Antimony selenide (Sb₂Se₃) has drawn much attention in the thin-film solar cells community because of its attractive properties such as appropriate band gap, high absorption coefficient, single-phase structure at room temperature, inert grain boundaries, low cost, and low toxicity^[1-27]. Moreover, there are only two elements in Sb₂Se₃, which means that its stoichiometric ratio can be controlled more easily than that of the ternary or quaternary compounds such as CuInGaSe and CuZnSnS(Se). The theoretical power conversion efficiency (PCE) has been predicted to be 26.87% for the Sb₂Se₃ solar cell^[2]. In the past decade the noticeable performance improvements have been achieved. The state-of-the-art Sb₂Se₃ solar cells have reached efficiencies of ≥10%^[3-5]. Various approaches have been developed to fabricate the Sb₂Se₃ solar cells, such as vacuum thermal evaporation^{[6, 7]}, rapid thermal evaporation (RTE)^{[8, 9]}, closed-space sublimation (CSS)^[10-12], vapor transport deposition (VTD)^[13-16], magnetron sputtering (MS)^[17-21], pulsed laser deposition^[22], hydrothermal deposition (HD)^{[3, 4, 23]}, chemical bath deposition (CBD)^[5], and spin coating^{[24, 25]}. Among these techniques, vacuum thermal evaporation is one of the techniques that is suitable for large-scale production with a moderate cost. Nevertheless, the PCE of the Sb₂Se₃ solar cells fabricated by the conventional vacuum evaporation are lower than 5%^{[6, 7, 28-30]}, which lags behind that of the device fabricated by the VTD, CSS, RTE, MS, and hydrothermal process.

In solar cell, the extraction and the recombination of the photogenerated charge compete against each other. Therefore, it is of crucial importance to suppress the charge recombination in the cell. In general, it is difficult to avoid the radiative inter-band recombination. What we can do, however, is to inhibit the trap-assisted Shockley–Read–Hall (SRH) recombination that occurrs in the bulk and at the interface. Hence, the interface passivation is one of the important research directions. For example, Al₂O₃, SiO₂, and SiN_x have been used to passivate the electrode interface in the Si solar cells^{[31, 32]}. 4, 4', 4"-tris(carbazol-9-yl)triphenylamine (TCTA) is a small organic molecular, which is often used as the host layer in organic electroluminescence diodes (OLEDs). TCTA has a wide optical bandgap of 3.3 eV. Its lowest unoccupied molecular orbital (LUMO) and highest occupied molecular orbital (HOMO) are located at −2.4 and −5.7 eV, respectively^[33]. The TCTA film fabricated by the vacuum thermal evaporation is very smooth and pinhole-free.

In this paper, we fabricated a Sb₂Se₃ solar cell by the vacuum thermal evaporation, and TCTA was used to passivate the indium tin oxide (ITO)-Sb₂Se₃ interface. By introducing a 3 nm-thick TCTA layer, the open-circuit voltage (V_oc) is improved from 0.36 to 0.42, and the PCE is raised from 3.2% to 4.3%.

Experimental details are given in Supplementary material. Thermogravimetric analysis showed that TCTA experienced a 0.5% weight loss at 440 °C. Under the vacuum condition, the decomposition temperature of TCTA is distinct from that under atmospheric pressure. Therefore, we cannot judge whether or not the TCTA can tolerate the vacuum annealing process at 280 °C. To gauge the impact of the TCTA on the Sb₂Se₃ film, a series of measurements were performed. Fig. S1 shows the absorption spectra of the Sb₂Se₃ film deposited on the glass with and without a TCTA layer. As can be observed, both of the films exhibit a broad absorption from the ultraviolet to the near-infrared region, and there is no discernible difference between the absorption spectra. A Tauc plot is given in the inset of Fig. S1, from which an optical bandgap of 1.1 eV was obtained. This agrees well with the previously reported results^{[11, 18, 22]}.

Fig. S2 shows the X-ray diffractions (XRD) patterns of the Sb₂Se₃ films (150 nm) deposited on the ITO substrate with and without a TCTA layer. As can be seen that both of the films show the (020) preferred orientation. The intensity of the (020) diffraction peak is increased from 1636 to 1789 by introducing a TCTA layer, which suggests a very slightly improvement of the crystal quality. Fig. S3 shows the top-view and the cross-sectional scanning electron microscopy (SEM) images of the films. It can be seen that there is no dramatical change after introducing a TCTA layer. Both of the films show a clear grain boundary.

To investigate the impact of the TCTA layer on the ITO-Sb₂Se₃ interface, we performed the X-ray photoelectron spectroscopy (XPS) measurements. The XPS peaks of In 3d_5/2 were given in Supplementary material as Fig. S4. As can be seen, there are two components in the In 3d_5/2 peak for ITO surface. The peaks located at 446.0 and 444.9 eV are assigned to In⁶⁺ and In⁵⁺, respectively. This is consistent with the view that a large amount of oxygen vacancies are presented in ITO, and a portion of indium are incompletely oxidized. However, for the ITO/Sb₂Se₃ (2 nm), the component of In⁶⁺ disappeared, and a new component appeared at 444.0 eV, which was ascribed to In⁴⁺. This result indicates that a portion of indium were reduced to lower valence states after an annealing treatment. In conjunction with the following results of the Sb 3d, it is speculated that a fraction of O were bonded with Sb. Fig. 1 shows the features of Sb 3d_3/2. It can be seen that there is only one component for the 50 nm-thick Sb₂Se₃ film. However, there are two components for the thinner Sb₂Se₃ films. The peak located at around 538.1 eV is originated from the Sb–Se, and the peak located at around 539.6 eV is ascribed to the Sb–O. During the annealing process, the chemical reaction occurred at the ITO–Sb₂Se₃ interface, which give rise to the Sb–O. The ratio of the Sb–O component is decreased from 41% to 17% by introducing a TCTA layer. This result indicates that the TCTA layer can inhibit the chemical reaction occurred at the ITO–Sb₂Se₃ interface, which may reduce the interface defect density. For the case of the thicker Sb₂Se₃ film, the surface is pristine Sb₂Se₃, and there is only one peak of Sb-Se. However, when the Sb₂Se₃ film is very thin, the surface composition is very complicated due to the diffusion and chemical reaction during the annealing treatment. From our previous secondary ion mass spectra measurements, it is known that there are Se, Sb, In, Sn, O, and Na (from glass substrate) elements on the surface^{[34, 35]}, and the surroundings of Sb–Se is very different from that for the thicker film. Therefore, the peak location of Sb–Se is different from that of the thicker film.

To access the influence of the TCTA layer on the charge-injection barrier we prepared the device with the architecture of ITO/TCTA (3 nm)/Sb₂Se₃ (50 nm)/NPB (5 nm)/Al, where, NPB is the N, N′-bis(naphthalen-1-yl)-N, N′-bis-(phenyl)benzidine. The device architecture and the molecular structure of NPB are shown in Fig. 2. The control device without a TCTA layer was also fabricated for comparison. The LUMO and the HOMO levels of NPB are located at −2.3 and −5.4 eV, respectively, which means that it can block both the electron and the hole injection from the Al electrode. Therefore, the charges in the device are mainly injected from the ITO electrode at lower bias voltage. Fig. 2 shows the current‒voltage (J‒V) characteristics of the device in the darkness. Here, the positive and negative bias were applied to the ITO electrode. Both the forward and the reverse current densities are decreased by introducing a TCTA layer, which suggests that TCTA can block both the electron and the hole injection from the ITO electrode. This can be ascribed to the higher LUMO level (−2.4 eV) and the lower HOMO level (−5.7 eV) of the TCTA.

Next, we prepared the solar cell with an architecture of ITO/TCTA (3.0 nm)/Sb₂Se₃ (50 nm)/C₆₀ (5 nm)/Alq₃ (3.0 nm)/Al, where Alq₃ is the tris(8-hydroxyquinolinato) aluminum. The device architecture and the molecular structures are shown in Fig. 3(a). The control device without a TCTA layer was also prepared for comparison. Here, C₆₀ served as an electron transport layer. An ultra-thin Alq₃ layer was used to passivate the cathode interface. The optimal thickness of the TCTA layer was 3.0 nm, and the increase in thickness lead to a deteriorated performance because of its resistance and hole-extraction barrier. Fig. 3(b) shows the J‒V characteristics of the devices under one-sun illumination (AM 1.5 100 mW/cm²), from which the short-circuit current density (J_sc), V_oc, fill factor (FF), and PCE were extracted. These parameters are listed in Table 1. By introducing a TCTA layer, the J_sc is increased from 18.02 to 19.75 mA/cm², the V_oc is raised from 0.36 to 0.42 V, the FF is elevated from 47% to 53%, and the PCE is enhanced from 3.2% to 4.3%. Fig. 3(c) shows the J‒V characteristics of the device measured in the darkness. It can be seen that the reverse current is about one order of magnitude smaller than that of the control device, which indicates a reduction in leakage current. The shunt resistance (R_sh) and the series resistance (R_s) were obtained from the differential resistance at zero bias and 0.9 V, respectively. The ideality factor (n) and the reverse saturation current density (J₀) can be obtained from the slope and the intercept of the curve of ln(J−V/R_sh) versue V−JR_s, respectively. Fig. S5 shows the plots of dV/dJ versue V and ln(J−V/R_sh) versue V−JR_s, which are extracted from the J‒V characteristics measured in the darkness. The fit parameters are summarized in Table 1. As can be seen, the R_sh is increased from 0.18 to 23.8 kΩ·cm² by introducing a TCTA layer, which suggests a suppression of the leakage current. The V_oc can be described as

1Voc=nkBTqln(JscJ0−1),

where k_B is the Bolzmann constant, and T is the absolute temperature^[36]. From Eq. (1) it is known that a lower J₀ means a higher V_oc. By introducing a TCTA layer, the J₀ is decreased from 1.0 × 10⁻² to 8.1 × 10⁻⁴ mA/cm², and the ideality factor is reduced from 2.05 to 1.61. The decrease of the J₀ is consistent with the raised V_oc, and the reduction in ideality factor suggests a suppression of the trap-assisted SRH recombination.

To reveal the origination of the performance enhancement we investigated the capacitance characteristics of the device. Fig. 4(a) shows the capacitance‒frequency (C‒f) characteristics of the device. It can be seen that both of the devices show a capacitance decrease as the frequency increases. The reason is that less and less charge states can response to the alternating current (AC) voltage as the frequency increases. In addition, the device capacitance is smaller than that of the control device. This result indicates a reduction of the defect state density, which is mainly caused by the anode passivation. According to the abrupt-junction approximation, the device capacitance can be described as

21C2=2(Vbi−V)qA2ε0εrNa,

where V_bi is the built-in potential, A is the device area, ε₀ is the vacuum permittivity, ε_r is the relative dielectric constant, and N_a is the carrier density^[37]. We can obtain the built-in potential from the intercept of the Moot-Schottky plot. Fig. 4(b) displays the C⁻²‒V characteristics of the device measured at a frequency of 100 kHz in the darkness. Both of the device exhibit an almost constant capacitance in the bias range −0.8 to 0 V, which indicates a complete depletion of the device. By introducing a TCTA layer, the V_bi is increased from 0.46 to 0.52 V, which is consistent with the raised V_oc.

To further elucidate the mechanism behind the device performance improvement, we measured the impedance spectroscopy of the device at a bias of 0.2 V in darkness. Fig. 5(a) shows the Nyquist plots, where Z′ and Z″ are the real and imaginary components of the impedance, respectively. The impedance data are fitted by Zview using an equivalent circuit model shown in Fig. 5(a), and the fit parameters are summarized in Table 2. There are three components in this model. R_s is used to describe the series resistance. R_ct and C_ct are connected in parallel to characterize the device at high frequency, which reflects the charge transport in the device. And R_r and C_r are connected in parallel to characterize the device at low frequency, which represents the charge recombination in the device. It can be seen that R_ct is increased from 3.0 to 5.6 Ω·cm² by introducing a TCTA layer, which is caused by the resistance and the hole-injection barrier introduced by the TCTA layer. However, R_r is significantly increased from 301 to 1846 Ω·cm², which suggests that charge recombination in the device is significantly suppressed. We also measured the transient photovoltage (TPV), and the results are shown in Fig. 5(b). By introducing a TCTA layer, the half-life is increased from 1.65 × 10⁻⁴ to 4.25 × 10⁻⁴ s, which suggests an inhibition of the charge recombination. This result is consistent with the above impedance analysis.

Let us now turn to the external quantum efficiency (EQE) spectra of the device. Fig. 6 shows the EQE spectra of the devices. Both of the devices demonstrate a wide spectral response up to 1000 nm. Both of the spectra display a valley at around 500 nm, which results from the weak light absorption in this region caused by the thin-film interference effect. The device with a TCTA layer has a higher EQE in 400‒1000 nm compared to that of the control device. The recombination at ITO-Sb₂Se₃ interface was suppressed by introducing a TCTA layer. Consequently, the charge extraction efficiency and the J_sc were improved, which gave rise to the increase in EQE. An EQE maximum of 74% was obtained at 435 nm on the device with a TCTA layer. And an EQE maximum of 72% was obtained at 400 nm on the control device. By integrating the EQE with the standard AM 1.5 solar spectrum, J_sc of 19.64 and 18.05 mA/cm² were achieved for the devices with and without a TCTA layer, respectively. These are very close to the values extracted from the J‒V measurements under illumination.

The repeatability and the stability of the device were investigated. We fabricated 69 solar cells. Fig. S6 displays the distribution of the PCEs, which shows a normal distribution with an average of 3.6 % and a variance of 0.47%. The device stability was tested without encapsulation. The PCE had a ~30% loss after a storage in air at 25 °C and 30% relative humility for 24 h. This problem is primarily caused by the organic materials, and can be partially overcome by encapsulation.

To clarify the role of the oxygen in Sb₂Se₃ film, we performed the following experiments. The ITO substrate was treated by oxygen plasma before device fabrication, which can increase the oxygen level on the ITO surface, and consequently increase the work function of the ITO anode. The increase in anode work function would enhance the hole collection and the device PCE. However, the device efficiency was decreased from 4.3% to 3.9% and from 3.2% to 2.8% for the devices with and without a TCTA layer, respectively, after the oxygen plasma treatment. The result is shown in Fig. S7 in Supplementary Material. This result is similar to that reported by Liu et al.^[38], who found that the device performance is deteriorated by adding oxygen in Sb₂Se₃ film near the anode. However, the PCE is increased by adding oxygen in Sb₂Se₃ film near the n-type layer.

The lower PCE of the Sb₂Se₃ solar cells fabricated by the conventional vacuum evaporation is mainly caused by two reasons: the first is the relatively higher defect density, which may lead to the trap-assisted recombination; and the second is the (020) oriented film, which is not favored for the carrier transport. To improve the device performance the innovations are required in the following directions. It is fundamental to regulate the film deposition kinetics to suppress the defect formation and realize the (001) oriented film. It is also important to select better electron and hole transport layers. In addition, increase of the carrier density by doping may further improve the efficiency of device.

We have demonstrated that a small organic molecule of TCTA can efficiently passivate the ITO–Sb₂Se₃ interface in the Sb₂Se₃ solar cell fabricated by the vacuum thermal evaporation. By adding a TCTA layer, the V_oc is raised from 0.36 to 0.42 V, and the PCE is significantly increased from 3.2% to 4.3%. The introduction of a TCTA layer does not significantly alter the properties of the Sb₂Se₃, and there are two major reasons for the enhancement of the device performance. First, the TCTA layer can block the electron diffusion from Sb₂Se₃ to ITO anode, and the interface recombination is therefore inhibited. Second, the chemical reaction between the ITO and Sb₂Se₃ is inhibited by introducing a TCTA layer, which may reduce the interface state density. The drawback is that the hole-extraction barrier is slightly raised by introducing a TCTA layer. This work provides a method to improve the performance of the Sb₂Se₃ solar cell.