Jingling
Liu‡
,
Kang
Gao‡
,
Hang
Cai
,
Xinyu
Wu
,
Xinsheng
Liu
,
Ke
Cheng
* and
Zuliang
Du
*
Key Laboratory for Special Functional Materials of Ministry of Education, National & Local Joint Engineering Research Center for High-efficiency Display and Lighting Technology, School of Materials Science and Engineering, and Collaborative Innovation Center of Nano Functional Materials and Applications, Henan University, Kaifeng 475004, China. E-mail: zld@henu.edu.cn; ck@henu.edu.cn
First published on 1st April 2023
Sn-free Cu2ZnGeSe4 (CZGSe) is emerging as a promising non-toxic and earth-abundant photovoltaic absorber material due to its attractive electrical and optical properties as well as its high theoretical conversion efficiency. Nevertheless, no photovoltaic device fabricated through the green electrodeposition process has yet been reported, likely due to the poor solubility of Ge-based salts and harsh electrodeposition conditions. Herein, we propose a GeSe-evoked synchronous strategy involving a Ge incorporation and selenization-regulated co-heating process of GeSe and Se, following electrodeposition of a Cu–Zn preformed layer. We experimentally found that the low-melting-point GeSe could promote the crystal growth and induce a high-quality bulk absorber layer and good back interface. In the GeSe-promoted sample, it was found that MoSe2 could ensure a good back quasi-Ohmic contact, and the band bending at the grain boundaries (GBs) was favorably inverted. Moreover, the depletion region width was also prolonged, and the deleterious CuZn near EF was passivated, leading to an increased carrier separation. In turn, a surprising progress in device performance was found, achieving a ground-breaking efficiency of 3.69%, and it could fill the bank of green electrodeposited CZGSe-based solar cells.
To overcome the Sn-associated issues, the substitution of Ge for Sn has shown enormous beneficial effects on kesterite-based devices, including favoring the crystallinity growth, reducing Sn-related defects, and improving the carrier lifetime.9–11 Inspired by this, Sn-free Cu2ZnGeSe4 (CZGSe) absorbers have begun to attract significant interest recently based on the following favorable aspects: (1) CZGSe has a higher Shockley–Queisser (SQ) limit of 33.3% due to its optimal band gap. Its trap-limited conversion (TLC) efficiency of 24.1% is also higher than that of Cu2ZnSnSe4 (CZTSe, 20.3%) and Cu2ZnSnS4 (CZTS, 20.9%);12 (2) CuZn defects are suppressed in CZGSe as compared to in Sn-based kesterite, as GeZn-induced recombination is lower than that of SnZn due to its deeper donor level;12–14 (3) the sub-band-gap absorption induced by disorders is less, because of the larger formation energies of stannite and kesterite phases (64 meV for CZGSe, 42 meV for CZTSe), leading to a low Eu value.12,13 A highest Voc of 744 mV in CZGSe-based solar cells was reported by Sahayaraj et al. with a PCE of 5.5%.15 A record efficiency of 8.5% was achieved through interface engineering.16
Despite its advantages over Sn-based kesterite solar cells, its efficiency remains low. This is mainly ascribed to defect complexation (CuZn–GeZnetc.) rooted in the similar redox activity of Ge to Sn, leading to a high nonradiative recombination in CZGSe.12,13,17 Besides, the band alignment between CZGSe and CdS is a cliff-type, which works against it in photovoltaic applications.16,18 Therefore, more efforts are required to fully understand and optimize this material. The phase impurities often serve as recombination centers, causing serious carrier recombination, and are thus detrimental to device efficiency. Several strategies have been produced to address this, such as vacuum-based magnetron sputtering/electron beam evaporation and organic solvents-based solution process.15,19–24 However, these strategies generally involve high costs, less security, and/or intricate deposition technologies. Thus, to explore low-cost, green, and simple new synthesis technology, optimizing the appropriate absorber conditions and purifying the phase impurities are the most important priorities at present.
In this work, high-quality CZGSe thin films were achieved through a simple combination of the green electrodeposition and selenization processes. In view of the poor solubility of Ge-based salts and harsh electrodeposition requirements, GeSe with a low melting point and high saturated vapor pressure at evaluated temperature was selected as the Ge source. A synchronous strategy of Ge incorporation and selenization was well designed through a co-heating process of GeSe and Se following the electrodeposition of a Cu–Zn preformed layer. Furthermore, the advantages of GeSe over GeSe2 for the film quality and device performance were addressed in detail.
Two schemes are proposed according to the different characteristics of Ge sources and that include solution preparation, electrodepositing a Cu/Zn prefabricated layer, Ge incorporation, and selenization, as illustrated in Fig. 1. The following grounds are considered for these procedures: (1) Ge is quite inappropriate due to its high melting point over 900 °C; (2) to avoid the introduction of heteroatoms, only selenides, such as GeSe2 and GeSe, are the right candidates; (3) experimentally, it was found that Ge is seriously insufficient obtained through a one-step co-heating of GeSe2 and Se. Therefore, two steps were involved for the GeSe2-based process; (4) taking into account the low melting point and easy sublimation properties as discussed below, GeSe is, undoubtedly, the most suitable candidate.29,30
To guide the experimental design, the thermodynamic properties were first investigated as shown in Fig. 2a and b. Under certain conditions, the vapor pressure determines the evaporation difficulty and relative contents of the corresponding products, deeply impacting the film quality. Thus, the temperature-dependent vapor pressures for GeSe, GeSe2, Ge, and Se were calculated according to the equation: logp = A − B/T, where p is the vapor pressure, A and B are constants and their specific values are given in the literature,29 and T is the absolute temperature. As shown in Fig. 2a, GeSe had a higher vapor pressure than that of GeSe2, second only to Se. Importantly, the vapor pressure of GeSe rapidly increased at 450 °C, below its melting point of 670 °C. Although the melting points of GeSe2 (700 °C) and GeSe were similar, their vapor pressures at elevated temperatures were far different, indicating GeSe evaporated more easily than GeSe2. Ge had almost no vapor below 600 °C, thus directly ruling it out. To experimentally confirm the calculated results, TGA was performed and the results are provided in Fig. 2b. GeSe showed weight loss with a starting point near 500 °C, below its melting point, while GeSe2 had a much higher starting point near 600 °C, consistent with the above calculations. Moreover, a lower evaporation temperature was also needed for our RTP experiment design. Taking this as a guide, the setting temperature should be close to or slightly above the evaporation temperature to ensure sufficient Ge vapor species.
Before the preparation of CZGSe thin films, the Cu/Zn preformed layer was first electrodeposited onto Mo-coated SLG as provided in Fig. 2 and S1.†Fig. 2c shows that the Cu/Zn preformed layer had a dense crystalline morphology, consistent with that of electrodeposited CZTSe in our previous work.25,26 From XRD, Cu–Zn alloys were observed, probably combined with Cu and Zn metal phases. Fig. S1† shows the EDX line scans, presenting clear Cu and Zn lines with a partial overlap, which could be assigned to Cu–Zn alloys.
The typical morphology and structure characterization of the absorber layers prepared with GeSe2 and GeSe (labeled as CZGSe-A and CZGSe-B, respectively) are provided in Fig. 3 and S2.† The optimal temperature for CZGSe-A was 500 °C, and for CZGSe-B it was 600 °C. The different optimal temperatures may be due to the different Ge participating in the growth mechanism. CZGSe-A involves Ge first, followed by selenization. If the selenization temperature is too high, GeSe2 easily evaporates again to cause the loss of Ge and the partial absorber layer to possibly decompose, similar to CZTSe.31 A higher selenization temperature was tried but failed to obtain the pure CZGSe phase. Nevertheless, in the CZGSe-B sample, GeSe and Se vapors were involved together. If the temperature is low, the GeSe evaporates inadequately, which can easily cause a lack of Ge. In addition, GeSe vapor can also form liquid GexSey with Se to promote the crystallization.24 A more comprehensive and in-depth study of the formation process will be followed up in the future. Moreover, the elemental compositions of the CZGSe-A and CZGSe-B absorber layers are provided in Table S1.† The CZGSe-A and CZGSe-B samples below refer to the optimized ones used without special instructions. The SEM and EDX line mapping analyses showed that CZGSe-B had a better crystallinity and a uniform well-compact morphology. An interfacial MoSe2 layer with a thickness of ∼100 nm was observed in CZGSe-B. A thin MoSe2 layer is well known to be able to convert Schottky into quasi-Ohmic contacts and facilitate carrier transport.32,33 Thus, the CZGSe-B sample had a high-quality bulk absorber layer and good back interface, guaranteeing better device performance.
Fig. 3 (a–d) Surface and cross-sectional SEM images, and EDX line profiles of CZGSe absorbers (a and b: CZGSe-A; c and d: CZGSe-B). (e and f) XRD patterns and Raman spectra of CZGSe-A and CZGSe-B. |
Fig. 3e shows the XRD patterns of the prepared CZGSe films. Three major peaks were observed in both samples and were assigned to the kesterite phase, indicating a single pure structure.21,24 Noticeably, two other peaks were present in the CZGSe-B sample, belonging to MoSe2, consistent with the observation of the morphological results. The Raman spectra for both were also recorded, as shown in Fig. 3f, with a 785 nm excitation wavelength, from which MoSe2 could be detected. Apparently, the two peaks at 177 and 204.5 cm−1, and one at 283 cm−1 were ascribed to the A and B modes of kesterite CZGSe, respectively.21,34 The other prominent peak at 242 cm−1 could be unequivocally assigned to the A mode of MoSe2,35 confirming the previous results. Ge-based and other possible binary or ternary compounds were not detected.
Grain boundaries (GBs), corresponding to the surface defects, highly affect the band structure and charge-separation characteristics.36,37 KPFM was performed to ascertain the carrier-separation characteristics. Fig. 4 reveals the KPFM results of the surface potential (SP) and GB characteristics for the CZGSe-A and CZGSe-B absorber layers. In the SP images, the blue and yellow regions represent the high and low SP values, respectively. From the Z-position and SP plots in Fig. 4c and f, it could be observed that the GBs for CZGSe-A had higher potentials than the intragrains (IGs), whereas there were lower potentials for CZGSe-B. The SP differences between the GBs and IGs for the CZGSe-A and CZGSe-B absorber layers were statistically analyzed by counting the different regions in Fig. 4g. A positive value of 45 mV was obtained for the CZGSe-A absorber layer, indicating a higher SP was dominant at/near the GBs. Nevertheless, a negative value of −38 mV for the CZGSe-B absorber layer indicated a lower SP at/near the GBs, coinciding with the Z-position and SP plots.
A higher SP at/near the GBs means a downward band bending at the GBs, indicating that local built-in potential exists at the GBs, acting as an electron pathway and/or hole barrier.38 Therefore, electrons are attracted to the GBs, and holes are repelled to the IGs in CZGSe-A. Yet, CZGSe-B brought about the opposite conclusions of upward band bending through the statistical analysis in Fig. 4g. The band bendings for CZGSe-A and CZGSe-B are illustrated in Fig. 4h. Previous studies have demonstrated that downward band bending at the GBs is favorable for the carrier separation and electron collection in CIGS- and CZTSSe-based absorber layers.28,38,39 CZGSe is similar to CZTSe, with undesirable surface defect CuZn predominating, mainly accumulated at/near the GBs. The downward bending band may cause the recombination of acceptor CuZn defects and the collected electrons, at the expense of some photogenerated electrons at the GBs. Moreover, recent reports have illustrated that upward band bending ensures a higher device performance.40,41 Therefore, the inversion of the band bending structure in CZGSe-B suggests it possessed a better surface junction to guarantee improved carrier separation.
A traditional configuration of SLG/Mo/CZGSe/CdS/i-ZnO/ITO/Ag grid solar cell device was fabricated, as presented in Fig. 5a. Apparently, the subsequent depositions of other functional layers (CdS, i-ZnO, and ITO) had no obvious impact on the CZGSe absorber layer. The statistic photovoltaic parameters for the CZGSe-A and CZGSe-B devices are given in Fig. 5b by accessing 20 individual cells. Due to the limitations of the electrodeposited process, each parameter showed large fluctuations. Nevertheless, the overall parameters for the CZGSe-B devices revealed a distinct increase, especially for Voc and FF, indicating that the CZGSe-B devices possessed a better film quality and surface interface, along with good reproducibility. The J–V characteristics of the champion CZGSe-A and CZGSe-B solar cells are presented in Fig. 5c, and the corresponding device parameters are summarized in Table 1. The CZTSe-A device presented a PCE of 2.14%, Jsc of 14.66 mA cm−2, Voc of 0.43 V, and FF of 33.78%. The CZGSe-B device exhibited a dramatically improved PCE of 3.69%, Jsc of 16.53 mA cm−2, Voc of 0.49 V, and FF of 45.57%. The CZGSe-B device showed a higher Voc value, likely considered to be due to accelerated carrier transport due to the depressed surface defects and/or Fermi level pinning. The higher FF in the CZGSe-B device was due to the better carrier transport path through the inversion of the band bending. Moreover, it is important to note that CZGSe has not yet been prepared by an electrodeposition process. Diode parameters were extracted from the J–V curves, as shown in Fig. S3† and Table 1. The high VOC deficit (Eg/q − VOC), indicated large nonradiative recombination losses at the interface.42 This suggest that deep-level/activated shallow-level defects at the CZTSSe/CdS interface still predominated, leading to a low efficiency. Besides, the large atomic fluctuation may have aggravated the compositional inhomogeneities and disorders, activating detrimental CuZn–GeZn deep-level defect clusters and increasing the antisite/shallow-level defects, thus leading to a deterioration in the device performance, as proved by the large A. Despite the low efficiency, we achieved the highest efficiency of the electrodeposited CZGSe and made an outstanding contribution to the development of green-processed CZGSe solar cells.
Devices | J sc (mA cm−2) | V oc (V) | FF (%) | E ff (%) | R s (Ω cm2) | R sh (Ω cm2) | E g/q–Voc (eV) | E g (eV) | E u (meV) | A | J 0 (A cm−2) |
---|---|---|---|---|---|---|---|---|---|---|---|
CZGSe-A | 14.66 | 0.43 | 33.78 | 2.14 | 1.2 | 490 | 1.02 | 1.45 | 40.5 | 2.64 | 6.7 × 10−6 |
CZGSe-B | 16.53 | 0.49 | 45.57 | 3.69 | 0.4 | 701 | 0.99 | 1.48 | 33.1 | 2.02 | 1.7 × 10−6 |
The EQE spectra in Fig. 5d show the improved photoelectric response in the overall response at 400–850 nm for the CZGSe-B device. These indicated a better carrier-collection capability benefiting from the enhanced bulk quality absorber and CZGSe/CdS interface. Besides, the bandgap energies for CZGSe-A and CZGSe-B were estimated to be 1.45 and 1.48 eV, as derived from the EQE curves in the inset of Fig. 5d, respectively. The Eg value of CZGSe-B was slightly higher than that of CZGSe-A, possibly owing to the suppression of defects/secondary phases within limits and the decrease in the nonstoichiometry of CZGSe-B.43 Furthermore, the Urbach energies (Eu) were used to analyze the tail states, which are highly related to the fluctuations of the electrostatic potential caused by charged defects.43 Here, Eu was obtained by fitting the reciprocal of the slope in the linear region, as presented in Fig. 5e. CZGSe-B showed a lower Eu value of 33.1 meV, indicating reduced cation disordering.43,44
C–V measurements of the CZGSe-A and CZGSe-B devices were performed to understand their internal limitations and to evaluate their p–n junction quality.45 The corresponding carrier density (NC–V) and depletion region width (Wd) could be derived from the C–V measurements, as depicted in Fig. 6a. The carrier density reflects the interface defect density contributed from the shallow/deep defects, while the depletion width represents the charge-separation ability. The carrier density of CZGSe-B was lower than that of CZGSe-A, indicating a reduced interface states density in CZGSe-B. Moreover, the depletion width of CZGSe-B (130 nm) was larger than that of CZGSe-A (120 nm), mainly ascribed to the reduced interface defects, indicating an improved p–n junction quality. Thus, the carrier separation could be enhanced, leading to a gradually improved FF. However, the interface states density in CZGSe was still high, possibly concurrently due to the complex shallow/deep defects at/near the interface and the lattice mismatch between CZGSe and CdS.12 Besides, the CZGSe and CdS had “cliff-like” band alignments, further aggravating the interfacial recombination, causing a large VOC loss.16,18
Admittance scans for CZGSe-A and CZGSe-B devices were also recorded, where the capacitance at high frequency related to the free carrier density, while that at low frequency referred to the sum of the deep traps and free carriers.46,47 The frequency ranged from 103 to 107 Hz in the dark, along with a 30 mV AC voltage and DC bias at zero during the measurement. As shown in Fig. 6b, the capacitance of the CZGSe-B devices showed less frequency dependence than that of the CZTGSe-A devices, indicating the lower trap densities in the CZGSe-B devices, consistent with the C–V and DLTS results.
EIS was performed under AM 1.5 G solar illumination with 100 mW cm−2 light intensity to characterize the carrier transport/recombination behavior for the CZGSe-A and CZGSe-B devices in Fig. 6c. The intercept of the x-axis corresponds to the series resistance (Rs), related to the bulk/contact resistances. The diameter of the semicircle corresponds to the recombination resistance (Rp) at the interface between CZTSe and CdS.43,48 The equivalent circuit in the inset was used to fit the Rs, Rp, and C values. The Rs values for the CZGSe-A and CZGSe-B devices were 30.2 and 13.7 Ω, respectively. The lower Rs for the CZGSe-B devices was mainly due to the better bulk quality and as MoSe2 was involved in the back contact. The better surface junction formed an inversion of the band bending structure at the GBs and stimulated a visible increase in the Rp values in the CZGSe-B devices. The minority lifetimes (τ) for the CZGSe-A and CZGSe-B devices were quantified as 6.9 and 12.8 μs from the equation τ = Rp × C,43 suggesting an improved carrier transport dynamic for the CZGSe-B devices. The improved carrier transport in the CZGSe-B device may be one of the main reasons for the VOC and FF improvements.
C-DLTS was further conducted to evaluate the defects density for the CZGSe-A and CZGSe-B devices, as displayed in Fig. 6d and e. The measurements were tested in the temperature range from 120 to 240 K, which is the characteristic interval for CuZn defects.38 An apparent peak at ∼180 K was observed in both samples, suggesting a defect energy level. The formation energies (Ea) and defect concentration (NT) could be calculated from Arrhenius plots by the linear fitting of the points near the peaks. Here, Ea was 0.294 and 0.226 eV for the CZGSe-A and CZGSe-B devices, respectively, consistent with the value of CuZn defects in CZGSe-based materials.14NT was 2.83 × 1013 and 1.83 × 1013 cm−3 for the CZGSe-A and CZGSe-B devices, respectively. It is worth noting that the CuZn defect was near the surface and close to EF, which means it could easily cause Fermi level pinning and act as a recombination center, which is not conducive to the device performance. Thus, it can be inferred that the reduced CuZn defect in the CZGSe-B devices would be reflected in the inversed electrical surface properties, mainly depressing interface recombination, leading to a better device performance.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d2nr07285h |
‡ These authors contributed equally to the work. |
This journal is © The Royal Society of Chemistry 2023 |