Gas Sensors Based on Exfoliated g-C₃N₄ for CO₂ Detection

Abstract

We report on the investigation of graphitic carbon nitride (g-C₃N₄) for carbon dioxide (CO₂) sensor applications. g-C₃N₄ is prepared by the thermal polycondensation of thiourea and sprayed onto a substrate with interdigitated electrodes. The resulting sensor device exhibited a high sensitivity to CO₂ molecules of ~200 ppm, a high responsivity of ~730 ms at 40 °C and a full recovery time of 36 s. Furthermore, a set of various characterization measurements demonstrated the excellent stability of both the g-C₃N₄ nanosheets and the fabricated gas sensor device. Meanwhile, density functional theory (DFT) calculations for the bulk and monolayer models, based on tri-s-triazine, revealed the optoelectronic properties of g-C₃N₄ and the interaction energy with CO₂, which is evaluated at −0.59 eV. This value indicates the very good affinity of g-C₃N₄ nanosheets to CO₂ molecules. Our findings shed light on the potential for g-C₃N₄ to be used for the development of high-performing gas sensor devices.

1. Introduction

Natural and industrial air pollutions due to toxic gas emissions constitute one of the major threats to the environment, causing serious hazards to human health such as respiratory and cardiovascular diseases [1]. Hence, there is an urgent need to reduce these emissions by issuing strict government regulations [2] while developing materials that can detect these gases before they reach the atmosphere. For gas detection, two-dimensional (2D) materials appear among the emerging solutions for gas sensing [3,4]. Indeed, due to their high surface area and good electronic and optical properties, 2D materials have attracted great attention in several domains such as photocatalysis, photovoltaics and gas sensing [5]. Among 2D materials, graphitic carbon nitride g-C₃N₄ (gCN) has shown several advantages for the development of gas sensors at a low processing cost and with low toxicity and high stability [6,7]. Indeed, gCN exhibits a planar structure similar to graphene, where C and N are covalently bonded, and adjacent layers are linked together by Van Der Waals forces. gCN has a tri-s-triazine cyclic structure, and the strong covalent coupling between C and N atoms offers this material very a high chemical and thermal stability under ambient conditions [8]. In addition, gCN can be used as a non-toxic organic semiconductor material possessing a bandgap energy of ~2.7 eV [9,10], making it a suitable catalyst for hydrogen evolution reaction, energy storage, contaminant degradation, CO₂ reduction and gas sensing [11]. In gas sensing applications, the performance of single-layer gCN remains attractive for two main reasons: (1) its intrinsic vacancies provide gCN with natural pathways for gas diffusion, unlike the artificial defects introduced in the other 2D materials [12,13]; (2) its large specific surface area offers gCN a high gas adsorption capacity [14]. gCN is easily synthesized by the thermal polycondensation process from affordable precursors such as thiourea (CH₄N₂S) [15], melamine (C₃H₆N₆) [16], dicyandiamide (C₂H₄N₄) [17] and urea (CH₄N₂O) [18], suggesting that it is one of the most promising candidates for industrial scaling up for various applications.

In the present work, a step-by-step fabrication and characterization of gCN are carried out to explore the routes of obtaining pure gCN to be used for gas sensing applications and especially CO₂ detection. Moreover, first-principles calculations are also performed to determine the optoelectronic properties of gCN responsible for its outstanding CO₂ detection.

2. Materials and Methods

Thiourea (CH₄N₂S, 99% purity, Alfa Aesar A12828.0E) and chloroform (CHCl₃, 67-66-3) were obtained from Fisher Scientific (Waltham, MA, USA). A crucible of 100 mL (Cole Parmer, Vernon Hills, IL, USA) and a furnace (Carbolite Gero, AAF 1100, Hope, UK) were used to synthesize gCN. Fourier transform infrared (FTIR) spectroscopy was conducted using KBr discs on an ABB Bomem FTLA2000 spectrophotometer. Thermogravimetric analysis (TGA) was carried out under an air atmosphere on a TA Instrument Q500 thermal analyzer. The crystal structure of thiourea was determined using X-ray powder diffraction (XRD) recorded on a Bruker-D4 diffractometer operating with Cu Kα irradiation. The vibration modes were recorded using a micro-Raman Renishaw spectrometer using UV light excitation (325 nm). A scanning electron microscope (SEM) Quanta 200 FEG and Transmission Electron Microscope (TEM) Tecnai from Thermo Fisher Sientific were used to examine the microstructure of the samples. A Retsch Milling Sieving Assisting machine was used for powder grinding, and a Palmsens 4 electrical measurements station was used for electrical measurements. A micro-probe system from the Nextron company (Basel, Switzerland) was used for gas detection tests (Figure 1). This system allows in-situ measurements of electrical properties under different environmental conditions: vacuum, temperature and controlled gas flow. The full sensing measurement set up was fully controlled by integrated software from Nextron. All detection tests were carried out in a gray room with a controlled relative humidity of 47%.

The gCN nanosheets were elaborated by the thermal polycondensation of thiourea (CH₄N₂S). For this, 45 g of thiourea was placed in a closed alumina crucible and heated at 550 °C in a muffle furnace for 2 h at a heating rate of 5 °C/min. After natural cooling to room temperature, the obtained compound in the form of yellow powder was subsequently crashed using a ball milling machine. The powder was then exfoliated using ultrasound for 3 h in chloroform solution. The resulting solution was sprayed to form a thin film onto a substrate with interdigitated electrodes using the spray coating technique. The mixture of chloroform and gCN powder at 5.3 × 10⁻² g/L concentration was sprayed at a 1 mL.mn⁻¹ flow rate for 40 min while heating the substrate at 400 °C. During the whole spraying process, the air pressure was maintained at 2 bar, and the spraying nozzle was placed 20 cm from the substrate.

3. Computational Method

The computed gCN has a hexagonal crystal structure, “space group P$\overline{6}$m2 (No. 187)”. The valence electron configurations selected for C and N atoms are 2s² 2p² and 2s² 2p³, respectively. The values of the relaxed lattice constants for bulk gCN are a = 7.135 Å and c = 7.040 Å. Figure 2a shows the 2 × 2 × 2 k-points used to obtain the band structure along the path H-K-Γ-M-L-H in the Brillouin zone. For the gCN monolayer (Figure 2b), 2 × 2 × 1 k-points were used, and a 15 Å vacuum space along the z-axis was added to isolate the gCN and prevent interactions between the adjacent layers.

In the structure of gCN, the C and N atoms are represented, respectively, in black and in blue. The gCN monolayer model is obtained by removing other layers from the unit cell of the bulk gCN along the c direction. The structure of gCN includes eight N atoms and six C atoms for each layer. The energy cutoff of the plane basis was set to 500 eV. The calculations were performed using the Perdew–Burke–Ernzerhof (PBE) exchange-correlation function with generalized gradient approximation (GGA) and the Heyd–Scuseria–Ernzerhof (HSE06) hybrid function, implemented in Cambridge Sequential Total Energy Package (CASTEP) simulation software (22.1.1 version, CDG, Cambridge, UK).

4. Results and Discussion

4.1. Material Characterization

Figure 3 shows the TGA weight loss of thiourea. The first degradation is observed at ~200 °C with two endothermic peaks caused by the sublimation of thiourea. In the temperature range of 225–400 °C, there is a slight weight loss due to the thermal condensation of thiourea into triazine (melamine) and tri-s-triazine (heptazine). Tri-s-triazine could form by the intermolecular rearrangement of triazine. A second degradation is observed around ~450 °C, which corresponds to the onset of polymerization of gCN. The polycondensation of thiourea gives several products before the complete formation of gCN, such as triazine and tri-s-triazine [19].

Figure 4a shows the XRD diagram of gCN. The main peaks are located at 27.25° and 12.99°, indicating the formation of pure gCN. The XRD peak shapes confirm the graphitic stacking structure, highlighted by the presence of two peaks characteristic of gCN bulk and gCN nanosheets. The intense diffraction peak at 27.25° corresponds to the (002) reflection of the graphitic interlayer stacking structure of gCN, and the weak diffraction peak at 12.99° is attributed to the (100) plane of the planar repeating units in gCN. The crystallite size d is estimated using Scherrer’s equation—d = k.λ/β.cosθ—to be ~9.28 nm and ~3.12 nm, respectively, for gCN bulk and nanosheets, where k is the shape constant (0.94), λ is the wavelength, θ is the Bragg angle and β is the peak full width at half-maximum. The XRD pattern of the polymeric gCN nanosheets indicates a graphitic structure with an interplanar stacking distance of aromatic units of ~0.33 nm.

The FTIR spectra of the elaborated gCN are provided in Figure 4b. Two main peaks at 809 cm⁻¹ and 887 cm⁻¹ show the presence of s-triazine units, caused by the out-of-plane bending vibrations of the triazine ring. The peaks appearing at 1249 cm⁻¹, 1328 cm⁻¹, 1413 cm⁻¹, 1467 cm⁻¹ and 1563 cm⁻¹ are ascribed to the stretching vibrations of the C–N bonds in tri-s-triazine. The peak at 1642 cm⁻¹ is due to the stretching vibration of the C=N bond in the heptazine units. The broadband absorption peak at 3169 cm⁻¹ shows the stretching vibration of the N–H bond of NH and NH₂ groups, presumably caused by the incomplete polycondensation of gCN. As illustrated in Figure 4b, the FTIR spectra of gCN nanosheets and bulk gCN have the same chemical structure.

Note that, under visible light excitation, the Raman spectra of gCN present a single broad response, which is due to the strong fluorescence and photoluminescence effects occurring in the graphitic gCN phase when excited by visible light [20,21]. Hence, to better apprehend the Raman response, Raman analyses are conducted using UV laser excitation. Figure 5 depicts the temperature-dependent Raman spectra of the gCN sample collected in the frequency range of 300–1800 cm⁻¹. The Raman response at room temperature of gCN is in good agreement with the reported results [21,22,23]. Moreover, one can note the presence of high intensity Raman modes at ~705 cm⁻¹, ~764 cm⁻¹, ~977 cm⁻¹, ~1113 cm⁻¹, ~1400 cm⁻¹, ~1483 cm⁻¹, ~1576 cm⁻¹ and ~1620 cm⁻¹.

The strong Raman vibration modes observed at 705 and 764 cm⁻¹ are related to the intraplanar bending vibrations of the C−N=C linked heptazine units. Moreover, the mode at ~977 cm⁻¹ was reported to be due to the symmetric N-breathing mode of the tri-s-triazine units [24]. These last modes are observed to present a blue shift as the temperature increases. Besides, the modes assigned to the D (1400 cm⁻¹) and G (1576 cm⁻¹) bands for the C sp² sites are also visible. The D band has an A_1g symmetry attributed to the breathing modes of C-sp² atoms in rings, and it is linked to structural defects, whereas the G band has E_2g symmetry involving the in-plane bond stretching motion of all C-sp² atoms and is linked to the degree of gCN graphitization [25,26]. As can also be seen, the temperature increase induced a weakening of the D and G Raman intensities ID and IG, respectively. To investigate the influence of the temperature change on the gCN structural distortions, the D and G bands’ intensity ratio ID/IG was monitored. Indeed, our findings show that when the temperature increased from ambient to 600 °C, the ID/IG ratio also increased. This is the signature of a high degree of structural distortion in the gCN. In addition, all vibrational modes are seen to downshift at lower frequencies with increasing temperature. Of particular interest, these results concur with a high thermal stability of our elaborated gCN nanosheets for temperatures as high as 600 °C since no gCN degradation was registered.

The microstructures of elaborated gCN were analyzed by SEM and TEM, and the respective micrographs are depicted in Figure 6.

The low and high magnification SEM images of dispersed nanosheets of gCN are given in Figure 6a,b, respectively. The nanosheet form is noticeable in these images, exhibiting a lateral size ranging from a few nanometers to a few micrometers. The TEM image shown in Figure 6c highlights the layered nanosheet morphology of the elaborated gCN. The size of the nanosheets is in the range of hundreds of nanometers. The stacking layers are evidenced; one can count four to five layers in Figure 6c.

4.2. Optical Properties

The band gap energy (E_g) of elaborated gCN is extracted from the absorption coefficient according to the model proposed by Tauc [27]:

$$\alpha h\nu =A{(h\nu -E_g)}^n,$$

(1)

where hν is the photon energy, E_g is the band gap energy, A is a constant, and n is an exponent indicating the optical absorption nature. Plotting (αhν)^1/2 against photon energy provides the indirect optical gap value. The band gap energy results are extracted here from the optical transmittance data of Figure 7a for the gCN bulk and monolayer and inserted in Figure 7b.

The gCN nanosheets exhibit higher transmittance in the full analyzed wavelength range of 300–1500 nm. The obtained band gap energy is 2.69 eV and 2.86 eV for the bulk and the nanosheets, respectively. These values are comparable to reported values [28,29,30].

4.3. Optoelectronic Properties

The band structures of gCN bulk and nanosheets, determined with different approximations, show that gCN has an indirect band gap with a minimum of the conduction band at point M of the Brillouin zone and the maximum of the valence band at point G of the Brillouin zone (e.g., Figure 8).

The computed band gap energies using GGA-PBE approximation show very low values of 1.15 eV and 1.32 eV for the gCN bulk and monolayer, respectively, which are surprisingly highly underestimated. However, using HSE06 approximation, the obtained band gap energies values are 2.47 eV and 2.72 eV for the gCN bulk abd gCN monolayer, respectively. These values are in agreement with the experimental results. We have gathered in the following

Table 1. gCN band gap energy comparison—experimental and theoretical.

System	Theoretical “DFT” Eg (eV)	Approximation	Experimental Eg (eV)
gCN bulk	1.36 [31]; 1.41 [32]; 1.15 [this work]	GGA-PBE	2.51 [33]; 2.59 eV [34]
	2.73 [31]; 2.71 [6]; [35],	HSE06
	2.68 [36]; 2.80 [37];		2.69 eV [this work]
	2.47 [this work]
gCN monolayer	1.59 [38]	GGA-PBE	2.70 [39], 2.71 [40]
	1.32 [this work]
	2.76 [41], 2.93 [42], 2.96 [43], 2.97 [38],	HSE06	2.86 eV [this work]
	3.20 [44], 3.22 [45], 4.02 [46]
	2.72 [this work]

various reported band gap energies of the gCN bulk and monolayer computed using both approximations to be compared to experimental values.

From this table, it is clear that the GGA-PBE approximation provides underestimated band gap energy values, as reported in various work, while the HSE06 approximation shows a very good agreement with the experimental results.

The total and partial densities of states (TDOS and PDOS) of the gCN bulk and monolayer are then computed using the HSE06 approximation and given in Figure 9a,b, respectively.

TDOS provides the overall states of both N and C atoms, whereas PDOS permits us to segregate both constituents. Indeed, our PDOS results show that the energy band states near the Fermi energy in the valence band are mainly occupied by the electrons of the N atoms, and only a small contribution of the electrons is derived from C atoms. In contrast, the conduction band is mainly constituted of atomic orbitals of both N and C with a large contribution from the C atom. This band is dominated by a strong hybridization of the electrons of the N and C atoms. Hence, both atoms could participate in gCN interactions with volatile gas, providing more possible sites for gas detection.

Moreover, we have determined the adsorption energy (E_ads) between the CO₂ gas molecule and the gCN monolayer using GGA-PBE approximation [19]. The obtained adsorption energy of the system is calculated using the following relationship [19]:

$$E_{ads}=E_{gCN-CO2}-\left(E_{gCN}+E_{gas}\right),$$

(2)

where $E_{gCN-CO2}$ is the total energy of the adsorbed CO₂ molecules by the gCN surface, and $E_{gCN}$ and $E_{CO2}$ are, respectively, the total energies of gCN and a CO₂ molecule. A positive adsorption energy value indicates that the reaction is endothermic and therefore unstable and energetically unfavorable, while a negative value concurs with energetically favorable adsorption as the reaction is exothermic and thus stable.

Table 2. Structural comparisons after adsorption of CO₂ at the surface of gCN and their adsorption energy.

	Before Optimization	Optimized Structure
gCN-CO2 configuration
Angle (degree)	180	179.42
Bond length (Å)	1.181	1.172/1.170
Distance gCN–CO2 (Å)	2.5	3.24
Eads (eV)	−0.59

shows the optimized structure of gCN while performing the adsorption calculations, including the distance between the CO₂ molecule and the gCN edge and the bond length.

Our finding revealed an adsorption energy value $E_{ad}=-0.59 \mathrm{eV}$. This result indicates that the reaction is exothermic and therefore stable and energetically favorable. This is consistent with the high surface stability of gCN against CO₂ molecules, which could lead to its high reactivity and is beneficial to enhancing gCN sensing performances. From the above, one can notice that the elaborated gCN nanosheets are highly suitable for CO₂ sensing.

4.4. Gas Sensing Tests

In this work, we have used a resistive gas sensor based on a charge transfer process, in which the sensor acts as a charge donor [47]. Once the sensor device is exposed to CO₂ gas, it is assumed that the charge transfer reaction occurs in different directions between the sensor and the adsorbed gases. This results in changes in the resistance of the sensor [48]. Adsorption-based sensors use the change in resistance of the g-CN nanosheet caused by the adsorption of CO₂. A physical adsorption takes place at the g-CN–CO₂ interfaces, leading to a reduction of CO₂ to ${\mathrm{CO}}_2^{-}$; thus, electrons exchange, modifying the electrical performances of the material as per the following reaction [49]:

$${\mathrm{CO}}_{2 \left(\mathrm{gas}\right)}+{\mathrm{e}}^{-}\leftrightarrow {\mathrm{CO}}_{2\left(\mathrm{ads}\right)}^{-},$$

To investigate the device recovery time dependent on temperature, we heated the gCN device to fully restore its base strength; then, we conducted CO₂ detection tests at 30 °C, 35 °C and 40 °C while monitoring the recovery time as shown in Figure 10.

From the plots of Figure 10, the recovery time as a function of temperature indicates the shortest recovery time of ~36 s achieved at 40 °C, while higher recovery times were registered at 30 °C and 35 °C, at ~117 s and ~50 s, respectively.

Hence, in the following, all our sensing tests were carried out with a 120 s dwell interval, which corresponds to a recovery time of 36 s achieved at 40 °C. A good recovery time was already reported to be considered as an important indicator of a high-performing gas sensor [50,51,52,53]. In this sense, the gCN nanosheets were tested for CO₂ detection to evaluate their stability, sensitivity and response. The gCN device was then exposed cyclically to three CO₂ concentrations at 40 °C—200, 600 and 1000 ppm—each for 60 s and maintained at 120 s as dwell time. To monitor the gCN response, a voltage sweep of 2 V at a 0.1 V step was conducted, and the change of the gCN resistance during the CO₂ exposition was collected. The result is illustrated in Figure 11.

As can be seen in Figure 11, for all the three concentrations, the gCN device showed a very fast response of about 730 ms once in contact with CO₂ and a very good recovery time of about 62 s, especially at the lower CO₂ concentration of 200 ppm. These results indicate the high CO₂ reactivity of the elaborated gCN device, as predicted by the adsorption energy calculations.

Moreover, to examine the sensitivity, we exposed (two cycles) the gCN device to CO₂, successively using the following concentrations: 200 ppm, 600 ppm and 1000 ppm. The sensor response is presented in Figure 12.

In this test, the device was repeatedly exposed to gas pulses for 60 s at each concentration, separated by periods of 120 s where the CO₂ stream was cut. The resulting resistance of the device gradually increased with increasing CO₂ gas concentrations. For instance, 5.26 × 10⁷, 5.85 × 10⁷ and 7.17 × 10⁷ kΩ corresponded to 200, 600 and 1000 ppm CO₂ in the first cycle and 6.86 × 10⁷, 7.19 × 10⁷ and 7.50 × 10⁷ kΩ in the second cycle, respectively. This increase in the resistance demonstrates that the gCN device exhibited a very good sensitivity to CO₂ concentration changes.

5. Conclusions

gCN nanosheets were successfully fabricated by the thermal polycondensation of thiourea. The experimental and theoretical investigations have demonstrated an indirect band gap of approximately 2.47 eV and 2.72 eV for bulk and monolayer gCN, respectively. The elaborated gCN was then deposited by spray coating and tested for CO₂ gas sensing. We have evaluated the adsorption of CO₂ into the gas sensor at −0.59 eV, highlighting the very good affinity of gCN nanosheets to CO₂ molecules. The sensing tests revealed that the gCN-based gas sensor has a high CO₂ detection capability of a few hundreds of ppm, high sensitivity and a fast recovery time of 36 s recorded at 40 °C. These findings are expected to provide insights for the low-cost incorporation of gCN to develop very promising, scalable and high-performance gas sensor devices.