A New Approach for Monitoring Sweat Ammonia Levels Using a Ventilated Capsule †

: Ammonium levels in sweat can potentially be used to measure muscle fatigue and to diagnose particular metabolic myopathies. To research the potential use of ammonia in sweat as a biomarker, a new real-time monitoring system is developed. This system consists of a capsule that is placed on the skin and ventilated with dry air. A metal-oxide gas sensor in the capsule detects the ammonia that is evaporated from sweat. The sensor system was built, and calibration experiments were performed. The sensors show good sensitivity from 27 mV/ppm to 1.1 mV/ppm in the desired measurement range of 1 to 30 ppm, respectively. A temperature and humidity sensor is integrated to compensate for temperature and humidity effects on the NH 3 sensor.


Introduction
Wearable sweat sensors offer new opportunities for continuously monitoring an athlete's condition unobtrusively. Sweat rate and sweat Na + and Clconcentrations may provide information about an athlete's hydration status and core temperature [1]. Measuring muscle fatigue would be another interesting aspect of monitoring an athlete's health status and monitoring the efficiency of training programs. To find a relationship between muscle fatigue and sweat constituents, more physiological research is required. Although lactate levels in the blood are a well-known indicator of muscle fatigue [2], it is known that lactate is a by-product of sweat gland metabolism as well [3], and no significant relationship between lactate levels in blood and lactate levels in sweat were found in the literature [4]. As an alternative, several researchers highlighted the possibility of using NH 4 + in sweat as a potential biomarker for muscle fatigue [5,6]. NH 4 + concentrations in sweat are relatively high (between 0.5 and 25 mM [7]). During very intense exercise, adenosine triphosphate (ATP) levels get very low, and inosine monophosphate (IMP) and NH 4 + are produced from adenosine monophosphate (AMP). This is part of the purine nucleotide pathway [8] and results in the accumulation of ammonia in the blood, which possibly results in an increase in NH 4 + concentrations in sweat during exercise as well. To research if NH 4 + levels in sweat are related to blood levels, and if sweat [NH 4 + ] can be used as a biomarker for muscle fatigue, reliable real-time NH 4 + monitoring systems are required. Moreover, monitoring of sweat (NH 4 + ) may have interesting clinical applications. Subjects with particular metabolic myopathies show abnormal concentrations of NH 3 during intense exercise. For example, subjects with myoadenylate deaminase (MAD) deficiency show very low ammonia variations in blood during ischemic exercise [9,10].
In this paper, we present a new way of monitoring NH 4 + levels in sweat to research the potential of the abovementioned applications. Previous research focused on ionic content measurement using potentiometry [6,11]. Since NH 3 (aq) is evaporating quickly, we propose to measure the NH 3 levels evaporated from sweat in a sensing capsule that is placed on the skin. Controlled airflow through the system is necessary to calculate the concentration in the sweat based on the concentration measured in the sensing capsule. Therefore, the sensor is integrated into a ventilated capsule system, which is commonly used by physiologists for measuring sweat rate [12]. This paper focuses on the design and development of the NH 3 measurement system.

Design
It is estimated that a sensor with a detection range between 0 and 100 ppm is required, by using data about NH 4 + in sweat found in the literature [7], sweat rate and volumetric flow rate of air through the capsule. Based on measurement range, price, and ability to integrate this sensor in a wearable, the MICS-5914 semiconductor metal oxide gas sensor (SGX Sensortech, Neuchâtel, Switzerland) was selected. This sensor consists of an Si/SiO 2 substrate with 4 gold-platinum electrodes. Two electrodes are used for heating the sensor and two are for performing the measurement. The metal oxide sensing layer including tungsten nano-plates is placed on top [13]. When the sensor is exposed to NH 3 , the resistance across the sensing electrodes will drop. Next to NH 3 , the sensor is sensitive to ethanol, hydrogen, propane and butane. These gases will not be formed during sweating in general conditions. Figure 1 shows a schematic overview of the entire measurement system. We use a series-series feedback circuit to read out the sensors (Figure 2a). This circuit has a linear response, which facilitates calibration. It also reduces noise, distortion, and variations due to temperature influences. The circuit includes the precision operational amplifier LTC2057 (Analog Devices, Norwood, MA, USA) and the MCP3421 (Microchip, Chandler, AZ, USA) analog to digital converter with differential input is used. These electronics with a single NH 3 sensor and a humidity and temperature sensor (HDC1080, Dallas, Texas Instruments) are placed at a small PCB (34 mm × 28 mm), that is designed to fit in a capsule (Figure 2b,c). The capsule can be taped to the skin of the athlete. The sensors are placed inside the capsule and the other supporting electronics of the readout circuit are placed at the other side of the PCB, outside the capsule. The capsule PCB is connected to a shield that contains the current source for the heater circuit, the connection to a power supply and a multiplexer to switch between different capsules. Data of four sensor capsules can be read out with an STM32F411RE microcontroller board (STMicroelectronics, Geneve, Switzerland) ( Figure 2d). It was chosen to place the sensors as close as possible to the collection surface, to prevent that NH 3 is absorbed by the materials of the capsule and the tubing before the NH 3 molecules are detected. A thin gold layer was sputtered on the inside of the capsule to prevent absorption in this area. The capsule itself is made of a 3D printed flexible photopolymer (Connex 3, Objet 350, Stratasys, Rehovot, Israel) and has a collection surface diameter of 26 mm. The airtight capsule will be placed at the skin with a double-sided acrylate adhesive (3M 1522, [14]) During the in situ experiments, the capsule can be ventilated with dry air at a flow rate between 0.2 and 1.2 L/min, depending on the type of exercise test that will be performed.

Sensor Characterization
Each of the MICS-5914 sensors has different resistance in air and the sensitivity will also differ [15]. Therefore, each individual sensor needs to be calibrated. A calibrated electrochemical NH3 sensor for air quality monitoring (S900, Aeroqual, Auckland, New Zealand) was purchased to calibrate the sensors. The sensor modules and the reference sensor are placed in a closed box with tubing to an Erlenmeyer flask. The air is circulated with a small fan through the glass and through the box. First, the sensor resistance must stabilize before NH3 can be measured reliably. Therefore, the sensors are started up 2 h before the measurement. One sensor is covered with aluminum foil, to keep an eye on the effects of temperature and humidity changes. An ammonia solution is made in the Erlenmeyer flask from household ammonia and demineralized water. With this solution, different NH3 concentration levels can be created in the closed box. Measurements of 30 to 40 min are performed at concentration levels of 3, 5, 10, 20 and 30 ppm NH3.
During sweating, humidity levels in the measurement capsule will increase. These humidity changes will influence the NH3 measurements. To quantify the influence of humidity, the Erlenmeyer flask is removed and beakers with different diameters with water are placed in the setup. In this way, different evaporation surfaces result in different humidity levels in the box.

Sensor Characterization
Each of the MICS-5914 sensors has different resistance in air and the sensitivity will also differ [15]. Therefore, each individual sensor needs to be calibrated. A calibrated electrochemical NH3 sensor for air quality monitoring (S900, Aeroqual, Auckland, New Zealand) was purchased to calibrate the sensors. The sensor modules and the reference sensor are placed in a closed box with tubing to an Erlenmeyer flask. The air is circulated with a small fan through the glass and through the box. First, the sensor resistance must stabilize before NH3 can be measured reliably. Therefore, the sensors are started up 2 h before the measurement. One sensor is covered with aluminum foil, to keep an eye on the effects of temperature and humidity changes. An ammonia solution is made in the Erlenmeyer flask from household ammonia and demineralized water. With this solution, different NH3 concentration levels can be created in the closed box. Measurements of 30 to 40 min are performed at concentration levels of 3, 5, 10, 20 and 30 ppm NH3.
During sweating, humidity levels in the measurement capsule will increase. These humidity changes will influence the NH3 measurements. To quantify the influence of humidity, the Erlenmeyer flask is removed and beakers with different diameters with water are placed in the setup. In this way, different evaporation surfaces result in different humidity levels in the box.

Sensor Characterization
Each of the MICS-5914 sensors has different resistance in air and the sensitivity will also differ [15]. Therefore, each individual sensor needs to be calibrated. A calibrated electrochemical NH 3 sensor for air quality monitoring (S900, Aeroqual, Auckland, New Zealand) was purchased to calibrate the sensors. The sensor modules and the reference sensor are placed in a closed box with tubing to an Erlenmeyer flask. The air is circulated with a small fan through the glass and through the box. First, the sensor resistance must stabilize before NH 3 can be measured reliably. Therefore, the sensors are started up 2 h before the measurement. One sensor is covered with aluminum foil, to keep an eye on the effects of temperature and humidity changes. An ammonia solution is made in the Erlenmeyer flask from household ammonia and demineralized water. With this solution, different NH 3 concentration levels can be created in the closed box. Measurements of 30 to 40 min are performed at concentration levels of 3, 5, 10, 20 and 30 ppm NH 3. During sweating, humidity levels in the measurement capsule will increase. These humidity changes will influence the NH 3 measurements. To quantify the influence of humidity, the Erlenmeyer flask is removed and beakers with different diameters with water are placed in the setup. In this way, different evaporation surfaces result in different humidity levels in the box. Figure 3a shows the raw data of the resistance change of the sensors during the NH 3 experiment. The temperature is 25 • C and the humidity level is 60%. The baseline resistance highly differs for each sensor (from 360 kΩ to 1140 kΩ). Furthermore, the sensor readings show a transient behavior. In the literature, it was found that metal oxide gas sensors have a fast response time compared to other types of gas sensors [16], and in experiments with ethylene, the MICS-5914 has a sorption and desorption time constant of τ s = 29 s and τ d = 43 s [13]. On top of that, the reference sensor shows a very similar transient response (Figure 3b). It is expected that the transient is mainly originating from the rising NH 3 concentration in air, since the NH 3 is evaporating from the flask, and it will take some time before an equilibrium at a certain concentration is reached. The small peaks between each concentration level originate from the moment that we change the flask and box is circulating air from the environment for a few minutes. During this moment, NH 3 and humidity levels will decrease, and sensor resistance will increase. When a new flask with an NH 3 solution is connected, the resistance will immediately decrease again.  Figure 3a shows the raw data of the resistance change of the sensors during the NH3 experiment. The temperature is 25 °C and the humidity level is 60%. The baseline resistance highly differs for each sensor (from 360 kΩ to 1140 kΩ). Furthermore, the sensor readings show a transient behavior. In the literature, it was found that metal oxide gas sensors have a fast response time compared to other types of gas sensors [16], and in experiments with ethylene, the MICS-5914 has a sorption and desorption time constant of τs = 29 s and τd = 43 s [13]. On top of that, the reference sensor shows a very similar transient response (Figure 3b). It is expected that the transient is mainly originating from the rising NH3 concentration in air, since the NH3 is evaporating from the flask, and it will take some time before an equilibrium at a certain concentration is reached. The small peaks between each concentration level originate from the moment that we change the flask and box is circulating air from the environment for a few minutes. During this moment, NH3 and humidity levels will decrease, and sensor resistance will increase. When a new flask with an NH3 solution is connected, the resistance will immediately decrease again. A calibration plot can be made using data from the MICS-5914 sensors (baseline resistance, R0 and sensor resistance, Rs) and the reference sensor measurements. This calibration plot is shown in Figure 4. A clear difference in sensitivity between sensors can be seen, which emphasizes that each individual sensor needs to be calibrated. In all experiments, it was found that the baseline sensor resistance drifts over time. This drift varies in each measurement. In further experiments, these measurements will be repeated multiple times, to test the stability of the sensors during NH3 measurements. A sensor with excellent stability would be beneficial if physiologists and clinicians want to know the absolute concentration levels in sweat. However, in studies about the relationship between NH3 and muscle fatigue, measuring relative changes in NH3 would already give valuable information. A calibration plot can be made using data from the MICS-5914 sensors (baseline resistance, R 0 and sensor resistance, R s ) and the reference sensor measurements. This calibration plot is shown in Figure 4. A clear difference in sensitivity between sensors can be seen, which emphasizes that each individual sensor needs to be calibrated. In all experiments, it was found that the baseline sensor resistance drifts over time. This drift varies in each measurement. In further experiments, these measurements will be repeated multiple times, to test the stability of the sensors during NH 3 measurements. A sensor with excellent stability would be beneficial if physiologists and clinicians want to know the absolute concentration levels in sweat. However, in studies about the relationship between NH 3 and muscle fatigue, measuring relative changes in NH 3 would already give valuable information.

Results and Discussion
When exposing the sensor to a changing humidity level from 38% to 50%, a clear drop in resistance, which varies between sensors (from 25% to 60%), can be seen. These changes are comparable to the changes measured by Krivec et al. [13]. Water molecules can be adsorbed at the sensor surface in a similar way as the volatile gas. These molecules reduce the baseline resistance and the sensitivity of the sensor because the available surface area for adsorption of NH 3 molecules decreases. The humidity measurements show the necessity of compensating for humidity influences during sweat NH 3 measurements. A preliminary physiological experiment was executed to research the environmental conditions (temperature and humidity) in the capsule during sweat experiments because sweat rate and skin temperature will influence humidity and temperature conditions. A hydrophobic PTFE membrane was placed between the sensor and the capsule opening, to protect the sensor against sweat droplets. The method and results can be found in the supplementary information. It was found that in this test, humidity levels increased from 8% to 30%, with increasing sweat rate. In this experiment, a dry airflow rate of 1.2 L/min was used to ensure that all sweat underneath the capsule evaporated. In future experiments, the flow rate needs to be well tuned because a balance needs to be found between having a well detectable [NH 3 ] in the capsule and minimizing humidity influences on the sensor resistance. The temperature underneath the capsule ranged between 34.1 • C and 36.3 • C. The varying temperature and humidity conditions confirm that it is important to integrate a temperature and humidity sensor in each capsule. When exposing the sensor to a changing humidity level from 38% to 50%, a clear drop in resistance, which varies between sensors (from 25% to 60%), can be seen. These changes are comparable to the changes measured by Krivec et al. [13]. Water molecules can be adsorbed at the sensor surface in a similar way as the volatile gas. These molecules reduce the baseline resistance and the sensitivity of the sensor because the available surface area for adsorption of NH3 molecules decreases. The humidity measurements show the necessity of compensating for humidity influences during sweat NH3 measurements. A preliminary physiological experiment was executed to research the environmental conditions (temperature and humidity) in the capsule during sweat experiments because sweat rate and skin temperature will influence humidity and temperature conditions. A hydrophobic PTFE membrane was placed between the sensor and the capsule opening, to protect the sensor against sweat droplets. The method and results can be found in the supplementary information. It was found that in this test, humidity levels increased from 8% to 30%, with increasing sweat rate. In this experiment, a dry airflow rate of 1.2 L/min was used to ensure that all sweat underneath the capsule evaporated. In future experiments, the flow rate needs to be well tuned because a balance needs to be found between having a well detectable [NH3] in the capsule and minimizing humidity influences on the sensor resistance. The temperature underneath the capsule ranged between 34.1 °C and 36.3 °C. The varying temperature and humidity conditions confirm that it is important to integrate a temperature and humidity sensor in each capsule.
After quantifying temperature and humidity effects in the expected conditions in the lab, the sensors will be tested in larger physiological experiments in which athletes will perform a high-intensity cycling exercise. In these experiments, we will validate the use of our NH3 ventilated capsule system for NH3 sweat measurement. Trends in NH3 levels in the capsule during exercise at a cycling ergometer will be researched. Based on the insights from these experiments, further improvements to the system will be made.

Conclusions
In this research, a new approach to measure NH3 levels in sweat is presented. A metal-oxide gas sensor is placed in a small ventilated capsule that will be placed on the skin. The sensor measures the NH3 molecules that evaporate from the sweat. At the same Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
Datasets are available from the corresponding author on request.