E.Co.Tech-electrochemical handheld breathalyzer COVID sensing technology

The organization of this section is as follows: (1) COMSOL™ simulation explaining the rationale behind the use of interdigitated electrode design, (2) Computational study to select EMIM[BF₄] as a suitable candidate for NO sensing, (3) Characterizing the NO binding interaction with EMIM[BF₄] for electrochemical gas sensing, (4) Translatability of the RTIL–NO interaction toward low power portable microelectronic prototype development, (5) Validation of the low power portable prototype for real-time NO detection in a clinical setting.

COMSOL™ simulation explaining the rationale behind using the interdigitated electrode design

For electrochemical sensing application, we use an interdigitated electrode design (Fig. 1A,B) to develop a planar capacitive sensor using RTIL as the transducer. An interdigitated electrode (IDE) offers advantage for gas sensing as it allows increased signal response due to electric field confinement. It helps in capturing the change in dielectric permittivity upon diffusion of the target gas. Moreover, gold is appropriate as the electrode material since it is electrochemically stable and possesses chemical inertness. We performed COMSOL (https://www.comsol.com/) simulation to replicate the electrode-electrolyte interface. Electrical parameters are applied to both the working electrode and the reference electrode (Fig. 1C,D). A constant AC potential of 10 mV with a DC bias is applied to the WE with respect to the RE to obtain enhanced sensing performance; the RE is grounded/insulated. The electric fields are confined within the RTIL–electrode interface boundaries. The electrolyte potential is maximum at the digits of the WE and varies from 0.01 to 0.001 V moving from WE to the RE in a smooth gradient manner for an input bias of 10 mV. Maximum current density is observed at the WE which influences the output response of the system. The surface current densities (Fig. S1) contributed by the electrode-electrolyte interface is dominant around the WE and equated to 289.83 A/m². This surface decays with a transient angle from WE to RE (as expected for a two-electrode system with no additional counter electrode) with no parasitic currents contributed by the electrodes going from WE to RE.

Computational study to select EMIM[BF₄] as a suitable candidate for NO sensing

A computational study has been done to visualize the interaction of the RTIL and NO. For this purpose, the structure of the RTIL is optimized using Hartree Fock, having a basis set of 6–31 g (d). The optimized structure of EMIM[BF₄] is depicted in Fig. S2. The optimized structure of EMIM[BF₄] suggests that there is a strong ionic interaction between EMIM⁺ and BF₄⁻ which makes this species stable. Then the RTIL NO interaction has been visualized by optimizing EMIM[BF₄] and NO. The optimized structure of EMIM[BF₄]-NO has been depicted in Fig. 2A. The result suggests the presence of strong non-covalent interaction between the O atom of NO and C3 and N1 of the imidazolium ring, which implies that the RTIL has strong affinity towards NO. As EMIM[BF₄] is an ionic species present in zwitter ionic form, it has the ability to interact with a variety of species having positive or negative polarity. We have also calculated the HOMO–LUMO energy of EMIM[BF₄], NO and EMIM[BF₄]-NO, depicted in Table 1.

We have calculated the HOMO–LUMO energy gap of RTIL and NO.

From the calculation, it has been found that the HOMO–LUMO energy gap has been substantially reduced (~ 30%) after the formation of EMIM[BF₄]-NO. This suggests that the interaction is feasible. We have also pulled out the thermodynamics data which shows that the electronic + thermal free energy of EMIM[BF₄]-NO is − 0.440047 Hartree whereas the electronic + thermal free energy of EMIM[BF₄] is − 0.421756 Hartree, which suggests the formation of EMIM[BF4]-NO is indeed thermodynamically feasible too.

We have also obtained the theoretical FTIR data of EMIM[BF₄]-NO and EMIM[BF₄] and the result is presented in Fig. 2B. The result shows standard FTIR peaks of EMIM[BF₄] at 841 cm⁻¹ depicting C–H bending, 1215 cm⁻¹ C–H stretching. Most other peaks are related to standard C–H stretching and bending, EMIM ring stretching etc. A substantial peak shift has been observed for EMIM[BF₄]-NO which suggests substantial interaction is taking place. This result leads us to carry forward our sensing mechanism to the next level.

Characterizing the NO binding interaction with EMIM[BF₄] for electrochemical gas sensing

RTILs are a new class of materials that can be utilized as a low-power, easy maintenance solution to develop a portable gas sensing system. RTIL’s are solvent free electrolytes consisting of cation/anion pairs. RTILs possess unique physicochemical properties such as high ionic conductivity, low volatility, high thermal stability, and wide electrochemical window which are advantageous from the perspective of gas sensing. All the electrochemical experiments have been carried out using an optimized DC potential of + 1.0 V. Experiments were performed under controlled N₂ flow. CA has been performed at three different potentials and chronoamperogram were plotted at 5 s for three replicates. We have optimized the operating potential to be + 1.0 V and time window to be 5 s as the interquartile range of variation is small and data is consistent for all replicates (Fig. S3).

Calibrated dose response is the most important characteristic which depicts the performance of the sensor. Dose-dependent response was studied using five different variations of NO concentration for the developed device, having the sensor attached. NO has been mixed with N₂ volumetrically to obtain four NO dilutions—50 ppb, 100 ppb, 150 ppb and 250 ppb. We have selected this sensing range according to the exhaled nitric oxide levels and their correlation to the underlying respiratory inflammation³⁵. Chronoamperometry is used to characterize the diffusion limited behavior of the RTIL modified electrode system. The potential was applied at 1 V for 30 s and the current at 5 s was extracted and plotted in the inset of Fig. 3.

The modified electrode–electrolyte interface acts as a semi-permeable layer, which creates a concentration gradient at the interface to allow easy diffusion of the target gas. The equation for Fick’s law of diffusion is as follows:

where J is the diffusion flux, D is the diffusion coefficient, x is the position, and C is the concentration.

For studying the diffusion characteristics using chronoamperometry as the transduction principle, we use the Cottrell equation as follows:

where, i is the current due to diffusion, (c_{j}^{0}) is the concentration of the diffused species, and t is time.

Chronoamperometric measurement records the cathodic current generated due to the diffusion of the NO molecules on the electrode–electrolyte interface. The transient cathodic diffusion current at 5 s with increasing concentration has been plotted in the inset. This transient current depicts the dynamic gas diffusion phenomenon across the electrode–electrolyte interface before the system reaches a steady state. Logarithmic scale fitting (log 3P1) was performed for the calibrated dose response and an R² value of 0.96 was obtained for the entire concentration range.

Evaluation of EMIM[BF₄] @IDE modified sensor specificity and selectivity for NO sensing in the presence of cross-reactive gases

We evaluated the selectivity and specificity of the modified electrode–electrolyte interface toward electrochemical sensing of NO over other environmental gases and VOCs that are present as cross-reactive molecules. The selectivity performance of NO at 100 ppb over nitrogen, carbon dioxide, methanol, and acetone are shown in Fig. 4. The fabricated sensor platform displays a distinguishable and specific CA response for the detection of NO. The average transient diffusion current at 5 s and 1 V was plotted, and it was observed that the average nonspecific signal for cross-reactive gases and vapors was 100 nA, whereas the specific signal for the detection of NO was 700 nA, which was more than ~ 3 times larger than the nonspecific signal.

Translatability of the RTIL–NO interaction toward low power portable microelectronic prototype development

COVID-19 is a humanitarian crisis. Governments, healthcare systems, regulatory agencies, research institutes, and industry players are working together to understand and address the challenge, support victims and their families and communities, and search for treatments and a vaccine. We believe that having an effective and rapid “decision support” device to complement existing diagnostic tests would enable much better containment of the virus. In consideration of these points, we developed a portable NO sensing prototype device in the form of a breath analyzer (see Supplementary Video SV1). The device leverages two microcontrollers that work collaboratively—one acting as a master controller to coordinate driver functions of the device and the other as a slave controller performing the aforementioned electrochemical analysis. The slave microcontroller performs computation on data received from the end-effector, which in this case is a sensor with an IDE coated with RTIL measuring EMIM[BF₄] as the target analyte (Scheme 2). The top end of the device has a single-use mouthpiece that is discarded after every subject breathes into the device. The device contains a slot near the top designed for a single-use mouthpiece that is discarded after every subject breathes into the device. The user initiates the electrochemical analysis sequence by pressing a “Test” button on the device itself. Upon initiating this sequence, the user is instructed to wait while the device begins a Chronoamperometry sequence in the manner described above and assesses baseline values of the target analyte as it exists in the current ambient environment. This sequence can be denoted as the “baseline sequence” and its results are stored locally for reference. When the baseline sequence terminates, the user is instructed to exhale into the device via the mouthpiece. Simultaneously, the device begins another identical Chronoamperometry sequence that assesses values of the target analyte over the interpreted breath as it passes over the sensor. We denote this sequence as the “stimulus sequence”. Six seconds are allotted for the breathing window over this stimulus sequence, but only three seconds are necessary in order to obtain an adequate result. Upon completion of the stimulus sequence, several algorithms present within the device firmware extract the 6-s diffusion current value from the stimulus sequence and compare it to the results similarly obtained by the baseline sequence. Note that, during both the baseline and stimulus sequences, several other sensors perform local measurements of the environment around the sensor IDE. Their data are also contributors to the final test readings, and we discuss these other sensors in the following section. At this point, the assessment is considered complete, but an optional step may be employed to wirelessly transmit assessment results to a database for storage or to a mobile application for easier visualization.

In this study, we evaluated the metabolic signals associated with coronavirus (SARS COV-2) such as up regulation of NO levels in breath. 84 human subjects have been tested, both in different use settings and within the infection cycle. It was found that the breath analyzer is superior to other techniques used for screening purposes. We believe that diagnostics tests do have an important role as “positive confirmatory test” and/or “negative confirmatory test when the clinical symptoms are not aligned with screening test outcome”. The signal obtained is first depicted as raw current value. Figure 5 represents the signal recorded by the breath analyzer device for all the patients. The graph depicts maximum measured current (pA) obtained during the breathing cycle. The subjects tested negative and positive are significantly different (p < 0.01).

Moreover, the data obtained for the breath sample collected from 84 human subjects was plotted as ratiometric sensor output where the ratio is calculated for the patient signal response with respect to the baseline (Fig. 6). A positive ratio provides strong evidence to support a COVID-positive subject while a negative ratio provides strong evidence to support the absence of the disease. A positive signal is the current value is greater than the threshold value set, therefore has a positive change in current. On the other hand, negative signal is the current value that is below the threshold set for the sensing NO, therefore has a negative change in current.

The breath analyzer determines a positive or negative result through a combination of weighted parameters, including both measured electrical responses, sample pressure and humidity. On the surface, with all other parameters held constant, an electrical charge of specific magnitude that occurs across the RTIL above a certain threshold is indicative of the presence of the virus in a patient’s body when a breath sample of the patient is analyzed by the device. However, there are several environmental variables that influence this response in a real-world application such as pressure, relative humidity, and temperature of both the patient’s breath sample and the surrounding ambient air that make this determination of the inorganic gas and metabolite concentration not so simple. The electrical current threshold mentioned above is subtle and easy to mistake if these parameters are not accounted for. Due to this, these devices contain several sensors that monitor the electrical charge across the RTIL along with the air pressure, relative humidity, and temperature of both the ambient air and the breath sample received from the patient. An algorithm present on the device takes these variables as inputs and passes them through several de-noising filters to maximize the signal-to-noise ratio. Due to the parameters that need to be accounted for, the data passes through these filters to extract the true normalized electrical signal. The slope, convexity, and final level of this signal are then analyzed to determine whether the breath sample contains a concentration of the specific metabolite and inorganic gas above the threshold. This algorithm was initially trained empirically and fine-tuned through accumulated experimental and clinical data. The algorithm uses laws of gas dynamics as well as other principles observed through internal experimentation on accumulated data. The statistical parameters as calculated by the data generated using the human subject-based testing have been summarized in Table 2.

Hence, this work is a first-time demonstration of a portable, low-power microelectronic platform for rapid and dynamic detection of NO levels in exhaled breath making it a suitable device for use in screening for the presence and absence of COVID-19. We have summarized the materials/methods/techniques that have been used for COVID-19 sensing along with our work (Table S1).