High-Responsivity Multilayer MoSe2 Phototransistors with Fast Response Time

There is a great interest in phototransistors based on transition metal dichalcogenides because of their interesting optoelectronic properties. However, most emphasis has been put on MoS2 and little attention has been given to MoSe2, which has higher optical absorbance. Here, we present a compelling case for multilayer MoSe2 phototransistors fabricated in a bottom-gate thin-film transistor configuration on SiO2/Si substrates. Under 650-nm-laser, our MoSe2 phototransistor exhibited the best performance among MoSe2 phototransistors in literature, including the highest responsivity (1.4 × 105 AW−1), the highest specific detectivity (5.5 × 1013 jones), and the fastest response time (1.7 ms). We also present a qualitative model to describe the device operation based on the combination of photoconductive and photogating effects. These results demonstrate the feasibility of achieving high performance in multilayer MoSe2 phototransistors, suggesting the possibility of further enhancement in the performance of MoSe2 phototransistors with proper device engineering.


Results and Discussion
Before fabricating MoSe 2 transistors, we first measure the optical absorbance of MoSe 2 crystals and mechanically exfoliated flakes on sapphire substrates across visible and near-infrared spectral ranges ( Fig. 1(a)). The MoSe 2 crystal is thicker than 100 μm and the thickness of exfoliated MoSe 2 flakes are in the range of 20-80 nm. Both samples show two excitonic absorbance peaks A and B at 1.55 eV and 1.82 eV, respectively, which is consistent with literature 20 . Next, multilayer MoSe 2 transistors are fabricated on SiO 2 /Si substrates. Figure 1(b) shows the optical microscopy image of a completed MoSe 2 transistor along with its schematic cross-section. The measured transfer curve of an MoSe 2 transistor in Fig. 1(c) shows asymmetric ambipolar behavior with strong n-type characteristic (MoSe 2 thickness (t) = 25 nm). For electron transport without light, the MoSe 2 transistor exhibits on/ off-current ratio (I on /I off ) of 10 5 and field-effect mobility (μ FE ) of 50.6 cm 2 V −1 s −1 extracted from μ FE = L(dI d /dV g )/ (WC ox V d ), where L is the channel length (5 μm), I d is drain current, V g is gate voltage, W is the channel width (27 μm), C ox is the oxide capacitance, and V d is the drain voltage (1 V). For hole transport without light, I on /I off of 10 4 and μ FE of 2.8 cm 2 V −1 s −1 are obtained. The n-type-dominant ambipolar behavior of MoSe 2 transistors with Ti/Au electrodes was also observed in literature 21,22 . The output curves in Fig. 1(d) show linear region at low V d suggesting decent contact properties. The transfer and output characteristics of an MoSe 2 transistor in Fig. 1(c,d) show the increase of I d with the power density of incident light.
The photocurrent (I ph ) of phototransistors based on transition metal dichalcogenides such as MoSe 2 is known to be dominated by photoconductive effect and photogating effect 23 . In photoconductive effect, photogenerated excess carriers increase conductivity resulting in increased current. The photocurrent component flowing between two electrodes by photoconductive effect is given by 24 where Δσ, E, D, Δn, q, μ, η, P in , h, ν, and τ are change in conductivity, electric field, depth of absorption region, change in carrier concentration, unit charge, carrier mobility, quantum efficiency, incident optical power, Planck constant, frequency of incident light, and carrier lifetime, respectively. The photoconductive component of I ph is proportional to areal power density of incident light P in and weakly depends on V g 23,24 . In photogating effect, one type of photogenerated carriers (electrons or holes) is trapped in localized states and the other type of carriers flows in the channel unrecombined. As this is equivalent to doping by the other type of carriers, photogating effect accompanies a shift of threshold voltage (V th ) 25 . As V th shifts, the drain current changes from I d to I d + ΔI d and it follows that 26 η λ  current, speed of light, respectively. Thus, the photogating component of I ph shows logarithmic dependence on P in and is roughly proportional to transconductance (g m ) 23,26 . Figure 2(a) shows I ph and g m as a function of V g for the same device in Fig. 1(c). The calculation of I ph (I ph = I light -I dark , where I light is I d in a detector with light), and g m (g m = dI d /dV g ) is based on the data in Fig. 1(c) at P in = 18 mWcm −2 . The similarity between I ph and g m suggests that photogating effect dominates the photoresponse of MoSe 2 transistors. In the inset of Fig. 2(a), the change in V th (ΔV th ) for electrons and holes is shown as a function of P in . The increasing change of V th with increasing P in also suggests the domiant role of photogating effect in our MoSe 2 phototransistors. However, the dependence of I ph on V g in Fig. 2(b) through (d) suggests that each effect dominates I ph at different range of V g . Figure 2(b) through (d) show I ph of our MoSe 2 phototransistor in an on-state for electrons (at V g = 40 V), an off-state (at V g = 0 V), and an on-state for holes (at V g = −40 V) as a function of P in in sequence. I ph is calculated based on the data in Fig. 1(c). The I ph in an on-state for electrons (at V g = 40 V) and for holes (at V g = −40 V) shows logarithmic dependence on P in suggesting the dominant role of photogating effect. However, the linear dependence of I ph in an off-state suggests the dominant role of photoconductive effect. Such a distinct dependence of I ph on V g regime was also observed in phototransistors based on The performance of an MoSe 2 transistor as a photodetector can be evaluated by responsivity (a measure of the electrical response to light) and specific detectivity (a measure of detector sensitivity) 31 . Responsivity (R) is given by R = (I light − I dark )/(P in A), where A is the area of the detector. Under the assumption that shot noise from I dark is the major contributor to the total noise, specific detectivity (D*) is given by 32 D* = RA 1/2 /(2qI dark ) 1/2 . Figure 4(a,b) show the calculated R and D* of the MoSe 2 phototransistor at different P in and V g . Maximum R of 1.4 × 10 5 AW −1 As R and D* increase with decreasing P in , the enhancement of R and D* in this work may be due to the low P in compared to that in literature. However, even at comparable P in in the range of 18-54 mWcm −2 , the maximum R and D* in this work (R = 519 AW −1 and D* = 1.3 × 10 12 jones) are about twice as high as those in literature. In Fig. 4(a,b), the overall dependence of R and D* on P in and V g is consistent with literature 13 . R increases as P in decreases or V g increases, while D* increases as P in or V g decreases. As P in increases, more holes fill shallow trap states where lifetime is short. This results in faster recombination hence R decreases. When P in increases, D* also decreases as R decreases and I dark remians unchanged. When V g increases, electrical doping at higher V g reduces contact resistance resulting in higher photocurrent and R. However, as V g increases, I dark also increases, which degrades D*.
It needs to be mentioned that our MoSe 2 transistors show wide device-to-device variation of μ FE , R, and D* (Table S1 in Supplementary Information). Such wide device-to-device variation is commonly observed in the transistors based on transition metal dichalcogenides such as MoSe 2 presumably because of the variation of intrinsic defects in crystals 33 . While it is very difficult to pinpoint the origin of high performance in the best device, the correlation between R and μ FE in this work (Fig. S1 in Supplementary Information) suggests that the enhanced optoelectronic properties may be related to the enhanced electrical performance of our MoSe 2 device. It is also supported by the fact that our MoSe 2 device shows the highest mobility among MoSe 2 transistors in Table 1.
We also note the negligible correlation between the optoelectronic properties of MoSe 2 devices and MoSe 2 thickness. This may seem counterintuitive because the width of energy bandgap changes for thin MoS 2 crystals (< ~4 nm in thickness) 4 and light absorption depends on MoSe 2 thickness. Yet, because the thickness of our MoSe 2 flakes ranges from 20 nm to 80 nm, we expect negligible differences in energy bandgap in our MoSe 2 devices. On the other hand, we expect higher responsivity for devices with thicker MoSe 2 as more light is absorbed in thicker MoSe 2 . However, the responsivity shows negligible correlation with thickness of MoSe 2 flakes in this investigation (Fig. S2 in Supplementary Information). This may be due to the variation of intrinsic materials quality overshadowing the effect of thickness. The mobility and detectivity in Fig. S2 also show negligible correlation with the thickness of MoSe 2 flakes, supporting this argument.
To explore the response time of our MoSe 2 phototransistors, we measure the time-resolved photoresponse of our MoSe 2 phototransistors for multiple illumination cycles. Figure 4(c) shows the result for the same device in Fig. 1(c). The incident laser with a power density of 54 mWcm −2 is modulated with a square wave at 100 Hz at V g = 15 V and V d = 10 V. The nearly identical response for multiple cycles suggests the overall rebustness and reproducibility of our MoSe 2 phototransistors. From a zoomed-in region in Fig. 4(d), we obtain rise time of 1.7 ms and fall time of 2.2 ms. (Rise time is calculated as the time taken by current to increase from 10% to 90% of the maximum current. Fall time is calculated as the time taken by current to decrease from 90% to 10% of the maximum current.) This is the fastest response time of MoSe 2 phototransistors ever reported in literature, which ranges from 5 ms to 400 ms 13,[15][16][17][18][19] . It is intriguing that our MoSe 2 phototransistors exhibit high responsivity and fast response time. Because the long lifetime of carriers in photogating effect suggests slow response to light, the fast response time in our MoSe 2 device may be related to the characteristics of trap states. One possibility is that trap states in our MoSe 2 device may have shorter lifetime and higher density than those in literature. Then,  while the shorter lifetime of trap states could provide fast response, the higher density of trap states could provide higher doping enhancing responsivity. However, we may only speculate at this stage and further investigation is needed on the characteristics of trap states including the distribution of trap energy, trap density, trap lifetime, and carrier capture probability. Table 1 compares μ FE , R, D*, and response time of MoSe 2 phototransistors in literature. Because the measurement conditions, such as V g , V d , P in , and excitation energy, can influence the device performance, comparable measurement conditions with those in literature are used in this work. Our MoSe 2 phototransistors exhibit the best performance in terms of μ FE , R, D*, and response time, demonstrating the feasibility of achieving high responsivity and fast response time in multilayer MoSe 2 phototransistors. Future work combining controlled growth of materials with optimized device architecture and processing will further enhance the performance of MoSe 2 phototransistors.

Conclusions
We report high-responsivity multilayer MoSe 2 phototransistors with fast response time fabricated with mechanically-exfoliated MoSe 2 flakes on SiO 2 /Si substrates. Our MoSe 2 phototransistors exhibit asymmetric ambipolar behavior with strong n-type characteristic. Without light, high on/off-current ratio of 10 5 and field-effect mobility of 50.6 cm 2 V −1 s −1 are obtained for electrons. Under 650-nm-laser, our MoSe 2 phototransistor exhibits the best performance among MoSe 2 phototransistors in literature including high responsivity (1.4 × 10 5 AW −1 ), high specific detectivity (5.5 × 10 13 jones), fast rise time (1.7 ms) and fast fall time (2.2 ms). The dependence of photocurrent on gate voltage and optical power density suggest that photocurrent is dominated by photogating effect in on-state and by photoconductive effect in off-state. These results demonstrate the feasibility of achieving high-performance multilayer MoSe 2 phototransistors, providing potentially important implications on using MoSe 2 phototransistors for a variety of applications including touch sensor panels, image sensors, solar cells, and communication devices.

Methods
Device fabrication. Multilayer MoSe 2 flakes were obtained by gold-mediated mechanical exfoliation 34 from bulk MoSe 2 crystals (2D Semiconductors) and transferred to highly doped p-type Si wafer with thermally grown SiO 2 (300 nm). The thickness of MoSe 2 flakes measured by atomic force microscope (AFM, Park Systems XE-100) existed between 20 nm and 80 nm. To form source and drain electrodes (100 μm × 100 μm) on top of MoSe 2 flakes, Ti (20 nm) and Au (50 nm) deposited by electron-beam evaporation were patterned using photolithography and etching. The device was then annealed at 200 °C in a vacuum tube furnace for 2 hours (100 sccm Ar and 10 sccm H 2 ) to remove resist residue and to decrease contact resistance.
Device characterization. Optical absorbance of MoSe 2 was measured by UV-visible spectroscopy (Perkin-Elmer Lambda 35). Electrical characterizations were carried out with current-voltage (I-V) measurements (Agilent 4155 C Semiconductor Parameter Analyzer) at room temperature. The photoresponse of MoSe 2 phototransistors was measured with a 650-nm-laser (beam size of 3 mm) at different power densities (0.027, 18 and 54 mW cm −2 ). Dynamic on/off switching was conducted using a function generator (Tekronix AFG310).