Bioinspired soft electroreceptors for artificial precontact somatosensation

Artificial haptic sensors form the basis of touch-based human-interfaced applications. However, they are unable to respond to remote events before physical contact. Some elasmobranch fishes, such as seawater sharks, use electroreception somatosensory system for remote environmental perception. Inspired by this ability, we design a soft artificial electroreceptor for sensing approaching targets. The electroreceptor, enabled by an elastomeric electret, is capable of encoding environmental precontact information into a series of voltage pulses functioning as unique precontact human interfaces. Electroceptor applications are demonstrated in a prewarning system, robotic control, game operation, and three-dimensional object recognition. These capabilities in perceiving proximal precontact events can lenrich the functionalities and applications of human-interfaced electronics.


Fig. S1
The physical model of a set of charged finite-sized planes.
To begin with, the physical model of a set of charged finite-sized planes is established (Fig. S1).We assume that there have n finite-sized planes and they have the same dimensions a and b along the x and y directions.All planes are centered at (x, y)=(0, 0) and located at positions z1, z2, …, zn with the surface charge densities σ1, σ2, …, σn.The electric potential (ϕ) at an arbitrary point r=(x, y, z) is Where ε(r) is the permittivity constant of the material at position r.
The electric field (E) is When integrating over the x' and y' directions, where x=y=0, the electric field along the z direction is 2: The simulation of the electroreceptor According to the physical model above and Fig. 2a, we assume that there is a space coordinates system located on the center of organo-hydrogel electrode.The equivalent charge planes of the ionic electrode, the elastomeric electret and external object are located at z1, z2, and z3 positions with the same size, respectively.The electric potential of the electrode at the position (0, 0, z1) is Where σ1, σ2 and σ3 are the surface charge density of external object, the elastomeric electret and the ionic electrode, respectively; z represents the separation distance between the electroreceptor and external object; ε1 is the permittivity of the elastomeric electret.
Under the open-circuit condition, σ3=0, therefore the open-circuit potential value of the electrode could be simplified as: The numerical calculation results of this model will be commendably consistent with the finite element simulation results of COMSOL.This has been adequately proven by our previous works (41,42).The analytic calculations of the equations above are extremely complicated.Therefore, in this work, COMSOL simulation was selected to visually reveal the variation tendency of the electrode potential when an object approaches the electroreceptor.

Supplementary Text 2
The optimization of electroreceptor matrix for higher resolution The resolution of the electroreceptor array here could be defined as the number of electroreceptor pixels per unit area.Obviously, the higher density of the pixels leads to a higher resolution and better profile recognition.However, there exists the edge effect for the electrorecetpor.The electric field lines of the electroreceptor units interfere with each other according to the superposition theorem, which hence affects the voltage output of every single electroreceptor.We previously carried out work on array triboelectric nanogenerators (TENG), and studied the edge effects of electrostatic filed theoretically (41).According to this work, the edge effect is highly related to the geometry structure and distribution of each unit (41).
Assuming the length of each square unit is represented by L, and the space between two adjacent units is W.Then, the W/L ratio has a direct effect on the electroreceptor unit's output.The output voltage of the electroreceptor unit will be decreased as reducing W/L ratio.
Clearly, a small W/L implies a small distance between the units, such that the electric fields originating from the edges of units interfere significantly with each other, thereby generating distortion electric field distributions.This is easy to be understood that the edge effect will be suppressed when the units are sparsely distributed.This trend has been demonstrated in our last study as shown in the above figures.On the other hand, as W/L increases, the spacial resolution will be reduced, as the too-sparsely distributed arrays will be unable to identify complex geometric shapes.Therefore, there seems to be a trade-off between the dense unit distribution and high voltage resolution.However, the spatial resolution can still be improved by increasing the unit numbers and reducing the unit sizes, but cautions should be paid, that the W/L ratio should be kept to be around a certain optimum value to not exacerbate the edge distortion effect.According to the previous study, the optimized W/L ratio is around 0.7 (41).
To verify whether the resolution can be improved by geometry optimization and to prove the proposed technology that is available for the "deep learning", some necessary experiments have been carried out by varying the geometry and layout of the electroreveptor arrays: Take the 3×3 array as an example.We firstly designed and fabricated two 3×3 electroreceptor arrays whose total working area is 5×5 cm 2 .Two kinds of W/L ratios were adopted.The first one: W = 1 cm and L = 1 cm, W/L=1; the second one: W = 0.4 cm and L = 1.4 cm, W/L=2/7 (Fig. 4E).We used the same metal ball with the same radius, r= 5 cm, to approach the two electroreceptor array at the same height of 1 cm, respectively.It can be found that although the unit voltage of the array W/L=2/7 is larger than that of the one W/L=1, the voltage distinction between each unit becomes smaller.This means the voltage resolution is improved.By using the height distance-voltage relationship obtained in Fig. 2F, we calculated and fitted the ball radius using the two arrays.The results showed that the larger W/L=1 ratio leads to a more accurate fitted ball radius of 4.84 cm, much improved from the 8.24 cm when W/L is 2/7 (Fig. 4F).
Then, we further developed a 5×5 electroreceptor array with more and finer pixels.The W/L of which was set to W = 0.5 cm and L = 0.6 cm, W/L=5/6 (Fig. 4G).The fitted ball diameter has even further improved accuracy of 5.03 cm (Fig. S19).
Therefore, these data showed clearly that, by optimizing the W/L ratio, higher resolution can be achieved with finer and more electroreceptor units.
The ball radius was fitted by the measured data of the electroreceptor matrix through the following method: the voltage of each electroreceptor was experimentally recorded, and the distance of the ball surface right on top of each electroreceptor can be then calculated by fitted curves in Fig. 2F, where the target distance-voltage relationship can be fitted with experimental data.Due to the rotational symmetry of the ball shape, the 9 distance values of 3×3 electroreceptor units can be projected and averaged into 5 points on the projected circle.
Then the ball radius can be calculated and compared with real ball size, for evaluating the measurement accuracy.The simulation results reveals that the distance-sensitivity and voltage absolute value increase dramatically with the charge density σ1 of the approaching surface, but are insensitive to the charge density of the electret σ2.These results are consistent with the experiment results in Fig. 2D-E.As for metal object, when increasing the surface charge density of elastomeric electret, the surface charge density of metal will be gradually increased due to the electrostatic induction.Therefore, the variation of the voltage output in Fig. 2F is similar to Fig. 2D.To optimize the charge density of elastomeric electret, a contact-separation model triboelectric nanogenerator (TENG) was fabricated and the elastomeric electret was served as the triboelectrification layer of the TENG.The output of the TENG was also measured by Keithley 6517.According to the working principle of the TENG (Fig. S8(a)), the output of the TENG will be comprehensively affected by the equivalent surface charge density due to the electrostatic induction effect.Therefore, a higher output of the TENG represents a larger charge density of the elastomeric electret.Therefore, a higher output of the TENG represents a larger charge density of the elastomeric electret.
From Fig. S8(b), it could be found that when control the charging temperature and the SiO2 fraction of the elastomeric electret, the output of the TENG was increased with the growth of charging electric field intensity.However, when the electric field intensity was increased to 7.5 kV/mm, the output was extremely decreased.It could be caused by the breakdown of the elastomeric electret under a high electric field intensity.Therefore, the best charging electric field intensity is 3.75 kV/mm.Fig. S8(c) shows that as the SiO2 fraction increasing the output of the TENG was increased.However, when the SiO2 fraction exceed 2 wt.%, the output was slightly decreased.
It could be caused by the aggregation of the SiO2 nano-particles.Therefore, the optimal SiO2 fraction should be 2 wt.%.
The relationship between the charging temperature and the output of the TENG is illustrated in Fig. S8(d).It could be seen that the output of the TENG was increasing when the charging temperature increased, but when the temperature exceed 100 ℃, the output was slightly decreased.A higher charging temperature would aggravate the thermal motion of the molecules.Under this condition, the negative charges will have more difficulties to be captured by SiO2 nano-particles.Hence, the optimal charging temperature should be 100 ℃     As shown in Fig. S14, the voltage output of the electroreceptor was rise to 9.1 V and 0.43 V respectively when PTFE and paper approach.If the threshold voltage was set to 2 V, the virtual robot could be triggered by PTFE, but couldn't be triggered by paper (Fig. S14 (c-d)).
However, if we set the threshold voltage to 0.4 V, the virtual robot could be triggered by both PTFE and paper (Fig. S14 (b)).Although the output of each electroreceptor array is different, the characteristic profiles (sphere) are still the comparable.The training dataset of the "ball", "cone", "ellipsoid" and "face" category under different humidity.Here, four different surface charge densities are selected to mimic the four humidity conditions (10%, 30%, 50% and 70%), as the humidity will affect the surface charge densities.

Fig. S2 .
Fig. S2.Thermal-charging process of the elastomeric electret.(a) The schematic diagram and the parameter curve of the thermal charging process.(b) The photograph of the prepared elastomeric electret film which exhibits excellent flexibility and transparency of the elastomeric electret.

Fig. S3 .
Fig. S3.SEM characterization of the electret.SEM images of (a) SiO2 nanoparticles, (b) cross-section of elastomeric electret and (c) organo-hydrogel, from which homogeneous dispersion of the SiO2 NPs in PDMS matrix and the porous structure of organo-hydrogel could be clearly observed.

Fig. S7 .
Fig. S7.The simulation results of COMSOL.(a) the relationship between the output voltage of electroreceptor and σ1 (the surface charge density of object), (b) the relationship between the output voltage of electroreceptor and σ2 (the surface charge density of elastomeric electret).

Fig. S8 .
Fig. S8.The optimization of the charge density of the elastomeric electret.(a) The working principle of the TENG.(b) The output variations of the TENGs when control the charging temperature and the SiO2 fraction of the elastomeric electret, and change the charging electric field intensity of the elastomeric electret.(c) The output variations of the TENGs when control the charging temperature and the charging electric field intensity of the elastomeric electret, and change the SiO2 fraction of the elastomeric electret.(d) The output variations of the TENGs when control the SiO2 fraction and the charging electric field intensity of the elastomeric electret, and change the charging temperature the elastomeric electret.

Fig. S11 .
Fig. S11.The output of the electroreceptor under light or darkness conditions

Fig. S15 .
Fig. S15.An example as the human-machine interface.The voltage signal that accumulated by electrometer when the human approaching the electroreceptor and trigger the robot arm.

Fig. S16 .
Fig. S16.The selection of the threshold voltage.(a) The output of the electroreceptor when materials with different surface charge density approaching.(b) the virtual robot with threshold 2 (0.4 V) was triggered by approaching paper.(c) the virtual robot with threshold 1 (2 V) failed to be triggered by approaching paper.(d) the virtual robot with threshold 1 (2 V) was triggered by approaching PTFE.

Fig
Fig. S17.The photograph of the touchless keyboard.

Fig. S30 .
Fig. S30.The training dataset of the "ball", "cone", "ellipsoid" and "face" category.Surface potential of each samples was changed from 490 to 1600 V