Transition metal dichalcogenides (TMDs) have attracted considerable attention as flexible and highly sensitive photodetectors due to their exceptional electrical, optical, and mechanical properties1, 2. Molybdenum disulfide (MoS2) is the most promising TMD material described to date, and has been widely studied for stable and large-area thin film growth with 2H-phase3. Various approaches have been reported to realize high-performance photodetectors based on MoS2, such as increasing light absorption by integrating with nanomaterials or conventional 3D materials4, 5, generating built-in junctions by localized doping6, and 2D-2D van der Waals heterostructure construction7. Such systems can achieve photodetection in the visible spectral range with ultra-high responsivity (~ 106 A/W) and ultra-fast photoresponse (~ 3 ps)8, 9. Low bandgap materials have been employed to bring 2D TMD photodetection technology to the NIR/SWIR spectral regime and, especially, the telecom-band photodetection, such as 1310 nm and 1550 nm etc., has attracted great attention for the extensive application in LiDAR and optical telecommunication7. However, IR photodetection with TMDs remains impractical at present owing to issues with responsivity and response speed. Sub-bandgap detection is feasible in MoS2 with highly defective samples10 or with photothermal effects11. However, this is accompanied by a decrease in the rate of photodetection response to several hundred milliseconds or even slower12, 13 due to the relatively slow processes of defect-trapped carrier dynamics and thermal-assisted photoconductivity changes. Photodetection based on hot electron transfer from the metal to the semiconductor is an alternative way to realize sub-bandgap photodetection without sacrificing response speed. However, hot carrier transfer efficiency depends strongly on the energy of light14, 15, thus the hot electron photodetectors based on MoS2 at 1,550 nm have rarely been studied due to the negligible photocurrent. As such, there is considerable interest in expanding the photodetection wavelength to the sub-bandgap region and simultaneously achieving high responsivity and fast response speed for NIR/SWIR photodetectors.
Metal contact is an essential part of modern electronic and optoelectronic devices and has attracted great interest in 2D material research16, 17, 18, 19. Metal contact not only affects device performance electrically via the Schottky junction20, but can also impact performance as a consequence of changing optical properties due to plasmonic effects or the creation of a Fabry–Pérot (F–P) cavity. By reducing the lateral size of an electrode to the sub-micron scale to match the plasmon frequency, one can achieve increased sub-bandgap absorption, resulting in higher photoresponsivity in the sub-bandgap regime due to hot electron transfer21. Moreover, by modulating the 2D material thickness to match the F-P-cavity-like resonance, it becomes possible to realize a considerable increase in absorption and external quantum efficiency22, 23. However, the interface condition-dependent optical properties of 2D materials are not well-studied, and these may be strongly dependent on the electrode fabrication method. The interface between the deposited metal and 2D material is highly hybridized with many defect sites24, 25, which can affect the optical properties of 2D material in the sub-bandgap regime26. Therefore, the condition of this interface provides an additional knob to control the optical properties of the 2D heterostructure, and may thus offer a facile mechanism for sub-bandgap light detection.
Here, we demonstrate that the optical properties of MoS2/Au electrode heterostructures are greatly affected by these interface conditions, which are in turn sensitive to the electrode fabrication method. High IR absorption of up to 60% can be realized by using industrial metal electrode deposition methods, with strong light absorption observed over a broad wavelength range of 800–1,600 nm. The thickness of the MoS2 layer is critical because it determines the resonance frequency of the F–P-like cavity formed in the MoS2-Au heterostructure. We studied the NIR light-induced photocurrent generation mechanism using photocurrent mapping and determined that a MoS2 layer fabricated onto an Au electrode with the deposition method exhibited markedly higher photocurrent than transferred electrodes due to the strong IR absorption. Wavelengths of 1,310 nm and 1,550 nm can be detected when the light illuminates the MoS2/sputter-deposited Au (d-Au) electrode, with photoresponsivity of 35 mA/W at 1,310 nm and 17 mA/W at 1,550 nm and a fast response rate of 100–200 µs due to increased light absorption in the heterostructure. Our results highlight the critical role of interfacial engineering on optical absorption and photocurrent generation, and present MoS2 as a stable and promising material for telecom-band photodetection.
Sub-bandgap light absorption in MoS2 via engineering of the electrode interface
The interfacial condition of the MoS2/Au electrode depends heavily on the Au electrode fabrication method20. Metal atoms or clusters with high kinetic energy can collide with the MoS2 lattice at high speed during the metal deposition process and form covalent bonds with MoS2 or generate defects in the MoS2 lattice. This results in the formation of a hybridized interlayer at the MoS2/Au interface, as shown in Fig. 1a24, 25. From an electrical point of view, the hybridization causes a change in the Schottky barrier through Fermi-level pinning and determines the contact resistance and carrier transport behaviour20, 25. Additionally, optical properties such as absorption, which determines the number of carriers that can be generated in the material under illumination, are affected by the hybridization layer at the interface26. We studied the effect of various electrode fabrication methods on the absorption spectrum by fabricating three Au thin films on the same MoS2 flake using thermal deposition, sputter deposition, and electrode transfer methods, respectively (Fig. 1b). The samples being studied were fabricated using probe tip-assisted thin metal film transfer, and no adhesive polymer or liquid was used during the entire sample fabrication process. The detailed procedure is discussed in the Supplementary Information (Fig. S1). An ultra-flat interface without hybridization can be expected with the Au transfer method. However, hybridization of MoS2 and Au might occur when the interface is fabricated via deposition. We observed that the absorption spectrum in the NIR range differed greatly across the three MoS2/Au heterostructures (Fig. 1c). However, we saw similar absorption at the A, B, and C exciton peaks. We observed absorption of 70% in the sub-bandgap wavelength with the sputter-fabricated heterostructure, which was considerably higher than that of their transferred (24%) or thermal-deposited (50%) counterparts. This indicated that the electrode fabrication method affected the absorption properties, and that the increased sub-bandgap absorption observed in the sputter-fabricated system might be attributable to the hybridization layer at the MoS2/Au interface. The deposition rate during the sputter deposition process was 10 nm/min, which is much faster than thermal deposition (0.6 nm/min). This also greatly affects the interface properties, introducing appreciable mid-gap states in the TMD.
We verified our conjecture by using high-resolution transmission electron microscopy (HRTEM) to measure the cross-section of MoS2 with different Au thin film interfaces (Fig. 1d–f). The transferred Au (t-Au)/MoS2 heterostructure (Fig. 1d) exhibited a complete MoS2 lattice under the Au electrode, with a 6 \(\AA\)distance between the Au and S atoms of MoS2. In contrast, the thermal- and sputter-deposited Au/MoS2 interfaces exhibited intimate contact between the Au and S atoms (Fig. 1e and 1f), indicating the presence of covalent bonds. There were a few blurry regions of the upper MoS2 lattice where the S atoms could not be observed in Fig. 1e and 1f compared with Fig. 1d. This demonstrated the presence of interfacial defect states in the MoS2, as widely discussed in previous work showing that metal deposition can induce defects in the lattice and generate mid-gap states24, 27. The extent of variation in the MoS2 lattice under the Au electrode in the sputter-deposited Au (d-Au) condition was relatively greater than that of its thermally-deposited counterpart, because whole-lattice distortion was observed in approximately four layers of MoS2. This indicated that additional defect states were generated at the MoS2/d-Au interface. Therefore, stronger sub-bandgap absorption was possible. These HRTEM images demonstrated that the electrode fabrication method meaningfully affects the MoS2 lattice at the interface, thereby influencing the electronic and optical properties of the resulting device.
Absorption spectrum modulation by controlling cavity resonance with MoS 2 thickness.
The hybridized layer increases NIR light absorption, and sub-bandgap light absorption is augmented if a hybridization layer is placed atop MoS2 as well. We fabricated such a sample by depositing an ultra-thin layer of Au onto the MoS2/Au heterostructure. This ultra-thin Au layer had a negligible effect on the absorption spectrum, as can be observed via optical microscopy (Fig. 2a, inset; Fig. S2). Absorption in the sub-bandgap wavelength increased to approximately 70% with the addition of this ultra-thin Au layer (Fig. 2a). Furthermore, the absorption spectrum of a sample (Fig. S3) that had a similar configuration of layers but with a van der Waals interface rather than the hybridization interface of MoS2/Au exhibited a low absorption peak of 28%, offering further evidence that the hybridization layer is critical for sub-bandgap absorption. Using atomic force microscopy (AFM), we measured the thickness of the ultrathin Au film to be 2.3 nm (Fig. S4), and confirmed the ultra-flat surface of this film without nanoparticle formation, eliminating the possibility of plasmonic absorption.
We further investigated the thickness-dependent absorption spectrum in the sub-bandgap regime. We constructed six MoS2 layers with different thicknesses ranging from 33–63 nm (Fig. S5), which demonstrated broad absorption in the NIR range, with an absorption peak redshift associated with increasing thickness of the MoS2 layer (Fig. 2b). We observed NIR absorption > 60% for all six samples, where the center of the peak wavelength linearly shifted in proportion to the increase in thickness of the MoS2 layer (Fig. 2c). These results were in good agreement with the F–P-cavity-like features of this design, as calculated using the transfer matrix method (Fig. 2c, red line). This indicated that the sub-bandgap absorption peak could be modulated, and that this spectrum modulation was achieved by the MoS2 thickness. The high refractive index of MoS2 enabled the formation of an F–P-like cavity with a thickness on the order of 100 nm. This anomalously strong NIR absorption can be explained through two mechanisms. In the first mechanism (Fig. 2d), sub-bandgap absorption occurs at the hybridized MoS2/Au interface 19. This hybridized layer is highly n-type doped, which can be deduced from the transfer curve of the MoS2 field-effect transistor (FET) redshift after deposition of the ultra-thin Au layer (Fig. S6). Hence, the mid-gap state was occupied with electrons, and these electrons can be excited to the conduction band of MoS2 by NIR light excitation. In the second mechanism, an F–P-like cavity was formed in the MoS2/Au heterostructure (Fig. 2e) 22, maximizing the sub-bandgap in the NIR wavelength range due to the occurrence of destructive interference. The thickness of MoS2 should be in the range of 20–80 nm to generate destructive interference in the range of 800–1,600 nm, as observed from the simulation results shown in Fig. 2c. Anomalously strong NIR light absorption in the broadband wavelength range was made possible within the hybridized MoS2/Au layer owing to these two mechanisms.
We propose that this electrode hybridization-enhanced NIR absorption augmentation is common for other 2D materials and metal configurations29,36,37, because it originates from mid-gap-state absorption in the TMD/metal hybridization layer and most metal deposition inevitably induces Fermi-level pinning28 and generates mid-gap states in the interface. However, the levels of hybridization and absorption intensity are different, and this may depend upon the metal thin film absorption and mid-gap state formed at the interfacial layer. For example, we fabricated Cu and Pt thin films with a thickness of 2 nm via sputter deposition onto MoS2-Au and observed that both films exhibited absorption peaks in the NIR wavelength range but with different absorption values (Fig. S7). Further investigation is therefore needed to clarify the influence of different metal species on sub-bandgap absorption.
Mechanism of photocurrent generation via sub-bandgap energy photons
Strong NIR absorption was feasible in the MoS2 film through standard Au electrode fabrication methods. In this way, it is possible to generate carriers by NIR light excitation, which can be used for NIR photodetection. We fabricated a sample to explore the photocurrent generation mechanism at the MoS2/Au deposited electrode interface (Fig. 3a). The MoS2 layer was exfoliated onto a glass substrate, after which two Au electrodes were fabricated using sputter deposition and thin-film transfer methods. We used optical microscopy and photocurrent mapping from the glass side to enable illumination of the MoS2 layer on top of the Au electrodes (Fig. 3b). Photocurrent mapping was performed by recording the source-drain current (Ids) during a focused 808-nm laser beam scan across the sample, such that the light-induced current difference (i.e., the photocurrent) could be observed from the mapping data. The resulting I–V curve of the sample exhibited an asymmetric shape (Fig. S8) due to the built-in junction formed in the lateral direction of the MoS2 layer and different Schottky junctions at the Au electrodes and MoS2 interface. We measured a bias voltage of 0 V, demonstrating that the photocurrent can only be generated in the MoS2 layer on top of the Au electrodes (Fig. 3c, left panel). Photocurrent was only generated at the electrode edge region on the d-Au side, although we observed a more uniform photocurrent in the overall MoS2 on the t-Au electrode side. The photocurrent also had an opposite direction for the two electrodes, indicating that the direction of electron flow was consistently from Au to MoS2 in both cases. The mechanism of sub-bandgap photocurrent can lead to hot electron transfer and sub-bandgap absorption, as shown in Fig. 3d and 3e. Therefore, it is possible to generate electrons in the conduction band of MoS2 by sub-bandgap excitation via these processes. The electrons can be driven by the built-in potential formed in the MoS2/Au contact area. It can be observed from the line profile of the left panel of Fig. 3c, that the maximum photocurrent with d-Au was slightly higher than that with t-Au. However, the d-Au electrode exhibits approximately three-fold higher absorption (Fig. S9). This is because of the asymmetric built-in potential formed in the contact region due to the asymmetric electrode; the junction in the MoS2/t-Au layer was higher than that in the MoS2/d-Au layer (Fig. 3f left panel). This results in decreased quantum efficiency of carriers generated in MoS2/d-Au even though greater carriers were generated by NIR light absorption.
We applied a bias voltage of − 1 and + 1 V to the device to eliminate the effect of the built-in potential, and the photocurrent mapping results are shown in middle and right panel of Fig. 3c. The photocurrent greatly increased (by a factor of 10) after applying the bias voltage compared with that at 0 V, and only appeared on one side of the electrode. The MoS2/d-Au layer generated a photocurrent with a maximum value of 750 nA when Vds = 1 V, whereas a photocurrent of 400 nA was generated under a voltage bias of − 1 V on the t-Au side. The magnitude of the electric field was similar under biases of − 1 and + 1 V. The photocurrent exhibited a similar value for t-Au and d-Au under laser excitation at 455 nm (Fig. S10). However, the sub-bandgap photocurrent generation area differed considerably. Photocurrent can be generated in the entire region of MoS2 on top of Au, even at a 20-µm distance from the edge. However, photocurrent was generated only at the edge area on the d-Au side, with a width of 7 µm. The maximum photocurrent was observed at the edge of the electrode, and exponentially decreased with distance. This can be attributed to the short electron lifetime at the hybridized interface. A two-fold higher photocurrent was observed on the d-Au side relative to the t-Au side due to the larger carrier density in MoS2 on d-Au resulting from sub-bandgap absorption (middle and right panel of Fig. 3f; only carriers attributed to photocurrent are shown). Hot electron transfer was inefficient in the longer-wavelength regime, and the photodetection properties greatly worsened as the wavelength increased14,17. However, photocarrier generation was unaffected by the wavelength in the sub-bandgap absorption. Therefore, long-wavelength NIR detection by the deposited electrodes might be relatively efficient.
Photocurrent generation in MoS2 at telecommunication wavelength excitation.
We subsequently demonstrated that it is able to employ the MoS2/d-Au heterostructure to generate carriers in the conduction band of MoS2 using light wavelengths of 1,310 and 1,550 nm, thereby producing photocurrent. In this experiment, we used a thicker MoS2 to realize high absorption in the telecom-band wavelength (Fig. S11). Since wavelengths of 1,310 and 1,550 nm cannot be detected by charge-coupled devices, we configured a custom optical setup to generate NIR illumination at a precisely targeted spot under the optical microscope (Fig. S12). We observed a notable increase in current in the plus bias voltage (Vds from 0.5 to 2 V) when NIR light was projected onto the MoS2/d-Au (Fig. 4a). In contrast, we observed a negligible difference in the minus bias voltage (Vds from − 0.5 to -2 V), which is in close agreement with our previous photocurrent mapping results (Fig. 3g). Photocurrents of 20 and 35 nA were generated under excitation at 1,550 and 1,310 nm under a bias voltage of 2 V (Fig. 4b). For MoS2/t-Au, in contrast, we observed a similar I–V curve whether the sample was excited at 1,310 nm, 1,550 nm, or under dark conditions (Fig. 4c), with a significantly lower photocurrent in the time-dependent response (Fig. 4d). We assessed photoresponse speeds for our MoS2/d-Au sample at 1,310 and 1,550 nm by switching the illumination on and off using a mechanical chopper at a frequency of 1 kHz (Fig. 4e). We observed a fast response speed in the range of 100–200 µs at both wavelengths. The light power-dependent photocurrent for this sample at both wavelengths is measured and plotted in Fig. 4f. In this plot, \({I}_{ph}={P}^{\alpha }\), where \(\alpha\) evaluate the gain-induced photocurrent, was 1.07 and 1.1 for 1,310 and 1,550 nm, respectively, which indicated that photocurrent have linear dependence on the light power density and there was no extra gain during the photocurrent generation process29. Therefore, this system exhibits the potential for fast photoresponse.
We measured the time-dependent photoresponse of this system by illuminating the entire sample with wavelengths of 1,310 and 1,550 nm and a light power density of 11.6 mW/cm2 (Fig. S13). We saw a fast photoresponse (< 1 ms) followed by a slow response of several seconds in the photocurrent upon illumination at 1,310 and 1,550 nm. It can be deduced that the slow photoresponse was due to the bolometric effect13, whereas the fast response was due to sub-bandgap absorption (Fig. 4e). The device responsivity is calculated with \({R=I}_{ph}/P\), where Iph is the photocurrent and P is the power of incident light. For the fast response, responsivity was 17 and 35 mA/W at 1,550 and 1,310 nm, respectively. Finally, we compared the performance of NIR photodetectors fabricated from TMD materials via different methods. Our MoS2/d-Au system enabled NIR light detection at wavelengths of up to ~ 1,550 nm, substantially outperforming other designs, with sub-bandgap detection in the wavelength range of 800–1,100 nm (Fig. 4g). This design also exhibited a far faster photoresponse (Fig. 4h), while achieving finite responsivity. This exceptional device performance was due to the strong light absorption at the MoS2/Au interface and the low recombination rate at the MoS2 channel.
