The semiconductor metal oxide (SMO)-based gas sensor, considered the current workhorse of semiconductor-based chemiresistive gas sensor technologies, requires high temperatures to initiate the surface reactions which result in the sensing response, making it difficult to fabricate and prone to high mechanical instability. Therefore, alternatives at lower temperatures are desired, where 2D materials seem to hold the most promise. Even at ambient temperature, their sensitivity is extraordinarily large due to their extremely high surface-to-volume ratio. However, some ongoing issues still need to be resolved before gas sensors based on 2D materials can be widely used and commercialized. The alternative room temperature solutions involve optical signals, either by designing an nondispersive infrared (NDIR) sensor based on the Beer-Lambert law or by introducing an additional UV illumination to SMO sensors. In both cases, complementary metal oxide semiconductor (CMOS) integration is not feasible, which is why continued interest in 2D-material-based gas sensors persists.
A large set of materials for gas sensing and gas sensor designs are under investigation and find themselves at vastly different maturity levels of realizable development. Several advanced gas sensing technologies which have already been commercialized by industry include electrochemical (EC) sensors, catalytic pellistors (CP), thermal pellistors (TP), piezo-electric (PE) sensors, photo-ionization (PI) devices, optical infrared (IR) adsorption sensors, and semiconductor metal oxide (SMO) chemiresistors [1][2][3][4]. These technologies are typically divided into two categories: One whose detection mechanism is based on changing a material’s electrical behavior after adsorption (e.g., conductivity, field effect) and a second whose detection depends on an induced change in another property (e.g., thermal, optical) [5].
The semiconductor-based chemiresistive technology provides an option with the lowest cost, footprint, and power dissipation, mainly as a consequence of its successful integration with complementary metal oxide semiconductor (CMOS) fabrication techniques. These characteristics are necessary in order to enable sensing solutions for portable technologies as well as IoE and IoT integration, while integration with CMOS further ensures a means for very high and reliable reproducibility [6]. It is crucial for the volume manufacturing of commercial devices that there be minimal inter-device variances and that there is high confidence in the capability to produce a device with predictable attributes and highly manageable tolerances. It also observes a low power dissipation attributed to the catalytic pellistor which is also inexpensive to manufacture and has a comparatively small physical footprint. However, compared to the semiconductor metal oxide (SMO) sensor, this device has far worse selectivity, lesser sensitivity, and a slower reaction time. Piezoelectric, photo-ionization, and infrared sensors all offer excellent sensitivity, but they all have the drawback of high power consumption, which prevents portability and internet-of-things (IoT) integration.
In comparison to existing alternatives, the SMO sensor appears to provide the most benefits towards potential IoT applications. Previous research has shown that it has several advantages leading to its commercialization [7][8][9][10][11][12][13][14], particularly in terms of response time, sensitivity, and possibilities for portability and down-scaling. The high sensitivity of SMO sensors is, however, only made possible by applying very high operating temperatures, which may compromise their reliability and durability. With the goal of CMOS integration for a chemiresistive gas sensing solution, the SMO sensor has emerged as the industry standard. A semiconductor-based approach, which operates at ambient temperature and can be designed with established CMOS fabrication techniques is, therefore, highly desired. Several sensors and prospective sensing materials, including two-dimensional (2D) semiconductors such as graphene and transition metal dichalcogenides (TMDs), are being investigated in this direction [15]. The fabrication of these films, and devices based on these films, is not trivial. Nevertheless, investigations in the past few years have shown some promise in fabricating films of high quality for sensing and have made significant progress [16][17].
For chemiresistors, the electrical resistance of the sensor changes as a consequence of the adsorption of gas molecules on its surface. The sensitivity S towards a particular gas is, therefore, determined by how much the resistance of the film changes after exposure to this gas, usually starting from a baseline using operating conditions in pure ambient air. However, it should be noted that some studies use inert ambient, e.g., N2, as the baseline resistance, which is why it is often difficult to truly compare sensitivities across a broad set of experimental literature studies. From all of the available types of semiconductor-based gas sensors, the chemiresistor is the easiest to fabricate, since it only requires placing the sensing film on top of an insulating substrate and adding two metal contacts, which are often formed using interdigitated electrodes, as shown in the figure (Reprinted with permission from Cao et al. [18], CC-BY 4.0). The read-out circuit for the chemiresistive gas sensor is also quite straight-forward since only the current flowing through the resistor needs to be measured, which can be performed by placing a load resistor and reading the voltage across it [19][20]. The ease of chemiresistive gas sensor fabrication brings about the potential for their cost-efficient fabrication, ability of scaling, and eventual CMOS integration (Figure 1).
The field-effect transistor (FET) is the workhorse of the semiconductor industry and is the cornerstone of digital logic and many memory devices. In addition, the field effect, as implemented in a FET or a metal-oxide-semiconductor FET (MOSFET), has been extensively used in advanced bio-sensing technologies with the biologically sensitive field-effect transistor (BioFET) [25][26] and ion-sensitive field-effect transistor (ISFET) [27][28] designs. These devices provide a means to adjust the electrostatics of the channel layer using an electrode, which is suspended in an electrolyte solution, which is in contact with the gate dielectric. The sensitivity of the BioFET and ISFET structures are governed by how the charge accumulation changes on the gate dielectric and the selectivity is introduced by placing immobilized receptors on top the gate, which will only react to specific bio-molecules. However, generating such a structure, which has a high sensitivity and selectivity, for gas sensing is more challenging. In bio-sensors and BioFETs the electrolyte serves to keep undesired molecules out of the gate or channel regions. For gas sensors, on the other hand, exposing the channel to the ambient means it is exposed to any molecule which may come in its vicinity, shown in figure (Reprinted with permission from Cao et al. [18], CC-BY 4.0) (Figure 2).
The semiconductor industry is clearly dominated by silicon, and the introduction of new applications and technologies on silicon is usually a stepping stone towards mass production and adoption in the market. Over the years, many materials have attempted to replace silicon, including materials with higher charge carrier mobility, such as germanium, and various group III-IV materials. However, none have been successful in commercialization on a broad scale and have only made breakthroughs in certain niche markets [43]. The continued scaling of silicon appears to have reached saturation and sub-3 nm channels pose significant challenges due to increased variability and reliability issues, but also due to the limited carrier mobility at these reduced scales [44][45].
Beyond transistor scaling and digital logic, interest in 2D materials has intensified over the past years due to their potential usability in a wide range of applications. Due to the low dimensionality of these materials, they exhibit properties of relevance to several research fields from solid-state physics to low dimensional molecular chemistry. Therefore, even minor shifts in the chemical structure of the film’s surface can be felt in its bulk properties. This feature makes 2D materials ideal for catalysis [46][47][48][49] and sensing [50][51][52]. The range of application for gas sensing is immense and includes devices which are able to detect hazardous gases, organic vapors, and humidity: Devices frequently used for medical diagnostics, environmental monitoring, and safety and security [52]. TMDs including MoS2, MoTe2, WSe2, and SnS2 have already been widely studied for gas sensing in the FET configuration, primarily for the detection of nitrogen-containing compounds. In the chemiresistor configuration, a broader group of 2D materials have shown high potential for gas sensor development, including TMDs, boron nitride, black phosphorus, and MXenes [30][53].
A thorough review of the means by which 2D materials are synthesized is provided by Knobloch in [54]. Here, these processes are viewed from the perspective of CMOS integration. The initial discovery of 2D materials for electronic applications was enabled by the mechanical exfoliation of single or few layers of graphene from a layered bulk crystal graphite [55]. While this process is not scalable and does not lend itself to CMOS integration, it is still frequently used in lab-based research into 2D materials and devices due to the simplicity of the process and the reasonable quality of films which can be achieved. The process involves thinning down a thick layer of a 2D crystal by placing it on adhesive tape. Subsequently, by frequently folding and unfolding the tape, increasingly fewer layers remain. These layers are then transferred to a wafer, typically SiO2-on-silicon [54][56][57].
Chemical vapor deposition (CVD) is probably the most widely studied means of depositing a film and is a staple of the microelectronics industry and of a CMOS foundry. It is a relatively simple bottom-up growth process which offers flexibility of metal precursor and a relatively fast growth rate. Several studies have examined the growth of graphene, hexagonal Boron Nitride (hBN), and TMDs using CVD, while only a few studies have looked into growing black phosphorene with this method. This has to do with black phosphorus’ tendency to quickly oxidize in the presence of oxygen, requiring an oxygen-free growth environment. Furthermore, a precursor is required to grow the material. For MoS2, the precursor is the chalgogen sulfur, as shown in the above figure (Reprinted with permission from Shi et al. [58], CC-BY 4.0); for phosphorene, this would be phosphine, which is a highly toxic material [59] (Figure 3).
A typical alternative to CVD in a CMOS foundry is physical vapor deposition (PVD) which most often refers to thermal evaporation deposition or sputtering. PVD belongs to a family of synthesis processes which enable large-scale processing of 2D van der Waals (vdW) materials. The figure shows such a setup for the sputtering of MoS2 films from a target (Reprinted with permission from Pang et al. , CC-BY 4.0) [64]. There is no fundamental limit on the size or shape of the films which can be generated using this process, which has been used to produce thicker vdW films for decades [65][66], while the deposition of a few layers has also been recently demonstrated [67][68]. PVD also does not require the transfer of the grown material onto the desired substrate, as growth on any substrate is inherently possible (Figure 4).
Molecular-beam epitaxy (MBE) is a process by which epitaxial growth can be performed on a large scale in an ultra-high vacuum chamber with pressures in the sub 10−10 mbar range [72] and a typical MBE setup is shown in the above figure (Reprinted from Vegar Ottesen, CC BY 3.0, via Wikimedia Commons). In MBE, the precursor molecules form a film on a heated crystalline substrate, which only provides the crystallographic information for the formation of the new film [54]. Since the substrate does not provide any catalytic surface effects, MBE results in a direct in situ growth of vertically-stacked heterostructures [73] (Figure 5).
Figure 5. Schematic of the MBE setup.
MBE is a very powerful tool for growing high-quality crystalline 2D films and has been extensively used to realize many films, including graphene, TMDs, and elemental 2D materials [74]. In addition, MBE has been applied to grow vertical and lateral vdW heterostructures [75]. There are, however, several difficulties in integrating MBE within a CMOS technology flow. The high vacuum requirements and high process sensitivity to small variations are a particular concern for mass production. Therefore, the technology remains principally a research tool for studying the fundamental properties of various material systems [76].
Atomic layer deposition (ALD) is a method of thin film deposition which offers more control over film conformality and thickness than traditional CVD. A typical multi-step ALD flow is shown in the above figure (Mcat chem446, CC BY-SA 3.0, via Wikimedia Commons). Because it may be used to deposit technologically important oxides and nitrides, such as the gate oxide HfO2 and gate metal TiN, ALD has emerged as a fundamental technique in semiconductor processing for advanced nodes. This is primarily due to the self-limiting nature of the process. In simple terms, ALD requires at least two self-limiting steps, during which different gases are allowed to interact with the surface, in order to ultimately initiate the deposition of a single monolayer of a material. During each step, a surface catalytic reaction takes place, which ensures that the surface is covered with a specific precursor. This precursor then only reacts with the species which enters the chamber in the second step, thereby forming the desired film. In this way, ALD can be used to grow very precise thin films with excellent conformality and thickness control down to the angstrom level [77] (Figure 6).
Figure 6. Schematic of the cyclical two-step ALD process.
The ALD process also does not require very high temperatures, which means that it could be the solution for the BEOL CMOS integration of 2D materials [78]. There has already been a demonstration of the successful growth of ML and bulk MoS2 using ALD at 300 °C using MoCl5 and H2S precursors for Mo and S, respectively [79][80]. The major concern with ALD is that the deposition conditions, including the substrate material, have been shown to significantly impact the nucleation and growth of the films [81].
An additional concern with the integration of 2D materials with CMOS technology is the difficulty in patterning the films, which will be critically damaged if exposed to plasma etching. A recent study by Ahn et al. [82] showed that it was possible to simultaneously deposit and etch MoS2 layers using MoCl5 and H2S precursors at 400 °C. Essentially, the authors show a selective deposition process, whereby the Mo-precursor MoCl5 would not adsorb onto the SiO2 surface, while adsorbing onto the surface areas which were covered by aluminum, even after 400 ALD cycles.
The main sensing mechanism of SMO sensors is through the surface adsorption of oxygen ions, such as O2−, O−, and O2− [83]. The presence of the oxygen ions on the surface create a depletion region, which is then reduced when these ions react with gas molecules of interest. Therein lies the core of the sensing mechanism for SMOs: For n-type sensing materials (e.g., SnO2, ZnO, TiO2) the resistance will decrease or increase, depending on if it is exposed to reducing or oxidizing gases, respectively. The inverse is the case when a p-type SMO material is used (e.g., CuO, NiO, Cr2O3). For gas sensors based on 2D materials, however, the process does not require oxygen adsorption and the mechanism mainly follows the charge-transfer process [30]. This means that the sensing film will act as a donor or acceptor of charges from the adsorbed gas molecule during the charge transfer procedure. Since different gases are able to exchange charges with the 2D film, it is the amount of charge that is exchanged, leading to changes in the conductive behavior of the film, that can be used to classify the specificity of the gas sensor. It should also be noted that the adsorption of gas molecules on the monolayer surface of a 2D semiconductor typically results in a change in its band structure. The adsorbed molecule could introduce additional energy states, giving rise to a shift in the Fermi level.
These material systems and interactions with gas molecules are typically studied using ab initio calculations [84][85]. The interactions, when the 2D material is pristine and not defected, depend mostly on vdW interactions between the gas molecule and the film. However, these weak forces alone were unable to explain many observed changes in the electrical properties of 2D semiconductors under varying ambient conditions [84][85]. These changes are proposed to be induced by the interactions between the gas molecules and point defects in the 2D semiconductor. Many defects are noted at the edges and grain boundaries, while an atomic mono-vacancy can readily appear also in the surface of a crystalline film [86]. Several studies have taken to first principles simulations in order to understand the interactions between gas molecules and differently defected 2D semiconductor surfaces [84][87][88][89][90].The change in the conductivity of graphene and other 2D materials as a result of a changing make-up of the ambient environment is already proven, and publications in this direction are plentiful. Even as early as 2007, the Nobel laureates for graphene’s fabrication and characterization (i.e., Geim and Novoselov [55]) described and reported on graphene’s changing electrical properties due to exposure to NO2 and NH3 [91]. Specifically, they showed that the adsorbed molecules increased the charge carrier density of graphene, with paramagnetic molecules such as NO2 acting as an electron dopant. However, due to graphene’s lack of a band gap, researchers have been unable to develop a functional digital logic FET. As a result, other possible 2D materials have gained momentum, which exhibit the presence of a reasonably large band gap while also having tremendous potential for concurrent FET, optical, and sensing applications.
There has been significant interest recently in the application of 2D heterostructures in order to increase the sensitivity or selectivity of the thin 2D layers towards gas sensing [152][153]. The use of heterostructures is not new and has also been applied to SMO-based chemiresistive sensors in the past [154][155].
2D | Two Dimensional |
ALD | Atomic Layer Deposition |
BEOL | Back End of Line |
BTI | Bias Temperature Instability |
CET | Capacitive Equivalent Thickness |
CMOS | Complementary Metal Oxide Semiconductor |
EOT | Equivalent Oxide Thickness |
FEOL | Front End of Line |
FET | Field Effect Transistor |
HAADF | High Angle Annular Dark Field |
IRDS | International Roadmap for Devices and Systems |
MOCVD | Metal Organic Chemical Vapor Deposition |
PTCDA | Perylene Tetracarboxylic Dianhydrid |
RIE | Reactive Ion Etching |
SCTD | Surface Charge Transfer Doping |
SWCNT | Single Walled Carbon Nano Tube |
TEM | Transmission Electron Microscopy |
TLM | Transfer Length Measurement |
TMD | Transition Metal Dichalcogenide |