The basic structure of substrate Sb₂Se₃ solar cells was as follows: Soda lime glass (SLG)/Mo/Sb₂Se₃/CdS/ZnO/ITO/Ni-Al. Mo coated soda lime glass (SLG) substrates measuring 7.5 cm² × 2.5 cm² were used in this study. NiO or MoO _x HTLs were deposited between the Mo electrode and Sb₂Se₃. Thin HTL films of 15 nm thickness were deposited using e-beam evaporation. 500 nm thick Sb₂Se₃ layers were prepared by TE of crystalline/powder Sb₂Se₃ source material (Alfa Aesar, 99.99%) at a deposition rate of ∼15 Å/s. The substrates were maintained at a temperature of 300°C throughout the deposition. The Sb₂Se₃ films were subsequently subjected to a heat treatment at 300°C for 30 min in Ar atmosphere in a tube furnace to promote recrystallisation. For the CSS Sb₂Se₃ films, a compact seed layer was grown at 0.05 mbar N₂ for 5 min with a source temperature of 350°C, followed by a 30 min growth step at 13 mbar and a source temperature of 450°C to produce a compact and highly orientated grain structure. The substrate was then rapidly cooled with N₂. An n-type CdS buffer layer (∼60 nm) was deposited by chemical bath deposition followed by DC-pulsed sputtering deposition of an i-ZnO (∼35 nm) layer plus a transparent conductive window layer ITO (∼200 nm). Front contact grids comprising Ni (∼50 nm) and Al (∼1,000 nm) were deposited through a shadow mask by e-beam evaporation. Finally, 0.16 cm² cells were defined by mechanical scribing on each substrate.

Superstrate Sb₂Se₃ solar cells have the following configuration: SLG/ITO/CdS/Sb₂Se₃/Au with NiO or MoO _x HTLs deposited between the metal contact and Sb₂Se₃ absorber. The ITO layer was deposited by DC-pulsed sputtering and Sb₂Se₃ layers were grown by TE and CSS as detailed above. Finally, Au back contacts with an area of 0.07 cm² were deposited through a shadow mask by e-beam evaporation.

The crystal structures of Sb₂Se₃ were characterised by X-ray diffraction (XRD) with Cu Kα1 (1.54056 Å) radiation (Rigaku SmartLab SE). The surface morphology and cross-sectional images of Sb₂Se₃ films were taken by scanning electron microscopy (SEM, Tescan Mira 3 FEG-SEM). Optical spectroscopy measurements were performed using a Shimadzu UV-2600 spectrophotometer fitted with an integrating sphere. Kelvin probe force microscopy (KFPM) measurements were done using a KP Technology KP020 single point kelvin probe system fitted with a standard 2 mm Au tip.

Current-density vs. voltage (J-V) measurements of Sb₂Se₃ thin film solar cells were performed using an Abet Technologies solar simulator at 1-sun (100 mW/cm²) illumination equivalent to air mass 1.5 global spectrum with light power density calibrated using a Si reference cell.

Device simulation was carried out for both substrate and superstrate configuration Sb₂Se₃ solar cell using Solar Cell Capacitance Simulator (SCAPS 1-D), which is based on the solutions to Poisson’s equation and continuity equation for electrons and holes in the vertical heterostructure of multilayer thin film PV device (Burgelman et al., 2000). The input parameters of the solar cells were defined with the Sb₂Se₃, HTL and electron transport layer (ETL) semiconducting properties, including experimentally determined bandgaps, electron affinity, density of states (Zeng et al., 2016), mobility of charge carriers (Chen et al., 2017), acceptor/donor concentrations (Wang et al., 2015), and defect state density (Leijtens et al., 2016). Defects were introduced at the Sb₂Se₃/CdS interface to simulate realistic device performance.

Simulation analysis using SCAPS software was implemented to evaluate the performance of reference substrate and superstrate Sb₂Se₃ solar cells and those incorporating MoO _x and NiO as HTLs, subsequently referred to as samples Ref, MoO _x and NiO, respectively (see Table 1 for film properties). Figure 2 shows the J-V curves and corresponding box plots of J-V parameters of both Sb₂Se₃ device configurations with incorporated HTLs. Regarding the substrate devices, all device parameters are improved, with the exception of V _oc which shows a slight decrease for devices with a HTL (down from 0.423 V for the reference device to 0.408 and 0.411 V for MoO _x and NiO devices, respectively). However, devices with MoO _x HTL show evidence of roll-over behaviour. The roll-over phenomenon, which occurs near the V _oc in a light J-V curve, is due to Schottky energy barrier formed at the absorber/metal interface at a solar cell back contact (Eisenbarth et al., 2011; Hädrich et al., 2011). It acts as a reverse biased diode when the main junction is forward biased, blocking carrier transport for increasing forward bias, resulting in roll-over behaviour in light J-V characteristics. The baseline J _sc in the reference device was 29.9 mA/cm², rising to 31.2 and 31.3 mA/cm² in MoO _x and NiO devices, respectively. Addition of HTL films to the reference device demonstrated a notable increase in FF for substrate devices. The FF in the reference device was 47.0%, rising to a maximum of 55.6% and 56.0% in the MoO _x and NiO devices, respectively. The increase in J _sc and FF of devices with integrated HTL materials directly translates into improvements in power conversion efficiency, PCE [η = 5.9% (Ref), 6.7% (MoO _x ) and 7.2% (NiO)]. The current-blocking energy barrier at the back contact of the MoO _x substrate device could explain the lower PCE in comparison to the device with a NiO HTL. It is important to note that the results shown are not representative of the maximum conversion efficiencies that may be achieved with Sb₂Se₃, as we are focusing solely on the effect of the HTL, while using currently available materials parameters.

For superstrate Sb₂Se₃ solar cells, devices with an incorporated HTL showed an increase in J _sc of around 8% from 31.5 mA/cm² observed in the reference device to 33.5 and 34.0 mA/cm² in the devices with a MoO _x and NiO HTL, respectively. As a result of the improvement in J _sc , the PCE of solar cells with a HTL increased to 7.5% (MoO _x ) and 8.0% (NiO) from the reference value of 7.3%. Interestingly, no roll-over was seen in the J-V curve for the MoO _x device which could be related to the use of Au as metallic back contact rather than Mo in the substrate devices. The work function (WF) of a metal employed as a rear contact on a PV device plays an important role in facilitating hole extraction at the contact (Fleck et al., 2020). Typically, Au is reported to have a WF of 5.10 eV (Michaelson, 1977) and Mo has WFs ranging from 4.50–4.95 eV, depending on the preferred crystal orientation of the metal (Green, 1969; Michaelson, 1977; Hölzl and Schulte, 1979). To illustrate the effect of back contact metal WF on substrate/superstrate Sb₂Se₃ device performance, Figure 3 shows the dependence of J-V parameters on the WF of Mo and Au metals. It is apparent that the J-V parameters of all substrate devices are sensitive to variations in the value of Mo WF. In the Ref and MoO _x substrate devices, V _oc decreases monotonically with Mo WF where a significant drop is observed from 0.432 V to 0.422 V at WF 4.95 eV to 0.036 V and 0.093 V at WF 4.50 eV for Ref and MoO _x devices, respectively. This is a clear indication of an increasing back contact barrier with decreasing Mo WF. This phenomenon has been observed experimentally in Sb₂Se₃ solar cells previously (Liu et al., 2014; Li et al., 2017). The V _oc in the NiO device is less affected by the Mo WF, reducing from 0.422 V at WF 4.95 eV to 0.319 V at WF 4.50 eV. A similar trend is seen in J _sc , FF and η parameters for the substrate devices. However, a low Mo WF of 4.50 eV causes a notable decrease in FF of the MoO _x device (12.3%), compared to the Ref and NiO devices (26.0% and 32.4%).

In order to understand the improvement of the device performance with the introduction of HTLs, it is necessary to consider the energy band alignment at the interfaces at the back of the PV devices. Figure 4 shows the simulated energy band diagrams of substrate and superstrate Sb₂Se₃ devices incorporating NiO and MoO _x HTLs. Due to a small electron affinity (EA = 1.46 eV (NiO), 2.05 eV (MoO _x )) and large band gaps (E _g ∼3.80 eV (NiO), 3.50 eV MoO _x x)) in both HTL materials, a large potential energy barrier is formed at the back contact, reflecting electrons. This barrier minimises carrier recombination at the back interfaces with Sb₂Se₃ and improves conductivity at the back electrode. However, it is apparent that a non-negligible hole barrier of 0.26 and 0.29 eV is formed at the MoO _x /Sb₂Se₃ interface of the substrate and superstrate devices, respectively, which can manifest as J-V roll-over behaviour seen in the simulated MoO _x substrate device. Thus the SCAPS simulations indicate the incorporation of a MoO _x or NiO HTL into substrate and superstrate configuration Sb₂Se₃ solar cells increases device performance compared to a standard solar cell by the introduction of a high energy barrier to electrons at the rear absorber/metal interface.

100 nm thick films of MoO _x and NiO were deposited on SLG at room temperature to facilitate characterisation of the HTLs. Figure 5 shows surface morphology SEM images of the respective HTLs. The MoO _x film exhibits an amorphous, flake-like structure in comparison to a compact crystalline morphology observed in the NiO film. XRD patterns in Figure 6 confirm the amorphous and crystalline nature of the MoO _x and NiO films, respectively. All the diffraction peaks in the NiO thin film were identified and indexed to cubic NiO (JCPDS number 04-0835) and no diffraction peaks of other impurity phases were observed.

Supplementary Materials S1A shows the spectral transmittance and reflectance of the NiO and MoO _x films on SLG. Both HTLs are highly transparent in the visible and near-infrared wavelength region and their transmittance falls sharply at ultraviolet wavelengths. However, the amorphous MoO _x film has slightly lower transmittance/higher reflectance in the sub-600 nm wavelength region compared to the crystalline NiO film. The bandgap energy (E _g ) of the HTL films was calculated by extrapolation of the linear region of the Tauc plot to the x-axis, according to the relation (Tauc et al., 1966): α h ν 2 = A h ν − E g (1) where α is the absorption coefficient of the semiconductor material, h is Planck’s constant, ν is the frequency of the electromagnetic radiation and A is a constant of proportionality. The estimated E _g values of NiO and MoO _x films are 3.95 and 3.85 eV, respectively (see Supplementary Materials S1B). A HTL film thickness of 15 nm was incorporated into the superstrate/substrate device to ensure a conformal coating of the HTL. A HTL requires a thickness sufficient to preserve the desired material properties and not impede charge transport considerably which would detrimentally increase series resistance in the finished devices.

TE and CSS deposition techniques were employed for Sb₂Se₃ film growth on SLG/ITO/CdS superstrates. For TE, the SLG/ITO/CdS superstrates were heated to 300°C prior to Sb₂Se₃ deposition in order to promote the growth of preferred (hk1) crystal orientations while minimising (hk0) orientations (Zhou et al., 2015) (hk0) planes, specifically (120), have been found to be detrimental to carrier transport (Guo et al., 2018; Wen et al., 2018; Li et al., 2019). The (hk0)-oriented Sb₂Se₃ nanoribbons are stacked parallel to the ITO/SLG superstrate where conductivity is inhibited by electrically insulating VdW bonds between the stacked nanoribbons. A seed layer is used in Sb₂Se₃ films deposited via CSS. This seed layer has a high density of nucleation points for the second stage of growth during the CSS process, which improves uniformity, raising the average efficiency of devices (Hutter et al., 2018a). Transmittance and reflectance data for a representative TE Sb₂Se₃ film was used to determine the E _g from a Tauc plot, which gave a E _g value of 1.17 eV in good agreement with (Birkett et al., 2018), see Supplementary Materials S2A,B. XRD patterns for Sb₂Se₃ films deposited by TE and CSS are shown in Figure 7A. The peaks in both XRD patterns are sharp and well resolved indicating the polycrystalline nature of the Sb₂Se₃ thin films. The lattice planes are cross-referenced to JCPDS card no. 15-0861 confirming the formation of orthorhombic Sb₂Se₃ with space group Pbnm. Both XRD patterns show similar characteristics, exhibiting strong (211) and (221) peaks with minimal contributions from (hk0) planes. Figures 7B–E shows the top and cross-sectional SEM images of Sb₂Se₃ thin films deposited by TE and CSS. The different growth techniques result in contrasting Sb₂Se₃ film morphologies. TE produces Sb₂Se₃ films of uniform thickness of ∼500 nm and densely packed grains, confirming the good crystallinity of the films, consistent with the XRD results (Figures 7B,D). However, this deposition method did not form a conformal coating of the Sb₂Se₃ film across the entire superstrate with the presence of pinholes observed, see Supplementary Materials S3A.

Conversely, CSS-grown Sb₂Se₃ films have a rough surface morphology with exceptionally large grains in comparison to the TE films and the grains extend the full depth of the layer. Larger grains are a prerequisite for better device performance as charge mobility is faster along the Sb₂Se₃ ribbons than hopping between the ribbons (see Figures 7C,E). The CSS films also showed a degree of porosity but not to the extent observed in the TE films, Supplementary Materials S3B. The presence of pinholes in the Sb₂Se₃ films is detrimental to device performance as shunting pathways may be formed upon subsequent deposition of the Au back contact (Hutter et al., 2018b).

J-V measurements under 1-sun illumination (100 mW/cm²) were performed on Sb₂Se₃ devices in the standard superstrate configuration and devices incorporating MoO _x and NiO HTLs. The light J-V curves were fitted using a single diode model to extract the values of series (R _s ) and shunt (R _sh ) resistances. Figure 8 compares the statistical distribution of the key PV parameters for these devices, where a minimum of 10 cells of each device type were measured. On average, there was a slight increase in V _oc when a NiO HTL was incorporated into the CSS device structure. Using a NiO HTL layer increased V _oc to 0.226 V from values of 0.201 and 0.186 V for Ref and MoO _x devices, respectively. The mean J _sc of NiO cells was also enhanced to 15.94 mA/cm² compared to Ref (11.34 mA/cm²) and MoO _x (10.54 mA/cm²) cells despite a slightly lower average FF in the NiO devices. This translates into a higher mean NiO CSS device efficiency of 1.01% with Ref and MoO _x devices achieving efficiencies of 0.71 and 0.59% respectively. Notwithstanding the higher average R _s (2.6 Ωcm²) and lower R _sh (74 Ωcm²) values for NiO CSS solar cells compared to Ref (R _s = 2.9 Ωcm², R _sh = 119 Ωcm²) and MoO _x (R _s = 1.3 Ωcm², R _sh = 167 Ωcm²) cells, using NiO as a HTL increases performance by boosting J _sc in CSS Sb₂Se₃ superstrate devices compared to the standard and MoO _x based devices.

The average J-V parameters of TE Sb₂Se₃ superstrate devices followed a similar trend to those observed in the CSS devices [V _oc : 0.209 V (Ref) → 0.214 V (MoO _x ) → 0.288 V (NiO), J _sc : 1.94 mA/cm² (MoO _x ) → 8.05 mA/cm² (Ref) → 13.48 mA/cm² (NiO) ⇒ η: 0.10% (MoO _x → 0.72% (Ref) → 1.38% (NiO)]. It is worth noting that the mean FF of the Ref TE cells (38.0%) was higher in relation to the cells with a HTL (27.2% MoO _x , 34.6% NiO). This correlates to an increase in R _sh of 382 Ωcm² in Ref samples from R _sh values of 349 Ωcm² and 154 Ωcm² measured in MoO _x and NiO cells, respectively. In TE superstrate device configuration, the thin MoO _x film appears to form a more resistive layer compared to Ref and NiO devices (R _s : 10.8 Ωcm² MoO _x , 4.1 Ωcm² Ref and 6.6 Ωcm² NiO). Thus, overall device performance in MoO _x based solar cells is negatively impacted by low J _sc and high R _s which could be related to the amorphous nature of the MoO _x thin film and the presence of a current-blocking barrier at the back contact highlighted in device simulations. Despite lower FF in NiO based solar cells, device efficiencies exceed those of Ref and MoO _x TE devices due to improvements in V _oc and J _sc showing the benefit of using NiO as a HTL in superstrate Sb₂Se₃ solar cells.

Figure 9 shows the XRD patterns of substrate Sb₂Se₃ thin films deposited via TE and CSS. All diffraction peaks are in good agreement with the orthorhombic Sb₂Se₃ (JCPDS 15-0861), which presents in the form of (hk0), (hk1) or (hk2). No diffraction peaks of other impurity phases were observed. TE Sb₂Se₃ films on Mo and Mo/MoO _x substrates show (020) and (120) peaks compared to all other Sb₂Se₃ films. The presence of (020) and (120) crystal orientations in thin Sb₂Se₃ films adversely affects PV device performance (Leng et al., 2014; Yuan et al., 2016; Li et al., 2017). However, when using a NiO HTL in TE Sb₂Se₃ films, it can be observed that the intensity of the diffraction peaks of Sb₂Se₃ is dominated by (221) and (211) crystal plane orientations. Furthermore, when using the Mo/NiO substrate, Sb₂Se₃ film shows an increased peak intensity for the (002) orientation. Since h and k miller indices have a zero value, it indicates that the (Sb₄Se₆) _n ribbons grow perpendicular to the substrate surface (Li et al., 2019). For CSS Sb₂Se₃ films, Ref and MoO _x samples demonstrate a higher (002) peak intensity than NiO.

Figures 10, 11 show SEM images of Sb₂Se₃ films on Mo-coated SLG deposited by TE and CSS methods, respectively. The top-down SEM images of the TE films (Figures 10A–C) show a difference in morphology depending on the presence of the underlying HTL. The MoO _x sample exhibits larger Sb₂Se₃ grains than the Ref sample and the presence of pinholes in both samples is patently obvious. On the other hand, the Sb₂Se₃ grains in the NiO sample appear more angular in nature although pinholes are still present in the film. The dissimilarity in morphology is emphasised in SEM cross-section images of the TE Sb₂Se₃ films (Figures 10D–F). Voids at the absorber/Mo interface are apparent in the Ref TE sample whereas the MoO _x sample shows a homogenous film with large grains. For the NiO sample, the Sb₂Se₃ grains appear column-like with no voids at the Mo interface. The top-down SEM image of all types of CSS Sb₂Se₃ thin films (Figures 11A–C) show significantly larger grains compared to the TE films. However, Sb₂Se₃ film in the Ref sample is on average thicker (∼1,000 nm) than the MoO _x (∼550 nm) and NiO (∼700 nm), see Figures 11D–F. The NiO sample also has a smoother surface topography.

To quantify the difference in orientations between the substrate Sb₂Se₃ thin films, the texture coefficient (TC) of diffraction peaks of the samples was calculated based on the following equation (Zoppi et al., 2006): T C h k l = I h k l I 0 h k l 1 N ∑ N I h k l I 0 h k l (2) where I _(hkl) is the measured peak intensity of (hkl) plane and I ₀ (hkl) the intensity in the standard XRD pattern. N is the total number of reflections considered for the calculation. A diffraction peak with a relatively large TC value ( > 1) indicates a preferred orientation of the grain along this direction. Figure 12 shows the TC for Sb₂Se₃ thin films with HTLs deposited by (A) TE and (B) CSS. It is apparent from Figure 12 that NiO HTL plays a critical role in eliminating the detrimental (hk0) planes in the TE samples and at the same time, significantly increases absorber growth in planes, i.e., (211), (221) that are perpendicular to the substrate surface. This further supports the enhanced device performance in solar cells when NiO is used as the HTL. In CSS samples, this templating effect of HTLs is not observed as no (hk0) planes are grown in the Ref and MoO _x samples. MoO _x increases the growth of favoured crystal planes including (211), (221), and (002) compared to the Ref substrate sample whereas NiO appears to inhibit the growth of the preferential planes, which may be attributed to rendering the seed layer ineffective but further study will be required to fully understand the reason.

Figure 13 shows the variation in J-V parameters measured for a minimum batch size of 10 Sb₂Se₃ solar cells in substrate configuration deposited by TE and CSS incorporating HTLs. The use of MoO _x /NiO HTLs adversely affects all device parameters in CSS-based solar cells. This can be explained by lower average R _sh values of 55 Ωcm² and 47 Ωcm² determined for MoO _x and NiO device types, respectively, compared to 172 Ωcm² in the Ref devices. The reason for the reduction in R _sh of the substrate devices with a HTL is not obvious. Only working TE devices were achieved by incorporating a NiO HTL, which can be attributed to the templating effect of the NiO film which eliminated the deleterious (hk0) crystal planes and promoted the growth of preferred (211) and (221) planes. As highlighted in device simulations, the performance of substrate Sb₂Se₃ solar cells can be dependent on the WF of Mo back contact (see Figure 3). Mo metal typically has a WF in the range of 4.5–4.95 eV. KPFM measurements on Mo coated SLG prior to Sb₂Se₃ deposition determined the Mo WF to be 4.6 eV. According to simulations, device performance of Ref and MoO _x substrate devices is severely impacted at the observed Mo WF. Simulated NiO device performance is affected to a lesser degree.

R _s values for both TE and CSS substrate Sb₂Se₃ devices were significantly higher than their superstrate counterparts and had a detrimental effect on overall substrate device performance. This could be related to a non-optimal sputtered ITO layer in the substrate devices with a typical sheet resistance of ∼ 35 Ω/□ (Qu et al., 2016) compared to commercially available ITO-coated glass slides used in superstrate devices with sheet resistances of 8–12 Ω/□ (Sigma Aldrich).

Experimentally determined device parameters, such as R _s and R _sh and apparent doping density (N _A ) of the Sb₂Se₃ absorber, were incorporated into SCAPS simulations of TE/CSS Sb₂Se₃ devices in superstrate/substrate configurations in order to replicate the observed behaviour of the fabricated devices. For an accurate representation of the fabricated cells, the N _A value for the Sb₂Se₃ absorber in the CSS devices was set to a value previously determined for the same CSS deposition process used in this study with a Sb₂Se₃ absorber thickness of 1 μm (Phillips et al., 2019). An experimentally determined N _A value for a typical 500 nm thick TE Sb₂Se₃ absorber was used in TE device simulations (see Table 1 for TE/CSS Sb₂Se₃ film properties). Figure 14 shows device performance of the simulated TE/CSS Sb₂Se₃ devices with experimentally determined R _s , R _sh and N _A values. Similar trends are observed for all device parameters of the simulated and fabricated solar cells in both device configurations indicating the simulated devices are a reasonable representation of actual Sb₂Se₃ solar cells (see Figures 8, 13). However, in superstrate configuration, simulations overestimate all J-V parameters, indicating factors other than R _s , R _sh and N _A are influencing device performance. Material properties such as carrier lifetimes, defects and band tails states have been cited as having a detrimental effect on overall device performance (Chen and Tang, 2020). In that work, a number of bulk defects in Sb₂Se₃ were identified with energy levels within the Sb₂Se₃ bandgap ranging from 0.18–0.94 eV above the valence band maximum. For simulation purposes, a mid-gap donor defect (0.62 eV) was introduced for the Sb₂Se₃ bulk to reproduce realistic device performance (Wen et al., 2018; Ma et al., 2020). Chen and Tang (2020) also highlighted significant recombination occurring at the n-p interface which severely impacts both V _oc and J _sc . The presence of additional Sb₂Se₃ bulk defects and increased absorber/buffer interface defect concentration could account for the differences observed between the simulated and fabricated devices studied here.

In addition, it is worth noting actual superstrate devices which incorporate a MoO _x HTL under-perform in relation to standard simulated superstrate devices (see Figure 8). This decrease in performance is not observed in the fabricated substrate Sb₂Se₃ solar cells with a MoO _x HTL. This discrepancy can be accounted for by different processing conditions applied during deposition of substrate and superstrate devices. During deposition of Sb₂Se₃ layer on SLG/Mo/HTL substrate, the substrate temperature is maintained at 300°C which is sufficient to crystallise the MoO _x film, see Supplementary Materials S4. The crystallised MoO _x film consists of a mixture of MoO₂, MoO₃ and intermediate reduced oxide phases. The phase composition affects the electronic and optical properties of the MoO _x film, with MoO₂ content lowering the resistivity, transmittance and bandgap (Inzani et al., 2017). Simulations also show a roll-over in the J-V curves for superstrate Sb₂Se₃ devices in both configurations (see Figure 15), indicating the presence of a barrier to carrier transport at the back contact seen in simulated energy band alignments as previously discussed (Figure 4).