Tin Diselenide (SnSe2) Van der Waals Semiconductor: Surface Chemical Reactivity, Ambient Stability, Chemical and Optical Sensors

Tin diselenide (SnSe2) is a layered semiconductor with broad application capabilities in the fields of energy storage, photocatalysis, and photodetection. Here, we correlate the physicochemical properties of this van der Waals semiconductor to sensing applications for detecting chemical species (chemosensors) and millimeter waves (terahertz photodetectors) by combining experiments of high-resolution electron energy loss spectroscopy and X-ray photoelectron spectroscopy with density functional theory. The response of the pristine, defective, and oxidized SnSe2 surface towards H2, H2O, H2S, NH3, and NO2 analytes was investigated. Furthermore, the effects of the thickness were assessed for monolayer, bilayer, and bulk samples of SnSe2. The formation of a sub-nanometric SnO2 skin over the SnSe2 surface (self-assembled SnO2/SnSe2 heterostructure) corresponds to a strong adsorption of all analytes. The formation of non-covalent bonds between SnO2 and analytes corresponds to an increase of the magnitude of the transferred charge. The theoretical model nicely fits experimental data on gas response to analytes, validating the SnO2/SnSe2 heterostructure as a suitable playground for sensing of noxious gases, with sensitivities of 0.43, 2.13, 0.11, 1.06 [ppm]−1 for H2, H2S, NH3, and NO2, respectively. The corresponding limit of detection is 5 ppm, 10 ppb, 250 ppb, and 400 ppb for H2, H2S, NH3, and NO2, respectively. Furthermore, SnSe2-based sensors are also suitable for fast large-area imaging applications at room temperature for millimeter waves in the THz range.

Here, we unveil surface properties of SnSe 2 single crystals by means of surface-science experiments and density functional theory (DFT). We demonstrate that the stoichiometric SnSe 2 sample is chemically inert, while the presence of Se vacancies induces surface oxidation with the formation of a sub-nanometric SnO 2 skin. We also explore the capability of SnSe 2 to realize devices for sensors for detecting noxious gases and imaging applications with non-ionizing radiations. Especially, we show that chemical sensing is feasible only when the pristine SnSe 2 surface is transformed into an heterostructure of SnO 2 /SnSe 2 . Concerning photodetection, we report the design of broadband SnSe 2 -based photodetectors interplayed by synergistic effects of multiple mechanisms. Considerably, the effect of hot electrons in ultrashort channel devices under strong light coupling results in outstanding performance in term of high responsivity at THz frequency.

Materials and Methods
Single crystals of SnSe 2 were grown by Bridgman-Stockbarger (Figure 1c). Stoichiometric ratio of 1:2 was put on evacuated quartz ampoule. The growth was carried out in a vertical two-zone tube furnace. The obtained crystal was characterized with X-ray diffraction (XRD) on powders, as shown Figure 1c. From the XRD spectrum, we can conclude that the crystal structure is CdI 2 -type (space group P-3m1). The lattice parameters are a = 0.3804 nm and c = 0.6128 nm consistently with previous works [46][47][48][49]. We also carried out the XRD and Laue diffraction measurements on single crystals. Samples were exfoliated in situ by adhesive tape. The absence of contamination in grown single crystals is secured by the survey X-ray photoelectron spectroscopy (XPS) spectrum.
XPS experiments were carried out with synchrotron light at APE-HE beamline at the Elettra Synchrotron in Trieste, Italy.
High resolution electron energy loss spectroscopy (HREELS) experiments were performed with a Delta 0.5 spectrometer (Specs GmbH, Germany). Spectra were taken in specular geometry, with an impinging angle of 55 • with respect to the perpendicular direction to the surface. The impinging energy is 3.5 eV.
Theoretical methods are reported in Section S1 of the Supplementary Materials. Fabrication process and measurements of devices are reported in section S2 of the Supplementary Materials.

Chemisorption of O 2 and H 2 O on Bulk SnSe 2
The Raman spectrum of the grown SnSe 2 single crystal ( Figure 1d) shows the E g and A 1g modes at 109 and 184 cm −1 , respectively, congruently with previous reports [50][51][52]. The narrow (00l) diffraction peaks (Figure 1c) reveal the excellent crystallinity for our SnSe 2 crystals.
To model surface chemical reactivity, the differential enthalpy ∆H ads and the differential Gibbs free energy ∆G for the adsorption of water and oxygen at room temperature, as well as the decomposition energy ∆H dec for both gases were calculated. The possibility of different kinds of defects was energetically evaluated and the formation of one Se vacancy is particularly feasible (only 1.28 eV/Se). The influence of Se vacancies was explored from a single vacancy in the outermost surface layer (Figure 2a) up to larger concentration of Se vacancies. We also calculated the different possible positions for a second Se vacancy, finding that the most energetically favorable location (0.98 eV/Se) is to have the second vacancy in the next neighbor to the first vacancy. To model such a large number of vacancies, an outermost SnSe-like layer on the surface was also considered (Figure 2b). Calculations indicate that physisorption of molecular oxygen is feasible at all investigated surfaces, although it is more energetically favorable at Se vacancies (∆G = −26.3 kJ/mol) rather than on defects-free SnSe 2 (∆G = −3.2 kJ/mol). The subsequent decomposition of molecular oxygen is an exothermic process for SnSe 2 , SnSe 1.88 , and SnSe. However, in the case of SnSe 1.88 and SnSe the differential enthalpy of decomposition is much lower than the defect-free SnSe 2 (−135.7, −236.1, and −42.3 kJ/mol, respectively). Thus, the oxidation rate should be greater on defective surfaces of SnSe 2 . After the decomposition of a single oxygen molecule, we assess the oxygenation of the whole surface, corresponding to an atomic structure with oxygen atom attached to each surface Se atom, and the successive oxidation in a metastable surface SnSe 2 O 2 phase (see Figure 2c), whose lifetime is estimated to be < 1 ms and, consequently, its presence on the surface could be detected only with time-resolved experiments. Hence, first the Se atoms in the SnSe 2 O 2 migrate to occupy the Se vacancies formed in the subsurface region and then the oxygen atoms from the SnSe 2 O 2 oxidize the Sn atoms of the surface layer to form SnO 2 , see Figure 2d. The following chemical equation can describe this process as O 2 + SnSe 2 + SnSe 2-y → SnSe 2 O 2 + SnSe 2-y → SnO 2 + 2Se + SnSe 2-y → SnO 2 + SnSe 2-x (1) with x < y, where SnSe 2-y corresponds to Se defects in the substrate, which are partially or totally saturated by Se freed from top layer after formation of SnO 2 -skin. We calculated the necessary energy to heal the Se vacancy by extracting Se from the SnO 2 skin and the differential enthalpy for this reaction is −3.54 eV/Se for the oxidized surface of SnSe 2 and −2.97 eV/Se for oxidized surface of SnSe 1.88 . These values cause a preferential oxidation of Sn with respect to Se and, subsequently, the surface oxide layer should be modelled as a SnO 2 /SnSe 2 heterostructure.
Instead, the physisorption of water is energetically unfavorable on SnSe 2 . Although near the Se vacancy the energy barrier for water absorption decreases it always remains a metastable process with positive ∆G in the Se vacancy sites (∆G = +3.4 kJ/mol). Even in the defect sites, a subsequent decomposition of water is extremely unfavorable (∆H dec = +175.7 kJ/mol). This behavior can be understood because the water molecules interact with diselenides via the formation of non-covalent bonds between lone pairs of electrons on sp orbitals of oxygen and unoccupied orbitals of metal centers in substrate. In the case of SnSe 2 , some mismatch between the size of water molecule and lattice parameters of the substrate is not so favorable for the formation of the described non-covalent bonds. The contribution to the adsorption energy calculated from the energy cost of the substrate and molecule distortions decreases ∆H, making it lower than the T∆S contribution in the ∆G calculation. In the case of adsorption on an oxidized substrate, hydrogen bonds are established between water and substrate oxygen. The formation of these hydrogen bonds can occur at much broader range of positions of water on the substrate and, therefore, no contribution in ∆H comes from distortions of either substrate or water.

Experimental Validation of the Theoretical Model
Probing vibrational modes could afford additional information on surface chemical processes and, especially, physicochemical mechanisms ruling the formation of an oxide skin. In particular, high-resolution electron energy loss spectroscopy (HREELS) experiments on H 2 O-exposed tin selenides SnSex, with x ranging between 1 and 2 (SnSe, SnSe 1.4 , SnSe 1.7 , SnSe 2 ), indicate the lack of chemisorbed molecules resulting from the presence of H 2 O, this is evident from the absence of O-H streching at 410-420 meV (molecular water) and 450 meV (hydroxyl groups) in the spectra in Figure 3 [53]. These findings are consistent with the positive Gibbs free energy of adsorption (corresponding to energetically unfavorable water adsorption) in Table 1. For a better comparison, we report in Figure 3 the vibrational data obtained by exposing other chalcogenides to the same dose of water (10 5 L). Unlike the surface of SnSex, a stable adsorption of water molecules was found on PtTe 1.6 and, moreover, of hydroxyl groups on InSe. The absence of reactivity toward water of Snbased chalcogenides makes them suitable for catalysis (in particular, photocatalytic water splitting [38], and hydrogen evolution reaction [54]) and drug delivery [55] (also bearing in mind that neither Sn nor Se are toxic). , and SnSe (blue). We also report vibrational acquired in the same conditions for water-dosed InSe (red) and PtTe 1.6 (brown). The energy of the primary electron beam is 4 eV. Table 1. Differential enthalpy ∆Hads and differential Gibbs free energy ∆G for physisorption at room temperature and differential enthalpy of decomposition ∆H dec for molecular oxygen and water on the surface of bulk samples of SnSe 2 , SnSe 1.88 , and SnSe. For the decomposition of oxygen, we also report, in parenthesis, the differential enthalpy for the formation of an oxygenated surface and a SnO 2 -like layer.
The inspection of the excitation spectrum probed by EELS, extended up to the ultraviolet range of the electromagnetic spectrum (Figure 4b), enables monitoring the surface status with a technique with probing depth as low as (0.9 ± 0.1) nm in our experimental conditions [58], which is lower by more than 10 2 with respect to Raman spectroscopy and optical techniques. Specifically, the excitation spectrum of the as-cleaved SnSe 2 surface shows a main feature at 15.9 eV with a shoulder at 12.0 eV, ascribed to interband transitions from Se-4s core levels and, moreover, two weak losses at 7.5 and 26.8 eV. The excitation spectrum of the air-exposed SnSe 2 sample is dominated by an emerging broad mode centered around~18 eV, with two weak peaks at 7.5 and 26.8 eV, evidently insensitive to surface modification. Notably, polycrystalline SnO 2 films display the feature at 18.0 eV. Precisely, this feature was previously attributed to the sub-oxide SnO 2-x phases [59]. However, the inspection of density of states (DOS) in Figure 5 reveals that the mode at 18.0 eV is related to a single-particle transition starting from O-2s band in SnO 2 . The weak peaks at 7.5 and 26.8 eV are ascribed to interband transitions originated by Sn-5s and Se-3s levels, respectively. Further information on the surface properties was provided by the inspection of core levels by means XPS experiments performed on both pristine and powderized SnSe 2 single crystals. As a matter of fact, it is also important to assess the characteristics of the powderized material with the future implementation in devices in mind, for which the surface/volume ratio should be maximized in order to improve the performance. Figure 6 shows the Sn-3d and Se-3d core levels of the as-cleaved SnSe 2 single-crystal surface and for the same surface modified by O 2 dosage with a total dose of 10 5 L (1 L = 10 -6 Torr·s). The Sn-3d 5/2 core level of the as-cleaved sample has a binding energy (BE) of 486.8 eV (Figure 6a). Correspondingly, the Se-3d core levels exhibit a single peak with the J = 5/2 component located at BE = 54.1 eV, compatibly with previous reports for SnSe 2 [60] and with a lower BE compared to the case of SnSe (BE = 53.7 eV). Exposure to 10 5 L of oxygen and storage in air only caused a slight change in the core level of the Se-3d. A new component at BE = 54.7 eV in Se-3d core level arising from Se(0) segregation is observed [61]. The total spectral area of this new component is 5.4% for O 2 dosage and 2.6% for air exposure. Especially, from the Se-3d core-level spectra (Figure 6b) one can observe the absence of the SnO 2 component, which would be characterized by the J = 5/2 component at BE of~59-60 eV. Conversely, in the powderized single crystal, one spectral component arising from surface oxidation was found (Figure 7). Specifically, the new component in the Sn-3d 5/2 core level related to SnO 2 was observed at BE = 487.8 eV (54% of the total spectral area) [62,63]. Remarkably, even after powderization, no trace of SeO 2 is present, as indicated by the featureless Se-3d spectra in the range 59-60 eV [64]. This finding confirms our prediction that Se is only involved in a metastable oxide phase, which represents a precursor for SnO 2 formation. However, in the powderized sample, a different Se oxidation state is present, as revealed by a broad feature in the spectra. Precisely, we assign the component at 55.0 eV to Se −2 and the higher to Se −2+δ (0 < δ < 1) [65]. In the powderized sample, we estimate δ to be~0.15 ± 0.05 from the analysis of the survey XPS spectrum. Therefore, sub-stoichiometric SnSe 1.7±0.1 coexists with SnSe 2 . We also estimated the thickness of the SnO 2 surface layer by means of quantitative analysis of XPS data [66], finding a thickness of (0.8 ± 0.1) nm (~2.5 monolayers) without observable changes after an exposure of one week in air. It should be noted that previous reports indicated that the surface of sub-stoichiometric SnSe 2 (SnSe 1.71 ) grown by molecular beam epitaxy [67] is unstable, with the subsequent formation of SnO x and SeO x phases. Conversely, in our case, both SnO and SeO x are not present on the surface. The SnO 2 /SnSe 2 heterostructure is more sensitive to chemisorbed species with respect to pristine SnSe 2 . On the pristine surface, the absorption of water molecules generates local changes in charge density in proximity to the absorbed molecules on the surface layer due to a rearrangement of chemical bonds, with a charge transfer of 0.17 e-per water molecule (Figure 8a). Thus, we conclude that pristine SnSe 2 is unsuitable for humidity sensing. On the other hand, the absorption of H 2 O on SnO 2 /SnSe 2 is energetically favorable even above room temperature. We calculated the values of transferred charge from H 2 O to the SnO 2 skin that are 0.43 and 0.30 e − for one and two H 2 O molecules per supercell, respectively. Correspondingly, the density of states (DOS) is modified with a direct correlation with the coverage of the adsorbate (Figure 8b-d), thus indicating the suitability for humidity sensing, even at the lowest concentrations of H 2 O. Note that decomposition of water molecule on the SnO 2 /SnSe 2 heterostructure is an exothermic process (−121.4 kJ/mol) and the subsequent water splitting is unfavorable, favoring the reversibility of the process. This further supports the use of the self-assembled SnO 2 /SnSe 2 heterostructure for humidity sensing. . Panels (a-c) report the change of the charge density after adsorption of one water molecule on SnSe 2 , one water molecule on SnO 2 -skin-terminated SnSe 2 , and two water molecules on SnO 2skin-terminated SnSe 2 , respectively. Panel (d) represents the DOS of SnO 2 -skin-terminated SnSe 2 (black curve) and of the same system modified by the adsorption of one (red curve) and two (blue curve) water molecules. Fermi level is set at 0.

Gas Sensing
The evaluation of the stability of SnSe 2 -based systems at the temperature of 200 • C, used in a typical gas sensing experiment, was performed. For the surface model of bulk SnSe 2, trilayers of SnSe 2 with fixed lattice parameters were used. For free-standing bi-and monolayer optimization of both atomic position and lattice parameters was performed. This provides the contribution in the energetics of the adsorption from flexibility of free-standing few-layers. In order to make calculations more realistic, we also considered the presence of the Se vacancies in the top layer. The formation of the single vacancy turns the SnSe 2 to SnSe 1. 88 . Results of the calculations (see Table 2) demonstrate that physical adsorption on bulk and few-layers of SnSe 2 is unfavorable at 200 • C. Contrarily, the presence of Se-vacancies makes physisorption favorable. The decomposition of molecular oxygen is favorable in all cases except for monolayer SnSe 1.88 . This exception is due to the combination of distortions caused by the presence of vacancies and distortions created by formation of new Se-O chemical bonds. Note that this result is valid only for free-standing monolayers. In fact, the deposition of a free-standing monolayer on a substrate decreases the flexibility of SnSe x and makes the system closer to a SnSe x bilayer. Table 2. Differential enthalpy (∆H phys ) and Gibbs free energy (∆G) at an operational temperature of 200 • C for the physisorption of molecular oxygen at the surface of bulk, bilayer, and monolayer SnSe 2 and SnSe 1.88 . The differential enthalphy for oxygen decomposition (∆H dec ) is also reported. All energies are expressed in kJ/mol. The favorability of decomposition of oxygen molecule even at defects-free substrate of SnSe 2 could be a starting point for a possible oxidation in the defective areas of SnSe 2 (vacancies, edges, grain boundaries). The formation of SnO 2 skin on the surface of bulk SnSe 2 leads to a migration of Se-atoms to subsurface area with the passivation of Sevacancies in the sub-surface layers [68]. In the case of free-standing few layers of SnSe x , an unstable structure would be formed. Therefore, we exclude this configuration from further investigation. Precisely, the discussed unstable structure is related to few free-standing layers, but if we include the presence of the substrate, the formation of more ordered SnO 2 skin will occur. Thus, results obtained for SnSe 2 /SnO 2 could be extended to few layers of SnSe 2 deposited on a substrate.

System
Considering the possibility of the oxidation with the formation of a SnO 2 skin, we modeled the physisorption of different analytes on SnSe x and on the SnO 2 skin over SnSe 2 substrate (Figure 9). Results of the calculations (see Table 3) demonstrate that the adsorption of various molecules on SnSe x surfaces obeys the same principles as in the case of oxygen. Definitely, defect-free SnSe 2 surface is nearly inert for all considered analytes. The presence of Se vacancies turns the free energy in a negative value making the adsorption favorable. Similarly, the adsorption of all analytes on monolayer SnSe x is less favorable than the case of bilayer and bulk. In contrast to non-oxidized SnSe x substrates, the formation of SnO 2 -skin corresponds to a strong adsorption of all analytes. The formation of non-covalent bonds between SnO 2 -substrate and analytes corresponds to an increase of magnitude of the transferred charge (see Figure 8). Note that the magnitudes of the free energies of adsorption and transferred charges in the case of SnO 2 skin are much larger than in the case of adsorption on Se-vacancies (SnSe 1.88 surfaces). Additionally, in the case of the adsorption on the SnO 2 skin, the value of the transferred charge is strongly distinct for different analytes.  The theoretical model discussed above was confirmed by experiments. The sensing properties of SnO 2 /SnSe 2 heterostructures were measured by monitoring the resistance change upon exposure to different concentrations of gases. In Figure 10, the dynamic response-recovery curve with NH 3 (40 ppm) and NO 2 (1 ppm) is shown, with opposite signs in the response curve related to the opposite charge transfer, as depicted in the inset. Interestingly, sub-ppm limit of detection is feasible for both cases, with 250 and 400 ppb for NH 3 and NO 2 , with near-room-temperature operational temperatures.
The SnO 2 /SnSe 2 heterostructure exhibit significantly enhanced response and superb response/recovery characteristics, also for H 2 S, for which the lowest detection limit reaches 10 ppb with a response value of 2, which is below the acceptable ambient levels of 20-100 ppb set by the Scientific Advisory Board on Toxic Air Pollutants (USA) [69,70].
Similarly, H 2 detection is feasible using SnO 2 /SnSe 2 heterostructure [68] with a response of 3 with an operational temperature of 150 • C at a concentration of 100 ppm and a limit of detection of 5 ppm. Considering the quite reduced costs of raw elements, this could be relevant considering that in the growing market for H 2 -powered devices-i.e., fuel cells-there is a requirement for cheap H 2 sensors. Table 4 reports an overview of literature results on near-room-temperature sensing of H 2 , H 2 S, NH 3 , and NO 2 for SnSe 2 -and SnO 2 -based sensors, at their respective operational temperature and concentration. Sensitivities of 0.43, 2.13, 0.11, and 1.06 [ppm] −1 were reported for H 2 , H 2 S, NH 3 , and NO 2, respectively ( Table 5, also reporting the limit of detection).
The SnO 2 /SnSe 2 heterostructure exhibits higher performance in terms of high response to analytes, low detection limit, high selectivity, and low power consumption [69][70][71][72][73][74]. Therefore, based on both calculations and experiments, one can conclude that the SnO 2 /SnSe 2 heterostructure is a promising platform for gas sensing.

SnSe 2 -Based Sensors for Large-Area Imaging with Millimetre Waves
An efficient THz detector at room temperature needs two characteristics: high mobility and a channel with nanometer length in a FET. Alternatively, the electron heat-driven detection mechanism could be used [86], based on the specific geometry of the device and the thermal/electrical properties of the material. In fact, SnSe 2 has both a high thermoelectric power and a suitable bandgap [87], which are advantageous characteristics for the production of hot electrons and for photothermal conversion. To improve the THz photodetection performance of the SnSe 2 -based device, the relationship between channel length and photoresponse should be investigated. In FETs, an important issue concerning THz detection is that the channel length should be less than tens of nanometers at the expense of rectification ratio. As mentioned above, it is possible to exploit hot electrons to overcome this limit, related to the absorption efficiency of photons. Specifically, the performance of SnSe 2 -based THz detectors with a channel length of 6 µm and 100 nm, along with the simulated electric field intensity, were evaluated (Figure 11a). A short channel SnSe 2 device is reported in Figure 11a-i,ii. In Figure 11a-iii, one can notice how the reduction of the antenna gap increases the value of the square of the electric field intensity (E 0 2 ) in the channel by almost three orders of magnitude. To study the different photoelectric properties of the devices as a function of the channel length, the photocurrent, the response time, and the reactivity were measured simultaneously using two radiations 0.04 and 0.12 THz with a power density of 2.5 and 1 mW cm −2 , respectively. The photocurrent increases linearly under electrical bias as V DS from −0.1 to 0.1 (Figure 11b), due to the increased carrier drift-velocity and reduced carrier transit time. Moreover, the short channel devices show a photocurrent greater than three orders of magnitude at 0.04 THz, congruently with the theoretical predictions of Figure 11a-iii. By exploiting the antenna gap in short channel device, it is possible to concentrate the THz radiation in a very small spot, consequently improving the SnSe 2 intraband absorption capacity, thus hot carriers can be efficiently produced [88]. In Figure 11c, one can observe a good linearity over a wide range of the photocurrent, of the short channel device, as a function of the incident power. Using the equation I ph ∝ P β to fit the experimental data, β is~0.98 under positive or negative bias, which is an interesting value for high-contrast imaging. Even at the frequency of 0.12 THz, there is the same improvement effect, as evident from Figure 11d, confirming the broadband nature of THz field nano-focusing. Additionally, the response time at 0.12 THz for both short and long channel devices was also studied (Figure 11e). For the short channel device, τ rise =2.7µs and τ fall = 2.3 µs-i.e., a response time of approximately 16 times faster than in the long channel device (τ rise = 45.4 µs, τ fall = 46.5 µs). Obviously, the long channel device has a slower response time than the short channel device, making the short channel device advantageous in applications. The increase in V DS rapidly increases the bias voltage response to 0.12 THz (Figure 11f). For short channel devices, the maximum response is 2.5 A W −1 , which is approximately five times larger than graphene-based devices (20 V W −1 ) [89] and 12 times larger than black-phosphorus-based devices (7.8 V W −1 ) [90]. This indicates that short-channel SnSe 2 devices have superior performance respect to graphene-and black phosphorus-based devices. Exploiting the excellent performance of the SnSe 2 , a large area imaging device with THz transmission was created for identifying hidden objects. Using a radiation of 0.12 THz, the shape of the glue-jar in the box was clearly visible (see Figure 12b). Furthermore, it was possible to detect not only the shape of the glue-jar but also the position of the amount of the glue inside the jar. Another key feature of the THz detector is its stability. Differently from black phosphorus [92], SnSe 2 -based devices exhibited better stability, congruently with the ambient stability assessed in Figure 6 by surface-science techniques. The absence of noticeable modifications in the photocurrent at 0.12 THz radiation after a prolonged storage in air (extended up to one month, Figure 12a) corroborates the excellent stability of SnSe 2 -based optoelectronic devices.

Conclusions
Here, we elucidated the main features of chemical reactivity of SnSe 2 . While the stoichiometric single crystal is chemically inert to ambient gases, the presence of selenium vacancies drastically affects surface chemical reactivity. The SnSe 2 −x surface is transformed into SnO 2skin-terminated SnSe 2 , with a thickness of the SnO 2 skin estimated to be sub-nanometric.
Interestingly, while the self-assembled SnO 2 /SnSe 2-x heterostructure is an exceptional platform for detecting chemical species, as demonstrated for H 2 , H 2 O, NO 2 , NH 3 , and HS 2 , the pristine SnSe 2 is unable to detect the same species. Thus, our results highlight the pivotal role of Se vacancies in metal dichalcogenides, which can transform the system from ambient-stable to an ultrasensitive gas sensor by tuning the stoichiometry.
Moreover, SnSe 2 is particularly suitable for THz photodetection, based on hot electrons. The response speed and the reactivity of the device are significantly improved thanks to the short channel, which exploits the localization of the electrostatic field and the high thermoelectric value of SnSe 2 . Furthermore, the device has excellent stability, even when the uncapped active channel is exposed to air for long periods, thanks to the exceptional chemical inertness of the single stoichiometric SnSe 2 crystals. Accordingly, SnSe 2 -based photodetectors represent suitable and promising candidates for imaging applications for homeland security and quality controls. Funding: We acknowledge financial support from the Spanish Ministry of Science and Innovation, through project PID2019-109525RB-I00.

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