
According to the Hertzian contact theory, the relationship between the load and the contact area can be rewritten as Equation, when the sphere is in contact with the elastic half-plane.
where E 1 and E 2 are the Young's moduli of two objects, ν 1 and ν 2 are the Poisson's ratios of two objects, b is the contact radius between the sphere and plane, R is the radius of the sphere, and W c is the load on sphere. When the pyramid contacts the elastic half-plane ( Figure S1, Supporting Information), the relationship between the load, press-in depth (h), and the contact area (S c ) is where α is the taper angle of the pyramid conductive structure, E* is the constant related to Young's modulus and Poisson's ratio of two objects, W P is applied load to the interlocking structure of the hemisphere-pyramid array, and h c is the depth of contact area.
For the structure shown in Figure 2a, a load W is applied on the hemisphere-pyramid array interlocking structure. When the radius of the hemispherical structure is much larger than the height of the pyramid, the contact area of all pyramid microstructures can be calculated as Therefore, the relationship between the load and the contact area of the hemisphere-pyramid array interlocking structure is Figure 1. Design of the dual-function tactile sensor with linear pressure and temperature perception. a) Schematic illustration of the dual-function tactile sensor array, represented as e-skin. b) Enlarged view and structural component of the sensor unit. Displacement distribution diagrams as object contacting in c) 0 kPa, d) 5 kPa, e) 20 kPa, respectively. f ) Pyramidal conductive structure with a double-layer piezoresistive film. g) SEM image of a pyramidal conductive structure. h) The equivalent circuit model of the pressure sensor. i) SEM image of the contact condition between pyramid and hemispherical electrode. j) Coplanar electrode structure and its equivalent circuit. k) SEM image of one hemisphere, which is in an uncompressed state with four pyramidal conductive structures. l) Digital picture of as-prepared e-skin. Scale bar in (l) is 20 mm.
www.advancedsciencenews.com www.advintellsyst.com linear (formula 5). Furthermore, the above relationship is verified by finite-element analysis (FEA). Figure 2b illustrates the contact sequence between the hemisphere and the pyramid, in which the contact process is from the center to periphery with the black line, the red line, the blue line, and the pink line in sequence. The contact area is calculated from the simulation within applied pressure of 0-100 kPa and accumulated in Figure 2c. In brief, the contact area gradually increases in the center (Area 1) in the low-pressure range and undergoes saturation after the pressure exceeding 30 kPa (Figure 2c black square symbol). Analogously, it shows saturation above 60 kPa in Area 2. Furthermore, since the hemisphere is not in contact with the blue and pink parts in the initial state, the contact area at Area 3 and Area 4 starts to increases at 20 and 70 kPa, www.advancedsciencenews.com www.advintellsyst.com respectively. Finally, the total contact area is the sum of the above four areas, which increases linearly with the applied pressure in the range of 0-100 kPa (Figure 2c green square symbol). Figure 2d-g shows the surface stress distribution of the flexible hemispherical electrode under 100 kPa at the pyramid height of 70, 140, 210, and 280 μm, respectively. Apparently, a larger pyramid height results in an increase deformation of the hemisphere and greater stress. Nevertheless, the contact area decreases with increase in pyramid height at the same array spacing and load (Figure 2l). The contact area tends to be linear with applied pressure at all pyramid heights (Figure 2l). The linearity error is calculated to be 1.79%, 2.04%, 5.49%, and 8.28% for the four heights in sequence. In contrast, the higher heights are much easier to crack under greater stress, resulting in poor stability. Therefore, the pyramid height is set at 140 μm in consideration of both linear error and stability of structures. Furthermore, the distance between each pyramid microstructure has an important effect on the variation of the contact area. Figure 2h-k shows the contact feature between the hemisphere and the pyramid as the distance in the range of 200-500 μm. Three colors represent different contact status called "near, sliding and sticking." In detail, the yellow area represents the hemisphere is near the pyramid structure without contact, and the orange color area represents the two structures contacting together but sliding against each other, while the red area indicates that the two structures stick together tightly. Therefore, the contact area is calculated from orange and red area, and the results are shown in Figure 2m. Upon the same pressure, the larger distance leads to the larger proportion of the contact area between the hemisphere and the plane in the total contact area. The proportion gradually increases to 80% as the distance up to 500 μm. The slope of the contact pressure-contact area curve and the sensitivity increase with increasing spacing of the pyramid array. The linearity error increases from 0.79% to 9.65%. Finally, the distance of each pyramid structure is set at 400 μm.
To fabricate the sensor, the pyramid microstructure is reverse copied from a silicon template first ( Figure S2a, Supporting Information). Then, the carbon black/PDMS mixture is prepared via the same method reported by our group before. [25] The mixtures with carbon black concentrations of 20% and 3.5% are sprayed on the pyramid structure sequentially, forming the two-layer conductive film ( Figure S2b, Supporting Information). It can be seen that the pyramid structure is still very clear ( Figure S2b, Supporting Information). Second, the flexible hemispherical electrode is fabricated, as shown in Figure S3, Supporting Information. The conductive viscous gel is dropped on the coplanar electrode ( Figure S4a, Supporting Information). After precuring for 30 min, a chromium/copper (7/200 nm) composite film is e-beam evaporated sequentially to obtain a flexible hemispherical electrode ( Figure S4b, Supporting Information). A huge depression appeared at the top of the hemisphere under the tangential extrusion of the hemispherical electrodes ( Figure S4c, Supporting Information). However, the hemispherical electrodes are still not detached from the coplanar electrodes, indicating the reliability of the flexible hemispherical electrodes. The hemisphere surface is totally the wrinkled structure; however, there is an obvious transition layer at the edge which can reduce stress concentration ( Figure S4d,e, Supporting Information). From the cross-sectional view ( Figure S4f,g, Supporting Information), the flexible hemisphere is evenly dispersed. The height of the surface wrinkles is greater than the gap between the wrinkles, which leads to the formation of a stable conductive path under the pressure of the wrinkled surface. Finally, the tactile sensor array has 80 Â 80 Â 1.2 mm 3 size (Figure 1l), which will be used in the human-robot cooperation in the following.
Upon pressure, the relative displacement of hemisphere to pyramid is calculated to be 0-0.4 mm. The displacement of structure will change the contact area, resulting in current ratio (ΔI/I 0 ) change of the sensor (Figure 3a). The two curves coincide well with each other in the whole range (Figure 3a). As a result, the relationship between ΔI/I 0 and pressure in the range of 0-100 kPa is plotted in Figure 3b, in which it exhibits a good linear relationship, resulting in a formula of ΔI/I 0 = 11.5 P þ 21.5. The fitting error is only 1.8%, which may arise from the slip between the hemisphere and pyramid structure. The finding is anastomosis with the simulation results with only 3.32% deviation, indicating the theoretical model is quite precise. Finally, the sensitivity is calculated to be 11.5 kPa À1 in whole range of 0-100 kPa. The comparisons of proximity and pressure performance of sensors are detailed in Table 1. Some sensors have a large detection range with small sensitivity, some sensors have a high sensitivity but a narrow detection range, while others have a low linearity in the detection range. Both these achieved sensitivity and response range are much higher than those of reported tactile sensors. Figure 3c plots the typical current-voltage curves with applied pressure in the range of 0.5-90 kPa. These curves show a nearlinear relationship under the applied pressures, indicating a good ohmic contact between the pyramid conductive film and flexible hemispherical electrode. Furthermore, the sensor exhibits a capacity to identify 0.5 and 1 kPa (Figure 3c inset). Figure 3d depicts the response curve of the sensor upon 0.5 kPa, from which the response and recovery times are calculated to be both %50 ms. The response time is a little longer than that of the inorganic materials-based pressure sensor; however, it is still in the similar range to those of tactile sensor based on PDMS composite in piezoresistive mode. The relative slow response can be attributed to the usage of viscoelastic PDMS in the sensor. Furthermore, the continuous responses under 5, 10, 20, 30 kPa are plotted in Figure 3e. The stable and repeatable response under the all applied pressures indicates that the sensor has a practicable usage in a large range for pressure sensing. Finally, the stability of contact process is evaluated by a long-term cyclic loading (50 kPa) and unloading (0 kPa) process up to 1.8 Â 10 4 s (Figure 3f ). Specifically, the resistance changes in the initial 500 s and last 500 s of the test are compared, resulting in similar behavior at the two durations (Figure 3f inset). The quite good stability of the tactile sensor is significantly important for the practice usage in future.  Figure 4a inset shows a digital picture of a sensor unit of the as-prepared dual-function tactile sensor. The resistance value of the temperature sensor at room temperature (20°C) is 700 Ω, which is considered as the initial resistance. The resistance variations (ΔR/R 0 ) increase as the temperature increases without pressure (Figure 4a). The experimental data can be fit with a line (Formula 6), leading to a straight slope of 0.0032 without pressure ( Figure 4a). The performance is considered to be of good sensitivity, considering the temperature sensor based on the thin-film metal and at noncontact model. [57][58][59][60][61] ΔR=R 0 ¼ 0.0032T À 0.0642 (6) Figure 4b shows the continuous responses at 35, 45, 55°C, respectively. The response time is about 1.5 s at each temperature, while the recovery times are 60, 90, and 100 s at 35, 45, 55°C, respectively. The recovery time is not linear with the perceived temperature, which is due to the convective heat transfer between the temperature-sensitive material and environment. Therefore, it requires more time to recover to the same resistance of the sensor in the high-temperature environment (Figure 4b). Furthermore, the stable and repeatable response at all applied temperatures (Figure 4b) indicates the sensor has a practicable usage in a large range for temperature sensing. Figure 4c plots the resistance change with the temperature fluctuating in 20-80°C repeatedly, which exhibits a very stable output. We also compare the capacitance change in the last 2500 s with its initial 2500 s, resulting in the same level (Figure 4c inset). The stability of resistance plays an important role in the usage on ambient temperature perception.
In practical applications, the simultaneous and independent detections of multiple stimuli are necessary. Figure 4d shows the effect of pressure on temperature detection. The temperature sensor output fluctuates around 25°C (Figure 4d upside) under the applied pressure as the contact object is at 25°C. Figure 4e shows simultaneously detecting object temperature and contact pressure. The sensitivity of the pressure sensor decreases as the temperature increases. It is because the thermal expansion of the  piezoresistive film causes the resistance to increase, leading to ΔI/I 0 decreasing at 100 kPa. For practical applications, it is required to use the dual-function tactile sensor array for spatially resolved pressure and temperature measurements. Figure 5a shows a sensor array with a 10 and 20 g weights at 50°C placed on it. The heated weight comes into contact with the dual-function tactile sensor, causing the temperature sensed by the corresponding unit to increase ( Figure 5b). As shown in Figure 4d, the temperature sensor output increases slightly under the applied pressure, indicating that the temperature sensor output is determined by the temperature of the weight rather than the actual weight. Larger weights have a larger contact area, leading to the fact that temperature detected by the sensor is closer to the temperature of the object under the larger volume of the weight. Due to the convective heat transfer between the object and the outside air, the units surrounding the object can also sense the temperature. The larger the volume of the object, the more thermal radiation is received around the weight due to the large volume of the weight, resulting in an increase in the temperature of the adjacent units. For pressure sensors, the units below the object are compressed, leading to localized current signals of corresponding units (Figure 5c). In contrary, the surrounding units are minimally unaffected. In addition, the pressure sensor can also be applied for continuous mapping. Figure 5e shows that a finger presses with different pressures. The temperature sensor output increases to 30°C rapidly, when the finger just contacts the sensor. The pressure applied by the finger is about 2.5, 3.4, 5.9, 6.0 kPa in order (Movie S1, Supporting Information). Meanwhile, the temperature is saturated at 30°C (Figure 5d). Finally, we use the sensor array for continuous pressure mapping (Figure 5f ). Figure 5g shows the time series response of  the pressure sensor array when the alphabet "L" is written on it with a finger (Movie S2, Supporting Information). These results highlight the promising applications of the flexible piezoresistive pressure sensor in human-machine interfaces. Finally, the dual-function tactile sensor array (e-skins) is attached to the robot (AUBO, i10) for active safety control applications. The robotic arm is set to move from the distal end to the proximal end (Movie S3, Supporting Information). The motion of the arm has been controlled by the posture changes when it encounters the different environments (i.e., contact pressure or temperature). Figure 5h schematically shows the safety control strategy of the robot arm based on as-prepared e-skins, which is similar to the interrupt control commonly utilized in kinematic control. To this end, the motion of the robot arm changes when the output of the pressure or temperature sensor reaches the set value. Once the robot arm arrives the target position according to the control strategy, the active safety control cycle ends. To identify the tactile sensor applications, the different environmental temperatures are provided by an air gun, while the human palm press provides contact pressures. Figure 5i shows the sensor outputs both from inner joint angle and from e-skins at the same time. First, the rotation of joint 0 leads to the robotic arm moving forward. The e-skins on the robotic arm approach the air gun, which makes the environment temperature increase. The resistance of the temperature sensor increases, thereby the output voltage decreases (Figure 5i blue curve). However, the pressure sensor has a little fluctuation at same duration (Figure 5i red curve). Once the output temperature sensor decreases below 1.6 V, the robotic arm stopes to move forward. Joint 0 rotates in reverse, while joint 1 and joint 2 start to rotate away from the heat source. Furthermore, when the contact pressure is applied on the e-skin, the resistance of the pressure sensor decreases and the output voltage value increases, while the temperature sensor is not affected. Once the output pressure sensor increases larger than 0.6 V, only joint 0 rotates in reverse direction, but other joints do not change. The motion control behaviors based on the e-skin are highly significant for human-robot cooperation to make the collaboration more effective, precise, and safe in the future.

Conclusion
In summary, we propose a dual-function tactile sensor with linear pressure and temperature perception at low coupling. The interlocking structure between the hemisphere and the pyramid enables the tactile sensor to have good linearity in the range of 0-100 kPa with a sensitivity of 11.5 kPa À1 . The temperature sensitive layer is pasted on the upper part of the tactile sensor, whose sensitivity can reach 0.0032°C À1 . Finally, the fabricated e-skins are used for many practical purposes such as active safety control of robotic arms, which can realize the perception and timely avoidance of sudden changes in external temperature and object collisions.

Experimental Section
Mechanical and Electronic Simulations: The FEA method was used to design the sensor structure and determine the feature size of the proposed sensor. The structural mechanics simulation was realized by ANSYS (ANSYS 2021, ANSYS Inc. 2020). The change of the resistance was simulated using the commercial software COMSOL Multiphysics (COMSOL Multiphysics 5.5, COMSOL Inc. 2019).
Fabrication of the Pyramid Conductive Film: PDMS/CB mixed solution of 3.5 and 20 wt% was made according to ref. [25]. Pyramid PDMS microstructures were made according to ref. [62]. 3.5 wt% mixed solution was sprayed on the pyramid structure for 20 s first; then, the same procedure was then repeated using a 20 wt% mixed solution. Finally, the pyramid structure was put into the oven to cure at 80°C for 2 h to get a pyramid conductive film. SEM images of the pyramid and pyramid conductive film are shown in Figure S2, Supporting Information.
Fabrication of the Flexible Hemispherical Electrode: The mixed solution was put into a vacuum drying oven for 6 h to completely volatilize n-heptane. The composite was then introduced into a syringe and dispensed on the ordered copper electrode. Finally, Cr/Cu (10/200 nm) was evaporated on the hemisphere. The diagram of the corresponding fabrication process is shown in Figure S3, Supporting Information. The digital images and SEM images of the flexible hemispherical electrode are shown in Figure S4, Supporting Information.
Fabrication of Temperature Sensitive Layer: Spiral nickel electrodes (200 nm) were evaporated on PI films. The electrodes were then placed in a tube furnace for annealing at 385°C for 4 h.
Fabrication of the Device: Flexible substrates and sealing layer were fabricated by PDMS with specific Teflon molds. The flexible electrodes and the temperature-sensitive layer were adhered to the flexible substrate and pyramid conductive film by room temperature vulcanized (RTV) silicone rubber, respectively. The sealing layer and flexible substrate were bonded together by surface plasma treatment. The diagram of the corresponding fabrication process is shown in Figure S5, Supporting Information.
Characterization: Optical images were obtained using Digital Single Lens Reflex (D5600, Nikon). SEM images were taken with field emission scanning electron microscope (FESEM) (SU8010, HITACHI). The mechanical test was carried out with a measurement system containing a motorized test stand connected with a force gauge (ESM1500G, Mark-10). Different temperatures were provided by the thermostat. The electrical properties of the temperature sensor were characterized by an Inductance-Capacitance-Resistance (LCR) digital electric bridge (HG2810B, Huigao Electronic Co., Ltd.). The electrical properties of pressure sensor were characterized by a semiconductor parameter analyzer (Keiteley 2400, Tektronix Inc.). The dynamic properties were tested on a collaborative robot (i10, AUBO Robot Co., Ltd.).

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.