Sensing-range-tunable pressure sensors realized by self-patterned-spacer design and vertical CNT arrays embedded in PDMS

A pressure sensor design suitable for a broad sensing range with high sensitivity and good stability is highly desirable for the detection of various pressures and meeting the requirements of different applications. Herein, we report sensing-range-tunable piezoresistive pressure sensors realized by self-patterned-spacer design. In the sensors, the two CNT-array layers embedded in PDMS are separated by the proposed self-patterned spacer. With this structure, the realized sensors with large initial resistance designed show tunable response thresholds from 300 Pa to 6.5 kPa while maintaining high sensitivity, which are realized by controlling the spacer thickness and the CNT length. Besides, the vertical CNT arrays have a large specific surface area, which can dramatically change the resistance of the pressure sensors and lead to high sensitivity with nearly 50 kPa−1. Benefiting from the designs of the self-patterned spacer and the advantageous combination of CNTs and PDMS, the pressure sensors also exhibit a rapid response/relaxation time of 24/32 ms, and good long-term stability with durability test over 10 000 loading/unloading cycles. On the other hand, the realized pressure sensors with small initial resistance designed show a typical piezoresistive characteristic. For applications, the pressure sensors with large initial resistance designed are suitable for the anti-noise applications with pressure thresholds to filter unnecessary noise and save power consumption, while the pressure sensors with small initial resistance designed show the capability of detecting mechanical forces and monitoring human physiological signals. Moreover, the self-patterned design and fabrication method of the spacers also show potentials to be applied in the existing works to further enhance or adjust the performance of those pressure sensors, showing great flexibility. This design demonstrates great potentials to be applied in future advanced flexible wearable systems such as health monitoring, human–machine interaction and the Internet of Things.


Introduction
Skin is a vital organ of human body that can protect humans from harm. By mimicking the sensing capabilities of human skin, scientists have developed electronic skin (e-skin) that can sense factors such as pressure, 1-3 strain, 4-6 temperature, 7,8 and humidity, 9,10 showing great potentials in health monitoring, robotics, and the Internet of Things (IoTs). [11][12][13][14][15] As the integral part of e-skin, the pressure sensor can convert the sensed external pressure into electrical signals, which plays an important role in disease diagnosis and motion recognition. 16,17 Pressure sensors can work with different devices such as resistors, 1-3 capacitors, 18-20 piezoelectric, 21,22 and transistors. 23 Among them, the piezoresistive pressure sensor has been widely studied for its high sensitivity and low detection limit. [1][2][3] To meet the various sensing requirements, pressure sensors are also required to be sensitive in various ranges, particularly in high-pressure ranges. Scientists have taken measures to broaden the sensing range of pressure sensors to realize this goal in recent years. Most of them show a higher sensitivity (11.28-57 kPa À1 ) in the initial stage and a lower sensitivity (0.33-1.08 kPa À1 ) in the high-pressure ranges. 3,24,25 For ultrawide pressure ranges, X. Li et al. introduced a exible piezoresistive pressure sensor based on the polyurethane sponge coated with MXene sheets, showing sensitivities of 0.014 kPa À1 , 0.015 kPa À1 and 0.001 kPa À1 in the ranges of 0-6.5 kPa, 6.5-85.1 kPa and 85.1-237.5 kPa, respectively. 26 When the working range is extended to the order of megapascal, S. Doshi and E. Thostenson have reported a exible carbon nanotube-based pressure sensor for a ultrawide sensing range from 0.0025 to 40 MPa, while the sensitivity degenerated to 0.05 MPa À1 . 27 In general, their sensitivity will decrease with the increase in pressure, which is determined by the characteristics of piezoresistive pressure sensors themselves. In other words, the high sensitivity and broad working range are contradictory to each other. Moreover, in some applications, it is necessary for the pressure sensors to respond sensitively only aer exceeding a certain pressure, instead of responding constantly in a broad range, such as a exible keyboard with a certain threshold to avoid mis-touch and exible data acquisition systems, which need to avoid impact from unwanted noises and save power consumption. Therefore, pressure sensors that can adjust the sensing range and possess high sensitivity at the same time are highly desirable.
In this paper, we realize piezoresistive pressure sensors with tunable sensing ranges by self-patterned-spacer design. The vertical carbon nanotube (CNT) arrays are grown by a plasmaenhanced chemical vapor deposition (PECVD) method and the two CNT-array layers embedded in PDMS are separated by the patterned spacer. The pressure sensors with large initial resistance designed can change the response thresholds from 300 Pa to 6.56 kPa while keeping the high sensitivity simultaneously. These properties are realized by adjusting the thickness of the spacer and the length of the CNTs. Due to the large specic surface area of the vertical CNT arrays, the resistance of the pressure sensors changes dramatically within a very narrow pressure range, which can achieve a high sensitivity of about 50 kPa À1 . The proposed pressure sensors also exhibit rapid response/relaxation time of 24/32 ms, and good long-term stability with durability test over 10 000 loading/unloading cycles, while the realized pressure sensors with small initial resistance designed show a typical piezoresistive characteristic. For applications, the pressure sensors with large initial resistance designed are suitable for a exible keyboard with antimis-touch function and exible data acquisition systems with shielding-noise functions, respectively. On the other hand, the pressure sensors with small initial resistance designed not only show the capability of monitoring mechanical forces such as pressing, bending and torsion, but also can be attached to the skin to monitor human physiological signals such as clenching, arm bending, walking, running, drinking, coughing and speaking. Moreover, the specialty and advantage of this selfpatterned-spacer design is that the silicon wafer not only serves as a substrate for CNT growth, but also acts as a natural mask for the patterning of the spacer without affecting the original CNT arrays. This self-patterned spacer separation method is also generally applicable to the other existing works to enhance or adjust their sensor performance, showing great exibility. The sensing-range-tunable pressure sensors realized by self-patterned-spacer design work as a prototype and show great potentials in exible wearable electronics such as health monitoring, human-machine interaction and the IoTs.

Growth of vertical CNT arrays
The metallic multi-walled CNT arrays were synthesized by a PECVD method. First, we used thermal growth to form 1 mm silicon dioxide on a clean silicon wafer. Then, 2 nm Fe as a catalyst was deposited on the Si/SiO 2 substrate. A radio frequency (RF) PECVD apparatus (Kejing OTF-1200X-II-80SL) was used for CNT growth. Aer the chamber was evacuated to high vacuum, the sample in the chamber was heated to 700 C with 7 sccm argon (Ar) and 3 sccm hydrogen (H 2 ). Aer that, methane (CH 4 ) was injected into the chamber at a ow rate of 15 sccm and with the RF power xed at 250 W. During this process, CNTs precipitated from catalyst particles following tip growth theory. Aer the growth, the chamber was cooled down to room temperature with 7 sccm Ar. The length of vertical CNTs can be controlled by adjusting the growing time.
Pressure sensor fabrication Fig. 1 depicts the fabrication process of the pressure sensors realized by the vertical CNT arrays embedded in PDMS and selfpatterned-spacer design. The stable structure of CNT arrays embedded in PDMS was achieved by transfer process with three layers of PDMS. 28 Before fabrication, the PDMS (Dow Corning Sylgard 184) liquid was obtained by fully stirring the mixture of the base and curing agent in a 10 : 1 mass ratio, followed by the degassing process in a vacuum chamber for 20 min. Subsequently, the PDMS layer of about 400 mm thickness was spincoated onto a polyethylene naphthalate (PEN) substrate. Then, it was put into an oven at 70 C for 15 min to achieve a semicured state and connected with the second PDMS layer of about 100 mm thickness, which helped to improve the atness of the substrate. Similarly, aer 15 min curing at 70 C, the third PDMS layer with a thickness of about 10 mm was spin-coated onto the second layer to work as the transfer medium. Aer that, the prepared CNT array sample with an area of about 1.5 cm Â 3 cm was immediately inverted and gently placed into uncured PDMS. During the baking process in an oven 70 C for 2 h, the neat CNT arrays would partially sink into the third PDMS layer. At the same time, the liquid PDMS would be adsorbed into the dense CNT arrays under the inuence of the capillary wetting effect. Since then, the stable structure of CNT arrays embedded in the PDMS lm was formed.
As the spacer used for adjusting the initial contact state between the upper and bottom CNT layers, the fourth PDMS layer with a designed thickness was spin-coated on the place without CNT arrays. It is especially designed in this process that the silicon wafer, which is the substrate for CNT growth, plays the role of a natural mask to form a patterned spacer. Since the interfacial bonding between PDMS and CNT arrays is much stronger than that between the Si/SiO 2 substrate and CNT arrays, the silicon substrate can be peeled off easily from the exible CNT/PDMS lm aer 1 h of curing. During the fabrication process of pressure sensors, the edges of two prepared CNT/PDMS lms were cut off, and then the prepared CNT/ PDMS lms were interlaced and sealed together, forming a solid structure aer baking for 2 h. The copper wires were connected to the CNT arrays exposed at both sides of the sensor with silver waste. Aer curing, the electrodes were encapsulated by the PDMS liquid to enhance the connection of electrodes and the lms. Finally, the two PEN lms were peeled off and the pressure sensors were completed. The spacer fabrication method is called self-patterned-spacer design.

Characterization and measurement
The morphologies and structures of the vertical CNT arrays and CNT/PDMS lms were characterized using a eld emission scanning electron microscope (FE-SEM, ZEISS SUPRA®55). The Raman spectra were measured using a Raman spectrometer (Horiba Labram HR Evolution). To test the mechanical and electrical performance of the pressure sensors, the static and dynamic pressures were applied using a manually actuated pressure testing machine (ZHIQU Precision Instruments ZQ-21A) and an electrodynamic force tester (ZHIQU Precision Instruments ZQ-990B), respectively. During the testing, the current changes in the pressure sensors were measured using a semiconductor characterization system (Keithley 4200-SCS) and a high-speed digital multimeter (Keithley DMM6500) using instrument-control soware (Keithley KickStart-2.2.1). It should be stated that the maximum measurable resistance of the test equipment is 100 MU.

Microstructure characterization
The characterization of the vertical CNT arrays and CNT/PDMS lms can be seen from SEM images. Fig. 2a shows the top and cross-sectional SEM views of 30 mm CNT arrays grown by the PECVD method. From the clean surface, the grown CNTs show few impurities and good alignment. The CNT diameter is inversely proportional to the catalyst thickness. Since the thickness of the Fe catalyst is only 2 nm, the diameter of CNTs is generally less than 20 nm, resulting in a high density. The CNT arrays are interconnected by a large number of conductive paths, which contribute to a high electroconductivity. The top and cross-sectional SEM views of the transferred CNT/PDMS lms with 30 mm and 40 mm CNT arrays are shown in Fig. 2b and c, respectively. Due to the capillary wetting effect, PDMS are adsorbed into CNT arrays, forming a variety of microstructures. The morphologies of these microstructures are mainly determined by the penetration extent of PDMS into CNT arrays. When longer CNTs are used, inltrated PDMS will not affect the CNT shape on the surface, that is, the surface morphology of the lm will not be much changed. At the same time, the contact joints of CNT arrays are not damaged, so the CNT/PDMS lm can keep the good conductivity of CNTs, which also determines the piezoresistive characteristics of the sensors. As shown in Fig. 2d, the CNT arrays are almost entirely transferred to PDMS, demonstrating a high transfer efficiency. The Raman spectra of the vertical CNT arrays, pure PDMS and the transferred CNT/ PDMS lm are shown in Fig. 2e. The Raman spectra of CNTs include a D band at 1342 cm À1 , a G band at 1580 cm À1 , and a 2D band at 2691 cm À1 , which represent the amorphous carbon, the graphitized carbon and the stacking order of the nanosheets, respectively, 29,30 while pure PDMS has no obvious peak in the Raman shi ranging from 100 to 3000 cm À1 . All the bands of CNTs can be found in the Raman spectra of the transferred CNT/PDMS lm, indicating that there are exposed CNT tips on the surface of the lm. Comparing the Raman spectra of CNTs and the CNT/PDMS lm, the intensity ratio of the G band and D band increases aer transfer. This is because the root of the asgrown CNTs with better graphitization was transferred to the surface of the lm.

Performance of the pressure sensors
For the characterization of the sensor sensitivity, the relative change in resistance is dened as (R 0 À R)/R 0 , where R 0 and R are the resistances of the pressure sensor without and with applied pressures, respectively. Thus, the sensitivity was calculated as (R 0 À R)/R 0 /DP, where DP is the pressure applied to the sensors. As mentioned above, the performance of pressure sensors can be adjusted by changing the thickness of spacers, and their working mechanisms can be simply divided into two categories. When the thickness of the spacer is relatively thin, the upper and bottom CNT layers are in contact with each other at the beginning, resulting in a small initial resistance of the pressure sensors. As shown in Fig. 3a, these sensors have characteristic curves of typical piezoresistive pressure sensors, and the sensitivity decreases as the pressure increases. As comparison, the pressure sensor with a 17 mm spacer shows a sensitivity of 0.86 kPa À1 within the range of 0-0.4 kPa, higher than that of the one with a 15 mm spacer. This is because the thicker the spacer is, the fewer the initial contacts between the upper and bottom CNT layers are, and thus, the larger the initial resistance of the pressure sensor is. This causes a relatively large change in resistance when applying the same pressure, and thus, the sensitivity of the pressure sensor will be higher. When the thickness of the spacer is continuingly increased, the upper and bottom CNT layers are separated from each other at the beginning, and the initial resistance of the pressure sensors is very large ($100 MU). Only when a certain pressure is reached, the upper and bottom CNT layers will touch each other and the sensor will reach a small resistance ($kU). That is, the resistance of the sensor will change greatly within a very narrow pressure range, resulting in a very high sensitivity. Fig. 3b demonstrates that the response thresholds of these pressure sensors can be adjusted from 2.82 to 6.56 kPa by changing the thickness of the spacers with a sensitivity close to 50 kPa À1 , which is the maximum measurable limit of the test equipment we used. The inset shows the resistance change of the pressure sensor as the pressure increases aer the response threshold, which is similar to the pressure sensor with small initial resistance designed. Fig. 3c shows that the minimum response thresholds of the pressure sensors can reach 300 Pa. Similarly, the response thresholds of the pressure sensors can also be controlled by varying the length of the CNT arrays, as shown in Fig. 3d. In theory, as long as the suitable spacers are chosen, we can adjust the response thresholds to any wanted pressure value. As shown in Fig. 3e and f, the proposed pressure sensors also show a fast response/relaxation time, which are 24/32 ms and 16/16 ms for pressure sensors with smaller and larger initial resistance designed, respectively. In general, due to the considerable viscoelasticity of exible materials, the relaxation time of many pressure sensors is relatively long. 25,31 The realized pressure sensors can overcome this disadvantage by designing the spacer and the air layer in a conned space. Obviously, aer the pressure is removed, the arched upper CNT/PDMS lm will be restored quickly by the elastic force from this unique structure. Fig. 3g shows the representative durability test of the sensors under variable pressures. The performance of the pressure sensor shows no signicant regression under loading/ unloading cycles of over 10 000 times, demonstrating its good long-term stability. Compared to pressure sensors, whose active materials and exible materials are mixed together or two different layers fabricated by coating, spin-coating, sputtering and soaking, 32-36 the robust CNTs in this work are embedded in PDMS and the upper and bottom CNT/PDMS lms are separated by a spacer and an air layer in a conned space, leading to an advantageous combination of CNTs and PDMS while avoiding excessive squeezing. This design demonstrates good long-term stability of the CNTs and the pressure sensors simultaneously.

Applications of the pressure sensors
In practical applications, pressure sensors are expected to be sensitive enough and repeatable for both subtle and large pressures, so it is valuable for the pressure sensors to possess a response-range-adjustable design, so that they can t for the various application needs while still keeping the high sensitivity simultaneously. For the pressure sensors with large initial resistance designed, the upper and bottom CNT layers are separated from each other at the beginning, and the sensor is almost in an insulated state. The pressure sensor will not start working until a certain pressure is reached. To further visualize the switching characteristic of this pressure sensor, we built a complete circuit including a constant voltage source (Agilent E3620A), an LED bulb, and a pressure sensor applied with different pressures of 3 kPa (Fig. 4a) and 5 kPa (Fig. 4b), respectively. It can be found that when the applied pressure is 3 kPa, the LED bulb is at off state, since the upper and bottom CNT layers cannot contact each other, and the resistance of the sensor is extremely large. When the applied pressure is 5 kPa, the LED bulb emits bright light, since the upper and bottom CNT layers contact and the circuit is working. The response of the LED bulb under different pressures proves that the sensor will only start working when the applied pressure reaches its critical switching point.  The switching characteristic of the pressure sensors with large initial resistance designed makes them applicable in many advanced applications. We take the exible keyboard as an example for future integration of exible systems. At present, the response thresholds of most exible keyboards based on pressure sensors are very small, which subjects to mis-touch and outputs wrong information when hands move. 37,38 Here, the realized pressure sensors with large initial resistance designed will only be activated when the pressures reach the pressure thresholds of the sensors, thus avoiding the mistouches and noises. At the same time, the pressure sensors have good long-term stability, which are more suitable for strong-durability applications like exible keyboards. Besides, in exible IoT applications, massive data will be generated, transported, and stored. However, a lot of useless data will be collected and a large amount of energy will be consumed under the interference of noise. The realized pressure sensor shows a strong ltering effect due to its pressure threshold. When used together with other exible data acquisition systems, the sensors can lter unnecessary noises, reduce the amount and difficulty of data processing from the hardware level, and avoid unnecessary power consumption. What is more, the pressure sensors can act as alarm devices, especially for the health monitoring of the elderly or children.
The pressure sensors with small initial resistance designed have a different working mechanism from those with large initial resistance designed. As is observed, the former with a typical piezoresistive characteristic are suitable for health monitoring applications. The results in Fig. 5 show that the pressure sensor with small initial resistance designed can detect dynamic pressing (Fig. 5a), bending (Fig. 5b) and torsion (Fig. 5c), demonstrating its ability to detect external forces.
Furthermore, the pressure sensors with small initial resistance designed can be attached to the human body to detect the human motions. Fig. 6a shows the response of the pressure sensor versus st clenching motion, showing high signal-tonoise ratio. Fig. 6b shows the response of the pressure sensor attached to the arm joint to detect arm bending. The result is a differentiable curve with some small peaks derived from arm joints. Fig. 6c shows the response of the pressure sensor when  attached to a volunteer's heel to monitor foot motions. The result shows a trend in resistance changing direction different from other detections in Fig. 6. When the pressure sensor was tightly attached to a person's uneven heel, the sensor was initially in a compressed state. When walking or running, the feet would be lied, the compression state of the sensor would be relieved, and thus the resistance of the sensor would increase. Comparing the curves of walking and running, the resistance change of running is bigger than walking, and the frequency of running is obviously faster. We also stuck the sensor onto the throat to detect the information during drinking, coughing, and speaking. As Fig. 6d demonstrates, each motion of drinking causes a relative resistance change of about 0.1 for the pressure sensor. At the same time, a small peak appears in the middle of each drink. This is because person's throat has a slight adjustment aer drinking. In the coughing detection shown in Fig. 6e, the volunteer coughed at low and high frequencies alternately, with a change in intensity at the same time. When the frequency of coughing is low, the time interval between peak and peak is increased. When the intensity of coughing is increased, not only does the amplitude of the curve become larger, but also the time required for coughing increases. All the detailed information about coughing can be detected by the sensor, demonstrating its potential to distinguish different coughing habits. When the volunteer repeatedly spoke words of "sensor", similar patterns were clearly recorded by the pressure sensor, as shown in Fig. 6f. The pattern contains split peaks matched with the disyllable pronunciation of "sensor". Evidently, the pressure sensor is good at vocal detections for both motions and pronunciations, showing the potential for applications in articial throat, health monitoring and voice recognition.

Conclusions
In conclusion, we have proposed and realized sensing-rangetunable piezoresistive pressure sensors by self-patternedspacer design. The vertical CNT arrays are grown by a PECVD method and the two CNT layers embedded in PDMS are separated by the self-patterned spacer in the novel design. The pressure sensors with large initial resistance designed can change the response thresholds from 300 Pa to 6.56 kPa while maintaining a high sensitivity, by controlling the thickness of the spacer and the length of CNTs. Due to the large specic surface area of the vertical CNT arrays, the resistance of the pressure sensors changes dramatically within a very narrow pressure range, achieving a high sensitivity of about 50 kPa À1 . Beneting from the designs of the self-patterned spacer and the advantageous combination of CNTs and PDMS, the pressure sensors also exhibit a rapid response/relaxation time of 24/32 ms and good long-term stability with durability test over 10 000 loading/unloading cycles. While the realized pressure sensors with small initial resistance designed show a typical piezoresistive characteristic. Moreover, the self-patterned design and fabrication method of the spacers are simple and controllable, showing potentials to be applied in the existing works to further enhance or adjust the performance of those pressure sensors. With this design, many piezoresistive pressure sensors with high sensitivity in low pressures can completely shi their performance to certain higher-pressure ranges to meet various applications, showing great exibility. For some existing microstructured pressure sensors, this method can be applied to fabricate spacers before peeling off the mold to adjust the device performance without affecting the original microstructures using the mold, which plays the role of a natural mask like the silicon wafer in this work. In practical applications, the pressure sensors with large initial resistance designed are suitable for the anti-noise applications that require pressure thresholds such as exible keyboards to avoid mis-touch and exible data acquisition systems to lter unnecessary noise and save power consumption, respectively. However, the pressure sensors with small initial resistance designed not only show the capability of detecting mechanical forces, but also can be attached to the skin to monitor human physiological signals. As a prototype, the sensing-range-tunable pressure sensors show great potentials in various exible wearable electronics such as health monitoring, humanmachine interaction, the IoTs, and beyond.

Conflicts of interest
There are no conicts of interest to declare.