Laser direct writing and characterizations of flexible piezoresistive sensors with microstructures

Functional materials with high viscosity and solid materials have received more and more attentions in flexible pressure sensors, which are inadequate in the most used molding method. Herein, laser direct writing (LDW) method is proposed to fabricate flexible piezoresistive sensors with microstructures on PDMS/ MWCNTs composites with an 8% MWCNTs mass fraction. By controlling laser energy, microstructures with different geometries can be obtained, which significantly impacts the performances of the sensors. Subsequently, curved microcones with excellent performance are fabricated under parameters of f = 40 kHz and v = 150 mm·s-1. The sensor exhibits continuous multi-linear sensitivity, ultrahigh original sensitivity of 21.80 % kPa-1, wide detection range of over 20 kPa, response/recovery time of ~100 ms and good cycle stability for more than 1000 times. Besides, obvious resistance variation can be observed when tiny pressure (a peanut of 30 Pa) is applied. Finally, the flexible piezoresistive sensor can be applied for LED brightness controlling, pulse detection and voice recognition.


Introduction
With the development of flexible electronics, biomedical tests, intelligent robots, internet of things and so on, flexible pressure sensors have received extensive attention due to the important roles played in signal acquisition and conversion 1−15 . According to responding signals, flexible pressure sensors can be divided into piezoresistive sensors 3−6 , capacitive sensors 7−9 , piezoelectric sensors 10,11 and triboelectric sensors 12,13 . Among these flexible pressure sensors, flexible piezoresistive sensors have advantages including simple structures, high sensitivity, short response time, and so on 6,16,17 . Besides, continuous emerging of advanced materials including metallic nanowires 5,18−20 , carbon nanotubes 2,3,21,22 , graphene 6,23−25 and MXenes 4,26,27 benefits the performances of piezoresistive sensors. Consequently, flexible piezoresistive sensors with excellent performances have been increasingly investigated.
It has been reported that the performances of flexible piezoresistive sensors can be remarkably improved by introducing microstructures on substrates 6,17,21,23,28−32 . Up to now, the molding method has been regarded as the most efficient and widely-used method for fabricating flexible sensors with microstructures. For example, Park et al. 21 fabricated an electronic skin (E-skin) with microstructures by patterning the composites of polydimethylsiloxane (PDMS) and multiwall carbon nanotubes (MW-CNTs) with etched silicon molds. Bae et al. 23 fabricated a PDMS piezoresistive sensor with hierarchical microstructures using an etched and thermal oxidized copper mold. Inspired by skin epidermis microstructures, Pang et al. 28 developed a piezoresistive sensor by molding PDMS with abrasive papers and subsequently coating with reduced graphene oxide (rGO). Nie et al. 29 fabricated a piezoresistive E-skin with hierarchical mountainridges microstructures by patterning PDMS on a banana leave. However, the molding method is mainly limited to the formation of microstructures for materials with relatively low viscosity (e.g. pure PDMS), which is inadequate for solid materials (e.g. MXenes and porous graphene) and materials with high viscosity (e.g. composites of CNTs at a high mass fraction).
In this paper, a laser direct writing (LDW) method was proposed to fabricate flexible piezoresistive sensors with microstructures using a PDMS/MWCNTs composite (MPC) of an 8% MWCNTs mass fraction. Laser processing has advantages including high fabrication efficiency, low processing cost and high dimensional resolution 33−37 . Subsequently, surface geometry of microstructures can be adjusted by controlling laser energy, which significantly influenced the performances of flexible piezoresistive sensors. When laser parameters were set to be f = 40 kHz and v = 150 mm·s -1 , curved cone-like microstructures can be obtained. This sensor showed ultrahigh sensitivity (21.80 % kPa -1 ), wide detection range (over 20 kPa), short response/recovery time (~100 ms), high resolution (a peanut for 30 Pa) and good cycle stability (1000 times). Finally, this sensor was applied in LED brightness controlling, pulse detection and voice recognition.

Materials and methods
Material preparation MWCNTs were bought from Nanjing XFNANO Materials Tech. Co., Ltd. The specific parameters of the CVD MWCNTs were: external diameter of 10~20 nm, length of 10~30 μm, purity of more than 95%, tap density of 0.22 g/cm 3 , true density of ~2.1 g/cm 3 and electrical conductivity of more than 100 S/cm. The morphology of MWCNTs was imaged by a field emission scanning elec-tron microscope (FESEM) (SUPRA55 SAPPHIRE, Germany) and shown in Fig. S1(a) (see Supplymentary information). PDMS (SYLGARD 184) and the curing agent were supplied by Dow Corning Co. Ltd, USA. Firstly, PDMS was uniformly mixed with curing agent at a mass ratio of 10:1. Subsequently, MWCNTs were gradually added into PDMS under a continuous stirring to obtain uniform mixture of an 8% MWCNTs mass fraction. Next, the mixture was degassed for 10 minutes, uniformly coated on a 1 mm thick glass sheet and cured at 75 °C for 1 hour. Finally, the cured MPC was peeled off from the glass. The morphologies of PDMS and MPC were imaged by a scanning electron microscope (SEM) (JSM-IT500A, Japan), as shown in Figs. S1(b)−(c).

Mechanical characterizations
The mechanical properties including Youngs modulus and lateral contraction of both PDMS and MPC were evaluated by a tensile experiment. One end of a testing sample was fixed with a clamp and another end was stretched with a slider. The stretching force was measured by a force meter (Handpi HP-5N, China) and the size of the testing sample was measured by a digital vernier caliper. Detailed evaluations of mechanical properties are provided in Supplementary information S2.
Laser processing and morphology observation of microstructures An ultraviolet (UV) laser marking machine (SEAL-355-10S, JPT, China) was employed for LDW of MPC. The basic parameters of the UV laser were shown in Table S1. The experimental parameters for LDW were shown in Table S2. The surface morphologies of microstructures were characterized by a SEM and a super deep scene 3D microscope (VHX5000, KEYENCE, Germany).
Design and assembly of flexible piezoresistive sensor Flexible piezoresistive sensor consists of PI (polymide) tapes, interdigital electrodes and a MPC substrate, as shown in Fig. 1. Firstly, a Au film of 50 nm was deposited on a PET (polyethylene terephthalate) film of 50 μm using a magnetron sputtering apparatus (EXPLORER-14, USA). Secondly, the Au/PET film was patterned to fabricate interdigital electrodes by ablating the Au film using a UV laser marking machine. Then, the Au/PET film and MPC substrate were laser cut into certain size. Finally, Au/PET interdigital electrodes were assembled with the MPC substrate and packed with 50 μm thick PI tapes. Typical snake-like (SLI) and concentric-ring shape (CRI) separation areas were designed for fabricating interdigital electrodes.

Electrical characterizations
During the experiments, the sensor was fixed on a horizontal table. A pressure meter was employed to apply pressure to the top surface of the sensor. The electrical signals were measured by a digital source meter (Keithley 2400). When a pressure was applied, variation curves of the sensor including real-time resistance (R-T) and current-voltage (I-V) were acquired and recorded by computer. Besides, dynamic time resistance (R-T) curves were obtained to evaluate the response time and recov-ery time of the sensor.

Applications of flexible piezoresistive sensor
Flexible piezoresistive sensor was employed for LED brightness controlling, pulse detection and voice recognitions. For LED brightness controlling, the sensor was connected in series with an LED, and subsequently paralleled with a direct current power source of 12 V. For pulse detection, the sensor was fixed on the wrist of a tester to obtain resistance variation during pulse beats. For voice recognition, the sensor was attached on vocal cords of a tester to obtain resistance variation during speaking.

Results and discussion
Mechanical properties of materials The thickness of PDMS and MPC decreased with the increasing ε z , as shown in Fig. 2(a). Young's modulus of PDMS and MPC were evaluated by σ z -ε z curves, as shown in Fig. 2(b). Young's modulus of MPC was evaluated to be 3.31 MPa, which was higher than that of PDMS (0.95 MPa). It can be attributed to the higher mechanical stiffness of MWCNTs than that of PDMS 38 . Nevertheless, MPC still showed good flexibility for fabrication of flexible sensor.

Fabrication of microstructures
The LDW processing system of microstructures is shown in Fig. 3(a). A parallel laser scanning strategy (circles of 0.3 mm diameter, offsetting of 1 mm and line spacing of 0.05 mm) was proposed for LDW, as shown in Fig. 3(b). Microstructures were processed under 4 times of orthometric scanning. The schematic of laser processing is shown in Fig. 3(c). When a laser beam was focused on the surface of MPC, the exposed material was ablated due to photochemistry and photothermic effects.  Notably, the ablation can be controlled by adapting parameters including laser repetition frequency f and scanning speed v. Variation of laser power and pulse with laser repetition frequency is shown in Fig. S3. When a low laser repetition frequency f and a fast scanning speed v was adopted, low microstructures were obtained due to a fairly low laser energy density 33,39 , as shown in Fig. 3(d).
With the increase of laser repetition frequency f, the laser power and pulse time were accordingly increased, and the laser spot was enlarged, thus increasing the ablation depth and ablation width. When laser scanning speed v was decreased, the ablation depth and ablation width were also increased due to higher energy density. The curved microstructures can be obtained due to the Gaussian distribution of laser energy 40 , as shown in Fig. 3(e). When a higher laser repetition frequency f and a slower scanning speed v were adopted, a fairly high energy density was obtained, thus bringing about approximate  cylindrical microstructures with larger height, as shown in Fig. 3

(f).
Surface morphology of microstructures Surface morphologies of microstructures under different laser processing parameters were shown in Fig. 4 and Fig. S4. A lower laser repetition frequency or a faster scanning speed resulted in a lower material removal rate, so lower microstructures were observed, as shown in Figs. 4(c) and 4(f). Increasing the laser frequency or decreasing the scanning speed induced an elevated laser energy density, so curved cone-like microstructures were observed, as shown in Figs. 4(a), 4(b), 4(d), 4(e), 4(h) and 4(i). Application of a high laser frequency and a lower scanning speed resulted in an extremely high laser energy density, which caused destruction to microstructures. Therefore, approximate cylindrical microstructures with large height but poor surface were observed, as shown in Fig. 4(g). Besides, quantities of particles around microstructures were found, as shown in Figs.  S4(g, h), which further demonstrated that the microstructures were seriously destroyed due to an excessive high laser energy density.
Performances of the flexible piezoresistive sensor SLI and CRI interdigital electrodes were designed as Figs. 5(a)-5(b). Subsequently, the interdigital electrodes were fabricated by LDW and packed with MPC substrate by PI tapes, as shown in Figs. 5(c)-5(d). The interdigital electrodes were electrically connected by microstructures over the separation area. Near the separation area, electrical current flow along the path of interdigital electrode 1→ microstructure i→ substrate→ microstructure j→ interdigital electrode 2. The total resistance of the sensor R total can be calculated as follow: where R electrodes is the resistance of interdigital electrodes, R Ci and R Cj are the contact resistance between microstructures and interdigital electrodes, R Vi and R Vj are the volume resistance of microstructures, R substrate is the resistance of MPC substance, R row, k (k = 1, 2, … , n) represents resistance of each row. Therein, R electrodes can be ignored due to the excellent conductivity of the gold interdigital electrodes. Thus the total resistance is mainly determined by volume resistance and contact resistance. The equivalent resistance model of flexible piezoresistive sensors is shown in Figs. 5(e)-5(f). When a pressure was applied, the MPC substrate was deformed and the crosslink state of MWCNTs became closer 21 , which decreasing R Vi , R Vj and R substrate . Besides, the contact area between microstructures and interdigital electrodes was increased, thus significantly decreased R Ci and R Cj . As a result, R total decreased with the increasing pressure applied. Further, sensitivity S of flexible piezoresistive sensors was defined as followed: where ΔP is the variation in pressure applied on the sensor, R 0 is the resistance of flexible piezoresistive sensor without pressure, and ΔR is the variation of sensor resistance. The units of ΔP, R 0 and ΔR are kPa, kΩ and kΩ, respectively. Sensitivity properties of the sensors processed at   different laser parameters are shown as Fig. 6(a) and Fig.  S5. When laser parameters were set to Param 7 (f = 45 kHz and v = 100 mm·s -1 ), the sensor demonstrated low sensitivity, which can be attributed to the poor morphologies of the microstructures. When scanning speed was set to 150 mm·s -1 (Param 2, Param 5 and Param 8), sensors showed significantly higher sensitivity and wider detection range than the planar one. It can be concluded that the curved cone-like microstructures brought obvious variation in contact areas under pressure, and the deformation of microstructures increased with the increase of pressure. Especially, when laser processing parameters were set to Param 5 (f = 40 kHz and v = 150 mm·s -1 ), the sensor exhibited the highest original sensitivity of 21.80 % kPa -1 and a wider detection range of over 20 kPa, which was employed for further experiments. The performance comparison between sensors with different designed interdigital electrodes was shown in Fig. 6(b). The original resistances of sensors with CRI and SLI interdigital electrodes were set through a precompression. The resistances of both sensors showed similar trends with ΔP. Namely, the shape of isolation areas in interdigital electrodes hardly affected the sensitivity of flexible piezoresistive sensors. In view of the processing effi-ciency of interdigital electrodes, SLI interdigital electrodes were adopted for further experiments. Properties of the sensors fabricated with Param 5 and packed with SLI interdigital electrodes were shown in Figs. 6(c) and 6(d). As shown in Fig. 6(c), this sensor showed obvious and multi-linear resistance variation with increasing pressure, which indicated good sensitivity. As shown in Fig. 6(d), this sensor showed linear I-V curves under different pressure, which indicated good resistance stability. When dynamic load was applied, the resistance response was shown in Fig. 7 and Video S1. Under stepped load, the sensor showed obvious resistance variation, as shown in Fig. 7(a). Besides, resistance variation can also be observed while tiny pressure (a peanut, ~30 Pa) was dropped, and response time and recovery time werẽ 100 ms, as shown in Fig. 7(b). Figure 7(c) showed the cycle stability of the sensor at 7 kPa, the resistance variation kept stable after 1000 cycle times.

Applications of the flexible piezoresistive sensor
The sensor was applied for LED brightness controlling, pulse detection and voice recognition. In LED brightness controlling, as shown in Fig. 9(a) and Video S2, the initial resistance of sensors was high, so the series connected LED couldn't be lit up before pressing. The LED light up at pressing the sensor with a finger, as a result of Zhang, ref. 6 Wang, ref. 30 Wang, ref.  reduced sensor resistance. When the finger moved away, the resistance recovered and the LED turned dark again. As shown in Figs. 9(b)−9(d) and Video S3, a flexible piezoresistive sensor was attached on the wrist of a male tester of 24 years for detecting wrist pulse. During pulse beat, the resistance of the sensor decreased due to blood pressure on wrist surface, and the recorded wave and magnified signal were shown in Figs. 9(b) and 9(c). Three characteristic peaks called percussion wave (Pwave), tidal wave (T-wave), and diastolic wave (D-wave) can be observed, which indicates the incident wave from the heartbeat, reflected wave from the hands and reflec-ted wave from the lower part of the body, respectively 42−45 . The peaks of P-wave and T-wave (P1 and P2) were respectively related to the systolic and diastolic blood pressure. The delay time ΔT between P1 and P2 was employed to estimate arterial stiffness. Herein, ΔT was estimated to be ~280 ms, which is consistent with literature data of a healthy male in his 20s 44 . The application of the sensor for voice recognition was shown in Figs. 9(e)-9(g) and Video S4. When the tester pronounced letters, resistance of the sensor attached to vocal cords changed with the shock energy. For a certain letter, resistance changes were consistent during test  several times, as shown in Video S4. For the different vowels, the resistance variations varied differently, both in tendencies and amplitudes.

Conclusions
LDW method was applied for the flexible piezoresistive sensor fabrication, with microstructures on PDMS/ MWCNTs composites of an 8% MWCNTs mass fraction and interdigital electrodes on an Au/PET film. By controlling laser energy, microstructures with different geometries were obtained, which modulated the performances of flexible piezoresistive sensors. When laser processing parameters were set to f = 40 kHz and v = 150 mm·s -1 , the obtained curved microcones rendered the sensor continuous multi-linear sensitivity, with ultrahigh original sensitivity of 21.80 % kPa -1 and wide detection range of more than 20 kPa. The shape of the separation areas in interdigital electrodes hardly affected the sensitivity of the sensors. In dynamic responding, the sensors showed short response/recovery time of ~100 ms, high resolution to tiny pressure (a peanut of 30 Pa) and excellent cycle stability (1000 times). Therefore, the flexible piezoresistive sensor was applied in LED brightness controlling, pulse detection and voice recognition. Additionally, the LDW method exhibites advantages of simple process, high efficiency, low cost and general applicability for most materials. Our work provides a simple and versatile method for fabricating flexible piezoresistive sensors using solid materials (e.g. MXenes and porous graphene) or materials with high viscosity (e.g. composites of CNTs at a high mass fraction), where each part of the sensor including the microstructures and the interdigital electrodes can be efficiently fabricated without the mask.