Temperature Correction of Printed Na+, K+, and pH Sensors with PEDOT:PSS‐Based Thermistors toward Wearable Sweat Sensing

Temperature correction for sensors is a critical aspect of ensuring accurate measurements in wearable devices, because skin and sweat temperatures vary between 20 and 40 °C depending on individual and time. Here, this study reports on the temperature dependence and correction techniques of printed Na+, K+, and pH sensors toward wearable applications. The ion sensor array is fabricated using a cost‐effective printing method. To enable temperature correction, a printed thermistor of crosslinked poly(3,4‐ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) is monolithically integrated with the ion sensor array on a flexible plastic substrate. Temperature dependence of the potential response of the printed ion sensors exhibits a linear behavior with a slope of 1–2 mV °C−1 in the physiological skin temperature range of 20–40 °C. Applying temperature correction to the ion sensors, the maximum relative errors are reduced from 60% to 7.8% for the Na+ sensors and from 76% to 14.6% for the K+ sensors, while the maximum absolute error is reduced from 0.88 to 0.19 for the pH sensors, indicating the critical importance of temperature correction as a technology for wearable printed ion sensors.


Introduction
Printed electronics is a promising foundational technology for the Internet of Things (IoT) ecosystems due to its ability to manufacture eco-friendly, large-area, highly customizable, and additive sensor devices.An example of beneficial applications of printed DOI: 10.1002/adsr.202300106sensor technology is wearable sweat sensors for healthcare, medical, and sports purposes.Sweat contains crucial biomarkers, [1] including electrolytes, [2] metabolites, [3] and amino acids, [4] which can be sensed and analyzed in realtime in a cloud-based system to facilitate the prevention and diagnosis of heat stroke, heart disease, and lifestyle related diseases. [5]To enhance the precision and reliability of these wearable sweat sensors while ensuring a seamless sweat sampling process, interface technologies between the sensors and human bodies, such as microfluidic devices for sweat flow control and vapor prevention, [6] biocompatible hydrogel pads, [7] and iontophoresis technology, [4] have been introduced.
Temperature correction for sensors is a critical aspect of ensuring accurate measurements in wearable devices.The skin and sweat temperatures are known to vary between 20 and 40 °C depending on individual and time due to a wide range of factors, including activity level, [8] dietary intake, [8] body part, [8] and skin injuries. [9,10]As such, temperature correction capabilities for enzyme-based sensors, including glucose and lactate sensors, [5,11] were investigated.[14][15][16] Despite the extensive research on the ion sensors, there were few reports on the temperature dependence and correction of ion sensors fabricated based on printing methods, and the target ions of these sensors were ammonium and nitrate ions, [13,14] which are minor in sweat.In addition, there were few detailed discussions on the temperature dependence and correction procedures of wearable sensors for sodium (Na + ) and potassium ions (K + ), which are major electrolyte biomarkers in sweat. [1]Therefore, detailed research on the temperature correction of printed Na + and K + sensors is necessary to achieve accurate and reliable measurements in wearable devices.
In this study, we report on the temperature dependence and correction techniques of printed Na + , K + , and pH sensors toward wearable applications.We employed silver nanoparticle ink and carbon black paste, which are common conductive materials in printed electronics, as the electrode materials for the ion sensors. [2,17]The ion sensor array was fabricated using a large-area compatible printing method that did not require vacuum processing.To enable temperature correction, a printed thermistor of crosslinked poly (3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) [18,19] was monolithically integrated with the ion sensor array on a flexible plastic substrate.The response of the sensors was evaluated using a wireless potentiometric sensing system on a printed circuit board (PCB), a 3-V sheet-type battery, and a tablet user interface with a custom-developed application.Considerable linear temperature dependence in the output signals of the printed ion sensors was observed within the physiological skin temperature range of 20-40 °C.Taking this temperature dependence in account, we devised a procedure for temperature correction with the printed thermistor and applied it to readouts from the ion sensors.The maximum relative errors were reduced from 60% to 7.8% for the Na + sensors and from 76% to 14.6% for the K + sensors, while the maximum absolute error was reduced from 0.88 to 0.19 for the pH sensors, indicating that temperature correction for printed ion sensors plays a critical role not only in laboratory settings but also in wearable applications.

Design of Printed Ion Sensors and Thermistors
As illustrated in Figure 1a, the wireless sensor device was composed of a sensor, an integrated circuit on printed circuit board (PCB), and a 3-V lithium-ion sheet-type battery.The dimensions of the PCB were 21 mm × 25 mm × 2 mm, which was sufficiently compact for wearable applications.The PCB featured two flexible printed circuit (FPC) connectors, thereby facilitating the removal of the sensor and battery.The PCB comprised an analog circuit (buffer, filter, and divider), a Bluetooth Low Energy (BLE) wireless communication module with a built-in analog-to-digital converter (ADC), and a voltage regulator for ensuring stable power supply (Detailed circuit diagrams are presented in Figure S2, Supporting Information).The supply voltage to the analog circuit was 1.8 V.The resolution of the potential signal of the ADC was set to 0.6 mV, which is sufficiently small.Figure 1b depicts the system-level block diagram of the signal transduction, conditioning, processing, and wireless transmission paths.The data acquired from the sensors were transmitted and stored on a tablet equipped with a custom-developed application.The structure of the sensors is depicted in Figure 1c.We integrated Na + , K + , pH sensors, and a temperature sensor on a 100-μm-thick polyethylene naphthalate (PEN) film using printing methods at process temperatures below 140 °C (process flow presented in Figure S1, Supporting Information).
The ion sensors in this study were potentiometric sensors (voltage output) that measure the potential difference between the ion-selective electrode (ISE) and the reference electrode (RE).The ion sensors consisted of three ISEs (Na + , K + , H + -ISE) and one common RE (Figure 1c).The ISE and RE electrodes each featured a bilayer structure that comprised inkjet-printed silver wiring (to provide sufficient conductivity as the electrode) and a carbon black layer (important for stabilizing the potential), [2,17] as shown in Figure 1c.The carbon black layer of the ISE functioned as an ion-to-electron transducer playing a pivotal role in stabilizing the potential of the ISE. [20]The top layer of the ISE consisted of an ion-sensitive membrane that contained an ionophore, conferring selective responsiveness to the specific ion.Different ionophore materials were used for the respective target ions (Na + , K + , H + ).The reason for employing a carbon black layer on the RE was to obtain higher potential stability against chloride ions than with Ag/AgCl reference electrodes, [13] given that chloride ions constitute a major component of electrolytes in sweat. [1]The top surface layer of the RE was a liquid junction layer (LJL) formed via drop-casting, which comprised a polyvinyl butyral (PVB) film dispersed with NaCl to enhance potential stability of the RE. [7,21,22]y applying a water-repellent fluororesin bank around the carbon black layer, we defined the area of the ISM/LJL while preventing electrical conduction between sweat/skin and the silver electrode by covering parts of the silver electrode surface with the bank to form an insulating layer.
The temperature sensor was a thermistor that changed its resistance according to temperature, consisting of electrodes, a temperature-sensitive layer of a PEDOT:PSS-based composite, [18] and encapsulating layers.The encapsulating layer was shared with the bank/insulating layer of the ion sensor.Preventing mois-ture from penetrating the PEDOT:PSS layer through the encapsulating layer is a key requirement for stable temperature measurement as we reported previously. [19]Thus, we employed an additional encapsulating layer of 80-μm-thick Teflon tape on the top surface of the thermistor (Figure 1c) to ensure resistance to moisture, enabling reliable measurement even in water (Figure S7, Supporting Information).

Response of Printed Ion Sensors and Thermistors
The response of the sensors was monitored by immersing them in electrolyte solutions with controlled temperature using a hot/cold plate (the complete experimental setup is depicted in Figure S3, Supporting Information).Figure 2a-c shows the representative open circuit potential responses of Na + , K + , and pH sensors at 20 °C, respectively, with the electrolyte concentration set at physiologically relevant levels of 10-100 mm NaCl, 1-10 mm KCl, and 3-8 pH, whereas the general electrolyte concentrations contained in sweat are known to be 20-100 mm for Na + , 4-24 mm for K + , and 4-7 for pH. [23]The potential response E of the ion sensor is given by the Nernst equation: , where E 0 is the standard potential, R is the ideal gas constant, T is the temperature, F is the Faraday constant, and [X] is the concentration of ion X.Hence, the sensitivity is evaluated in a unit of V dec −1 .The calibration of the Na + , K + , and pH sensors was linear (Figure 2d-f), and the obtained sensitivity was 62.2 ± 1.9, 61.2 ± 2.5, and 57.3 ± 0.6 mV dec −1 (n = 5), respectively, which were close to the theoretical value of 58.2 mV dec −1 at 20 °C.E 0 was estimated to be ≈0.05V for the Na + sensors, ≈0.11 V for the K + sensors, and ≈0.36 V for the pH sensors.The hysteresis in response of each ion sensor was several mV under increase/decrease in respective ion concentration, representing the good reversibility (Figure S4, Supporting Information).The response to other ions (Na + , K + , H + , and NH 4 + ) than the target ion of each sensor was less than a few mV, confirming the good selectivity (Figure S5, Supporting Information).Potential change of the RE versus a commercial Ag/AgCl reference when the chloride ion concentration was changed from 10 to 100 mm was also less than a few mV (Figure S6, Supporting Information).
Figure 2g depicts the representative response of the temperature sensor in the physiological skin temperature range of 20-40 °C, exhibiting negative temperature coefficient (NTC) behavior as a thermistor.The sensitivity of the thermistor is defined by the temperature coefficient of resistance (TCR), given by the following equation: TCR where T is the temperature, T 0 is the reference temperature, R is the resistance at T, and R 0 is the resistance at T 0 .Figure 2h shows the resistance change relative to the reference resistance R 0 (herein, T 0 is 20 °C), and the estimated TCR was −0.70 ± 0.02% °C−1 (n = 5), which was comparable to the sensitivity reported in previous literature. [18,19]Figure 2i shows the resistance-temperature behavior.The B-coefficient was 690 ± 26 K (n = 5), estimated by the equation for resistance of NTC thermistors, The temperature sensitivity of the NTC thermistor was high enough to put the temperature resolution set-ting down to ≈0.13 °C in a system using a voltage divider circuit (shown in Figure S2, Supporting Information) and an ADC with a voltage resolution of 0.6 mV, representing an acceptable resolution setting for temperature correction as well as skin temperature measurement.The power consumption of the thermistors was estimated to be ≈3.6 μW, indicating the heat generated by the thermistor had almost no effect on measurements.

Influence of Temperature on Response of Printed Ion Sensors
The temperature dependence of potential response of ion sensors is typically considered negligible, with a value of ≈0.09 mV°C −1 according to the Nernst equation.However, empirical evidence suggests that a significant linear dependence exists with values of 1-3 mV °C−1 , [14,24] and the dependence is influenced by various factors, such as the material and geometry of the sensor.One explanation for this linear temperature dependence is that standard potential, E 0 , varies linearly depending on temperature, [14] which is not taken into account in the Nernst equation.The evaluation of this linear temperature dependence is of paramount importance, as it imparts an exponential temperature dependence to the inferred activity/concentration of analytes calculated from the measured potential. [14]igure 3a-c depicts the relationship between the open circuit potential and temperature of the Na + , K + , and pH sensors, respectively.In our ion sensors, a linear temperature dependence of the potential was observed.The sensitivity to temperature was −1.1 ± 0.08 mV °C−1 in the range of 10-100 mm for the Na + sensors, while it was −1.7 ± 0.09 mV °C−1 in the range of 1-10 mm for the K + sensors, suggesting that the sensitivity to temperature was independent of the concentration of the electrolyte solutions.For the pH sensors, the sensitivity was −2.4 ± 0.21 mV °C−1 at pH 8, −2.0 ± 0.17 mV °C−1 at pH 5, and −1.4 ± 0.16 mV °C−1 at pH 3, depending slightly on the concentration.These values were found to be comparable to previously reported values of 1-3 mV °C−1 , [14,24] indicating that temperature correction is an indispensable prerequisite for accurate measurement.Theoretical temperature dependence of ion sensitivity is 0.2 mV dec −1 °C−1 according to the Nernst equation, while for the Na + and K + sensors, no significant temperature dependence was observed (Figure 2d,e).For the pH sensors, a linear fit of 0.16 mV dec −1 °C−1 was found to be in approximate agreement with theoretical predictions (Figure 2f).

Temperature Correction of Printed Ion Sensors with Printed Thermistors
To achieve accurate measurements in environments where temperature fluctuations are present, a modification of the Nernst equation that accounts for temperature effects was implemented.Considering the linear temperature dependence of the E and the small temperature-dependent sensitivities (shown in Figure 3), we ignored the impact of temperature on sensitivity, and a temperature correction term was added to Nernst equation, yielding the following expression: where  is the slope of E with respect to temperature T,  is the sensitivity to concentration,  is the standard potential at T 0 , and [X] is the concentration of ion X.By rearranging Equation 1 with respect to concentration, we obtained the following formula that includes the temperature correction term: The calibration of the ion sensor at two concentration points was performed at 20 and 40 °C, respectively (for a total of four calibration points), to obtain , , and .Here,  was obtained by taking the average of the values obtained at the two different concentrations.T was obtained from the thermistor calibrated using the following equation: that was derived by combining the equation for resistance of negative temperature coefficient (NTC) thermistors, , where R 0 is the resistance at T 0 , and B is the B-coefficient, with the equation for the output voltage of the voltage divider, where R ref is the reference resistance (100 kΩ, see Figure S2, Supporting Information), and V DD is supply voltage.
Figure 4a-c shows the temperatures obtained using the printed thermistor along with the results of temperature correction for the printed ion sensor responses.In contrast to the concentrations without temperature correction, the corrected concentrations were very close to the true values, indicating the effectiveness of temperature correction.Although some noise and deviation from target temperatures of 20, 30, and 40 °C were observed due to the convection and temperature non-uniformity of the measured solutions, it was suggested that the NTC thermistor used for temperature correction was sufficiently sensitive to temperature, as real-time temperature measurements of the solutions were possible.Figure 4d-f presents the maximum errors from the true values in the measured concentrations for each of the five samples.By applying temperature correction, the relative errors (defined as |Measured value−True value| True value × 100%) were reduced from 60% to 7.8% for the Na + sensors and from 76% to 14.6% for the K + sensors, while the absolute errors (defined as |Measured value − True value|) was reduced from 0.88 to 0.19 for the pH sensors.These results implied the critical importance of temperature correction in the context of printed ion sensors in wearable applications.

Conclusion
We integrated ion and temperature sensors on a plastic substrate using printing methods to investigate the effectiveness of temperature correction for the ion sensors toward wearable sweat sensing.The temperature dependence of the open circuit potential of the printed ion sensors was found to exhibit a linear behavior with a slope of 1-2 mV °C−1 in the physiological skin temperature range of 20-40 °C.Applying temperature correction to the ion sensors, the maximum relative errors were reduced from 60% to 7.8% for the Na + sensors and from 76% to 14.6% for the K + sensors, while the absolute error was reduced from 0.88 to 0.19 for the pH sensors, indicating the critical importance of temperature correction as a technology for accurate measurements in wearable applications.Further detailed investigation and modeling of the factors affecting temperature dependence [25] as well as refinement of the temperature correction formula are expected to enable on-body, real-time, and accurate sweat data acquisition in an environment with temperature fluctuations, bringing wearable sweat sensors closer to practicality.

Figure 1 .
Figure 1.Images and schematic illustrations of the wireless ion and temperature sensor devices.a) Photograph of the wireless sensor device.b) Systemlevel block diagram of the wireless sensor device showing signal transduction (orange), conditioning (blue), processing, and wireless transmission (green) paths from sensors to the tablet user interface with custom-developed application.c) Schematic structure of the printed ion and temperature sensor.

Figure 2 .
Figure 2. Responses of the printed sensors.The open circuit potential responses of the a) sodium (Na + ) in NaCl (10-100 mm), b) potassium (K + ) in KCl (1-10 mm), and c) hydrogen ion (pH) sensors in McIlvain buffer solutions (3-8 pH).Data recording was performed continuously.d-f) The corresponding calibration plots of the ion sensors.g) The resistance response of the thermistor to temperature change (20-40 °C) in an electrolyte solution.Data recording was paused for 30 s when each changing to solution with different temperature setting.h) The relative resistance and i) resistance of the thermistor with respect to temperature.

Figure 3 .
Figure 3. Temperature dependence of the printed ion sensors.Measured open circuit potential versus temperature of the a) Na + sensors in NaCl solutions (10, 33, 100 mm), b) K + sensors in KCl solutions (1, 3.3, 10 mm), and c) pH sensors in McIlvaine buffer solution (3, 5, 8 pH).The dashed color lines represent linear fits.d-f) The corresponding sensitivity plots versus temperature of the ion sensors, along with Nernst equation (dashed black line).

Figure 4 .
Figure 4. Measured concentration without and with temperature correction (T-correction).Measured concentration of the a) Na + , b) K + , and c) pH sensors in electrolyte solutions with different concentrations or temperature, along with measured temperatures using the thermistor.Data recording was paused for 30 s only when each changing to a solution with a different temperature setting.The dashed lines represent the true values of the concentration.The maximum relative errors for the d) Na + and e) K + sensors.f) The maximum absolute error for the pH sensors.