Sputtering of Molybdenum as a Promising Back Electrode Candidate for Superstrate Structured Sb2S3 Solar Cells

Abstract Sb2S3 is rapidly developed as light absorber material for solar cells due to its excellent photoelectric properties. However, the use of the organic hole transport layer of Spiro‐OMeTAD and gold (Au) in Sb2S3 solar cells imposes serious problems in stability and cost. In this work, low‐cost molybdenum (Mo) prepared by magnetron sputtering is demonstrated to serve as a back electrode in superstrate structured Sb2S3 solar cells for the first time. And a multifunctional layer of Se is inserted between Sb2S3/Mo interface by evaporation, which plays vital roles as: i) soft loading of high‐energy Mo particles with the help of cottonlike‐Se layer; ii) formation of surficial Sb2Se3 on Sb2S3 layer, and then reducing hole transportation barrier. To further alleviate the roll‐over effect, a pre‐selenide Mo target and consequentially form a MoSe2 is skillfully sputtered, which is expected to manipulate the band alignment and render an enhanced holes extraction. Impressively, the device with an optimized Mo electrode achieves an efficiency of 5.1%, which is one of the highest values among non‐noble metal electrode based Sb2S3 solar cells. This work sheds light on the potential development of low‐cost metal electrodes for superstrate Sb2S3 devices by carefully designing the back contact interface.


Introduction
Binary compound antimony sulfide (Sb 2 S 3 ) has attracted a great deal of interest as a potential absorber candidate for DOI: 10.1002/advs.202303414photovoltaic applications due to its high absorption coefficient (1×10 5 cm -1 ), abundant raw materials, and environment-friendly characteristics. [1]Moreover, its suitable band gap (≈1.7 eV) makes its theoretical efficiency up to ≈30%. [2]However, there is a big gap between the recorded efficiency (8%) and the theoretical value. [3]omprehensive studies, such as device structures, film preparation, and interface engineering, have been well conducted.For most highly-efficient Sb 2 S 3 solar cells, they commonly use an organic hole transport layer (HTL) and/or noble Au electrode as the back contact materials (BCM).For example, the recorded power conversion efficiency (PCE) of Sb 2 S 3 solar cells in dye-sensitized and planar structures are 7.5% and 8% which are equipped with the PEDOT:PSS/PCPDTBT/Au and Spiro-OMeTAD/Au, respectively. [3,4]Besides, the P3HT is also fully exploited by the Papadopoulos group and achieves a PCE of 3.97% of Sb 2 S 3 solar cells. [5]The outstanding physical property, i.e., suitable band level of these organic HTL, and high work function (W F , 5.1 eV) and conductivity (4.52×10 7 S m -1 ) of Au, enables many groups to make great progress of Sb 2 S 3 solar cells as shown in Table S1 (Supporting Information).However, the organic HTL and/or noble metal obviously limit the stability of the devices and also increase the manufacturing cost.Hence, it is urgent to develop a low-cost and highly stable BCM for Sb 2 S 3 solar cells.Encouraging, carbon has a W F close to that of gold, which is also regarded as a candidate for low-cost electrode materials to replace Au. [6] But, the current carbon-based Sb 2 S 3 solar cells is still restricted by the high square resistance of carbon electrode, due to its porosity characteristics. [7]Therefore, it is necessary to develop a cheap, stable, and highly conductive back electrode for Sb 2 S 3 solar cells.
Mo, a kind of abundant metallic element (1.5 ppm) and lowscost (< 0.1% of that of Au), has attracted extensive attention in the field of optoelectronic devices and energy storage due to its excellent electrical conductivity (1.87×10 7 S m -1 ) and high toughness. [8]In particular, Mo shows a great application potential in the photovoltaic device by taking advantage of its high thermal stability and flexibility.For example, state-of-the-art thin-film solar cells, such as copper indium gallium selenide (CIGS) and copper zinc tin sulfoselenide (CZTSSe) routinely adopt Mo as a bottom electrode, then assembling into a substrate device. [9,10]urthermore, Mo is also applied in substrate Sb 2 S 3 thin film solar cells with Mo/Sb 2 S 3 /CdS/ZnO/AZO/Al structure, achieving a moderate efficiency of 3.75%. [11]14][15][16][17] It is suggested that Mo is a potential electrode material for substrate Sb 2 S 3 thin film solar cells, in which Mo is directly sputtered on an inert glass.This deposition process is harsh with high sputtering power and high temperature.Hence, to improve the buried interface of Mo and Sb 2 S 3 , modification of Mo electrode, including the regulation of its adhesion, conductivity, and surficial property, can be easily realized by experiencing the Mo electrode under high pressure, high-temperature in situ deposition, and violent post-sulfurization/selenization.However, up to now, Mo does not appear as the back electrode in the superstrate structured Sb 2 S 3 solar cells, in which a simple architecture of TCO/ETL/Sb 2 S 3 /BCM is used (TCO: FTO, ITO; electron transport layer (ETL): TiO 2 , CdS, SnO 2 ; BCM: Spiro-OMeTAD and/or Au).The reason is that Mo metal possesses high thermodynamic stability and melting point, so it is inapplicable for the traditional chemical coating method or evaporation deposition route.By contrast, the magnetron sputtering strategy can perfectly deposit the Mo layer with a uniform morphology and is more suitable for industrial production.While the Sb 2 S 3 film will suffer from the violent bombardment of high-energy particles during the sputtering process and greatly deteriorate the surface structure of the Sb 2 S 3 layer.The damaged surface inevitably introduces a mass of defect sites and then leads to severe recombination, which is a great challenge for superstrate structured Sb 2 S 3 solar cells.This is the main consideration about the application of sputtering metal electrode for a superstrate solar cell and still pending for the non-noble metal deposition.In addition, Mo will produce serious band alignment mismatches if it is directly combined with the Sb 2 S 3 absorption layer, due to the low WF of Mo and deep valence band maximum (VBM) of Sb 2 S 3 .Therefore, how to solve the band-level mismatch between Mo and absorption layer as well high-energy Mo beam bombardment are vital to improve the PCE of low-cost Sb 2 S 3 solar cells.
In this work, we deposit Mo as the back electrode of a superstrate structured Sb 2 S 3 solar cell by magnetron sputtering for the first time.And evaporation of a glue layer of Se is skillfully introduced between Sb 2 S 3 and Mo electrodes to alleviate the high-energy particle penetration during the sputtering process, namely the soft-loading of Mo.The trade-off between the shield function and carrier transport barrier of the Se layer is studied in detail by adjusting its thickness.In addition, an ultrathin Sb 2 Se 3 buffer layer is detected on the Sb 2 S 3 surface after the post-annealing, attributing to trace residue of the Se shield layer, which logically up-shifts VBM of Se-modified Sb 2 S 3 surface and then significantly improves hole collecting efficiency.However, due to the low W F of the Mo electrode, small band bending hinders the hole transportation toward the Mo electrode, leading to a serious trailing state which results in an obvious cross-over effect of the current density-voltage (J-V) curve. [18]Hence, we skillfully sputter a pre-selenide Mo target which is aimed to insert a MoSe 2 buffer layer between Sb 2 S 3 and Mo.Hereafter, composite layers of MoSe 2 and Mo are sputtered on the Sb 2 S 3 film.This MoSe 2 buffer layer effectively alleviates the energy level mismatch between Mo and Sb 2 S 3 and then accelerates the transport of carri-ers and improves the efficiency of the device.Finally, a new device structure is assembled as FTO/CdS/Sb 2 S 3 /MoSe 2 /Mo and delivers a champion efficiency of 5.1% which is the highest PCE of non-noble metal electrode-based Sb 2 S 3 .

Superstrate Structured Sb 2 S 3 Solar Cells with Mo Electrodes
Mo exhibits tremendous potential as electrodes in thin film solar cells by enjoying its high thermal stability, conductivity, and flexibility.It is often used as a bottom electrode either in CIGS, CZTSSe, or Sb 2 (S x ,Se 1−x ) 3 (0≤x≤1) with substrate structures so far. [19,20]In this work, Mo is used as an electrode in superstrate structured Sb 2 S 3 solar cells for the first time, displaying a simple device architecture of FTO/CdS/Sb 2 S 3 /Mo (Figure 1).As a preliminary attempt, the Mo electrodes were deposited directly onto the as-annealed FTO/CdS/Sb 2 S 3 film by magnetron sputtering (150 °C, 100 W, 2 h), where the active area of 0.09 cm 2 is defined by a mask.For more sputtering details, please refer to the Experimental Section.
First, we conducted a scanning electron microscopy (SEM) to explore the morphologies of the as-prepared Sb 2 S 3 films.It can be observed in Figure S1 (Supporting Information) that the annealed Sb 2 S 3 films (crystalline phase, referred to as c-Sb 2 S 3 ) derived from the hydrothermal route exhibit a flat and uniform morphology which is the prerequisite for the subsequent sputtering deposition of Mo electrodes.As shown in Figure 2a of the crosssectional SEM image, a high-quality Mo electrode is directly deposited on the c-Sb 2 S 3 thin film by a facile magnetron sputtering, exhibiting a high uniformity and flatness attached with typical vertical columnar grains.It is well known that such a sputtering process is widely adopted to grow compact, uniform, tough, and highly conductive Mo electrodes, and displaying a similar columnar morphology when compared with the counterpart deposited on a glass substrate (Figure S2, Supporting Information).This is well consistent with Mo electrodes in substrate structured Sb 2 S 3 , as well as CIGS, CZTSSe solar cells. [21]The conductivity, one of the most appealing characters of the metal electrodes, is further explored on devices by a four-point probe test (Figure 2b).It is concluded that the square resistance of the Mo deposited on Sb 2 S 3 surface (5.6 Ω □ -1 ) is not distinctly different from that on the glass substrate (5.4Ω □ -1 ) in the same thickness of ≈900 nm under the standard sputtering process.Unexpectedly, the Moattached Sb 2 S 3 solar cells shows an awful device performance (Figure 2c), although the excellent characteristic of the Mo electrode is presented above.Almost no photoelectric response is observed in this case which may be caused by the high-energy particles of Mo penetration to form shunting paths.In order to reveal the underlying reason, it can be seen from the topography of Figure 2a that Mo deposition is in good contact with the surface of c-Sb 2 S 3 layer, but can also be embedded at the grain boundary owing to the tiny holes.Based on this, we suggest that high energy particles of sputtered Mo particle would penetrate the c-Sb 2 S 3 layer through the grain boundary or the void created by grain fusion, and generating shunt paths.In order to mitigate the damage to the as-prepared Sb 2 S 3 film caused by the high energy particles, the sputtering power is reduced from 150 to 60 W.However, it still cannot avoid from device failure which can be seen in J-V curve of Figure S3 (Supporting Information).Hence, no devices are exempt from shunting once the Mo metal is directly sputtered on the as-annealed Sb 2 S 3 films.
As discussed above, the possible shunt path maybe created at grain boundary of the annealed Sb 2 S 3 film.Hence seeking of a dense film gives the top priority to prevent the device breakdown against the energetic particles beam during Mo sputtering.As shown in Figure S4 (Supporting Information), we can find that the as-prepared Sb 2 S 3 precursor by hydrothermal method, exhibiting with amorphous property (referred to as -Sb 2 S 3 ), is obviously denser than the c-Sb 2 S 3 polycrystalline film.Therefore, the Mo electrode is optimally deposited on -Sb 2 S 3 , which shows a tight contact with Sb 2 S 3 precursor (Figure 2d).Sequentially, the samples are integrally annealed under N 2 atmosphere to induce the grain growth of Sb 2 S 3 and then consequentially reduce the bulk recombination.As displayed in Figure 2e, large grains are observed in c-Sb 2 S 3 film after post-annealing, and the tight back contact interface is still reserved which is benefit for the effective carrier collection of Mo electrode.In this case, the Sb 2 S 3 device performance has greatly improved from the initial resistance state to an obvious photoelectric response, which can be seen in Figure 2f.For example, the open circuit voltage (V oc ) is boosted to 0.294 V, and current density (J sc ) is enhanced to 10.4 mA•cm −2 .But, the PCE of 0.97% is still poor, featured with relatively low V oc and fill factor (FF), indicating the fatal destruction of high-energy Mo particles cannot be completely avoided.Therefore, it is urgent  to introduce a protective layer to solve the device shunting during sputtering Mo electrode.

Insertion of a Selenium Multifunctional Layer between Back Contact Interface
Selenium is an excellent semiconductor material, possessing charming features of stable single-element component, high saturated vapor pressure (1.1 Pa @ 250 °C) and low melting point (220 °C), which is extensively used in the preparation and photoelectric property regulation of chalcogenide compounds. [22]With regard to Sb 2 (S, Se) 3 system, i) Sb 2 S 3 and Sb 2 Se 3 are isomorphic with same space group and similar cell parameters, indicating that S atom of Sb 2 S 3 is easily replaced by Se atom.ii) It also bestows an excellent ability of bandgap engineering by facilely changing S/Se ratios.Then Sb 2 Se 3 generated on the surface of Sb 2 S 3 can facilitate the hole transport toward BCM by upshifting the deep VBM of Sb 2 S 3 .iii) Se layer can be used as a barrier layer to prevent the penetration of high-energy particles, realizing soft landing of Mo during sputtering.iv) MoSe 2 may be thermodynamically formed by reacting Se with Mo, and then effectively alleviates the energy level mismatch at back contact interface, which can accelerate the transport of carriers and improve the device PCE.
In order to verify the above multi-merits of Se, a glue Se layer is then employed by flash evaporation prior to the Mo electrode construction for Sb 2 S 3 solar cells, please refer to left sketch of Figure 1.The as-evaporated Se film on Sb 2 S 3 surface shows a cotton-like morphology with porous features and a thickness of 460 nm (Figure 3a,b), which can be attributed by the low-energy Se atoms during the flash evaporation.After sputtering of Mo electrodes, a Mo layer with rice grains is then coated on Sb 2 S 3 precursor, also leaving a tiny porous selenium (≈76 nm) between them (Figure 3c).However, after annealing, a close connection of Mo and c-Sb 2 S 3 is observed in Figure 3d, in which the residual Se is consumed by reacting with Sb 2 S 3 during postannealing (discussed in the following section).Such a resulted perfect back contact interface is benefit for the hole collection toward Mo electrodes.Here, we propose a sputtering growth model of Mo electrodes with the assist of an additional Se layer.As shown in Figure 3d, the sputtered Mo particle with high energy first approaches Sb 2 S 3 surface and then collides with the cottonlike Se layer, where this scatter-shot will speed down the Mo particles and realize a soft loading of Mo without damage of Sb 2 S 3 .The similar energy absorption effect is observed in the porous materials, which also confirms that the as-prepared porous selenium layer could effectively absorb the high energy of the sputtering Mo particles and serves as an important buffer layer.Concurrently, a large proportion of Se is scattered out from Se glue layer, leaving a trace residue of Se (Figure 3c).After complete coverage of the bottom layer of Mo, the upper Mo will then be deposited smoothly with a thickness of 800 nm within 2 h sputtering.This growth model is also confirmed by the structure of as-deposited Mo electrode, as shown in Figure S10 (Supporting Information).The XRD result indicates the peak intensity of Mo at 2 = 41.5°decreasessignificantly after adding 0.1 g Se prior to Mo deposition.This is due to fact that the Mo atoms with lower energy, as indicated in the second stage of growth model of Figure 3e, is not sufficient to migrate and induce metal recrystallization and finally degrade its crystallinity.Then the conductivity of as-deposited Mo is tested by four-point probe test in Figure S5 (Supporting Information) in consideration of the distinctive morphology and crystalline of Mo W/O or W/ selenium glue layer.Encouragingly, the Mo electrode coated with the help of Se layer still exhibits a relatively low square resistance of 6 Ω □ -1 which is comparable with the counterpart of Mo on glass.In brief, the additional Se glue layers not only speed down the sputtered Mo atoms and form a benign contact interface, but also maintain an excellent conductivity property, which is essential for the vigorous sputtering of metal electrodes in superstate Sb 2 S 3 solar cells.
To further reveal the effect of different thicknesses of Se layer on the quality of Mo back contact, various dosage of Se powder is used as 0.01, 0.1, and 0.3 g, labeled as 0.01-0.3g Se-layer hereinafter.Similar surface morphologies of Se layer are observed in both Figure 3a S3 (Supporting Information).The device with 0.01 g Se shows J sc of 9.9 mA•cm −2 , V oc of 0.578 V, FF of 36.2%, and a PCE of 2.1%.The PCE of 0.01 g Se-based device significantly outperforms that of 0g-Se based device by 116%, especially for the improved V oc .It implies the Se layer can prevent the penetration by speeding down the Mo particles, as indicated in the above growth model.When the mass of Se increases to 0.1 g, the overall device metrics of the Sb 2 S 3 solar cells are improved obviously, and delivering a PCE of 4.2%, along with J sc of 13.6 mA•cm −2 , V oc of 0.699 V, and FF of 44.1%.However, further increasing the feeding dosage of Se (0.3 g) don't improve device performance, mainly reflecting in a significant reduction in J sc .To further confirm the device difference, as well as repeatability, the statistic boxplots of device parameters collected from 40 devices are exhibited in Figure S7 (Supporting Information).It is worth mentioning that 0.01 g Se devices have a large V oc deviation, which is also related to the coverage of Se layer as discussed in the previous growth model.All device metrics, including V oc , J sc , FF, and PCE gain a remarkable improvement after employing a 0.01 g Se-layer.
To disclose the function of Se layer with various thickness, the cross-sectional morphology of device is revealed in Figure S8 (Supporting Information).As for 0.01 g Se-layer case, the assputtered Mo electrodes before or after annealing are all uniform and flat, featuring with a typical columnar grain, which is similar to the sample without a Se shield.Such a thin Se layer can alleviate the high-energy Mo particles bombardment to a certain extent, but it still cannot be completely prevented, both of which leads to a moderate V oc of 0.577 V.With the increase of Se mass to 0.1 g, there is an amorphous Se layer ≈76 nm appeared between the Mo electrode and the absorption layer after sputtering Mo on it (Figure 3c), which is one sixth the as-evaporation Se layer before sputtering.This again verifies the self-sacrificing function of Se layer during the sputtering process, and endowing an appreciable soft-loading deposition.And, after annealing (Figure 3d), the amorphous Se layer disappears and the back contact of the Sb 2 S 3 device is obviously improved.The reaction of Se at back contact interface will discuss in the coming section.Once the Se mass increases to 0.3 g, however, a thick residual Se layer obviously appears between Sb 2 S 3 and Mo electrode even after post-annealing (Figure S8c,d, Supporting Information), which leads to a large gap between Sb 2 S 3 layer and Mo electrode and then deteriorates the hole transportation toward Mo electrode.The resulted imperfect interface produces a large contact resistance, which pulls down both J sc and FF.This is also verified by the external quantum efficiency (EQE) as shown in Figure 4b.It shows an inferior response of 0.01g-based devices compared with the 0.1 g counterparts in the whole spectrum, owing to the short-circuited channel captures the carriers.As for 0.3 g Se case, the response strength is also significantly lower than that of 0.1 g, because thick Se layer is not conducive to carrier transport between absorber and electrode due to the large gap as described above.In addition, a slight shift of absorption edge is observed when different dosage of Se is used.To clear it, the band gap of Sb 2 S 3 film is determined by plotting the curve of EQE versus Energy.As shown in Figure S9 (Supporting Information), both 0.01 and 0.1 g Se-based samples possess an almost identical band gap of ≈1.69 eV, which is close to that of untreated Sb 2 S 3 film (0g-Se).And the 0.1g-Se case is then referred as Sb 2 S 3 due to the same bandgap comparable with pristine Sb 2 S 3 .But, a massive dose of Se (0.3 g) observably narrows the band gap of Sb 2 S 3 to 1.64 eV.So, the residual Se after sputtering of Mo should be kept in mind as it will sequentially react with Sb 2 S 3 due to the low formation enthalpies of Sb 2 (S,Se) 3 alloy. [23]re, the second role of Se will be studied, according to it also bestows an excellent regulation ability of crystalline structure and bandgap of Sb 2 S 3 by facilely changing S/Se ratios.To verify this, the X-ray diffraction (XRD) curves with respect to Se layer-modified devices are exhibited in Figure S10 (Supporting Information).Clearly, all the diffraction peaks are well in line to the standard pattern of Sb 2 S 3 (PDF#42-1393) except the peaks for FTO and Mo substrate.And the Sb 2 S 3 films show preferred orientation of (211).The close inspection found that the peak of (211) move to a low 2 side with the increase of Se mass (Figure 4c), along with the change of intensity of each diffraction peak, which can be explained by the lattice expansion induced by the replacement of S 2− ion (1.84 Å) with larger Se 2− ion (1.98 Å).Then the resulted Sb 2 Se 3 will improve the back contact interface by up shifting the deep VBM of pristine Sb 2 S 3 surface.Deng et al. also observed the similar phenomenon that the addition of Sb 2 Se 3 layers prepared by rapid thermal evaporation serves as a hole transport layers for superstrate Sb 2 S 3 solar cells, which obviously reduces the back-contact barrier and improve hole collection yield by enjoying the matched energy-level alignments. [2]On the other hand, in order to reveal the impact of post-annealing on the orientation Sb 2 S 3 thin films, the texture coefficient (TC) of the main (hk0) and (hk1) planes are calculated.As shown in Figure 4d, the TC value of the (hk1) in the Sb 2 S 3 thin films increases after inserting 0.1g-Se layers, while the (hk0) shows a opposite trend.This can be explained by top-down growth model induced by the as-generated Sb 2 Se 3 seed layer.In our previous work, we propose two growth model affecting orientation of Sb 2 S 3 films, such as substrate-induced (bottom-up) and surface-induced growth (also denote as top-down growth). [6]ere, we verify the results by comparing it to bottom-up growth mode, as shown in Figure S11 (Supporting Information).It can be seen that although the same precursor is adopted and then the main composition is undoubtedly identical (Figure S11a, Supporting Information), the preferred orientation is distinctly different (Figure S11b, Supporting Information).Applied the topdown growth model, the as-prepared Mo electrode was heated in advance during annealing with the inversion method which leads the preferential formation of the seed layer of Sb 2 Se 3 as validated in Figure S12 (Supporting Information).When the heat treatment time is as short as 10 s, a small amount of Sb 2 Se 3 with preferred (221)-oriented plane is formed, and the strength of Sb 2 S 3 gradually becomes stronger as the annealing duration becomes longer.Hence, the vertical [Sb 4 Se 6 ] n standing on the bare Sb 2 S 3 layer functions as a seed layer which helps to induce the (hk1)-preferred orientation of Sb 2 S 3 films getting away from the bound of substrate.It is well known that the (hk0)-oriented grains manifests (Sb 4 S 6 ) n ribbons lay flat on the substrate, while the counterpart of (hk1)-oriented grains stack more vertically on the substrate which is beneficial to carrier transport along the perpendicular direction.Hence, the alloyed Sb 2 S 3 thin film with the favored (hk1) orientation will render a more effective separation of photo-induced carrier in the corresponding device.In preliminary summary, the introduction of a multifunctional Se layer can kill two birds with one stone.It not only enables the soft-loading of Mo electrode, but also forms an effective Sb 2 Se 3 layer to boost carrier transport, which means that the sputtering of Mo electrode for superstrate structured Sb 2 S 3 solar cells is feasible.

Skillful Design of MoSe 2 Buffer Layer for Sb 2 S 3 Solar Cells
Although Mo can be successfully deposited as electrode for superstrate structured Sb 2 S 3 solar cells with the help of Se layer, the device performance is still unsatisfactory.For example, Figure 4a shows the J-V characteristic curve of the optimized Sb 2 S 3 device with 0.1g-Se layer, displaying an obvious roll-over effect (current saturation at high forward bias) under illumination.It reflects a serious hole-collection barrier at back contact in Mo-attached superstrate Sb 2 S 3 device. [24]Hence, seeking or modification of BCM with a suitable W F are of primary task to build a preferred back contact interface.With regard to substrate chalcogenide device, pre-selenization or post-selenizaton is always conducted to generate a MoSe 2 buffer layer on the surface of Mo bottom electrode, which can significantly improve the interfacial energy band alignments. [25,26]For example, Li et al. introduce an additional selenization process, forming a thin MoSe 2 layer on the Mo surface, which was found to effectively improve the bottom Mo electrode contact and alleviate the roll-over effect of substrate Sb 2 Se 3 thin film solar cells. [27]Inspirited by it, we first examine if the additional Se layer would convert Mo into MoSe 2 during the postannealing.Unfortunately, whether it's 0.01, 0.1, or 0.3 g fed, there is no MoSe 2 peaks observed in the XRD pattern (Figure S10, Supporting Information).The original conception is that the residual Se multifunctional layer may not only form a surficial Sb 2 Se 3 , but also generate MoSe 2 between the Sb 2 S 3 and Mo.However, the limitation of the annealing temperature is fixed as low as 350 °C mainly considering of the low melting point of Sb 2 S 3 . [28]Such a low temperature is not enough to form MoSe 2 , which is again confirmed by the absence of MoSe 2 in control Mo film annealed under Se atmosphere at the same temperature (Figure S13, Supporting Information).So how to successfully synthesize MoSe 2 and incorporate it into the Mo-based Sb 2 S 3 device is very significant for alleviating the roll-over effect.
Herein, we skillfully insert a MoSe 2 buffer layer between Mo and Sb 2 S 3 films.To simplify the sputtering procedure, a preselenized Mo target is prepared by sequentially evaporating Se films on the surface of Mo target and then forming MoSe 2 through post-annealing at the high temperature of 500 °C for 20 min.As shown in Figure 5a of Raman spectra of the preselenized Mo target, the peaks located at 165 and 285 cm −1 corresponds to two weak vibration modes of E 1 g and E 2 g of MoSe 2 while 242 cm −1 is assigned to the strong vibration mode of A 1 g, of MoSe 2 . [29]These results confirm the formation of MoSe 2 on the surface of Mo target.In order to uncover the sequential sputtering of MoSe 2 and Mo, the pre-selenzied target is sputtered on glass for 1 min under the same conditions as Mo target.The Raman spectra of as-sputtered film only appears a main peak of MoSe 2 at 242 cm -1 (Figure S14, Supporting Information) indicating the priority sputtering of surficial MoSe 2 .After sputtering with 10 min, the XPS results display the disappearance of Se element and detect only Mo peak (Figure S15, Supporting Information).It means that during sputtering, an ultra-thin MoSe 2 is first sputtered out followed by Mo electrode deposition.The electrode thickness with different deposition times was measured from cross-sectional SEM and shown in Figure S16 (Supporting Information).It is obviously that the electrode thickness is almost linear with the sputtering duration, and the calculated deposition rate is ≈1.3 Å s -1 .When the layer of MoSe 2 (< 78 nm) is finished sputtering, it is followed by the deposition of the Mo electrode and thus the MoSe 2 /Mo composite is successfully constructed for superstrate Sb 2 S 3 solar cells.
In order to clarify the back contact compound, we also use super glue (≥ 500 g/25 mm) to peel the back electrode of Mo from Sb 2 S 3 surface for XPS testing.The XPS spectra of stripped electrode surface is recorded, and the Se 3d core levels is then shown in Figure 5b,c.It can be seen that the response of Se signal is not observed when the target doesn't experience preselenization, whereas the pre-treated target produces an intense response of Se peak, in which the peak located at binding energies of 54.50 is ascribed to Se 2− of MoSe 2 .The Raman spectra also confirmed the successful formation of MoSe 2 on the surface of Mo electrode as shown in Figure S17 (Supporting Information).Combining with the electrodes with or without MoSe 2 , the band alignment of Sb 2 S 3 planar solar cells is then sketched in Figure 5d, in which the band energy level data are referred from the literatures. [11,28]Compared with Mo electrode without MoSe 2 , the VBM of Mo electrode with MoSe 2 is found to be closer to the VBM of Sb 2 S 3 which observably reduces the hole extraction barrier, as well as nonradiative recombination at the Sb 2 S 3 /Mo interface.Thus, such a cascade band alignment delivers a lower contact resistance at back interface and promotion of hole collection efficiency.It is noticed that although the introduction of MoSe 2 ca be facilely realized by pre-salinization of Mo bottom electrode at high temperature and long duration (i.e., 550 °C for 30 min proposed by Deng et al.) [11] in the substrate-based Sb 2 S 3 solar cells, the Se-modification of Mo top electrode is still full of challenges due to low melting point of Sb 2 S 3 .Hence the pre-selenization of Mo target developed in this work provides a potential regulation path for Mo/Sb 2 S 3 interface.
To assess the impact of band alignment engineering, the device performances are tested by the J-V in Sb 2 S 3 devices with MoSe 2 buffer layer as shown in Figure 5e.Notably, the roll-over effect completely disappears at the high voltage region, indicating the modified back contact electrodes remarkably reduce the hole collection barrier after the introduction of MoSe 2 layer.And the device parameters of the representative devices are shown in Table S4 (Supporting Information), indicating a remarkable improvement of PCE from 4.2% to 5.1%, delivering a V oc of 0.709 V, a J sc of 14.7 mA•cm −2 , and FF of 48.6%.Typically, the device displays a negligible hysteresis, confirmed by the almost same J-V curves measured under both reverse and forward scan directions (Figure S18, Supporting Information).Such facile engineering of the back contact interface leads to an obvious increase in V oc and FF and consequentially boosts PCE.In order to demonstrate the superiority of the MoSe 2 layer, the EQE test result along with the integral curve corresponding to the champion device are also given in Figure 5f.It shows an absorption value up to 87.4% at 540 nm, which is higher than Sb 2 S 3 solar cells without MoSe 2 , which manifests that the additional MoSe 2 is conducive to carrier separation and transport and improves the overall efficiency of the device.To further illustrate it, the J-V test in the dark (Figure 5g) shows the lowest leakage current of 1.2E-7 mA•cm −2 in the MoSe 2 equipped device, indicating a decrease in leakage current compared with that without MoSe 2 .Generally, the leakage current here is the result of the Shockley-Read-Hall recombination of carriers in the depletion layer or interface. [7]As the fabrication procedure of CdS/Sb 2 S 3 junction is identical, the relatively smallest leakage current of the MoSe 2 modified device can be ascribed to the formation of benign back contact interface which is the reason for the increase in V oc .Besides, the dependences of V oc and J sc on light intensity are also recorded to investigate the recombination kinetics of the solar cell.Figure 5h shows the V oc measured as a function of the light intensity and their dependence is linearly fitted by the following Equation (1): [30] Voc = n kT q ln (I) where I is the light intensity, q refers to elementary charge, T is absolute temperature, k is the Boltzmann constant, C is a constant, and n is the ideal factor connected with recombination.
According to the slope of the fitted line, the values of n are determined to be 2.184 and 1.425 for the Sb 2 S 3 solar cells without and with MoSe 2 , respectively.The lower n of the MoSe 2 layer with Sb 2 S 3 device indicates the monomolecular trap-assisted Shockley-Read-Hall (SRH) recombination at the back contact interface can be effectively inhibited.In addition, as depicted in Figure 5i, the function between J sc and light intensity can be studied according to Equation (2): where I is the light intensity and  represents power-law component. [31]It can be concluded from the fitting line that the values of  are determined to be 0.8 and 0.97 for the pristine Sb 2 S 3 and MoSe 2 -basd Sb 2 S 3 solar cells, respectively.This larger  value means more efficient carrier collection.In brief, the favorable modification of back contact interface by an additional MoSe 2 layer significantly improve the performance of Mo-based superstrate Sb 2 S 3 solar cells.
Finally, the normalized PCE of the best device (unencapsulated) as a function of the expose time in ambient conditions with 45% humidity, ≈26 °C room temperature is shown in Figure S19 (Supporting Information).Specifically, the device still maintains 90.5% of the initial PCE after storing for 13 months without encapsulation, indicating superior device stability with Mo electrodes.Such a Mo electrodes, combining with the careful interface engineering, enables to achieve a low-cost and full-inorganic superstrate Sb 2 S 3 solar cell.

Conclusion
In this work, we have successfully applied a low-cost Mo electrode in superstrate structured Sb 2 S 3 solar cells for the first time, breaking the limitation of only noble-Au metal electrode.An additional Se layer is also skillfully used to softly load Mo particles, as well as an adhesive layer to form Sb 2 Se 3 to improve the device back contact due to the deep VBM of Sb 2 S 3 surface.To further overcome the roll-over issue, we artfully design a pre-selenized Mo target and then incorporate it into the device.Finally, the device with a MoSe 2 -modified Mo electrode has achieved an impressive PCE of 5.1%, which is among the highest PCE of Sb 2 S 3 solar cells equipped with metal electrode except for Au.This research provides a potential low-cost Mo electrode for superstrate structured Sb 2 S 3 solar cells, also reveals a universal interface engineering strategy for other low-cost and highly stable electrode deposition.O, AR, Macklin), and selenium (Se, 99.999%, Aladdin), all the chemicals were used as received without further purification.

Experimental Section
Fabrication of P-N Junction: The P-N junction structure of the Sb 2 S 3 planar solar cell was FTO/CdS/Sb 2 S 3 .To build a P-N junction of CdS/Sb 2 S 3 , the FTO conductive glass as device substrate was ultrasonically cleaned with detergent, acetone, anhydrous ethanol, and deionized water in sequence for 1 h.After drying, the substrate was treated by UV ozone cleaner for 20 min.The CdS layer was then deposited using chemical bath deposition method carried out at 65 °C for 15 min in a solution of 100 ml of deionized water, 0.15 mmol of Cd(NO 3 ) 2 •4H 2 O, 9.4 mmol of CH 4 N 2 S, and 0.19 mol of NH 3 •H 2 O.To improve the crystallinity, the asprepared CdS was further treated with CdCl 2 .In this process, as-prepared CdS was spin-coated with CdCl 2 (40 μl, 0.05 g CdCl 2 is soluble in 10 mL methanol) at 2500 rpm for 30 s. Subsequently, the CdS substrate was annealed in air at 380 °C for 5 min.Afterward, the hydrothermal method was adopted for the deposition of Sb 2 S 3 as an absorber layer according to the previously reported recipe.The Sb 2 S 3 film was synthesized by using Na 2 S 2 O 3  O were added into a Teflon tank (100 ml) of an autoclave that contained 80 ml of ultrapure water, which was then heated at 120 °C for 12 h.After the hydrothermal reaction, the sample was washed with ultrapure water and dried on a hotplate at 70 °C for 10 min under atmospheric pressure.Finally, the unannealed P-N junction CdS/Sb 2 S 3 was obtained.
Solar Cells Fabrication: The FTO/CdS/Sb 2 S 3 /Mo structure, Mo was deposited directly on the as-prepared FTO/CdS/Sb 2 S 3 (before and after annealing) by magnetron sputtering (150 °C, 100 W, 2 h) as a back electrode, which could be seen in Figure 1.Finally, the unannealed samples were further annealed at 350 °C for 10 min in a nitrogen atmosphere to improve its quality.To insert a Se functional layer, Se (0.01, 0.1, and 0.3 g) were evaporated on Sb 2 S 3 surface before Mo sputtering.The evaporation conditions were 1 Pa pressure, 10 cm of source-substrate distance, and complete evaporation of powder source for 15 s under 120 A of current.As for MoSe 2 layer, the Mo target was first treated with Se (0.005 g) vaporization (consistent with the above Se evaporation conditions), and then annealed on the hot table under N 2 atmosphere at 500 °C for 20 min.Sequentially, the pre-selenzied Mo was sputtered under the same procedure with the pristine Mo sputtering.
Devices Characterizations: The crystal structure of Sb 2 S 3 film was characterized by X-ray diffraction (XRD, D/Max-rA).The microstructure was observed using high-resolution field emission scanning electron microscopy (FE-SEM).The surface composition was characterized by Xray photoelectron spectroscopy (XPS, Shimadu, AXIS SUPRA+).The J-V curves of the solar cells were characterized using a Keithley 4200-SCS semiconductor characterization system equipped with an AAASAN-EI ELECTRIC solar simulator (XES-40S1).The solar simulator illumination intensity was calibrated by a monocrystalline silicon reference cell (PV Measurements, PVM959) calibrated by the National Renewable Energy Laboratory (NREL).The square resistance of the Mo was observed using four probe test (RTS-8).The EQE of the devices was performed by a solar cell quantum efficiency measurement system (Model QEX10).

Figure 1 .
Figure 1.Schematic diagram of the back contact construction process for a superstrate Sb 2 S 3 solar cell.Right sketch is the magnetron sputtering of Mo electrode, and left sketch is the thermal evaporation of Se layer.

Figure 2 .
Figure 2. a) The cross-sectional images of Mo deposited directly on c-Sb 2 S 3 films.b) Square resistance of Mo electrode deposited on c-Sb 2 S 3 layer or glass substrate.c) The J-V curve of corresponding Sb 2 S 3 solar cell.The cross-sectional images of Mo deposited on d) -Sb 2 S 3 precursor films, and e) following a post-annealing.f) The J-V curve of corresponding Sb 2 S 3 solar cell with an integral post-annealing.

Figure 3 .
Figure 3.The a) surface and b) cross-sectional images of 0.1 g Se coated Sb 2 S 3 thin films.The cross-sectional images of Sb 2 S 3 thin films attached with 0.1 g Se layer and Mo electrode c) before and d) after annealing.e) The growth model of Mo electrode by sputtering with the assist of an additional Se layer.

Figure 4 .
Figure 4. a) J-V and b) EQE curves of the Sb 2 S 3 solar cell modified with 0.01, 0.1, and 0.3 g Se-layer.c) Corresponding XRD patterns ≈2 range 28.5-30.0°show a shift in the peak of (211).d) The texture coefficients (TC) deduced from XRD pattern.
,b and Figure S6 (Supporting Information), exposing of a porosity trait.While different thicknesses of Se layers (190-800 nm) are detected from the cross-sectional SEM images with feeding different dosage of Se powder source.To optimize the Se layer thickness, the J-V curves of the Sb 2 S 3 solar cells with 0.01, 0.1, and 0.3 g Se-layer were measured under simulated AM 1.5 (100 mW•cm −2 ) sunlight, as shown in Figure 4a.More device metrics can be found in Table

Figure 5 .
Figure 5. a) Raman spectra of the pre-selenization Mo target.The XPS curves of the Mo electrode peeled by tape, which is sputtered from b) pristine c) pre-selenized Mo target.d) The band alignment diagram of the Sb 2 S 3 device.The e) J-V curve and f) EQE spectra of the champion Sb 2 S 3 device with pre-selenization Mo target.g) Dark J-V characteristics of the Sb 2 S 3 thin-film solar cells, and h) V oc and i) J sc as a function of light intensity.