GaN/PDMS-based opto-electro-mechanical tactile sensors

Abstract

Tactile sensors are crucial in robotics and medical diagnostics, requiring precise real-time detection. However, the development of a compact sensor that can measure force across a wide range, with high resolution and rapid response along three axes, remains extremely limited. Herein, an opto-electro-mechanical tactile sensor is reported, utilizing a monolithically integrated GaN-based optochip with a fingerprint-patterned polydimethylsiloxane (PDMS) film. The sensor exhibits a linear response over a broad measurement range of ±100 mN for shear force and 0–200 mN for normal force, with a detection resolution of 0.07 mN. It also demonstrates fast response and recovery times of 0.85 ms and 0.82 ms, respectively. Experimental verification of its application in surface topography scanning and organ lesion assessment highlights its potential for enhancing robotic perception and medical diagnosis.

Introduction

With the rapid advancement of sensing technology, research efforts have increasingly focused on endowing robots with anthropomorphic sensing capabilities to tackle complex and delicate tasks, such as handling fragile objects^1,2, perceiving object shape and texture^3–5, and executing precise medical procedures^6–9. In particular, tactile sensing units, developed based on various sensing principles, such as resistive^10,11, piezoelectric^12,13, capacitive^14,15 and optical means^16–20, serve as a pivotal bridge enabling these tasks through object recognition, localization, and manipulation.

Recently, numerous advanced designs incorporating microstructures, such as pyramids^21,22, interlocking^23,24, and bionic fingerprint structures^25–27, have been introduced to enhance the accuracy, sensitivity, and versatility of tactile sensors. However, as robotics progresses towards higher levels of intelligence, the capability to precisely measure forces along multiple axes becomes essential for implementing refined control strategies and ensuring safety in robotic operations²⁸. Single-dimensional force sensing is inadequate in addressing the complexities associated with diverse application scenarios. In the pursuit of further developing multi-dimensional force-sensing units, there is a growing concern regarding signal crosstalk across multi-axis detection²⁹. Additionally, the analysis of shear and normal forces from the crosstalk signal necessitates complex signal processing algorithms, which increases system complexity and expenses^30,31. This challenge is further compounded in medical diagnostics and robotic operations, where maintaining signal integrity in complex electromagnetic environments is paramount to preventing diagnostic errors and operational failures^32,33. Optical sensing modalities inherently offer distinct advantages, including immunity to electromagnetic interference, rapid response characteristics, and high accuracy^34,35. However, conventional optical tactile sensing systems necessitate bulky external illumination sources and complex optical demodulation components, significantly hindering miniaturization efforts and system integration. These spatial and architectural limitations present substantial challenges to the compact form-factor requirements of next-generation robotic and biomedical platforms.

On-chip photonic integration emerges as an effective approach for enhancing system compactness and robustness by assembling functional components on a single platform^36,37. Due to its high efficiency, long lifespan, and high stability, GaN semiconductors and their alloys are promising candidates for developing optical devices based on the monolithic integration approach³⁸. Although tactile sensing based on GaN optoelectronics has been demonstrated³⁹, considerable challenges remain in achieving a linear sensing response across a wide measurement range along three axes, while maintaining resolution and minimizing crosstalk.

To overcome these limitations, this work introduces a miniaturized opto-electro-mechanical tactile sensor capable of detecting shear and normal forces. The sensor incorporates a GaN-based optochip with a double-side patterned polydimethylsiloxane (PDMS) film. The cross-shaped optochip monolithically integrates a light emitter and four photodetectors, laying the foundation for the detection of shear and normal forces. The incorporation of a fingerprint structure on the PDMS film efficiently converts force stimuli into optical signals sensed by the photodetectors. The optoelectronic properties of the optochip, as well as the responses of the sensor to shear and normal forces, are characterized to verify its effectiveness. Furthermore, the sensor is applied in surface topography scanning to demonstrate its potential use in the fields of robot perception and medical diagnosis.

Results and Discussion

Inspired by the structural characteristics of human finger skin, a biomimetic tactile sensor is designed. The tactile sensor primarily consists of a GaN-based optochip, which is responsive to external contact media, and a PDMS film layer with a top surface patterned with a fingerprint-like structure, as shown in Fig. 1a. The PDMS film serves to emulate the epidermis of human finger skin, facilitating the transmission of external stimuli. Meanwhile, the optochip mimics the functionality of Meissner corpuscles within the skin, sensing stimuli and generating corresponding signals. The photographs of the optochip after laser cutting is presented in Fig. 1b. Figure 1c shows the schematic layout of the optochip, which comprises a centrally located cross-shaped LED, surrounded by four symmetrically arranged independent PDs. The resultant sensor based on the integration of the optochip with the patterned PDMS is shown in Fig. 1d.

Fig. 1. Structure of the GaN/PDMS-based opto-electro-mechanical tactile sensor.

a Schematic diagram of the skin-inspired tactile sensor and the layout of the optochip. b Optical image of the optochip after laser shaping. c Optical image showing the distribution of LED and PDs on the optochip. d Optical image of the resultant sensor

The configuration of the optochip, regarding the positioning of the PDs and the definition of the measurement axes, are depicted in Fig. 2a. The sensing principle of the tactile sensor is illustrated in Fig. 2b. Initially, without force being applied to the sensor, the sapphire of the optochip, with a refractive index of 1.78, is mostly exposed to air. Light incident at angles exceeding the critical angle of 34.2° undergoes total internal reflection and is coupled to the photodiodes (PDs). When the shear or normal force is applied, the PDMS film deforms, thereby increasing its contact area with the sapphire. Given that PDMS possesses a refractive index of 1.44, the critical angle at the sapphire/PDMS interface rises to 53.5°, resulting in a reduction in the reflected light intensity.

Fig. 2. Sensing principle and simulation.

a Schematic diagram showing the positions of PD_1-4 on the optochip with respective to three axes. b Schematic diagrams of the sensing principle of the sensor. c Simulation results of the for varying force magnitudes. The color scale bar represents the magnitude of displacement of the PDMS film. d Simulation results showing the contact area with varying fingerprint sizes and shear force

As the patterned PDMS serves as a key component in converting applied forces into contact changes at the sapphire boundary, its structural parameters are investigated. The bottom pyramid pattern of the PDMS is designed to align with the dimensions of the underlying cross-shaped optochip, and the size of the fingerprint structure on the top of the PDMS film could be an important factor in affecting the sensing performance. The effect of varying the size of the fingerprint structure on the contact area under applied shear force is investigated through 3D simulation using finite element method. The simulation model is constructed identical to experimental device with a double-side patterned PDMS film and an optochip. The integrated displacement (URES) is calculated as the vector sum of the displacement components of the PDMS film in each direction, which represents the total displacement of a point on an object subjected to multiple forces or stresses. In other words, the displacement profile can be used to represent the contact area between the PDMS film and the sapphire interface. For instance, as the tangential force increases, the displacement of the PDMS film increases, leading to a larger contact area with the sapphire interface, as shown in Fig. 2c.

To provide a comprehensive understanding, the contact areas between PDMS and sapphire are computed for different fingerprint sizes (K) across a force range of 10–150 mN, with increments of 20 mN. For quantitative illustration, the contact areas are derived from the simulation maps through normalization, defined as the ratio of the PDMS-sapphire contact area to the area of the PD (See Supporting Information S2) and plotted in Fig. 2d. For the PDMS films with larger K values of 400–600 μm, the contact area remains smaller throughout the applied force. It is observed that reducing K can effectively increase the ability of structural deformation of the PDMS film to force, resulting in an increased contact area. However, as K decreases below 200 μm, the contact area between the PDMS and the sapphire becomes more easily saturated under minimal force application. In other words, smaller K values yield higher sensitivity but a narrower force response range. Therefore, the PDMS film with K = 300 μm is selected to strike a balance between sensitivity and response range. Furthermore, with the fingerprint size K fixed at 300 μm, different fingerprint spacings are evaluated. It is found that sensor responsiveness improves with a densely packed arrangement of the fingerprint pattern (see Supporting Information S3).

Properties of GaN optochip

The properties of the on-chip LED and PD are investigated, focusing on both electrical and optical characteristics. Figure 3a presents the current-voltage (I–V) curve of the LED, revealing a forward-biased voltage of 2.47 V at a current of 10 mA. By analyzing the reciprocal slope of the linear region, the resistance is determined to be approximately 9.2 Ω. The inset in Fig. 3a illustrates that the light output power of the LED is directly proportional to the bias current. Figure 3b plots the emission spectrum of the LED at a current of 10 mA, which exhibits a peak at around 447 nm, alongside the absorption spectrum of the PD. The spectral overlap between 420 nm and 445 nm indicates that the PD, utilizing the same InGaN/GaN MQWs, can effectively respond to the LED emission. Figure 3c shows the photocurrent-voltage curve of the PDs, measured under different LED currents (I_LED). As I_LED increases from 0 mA to 10 mA, the photocurrent rises from 10^-9A to 10^-5A. Moreover, Fig. 3d plots the photocurrents of PD_1-4 measured at varying I_LED and shows a decrease of 26.3% in photocurrent level when transiting from pressed to released states of the sensor, demonstrating the effective response of the PDs to changes in LED intensity.

Fig. 3. Properties of the GaN optochip.

a I–V characteristic of the LED. The inset shows the light output power versus I_LED. b Emission and absorption spectra of the on-chip LED and PD. c Plot of I_PD as a function of voltage. d Plot of photocurrents of the PD_1-4 versus LED currents

Performance of Tactile Sensor

The sensing performance of the tactile sensor is investigated using the experimental setup shown in Fig. 4a. The sensor is fixed on a motorized displacement stage. The LED is driven by a sourcemeter (Keithley 2450), which provides a constant current to the LED. The photocurrents from the PDs are measured using multimeters (Keithley DMM6500) with a measurement resolution of 0.05 nA. During the measurement, the magnitudes of the shear and normal forces applied to the sensor are calibrated using two commercial force sensors (DaySensor DYX-306) with a measurement resolution of 1 mN. Simultaneously, the photocurrents from the four PDs are recorded. The photocurrent changes caused by force variations along the X-, Y-, and Z-axes are denoted as I_X, I_Y, and I_Z, respectively, and are defined as

$$I_X={\Delta I}_{\mathit{PD}3}-{\Delta I}_{\mathit{PD}1}$$

$$I_Y={\Delta I}_{\mathit{PD}4}-{\Delta I}_{\mathit{PD}2}$$

$$I_Z={\Delta I}_{\mathit{PD}1}+{\Delta I}_{\mathit{PD}2}+{\Delta I}_{\mathit{PD}3}+{\Delta I}_{\mathit{PD}4}$$

where ΔI_PD1 = I_PD1 - I_{PD1_0}, I_PD1 represent the real-time reading of photocurrent of PD₁, I_{PD1_0} represents the photocurrent corresponding to PD₁ when no force is applied. The same applies to the other PDs.

Fig. 4. Experimental setup and optimization of the optochip.

a Schematic diagram of the experimental setup for the force measurements. Photocurrent response measured along X-negative axis from 0 to 100 mN in 10 mN increments (b) before and (c) after cutting of the optochip. I_X/Y as a function of shear force (d) before and (e) after cutting of the optochip

The sensing performances of the sensors incorporating optochip before and after laser cutting are investigated. Under a preload of constant normal force of 120 mN, shear force tests are conducted on the sensors along the negative X-axis, with force increments of 10 mN. The test results for PD_1-4 are shown in Fig. 4b, c. When shear force is applied in the direction of PD₃, the PDMS film deforms laterally towards PD₃. The contact area between PD₃ and the PDMS increases, while the contact area between PD₁ and the PDMS decreases. Consequently, the photocurrent of PD₃ increases while that of PD₁ decreases, and the photocurrents of PD₂ and PD₄ on the Y-axis change synchronously. The sensing performance curves of the sensor, based on the optochip before and after cutting, are plotted in Fig. 4d, e. The sensitivity of the sensor, based on the post-cut optochip, increased from -13.7 nA/mN to -29.4 nA/mN. The maximum variation of I_Y on the Y-axis decreased from 0.15 μA to 0.1 μA, indicating a 33.3% reduction in Y-axis crosstalk. Such enhancement can be attributed to the cross-shaped designed of the optochip, which better confines the light into the sensing region and also reducing the phenomenon of external light diffusion.

After identifying the effectiveness of the cross-shaped optochip, the sensor is characterized. The PDMS sensing films with close-packed fingerprint sizes of K = 200 μm, 300 μm, and 400 μm are prepared through identical molding process (See Supporting Information S1). The photocurrent responses of the optochip incorporating three PDMS films are measured and plotted in Fig. 5a–c. Under a normal force of 120 mN, a shear force is applied in the PD₃ direction in 10 mN increments to test their force-sensing performance. For the sensor with the PDMS film with K = 200 μm, the linear measurement range is 80 mN, while it reaches 100 mN for K = 300 μm and K = 400 μm. Comparing the slopes of the fitted curves shows that as the fingerprint size K decreases, the sensitivity of the sensor increases while the linear measurement range decreases, which is consistent with the simulation results in Fig. 2c. Secondly, for K = 300 μm, PDMS films with different spacings of fingerprint pattern were prepared, and their minimum detection limits are tested, as shown in Supporting Information S4. The step graph comparison indicates that as the fingerprint spacing decreases, the photocurrent waveform becomes more stable, and the sensitivity of the sensor gradually increases, consistent with the simulation results in Supporting Information S3. Finally, the sensing film with a fingerprint size of K = 300 μm is chosen.

Fig. 5. Force sensing performance.

Force sensing performance of the device with PDMS sensing film fingerprint sizes of (a) K = 200 μm, (b) K = 300 μm, and (c) K = 400 μm. I_X/Y as a function of shear force along (d) X- and (e) Y-axes. f I_Z as a function of normal force along Z-axis

Through the optimization, the sensor adopting the cross-shaped optochip and a PDMS film with K = 300 μm is further characterized. Figure 5d illustrates the correlation between I_X and the shear force range of ±100 mN along the X-axis. Linear regression analysis reveals an R² of 0.974 and a sensitivity of 30.9 nA/mN. Similarly, for the Y-axis, within the sensing range of ±100 mN, an R² of 0.988 and a sensitivity of 28.3 nA/mN were determined, as shown in Fig. 5e. The photocurrent of the Y-axis shows minimal variation during measurements along the X-axis, and vice versa, indicating that the inter-axis crosstalk is highly suppressed. The sensing range for normal force is depicted in Fig. 5f, spanning from 0 to 200 mN, with an R² of 0.977 and a sensitivity of -154.8 nA/mN.

The dynamic response of the sensor is characterized by applying a shear force along the X-axis, where PD₃ is positioned. Experimental configuration for transient response measurements is provided in Supporting Information S5. Figure 6a illustrates the transient response of the sensor to an applied shear force of 100 mN, revealing a response time of 0.85 ms and a recovery time of 0.82 ms. The response time is defined as the duration for photocurrent decay from 90% to 10% of its peak value during the falling edge, while recovery time denotes the duration required for photocurrent to return from 10% to 90% of baseline during the rising edge. In addition, investigations under varying shear force magnitudes confirm consistent sub-millisecond response/recovery times across all tested conditions (See Supporting Information S6). Figure 6b presents the photocurrent response of sensor when applying dynamic force ranging from 0 mN to 60 mN in 10 mN step. The response curve shows a symmetrical distribution of stepwise changes, indicating high stability and repeatability of the sensor. Figure 6c depicts the dynamic response of the sensor to varying shear force levels of 30 mN, 60 mN, and 90 mN, with corresponding photocurrent changes of approximately 0.41 μA, 0.75 μA, and 1.00 μA, respectively. Figure 6d depicts the photocurrent variations of sensor in response to force variations, revealing that a distinct stepwise change of 1 nA in photocurrent can be observed. Based on the linear correlation between the photocurrent changes of the sensor and the magnitude of the applied force, its measurement resolutions in response to shear and normal forces are determined to be approximately 0.07 mN. The measurement response in response to normal force can be found in Supporting Information S7.

Fig. 6. Dynamic performance and durability of the tactile sensor.

a Plot of the transient response of the sensor. b Dynamic response of the sensor when subjected to force increments and decrements of 10 mN. c Photocurrent curves measured under varying shear force conditions. d Detection limit of the sensor measured with increasing shear force. e Response of the sensor when subjected to a 3500-cycle test of the X-axis shear force within the range of 0–100 mN. f Enlarged view of the test results presented in (e)

The stability of the sensor is evaluated through repetitive cyclic testing. As shown in Fig. 6e, the sensor is subjected to a shear force of 100 mN, which is applied and released over more than 3500 cycles. The photocurrent curves illustrate a high degree of reproducibility across multiple cycles. A partially enlarged view in Fig. 6f shows that the photocurrent variations of ΔI_PD3 and ΔI_PD1 exhibit opposing trends during the test, consistent with the sensing response shown in Fig. 2b. The experimental results of the repeatability test for the Y- and Z-axes directions can be found in Supporting Information S8.

Table 1 presents a comparative analysis of the tactile sensor developed in this study with previously reported tactile sensors. The sensor exhibits superior performance in terms of measurement range, resolution, sensitivity, and response time. Notably, it consistently offers a linear response over a wide measurement range, coupled with high resolution, along all three axes. Moreover, the GaN optochip employs a monolithic integration design with wafer-scale manufacturing techniques, significantly reducing device footprint and manufacturing costs. While the prospect of further thickness reduction of the PDMS layer holds promise for expanding application scenarios, direct thinning may compromise its deformability, thereby limiting the operational force range. Subsequent research will explore alternative elastomer formulations and microstructural engineering strategies to achieve thinner device profiles without sacrificing sensing performance or structural robustness.

Table 1.

Comparison of shear sensing devices/systems

Mechanism	Range (X-, Y-, Z-axis)	Resolution (X-, Y-, Z-axis)	Sensitivity	Response time	Characteristics
Hall effect40	19.5 mN in X- and Y-axes	0.2 mN in X- and Y-axes	6.63 µT/mN	73 ms	Non-Linear
Magnetoresistance41	7.8 mN	0.33 mN	9.62 mN/mV	—	Non-Linear
Piezoresistive42	100 mN, 100 mN, 150 mN	5.4mN in X- and Y-axes	25.76 N-1	112 ms	Linear
Piezoresistive43	0.6 N, 0.6 N, 0.3 N	50 mN, 50 mN, 34 mN	12.1 kPa-1	3.1 ms	Non-Linear
Piezoresistive44	90 mN, 90 mN, 1.8 N	30 mN, 30 mN, 10 mN	7.73%/N	—	Linear
GaN optochip39	20 mN, 20 mN, 50–200 mN	2 mN, 2 mN, 2 mN	—	—	Non-Linear
Optoelectronic chip45	15 mN, 15 mN, 220 mN	1 mN, 1 mN, 1 mN	−7.42 nA/mN	3.4 ms	Non-Linear
This work	±100 mN, ±100 mN, 200 mN	0.07 mN, 0.07 mN, 0.07 mN	30.9 nA/mN	0.85 ms	Linear

Applications

The utility of the tactile sensor is explored for its potential applications in morphology scanning. The process of topography scanning is analogous to the touch-sensing mechanism of the human fingertip, whereby the sensor is slid over the surface of the sample to discern and capture its topography. The scanning of acrylic patterns comprising 1D semi-cylindrical ridges and 2D hemispherical arrays is illustrated in Fig. 7a–c. The scanning intervals along the X- and Y-axes are set at 1 mm. Topographic maps are constructed based on the photocurrent data acquired from uniaxial and biaxial scanning, as shown in Fig. 7d, e. These maps provide a visual representation of the scanned surface, where the intensity variations of I_X and I_Y within the maps directly correspond to the intricacies and variations present in the scanned samples. The superposition of data from dual-axis scanning provides an enhanced depiction of the surface topographical characteristics.

Fig. 7. Applications in morphology scanning and medical diagnostics.

a Optical image of the experimental setup for scanning measurements. Optical images of the (b) 1D and (c) 2D acrylic patterns for scanning. Photocurrents maps measured from (d) 1D and (e) 2D acrylic patterns along different axes. (f) Optical images showing the measurement on porcine kidney. g Photocurrent maps measured from healthy porcine kidney and porcine kidney with a simulated nodular along different axe

In the field of medical diagnostics, the tactile sensor is employed to emulate the functionality of surgical palpation probes to detect sample profiles. Owing to the inherent invisibility of diseased nodules from the superficial layer, tactile sensing techniques have emerged as a feasible solution for assessing the size and location of the nodules. To simulate a pathological condition, a resin pellet with a diameter of 5 mm is used to serve as a diseased nodule and embedded into a fresh porcine kidney. Subsequently, the sensor is gently applied to the surface of a porcine kidney for palpation experiments, as shown in Fig. 7f. Palpations of both treated and untreated porcine kidneys are conducted in a 10 × 10 mm² area along the X- and Y-axes. The resulting photocurrent maps are illustrated in Fig. 7g, which depict the morphology of the scanned samples. Notably, when palpating a healthy kidney, minimal variations in photocurrent below 0.006 μA are observed. Conversely, in the presence of the mimic diseased nodule, a significant alteration in photocurrent, reaching up to 0.295 μA, is recorded. The combined scanning outcomes along both axes facilitate a refined and precise morphology of the nodule, thereby demonstrating the promising diagnostic potential of the tactile sensor.

Conclusion

In summary, an opto-electro-mechanical tactile sensor has been developed, comprising a monolithically integrated GaN-on-sapphire optochip with a fingerprint-patterned PDMS film. The optochip, which incorporates an LED and four independent PDs, is capable of detecting optical changes that arise from the force-induced deformation of the PDMS film at the sapphire interface, enabling force sensing along different axes. The cross-shaped chip geometry, achieved through laser cutting, serves to enhance sensitivity and effectively mitigate crosstalk interference between the measurements along the X- and Y-axes. The optimized sensor exhibits three-axis force sensing capability with a wide measurement range, a high resolution of 0.07 mN, a fast response/recovery time of 0.85 ms/0.82 ms, as well as high repeatability and stability. Moreover, experimental evaluations in surface morphology scanning and medical palpation demonstrate the potential practicality of the sensor.

Materials and methods

Fabrication of GaN optochip

A 6-µm-thick GaN-based epitaxial structure is grown on a 4-inch c-plane sapphire substrate using metal-organic chemical vapor deposition, comprising an unintentionally doped GaN, n-GaN, InGaN/GaN multiple quantum wells (MQWs), and p-GaN. A 120-nm-thick indium tin oxide layer (ITO) is deposited on the p-GaN surface, serving as a current spreading layer. The mesas of a light-emitting diode (LED) and four photodetectors (PDs) are patterned through photolithography, and the unmasked areas are etched to reveal the n-GaN using inductively coupled plasma etching. Electrodes are fabricated using a combination of photolithography, electron beam evaporation, and lift-off techniques. A 360-nm-thick SiO₂ passivation layer is coated via plasma-enhanced chemical vapor deposition, followed by the deposition of a 3.2-µm-thick SiO₂/TiO₂ distributed Bragg reflector (DBR) using an optical thin-film coater. The 2.4-µm-thick metal pads are formed using a combination of photolithography and electron-beam evaporation. Subsequently, the sapphire thickness is polished down to 150 µm and the fabricated wafer is subjected to a laser dicing process, which separates it into individual chips.

Preparation of PDMS films

The PDMS gel (Sylgard 184, Dow Corning) is formulated by mixing prepolymer and curing agent in a 10:1 ratio and then poured over the two resin molds with inversed patterns of pyramid and fingerprint structures. After curing at 85 °C for 2 h, the double-side patterned PDMS film with an overall size of 7 × 7 × 1.5 mm³ is formed. Detailed fabrication procedures can be found in Supporting Information S1.

Assembly of tactile sensors

The GaN optochip is flip-chip bonded onto a printed circuit board (PCB) to establish electrical connections. Subsequently, the patterned PDMS film is aligned and fixed above the optochip.

Author contributions

Conceptualization, R.Y.H., T.X.C., L.Z., and K.H. Li.; methodology, R.Y.H., T.X.C., L.Z., and K.H. Li.; Design and fabrication, R.Y.H., T.X.C., and K.H. Li.; data curation and analysis, R.Y.H., T.X.C.; supplemental materials, R.Y.H., T.X.C.; writing—original draft, R.Y.H., T.X.C., L.Z., and K.H. Li.; writing—revision & editing, R.Y.H., T.X.C., L.Z., and K.H. Li.; funding acquisition, L.Z., and K.H. Li.; R.Y.H., and T.X.C. contributed equally to this work.

Data availability

The data supporting plots within this paper and other findings of this study are available from the corresponding author upon request.

Conflict of interest

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41378-025-00998-0.