Interface designed MoS2/GaAs heterostructure solar cell with sandwich stacked hexagonal boron nitride

MoS2 is a layered two-dimensional semiconductor with a direct band gap of 1.8 eV. The MoS2/bulk semiconductor system offers a new platform for solar cell device design. Different from the conventional bulk p-n junctions, in the MoS2/bulk semiconductor heterostructure, static charge transfer shifts the Fermi level of MoS2 toward that of bulk semiconductor, lowering the barrier height of the formed junction. Herein, we introduce hexagonal boron nitride (h-BN) into MoS2/GaAs heterostructure to suppress the static charge transfer, and the obtained MoS2/h-BN/GaAs solar cell exhibits an improved power conversion efficiency of 5.42%. More importantly, the sandwiched h-BN makes the Fermi level tuning of MoS2 more effective. By employing chemical doping and electrical gating into the solar cell device, PCE of 9.03% is achieved, which is the highest among all the reported monolayer transition metal dichalcogenide based solar cells.

Two-dimensional (2D) materials provide rich physics in designing of new optoelectronic devices [1][2][3][4] . Because of the low light absorbance of the atomic thin 2D materials 5,6 , the external semiconductor is usually incorporated to improve the performance of 2D material based devices [7][8][9] . Photodetectors based on monolayer graphene have been reported to show photo gain as high as ~10 8 and photo responsivity as high as ~10 7 A/W through the enhanced light absorption with covering semiconductor quantum dots on graphene 8 . Forming 2D materials/bulk materials heterostructure junctions is an alternative choice to obtain high performance optoelectronic devices as the bulk semiconductor can fully absorb incident light 10,11 . As the first discovered 2D material with many fascinating electrical and optical properties, graphene and its heterostructures have attracted much attention for solar cells worldwide [12][13][14][15] . Power conversion efficiency (PCE) of solar cells based on graphene/Si system has been improved from 1.65% to 15.6% since the first reported graphene/Si heterostructure solar cell in the year 2010 16,17 . Recently, we have reported graphene/GaAs solar cell with PCE of 18.5% 18 . On the other hand, single layer 2D molybdenum disulfide (MoS 2 ) is semiconductor with a direct band gap of 1.8 eV 19 . MoS 2 with thickness less than 1 nm can absorb 5-10% incident light 20 . Also, MoS 2 can be synthesized with large area by chemical vapor deposition (CVD) method [21][22][23] . Based on the abovementioned merits, the MoS 2 /bulk semiconductor system offers a new platform for optoelectronic device design. It has been reported that MoS 2 / Si heterostructure solar cell has an efficiency of 5.23% with the assistance of aluminum deposition on MoS 2 6 . However, much more work on MoS 2 /semiconductor heterostructure is highly desirable both for the fundamental research interest and the potential photovoltaic application. Among all the bulk semiconductors, GaAs has a suitable direct band gap of 1.42 eV and high electron mobility (8000 cm 2 V −1 s −1 at 300 K) 24 , which makes itself one of the best candidates for high performance solar cells 25,26 . Tunable Fermi level is one of the unique physical properties of 2D materials, which can be finely tuned by chemical doping or electrical gating [27][28][29][30] . Different from the conventional bulk p-n junctions, there is static charge transfer between 2D materials and bulk semiconductor, which could severely lower the Fermi level difference between bulk semiconductor and 2D material 31 , and lead to a decreased junction barrier height. The photovoltaic performance of the heterojunction is greatly influenced by the junction barrier height, which means suppressing the static charge transfer between 2D materials and semiconductor substrate are highly desirable. Herein, we introduce 2D hexagonal boron nitride (h-BN) into the MoS 2 /GaAs heterostructure to suppress the static charge transfer. More importantly, the inserted h-BN layer makes the tuning of Fermi level of MoS 2 more effective, which greatly improves the performance of solar cells. Based on the interface band structure designing and Fermi level tuning of MoS 2 , 9.03% of PCE has been achieved.

Results
Physical design of the MoS 2 based solar cell. The schematic electronic band structure of the independent MoS 2 and GaAs is shown in Fig. 1a. The electron affinity (energy gap between vacuum level and the bottom level of conduction band E C-MS ) of MoS 2 (χ MS ) is 4.0 eV 32 , and the band gap of MoS 2 is 1.8 eV. As the measured sheet resistance of MoS 2 is in the range of 10 4 -10 6 Ω /□ , indicating the Fermi level of MoS 2 (E F-MS ) locates near the middle of the band gap. The electron affinity of GaAs (χ GA ) is 4.07 eV. The Fermi level of GaAs (E F-GA ) used in this study locates around the bottom level of conduction band (E C-GA ) because the n-type doping concentration is around 10 18 cm −3 . When MoS 2 touches with GaAs, due to the Fermi level difference, some of majority electrons of GaAs inject into MoS 2 , shifting E F-MS by Δ E F-MS , as shown in Fig. 1b, which can be quantitatively expressed as: where Δ n is the change of electron concentration in MoS 2 which affected by the injected electrons from GaAs, n i is the intrinsic carrier concentration in MoS 2 , k is the Boltzmann constant and T is the absolute temperature. The barrier height (Φ barrier ) of the MoS 2 /GaAs heterojunction can be presented as: It is very clear that suppressing the static charge transfer during the formation of MoS 2 /GaAs heterojunction can result in higher Φ barrier . We propose a device by inserting 2D h-BN into MoS 2 /GaAs Schottky diode as the interface layer to suppress the static charge transfer. h-BN is one of 2D materials with a band gap of 5.9 eV and dielectric constant of 4.0 33 . The electronic band alignment of MoS 2 /h-BN/GaAs heterojunction can be seen in Fig. 1c. As h-BN has a negative electron affinity 34 , the electron transfer from GaAs to MoS 2 is suppressed during the formation of the MoS 2 /h-BN/GaAs heterojunction. As a result, Δ E F-MS is reduced and Φ barrier of the junction is lifted up. Under illumination, photo generated excess electrons and holes are collected by GaAs and MoS 2 , respectively. As shown in Fig. 1d, transport of holes from GaAs to MoS 2 is almost unaffected after inserting the ultrathin 2D BN layer, which dominates the power conversion from light to electricity. In other words, the open circuit voltage (V oc ) of the solar cell can be increased by the inserted h-BN while short circuit current density (J sc ) stays almost unchanged, thus, solar cell with a better performance can be expected. Fig. 2 shows the schematic fabrication processes of MoS 2 /GaAs and MoS 2 /h-BN/GaAs Schottky junction based solar cells. After removal of the native oxide on the GaAs substrate, Au with a thickness of 60 nm was evaporated on the rear surface of GaAs forming ohmic contact. Then the front surface of GaAs was cleaned with dilute HCl aqueous solution. Front surface passivation is achieved with remote NH 3 plasma treatment for 5 min with power of 120 Watt and frequency of 27.5 MHz. After the passivation treatment, h-BN and MoS 2 in sequence or MoS 2 alone is directly transferred onto the front surface of GaAs substrate, followed with the deposition of front Au contacts (60 nm) with mask. Inset in Fig. 2 shows the digital photographs of the typical MoS 2 /GaAs and MoS 2 /h-BN/GaAs heterojunction based solar cells, where the thin line shape of the active area can be seen, which is designed for efficiently current collection based on the high resistance of monolayer MoS 2 . Fig. 3a shows the schematic cross section structure of the MoS 2 /h-BN/GaAs heterojunction solar cell, which is composed of rear Au contact, GaAs substrate, h-BN layer, MoS 2 layer and the front Au contact from bottom to top. The active area of the device is defined with the opened window in the front Au contact, as shown in Fig. 3b. The width of the active area is 120 μ m and the length is 5 mm, making the active area 0.6 mm 2 . As the thickness of the front Au contact is 60 nm, no light can be absorbed by the device in the Au shadowed area, which guarantees the precise active area. The high resolution transmission electron microscopy (HRTEM) image of the MoS 2 is shown in Fig. 3c, which shows the six fold symmetry nature of the MoS 2 . The inset of the Fig. 3c shows the HRTEM image of MoS 2 layer, which clearly indicates the CVD grown MoS 2 is monolayer. The electron diffraction pattern can be seen in Supplementary Information Fig. S1, which also implies the monolayer nature of MoS 2 . Fig. 3d shows the absorption spectrum of the CVD grown MoS 2 , where three absorption peaks corresponding to 436 nm, 619 nm and 662 nm can be seen in the wavelength range of 350-800 nm. The peak of the absorbance locates at 436 nm is 8.9%, which is in agreement with the reported absorbance of the monolayer MoS 2 20 . The Raman spectrum of the MoS 2 on Si/SiO 2 substrate is shown in Fig. 3e. The Raman peaks corresponding to E 1 2g and A 1g modes of MoS 2 locate at 384.3 cm −1 and 404.5 cm −1 , respectively, indicating the grown MoS 2 is monolayer 35 . Fig. 3f presents the Raman spectrum of h-BN, where the 1371 cm −1 peak indicates it is monolayer 36 . The digital photographs of transferred h-BN on Si/SiO 2 substrate, optical microscopy image and atomic force microscopy image of MoS 2 and optical microscopy image of h-BN can be seen in Supplementary  Information Fig. S2, where can be seen that the homogeneity of CVD grown MoS 2 and h-BN is good. Fig. 4a shows the dark current density-voltage (J-V) curves of the MoS 2 /GaAs and MoS 2 /h-BN/GaAs heterojunctions, both of which show good rectifying characteristics. It is noteworthy that in this study if no mentioned, GaAs substrate is n-type doped. We also test the J-V curve of MoS 2 /p-GaAs, which shows bad rectifying characteristics as presented in Supplementary Information Figure S3. The threshold voltage (the voltage needed to reach a current density of 2 mA/cm 2 here) for the MoS 2 /GaAs heterojunction is 0.41 V, while the value for the MoS 2 /h-BN/GaAs hetrostructure is 0.52 V, suggesting that Φ barrier is increased by the interlayer h-BN. The value of Φ barrier can be deduced through fitting of dark J-V curves as expressed by:

Basic properties of the MoS 2 /h-BN/GaAs heterostructure solar cell.
where K is the Boltzmann constant, N IF is the junction ideality factor and q is the value of electron charge. Based on thermionic-emission theory, saturation current density J 0 can be described as: where A * is the effective Richardson's constant of n-type GaAs (8.16 A/k•cm 2 ) 37 . Based on equations (3) and (  to higher FF while higher R s leads to lower FF. For the MoS 2 /h-BN/GaAs device, even with increased R s , the decreased N IF increases the FF compared with the value of MoS 2 /GaAs device. The electrical properties mentioned above indicate the importance of the interface recombination condition for the MoS 2 /GaAs and MoS 2 /h-BN/GaAs devices. Here transient photoluminescence (PL) is employed to investigate the kinetics of the photo generated carriers near the interface. Fig. 4d shows the transient PL decay curves for bare GaAs substrate, and the same MoS 2 /GaAs and MoS 2 /h-BN/GaAs heterostructure devices with PCE of 4.82% and 5.42% respectively. The decay curves show double channel dependent behavior, corresponding to a fast decay channel (in the range of 1 ns to 2 ns) and a slow decay channel (after 2 ns). The wavelength of the excitation laser is 450 nm, and the absorption depth is close to the surface of GaAs (about 50 nm). The quick decay range in the first nanosecond is related to carrier kinetics at surface or interface, and the subsequent slow decay range is dominated by the bulk recombination processes. PL decay time constants are deduced by exponentially fitting the PL intensity decay curves in the fast and slow decay ranges as shown in Fig. 4e,f, respectively. The fitted PL decay time constants in the fast decay range for bare GaAs, MoS 2 /GaAs and MoS 2 /h-BN/GaAs are 0.97 ns, 0.59 ns and 0.52 ns, respectively, and the values in the slow decay range are 1.92 ns, 1.89 ns and 1.85 ns, respectively. For the bare GaAs, photo generated excess carriers recombine with emission of photons or phonons. In MoS 2 /GaAs heterostructure, besides the process mentioned above, parts of the excited holes in GaAs are separated by the heterojunction and collected by MoS 2 . The separated electrons and holes cannot participate in the radiation recombination process. Thus PL decay time constant is decreased. In the MoS 2 /h-BN/GaAs heterostructure, Φ barrier is increased, which leads to higher speed of carrier separation process and even shorter PL decay time constant. Recombination will take place when electrons produced in GaAs cross the interface. The recombination rate is influenced by the carriers crossing time, which corresponds to the PL decay time constant in the fast decay range. For the device with interlayer h-BN, crossing time is shortened, resulting in the lowered interface recombination rate and lower value of N IF . For the PL decay in the slow decay range, similar time constants imply that the PL decay process is dominated by the bulk recombination properties.  to 20.8 mA/cm 2 . The FF values are 46.6% and 53.7%, and the PCE values are 5.38% and 7.15% for the undoped and doped devices, respectively. The increase of the FF after doping is related to the decrease of R s , as shown in Fig. 5b, the value of R s without doping of MoS 2 is 56.0 Ω , while R s is decreased to 45.9 Ω after doping. In addition, to explore the stability of the doped MoS 2 /h-BN/GaAs solar cell, we test the variation of the PCE values in 50 hours under AM1.5G illumination, as seen in Fig. 5c. The device was sealed by polymethyl methacrylate (PMMA) through spining-on-coating. The starting PCE value is 6.73%, while after illumination for 50 hrs, the PCE increases to 6.96%. Considering the light induced degradation of crystaslline silicon solar cell is usually happened in the first 24 hrs under AM1.5G illumination, it can be safely concluded the stability of MoS 2 /h-BN/GaAs solar cell under illumination is good with suitable encapsulation.

Further improvement of the MoS 2 /h-BN/GaAs heterostructure solar cell by electrical gating.
As a atomic thin 2D semoconductor, the Fermi level of MoS 2 can be finely tuned with gating effect 38 .
Here we employ PEO based ion polymer as the top gate electrode 39 , and the schematic structure of the field effect MoS 2 /h-BN/GaAs solar cell is shown in Fig. 6a. Ion gate is directly covered on the surface of AuCl 3 doped MoS 2 . Negative voltage is applied on the ion gate and the rear Au contact is connected to the ground. Fig. 6b shows the J-V curves of the field effect solar cell under AM1.5G illumination. When gate voltage (V gate ) equals to − 0.5 V, V oc of the solar cell is increased from 0.64 V to 0.72 V. Meanwhile, J sc is slightly increased from 20.2 mA/cm 2 to 20.7 mA/cm 2 , which might be attributed to the enhanced efficiency of charge seperation with gating. FF is increased from 53.1% to 54.9% and PCE is improved from 6.87% to 8.27%. when V gate equals to − 1.0 V, the obtained values of V oc , J sc and FF are 0.76 V, 21.1 mA/cm 2 and 56.3%, respectively. And the final PCE is 9.03%. Fig. 6c shows the dark J-V curves with different V gate . The threshold voltage is increased when the negative V gate increases. By fitting of the dark J-V curves shown in Fig. 6c, values of Φ barrier can be obtained and shown in Fig. 6d, where the values of V oc corresponding to different V gate are also shown. Fig. 6d discloses that the increased V oc is mainly attributed to the improved Φ barrier under gating effect. The

Discussion
MoS 2 /GaAs heterostructure based solar cell is investigated. Different from the traditional p-n junctions and metal/semiconductor Schottky junctions, the charge transfer between MoS 2 and the GaAs substrate can greatly influence the position of Fermi level in the 2D material, which leads to a much lower barrier height than the ideal value originated from the Fermi level difference. The barrier height is a key factor for the electrical properties of electronic and optoelectronic devices. Thus, suppressing or preventing the charge transfer during the formation of the 2D material based heterojunctions is highly desired to achieve high performance devices. Herein, we demonstrated the performance of MoS 2 /GaAs based heterostructure solar cell is improved by inserting interlayer h-BN. The inserted h-BN layer can suppress the electron injection from n-type GaAs into MoS 2 during the junction formation, while does not affect the hole separation and collection processes according to the electronic band structure of h-BN. Thus, PCE is increased from 4.82% to 5.42% after inserting the BN layer as higher barrier height and V oc can be achieved. Furthermore, by employing chemical doping and electrical gating into the solar cell device, PCE of 9.03% is achieved, which is the highest among all the reported monolayer transition-metal dichalcogenide-based solar cells. This physical picture and technique could be extended into other 2D materials/semiconductor heterostructure based electronic and optoelectronic devices.

Methods
Monolayer h-BN was grown on copper substrate with B 3 N 3 H 6 as the precursor at 1000 o C for 30 min 40 . Single layer MoS 2 film was grown on Si/SiO 2 substrate in a quartz tube with CVD method 41 . MoO 3 powder and sulfur powder (99.9%, both bought from Aladdin) was used as the precursor. Growth temperature was set at 650 o C. 60 nm Au was thermally evaporated on back surface of GaAs to form rear contact. GaAs substrate was cleaned by dipping the samples into 10%wt HCl solution for 5 min followed with DI water rinse. Surface passivation of GaAs was realized by remote NH 3 plasma treatment for 5 min with 120 Watt 27.5 MHz RF generator. h-BN was transferred onto the GaAs substrate using PMMA as the sacrificing layer. After PMMA spun-on, MoS 2 on Si/SiO 2 substrate was immersed into deionized water to lift-off the PMMA-MoS 2 films. After transferring, PMMA was removed by immersing the samples into acetone for 20 min.
MoS 2 and h-BN were characterized by Raman spectroscopy (Renishaw inVia Reflex) with the excitation wavelength of 532 nm. The microstructure of MoS 2 was examined by HRTEM (Tecnai F-20 operating at 200 KV). Atomic force Microscopy (AFM) characterization was performed using Veeco dimension 3100 system. The MoS 2 /h-BN/GaAs solar devices were tested by Agilent B1500A system with a solar simulator under AM1.5G condition. It is noteworthy that the illumination intensity was calibrated with a standard Si solar cell. Transient PL measurements were used to evaluate the charge recombination and separation behaviors at the interfaces of MoS 2 /h-BN/GaAs heterojunction. The excitation light source (PicoHarp 300 system) was a 450 nm pulsed laser with 1 MHz repetition rate and 50 ps pulse duration with power of 10 μ W. The diameter of the excitation laser spot was 10 μ m. The PL signal with wavelength shorter than 1100 nm was collected by a multimode optical fiber and recorded by a Horiba Jobin Yvon iHR550 spectrometer. All spectra were collected until the peak value reaching 5000 counts.