The role of traps in the photocurrent generation mechanism in thin In-Se photodetectors

Due to the excellent electrical transport properties and optoelectronic performance, thin indium selenide (InSe) has recently attracted attention in the field of 2D semiconducting materials. However, the mechanism behind the photocurrent generation in thin InSe photodetectors remains elusive. Here, we present a set of experiments aimed at explaining the strong scattering in the photoresponsivity values reported in the literature for thin InSe photodetectors. By performing optoelectronic measurements on thin InSe-based photodetectors operated under different environmental conditions we find that the photoresponsivity, the response time and the photocurrent power dependency are strongly correlated in this material. This observation indicates that the photogating effect plays an imporant role for thin InSe flakes, and it is the dominant mechanism in the ultra-high photoresponsivity of pristine InSe devices. In addition, when exposing the pristine InSe photodetectors to the ambient environment we observe a fast and irreversible change in the photoresponse, with a decrease in the photoresponsivity accompanied by an increase of the operating speed. We attribute this photodetector performance change (upon atmospheric exposure) to the decrease in the density of the traps present in InSe, due to the passivation of selenium vacancies by atmospheric oxygen species. This passivation is accompanied by a downward shift of the InSe Fermi level and by a decrease of the Fermi level pinning, which leads to an increase of the Schottky barrier between Au and InSe. Our study reveals the important role of traps induced by defects in tailoring the properties of devices based on 2D materials and offers a controllable route to design and functionalize thin InSe photodetectors to realize devices with either ultrahigh photoresposivity or fast operation speed.

Abstract: Due to the excellent electrical transport properties and optoelectronic performance, thin indium selenide (InSe) has recently attracted attention in the field of 2D semiconducting materials. However, the mechanism behind the photocurrent generation in thin InSe photodetectors remains elusive. Here, we present a set of experiments aimed at explaining the strong scattering in the photoresponsivity values reported in the literature for thin InSe photodetectors. By performing optoelectronic measurements on thin InSe-based photodetectors operated under different environmental conditions we find that the photoresponsivity, the response time and the photocurrent power dependency are strongly correlated in this material. This observation indicates that the photogating effect plays an imporant role for thin InSe flakes, and it is the dominant mechanism in the ultra-high photoresponsivity of pristine InSe devices. In addition, when exposing the pristine InSe photodetectors to the ambient environment we observe a fast and irreversible change in the photoresponse, with a decrease in the photoresponsivity accompanied by an increase of the operating speed. We attribute this photodetector performance change (upon atmospheric exposure) to the decrease in the density of the traps

Introduction
The isolation of ultrathin two-dimensional (2D) semiconducting materials, such as single-layer transition metal dichalcogenides (TMDCs) and few-layer black phosphorous (bP), has attracted large attention due to their potential applications in next-generation electronic and optoelectronic devices. [1][2][3][4][5] The large surface-to-volume ratio of these 2D materials, which on one side makes them very tunable and sensitive to external stimuli, on the other side can make these materials and the devices based on them extremely vulnerable to environmental degradation. [6][7][8][9][10][11][12] For example, when few-layer bP is exposed to the air, a fast degradation of the material occurs through a photooxidation process that leads to a reduction of the performances and the eventual failure of devices based on bP. [6][7][8] Similar environmental degradation phenomenon also has been observed on thin gallium selenide (GaSe), 9 gallium telluride (GaTe), 13 and even CVD (chemical vapor deposition)-grown single-layer MoS2 and WS2, two members of the TMDC family. 14 Moreover, the mobility of 2D semiconducting transistors kept in vacuum or encapsulated with boron nitride is typically more than one order of magnitude larger than the mobility measured in the air. [15][16] In the case of single-layer and bilayer MoS2, among the various reports, the device mobility at room temperature ranges from 0.1 cm 2 V -1 s -1 in air to tens of cm 2 V -1 s -1 in vacuum or with top/bottom deposited protection materials. [17][18][19] All these observations can be explained by the presence of defects in the materials, such as chalcogen vacancies in the layered metal chalcogenides. [20][21][22] These defects can act as preferential sites for physical/chemical adsorption of environmental species (that can initiate the degradation process of 2D materials) and/or may introduce additional scattering of the carriers (that could act as harmful active traps in working devices).
Indium selenide (InSe), an n-type semiconductor which belongs to the III-VIA family, has recently attracted large attention because of its extraordinary charge transport properties, superior mechanical flexibility and strong lightmatter interaction. [23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41] Various groups reported on transistors based on thin InSe fabricated on different substrates (SiO2/Si, hexagonal Boron Nitride (h-BN), Poly(methyl methacrylate) (PMMA)) with mobility values as large as 3700 cm 2 V -1 s -1 at room temperature and ~13000 cm 2 V -1 s -1 at 4 K. 23,[25][26][27][28] The bandgap of 1.3 eV in bulk that becomes larger than 3 eV for an InSe single-layer makes this material interesting for broadband photodetection from the near-infrared to the near ultraviolet region of the electromagnetic spectrum. [30][31][32][33][34][35][36][37][38][39][40] Various photodetectors based on thin InSe flakes (as the active channel part), including metal-semiconductormetal (M-S-M) geometry and graphene based van der Waals heterostructures, have been reported in literature with responsivities going from 0.035 A W -1 to ultrahigh values of ~10 7 A W -1 and detectivities up to ~10 15 Jones 30-33, 35-38, 42 which are among the best performances reported for 2D photodetectors. 33,40 See Supporting Information, Table S1, for a comparison between the reported values in the literature for the figures of merit of InSe photodetectors. Such a large scattering in the responsivity values has also been observed for other chalcogenide-based 2D materials. For example in MoS2 photodetectors, traps states due to sulfur atomic vacancies influence the photocurrent generation giving rise to photogain and introducing Fermi level pinning at the metal-semiconductor (M-S) interface. [43][44][45][46][47][48][49] Similarly, in the case of InSe, the defects related to In adatoms and Se vacancies, which can be related to the In-rich atmosphere in which high-quality InSe crytals are grown, 50-51 may play important roles during the photocurrent generation process in thin InSe photodetectors. 20,[50][51][52] Interestingly, the presence of defects in InSe crystals, especially selenium vacancies (VSe), 22 is predicted to promote the physical adsorption and chemical dissociation of O2/H2O molecules and VSe can act as preferential sites for the adsorption of these molecules. 20,[52][53][54][55] This phenomenon can help in explaining the performance degradation reported in various works on thin InSe devices. 20,33,51,54,56 Recently, both theoretical calculations and experimental scanning transmission electron microscopy (STEM) reported that the selenium vacancies VSe in InSe crystals can be passivated by chemical dissociation of O atoms at these sites. 20,22,51 Similarly, Po-Hsun et. al. demonstrated that the change of Raman and X-ray photoelectron spectra (XPS) of pristine InSe flakes exposed to ambient air is comparable to that observed when exposing InSe to dry oxygen atmosphere. The authors attribute these observations to the formation of a surficial InSe1-xOx layer that encapsulate the InSe beneath and promote a long-term stability of thin InSe devices. 26 Despite the large amount of work on InSe photodetectors, an important question that remains unanswered is how this passivation influences the optoelectronic properties of photodetectors based on thin InSe.
In this work, by comparing optoelectronic measurements obtained in InSe photodetectors operated under various environmental conditions, we propose a model to explain the photocurrent generation mechanism in InSe devices and how it is influenced by the environment. According to our observations, the optoelectronic properties of thin InSe photodetectors are stable over a long time of more than a few weeks when the devices are stored in vacuum This is the authors' version (pre peer-review) of: Q Zhao, et al. Materials Horizons, 2020 https://doi.org/10.1039/C9MH01020C condition. When exposing the devices to ambient conditions, we observe a fast decrease of the responsivity accompanied by an increase in the operation speed (reduction of the response time) of the device. The high photoresponsivity observed in pristine thin InSe photodetector can be attributed to a strong photogating effect, which mainly originates from traps for the minority carriers (holes). After exposing the device to the air, the passivation of the chalcogen vacancies VSe by oxygen induces a decrease in photocurrent, which can be explained by a decrease in the number of traps and a quenching of photogain and photogating in the system. Decreasing the density of traps in the 2D material has also a secondary effect on the devices, by reducing the Fermi level pinning we observe an increase of the Schottky barrier at the metal-semiconductor (M-S) interface consistent with the Schottky-Mott rule, as confirmed by scanning photocurrent measurements. Our results provide a further understanding of photocurrent generation mechanism in photodetectors based on thin InSe and can pave the way in utilizing this novel material in high-performance electronics and optoelectronics.

Results and discussions
Thin InSe photodetectors are fabricated by deterministic placement of InSe flakes (thicknesses going from ~5 nm to ~20 nm) mechanically exfoliated from an InSe bulk crystal grown by Bridgman method. 57 By characterizing the InSe crystal with transmission electronic microscope (TEM) and Raman spectroscopy, we find that the phase of the crystal is ε-type, see Figure S1 of the Supporting Information and our previous work. 36 Figure 1a   we find that the thinnest region of the flake, which forms the channel in the final device, is 9.1 nm thick corresponding to approximately 11 layers. 23 Just after the fabrication of the device, we performed Raman spectroscopy measurements of the InSe flake. Figure   1c shows the Raman spectrum acquired on the region of the flake located above the gold electrode (indicated in Figure 1b). The spectrum shows three prominent peaks centered at 116 cm -1 , 178 cm -1, and 227 cm -1 . These three peaks are due to vibrational modes of InSe and they can be attributed respectively to A'1, E'' and A'1. A fourth peak is visible at 200 cm -1 and is due to A''2 (see Figure S1b), which is sensitive to the crystalline phase. 36 Since the degradation of 2D materials is a common phenomenon, we test the stability of the InSe flake by repeating the Raman measurements after two weeks of exposure to ambient conditions. As can be seen from Figure 1c and the degradation of the top ∼1-2 layers. Therefore, these Raman measurements indicate that the structure of mechanically exfoliated thin InSe flakes does not degrade completely upon exposure to ambient air, a different behavior from thin black phosporous or GaSe. 6,9 To investigate the properties of pristine InSe photodetectors, we perform optoelectronic measurements on a freshly prepared device kept in vacuum (sample #1 in Figure S3). We carried out the measurements in a homemade vacuum probe-station connected to a turbomolecular pumping station capable of reaching a base vacuum level of ∼1 × 10 -6 mbar. All the measurements presented in this article were performed at room temperature (T = 300 K). Figure 2a shows the transfer curve of the device, that is the source-drain current recorded while slowly changing the gate voltage VG. This measurement was performed with a bias voltage VDS of 1 V and keeping the device in dark conditions. The device shows negligible current at negative gate voltages (off state, VG < VT) and starts to conduct current for positive voltages (on state, VG > VT). These observations indicate that the InSe channel is n-doped and from the plot we extract a threshold voltage VT = 10 V, which is in agreement with previous reports on InSe transistors. 23,[25][26][27][28] From this transfer curve we can also estimate the electron mobility using the equation: . ( From the data of Figure 2b, we find S ~ 10 V/decade, a value that is much larger than the ideal MOSFET (metaloxide-semiconductor FET) subthreshold swing of 60 mV/decade and that is comparable to values reported previously for atomically thin high-gain photodetectors. 43 This large value for S points to the existence of trap levels in the device. 61 Additional electrical measurements of this device can be found in Figure S4 of the Supporting Information.
After the characterization of the InSe photodetector in dark conditions, we test its optoelectronic properties and response to external illumination. The blue curve in Figure and where Iph is the photocurrent calculated as (Ilight -Idark), P is the external illumination density, A is the active area of the photodetector (that we assume equal to the InSe channel area) and e is the elementary charge. Note that the formula to calculate the detectivity of Eq. 4 assumes that the photodetector is limited by shot noise. 46 In Figure 2c we show these two figures of merit R and D* as a function of gate voltage, calculated using the data of In the following, we focus on the influence of air on the performances of InSe photodetectors. Just after the device fabrication, a pristine InSe photodetector (sample #2 in Figure S3) was stored in high vacuum (pressure ~10 -6 mbar) and the optoelectronic performance was measured. To characterize the device in this case we studied its response in time to external square-wave modulated illumination. The red curve in Figure 3a shows , while in the case of 225 mW μm -2 the current reaches a value of ~25 nA and the responsivity is 0.11 A W -1 . This increase of the responsivity for decreasing illumination power densities points to the presence of traps for minority carriers (holes) in the device and we will discuss in more detail the power dependent measurements later in the article. [46][47] The slow response of the photocurrent to external illumination shown in Figure 3a can be quantitatively characterized by the response time, which we estimate using the 10%-90% criterion. Here the response time is defined as the time that it takes for the current to increase from 10% of the saturation value to 90%. From the current vs.
time traces (I-t) at 450 mW μm -2 we estimate a rising time τon = 77 s and a decay time τoff = 3.2 s. Figure 3b shows the data just discussed plotted in semi-logarithmic scale, where the small dark current of ~400 pA is visible. After carrying out the measurements in vacuum we exposed the device to air in ambient conditions for approximately 20 hours and then we repeated the I-t measurements. The green curve in Figure 3b is the corresponding I-t measured in air with the same illumination parameters used for the I-t in vacuum. As can be seen from the plot, both the dark current Idark and the current under illumination Ilight recorded in air conditions are much smaller than the initial values recorded in vacuum and the photocurrent Iph becomes ~1.2 nA, a ~40 times reduction. At the same time, we also observe a decrease of both the rising and decay time of the device, which becomes smaller than ~60 ms, approximately 2 or 3 orders of magnitude smaller than the initial values. Subsequently, after measuring the device in the air, we evacuated the chamber reaching again a pressure of ~10 -6 mbar and we carried out the same optoelectronic characterization. From the blue curve shown in Figure 3b, one can see that both Idark and Ilight remain at the same level recorded in the air indicating that the device underwent an irreversible transformation after being exposed to air for ~20 hours. The extracted responsivity is 0.002 A W -1 and both the rising and the decay time are equal to ~40 ms. From these measurements, we find that by exposing the device to air, the InSe photodetector goes from an initial state in with larger responsivity but slower time response to a final state with smaller responsivity and faster time response.
In a photodetector, important information about the mechanism behind the photocurrent generation can be extracted from the power dependency of the photocurrent. 9, 64 During the I-t measurements discussed above, we also measured the photocurrent for different light intensities going from 45 mW μm -2 to 450 mW μm -2 . For each value of light intensity, we extract the maximum photocurrent and we plot the results in Figure 3c. The photocurrent as a function of the illumination power density is shown in a log-log representation and the three datasets correspond to the three states of the device (1 pristine in vacuum, 2 after 20 hours of exposure to air, 3 in vacuum after being exposed to air). From the graph, one can see that each of the three datasets follows approximately a straight line, indicating that the photocurrent and the illumination power are connected by a power law. This relation can be expressed by the equation: where α is the dimensionless exponent of the power law and b is a parameter related to the photodetector responsivity. The value of the exponent α provides the information of traps present in the photodetecting system.
In fact, in an ideal trap-free photodetector, the exponent α is equal to 1 meaning that the photocurrent scales linearly with the illumination power and the responsivity is constant as a function of power. When trap states (for minority carriers) are present in the system α becomes smaller than 1 and the responsivity depends sub-linearly on the illumination power (as for high powers most of the traps are already filled in and further illumination power cannot effectively increase the photogain), effectively decreasing for higher illumination powers. As can be seen from Figure 3c, by fitting the data to equation (5) we find that the InSe photodetector in its pristine state in vacuum (1) is characterized by α = 0.77. This value increases to 0.94 after that 20 hours of exposure of the device to air (2).
The final measurements performed on the device in vacuum after being exposed to air (3) show a value for α of 1.0. The observed evolution of α indicates that in pristine InSe photodetectors trap states play a role in the photocurrent generation process and that these trap states can be modified and eventually irreversibly removed after exposing the InSe photodetector to air.
The increase of α in InSe photodetectors exposed to air, discussed above for a single device, has been observed in all the five investigated Au-InSe-Au devices (see Figure S5 of the Supplementary Information). Moreover, a similar increase in α has been observed also in graphite-InSe-graphite devices (see supporting information Figure S7).
Importantly, this common behavior is independent on the initial parameters of the measured devices such as responsivity, α or response time. This fact has been illustrated in Figure 4a-b where we collect the results of five different devices (shown in Figure S3) measured in the pristine state and after several exposure times to air. Figure   4a shows a semi-logarithmic graph of the decay time of the five devices plotted as a function of the exponent α, In this log-log plot, the decay time of the five different InSe photodetectors is plotted versus the photoresponsivity extracted for each device. In this case, we measured responsivity values going from 10 -3 A W -1 to 10 2 A W -1 corresponding to decay times going from less than 0.1 s to more than 1000 s. The graph shows a correlation between the two with the data-points following approximately a straight line, which indicates that the two variables are connected by a power law. The black dashed line represents a power law with an exponent equal to 1, corresponding to a linear relationship between the responsivity and the response time. As can be seen from the plot, for large values of both R and τoff, the data-points follow perfectly the linear relation, suggesting that the photogating effect is the dominant mechanism for InSe photodetectors with large responsivities. In fact, in photogating dominated devices the minority carriers get trapped in long-lived charge traps which limits in an effective way the response time of the device (they are typically slower than photoconductive devices) but provides an external source of photogain (when the channel drift time of the charge carriers is much shorter than This is the authors' version (pre peer-review) of: Q Zhao, et al. Materials Horizons, 2020 https://doi.org/10.1039/C9MH01020C the charge trapping time the device presents a photogain proportional to τoff/τdrift). 43,46 For lower values of R and τoff, the data-points start to deviate from the linear relation and present larger scattering, indicating that the photocurrent generation mechanism in these photodetectors is less dominated by photogating.
After exploring the correlation between α, R and τoff we discuss a last set of experiments that highlight the change in the electronic structure of InSe after the exposure to air. After the initial optoelectronic characterization of device #1 in vacuum (transfer curves and photocurrent power dependency at 405 nm, see Figure S4 in the Supporting Information), we expose the device to air in dark environment for 2 hours and then we evacuate again the vacuum chamber and perform the optoelectronic measurements. We repeat these measurements during air-vac-  Figure S6 of the Supplementary Information. In a second experiment, we study the evolution of the spatially resolved photocurrent in device #4 using scanning photocurrent microscopy.
Briefly, in this technique we focus a 650 nm laser in a ∼1 um 2 circular spot onto the surface of the InSe photodetector and move it across the sample while recording for each position the source-drain current and the intensity of the reflected light (see Figure S8 of the Supporting Information). Figure 4d shows the average current recorded in device #4 while scanning the laser spot from the source to the drain electrode in steps of 0.5 µm in pristine conditions (blue curves) and after 10 days of exposure to air (red). The line-profile of the pristine device at VDS = 1 V shows a broad and high photocurrent intensity region over the entire InSe channel area. On the contrary, the aged device presents a strong and sharp photocurrent peak located near the source contact at the end of the device channel, which is consistent with the presence of Schottky barriers at the InSe/gold contacts. [65][66] See figure   S8 of the Supporting Information for the measurements under applied bias of VSD = -1 V.
Before discussing a model that can explain the photocurrent generation mechanism in thin InSe photodetectors it is instructive to summarize the experimental observations presented above. In summary, we studied five different InSe photodetectors in their freshly fabricated pristine state, and after the exposure to ambient conditions.
1. From electrical measurements in a field effect transistor configuration, we find that pristine InSe photodetectors are n-type doped (Figure 2a). On the other hand, the minority carriers traps with density Pt are expected to play a dominant role in the photocurrent generation process since they can generate photogain. 20,[51][52]55 In dark conditions the InSe channel has a density of free electrons n (located in CB) and of free holes p (VB) and, as a consequence of the position of the Fermi level, there is a small density of trapped holes pt (meaning that most of the trap levels are occupied by electrons, density nt). A free hole in VB (free electron in CB) has a certain probability of getting trapped by a trap level with rate τt -1 . The opposite process also is possible and a trapped hole (electron) can jump from a trap level to VB (CB) with rate τd -1 . In general, the trapping and detrapping rates τt -1 and τt -1 can be different and there can be also a difference between the rates relative to holes and electrons.

The optoelectronic figures
The band diagram of Figure 4e can be used to explain the current flowing through the device in dark conditions and serves as a basis to understand the photogeneration of current. When we illuminate the InSe photodetectors with external illumination, electron-hole pairs are generated and the density of free holes p and free electrons n increases. The separation of these electron-hole pairs due to the electric field related to the source-drain voltage gives rise to a photocurrent. This process that is usually called "photoconducting effect" (PC) and does not show photogain. In fact, the maximum responsivity achievable in a photodetector working solely by PC for illumination at 405 nm is ∼0.33 A W -1 . In absence of active traps in the photodetector, PC is typically the dominant photocurrent generation mechanism. On the other hand, if hole traping levels are present in the system, each photogenerated hole has a certain probability (related to τt -1 ) of getting trapped in one of these levels for an average time τd. This trapping process for minority carriers can give rise to "photogating effect" (PG), a process that can show photogain and can give responsivities much larger than ∼0.33 A W -1 at 405 nm. The magnitude of the photogain is related to the ratio between the trapping time and the drift time (the trapping time can be many orders of magnitude larger than the drift time). 43,46 In the case of a pristine InSe photodetector measured in vacuum (stage 1), the large responsivity values can be explained by PG. This mechanism also explains the large response times recorded for pristine photodetectors as in the PG mechanism the reponse speed is limitted by the trapping time. The power dependency of the photocurrent also confirms this scenario. In fact, a photodetector without trap levels is expected to be characterized by a value of the exponent α equal to 1 while in a photodetector containing active traps, α is smaller than 1. For the pristine InSe devices, dominant by PG effect, the interface between gold and InSe is characterized by small Schottky barriers, consistent with the scanning photocurrent measurements of device #4, and the alignment between the Fermi level of InSe and of the gold electrodes is determined by the Fermi level pinning to the trap levels. 49,67 The right panel of Figure 4e, schematise the evolution in the band structure after the exposure of InSe photodetectors to air (stage 2). In this case, we propose that a change in the trap levels induce a strong change in the photocurrent generation process. In fact, the interaction between the selenium vacancies and the oxygen or water molecules present in air leads to a passivation of the defects, as predicted by theoretical calculations, 20     The relationship between decay time and photoresponsivity in various InSe photdetectors. All the dots in the same shape and color were measured with the same device after different air exposure times. c) Photocurrent exponent α (left axis) and threshold voltage VT (right axis) recorded in a device (#1) exposed to air as a function of time. d) Spatially resolved photocurrent at Vds = 1 V of device #4 recorded just after fabrication (blue) and after 10 days in air (red). e) Schematic band diagram of InSe in dark conditions with traps and Fermi level pinning (1, left) and in absence of traps with the alignment predicted by the Schottky-Mott rule (2, right). Depicted there are the valence (VB) and conduction band (CB) of InSe, the gold elecrodes and the Fermi energy (EF) and a set of hole trapping levels (whose density is Pt) and electron trapping level (density Nt). The density of free electrons, free holes, trapped electrons and trapped holes are respectively n, p, nt. and pt.  The Raman spectra of thin InSe flake on Au. The peak centered at 200 cm -1 gives the confirmation that the InSe is ε phase.

Figure S2
: Raman spectra (532 nm laser, power density 0.11 mW μm -2 ) of the InSe flake shown in Figure 1(b) on SiO2/Si substrate recorded in the pristine state and after two weeks in air conditions. The Raman characterization on SiO2/Si and Au/SiO2/Si indicates that the thin InSe flake in the air is structurally stable.    Figure S3) before and after air passivation. α (a), photocurrent (b), and decay time (c) of five InSe photodetectors with pristine state and after air passivation under the same measurement conditions. All the tested InSe photodetectors share the same manner when exposed to air. Figure S6: Time-dependent performance of the InSe photodetector (#1) when exposed to air. The gate-dependence of the InSe photodetector in dark (a), photoresponsivity (b) and detectivity (c) of 405 nm at 0.92 W m -2 illumination after exposed to air 0 h, 2 h, 11 h. The evolution of photocurrent (at 906 W m -2 ), rising and decay time (at 906 W m -2 ), and photocurrent -illumination intensity dependence of 405 nm light as a function of exposure time in air.

Figure S7
: Time-dependent performance of a graphite-InSe-graphite photodetector when exposed to air. Figure S8: a) Optical image of device #4. b) Intensity of the laser reflection from the sample recorded during the scanning photocurrent measurements of panels c and d. c-d) Spatially resolved photocurrent at Vds = ±1 V of device #4 recorded just after fabrication (blue) and after 10 days in air (red).

TOC graphics:
The role of traps in the photocurrent generation mechanism in thin InSe photodetectors