Distinctive Deep-Level Defects in Non-Stoichiometric Sb2 Se3 Photovoltaic Materials.

Characterizing defect levels and identifying the compositional elements in semiconducting materials are important research subject for understanding the mechanism of photogenerated carrier recombination and reducing energy loss during solar energy conversion. Here it shows that deep-level defect in antimony triselenide (Sb2 Se3 ) is sensitively dependent on the stoichiometry. For the first time it experimentally observes the formation of amphoteric SbSe defect in Sb-rich Sb2 Se3 . This amphoteric defect possesses equivalent capability of trapping electron and hole, which plays critical role in charge recombination and device performance. In comparative investigation, it also uncovers the reason why Se-rich Sb2 Se3 is able to deliver high device performance from the defect formation perspective. This study demonstrates the crucial defect types in Sb2 Se3 and provides a guidance toward the fabrication of efficient Sb2 Se3 photovoltaic device and relevant optoelectronic devices.

Introduction

In semiconducting materials, point defect with deep energy level and large capture cross section possesses high probability to induce trap‐assisted Shockley‐Read‐Hall (SRH) recombination.[ , , ] It is also recognized that SRH recombination degrades the solar cell performance by inducing open‐circuit voltage (V OC) loss,[ ]which is the major challenge in the materials synthesis for high performance solar cells. Therefore, the identification and passivation of deep‐level defects arouse intense interests. The continuous efforts in the field, from Cu(In,Ga)Se2 and CdTe to perovskite solar cells, have upgraded the solar cell fabrication technologies and new understandings in the semiconducting materials.[ , , , ]

As a class of emerging photovoltaic material, quasi‐1D antimony triselenide (Sb2Se3) has drawn increasing interests due to its suitable band gap (1.1–1.3 eV), large light absorption coefficient (>105 cm−1 in visible and near‐infrared region), earth‐abundant storage, low toxicity and benign grain boundaries without dangling bonds along (Sb4Se6)n ribbons.[ , , ] Recent years, power conversion efficiency (PCE) of Sb2Se3 has made a great progress, Li et al. achieved the highest PCE of 9.2%,[ ] indicating the great potential for future application upon further efficiency improvement. In some well‐prepared devices, the short‐circuit current density (J SC) and fill factor (FF) can reach decent values. However, all reported V OC of Sb2Se3 solar cells is below 500 mV, corresponding to a serious V OC loss of 700–800 mV.[ , , ] Therefore, the large V OC deficit has become the major limiting factor for essential efficiency improvement. In principle, to improve the V OC, it is critical to suppress the SRH recombination induced by deep‐level defect.[ , ]

The theoretical studies have pointed out that there are complicated deep‐level defects derived from the distinctive quasi‐1D structure in Sb2Se3. [ , , , , , ] Practically, it is difficult to serve as guidance for the solar cell fabrication since the calculations consider all possible defects regardless of the materials processing. The materials synthesis usually influences the defect formation. Therefore, it is an urgent task to probe the defect characteristics experimentally. In this regard, there are three issues need to be resolved: (1) high‐purity sample should be prepared for clear and reliable detection, (2) the defect should be correlated with device performance, which requires specific detection technique, and (3) the relationship between defect and carrier dynamics is required to be established to sophistically control carrier transport and suppress recombination.

Here we figure out the critical issues by fabricating high purity Sb‐rich and Se‐rich Sb2Se3 films, in which only the Sb, Se and Sb2Se3 are applied as the precursor materials, so as to gain films without impurities derived from extrinsic elements. The deep‐level transient spectroscopy (DLTS) measurement is based on complete device, which perfectly establish the defect properties and device performance. We further investigate the correlations between the defect properties and carrier lifetime which is characterized by transient absorption spectroscopy (TAS). Therefore, we are able to finally establish the correlation between materials processing, defect properties, carrier recombination dynamics as well as the device performance, and in turn provide new understanding to manufacture Sb2Se3 for efficient device applications.

Results and Discussion

In brief, we prepare Sb‐rich and Se‐rich Sb2Se3 films by using Sb or Se with Sb2Se3 powders as precursor materials in a dual‐source thermal evaporation deposition (see details in Experimental Section). Finally, the atomic ratio of Se/Sb for Sb‐rich and Se‐rich Sb2Se3 are 1.37 and 1.56 measured by energy dispersive X‐ray spectroscopy (EDX), respectively (Table S1, Supporting Information). SEM images (Figure  , Figure S1a,b, Supporting Information) indicate that both Sb‐rich and Se‐rich films show compact and pin‐hole free morphology. Furthermore, the atomic force microscopy (AFM) images also display the similar roughness (root mean square, R q) for Sb‐rich and Se‐rich Sb2Se3 films (Figure S2, Supporting Information), which are 25.4 and 25.9 nm, respectively. Notably, all diffractions of XRD patterns (Figure 1c) are consistent well with [hk1] preferentially oriented orthorhombic Sb2Se3 (JCPDS No. 15‐0861),[ ] and no noticeable diffraction peaks of impurities such as elemental Sb, Se and oxides are observed.

Figure 1

a,b) Surface SEM images of Sb‐rich and Se‐rich Sb2Se3 films. c,d) XRD patterns and Raman spectra of Sb‐rich and Se‐rich Sb2Se3 films. e,f) XPS spectra of Sb 3d and Se 3d for Sb‐rich and Se‐rich Sb2Se3.

To further confirm the purity of the samples, we conduct the Raman spectroscopy and X‐ray photoelectron energy spectroscopy (XPS). Raman spectra (Figure 1d) demonstrate that both Sb‐rich and Se‐rich films only display three characteristic peaks at 153 (B g), 190 (A g), and 209 (A g) cm−1, which are indexed to the symmetric vibration of Sb—Se bond in Sb2Se3 respectively.[ ] Obviously, there are no peaks belonging to Sb2O3 detected.[ ] In addition, the XPS spectra (Figure 1e,f) appear two peaks at high binding energy of 539.3 and 529.9 eV, which are ascribed to the Sb 3d 3/2 and 3d 5/2 of Sb2Se3. The low binding energy at 55.6 and 54.7 eV are corresponding to Se 3d 3/2 and 3d 5/2 of Sb2Se3,[ ] respectively. Combining the XRD and XPS results, we confirm that the as‐prepared Sb2Se3 films are free of impurities such as elemental Sb, elemental Se and oxides.[ ]

Furthermore, to determine the band gap and energy levels of the as‐synthesized Sb2Se3 films, we conduct ultraviolet–visible (UV–vis) absorption spectroscopy and ultraviolet photoelectron energy spectroscopy (UPS). UV–vis spectra (Figure S1d, Supporting Information) demonstrate that Sb‐rich and Se‐rich Sb2Se3 share an identical band gap of 1.20 eV. Then we measured secondary electron cutoff and valence band position through UPS spectra (Figure S1e,f, Supporting Information). Therefore, we can calculate the work function (Fermi level) of Sb‐rich and Se‐rich Sb2Se3 which are −4.64 and −4.67 eV, respectively. Combined with the band gap of 1.20 eV, we can calculate that the conduction band minimum (CBM) and valence band maximum (VBM) position of Sb‐rich Sb2Se3 are −4.25 and −5.45 eV, while the CBM and VBM of Se‐rich Sb2Se3 are −4.26 and −5.46 eV, respectively.

To further identify the conductivity type of Sb2Se3, we also carry out Hall effect measurement. The as‐prepared Sb2Se3 films with 300‐nm thickness and 1‐cm2 area are selected as samples to be examined at 300 K. And the results are summarized in Table S2 (Supporting Information). The Sb‐rich and Se‐rich Sb2Se3 show similar carrier concentration (≈1016 cm−3) and both possess negative Hall coefficient. The negative Hall coefficient indicates the majority carriers of Sb2Se3 are electrons, namely, N type conductivity.

The DLTS characterization is based on complete device structure, it is thus necessary to examine the photovoltaic performance of the films to confirm the quality at device level. The solar cell with a superstrate configuration of FTO/CdS/Sb2Se3/Spiro‐OMeTAD/Au was fabricated (Figure S1c, Supporting Information). The current density–voltage (J–V) characteristics for optimal Sb‐rich and Se‐rich devices were examined under standard AM1.5G illumination (Figure  ). The statistical V OC, J SC, FF, and PCE of 40 devices each for Sb‐rich and Se‐rich Sb2Se3 are illustrated in Figure 2c–f. Strikingly, Se‐rich Sb2Se3 device shows more efficient performance than Sb‐rich Sb2Se3 devices, especially the remarkable enhancement of V OC and FF. The excellent performance reproducibility for Se‐rich devices suggests the reliability of the device fabrication. Finally, the best Se‐rich device delivers a PCE of 6.0%, with V OC of 390 mV, J SC of 25.86 mA cm−2 and FF of 59.21%. In contrast, the Sb‐rich device only delivers a maximum PCE of 3.5%, with V OC of 300 mV, J SC of 23.39 mA cm−2 and FF of 49.79%, respectively.

Figure 2

a) J–V curves of the champion Sb‐rich and Se‐rich Sb2Se3 solar cells. b) EQE characteristics and the corresponding integrated current density for optimal Sb‐rich and Se‐rich Sb2Se3 devices. c–f) Statistical V OC, J SC, FF, and PCE of Sb‐rich and Se‐rich devices.

Moreover, we also perform electrochemical impedance spectroscopy (EIS) to study the resistance of Sb2Se3 device (Figure S3, Supporting Information). The charge transfer resistance (R s) and charge recombination resistance (R rec) for Sb‐rich and Se‐rich Sb2Se3 can be extracted from the arc of Nyquist plot according to corresponding equivalent circuit. The R s for Sb‐rich and Se‐rich Sb2Se3 devices are 3.81 and 2.66 Ω cm2, while the R rec for Sb‐rich and Se‐rich Sb2Se3 devices are 42.66 and 126.12 Ω cm2, respectively. In principle, the reduced R s reflects the improved interface, and the enhanced R rec indicates the inhibited trap‐assisted SRH recombination. Therefore, it suggests that Se‐rich device possesses better interface and lower bulk SRH recombination rate.

The external quantum efficiency (EQE, Figure 2b) is also carried out. The EQE results indicate that both kinds of devices show photoelectron response over a wide range from 300 to 1100 nm. In addition, the notch at 400–500 nm wavelength is associated with the parasitic absorption of CdS substrate.[ ] In particular, the increased response at long wavelength (500–1000 nm) for Se‐rich device is attributed to the improved electrical quality of Sb2Se3 film which will be analyzed from defect and carrier transport perspective later. Furthermore, the integrated J SC of Sb‐rich (22.98 mA cm−2) and Se‐rich (26.26 mA cm−2) devices obtained from EQE accords well with that of J–V characteristics. Notably, the absorption onset for Sb‐rich and Se‐rich Sb2Se3 films calculated from differential EQE (Figure S4, Supporting Information) both are 1010 nm, which is corresponding to 1.23 eV band gap according to E g = hc/λ (where E g, h,c, and λ are respective for band gap, Planck constant, light speed and wavelength), and the result echoes the UV–vis spectra.

We then perform DLTS characterizations, which are selected with optical and electrical dual‐pulse mode. Multiple pulse voltages are applied thus to realize repeatable and convincing results. In addition, as for N type Sb2Se3, the positive peaks and negative peaks represent electron and hole traps, corresponding to donor and acceptor defects respectively. Therefore, as shown in Figure  , there are one electron trap (E1) and one hole trap (H1) observed in Sb‐rich Sb2Se3 sample, while there are two hole traps (H2 and H3) identified in Se‐rich Sb2Se3. The detailed defect information including trap level (E T), capture cross section (σ) and trap density (N T) extracted from DLTS signals are summarized in Table  . In detail, the trap E1 lies at 0.57 eV below CBM, and the trap H1 is at 0.63 eV above VBM in Sb‐rich Sb2Se3 (depicted in Figure 3c). Interestingly, considering the band gap of 1.2 eV, they exactly lie at the same position (mid‐gap) in forbidden band, even though one of them manifests donor characteristic, the other displays acceptor property. Furthermore, the trap E1 and H1 also show similar trap density and capture cross section. Interestingly, a theoretical study of Savory and Scanlon revealed SbSe antisite defect in Sb‐rich Sb2Se3 is amphoteric and possesses equivalent capability of capturing electron and hole simultaneously, resulting in similar trap energy level, capture cross section and trap density for electron and hole.[ ] Accordingly, we can reasonably assign the two trap signals as one amphoteric defect. Furthermore, SbSe defect shows much high concentration (at ≈1014 cm−3 level), owing to its quite low defect formation energy compared with formation enthalpy of Sb2Se3 phase. Therefore, this defect plays crucial role in carrier recombination in solar cells.

Figure 3

a,b) Dual‐pulse mode DLTS signals under variable pulse voltage (0.2–0.6 V) for Sb‐rich and Se‐rich Sb2Se3 devices. c,d) Schematic diagram of energy band and defect level of Sb‐rich and Se‐rich Sb2Se3. E C, E V, and E F stand for CBM, VBM, and Fermi level, respectively.

Table 1

Summarized defect information obtained from DLTS signals of Sb‐rich and Se‐rich Sb2Se3

Sb2Se3	Trap	E T [eV]	σ [cm2]	N T [cm−3]
Sb‐rich	E1	E C−0.57 ± 0.04	(0.04–3.89) × 10−15	(0.05–2.68) × 1014
	H1	E V+0.63 ± 0.07	(0.03–1.20) × 10−15	(0.11–2.10) × 1014
Se‐rich	H2	E V+0.55 ± 0.06	(0.01–3.57) × 10−15	(0.23–2.09) × 1013
	H3	E V+0.66 ± 0.07	(0.03–5.03) × 10−15	(0.07–2.51) × 1013

In Se‐rich Sb2Se3 sample, the traps H2 and H3 that locate at 0.55 and 0.66 eV above the VBM (Figure 3d) are respectively assigned to VSb vacancy and SeSb antisite defects according to the calculations,[ , , ] since their rather low formation energy and suitable transition level in Se‐rich Sb2Se3 sample. It should be noted that all the atomic positions Sb1, Sb2, Se1, Se2, and Se3 may generate the respective VSb, SeSb, and SbSe defects in each sample (Figure  ), since the same kind of defect at different atomic positions possess quite close formation energy and transition level in theoretical calculations.[ , , ]

Figure 4

a,b) The perfect lattice of quasi‐1D Sb2Se3. c–e) The possible position of SbSe2, VSb2, and SeSb2 defects in Sb2Se3 lattice respectively.

Evidently, no matter whether SbSe defect in Sb‐rich or VSb and SeSb defects in Se‐rich, all of them are ultra‐deep‐level defects with active energy much higher than 0.025 eV.[ , ] Therefore, they are hardly ionized but act as recombination centers, resulting in trap‐assisted SRH recombination, it is thus detrimental to carrier lifetime. In particular, the amphoteric SbSe defect in Sb‐rich Sb2Se3 possesses a similar ability of trapping electron and hole, which means the comparative possibility of pinning electron and hole quasi‐Fermi level, while the VSb and SeSb in Se‐rich are only associated with hole trapping. Furthermore, the higher defect density of SbSe compared with VSb and SeSb would generate more severe decline of carrier lifetime (τ) according to SRH model (τ ∝ (σ N T)−1).[ , ]

To figure out the effect of deep‐level defect on the carrier transport dynamics, we then carry out TAS characterizations. The time window of time‐resolved transient spectra is of 0–2000 ps. A 360 nm pulse laser is selected as source to illuminate the Sb2Se3 films deposited on soda lime glass. According to the TAS mapping (Figure  ), we find that there are only one absorption peaks near 690 nm wavelength detected in both films, which belongs to the characteristic absorption peak of Sb2Se3. Thereout, we obtain the decay kinetics of two films tracked at 690 nm wavelength (Figure 5c). Additionally, due to the absence of electron donor and acceptor layer, the decay kinetics reflects the whole relaxation of exciton in Sb2Se3 to ground state through recombination.[ ] Finally, the kinetics is fitted by biexponential decay model, and the results are listed in Table S3 (Supporting Information). The carrier lifetime for Sb‐rich and Se‐rich Sb2Se3 are 7.66 and 16.15 ns, respectively. The increased carrier lifetime of Se‐rich Sb2Se3 exactly suggests the less defect induced SRH recombination compared with Sb‐rich. Hence, TAS study consolidates the deep‐level defect analysis and interprets the enhanced photovoltaic performance and photoelectric response (Figure 3b) in Se‐rich device.

Figure 5

a,b) TAS mapping of Sb‐rich and Se‐rich Sb2Se3 films deposited on glass substrate. c) Transient decay kinetics (scatter) and fit (solid lines) monitored at 690 nm wavelength for Sb‐rich and Se‐rich Sb2Se3 films.

To further study effects of Sb2Se3 composition on deep‐level defects, we also prepare Sb2Se3 films with Se/Sb atomic ratio of 1.50 (stochiometric) and 1.60 (high Se‐rich) by delicately tuning the addition of Se or Sb during thermal evaporation (Table S4, Supporting Information). Then we conduct DLTS and TAS measurements for stoichiometric and high Se‐rich Sb2Se3 (Table S5, Figures S5 and S6, Supporting Information). The DLTS results demonstrate that stoichiometric and high Se‐rich Sb2Se3 possess similar trap states with Se‐rich (Se/Sb ≈1.56). They all exhibit two hole traps H2 and H3 corresponding to VSb and SeSb defect. Additionally, the trap density of stoichiometric Sb2Se3 is much approximate to Se‐rich condition, correspondingly, whose carrier lifetime (15.58 ns) gained from TAS is also close to that of Se‐rich (16.15 ns). It implies that the variation of element composition within a certain range, i.e., close to the stochiometric and Se‐rich Sb2Se3, may not lead to essential changes in the deep‐level defect types. However, the trap density of VSb and SeSb are increased with the higher Se/Sb atomic ratio, which induce reduced carrier lifetime.

To establish the correlation between SRH recombination, V OC loss and defect properties, we examine the dark J–V curves of optimal Sb‐rich and Se‐rich devices (Figure S7, Supporting Information). In comparison, the Se‐rich Sb2Se3 device delivers lower ideality factor (A = 1.47) and reverse saturate current (J 0 = 2.15 × 10−4 mA cm−2) than that of Sb‐rich (A = 1.62, J 0 = 7.36 × 10−3 mA cm−2). In principle, A and J 0 are strongly related to the rectification of diode. The ideality factor value falls into 1 to 2, suggesting the hybrid recombination mechanism from interface and bulk simultaneously.[ ] Besides, the reduced A value for Se‐rich device suggest the more efficient suppression of trap‐assisted SRH recombination in bulk.[ ]

In addition, since the V OC is inversely associated with J 0 in line with Equation (1), the reduced J 0 in Se‐rich device also contribute to enhanced photovoltaics performance. where k B, T, and q are Boltzmann constant, temperature, and elementary charge, respectively. In practice, J 0 consists of two parts, one is the leakage current caused by interface recombination (JIF 0), and the other is leakage current caused by SRH recombination in space charge region (SCR, JSCR 0) according to Equation (2).[ ]

Furthermore, the JSCR 0 is inversely proportional to carrier lifetime (Equation (3), the N C/V, F m, and E g are state density in conduction (valence) band, the electric field at the position of maximum recombination and band gap.).[ ] Accordingly, we can draw a conclusion that V OC is positively related to carrier lifetime. It suggests the decrescent carrier lifetime stemming from the defect‐induced SRH recombination is the predominant cause for V OC loss. Notably, Se‐rich Sb2Se3 possesses relatively lower defect concentration and prolonged carrier lifetime when compared with Sb‐rich once, indicating that the Se‐rich Sb2Se3 is more promising for obtaining high efficiency device in terms of increasing the carrier lifetime and resultant V OC.[ ]

Conclusion

In summary, we reveal the stoichiometry depended defect properties in Sb2Se3 films. We confirmed the purity of the as‐synthesized Sb2Se3 films in terms of crystal structure, compositional elements as well as the chemical bond, which provides solid ground for defect analysis. We experimentally disclose the theoretically predicted amphoteric SbSe defect in Sb‐Sb2Se3, enriching the fundamental understanding in the defect study. We also provide new perspective in different challenges in the Sb‐rich and Se‐rich Sb2Se3 based solar cells. In Se‐rich Sb2Se3 solar cell, it has relatively low defect concentration and shallower energy level, showing promise in the efficiency improvement. However, although the Sb‐rich Sb2Se3 sample possesses two kinds of defect, they are actually coming from one kind of element substitution. Practically, it simplifies the defect engineering in materials synthesis.

Experimental Section

Preparation of Sb2Se3 Films

Sb2Se3 films were deposited on substrate using thermal evaporation deposition under a pressure of 5 × 10−4 Pa. Sb2Se3 (99.99%, zhongnuoxincai) powder was used as evaporation source, and an amount of Sb or Se (99.999%, Sinopharm) powder was used as co‐evaporation source to tailor the composition of Sb2Se3 film. The Sb2Se3 vapor was deposited on substrate (315 °C preheated) with an evaporation rate of 4–6 nm s−1. The final thickness of films was controlled around 300 nm. Finally, the as‐deposited films were post‐annealed at 380 °C for 8 min on a preheated hot plate in a N2‐filled glove box.

Fabrication of Sb2Se3 Solar Cells

The FTO glass (TEC‐A7) was cleaned by DI water, isopropanol, acetone, and ethanol firstly, and then was cleaned for 15 min by UV ozone. Next, a 60‐nm CdS ETL was deposited on FTO by CBD method.[ ] Subsequently, the Sb2Se3 film was deposited on FTO/CdS substrate by thermal evaporation method as described above. Afterwards, 90‐nm thick Spiro‐OMeTAD (doped with 520 mg mL−1 Li‐TFSI acetonitrile solution) layer was selected as HTL.[ ] Finally, the 60‐nm Au back electrode was evaporated on the HTL under a pressure of 5 × 10−4 Pa. The active area was defined as 0.04 cm−2 by a metallic mask.

Characterization of Films and Devices

The SEM images of Sb2Se3 thin films were characterized by FE‐SEM (Hitachi SU8220) equipped with an EDS (Bruker) module. The AFM topographics for Sb2Se3 films were characterized by Bruker Dimension Icon. The crystal structure was measured by XRD (Bruker Advance D8 diffractometer) with Cu Kα radiation (λ = 1.5406 Å). Raman spectroscopy was carried out using a Renishaw Raman spectrometer with 532 nm laser excitation. XPS spectra of the films were conducted on a Thermo Scientific K‐Alpha+ instrument equipped with a monochromatic Al Kα X‐ray source. UV−vis spectrophotometer (SOLID3700) was used to measure the light absorption properties of the Sb2Se3 films. UPS spectra was conducted via a PHI5000 VersaProbe III (Scanning ESCA Microprobe) SCA (Spherical Analyzer), and a He (I) discharge lamp (21.2 eV) was selected as the UV light source. The work function was calculated by subtracting the secondary electron cutoffs from the excitation light energy (21.2 eV). The Hall effect measurement was exerted by Ecopia Hall Effect Tester (HMS‐7000) at 300 K. TAS of Sb2Se3 films were measured by a commercial Helios setup from Ultrafast System. Additionally, the pump and probe laser pulses were generated by frequency doubling the fundamental output (Coherent Vitesse, 80 MHz, Ti‐sapphire laser) and through white light generated with a sapphire plate, respectively. The decay kinetics were fitted by biexponential model y = ΣA iexp(−x/t i), and carrier lifetime (τ) was obtained by τ = ΣA i t i 2/ΣA i t i (i = 2). A Keithley 2400 apparatus along with solar‐simulated AM 1.5 G sunlight (a standard xenon‐lamp‐based solar simulator (Oriel Sol 3A)) was used to examine the J–V characteristics. The solar simulator was calibrated before test by a standard monocrystalline silicon (Oriel P/N 91150 V, with a KG‐5 visible color filter), calibrated by NREL. The EQE values of the devices were measured using an illumination system with a single source (halogen lamp) in connection with a monochromator (Model SPIEQ200). EIS was performed at a bias of 0.2 V in dark with the frequency ranging from 1 MHz to 1 Hz based on Zahner Mess System PP211 electrochemical workstation. The DLTS measurement was performed via a Phystech FT‐1230 HERA‐DLTS system equipped with a 10‐mW red laser (635 nm wavelength). The modified Boonton 7200 capacitance meter was used to measure dynamic capacitance. A complete device with area of 0.01 cm−2 was selected to test. The temperature scan range was from 120 to 420 K at 2 K heating intervals. The dual pulse mode was set as electrical (pulse voltage) and optical (laser excitation), and they were exerted and removed simultaneously. In specific, the reverse bias, pulse voltage, pulse width (electric & optical), and period width were −0.4 V, 0.2–0.6 V, 10 ms, and 100 ms, respectively.

Conflict of Interest

The authors declare no conflict of interest.

Supporting Information

Click here for additional data file.