Persistent photoconductivity in two-dimensional Mo_1−x W _x Se₂–MoSe₂ van der Waals heterojunctions

Abstract

Van der Waals (vdW) heterojunctions consisting of vertically-stacked individual or multiple layers of two-dimensional layered semiconductors, especially the transition metal dichalcogenides (TMDs), show novel optoelectronic functionalities due to the sensitivity of their electronic and optical properties to strong quantum confinement and interfacial interactions. Here, monolayers of n-type MoSe₂ and p-type Mo_1−x W _x Se₂ are grown by vapor transport methods, then transferred and stamped to form artificial vdW heterostructures with strong interlayer coupling as proven in photoluminescence and low-frequency Raman spectroscopy measurements. Remarkably, the heterojunctions exhibit an unprecedented photoconductivity effect that persists at room temperature for several days. This persistent photoconductivity is shown to be tunable by applying a gate bias that equilibrates the charge distribution. These measurements indicate that such ultrathin vdW heterojunctions can function as rewritable optoelectronic switches or memory elements under time-dependent photo-illumination, an effect which appears promising for new monolayer TMDs-based optoelectronic devices applications.

I. INTRODUCTION

Two-dimensional (2D) layered semiconductors, especially transition metal dichalcogenides (TMDs), have emerged as exciting and versatile materials when their thickness is reduced to a monolayer or few-layers due to the emergence of quantum confinement effects and strong interfacial interactions. ^1,2 Stacking different 2D semiconductors into van der Waals (vdW) heterojunctions creates a diverse palette of new, artificially-structured, layered materials with tunable optoelectronic properties depending on the stacking order, relative orientation angle, and atomic registry between the layers. Recently, p–n vdW heterojunctions have been constructed by either stacking or epitaxially growing different layered materials on top of one another or growing them laterally, and have demonstrated desirable optoelectronic functionalities such as photodetectors, photovoltaics, light-emitting diodes, and photodiodes. ^3–11 Persistent photoconductivity (PPC), an interesting optoelectronic effect where enhanced electrical conductivity persists after the removal of light illumination, ^12–14 has so far received less attention, but has obvious potential for rewritable photoresponsive switches and memory elements. ¹⁵

PPC has been primarily observed in compound semiconductors and layered structures. ^16–18 However, 2D materials have suitable properties for PPC that include strong light-matter interaction, high carrier mobility, and gate tunability. ¹⁵ While graphene has low photoresponsivity due to its intrinsic metallic property, hybrid graphene structures utilizing quantum dots or MoS₂ atomic layers exhibit interesting PPC properties. ^15,19 TMDs, especially MoS₂ and MoSe₂, are highly photoresponsive materials with strong optical absorption and band gaps in the visible spectrum. vdW heterostructures constructed from these and other TMDs are good candidates for photoresponsive hybrid materials with charge exchange across the atomically sharp interfaces governed by quantum tunneling transport.

Doping or alloying have proven very effective ways to tune the optical and electrical properties in TMD monolayers. ^20,21 Recently, we found that isoelectronic substitution of Mo by W in the monolayer MoSe₂ lattice can switch it from n-type to p-type behavior. ²² The resulting alloy, i.e., Mo_1−x W _x Se₂, has almost the same lattice constant as intrinsic MoSe₂, but exhibits a tunable band gap with doping concentration, providing a tunable candidate for the construction of vdW heterojunctions.

In this study, vdW heterojunctions are fabricated by transferring and stacking monolayer single-crystals of MoSe₂ and Mo_1−x W _x Se₂ that were grown by vapor transport methods. The excellent heterojunction quality is assessed by aberration-corrected transmission electron microscopy (TEM), and strong interlayer coupling verified by low-frequency (LF) Raman spectroscopy and photoluminescence (PL). Electrical measurements are used to characterize the p–n junction formed by the stacked MoSe₂ and Mo_1−x W _x Se₂, which are also found to exhibit extremely high photoresponsivity and giant PPC at room temperature. Under time-dependent photo-illumination, these heterojunctions can function as rewritable optoelectronic switches or memory elements.

II. EXPERIMENTAL

A. Material growth

Crystalline monolayers of MoSe₂ and Mo_1−x W _x Se₂ were synthesized using a low pressure chemical vapor deposition (CVD) approach that is similar to those described previously. ²³ The synthesis was conducted in a tube furnace CVD reactor equipped with a 2 in. quartz tube. In a typical run, the growth substrates, i.e., Si wafer with 250 nm SiO₂ (SiO₂/Si) cleaned by acetone and isopropanol, were placed face down above an alumina crucible containing ∼0.2 g of MoO₃ powder (for the growth of Mo_1−x W _x Se₂, a mixture of MoO₃ and WO₃ powder was used), which was then inserted into the center of the quartz tube. Another crucible containing ∼1.2 g Se powder was located at the upstream side of the tube. After evacuating the tube to ∼5 × 10⁻³ Torr, flows of 40 sccm (standard cubic centimeter per minute) argon and 4 sccm hydrogen gas were introduced into the tube, and the reaction was conducted at 780 °C (with a temperature ramping rate of 30 °C/min) for 5 min at a reaction chamber pressure of 20 Torr. At 780 °C, the temperature at the location of Se powder was ∼290 °C. After growth, the furnace was cooled naturally to room temperature.

B. Monolayer crystal transfer and heterojunction fabrication

For the heterojunction fabrication and TEM sample preparation, poly(methyl methacrylate) (PMMA) was first spun onto the monolayer crystals on the SiO₂/Si substrate at 3500 rpm for 60 s. The PMMA-coated substrate was then floated on 1 M KOH solution that etched silica epilayer, leaving the PMMA film with the monolayer crystals floating on the solution surface. The film was transferred to deionized water several times to remove residual KOH. For TEM samples, the washed film was captured on a Si TEM grid covered by a 50 nm-thick amorphous SiN film with 2 μm windows. For heterojunction fabrication, the film with the monolayer MoSe₂ was stacked onto the substrate with the monolayer Mo_1−x W _x Se₂. The PMMA was removed by acetone and baking at 300 °C in ∼30 Torr Ar/H₂ (95%/5%) flowing for 2 h.

C. Device fabrication

Electron beam lithography (FEI DB-FIB with Raith pattern writing software; FEI Company, Hillsboro, Oregon) was used for monolayer MoSe₂–Mo_1−x W _x Se₂ heterojunction device fabrication. A layer of PMMA 495A4 was spun-coat on the SiO₂ (250 nm)/Si substrate with monolayer MoSe₂–Mo_1−x W _x Se₂ heterojunctions, followed by a 180 °C annealing. After pattern writing, development, and lift off, a 10 nm layer of Ti followed by a 50 nm layer of Au was deposited using electron beam evaporation.

D. Characterizations

The morphologies of the monolayer MoSe₂ and Mo_1−x W _x Se₂ crystals were characterized using optical microscopy (Leica DM4500 P, Wetzlar, Germany), scanning electron microscopy (SEM, Zeiss Merlin, Oberkochen, Germany), and atomic force microscopy (AFM, Bruker Dimension Icon, Billerica, Massachusetts). The atomic structures of monolayer MoSe₂ and Mo_1−x W _x Se₂ were investigated via annular dark field (ADF) imaging using an aberration-corrected scanning transmission electron microscope (STEM) (ADF-STEM, Nion UltraSTEM™ 100, Kirkland, Washington) operating at 100 kV, using a half-angle of the ADF detector that ranged from 86 to 200 mrad.

Raman measurements were performed using a micro-Raman system (JobinYvon Horiba, T64000, Edison, New Jersey) based on a triple spectrometer equipped with three 1800 groves/mm gratings and a liquid nitrogen cooled charge-coupled device (CCD) detector. The Raman spectra were acquired under a microscope in backscattering configuration using 532 nm laser excitation (0.1 mW laser power). The excitation laser was focused to a ∼1 μm spot using a microscope objective (100x, numeric aperture, N/A = 0.9).

PL measurements were conducted using a home-built micro-PL setup, which included an upright microscope coupled to a spectrometer (Spectra Pro 2300i, Princeton Instruments, Acton, Massachusetts, f = 0.3 m, 150 grooves/mm grating) equipped with a CCD camera (Pixis 256BR, Princeton Instruments). The PL was collected through a 100x objective.

The electrical properties and photoresponse of the monolayer flakes and heterojunctions were measured in vacuum (∼10⁻⁶ Torr) under a probe station using a semiconductor analyzer (Keithley 4200, Keithley Instruments, Cleveland, Ohio) and a laser driven white light source with a power density of 64.42 mW/cm² from 400 to 800 nm.

III. RESULTS AND DISCUSSION

Figure 1(a) shows an optical micrograph of the as-grown MoSe₂ crystalline flakes and Fig. 1(b) shows a SEM image of the as-grown Mo_1−x W _x Se₂ flakes on SiO₂/Si substrates. The two types of flakes show the same triangular shape and similar sizes ranging from tens to a hundred micrometers. AFM images were used to measure the thickness of these crystals (∼0.7nm as in Fig. 1(c)), and correspond to monolayers of MoSe₂ and Mo_1−x W _x Se₂. This indicates that large-sized, uniform monolayer MoSe₂ and Mo_1−x W _x Se₂ were synthesized. To make MoSe₂–Mo_1−x W _x Se₂ heterojunctions, the as-grown monolayer MoSe₂ flakes were transferred from the substrate and stacked onto the as-grown monolayer Mo_1−x W _x Se₂ flakes. The optical micrograph [Fig. 1(d)] and SEM images [Figs. 1(e–f)] show that many stacked flakes are formed with arbitrary interlayer rotation angles, and the overlapping region can be clearly distinguished (as indicated by solid red arrows).

FIG. 1. Morphologies of monolayer MoSe₂, Mo_1−x W _x Se₂, and stacked flakes. (a) Optical micrograph of the monolayer MoSe₂ flakes grown on SiO₂/Si substrate. (b) SEM image of the monolayer Mo_1−x W _x Se₂ flakes grown on SiO₂/Si substrate. (c) AFM image of monolayer Mo_1−x W _x Se₂ flakes. Inset is the height profile along the solid blue arrow. (d–f) Optical micrograph and SEM images of as-grown monolayer MoSe₂ flakes transferred and stacked onto as-grown monolayer Mo_1−x W _x Se₂ flakes. The solid red arrows indicate the overlapping (junction) region.

Figure 2(a) shows a typical atomic resolution ADF-STEM (Z-contrast) image of an as-grown monolayer of Mo_1−x W _x Se₂. A hexagonal lattice structure with a lattice constant a = 0.329 nm is clearly displayed (see also the fast Fourier transform [FFT] pattern in the inset), which is the same as pristine monolayer MoSe₂. The atoms in Fig. 2(a) show different brightness since the intensity of the image is directly related to the atomic number of each atom. Figure 2(b) shows an intensity profile along the solid red arrow in Fig. 2(a), which typically represents the three types of sites in the monolayer Mo_1−x W _x Se₂. The Se sites show slightly higher intensity than the Mo sites because each Se column contains two atoms, while the sites showing the highest intensity correspond to Mo substituted by W. Therefore, the monolayer Mo_1−x W _x Se₂ shows alternating atomic arrangement of Mo and Se sites in the hexagonal rings, with the Mo atoms partially and randomly substituted by W atoms. According the ADF-STEM results, the local concentration of W is ∼18%. Figure 2(c) shows an ADF-STEM image of a monolayer of MoSe₂/Mo_1−x W _x Se₂ stack. Periodic Moiré patterns are observed due to interlayer rotation, and the FFT pattern [inset of Fig. 2(c)] shows that the two lattices are misaligned by ∼15°. The Moiré pattern also indicates that the interface between the two monolayers is clean and atomically sharp.

FIG. 2. Atomic structure of monolayer Mo_1−x W _x Se₂. (a) Atomic-resolution ADF-STEM image of monolayer Mo_1−x W _x Se₂. Inset is the corresponding FFT pattern. (b) Intensity profiles along the solid red arrow in (a). (c) Atomic-resolution ADF-STEM image of the overlapping (junction) area of two stacked monolayer MoSe₂ and Mo_1−x W _x Se₂. Inset is the corresponding FFT pattern, indicating a 15° of interlayer rotation.

The optical properties of monolayer MoSe₂, Mo_1−x W _x Se₂, and their stacked heterojunctions [Fig. 3(a)] were studied using Raman and PL spectroscopies at room temperature, using 532-nm laser excitation. Three typical Raman modes, i.e., the dominant A_1g mode (out-of-plane), and the weak in-plane E_1g and ${\rm{E}}_{2{\rm{g}}}^1$ modes, show up in the Raman spectra of both monolayer MoSe₂ and Mo_1−x W _x Se₂ [Fig. 3(b)]. ²⁴ The enlarged Raman spectrum [inset in Fig. 3(b)] indicates that the A_1g mode of monolayer Mo_1−x W _x Se₂ is shifted from 240.9 (monolayer MoSe₂) to 242.1 cm⁻¹, which is consistent with previous reported monolayer Mo_1−x W _x Se₂ alloys. ²⁴ Figure 3(c) shows the room temperature PL spectra of individual monolayers of MoSe₂ and Mo_1−x W _x Se₂. The as-grown monolayer of MoSe₂ on SiO₂/Si exhibits a single emission band peaking at ∼1.528 eV [Fig. 3(c), solid black curve], corresponding to radiative recombination of the A-exciton in monolayer MoSe₂. ²⁵ The A-exciton emission in the as-grown monolayer of Mo_1−x W _x Se₂ on SiO₂/Si shows a similar band shape, but exhibits a blue-shift of the emission peak to ∼1.554 eV [Figs. 3(c) and 3(d), solid red curves] compared with MoSe₂. The blue shift can be related to the increase of band gap energy due to W-substitution. Interestingly, compared with the as-grown monolayer MoSe₂ and Mo_1−x W _x Se₂, the transferred monolayers [e.g., the MoSe₂ flake shown in Fig. 3(a)] both exhibits blue shift of the emission spectra, peaking at ∼1.577 eV [Figs. 3(c) and 3(d), solid green curve, obtained from the spot 1 in Fig. 3(a)] and ∼1.603 eV [dashed red curve in Fig. 3(c)], respectively. Such a blue shift can be attributed to the different interaction of the as-grown and transferred monolayers with the substrate.

FIG. 3. Optical properties of monolayer MoSe₂ and Mo_1−x W _x Se₂. (a) Optical micrograph of CVD-grown monolayer MoSe₂ (larger flake) transferred and stacked onto CVD-grown monolayer Mo_1−x W _x Se₂ (smaller flake). (b) Raman spectra of the as-grown monolayer MoSe₂ (black curve) and Mo_1−x W _x Se₂ (red curve) with 532 nm laser excitation. Note that the spectra were offset for clarity. (c) Normalized PL spectra of the as-grown monolayer MoSe₂ (solid black curve) and Mo_1−x W _x Se₂ (solid red curve), and the CVD-grown monolayer MoSe₂ (green curve) and Mo_1−x W _x Se₂ (dashed red curve) after transfer with 532 nm laser excitation. (d) PL spectra obtained from spot 1 (transferred CVD-grown monolayer MoSe₂), 2 (as-grown monolayer Mo_1−x W _x Se₂), and 3 (overlapping region) labeled on (a), represented by solid green, red, and blue curves, respectively. (e) LF Raman spectrum (solid blue curve) obtained from the overlapping area in (a). The solid red curve is the calculated LF Raman spectrum.

To study the interlayer coupling in the heterojunction composed of MoSe₂ and Mo_1−x W _x Se₂, monolayers, we measured the PL from the overlapping region of the two flakes shown in Fig. 3(a). The PL intensity from the heterojunction [solid blue curve in Fig. 3(d)] is significantly quenched and the emission band is peaked at a lower energy (i.e., ∼1.507 eV) compared with the individual monolayer MoSe₂ [solid green curve in Fig. 3(d)] and Mo_1−x W _x Se₂ [solid red curve in Fig. 3(d)]. The significant PL quenching and red-shifted emission band suggest that the emission from the heterojunction area originates from interlayer charge recombination, ⁴ and the emergence of such emission also indicates strong interlayer coupling in the stacked heterojunction.

LF (below 50 cm⁻¹) Raman spectroscopy contains rich information regarding the LF shear and breathing modes associated with in-plane and out-of-plane interlayer vibrations of the whole layers, which is an effective and sensitive way to understand the vdW interactions and coupling between layers in stacked 2D crystals. ²⁶ Figure 3(e) shows the LF Raman spectrum obtained from the stacked MoSe₂/Mo_1−x W _x Se₂ region in Fig. 3(a). Strong, narrow peaks associated with LF shear (at 19.1 cm⁻¹) and breathing (at 32.9 cm⁻¹) modes are observed, which is more evidence for strong interlayer coupling in the heterojunction. According to our previous study on bilayer MoSe₂, when the two layers are stacked at exactly 60° interlayer rotation (i.e., 2H stacking configuration), the LF spectrum shows a dominant narrow shear mode peak at 19 cm⁻¹ and a very weak, broad feature at ∼34 cm⁻¹ related to the breathing mode. ²⁶ In the current stacked MoSe₂/Mo_1−x W _x Se₂ shown in Fig. 3(a), the interlayer rotation is measured to be 57.3°, slightly shifted from perfect 2H stacking, and this leads to significant enhancement of the breathing mode. Such a change in the LF Raman spectrum is attributed to the periodic arrangement of patches with different stacking configurations (i.e., high-symmetry 2H, AB′, A′B stacking and unaligned bilayers) as a result of the slight deviation in the interlayer rotation from 60° (i.e., perfect 2H stacking). Details of this study are presented elsewhere. ²⁷

To investigate the optoelectronic properties of the MoSe₂/Mo_1−x W _x Se₂ heterojunction, a device was made by patterning source–drain contacts (Ti/Au) on both flakes while using the highly-doped Si substrate as the back-gating electrode [Fig. 4(a)]. Our previous study has already demonstrated that MoSe₂ monolayers showed n-type conduction, which switched to p-type in Mo_1−x W _x Se₂ monolayers. ²² Indeed, the output (I _ds–V _ds) curve (at V _bg = 0 V) from the MoSe₂/Mo_1−x W _x Se₂ heterojunction region clearly shows a typical rectifying behavior, with current only being able to pass through the device when the p-type Mo_1−x W _x Se₂ is forward biased [Fig. 4(b), solid black curve]. To verify that the rectifying behavior originated from the junction, we also measured output curves of the individual MoSe₂ and Mo_1−x W _x Se₂ flakes in Fig. 4(a). The output curves show linear regions [inset of Fig. 4(c)], indicating Ohmic contacts between the electrodes and flakes. When shown on logarithmic scales, the output curves show very symmetrical behaviors [Fig. 4(c)], demonstrating that the rectifying behavior originated from the p–n heterojunction formed by stacking of Mo_1−x W _x Se₂ and MoSe₂ monolayers. The photoresponse of the heterojunction shown in Fig. 4(a) was studied using a white light lamp with power density of 64.42 mW/cm² as the illumination source. Under white light illumination, the I _ds (at V _bg = 0 V) of the heterojunction was significantly increased by ∼3 order of magnitude compared to the dark current, with a photoresponsivity of ∼13.2 A/W at V _ds = −5 V [Fig. 4(b), solid red curve]. This result demonstrates that the stacked heterojunction have a very good photoresponse. However, after removal of light source, the current in the p–n junction did not immediately decrease back to that originally measured for the dark state; instead, it experienced a fast decay in the first hour and then decreased very slowly, persisting with enhanced photocurrent for more than 2 days [Fig. 4(d)]. The higher conductivity state activated by light exposure persisted for a long time (several days) after the removal of light, showing a giant PPC. The current decay can be fitted exponentially, with a time constant τ of ∼12.2 h [Fig. 4(e)]. Interestingly, the PPC was not observed from single monolayers of MoSe₂ or Mo_1−x W _x Se₂, which show fast response time as the light is switched on and off [Fig. 4(f)], typical for the photoresponse behavior of 2D TMDs. It can be concluded that the giant PPC comes only from the junction area. The giant PPC in the monolayer MoSe₂–Mo_1−x W _x Se₂ heterojunction may be due to the following two reasons: (i) the existence of potential barriers or surface barriers on the interface of the heterojunction that spatially separate the photo-generated charge carriers. ¹⁸ (ii) The possible existence of traps at the interface of the heterojunction that extend electron-hole recombination lifetime. The detailed mechanism behind the PPC behavior is under further study.

FIG. 4. PPC and photoresponsive switching of the MoSe₂–Mo_1−x W _x Se₂ heterojunction. (a) Optical micrograph of a CVD-grown monolayer MoSe₂ (upper flake) transferred and stacked onto a CVD-grown monolayer Mo_1−x W _x Se₂ (lower flake). The overlapping region of the two flakes forms a heterojunction. Electrodes 1 and 2 were patterned on MoSe₂ while 3 and 4 on Mo_1−x W _x Se₂. (b) I _ds–V _ds (with I _ds shown in absolute values) curves of the heterojunction area shown in (a) on logarithmic scales at zero back-gate voltage in dark (solid black curve) and under white light illumination (solid red curve). (c) I _ds–V _ds (with I _ds shown in absolute values) curves of the individual monolayer MoSe₂ (solid blue curve) and Mo_1−x W _x Se₂ (solid red curve) shown in (a) on logarithmic scales at zero back-gate voltage in dark. Inset is the corresponding I _ds–V _ds curves on linear scale. (d) I _ds–V _ds curves of the heterojunction at zero back-gate voltage in dark (solid black curve), under white light illumination (solid red curve), and after removal of the illumination for 5 min (solid blue curve), 10 min (solid orange curve), 20 min (solid green curve), 30 min (solid magenta curve), 40 min (solid yellow curve), and 3000 min (solid pink curve). (e) Decay of the photocurrent after removal of white light source (solid squares), which was fitted exponentially (red curve) with τ = 12.2 h. (f) Time-resolved photoresponse at a bias voltage of −5 V of the individual Mo_1−x W _x Se₂ flake shown in (a). (g) Photocurrent induction and switching operation of the heterojunction. The gray-shaded areas indicate the presence of white light illumination, during which the current gradually increased to saturation level at V _ds = −5 V. The green dashed vertical lines indicate the application of gate pulses (100 V), where the current was recovered to the initial level.

The photoresponse of the MoSe₂–Mo_1−x W _x Se₂ heterojunction was further investigated by studying time-resolved current change at V _ds = −5 V. Figure 4(g) shows the result of three photo-illumination cycles. Upon illumination, the photocurrent (−I _p) increased gradually and saturated at a certain level after ∼120 s (at zero back-gate voltage). After switching the light off, the device showed PPC with slow decay of current (with a time constant τ of ∼12.2 h); however, the current could be rapidly and fully recovered to the initial level (dark state) by applying a gate bias pulse at 100 V, which equilibrated the charge distribution. Repeating these steps again resulted in the same behavior, showing the robustness of this heterojunction. Such behavior indicates that the MoSe₂–Mo_1−x W _x Se₂ heterojunction with PPC can potentially be designed to function as photoresponsive memory devices.

In summary, vdW heterostructures were fabricated by transferring and stacking as-grown crystalline monolayers of MoSe₂ onto as-grown monolayers of Mo_1−x W _x Se₂. These heterostructures showed strong interlayer coupling as proven by the emergence of interlayer charge transfer PL and the presence of LF Raman modes. When illuminated with white light, the heterojunction shows extremely high and persistent photocurrent lasting days, which appear applicable for rewritable optoelectronic switch and memory applications involving time-dependent photo-illumination. This MoSe₂–Mo_1−x W _x Se₂ heterojunction can potentially lead to new monolayer TMDs-based optoelectronic devices.