**Silver Nanomaterials as Electron Mediators in a Bio-Electronic Tongue Dedicated to the Analysis of Milks. The Role of the Aspect Ratio of Nanoparticles vs. Nanowires †**

**Coral Salvo-Comino 1,2, Clara Perez-Gonzalez 1,3, Fernando Martin-Pedrosa 2,3, Cristina Garcia-Cabezon 2,3,\* and Maria Luz Rodriguez-Mendez 1,2,\***

	- claraperez.biotecnologia@gmail.com (C.P.-G.)

**Abstract:** The integration of silver nanomaterials as electron mediators in electrochemical biosensors can be crucial to improve the affinity with biomolecules and the electrochemical response. In this work, two voltammetric bioelectronics tongues (bioET) formed by biosensors based on the combination of enzymes with silver nanoparticles (AgNPs) (bioET-1) or silver nanowires (AgNWs) (bioET-2) have been developed and used to analyze milks. Each array was formed by four biosensors formed by enzymes (glucose oxidase, galactose oxidase, β-galactosidase and a blank), capable to detect compounds usually found in milks. Principal component analysis (PCA) has revealed the ability of both biosensor systems to discriminate between milk samples with different fat contents, but with some differences, attributed to the structure employed in the detection.

**Keywords:** silver nanowires; silver nanoparticles; electronic tongue; electrochemical biosensor; enzymes; milk

#### **1. Introduction**

Milk is an essential component of human diets, although its composition varies depending on the brand, storage period, animal origin and the components that make milk up. As a result, it is vital to assess the composition and quality of milk from the time it is obtained to the time it is consumed [1].

The implementation of biosensors in food industry has taken an essential role due to the fact that these devices are able to provide qualitative information with high specificity and selectivity with the advantages of being portable, miniaturizable, cheap, stable, fast and show effective on-line response [2,3].

The use of nanomaterials in electrochemical sensors and biosensors has attracted researchers' attention. However, using specialized sensors in the analysis of complicated matrices to determine specific parameters, is insufficient to generate useful data. The development of arrays of sensors may be the solution to avoid irrelevant information. For this reason, the implementation of electronic tongues (ET) may be the solution to determine important parameters and properties of complex samples and to be able of discriminate between them [4]. Metallic nanostructures are excellent sensing materials due to their high surface area, high aspect ratio, electrical conductivity and electrocatalytic characteristics, which provide good sensing properties for the detection of a wide range of analytes. Moreover, it has been demonstrated that their morphology is essential on the electrochemical response and their ability to improve the electron mobility.

**Citation:** Salvo-Comino, C.; Perez-Gonzalez, C.; Martin-Pedrosa, F.; Garcia-Cabezon, C.; Rodriguez-Mendez, M.L. Silver Nanomaterials as Electron Mediators in a Bio-Electronic Tongue Dedicated to the Analysis of Milks. The Role of the Aspect Ratio of Nanoparticles vs. Nanowires. *Chem. Proc.* **2021**, *5*, 30. https://doi.org/10.3390/CSAC2021- 10554

Academic Editor: Núria Serrano

Published: 1 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Furthermore, due to their large specific surface area and high surface free energy, metallic nanomaterials may strongly adsorb biomolecules maintaining their bioactivity due to their biocompatibility [5].

This research aims to create a bioelectronic tongue (bioET) for milk analysis. Two electrochemical biosensor arrays in which AgNPs (bioET-1) or AgNWs (bioET-2) have been deposited as enzyme supports, were created and tested for their capacity to differentiate between milk samples with variable fat and nutritional content. For this purpose, unsupervised (PCA) multivariate classification methods were used to assess the bioETs' performance. Finally, the discrimination abilities of both sensor arrays were evaluated to determine which silver nanomaterial provided the best results.

#### **2. Results and Discussion**

#### *2.1. Sensors Development*

Metallic nanomaterials previously synthetized were deposited onto BDD substrates by a cast of 25 μL suspension of AgNPs or AgNWs. After drying, 25 μL of the correspondent enzyme (5 mg/mL) was drop-casted onto the modified electrode. Finally, 25 μL of Nafion® was drop-casted and let dry overnight at room temperature.

#### *2.2. Milk Samples*

The samples collection consisted of 6 groups of milks purchased in the supermarket, including milks with different fat content (skimmed, semi-skimmed and whole), and different nutritional content (classic and lactose-free).

#### *2.3. Electrochemical Characterization*

The electrochemical responses of the bioETs were studied by cyclic voltammetry in a 50% diluted milk 0.1 M KCl solution.

Figure 1 illustrates the electrochemical response of the developed sensor arrays with the bioETs for one type of milk (Classical Whole milk). As it can be observed, when the biological material is immobilized on the modified electrode surfaces, the increase in the intensity of the cathodic and anodic peaks is more pronounced in the presence of nanowires in the AgNWs bioET.

The results demonstrate that silver nanowires outperform silver nanoparticles in terms of electrocatalytic activity because the electrical conductivity and the electrocatalytic properties of the developed sensors have been enhanced when AgNWs are used as electron mediators. Furthermore, because AgNWs have a high surface-to-volume ratio, they have more surface active sites, allowing for improved enzyme immobilization [5].

As is well known, the ultimate electrocatalytic activity of electrode materials was impacted by surface shape, nanoparticle size and fabrication technique. For this reason, it can increase electrode surface area depending on the chosen morphology, but mass transport behavior of silver nanoparticles, for example, is difficult to define because nanoparticles are randomly spread or aggregated, whereas nanowires, increased in high proportion the electrical conductivity and mass transport in comparison with them [6].

Multivariate data analysis was used to evaluate the responses of the proposed bioETs in order to discriminate between different types of milks.

Data were evaluated using principal component analysis for this purpose (PCA).

**Figure 1.** Electrochemical response of the sensors of AgNPs bioET (black) and AgNWs bioET (red) in a 50% diluted Classical Whole Milk in 0.1 M KCl solution, (**a**) sensors formed without enzyme, (**b**) biosensors with glucose oxidase immobilized, (**c**) biosensors with galactose oxidase immobilized and (**d**) biosensors with lactate dehydrogenase immobilized.

Figure 2 shows the scores plot of both bioETs. As Figure 2a shows the accumulated explained variance for the two first components of AgNWs bioET was distributed in 49% (PC-1) and 30% (PC2) and for the AgNPs biET (Figure 2b) was 49% (PC1) and 19% (PC2). As it can be shown in the figure, the scores plot of the array that uses nanoparticles displays certain overlapping between Classical Skimmed milks and Classical Whole, appearing in the same quadrant, it is obvious that the AgNPs bioET presents difficulties to effectively discriminate milks based on their fat content. However, in both cases a clear discrimination of milks according the nutritional contents was achieved. This result indicates that the AgNWs bioET provides a higher proportion of the explained variance for the same number of PCs.

These findings can be explained by the fact that immobilizing the enzyme on nanomaterials improves the enzyme's catalytic effectiveness greatly due to its operational stability. However, it has been demonstrated that the size, morphology and charge distribution of the nanomaterial can variate the effects on enzyme structure and corresponding activity [7]. When nanowires are used as support for enzyme immobilization, different reactivity and good enzyme immobilization is obtained through easier interactions between enzyme and material surface due to providing higher surface/volume ratio in comparison with nanoparticles, thus could be the reason why the discrimination capability of the bioET is higher when AgNWs are used as electrocatalytic material.

**Figure 2.** PCA score-plot analyzed using a 4 sensors array (**a**) AgNWs bioET or (**b**) AgNPs bioET, of the 6 milks of different fat content and nutritional characteristics: Classical Skimmed (black), Classical Skimmed Lactose free (red), Classical Whole (blue), Classical Whole Lactose free (pink), Classical Semi Skimmed (green) and Classical Semi Skimmed Lactose free (dark blue).

#### **3. Conclusions**

In this work, two bioelectronics tongues modified with AgNWs or AgNPs has been developed to discriminate between milks with different nutritional composition. The electrochemical responses based on cyclic voltammetry, of the two different bioETs have been crucial to evaluate the influence of the morphology of the conductive material. Furthermore, the statistical analysis based on PCA loading plots have demonstrated the high capability of the sensors arrays to discriminate between milks.

Finally, it can be concluded that the use of AgNWs could be a better choice because their excellent electrocatalytic properties.

**Funding:** This research was funded by Ministerio de Ciencia Innovación y Universidades-FEDERPlan Nacional (RTI2018-097990-B-100), Junta de Castilla y Leon-FEDER VA275P 18, Infraestructuras Red de Castilla y León (INFRARED) UVA01 and MINECO (BES-2016-077825).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Proceeding Paper* **Development of a Bioelectronic Tongue Modified with Gold Nanoparticles for Dairy Analysis †**

**Clara Pérez-González 1,2, Coral Salvo-Comino 2,3, Fernando Martin-Pedrosa 1,2, Cristina García-Cabezón 1,2,\* and María Luz Rodríguez Méndez 2,3,\***


**Abstract:** The objective of this work was to create an all-solid-state potentiometric bioelectronic tongue with an array of polymeric membrane-based sensors, which could then be used in the dairy sector. Membranes were modified with gold nanoparticles, and enzymes were covalently linked to the sensor's surface to create an array of sensors with greater sensitivity. The responses of the sensors modified with gold nanoparticles and covalently associated enzymes, showed higher sensitivities. Moreover, the developed bioelectronic tongue was able to perform the discrimination of milks with different nutritional characteristics by applying principal component analysis. In addition, the results obtained showed that by applying partial least squares analysis, the system could be used as a prediction system for different chemical parameters (such as acidity, proteins, lactose, etc.).

**Keywords:** electronic tongue; biosensor; potentiometric; dairy industry; nanoparticles

#### **1. Introduction**

The concept of electronic tongues has expanded rapidly during recent years due to their large potential. E-tongues are based on sensor arrays with low selectivity and high cross-selectivity between multiple sensors [1]. Unlike other analytical methods, this type of device allows the acquisition of chemical information from different matrixes by applying suitable multivariate statistical, qualitative or quantitative data processing techniques, thereby targeting issues to overcome drawbacks, such as the requirement to pretreat samples, noise issues, and collinearity between variables [2].

E-tongues can implement a range of transduction principles, electrochemical sensors (potentiometric, amperometric, voltammetric, or impedimetric sensors) being the most common sensors applied in the development of e-tongues. Potentiometric sensors are based on the measurement of the differences in the interface potential created across a selective membrane. The interaction between the electrode and the solution determines this potential, which is related to the physicochemical properties of the solutions under inquiry [3].

The objective of this work was to design an all-solid-state potentiometric bioelectronic tongue (bio-ET) dedicated to the dairy sector, using an array of biosensors based on polymeric membranes operating in parallel. The membranes were modified with gold nanoparticles (AuNPs) to create an array of sensors with greater sensitivity [4]. Moreover, to further improve the sensor's selectivity, enzymes, including galactose oxidase, urease, and lactate dehydrogenase, were covalently attached to the PVC surface. The bio-ET has been applied to the analysis of milk samples with different nutritional contents.

**Citation:** Pérez-González, C.; Salvo-Comino, C.; Martin-Pedrosa, F.; García-Cabezón, C.; Rodríguez Méndez, M.L. Development of a Bioelectronic Tongue Modified with Gold Nanoparticles for Dairy Analysis. *Chem. Proc.* **2021**, *5*, 31. https://doi.org/10.3390/CSAC2021- 10553

Academic Editor: Núria Serrano

Published: 1 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

#### **2. Materials and Methods**

The electronic tongue was constructed by combining sensors based on polymeric matrices that differ in composition. The polymeric membranes were made of PVC [poly(vinyl chloride)] mixed with an additive (oleyl alcohol), a plasticiser [(bis(1-butylpentyl) adipate, tris(2ethylhexyl)phosphate or 2-nitrophenyl-octylether] and gold nanoparticles creating an array of 27 sensors. Each of the polymeric mixtures were applied on solid conducting silver-epoxy supports. The bio-ET was composed of the sensor array, an Ag/AgCl reference electrode, and a data-collecting multiplexer.

The effect of gold nanoparticles and the bioelectronic tongue performance was tested by analysing six standard solutions at concentrations ranging from 1 × <sup>10</sup>−<sup>3</sup> to 1 × <sup>10</sup>−<sup>1</sup> mol/L. When applying the multivariate analysis technique, a set of sensors was selected for the construction of a bio-ET that was applied to twelve commercial milk samples. Potentiometric signals obtained from the array of sensors were processed using principal component analysis (PCA). Using partial least squares (PLS), correlations between the responses of the sensors and the chemical parameters were established.

#### **3. Results**

The results obtained in this research indicate that sensors with a higher percentage of gold nanoparticles in their composition showed higher sensitivities towards compounds found in milk. An example of the sensor behaviour is shown in Figure 1. This behaviour could be due to an increase in the sensitivity of the sensors towards specific ions on the interface of the polymeric membrane when AuNPs are included in the matrix composition.

**Figure 1.** Each sensor's response to glucose at increasing concentrations (blue = 10−<sup>4</sup> M, green = 10−<sup>3</sup> M, and orange = 10−<sup>2</sup> M) depending on the percentage of AuNPs applied to the membrane matrix.

Moreover, the sensor that combined gold nanoparticles with enzymes showed a greater ability to differentiate between increasing concentrations of products of interest found in milk, such as urea, lactic acid, galactose, etc., with deviations of voltages up to 65 mV between different samples.

Furthermore, using statistical analysis (PCA), the constructed bioelectronic tongue was able to classify milk with various nutritional features, resulting in four distinct groups that were also sorted according to the fat content of the samples (Figure 2).

**Figure 2.** Classification of the milk samples studied according to the score diagram of the principal component analysis.

Additionally, the study's findings revealed that using partial least squares analysis (PLS), with regression coefficients above 0.85 for three variables in the physio-chemical parameters studied, the developed bioelectronic tongue could be used as a prediction system to determine the parameters, such as density, acidity, lactose, or fat content, of future milk samples (Table 1).

**Table 1.** Correlation parameters resulting from the regression of partial least squares analysis (PLS).


#### **4. Conclusions**

In this work, a bioelectronic tongue was developed and used to predict the chemical characteristics of milk. Sensors with higher concentrations of gold nanoparticles in their composition showed greater sensitivity towards the compounds of interest in milk (such as lactic acid, galactose, and urea). The system, using nine potentiometric sensors, could be successfully used to discriminate between milks, applying PCA based on their nutritional content. The bio-ET was successfully used to predict the acidity and density, in addition to the protein, lactose, and fat content of the milk, with low errors and high correlation coefficients for three factors. This device could be adapted for its implementation in the dairy industry.

**Funding:** MICINN-FEDER Plan Nacional (RTI2018-097990-B-100), Consejería de Educación Junta de Castilla y Leon- FEDER VA275P18 and «Infraestructuras Red de Castilla y León (INFRARED)» UVA01.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Abstract* **Core Modulation of Porphyrins for Chemical Sensing †**

**Karolis Norvaiša 1,\* and Mathias Otto Senge 1,2**


**Abstract:** The inner core system of metal-free ('free base') porphyrins has continually served as a ligand for various metal ions, but it was only recently studied in organocatalysis due its highly tunable basicity. Highly conjugated porphyrin systems offer spectrophotometric sensitivity toward geometrical and/or electronic changes and, thus, utilizing the porphyrin core for the selective detection of substrates in solution offers significant potential for a multitude of applications. However, solvation and dilution drastically affect weak interactions by dispersing the binding agent to its surroundings. Thus, the spectroscopic detection of N–H···X-type binding in porphyrin solutions is almost impossible without especially designing the binding pocket. Here, we present the first report on the spectroscopic detection of N–H···X-type interplay in porphyrins formed by weak interactions. Protonated 2,3,7,8,12,13,17,18-octaethyl-5,10,15,20-tetrakis(2-aminophenyl) porphyrin contains coordination sites for the selective binding of charge-bearing analytes, revealing characteristic spectroscopic responses. While electronic absorption spectroscopy proved to be a particularly useful tool for the detection of porphyrin–analyte interactions in the supramolecular complexes, X-ray crystallography helped to pinpoint the orientation, flexibility, and encapsulation of substrates in the corresponding atropisomers. This charge-assisted complexation of analytes in the anion-selective porphyrin inner core system is ideal for the study of atropisomers using high-resolution NMR, since it reduces the proton exchange rate, generating static proton signals. Therefore, we were able to characterize all four rotamers of the nonplanar 2,3,7,8,12,13,17,18-octaethyl-5,10,15,20-tetrakis(2 aminophenyl) porphyrin by performing 1D and 2D NMR spectroscopic analyses of host-guest systems consisting of benzenesulfonic acid (BSA) and each porphyrin atropisomer. Lastly, a detailed assignment of the symmetry operations that are unique to porphyrin atropisomers allowed us to accurately identify the rotamers using NMR techniques only. Overall, the N–H···X-type interplay in porphyrins formed by weak interactions that form restricted H-bonding complexes is shown to be the key to unravelling the atropisomeric enigma.

**Keywords:** porphyrins; sensing; atropisomers; NMR; UV–vis; nonplanar

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/CSAC2021-10417/s1.

**Citation:** Norvaiša, K.; Senge, M.O. Core Modulation of Porphyrins for Chemical Sensing. *Chem. Proc.* **2021**, *5*, 32. https://doi.org/10.3390/ CSAC2021-10417

Academic Editor: Nicole Jaffrezic-Renault

Published: 30 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

### *Proceeding Paper* **Nanostructured Bismuth Electrodes for Non-Enzymatic Paracetamol Sensing: Development, Testing, and Computational Approach †**

**Mallikarjun Madagalam 1, Federica Catania 1, Mattia Bartoli 2,3,\* , Alberto Tagliaferro 3,4 and Sandro Carrara <sup>5</sup>**


**Abstract:** In this work, new Screen Printed Carbon-paste Electrodes (SPCEs) were developed through deposition of nanostructures of HO–BiONO3 synthesized with or without surfactant additions. We performed a cyclic voltammetry study showing the improvement in performance of bismuth tailored electrodes for paracetamol sensing compared with bare SPCE. A computation study was also performed for investigating the interaction between paracetamol and bismuth species during the electron transfer process enlighten the preferential sites of interaction on the surface of modified SPEs.

**Keywords:** paracetamol; bismuth; SPCE

#### **1. Introduction**

Paracetamol is among the most studied emerging pollutants [1] and one of the most used antipyretic drugs in the world [2]. Accordingly, the development of a new trustworthy paracetamol detection system is a key pillar in the sensors field for personalized therapy and for developing environmental monitoring. Among all available approaches, electrochemical techniques are the most useful considering production-cost, sensibility and reproducibility of sensing technologies [3]. The development of new electrode materials for paracetamol detection has attracted more interest in recent years. To improve the sensing performance, several studies used modified carbon electrodes tailored with several nanostructured materials [2,3]. Recently, Madagalam et al. [4] reported a new SPCE decorated with bismuth sub-nitrate as an effective electrochemical sensor for paracetamol detection. The choice of bismuth was driven by the great tuneability of bismuth species [5], together with its remarkable electrochemical performances [6,7].

Here, we move a step forward in the comprehension of the paracetamol detection by using bismuth tailored SPCEs through a computational study. Accordingly, we report a solid approach for the calculation of electron transfer rate with new structural information of transitional state geometry of paracetamol–bismuth sub-nitrate system.

#### **2. Materials and Methods**

#### *2.1. Materials*

Bismuth nitrate pentahydrate (Bi(NO3)3·5H2O), 1-butanol, and Polyethylene glycol (PEG) were used for material synthesis at Politecnico di Torino, Italy. Phosphate buffer

**Citation:** Madagalam, M.; Catania, F.; Bartoli, M.; Tagliaferro, A.; Carrara, S. Nanostructured Bismuth Electrodes for Non-Enzymatic Paracetamol Sensing: Development, Testing, and Computational Approach. *Chem. Proc.* **2021**, *5*, 33. https://doi.org/ 10.3390/CSAC2021-10427

Academic Editor: Nicole Jaffrezic-Renault

Published: 30 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

saline (PBS), DI water, and paracetamol tablets made into powder were used for the preparation of material suspensions and electrolytic solution at EPFL, Neuchatel, Switzerland.

#### *2.2. Methods*

#### 2.2.1. Material Synthesis and Modification of SPCEs

HO–BiONO3 was synthesized According to Liu et al. [8] and by the modified synthesis reported by Madagalam et al. [4]. SPCEs were prepared according to the procedure described in [4]. A SPCEs consisting on a working electrode and a counter electrode in carbon, and a reference electrode in silver/silver-chloride is presented in this study.

#### 2.2.2. Electrochemical Measurements

Cyclic Voltammetry measurements (CV) were performed under atmospheric conditions using AutoLab potentiostat and run accordingly with the procedure described in [4]. For an electrochemically reversible electron transfer process involving freely diffusing redox species, the Randles–Sevcik equation [9] describes the linear increase of the peak current *ip* (A) with the square root of the scan rate ν (V/s):

$$\dot{a}\_p = 0.466 n FAC \left(\frac{nF\nu D}{RT}\right)^{\frac{1}{2}},\tag{1}$$

where *n* is the number of electrons transferred in the redox reaction, *A* (cm2) is the electrode geometric surface area, *D* (cm2/s) is the diffusion coefficient of the oxidized analyte, and *C* (mol/cm3) is the bulk concentration of the analyte. The Randles–Sevcik equation (Equation (1)) is used to deduce the analyte behavior identifying the two scenarios when it is freely diffusing in solution or adsorbed on the surface of the electrode.

Kinetic parameters were calculated using the Laviron model [10], where the peak potentials are described as follows, with the cathodic peak potential being:

$$E\_{p\varepsilon} = E^0 - \left(\frac{RT}{anF}\right) \ln\left[\frac{a}{|m|}\right] \tag{2}$$

and the anodic peak potential being:

$$E\_{\rm pr} = E^0 + \left(\frac{RT}{(1-\alpha)nF}\right) \ln\left[\frac{(1-\alpha)}{|m|}\right] \tag{3}$$

with *m* = (*RT/F*) (k/nν). *R* is the universal gas constant, *n* is the number of electrons involved in the redox reaction, *T* is the absolute temperature in Kelvin, *F* is the Faraday constant, *ν* is the scan rate, and *E*<sup>0</sup> is the surface standard potential. The kinetic rate constant *k* was calculated by substituting α value into the following equation:

$$
\ln(k) = \pi \ln(1-\pi) + (1-\pi)\ln\alpha - \ln\left(\frac{RT}{nF\nu}\right) - \pi(1-\pi)\frac{nF\Delta E\_p}{RT} \tag{4}
$$

#### 2.2.3. Modelling

Computational simulations were run with HyperChem (HC) software using the following procedure.

A 3D structure is obtained in HC by the function *model build* for both paracetamol and Bi clusters. Then, the simulations started to evaluate the electronic properties and the geometry information required in the generalized Marcus model. Therefore, the potential energy must be minimized with respect to the system coordinated by means of geometry optimization. This is done by evaluating the potential energy as function of molecules coordinates according to a hybrid quantum/molecular mechanics approach (MM+) [11]. Then, a second simulation was run in line with the neglect of differential diatomic overlap

approximation (PM3 basis set) [12]. Two different Bi clusters model the polymeric layer as shown in Figure 1.

**Figure 1.** BiONO3 dimer on the left and BiONO3 trimer on the right.

The dimer was assumed to be the minimum unit allowing the electron transfer (ET). Indeed, ET is depicted through two Bi–C "fake bonds" as shown in Figure 2 by supposing that each couple exchanges just one electron and the two charged particles are simultaneously acquired from paracetamol [1].

**Figure 2.** Transition state model: the double dashed lines on paracetamol defined the chemical bonds involved in the reaction while the single ones defined the Bi–C "fake bonds" through which the ET occurs.

The Figure 2 represents one of the hypothesized transition states (HTS) geometry. Indeed, the transitional state (TS) is identified as a unique crossing point in the potential energy surface (PES) defining the reaction coordinates of both reactants and products [13]. However, according to Tachiya et al. [14] the TS is not uniquely defined since it depends on the probability to achieve a certain value of the electrostatic potential distribution. For that reason, it is necessary to talk about "hypothetical" TS. Moreover, it is observed after each simulation that the interaction distance between each Bi–C couple is not equal:, the system is asymmetric, thus the average between the interaction distance is kept within the model according to the procedure reported in [14]. Afterwards, the simulations were repeated for the Bi trimer. The geometry optimization was performed two times in both cases by cutting the "fake bonds" before starting the second simulation. The solvation effect arising from the electrolyte wetting the electrode at the interface with the organic compound was considered as well. It was modelled by adding water molecules surrounding paracetamol and it was established that the ET would be ensured by a minimum amount of five molecules. Finally, the HTS in which the nitric functional groups in both dimer and trimer were replaced by hydroxyl group by supposing that the functionalized electrode undergone a cleaning pre-treatment in H2SO4 was simulated. The electron transfer rate constant in a CdS-Phenol system was analyzed to validate our model. According to Serpone et al. [15], we simulated the electrochemical redox process between a CdS powder and a phenol molecule in solution by defining a box containing water molecules, as shown in Figure 2. All of them were mirrored outside the box by reproducing the system in solution. The obtained ET rate constant was *<sup>k</sup>* = 6.42·10−<sup>5</sup> <sup>s</sup><sup>−</sup>1, in compliance with the literature constant (*<sup>k</sup>* = 5.17·10−<sup>5</sup> <sup>s</sup><sup>−</sup>1).

#### **3. Results and Discussion**

#### *3.1. Electrochemical Measurements*

SPCEs modified sensors were tested by Madagalam et al. [4] through CV and main data are summarized in Table 1.

The peaks detected during CV without and with paracetamol were distinguishable and confirmed that we observed higher peak currents when there was paracetamol in the electrolytic solution. PEG-coated HO–BiONO3 sensor was found to have a potential shift of −75 mV, whereas HO–BiONO3 sensor has a shift of −26 mV due to Nernst effect compared to the bare SPCE sensor. PEG-coated HO–BiONO3 sensor was also found to have a higher oxidation peak current of 52.1 ± 3.0 μA. To study the behavior of the electrochemical system, CVs were run by varying the scan rate (50 mV/s–300 mV/s) while determining 1 mM paracetamol in 0.1 M PBS at pH 7. It was observed that the peak position changed by increasing the scan rate while the redox current increased linearly with the square root of the scan rate. From these observations and Equation (1), it was possible to give an indication that the electrochemical system was a freely diffusing quasi-reversible system. By plotting peak positions (Epa, Epc) versus ln(ν), a linear variation was observed and peak-to-peak separation (ΔEp) increased linearly with ln(ν). According to Equations (2) and (3), it was found that n is approximately equal to 2 meaning that two electrons were participating in the redox reaction of paracetamol as reported in [1]. The α values are listed in Table 1 for different sensors together with the k values obtained at a scan rate of 100 mV/s for three different sensors. PEG-coated HO–BiONO3 sensor had a higher rate constant of 42.0 ± 9.8 ms−<sup>1</sup> with lower <sup>Δ</sup>Ep of 243 ± 10 mV compared to other sensors. This is a big advantage compared to the SPCE sensor since a higher rate constant gives rise to faster electrochemical reaction and lower ΔEp indicates a higher possibility of a reversible reaction (low resistance). Quantitative studies run by CVs measurements showed a higher sensitivity for PEG-coated HO–BiONO3 compared with bare SPCE due to the nanostructure deposited on the surface [4].

**Table 1.** Main experimental outputs from electrochemical measurements of bare and tailored SPCEs.


#### *3.2. Computational Evaluation*

The electron transfer rate constant values coming from the computational simulations are presented in the following sections by underlying the affecting parameters on the transition state for both the HO–BiONO3 dimer and trimer cases. A geometrical scheme as shown in Figure 3 acts as reference to describe the variations in the hypothetical transition state conformation depending on the functional groups bonded to the Bi active atoms. According to the assumption that paracetamol interacts with Bi cluster through its two carbon atoms, we can define:


**Figure 3.** Paracetamol geometrical description on the left showing the symmetrical axis *C2* (right); dihedral angle is defined as the angle between the plane *A* (green) on which the paracetamol carbon ring lies and the plane *B* (orange) passing through the Bi atoms (center); description of paracetamol plane rotation of an angle β with respect to the symmetric axis (left).

#### 3.2.1. HO–BiONO3 Dimers

All the results are shown in Table 2. The different cases are labelled with the functional groups bonded to the two Bi atoms.


**Table 2.** Electron transfer rate constant and activation energy for Bi dimers.

We firstly evaluated the characteristics of symmetric dimers with two nitric or hydroxylic groups as shown in Figure 4. In both cases, the electron transfer rate constant is of the order of magnitude of 10−<sup>3</sup> s−<sup>1</sup> a factor of 2 higher in the hydroxylic groups case. The activation energy difference is about 15%. This is reasonable, as the dihedral angle was slightly greater in NO3-NO3 than OH-OH by fixing the angle β, as can be seen in Figure 5. However, paracetamol approaching is hugely affected by the two functional groups. Indeed, the NO3 group bonded with Bi (a) was above the Bi plane and moved closer to the paracetamol nitro groups by leading to a distortion of the Bi-O bond on that side. The surrounding water molecules improved that distortion and emphasized the steric hindrance. On the other hand, having four identical hydroxylic groups ensured a less steric effect since no strong repulsions between electron clouds can arise in that case.

A second comparison was made between the HTS associated to the two specular Bi-dimers having an OH group and a NO3 alternatively bonded to each Bi atom. In this case, a larger dihedral angle resulted when NO3 was bonded to the right Bi since the interaction with the paracetamol OH-group lead the nitric group position above the Bi plane. This also resulted in an ET rate constant equals to 6.9 · <sup>10</sup>−<sup>3</sup> <sup>s</sup>−<sup>1</sup> which is smaller than the NO3-OH by a factor of 4. Regarding the activation energy, a difference of only 4% was obtained since the structures involved are the same in terms of atoms. From these simulations, we realized that having hydroxyl groups could enhance the sensing properties of the functionalized layer despite the steric effect arising from neighboring nitric groups.

**Figure 4.** Molecular conformation of the HTS associated to the dimer case NO3–NO3 on the left and OH–OH on the right and two bismuth atoms (named **a** and **b** respectively).

**Figure 5.** Overlapped molecular structure of HTS associated to the dimer NO3–NO3 (blue planes) and OH–OH (yellow planes): *α* angles comparison at fixed β.

#### 3.2.2. HO–BiONO3 Trimer

All the results are shown in Table 3. The different cases are labelled with the functional groups bonded to the three Bi atoms.

**Table 3.** k and activation energy for Bi trimers.


When we simulated the HTS associated to the trimer cases, we introduced a further assumption by distinguishing among the reacting Bi atoms and the non-reacting ones since just two Bi atoms are supposed to interact with the paracetamol C atoms. Based on that, we observed that the steric effect was more evident in the NO3-OH-NO3 structure due to the orientation of the nitric group bonded with the non-reacting Bi atom towards the nitro group of the organic compound. On the contrary, no distortion occurs in the NO3-NO3-NO3 because the functional groups bonded with reacting Bi lie in the same plane. However, lower activation energy of the transition state made the NO3-OH-NO3 more energetically favorable to the ET process and further confirmed that the presence of some OH groups should improve the efficiency of the ET process. Indeed, the ET rate constant was three orders of magnitude greater in NO3-OH-NO3 case than the NO3-NO3-NO3 case.

Finally, the specular case for the trimer was simulated by replacing the two nitric groups with hydroxyl as shown in Figure 6. We observed that NO3-OH-OH was strongly affected by the orientation of NO3 above the Bi plane towards the paracetamol nitro group that is also above the Bi plane. Nevertheless, no distortion occurred, and it can be seen in Figure 7 that the dihedral angle was qualitatively the same as for the OH–OH–NO3 case. These also let to ET values in the two specular trimers with a difference less than 2% and geometries with very similar the dihedral angles. Indeed, we obtained a value of 9.16 eV for NO3–OH–OH and 9.87 eV for OH–OH–NO3.

**Figure 6.** Molecular conformation of the HTS associated to the dimer case NO3–OH–OH on the left and OH–OH–NO3 on the right and three bismuth atoms (named **a**, **b** and **c** respectively).

**Figure 7.** Overlapped molecular structure of HTS associated to the dimer OH–OH–NO3 (blue planes) and NO3–OH–OH (yellow planes): *α* angles comparison at fixed β.

#### **4. Conclusions**

As clearly shown from the data, the electrochemical sensing boost observed for HO– BiONO3 tailored SPCEs is likely due to defective sites on HO–BiONO3 particles sitting on the surface of the electrode. The hydroxyl functionalities played a relevant role in the paracetamol–bismuth interaction that defines the geometry of the transitional state as showed by the computational study. Accordingly, a more polar surface induced a better interaction with paracetamol.

This study has firstly described in a comprehensive way the effect of bismuth subnitrate chemistry in an electrochemical system considering both empirical and experimental data sets, leading to the foundation for a rational design of bismuth modified SPCEs.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/CSAC2021-10427/s1.

**Author Contributions:** Conceptualization, A.T., S.C. and M.B.; formal analysis, M.M. and F.C.; resources, A.T. and S.C.; data curation, M.M. and F.C.; writing—original draft preparation, M.M., F.C. and M.B.; writing—review and editing, M.M., F.C., M.B., A.T. and S.C.; supervision, M.B.; project administration, A.T. and S.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Abstract* **Plasmonic Hydrogel Nanocomposites with Combined Optical and Mechanical Properties for Biochemical Sensing †**

**Bruno Miranda 1,2,\* , Rosalba Moretta <sup>1</sup> , Selene De Martino 3, Principia Dardano <sup>1</sup> , Ilaria Rea <sup>1</sup> , Carlo Forestiere <sup>2</sup> and Luca De Stefano <sup>1</sup>**


**Abstract:** Localized surface plasmon resonance (LSPR) and metal-enhanced-fluorescence (MEF) based optical biosensors exhibit unique properties compared to other sensing devices that can be exploited for the design point-of-care (POC) diagnostic tools [1]. Plasmonic devices exploit the capability of noble-metal nanoparticles of absorbing light at a well-defined wavelength. The increasing request for wearable, flexible and easy-to-use diagnostic tools has brought to the development of plasmonic nanocomposites, whose peculiar performances arise from the combination of the optical properties of plasmonic nanoparticles and mechanical properties of the polymeric matrix in which they are embedded [2,3]. An optical platform based on spherical gold nanoparticles (AuNPs) embedded in high molecular weight poly-(ethylene glycol) diacrylate (PEGDA) hydrogel is proposed. PEGDA hydrogel represents a biocompatible, flexible, transparent polymeric network to design wearable, 3D, plasmonic biosensors for the detection of targets with different molecular weights for the early diagnosis of disease. The swelling capability of PEGDA is directly correlated to the plasmonic decoupling of AuNPs embedded within the matrix. A study on the effect of swelling on the optical response of the PEGDA/AuNPs composites was investigated by using a biorecognition layer/target model system. Specifically, after the in situ chemical modification of the AuNPs within the hydrogel, the interaction biotin-streptavidin is monitored within the 3D hydrogel network. Additionally, metal-enhanced fluorescence is observed within the PEGDA/AuNPs nanocomposites, which can be exploited to achieve an ultra-low limit of detection. LSPR signal was monitored via transmission mode customized setup and MEF signal was detected via fluorescence and confocal microscopes. Label-free (LSPR-based) and fluorescence (MEF-based) signals of a high molecular weight target analyte were successfully monitored with relatively high resolutions and low limits of detection compared to the standard polymeric optical platforms available in the literature. The optimized platform could represent a highly reproducible and low-cost novel biosensor to be applied as a POC diagnostic tool in healthcare and food monitoring applications.

**Keywords:** optical biosensors; flexible hybrid materials; disease early diagnosis; nanofabrication techniques; nanocomposite materials; LSPR-based biosensors; metal-enhanced fluorescence

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/CSAC2021-10467/s1.

**Author Contributions:** Conceptualization, L.D.S., I.R., and P.D.; methodology, B.M., C.F., S.D.M. and R.M.; validation, B.M. and C.F.; formal analysis, B.M.; investigation, B.M., and R.M.; resources, L.D.S.;

**Citation:** Miranda, B.; Moretta, R.; Martino, S.D.; Dardano, P.; Rea, I.; Forestiere, C.; Stefano, L.D. Plasmonic Hydrogel Nanocomposites with Combined Optical and Mechanical Properties for Biochemical Sensing. *Chem. Proc.* **2021**, *5*, 34. https:// doi.org/10.3390/CSAC2021-10467

Academic Editor: Elisabetta Comini

Published: 30 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

writing—original draft preparation, B.M., and R.M.; writing—review and editing, C.F., I.R., and P.D.; supervision, L.D.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data that support the findings of this study are available from the corresponding author upon reasonable request.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Proceeding Paper* **Statistical Analysis for Selective Identifications of VOCs by Using Surface Functionalized MoS2 Based Sensor Array †**

**Uttam Narendra Thakur \*, Radha Bhardwaj and Arnab Hazra \***

Department of Electrical & Electronics Engineering, Birla Institute of Technology and Science (BITS)-Pilani, Vidya Vihar, Pilani 333031, India; radikabhardwaj.rb@gmail.com

**\*** Correspondence: unthakur08@gmail.com (U.N.T.); arnabhazra2013@gmail.com (A.H.)

† Presented at the 1st International Electronic Conference on Chemical Sensors and Analytical Chemistry, 1–15 July 2021; Available online: https://csac2021.sciforum.net/.

**Abstract:** Disease diagnosis through breath analysis has attracted significant attention in recent years due to its noninvasive nature, rapid testing ability, and applicability for patients of all ages. More than 1000 volatile organic components (VOCs) exist in human breath, but only selected VOCs are associated with specific diseases. Selective identification of those disease marker VOCs using an array of multiple sensors are highly desirable in the current scenario. The use of efficient sensors and the use of suitable classification algorithms is essential for the selective and reliable detection of those disease markers in complex breath. In the current study, we fabricated a noble metal (Au, Pd and Pt) nanoparticle-functionalized MoS2 (Chalcogenides, Sigma Aldrich, St. Louis, MO, USA)-based sensor array for the selective identification of different VOCs. Four sensors, i.e., pure MoS2, Au/MoS2, Pd/MoS2, and Pt/MoS2 were tested under exposure to different VOCs, such as acetone, benzene, ethanol, xylene, 2-propenol, methanol and toluene, at 50 ◦C. Initially, principal component analysis (PCA) and linear discriminant analysis (LDA) were used to discriminate those seven VOCs. As compared to the PCA, LDA was able to discriminate well between the seven VOCs. Four different machine learning algorithms such as k-nearest neighbors (kNN), decision tree, random forest, and multinomial logistic regression were used to further identify those VOCs. The classification accuracy of those seven VOCs using KNN, decision tree, random forest, and multinomial logistic regression was 97.14%, 92.43%, 84.1%, and 98.97%, respectively. These results authenticated that multinomial logistic regression performed best between the four machine learning algorithms to discriminate and differentiate the multiple VOCs that generally exist in human breath.

**Keywords:** breath analysis; surface functionalized MoS2; classification; discrimination

#### **1. Introduction**

In the field of medical diagnostic and health care systems, breath analysis has gained a lot of interest for the noninvasive detection of diseases and monitoring of health parameters [1,2]. More than 1000 volatile organic components (VOCs) are present in exhaled breath, but only some of them are considered disease markers [3,4]. In this context, the selective detection of the different VOCs using smart sensor systems has a high demand for efficient breath analysis. Selective detection can also be achieved using suitable pattern recognition algorithms on sensor signals. For the early detection of disease, the combination of a highly selective sensor and an effective machine learning algorithm is required. Diagnostics through breath is less time consuming compared to the clinical process and, at the same time, it is cost-efficient as does not require well-trained professionals and the sensors are less costly [5,6].

Chemiresistive sensors typically recognize a target VOC by changing their resistance depending upon the adsorption-desorption properties of the analyte to the detecting layer surface. An extensive variety of materials are used for VOC sensing, including thin metal

**Citation:** Thakur, U.N.; Bhardwaj, R.; Hazra, A. Statistical Analysis for Selective Identifications of VOCs by Using Surface Functionalized MoS2 Based Sensor Array. *Chem. Proc.* **2021**, *5*, 35. https://doi.org/10.3390/ CSAC2021-10451

Academic Editor: Elisabetta Comini

Published: 30 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

films [7], metal oxides [8–10], polymers [11], etc. Accessible surface functionalization possibilities, high surface area to volume ratio, increased flexibility, high sensitivity, and tunable bandgap make two-dimensional molybdenum disulfide (MoS2) an encouraging channel material to sense the VOC [12,13].

A pattern recognition algorithm also plays an essential role in the detection of VOC. A suitable classifier is required to achieve an effective classification rate in VOC sensing based on the sensor data. Different algorithms such as partial least squares discriminant analysis [14], canonical discriminant analysis [15], k-nearest neighbor [4,16], Discriminant function analysis [17], support vector machine [18], random forest [19], logistic regression [20], etc. were reported in the literature. In some of the reported literature, different types of neural network classifiers were used [21–24].

In the current study, we used principal component analysis (PCA) and linear discriminant analysis (LDA) to visualize the data in lesser dimensions compared to the original extent. Furthermore, four different supervised algorithms, k-nearest neighbor (kNN), decision tree, random forest, and multinomial logistic regression, were implemented to identify the best-suited algorithm based on their performance.

#### **2. Material and Methods**

#### *2.1. Preparation of MoS2 and Noble Metal Nanoparticles Solutions*

All materials MoS2 (Chalcogenides, Sigma Aldrich, St. Louis, MO, USA), gold (III) chloride (AuCl3, 99%, Sigma Aldrich), palladium chloride (PdCl2, 60%, Molychem, Mumbai, India), and chloroplatinic acid (H4PtCl6xH2O, 40%, Molychem) were analytical grade and used without any further purification. The 0.2 wt% MoS2 solution was prepared in deionized water and stirred for 1.5 h at room temperature to maintain homogeneity. Similarly, 0.1 MM aqueous solutions of noble metal nanoparticles (Au, Pd, Pt) were prepared by adding corresponding metal salts in deionized water with continuous stirring and dropwise diluted HCl was also added to obtain stable and uniform nanoparticles at room temperature.

Au, Pd and Pt nanoparticle-loaded MoS2 samples were prepared using the spray coating technique. Firstly, MoS2 solution was spray coated on washed SiO2/Si substrate and dried at room temperature. In the final step, nanoparticle solutions were spray-coated on previously deposited MoS2 and dried at room temperature.

Thermal annealing was performed for 4 h at 250 ◦C to provide crystallization and thermal stability in all 4 samples (MoS2, Au-MOS2, Pd-MoS2, and Pt-MoS2). The MoS2 flakes were coated by very tiny Au NPs (12 nm), whereas the larger Pd NPs (54 nm) were deposited, maintaining a consistent spacing. Pt nanoparticles on the MoS2 surface had the biggest size (77 nm) of the three.

#### *2.2. Fabrication of Sensors*

Au source and drain electrodes of 150 nm thickness were deposited on all four samples by using an electron beam evaporation unit. Sensors were then placed into a sensor holder, and further sensing performance was studied.

The sensor holder was placed in a glass sealed sensing chamber of 650 mL on a heating plate. The sensing performance of prepared sensors was examined by a static mode sensing setup where VOCs were injected using micro syringes (Hamilton micro syringe). The sensor was recovered by flowing 450 SCCM synthetic air by using a mass flow controller. The amount of injected VOC was calculated by using the formula: C (ppm) = 2.46 × (V1D/VM) × 103, where D (gm/mL), M (gm/mol), and V (L) represent density of the VOC, molecular weight of the VOC and volume of vaporization chamber, respectively [13,25,26]. Seven different VOCs, i.e., acetone, 2-propanol, benzene, ethanol, methanol, toluene, and xylene were tested during the study. Using a Keithley 6487 source meter, sensing performance was recorded applying 1 V constant bias. The sensitivity of the sensor was calculated by formula; Ra − Rv/Ra × 100 where Ra and Rv were the resistances of the sensor in the air and in target VOC.

To read the generated output of sensors stored in CSV file a python script was used. All the algorithms, analysis, and plotting were performed on Python 3.7 and Jupiter notebook as a platform.

#### **3. Results and Discussion**

#### *3.1. VOC Sensing*

As a reference ambient, synthetic air was used to perform the gas sensing measurements of the four different sensors: pure MoS2, Au-MoS2, Pd-MoS2, and Pt-MoS2. Figure 1 shows the change in resistance (MΩ) with respect to time at 50 ◦C. In the presence of VOCs, as the exposer time increased, the resistance offered by the sensor decreased. This decrease in resistance confirms that the sensor is an n-type property. In the presence of seven distinct VOCs, i.e., acetone, 2-propanol, benzene, ethanol, methanol, toluene, and xylene, four different sensors, i.e., pure MoS2, Au-MoS2, Pd-MoS2, and Pt-MoS2, were observed and stored for further processing of data.

**Figure 1.** Change in resistance offered by sensors (**a**) MoS2 (**b**) Au-MoS2 (**c**) Pd-MoS2 (**d**) Pt-MoS2 with respect to time in presence of 7 VOCs.

#### *3.2. Data Analysis*

Figure 2 describes the influence of the volatile organic components (VOCs) on the outcomes of two-dimensionality reduction techniques: principal component analysis and linear discriminant analysis. The measurement parameters were kept constant during the experiment. The operating temperature was 50 ◦C, the response was taken up to 600 s. and the sample concentration was 100 ppm.

The response obtained by the four different sensors for seven different VOCs was used for principal component analysis (PCA). The three-dimensional plot between the first principal component (PC1), second principal component (PC2), and third principal component (PC3) is represented in Figure 2. As we have four independent variables (sensor responses), the maximum principal component obtained was four. Therefore, we have considered only the first three principal components contributing the most to the explained variance in this analysis. The total explained variance was 93.58%, in which PC1 contributed 52.52%, PC2 contributed 30.91%, and PC3 contributed 10.14%. All seven VOCs had their compact clusters, and they separated. Still, the separation between the cluster of acetone/2-propanol and benzene/toluene was quite small, which increased the possibility of misclassification.

**Figure 2.** Scatter plot from the exposure of four sensors to seven VOCs in (**a**) PCA (**b**) LDA.

Taking account of the problem of discrimination between the different VOCs, a linear discrimination analysis was also performed. The same sensor response vector was used in the linear discriminant analysis (LDA). Figure 2b shows that the employment of the classifier allowed the discrimination of all seven VOCs. Thus, LDA is highly efficient for investigating the VOCs based on the sensor response. A three-dimensional plot is shown in Figure 2b, which clearly depicts the performance of the LDA on the raw data (sensor response vector). The different VOCs were densely clustered within their groups, and they were well separated from each other. So, there was a significantly lower probability of misclassification among the VOCs. The 2-propanol was slightly more stretched along the second linear discriminant function (LD2) axis, and xylene was along the third discriminant function (LD3). The three discriminant functions, LD1, LD2, and LD3 contributed 71.22%, 27.42% and 1.21%, respectively, the total resultant explained the variance for the classifier becoming 99.85%.

#### *3.3. VOC Identification*

The previously discussed LDA and PCA plot gives only the visual representation of the separation of VOCs based on the sensor response. The goal of the sensor setup is to design a generalized model based on the known data during the training phase and try to predict the class when an unknown data sample is encountered.

The supervised algorithm was performed in the current work to determine the VOCs; four different machine learning algorithms such as k-nearest neighbor (kNN), decision tree, random forest, and multinomial logistic regression were used to identify those seven VOCs. The normalized sensor response was fed to the algorithms, and the whole data set was divided into training testing data with 70% and 30%, respectively. The data set consisted of 4200 measurements of each sensor, with each class containing 600 data vectors and seven classes. So, 2940 vectors were used to train the model, and the remaining 1260 vectors were used to test the model. For identification of VOCs, above 84% was the classification accuracy for every classifier with an accuracy of 97.14%, 92.43%, 84.1%, 98.97% for kNN, decision tree, random forest, and multinomial logistic regression, respectively. A confusion matrix is used to calculate the classification accuracy, and the confusion matrix furnishes the observation into what components were mistakenly classified. Figure 3a shows the confusion matrix of kNN where 11 samples of toluene were classified as xylene and 10 samples of benzene were wrongly predicted as ethanol. Figure 3b is a representation of the confusion matrix obtained from the decision tree algorithm. The confusion matrix of the random forest and multinomial logistic regression are shown in Figure 3c,d, respectively. In multinomial logistic regression, only 12 benzene samples were identified as acetone, and one sample of xylene was identified as toluene.

**Figure 3.** Confusion matrix of (**a**) k-nearest neighbor, (**b**) decision tree, (**c**) random forest, and (**d**) multinomial logistic regression.

#### **4. Conclusions**

The ability of a surface-functionalized MoS2 sensor to distinguish between the various VOCs was appraised by PCA and LDA, in which LDA laid out the excellent separation between the VOCs. Further, to evaluate the effectiveness of the sensor output to identify the VOCs, four different machine learning-based (supervised) classification algorithms were implemented, and among them, the k-nearest neighbor and multinomial logistic regression were performed outstandingly with an accuracy of 97.14% and 98.97%, respectively. Thus, high selectivity and accuracy authenticate that the system discriminates and differentiates the multiple VOCs that generally exist in human breath.

#### **Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/CSAC2021-10451/s1.

**Author Contributions:** U.N.T. is responsible for the experiment and analyzed data. R.B. is responsible for the research activity with fabrication of sensors and their characterization and conceived the experiment. A.H. designed the research. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported in part by SPARC grant (SPARC/2018-2019/P1394/SL), Ministry of Human resource development (MHRD), Govt. of India and Department of Biotechnology grant (Letter No. BT/PR28727/NNT/28/1569/ 2018).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

