Bidimensional Engineered Amorphous a-SnO2 Interfaces: Synthesis and Gas Sensing Response to H2S and Humidity.

Two-dimensional (2D) transition metal dichalcogenides (TMDs) and metal chalcogenides (MCs), despite their excellent gas sensing properties, are subjected to spontaneous oxidation in ambient air, negatively affecting the sensor's signal reproducibility in the long run. Taking advantage of spontaneous oxidation, we synthesized fully amorphous a-SnO2 2D flakes (≈30 nm thick) by annealing in air 2D SnSe2 for two weeks at temperatures below the crystallization temperature of SnO2 (T < 280 °C). These engineered a-SnO2 interfaces, preserving all the precursor's 2D surface-to-volume features, are stable in dry/wet air up to 250 °C, with excellent baseline and sensor's signal reproducibility to H2S (400 ppb to 1.5 ppm) and humidity (10-80% relative humidity (RH)) at 100 °C for one year. Specifically, by combined density functional theory and ab initio molecular dynamics, we demonstrated that H2S and H2O compete by dissociative chemisorption over the same a-SnO2 adsorption sites, disclosing the humidity cross-response to H2S sensing. Tests confirmed that humidity decreases the baseline resistance, hampers the H2S sensor's signal (i.e., relative response (RR) = Ra/Rg), and increases the limit of detection (LOD). At 1 ppm, the H2S sensor's signal decreases from an RR of 2.4 ± 0.1 at 0% RH to 1.9 ± 0.1 at 80% RH, while the LOD increases from 210 to 380 ppb. Utilizing a suitable thermal treatment, here, we report an amorphization procedure that can be easily extended to a large variety of TMDs and MCs, opening extraordinary applications for 2D layered amorphous metal oxide gas sensors.

Two-dimensional (2D) layered transition metal dichalcogenide (TMD) and metal chalcogenide (MC) semiconductors, with near atomic-scale thickness, have been extensively proposed in the past decade as alternative materials for traditional nanocrystalline metal oxides (MO) for gas sensing applications.[1−4] Key advantages of these interfaces are represented by their high surface-to-volume ratios,[5] the direct-to-indirect band gap transition,[6,7] the occurrence of chemical terminations like edges, boundaries, and surface vacancies,[8−10] and the engineered functionalities by metal nanoparticle decoration or substitutional doping.[11,12] Despite these features, a substantial disadvantage of TMDs and MCs, adversely affecting sensors’ signal reproducibility, is represented by their intrinsic thermodynamic instability (ΔG < 0), leading to spontaneous oxidation in dry-/wet-air laboratory conditions.[13,14] In details, the displacement of sulfur, selenium, and tellurium atoms, operated by ambient O2 in MoS2 and WS2 sulfides,[15,16] MoSe2, WSe2, InSe, GaSe, and SnSe2 selenides,[17−20] and MoTe2 and WTe2[21,22] tellurides, stimulates the nucleation over step edges of amorphous oxidized states, which proceeds through basal planes, eventually passivating all the flake’s surface. This phenomenon is further enhanced when the sensor’s operating temperature (OT) is increased in the range of 25–150 °C to compensate for irreversible adsorption of gas molecules, as frequently experienced in metal oxide and 2D layered sensors.[23,24]

Spontaneous oxidation of chalcogenides represents, indeed, an excellent opportunity to synthesize new kinds of interfaces comprising thin layers of amorphous metal oxides grown over 2D layered crystalline materials, yielding a-MO/TMD and a-MO/MC heterostructures with unexpected applications in the field of catalysis and gas sensing.[25] On this account, we recently demonstrated by means of experiments and theory that it is possible to synthesize a-SnO2/SnSe2 heterostructures to detect NO2, H2, NH3, and humidity[26,27] by controlled oxidation in air of 2D exfoliated SnSe2 layers. Besides their excellent gas sensing response, these 2D amorphous/2D crystalline heterostructures unfortunately retain some remarkable limitations. The difficulty of controlling the final thickness of the growing a-SnO2 oxide over the 2D crystalline platform and the risk that the a-SnO2 film is not self-passivating, i.e., not protecting the underlying 2D layer from further oxidation, make their practical exploitation challenging.[28]

Departing from liquid-phase exfoliated SnSe2 layers (10–30 nm thick) and controlling the oxidation in air at 250 °C for two weeks, at temperatures below the crystallization temperature of a-SnO2 oxide (i.e., ≈280 °C), we show for the first time that the oxidation process of 2D SnSe2 can be successfully driven to the “core” of the flakes, yielding single-phase, fully amorphous 2D a-SnO2, which is stable up to 250 °C and sensitive to H2S gas (400 ppb to 1.5 ppm) and to humid air (10–80% RH (relative humidity) (RH @ 25 °C)) at a 100 °C operating temperature. Layered amorphous metal oxide sensors (LAMOS) like a-SnO2 can be easily manufactured in a thin-film form by standard spin-coating deposition techniques representing a new interface for chemoresistive gas sensing applications. The amorphization thermal oxidation process here validated can be extended to a large variety of TMDs and MCs, opening new opportunities for “LAMOS” interfaces with unexplored surface-science capabilities, probably well beyond gas sensing applications.

Results and Discussion

Thermal stability of exfoliated SnSe2 flakes has been preliminarily investigated by simultaneous thermogravimetric (TG) and differential thermal analysis (DTA) techniques in air and nitrogen atmospheres by heating as-exfoliated SnSe2 at 5 °C/min, to maximize the gain of the TG signal, to 1050 °C as shown in Figure a,b. In the range of 200–600 °C (Figure a), a weight increase of approximately 3.0% is recorded in static air corresponding to the onset of an exothermic peak of the DTA signal located at 340 °C. This result (i) is congruent with previous theoretical investigations predicting the formation of an intermediate SnSe2O2 oxide[26,29] and (ii) rules out any sublimation of SeO2 species (eventually associated with a weight loss in Figure a), as previously found for SnSe2 powders.[30] In the temperature range of 600–800 °C, the measured −44.8 ± 0.7% weight loss in air well agrees with the theoretical weight loss of −45.5% corresponding to the complete oxidation of SnSe2 to SnO2 (maximum rate at 629 °C).

Figure 1

Thermogravimetric (TG) and differential thermal analysis (DTA) plots of as-exfoliated SnSe2 flakes heated in air (a) and nitrogen (b) atmospheres at 5 °C/min from 25 to 1050 °C. Black and red lines refer to TG and DTA signals, respectively.

Heating in a N2 atmosphere in the 200–1000 °C range, as shown in Figure b, inhibits any weight gain, confirming the absence of substantial oxidation phenomena in a nitrogen atmosphere with increasing temperature. The measured weight losses in N2 of −22.4% (at 612 °C) and −67.9% (at 835 °C) can be further attributed to the conversion of SnSe2 to Sn2Se3[31,32] and the complete removal of Se and partial sublimation of Sn atoms, respectively,[31] as reported in the literature.

The surface chemical composition of annealed SnSe2 at 250 °C for two weeks has been investigated by XPS analysis. Figure a–c shows the detailed XPS Sn 3d, O 1s (b), and Se 3d (c) core-level spectra of SnSe2 flakes after two weeks of annealing in static air at 250 °C. Deconvolution of the Sn 3d5/2 core-level spectrum (Figure a) and quantitative analysis of the phases’ composition shown in Table highlight the complete oxidation of the surface, as demonstrated by the occurrence of two contributions ascribed to (i) stoichiometric SnO2 (orange line) with maximum peak intensity of the j = 5/2 component centered at 487.4 eV[33] covering approximately 97% of the whole spectral area and (ii) defective SnO2– (green line) at 486.6 eV[34] representing nearly 3% of the entire signal.

Figure 2

Deconvoluted Sn 3d (a), O 1s (b), and Se 3d (c) core-level spectra of SnSe2 flakes after two weeks of annealing at 250 °C. Raw data (empty gray circles) and cumulative fits (blue lines) are reported. (d) Survey XPS spectra of SnSe2 exfoliated nanosheets measured before (black) and after annealing (blue). Shaded gray areas highlight O 1s, Sn 3d, and Se 3d core levels.

Table 1

Relative and Cumulative Surface Atomic Concentrations (at. %) of Sn, O, and Se Elements

	Sn 3d5/2	O 1s	Se 3d5/2
single spectral area [arbitrary units]	1417	6143	245
relative percentages (with respect to cumulative counts of Sn 3d, O 1s, and Se 3d)	18%	79%	3%

The lack of any signal at 486.1 eV,[29,35] corresponding to the vertical dashed line of Figure a, confirms the absence of Sn–Se chemical bonds in the annealed sample. These results are congruent with O 1s and Se 3d core-level spectra of Figure b,c respectively. Specifically, the O 1s spectrum can be decomposed into three signals. The main peak, representing ≈50% of the total spectral area, is centered at 531.3 eV corresponding to Sn–O.[36−38] Those located at 532.2 and 532.7 eV are ascribed to SeO2[39,40] (≈10%) and adsorbed H2O[41] (≈40%), respectively. Finally, the Se 3d core-level spectrum (Figure c) comprises three contributions at 59.3, 56, and 54.6 eV, associated to SeO2,[41] metallic Se,[42] and SnSe2–,[18] respectively.

According to Table , having set at 100% the cumulative spectral areas of Sn 3d, O 1s, and Se 3d, net of the elemental sensitivity of the XPS technique,[43] the relative atomic percentages of Se, Sn, and O yield Se 3d ≈ 3%, Sn 3d ≈ 18%, and O 1s ≈ 79%. In particular, the elemental ratio of O:Sn is found to be close to 2:1, which supports the occurrence of the SnO2 phase.

Notably, the negligible contribution of the Se 3d signal in the annealed sample is confirmed in the survey spectra of Figure d exhibiting (compared by electronic magnification of the spectra in the red circles) the vanishing of the Se 3d signal in the annealed sample (blue line) with respect to the exfoliated SnSe2 one (black line). It may be concluded that after controlled thermal treatment, the gas-responding surface of the annealed SnSe2 flakes comprises almost stoichiometric SnO2 and a negligible amount, approximately close to the instrumental resolution of the XPS equipment (±1%), of SeO2 phases.

The amorphization process of SnSe2 at different annealing times and temperatures has been investigated by the grazing incidence (GI)-XRD technique over spin-coated and annealed thin films deposited over silicon substrates. According to Figure , as-exfoliated SnSe2 exhibits a major diffraction peak (black line) corresponding to the (001) plane of SnSe2 at 2θ = 14.4° (ICDD card no. 96-154-8806). At 250 °C, with proceeding the annealing time from one week (blue line) to two weeks (magenta line), the (001) peaks almost disappear (see also the inset of Figure ), confirming the effectiveness of the amorphization process. The two-weeks annealed sample at 250 °C (magenta), further annealed for an extra week at 280 °C, retains its amorphous structure (green), highlighting no substantial recrystallization phenomena of amorphous a-SnO2 into crystalline SnO2 as it will be further discussed in the HRTEM characterization.

Figure 3

Grazing incidence (GI)-XRD diffraction patterns of as-exfoliated SnSe2 flakes (a) and SnSe2 annealed at 250 °C for one week (b), 250 °C for two weeks (c), and 250 °C for two weeks + one extra week at 280 °C (d).

The microstructural evolution of SnSe2 flakes annealed at different times/temperatures, over selected flake’s regions (Figure S1), has been characterized by HRTEM microscopy and is shown in Figure . As-exfoliated SnSe2 (Figure a) exhibits a fully crystalline structure (SAED1), corresponding to the inner region of the flakes, with an interplanar distance of 0.30 nm (Figure e), congruent with the (001) crystallographic plane of SnSe2 determined by GI-XRD. Notably, an amorphous edge (SAED2), extending by approximately 8 nm inside the flake’s terrace, is also detectable (Figure e), which is congruent with an oxidation process mechanism advancing from the outside to the inside of the flakes, as extensively reported for TMD and MD materials.[13,18] Despite our previous research demonstrating no significant edge oxidation phenomena of liquid-phase exfoliated SnSe2,[28] in this case, we attribute the step-edge amorphization process of SnSe2 to the combined action of different sonicating conditions and the use of a different solvent (here NMP). With annealing at 250 °C for one week (Figure b,f), the degree of crystallization decreases compared to the as-exfoliated sample, as confirmed by the formation of halos in SAED patterns (i.e., compare SAED1 of Figure a with SAED2 of Figure b). The onset of an amorphization phenomenon is confirmed in Figure f where a patchwork of crystalline/amorphous phases is clearly displayed. With annealing for two weeks at 250 °C, crystalline domains of the parent SnSe2 completely disappear, as exhibited in Figure c,g, confirming the completeness of the amorphization process. To conclude, we also tried to investigate the recrystallization mechanism of amorphous a-SnO2 into crystalline SnO2, as shown in Figure d,h. We found that by an extra week of annealing at 280 °C, crystalline domains are initially formed on step edges, as shown in Figure d with corresponding interatomic plane distances of 0.33 nm (Figure h), attributed to the (110) plane distances of tetragonal rutile SnO2.[44] It turns out that the recrystallization of a-SnO2 into crystalline SnO2 proceeds from the outside to the inside of the flake, considering that no nucleating SnO2 crystallites are visible inside the flakes as shown in Figure d,h. The limited extension of the crystalline domains with respect to the amorphous ones shown in Figure d,h may also explain the lack of any diffraction peak attributed to crystalline SnO2 in the GI-XRD pattern of Figure .

Figure 4

HRTEM pictures of (a,e) as-exfoliated SnSe2 flakes and SnSe2 (b,f) annealed at 250 °C for one week, (c,g) annealed at 250 °C for two weeks, and (d,h) annealed at 250 °C for two weeks + one extra week at 280 °C. SAED patterns corresponding to the identified regions are shown in the insets.

In conclusion, the whole amorphization process here presented, comprising the recrystallization of a-SnO2 to SnO2, possibly represents interesting evidence of a nearly topotactic transformation of a 2D SnSe2 metal chalcogenide into a 2D a-SnO2 metal oxide, which includes loss of selenium and oxygen gain so that the final a-2D structure, retains the same bidimensional feature of the original material. This process represents a matter worthy of further investigation, eventually constituting a promising route to synthesize new metal oxide a-2D interfaces for gas sensing applications. Combining XPS, XRD, and HRTEM observations, it may be concluded that by annealing in air for two weeks at 250 °C, approximately 30 °C below the onset of the recrystallization temperature of a-SnO2, the complete oxidation/amorphization of exfoliated 2D SnSe2 flakes into 2D a-SnO2 is achieved.

Gas Sensing Response to H2S and Humidity

The best operating temperature (OT) for H2S and H2O of the two-weeks/250 °C annealed a-SnO2 sample has been identified in light of two main features of the sensor’s response: (i) sensor’s signal amplitude, as represented by the relative response ratio (RR = Rair/Rgas), and (ii) recovery of the baseline resistance after gas desorption (BLR, i.e., the resistance in air at equilibrium). Tests have been carried out in the OT range of 25–150 °C in dry-air carrier gas exposing the film to 1 ppm H2S and 40% relative humidity (40% @ 25 °C), as shown in Figure a,b respectively. By increasing the operating temperature, the sensor shows a monotonic decrease in the BLR, indicating a semiconducting behavior with an n-type response to both H2S and humid-air reducing gases, consistent with preliminary results on a-SnO2/SnSe2 interfaces[28] and metal oxide SnO2 sensors.[45]

Figure 5

Electrical responses of the two-weeks/250 °C annealed a-SnO2 sensor in dry air at different OTs (from 25 to 150 °C) to (a) 1 ppm H2S and (b) 40% RH (RH @ 25 °C). (c) Baseline resistance evolution in dry air at a 100 °C OT of (i) as-exfoliated SnSe2 (yellow region) and (ii) as-annealed (two weeks/250 °C) a-SnO2 (green region) and (iii) a-SnO2 thin-film baseline resistance randomly measured during one-year conditioning at 100 °C (blue region).

According to Figure a, in the temperature range of 25–75 °C, the sensor displays a low signal response to 1 ppm H2S and no recovery of the BLR (dashed black lines in the figure) even after 2 h of dry-air purge. With increasing the OTs, both the sensor’s signal and desorption kinetics improve. The BLR is fully recovered, departing from the 100 °C OT, topping the best sensor’s signal response at 150 °C. Poor recovery of the BLR in the temperature range of 25–75 °C, shown in Figure a, is a typical feature of a large variety of 2D TMD/MC gas sensor interfaces operated at near-room temperatures.[46] On this account, to improve gas desorption rates and BLR recovery, UV–vis light irradiation or increasing the sensor’s operating temperature in the 75–150 °C range has been proposed for both traditional metal oxide[47,48] and 2D TMD/MD sensors.[49,50] These strategies indeed show remarkable limitations. UV-light irradiation of 2D TMDs/MDs requires almost monolayer thin interfaces (i.e., <5–7 nm),[46,51,52] whereas thermal heating at higher temperatures stimulates fast surface oxidation, hampering BLR reproducibility over the long run. On this account, a-SnO2 annealed at 250 °C can be safely operated in the temperature range of 100–150 °C without any risk of further degradation or recrystallization, providing an effective solution for fast BLR recovery and improved sensor’s signal reproducibility. Water vapor at 40% RH (RH @ 25 °C) behaves like H2S, as shown in Figure b, with improved sensor’s signal and BLR recovery in the temperature range of 100–150 °C (see also dynamic humidity responses in the range of 25–75 °C in Figure S4). Notably, a complete recovery of the baseline is also recorded at 25 °C, possibly on account of a protonic (H+) Grotthuss chain-like conduction mechanism induced by physisorbed water at lower temperatures.[53] Departing from the 100 °C OT, most of the physisorbed water is removed[54] and water vapor responds as a reducing gas as previously reported.[55]

Baseline resistance reproducibility over the long run, under sustained OTs and dry/wet conditions, also represents a key issue for the exploitation of this new kind of interface. Figure c shows the evolution of the BLR in dry air recorded at a 100 °C OT of (i) as-exfoliated SnSe2 (yellow region), (ii) two-weeks/250 °C annealed SnSe2 (green region), and (iii) a-SnO2 thin films after one-year conditioning at 100 °C (blue region). The amorphization process, producing truly stable a-SnO2 oxide, sharply increases the BLR in dry air (yellow-green region), which finally stabilizes, exhibiting excellent reproducibility and stability (±5% of the BLR variation) over the long run (blue region). One-year recordings of the electrical resistance of a-SnO2 to 1 ppm H2S gas and 40% humidity, shown in Figure S5, confirmed a remarkable reproducibility with an associated uncertainty of the sensor’s signal to H2S gas and humidity as low as ±0.1 and ±0.2, respectively.

Figure a,b shows the dynamic resistance changes at a 100 °C OT of a-SnO2 to H2S (400 ppb to 1.5 ppm range) and H2O (10–80% RH range, RH @ 25 °C) in dry-air carrier gas, respectively. Both H2S and H2O at a 100 °C OT yield strong interactions with the a-SnO2 surface, indeed with excellent recovery of the BLR following each step of gas/humidity purge. The effect of the 40% RH background to the dynamic H2S response shown in Figure c (see also Figure S6 at 60 and 80% RH backgrounds) highlights that humid water decreases the BLR, though preserving a satisfactory H2S gas dynamic modulation.

Figure 6

Dynamic electrical responses in dry air at a 100 °C OT to (a) H2S (400 ppb to 1.5 ppm) and (b) H2O in the range of 10–80% RH (RH @ 25 °C). (c) Dynamic electrical responses of a-SnO2 at a 40% RH background and increasing concentrations of H2S (400 ppb to 1.5 ppm). (d) 40% RH humidity cross-response to 1 ppm H2S at a 100 °C OT: (i) first step in dry air and 1 ppm H2S; (ii) second step at a 40% RH background and 1 ppm H2S; (iii) third step, equivalent to (i), to check for short-term repeatability. (e) Log/log calibration plots at different RH values from 0 to 80% (as to the arrow) to increasing concentrations of H2S, measured at a 100 °C OT (associated standard deviations calculated over a set of five consecutive measurements); (f) adsorption and desorption times of H2S (1 ppm) to increasing RH values as to the red arrow.

The assessment of the sensor’s signal variations when both humidity and H2S compete at the same time over the sensor’s surface represents another key issue of the sensor’s performance. This feature, known as humidity cross-response (CR) to H2S sensing, is shown in Figure d. The test comprises (i) a first step in dry air and 1 ppm H2S, (ii) a second step at a 40% RH background and 1 ppm H2S, and (iii) a third step, equivalent to (i), to check for short-term repeatability. Comparing the sensors’ signals in dry (i) and wet conditions (ii), no relevant changes of the RRs (RR = Ra/Rg) are displayed, considering that 1 ppm H2S yields an RR of 2.4 ± 0.1 in dry air and an RR of 2.3 ± 0.1 in humid air (40% RH), provided an associated sensor’s signal uncertainty of ±0.1 (measured over a set of five consecutive measurements). Moreover, no substantial changes in the electrical response are displayed comparing panels (i) and (iii), demonstrating an excellent short-term repeatability.

The humidity cross-response (CR) to H2S sensing in the whole gas/humidity concentration range at a 100 °C OT, as represented by the log–log calibration plots of the sensor’s signal (i.e., RR = Ra/Rg) vs H2S gas concentrations and different RH values, is shown in Figure e. Increasing relative humidity from dry conditions to 80% RH (as to the direction of the arrow), sensor’s signal amplitude decreases. Specifically, at 1 ppm H2S, sensor’s signal amplitude decreases from an RR of 2.4 ± 0.1 at 0% RH to an RR of 1.9 ± 0.1 at 80% RH. Moreover, congruent with this tendency, water vapor has a negative effect on increasing the H2S limit of detection (LOD). By numerical extrapolation of the calibration lines of Figure e (according to the methods in Supporting Information, Section S3), the theoretical LOD increases from 210 ppb at 0% RH to 380 ppb at 80% RH, confirming the inhibiting effect of water vapor upon H2S sensing. The antisynergistic interaction of water vapor upon H2S detection is also confirmed in Figure f and Figure S7, showing the adsorption/desorption times of 1 ppm H2S as a function of the humidity content. Increasing RH% (as to the direction of the red arrow), the adsorption time of 1 ppm H2S increases from 9 (@ 0% RH) to 20 min (@ 80% RH), while the sensors’ signal amplitude decreases from an RR of 2.4 ± 0.1 at 0% RH to an RR of 1.9 ± 0.1 at 80% RH. As a concluding remark, the adsorption–desorption times here reported (order of minutes) are in most cases much longer than those frequently reported in the literature (order of seconds). On this account, it should be noted that response times are mostly dependent on the experimental conditions like the humidity content and gas fluid dynamics inside the test cell. As to the latter, the theoretical residence time of the gas (TRT) given by the ratio between the cell volume [cm3] and the gas flow rate [cm3/min] may significantly differ from the mean residence time (MRT), representing the actual time to completely fill/empty the test cell. In a previous paper utilizing NO2 as a marker gas, we found that the characteristic MRT of our experimental setup is between 4 and 5 min[28] with respect to a TRT of only 1 min. Humidity as well influences response time. As shown in Figure f, the response time to 1 ppm H2S almost doubles from dry to 80% RH wet conditions. In conclusion, when comparing response times of different sensors/gases, the experimental setup, the humidity content, and operating temperatures should always be considered and eventually normalized.

Gas sensing relative responses (RR = Ra/Rg) of selected sensors, calculated by normalizing literature data to 1 ppm H2S in dry air and at different OTs, are compared in Table . In addition to traditional porous metal oxide sensors[56] and metal oxide heterostructures, which guarantee the most favorable catalytic efficiency, a-SnO2 performs better than traditional 2D TMD/MD[1−4] sensors operated in the same temperature range (150–200 °C). 2D n-n/p-n heterostructures[1−4] operating at room temperature show RRs slightly smaller than that of a-SnO2, excluding n-n SnSe2/SnO2[29] with an associated sensor’s signal as high as 7.5 to 1 ppm H2S gas. Regarding n-n/p-n 2D heterostructures, which can be classified as “decorated” interfaces since they almost comprise 3D crystalline metal oxide nanoparticles grown over 2D TMD/MD flakes, it is not usually appreciated that the naked 2D TMD/MD surfaces of the heterostructure, i.e., the ones not protected by the “decoration”, are likely to be oxidized in dry/wet air in the long run, negatively affecting the reproducibility of the electrical signal.

Table 2

Comparison of the H2S Gas Sensing Performances of Different Sensors’ Interfaces Obtained by Normalizing Literature Data to 1 ppm H2S in Dry Air

sensing materials	H2S [ppm]	response Ra/Rg [−]	OT [°C]	ref.
3D metal oxide[56]
SnO2 porous NF	1	14.3	350	(45)
ZnO (thin film)	1	4.3	330	(57)
WO3	1	4.5	330	(58)
CuO	1	1.8	135	(59)
p-Co3O4	1	2.0	210	(60)
3D metal oxide heterostructures
MoO3/SnO2	1	9	115	(61)
Cu2O/CuO	1	6.3	95	(62)
rGO/WO3	1	7.7	330	(58)
ZnO/CuO	1	6.7	25	(63)
PdRh ZnO-HC	1	2.9	260	(64)
MoO3/WO3	1	14.0	250	(65)
2D transition metal dichalcogenides (TMDs) and metal chalcogenides (MCs)[1−4]
p-type WS2	1	1.1	200	(66)
n-type WS2	1	1.2	150	(67)
MoSe2	1	1.2	200	(68)
SnSe2	1	1.8	200	(29)
2D n-n/p-n heterostructures[1,2,4]
SnO2/SnSe2	1	7.5	25	(29)
CuO/MoS2	1	1.1	25	(69)
Ag-MoSe2/rGO	1	1.2	25	(70)
SnSe2/WO3	1	1.3	25	(71)
2D LAMOS (this work)
a-SnO2	1	2.4	100	this work

Theoretical Model of H2S and Humidity Adsorption

To support the experimental results, we carried out DFT atomistic simulations of H2S molecule adsorption utilizing the same theoretical model of water molecules anchoring on the amorphous a-SnO2/SnSe2 nanosheet (NS).[28,72,73] Three stable geometries, all chemisorbed, whose structure and energetics are shown in Figure (1–3) and Table , respectively, are found according to the theoretical procedure described in the Supporting Information. Specifically, Figure exhibits three configurations with broken H–S bonds (two O–H bonds are similarly formed), supporting previous findings over crystalline SnO2 that molecularly adsorbed H2S attack is not favored.[74] The first structure (1, see Figure , top) is characterized by a S atom bound both to one Sn (dSn–S = 2.47 Å) and to one O atom (dSn–O = 1.68 Å). This mechanism is exothermic by 1.38 eV (following eq S1 in the SI). Bader analysis confirms what is expected on the presence of a newly formed S–O bond, that is, a slightly positively charged sulfur atom (+0.19) because of the larger electronegativity of oxygen.

Figure 7

The three most stable optimized structures of H2S anchored on top of the a-SnO2 as obtained by DFT calculations on top of AIMD-calculated trajectories (see Computational Methods in Supporting Information, Section S4 for details) (yellow, S; mauve, Sn; red, oxygen; white, H atoms).

Table 3

Main Structural and Thermodynamic Data of the Three Most Stable Anchoring Mechanismsa

			Bader charge [charge units]
adsorbate	structure	Eads [eV]	S	H(1)	H(2)
H2S	1	–1.38 (−0.28)	+0.19	+0.58	+0.60
H2S	2	–4.52 (−3.42)	–0.71	+0.60	+0.64
H2S	3	–3.65 (−2.55)	–0.72	+0.60	+0.59

See Figure , 1–3. Eads is in eV, and the Bader charge is in elementary charge units. Eads are values obtained by combining AIMD+DFT approaches for Esurf_H2S and (in brackets) values obtained still by combining AIMD+DFT approaches for the calculation of both Esurf_H2S and Esurf terms in eq S1 in Supporting Information, Section S4, which is the same for C2, i.e., the most stable anchoring mechanism of H2O as described in ref (28). A Bader charge of >0 indicates charge transfer from the adsorbed species to the a-SnO2 surface. A Bader charge of <0 indicates electron charge transfer from the a-SnO2 surface to the adsorbed species.

The second structure (2 in Figure , middle) is the most thermodynamically stable (Eads = −4.52 eV) and is characterized by a sulfur atom bound to two Sn atoms. Such two Sn–S bonds are 2.44 and 2.46 Å, respectively, showing that the enhanced stability may be correlated with the shorter Sn–S bond length. In this case, the Bader analysis reveals an enhanced electron localization on S (−0.71), a fingerprint of the more marked electronegativity of S (compared to that of Sn, i.e., 2.58 vs 1.96).[75] The last structure (3 in Figure , bottom), where the S atom is three-fold coordinated with Sn atoms (dSn–S = 2.46, 2.55, and 2.94 Å), is still markedly stabilized, compared to the reactants, with an adsorption energy of −3.65 eV and charge distribution on S (−0.72) very close to that in 2. Regarding the H(1) and H(2) hydrogen adsorption modes, deriving from the rupture of the H2S molecule, the Bader charge associated to the formation of O–H bonds with the O atoms at the surface is similar for all the three structures with a positive sign confirming the direction of the charge transfer from the adsorbed species to the a-SnO2 surface.

An overall schematization of the most stable H2S adsorption configuration over the optimized initial clean a-SnO2 surface shown on the right-hand side of Figure a confirms the occurrence of a homolytic dissociation of H2S with the formation of a rooted S atom, two-fold coordinated with Sn lattice atoms indicated as (S)2Sn, and two rooted hydroxyls group (i.e., (OH)O).

Figure 8

Schematization of the adsorption attack of H2S and H2O over the optimized clean a-SnO2 surface. The left-hand side refers to the optimized initial clean a-SnO2 system, and the right-hand side refers to the adsorption modes of H2S (a) and H2O (b). Yellow, S; mauve, Sn; red, oxygen; white, H atoms.

In the same fashion, as shown in the right-hand side of Figure b following a heterolytic rupture of OH bonds in H2O,28 water vapor chemisorbs over the a-SnO2 surface with the formation of a rooted hydroxyl, two-fold coordinated with surface Sn atoms indicated as (OH)2Sn, and one rooted hydroxyl group (OH)O. Corresponding energetics and Bader charges are shown in the last line of Table , referring to the most stable H2O attack (specifically C2 in ref (28)).

Gas Sensing Mechanism

At a 100 °C operating temperature, the adsorption of H2S gas and H2O over a-SnO2 is experimentally transduced by a decrease in the sensor’s resistance, as displayed in Figure a,b. In our case, we consider a clean a-SnO2 amorphous surface where both H2S and H2O compete, as depicted in Figure a,b, at the same time over the same a-SnO2 adsorption sites, according to the following reactions:

Compliant to theoretical computations, the positive (+) overall Bader charge balance of H2S adsorption (see structure 2, Table ), derived from the balance of nδ+, mδ–, and hydroxyl–hydrogen atomic charge distribution as schematized by reaction , provides an electron enrichment of the a-SnO2 surface, which is congruent to the decrease in the sensor’s resistance with H2S gas as shown in Figure a. Conversely, according to reaction , the negative (−) overall Bader charge balance of H2O adsorption (see structure C2, Table (28)) confirms the electron depletion of the surface, which conflicts with the decrease in the sensor’s resistance recorded in Figure b. This apparent contradiction can be explained accounting for the lower electron affinity of the rooted hydroxyls (OH)0,which are easily ionized according to reaction , providing extra electrons (e′) that outweigh the negative charge depletion operated by water adsorption.

Fundamental investigations on the temperature interaction of water vapor over crystalline SnO2 utilizing operando DRIFT spectroscopy[76] confirmed, opposite to our dissociative adsorption mechanism, that neither physisorption nor dissociative water adsorption occurs over the surface of crystalline SnO2 in the temperature range of 100–150 °C. On this account, theoretical vs experimental conditions should always be considered. Our theoretical approach refers to a clean a-SnO2 amorphous interface, while DRIFT experiments specifically apply to sol–gel-prepared crystalline SnO2. Overall, it may be concluded that regardless of (i) the preparation conditions, (ii) the crystalline or amorphous nature of the interface, and (iii) the 2D or 3D geometry of the platform, in all cases, humidity substantially affects the resistance of the device. Specifically, it seems that the 2D layered nature of TMDs and MDs does not gain substantial advantages to improve humidity cross-interference compared to traditional metal oxide sensors. However, theoretical and practical investigations of the water vapor adsorption mechanism over 2D layered materials are still young, while effective practical strategies to promote selectivity have not been implemented yet.

It is our opinion that theoretical and experimental investigations at lower operating temperatures (OTs) are needed, considering that both metal oxide and 2D TMD/MD layered gas sensors are increasingly operated between room temperature and a 100 °C OT. On this account, the amorphous seamless texture of the a-SnO2 sensor represents an ideal platform for operando DRIFT spectroscopy measurements considering the absence of crystalline planes, an opportunity that rules out any influence of the preparation conditions on the adsorption mechanism of different molecules.[77]

Figure finally illustrates a possible effect of gas/water adsorption over layered a-SnO2 and the conduction mechanism of nanosheet networks. Departing from few-layer gray-colored a-SnO2 flakes, representing the situation in dry air (Figure a), as soon as H2S gas or H2O adsorbs over a-SnO2, the flake’s surface charge carrier concentration increases, corresponding to the yellow-colored regions of Figure b. Depending on the thickness of the stacked flakes, the extension of the injected regions may be limited to surface layers, leaving the inner flakes unaffected (inner gray region of Figure b), or eventually extended to the core (fully injected flakes, not shown here). Remarkably, by controlling the liquid-phase exfoliation procedure,[78] the thickness of the flakes can be easily tailored to cover all the possible conduction regimes, from few-nanometer fully injected thin layers to partially injected thicker ones, therefore effectively modulating the gas response. Lastly, we have reported in Figure c the schematization of a spin-coated thin film, comprising a disordered network of a-SnO2 nanosheets, forming localized intersheet junctions (highlighted in red in Figure c), and enabling charge transfer across the layers. In this case, the electrical conduction model is represented as an arrangement of in-series pairs of resistances where RS and RJ represent the sheet and junction resistances, respectively, as recently brilliantly discussed.[79] Considering that amorphous oxide semiconductors are generally characterized by a high electron mobility (≈10 cm2/(V s))[80] exceptionally topping the field-effect mobility of ≈100 cm2/(V s) for amorphous SnO2,[81] a junction-limited conduction mechanism where RJ ≫ RS prevails. This model addresses the formation of Schottky barriers between the flakes, modulated by the nature and composition of the adsorbing gas in the same way as the conduction mechanism of loosely sintered metal oxide nanoparticles in traditional chemoresistive sensors.

Figure 9

Schematization of a possible gas/water vapor adsorption mechanism over few flakes a-SnO2 and the conduction mechanism of nanosheet networks. (a) Few-layer a-SnO2 in dry air; (b) few-layer a-SnO2 after exposure to H2S or H2O (yellow regions represent the charge injected zones and the gray inner regions the not injected ones); (c) network morphology of spin-coated a-SnO2 flakes forming localized intersheet junctions (colored in red) with the schematization of the current transfer equivalent circuit between the sheets (RNS = sheet resistance and RJ = junction resistance).

Conclusions

We have reported an innovative and simple procedure to synthesize “LAMOS”, specifically layered amorphous a-SnO2 metal oxide sensors, by annealing liquid-phase exfoliated 2D SnSe2 in air for two weeks at 250 °C. The oxidation process of 2D SnSe2 here validated, carried out at temperatures below the crystallization temperature of SnO2 (280 °C), enables the spontaneous substitution of sulfur with oxygen atoms in 2D SnSe2. Such layered amorphous a-SnO2 flakes, which are stable up to 250 °C, preserve all the geometrical features of their 2D precursor counterparts. Thin-film sensors of amorphous a-SnO2 flakes, fabricated by spin-coating over patterned electrodes, are sensitive to H2S and humidity at a 100 °C operating temperature, with excellent baseline resistance recovery and sensor’s signal reproducibility over one-year deployment. We also found that the electrical response to H2S and humidity of a-SnO2 is like that of crystalline SnO2 microporous metal oxides, with associated humidity cross-effects on H2S sensing and a reduced sensor’s signal amplitude with increasing the humidity content. We also demonstrated the hindering effect of water vapor upon H2S sensing by a combined DFT+AIMD computational approach, highlighting that both H2O and H2S compete at the same time, over the same a-SnO2 adsorption site, according to a dissociative chemisorption mechanism. We additionally indicated a possible conduction mechanism of the a-SnO2 thin-film device, theorizing the formation of Schottky barriers between the flakes, modulated by the nature and composition of the adsorbing gas, in the same way as the conduction mechanism of loosely sintered metal oxide nanoparticles in traditional chemoresistive sensors.

In conclusion, we have validated an effective strategy to offset typical drift electrical signal phenomena in 2D TMD/MD sensors induced by spontaneous degradation in dry/wet ambient conditions of the sensor’s surface. On this account, we validated a “core” oxidation/amorphization synthesis of pristine 2D SnSe2 chalcogenide flakes to yield a-SnO2 gas sensors. Remarkably, this methodology can be extended to a large variety of TMDs and MCs, opening new opportunities for “LAMOS” interfaces with unexplored surface-science capabilities, probably well beyond gas sensing applications.