1. Energy Band Structure of Graphene

Graphene is a two-dimensional (2D) material with a monolayer of sp²-bonded carbon atoms. Electrons are delocalized and free to move in this 2D sheet. The valence and conduction bands of graphene are conical valleys, which are attached at Dirac points. The energy dispersion curve is linear around the Dirac points, leading to an extremely large mobility of the charge carriers. Pristine graphene is a zero bandgap semimetal, and its Fermi level is located at the Dirac point, corresponding to zero conductivity at T = 0 K under vacuum ^[1]. In practical situations, graphene possesses defects that cause a p-type doping ^[2]. The zero or quasi-zero bandgap of pristine graphene can be a hurdle to use it as a sensing layer ^[3]. However, the electron transfer with target gas molecules leads to the Fermi level shift. More precisely, when graphene reacts with electron-donor or electron-acceptor gas molecules, the zero conductivity disappears. Moreover, doping or defects can also shift the Fermi level, which makes graphene become a p-type or n-type semiconductor. Due to its energy band structure, graphene is very sensitive to Fermi level changes. Its conductivity can be easily modified by doping or a back gate voltage when the graphene sheet is fabricated as a field effect transistor sensor ^[4].

3. Preparation Methods of Graphene

Historical graphene was first prepared by mechanical exfoliation of graphite [4] ^[9]. For example, highly oriented pyrolytic graphite (HOPG) is attached to the photoresist layer over a glass substrate. Using scotch tape, thick flakes of graphite are repeatedly peeled off. Thin flakes left in the photoresist are released in acetone. The target substrate (such as SiO₂/Si wafer) is dipped in the acetone and then washed with plenty of water and propanol, which removes mostly thick flakes. In this case, some thin flakes are captured on the target substrate’s surface due to van der Waals and/or capillary forces. Although mechanical exfoliation can provide high-quality single-layer graphene, it is not a scalable technique, and other techniques were proposed.

Chemical vapor deposition (CVD) holds great promises for large-scale production of high-quality graphene ^[10],^[11]. Conceptually, graphene CVD is a simple technique: it involves the decomposition of hydrocarbon precursors (such as CH₄) on catalytic metal substrates at high temperature (about 1000 °C) in a controlled atmosphere or atmospheric pressure. Cu, Ni, Ti, Ru, Pd, Pt, Ir, Co, Au, and Rh can be used as the catalytic metal substrates ^[12]. Among them, Cu is a very attractive catalyst and extensively chosen, since very low carbon solubility of Cu allows self-limited graphene growth, leading to highly homogeneous graphene sheets.

Graphene has also been grown on the surface of single-crystal silicon carbide (SiC) by thermal decomposition ^[13]. Namely, Si sublimation from SiC surface yields single-layer or multi-layer graphene structure at the graphene–SiC interface. However, it is quite difficult to exfoliate or transfer from the SiC substrate.

For large-scale production of low-cost graphene, one of the most popular approaches is the use of strong oxidizing agents to obtain GO and then convert GO to reduced graphene oxide (rGO) by thermal treatment or chemical reduction. In 1958, Hummers ^[14] reported a method for the GO synthesis by using KMnO₄ and NaNO₃ in concentrated H₂SO₄. This method has broadly been used in gas-sensor fabrication because of its high efficiency. However, it still has a few drawbacks. For instance, the oxidation procedure releases toxic gasses and the control of the graphene oxide sheets is difficult. Recently, several modified methods were proposed in the literature to improve the control of the GO and rGO synthesis ^[15],^[16]. It is worth noting that the reduction in GO is not complete, and there are defects and OH groups in the final rGO ^[17],^[18].

4. Functionalized Graphene NH₃ Sensors

Pristine graphene without bandgap behaves like a semimetal ^[19]. It is not a good choice for the NH₃ detection, because it does not possess any functional groups or defects. Monocrystalline graphene of mechanical exfoliation is well known as pristine graphene. However, it includes few isolated point defects. When NH₃ molecules are attached on the isolated point defects, the resistance of the pristine graphene sensor does not significantly change. Since the donor electrons of NH₃ can transfer through low resistance pathways around the isolated point defects ^[20]. On the contrary, functionalized graphene is one of the most promising materials for the NH₃ sensor. Thanks to the abundant defects (grain boundaries, line defects, or a few point defects) and functional groups on graphene, they strongly adsorb NH₃ molecules. Around such defects, no low-resistance pathways exist. So, the resistance change caused by NH₃ adsorption is significant, leading to a higher sensitivity.

Non-functionalized graphene is nonselective to gases ^[21]. A variety of gas molecules can be adsorbed on it to give similar resistance changes. For p-type graphene, the adsorption of a reducing gas (either NH₃ or H₂S) donates electrons to the graphene and depletes the concentration of holes, increasing the graphene resistance. The adsorption of an oxidizing gas (either NO₂ or SO₂) accepts electrons from the graphene and enhances the concentration of holes, decreasing the graphene resistance. As a consequence, different gases may generate the same sensing signals ^[22]. Indeed, the first non-functionalized graphene sensor can adsorb not only NH₃, but also CO, H₂O, and NO₂ in the same conditions. Ultraviolet or infrared radiation is able to make graphene sensitive to different types of gases, allowing for certain selectivity. However, the light adsorption efficiency of graphene is weak ^[23], ^[24]. The selective capabilities of graphene can be easily enhanced by grafting functional groups on its surface. This strategy is followed by most of the researchers.

The functionalization (generating specific links and increasing the adsorbing sites) is a viable method to improve chemical inert and poor selectivity of graphene. The graphene functionalization can be classified to covalent and non-covalent strategies ^[25]. The former transforms graphene’s π-orbitals from sp² into sp³, disrupting graphene electronic and mechanical properties. Whereas, the latter provides functional groups to the graphene surface by electrostatic or π–π interactions, enhancing the graphene bonding ability and simultaneously preserving the graphene original properties (the high carrier mobility and favorable noise characteristics). It is, thus, important to optimize and improve non-covalent functionalization techniques.

Over the past few years, a number of functionalization procedures have been developed. Various materials, from metallic nanoparticles and metal oxides to organic molecules and conducting polymers, are used to accomplish the graphene functionalization for promoting a high sensitivity and great selectivity to NH₃ sensing. According to the interaction nature between the target molecule and the receptor attached to graphene (or rGO), sensing reactions are distinguished to physisorption and chemisorption. The physisorption depends on the van der Waals forces of attraction with low binding energy, leading to full and fast recovery, but low sensitivity and poor selectivity. The chemisorption depends on the formation of chemical bonds between NH₃ molecules and the sensing layers, resulting in good selectivity, but slow and incomplete recovery. For certain sensing layers, physisorption and chemisorption can happen at the same time. Both adsorptions can cause charge transfer, yielding p-doping or n-doping of the graphene film.