Hot Electron Dynamics in MoS2/Pt Van Der Waals Electrode Interface for Self‐Powered Hot Electron Photodetection

Hot electron photodetection provides a powerful platform for photosensing beyond the bandgap of a semiconductor. High‐performing hot electron photodetection has been reported in 2D transition metal dichalcogenide material‐based devices without the support of plasmonic metal nanostructures but with planar metal electrodes. However, the mechanism driving hot electron dynamics in 2D transition metal dichalcogenide devices has not been explored. Here, we uncover the hot electron transfer in MoS2 and Pt van der Waals (vdW) metal electrodes by transient reflection spectroscopy, revealing a sub‐picosecond transfer time of hot electron and a decelerated recombination process in MoS2 at the below bandgap photoexcitation compared to the pristine MoS2. With an independent photocurrent mapping, the ultralong diffusion is revealed in MoS2/vdW metal electrode and a self‐powered near‐infrared (NIR) photodetector is demonstrated with a high responsivity of 6 mA W−1 and detectivity of 9 × 109 Jones at a wavelength of 1062 nm by integrating Pt and Ag asymmetric vdW electrodes into MoS2. The results will pave the way for the next generation of hot‐electron‐based self‐powered optoelectronic devices.

enabled through hot electron transfer from vdW metal electrode by the sub-bandgap photoexcitation. The hot electron transfer time is revealed to be 143 fs and the lifetime of the transferred electron in MoS 2 conduction band is longer than 800 ps. MoS 2 and the two vdW metal electrodes (i.e., Pt and Ag) with large work function differences were integrated by the van der Waals method to construct a photovoltaic device and exploit the hot electrons. We demonstrated exceptional self-powered NIR photodetector working at 1062 nm, performing responsivity of 6 mA W −1 . The hot electron dynamics at the MoS 2 /Pt vdW electrode interface provide new insights into the design of hot electron photo detector and boost the further development of NIR photo detection technologies.

Hot Electron Dynamics in MoS 2 /Pt Van der Waals Electrode
We performed TR spectroscopy to investigate the hot electron transfer from Pt to MoS 2 and the electron decay process in MoS 2 . The detailed sample information is provided in Figure S1 in the Supporting Information. TR spectroscopy can reveal the excited carrier dynamics in a semiconductor material by generating excited carriers in the conduction and valence bands with a pump beam, subsequently revealing the excited carrierinduced reflection change (∆R) of a material with a broadband white light probe beam at different time delay ∆T. Thus, it is possible to reveal the pump induced carrier density change in MoS 2 conduction band and valence band within femtosecond time range (Figure 1a,b). First, the time-dependent ∆R signal of the pristine MoS 2 was measured at an excitation wavelength of 450 nm (2.75 eV). The photobleaching (PB) signal and photoinduced absorption (PIA) signal of MoS 2 A and B excitons can be observed from the ∆R mapping data, which arise from the carrier density increase in the A and B excitons by the pump beam [26] (Figure 1c). From the spectrum data (Figure 1d), we confirmed that the PB signal from the ∆R data (≈660 and 610 nm for PB signals of A and B excitons, respectively) agrees well with the reflection spectrum in the top panel. The PB and PIA kinetics for the A exciton exhibited the same decay behavior (Figure 1e). Decay time constants of 25 and 142 ps were obtained using the biexponential function, which are attributed to the defect trapping and exciton recombination processes, respectively, as shown in Figure 1b.
At an excitation wavelength of 800 nm (1.55 eV), we further explored the sub-bandgap light-excited carrier dynamics in MoS 2 ( Figure 1f) and the MoS 2 /Pt interface (Figure 1i). An optical microscopy (OM) image and configuration of the measured sample are shown in Figure S 1a,b in the Supporting Information. The Pt vdW metal electrode was fabricated and transferred via a probe-tip-assisted metal film transfer method, [27] and the detailed fabrication process is discussed in the Experimental Section. In the pristine MoS 2 , no PB and PIA signals for the A exciton were observed (Figure 1g), owing to the absence of a band-filling effect at an excitation wavelength of 800 nm. However, in the MoS 2 /Pt heterostructure, the PIA signal is clearly observed, which is attributed to the band-filling of the A exciton band of MoS 2 by hot electron transfer from Pt to MoS 2 ( Figure 1j). The electron density in the A exciton state of MoS 2 increases via hot electron transfer with 800 nm pump excitation, leading to the unusual band filling of the PB signal at 660 nm and the PIA signal at 685 nm for the A exciton. Note that the ∆R intensity in PB for A exciton and PB and PIA for B exciton, are reduced compared to the direct band transition with high energy pump in pristine MoS 2 (Figure 1c) because the hot electron transfer is less efficient compare to the direct band transition. [28] Moreover, the ultrafast carrier rise time of the A exciton was observed to be within 143 fs (Figure 1h), which is comparable with the hot carrier transfer speed from the metal to semiconductor obtained with time-resolved photoemission electron microscopy. [29] The decay process of the carriers transferred from Pt to MoS 2 was studied using the kinetics of the PIA signal ( Figure 1k). The carrier lifetime was prolonged compared with that of the pristine sample excited by high pump energy (Figure 1e). The ∆R signal remained almost consistent until 100 ps and slowly decayed until 800 ps to 60% of its maximum signal. The electron lifetime is much slower than the τ 2 (Figure 1e), which is the nonradiative recombination time of the interband transition generated carriers. We can infer that two decay processes, electron trapping to the defect states and nonradiative recombination, are suppressed. The possible scenarios are discussed as follows. (i) The integration of the vdW metal electrode into the MoS 2 layer does not introduce any defect sites in the MoS 2 lattice, so MoS 2 on top of Pt remains of high quality after electrode fabrication. In comparison, the deposited Pt/ MoS 2 contact shows that the carrier lifetime is greatly decreased by 1.3 ps, owing to the strong trap state produced during the deposition process ( Figure S2, Supporting Information). (ii) Pristine MoS 2 is typically an n-type material, and the dominant defect formed in MoS 2 is the hole trapping site. [30] Therefore, carriers excited at 450-nm pump excitation rapidly decay to the hole-trapping sites, [31] whereas hole trapping does not occur in MoS 2 , and the rapid decay component disappears at 800 nm excitation. (iii) The electrons transferred from Pt to MoS 2 remain as free electrons in MoS 2 , whereas the photoexcited carriers in pristine MoS 2 exist as strongly bound excitons. Recombination is discouraged by the charge of non-neutrality, making it challenging to find a hole to recombine. An anomalously long carrier lifetime can provide many opportunities for utilizing hot carriers, from light energy harvesting to NIR photodetection applications.

Photovoltaic Device Fabricated by the Asymmetric Van der Waals Contact
We demonstrated an NIR photodetector in the self-powered mode. Figure 2a shows the device configuration and band alignment. The top panel shows a cross-sectional schematic of the device ( Figure S3, Supporting Information). The device consisted of a mica substrate, MoS 2, and two asymmetric vdW metal electrodes, Pt and Ag. Considering the band alignment, the opposite Fermi level shift on the MoS 2 side is expected because of the large work function difference between the two vdW electrodes. Indeed, after contact between Pt and MoS 2 , electrons were transferred from MoS 2 to Pt. The Fermi level www.advmatinterfaces.de downshifts due to the low electron density at MoS 2 , inducing p-type MoS 2 at the interface (left panel of Figure 2a). In contrast, following contact, electrons in Ag are transferred to MoS 2, and the electron density in MoS 2 increases, resulting in more N-doped MoS 2 at the interface (right panel of Figure 2a). Therefore, a P-N junction was created in the lateral direction  Figure 2a). We observed that the van der Waals contact with nanometer scale flatness ( Figure S4, Supporting Information) plays an important role in the construction of the P-N junctions. [32] No organic residue contaminants exist at the interface between the metal electrode and MoS 2 . Consequently, the optimal van der Waals interface between the MoS 2 and the Pt (and Ag) electrode can be formed, leading to efficient charge transfer. Moreover, due to the absence of chemical bonds, a high Schottky barrier and a strong built-in junction in the channel can be expected. The Pt and Ag electrodes provide a large built-in potential due to the p-and n-doping effects on MoS 2 , respectively. Using the probe-tip-assisted metal film transfer method, the 120-nm-thick Au layer serves solely to transfer the Pt (or Ag) film to the top of the MoS 2 film. We further conducted the Kelvin probe force microscopy (KPFM) of MoS 2 contacted with Ag and Pt electrodes ( Figure S5, Supporting Information). The large surface potential difference of 320 mV between MoS 2 on Ag and MoS 2 on Pt demonstrating a strong P-N junction was generated in the channel region.
The optical microscopy (OM) image and output curve of the as-fabricated device are shown in Figure 2b,c, respectively. The OM images and photocurrent data were measured from the transparent mica substrate side. Benefiting from the P-N junction formed in MoS 2 , robust rectifying behavior with a forward/reverse rectification ratio of 10 3 was observed. An ideality factor of n = 1.1 was also achieved. Under white-light illumination, an open-circuit voltage of 120 mV and a large photocurrent of 4.7 nA at a bias voltage of 0 V were measured, which is comparable to a previous report. [33] Furthermore, the on/off ratio of the device at a bias voltage of 0.5 V, which reaches up to 10 2 , ascribed to the low current density under the reverse bias condition and indicates the formation of a strong built-in potential in the device. The excellent photovoltaic behavior of the device suggests the possibility of self-powered photodetection.

Self-Powered NIR Photodetection
The photoresponse properties of the photovoltaic device were investigated under a laser illumination of 1062 nm. In the output curve (Figure 3a), a significant spontaneous photocurrent was obtained, especially when the external bias was zero. At zero bias voltage, no current flowed in the dark condition, whereas a photocurrent of 10 pA or higher was produced with 1062 nm laser illumination (2 mW cm −2 ). Then, we examined the self-powered photodetection performance with timedependent photocurrent measurement (Figure 3b), observing a stable photocurrent of ≈20 pA under a bias voltage of 0 V and light illumination of 1062 nm at a fluence of 2 mW cm −2 . An on/off ratio of the photocurrent up to 10 3 was achieved because of the low dark current in the self-power detection mode. The rapid photocurrent response of 10 µs was demonstrated (Figure 3c), attributed to the strong built-in electric field and negligible photoconductive gating effect. [34] To further investigate the photocurrent generation mechanism from our device, the photocurrent mapping was conducted with a 1062-nm laser under a bias voltage of 0 V (Figure 3d,e). It demonstrates the MoS 2 on top of Pt electrode is the main region for creating photocurrent in the device. Due to negligible absorption of the MoS 2 under the sub-bandgap region, [35] it is difficult to have considerable photocurrent by photoinduced electron-hole generation. [36] However, the hot electron transferred from Pt to MoS 2 conduction band are able to contribute to the photocurrent in our devices, to generate efficient photocurrent under the 1062 nm wavelength. Strikingly, even when the focused light with a laser spot size of 2 µm illuminates the channel region, the photocurrent is generated across a distance of 18 µm, from the channel in the whole MoS 2 area on top of the Pt electrode. We could set the effective area of the device as the region of MoS 2 on top of Pt. The photoresponsivity of 6 mA W −1 and the detectivity of 9 × 10 9 Jones are achieved, which can be calculated using R = I ph /P in , and D* = RA 1/2 /(2eI dark ) 1/2 , respectively, where R is the responsivity, I ph is the photocurrent, P in is the incident light power, A is the effective area, e is the electron charge, and I dark is the dark current. The external quantum efficiency of 0.7% can be obtained using EQE = hcR/eλ, where h is the Planck's constant, c is the speed of light, and λ is the wavelength of light. Need to mention that the hot carrier transfer also may happen in the Ag/MoS 2 interface. However, the absorption of Ag is much lower than Pt and the built-in potential in the channel region blocks the electron flow, thereby demonstrating negligible photocurrent. For step II in Figure 4, it is abnormal for 2D semiconductor materials to have a long carrier diffusion length compared to the result from transient absorption measurement, in which diffusion length is ≈1-2 µm. [37] The relatively long carrier diffusion length can be attributed to the prolonged lifetime of electrons transferred. Moreover, electrons in our device exist as unbound free carriers, which are different from the strongly bound excitons measured in the previous study. The superior diffusion length of the carrier has also been reported in other materials, such as perovskite and silicon, which are measured through conductive atomic force microscopy or transient

www.advmatinterfaces.de
scanning electron microscopy. [38,39] The small exciton binding energy, long carrier lifetime, and high diffusion coefficient enabled the superior diffusion length in the above results. We speculated that two processes were responsible for the increased diffusion coefficient of transferred hot electrons. (I) The hot electron has high kinetic energy and can ballistically transport over 150 nm within 20 fs [40] . (II) After thermalization, the hot electron remains a free carrier instead of undergoing exciton formation, and thus can move freely at a high speed without the restraint of the strong binding energy. [41]

Conclusion
The hot electron dynamics at the MoS 2 /van der Waals Pt metal electrode interface were observed using transient reflection spectroscopy. Under sub-bandgap light excitation, carrier generation of the A exciton band in MoS 2 is observed owing to hot electron transfer from Pt to the MoS 2 conduction band. The ultrafast hot electron transfer time was 143 fs. A prolonged electron lifetime (>800 ps) was obtained in the MoS 2 region on top of Pt, which is longer than the interband transition-generated photocarriers in pristine MoS 2 . A high-performance photovoltaic device based on MoS 2 was fabricated using asymmetric vdW electrodes, Ag/Au, and Pt/Au. A high rectification ratio of 10 3 was obtained because of the strong built-in potential in the lateral direction resulting from the n-and p-doping of MoS 2 by the electrodes. The sub-bandgap photocurrent response was examined using a 1062-nm laser light, which was illuminated from the transparent electrode side under a bias voltage of 0 V. A photoresponsivity of 6 mA W −1 and detectivity of 9 × 10 9 Jones were achieved in the self-power operating mode. Photocurrent mapping indicates that MoS 2 on the Pt electrode is an effective area for photocurrent generation. The entire MoS 2 layer on top of Pt generated the photocurrent, even when the focused light was illuminated at a position with a distance of 18 µm from the electrode edge. Our results pave the way for hot electron-based self-powered optoelectronic applications which require lightweight and low-power consumption, especially for the nanoand biotechnologies.

Experimental Section
Van der Waals Electrode Integration into MoS 2 : Both the probe-tipassisted thin metal film transfer and template strip methods were used to fabricate vdW electrodes on MoS 2 ( Figure S3, Supporting Information). [27] The Ag/Au film was deposited by thermal evaporation on a SiO 2 /Si substrate with thicknesses of 10 nm for Ag and 120 nm for Au (Figure 2a). For the Pt/Au film, Pt was deposited by electron beam evaporation with a thickness of 10 nm and Au was subsequently deposited with a thickness of 120 nm on top. The film can be easily torn off from the substrate by pushing one edge of the metal film because the metal films of both Ag/Au and Pt/Au cut by a probe tip (ST-10-1, GGB) are very lightly stuck on the SiO 2 /Si substrate ( Figure S3b, Supporting Information). Then, the film on top of the designated location was transferred with the XYZ translation stage under an optical microscope at the ambient condition. The Ag/Au and Pt/Au thin films were transferred onto the MoS 2 bulk flake with the thickness of 20-40 nm recognized by the color. The roughness of the exfoliated metal film was determined by the SiO 2 /Si substrate with an RMS roughness of ≈200-300 pm, which was ≈1.2 nm for Au and 0.2 nm for Pt ( Figure S4, Supporting Information), forming an ideal vdW interface with MoS 2 . The top Au with a thickness of 120 nm was used to protect the bottom contact part during the mechanical transfer process.
Device Characterization: The output curve and time-dependent photocurrent were measured using a source-meter unit (Keithley www.advmatinterfaces.de 2636 B). For the NIR light detection measurement, a 1062-nm laser (Lambda Beam, RGB photonics) was used with a power density of 2 mW cm −2 . The roughness of the Pt and Ag surfaces was determined using atomic force microscopy (AFM 5000 II). Moreover, the photocurrent mapping data were measured using a photocurrent mapping system (XPER PC, NANOBASE) by steering the beam with the galvo mirror set, and the spontaneous photocurrent was measured in the AC mode using the current preamplifier (SR570) and lock-in amplifier (SR830).
Transient Reflection Measurements: The Ti:sapphire amplifier (LegendElite, Coherent) operating at a center wavelength of 750 nm, repetition rate of 1 kHz, and pulse duration of 25 fs was divided into two beams. One operated an optical parametric amplifier (TOPAS Prime, Light Conversion), which was used as the pump beam. The other focused on a nonlinear crystal and generated a white light continuum with a wavelength range of 500-800 nm, which was used as the probe beam. Pump wavelengths of 450 and 800 nm were used to generate photoexcited carriers above and below the bandgap of MoS 2 . The energy per pulse was 10 and 5 pJ, respectively. The 50× objective lens focused the pump and probe beams onto the sample and collected the reflected probe beam. The differential absorption was measured using a TR spectrometer (Helios, Ultrafast Systems).

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.