Development of piezoresistive poly(ɛ-caprolactone) sensor with unique flexibility and sensing performance for rehabilitative application

ABSTRACT

Piezoresistive sensor is applied extensively in rehabilitative monitoring, to help with patients’ physical and mental health recovery. However, the manufacturing complexity of hierarchical structure in sensing element remains a major limitation to piezoresistive sensor. Here, we report a scalable and simple method for the preparation of piezoresistive sensing element. Based on biocompatible poly(ɛ-caprolactone) (PCL), uniaxial cold stretching resulted in a flexible sheet-like element with unique hierarchical geometries comprising highly aligned multiscale ridges on sheet’s double sides. This leads to higher mechanical elasticity of the hierarchitectured PCL element and enhanced resistance against plastic deformation due to the improved Young’s modulus. Further assembly of the piezoresistive sensor demonstrates a wide detection range of 25–4.7 × 10⁴ Pa with rapid response and the high sensitivity of 1.37 kPa⁻¹ at small pressure loadings. The sensor shows successful application in the monitoring physiological signals including pulse, speaking, face expression and limb motion.

GRAPHICAL ABSTRACT

KEYWORDS:

• Flexible electronics
• piezoresistive sensor
• PCL

Introduction

Rehabilitation can help patients suffering from physical impairments or disabilities, to relearn the lost skills for independent and improved life quality [1]. In this paradigm, rehabilitative monitoring is required for the patients in exercising their limbs and restoring self-confidence to get rid of disease influence [2]. Flexible electronics that are widely used in the detection of motion, physiological indices and diseases (e.g. Parkinson’s syndrome, stroke and multiple sclerosis [3]) can have applications in providing real-time whole-day monitoring with accurate signals and early alarming. Compared with the traditional electronic devices, such flexible ones are advanced in the portability, flexibility, scalability, and unique deformability [4]. These flexible electronics are commonly involved sensors for the specific detection of pressure, temperature, humidity and mechanical strain. Among these requirements, mechanical sensing is a key function of flexible electronics to be applied in health monitoring, including the measurement of various mechanical strains such as body movement and physiological actions (e.g. pulse, breathing and heartbeat) [5]. For example, wearable sensors can be used to detect human pulse waveforms including the characteristic systolic blood pressure pea, diastolic blood pressure peak, and diastolic blood pressure, in order to monitor the cardiovascular health status in real time [6]. These functions are achieved by the electronic devices to convert mechanical strain into electrical signals through piezoresistive, capacitive and piezoelectric sensing principles [7].

Piezoresistive sensors that are widely investigated have advantages in easy fabrication, low energy consumption, straightforward read-out connections, and broad detection ranges of loadings [8,9]. By definition, piezoresistive effect refers to change of the resistivity caused by external force applied to material. Traditional piezoresistive materials, for example the metal and semiconductor, are usually rigid, and when applied to flexible electronics, it is necessary to incorporate additional structure design for stretchable capability [10]. To reduce the complexity of manufacturing process, an alternative dopes flexible polymeric material with conductive particle. This results in the composite possessing piezoresistive properties, because of the formation of percolation networks and tunnelling effect [11]. In order to achieve the high sensitivity, wide range of detection and rapid response for healthcare application, different surface structures including dome, pillar, pyramid and hierarchical microstructure are designed and engineered for highly piezoresistive performance [12]. The deformation of surface structure, rather than the substrate bulk, results in significant changes in contact resistance, leading to enhanced sensing properties than the change of bulk resistance does [13].

For example, it is found that surface construction of microstructures can improve the detection range of a piezoresistive sensor. This is evidenced by the bioinspired design of hierarchical structure, which results in a piezoresistive sensor obtaining the wide pressure regime up to ~12 kPa [14]. Through hierarchical interlocking of a hemispherical array with thin column, an ion-electronic flexible piezoresistive sensor shows a upper detection limit of 485 kPa [15]. Meanwhile, by the control of controllable graphene-nanowalls wrinkle, the lower detection limit of a piezoresistive sensor can be reduced to a very low pressure of 0.2 Pa [16]. On the other hand, a pillar array prepared from polydimethylsiloxane (PDMS) and coated using silver nanowires results in a sensitivity of 20.08 kPa⁻¹ for the piezoresistive sensor [17]. The further research by employing a design of pleated structure with 2D MXene, generates a flexible and even highly sensitive (up to 148.26 kPa⁻¹) piezoresistive sensor [18]. In other studies, design of surface structure is found to make faster response of the piezoresistive sensing performance. Taken a conductive micro-gold word sensor for example, it shows a fast response time of 20 ms based on PDMS/carbon nanotube composite [19]. Similarly, the design of a piezoresistive sensor based on an interlocked graphene microarray obtains a comparative response time of 19 ms [20]. In another study, using a layered pyramid microarray structure, an ultra-low response time of 10 ms is reported [21].

Great efforts are made in order to reduce manufacturing complexity and cost, as the most design and engineering of advanced surface structure associates often with complicated, expensive, and time-consuming procedures [22]. Piezoresistive surface structure are mainly from lithography, natural material and sacrificial template [12], of which lithography technology obtains more and more attentions in the application in flexible electronics. For example, femtosecond laser surface micropatterning [23] and direct laser writing are applied to engineer single-stage microstructure, followed by laser grid marking creating further layered microstructure, which makes significant improvements to the sensitivity of flexible piezoresistive sensors [24]. Such lithographic method is advanced in accurately controlling the structure size, spacing and geometry at micro and nano scales. However, it is time-consuming, requires high cost and has inherent difficulty in realizing complex and large-area fabrication [25]. Recently, natural materials from animals and plants have been reported as templates for the development of piezoresistive surface structure, without needs of expensive and complex instruments. For instance, by replicating the inverted microstructure on Ginkgo biloba surface followed by MXene(Ti₃C₂T_x) spray coating and selectively etching, an ultra-sensitive pressure sensor is developed (Sensitivity: 403.46 kPa⁻¹) [26]. However, templates from the natural materials are significantly limited to the available sources, where the structure geometric shape, dimension and morphology cannot be controlled or changed, making it difficult to be standardized for reproducible and massive preparation [27].

Current solutions have not fully addressed the above problems related to piezoresistive structures for flexible electronic application. In addition, towards healthcare applications, the piezoresistive sensor is also expected to be biocompatible and environmental friendly. In this study, we developed a scalable and simple method for engineering double-sided hierarchical sheet. This sheet is based on biocompatible and degradable composite, showing unique piezoresistive properties and healthcare applications.

Materials and methods

Materials

PCL (MW = 80,000), conducting state carbon black (CB) and PDMS (Sylgard 184) were purchased from Sigma-Aldrich (U.S.A.), TIMCAL Ltd (Switzerland) and Dow Corning Co. Ltd (U.S.A.), respectively. Dichloromethane (DCM), methanol (MeOH), ethanol (EtOH) and sodium hydroxide (NaOH) were from Sinopharm Chemical Reagent Co. Ltd (China).

Fabrication of piezoresistive sensor

PCL solution was obtained using a mixed DCM/MeOH (9:1, v/v) solvent and casted into film (Named as cPCL) using glass petri dishes (40 mL per dish, 150 mm in diameter, AN-5058558, Normax Biomed Ltd, Portugal). The dish was covered with aluminium foil (Pore density: 1.1 mm cm⁻²) and put in a fume hood (21°C) for three days. Hierarchical surface structure was induced by uniaxial stretching of the cPCL using a tensile machine (HYC‐2011, Hongjin, China). The stretched film was named as µPCL.

Surface modification of µPCL was by alkaline treatment (1 M NaOH, 24 h), followed by CB coating (0.1 wt % dispersed in EtOH). The hydrolysed and CB coated PCL were named as mPCL and µPCL/CB, respectively.

Flexible electrodes were prepared from PDMS sheets (thickness: ~750 µm) and coated with gold (Au) using an ion sputter apparatus (SBC-12, KYKY Technology CO., LTD., China). The piezoresistive sensor was assembled by sandwiching µPCL/CB with a pair of coplanar PDMS/Au electrodes (Figure 1a).

Figure 1. Fabrication of piezoresistive µPCL/CB sensing sheet. (a) Scheme illustrating design and engineering of piezoresistive PCL sensing sheet and the sensor for rehabilitative application. PCL, poly(µ-caprolactone); CB, carbon black; cPCL, cast PCL film; µPCL, uniaxially stretched PCL film; and µPCL/CB, CB-coated µPCL. Scale bar, 1 cm. (b) Representative SEM images of cPCL. Dashed red line, PCL crystal zone; dashed blue line, PCL amorphous zone; dashed green box, open pore. Scale bar, 100 µm. (c) Representative SEM images of µPCL. Dashed green box, open pore; yellow arrow, macroridge; purple arrow, macrogroove; and double-headed white arrow, uniaxial stretching direction. Scale bar, 100 µm. (d) Surface wettability. CA, contact angle of water. n = 4. (e) XRD spectra. mPCL, modified µPCL by alkaline hydrolysis. (f) FTIR spectra.

Material characterisations

Sample after gold coating was examined for surface topography using Scanning Electron Microscope (SEM, S-4300, Hitachi, Japan).

Chemical groups of the sample were examined using Fourier Transform Infrared Spectroscopy (FTIR, Nicolet iS50, Thermo Scientific, U.S.A.). Wavenumbers ranged from 400 to 4000 cm⁻¹ were carried out. Functional groups were identified by comparing the sample spectra with those reported in published literature [28].

Phase structure of the sample was examined using Diffraction of X-rays (XRD, Miniflex600, Rigaku, Japan). Cu-Ka radiation was employed in atmospheric conditions at wavelength of 1.5418 Å and 2θ values of 10–80°. Data were collected at a step width of 0.2° with duration of 1.2 s. Phase structure was identified by comparing the testing diffraction patterns to the ICDD standards (JCPDS).

Surface wettability of the sample was examined using the build-in function of Contact Angle Analyzer (SDC-200S, ShengDing, China). A droplet of 2 µL was used, and the image of droplet was taken 5 s after contacting the sample surface. On each sample, the droplet was imaged along two perpendicular directions at more than three random sites. Five samples were used for each group.

Mechanical characterisations

Mechanical properties were examined using the Tensile Testing Machine. Sample thickness was measured using Digital Micrometre (0–25 mm, Meinaite, Germany). A tensile speed of 10 mm min⁻¹ and loading force of 100 N were used. Data were analysed using the Origin software (Origin9.1, OriginLab, U.S.A.). For the sample with yield phenomenon, the low yield point was used to determine the yield stress and strain, and for the left an offset strain of 0.5% was used to determine the yield stress and strain. Five samples of each group were measured.

Electrical characterisations

Current change to applied pressure was examined as described previously [29]. Briefly, a mechanical testing machine with controlled force-time characteristics (ZQ-990LA, Zhiqu Precision Instrument Co., Ltd, China) was to provide pressure loading. The pressure (P) is determined as p = F/S (S = 1.5 × 1.5 cm²). Meanwhile, Semiconductor Parameter Instrument (Keithley Sourcemeter 2634B, U.S.A.) was used to record the current-time characteristics every 50 ms at an applied voltage of 1 V. Sensitivity of the assembled sensor was determined by S = (ΔI/I₀)/ΔP.

To evaluate the concurrency of current and pressure changes, the samples were subjected to slow (loading speed: 10 mm min⁻¹) and fast loading (loading speed: 80 mm min⁻¹) at different pressures of 1 and 10 kPa. Finally, to test the responsive stability of assembled PCL sensor, a cycle loading of 1.3 kPa was applied at the frequency of 1.5 Hz, with a constant voltage of 1 V. In other testing, the voltage was kept at 1 V unless specific statement.

Healthcare applications of the assembled µPCL/CB sensor

To evaluate the detection limit, a small weight (500 mg, ∼25 Pa) was applied to generate pressure loading. In another testing to evaluate the capability of non-contact detection, breeze was generated towards the sensor by an ear sucking ball. The electric signal collection of on-body measurement was in consent with the tester. All tests were in accordance with the laws and with the approval of the Medical Ethics Committee of University. In the testing of human pulse, smiling, voice and limb movement, the sensor was attached to face, cords, and limbs including finger, wrist and elbow. In all testing, the applied voltage was 1 V.

Statistical analysis

Data was expressed as mean ± S.D. unless otherwise specified. Statistical comparisons between two groups were performed based on the student’s paired test with two-tailed distribution using GraphPad Prism 8.0.2 (GraphPad Software Inc., U.S.A.). A value of p > 0.05 is considered as no significance (NS), and p < 0.05 represents significance, where *, ** and *** corresponding to p < 0.05, p < 0.01 and p < 0.001, respectively.

Results and discussion

Piezoresistive thin-film design and fabrication

Keeping the rapidly increased ageing population of society in mind, body signals generated from mechanical force can be recorded for accurate diagnosis, early warning, and real-time all-day monitoring are critical. Figure 1a illustrates the design and fabrication of PCL piezoresistive sensor for healthcare and rehabilitation application. By sandwiching µPCL/CB with a pair of coplanar electrodes, the sensor is assembled. Here, PCL is selected as the sensing element because of its unique biocompatibility and environment degradability. It can be observed that µPCL/CB not only shows the flexibility and self-standing properties but also is designed to possess piezoresistive response to pressure loading. To the best of our knowledge, this PCL-based sensing element has not been reported previously.

As shown in Figures 1b,c, µPCL is obtained from cPCL which shows unique crystal morphologies on its front and back surfaces. After uniaxial stretching, µPCL shows distinct surface topographies. On its front surface, prominent ridges and grooves are observed which orientate preferentially along the stretching direction (Figure 1c) and is consistent with our previous reports [30]. During uniaxial stretching, formation of the ridges is because of the PCL crystals experienced recrystallisation and alignment whilst the grooves are formed from the PCL amorphous regions which experienced more deformation [31]. On the back surface of µPCL, similar findings are observed. Uniaxial stretching enlarges the pores on film back surface. Consistently, the small crystal structures on film back surface correspond to the resulted prominent ridges on front surface.

Figure 1d demonstrates anisotropic surface wettability due to the presence of surface ridges/grooves, which disappears after the CB coating. It is found that the alkaline hydrolysis increases surface hydrophilicity, and the CB coating contributes to water repellency leading to finally significant hydrophobicity. Figure 1e shows the XRD patterns of µPCL/CB reserved the PCL characterised crystal peaks (2θ of 21.4° and 23.7°) [32]. However, due to the minimum amount of CB coating, the carbon peak cannot be observed. Meanwhile, FTIR examination confirms no chimerical property changes to either PCL or CB (Figure 1f).

Mechanical properties of the µPCL/CB as sensing element were further investigated (Figure 2 and Table 1). By uniaxial stretching, the mechanical performance of µPCL is improved compared with cPCL. The yield stress and strain of µPCL are 553.8% and 343.5% higher than those of cPCL, respectively. During the stretching process, this is obtained due to the strain-induced strengthening effect leading to simultaneous enhancements in yield stress and strain. In contrast, the yield stress (less than 4 MPa) and elastic strain (less than ~6.2%) of cPCL are small and not suitable for piezoresistive investigation. Meanwhile, the E of µPCL is about 1.9 times of cPCL, which suggests the potential of better structure response and deformation recovery upon pressure loading. Such enhanced mechanical properties are necessary for µPCL/CB as the sensing element towards long-term and stable use in electronic devices. Because healthcare sensor meets various external loadings easily and tends to have permanent structure failure. Here, after uniaxial stretching the ultimate stress of µPCL is 254.3% increased, with an indication of better capability to withstand large force loading before failure. Compared with µPCL there are no significant differences in both stress-strain behaviours and mechanical indexes of mPCL and µPCL/CB. This indicates surface modifications of alkaline hydrolysis and CB coating have not caused obvious changes to mechanical properties.

Figure 2. Effects of manufacturing factors on material stress-strain curves. (a) cPCL. (b) µPCL. (c) mPCL. (d) µPCL/CB. Gray zones, enlarged parts of the elastic deformation. n = 3.

Table 1. Effects of manufacturing process on mechanical properties. (n = 3).

Samples	E* (MPa)	σy* (MPa)	ɛy* (%)	UTS* (MPa)	Break strain (%)
cPCL	60.0 ± 7.1	3.9 ± 0.4	6.2 ± 0.8	8.1 ± 0.3	946.2 ± 90.2
μPCL	112.4 ± 15	25.5 ± 5.6	27.5 ± 4.5	28.7 ± 6.9	136.8 ± 8.8
mPCL	116.5 ± 3.6	27.6 ± 2.8	30.2 ± 6.9	28.4 ± 3.0	130.8 ± 17.0
μPCL/CB	107.2 ± 8.4	28.1 ± 1.9	34.1 ± 2.1	30.9 ± 1.8	144.3 ± 4.7

*E, Young’s modulus; σy, Yield stress; ɛy, Yield strain; and UTS, Ultra tensile stress.

Sensing performance and mechanism of the µPCL/CB sensor

To explore good response performance, sensor assembled with the front, back and double-surface hierarchitecture of µPCL/CB are investigated, and the sensitivity is defined as S = ((I−I₀)/I₀)/P. It is found that the sensor based on front single-layered sensing surface shows a maximum sensitivity of 1.67 kPa⁻¹ at low loadings less than 1 kPa (Figures 3a,b). This is much higher than that based on back single-layered sensing surface (0.24 kPa⁻¹). The sensor assembled with double-face sensing hierarchitecture makes a compromised maximum sensitivity of 1.37 kPa⁻¹. This suggests that the sensitivity of a piezoresistive sensor from double-layered electrodes may not be the simple superposition of single-layered performances.

Figure 3. Electrical response and mechanism of the μPCL/CB sensing sheet. Different sensors are assembled by attaching μPCL/CB front (front), back (back) and double (double) faces to electrodes, respectively. (a) Relative current curves. (b) Sensitivity. (c) Scheme illustrating changes of the surface hierarchitecture of μPCL/CB in response to different applied loadings. (d) Relative current curves of the sensors based on μPCL/CB and cPCL/CB in response to different pressures. (e) Responsive repeatability of the μPCL/CB sensor at different pressure loadings.

The pressure detection range of µPCL/CB double-layered sensor can be up to 47 kPa, probably because of the double-layered hierarchitectures deforming at the same time upon loadings. This results in the larger resistance to structural deformation and thus greater changes of relative conductivity. In contrast, those of the front and back single-layered sensors are only 1.5 kPa and 3 kPa, respectively. This indicates that they are not an ideal choice for piezoresistive application. Compared with the published PDMS-based hierarchitecture sensor, the µPCL/CB double-layered-sensor is advanced in the higher maximum sensitivity (vs. S of 1.2 kPa⁻¹) and wider sensing range (vs. 0–25 kPa) [33]. With increased pressure loadings, sensitivity of the µPCL/CB sensor decreases, with an S greater than 1.37 kPa⁻¹ for the loadings within 0–2.5 kPa, around 1 kPa⁻¹ within the loadings of 2.5–10 kPa, higher than 0.5 kPa⁻¹ in the loadings of 10–20 kPa, and above 0.1 in the loadings of 20–47 kPa. This is probably attributed to the micro ridges as major responsive structure at small loadings and the macro ridges at large loadings (Figure 3c). In agreement with the above observations, Figure 3d shows the sensor based on cPCL/CB which lacks the stretching-induced surface hierarchitecture and is failure in response to pressure loadings.

When the µPCL/CB sensor receives dynamic loadings, Figure 3e shows the responsive current changes among 0.48–19.52 kPa. The pressure increases results in enhanced current response, with an indication of different deformation degrees of the hierarchitecture under gradually increased loadings. At both high and low pressures, the sensor demonstrates good responsive stability to cycle loadings and unloadings. This is attributed to the abundant ridges which upon loading share the pressure at micro and macro scales and thus obtain good recoverable capability with repeatable response (Figure 3c).

To evaluate the durability of as-fabricated PCL sensing hierarchitecture, we tested the pressure loading at different frequencies (Figure 4a), and the results of relative current changes show a stable response. Further relative current changes recorded when the repeated loading and unloading are applied for 12,000 cycles demonstrate a long-term stability of responsive capability (Figure 4b). The µPCL/CB sensor displays two distinct periods of response stability, with the large current changes within 7,000 cycles and lower ones over 7,000 cycles. This is probably because of the PCL hierarchitecture, where after 7000 cycles the microscale ridges may become deformed permanently and decrease the relative current change. However, there are still a large number of macroscale ridges that can make elastic deformation continuously and contribute to the stable response after 7000 cycles, suggesting the good dynamic durability and stability of µPCL/CB sensor for long-term application.

Figure 4. Electromechanical performances of the μPCL/CB sensing sheet. (a) Relative current change of the μPCL/CB sensor in response to at different frequencies at a pressure loading of 3 kPa. (b) Durability of the μPCL/CB sensor in response to cycle loadings (1.3 kPa, 0.8 hz). (c) I-t and P-t curves at different pressures and loading speeds.

We further apply pressure at different speeds to evaluate the coincidence between loading and piezoresistive response (Figure 4c). Both the slow of 10 mm min⁻¹ and high of 80 mm min⁻¹ rates of pressure loadings can result in the µPCL/CB sensor with immediate electrical response. It is found consistently that the higher-pressure loading generates the larger relative current response. At a low pressure of 1 kPa, the higher speed of loading results in a larger relative current change than the lower one. This is similarly observed at the higher pressure loading of 10 kPa, and can be attributed to the induced overloading of pressure at the high speed. These results suggest that the µPCL/CB sensor is sensitive to pressure overloading. During the initial period of loading at 1 kPa, the responsive curves of relative current are above those of the pressure loading curves at the small loading speed of 10 mm min⁻¹. When the loading speed increases to 80 mm min⁻¹, the two kinds of curves become overlapped ideally. Such phenomenon indicates the capability of µPCL/CB sensor for the rapid loading response, and has also been observed for loadings at the high pressure of 10 kPa.

Healthcare applications of µPCL/CB sensor

Further exploration of the µPCL/CB sensor is for healthcare applications based on its unique sensitivity, fast response and good stability. By using a weight of 500 mg, it demonstrates the response to minimum loading and unloading, where the lower detection limit can be small to ~24 Pa (Figure 5a). In combination with Figures 3a,b, the detection range of µPCL/CB sensor can be ~24 Pa to 47 kPa. This performance is comparable to that obtained based on graphene microarrays (1 Pa to 32kPa) [20]. Non-contact small loading, for example breeze from the ear ball, can also be detected by the µPCL/CB sensor as shown in Figure 5b. The contactless pressure loading results in sharp current change, indicating a rapid response by the sensor while the quick current disappearance indicates the capability of fast returning to its initial state. Probably because of the hierarchitectural design, micro ridges on the sensing element of µPCL/CB are easily deformed and recovered when a pressure is applied in the absence of physical contact.

Figure 5. Healthcare monitoring applications of the μPCL/CB sensor. (a) Detection of tiny loadings. Inset: optical image of the weight loading at 25 Pa. (b) Non-contact pressure detection. Inset: optical image of airflow loading by an ear ball. (c) Physiological pulse detection. Inset (left): featured three peaks of human pulse; and inset (right): optical image of curved surface fitting to human wrist. (d-f) Detection of limb motion. Insets: optical images of figure bending (d), Wrist motion (e) and elbow motion (f) at different degrees. (g) Detection of facial expression. Insets: optical images of curved surface fitting to human facial configuration. (h, i) detection of throat motion and sound recognition when spoken ‘thank you’ (h) and ‘xiexie ni’ (i). Insets: optical images of curved surface fitting to human throat.

The pressure related to daily activities of human beings can be categorised into three regimes, including a subtle-pressure of 1–10³ Pa (e.g. human skin sensing), a low-pressure of 10³–10⁴ Pa (e.g. gentle manipulation of items) and a medium-pressure of 10–100 kPa (e.g. blood pulse, and joint movements) [34]. In line with the response to tiny pressure, the µPCL/CB sensor can be applied to detect the pulses of human body (medium loadings, Figure 5c). By attaching the sensor closely on the curved surface of tester’s wrist, it can real-time detect the pulses, with the characteristic pulse peaks of ‘P’ (percussion), ‘T’ (tidal) and ‘D’ (diastolic). Since the pulse provides important vital signs of human body and health condition [35], these results thus indicate the potential of µPCL/CB sensor for physiological health detection. Figures 5d–f show the application of µPCL/CB sensor in detecting limb motion. When the finger bends, it can detect the motion and distinguish the amplitudes of blending. Its responses with the increased relative current change when the bending amplitude is increased. The complete blending of finger generates a maximum loading to the sensor and subsequently slack contact with the sensor. This results in an instantaneous strong response and then reduced responsive strength. The µPCL/CB sensor similarly demonstrates the capability of detecting blending of the wrist and elbow. These applications can be used for rehabilitative monitoring.

Finally, the µPCL/CB sensor shows its capability in the detection of facial expression and speech recognition. As shown in Figure 5g, by attaching to the tester’s face, the sensor exhibits consistent output of the relative current change when smiling expression is made during the test. Different durations of the smiling correlate to the relative peak width of output signals. Such application is useful for the patients who have problems with muscle control after a stroke by allowing the real-time detection and signal feedback. In another case, speech recognition, the basis of human-computer interaction, is demonstrated using the µPCL/CB sensor by attaching to the curved surface of the larynx (Figures 5h,i). The sensor can sense vibration of the vocal cords when people speak and distinct the different characteristic peaks of ‘Thank you’ and ‘Xie xie ni’. The term of ‘Thank you’ exhibits only two stress nodes, while Chinese term shows three stress nodes. Such difference indicates good performance of the sensor for speech recognition which may be applied for human–computer interaction.

Conclusion

Here, a scalable and simple method is developed for engineering double-sided piezoresistive surface. By uniaxial cold stretching, biocompatible PCL is fabricated into the sheet-like element with unique hiearchitecture, obtaining highly aligned multiscale ridges and grooves. The hiearchitectured element shows enhancement in elasticity and Young’s modulus, leading to higher mechanical resistance against plastic deformation and improved sensing range. This was demonstrated by the assembly of piezoresistive sensor, with a wide detection range from 25 Pa to 4.7 × 10⁴ Pa. The sensor can monitor different physiological signals of pulse, speech, facial expression and limb motion. This manufacturing method is applicable for polymeric materials. The as-fabricated PCL sensor has potential in future medicine for rehabilitation monitoring.

Author contributions

Jiabin Liu: Methodology, Visualization, Software, Validation, Investigation, Formal analysis, Writing – original draft, review and editing. Fang Zhong: Methodology, Visualization, Investigation, Formal analysis, Writing – original draft. Peng Liu: Methodology, Discussion. Pan Xu: Methodology, Investigation. Xiaoxi Long: Methodology, Visualization, Discussion. Chao Ma: Resources, Supervision, Funding acquisition, Discussion. Song Liu: Conceptualization, Methodology, Discussion. Nan Lin: Methodology, Discussion. Zuyong Wang: Conceptualization, Resources, Supervision, Project administration, Funding acquisition, Discussion, Writing – review and editing.

Disclosure statement

A patent for the preparation technique described in this article has been granted (CN202111248089.X).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/10667857.2024.2332040.

Additional information

Funding

This research was supported by the Outstanding Youth Scientist Foundation of Hunan Province [2020JJ2001], Hunan Provincial Technology Innovation Platform and Talent Program [2017XK2047], and Fundamental Research Funds for the Central Universities of P R China [531107050927]. Dr Z Wang received financial support from Hunan University for the Yuelu Young Scholars [JY-Q/008/2016].