Photostability of Monolayer Transition-Metal Dichalcogenides in Ambient Air and Acidic/Basic Aqueous Solutions

We report on the photostability of monolayer (1L) transition-metal dichalcogenides (TMDCs) in air and in aqueous solutions, as probed using photoluminescence spectroscopy. 1L-WSe2 was readily degraded under continuous irradiation of visible light in aqueous solutions, whereas 1L-MoS2 was relatively stable in both ambient air and aqueous solutions. The stability difference between these two materials was mainly ascribed to the oxidization reaction at the interface of 1L-TMDCs and the O2/H2O redox system induced by both band alignment and photogenerated holes. This interpretation was strongly supported by the observation of the lower degradation rate of 1L-WSe2 in the dark and in degassed water with a lower concentration of dissolved oxygen compared with the degradation rate of 1L-WSe2 in distilled water. Furthermore, the degradation rate was also nearly proportional to the number of photogenerated carriers. The degradation rate under acidic conditions was smaller than that under the basic conditions. The results are attributed to the oxidation/reduction potential of 1L-WSe2 and to the dissolution reaction of degraded species, both of which are strongly pH-dependent.


■ INTRODUCTION
Monolayers (1L) of transition-metal dichalcogenides (TMDCs) MX 2 , where M is a transition metal (typically Mo or W) and X is a chalcogen (typically S, Se, or Te), have emerged as new direct-gap semiconductors with truly twodimensional structures. Because of the striking change in their electronic structures from indirect (>2L) to direct band gap (1L) depending on their number of layers, 1−3 1L-TMDCs show interesting optical and electrical properties which have strong potential for use in future electronic and optoelectronic devices such as phototransistors, 4−9 photodetectors, 10−12 lightemitting devices, 13,14 light modulators, 15 and solar cells. 16−18 Tunable photoluminescence (PL) has thus far been demonstrated using chemisorption and physisorption, 19−22 covalent bonding, 23,24 and electrostatic methods. 25,26 The electronic properties have also been studied using chemical doping and defect engineering. 27−29 Additionally, the applications of 1L-TMDCs as gas sensors or biosensors and photocatalysts for water splitting have been reported. 30−35 Electrochemical etching or oxidation reactions of atomically thin TMDCs under photoirradiation 36,37 or ozone exposure 38 conditions were also studied. These previous studies suggest that deep understanding of the influence of surface reactions on the physical properties and stability of 1L-TMDCs is important for realizing their application in various environments such as ambient air or wet conditions. However, knowledge of the stability of 1L-TMDCs under such conditions is limited.
In this paper, we report on the stability of the most typical 1L-TMDCs1L-WSe 2 and 1L-MoS 2 in ambient air and acidic/basic aqueous solutions under visible-light irradiation. The PL intensities of 1L-WSe 2 distinctly decreased in acidic/ basic aqueous solutions, whereas MoS 2 was stable under the same conditions and its PL intensity showed a negligible decrease. The PL spectral shape did not show a remarkable change with increasing irradiation time for either 1L-WSe 2 or 1L-MoS 2 , suggesting that the change in the PL intensity was not caused by photoinduced carrier doping but was mainly attributable to photoinduced degradation. We confirmed that the degradation rate was lower in the dark and in degassed water than in distilled water, suggesting that both light and dissolved oxygen are necessary for photodegradation reactions. Furthermore, 1L-WSe 2 under acidic and neutral conditions showed a more moderate degradation rate than that under the basic condition. These results suggest that the electrochemical reaction with 1L-TMDCs and photogenerated holes in the presence of oxygen and water is the key mechanism of the degradation. This mechanism is effective for 1L-WSe 2 and ineffective for MoS 2 because of the difference in the relative alignment among the oxidation/reduction potentials of O 2 / H 2 O and the levels of the valence and conduction bands of each 1L-TMDC.
■ RESULTS AND DISCUSSION Figure 1 shows the schematic of the optical system for the PL measurements in air or aqueous solutions. 1L-TMDCs prepared on a transparent glass substrate ( Figure S1) were irradiated with visible wavelength light (580 nm) with a constant power density (800 W cm −2 ). The signal was recorded using a silicon charge-coupled device (CCD) detector equipped with a spectrometer (see the Experimental Methods section for details).  Figure 2a, the PL intensity decreased slightly within 90 min under photoirradiation (580 nm, 800 W cm −2 , which is in the linear region shown in Figure  S2). By contrast, 1L-MoS 2 showed a negligible change in PL intensity ( Figure 2c). The normalized PL spectra show an undetectable shape change in the spectra of both 1L-WSe 2 and 1L-MoS 2 (Figure 2b,d). Because the optical absorption coefficients of 1L-TMDCs at 580 nm are similar (1 × 10 5 to 2 × 10 5 cm −1 ), 39,40 we expected similar numbers of electrons and holes to be photogenerated in these samples. Thus, these results imply that the stability of 1L-WSe 2 is slightly lower than that of MoS 2 in ambient air under photoirradiation.
Because the properties of 1L-TMDCs are known to be strongly modified by electrochemical effects at the surfaces, 22,41 we next studied their stabilities in water under light irradiation. Under these conditions, water molecules are adsorbed onto the surface and electrochemical effects because the oxygen/water redox couple becomes important. 41 We first performed PL measurements with the 1L-TMDCs in distilled water and then with the 1L-TMDCs in acidic/basic solutions. As shown in Figure 3a,c, the PL intensity of 1L-WSe 2 ( Figure  3a) in aqueous solution rapidly decreased with increasing irradiation time, whereas the PL intensity of 1L-MoS 2 ( Figure  3c) was stable under all of the investigated conditions. The main decrease of the 1L-WSe 2 PL intensity under neutral (distilled water) and basic conditions occurred within the first 10−15 min; the decrease then slowed, likely because of the increased concentration of the degradation products from 1L-WSe 2 . Under the acidic condition, as shown in Figure 3a, 1L-WSe 2 exhibited a relatively moderate degradation rate, in contrast to the fast degradation observed under other conditions. The original PL intensities decreased by approximately 15, 40, and 55% within 10 min in acidic (pH = 2), neutral (pH = 7), and basic (pH = 12) conditions, respectively. The normalized PL spectra of 1L-WSe 2 and 1L-MoS 2 show almost no change of the spectral shape within the irradiation time as shown in Figure 3b,d. We also confirmed the photoinduced degradation of 1L-WSe 2 in water directly by observing optical and PL images before and after the photoirradiation ( Figure S3).
To elucidate the mechanism of the material-dependent degradation behavior and the differences among degradation rates of the materials in different environments, we consider O 2 /H 2 O redox couple and the electrochemical reaction at the interface between aqueous solutions and 1L-TMDCs. The effects of O 2 and H 2 O molecules on the optical and electrical properties of the 1L-TMDCs were reported. 22,41 The transfer of photogenerated electrons or holes may be a main factor related to the degradation or decomposition of the 1L-TMDCs in an O 2 /H 2 O system, depending on the band alignment of the semiconductors with the oxidation/reduction potentials of O 2 / H 2 O. In general, photogenerated holes (h + ) can facilitate the oxidation of compound semiconductors (AB) 42 and destruct their original structures where z represents the number of holes and solv is solvent. In the current study, A and B are the transition metal and chalcogen that compose the semiconductors, respectively. The band alignment of 1L-WSe 2 relative to vacuum is shown in Figure 4a. The conduction-band minimum is near −3.6 eV, which is much higher than the H + /H 2 reduction potential (ca. −4.83 eV). 43−47 Thus, some of the photogenerated electrons can reduce H + to H 2 . This process can also occur for 1L-MoS 2 (Figure 4b), which has a conduction-band minimum at approximately −4.1 eV; 47,48 however, the smaller energy difference from ϕ(H + /H 2 ) results in a lower electron driving force/rate to reduce H + to H 2 in comparison with 1L-WSe 2 , as reported in previous studies. 49 For 1L-MoS 2 , the valence-band maximum is at approximately −6.0 eV which is lower than the potential of O 2 /H 2 O and some of the photogenerated holes may oxygenate H 2 O to O 2 . In this case, fewer excess holes remain and MoS 2 itself is not likely to be degraded within a short time. Thus, we can explain the difference in the degradation behaviors between 1L-MoS 2 and 1L-WSe 2 . Specifically, the change in pH results in a change of the redox potentials of the O 2 /H 2 O couple tested in this study (from −5.54 to −4.95 eV) between pH 2 and 12, and the oxidation and reduction potentials for 1L-WSe 2 are also shifted. 42 When H + (or OH − ) is directly involved in the oxidization/reduction reaction, as in eq 3, the relative potentials between O 2 /H 2 O and 1L-WSe 2 will change, 42 becoming one of the factors influencing the pHdependent degradation effect. Furthermore, WO 3 will be involved in a more complicated pH-dependent reaction in aqueous solution because of its thermodynamic instability. 54,55 To further understand the pH dependence, we also consider the second reaction of the oxidation product whose reaction rate also strongly depends on the solution pH. According to eqs 3 and 4, the pH dependence of the photodegradation rate can be understood as follows: under the basic condition, the reaction resulting in the dissolution of WO 3 should become faster than that under the acidic condition. Thus, the fast degradation under the basic condition is attributable to both the enhanced oxidation reaction and the dissolution reaction of WSe 2 and WO 3 , which generate more defects and edges 56−59 in the 1L-WSe 2 flake. This hypothesis is supported by the pH dependence of the decrease of the PL intensities ( Figure 3). For the PL intensity under the acidic condition, the time trace is well fitted by a single-exponential function. By contrast, the trace corresponding to the basic condition shows a much faster initial degradation that cannot be explained by a single reaction, implying involvement of an additional electrochemical process with a different time scale, as expressed in eq 4. WSe 2 degradation species differ among the acidic (WO 3 and SeO 4 2− ), neutral, and basic conditions (WO 4 2− and SeO 4 2− ). 60 This observation is consistent with the recently reported pH-dependent Gibbs free energy of WSe 2 . 60 For samples under acidic conditions, we can simply set the degradation rate equation as dA/dt = −αA, where A is the undegraded area of 1L-WSe 2 at time t and α is the rate constant. Solving this equation gives A = A 0 e −αt , where A 0 is the original area of 1L-WSe 2 . The undegraded area of 1L-WSe 2 decreases exponentially, and the decrease in PL intensity can be well reproduced using a single-exponential function, as shown in Figure 3.
To further confirm the photoinduced degradation effects, we measured the PL properties of the samples kept in the dark and in degassed water. The samples kept under dark conditions were not illuminated except during the measurements. As shown in Figure 5, the degradation reactions of 1L-WSe 2 in both dark (not degassed) (∼46%) and degassed (with light) conditions (∼38%) for 30 min are much slower than that of 1L-WSe 2 in distilled water (not degassed, with light) (∼82%), which supports our hypothesis that the degradation is mainly induced by the electrochemical effects of the photogenerated holes and dissolved oxygen in water. 61,62 The effect of photogenerated carrier density on the degradation rate was also confirmed. As shown in Figure 6,

ACS Omega
Article the degradation rates of 1L-WSe 2 under all conditions (acidic, neutral, and basic) are nearly proportional to the density of photogenerated electrons/holes, which strongly supports the hypothesis that degradation occurs mainly from photoinduced effects. The wavelength-dependent degradation rate, as shown in Figure S4, further supports this hypothesis. Moreover, under the acidic condition, the degradation rate was almost one order of magnitude smaller than that under the basic condition, which also supports the aforementioned hypothesis. By extrapolation, the degradation rate is between 0.1 × 10 −3 and 0.3 × 10 −3 s −1 when the photocarrier density is 0 ( Figure  6), consistent with the degradation rate in the dark (Figure 5), where a slow degradation with a rate of ∼0.3 × 10 −3 s −1 is observed.

■ CONCLUSIONS
In summary, we studied the stability of 1L-WSe 2 and 1L-MoS 2 in ambient air, distilled water, and acidic/basic aqueous solutions under light irradiation. In contrast to the good stability of both materials in ambient air, 1L-MoS 2 was found to be much more stable than 1L-WSe 2 under aqueous conditions. These results were consistently attributed to electrochemical reactions involving photoinduced electrons/ holes at the interface of 1L-WSe 2 and the O 2 /H 2 O redox system. A much lower degradation rate was observed for 1L-WSe 2 under acidic conditions than under basic conditions, which we attributed mainly to the pH dependence of the oxidation/reduction potential of WSe 2 and to the further reaction of the oxidation species. We also confirmed that the degradation rates of 1L-WSe 2 in both degassed water and dark conditions are much lower than those of 1L-WSe 2 in distilled water. The slower degradation in degassed water and dark conditions supports our hypothesis that photoinduced electrons/holes and dissolved oxygen are the main contributors to the degradation of 1L-WSe 2 . Our results may provide insights into the effects of photogenerated electrons/holes and electrochemical phenomena at the interface of atomically thin semiconductors and an O 2 /H 2 O redox system on their stability, which is important for their applications in realistic environments because these effects are unavoidable when a monolayer material with a large specific surface area is in a practical environment with finite humidity or in solution.

■ EXPERIMENTAL METHODS
1L-TMDC samples were prepared from a bulk single crystal (hq graphene) using a standard mechanical exfoliation method. The thickness of an exfoliated TMDC was first determined via image contrast using an optical microscope, and the Raman and PL spectra were then used to confirm the thickness. The exfoliated 1L-TMDCs were transferred onto a transparent glass substrate using polydimethylsiloxane, and the substrate with 1L-TMDCs was then tightly bound with a glass tube to compose a liquid reservoir in which sufficient liquid could be kept over the sample for the optical measurements in aqueous solutions. H 2 SO 4 and NaOH were used as the pH adjustors to attain acidic, neutral, or basic aqueous solutions. Degassed water was prepared using argon gas bubbling to reduce the dissolved oxygen concentration.
Optical measurements were performed through the transparent glass substrate using a homemade optical measurement system with a supercontinuum light source (20 ps pulse duration and 40 MHz frequency) equipped with a monochromator ( Figure 1). PL spectra were recorded with the samples in ambient air, in water, and finally in acidic and basic aqueous solutions using a liquid nitrogen-cooled silicon CCD detector. The measurements were performed at regular intervals (5 min) with a signal acquisition time of 5 s. During the measurements, samples were kept illuminated by light with a power density of 800 W cm −2 , which is in the linear region as shown in Figure S2, except for the experiment under dark conditions. The amount of water and aqueous solutions was kept the same, and all the measurements were performed at room temperature.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsomega.9b01067.
Fundamental Raman and PL spectra of 1L-WSe 2 and 1L-MoS 2 used in this study; excitation power dependence of PL intensity of 1L-WSe 2 and the wavelength dependence of the degradation rate in the distilled water condition; optical and PL images of 1L-WSe 2 before and after photoirradiation (PDF)

ACS Omega
Article ■ ACKNOWLEDGMENTS This research was supported by JSPS KAKENHI grant numbers JP16H00911, JP15K13337, JP15H05408, JP15K13500, JP16H00910, JP16H06331, JP17K19055, by the Asahi Glass Foundation, by JST CREST (JPMJCR16F3, JPMJCR18I5), by the Research Foundation for Opto-Science and Technology, and by the Nakatani Foundation.