A Novel Real-time Carbon Dioxide Analyzer for Health and Environmental Applications

Abstract

To be able to detect carbon dioxide (CO₂) with high accuracy and fast response time is critical for many health and environmental applications. We report on a pocket-sized CO₂ sensor for real-time analysis of end-tidal CO_2, and environmental CO₂. The sensor shows fast and reversible response to CO₂ over a wide concentration range, covering the needs of both environmental and health applications. It is also immune to the presence of various interfering gases in ambient or expired air. Furthermore, the sensor has been used for real-time breath analysis, and the results are in good agreement with those from a commercial CO₂ detector.

1. INTRODUCTION

A capability that can detect carbon dioxide (CO₂) with high accuracy and with fast response time is critical for many health and environmental applications [1–15]. For example, measuring CO₂ levels in breath at the end of expiration, known as end-tidal CO₂ (EtCO₂), allows for non-invasive evaluation of systemic metabolism, perfusion, ventilation, and cardiac output, which provides doctors and patients with a non-invasive method to diagnose asthma, Chronic Obstructive Pulmonary Disease (COPD), and cardiovascular diseases [5, 6]. Similarly, monitoring of indoor CO₂ levels allows for assessment of indoor air quality (IAQ). Higher levels of indoor CO₂ are associated with increased prevalence of certain mucous membrane and sick building syndrome (SBS) symptoms [15]. Infrared detection technology is currently utilized for measuring CO₂ in breath and in air. While useful, this technology experiences strong interference from humidity that is present both in breath and in air. Moreover, the infrared approach requires special sample pretreatments in order to reduce the humidity, which further adds to the cost of the device technology and limits its usefulness for applications in clinical settings. In the case of indoor environmental CO₂ sensing, the use of infrared technology is hampered by the interference of environmental humidity making detection of CO₂ levels inaccurate. There is a need, therefore, for developing a compact, low-cost, easy-to-use, and accurate CO₂ sensor for tracking CO₂ in human breath and for monitoring indoor air quality [16 – 17].

An alternative to infrared sensing is a detecting method based on colorimetry, which identifies CO₂ based on the change of color of a pH-sensitive indicator [18 – 23]. Compared to the infrared CO₂ sensors, the colorimetric approach has several potential advantages, including simplicity, miniaturization, low cost, and immunity to humidity changes, thereby making colorometric sensors an attractive technology. While these sensors may show great promise, their response and recovery time are too slow for breath-by-breath analysis, and their detection limits and reversibility are insufficient to ensure accurate detection of CO₂ in the environment. In order to solve these problems, many attempts, including pre-treatment of the sensing materials have been undertaken to reduce cost and ensure high performance of colorimetric CO₂ sensors. These improvement activities, however, often result in more complex instruments and sensor preparation methods.

In a previous publication, we introduced a portable breath analyzer for the determination of the expired CO₂. The device features a colorimetric sensor that could analyze breath CO₂ concentration accurately, and a fluidic system for efficient delivery of breath sample to the sensing element. A 3D model was created to simulate the sample flow and reaction of CO₂ with the sensing materials and color changes associated with the chemical reactions [24]. Despite the success, the sensor response was slow and semi-reversible, which is not suitable for breath-by-breath CO₂ analysis as needed for capnography, and moreover, which is insensitive to real-time monitoring of CO₂ in air. In the present work, we report a CO₂ sensor with a response time as fast as ~0.1 sec., which enables breath-by-breath analysis (see more details in Experimental and Results Section). Additionally, the sensor has a wide dynamic range (with the CO₂ concentration up to 11.5%), and low detection limit of a few tens of ppm, which are suitable for indoor air quality monitoring.

2. EXPERIMENTAL

2.1. Reagents and sensor preparation

The colorimetric CO₂ sensor presented in this paper was prepared by coating a sensor chip with a solution containing HCO₃⁻/CO₃²⁻ buffer and m-cresol purple as the sensing element [19, 24]. All the reagents used in this work were analytical grade and purchased from Sigma-Aldrich (St. Louis, MO, USA). As described in our previous work, when an ambient air sample or warm breath sample was brought into contact with the sensor chip, the pH of the sensing material decreased, which led to a color change [24]. This color change was detected with an optoelectronic detection system (see details below).

Note that several differences are stayed in the present work with respect to previous work [24]. Regarding the sensor: a- The sensor substrate uses a fluorinated hydrophobic membrane instead of transparent polyethylene-based plastic; b- The hydrophobic membrane has microstructures and surface properties that provide faster water desorption and significantly improved sensor time response; c- The new indicator, m-cresol purple, provides a wider dynamic range of CO₂ % concentration due to its lower pka. Compared with thymol blue [24], m-cresol purple changes color more sensitively under weakly alkaline to neutral conditions (pH value: 9.0 – 6.0), which typically corresponds to the range of normal breath CO₂ levels (1.0% – 11.5%) [25 – 27] (Fig. 1).

Figure 1.

Titration curves for the HCO₃⁻/CO₃²⁻ buffer system modified with thymol blue and m-cresol purple. The color changes in the CO₂ sensor were due to the pKa value (for thymol blue: color changes from blue (pH > pKa) to yellow (pH < pKa); for m-cresol purple: color changes from purple (pH > pKa) to yellow (pH < pKa)). Compared with thymol blue, m-cresol purple changes color more sensitively under neutral to weakly alkaline conditions, which provides a larger dynamic range for the breath CO₂ detection.

Regarding the device: a- For breath applications, the gas flow is designed that both exhalation of breath samples, as well as inhalation of clean air are enabled through the device. The later enables in-situ sensor regeneration; b- The device’s gas flow component has larger diameter (~20 mm) than the device published in Ref. 24 (few mm), which significantly reduces the resistance to breathing, enabling breath-by-breath analysis.

2.2. Device description

The CO₂ sensor chip was inserted into a detection chamber, which included an optoelectronic detection system consisting of a red LED (wavelength = 633 nm, LEDtronics. Inc., CA, USA) as a light source and a photodiode (OSRM GmbH, Germany) as a light detector (Fig. 2a). The response of the sensor was characterized by measuring the change in the intensity of transmitted light caused by the interaction of CO₂ with the sensing element (Fig. 2b). Gas sample, from either the ambient air or a flow containing simulated breath was directed to the sensor detection chamber with a pump. Real breath test was also conducted by asking a volunteer to breathe into the device via a mouthpiece. The device also contains an electronic circuit, which collects and processes the data from the optoelectronic detection system, and then wirelessly sends the data to a smartphone via a Bluetooth chip.

Figure 2.

(a) Pocket-sized CO₂ sensor device. It has inlets and outlets that allow the detection and analysis of both breath and environmental air samples. The key-sensing element is a CO₂ sensor chip, which can be inserted into or remove from the device. The device also features an electronic circuit for signal processing and Bluetooth for wireless transmission of data. (b) The sensor chip changes color from purple to yellow after exposure to CO₂ and the concentration of CO₂ is determined by measuring the change in intensity of transmitted light for the sensing element.

2.3. Device characterization and validation

Simulated breath samples

In order to investigate the response of the CO₂ sensor to real breath, a humidified CO₂ gas mixture was employed to simulate the expired air. The simulated expired breath samples were prepared by first mixing 80% N₂ + 20% O₂ air with different amounts of CO₂ (Praxair. Inc.), ranging from 0.03% to 11.5%. The CO₂ mixtures were then pumped through a sealed water system immersed in a thermostatic water bath (Thermo Scientific, USA) at 35°C to generate 35°C and 100% relative humidity. The simulated expired gas samples and ambient air were then introduced into the breath analyzer alternately at a flow rate of 6L/min to simulate expiration and inspiration processes as shown in Fig. 3a.

Figure 3.

Schematic diagram of calibration setup for investigating the response of the CO₂ sensor to (a) breath samples and (b) trace CO₂ in air gas samples. (c) Schematic diagram of the laboratory setup for validating the performance of the CO₂ sensor using real breath samples from volunteers and a commercial breath CO₂ analyzer.

Simulated environmental samples

For environmental CO₂ analysis, the CO₂ gas mixtures were prepared by mixing ultrahigh purity air with CO₂ to simulate indoor air with CO₂ concentration ranging from 0 ppm to 1350 ppm. The simulated indoor air samples and ultrahigh purity air were then alternately introduced into the CO₂ analyzer at a flow rate of 6 L/min as shown in Fig. 3b.

Sensor signal measurement

The change in light intensity of the CO₂ sensor was used to characterize the color change upon exposure to alternating sampling and purging periods of fixed time. The change in light intensity of the CO₂ sensor is evaluated as sensor signal from the baseline signal. This means that baseline shifts are corrected, which renders relatively small sensor response dispersion error at a given CO₂ concentration (see below for more details). The change in transmitted light of the CO₂ sensor is:

$$\mathrm{\Delta }I(t)=I(t)-I(0),$$

where I(0) is the light intensity prior to the exposure of the sensor surface to the gas samples, I(t) is the light intensity at time t during the sampling process. In addition, the performance of this device was further validated using a commercial breath CO₂ analyzer (capnography analyzer from VacuMed, CA) and real breath samples from volunteers (Fig. 3c).

Cross-sensitivity analysis

The interference of other chemicals present in expired air and atmosphere, such as ethanol, acetone, acetonitrile, and NH₃ was investigated by introducing the humidified gas mixtures containing pure N₂ and the interfering gases into the CO₂ analyzer. The response of the CO₂ sensor exposed to the interfering gases was compared with the response of the sensor exposed to 1% CO₂ gas mixture.

3. RESULTS AND DISCUSSION

3.1. Sensing mechanism of the CO₂ sensor

The basic sensing principle of the new CO₂ sensor is based on the adsorption and desorption processes of CO₂ in the sensor during the test [24]. Briefly, during the sensing process, as gas samples (either expired breath or environmental air) were introduced into the device, CO₂ was adsorbed by the HCO₃⁻/CO₃²⁻ buffer system in the sensor, and the pH of the sensing system decreased. However, during the purging process, the clean air flowed over the sensor, and the CO₂ molecules were dissociated from the sensor, resulting in the recovery of pH to the initial value. As described in Section 2.1, m-cresol purple was used as the pH indicator of the sensing system. Since the variation of pH in the sensor was correlated to the concentration of CO₂ in the gas flow, the concentration of CO₂ in either the expired breath or environmental air can be determined via monitoring the color change of m-cresol purple (mCP). The chemical reaction process is as follows [28, 29]:

$${CO}_2(g)\underset{{k_0}^{\prime }}{\overset{k_0}{\rightleftarrows }}{CO}_2(aq)$$ (R1)

$${\mathit{mCP}}^{2-}(\mathit{purple})+H_{2}O+{CO}_2(aq)\rightleftarrows {\mathit{mCP}}^{2-}(\mathit{purple})+H_2{CO}_3\rightleftarrows {\mathit{mCPH}}^{-}(\mathit{yellow})+{\mathit{HCO}}_3^{-}$$ (R2)

where CO₂(aq) is the dissolved CO₂ in the sensor, mCP²⁻ is the deprotonated form of m-cresol purple and mCPH⁻ is the protonated form of m-cresol purple in the HCO₃⁻/CO₃²⁻ buffer solution.

3.2. Characterization of the CO₂ sensor

Breath CO₂ analysis calibration

Fig. 4a shows the response of the CO₂ sensor exposed to the alternating atmospheres of dry air and humidified 5% CO₂ gas mixture. As described earlier, when the CO₂ gas sample was introduced into the device, water vapor condensed onto the sensor surface and CO₂ was adsorbed by the HCO₃⁻/CO₃²⁻ buffer solution in the sensor, which decreased the pH level of the sensing system, and changed the color of the pH sensing probe from purple to yellow (Fig. 2b). This color change was measured as a change in the transmitted light intensity, which was correlated to the CO₂ concentration in the gas sample. Conversely, when CO₂-free dry air was introduced into the device to simulate the inspiration process, CO₂ was stripped from the sensor and the color changed back from yellow to purple, which was also recorded by the light intensity change. The CO₂ concentration in the simulated breath sample can be determined from the maximum intensity change in each simulated breathing cycle. It is worthy to notice, the sensor signal noise at the maximum response level is generated by a current inability of the analog-to-digital converter of the microcontroller used in the detection system to smoothly manage small voltage signal changes. The problem is currently being corrected from hardware design standpoint. In addition, Fig. 4b shows the maximum light intensity change vs. the CO₂ concentration in the gas samples. It can be observed that the maximum change of light intensity increases with CO₂ concentration from 0.03% to 11.5% with a Langmuir-like behavior (R² = 0.9986). This concentration range sufficiently covers the need of breath CO₂ analysis for medical applications, which typically ranges between 2.5 and 5.0% in healthy individuals [30].

Figure 4.

(a) Response of the CO₂ sensor exposed to the alternating atmospheres of dry air and artificial expired air containing 5 % CO₂. The light intensity increases rapidly when the sensor was exposed to CO₂-containing atmosphere and returns to the initial value rapidly when the exposed to CO₂-free atmosphere. (b) The relationship between the maximum intensity change (indicated by ΔV, the output of the photodiode) and CO₂ concentration. The light intensity increased with the CO₂ concentration. Response vs. concentration to Langmuir-like equation is shown in the figure insert.

Environmental CO₂ analysis calibration

Fig. 5a shows the response of the CO₂ sensor exposed to the alternating atmospheres of ultrahigh purity air and simulated environmental air samples. The light intensity increased as CO₂ interacted with the sensing system and reduced gradually as pure air passed through the device. Fig. 5b shows the relationship between the maximum change of light intensity in each sampling-purging cycle and CO₂ concentration. The intensity change increased with CO₂ concentration with a Langmuir-like behavior (R² = 0.9965). The detectable range of the sensor to the ambient CO₂ was found from few tens of ppm to few thousand ppm levels, which covers the needs for environmental CO₂ detection [31].

Figure 5.

Application of CO₂ sensor for environmental sensing: (a) Response of the CO₂ sensor exposed to the alternating atmospheres of simulated environmental air samples. (b) The light intensity changes increases with CO₂ concentration in the simulated environmental air sample. Response vs. concentration fitting to Langmuir-like equation is shown in the figure insert.

3.3. Cross-sensitivity study

In order to investigate the interference effects of other chemicals that might exist in the expired breath and ambient air, humidified gas mixtures containing pure N₂ and different interfering gases were introduced into the CO₂ device at a flow rate of 6L/min for 30s. As shown in Table 1, compared to the response of the sensor exposed to a 1% CO₂ gas sample, the interference of the other gases is within 12%, which can be considered negligible in breath CO₂ analysis as the CO₂ levels in breath typically ranges from 3.5 to 5.5 %. However, interference from chemicals such as NH₃ needs to be corrected when the CO₂ sensor is used for environmental CO₂ monitoring.

Table 1.

Cross-sensitivity of CO₂ sensor for other gases in the expired and environmental air

Gas	ΔImax (V)	Maximum Interference (%)
1% CO2	0.170	/
Pure N2	0.010	5.8
165 ppm ethanol	0.014	8.2
20 ppm acetone	0.013	7.6
21 % O2	0.017	10.0
250 ppb NH3	0.017	10.0
100 ppb acetonitrile	0.019	11.2

In addition to the above-mentioned studies, the effect of ambient humidity on breath CO₂ analysis was evaluated. For relative humidity ranging between 20 and 85 %, the sensor response to the expired CO₂ remained constant within 10% variation (not shown). Some interference effect was detected for the relative humidity levels higher than 85%. However, the responses of the sensor to humid air and humid air with CO₂ mixture (e.g. simulated breath at 35°C, and 100% humidity) are different. Therefore the effect of extremely high ambient humidity could be corrected by use of an internal humidity sensor in the CO₂ analyzer.

3.4 Real breath analysis

In order to investigate the application of the CO₂ sensor to real breath analysis, including breath patterns and ventilation rate, the CO₂ sensor was calibrated using breath samples in combination with flow rate measurements from an in-situ flow meter. The breath patterns of two volunteers were also analyzed. The volunteers included one healthy person and one asthma patient from the VA Medical Center, Phoenix, AZ. They were asked to breathe normally through the mouthpiece. The end-tidal CO₂ values (the maximal concentration of CO₂ at the end of expiration) were calculated and compared with the values assessed from a capnogram recorded with a commercial breath CO₂ analyzer (Fig. 6). The “capnogram” refers to a real-time waveform of CO₂ concentration in the process of respiration with gas samples taken from a sidestream port located a few millimeters away from the colorimetric sensor as shown in Fig. 3c. From Fig. 6a and 6b, it can be seen that the end-tidal CO₂ values measured by the new CO₂ analyzer were comparable to the values assessed by the commercial analyzer for both the healthy person and asthma patient. In addition, the capnogram obtained from the new CO₂ analyzer also showed good agreement with the commercial device. The 90% response time of the CO₂ analyzer was less than ~150 ms, which meets the requirements for breath-by-breath CO₂ analysis (< 200 – 300 ms). Furthermore, the new CO₂ analyzer showed faster response than the commercial analyzer, which had a fraction of a second delay in gas detection due to the aspiration of the gas sample from the sample site through the sampling tube (around 3 ft. long) and into its detection unit (infrared detection chamber) (Fig. 3c). It was evident that time response improvements were achieved in the new CO₂ analyzer due to the in-situ location of the sensor chip in the mainstream of the breath sample flow.

Figure 6.

Comparison of the CO₂ sensor by this work and a commercial infrared CO₂ analyzer. Both sensors were exposed to real breath samples from (a) a healthy person and (b) an asthma patient. The capnograms obtained from the present CO₂ sensor and the commercial device are in good agreement with each other.

4. CONCLUSIONS

This paper describes the development of a pocket-sized CO₂ sensor for personal healthcare and environmental monitoring. The device has an integrated circuit and an optical-based detection chamber, which includes a colorimetric CO₂ sensor. The CO₂ sensor shows reversible response to CO₂ and a dynamic detection range from ~50 ppm to 11.5%. When compared to a commercial infrared capnography detector, the CO₂ analyzer shows fast response with a 90% response time of less than 150ms, which indicates that the CO₂ analyzer is suitable for breath-by-breath analysis. In addition, the CO₂ sensor shows accurate detection of CO₂ levels after non-linear calibration of the sensor response. The interference study indicates that other chemicals typically present in the expired breath and ambient air have negligible effects on the response of the CO₂ sensor. Compared with traditional CO₂ measurement technology, the new pocket-sized CO₂ analyzer allows for simple, fast, and accurate assessment of breath CO₂ levels, and breath CO₂ profiles for patients with COPD and asthma, as well as CO₂ levels in ambient air.

Highlights.

• A novel real-time carbon dioxide (CO₂) sensor for health and environmental applications is developed.

• The sensor shows fast and reversible response to CO₂ over a wide concentration range.

• The sensor is immune to the presence of various interfering gases in ambient or expired air.

• The sensor can be used for real-time breath analysis, and the results are in good agreement with those from a commercial CO₂ detector.

Biographies

Di Zhao received her BS and MS degree in Materials Science and Engineering from Tongji University in 2006 and 2009, respectively. She is currently working towards her PhD in Chemical Engineering at Arizona State University under the supervision of Dr. Erica Forzani. Her research interests focus on novel colorimetric sensors for the detection of gas and biomarkers in human breath.

Dylan Miller is a Graduate Research Assistant in the Center for Bioelectronics and Biosensors, The Biodesign Institute, at ASU. He received his BS and MS degree in Mechanical Engineering at Arizona State University. His current research is focused on the mechanical design of wireless sensor devices with applications in mobile health and environmental safety.

Xiaojun Xian is Assistant Research Scientist of Center for Bioelectronics and Biosensors, The Biodesign Institute, at ASU. He received his PhD and BS in Physical Chemistry in Peking University, China. He joined Arizona State University in 2009 as Postdoctoral Research Associate and then worked as Assistant Research Scientist. His current research interests include development of mobile health sensor for chronic diseases management and environmental chemical sensor for personal exposure assessment.

Francis Tsow is Assistant Research Professor of the Center of Bioelectronics and Biosensors, the Biodesign Institute, at Arizona State University (ASU). He received his PhD in Electrical Engineering at Arizona State University. His current research focus on novel science and technology particularly those related to sensing applications.

Erica Forzani is Assistant Professor of the School for Engineering of Matter, Transport and Energy, Ira A Fulton Schools of Engineering at Arizona State University (ASU); and Deputy Director of Center for Bioelectronics and Biosensors, at the Biodesign Institute, ASU. She received her PhD in Chemistry and BS in Clinical Chemistry and Biochemistry in Cordoba National University, Argentina. She joined Arizona State University in 2003 as research associate of the Department of Electrical Engineering, where she later worked as Assistant Research Professor. Her current research is focused on science and technology of novel sensors and integration of sensors into wireless, non-invasive and inexpensive sensor devices with applications in mobile health, and environmental health & safety.