Effect of Heterojunction Characteristics and Deep Electronic Levels on the Performance of (Cd,Zn)S/Sb₂Se₃ Solar Cells

Abstract

Antimony selenide (Sb₂Se₃) is an Earth-abundant and non-toxic material that stands out as a promising absorber for the fabrication of thin film solar cells. Despite significant advancements in recent years, all the devices reported in the literature exhibit open-circuit voltages well below the theoretical value. Identifying the factors contributing to this low voltage is an essential step for increasing the efficiency beyond the recently attained 10% milestone and moving closer to the theoretical limit. In this paper, we present the results of an in-depth analysis of a Sb₂Se₃ solar cell in the common superstrate configuration. By making use of current density–voltage characteristic as a function of both temperature and wavelength, capacitance–voltage measurements, and admittance spectroscopy, we ascribe the low open-circuit voltage to the presence of a potential barrier within the absorber material near the junction interface Furthermore, it was observed that the junction behavior in the dark and under illumination changes, which is compatible with the presence of deep electronic levels connected with intrinsic point defects.

1. Introduction

Antimony chalcogenides, such as Sb₂S₃, Sb₂Se₃, Sb₂(S,Se)₃ [1], and CuSb₂Se₃ [2], are gaining attention within the research community as potential and promising absorber materials for thin film solar cells. In particular, antimony selenide (Sb₂Se₃) exhibits several advantageous characteristics that make it suitable for this role. Sb₂Se₃ is generally considered to be an indirect bandgap semiconductor. Previous studies have reported an indirect bandgap of approximately 1.04 eV and a direct bandgap of about 1.17 eV at room temperature. Nevertheless, Sb₂Se₃ exhibits a strong absorption coefficient, which is advantageous for photovoltaic applications. The high absorption coefficient of Sb₂Se₃, despite its indirect bandgap, can be justified by the following considerations:

• Quasi-1D Crystal Structure: strong orbital interactions inside the ribbon enhance light absorption.
• Near-Direct Optical Transitions: a close-to-direct bandgap (~1.1–1.2 eV) boosts absorption.
• p-d Orbital Hybridization: strong Se 4p–Sb 5s interactions improve optical transitions.
• Short Exciton Diffusion Length: charges are generated close to where light is absorbed.
• High Optical Density of States: enhancement of photon absorption [1].

These factors make Sb₂Se₃ behave more like a quasi-direct bandgap material with an E_g of 1.17 eV [3] that lies within the range for maximum efficiencies in the Shockley–Queisser (SQ) model [4].

Additionally, the innocuous nature and abundance of antimony selenide on the Earth’s crust suggest great potential for commercialization [5,6]. Sb₂Se₃ has a high absorption coefficient (α > 10⁵ cm⁻¹), which allows for thinner absorber layers; furthermore, the absence of polymorphism and a low melting point (885 K) facilitate efficient synthesis by different deposition techniques, such as Chemical Bath Deposition (CBD), Close-Spaced Sublimation (CSS), Vapor Transport Deposition (VTD), Rapid Thermal Evaporation (RTE), Ion Vapor Deposition (IVD), and Radio-Frequency Magnetron Sputtering (RFMS) [1,7,8]. On the other hand, the deposition of antimony selenide presents challenges, particularly regarding its sensitivity to the substrate [9], its tendency to form defects [10,11], and the occurrence of stoichiometric deviations [12]. These negative issues are reflected in low and highly anisotropic mobility, as well as self-compensation, which constitutes an impediment to extrinsic doping. State-of-the-art devices have partially addressed these challenges by careful control of the preferential orientation [13], band offset tuning [14], passivation of interface defects [15], and reduction of contact barriers [16]. These efforts led to the current record efficiency of 10.57% [17]. Research efforts were especially focused on enhancing Fill Factor (FF) and short-circuit current density (J_sc), which are both now over 70% of their SQ limits. However, no similar progress was achieved in increasing the V_oc, which represents a notable problem for further efficiency improvements. Indeed, many reported devices in the literature exhibit a V_oc close to 50% of the SQ limit [18]. Understanding the causes behind this V_oc deficit is imperative for increasing the efficiency of Sb₂Se₃ solar cells.

This study employs sputtering and closed-space sublimation (CSS) for thin-film deposition, which offers key advantages over conventional methods. Sputtering, a PVD technique, ensures high film uniformity, precise thickness control, and strong adhesion, making it ideal for depositing high-purity layers. Compared to solution growth methods, it provides superior structural integrity. On the other hand, CSS is a highly efficient method for Sb₂Se₃ deposition, offering high deposition rates, scalability, and excellent crystallinity. It promotes better grain growth and reduces grain boundary defects, improving charge carrier transport. Moreover, low-temperature CSS is considered a cost-effective and energy-efficient choice for industrial solar cell production.

The ZnO/CdS-(Cd, Zn)S combination in Sb₂Se₃ solar cells provides key advantages in terms of optical, electrical, and structural properties, ensuring efficient charge transport and stability. ZnO, as an HRL (High Resistive Layer), offers high transparency (~3.3 eV bandgap), high resistivity, which is essential for reducing leakage currents, and strong chemical stability, outperforming alternatives like SnO₂ in balancing optical and electrical properties.

Using CdS-(Cd, Zn)S as a buffer layer provides energy gap tuning, low interface defects, and good stability, making it more reliable than alternatives like ZnS or In₂S₃ and MgZnO despite cadmium toxicity concerns. Overall, ZnO and CdS-(Cd, Zn)S remain technologically viable choices that ensure the efficient charge extraction, scalability, and high performance of Sb₂Se₃ solar cells.

In this work, different optoelectrical measurements were carried out on Sb₂Se₃-based solar cells [19,20] to explain some unusual features of the J-V characteristic, namely, the presence of a “kink” and “crossover”. Both features may be justified by the existence of a potential barrier at the interface between the n-type electron transport layer (ETL) and the p-type absorber layer. Capacitance measurements reveal that the positive charge density in the region of the junction follows a linear profile rather than the typical step function. Furthermore, temperature-dependent admittance measurements show the presence of two deep levels, likely connected with intrinsic defects. As a fallout of this comprehensive assessment, we propose a possible path towards the improvement of photovoltage (V_oc) and Fill Factor (FF) aimed at pushing up the power conversion efficiency (PCE) of Sb₂Se₃-based solar cells.

2. Experimental Section

Solar cells were fabricated in a superstrate configuration (Figure 1a), with ITO and ZnO as the front contact. CdS and Cd_0.85Zn_0.15S ((Cd,Zn)S) serve as the ETLs and, right after, antimony selenide (Sb₂Se₃) was deposited by Close-Spaced Sublimation (CSS) (details on growth parameters and technical specifications can be found in previous publications [9,20]). The back contact was an amorphous thin film composed of iron, sulfur, and oxygen; we refer to this layer as Fe-S-O, and it was coated by an Au layer. All layers, except Sb₂Se₃, were deposited using either r.f. magnetron sputtering (ZnO, CdS, (Cd,Zn)S) or d.c. magnetron sputtering (ITO, Fe-S-O, Au). The current–voltage (J-V) characteristics of the solar cells were assessed using a continuous LOT-Oriel solar simulator (Oriel, Irvine, CA, USA) equipped with an air mass AM1.5G filter. The simulator was equipped with a 600 W Xenon lamp (Oriel, Irvine, CA, USA) that gives a 1 kW/m² light power density. To ensure accuracy, a calibrated pyranometer served as a reference, and the measurements were conducted at the standard temperature of 298 K. For the intensity-dependent current–voltage measurements, neutral density optical filters with optical densities (OD) of 3, 2, 1.3, 1, and 0.3 were employed.

J-V curves were taken under illumination, using a KEITHLEY 2400 (Keithley, Solon, OH, USA) source meter, between 233 K and 353 K in steps of 20 K. The antimony selenide-based solar cell was placed inside a thermostatic chamber and illuminated through an optical fiber coupled with a Linos photonics LQX 1000 lamp (Qioptiq Photonics, Gottingen, De). To achieve illumination in different spectral ranges, a combination of band-pass optical filters in the range from 1000 nm to 500 nm was employed.

Capacitance–voltage (CV) measurements in the dark and at room temperature (RT) were carried out with a Keithley 4200-SCS semiconductor characterization system, selecting the highest available frequency (10 MHz) to limit the contribution of deep levels to the capacitance while varying the applied reverse bias within the range of from 0.4 V to 2 V. For admittance spectroscopy, experiments were carried out in the dark and under light using a Hewlett Packard 4284A (Hewlett Packard Enterprise, Palo Alto, CA, USA) PRECISION LCR METER by varying the frequency between 100 Hz and 1 MHz, and the temperature in the range 273 K–353 K.

To determine the energy gap of CdS and (Cd,Zn)S, UV–vis absorption measurements were acquired with a Varian 2390 UV-VIS-NIR (Varian Inc., Palo Alto, CA, USA) in the range between 300 nm and 1500 nm using a transmission configuration.

The (scanning) transmission electron microscopy ((S)TEM) investigations were carried out in an aberration-corrected Titan Themis G2 200-type microscope ((Thermo Fisher Scientific, Waltham, MA, USA)) operating at 200 keV equipped with an X-FEG gun, a 4 k × 4 k CETA 16 CMOS camera (ThermoFisher Scientific), and a HAADF detector (E.A. Fischione Instruments (Fischione Instruments, Inc., PA, USA)). For the energy dispersive X-ray (EDX) mapping, a Super-X detector was used. The cross-sectional TEM samples were prepared by focused ion beam (FIB) technique using a ThermoFisher Scios 2 Dual Beam microscope with EasyLift^TM nanomanipulator. The TEM lamella was immediately moved to the TEM after preparation in FIB in order to reduce the contamination with air. After the TEM experiments, the lamella was stored in a vacuum.

3. Results and Discussion

3.1. The Sb₂Se₃ Solar Cell

Figure 1a illustrates a non-scaled diagram of the structures of the Cd_0.85Zn_0.15S/Sb₂Se₃-based solar cell. The key photovoltaic parameters, extracted from the J-V characteristic presented in Figure 1b, are reported in

Table 1. Main photovoltaic parameters of the investigated solar cell. V_oc is the open circuit voltage, V_X is the crossover voltage, J_sc is the current density, FF is the fill factor, PCE is the power conversion efficiency, R_sh is the shunt resistance, R_S is the series resistance, n is the ideality factor.

Voc [mV]	365 ± 6.0	PCE [%]	4.36 ± 0.3
Vx [mV]	375 ± 6.0	Rsh [Ωcm2]	378 ± 16
Jsc [mA/cm2]	17.1 ± 0.3	RS [Ωcm2]	3.6 ± 0.1
FF [%]	49 ± 3.0	n	2.1 ± 0.1

.

To evaluate the series resistance (R_s) and the ideality factor (n) of the solar cell, a manipulation of the J-V expression, neglecting the saturation current and supposing R_s << R_sh, yields the following [14]:

$${\displaystyle \frac{dV}{dJ}}={\displaystyle \frac{nKT}{q}}{\left(J+J_{ph}\right)}^{-1}+R_S$$

(1)

Here, n represents the diode ideality factor and J_ph is the current attributed to the photogenerated carriers, which can be approximated by the short-circuit current J_sc.

From the plot in Figure 2a, it is possible to extract the series resistance R_S while, from the plot of dJ/dV in Figure 2b, one can obtain the shunt resistance R_sh (both values reported in Table 1) once the value of R_S is known from the fit of Figure 2a [21].

$${\displaystyle \frac{dJ}{dV}}={\displaystyle \frac{q{J}_0}{nKT}}{e}^{\left[{\displaystyle \frac{q\left(V-J{R}_S\right)}{nKT}}\right]}+{\displaystyle \frac{1}{R_{sh}}}$$

(2)

For the subsequent analysis of the cell properties, it is important to note that the interface between the p-type Sb₂Se₃ and the n-type (Cd,Zn)S region is not sharp but rather graded due to layer intermixing, which helps minimize the lattice mismatch between the two materials. According to relation (3), the lattice mismatch (δ) between Sb₂Se₃ and Zn_0.15Cd_0.85S is indeed large [22]:

$$\delta (\%)={\displaystyle \frac{2\cdot \left|a_{\left(\mathrm{Cd},\mathrm{Zn}\right)\mathrm{S}}-b_{S{b}_{2}S{e}_3}\right|}{a_{\left(\mathrm{Cd},\mathrm{Zn}\right)\mathrm{S}}+b_{S{b}_{2}S{e}_3}}}\cdot 100=2.63\%$$

(3)

The lattice parameters of (Cd,Zn)S and Sb₂Se₃ are denoted as $a_{\left(\mathrm{Cd},\mathrm{Zn}\right)\mathrm{S}}$ and $b_{S{b}_{2}S{e}_3}$, respectively.

The lattice parameter of (Cd,Zn)S (with 15% atomic concentration of Zn) is estimated using Vegard law [23,24] with $a_{CdS}=4.14 \AA$ and $a_{ZnS}=3.82 \AA$ [25,26].

$$a_{Z{n}_{0.15}C{d}_{0.85}S}=0.15a_{ZnS}+0.85a_{CdS}=4.092\AA$$

(4)

The lattice parameter of Sb₂Se₃ to be considered is $b_{S{b}_{2}S{e}_3}$ = 3.9858 Å, as for our growth conditions the Sb₂Se₃ preferential orientation is [001] direction, i.e., parallel to the growth axis [9].

The nominal concentration of 15 at% Zn inside the CdS crystal structure was confirmed by taking the E_g of (Cd,Zn)S (2.67 ± 0.01 eV) and CdS (2.48 ± 0.01 eV) from a Tauc plot (see Figure 3). With the assumption that the energy gap varies linearly with the Zn concentration between the E_g of CdS (0%) and that of ZnS (100%), and taking the E_g of ZnS = 3.6 eV [27], we estimated the at% of Zn in CdS to be (17 ± 2)%, which is consistent with the present 15 at% value.

A lattice mismatch exceeding 1% normally introduces a significant density of crystallographic defects at the interface, which provides energy levels acting as recombination centers. This well-known fact limits the number of photocarriers that actually reach the device electrodes. Such a phenomenon was previously observed in CdS/CdTe-based solar cells and pre-vented the formation of an adapting layer [28], which gradually reduced the effective mismatch. Similarly, in antimony selenide-based solar cells, the introduction of such an adapting layer proved to be essential to circumvent the lattice mismatch issue.

The introduction of (Cd,Zn)S at the junction, besides decreasing the lattice mismatch, has two further advantages. First, as previously shown [9], the introduction of a Zn_0.15Cd_0.85S layer allows the growth of antimony selenide with preferential grain growth orientation along the [001] direction. Second, as demonstrated in ref. [14], a zinc concentration ranging from 15 at% to 20 at% within CdS enables the achievement of a spike-like type conduction band offset (CBO) at the interface with Sb₂Se₃, which is sufficiently small so as to not induce significant effects on the photovoltaic parameters of the cell.

The hypothesis regarding the existence of the intermixing layer between (Cd,Zn)S and Sb₂Se₃, based on the experience with the CdS/CdTe system, has been experimentally verified using HRTEM and EDS measurements in cross-section on the complete Sb₂Se₃-based cell. Figure 4b shows the EDS spectrum related to the cross-section highlighted in Figure 4a.

In Figure 4c–e,h, an interdiffusion of selenium and antimony atoms into the sulfide layer is observed, which forms an adaptive layer with an estimated thickness of about 20 nm in accordance with the EDS profile spectrum depicted in Figure 4b. The interdiffusion of Se and Sb into the sulfide layer occurs during the CSS deposition of Sb₂Se₃ at temperatures between 350 and 400 °C. This temperature range is sufficient to promote diffusion at the (Cd, Zn)S/Sb₂Se₃ interface, leading to the formation of an intermixing layer [29]. The observed interdiffusion is uniquely produced during the CSS growth as no additional heating is applied after deposition. Additionally, a non-negligible amount of oxygen (10 to 20 at%) seems present outside the ZnO layer, including in the Sb₂Se₃ layer. This amount cannot be explained by assuming the presence of oxygen during deposition since the starting vacuum of all the different deposition processes is less than 10⁻⁶ mbar, and the oxygen presence has been shown to be undetectable with an oxygen vapor pressure that is more than one order of magnitude greater [30]. A hypothesis regarding this anomalous experimental result is that the cross-sectional ca. 30 nm thin TEM lamella with a large free surface tends to oxidize during sample transfer, and, thus, the oxygen is not actually present in the subsequent layers, especially not in the Sb₂Se₃. After the TEM investigation, we put the samples into a vacuum (10⁻⁶ mbar); however, the measured amount of oxygen was 1.5 times higher on the second day due to the double transfer that exposed the sample to contamination by air.

TEM measurements on the Sb₂Se₃ do not show any spurious secondary phases of antimony oxides in accordance with the results obtained in Figure 5.

3.2. Kink Effect in the J-V Characteristic

As clearly shown in Figure 1b, the current–density characteristic for the polycrystalline Sb₂Se₃-based solar cell exhibits different behavior in the dark and under illumination. Specifically, under illumination, a kink is observed at high voltages in the first quadrant of the forward J-V characteristic. This behavior is attributed to the presence of an energy barrier within the absorber layer, which can be explained by the two-diode model in series configuration (Figure 6a) [31,32]. In the first quadrant of the J-V characteristic, the main junction barrier is completely flattened by the voltage applied to the contacts. In this situation, holes must exit from the negative pole and electrons from the positive pole, contrary to what happens in the fourth quadrant, where the built-in electric field of the junction pushes the photogenerated carriers in the opposite direction.

In the case of the (Cd,Zn)S/Sb₂Se₃ junction, for a forward bias above V_oc, this is certainly true for electrons, but holes tend to accumulate at the interface of the p-side of the junction. This accumulation is a result of the potential step caused by the offset between the valence bands of (Cd,Zn)S and Sb₂Se₃ (see Figure 7).

These materials possess quite different band gaps, and the low conduction band offset (CBO) leads to a big cliff-like discontinuity in the valence band, even under the conditions of the first quadrant of the J-V characteristic, as explained in [33]. Therefore, a fraction of the holes, partly photogenerated and partly injected, are trapped in the Sb₂Se₃ layer near the interface with (Cd,Zn)S. The trapped holes, unable to move freely, promote recombination with electrons, causing a noticeable decrease in current, which is seen as a kink in the J-V characteristic.

In the dark, the phenomenon is much less evident, as the resistance offered by the materials only allows the flow of a weaker current and, accordingly, the number of trapped holes is also lower.

To determine the barrier height, we performed temperature-dependent J-V characteristics at different temperatures, and we assumed that the barrier height as a function of the temperature follows a thermally activated behavior, as reported in ref. [34]:

$$R_{S}T=C e^{{\displaystyle \frac{{\Phi }_B}{KT}}}$$

(5)

where R_S is the total resistance comprising the resistance due to the materials constituting the solar cell, the resistances arising from contacts R₀, and the resistance of the additional barrier D_B (see Figure 6a). C is a constant independent of the temperature T, and Φ_B is the barrier height. If R₀ is sufficiently small compared to the barrier resistance, and, therefore, negligible, then the temperature dependence of R_s is primarily attributed to D_B, which is the circuital representation of the barrier due to charge trapping.

The total series resistance (R_s) was determined by linearly fitting the J-V characteristics, at different temperatures, around V_oc. The inverse of the fitting slope represents the value of R_s. The barrier height was estimated to be 306 ± 9 meV from the slope of ln(R_sT) vs. 1/T (Figure 8a) [35,36,37]. As reported in Figure 6b, the barrier height varies with the wavelength of the incident light. This happens because the penetration length depends on the wavelength and so does the thickness of the material where the photogenerated holes pile up [38]. This experimental fact may be appreciated from Figure 8b, where the barrier height is plotted as a function of the light penetration depth (remembering that approximately 82% of light is absorbed within a thickness of two penetration lengths, according to the Beer–Lambert–Bouguer law).

This result has two consequences: first, it completely rules out the dependence of the kink on potential barriers at the contacts, and, second, it shows that the barrier, due to trapped holes, is at its maximum at the junction region near the interface and then decreases into the p-type material.

3.3. Crossover Effect in the J-V Characteristic

From Figure 1b, it is evident that the J-V characteristic under light is not a simple translation of the dark J-V characteristic, as in an ideal junction. The two characteristics show a crossover, indicating a potential-dependent photocurrent [39,40,41]. It is typically assumed that the photogenerated current primarily depends on the intensity of incident light, with only a minor dependence on the voltage V. However, in this case, the dependence on V is not sufficiently weak to be disregarded. Consequently, we can describe the current flowing in the solar cell as follows (superposition principle) [33]:

$$J_L\left(V,g\right)={J_D}_{sc}\left(V,g\right)+J_{ph}\left(V,g\right)$$

(6)

where J_L(V,g) is the total flowing current, J_DSC(V,g) is the current flowing in the diode (p-n junction), and J_ph(V,g) is the current generated by light. Here, V is the applied bias and g is the photocarrier generation rate. The net current of the diode without the contribution of light-induced generation can be expressed, considering the dark current, as follows:

$${J_D}_{sc}\left(V,g\right)=J_j\left(V,g\right)-J_{dark}$$

(7)

Here, J_j(V,g) represents the injection current of the diode, and J_dark is the current of the unilluminated diode, which, being minimally dependent on the potential, can be considered as constant. Consequently, the difference between J_L(V,g) and J_dark can give some information about the crossover:

$$\mathsf{\Delta }=J_L\left(V,g\right)-J_{dark}=\left[J_j\left(V,g\right)-J_{dark}\right]+\left\{J_{ph}\left(V,g\right)-J_{dark}\right\}$$

(8)

From the first term (in square brackets) on the right-hand side of Equation (8), we can infer that, if the current injected by the junction also depends on the illumination, then, in Δ, this term predominates, and the crossover essentially depends on this. On the contrary, if photogeneration depends on the potential, then the second term (curly brackets) predominates. The difference ∆ has a trade-off at the point V_x. In our Sb₂Se₃-based cell, the p-n junction can be considered as a hole-blocking heterojunction, attributed to the electron affinity of the n-type layer ((Cd,Zn)S) and the doping level of the p-type layer (Sb₂Se₃). As mentioned earlier, this implies an accumulation of positive charge in the p-type region of the junction and, consequently, (J_j(V,g) − J_dark) > 0. This remains valid even when considering the intermixing layer, which reduces the lattice mismatch but does not fully adapt the offset in the valence bands between (Cd,Zn)S and Sb₂Se₃. Concerning the presence of the crossover, there are, however, two more questions to be discussed. First, it should be considered that the potential we observe at the contacts is not the built-in potential V_bi but rather the potential that drops on the resistances of materials and contacts. We can denote this potential as V_bi,c to indicate that it is the potential observed at the contacts.

Secondly, as mentioned earlier, we need to consider that, in the (Cd,Zn)S/Sb₂Se₃ cell under moderate forward bias, only electrons can flow through the p-type contact, while holes are partially blocked by the barrier between the valence band of (Cd,Zn)S and Sb₂Se₃. The photocurrent is, therefore, voltage-dependent but cannot become zero until all electrons and holes have reached their respective contacts. Nevertheless, the presence of positive charge accumulation in the junction region under light implies a change in the internal electric fields. This, in turn, means that the diode current becomes dependent on the light-induced generation of charge carriers. This influences the point where the photocurrent and the dark current intersect. Two cases can be distinguished:

• In the case where the doping density within the CdS + (Cd,Zn)S layer is quite low, in the order of 10¹³ cm⁻³, the electric field primarily drops in the n-type region and the p-type region will be only slightly depleted. In this region, there will be a relatively small accumulation of photogenerated positive charges. Therefore, a higher potential will be required to reach the crossover voltage, which is typically greater than V_bi,c.
• In contrast, when the n-type region of the junction is more heavily doped than the p-type region, the latter experiences greater depletion. This results in a more pronounced bending of the energy bands in this region. Consequently, there is an increased accumulation of positive charges in this area, leading to a crossover point that is lower than V_bi,c.

To understand which of the two situations we are in, we need to assess how much V_bi,c depends on the series resistances of the device.

Considering the series resistance values of our solar cell (R_S ≈ 3.6 Ω ∙ cm²), the relative voltage drop (J∙R_S) across the materials is not significant, especially when the current density is very low, i.e., the voltage is close to V_oc and/or the crossover point V_x. Therefore, V_bi,c is approximately equal to V_bi.

In summary, the condition where V_x < V_bi arises because of the higher doping level in Sb₂Se₃ compared to (Cd,Zn)S. It is now necessary to address the question of how V_x is influenced by light (generation). The variation of V_x and V_oc as a function of the wavelength of the incident light has been illustrated in Figure 9a. Measurements were conducted using bandpass filters centered at wavelengths of 650 nm, 750 nm, 850 nm, and 950 nm. The V_x and V_oc values were then extracted from the standardized JV curves normalized for an equal number of absorbed photons.

The graph (Figure 9a) indicates that V_x is always slightly higher than V_oc, aligning with the definition of V_x. However, the difference between V_x and V_oc seems to widen as the wavelength decreases (Figure 9b). Considering the penetration depth of the light into the Sb₂Se₃ layer, carriers generated by a longer wavelength must cover a greater distance to reach the electrodes (Figure 8b), amplifying the impact of recombination effects. This phenomenon results in a diminished photovoltage. The graph in Figure 9a shows a non-uniform dependency of V_x on light generation.

The accumulation of photogenerated charges at the junction varies with the wavelength of the absorbed light. Holes generated by short-wavelength light are near the junction, allowing for more effective accumulation compared to holes generated farther away by long-wavelength light. Explaining this analytically is challenging, but the observed trend as a function of absorbed light aligns with what is depicted in Figure 9a,b. Additionally, the slight increase in V_x values compared to V_oc supports the accuracy of the V_oc ≤ V_x ≤ V_bi relationship.

3.4. Capacitance–Voltage Characterization

Capacitance–voltage measurements, in the dark and under light, were performed to obtain the built-in potential and the behavior of the spatial charge density. The capacitance of the p-n junction can be expressed for any diode as follows [42]:

$$C=constant·{\left(1+{\displaystyle \frac{V}{V_{bi}}}\right)}^{-\beta }$$

(9)

Here, C represents the capacitance of the junction, V denotes the reverse bias, V_bi stands for the barrier potential, and β is a coefficient accounting for the dependency of C on the reverse bias V. For an ideal junction with an abrupt charge density step profile, β = 1/2. β can be determined from the slope of the linear fit of ln(C) vs. ln(V) in dark and light condition (see Figure 10). This procedure enables the determination of the spatial distribution of a space charge within the junction region. The values of β result in ${\displaystyle \frac{1}{2.94}}$ and ${\displaystyle \frac{1}{3.06}}$ in dark conditions and under light, respectively. Both values are very close to ${\displaystyle \frac{1}{3}}$ and suggest a linear dependence of C³ vs. V, which corresponds to a linear dependence of the space charge density within the space charge region, according to the C-V theory for a linearly graded junction reported in [43,44].

The relationship between depletion region capacitance and applied voltage is thus as follows:

$${\displaystyle \frac{1}{C^3}}={\displaystyle \frac{12\left({\mathrm{V}}_{bi}-V\right)}{q\left(\delta \rho \right)A^3{\left({ϵ}_0ϵ\right)}^2}}$$

(10)

where V represents the applied voltage, V_bi is the built-in potential, q is the electron charge, $\left(\delta \rho \right)$ indicates the impurity gradient expressed in cm⁻⁴, A denotes the active area of the diode, ϵ is the relative dielectric constant (assumed to be 15.1 [45,46]), and ϵ₀ represents the vacuum dielectric constant. Equation (10) is indeed confirmed by the plot in Figure 11 that reports the plot of 1/C³ vs. applied bias, and the linear fit in the negative bias region can be used to determine the built-in potential from the intercept with the V-axis. V_bi was found to be (0.53 ± 0.05) V and (0.82 ± 0.05) V under light and dark conditions, respectively [35,36,37]. The result reveals a higher built-in potential in dark conditions compared to that under light. This discrepancy contradicts classical theory, as these two values are expected to be equal and suggest the presence of an illumination-linked potential. Indeed, when we consider the height (0.3 V) of the barrier attributed to the hole-blocking heterojunction and add it to V_bi under illumination (0.53 V), we obtain a value very close to the V_bi (0.83 V) in the dark. This confirms that the first term on the left side of Equation (8) is significantly influenced by g, the photocarrier generation rate, as corroborated by the presence of the crossover discussed above.

3.5. Deep Levels

In the dark, the plot of the J-V curve on a log–log graph exhibits three distinct linear regions [47,48,49,50]:

• Ohmic region ($J\propto V$);
• Trap-filling-limit region ($J\propto V^{n>2}$);
• Child region ($J\propto V^2$).

In the presence of traps, increasing the applied voltage results in an increased injected current, which will fill empty traps. The trap density ($N_t$) can be determined from the trap-filling-limit (V_tfl) voltage, which is the intersection between the linear fits of the ohmic and trap-filling-limit regions of Figure 12. Once the intersection is known, the trap density can be calculated from the following relation [21,51]:

$$N_t={\displaystyle \frac{2ϵ{ϵ}_0{V}_{tfl}}{q{L}^2}}$$

(11)

where ε is the relative dielectric constant (assumed equal to 15.1 [45,46]), ε₀ is the vacuum dielectric constant, q is the electron charge, and L is the antimony selenide thickness [52].

With an estimated trap density of N_t ≈ 1.1 ∙ 10¹³ cm^-3, it can be then assumed that all traps in the Child region are filled.

Consequently, the relationship governing the current–voltage trend in this region follows the Mott–Gurney law [48].

$$J_C={\displaystyle \frac{9}{8}}ϵ{ϵ}_0{\mu }_h{\displaystyle \frac{V_C-{\mathrm{V}}_{bi}}{L^3}}$$

(12)

Here, J_C and V_C are the values at the intercept between the linear fits of trap-filling-limit and Child regions. V_bi is the built-in voltage in the dark, and µ_h denotes the hole mobility. Solving Equation (12) yields µ_h ≈ 8.9 cm² V⁻¹ s⁻¹.

Utilizing the resistivity measured in our previous work for antimony selenide in the dark [9], and applying the well-known relationship, we obtain the following:

$$\sigma ={\displaystyle \frac{1}{\rho }}=qp\mu$$

(13)

A free hole density p of 1.4 $\times$ 10¹⁴ cm⁻³ is obtained. The difference of just one order of magnitude between the free hole density and the density of traps may explain the rather low performances of the solar cell.

To assess the energy position of the traps, admittance spectroscopy measurements in the dark and under illumination at steps of 20 K from 193 K to 353 K were carried out. Notably, the curves at the lowest temperatures exhibit an inflection point for the capacitance at high frequency, which again confirms the presence of deep traps [53] as shown in Figure 13.

By evaluating the frequency ω₀ at which the inflection points occur and considering the following relationship [54]:

$${\displaystyle \frac{{\omega }_0}{T^2}}=2{\xi }_0\exp \left(-{\displaystyle \frac{E_a}{KT}}\right)$$

(14)

the activation energy of the deep levels can be determined, as shown in Figure 14.

Two deep levels acting as traps were identified at (183 ± 23) meV and (324 ± 39) meV in the dark and under light, respectively. It is important to note that this is only an estimate since the C-ω curves were obtained at relatively large d-factor values and that, with this procedure, it is not possible to determine the exact position of these traps, whether in the p-type or the n-type region. Anyway, from the curves shown in Figure 13 and Figure 14, we can deduce that these energies represent the limits of a narrow energy band corresponding to a distribution of traps. This conclusion is supported by the absence of an evident inflection in each curve, as the decrease in capacitance is not sharp, which points to the overlapping effects of multiple trap levels close in energy. This band of traps may explain why two different energy levels show up in the dark or under light (Figure 14). This phenomenon can be explained by the band bending occurring at the intermixing layer between (Cd,Zn)S and Sb₂Se₃. Indeed, because of the extended trap levels and the variable Fermi and quasi-Fermi levels in the dark and after illumination, only energy-related fractions of the traps will be filled. The literature [55] reports numerous levels within the Sb₂Se₃ energy gap; most of them are categorized as deep donors, with only a few classified as acceptor-like defects. Among these, the antimony vacancy (V_Sb) corresponds to an energy level located at 180 meV above the top of the valence band [55,56], which is consistent with the lower extreme of the trap band detected by admittance measurements (Figure 13 and Figure 14). If V_Sb [55,56] is the actual point defect acting as a trap then one is led to conclude that the trap band is localized in the Sb₂Se₃ layer.

Preventing defects can reduce recombination, leading to a higher open-circuit voltage (V_oc) and fill factor (FF), while minimizing the lattice mismatch improves carrier transport and short-circuit current density (J_sc). Based on trends in similar thin-film technologies, efficiency gains of 20–30% are expected with effective defect passivation and optimized interfaces.

4. Conclusions

This study aims to improve the efficiency of Sb₂Se₃-based solar cells by optimizing DC/RF sputtering for the ZnO/CdS-(Cd,Zn)S and back-contact layers, along with the CSS for Sb₂Se₃, in order to improve film quality, interface engineering, and defect reduction. Compared to other Sb₂Se₃ studies, this approach offers significant advantages. Unlike chemical bath deposition, electrodeposition, hydrothermal deposition, and thermal evaporation, it ensures superior crystallinity, lower defect density, and improved charge transport. When considering alternative absorbers, like CdTe, CIGS, and perovskites, Sb₂Se₃ stands out for its earth abundance, non-toxicity, and long-term stability. Although CdTe and CIGS offer higher efficiencies, they face material scarcity and toxicity concerns, while perovskites suffer from stability issues, particularly under prolonged exposure to moisture and heat, which hinder their large-scale deployment. Despite the actual lower efficiencies (~10–11%) of Sb₂Se₃, advancements in passivation and heterojunction engineering can bridge this gap. This work contributes to the field by proposing novel scalable, reproducible deposition methods, implementing defect passivation strategies, and demonstrating industrial viability. With continued progress, Sb₂Se₃ solar cells can achieve higher efficiencies and become a commercially viable and sustainable thin-film photovoltaic technology.

In particular, we discussed the non-standard J-V characteristic of (Cd,Zn)S/Sb₂Se₃ heterojunctions, particularly focusing on features such as kink, crossover, and photovoltage losses. The presence of a kink can be attributed to the accumulation of positive charges in the junction region, arising from a non-perfect alignment of the energy bands with the (Cd,Zn)S. On the other hand, the occurrence of a crossover is likely due to a small built-in voltage resulting from the limited density of holes in Sb₂Se₃. Capacitance measurements suggest that the (Cd,Zn)S/Sb₂Se₃ junction is not sharp but rather presents a linear charge distribution. Two deep levels sensitive to illumination conditions were identified through admittance measurements, and there are indications that V_Sb is involved as one of the traps.

Higher power conversion efficiencies in Sb₂Se₃-based solar cells can be expected by properly addressing some major issues:

• Given the imperfect alignment of the energy bands with (Cd,Zn)S, it is imperative to explore alternative n-type partners, such as Zn(S,O), TiO₂, MgO₂, Cd₂Sn0₄, and others, in an attempt to minimize the interface barrier. A possible benefit of such attempts could be the substitution of cadmium-based semiconductors with environment-friendly materials and, at the same time, enhanced transparency in the ETL.
• Replacing cadmium-based materials in Sb₂Se₃ solar cells offers significant environmental benefits by reducing toxicity and minimizing contamination risksEliminating cadmium simplifies manufacturing, enhances worker safety, and facilitates recycling by avoiding hazardous waste management. Alternative materials, such as Zn(O,S), ZnMgO, or low-Cd (Cd, Zn)S, can provide suitable band alignment while reducing environmental impact, though interface engineering remains crucial to maintaining performance. As sustainability becomes a key factor in solar energy adoption, cadmium-free Sb₂Se₃ solar cells align with stricter environmental policies and market demands, enhancing their long-term commercial viability.
• The correlation between the deep levels and intrinsic defects offers the opportunity to adjust the Sb₂Se₃ growth parameters and post-deposition treatments. This adjustment aims to minimize or eliminate intrinsic defects responsible for the formation of deep levels. Preventing deep levels acting as traps for charge carriers would enhance free charge accumulation in the junction region, thereby improving the built-in potential, photovoltage, and photocurrent.

It is crucial to carefully consider the lattice mismatch between the ETL layer and Sb₂Se₃ to mitigate the formation of the interfacial states responsible for recombination centers. These recombination centers can significantly reduce the internal quantum efficiency, leading to adverse effects on photovoltaic conversion efficiency.

The insights gained in this study can help to improve the efficiency of Sb₂Se₃-based solar cells, stimulate a broader research effort, and attract commercial attention.

Indeed, the transition from laboratory-scale to industrial applications for Sb₂Se₃-based solar cells using DC/RF sputtering and close-spaced sublimation (CSS) involves key considerations for scalability, cost-effectiveness, device stability, and compatibility with existing production infrastructure.

Low-temperature sputtering and CSS are scalable techniques with low energy consumption and high deposition rates, making them suitable and cost-effective for large-area solar cell production.

Reducing deep-level defects and lattice mismatches is crucial for ensuring long-term device performance and stability. The techniques are compatible with existing solar cell production lines, hence reducing initial investment and speeding up commercialization.

Efficiency improvements and optimization will help Sb₂Se₃-based solar cells compete with CdTe and CIGS. In summary, Sb₂Se₃-based solar cells have strong potential to become commercial players.