Adaptable thermoresponsive polymer for long-term electrical coupling in plant electrophysiology monitoring

Abstract

Electrophysiological signals provide valuable insights into plant health, facilitating measures to enhance crop productivity. Despite advances in measurement methods, long-term (>1 day) acquisition techniques remain limited, hindering continuous monitoring. Current long-term techniques rely on invasive electrodes, as noninvasive electrodes fall short in operational duration and conformability. Here, a capacitively coupled electrode with an adaptable coupling layer is developed for noninvasive, month-long electrophysiological monitoring on diverse plants. The adaptable coupling layer is formed by in situ sol-gel transition, followed by dehydration of thermoresponsive hydrogel on plants, achieving high conformability to complex surfaces and stable electrical coupling. For 1 month on trichome-covered plant surfaces, the electrode maintains a high signal-to-noise ratio comparable to a gold-standard noninvasive electrode, which typically lasts a few hours. Long-term monitoring reveals drought-specific signal features that correlate with plant water status. Physiological investigations indicate an essential role of calcium and reactive oxygen species, highlighting the potential of our electrode in generating biological insights and inspiring plant sensing innovations.

An electrode enables month-long, noninvasive electrophysiology on diverse plants, unlocking insights into plant health.

INTRODUCTION

With the pressing need to improve agricultural productivity, it is critical to monitor the health and development of plants. Through probing various signaling pathways of plants, including physical signals (electrophysiology, hydraulic waves, etc.) and chemical signals [ions, phytohormones, reactive oxygen species (ROS), etc.], we can understand how plants develop and respond to environmental changes or external stresses (1). In particular, electrophysiological (EP) signals (2, 3) offer unique advantages for real-time plant health monitoring. First, EP signals are fast responsive to stress exposure, usually within one to tens of seconds (4). The quick response of EP signals offers the potential of detecting early responses in plant signaling cascades (5). Furthermore, EP signals travel fast at speeds between a few centimeters per minute (6) to tens of centimeters per second (7) over long distances [e.g., from one leaf to several other leaves (8) and from root to shoot (9)] to trigger systemic responses. This enables effective whole plant tracking by attaching a small number of sensors sparsely. EP signals can be measured noninvasively from the plant surface. The noninvasive form factor reduces contamination or damage to plants, which is critical for practical application. In addition, EP signals experience changes under a wide variety of stress conditions, including both biotic [e.g., bacteria (10) and pests (11)] and abiotic stresses [e.g., drought, salt (12), and nutrient deficiency (13)]. This allows potential detection of various stressors using a single sensor modality. However, an associated problem is the poor discrimination of stressors because EP signals are a combination of electrical activities within plants (6) and they can have similar response patterns as common downstream outcomes under different stressors. Nevertheless, recent research has elucidated stimulus-specific quantitative difference in the temporal dynamics of EP signals (14), and advanced data analysis and machine learning methods have shown promise in accurate classification of stress conditions (12, 13). Combined with environmental data and advanced data analytics (6), EP signals hold great promise in in situ plant health assessment and diagnosis. For real-time assessment of plant health, long-term continuous monitoring of plant EP signals is required.

Conventional electrodes used for long-term monitoring of EP signals are invasive metal electrodes that are directly inserted into plant stems (4, 12). These invasive electrodes result in permanent damage to plants. To avoid plant wounding and minimize associated signal interference, minimally invasive microneedles and noninvasive electrodes, such as ionically conductive hydrogel electrodes or electronically conductive dry electrodes, are being increasingly used (15, 16). Although microneedle electrodes are capable of long-term EP monitoring for up to 7 days, their application is limited to stems, petioles (12), or thick leaves (17) to ensure reliable mechanical anchoring and for plants to survive the microdamages. However, monitoring on soft and fragile leaves is important, first, because leaves display photoelectric behaviors that are closely linked to photosynthesis (18, 19) and, second, because model plants used by biologists (e.g., Arabidopsis thaliana and Nicotiana benthamiana) and common agricultural crops (e.g., tomato and lettuce) do not have thick leaves. Therefore, microneedle electrodes have limited capability to reveal the rich EP dynamics on diverse plants. Wet electrodes (20, 21), with good conformability on plant surfaces, are used as the gold standard for short-term signal measurement (22). However, their working duration is limited by dehydration to within a day. Rehydration of wet electrodes to restore signal measurements often results in unrepeatable and unstable signals, impairing continuous monitoring (23). On the other hand, dry electrodes (24) are typically restricted to flat and smooth plant surfaces due to poor conformability on hairy or rough surfaces. Noticeable improvement in the conformability of dry electrodes has been made through various processing methods, such as direct vapor printing (25) or hydroprinting (26, 27). However, these methods involve exposing the whole plant or part of the plant to a specific gas or liquid environment, making them a less scalable approach. It also remains unclear whether these treatments affect the long-term EP signals of plants. Moreover, hydroprinting is restricted in the location that electrodes can be applied to; petioles and stems on plants with dense leaves are challenging to attach electrodes. Therefore, we aim to develop an electrode that can conform to complex plant surfaces while maintaining long-term stability in measuring plant EP signals, through a simple and versatile application process.

To achieve this goal, we draw inspiration from the capacitive measurement of biopotentials on the human skin. In human-wearable applications, using a high–input-impedance amplifier, electrodes can capacitively couple with the skin through a dielectric material (e.g., ceramics, fabrics, and polymers) for accurate and long-term biopotential measurements (28). Similar to human tissues, plant tissues are rich in electrolytes and have high water content (29, 30). Plants also have a protective and electrically insulating cuticle on their epidermis, much similar to the human stratum corneum, whose high electrical impedance is undesirable for biopotential signal acquisition (31). On the basis of the similar structural and electrical properties of human skin and plant tissues, we hypothesize that through a thin and conformal dielectric layer on the cuticle, conductors can capacitively couple with the plant for long-term stable EP measurement (fig. S1). Endowing the dielectric layer with adaptable properties through stimulus-responsive materials may help realize high conformability through a simple application process.

Here, we present a capacitively coupled electrode that enables high-quality long-term monitoring of plant EP signals. The electrode consists of a flexible conductor with an adaptable coupling layer, formed through the dehydration of a thermoresponsive hydrogel applied to the plant surface. Thermoresponsive hydrogel is used for its sol-gel transition and amphiphilicity to achieve a conformable interface on morphologically and chemically complex plant surfaces (e.g., rough, hairy, and superhydrophobic). The dehydrated form retains the high conformability where it adapts to the morphology of applied surfaces, allowing effective and stable capacitive coupling for long-term signal acquisition. The use of both wet and dehydrated forms of thermoresponsive hydrogel during electrode assembly and function effectively combines the benefits of conventional wet and dry electrodes. The capacitively coupled electrode is not to be confused with capacitance sensors that measure leaf capacitance in a parallel plate configuration (32, 33). Our electrode adopts the concept of capacitive coupling at the electrode-plant interface for biopotential measurement, whereas capacitance sensors focus on the dielectric properties of leaves. Using the capacitively coupled electrode, stable EP signals with high signal-to-noise ratios (SNRs), comparable to the gold standard of short-term noninvasive electrodes, were acquired over a month. Compared to existing dielectric materials, thermoresponsive polymer offers a unique combination of desirable properties for a capacitive coupling layer, including plant-safe yet robust electrode assembly, high signal stability and fidelity, and versatile and facile application. In addition, we unraveled the detailed mechanism of capacitive coupling on plants and correlated electrode impedance with SNR, laying the foundation for on-plant electrical sensor development. The utility of the capacitively coupled electrode was demonstrated through a 28-day continuous monitoring under changing light cycles, a 14-day drought stress and recovery simulation, and a 9-day salt stress simulation. We observed previously unreported EP dynamics and found the correlation between long-term EP signals and drought stress conditions. Physiological investigations indicate the essential role of Ca²⁺ and ROS in modulating such EP responses.

RESULTS

On-plant assembly of capacitively coupled electrode

The design of the capacitively coupled electrode includes a thin and adaptable coupling layer between the conductor and plant surface, formed from the dehydration of a thermoresponsive hydrogel (Fig. 1A). As shown in the equivalent circuit model, the capacitive coupling involves both the coupling layer and the plant epidermis (Fig. 1A, top). A simple way to understand this model is by viewing the conductor and the plant internal tissue as the conductive ends of a capacitor, with the coupling layer and plant epidermis as the capacitive layer between the two ends. For comparison, signal transduction through conventional wet and dry electrodes is illustrated in Fig. 1B. Wet electrodes conduct electricity mainly via ionic movement in the gel, and their impedance is primarily affected by gel resistance and the interface with plants (34). On the other hand, signal transduction by dry electrodes is directly through contact with the plant epidermis, where resistance and capacitance of the contact interface are affected by electrode conformability and the presence of air gaps (34).

Fig. 1. On-plant assembly of capacitively coupled electrode for long-term plant electrophysiology monitoring.

(A) Schematics and equivalent circuit model for capacitively coupled electrode (C_coupling and R_coupling: capacitance and resistance of adaptable coupling layer; C_epi and R_epi: capacitance and resistance of plant epidermis; R_plant: resistance of internal plant). Electrode assembly process is shown at the bottom: Thermoresponsive hydrogel is applied on plants in solution state (stored at low temperature) and gelates within 40 s (upon equilibrating at room temperature). The capacitively coupled electrode is fully assembled when the thermoresponsive hydrogel dehydrates over 1 day to form the adaptable coupling layer. (B) Schematics and equivalent circuit models for conventional wet and dry electrodes showing their limitations in long-term monitoring (C_i and R_i: double layer capacitance and charge transfer resistance at conductor/gel interface; R_gel: resistance of gel; C_c and R_c: contact capacitance and resistance at conductor/plant interface). (C) Photographs showing the conformal application of capacitively coupled electrodes on hairy leaf (top) and vertical stem (bottom) of a tobacco plant (N. benthamiana). (D) Cross-sectional scanning electron microscopy (SEM) image of a piece of standalone adaptable coupling layer peeled off from a hairy tobacco leaf. The replicated hairy morphology demonstrates high conformability of the coupling layer. (E) Comparable SNR between the capacitively coupled electrode and gold-standard agar gel electrode over 31 days (agar gel was freshly applied before every measurement). Mean ± SD; N = 6 plants.

In this work, the thermoresponsive polymer used is made of an amphiphilic copolymer of poly(ethylene glycol) (PEG), poly(propylene glycol) (PPG), and polycaprolactone (PCL), known as EPC. EPC polymer was selected as the model material, as the sol-gel transition properties and adhesion strength of EPC hydrogel were previously optimized for plant application (21). The plants used were N. benthamiana (tobacco) (unless otherwise stated), which have soft leaves covered by dense trichomes, to demonstrate the high conformability of our electrodes. EPC hydrogel exhibits sol-gel transition when the temperature rises (fig. S2) (35). For the application on plants, EPC aqueous solution stored at ~4°C was applied onto plants through a dropper and allowed for in situ gelation on plants (Fig. 1A, bottom) (21). The application of cold solution was found to be nondamaging to plants (fig. S3 and movie S1), agreeing with a previous report (8). Flexible conductors were then gently placed on the hydrogel to complete the electrode assembly. Carbon nanotube and bacterial cellulose (CNT-BC) composite films with a sheet resistance of (501 ± 135) × 10⁻³ ohms square⁻¹ and a thickness of 22 ± 6.3 μm (fig. S4) (detailed fabrication method in Materials and Methods) were used as conductors for their lightweight (36) and chemical stability (37) properties (importance of being lightweight in note S1). Last, the coupling layer between the conductor and plant forms from the natural dehydration of the thermoresponsive hydrogel in the environment of plant growth for a day (Fig. 1A, bottom).

The use of thermoresponsive hydrogel during application allows the capacitively coupled electrode to be assembled on complex surfaces such as hairy, superhydrophobic leaves and vertical stems (Fig. 1C, fig. S5, and movie S2). The coupling layer adapts to the plant topography (Fig. 1D) and allows high electrode conformability (fig. S6). The thin coupling layer, with a mean thickness of 51.9 μm (fig. S7A), provides a short coupling distance for effective signal acquisition. In addition, the thermoresponsive polymer penetrates the porous conductor film, forming a tightly bonded composite material that enhances electrode integrity (fig. S7B). Furthermore, to examine the long-term biocompatibility of capacitively coupled electrodes, two metrics, open stomata percentage and chlorophyll content, were used to assess the effects of electrode application on plant transpiration and photosynthesis (38, 39). Both metrics remain comparable over 1 month between leaves with and without the electrodes (fig. S8), implying negligible impact on plant physiological activities by long-term attachment of the capacitively coupled electrodes.

Superior signal acquisition compared to conventional electrodes

To test the ability of the capacitively coupled electrode to acquire plant signals, we compared SNR under flame stimulus against agar gel electrodes. As flame stimulus produces strong and repeatable EP signals (40), it was used for SNR calculation to compare performance between electrodes in this study (detailed procedure in Materials and Methods, Comparison of electrode performance on plants). Currently, Ag/AgCl electrodes with a small amount of salt-doped agar gel for contact with plant surface stand as the gold standard for noninvasive short-term measurement (22). For comparison, the capacitively coupled electrodes were tested over 31 days with agar gel electrodes freshly applied before each signal measurement (fig. S9, A and B). All capacitively coupled and agar gel electrodes applied on six plants record flame-induced variation potentials (3, 41), with highly similar signal waveforms and amplitudes (fig. S9C), and the capacitively coupled electrodes acquire a comparable SNR to freshly applied agar gel electrodes over a month (Fig. 1E and fig. S9D). For longer-period measurement, high-quality signals can be recorded on a peperomia plant after 50 days of electrode application (fig. S10B). These results indicate the high signal quality and long-term stability of the capacitively coupled electrode. The universality of the electrode for application on diverse plant surfaces is also shown by the high-quality signals acquired after a month of application on eight plant species (fig. S11).

Capacitively coupled electrodes display superiority in noninvasiveness, conformability, signal quality, applicable sites on plants, and long-term monitoring, as compared to conventional electrodes (Fig. 2A and rating criteria of radar plot in table S1). To quantify such advantageous performance, we first benchmarked our capacitively coupled electrodes against noninvasive wet electrodes: freshly applied agar gel and EPC hydrogel electrodes. Solid-state hydrogels were not included in this comparison, as their poor conformability on hairy surfaces has shown to impair signal quality (21). Wet electrodes, with high water content, undergo dehydration over time, which leads to a large EP signal drift. As seen from Fig. 2 (B and C), when attached to plant leaves, capacitively coupled electrodes display the smallest drift amplitude within 7 hours (33 ± 20 mV for capacitively coupled electrodes, 93 ± 5.3 mV for EPC hydrogel electrodes, and 156 ± 13 mV for agar gel electrodes). The reduced drift of capacitively coupled electrodes also explains the higher SNR observed in Fig. 1E (detailed explanation in fig. S12). In addition, the positive drift of capacitively coupled electrodes, unlike the negative drifts recorded from agar gel and EPC hydrogel electrodes (Fig. 2B), might come from true plant EP activity instead of electrode instability, which is inconclusive from the on-plant drift characterization. The potential change rate was further determined from the first-order derivative of the potential readings. As acute stress–induced signals last up to around 2 hours (8, 21, 42), it is important to maintain drift free within at least a 2-hour window. In fig. S13, it is challenging to identify a 2-hour drift-free period from wet electrodes. Only the capacitively coupled electrodes, with the smallest potential change rate, bring about a promising solution for recording true plant signals. In addition, although wet electrodes with antidehydrating properties have been developed, the techniques used, e.g., adding high salt content or strong humectants, may cause harm to interfacing plants. To examine this, we attached antidehydrating hydrogels doped with 0.188 M CaCl₂ and solvent exchanged with glycerol, respectively, to a choy sum leaf. Severe leaf yellowing was observed within 3 days in both cases (fig. S14). Therefore, antidehydrating hydrogels pose safety concerns to plant health and are not suitable for long-term application on plants.

Fig. 2. Superior signal acquisition performance of capacitively coupled electrode compared to conventional electrodes.

(A) Radar plot for comparison of electrode properties among wet, dry, invasive, and capacitively coupled electrodes, with rating criteria shown in table S1. (B) Plant leaf potential measured by capacitively coupled electrode showing the lowest drift as compared to wet electrodes (EPC hydrogel and agar gel electrodes) over 7 hours. Shadows stand for SD. (C) Quantified amplitude (mean ± SD) of potential drift over 7 hours. (D) Optical microscopic image showing poor conformability of dry electrode (CNT conductor strip) on hairy tobacco leaf. (E) Potential readings of on-leaf electrodes during mechanical disturbance, showing reduced motion artifact by capacitively coupled electrode. Dry electrode was adhered on leaf using tape. (F) Long-term on-plant measurement shows noisier signals from dry electrode than capacitively coupled electrode. SNRs (mean ± SD) were calculated from responses to the light-on event over 6 days. Gray background indicates night, and white background indicates day. (G) Long-term on-plant signals recorded using invasive electrode in stem and capacitively coupled electrode on leaf. The zoom-in graph shows higher intensities of light-triggered EP signals measured by on-leaf electrode. Reference electrodes used are the same as recording electrodes, fixed on/in stem. (H) Photograph showing extensive tissue damage on leaf caused by invasive electrode.

Dry electrodes, with long-term monitoring capabilities, are restricted to use on flat surfaces due to poor conformability on plants with complex topography (Fig. 2D). We used CNT conductive strips fixed on hairy tobacco leaves through adhesive tapes as the dry electrode for comparison with the capacitively coupled electrode. Under mechanical disturbance, such as touch and wind, the capacitively coupled electrodes maintained a stable signal while dry electrodes experienced larger motion artifacts (Fig. 2E). During the long-term on-plant monitoring, dry electrodes recorded noisy signals and prevented accurate analysis (Fig. 2F and fig. S15). The large signal fluctuations likely come from leaf movements induced by atmospheric disturbances and/or as a natural behavior under changing light (43, 44). In contrast, the highly conformable and adhesive capacitively coupled electrode recorded stable and clear signals. By comparing the light-triggered transient potential changes across 6 days, the capacitively coupled electrode recorded a higher SNR of 45.9 ± 0.5 dB, as compared to 33.7 ± 7.1 dB by the dry electrode (more replicates with SNR calculation in fig. S15).

Last, compared to minimally invasive microneedles and invasive metal electrodes, our capacitively coupled electrodes display higher biocompatibility and signal stability and enable the measurement of signals that cannot be captured invasively. Despite being minimally invasive, microneedles still cause irreversible damage to soft leaves, including yellowing and wrinkling (fig. S16). This is in stark contrast with the negligible influence on plant physiology by the capacitively coupled electrode (fig. S8). On the other hand, invasive Ag/AgCl electrodes result in unstable signal acquisition on tobacco petioles, presumably due to mechanical and chemical instabilities (fig. S17). Conventionally, to ensure sufficient mechanical stability, invasive electrodes are usually inserted into sturdy tissues such as stems (fig. S18A). However, we observed that such a testing configuration failed to capture light-triggered transient potential changes effectively (Fig. 2G). In contrast, our capacitively coupled electrodes when attached to leaves could capture such signals in high intensity. This difference is likely because photoresponsive components are mostly present in leaves and light-triggered potential changes are local signals (45). The ability to monitor light-triggered EP variations opens the possibility of studying light-related physiological processes (46, 47), which are instrumental in plant development. In addition, the long-term application of invasive electrodes on the soft leaves of small plants is deemed unsuitable due to extensive tissue damage (Fig. 2H), whereas noninvasive capacitively coupled electrodes provide a stable attachment on leaves (fig. S18B).

In summary, capacitively coupled electrodes extend the working duration of wet electrodes without inducing adverse physiological effects. They offer enhanced mechanical stability and conformability over dry electrodes and can be applied on more diverse organs than invasive electrodes, potentially generating previously unidentified biological insights. None of the existing technologies can achieve all the desirable properties simultaneously, highlighting the unprecedented sensing capability of the capacitively coupled electrode.

Capacitive coupling as the signal acquisition mechanism

To understand the signal acquisition mechanism of capacitively coupled electrodes on plants and to examine the long-term electrical stability, we used electrochemical impedance spectroscopy (EIS) to measure electrode impedance on plants over a month (Fig. 3A, on-leaf). The electrodes (0.09 cm²) were placed 1 cm apart from each other on the leaf surface. In addition, we tested a laminate configuration, where two conductors (0.3 cm²) sandwiched a thermoresponsive hydrogel or the corresponding coupling layer formed after dehydration (Fig. 3A, laminate). This is to exclude effects from plants and observe changes solely in the thermoresponsive hydrogel or the coupling layer. Impedance measurements are compared at 0.1 Hz, as EP signals have low signaling frequencies (fig. S19). Furthermore, to visualize how dehydration affects electrical conduction, we started the test immediately after the application of the thermoresponsive hydrogel, covering the whole wet-to-dry transition process. Three impedance stages are revealed in this month-long test: bulk dehydration, resistive to capacitive transition, and aging of the coupling layer/plant interface (Fig. 3, A and B).

Fig. 3. Capacitive coupling with adaptable coupling layer on plants.

(A) Impedance of on-leaf (N = 8 plants) and laminate (N = 5) setup, Z_on-leaf and Z_laminate, respectively, at 0.1 Hz over 31 days. t = 0 when the thermoresponsive hydrogel solution was first applied on plants. (B) Schematics of three transformation stages of the electrode on plants over 31 days: bulk dehydration during electrode assembly, resistive to capacitive transition (when capacitively coupled electrode is fully assembled), and aging of coupling layer/plant interface accounting for long-term changes. (C) Large decrease in water content and thickness of hydrogel in the first 4 hours, indicating bulk dehydration. (D) Nyquist plot of Z_laminate for a representative sample during the first 4 hours, indicating a decrease in intrinsic gel impedance during bulk dehydration. (E) Nyquist plot of Z_laminate for a representative sample, indicating change from resistive to capacitive conduction. h, hours; d, days.(F) Nyquist plot of day-7 Z_on-leaf for a representative plant with equivalent circuit model, confirming capacitive coupling of electrode with plant (CPE_epi and R_epi: constant phase element and resistance of plant epidermis; CPE_coupling and R_coupling: constant phase element and resistance of the adaptable coupling layer). CPE was used to indicate imperfect capacitance. (G) Long-term Z_on-leaf with various combinations of coupling layer degradability and plant leaf condition (N = 3), showing that both degradation of the coupling layer and plant leaf wilting contribute to Z_on-leaf increase. (H) SNR of plant EP signal with the corresponding electrode-plant system impedance over a month, showing that high SNR (>45 dB) was maintained despite the decreasing SNR with increasing impedance (symbol shapes represent individual plants).

In the initial stage, impedance at 0.1 Hz of both on-leaf and laminate set-ups, Z_on-leaf and Z_laminate, respectively, experiences a drop (Fig. 3A) due to dehydration of the hydrogel, characterized by decreases in water content and thickness (Fig. 3C and fig. S20). The Nyquist plot of Z_laminate reveals a decrease in the bulk resistance of the hydrogel in the first 4 hours, indicated by the decreasing x-intercept value (Fig. 3D and fig. S21, A and B). This is due to the reduction in gel thickness and the concurrent increase in ionic conductivity (fig. S21C), driven by the rise in ionic concentration with water loss. The longer dehydration duration of the on-leaf setup (1 day; read from Fig. 3A) as compared to the laminate setup (4 hours) might stem from the humid microclimate around plants and/or the presence of plant secretion. The main signal transduction mechanism in the hydrogel layer before dehydration is via ionic resistive charge migration. During this dehydration stage, the hydrogel is a wet electrode, which is nonideal for recording EP signals due to the large signal drift (Fig. 2, B and C). For stable EP signal recording, the electrode was used after the bulk dehydration stage (1 day).

As the thermoresponsive hydrogel undergoes further dehydration to form the coupling layer, an increase in Z_on-leaf and Z_laminate is observed (Fig. 3A). After 4 hours, the Nyquist plot of Z_laminate gradually forms a semicircle (Fig. 3E and fig. S22A), with a phase shift observed in the Bode plot (fig. S22B). The strong presence of the semicircle at high frequencies on day 7 reveals the intrinsic capacitive nature of the coupling layer. Furthermore, the Z_on-leaf observed phase shifts toward more negative values at low frequencies below 10 Hz from day 1 to day 7 (fig. S22C), indicating that the capacitive contribution is becoming more pronounced. To confirm the presence of capacitive coupling between the electrode and the plant, we conducted circuit modeling for Z_on-leaf measured on day 7. The equivalent circuit models a series of two circuits of a resistor (R) in parallel with a constant phase element (CPE), with the two R-CPE pairs corresponding to the plant epidermis and the coupling layer, respectively (Fig. 3F and fig. S23). In this circuit model, the impedance of plant internal tissues is considered to be negligible as compared to other components (48). A low chi-square (𝛘²) value is achieved across all samples, indicating a good fit (table S2). CPEs, representing nonideal behavior of double-layer capacitance, are used for fitting to account for any dispersity of current (46, 47). The impedance of CPE is expressed as Z(CPE) = 1/[Q(jω)ⁿ], with Q representing the capacitance value and n reflecting the current dispersion. When n = 1, CPE is equivalent to a perfect capacitor. The presence of CPE_epi and CPE_coupling in the circuit model reflects the capacitive behavior of the epidermis and adaptable coupling layer, which facilitates capacitive coupling between the electrode and plant tissues during signal acquisition. The adaptable coupling layer achieved an average equivalent coupling resistance and capacitance of 4.58 × 10⁶ ohms || 15.9 nF (details in table S2) (49). Compared to a cotton-based capacitive electrode used for on-skin measurement (200 × 10⁶ ohms || 20 pF) (50), the lower resistance and higher capacitance of our electrode allow minimal signal attenuation at low frequencies. However, it should be emphasized that there is mixed coupling where both resistive and capacitive elements play a role, which is commonly seen in capacitive electrodes (50, 51).

After prolonged on-leaf attachment (day 7 onward), an increase in Z_on-leaf is observed (Fig. 3A). At 0.1 Hz, the increase in Z_on-leaf should arise from changes at the coupling layer/plant interface, as Z_laminate remains constant after 7 days (Fig. 3A). To find out the origin of such changes at the interface, we first note that Z_laminate continues to record impedance changes at frequencies above 1 Hz (fig. S24), which points to the possibility of intrinsic property change of the coupling layer, affecting the coupling layer/plant interface. Degradation of the coupling layer was therefore investigated, considering the presence of biodegradable PCL components in the thermoresponsive polymer used. For comparison, a nondegradable coupling layer (formed from PEG-PPG thermoresponsive polymer) was used. Moreover, visual observation during Z_on-leaf measurements suggested a possible contribution from tobacco leaf wilting (fig. S25A)_. To investigate the effects of plant wilting, we used long-living Peperomia obtusifolia (peperomia) plants as a nonwilting control plant (fig. S25B). Direct impedance measurements of both tobacco and peperomia leaves show a negligible change in intrinsic tissue impedance over 4 weeks, even when tobacco leaves experienced severe wilting (fig. S26). This excludes the contribution of internal leaf impedance to the observed increase in Z_on-leaf. Hence, plant wilting was suspected to have impaired the conformal contact between the plant surface and the coupling layer. As shown in Fig. 3G and fig. S27, only nondegradable electrodes applied on nonwilting plants do not display an increasing impedance after 14 days. It can therefore be concluded that the plant interface is affected by both the coupling layer degradation and leaf wilting, likely due to the deterioration of the conformal interface.

Capacitively coupled electrodes on tobacco plants maintained a high SNR over 1 month (Fig. 1E) despite the increase in impedance. For our setup with an amplifier input impedance of 500 × 10⁶ ohms, SNR is impaired when Z_on-leaf exceeds 1 × 10⁹ ohms (fig. S28). Correlation studies of impedance and SNR were therefore conducted to elucidate the requirements in impedance for high-SNR signal acquisition (details in note S2). Impedance was measured with the same setup as for SNR measurement, between an electrode on a leaf and a Ag/AgCl reference electrode in the soil. The SNR decreases as impedance increases (Fig. 3H), which aligns with a previous report on neural electrodes (52). Nevertheless, the increasing impedance over time only causes slight SNR drops within experimental variations (note S2). Overall, our capacitively coupled electrode’s impedance is sufficiently low for all the analyses conducted in this work and can acquire signals with a high SNR above 45 dB for a month (Fig. 3H).

Advantages of thermoresponsive polymer for coupling layer formation

Because capacitive coupling is a general principle for surface electrodes, any dielectric material may serve as a coupling layer on plants. To explore the unique advantages of thermoresponsive polymer, we compared the physicochemical properties and sensing efficacy of alternative coupling materials, including acrylic adhesives, silicone elastomers, and aqueous polymer solutions, to those of the thermoresponsive polymer. Overall, thermoresponsive polymer achieves the most robust yet plant-safe electrode assembly, the most versatile electrode application, and the highest signal quality for long-term monitoring as compared to other coupling materials.

Acrylic adhesive tapes are commonly used as substrates and dielectrics in stretchable devices (53) and for robust adhesion on the human skin (54). We tested acrylic tapes of two thickness: 0.4 mm (thick acrylic) and 0.1 mm (thin acrylic). Similar to dry electrodes based on CNT conductor strips (Fig. 2D), these tapes showed poor conformability on hairy plant surfaces (fig. S29). Although compression improves conformability, it risks mechanical damage to the cuticle or epidermal cells of soft leaves. Moreover, even with pressure applied during tape attachment, the electrodes failed to capture signals with high fidelity (Fig. 4A and fig. S30), likely due to their limited conformability and large thickness. In addition, the acrylic tapes caused mechanical damage to the applied plants upon removal (Fig. 4B), preventing further usage of the affected leaves. In contrast, the thermoresponsive polymer allowed intact removal without causing any visible damage to the leaves (Fig. 4C and movie S3).

Fig. 4. Advantages of thermoresponsive polymer over other capacitive coupling materials for on-plant electrophysiology.

(A) Potential readings from the following electrodes: CNT conductor films with thick acrylic tape (0.4 mm), thin acrylic tape (0.1 mm), and thermoresponsive polymer, respectively, benchmarked against agar gel–Ag/AgCl electrodes, showing poor signal fidelity of acrylic tapes. Electrodes had shuffled placement on leaves to minimize location variability, and signals were generated by flame wounding a neighboring leaf. (B) Photograph showing mechanical damage upon peeling of acrylic tape from a tobacco leaf. (C) Photograph showing clean removal of thermoresponsive polymer layer from a tobacco leaf. (D) Photographs of leaves infiltrated with polydimethylsiloxane (PDMS), taken after the removal of surface PDMS. (E) Fourier transform infrared (FTIR) spectra of pristine leaves and leaves treated with PDMS (surface PDMS removed), confirming infiltration of PDMS components by the -Si-CH₃ peaks. (F) Potential readings from CNT conductor films coupled with Ecoflex and thermoresponsive polymer, respectively, benchmarked against agar gel–Ag/AgCl electrodes, showing poor signal stability of Ecoflex coupling layer. (G) Poor permeability of Ecoflex compared to thermoresponsive polymer. (H) Photographs showing the application process of thermoresponsive hydrogel and PEG solution on vertical tobacco stems, highlighting the robust electrode assembly by thermoresponsive hydrogel. (I) Higher shear adhesive strength of thermoresponsive hydrogel compared to agar gel, PEG, and poly(vinyl alcohol) (PVA) solutions. (J) Higher adhesive strength of thermoresponsive polymer compared to agar, both in dry state.

Silicone elastomers are another common type of substrates and dielectrics in bioelectronics. Polydimethylsiloxane (PDMS) and Ecoflex were selected as representatives in this study. Preformed silicone thin films will have poor conformability on hairy plant surfaces similar to dry electrodes. Therefore, on-plant in situ curing was conducted, mimicking the sol-gel transition of thermoresponsive polymer. PDMS precursor takes ~1 day to be completely cured at room temperature. Consequently, during application, PDMS precursor flowed uncontrollably on nonflat surfaces, making precise and localized application challenging (fig. S31). Moreover, PDMS precursor infiltrated into the leaf tissues before complete curing, which may negatively impact plant physiology. This infiltration is shown by the transmittance change of plant leaves (Fig. 4D) and the presence of -Si-CH₃ peaks in the Fourier transform infrared (FTIR) spectroscopy spectra of leaves (Fig. 4E) after removal of cured PDMS from the surface. In contrast, the thermoresponsive polymer did not infiltrate into plants, presumably due to its high molecular weight (M_n ~ 70 kDa) (21). On the other hand, Ecoflex is able to cure within 5 min, enabling in situ formation of coupling layers in a way similar to the thermoresponsive polymer. However, the high viscosity of its precursor and fast curing prevented the formation of a thin layer. This may have contributed to the high occurrence of electrode failure (fig. S32B) and poor signal stability and fidelity (Fig. 4F and fig. S32C) during a long-term test. Furthermore, Ecoflex has a much lower water vapor permeability (125 g m⁻² day⁻¹) than thermoresponsive polymer (488 g m⁻² day⁻¹) (Fig. 4G), which may affect plant health in the long-term by hindering water vapor exchange.

Last, aqueous polymer solutions were explored for in situ formation of capacitive coupling layers because thermoresponsive polymer was applied on plants as an aqueous solution. Poly(vinyl alcohol) (PVA) and PEG were selected as common water-soluble polymers. Agar gel, the gold standard for wet electrodes, was also explored with its dehydrated form as a coupling layer. In terms of electrode attachment, the sol-gel transition of thermoresponsive hydrogel allows a more robust electrode assembly on nonhorizontal surfaces, including vertical stems (Fig. 4H, fig. S33, and movies S2 and S4 to S6). This is quantified by the higher adhesive strength of thermoresponsive hydrogel as compared to other polymer solutions (Fig. 4I). Furthermore, although agar gel is able to maintain attachment on vertical surfaces (movie S4), the corresponding coupling layer after dehydration adheres much weaker than thermoresponsive polymer (0.978-kPa versus 32.1-kPa adhesive strength) (Fig. 4J), preventing robust electrode attachment. The strong attachment of electrodes using thermoresponsive polymer is further demonstrated in movie S7. In addition, agar gel did not allow a conformal attachment on superhydrophobic plant surfaces (fig. S5). To test signal quality, we applied electrodes assembled from the polymer solutions on horizontal leaves with extra care to prevent detachment (fig. S34, A and B). All polymer-coupled electrodes show the ability to measure signals over a month, with comparable SNR to the gold standard (fig. S34C). Nevertheless, the electrodes with thermoresponsive polymer as the coupling layer show the best stability and reproducibility among all electrodes (figs. S34D and S35).

Confirming the advantages of the thermoresponsive polymer, we then explored its universality to couple with different conductors. In addition, the need for a salt dopant in thermoresponsive polymer was also explored, as the capacitively coupled electrode does not rely on ionic conductivity for signal transduction. The SNR of three conductors (Au, CNT, and Ag/AgCl), bridged by an adaptable coupling layer with and without salt dopant, was compared. All conductor-dopant combinations maintained similar and high SNR over a month (figs. S36 and S37). These results show that the type of conductors that can be effectively bridged by thermoresponsive polymer is versatile, and ionic dopants are not compulsory for effective capacitive coupling, eliminating the risk of chronic salt stress.

Month-long plant electrophysiology study using capacitively coupled electrodes

We used capacitively coupled electrodes for long-term continuous monitoring of plant EP signals under controlled and varying environmental conditions in an attempt to uncover previously unidentified biological insights. First, to showcase the long-term reliability of capacitively coupled electrodes, we conducted 28-day continuous monitoring on four plants and observed EP patterns in response to daily light/dark intervals. The testing chamber was equipped with white light-emitting diodes (LEDs) to artificially introduce day and night periods, and the light/dark cycle was switched from 16 hours/8 hours to 8 hours/16 hours and then back to 16 hours/8 hours (fig. S38A). As shown in fig. S38 (B to D), there are observable changes in the plant EP circadian cycles when the light/dark cycle is changed. The constant potential readings from control electrodes (dark gray lines; electrode design in fig. S38E) prove that the measured signals are true plant EP signals rather than electrode artifacts induced by light/dark transitions. In addition, the signals acquired are stable with minimal drift or noise. Notably, the long-term attachment of the capacitively coupled electrodes eliminates confounding factors associated with repeated reattachment of electrodes, such as electrode displacement and changes in leaf surface properties, which can hinder reliable signal analysis over long term.

Because changing light/dark cycles may present stress to plants through disrupting their internal circadian clocks (46), we further analyzed the above 28-day signals aiming to uncover trends and patterns. We found that transitioning the cycle from 16 hours/8 hours to 8 hours/16 hours, and vice versa, resulted in a reduction in the amplitude of light-induced potential change (fig. S38F). The amplitude decrease is more pronounced when the light phase is shortened (16 hours/8 hours to 8 hours/16 hours) (fig. S38G) than it is lengthened (8 hours/16 hours to 16 hours/8 hours). There is also an overall decreasing trend of the amplitude, likely reflecting stress responses to fluctuating light/dark cycles.

Drought, being one of the major abiotic stressors, greatly affects crop quality and productivity, posing serious challenges to global food security (55). Drought stress in plants can be reflected in EP signals (5, 56, 57), and EP signals enable monitoring under extreme drought conditions (5), which is challenging with common water status assessment methods such as leaf turgor sensors (58, 59) or pressure chambers (60). However, existing EP-based drought stress monitoring studies are mainly conducted using invasive electrodes in stems, which may not capture critical physiological changes occurring in the leaves, and it remains unclear whether the long-term tissue damage from electrode insertion affects the plants’ response toward drought stress. Here, drought stress and recovery were simulated on tobacco plants over 14 days with daily 16-hour/8-hour light/dark cycle. Continuous EP signals were recorded using the capacitively coupled electrodes (Fig. 5A and raw data in fig. S39), and daily visual inspection and soil water content measurement set the benchmark for the water status of plants. Stomata closure was microscopically observed before macroscopic wilting, suggesting an early drought stress response (fig. S40). For easy reference, we define the day of visible wilting as “drought” and the day before as “predrought” (Fig. 5A).

Fig. 5. Correlation of plant drought stress and long-term electrophysiology monitored by capacitively coupled electrodes.

(A) Procedure of drought stress and recovery experiment and photographs of plants under testing. Predrought is defined as 1 day before plants display wilting symptoms. (B) Amplitude of light-induced potential change for plants under 9 days of escalating drought (red) benchmarked against well-watered control plants (gray), showing an increase followed by decrease upon wilting. a.u., arbitrary units. (C) Light-induced potential change before and after inhibition of Ca²⁺ channels in the same leaf. N = 3. (D) Light-induced potential change before and after inhibition of ROS production in the same leaf. N = 3. (E) Confocal images of intracellular Ca²⁺ (top) and apoplastic ROS (bottom) during drought simulation, with quantification of the mean fluorescence intensity per region of interest (ROI) shown on the right. BF, bright field. Fluo-3 AM and 2′,7′-dichlorodihydrofluorescein diacetate (H₂DCFDA): dyes used for visualizing Ca²⁺ and ROS, respectively. Scale bar, 50 μm. CK, control plants; DR, drought-stressed plants. (F) Plant EP signals under the light and dark conditions with starting potential aligned for all samples on each day. N = 6. (G) Mean potential calculated from signals in (F) 1 hour after light on or light off, showing opposing trends with time under light and dark conditions. Dark mean potential changes before visual symptom onset on day 7. Solid lines in (C), (D), and (F) are mean values, and shadows represent SD. Box plots in (B), (E), and (G) display first quartile, median, and third quartile. Whiskers are drawn at 1.5 interquartile range distance. Means are denoted by empty squares. *P < 0.05; ***P < 0.001; ****P < 0.0001; n.s., not significant. Two sample t test.

First, we observed an increase in the amplitude of light-induced potential during predrought (Fig. 5B). The amplitude is correlated with the water status of plants, evidenced by the amplitude of the well-watered control plants, which maintains relatively constant over time with slight decrease within 2 days after watering (Fig. 5B and figs. S41 and S42). To further investigate how light-induced potential change is linked to physiological changes under early drought stress, we examined the intracellular Ca²⁺ and apoplastic ROS levels during drought stress. Light-induced potential change is triggered mainly by the influx of Ca²⁺ into the cell upon light stimulus with contribution of ROS waves (14, 61). The functional relationship between light-induced potential change and Ca²⁺ is supported by a substantial reduction in the amplitude of light-induced potential change upon inhibition of Ca²⁺ channels by GdCl₃, which blocks Ca²⁺ influx (Fig. 5C). On the other hand, the functional relationship with ROS is shown by a slight decrease in the amplitude upon inhibition of nicotinamide adenine dinucleotide phosphate (NADPH) oxidases, which produces apoplastic ROS (Fig. 5D). Thus, we conclude that the amplitude of light-induced potential change is mainly contributed by the Ca²⁺ influx, while ROS production serves a supportive role. Using confocal imaging (Fig. 5E), we observed an increase in both Ca²⁺ and ROS levels during predrought. A higher intracellular Ca²⁺ can boost ROS production, which, in turn, activates more Ca²⁺ channels (62, 63). The slight increase in Ca²⁺ and ROS during predrought may have therefore triggered a positive feedback loop that enhances the Ca²⁺ influx upon light stimulus, thereby increasing the amplitude of light-induced positive change. Under drought, when plants are wilted, a large increase in Ca²⁺ with persistent high ROS is observed (Fig. 5E). The high level of basal intracellular Ca²⁺ may have limited further Ca²⁺ influx during light onset (64, 65), and wilting of leaves may have impaired cellular functions associated with light-responsive EP signaling, leading to the decreasing amplitude of light-induced potential change, as observed under drought (Fig. 5B). A comparison between well-watered control plants tested on the same day and plants undergoing drought simulation proves that the measured increases in Ca²⁺ and ROS levels are due to drought stress (Fig. 5E and fig. S43).

Besides the transient responses to light, long-term signal features also show correlation with plant water status. The potential readings under light and dark conditions were processed separately with daily normalization of the initial potential readings to zero. The change in signals over the four different conditions (normal, predrought, drought, and recovered) is shown in Fig. 5F. An opposing trend between the potential levels under light and dark conditions is observed when plants experience escalating drought. To quantify such changes, we calculated the mean potential 1 hour after the start of each light/dark period to remove the influence of transient response toward light. Results show an increase in dark mean potential and a decrease in light mean potential (Fig. 5G). This phenomenon is particularly interesting, as it reveals that plants modulate their physiology under stress differently in the day and at night, reflecting the influence of circadian clock on drought adaptation (66, 67). The change in mean potential was further proved to be due to drought stress when compared with well-watered control plants (fig. S44). Noticeably, the change in light mean potential (decreasing after day 6) happens only when plants start to wilt. In contrast, dark mean potential shows an early change before wilting (increasing after day 4) and a close correlation with soil water content (Fig. 5G). Despite plant-to-plant variations, all individual plants show similar trends for the three identified signal features: amplitude of light-induced potential change and mean potential under light and dark conditions (fig. S45 and note S3). A repeated round of drought simulation further confirms the strong correlation between the signal features and drought stress (fig. S46).

To assess whether the observed signal changes were unique to drought, we also examined plants under moderate salt stress (soil immersion in 0.15 M NaCl for 1 week) (fig. S47). Both stresses impose osmotic pressure on plants, while salt stress additionally introduces ionic toxicity (68). Over the 1 week of salt exposure, no obvious necrosis, typically associated with ionic toxicity, was observed (fig. S47A), likely due to the stronger tolerance of mature plants and the relatively short duration of salt exposure. Nonetheless, older leaves began to wilt from the third day of salt application (fig. S47A), indicating that osmotic stress was successfully induced by salt stimulation, allowing it to serve as an osmotic comparison to drought stress. In contrast to drought stress, salt stress does not elicit an increase in the amplitude of light-induced potential change before leaf wilting (fig. S47B, left). This is explained by the lack of increase in intracellular Ca²⁺, although a slight increase in apoplastic ROS is observed (fig. S48). Despite the difference before wilting appears, both stresses induce a similar decrease in amplitude once wilting occurs (fig. S47B, left). Furthermore, salt stress does not result in obvious changes to the mean potentials (fig. S47B, middle and right) or the opposing trends between light and dark conditions. This contrasts sharply with drought stress, where early electrical responses and circadian influences are prominent. Given that our electrodes are applied on leaves, the absence of obvious potential changes might be due to stronger osmotic adjustment occurring at roots than shoots under salt stress (69), while drought stress induces stronger shoot responses (66, 70). The above signal features and physiological insights, resulting from the unique capability of the capacitively coupled electrode in measuring long-term on-leaf EP signals, set a foundation for future biological study and discoveries.

DISCUSSION

In summary, we presented a capacitively coupled electrode that uses thermoresponsive polymer for the formation of an electrical coupling layer to improve the long-term signal quality and stability of noninvasive electrodes in measuring plant EP signals. The electrode allows universal conformable attachment on different types of plant surfaces. The mechanically and electrically stable coupling layer between the conductor and plants enables month-long capacitive coupling, minimizing signal drift or noise. Although capacitive measurement of biopotentials is widely explored for on-skin applications, our work first demonstrates this working principle for on-plant EP monitoring. Furthermore, we propose the use of thermoresponsive polymer as a capacitive coupling layer and showcase its unique combination of advantages in applicability, biocompatibility, signal quality, reproducibility, and stability, which cannot be simultaneously achieved by other polymer coupling layers. This strategy stretches the limit of sensing capabilities of existing plant EP electrodes, offering unprecedented versatility, signal fidelity, and monitoring duration.

Current research on surface electrophysiology sensors, both on animal tissues and on plants, focuses on developing conductive conformable materials. Conformability usually comes from reduced modulus, high stretchability (71, 72), and shape-morphing capabilities (73–75). Engineering these properties in conductive materials or vice versa is fundamentally challenging. In this work, we circumvent the stringent requirement in electrical conductivity for biointerfacing materials using the strategy of capacitive coupling. This strategy not only enables the use of nonconductive materials with unique properties but also provides more design freedom in material development. We also show that such a capacitive coupling strategy applies to many other polymer and conductor systems, paving the way for functional material and device innovations in sensor and biohybrid applications on plants (76).

The preliminary analysis and physiological investigation of plant EP signals under various stresses in this work promise a more fundamental understanding of plant signaling and may lead to innovative plant monitoring technologies. For instance, because of the unique capability of the capacitively coupled electrode in monitoring long-term light-responsive EP signals on leaves, we observed shifting EP circadian cycles when switching light/dark conditions, manifested in the decreasing amplitude of light-induced potential changes. The amplitude of light-induced potential changes also displayed an obvious correlation with plant water status and is tightly linked to the Ca²⁺ and ROS levels in leaves. Moreover, circadian effect on drought-stressed plants was observed, where EP signals under light and dark conditions changed in opposing directions. While our observations are preliminary, the capacitively coupled electrode provides a reliable platform for further in-depth study. Possible future directions include systemic long-term study to identify signal features under salt stress and to unravel the underlying mechanisms leading to different responses to drought and salt stresses. Presence of “electrical memory” of plants where the EP dynamics is temporarily or permanently altered by previous stresses could also be explored.

Further improvement of the capacitively coupled electrode can be made in both the adaptable coupling layer and the conductor. First, the thermoresponsive EPC polymer can be modified with various ionic side groups (77) to balance dielectric properties with conductivity, for maximal capacitance as a capacitive interlayer. More fundamental physical aspects of capacitive coupling should be investigated, such as how charge migration and polarization affect signal transduction. Furthermore, the adaptable coupling layer can be engineered with stretchability to morph together with growing plants. This will enable the monitoring of fast-growing plants and even life-long monitoring throughout the growth cycle. To this end, impedance stability should be further improved, possibly using a nondegradable coupling layer, which has shown more stable impedance beyond a month (Fig. 3G). In terms of the conductor, future research on improving light transparency and gas permeability will be needed to make the sensor imperceptible to plants. In addition, the development of advanced computational methods is required for processing large datasets and accurately identifying stress signals to enable real-time, long-term plant health monitoring. Together, these technological advancements will provide plant biologists, ecologists, farmers, etc., with capable tools to measure and analyze previously inaccessible signals and to study how plants adapt and respond to the environment. These signals and studies will generate insights that help address issues in food security, biodiversity, and climate change.

MATERIALS AND METHODS

Materials

Preparation of polymer gels as coupling layer

To prepare agar gel, agar powder (Sigma-Aldrich, A1296) was weighed and added in 10 mM KCl at 0.5% (w/v). The mixture was heated and stirred at ~90°C until complete solvation and then cooled to room temperature with magnetic stirring at 300 rpm. The resultant mixture is a viscous translucent liquid, stable at room temperature for months. The detailed protocol for the synthesis of EPC polymer can be found in our previous paper (21). Synthesis of PEG-PPG polymer is the same as EPC polymer except without adding PCL-diol. EPC and PEG-PPG hydrogel solutions were prepared by adding 10 wt % of lyophilized copolymers in 10 mM KCl and stored at 4°C for 3 days to dissolve and equilibrate.

Preparation of polymer liquid precursors as coupling layer

PVA solution was prepared by dissolving PVA (average M_w: 13,000 to 23,000 and 98% hydrolyzed; Sigma-Aldrich) at 20 wt % in 10 mM KCl, under stirring at 90°C. PEG solution was prepared by dissolving PEG (average M_n: 35,000; Sigma-Aldrich) at 20 wt % in 10 mM KCl at room temperature on a shaking bed. Ecoflex 00-35 FAST (Smooth-On) was used for its fast cure within 5 min to facilitate electrode application on plants. Precursors A and B were mixed at a 1:1 ratio, dropped on the plant, and capped with a conductor immediately to ensure liquid-phase contact and conformal curing.

Preparation of conductor strips

Au strips were prepared through thermal evaporation of 70-nm Au on flexible polyimide (PI) thin films (~30-μm thickness). A 5-nm-thick Cr layer was first deposited to enhance adhesion between Au and PI. The coated thin films were cut into desired dimensions (usually 2 mm in width and 5 to 10 cm in length) for use. Ag/AgCl conductor strips were prepared by coating Ag/AgCl paste (Dycotec Materials, DM-SIP-3450) on one tip of Au strips and drying in a 60°C oven. To prepare CNT conductor strips, CNT-BC composite films were prepared at 80 wt % CNT and 20 wt % BC. Removal of impurities in powdered CNT (BuckyUSA, Bu-202) was done by immersion in 3 M aqueous HCl for 12 hours and filtration with DI water 5 times. 64 mg of CNT and 26 mg of BC were dispersed in deionized (DI) water at 0.2 mg/ml and mixed using a blender. Vacuum infiltration of the dispersion was conducted using a 9-cm filter membrane to obtain the thin CNT-BC composite film. The film was cut into 3 mm–by–3 mm pieces and pasted on one tip of Au strips using Ag paint (Electrolube, SCP03B), followed by 60°C drying in the oven.

Preparation of invasive electrodes

The Ag/AgCl invasive electrodes were prepared by electrochemical chlorination on Ag wires (100-μm diameter). dc bias (1.5 V) was applied across a Ag wire (+) and a Pt counter electrode (−) in 0.1 M HCl for ~100 s (22).

Fabrication of microneedles

The microneedles are fabricated with ultraviolet-cured polyacrylate precursor (Anycubic, Standard Resin) in PDMS mold with needle dimension of 0.8 mm in height, 0.4 mm in base diameter, and a substrate thickness of 1 mm.

Preparation of PVA hydrogel

PVA (average M_w: 146,000 to 186,000 and 87 to 89% hydrolyzed; Sigma-Aldrich) was dissolved in DI water at 10% (w/v) by heating to 90°C under constant stirring until a clear solution was obtained. For PVA hydrogel with salt, 1.25 g of CaCl₂ was added to every 6 g of PVA (equivalent to ~0.188 M CaCl₂). The prepared PVA solution was poured into 24-well Corning Falcon Cell Culture Plate (Sigma-Aldrich, CLS353047) and subjected to one freeze-thaw cycle (freezing at −80°C for 24 hours followed by thawing at room temperature). For preparation of PVA hydrogel with glycerol, PVA hydrogel prepared with DI water was immersed in glycerol for solvent exchange for 1 day.

Material characterization

Rheology

Rheological studies of thermoresponsive EPC hydrogel were conducted using TA Instruments DHR-3, Research Instruments Inc. All amplitude sweeps were conducted with an oscillation frequency of 5 rad s⁻¹ from 0.1 to 3000%, at 25°C. Temperature ramp was conducted from 0° to 60°C at an oscillation frequency of 5 rad s⁻¹, 1% oscillation amplitude, and temperature ramp rate of 3°C min⁻¹.

Water content measurement

The water content was measured with thermal gravimetric analysis (TGA Q500) by calculating the weight change after a temperature ramp from 27° to 100°C at a ramp rate of 10°C/min and equilibration at 100°C for 10 min (typical measurement shown in fig. S20).

Measurement of thermoresponsive polymer thickness changes with time

The thickness of thermoresponsive hydrogel droplets on glass slides at specific time points postdroplet application was measured using a confocal microscope (Leica, DMi8 Inverted Microscope). To prepare the sample, thermoresponsive hydrogel solution (10% EPC and 10 mM KCl) was dropped on a glass slide using a plastic dropper, and a PI film of similar size to the droplet was gently placed on the droplet. The addition of a PI film was to mimic the attachment of electrodes on plants while ensuring transparency for microscope observation. Six samples were prepared and measured separately. The thickness of the droplet was measured by recording the z values (stage height) of the top and the bottom surfaces of the droplet (bright-field mode) and taking the difference. After each observation, the samples were placed in the plant growth environment to approximate the dehydration conditions. Nevertheless, because of substrate (glass versus plants) and capping material (PI film versus PI-Au-Ag paint-CNT laminates) differences, the thickness change measured in this way inevitably differed from the true value of applying electrodes on plants.

FTIR spectroscopy of leaves

The leaves treated with PDMS were cut off the plants, and PDMS was removed from the surface of the leaf. The samples were measured using Perkin Elmer Frontier under attenuated total reflectance mode.

Shear adhesion test

For adhesive strength testing of thermoresponsive hydrogel, agar gel, PVA, PEG solutions, or Ecoflex precursor, a three-dimensional (3D) printed polylactic acid mold (1-cm² area) was adhered to the wet gel or solutions applied on cut tobacco petioles. Ecoflex 00-30 was used in place of Ecoflex 00-35 FAST due to its longer curing time, making it more suitable for handling during adhesion strength testing. Precursors A and B were mixed at a 1:1 ratio, applied on a cut tobacco petiole, and then adhered by a mold, before testing for the adhesion strength. For adhesive strength testing of dried thermoresponsive polymer and agar, a CNT-BC film was first superglued onto the 3D-printed mold. This assembly was then applied to the gel on tobacco petioles for dehydration over 1 day before testing. All shear adhesion tests were carried out with MTS Criterion Model C42 with a strain speed of 50 mm/min.

Water vapor permeability test

Thermoresponsive polymer film was first prepared by dehydrating thermoresponsive hydrogel in a silicone mold. Ecoflex was prepared by dropcasting the precursor on a petri dish. The prepared films were secured to cover 10-ml glass vials containing 9 ml of DI water, and the permeability is calculated by W_loss/(A·t) where W_loss is the weight of water loss in the container, A is the area of the vial mouth, and t is the evaporation time.

Observation of cross section and thickness of thermoresponsive polymer

Thermoresponsive hydrogel was dropcasted on plant surfaces for dehydration over a day to form the adaptable coupling layer. The standalone adaptable coupling layer was obtained by peeling the layer off the plant surface. Cross-sectional samples were prepared by liquid nitrogen freezing and fracture. Samples were sputtered with 10-nm Au before scanning electron microscopy (SEM) observation. Surface morphology and cross-sectional view of thermoresponsive polymer were examined with field-emission SEM (FESEM; JEOL, JSM-7600F) at an accelerating voltage of 5 kV.

Observation of interface between thermoresponsive polymer and CNT-BC film

Thermoresponsive hydrogel was applied on PDMS substrate, and CNT-BC film was applied on top of the hydrogel layer for drying to mimic the application on plants. After complete drying, samples were removed from PDMS substrate and sputtered with 10-nm Au before SEM observation. Cross-sectional view was taken at the edge of the sample with FESEM (JEOL, JSM-7600F) at an accelerating voltage of 5 kV.

Sheet resistance measurement

The sheet resistance of the CNT-BC thin-film conductor was measured using a four-point probe system (AiT Equipment, CMT-series) by Jandel Engineering. The thin film conductor was pasted on a PI tape to ensure flatness. Six samples of dimension 8.0 mm by 20.0 mm were prepared and placed on a glass slide during measurement. Three positions on a single sample were measured. The four-point probe had a 100-g load on each needle, with a tip radius of 100 μm and needle-to-needle spacing of 1.00 mm.

Thickness measurement of CNT-BC conductor

The thickness of the CNT-BC thin-film conductor was measured using a thickness gauge with a resolution of 0.001 mm.

Electrode characterization

Temperature measurement of thermoresponsive hydrogel during application

Thermal imaging camera (Fluke, Ti480 PRO Infrared Camera) was used for recording the entire gel application process on a tobacco plant leaf. An agar gel electrode was preapplied on the same leaf to record any EP responses from the cold solution application.

Impedance measurement

Potentiostatic EIS was conducted from 10⁵ to 10⁻¹ Hz at an amplitude of 50 mV using Electrochemical Workstation Zennium (ZAHNER-elektrik GmbH & Co. KG). Z_on-leaf is the impedance of two capacitively coupled electrodes on a tobacco leaf with a 1-cm distance (n = 8 individual plants). The EPC hydrogel was applied freshly onto the surface of a tobacco leaf through drop casting, and a conductor with an area of 0.09 cm² (a square electrode with a side length of 0.3 cm) was positioned on top of the hydrogel for assembly of the capacitively coupled electrode. Z_laminate is the impedance of thermoresponsive polymer between two Au-PI conductor strips (n = 5) with a width of 0.3 cm. The contact area between the EPC hydrogel and the Au-PI conductor films was maintained at 0.3 cm², with the conductor strips overlapping over a length of 1 cm. Experimental results were fitted with equivalent electrical circuit models using ZView4 for further analysis.

Biocompatibility assessment I: Stomata observation

Capacitively coupled electrodes were applied to the adaxial leaf surfaces of tobacco plants that were around 7 weeks old. Each plant had one leaf attached with the electrode, and a total of seven plants was used. Next, plants were cultured under normal growth conditions for 4 weeks before microscopic observation. Leica DM2700 M Upright Materials Microscope was used in reflective mode under white light. Leaves to be observed were cut off from the mother plants and flattened and fixed on a glass slide using tapes with the abaxial surface facing up. Abaxial surfaces were observed because stomata are larger in number on abaxial surfaces than on adaxial surfaces. Microscope images at ×20 magnification on the leaf surface above the electrode and away from the electrode but within a 1-cm distance were captured, respectively (fig. S8B). The numbers of open and closed stomata on each microscopic image were counted, and the percentage of open stomata out of the total number of stomata was calculated and plotted.

Biocompatibility assessment II: Chlorophyll content measurement

The chlorophyll content of plants was measured using a handheld chlorophyll meter (LCPM-A10), which calculates soil plant analysis development (SPAD) from 0.00 to 99.9%. The meter used an optical method, which measures the intensity of light transmitted or reflected at two wavelengths (650 and 940 nm). For long-term measurements, the electrodes were not removed throughout the measuring period of 1 month. To prevent interference from the electrode on the optical detection by the meter, measurement of chlorophyll content of electrode-covered leaves was conducted beside the electrode on the same leaf.

SEM and energy-dispersive x-ray of invasive electrodes

SEM observation and energy-dispersive x-ray analysis were conducted on Ag/AgCl invasive electrodes using FESEM (JEOL, JSM-7800F) before and after insertion in tobacco petioles for a month.

Optical microscopy of plants with invasive electrodes

Microscopic images of plants with poked holes by invasive electrodes were captured using a Leica DM2700 M Upright Materials Microscope in reflective mode under white light.

Plant cultivation and maintenance

Plants were cultivated in an air-conditioned room with temperature and humidity maintained at 24° to 25°C and 50 to 60%, respectively. All plants were subjected to 16-hour light and 8-hour dark conditions and were watered regularly. N. benthamiana seeds used were obtained from the Temasek Life Sciences Laboratory, Singapore, and the remaining plants were purchased from a local market.

Plant electrophysiology testing

The general procedure

USB-2610 Series DAQ (Smacq Technologies) was used for data acquisition. The data sampling rate was 10 Hz for short-term testing of less than 6 hours and 5 Hz for longer-term testing to reduce data size. Recording electrodes were usually applied in the middle of leaves while avoiding the midrib. Reference electrodes for testing less than 24 hours were Ag/AgCl electrodes (Shanghai Jingchong Electronic Technology Development Co. Ltd.) inserted in moist soil. For testing longer periods, capacitively coupled electrodes were applied at the bottom of stems to provide stable potential. This was done because prolonged insertion of Ag/AgCl electrodes in soil caused leakage of electrolytes. A Faraday cage made of copper mesh enclosed the entire testing system including plants, electrodes, and DAQ for shielding electromagnetic interference. The ground was connected to the cage. A laptop connected to DAQ was placed outside the cage for data visualization and storage. White LED lights were equipped on the top of and outside the cage to provide timed lighting. To apply electrodes composed of fresh or dry hydrogels and polymers, the liquid precursor was dropped on plants, followed by gently laminating the conductor strip on top of the liquid droplet. Plants were then put back into the growth room for gel/polymer solution to dry. Fresh gel–based electrodes were applied right before testing. Raw signals came in the form of electric potential as a function of time. Further processing was performed using MATLAB and Origin software. Specific processing methods for each experiment are elaborated in separate sections.

Comparison of electrode performance on plants

Electrodes made of different combinations of conductor materials and gel/polymer compositions were compared in terms of signal amplitude, background noise level, and SNR, on days 3, 7, 14, 21, and 31 from the day of electrode application. On each day, stress-induced plant EP signals were triggered by flame wounding a distant leaf from the testing leaf for 3 s, and signal recording usually lasted for less than 2 hours for each wounding event. On the testing leaf, all electrodes under comparison were attached at a minimal distance possible from each other without touching (<1 cm). This was to minimize the signal difference coming from the location effect. For tests involving agar gel electrode, a fresh electrode was attached to the plant before each signal recording. Eight plants of the same age (~7 weeks) were used in a single batch of experiments. To further reduce the location effect, the positions of electrodes on each plant were rotated to make sure no two plants had the same spatial arrangement of electrodes. Applying all electrodes on a single leaf minimized plant-to-plant variation, for a more accurate comparison of electrode performance. For all signal acquisition, Ag/AgCl reference electrode inserted in the soil was used. Signal amplitude was determined as the difference between the maximal potential and the minimal potential of stress-induced electric potential perturbations. The background noise level was calculated as the root mean square difference between the raw signal and smoothed signal (averaging 20 points), in the interval bound by maximal potential and minimal potential (usually ~1000 s). SNR was then calculated using the following equation: SNR = (A_signal/A_noise)² where A_signal is the signal amplitude and A_noise is the background noise level.

Circadian cycle monitoring

Setup

Electrodes, plants, and the testing setup were prepared according to the general procedure stated above. Deviations and detailed implementation include the following: (i) Electrodes were applied at the intersection between petiole and leaf, where leaf growth was minimal. This was to minimize growth-induced electrode instability. (ii) For each plant, two electrodes were applied on the seventh and eighth youngest leaf, respectively, sharing one reference electrode at the bottom of the stem. Four plants were tested. (iii) Control electrodes were prepared by attaching the capacitively coupled electrodes to an “artificial plant” mimicking the material composition and structure of plants (fig. S38E). Poly(ethylene) film resembled the insulating cuticle, holes poked by needles resembled open stomata, and the underlying agar gel with electrolyte mimicked conductive and watery plant tissues. This artificial plant was soaked in 0.5 M NaCl throughout the test, and evaporation was minimized through careful sealing of the container. A Ag/AgCl electrode was inserted into the electrolyte to serve as the reference. (iv) The light/dark cycle was changed between 16 hours/8 hours and 8 hours/16 hours as described in the Results section: “Month-long plant electrophysiology study using capacitively coupled electrodes.”

Signal processing

The amplitude of light-induced potential change is calculated by taking the maximum value within 10 min from the start of light (fig. S38F).

Drought stress simulation and monitoring

Setup

Electrodes, plants, and the testing setup were prepared according to the general procedure stated above. Plants were subjected to natural drying over 9 days before recovery of plants by rewatering on day 9.

Soil water content measurement

A handheld soil data logger (RS-TRREC Soil Analyser and Data Logger RS-TRREC-N01-1-EX), adopting the time-domain reflectometry (TDR) principle, was used for measuring the water content present in the soil. The soil moisture sensor was inserted directly into the soil of one potted plant for estimation of the soil water content of six plants.

Signal processing

The same calculation for amplitude of light-induced potential change in the circadian cycle monitoring was used. For all plants, the starting potential when the lights were on/off on each day was aligned to zero, and only data points 1 hour after the lights on/off were processed. For light conditions, mean potential values were calculated across the remaining 15 hours. For dark conditions, mean potential values were calculated across the remaining 7 hours.

Stomata imaging of leaves

Stomatal closure in response to drought stress was assessed using 4- to 5-week-old N. benthamiana plants. Fully expanded leaves were infiltrated with 20 μM propidium iodide solution and incubated in the dark for 10 min to stain the stomatal apertures. Following incubation, leaves were excised and briefly rinsed with distilled water to remove excess dye. The stained leaves were mounted on microscope slides with coverslips, ensuring no air bubbles were trapped, and immediately observed using a customized Nikon Ti2-E inverted microscope equipped with a spinning disk confocal (SDC) confocal scanner unit (CSU) module (Gataca Systems) to visualize the stomatal aperture at the following settings: excitation at 488 nm and emission at 505 to 545 nm (green fluorescence); excitation at 561 nm and emission at 600 to 640 nm (red fluorescence). All images were analyzed with Fiji (ImageJ). The width and length of the stomatal aperture were measured, and the stomatal aperture index was calculated by division of the aperture width through the length.

Salt stress simulation and monitoring

Setup and signal processing

Electrodes, plants, and the testing setup were prepared according to the general procedure stated above. Plants were subjected to salt stress by watering 0.15 M NaCl salt solution in the tray for natural uptake by plants. Similar signal processing as drought stress was conducted.

Soil conductivity measurement

A handheld soil data logger (RS-TRREC Soil Analyser and Data Logger RS-TRREC-N01-1-EX), adopting the TDR principle, was used for measuring the soil conductivity. The soil conductivity sensor was inserted directly into the soil of one potted plant for estimation of the soil conductivity of all three monitored plants.

Investigation of intracellular Ca²⁺ and apoplastic ROS

Microscopic quantification

Apoplastic ROS in leaf tissues of 4- to 5-week-old N. benthamiana plants were visualized using the fluorescent dye 2′p7′pdichlorodihydrofluorescein diacetate (H₂DCFDA). Leaves were infiltrated with 20 μ LH₂DCFDA and incubated in the dark for 7 min. After incubation, leaf sections were excised to the appropriate size and mounted on glass slides. Coverslips were placed on top to seal the samples, ensuring that no air bubbles were trapped.

Intracellular Ca²⁺ ions were detected using the fluorescent Ca²⁺ indicator Fluo-3 AM (MedChemExpress, HY-D0716). Leaf samples were stained with 10 μM Fluo-3 AM for 2 to 3 hours in the dark and subsequently washed twice with distilled water before imaging.

Microscopy was performed using a customized Nikon Ti2-E inverted microscope equipped with an SDC CSU module (Gataca Systems). Fluorescent signals were collected with a scientific complementary metal–oxide–semiconductor (sCMOS) camera (Hamamatsu, ORCA-Fusion) under 60× and 100× objective lens, and all image acquisitions were controlled by MetaMorph software (Molecular Devices). Excitation/emission wavelengths were set at 488/510 nm for H₂DCFDA and Fluo-3 AM, and 639/710 nm for chlorophyll autofluorescence. All images were analyzed with Fiji (ImageJ). Regions of interest (ROIs) were manually selected for each image, and mean fluorescence intensity values were quantified.

Inhibition of Ca²⁺ channels and ROS

Calcium channels were inhibited by injecting 2 mM GdCl₃ into the abaxial surface of the leaf that was attached with an EP recording electrode. ROS was inhibited by injecting 20 μM diphenyleneiodonium chloride (DPI) into the abaxial surface of leaf with electrode. DPI blocks NADPH oxidases from producing ROS.

Supplementary Materials

The PDF file includes:

Supplementary Notes S1 to S3

Figs. S1 to S51

Tables S1 and S2

Legends for movies S1 to S7

References

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S7