High‐Performance and Lithography‐Free Au/WS2/Ag Vertical Schottky Junction Solar Cells

Due to their extraordinarily large optical absorption coefficients, transition metal dichalcogenides (TMDs) are gaining more and more attention for photovoltaic applications. Improving the device performance of a TMD solar cell requires an optimal device architecture and reliable fabrication processes. Metal/WS2‐multilayer/metal heterojunctions are fabricated using lithography‐free processes. 20 nm thick WS2 flakes are exfoliated on template‐stripped Ag bottom electrodes, and then 10 nm thick Au top electrodes with a diameter of 2 µm are evaporated on the WS2 surface using holey carbon films as shadow masks. Current‐sensing atomic force microscope measurements reveal that the Au/WS2/Ag devices exhibit prominent rectifying characteristics, indicating the formation of Schottky diodes. The power conversion efficiency of the Schottky junction is as high as 5.0%, when illuminated by a light‐emitting diode with a peak wavelength of 625 nm and a power density of 2.5 mW cm−2. These devices also possess broadband and incident‐angle‐insensitive absorption capability due to the very large refractive indices and extremely small thickness of the WS2 flakes. The simple fabrication procedures proposed in this work demonstrate high‐performance and high‐yield TMD photovoltaic devices.

DOI: 10.1002/admi.202300031 mobility of TMDs have prompted an intensive study to develop various optoelectronic devices, such as quantum emitters, [3] light-emitting devices (LEDs), [4] photodetectors, [5,6] and solar cells. [7][8][9][10][11][12][13][14][15] Photocurrent and output voltage must be increased to improve the power conversion efficiency (PCE) of a solar cell. Therefore, the essential parameters for evaluating the performance of a solar cell are the short-circuit current density (J SC ) and the open-circuit voltage (V OC ), which indicate the maximum current density and output voltage under light illumination, respectively.
Numerous studies have concentrated on the fabrication and characterization of TMD-based solar cells. [7][8][9][10][11][12][13][14][15] TMDs have extraordinarily high absorption coefficients; hence, the photocurrent anticipated from the optical absorption in TMD layers is substantially greater than that from conventional SC thin films of the same thickness. [1][2][3]7] Furthermore, the superior mechanical strength of TMDs motivates intensive research efforts to develop semitransparent and flexible photovoltaic devices using thin TMD layers. [12,15] However, the measured J SC of TMD-based solar cells is considerably lower than the value predicted from the optical absorption in the active layers. [8][9][10][11][12][13][14][15] Even though TMD heterostructures are composed of identical materials with similar thicknesses, the V OC and J SC values reported in the literature vary significantly. [9] In most cases, TMD heterostructures are fabricated by stacking randomly shaped, small-sized exfoliated flakes. Subsequent electrode preparation with electron-beam lithography and lift-off is time-consuming and labor-intensive. Chemicals and solutions used in these processes may contaminate TMD flakes. As a result, the overall yield of the device is very low. The lack of reliable accumulated data from nearly identical devices makes optimizing materials and device architectures extremely difficult. Attaching electrodes to the active TMD layers often produces quasi-vertical heterojunctions in which electric current flows not only perpendicular to the heterointerfaces but also laterally across the thin TMD layers. [9] This lateral transport through the thin TMD layers may limit the collection of photogenerated carriers and increase the series resistance of the solar cell. [8][9][10] This suggests that a combination of optimal device architecture and reliable fabrication Due to their extraordinarily large optical absorption coefficients, transition metal dichalcogenides (TMDs) are gaining more and more attention for photovoltaic applications. Improving the device performance of a TMD solar cell requires an optimal device architecture and reliable fabrication processes. Metal/WS 2 -multilayer/metal heterojunctions are fabricated using lithography-free processes. 20 nm thick WS 2 flakes are exfoliated on template-stripped Ag bottom electrodes, and then 10 nm thick Au top electrodes with a diameter of 2 µm are evaporated on the WS 2 surface using holey carbon films as shadow masks. Current-sensing atomic force microscope measurements reveal that the Au/WS 2 /Ag devices exhibit prominent rectifying characteristics, indicating the formation of Schottky diodes. The power conversion efficiency of the Schottky junction is as high as 5.0%, when illuminated by a light-emitting diode with a peak wavelength of 625 nm and a power density of 2.5 mW cm −2 . These devices also possess broadband and incident-angle-insensitive absorption capability due to the very large refractive indices and extremely small thickness of the WS 2 flakes. The simple fabrication procedures proposed in this work demonstrate high-performance and high-yield TMD photovoltaic devices.
For solar cell applications, researchers are interested in metal/TMD Schottky junctions as well as TMD-TMD heterojunctions. [12][13][14][15] Recent research has shown that vertical transport in TMD Schottky junction solar cells enables a very efficient collection of photogenerated carriers. [14,15] For the fabrication of TMD Schottky devices, thick multilayers enhance optical absorption in the active TMD region. [12][13][14][15] Furthermore, substantial tunneling current through the sub-nanometerthick TMD monolayers may lower the measured PCE. [9] A high recombination rate at the metal/SC interface can reduce the PCE of a Schottky diode solar cell. Thus, continued theoretical and experimental studies are highly required to improve metal/ TMD contact characteristics. [16][17][18][19][20] Simple and reproducible fabrication processes are crucial for accumulating experimental data on the photovoltaic characteristics of TMD Schottky diodes.
Researchers recently developed metal-assisted exfoliation (MAE) technique. [21][22][23] The high binding energy at the TMD/metal interface enables high-yield delamination of TMD thin flakes. [21] The template-stripping method is utilized to prepare extremely flat and clean metal layers for MAE. [21][22][23] TMD/metal structures, prepared by the MAE and template-stripping techniques, can be used to fabricate electrical and optoelectronic devices, where the metal layers work as electrodes. Moreover, TMDs with a few tens of nanometers thickness can exhibit selfhybridized exciton-polaritons (EPs). [24][25][26] Huge exciton binding energies and high refractive indices of TMDs can lead to the formation of EPs at room temperature without external cavities. With the aid of EPs, the integration of TMDs with metal electrodes can modify the spectral response of TMDs and increase the optical absorption over broad wavelength ranges. [25,26] We should thoroughly investigate the solar cell application potential of the MAE-fabricated TMD/metal structures.
In this work, we fabricated Au/WS 2 /Ag vertical Schottky diodes and investigated their photovoltaic characteristics. WS 2 multilayer flakes and template-stripped Ag electrodes were integrated using the Ag-mediated MAE method. On the WS 2 surfaces, Au top electrodes (diameter: 2 µm) were evaporated using holey carbon films as shadow masks. The current-voltage characteristics of Au/WS 2 /Ag devices, obtained by a current-sensing atomic force microscope, exhibited clear rectifying behavior, indicating the formation of Schottky diodes. Under the illumination of LEDs with various wavelengths, the V OC and J SC of vertical devices were investigated. Optical measurements and simulations of the Au/WS 2 /Ag device were also performed to understand the photocurrent characteristics. All these results enable us to evaluate the photovoltaic performance of the Au/WS 2 /Ag vertical Schottky diodes fabricated by the newly proposed lithography-free processes.

Fabrication of Au/WS 2 /Ag Schottky Junctions
Au/WS 2 /Ag structures were fabricated using template-stripping and Ag-mediated exfoliation techniques. 200 nm thick Ag films were deposited on SiO 2 /Si wafers using an electronbeam evaporator with a base pressure of 10 −7 Torr. As shown in Figure 1a, a slide glass was attached to an Ag-coated SiO 2 /Si wafer with UV-curable epoxy (NOA63, Norland Products). Then the sample was exposed to radiation by a 395 nm flashlight (U2, Tattu) with a power density of 3.5 J cm −2 for 1 min. After the curing, the Ag thin films were peeled off from the SiO 2 /Si wafers. The surface roughness of the template-stripped Ag layer was as small as 1 nm. By employing blue tape (SPV224, Nitto) and a WS 2 bulk crystal (HQ Graphene), thin WS 2 flakes were exfoliated on the Ag layer to obtain the WS 2 /Ag samples ( Figure 1b). The Raman spectra of the samples were obtained in order to examine their interfacial properties ( Figure S1, Supporting Information). The MAE-exfoliated WS 2 flake was strained on the Ag layer, indicating that there was no serious concern about an oxide layer formation at the WS 2 /Ag interface in our sample. [23] Using holey carbon films with square arrays of holes (diameter: 2 µm and period: 4 µm) (CFT224C-50, EMS) as shadow masks during evaporation, 10 nm thick Au top electrodes were formed on the WS 2 surface ( Figure S2, Supporting Information). These simple lithography-free processes enable high-yield fabrication of Au/WS 2 /Ag vertical junction devices ( Figure 1c).

Optical Characterizations and Calculations
Optical reflectance spectra of flakes were measured using an optical microscope (OM) (LV100, Nikon) with a white LED light source (Solis-3C, Thorlabs). [26] The spectra from a chosen area of several µm 2 were obtained by collecting the reflected light from the sample surface using a 50 µm diameter optical fiber (M50L02S-A, Thorlabs) and a spectrometer (Maya 2000 Pro, Ocean Optics). The optical characteristics of the Au/WS 2 /Ag structures were simulated using COMSOL Multiphysics, a commercial software. The optical constants of Au and Ag [27] and WS 2 multilayers [28] were obtained from the literature. In the simulation model, the incident power density is fixed at 1 W m −2 Figure 1. Schematic illustrations of a) the template-stripping method to prepare ultraflat Ag layers, b) the Ag-mediated exfoliation process to obtain thin WS 2 flakes on the Ag layer, and c) an Au/WS 2 /Ag vertical heterojunction.
www.advmatinterfaces.de (0.1 mW cm −2 ) for simplicity. The use of Floquet periodic boundary conditions enabled the calculation of cases with various incident angles and polarizations.

Current-Sensing Atomic Force Microscopy Measurements
The surface morphology and the local current of the samples were simultaneously measured using a current-sensing atomic force microscopy (c-AFM) system (XE-100, Park Systems) in a glove box. [29] For transport measurements with nanoscopic spatial resolution, a wear-resistant highly doped diamond-coated tip with a spring constant of 0.5 N m −1 and a resonance frequency of 20 kHz (CDT-CONTR, Nanosensors) was employed. An infrared laser diode (wavelength: 830 nm) and a positionsensitive photodiode were used to detect the deflection of the cantilever. As photocurrent measurement light sources, blue (M470L5, Thorlabs), green (M530L4, Thorlabs), orange (M617L5, Thorlabs), and deep red (M680L4, Thorlabs) LEDs were utilized. The illumination intensity of the visible-light LEDs was adjusted using a USB-controlled constant current driver (UPLED, Thorlabs) and measured using a photodiode power sensor (S170C, Thorlabs). All the measurements were performed in the glove box purged with N 2 gas to avoid artifacts caused by ambient-gas adsorption. [29] Figure 2a consists of several regions with distinct colors. It is well known that the interference and absorption of incident light determine the apparent color of the flake. [24][25][26][29][30][31] The AFM measurements allow the estimation of the thickness of each region: the central region of the flake has a thickness of 20 nm, and the thickness of the Au topelectrode is 10 nm (Figure 2b). The diameter of the Au electrode was chosen to be 2 µm, smaller than the typical flake size. Consequently, many top electrodes can be prepared on the flakes with identical thicknesses. Electrical characterization of nearly identical devices allows us to extract key device performance factors, such as electrode material, electrode thickness, flake thickness, and so on.

Results and Discussion
For photovoltaic characterizations, 10 nm thick Au films were chosen as semitransparent top electrodes, whose sheet resistance could be as small as 50 Ω. [32] As shown in Figure 2a, the region containing the Au electrode has a similar color as the surrounding region. This indicates that the Au electrode does not significantly decrease the optical absorption in the WS 2 flake underneath it. Also, it should be noted that the Au electrode has a smooth surface, despite its very small thickness. Due to the poor wettability of Au to many materials, the growth of continuous Au thin films with a thickness ≤10 nm is extremely challenging. [33] As shown in Figure 2b, our Au electrodes on WS 2 flakes have a very smooth morphology in spite of their small thickness (10 nm). TMDs allow the growth of continuous flat Au thin films with metallic behaviors, even at thicknesses down to 3-4 nm. [33] Thus, the 10 nm Au films can be used as semitransparent electrodes for our devices.   illumination. The bottom Ag electrode was grounded to the metal sample disk using Ag paste, and the external bias voltage was applied to the Au top electrode via the conducting tip of the c-AFM system ( Figure S3a, Supporting Information). In a dark state, the device exhibits rectifying transport characteristics. Similar diode-like behaviors have been reported for metal/TMD-multilayer/metal devices. [12][13][14][15] It has been reported that the Schottky and Ohmic contacts can be formed at the Au/WS 2 and Ag/WS 2 interfaces, respectively, due to notable differences between the work functions of Au and Ag. [14,18] As written above, several devices can be fabricated in the same flake, and their transport characteristics are very similar, demonstrating the capability of our technique to fabricate high-yield and high-quality Schottky junctions ( Figure S3b,c, Supporting Information).
The exponential increase of J with respect to V is evident from the linear relationship between log|J| and V ( Figure S4a, Supporting Information). Based on the thermionic emission law, [12][13][14][15]34] the Schottky barrier height (SBH) and ideality factor (n) of the Au/WS 2 /Ag device were estimated to be 240 meV and 1.73, respectively ( Figure S4a, Supporting Information). Murali et al. estimated the SBH of 270 (or 290) meV at Au/WS 2 interface with WS 2 flakes with thicknesses of 40-60 nm, [19] which is very similar to the SBH estimated from our device. Due to series resistance, the current-voltage characteristics of diodes at high bias voltages often deviate from the exponential relationship between the current and voltage. In c-AFM measurements, the series resistance is caused by the current spreading from a nanoscale contact to a sample; hence, the tip radius affects the series resistance. [34] In our experiments, the top Au electrode (diameter: 2 µm) collects the current from the WS 2 flake. Consequently, log|J| increases linearly up to the maximum V in the measurement range (300 mV), suggesting that the series resistance is very small. The J-V curves exhibit no hysteresis when the applied voltage is swept in the forward and reverse directions.
Under illumination, the J-V curves are moved downward, indicating that our device possesses photovoltaic effects (Figure 3a). The orange LED light source (M617L5, Thorlabs) had a peak wavelength of 625 nm and a bandwidth of 16 nm. The incident angle of the LED was 60° and the optical power density (P opt ) was varied from 0.5 to 2.5 mW cm −2 . The shunt resistance (R sh ) estimated from light J-V characteristics is almost identical to that from dark state characteristics ( Figure S4b, Supporting Information). As increasing P opt , both the short-circuit current (J SC ) and open-circuit voltage (V OC ) increase (Figure 3a). At P opt = 2.5 mW cm −2 , we measure a V OC of 170 mV, a J SC of 1.45 mA cm −2 , a fill factor of 0.51, and a PCE of 5.0%. The PCE measured from our Schottky diode is comparable to a maximum single wavelength PCE of 3.4% from their vertical heterojunctions of graphene/WSe 2 (9 nm)/MoS 2 (3 nm)/Au under 740 W cm −2 of 633 nm laser illumination. [10] Compared with the p-n junction solar cell, the Schottky diode solar cell has a relatively small V OC , which is limited by the SBH. [12][13][14][15] It is well known that the Schottky contacts with thinner WS 2 flakes have larger SBH, owing to the smaller electron affinities of the thinner WS 2 . [2] (We observed lower current from the thinner WS 2 region, which could be attributed to the thicknessdependent SBH change, as shown in Figure S5, Supporting Information.) However, absorption loss significantly limits J SC and resulting PCE. The PCE of a Schottky solar cell with a few layers of TMD flakes was only 0.7%. [12] From the slope of the measured V OC versus ln(J SC ) plot obtained by the orange LED (Figure 3b), n can be estimated to be 1.12. [14] As comparison, the photovoltaic characteristics of our devices were also studied using a white LED ( Figure S6, Supporting Information). The estimated n from the white light illumination was 1.04. These n values are smaller than that extracted from the dark-state J-V plot (1.73) ( Figure S4a, Supporting Information). A similar light-induced reduction of n was reported in Au/MoSe 2 /Ag Schottky junction photodetector. [17] At present, a clear understanding of the discrepancy is lacking, and additional studies are needed to clarify the underlying mechanism. The measured J SC follows a power law as a function of P opt : J SC ∝ P opt α (Figure 3c). The estimated α was 0.82 less than 1, which could be originated from poor carrier collection. [8] The evaporation of metal electrodes in a vacuum may create defects at the metal/WS 2 interface, [16] thereby limiting the carrier collection. Figure 4a shows the wavelength-dependent light J-V characteristics of an Au/WS 2 (20 nm)/Ag device. As light sources, four different LEDs were used: blue (peak wavelength: 470 nm, bandwidth: 28 nm, M470L5, Thorlabs), green (peak wavelength: 522 nm, bandwidth: 35 nm, M530L4, Thorlabs), orange www.advmatinterfaces.de (peak wavelength: 625 nm, bandwidth: 16 nm, M617L5, Thorlabs), and deep red (peak wavelength: 680 nm, bandwidth: 22 nm, M680L4, Thorlabs). The beam spot was located at the same position on the sample for the different LEDs, since the fiber-coupled LED light was sent to the sample using an identical focusing lens. All the light J-V curves were obtained at P opt of 2.5 mW cm −2 by adjusting the current supplied to the LEDs. The illumination of the Au/WS 2 /Ag Schottky junctions using four different colored LEDs shifts their J-V curves downward, revealing the photovoltaic effects. The measured V OC of 170 mV and J SC of 1.45 mA cm −2 were the largest under the orange LED illumination. The electric output power (P el ) can be estimated from the light J-V plots, as shown in Figure 4b. The maximum P el values (corresponding PCE) are 0.13 (5.0%), 0.071 (2.8%), 0.059 (2.4%), and 0.035 (1.4%) mW cm −2 for the orange, green, blue, and deep red LED light sources, respectively. The average PCE for the four LEDs is 2.9%. The measured J SC and PCE of our devices well exceed those of Au/WS 2 /Ag Schottky diodes reported earlier. [14] The measured (Figure 5a) and calculated (Figure 5b) optical reflectance spectra of an Au(10 nm)/WS 2 (20 nm)/Ag(200 nm) device under normal incident light agree with each other. Optical simulations of our devices were performed using COMSOL Multiphysics ( Figure S7a, Supporting Information). In the reflectance spectra, notable dips appear at λ = 595 and 645 nm. These dip positions are different from the exciton resonance wavelengths of bulk WS 2 : 625 and 520 nm for the A and B excitons, respectively. [25,26,28] This suggests that the optical properties of our TMD multilayer flakes cannot be fully accounted for by the exciton-related characteristics alone. It has been reported that strong coupling between excitons and photons and subsequent EP formation can be observed in TMD multilayer flakes even without external cavities. [24][25][26] In the reflectance spectra of TMD flakes, thickness-dependent dips in addition to the exciton resonance dips are clear signatures of the self-hybridized EPs. [24][25][26] We recently reported that substantial phase shifts of light at the WS 2 /Au interface could cause EPs to occur even in sub-10 nm thick WS 2 flakes on Au layers. [26] Comparing WS 2 /Ag samples with and without the Au top electrodes reveals that the presence of the Au layer significantly modifies the reflectance spectra ( Figure S8, Supporting Information). Consequently, the reflectance dips at λ = 595 and 645 nm can be attributed to the thickness-dependent EP resonance of Au/WS 2 /Ag structures. [25,26] In contrast, the weak dip at λ = 520 nm is originated from the exciton resonance of bulk WS 2 . [25,26,28] The reflectance spectra in Figure S8 (Supporting Information) suggest that the EP formation can reduce the optical reflectance of our device over broad wavelength ranges. [25,26] As shown in Figure 5b, the averaged power loss density at each position (z) and a specific wavelength (λ) can be obtained from the following equation, Q e (z,λ) = nκωε 0 |E(z,λ)| 2 , in which n, κ, ω, and ε 0 represent the real and imaginary part of the refractive index, input light frequency, and vacuum permittivity, respectively. E(z,λ) is the calculated total electric field under a fixed input power density of 1 W m −2 . The distribution of Q e (z,λ) clearly shows that the WS 2 flake, rather than the top and bottom electrodes, can absorb most of the incident light. The electric field distributions were obtained from the optical simulations, as shown in Figure S7b,c (Supporting Information). The field intensity indicates that the high reflectivity of the sample leads to the formation of standing waves perpendicular to the sample surface ( Figure S7b, Supporting Information). It can be noted that the overall electric field in the air at λ = 595 and 645 nm has a significantly lower intensity than the field in other wavelength ranges. The Q e (z,λ) distribution in Figure 5b suggests that large absorption in the Au/WS 2 /Ag device drastically decreases the reflected light intensity in the air. are 21% and 9%, respectively, for the entire visible wavelength range. The bottom Ag layers work well as back reflectors, as reported by others' works. [10,14] The strong electric field in the extremely thin flake raises the electric field in the adjacent metal layers, as shown in the calculated electric field distribution ( Figure S7c, Supporting Information). Consequently, the absorption peaks of the electrodes are similar to those of the WS 2 flake.
Because the cantilever used for the c-AFM measurements was located above the sample, the light J-V characteristics in Figure 4a,b were obtained at a large incident angle of θ = 60°.

www.advmatinterfaces.de
The largest (smallest) J SC of 1.45 (0.42) mA cm −2 was obtained from the orange (deep red) LED with a peak wavelength of 625 (680) nm. The measured J SC values from the green (peak wavelength: 522 nm) and blue (peak wavelength: 470 nm) LEDs were 0.85 and 0.72 mA cm −2 , respectively. These wavelengthdependent J SC can be well explained by the calculated A[WS 2 ] in Figure 6c.
While varying θ from 0° to 30°, the amount of absorption at each layer (A[Au], A[WS 2 ], and A[Ag]) does not change much (Figure 6a,b). When increasing θ up to 60°, a reduction of the absorption can be noticed (Figure 6c). Figure 6d shows  According to the Fresnel law, incident electromagnetic (EM) waves on multilayers undergo phase shifts upon reflection and transmission at the interfaces of two different media. [31] Since TMD and metal have complex refractive indices with very large real and imaginary parts in the visible spectrum, the reflection/transmission-induced phase shifts of EM waves at the TMD/metal interfaces, rather than the phase difference induced by the propagation in thin TMD layers, dominantly determine optical interference of our metal/TMD/metal structures. [31] The reflection and transmission phase shifts depend on the polarization and θ of incident light. As a result, the reflectance and absorption spectra of our Au/WS 2 /Ag devices can vary with respect to angle and polarization (Figures S9, S10a-d, and S11a-d, Supporting Information). At θ = 30°, two distinct polarizations (S and P) affect the reflectance and absorption spectra (Figures S9 and S10a-d, Supporting Information). At θ = 60°, the polarization-dependence becomes evident (Figures S9 and S11a-d, Supporting Information). S-polarized light significantly increases the total reflectance at high incident angles, whereas P-polarized light decreases the total reflectance, especially at short wavelengths ( Figure S9, Supporting Information). Large reflectance leads to absorption loss: the overall A[WS 2 ] at θ = 60° under S-polarized light illumination is less than that under P-polarized light illumination ( Figure S10c,d, Supporting Information). The measured light J-V plots support the calculated polarization-dependent absorption spectra ( Figure S12, Supporting Information). The average absorption under P-and S-polarized light illumination corresponds to the absorption under unpolarized light illumination. Consequently, high incident angles do not severely decrease A[WS 2 ], as shown in Figure 6d. There are daily, seasonal, and site-specific variations in the solar angle of incidence. Therefore, angle-insensitive absorption is beneficial for photovoltaic devices operating outdoors.

Conclusion
We fabricated Au/WS 2 -multilayer/Ag Schottky junctions, which exhibited a high V OC of 170 mV, a large J SC of 1.45 mA cm −2 , a www.advmatinterfaces.de fill factor of 0.51, and a PCE of 5.0%, when illuminated by an LED with a peak wavelength of 625 nm and an optical power density of 2.5 mW cm −2 . The 10 nm thick Au top electrodes and highly reflective Ag bottom electrodes allowed WS 2 flakes to absorb most of the incident light. Furthermore, the large refractive indices and ultrathin thickness of the WS 2 multilayers enabled broadband and angle-insensitive absorption, which is highly desirable for photovoltaic device operation. We successfully demonstrated high-performance and high-yield vertical Au/WS 2 -multilayer/Ag Schottky junctions using lithographyfree processes.

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