Charge Redistribution Mechanisms in SnSe2 Surfaces Exposed to Oxidative and Humid Environments and Their Related Influence on Chemical Sensing.

Tin diselenide (SnSe2) is a van der Waals semiconductor, which spontaneously forms a subnanometric SnO2 skin once exposed to air. Here, by means of surface-science spectroscopies and density functional theory, we have investigated the charge redistribution at the SnO2-SnSe2 heterojunction in both oxidative and humid environments. Explicitly, we find that the work function of the pristine SnSe2 surface increases by 0.23 and 0.40 eV upon exposure to O2 and air, respectively, with a charge transfer reaching 0.56 e-/SnO2 between the underlying SnSe2 and the SnO2 skin. Remarkably, both pristine SnSe2 and defective SnSe2 display chemical inertness toward water, in contrast to other metal chalcogenides. Conversely, the SnO2-SnSe2 interface formed upon surface oxidation is highly reactive toward water, with subsequent implications for SnSe2-based devices working in ambient humidity, including chemical sensors. Our findings also imply that recent reports on humidity sensing with SnSe2 should be reinterpreted, considering the pivotal role of the oxide skin in the interaction with water molecules.

After the advent of graphene,[1−3] van der Waals semiconductors are attracting considerable attention, owing to their application capabilities that are often complementary to those of graphene,[4−6] with the subsequent prospect of novel disruptive technologies in different technological areas.[4,7,8] This class of materials is characterized by weak van der Waals bonds between layers enabling their cleavage by mechanical[9] and liquid-phase[10] exfoliation. Among van der Waals semiconductors, several materials show serious drawbacks, limiting their technological potential. Specifically, MoS2 and WS2 display intrinsic electron mobility as low as some tens of cm2 V–1 s–1 at 300 K;[11] black phosphorus rapidly degrades in air due to surface oxidation;[12,13] GaSe is affected by both environmental and laser-induced degradation;[14,15] and PdSe2[16] has a limited commercial potential, due to the constantly growing price of Pd ($2000–2400/oz), which nearly doubled from 2019 to 2020.

Tin diselenide (SnSe2) is a van der Waals semiconductor with a CdI2-type crystal structure,[17] belonging to the P3̅m1 space group, with tin (Sn) atoms interweaved between two hexagonally packed atomic layers of selenium (Se) (see the atomic structure in Figure S1a,b).[18,19] SnSe2 shows its high intrinsic electron mobility (462.6 cm2 V–1 s–1 at 300 K[20]) and ultralow thermal conductivity (3.82 W m–1 K–1[20]). It displays pressure-induced periodic lattice distortion, and moreover, it enables novel device functionalities being a phase change memory material; i.e., its atomic structure can reversibly switch from amorphous to crystalline upon laser heating, with consequent remarkable variations in optical reflectivity. Because of these peculiarities, SnSe2 has high application capabilities in numerous fields, including photocatalysis,[21,22] superconductivity,[23,24] Li-ion[18,25,26] and Na-ion[18,26,27] batteries, photodetection,[28] saturable absorbers for eye-safe lasers,[29] and thermoelectricity.[30−32] Furthermore, SnSe2 was used as a co-catalyst for hydrogen evolution reaction.[33]

However, all Sn-based chalcogenides are usually affected by rapid surface degradation with the emergence of tin oxide phases.[34,35] Additionally, the oxidation of starting element Sn during the synthesis can also influence the transport properties of the resulting crystal. Therefore, technological exploitation of Sn-based chalcogenides remains particularly challenging. Especially, the stability of SnSe2-based devices in the ambient atmosphere is related to the chemical reactivity of its surface.

Recently, it has been shown that, though stoichiometric SnSe2 shows outstanding chemical stability under ambient conditions, the presence of Se vacancies drastically affects surface chemical reactivity.[36] The SnSe2– surface is transformed into SnO2 skin-terminated SnSe2, with the thickness of the SnO2 skin estimated to be subnanometric.[36] Unexpectedly, the self-assembled heterostructure formed by exploiting the natural interaction with air is particularly appropriate for ultrasensitive gas sensing, as demonstrated for NO2 and H2 with sensitivities of (1.06 ± 0.03) and (0.43 ± 0.02) ppm–1.[36] Remarkably, such sensors are effective under dry air conditions, while previously devised SnSe2 sensors used N2 as the carrier gas.[37,38] Moreover, the NO2 sensitivity of the SnO2–SnSe2 heterostructure is significantly higher compared to those of sensors based on other van der Waals semiconductors and their heterostructures.[39,40]

Remarkably, the oxide skin plays a pivotal role in NO2 and H2 sensing, congruently with the abundant literature on SnO2-based sensors.[41−50] The modulation of resistivity upon gas adsorption is strictly connected to charge distribution in the sensing material, ultimately related to the formation of surface dipoles at the SnO2–SnSe2 heterojunction arising from local charge redistribution. Thus, to understand the conduction mechanism ruling chemical sensing, it is crucial to shed light on charge redistribution at the SnO2–SnSe2 heterostructure by measuring work-function changes. Furthermore, sensing experiments in ref (36) were carried out in dry air; thus, stability in a humid environment remains unexplored, although real conditions mandatorily require sensors to work in a changing humidity background[51,52] (not only humidity sensors[53,54]). Despite the relevance of the influence of the humid environment for practical applications, surprisingly it has been scarcely investigated, although previous reports indicated a decrease in resistance under exposure to a humid atmosphere,[51,52] which represents an unambiguous fingerprint that H2O behaves as a reducing gas in the interaction with the SnO2 surface.

In addition, the interaction with water is relevant also for understanding the stability of any other SnSe2-based (opto)electronic device[55] working in ambient humidity, as well as the eventual environmental doping effects in transport properties.[56] Actually, recently different groups have reported that SnSe2 is extremely sensitive to humid environments,[37,38,57] with the possibility of using it in humidity-sensing devices.

Here, we unveil the surface properties of SnSe2 single crystals and their modifications in oxidative and humid environments by means of surface-science experiments and density functional theory (DFT). Definitely, surface oxidation induces an increase in the work function of 0.4 eV, owing to the charge transfer between the substrate and the SnO2 skin of 0.56 e– per SnO2 unit. As opposed to previous reports,[37,38,57] the pristine SnSe2 surface is inert to water at room temperature, while the SnO2–SnSe2 heterostructure displays notable sensitivity to humidity.

The presence of the SnO2 skin in the SnSe2 surface exposed to oxidative environments was ensured by both microscopic evidence from low-energy electron microscopy (LEEM) (Figure S5) and vibrational experiments from high-resolution electron energy loss spectroscopy (HREELS) (Figure S6).

The analysis of the variation of work function ΔΦ probed by LEEM could provide important insights into charge redistribution arising from surface oxidation (Figure a), as the total reflectivity threshold in electron backscattering (the MEM–LEEM transition, where MEM stands for mirror electron microscopy) represents a direct measurement of the variation of the surface potential.[58] Explicitly, we find ΔΦ to be 0.23 eV for the SnSe2 surface modified by exposure to 700 L of O2 at room temperature, while air exposure for 15 min induces a further shift in the work function, resulting in a total increase of 0.40 eV. The observed value of ΔΦ can be explained by considering the activation of surface dipoles, due to charge transfer at the interface from substrate to adsorbed oxygen atoms. The electronegativity of oxygen makes its adsorption generally associated with a charge transfer from the substrate to the adsorbate layer, with a subsequent increase in the work function.[59] Considering that the work function of the pristine SnSe2 single crystal is ∼4.6 eV,[60] while that of SnO2 is known to be ∼4.9 eV[61] (although its value can be tuned by reduction reactions[62]), both the sign and the magnitude of the experimental value of ΔΦ are consistent with surface oxidation, involving the formation of a subnanometric SnO2 skin. We can infer that previous experimental studies reporting a work function of SnSe2 of (5.0 ± 0.1) eV[63,64] could be affected by surface oxidation, which generates a self-assembled SnO2–SnSe2 heterostructure with an increased work function. To verify this statement, we calculated ΔΦ for the oxidation of the pristine SnSe2 surface, finding a value of 0.52 eV in qualitative agreement with experimental measurements. We also note that, in the air-exposed sample, variations in the I–V curve associated with electron diffraction from a surface with crystalline order[65] are suppressed, due to the formation of a disordered surface oxide phase.

Figure 1

(a) LEEM I–V curves at the MEM–LEEM transition for the as-cleaved sample (black), after a dose of 700 L of O2 (blue), and after air exposure for 15 min (pink). The shift of the MEM–LEEM transition, characterized by the sharp decrease in intensity, indicates an oxidation-induced modification of the surface potential. (b) Changes in charge density after the formation of the interface between the SnSe2 substrate and SnO2 skin. Sn, Se, and O atoms are represented as dark blue, light green, and red balls, respectively.

Complementary information about the electronic properties of the SnO2–SnSe2 heterostructure was achieved by comparing the surface excitation spectrum probed by electron energy loss spectroscopy (EELS) with the theoretical density of states (DOS) (section S4 of the Supporting Information). The impact of defects on the DOS is assessed in section S5.

To estimate the amount of charge transfer between the SnSe2 substrate and the SnO2 skin, we calculated the charge density distribution of (i) one SnO2 layer over two layers of SnSe2 (to model the SnO2–SnSe2 heterostructure), (ii) a free-standing SnO2 single unit, and (ii) a bilayer of SnSe2. Then, we calculated the difference between the charge densities of the whole SnO2–SnSe2 interface and those one of its components (single SnO2 unit and bilayer SnSe2). The obtained charge density difference (Figure b) illustrates charge redistribution following the formation of the SnO2–SnSe2 interface. The integration of the charge density difference along the c axis provides information regarding the charge transfer between the SnSe2 substrate and the SnO2 skin. Note that the formation of the SnO2–SnSe2 interface provides changes in the charge density difference in not only the outermost SnSe2 layer but also the subsurface area, namely the second SnSe2 layer. Definitely, the charge transfer is estimated to be 0.56 e– per SnO2 unit.

While the adsorption of O2 with further decomposition is energetically favorable on SnSe2 (negative values of ΔG and ΔHdec), as well as on SnSe1.88 and SnSe (Table ), our theoretical model indicates that water does not adsorb on SnSe2. The energy cost for water adsorption is decreased in the presence of Se vacancies (SnSe1.88) down to ∼3 kJ/mol, although water adsorption (as well as decomposition) remains energetically unfavorable. Similarly, SnSe also shows outstanding chemical inertness toward water.

Table 1

Differential Enthalpies (ΔHads), Differential Gibbs Free Energies of Physisorption (ΔG), and Differential Enthalpies of Decomposition (ΔHdec) for Molecular Oxygen and Water on Pristine SnSe2, SnSe1.88, and SnSe Surfacesa

		physisorption		decomposition
surface	adsorbant	ΔHads (kJ/mol)	ΔG (kJ/mol)	ΔHdec (kJ/mol)
SnSe2	O2	–17.46	–3.16	–42.28 (−161.58/∼−40.2)
	H2O	–13.27	18.03	220.91
SnSe1.88	O2	–37.58	–26.28	–135.67 (−99.05/–406.65)
	H2O	–27.93	3.37	175.61
SnSe	O2	–11.53	–0.23	–236.03 (−323.10/95.4)
	H2O	–8.12	23.18	82.22
SnO2 skin	H2O	–119.70	–106.67	–121.31

For oxygen decomposition, the table also displays the differential enthalpy of the oxidation of the whole surface with formation of SnO and SnO2-like layers (in parentheses).

Considering that the yield of chemical reactions also depends on the probability of the interactions between reactants, we calculated Langmuir adsorption isotherms (Figure S12). Specifically, the combination of thermodynamic and kinetic calculations evidences that the largest part of the SnSe surface will be oxidized under experimental conditions (72% and 75% for SnSe2 and SnSe1.88, respectively).

On the contrary, the saturation coverage for water at room temperature is just 0.01 ML (with ML being monolayer) for SnSe2 and SnSe1.88, while the full coverage (1 ML) is reached upon exposing the SnO2 skin to only 5 × 10–3 L of H2O below 500 °C, thus evidencing the aptness of the SnO2–SnSe2 interface for ultrasensitive humidity sensing. The increase in temperature corresponds to a decrease in the sticking coefficient, with monolayer saturation reached at 0.05 and 10 L at 500 and 800 °C, respectively. Thus, the SnO2–SnSe2 heterostructure remains rather sensitive even at high operational temperatures.

Therefore, the SnO2–SnSe2 heterostructure shows superior chemical reactivity toward ambient species with respect to SnSe2. On the pristine SnSe2 surface, the local rearrangement of chemical bonds around each adsorbed water molecule is the origin of a redistribution of the charge density in the surface layer of SnSe2 with a charge transfer of 0.17 e– per water molecule (Figure a). Correspondingly, water adsorption on SnSe2 and SnSe1.88 surfaces is energetically unfavorable for temperatures above 124 and 264 K, respectively (Figure S11). Hence, we conclude that pristine SnSe2 is stable in a humid environment and, consequently, is unsuitable for humidity sensing, contrarily to conclusions in refs (37), (38), and (57) On the contrary, adsorption of H2O on the SnO2–SnSe2 heterostructure (Figure b) is energetically favorable even above room temperature (Figure S11). The values of charge transferred from H2O to the SnO2 skin are 0.43 and 0.30 e– for one and two H2O molecules per supercell, respectively. Correspondingly, DOS (Figure c) is modified with a direct correlation with the coverage of the adsorbate, hence proving the appropriateness for humidity sensing also at low concentrations of H2O.

Figure 2

Changes in charge density after adsorption of one water molecule on (a) SnSe2 and (b) SnO2 skin-terminated SnSe2. Panel c represents the DOS of SnO2 skin-terminated SnSe2 (black) and of the same system modified by the adsorption of one (red) and two (blue) water molecules. The Fermi level is set at 0. Panel d shows the response of the SnSe2–SnO2 heterostructure to 20% relative humidity (RH) at an operational temperature (OT) of 150 °C (note that the average residence time of the gas in the cell is approximately 10 min).

Note that decomposition of a water molecule on the SnO2 skin-terminated SnSe2 is an exothermic process (see Table ), although the energy gain from this process is moderate (−121.31 kJ/mol) and further water splitting is unfavorable, supporting the possible reversibility of the process.

The SnO2–SnSe2 heterostructure was tested as a humidity sensor (Figure d) at an operational temperature of 150 °C. Our devised humidity sensor exhibited (i) full recovery of the baseline resistance after water desorption and (ii) high sensitivity to water molecules, measured as the relative response (RR, the ratio between the resistance in dry air, Ra, and RH, the resistance in a humid environment), and an experimental limit of detection (LOD) in terms of relative humidity (RH) as low as 20% (Figure d).

Recently, different authors[37,38,57] have reported the outstanding performances of SnSe2 in humidity-sensing devices. Our findings elucidate the key role of the surface oxide skin in the interaction with a humid environment. On the contrary, in refs (37), (38), and (57), surface oxidation was not assessed; thus, the mechanism ruling humidity sensing discussed therein should be reinterpreted.

Theoretical results were validated by surface-science techniques. In particular, HREELS experiments on water-dosed Sn-based selenides (SnSe, SnSe1.4, SnSe1.7, and SnSe2) indicate the absence of chemisorbed water-derived species, as indicated by the lack of O–H stretching at 408–425 meV (molecular water) and 445–460 meV (hydroxyl groups) in spectra in Figure (see ref (66) for more details). These findings are consistent with the positive differential Gibbs free energy of adsorption (corresponding to energetically unfavorable water adsorption) in Table . For the sake of comparison, we report in Figure also vibrational data obtained after exposure to the same dose of H2O (105 L, with 1 L = 1 × 10–6 Torr s) at room temperature the surface of other metal chalcogenides, which instead enable the stable adsorption of water molecules (PtTe1.6) and hydroxyl groups (InSe).

Figure 3

HREELS spectra in the region of the O–H band acquired after exposure to 105 L of H2O at room temperature the surfaces of different Sn-based chalcogenides: SnSe2 (orange curve), SnSe1.7 (black), SnSe1.4 (green), and SnSe (blue). To provide a straightforward comparison, the figure also displays data for H2O-dosed InSe (red) and PtTe1.6 (brown) surfaces (105 L at room temperature). The impinging energy is 4 eV.

The absence of reactivity toward water of Sn-based chalcogenides makes them suitable for catalysis (especially, photocatalytic water splitting[22] and hydrogen evolution reaction[67]) and drug delivery[68] (also considering that neither Sn nor Se is toxic). Congruently, SnSe2 was used as a co-catalyst in combination with TiO2 for hydrogen evolution reaction.[33]

Further information about the surface chemical bonds is gained by the inspection of core levels via X-ray photoelectron spectroscopy (XPS) experiments. Figure shows the Sn-3d and Se-3d core levels of the SnSe2 single-crystal surface cleaved in ultrahigh vacuum and for the same surface modified by O2 and H2O dosage with a total dose of 105 L. The Sn-3d5/2 core level in the as-cleaved surface displayed a binding energy (BE) of 486.8 eV (Figure b). Congruently, the Se-3d5/2 core level had a single component at a BE of 54.1 eV, in agreement with previous results for SnSe2[69] and with a shift of +0.4 eV compared to the case of SnSe. Surface treatments, i.e., 105 L of O2 and H2O exposure, induce only slight changes in Se-3d core levels. A novel doublet appeared in Se-3d (BE = 53.7 eV for 3d5/2), whose total spectral area is 5.4% (for O2 dosage) and 2.6% (for air exposure), arising from Se(0) segregation.[70] In particular, from the analysis of Se-3d core-level spectra (Figure c), we can infer the absence of O–Se–O bonds, which would have a BE of ∼59–60 eV.[71−73] Congruently, the intensity of the O-1s peak is especially small in SnSe2 exposed to both an oxidative and humid environment (Figure a); thus, we can evaluate the amount of oxygen to be <0.04 ML, due to a particularly weak sticking coefficient for oxygen adsorption at 300 K on SnSe2, with the O2 sticking coefficient being <10–5.

Figure 4

(a) O-1s, (b) Sn-3d, and (c) Se-3d core levels for the pristine surface of SnSe2 cleaved in situ under ultra-high-vacuum conditions and its alteration after exposure to oxidative (105 L of O2) and humid (105 L of H2O) environments at room temperature. The photon energy is 800 eV. We also report in each panel the corresponding spectrum for SnO2–SnSe2– exposed to a humid environment at room temperature, with x estimated to be 0.29.

On the contrary, we observed quite distinct peaks in SnO2–SnSe2– exposed to a humid environment (outermost spectra in the various panels of Figure ), and a Sn-3d doublet with a J = 5/2 component is present at a BE of 487.8 eV, due to SnO2 (relative amplitude of 54%), which is consistent with previous reports for this system.[74,75] Remarkably, no trace of O–Se–O bonds is present, as suggested by the lack of Se-3d components at 59–60 eV.[72] This result confirms theoretical expectations that Se is involved in only a metastable oxide phase, which represents a precursor for SnO2 formation. Nevertheless, a broad spectral component in Se-3d suggests a different oxidation state for Se. In particular, the peak at 55.0 eV is ascribed to Se2–, while that one at a higher BE should be attributed to Se–2+α (0 < α < 1).[76] The analysis of survey XPS spectra enables us to evaluate α as ∼0.145, corresponding to substoichiometric SnSe1.71. Congruently with the results in ref (36), oxidation is feasible only in substoichiometric SnSe2–, while perfectly stoichiometric SnSe2 is robust in oxidative environments, thus evidencing the pivotal role of Se vacancies in surface oxidation.

The O-1s spectrum for the SnSe2 surface exposed to a humid environment shows new components arising from −OH groups (relative amplitude of 45%) and H2O (relative amplitude of 9%) at BEs of 531.6 and 533.6 eV, respectively.[77−79] We also exposed the SnO2–SnSe2 heterostructure to the humid environment, with the corresponding O-1s spectrum displaying the SnO2 component (relative amplitude of 50%) at a BE of 531.5 eV,[80,81] overlapped with the −OH component.

In conclusion, we investigated (i) the modifications of surface properties once pristine SnSe2 assumes a subnanometric SnO2 skin upon interaction with oxidative environments and (ii) the subsequent implications for chemical sensors. Definitely, the oxidation process has a direct effect on the work function, which is increased by 0.4 eV, owing to the charge transfer between the substrate and the SnO2 skin of 0.56 e– per SnO2 unit. Though the SnSe2 surface is inert to water at room temperature, upon surface oxidation the SnO2–SnSe2 interface shows a remarkable sensitivity to humidity. The charge transfer from H2O to the SnO2 skin is estimated to be 0.43 and 0.30 e– for one and two H2O molecules per supercell, respectively. Correspondingly, the DOS is correlated with water coverage, hence proving the aptness for humidity sensing also at low concentrations of H2O. Definitely, our findings prove the significant influence of humid environments on the electrical response of the SnO2–SnSe2 heterostructure. Moreover, recent reports regarding the use of SnSe2 in humidity sensors should be reconsidered with regard to the physicochemical mechanism.

Methods

Theoretical methods are described in section S8.

The single crystals were grown by the Bridgman–Stockbarger method, according to the procedure described in refs (36) and (82) (see also section S1). Their crystalline quality was secured by X-ray diffraction (XRD) (Figures S1c and S2). The analysis of the XPS survey spectrum proves the absence of contaminants in bulk crystals (Figure S3). Samples were exfoliated in situ for surface-science investigations, by using scotch tape. Gas dosage was carried out at a partial pressure of 10–4 mbar.

XPS experiments were carried out at the APE-HE beamline at the Elettra-Trieste synchrotron. Core-level measurements were performed with an Omicron EA125 hemispherical electron energy analyzer, with the sample at room temperature and in normal emission. Linearly polarized light formed an angle of 45° with respect to the perpendicular direction of the surface. After the subtraction of a Shirley background, Sn-3d core-level spectra were analyzed by using a Gaussian line shape convoluted with a Doniach–Sunjic function,[83] while Se-3d and O-1s were fitted by Voigt line shapes.

HREELS experiments were performed with an Ibach-type spectrometer. The primary electron beam energy was 3.5 eV. HREELS spectra were recorded under specular conditions.

Measurements of LEEM images (Figure S4), EELS (Figure S6), and work-function changes ΔΦ (Figure a) were carried out at the soft X-ray beamline Nanospectroscopy at Elettra-Trieste synchrotron, using an energy-filtered LEEM–PEEM microscope with a spatial resolution of 10 nm. Specifically, measurements of ΔΦ were carried out by varying the electron beam energy across the total electron reflectivity threshold. This threshold is commonly termed the MEM–LEEM transition, which is characterized by a steep decrease in intensity as a function of a bias voltage applied to the sample (start voltage) as a decelerating potential. The ΔΦ value is identified by the shifts in the bias potential corresponding to the MEM–LEEM transition.

The gas sensing response to humidity at an operating temperature of 150 °C was determined by a volt–amperometric technique, as reported in ref (36). The RH air stream at 20% RH was obtained by mixing dry with saturated water-vapor air. In the analysis of the gas response, the relative response (RR) is defined as the ratio between the measured electrical resistance in dry air (Ra) and that under 20% RH (RH).