Introduction

Panoramic perception, a technology enabling comprehensive information sensing, has become an intriguing topic of research. Due to the short wavelengths of electromagnetic waves, current information perception methods using electromagnetic waves (such as vision and lidar) demonstrate high directionality, but the field of view of these devices remains very limited. Achieving 360-degree information acquisition with a single-stack device is challenging and generally requires sensor arrays1. As a crucial modality for perceiving the physical world, acoustic waves offer extra advantages for panoramic perception due to their longer wavelengths. Acoustic perception, especially directional acoustic sensing, has found applications in numerous fields including national defense security2,3, environmental protection4,5,6, human-computer interaction7,8,9, and healthcare10,11,12.

Researchers have developed various acoustic sensors based on different transduction mechanisms, including piezoresistive13,14, capacitive15,16, piezoelectric17,18,19,20, optical21,22, and triboelectric principles23,24. These sensors predominantly operate through the principle of acoustic pressure detection using diaphragms. Recently, Pezone et al. proposed wafer-scale integrated MEMS microphones, which addressed performance issues arising from mass and stiffness increases in large diaphragms and exhibited high mechanical compliance (0.081–1.07 μm/Pa)25. Thus, the transition from rigidity to flexibility represents a significant research direction in acoustic sensors, aiming to achieve enhanced sensitivity across a wider frequency spectrum. Wang et al. developed a piezoelectric acoustic sensor by utilizing low elastic modulus ferroelectric materials, achieving a high sensitivity of 47.43 mV Pa−1 cm−2 and a high frequency resolution of up to 0.1 Hz26. The majority of these sensors operate on the principle of diaphragm-based acoustic pressure detection. Directional acoustic sensing is generally achieved by detecting differences in acoustic waves, that is, by using microphone arrays to measure the pressure differences and time delays at multiple points27,28,29. Owing to the non-directionality of acoustic pressure, these sensors exhibit inherent limitations. The magnitude of the acoustic pressure gradient difference is closely related to the spatial distance and acoustic frequency. For the smaller distance and the lower frequency, the gradient difference is smaller and the measurement accuracy is lower. Itani et al. utilized a self-distributing wireless microphone array across a large area to achieve localization and separation of speech from different 2D regions30. Therefore, existing acoustic pressure-based acoustic sensors struggle to achieve accurately measurement of the direction of low-frequency acoustic waves while being miniaturized31,32.

In contrast to diaphragm-based methods that detect acoustic pressure, many animals (such as spiders and mosquitoes) use hair to sense acoustic flow33,34. Gong et al. summarized the mechanisms of biological flow perception, demonstrating that hair can efficiently capture mechanical signals35. Due to the action of fluid viscous forces, the highly damping ciliary structures can move with the fluid flow. Since flow is a vector that inherently contains information about the direction of acoustic propagation, this type of acoustic flow sensing not only detects acoustic with high sensitivity but also has an intrinsic directivity to the direction of acoustic. Zhou et al. demonstrated the potential of natural spider silk for detecting acoustic flow and proved that spiders can use webs several times larger than themselves to remotely detect and locate the acoustic waves emitted by prey or predators36,37. While these findings demonstrate that natural spider silk can couple acoustic flow with near-maximum physical efficiency, which is significant for understanding the auditory mechanisms of animals and developing new types of acoustic sensors, natural spider silk lacks the capability for large-scale standardized production. Current research on spider-web structures mainly focuses on pressure and strain sensing. For instance, Dong et al. prepared a spider-web-like polyimide aerogel, achieving highly sensitive “multi-touch” stress detection38. Wu et al. fabricated spider-web-like hydrogel sensors using 3D printing, enabling intelligent gesture recognition39. To date, the use of photolithography to fabricate bio-inspired web-like structures for low-frequency directional acoustic detection has not been reported.

Inspired by this auditory mechanism, we propose an acoustic sensing device based on a self-supporting biomimetic web-like structure (WS). This web-like acoustic sensor (WAS) is compatible with standard semiconductor manufacturing processes, and exhibit exceptional acoustic sensing capabilities within the 10–2000 Hz range. Benefiting from the optimized structure and the acoustic flow sensing mechanism, the device exhibits remarkable sensitivity of 9.36 mm/s/Pa @100 Hz, high linearity (R2 > 0.99), fine frequency resolution (0.05 Hz) and excellent directional acoustic detection capabilities. It can achieve acoustic source localization without relying on conventional large-scale microphone arrays. This offers a novel approach for advanced miniaturized bio-inspired acoustic sensing and positioning devices. Therefore, it has significant application potential in various fields such as heart acoustic detection, drone detection, underwater acoustic detection, and disaster monitoring that require low-frequency, high-sensitivity, small-size, and directional acoustic sensing. Furthermore, the device can also be utilized as a microfluidic or micro-vibration sensor.

Results

Concept of the WS

Spider webs, with their complex structure, not only excel in performing the primary function of hunting, but can also highly respond and effectively capture weak acoustic signal driven by the frictional force generated by the movement of surrounding air particles during sound propagation (Fig. 1a)40,41,42 The spider web is a highly efficient acoustically coupled antennas, which amplify the information of air particle movement into signals that spiders can interpret43,44,45. This mechanism is different from the well-known signal transmission caused by direct contact between animals and silk46,47, and the acquisition of sound information through eardrums10,48. Inspired by this mechanism, we constructed an artificial high-efficiency acoustic flow coupling microstructure modeled on a spider web, and utilized optical interference for data transmission.

Fig. 1: Schematic diagram of the WAS inspired by the auditory system of orb-weaving spiders.
figure 1

a Spider web is a highly sensitive acoustic antenna that detects airborne flow caused by acoustic signals. b Schematic diagram of the simplified model of a spider web

Full size image

Modelling and theoretical analysis of the WS

Spiders can dynamically adjust the tension and structure of their webs, optimizing their ability to detect weak acoustic waves49,50. As shown in Fig. 1b, we propose a parametric simplified model of a spider web structure, which composes of a hub diaphragm, radial threads and spiral threads. The structural model in Fig. 1b includes seven design parameters, which are L, d, wr, ws, Ns, Nr, and T. The length of all radial threads, which corresponds to the theoretical minimum diameter of the entire WS, is denoted as L. The diameter of the central hub diaphragm (used for subsequent information retrieval) is represented by d. The linewidths of radial and spiral threads are wr and ws, respectively. The number of radial and spiral threads are Nr and Ns, respectively. The overall thickness of the WS is T. Meanwhile, the entire WS exhibits strict periodicity. The angle between adjacent radial threads remains consistent, and the distance between adjacent spiral threads is uniform.

Under the action of airborne acoustic flow, the WS undergoes motion and can be modelled using the structural dynamics equation. Fig. S1 represents the position of the WS in the coordinate axes. The detailed modeling process and theoretical analysis of the WS can be found in the Supplementary Information. Based on the approximate analysis, the WS move with the fluid with the same amplitude, which means that the fluid motion is almost fully faithful across a wide range of frequencies. Using the partial differential equation to directly calculate the motion of the 2D WS in a plane-wave acoustic flow field is nearly impossible. Therefore, we investigated the relationship between the structure dimensions of the WS and its acoustic response performance using Finite element analysis (FEA). The material of WS is polyimide (PI), which is low density, high strength, and compatibility with the standard manufacturing processing (E = 3.1e9 Pa, ρ = 1300 kg/m3, and Poisson’s ratio=0.37).

When acoustic flow acts on the WS, it undergoes motion. As shown in Fig. 2a, b, it is the velocity and displacement of the WS (the parameters of the model are provided in supplementary Table S1) under a 0.1 Pa planar acoustic wave, respectively. The deformation in the figure is magnified 1500 times. The amplitude of displacement of the WS in the acoustic field is slight, so that the effect of the nonlinear effect can be neglected. In Fig. 2c, d, and e, the FEA simulations of the acoustic response properties of the WS to the diameter L, thickness (T) and linewidth (wr and ws) are displayed. The simulated results indicate that the acoustic response performance of the WS is improved for the increasing diameter, and decreasing thickness and linewidth. Specifically, within a larger frequency range, Vsilk/Vair ≈ 1. The L and T have notable impact on the WS’s acoustic response, while the linewidth has small effect. The acoustic response performance of the WS with different numbers of radial and helical threads is displayed in Figs. S2 and S3. For the increasing Nr, the WS’s acoustic response performance is improved. However, the effect of Ns (within the range of 1 to 15) on the WS’s acoustic response is very small. Increasing the diameter of hub diaphragm, the WS’s acoustic response performance will deteriorate, as shown in Fig. S4. In the simulation, only one parameter is changed each time, while the influence of WS stress on the acoustic response performance is ignored.

Fig. 2: FEA results of the WS.
figure 2

a Vibration velocity and (b) deformation of the WS with radial threads of L = 10 mm, wr=ws = 3 μm, Nr = 24, Ns = 15, T = 1 μm, d = 200 μm under 0.1 Pa acoustic pressure at 100 Hz. c Frequency response of the WS with various length (L) of radial threads (d) Various thicknesses (T). e Various linewidth

Full size image

Fabrication of the WS

The ultra-thin WS with large size and narrow linewidth can achieve better acoustic response performance. However, such WS is extremely fragile, poor stability and difficult to process. Considering these factors, a WS with L = 10 mm, wr=ws = 4 μm, Nr = 24, Ns = 15, T = 2 μm, d = 200 μm is fabricated to verify that the proposed WS with excellent acoustic performance. To facilitate subsequent clamping and fixation for acoustic response performance testing, a 10 mm square frame was externally reserved around the WS. The detailed manufacture process of the WS based on the lift-off/deep silicon etching process is shown in Fig. S5. Figure 3a presents a photograph of the WS. As shown in Fig. 3b, it provides the center microscopic image of WS, and the measured d of the WS is about 196 μm. The thickness and linewidth of the WS are characterized by using an atomic force microscope (AFM), as shown in Fig. 3c–f. The thickness of the WS is about 1.96 μm, and the linewidth is about 3.7-4.1 μm. The dimensional deviations are mainly caused by over-etching.

Fig. 3: Macroscopic and microscopic geometry of the self-supporting WS.
figure 3

a Optical image of the WS. b Microscopic image of the WS with 10x objective lens. c AFM images of the hub diaphragm edge in the WS. d Height profile of the hub diaphragm edge indicates that the T is approximately 1.96 μm. e AFM images of one thread in the WS. f Height and width profile of a thread indicate the upper width being about 3.7 μm and bottom width being about 4.1 μm

Full size image

Acoustic sensing ability of the WS

To characterize the acoustic sensing performance of the WS, we built the acoustic property measurement device as shown in Fig. 4a. The acoustic flow field is generated by a speaker located at 2 m away in the orthogonal direction of the WS. The acoustic-induced the WS motion is measured by optical interference. The WS is fixed on a three-dimensional precision displacement stage to ensure accurate positioning in the vertical and horizontal directions. The acoustic pressure near the WS is measured by a standard pressure microphone. All experiments are performed in a quiet environment.

Fig. 4: Acoustic response performance of the WS.
figure 4

a Schematic diagram of experimental setup for testing the WS. Inset is the laser beam focused on the WS’s hub diaphragm. b Velocity response of the WS at 500 Hz for different acoustic pressure. c Time response and the corresponding frequency spectrum of the WS for the 500 Hz acoustic wave at 0.05 Pa. d Frequency response of the WS with various sound pressure levels. e Velocity versus acoustic pressures at various frequencies. f Time response and the corresponding frequency response spectrum of the WS to dual frequency acoustic signals at 500.00 Hz and 500.05 Hz. g Time response and the corresponding frequency response spectrum of the WS to acoustic signals at 499.95 Hz, 500.00 Hz and 500.05 Hz. h Schematic diagram of directionality measurements. i Time response of the WS to the incident acoustic wave at 500 Hz in different directions. j Theoretical and measured directionality of the WS at 500 Hz, the blue curve is the theoretical results and the red dots are the measured results. All measurements are performed in the hub diaphragm of the WS. Error bands (d) and bars (e) represent the s.d. of the results of at least three replicates

Full size image

The performance of the WS is tested by using 500 Hz acoustic signals at different acoustic pressure. As shown in Fig. 4b, the time domain velocity output of the WS is typical 500 Hz sinusoidal alternating signals at different acoustic pressures, and the velocity increases as the input acoustic pressure increases. The time-domain signal is fast Fourier transformed with a distinct amplitude peak at 500 Hz, indicating that the velocity output contains mainly 500 Hz frequency signals, consistent with the input acoustic signal as shown in Fig. 4c. For other acoustic frequency signals, the time response of the WS and the corresponding frequency domain response are shown in Fig. S6. The locations of the amplitude peak in the frequency domain are consistent with the input signal. For different acoustic pressures, the measured frequency response of the WS is shown in Fig. 4d. The velocity response at different acoustic pressure at several representative frequencies is shown in Fig. 4e. According to Fig. 4e the measured velocity sensitivity of the WS is 3.70 mm/s/Pa at the acoustic frequency of 500 Hz. Meanwhile, the maximal velocity sensitivity of the WS is 9.36 mm/s/Pa @100 Hz. The correlation index R2 of the WS at different frequencies is greater than 0.99, which indicates that the designed WS has excellent linearity. The experimental results deviate from the simulation results because the effect of stress on the acoustic response is not considered during the FEA, but this does not affect our use of finite element simulation to guide the structural design of the WS.

Since actual acoustic signals typically have multiple frequency components, the detection resolution of the acoustic transducer is critical for high-fidelity acoustic signals sense. The frequency resolution of the WS is verified by using dual-frequency (500.00 Hz and 500.05 Hz) and triple-frequency (499.95 Hz, 500.00 Hz, and 500.05 Hz) acoustic signals. The detected time response and the corresponding frequency spectra are shown in Fig. 4f, g. In Fig. S7, the frequency resolution of the WS for acoustic signals around 1000 Hz is also demonstrated. It indicates that the WS is able to detect acoustic signals with complex frequencies with a fine resolution of 0.05 Hz. This capability is derived from the web-like structure of the WAS and is superior to most acoustic sensors published in previous studies (e.g., 0.5 Hz in ref. 51, 0.1 Hz in ref. 52).

Furthermore, the directionality and stability are also important performance of the WS for their applications. Unlike the traditional pressure sensor, the WS operates as a velocity sensor, where velocity is a vector inherently carrying directional information. Therefore, the WS can serve as an excellent vector acoustic sensor. As shown in Fig. 4h, the acoustic source was moved in 15° increments and maintains constant sound frequency and intensity during the experiment process. Simultaneously, the time-domain spectra of sound signals retrieved from the WS are recorded, as depicted in Fig. 4i, j. Taking 500 Hz acoustic signal as an example, when the angle is changed from 0 to 90°, the velocity of the WS is reduced from 1.13 mm/s to 0.12 mm/s. Therefore, the velocity response is a reduction of 89.4%. The responses of directionality for other acoustic frequencies are illustrated in Fig. S8. These results illustrate the excellent directional acoustic sensing capability of the designed WS. To further elucidate the WS’s long-term stability, an evaluation is conducted in Fig. S9 for 17,000 cycles at 2 kHz and 0.1 Pa acoustic pressure.

Discussion

When the linewidth and thickness of the WS are sufficiently small, mechanical forces, such as bending stiffness and inertia, are dominated by fluid forces related to air viscosity. Consequently, the WS is primarily constrained to move in unison with the surrounding viscous air. As a result, the WS has very high efficiency to be coupled to that of the acoustic flow as shown by the red line in Fig. 5a. Additionally, the WS show a low specific acoustic impedance Zs = 106.84–4490.36 Pa*s/m, and the Zs is calculated by dividing the acoustic pressure by the particle velocity as shown by the blue line in Fig. 5a. The WS’s exceptional mechanical characteristics, including high coupling efficiency and Ultra-low Zs, enabled the sensor to exhibit high mechanical compliance (Cm = 23.6–0.016 μm/Pa) as shown in Fig. 5b. The Cm of the commercial MEMS microphone is 1.3 nm/Pa (MP23DB01HP). Additionally, the Cm of the WS is more than 10–1000× higher than the silicon, graphene, metals reported in the state-of-the-art diaphragms, to our knowledge27,53.

Fig. 5: High coupling efficiency verification, mechanical compliance testing and MDP testing of the WS.
figure 5

a Coupling efficiency of the WS and acoustic flow (red line), Specific acoustic impedance of the WS (blue line). b Mechanical compliance of the WS. c NSD of the WS, the red dot line is the noise spectrum used to calculate MDP. d Measured MDP of the WS

Full size image

The overall noise of the WS, with the speaker turned off, is measured using the same setup in Fig. 4a. The amplitude of the noise spectrum is divided by a frequency interval of 0.916 Hz to obtain the noise spectral density (NSD), as shown in Fig. 5c. The main characteristics of the WS overall noise are optoelectronic noise, thermo-mechanical noise and environmental noise. The WS’s whole noise is dominated by the thermo-mechanical noise above ~300 Hz. The room’s environmental noise dominated the frequency band below ~300 Hz, which could be reduced in an acoustic isolation room. The minimal detectable acoustic pressure (MDP) is defined as PMDP = UNSD/Sm, where UNSD is the whole NSD and Sm is the measured frequency response. For most sensors, the higher the mechanical sensitivity (Sm), the lower the MDP, but high sensitivity does not mean that the microphone can detect faint sounds, as its MDP also depends on the noise level. As shown in Fig. 5d, the WS has high sensitivity below ~300 Hz, but the measured MDP is lower. The possible factor is that WS is susceptible to the environmental noise due to its higher Sm. Supplementary Table S2 presents a detailed comparison of key parameters (Cm, MDP) for representative acoustic sensors reported in the literature, thereby substantiating the superior performance of the WS.

To demonstrate the application of the WS as an acoustic sensor, we measure the response of human voice of the WS. Male and female participants repeat six simple words (“Hello world,” “One,” “Two,” “Three,” “Four,” “Five”). Using the real-time fast Fourier transform analysis, we examined the voice waveforms and spectrograms of speakers. The acoustic waveforms (top) and spectrograms (bottom) of a female speaker are shown in Fig. 6a (standard microphone) and Fig. 6b (the WAS sensor). The acoustic waveforms and spectrograms of a male speaker are shown in Fig. 6c (standard microphone) and Fig. 6d (the WAS). The waveforms and spectrograms obtained from the WAS are closely matching the corresponding airborne acoustic signals measured by the standard microphone. Therefore, the designed WAS is high-fidelity capability in detecting sound signals. The subtle differences between the waveform and spectrogram of the acoustic waves recorded by the standard microphone and the WAS are mainly due to the frequency response differences between the two sensors. The WAS, designed for low-frequency directional sensing, exhibits much higher sensitivity at low frequencies than the standard microphone. However, as shown in Fig. S10, the WAS exhibits attenuation at frequencies above 1000 Hz until it is nearly zero, whereas the standard microphone maintains a flat response across this range. Notably, the acoustic response range of the WAS can be extended by optimizing its structural parameters, such as reducing the thickness and linewidth of the WS, as illustrated in Fig. 2. Consequently, there will be some differences in the recorded waveforms of acoustic waves by the two sensors.

Fig. 6: Application of the WS for voice recognition and micro-vibration measurements.
figure 6

Pronunciation of a female speaker recorded by the reference microphone (a) and the WAS (b). Pronunciation of a male speaker recorded by reference microphone (c) and the WAS (d). e Schematic diagram of the micro-vibration measurement, where knocking and vibration are performed separately. Time-domain waveforms (f) and spectrograms (g) of the response of the WS to finger-knocking at different distances. Time-domain waveforms (h) and corresponding frequency-domain (i) of the response of the WS to vibration of the phone

Full size image

Leveraging the WS’s excellent information sensing capability, we also demonstrated its potential as a micro-vibration sensor. Figure 6e illustrates the schematic diagram of vibration test, where micro-vibrations originate from the finger knock at different locations and vibrations from a phone. The experimental results indicate that the WS can respond to finger-knocking at different locations, as shown in Fig. 6f, g. Subsequently, placing the phone near the WS, as shown in Fig. 6h, i, the WS clearly capture the vibration’s information from each phone vibration. These experimental results validate the WS’s ability to sense micro-vibrations.

Conclusion

In summary, we designed and fabricated the bio-inspired web-like structure to sense acoustic flow. The capability of WS as a transducing medium for manufacturing acoustic sensors are demonstrated theoretically and experimentally. Benefitting from the acoustic flow sensing mechanism, the WS exhibits exceptionally ultra-high mechanical compliance, high sensitivity, inherent directivity toward acoustic field orientation, outstanding detection resolution and high linearity with a correlation coefficient exceeding 0.99. These characteristics make the WS suitable to detect faint acoustic information and voice recognition. Furthermore, due to the WS’s compatibility with standard fabrication methods, it offers advantages in large-scale production and cost-effectiveness. This study provides a novel approach for developing the next generation of acoustic flow sensors, offering considerable potential for applications in panoramic perception.

Materials and methods

The WS fabrication

The fabrication processing of the WS can be divided into three main steps, which are (i) growth and etching of SiO2; (ii) patterned etching of PI; and (iii) deep silicon etching on the backside.

(i) The wafers are successively sonicated using acetone and isopropyl alcohol for 5 min to clean off the impurities on the wafer surface, and then is wash with deionised water and dried. The SiO2 grow on the wafer surface by thermal oxygen at a thickness of about 800 nm. Photoresist is applied to the wafer surface, and the wafer is exposed and developed by using an MA8 photolithography machine. The SiO2 on the wafer is etched using a P5000 device. After etching completion, the silicon wafer is organically cleaned using organic solvents. (ii) Spin-coating the adhesive on the wafer surface by a levelling machine and then spin-coating the PI. The PI is cured at high temperature and naturally cooled to room temperature. Photoresist is applied again to the silicon surface, and the coated wafer is exposed using an MA8 photolithography machine. After development, the PI layer on the silicon wafer is etched using a P5000 equipment. Upon completion of etching, the silicon wafer undergo organic cleaning and plasma cleaning, achieving patterned etching of the PI. (iii) Photoresist is applied to the backside of the silicon wafer, and the coated wafer was exposed and developed. The SiO2 layer on the backside of the silicon wafer is etched. A protective layer is applied to the front side, and scribing is performed on the backside. Deep silicon etching is carried out from the backside of the silicon wafer to create a suspended film. The sample is then heated and soaked in alcohol and acetone to remove the photoresist. Finally, the silicon wafer is immersed in an HF acid solution for 2-3 minutes, completing the preparation of the WS. Fabrication steps is shown in Fig. S5.

Characterization and acoustic measurement

A commercial loudspeaker (EDIFIER R2000DB) is employed as the source of the acoustic signals. The frequency and intensity of the acoustic signals are controlled by using a computer and audio test software. We record acoustic pressure using a standard microphone (AWA-14435). The microphone output is amplified by a charge amplifier and transmitted to a computer after collection by a data acquisition card (DAQ, 1706U). Acoustic-induced the WS motion is measured using a laser vibrometer (SOPTOP LV-S01). To avoid interference from external vibrations and noise, all experiments are conducted in a quiet environment, and the testing equipment and the WS are installed on a sturdy optical base.

FEA modelling

The acoustic response of the WS is simulated by finite element methods. The material parameters for the PI film are set as follows: E = 3.1e9 Pa, ρ = 1300 kg/m³, and Poisson’s ratio = 0.37. The sound field is a plane wave with an amplitude of 0.1 Pa, and the frequency step size is 20 sampling points per 10 times frequency.