Humidity Stable Thermoelectric Hybrid Materials Toward a Self‐Powered Triple Sensing System

Highly sensitive and humidity‐resistive detection of the most common physical stimuli is of primary importance for practical application in real‐time monitoring. Here, a simple yet effective strategy is reported to achieve a highly humidity‐stable hybrid composite that enables simultaneous and accurate pressure and temperature sensing in a single sensor. The improved electronic performance is due to the enhanced planarity of poly (3,‐4ethylenedioxythiophene) (PEDOT) and charge transfer between PEDOT:polystyrene sulfonate (PEDOT:PSS) and multi‐walled carbon nanotubes (CNTs) by strong π–π interaction. The preferred electronic pathway induced by a robust morphology in the hybrid composite is responsible for the high humidity stability. This study also demonstrates that the sensor has tremendous potential for intelligent object identification with a high level of 97.78% accuracy. Together with the position‐detection capability of a triboelectric nanogenerator (TENG), advantages for potential industrial applications of the triple sensing system in terms of intelligent classification without seeing are foreseen.


Introduction
Integration of organic electronics enables significant advances in artificial intelligence technologies that directly interact with controls.Moreover, the most heavily explored materials are Sibased or other types of inorganic elements, which require high manufacturing costs and lack flexibility.Human skin is a natural sensor capable of perceiving subtle and local variations in temperature and pressure simultaneously and precisely.Taking inspiration from this, a single component with dual functionalities has constituted a more sought-after strategy in recent years.Despite rational choices or elaborate designs, active elements that can satisfy two requirements at the same time are still limited.More importantly, highly sensitive detection of input stimuli remains one grand challenge for the realization of bimodal sensors due to the multiple-feedback interference originating from the decoupling analysis. [13]Therefore, developing a single device that can also transduce temperature and pressure into two separate output signals is highly desirable.
In the search for a single sensor that can achieve the properties mentioned above, PEDOT:PSS, distinguished by mechanical flexibility, environmental friendliness, and cost-effectiveness, represents a promising candidate. [14]Benefiting from the intrinsic thermoelectric effect and high conductivity, such a conducting polymer endows sensors with temperature-sensing capability by generating different thermovoltages as responses to the applied temperature gradients. [8]8b] As the interference between the different response signals is almost eliminated here, the need for extra decoupling analysis is eliminated, thus facilitating faster and more effective sensing of temperature and pressure.Note that porous microstructure only provides the function to accommodate the pressureinduced volume changes, while the PEDOT:PSS polymer is responsible for resulting resistance changes as PEDOT:PSS is the only conductive element involved. [8]] One aspect neglected is humidity stability under service conditions since humidity instability would exert detrimental effects on actual implementations. [17]enerally speaking, as a typical mixed ionic-electronic conductor, PEDOT:PSS is susceptible to humidity evolution due to the high hygroscopic effect originating from PSS moieties, thus interrupting the electrical properties detected for the dual-parameter sensor. [18]Depending on the processing conditions, this feature might be seriously prominent.
In this work, we demonstrate a robust, environmentally friendly, and cost-effective fabrication of pressure-temperature dual sensors by a simple dip-coating of the highly conductive active components onto preformed microstructured polydimethylsiloxane (PDMS) sponge, which serves as a supporting frame ensuring good mechanical flexibility.The optimization of electrical property PEDOT:PSS-based thermoelectric (TE) material is enabled by facile additive engineering of DMSO doping and multi-walled carbon nanotubes (CNTs) integration.As a secondary dopant, DMSO used here allows the manipulation of the temperature-dependent electrical properties, thus favoring the decoupling of the enhancements of electrical conductivity from the drop in the electronic Seebeck coefficient.Taking the advantages of 1D network and its intrinsic high electron mobility, [19] the CNTs facilitate further improved electrical conductivity and more critically, give rise to unprecedented humidity stability through controlling the phase segregation of PEDOT:PSS and providing distinct charge carrier transport pathways.To verify such a synergistic effect, multidimensional characterizations have been conducted by grazing-incidence wide/small-angle Xray scattering (GIWAXS/GISAXS), Raman, Fourier-transform infrared (FTIR), and alternating current (AC) impedance spectroscopy.Consequently, the resulting sponge sensor demonstrates high humidity stability, sensitive temperature, and pressure detection capability with largely suppressed or circumvented cross-talk.With well-preserved linear relationships in the current-voltage characteristics, the systematic variations of linear slope and voltage offset are indicators of pressure and temperature gradient that the sensor is exposed to, respectively.Furthermore, combined with the newly emerging technology called triboelectric nanogenerator (TENG) based on the coupling effect of triboelectrification and electrostatic induction, [20] the 6-pressuretemperature sensors-based array and 2-triboelectric nanogenerator (TENG) sensors are integrated into the soft grippers with a robotic arm to simulate an intelligent identification system.Based on independent triboelectric and piezoresistive effects, the grippers can detect their location as well as identify the target objects, respectively.By applying a machine learning (ML) tool (Figure S1, Supporting Information), the recognition of objects can be achieved with a high accuracy of 97.78%, rendering our sensor promising for practical applications in artificial intelligence and robotics systems.

P_D_CNTs@PDMS Sponge
The optimum composition of the hybrid PEDOT:PSS composite in this work is 9%DMSO-PEDOT:PSS-30%CNTs.This optimum hybrid composite is denoted as "P_D_CNTs" in the following unless otherwise specified.The P_D_CNTs@PDMS sponge is constructed by coating the obtained PDMS sponge with PE-DOT:PSS/CNTs hybrids.Among all chemicals used (Figure 1a), the conjugated polymer PEDOT:PSS and inorganic CNTs together constitute the active layer, in which the former serves as thermoelectric material providing electronic thermovoltage, while the latter can not only increase the conducting paths but can also contribute to enhanced thermal stability and humidity resistance.DMSO, a commonly used polar solvent, is capable of improving the conductivity of PEDOT:PSS and leading to an absence of cross-talk in the pressure-temperature sensors.Zonyl FS-300 functions as a surfactant to enhance the dispersion of CNTs in the mixture emulsion and promote the interactions between the parent PDMS sponge and active layer.The digital photographs (Figure 1b) show that CNTs cannot be dispersed in pure DI water, in contrast, the CNTs dispersion treated with Zonyl FS-300 surfactant or acid is quite stable and no visible sediment occurs after 24 h.PDMS mainly acts as a porous framework to support the active layer and endows the whole system with good flexibility and mechanical compressibility.The synthetic procedures involved are schematically illustrated in Figure 1c.Beginning with the commercially available saccharose cube as a bio-based structure-directing agent, the porous PDMS sponge is obtained after PDMS soaking, thermal curing, and water washing steps.The porosity of the virgin PDMS sponge is estimated to be (79.7 ± 3.2)%, showing a porous and lightweight feature (Figure S2, Supporting Information).Subsequently, the incorporation of the active layer onto the resulting PDMS microstructures generates the target P_D_CNTs@PDMS sponge, as evidenced by a color change from white to dark (Figure S3, Supporting Information).Scanning electron microscopy (SEM) demonstrates that the as-obtained P_D_CNTs@PDMS sponge has a 3D highly porous structure on a microscale (Figure 1d), making it sensitive toward pressure loading.According to energy dispersive spectroscopy (EDS) mapping results, the homogeneously distributed S and F elements verify that the active layer has been successfully anchored on the PDMS sponge surface (Figure S4, Supporting Information).The P_D_CNTs@PDMS sponge deforms when pressure is applied, but recovers to its initial state after strain release (Figure 1e), suggesting good flexibility and compressibility.Typically, the resistance is defined by the linear slope of the I-V curve, while the pressure-induced deformation is positively proportional to the resistance change of the active layers (Figure 1f).Based on the thermoelectric mechanism, the applied temperature gradient (ΔT) gives rise to a voltage offset (V therm ), which is quantified by V therm = ΔT × S T , where S T denotes the Seebeck coefficient.Accordingly, it can be read out from the intercept of the I-V curve on the voltage axis (Figure 1g).By transducing external stimuli into separate electrical signals using independent thermoelectric and piezoresistive effects, simultaneous temperature and pressure monitoring are theoretically achievable in the P_D_CNTs sponge (Figure 1h).

Performance Optimization and Primary Measurement of the P_D_CNTs Composite
Having established the design concept, the P_D_CNTs@PDMS sponge is next explored to demonstrate that it is capable of discriminating the coupled thermal and pressure stimuli without any cross-talk.First of all, thick films by drop-casting are characterized.Intrinsically, a pristine PEDOT:PSS film suffers from poor conductivity as electrically conducting PEDOT-rich domains are surrounded by an excessive amount of insulating PSS-rich domains due to the electrostatic interaction, leading to a lack of intrachain and interchain interactions.DMSO additive engineering is a simple yet effective strategy to weaken such electrostatic interaction.As a consequence, it considerably improves the conductivity by two orders of magnitude (Figure S5a, Supporting Information).Simultaneously, it is also beneficial for the phase separation in the P_D film, as evidenced by the appearance of more pronounced fiber-like elongated nanostructures (Figure S6b, Supporting Information).By stark contrast, the pristine film presents to be well-entangled without discernible phase boundaries (Figure S6a, Supporting Information). [21]As indicated by further slightly decreased sheet resistance with increasing CNTs contents, the selective CNTs incorporation improves the conductivity of P_D_CNTs film further due to its highly conductive nature (Figure S7a, Supporting Information).Additionally, the favorable - interactions between CNTs and PE-DOT segments (Figure S8b, Supporting Information), and CNTsinduced negative work function shift also facilitate the charge carrier transport and thus ensure the higher electrical conductivity of the P_D_CNTs film (Figure S9, Supporting Information).When constant pressure is applied, the CNTs-loaded sponge also shows a strongly improved sensitivity (Figure S10, Supporting Information).
GIWAXS measurements are performed to gain insights into the effect of DMSO and CNTs on PEDOT:PSS on the molecular level.The exemplary 2D GIWAXS data of the P_D_CNTs film is shown in Figure 2a, that GIWAXS data of the other samples can be found in Figure S11a,b (Supporting Information).In contrast to the pristine film, the PEDOT - stacking peaks exhibit a high-q shifting and narrowing for the P_D and P_D_CNTs films (Figure 2c), which can be interpreted as enhanced planarity induced by the increased proportion of quinoid conformation (Figure 2b). [22]As the rate-limiting step, the interchain charge transfer is highly dependent on the - stacking distance. [23]herefore, the more compact packing mode identified in the P_D and P_D_CNTs films will create more effective pathways for the charge carrier transport (Figure S11c and Table S1, Supporting Information).Compared to the pristine film, a new intensity hump centered ≈0.6-0.7 Å −1 appears, which can be tentatively visualized as alternate PSS lamella stacking with PEDOT sandwiched in between, [24] further corroborating the enhanced phase separation.Upon CNTs loading, the Seebeck coefficient of the P_D_CNTs films slightly increases up to a value of ≈23-24 μV K −1 at a CNTs content of 20-40 wt.%, after which it shows a decreasing tendency (Figure S7b, Supporting Information).The initial improvement is undoubtedly caused by the inherently high Seebeck coefficient of CNTs (Figure S12, Supporting Information), while self-aggregation of CNTs might be induced by the excessive CNT presence, and thus lead to a detrimental effect (Figure S13f-j, Supporting Information).Note that the frequently used strong acids are comparable to Zonyl FS-300 in dispersing the CNTs (Figure 1b), however, the Seebeck coefficients of the resultant P_D_CNTs films deteriorate irrespective of the CNTs contents (Figure S14, Supporting Information).
By tracking the changes of broadened shapes and weakened vibration of all FTIR peaks, FTIR spectra further reveal multiple intermolecular interactions among the PEDOT:PSS, Zonyl, PDMS, and CNTs (Figure 2d,e). [25]The pristine film displays two characteristic vibrations at 1119 and 1130 cm −1 originating from the C─O─C bond stretching of the ethylenedioxy group [26] and in-plane skeleton vibration of aromatic ring, [27] respectively (Figure 2d), in which the former vanishes in the P_D and P_D_CNTs films.This phenomenon is most possibly triggered by the strong presence of - interaction between CNTs and PEDOT:PSS, which significantly alters the chemical environment surrounding PEDOT.Besides, Raman spectroscopy reflects the alterations associated with the PEDOT resonant structure.By monitoring the C  ═C  symmetric stretching vibration located at 1440 cm −1 , the peak broadening observed in P_D and P_D_CNTs films indicates the transformation of the chain conformation from benzoid to quinoid, [28] consistent with the GI-WAXS analysis. [22]Additionally, the existence of interfacial - interactions and the occurrence of charge transfer between the PEDOT:PSS and CNTs are manifested as the typical blueshift of C  ═C  asymmetric stretching vibration, from 1579 cm −1 in the P_D film to 1584 cm −1 in the P_D_CNTs film. [29]As evidenced by the emergence of a prominent D * peak at 2707 cm −1 in the Raman spectra (Figure 2f) and by the presence of the C─H alkyl chain functional group in the FTIR spectra (Figure S15c, Supporting Information), CNTs are indeed incorporated into the P_D_CNTs film.Despite this, the strongest Raman peak can be easily identified to be the C  ═C  band, indicating that it is the DMSO treatment rather than the CNTs addition that almost plays the dominant role in determining the electrical performance in the P_D_CNTs film. [30]ith the aim to evaluate the pressure-and temperaturesensing capability of the P_D_CNTs hybrid composite without interference, we measure the temperature (T)-dependent conductivity () and temperature-gradient (ΔT)-dependent Seebeck coefficient.Two different tendencies can be distinguished for  versus T (Figure 2g): the positive correlation in the pristine film is typical of a hopping-dominated semiconductor regime; the almost constant  in the P_D film and P_D_CNTs film implies metal conduction also participates in the charge transport but has a different contribution compared with hopping conduction.Further comparison ascertains the major role of DMSO in such temperature insensitivity (Figure S16, Supporting Information), in good agreement with a previous report. [31]Noticeably, the extended temperature window within which  remains almost unchanged ranges from 25 to 100 °C.In addition, all samples show nearly negligible changes in the Seebeck coefficients compared to the pure CNTs film (Figure 2h; Figure S12, Supporting Information).Additionally, there is only a slight mass loss of ≈4% below 100 °C for the P_D_CNTs composite, demonstrating good thermal stability (Figure S17, Supporting Information).As such, the temperature-independent resistance and Seebeck coefficient allow for simultaneous sensing of both pressure and temperature stimuli without any cross-talk by using current and voltage change, respectively, within a wide temperature range.
Next, a homemade experimental setup is implemented to test the P_D_CNTs@PDMS dual-parameter sensor (Figure S18, Supporting Information).Upon compression along the axial direction, the resistance of the sponge shows a negative correlation with biased pressure, as illustrated by larger slopes of the linear I-V curves (Figure 3a; Figures S19 and S20, Supporting Information).When applying different temperature gradients across the bulk of the sponge, the V therm prominently increases from 0 to 1.33 mV as ΔT elevates from 0 to 58 K, as indicated by the shifts of the voltage axis intercept (Figure 3b). Figure 3c depicts the electrical response of this sensor upon subjecting it to coupled temperature and pressure stimuli.All I-V curves still display well-defined linear relationships but with varied slopes and voltage axis intercepts.By changing the temperature at a fixed pressure value, the sensor generates different thermovoltages with identical resistance, as demonstrated by parallel straight lines with distinct voltage intercepts shown in Figure 3c.On the other hand, if the applied temperature gradient is constant, the resistance decreases with pressure, as revealed by the slope changes of the I-V curves.These observations further indicate that the P_D_CNTs@PDMS sponge sensor has likely similar transport mechanisms and microscopic morphology as the corresponding thick free-standing drop-cast film.Therefore, with the active layer coated on the deformable PDMS sponge, our sensor enables independent temperature-and pressure sensing, in which the temperature and pressure information can be simultaneously extracted from the intercept and slope of I-V curves, respectively.Figure 3d further demonstrates the output thermovoltages (V therm ) and resistance changes (defined by (R-R 0 )/R 0 ) in realtime.Initially, no pronounced variation occurs for V therm and (R-R 0 )/R 0 as no ΔT or ΔP is applied in this period (d-1).During the introduction of a temperature differential across the sponge, V therm monotonously increases up to 0.27 mV at ΔT of ≈11.6 K, but the resistance stays virtually constant (d-2).Following this, V therm slowly decreases and eventually stabilizes at 0.14 mV at ΔT of ≈6 K, accompanied by nearly unchanged (R-R 0 )/R 0 (d-3 and d-4).In the subsequent d-5 regime, a 20 kPa pressure is gradually applied while ΔT is maintained at 6 K.As a consequence, (R-R 0 )/R 0 decreases proportionally, but V therm remains almost stable.Ultimately, no noticeable changes can be observed for (R-R 0 )/R 0 (d-6).
Furthermore, we validate the dual-sensing feasibility of this P_D_CNTs@PDMS sponge in practical applications.It is first attached to the Bluetooth speaker so that its responsiveness to human voice signals can be evaluated.As displayed in Figure 3e, Chinese poetry can be captured and visualized in the form of an image.Every character has its own characteristic response signal, as demonstrated by the different heights and durations of the spikes.Moreover, pauses between sentences can be perceived and distinguished as well, making this sensor a potential candidate for sound recognition.Under an external sound force, the increased resistance can be attributed to the decreased contact area of conductive units due to the slippage of the active layers.When using English words as the sound source, we can observe similar characteristic waveforms (Figure S21a,b, Supporting Information), further corroborating the high-pressure sensitivity of the P_D_CNTs@PDMS sponge sensor.Also, this sensor can be easily used to monitor large human body motions, such as index finger bending, elbow bending, and knee joint walking movements (Figure S22, Supporting Information).As the above physiological signals are mainly detected based on the pressure modulations they give rise to, it further proves the sensitivity, reliability, and real-time transduction ability of this device in pressure sensing.Next, this sensing device is also used to track the time required for different volumes of hot water (≈95 °C) to cool down to room temperature (≈23 °C).As shown in Figure 3f, it takes at least 26, 54, and 72 min for 20, 50, and 80 mL of hot water to cool down to ambient conditions, respectively.Additionally, the demo of a user approaching the P_D_CNTs@PDMS sponge sensor through non-contact small temperature difference sensing appears in Video S1 (Supporting Information).At a temperature difference of ≈1.2 K, the P_D_CNTs@PDMS sensor shows a response time of 3.1 s during the heating, and a relaxation time of 20.3 s under natural air cooling (Figure S23, Supporting Information).As demonstrated in Figure S24 (Supporting Information), the P_D_CNTs@PDMS sponge sensor can withstand over 2500 cycles of repeated compressing and releasing cycles, suggesting the excellent repeatability and durability of the P_D_CNTs@PDMS sponge sensor.

Humidity Stability of the P_D_CNTs@PDMS Composite
High environmental stability, particularly humidity stability, is highly desired for the practical application of a pressuretemperature dual sensor in real-time monitoring.However, it is acknowledged that an elevated humidity level would promote ionic conduction in polyelectrolytes, which is especially true for PEDOT:PSS, a typical mixed ionic-electronic conductor.As a consequence, it exerts a detrimental effect on pressuretemperature sensing by interfering with the characteristic electrical signals mentioned above.In the present work, the combination of DMSO and CNTs can significantly eliminate such a humidity effect of PEDOT:PSS, which is investigated in more depth by impedance spectra.Figure 4a,b compares the Nyquist and Bode plots for pristine, P_D, and P_D_CNTs films at the lowest (10% RH) and highest humidity (81% RH) conditions, and the corresponding impedance spectra at different humidity levels can be found in Figure S25 (Supporting Information).Not surprisingly, the pristine film at low humidity behaves like a pure electronic conductor, as represented by an inductive line with a nearly constant real part of the impedance over the entire frequency range. [32]This is because the water content is insufficient for H + ions to be dissociated from PSSH groups to be mobile. [33]The Bode plots further demonstrate that the resistance almost remains constant except for the small deviations in the high-frequency range (>10 6 Hz).At high humidity, the coexisting electronic and ionic contributions can be identified in the Nyquist plot, in which the great semicircle within the lowerfrequency region arises from electronic conduction, while the depressed semicircle in the high-frequency region represents the ionic conduction. [34]The corresponding Bode plot demonstrates to be a slight decrease in the magnitude of impedance as the frequency increases, even before the break frequency, which further confirms the ionic contribution to conductivity.For the P_D film, the Nyquist plot at high humidity is typical of an R-C parallel circuit, implying that the ionic pathway becomes significant compared to that at low humidity, but the diffusion coefficient of ions is still pretty low.This finding can be rationalized by the DMSO secondary dopant, which was documented to increase the electronic conducting contribution but obscures the ionic conducting contribution for PEDOT:PSS. [35]Besides, the capacitive reactance becomes effective in total impedance above the break frequency (>10 5 Hz) as revealed in the Bode plot.No noticeable alternations can be found between low and high humidity conditions in the Nyquist and Bode plots of P_D_CNTs film, indicating the resistive element dominantly determines the impedance spectra.The primary electrical transport of P_D_CNTs film, in comparing the different processing conditions as indicated together with their model fits (lines).The inset shows three main PEDOT domain structures, denoted as large (l), medium (m), and small (s), where each was defined by a form factor with respectively correlated domain radius (r l , r m , r s ).f) Characteristic interdomain distance and domain size of small-sized, middle-sized, and large-sized structures.g) Domain size distributions extracted from the GISAXS modeling results by the superposition of the Gaussian distributions of the form factors. h) Schematic of carrier transfer such as hole (h + ) and ion (H + ) transport in PEDOT:PSS films at high humidity.The dashed red arrows illustrate the electronic contribution, in which the hole transport proceeds via the PEDOT domains.The solid red arrows represent the ionic contribution that predominantly occurs with H + cation transport through the PSS matrix.i) Generated voltage and j) resistance of the P_D_CNTs@PDMS sponge under no pressure at different humidity conditions.sharp contrast, is the electronic pathway.To track the conductivity evolution in a quantified manner, the electronic (R e ) and ionic (R i ) resistances are further extracted by equivalent circuit fits and plotted as a function of humidity level (Figure 4c).Due to the humidity-dependent swelling behavior, the R e of the pristine and P_D films increases with the humidity, which is closely related to the enlarged distances between neighboring PEDOT-rich domains caused by the water absorption in PSS shells. [36]Remarkably, the R e of P_D_CNTs film is considerably stable over the entire humidity window, and the minimum waterhealing efficiency also corroborates this (Figure S26, Supporting Information).
The most plausible explanation for such humidity stability is inner morphology changes triggered by the DMSO and CNTs addition, which are well-documented to be responsible for the hole or ionic transport pathways in mixed conductors.To confirm this hypothesis, GISAXS measurements are performed to access the phase separation structure of these films in terms of the structure sizes and spatial correlations. [37]Obviously, the incorporation of the high-electron-density CNT phase enhances lateral diffuse scattering in the case of the P_D_CNTs film (Figure 4d) compared to the pristine and P_D films (Figure S27a,b, Supporting Information).To explore the effect of CNTs on the spatial arrangements of PEDOT domains, the horizontal line cuts are performed from the 2D GISAXS data at the material-characteristic Yoneda peak of PEDOT.They are plotted in conjunction with the corresponding model fits (Figure 4e).Three cylindrical form factors (the inset in Figure 4e), are needed for data modeling, thus structural information including the PEDOT domain radii, domain size distributions, and interdomain distances is obtained. [38]ompared to the pristine film, the medium and large PEDOT domains for the P_D film increase in dimensions, accompanied by a reduction in corresponding distances (Figure 4f).The small PEDOT domains show slight shrinkage in size, while no reliable structure factors could be extracted for them, meaning that they do not have a well-defined nearest neighbor distance.Inversely, the subsequent CNTs introduction decreases the domain size but barely alters corresponding separation distances in the P_D_CNTs film.The small-sized PEDOT domains were reported to be crucial for the overall interdomain electrical conductivity within PEDOT:PSS films. [39]Figure 4g displays the PEDOT domain distribution given by the superposition of the Gaussian distributions.The highest peak intensity can be easily assigned to small-sized PEDOT domains, indicating the largest amount of small-sized PEDOT domains.Accordingly, the volume fraction is estimated (Figure S28, Supporting Information).Upon the DMSO doping, the large PEDOT domains decrease in volume while small and medium PEDOT domains increase in volume.After the CNT introduction, however, a drastic shift in the volume fraction suggests that the small domains constitute most of the P_D_CNTs film.These lead us to deduce that the disintegration of the large-and medium-sized PEDOT domains produces finer and denser distributed PEDOT domains, which are beneficial for a rapid charge flow.A schematic of the morphological changes and associated charge carrier transfer is proposed in Figure 4h.In contrast to the pristine film, DMSO treatment enhances the phase continuity of conductive-PEDOT, thus facilitating electronic transport.35a,36] As such, the higher level of densification of PE-DOT regions in the P_D_CNTs sample contributes to more beneficial electronic conduction.In addition, the presence of CNTs frustrates ion migration, endowing electronic conduction more favored over ion conduction.Furthermore, electron transfer can also occur between PEDOT and CNT as well-connected CNTs can serve as another electronic pathway, so electronic conduction is preferred rather than mixed conduction.Thus, the hybrid P_D_CNTs film is moisture-insensitive.This is also the case for the P_D_CNTs@PDMS sponge, as demonstrated by the stability of its electrical properties (Figure 4i,j; Figure S29, Supporting Information).At a given ΔT of ≈25 K, the generated Seebeck voltage by the P_D_CNTs@PDMS sponge is negligible when the humidity level is elevated from 10% to 81%.This phenomenon further evidences the dominant role of the electronic Seebeck effect, while the widely reported ionic Seebeck effect in pristine PE-DOT:PSS is suppressed herein.During the humidity ramping, a slight variation in resistance is observed.35b] Benefiting from such a humidity-independent electrical performance, our P_D_CNTs@PDMS sponge is advantageous for reliable and flexible electronics in ambient conditions.

Object Detection, Identification, and Sorting
Based on the preliminary results, we then turn our attention to the practical application of our P_D_CNTs@PDMS sensor in an intelligent identification system.Therefore, we affix the inner side of the soft grippers with six P_D_CNTs@PDMS sponge sensors for intelligent grasp and further attach the two bottom sensors with single-electrode-mode PDMS-based TENG as electroreceptors for object location detection (Figure 5a).By virtue of their high sensitivity and excellent flexibility, the P_D_CNTs@PDMS sponge sensors can be integrated into the soft grippers to collect object-characteristic contact information by monitoring the pressure-induced electrical signals.Enabled by the electrostatic induction effect from TENG, the varying electric field introduced by the external object causes the collected output signals, which is responsible for this proximity sensing capability (Figure S30, Supporting Information) that presents great practicality in the non-contact object detection system even in a dark environment.Moreover, a new data acquisition circuit is specially designed to collect multi-channel data generated by TENG (Figure S31, Supporting Information).When robotic grippers cruise through one object in the wide-open state, the varied potential from two TENGs is generated as two peaks which can be used to anchor the location of the target objects (Figure 5b; Video S2, Supporting Information).When scanning on a clean experimental table, however, the output signals remain almost unchanged due to the absence of relative potential (Video S3, Supporting Information).During the grasping process, objects with different shapes and materials lead to specific resistance signals delivered by our P_D_CNTs@PDMS sensor, thus allowing the system to identify them by ML technology.18 common objects are selected as experimental targets for the grasping process (Figure 5c), and six P_D_CNTs@PDMS sensors record the corresponding output signals (Figure 5d).To enrich the signals datasets, these objects are grasped by the robotic gripper 50 times separately, and collected output signals are divided into training and test groups with a ratio of 7:3, respectively.By random finger pressing of six electrodes, the output voltage in our system is characteristic of quasi-square wave signals (Figure S32, Supporting Information) rather than pulse signals seen in other previous works, [5e,40] which endows another feature in the ML algorithm.Note that the force applied to each sensor by the finger also reflects a more realistic scenario, as it closely resembles the unpredictable nature of human-machine interactions.Additionally, in a real grasping process, the grasping position has a big impact on output signals, namely, the output of the corresponding electrodes for the same object varies as well because of different contact positions along the grippers (Figure S33, Supporting Information).The complex signal behavior exhibited by the same object during a random grasping process underscores the necessity of implementing ML technology to facilitate more intelligent grasping.Figure 5e depicts a schematic illustration of the machine learning technology used in our work, which includes signal acquisition, dataset formation, deep learning, and object recognition.Based on all acquired signals, the best accuracy of 97.78% can be achieved (Figure 5f), demonstrating a high accuracy of object identification in our sensor system-based intelligent grasping process.Additionally, the training identification accuracy obtained can be beyond 98% after 40 times of training cycles (Figure S34, Supporting Information).By combining object identification and location detection, our sensor system shows great potential in the intelligent sorting system.As a demonstration, the first step is perceiving the potential position of objects by generated TENG signal peaks when the wide-open-state grippers cruise through the experimental table that is divided into six areas; the robotic grippers will grasp and identify objects in the marked location (Video S4, Supporting Information).Additionally, our triple sensing system shows the potential of self-powered capability via the TE and TENG generators to deliver respectable power (Figure S35, Supporting Information).

Conclusion
In summary, we propose the design concept for a simple but robust fabrication of a P_D_CNTs temperature-pressure dualparameter sensor, which relies on the uniform coating of highlyconductive thermoelectric components onto a 3D porous framework.The unique combination of thermoelectric and piezoresistive mechanisms endows our sensor with simultaneous but independent dual-parameter sensing capability without prominent cross-talk.Additionally, this work not only focuses on a critical challenge in developing high-robustness sensors against environmental humidity based on hydrophilic conducting polymer but also offers high sensitivity and a wider temperaturedetection window.In addition to real-time monitoring of decoupled temperature and pressure stimuli, our sensor is also capable of sound visualization and human motion detection.More importantly, the triple sensing capability and self-powered ability are evidenced by their integration with PDMS-based TENG for application in intelligent object identification.Our facile strategy toward low-cost, flexible and scalable, high-sensitive electrodes holds great promise for more effective human-machine interaction and perception feedback for soft robotics.

Figure 1 .
Figure 1.Fabrication process and decoupled sensing mechanism of the sensor.a) Starting materials.b) Physical appearance of CNTs dispersion with 0 and 24 h aging hours i) w/o any treatment, ii) w/ acid treatment, iii) w/ Zonyl FS-300 treatment.c) Schematic illustration of the fabrication process of the P_D_CNTs@PDMS sponge.d) SEM image of the P_D_CNTs@PDMS sponge and its corresponding elemental mapping images for C, F, O, Si, and S. e) Photographs of the P_D_CNTs@PDMS sponge state before (initial), under (I and II), and after (relaxed) finger pressing.f-h) Schematic diagram of the pressure, temperature, and coupled pressure and temperature (P-T) stimuli sensing mechanism.

Figure 2 .
Figure 2. Characterization and electrical performance of P_D_CNTs composite.a) 2D GIWAXS data of the P_D_CNTs film.b) Schematic diagram introducing the PEDOT -- stacking packing in crystalline domain.c) 1D scattering profiles obtained from radial cake cuts of the 2D GIWAXS data along the out-of-plane (filled circle) and in-plane (hollow circle) directions.The light-pink and pale-turquoise rectangular boxes highlight the scattering contribution from the Kapton window and two modes of lamella-stacked PSS, respectively.The missing data are caused by the detector gap.d) FTIR spectra of different PEDOT:PSS films and pure CNTs film.e) FTIR spectra of the pure PDMS, P_D@PDMS, and P_D_CNTs@PDMS sponges.f) Raman spectra of pristine, P_D, P_D_CNTs, and pure CNTs films.The inset depicts the enlarged image of the C  ═C  symmetric stretching vibration.g) Normalized conductivity of different PEDOT:PSS films as a function of temperature.h) Temperature-gradient dependence of Seebeck coefficient of different PEDOT:PSS films.

Figure 3 .
Figure 3. Pressure-and thermal-sensing performance of a P_D_CNTs@PDMS sponge sensing device.a-c) Measured I-V curves of a P_D_CNTs@PDMS sponge sensor under different pressures, temperature gradients, and combinations of pressure and temperature gradients.d) Real-time monitoring of output voltage and relative resistance change of a P_D_CNTs@PDMS sponge sensing device under temperature and pressure variations.e) Recognition signals of Chinese poetry.f) Generated voltage as a function of time, for hot water with different volumes.

Figure 4 .
Figure 4. Enhanced humidity stability of P_D_CNTs@PDMS sponge.a) Nyquist and b) Bode plots at minimum (10%) and maximum (81%) humidity conditions.c) The electronic and ionic resistance determined from the impedance spectra by equivalent circuit fit analysis.d) 2D GISAXS data of the P_D_CNTs film.The dashed indigo arrow represents the position of the horizontal line cut.e) Horizontal line cuts from the 2D GISAXS data (symbols)comparing the different processing conditions as indicated together with their model fits (lines).The inset shows three main PEDOT domain structures, denoted as large (l), medium (m), and small (s), where each was defined by a form factor with respectively correlated domain radius (r l , r m , r s ).f) Characteristic interdomain distance and domain size of small-sized, middle-sized, and large-sized structures.g) Domain size distributions extracted from the GISAXS modeling results by the superposition of the Gaussian distributions of the form factors. h) Schematic of carrier transfer such as hole (h + ) and ion (H + ) transport in PEDOT:PSS films at high humidity.The dashed red arrows illustrate the electronic contribution, in which the hole transport proceeds via the PEDOT domains.The solid red arrows represent the ionic contribution that predominantly occurs with H + cation transport through the PSS matrix.i) Generated voltage and j) resistance of the P_D_CNTs@PDMS sponge under no pressure at different humidity conditions.

Figure 5 .
Figure 5. Location detection and smart identification.a) The channel arrangement of integrated P_D_CNTs sensor array and TENG electroreceptors for data collection.b) Smooth-processed voltage signals from the object scan process.c) Photographs of 18 types of target objects to be grasped and recognized.d) 3D plots of output signals from P_D_CNTs sensors corresponding to different objects.e) Schematic illustration of components of the recognition system, including sensors, signal processing, and machine learning classifiers.f) Confusion matrix of the verifying results for the 18 objects with 6 P_D_CNTs sensor channels.