1 Introduction

Considering comprehensive factors such as element abundance, cost, efficiency, environmental protection, and stability, the research and development of novel photovoltaic materials have never stopped [1,2,3]. The unique electronic structure of the Pb2+ plays a very critical role in the outstanding photovoltaic performances of perovskite materials [4,5,6,7,8], and it could be replaced by a suitable metal or multi-metal combination by a green chemical synthesis strategy, for the construction of environmentally friendly and high-quality inorganic photovoltaic thin films with ideal bandgaps, which have shown great application potential [9,10,11,12,13,14]. Particularly, isoelectronic Bi3+ possesses the same electron arrangement (6s26p0) and similar ionic radius with Pb2+. From the point of view of the principle of charge balance (2Pb2+ → Ag+  + Bi3+), the lead-free Cs2AgBiBr6 double perovskite material based on the metal combination of Ag+ and Bi3+ has a good 3D structure, high absorption coefficient (≈ 105 cm−1), long carrier lifetime, and high temperature/humidity stability, which provide the possibility and diversity for designing more ideal light-absorbing materials [10, 12, 15]. However, most of the reported Bi-based inorganic double perovskite materials show indirect bandgaps, and it is difficult to achieve the ideal bandgap values for single-junction solar cells [16].

In recent years, the design and performance optimization of perovskite-like Cu-Ag-Bi-based ternary halide photovoltaic materials with suitable bandgaps, such as CuBiI4 and Ag1−3xBi1+xI4, have attracted the attention of researchers [17,18,19]. However, due to the intrinsic influence of its own structure, the further improvement in both efficiency and stability is limited. In 2021, Rosseinsky's group [20] first reported a novel Cu2AgBiI6 multi-halide photovoltaic material inspired by the design of the double perovskite structure. The material has a direct bandgap of 2.06 eV, an ultra-high optical absorption coefficient of 1.0 × 105 cm−1, a substantial charge-carrier mobility of 1.7 cm2V−1 s−1, and a long photoluminescence lifetime of 33 ns. However, the power conversion efficiency (PCE) is only 0.43% due to the large bandgap and poor film quality. Recently, it was found that the bandgap has little change by varying Cu+ ion content under the premise of keeping the Ag/Bi atomic ratio (1:1) [21]. Chang et. al. [22] introduced pyridine additives to improve the film quality and got a significant increase in PCE to 1.00% for Cu2AgBiI6 solar cell devices. And then, the crystallization kinetics of the film was controlled by adding 1.5 vol% HI, which improved the surface coverage and large crystallinity, resulting in an enhanced PCE of 1.3% [23]. As far as we know, the best efficiency of up to 2.39% is achieved by the hot-casting process [24]. Although such a new Cu-Ag-Bi-I-based compound has shown great potential as high-quality photovoltaic material, the current reported bandgap of ~ 2.0 eV still greatly deviates from the ideal bandgap [20,21,22, 25], which is still subject to structural design limitations with Ag/Bi ratio maintaining at 1:1. Another main problem lies in that the non-redox compound reaction among CuI, AgI, and BiI3 is prone to problems such as the random growth of secondary phases and poor film quality [20, 22,23,24].

While both free from the influence of lead and beyond the design concept of traditional double perovskite structure (Ag/Bi = 1:1), how to play the role of elemental chemistry to tune the optoelectronic properties of such Cu-Ag-Bi-I based compounds is a huge challenge for researchers. The electron transition from the electron-rich halogen to metal plays an important role in the spectral absorption of ionic crystals. The electron transition from I− to Bi3+ is easier than that from I− to Ag+ due to the higher oxidation number (III) of Bi3+. Thus, Bi-I coordination bond can absorb light with a longer wavelength, while the Ag-I coordination bond can absorb a shorter light. As a result, a smaller fraction of silver bismuth and a larger fraction of bismuth octahedra would be beneficial to reduce the bandgap. The fabrication of films based on metal elemental surface reaction is a kind of redox reaction with elemental material sputtered film as raw material, which is beneficial to tune the element ratio and improve the quality of film formation [17, 26,27,28].

In this work, the intelligently designed lead-free inorganic CuaAgm1Bim2In absorber layer and the hole transport layer (CuI) were simultaneously fabricated by one-step gas–solid-phase diffusion-induced direct metal surface elemental reaction (DMSER) of copper-silver-bismuth-iodine at a low temperature. The bandgap of CuaAgm1Bim2In could be reduced to 1.78 eV by modifying the sputtered Cu/Ag/Bi metal film thickness due to the destroy of the Ag/Bi ratio at 1:1. The CuAgBi2I8 solar cell with an FTO/TiO2/CuaAgm1Bim2In/CuI/carbon structure gained a champion efficiency of 2.76% for the single-junction solar cells. This approach of gas–solid-phase diffusion-induced elemental reaction is very simple and convenient, which presents a potential pathway to large-scale industrial production of Cu-Ag-Bi-I photovoltaic film materials.

2 Experimental and Characterization

2.1 Materials

The FTO substrates (1.5 × 1.5 cm2, 8 Ω) were purchased from Shanghai Zaofu New Materials Co., Ltd. Copper target (99%), bismuth target (99%), silver target (99%)) were purchased from Beijing Zhong Cheng Company. Acetone (C3H6O, 99.5%), isopropanol (C3H8O, 99.5%), anhydrous ethanol (C2H6O, 99.5%), and nitric acid (HNO3, 68%) were purchased from Sinopharm Chemical Reagent Co., Ltd. Carbon paste was obtained from Shanghai Mater Win New Materials Co., Ltd.

2.2 Fabrication of CuaAgm1Bim2In/CuI Thin Film

First, the FTO was ultrasonic cleaned with detergent, deionized water, acetone, isopropanol, and anhydrous ethanol for 30 min, then blow-dried with N2, and irradiated with ultraviolet ozone for 30 min to remove organic matter on the surface. Second, 180 nm of Bi, 90 nm of Cu, and 60 nm of Ag thin films were deposited on the treated FTO substrates by a magnetron sputtering instrument under the Ar2 atmosphere. Third, 0.1 g of I2 particle and the sputtered Bi/Cu/Ag films were put in a sealed Teflon-lined stainless autoclave and heated at 100°C for 24 h. Finally, a gray-black CuAgBi2I8/CuI thin film was obtained on the FTO substrate. Cu0.7AgBi2I7.7 and Cu0.6AgBi2I7.6 thin films were fabricated by using the same procedure except for the thickness of Bi, Cu, and Ag. The Bi, Cu, and Ag thickness of Cu0.7AgBi2I7.7 films is 180, 60, and 60 nm, respectively. The Bi, Cu and Ag thickness of Cu0.6AgBi2I7.6 films is 120, 60, and 60 nm, respectively.

2.3 Preparation of CuaAgm1Bim2In Thin Film

The pure CuaAgm1Bim2In film was obtained by etching CuI with 0.5 mol L−1 HNO3 for about 14 s.

2.4 Fabrication of FTO/c-TiO2/m-TiO2/CuaAgm1Bim2In/CuI/C Solar Cell

The commercial FTO substrates with etched line were ultrasonically cleaned with detergent, deionized water, acetone, isopropanol, and anhydrous ethanol for 30 min in subsequence, then blow-dried with N2, and irradiated with ultraviolet ozone for 30 min for removing organic on the substrate surface. 80 μL of c-TiO2 precursor was spin coated on the treated FTO substrates at the speed of 7000 rpm for 30 s, followed by annealing in air at 500°C for 2 h. The mesoscopic TiO2 (m-TiO2) layer was deposited by spin-coating the TiO2 colloid (30 s, 2000 rpm) and annealing it in air at 450°C for 30 min. After the fabricated TiO2 layer was treated by UV-ozone for 30 min, three metal monolayer films of Bi, Cu, and Ag metal thin film were sequentially deposited using a magnetron sputtering instrument under Ar atmosphere. And then, the sputtered metal films and iodine pellets were put into a sealed Teflon-lined stainless autoclave and heated at 100°C for 24 h to afford a structure of FTO/c-TiO2/m-TiO2/CuaAgm1Bim2In/CuI thin film. Finally, the commercial low-temperature conductive carbon paste was scraped onto the above prepared film with a blade and heated at 100°C for 10 min to prepare electrode with the active area of 0.04 cm2.

2.5 Characterizations

The X-ray powder diffraction (XRD) patterns were obtained by using an X-ray diffractometer (Bruker, D8-Advanced) with Cu Kα radiation in a step size of 0.02°. Raman spectra were recorded by Raman spectrometer (Renishaw In Via). The X-ray photoelectron spectra (XPS) and ultraviolet photoemission spectra (UPS) were measured via an X-ray photoelectron spectrometer (ESCALAB Xi, Thermo Fisher Scientific). The top-view and cross-sectional morphologies, and element component were investigated through scanning electron microscopy (SEM) with an energy-dispersive X-ray (EDS) (Nova NanoSEM 50). The absorption curves were measured by a UV–vis-NIR spectrophotometer (Cary 5000 UV–Vis, Varian). The time-resolved photoluminescence (TRPL) was collected by a fluorescence spectrophotometer equipped with an excitation wavelength of 465 nm. The TSPV signals were collected by an oscilloscope (Tektronix TDS 3054C, 500 MHz) when the film was irradiated by a 355-nm pulsed laser. Current density–voltage characteristic (J-V) curves were measured in air under AM 1.5G sunlight generated by a solar simulator.

3 Results and Discussion

3.1 Fabrication and Characterization of CuaAgm1Bim2In Films

A series of CuaAgm1Bim2In quaternary compound films were prepared by the gas–solid-phase reaction method of copper-silver-bismuth-iodine at low temperature (Fig. 1a). In short, the definite thicknesses of Bi, Cu and Ag metal layers (the total metal thickness 240–330 nm) were successively sputtered on FTO glass in sequence (Figs. S1-S3), which reacted with excess iodine gas to afford Cu0.6AgBi2I7.6, Cu0.7AgBi2I7.7, and CuAgBi2I8 films at the bottom and CuI upper layer in a sealed Teflon-lined stainless autoclave. This pure phase CuAgBi2I8 powder was prepared by spin coating and annealing of iodide solution. The obtained samples were characterized by XRD. Figure 1b shows the XRD patterns of CuaAgm1Bim2In/CuI films, CuAgBi2I8 powder and the calculated structure by material studio (MS). These peaks at 22.8°, 25.2°, 29.7°, 41.5° should be assigned to the F-43 m cubic lattice of CuI (JCPDS No. 96–101-1240), indicating that a layer of CuI forms on the surface of CuAgBi2I8 film in the gas–solid-phase redox reaction, while the peaks at 12.6°, 24.1°, 25.5°, 29.1°, 38.6°, 42.8°, 49.7°, and 51.5° should be attributed to (111), (311), (222), (400), (333), (440), (622), and (444) planes of the relevant cubic (Fd \(\overline{3 }\) m) phases of the CuAgBi2I8 quaternary compound. The structure of CuAgBi2I8 was obtained by replacing half of the Cu atoms with Ag atoms in CuBiI4 structure (ICSD No. 71533) (Fig. 1c), which is distinct with the alternating layer structure of Cu2AgBiI6 with both octahedral and tetrahedral sites filled with Cu+, Ag+, Bi3+, or I− ions [20].

Fig. 1
figure 1

a Schematic of CuaAgm1Bim2In/CuI by the gas–solid-phase diffusion-induced direct metal surface elemental reaction (DMSER) of copper-silver-bismuth-iodine at a low temperature through tuning the Cu/Bi/Ag ratio. b XRD patterns spectra of CuaAgm1Bim2In/CuI, CuAgBi2I8 powder and calculated CuAgBi2I8 structure by MS software. c Crystal structure diagram of CuAgBi2I8 obtained by replacing Cu with Ag in the CuBiI4 structure. d HR-TEM image of CuAgBi2I8 and FFT pattern from red region and yellow region of CuAgBi2I8

Full size image

To clarify the structure difference between CuAgBi2I8 and Cu2AgBiI6, the upper CuI layer was etched by diluted HNO3 firstly. As shown in cross-sectional SEM images (Figs. S4-S6), the CuaAgm1Bim2In/CuI bilayer structure can transfer to single CuaAgm1Bim2In layer after removing CuI by diluted HNO3 etching. Compared with the reported Cu2AgBiI6 structure [18, 29, 30], the typical peaks of CuAgBi2I8 at 2θ = 24.1°, 42.8°, and 51.5° shifted gradually to lower degrees (Fig. S7a), further confirming that the structure of CuAgBi2I8 is different from that of Cu2AgBiI6. The high-resolution transmission electron microscopy (HR-TEM) image of CuAgBi2I8 is shown in Fig. 1d. The clearly interplanar lattice spacings of 0.216 and 0.230 nm are indexed to crystal planes (440) and (333) of CuAgBi2I8, respectively. Therefore, the CuAgBi2I8 compound was successfully obtained. In the same preparation process, we regulate the ratio of Cu/Bi/Ag to prepared Cu0.6AgBi2I7.6, Cu0.7AgBi2I7.7 compound film. According to the XRD pattern of CuaAgm1Bim2In films before and after etching (Fig. 1b and S7b), the peak positions of Cu0.6AgBi2I7.6, Cu0.7AgBi2I7.7, and CuAgBi2I8 have not changed except for the relative strength. Moreover, we further carried out a series of control experiments and compared the binary and ternary compounds with CuAgBi2I8 compound by XRD patterns (Fig. S8). The results show that there are no other phases in the absorption layer, and the absorption layer should be a single phase.

XPS was carried out to further investigate the composition of the CuaAgm1Bim2In film after etching. The full XPS survey shows that all the expected elements Cu, Ag, Bi, and I were observed on the CuAgBi2I8 film after etching in Fig. S9. From the XPS spectra of Cu 2p of CuAgBi2I8 film and superficial CuI (Fig. S10), the two peaks with binding energies of 952.2 eV (2p1/2) and 932.2 eV (2P3/2) were observed obviously. The splitting of the spin orbit components is 20.1 eV, which is in accord with Cu+ [31, 32]. Furthermore, the Cu 2p peak positions of CuAgBi2I8 film are consistent with CuI, indicating that Cu+ is present in both CuI layer and CuAgBi2I8 layer. The Bi 4f5/2 and Bi 4f7/2 peaks at binding energies of 159.0 and 164.3 eV (Fig. S11) are associated with the iodide-bound bismuth (III) [33]. The Ag 3d spin orbitals (3d5/2 at 368.2 eV and 3d3/2 at 374.2 eV) with doublet separation of 6.0 eV can be assigned to Ag(I) (Fig. S12) [32]. Interestingly, the I core-level spectra show that the I 3d3/2 and 3d5/2 binding energies of CuAgBi2I8 shift 0.4 eV to the left in comparison to the CuI, revealing of the iodide-bound bismuth (Fig. S13). From the Raman spectra of Cu0.6AgBi2I7.6, Cu0.7AgBi2I7.7 and CuAgBi2I8 thin films (Fig. S14), the observed peak at 111.1 and 151.7 cm−1 should be assigned to CuaAgm1Bim2In [20]. Interestingly, the peak position at 151.7 cm−1 never changes as the proportion of Cu/Ag/Bi elements, providing the further proof that the modified CuaAgm1Bim2In quaternary compound should have the same crystal structure.

The evolution mechanism of CuAgBi2I8/CuI bilayer structure was studied by a series of controlled experiments. From the XRD patterns of CuAgBi2I8 compounds at different reaction time (Fig. S15), the typical peaks of CuBiI4, CuI, BiI3, and AgI could be found after one hour of reaction. When the reaction time was extended to 6 h, these peaks of CuAgBi2I8, CuBiI4, CuI, BiI3, and AgI coexisted on the XRD pattern. After 24 h of reaction, the pure CuaAgm1Bim2In/CuI bilayer structure was formed. The proposed reaction equation is as follows:

$${\text{2Ag }}\left( {\text{s}} \right){\mkern 1mu} + {\text{ I}}_{2} \;\left( {\text{g}} \right) \to {\text{2AgI }}\left( {\text{s}} \right)$$
(1)
$${\text{2Cu }}\left( {\text{s}} \right){\mkern 1mu} + {\text{ I}}_{2} \;\left( {\text{g}} \right) \to {\text{2CuI }}\left( {\text{s}} \right)$$
(2)
$${\text{Cu }}\left( {\text{s}} \right){\mkern 1mu} + {\text{ Bi }}\left( {\text{s}} \right){\mkern 1mu} + {\text{ 2I}}_{2} \;\left( {\text{g}} \right) \to {\text{CuBiI}}_{4} \;\left( {\text{s}} \right)$$
(3)
$${\text{2Bi }}\left( {\text{s}} \right){\mkern 1mu} + {\text{ 3I}}_{2} \;\left( {\text{g}} \right) \to {\text{2BiI}}_{3} \;\left( {\text{s}} \right)$$
(4)
$${\text{AgI }}\left( {\text{s}} \right){\mkern 1mu} + {\text{ BiI}}_{3} \;\left( {\text{s}} \right){\mkern 1mu} + {\text{ CuBiI}}_{4} \;\left( {\text{s}} \right) \to {\text{CuAgBi}}_{2} {\text{I}}_{8} \;\left( {\text{s}} \right)$$
(5)

The standard electrode potentials of Cu, Ag, and Bi are given by: E° (Ag+/Ag) = 0.80V, E° (Cu+/Cu) = 0.52V and E° (Bi3+/Bi) = 0.31V, respectively. Thus, the order of Bi, Ag, and Cu metal activation is Bi > Cu > Ag. Since Bi has the highest metal activation, we put it at the bottom to ensure that the reaction is complete. Due to the moderate activation of Cu, we tend to place the Cu under the Ag layer. As a result, a layer of CuI was formed on the CuAgBi2I8 indicating that there is a phenomenon of outward diffusion of Cu and inward diffusion of Ag driven by potential difference. The outer layer of Ag and the middle layer of Cu can firstly react with I2 gas to form AgI and CuI, respectively. Secondly, Bi and the extra Cu could further react with I2 gas to afford CuBiI4 during the atomic mutual diffusion. Thirdly, the BiI3 was formed by the reaction of the residual Bi and I2. Finally, the generated BiI3, AgI, and CuBiI4 were combined to form quaternary CuAgBi2I8 compound film at the bottom. Therefore, the CuAgBi2I8/CuI bilayer structure was formed mainly by the atomic mutual diffusion of Cu, Bi and Ag during the redox reaction.

To determine the two-layered structure, scanning electron microscopy with energy-dispersive X-ray (SEM–EDX) was employed in characterization of morphology and composition analysis. As shown on the top-view SEM images (Fig. 2a, b, c), with the increase in the proportion of Cu in the compound, the upper grains gradually become larger and fuller, indicating the increase in Cu content can promote CuI growth. On the cross-sectional SEM images (Fig. 2d, e, f), the continuous and dense CuaAgm1Bim2In  and CuI bilayer structure can be obviously observed. The thickness of the top and the bottom layers was measured, about 550 and 480 nm for Cu0.6AgBi2I7.6/CuI, about 490 and 630 nm for Cu0.7AgBi2I7.7/CuI, and about 690 and 800 nm for CuAgBi2I8/CuI film, respectively. Take CuAgBi2I8 as an example, each layer of metal is about 60–180 nm, and the total sputtering thickness of Cu, Ag, and Bi is over 300 nm. On the one hand, to react with these metals, at least 600 nm of stacked iodine atoms is required since iodine atoms are approximately twice the total amount of Cu-Ag-Bi metals in terms of the atomic ratio of CuAgBi2I8 chemical formula. Therefore, the total thickness of Cu-Ag-Bi-I compounds may reach about 900 nm. On the other hand, the density difference between the metal and the compound must be taken into the account. Metals usually are denser than compounds. For example, Cu and CuI densities are about 9.0 and 5.6 g cm−3, respectively. And Bi and BiI3 densities are about 9.8 and 5.8 g cm−3, respectively. If the density of the metal is 1.6 times that of the compound, the total thicknesses of Cu-Ag-Bi-I compounds are roughly estimated to be 1440 nm, which is consistent with the observed thickness of CuaAgm1Bim2In/CuI films (1490 nm). Therefore, the thicknesses of Cu-Ag-Bi-I compounds and CuI are both over 400 nm.

Fig. 2
figure 2

Top-view and cross-sectional SEM images. a, d for Cu0.6AgBi2I7.6, b, e for Cu0.7AgBi2I7.7, c, f for CuAgBi2I8 thin films on FTO substrates, respectively. SEM mapping images of Cu, Ag, Bi, and I elements for g Cu0.6AgBi2I7.6, h Cu0.7AgBi2I7.7, and i CuAgBi2I8

Full size image

The average composition of Cu0.6AgBi2I7.6, Cu0.7AgBi2I7.7, and CuAgBi2I8 was calculated statistically by SEM–EDX for no less than three regions (Tables S1-S3). The Ag/Bi ratio in all CuaAgm1Bim2In is closed to 1:2. The SEM–EDX mapping images show a uniform distribution of Cu, Ag, Bi, and I elements (Fig. 2g, h, i). A small amount of Ag+ remained in the upper CuI layer. It is worth noting that the iodine content detected by EDX was lower than the expected value, which is attributed to the loss of iodine under the strong electron beam irradiation [23]. To further demonstrate the chemical composition, we scraped the powders of CuaAgm1Bim2In (after etching) from FTO glass for inductively coupled plasma (ICP) and XPS tests. ICP results show that the atomic percentage of Cu, Ag, Bi, and I in CuAgBi2I8 is 9.70%, 8.41%, 17.4%, and 64.5% (Table S4), respectively. Thus, the composition of CuAgBi2I8 sample should be Cu1.1Ag1.0Bi2.1I7.7 by ICP measurement. At the same time, the compositions of Cu0.7AgBi2I7.7 and Cu0.6AgBi2I7.6 samples were measured to be Cu0.69Ag1.0Bi1.8I8.1 and Cu0.58Ag1.0Bi2.0I8.1, respectively, which are in line with the SEM–EDX results. The semi-quantitative analysis results of XPS also show that the atomic ratio of Ag to Bi is close to 1:2, which further proves the correctness of these compositions of CuaAgm1Bim2In (Table S5). Therefore, the CuaAgm1Bim2In/CuI two-layered structure on the FTO was prepared successfully. The obtained CuI layer can be as a natural hole transport material of CuaAgm1Bim2In absorption layer, which have excellent interfacial contact with the bottom layer [34].

3.2 Optical Properties of CuaAgm1Bim2In Films

The optical properties of CuaAgm1Bim2In films after etching were characterized by UV–vis absorption spectra, TRPL decay curves, and transient surface photovoltage (TSPV) curves. As shown in Fig. 3a, with the increase in Cu content, the absorption edge shows red shift. According to the Tauc plot (Fig. 3b), the lowest value of bandgap is about 1.78 eV, 0.28 eV lower than the reported 2.06 eV [20]. Herz et al. [21] once tried to adjust the ratio of Cu by controlling the ratio of Ag/Bi to 1:1 and found that the bandgap only shifts by 0.05 eV from 2.05 eV for x = 0 (AgBiI4) down to 2.00 eV for x = 0.6 (Cu6AgBiI10). Thus, Cu content regulation cannot lead to such a large bandgap shift (0.28 eV). The modification of the ratio of Ag to Bi (atomic ratio Ag/Bi = 1:2) should play a key role on the bandgap regulation strategy. TRPL decay measurements show that the average carrier lifetime (τave) increased from 81.3 ns for Cu0.6AgBi2I7.6 film to 201 ns for CuAgBi2I8 film (Fig. 3c, Table S6). The PL lifetime of CuAgBi2I8 film is about 6 times that of the reported Cu2AgBiI6 film, indicating that the CuAgBi2I8 film has more advantages in charge separation. TSPV measurements of CuaAgm1Bim2In films were carried out by using a 355-nm pulsed laser with a pulse width of 4 ns in according to the reported method [35]. A positive signal means that these CuaAgm1Bim2In films have n-type semiconductor properties due to the accumulation of photoinduced electrons at the surface (Fig. 3d) [17, 35]. The photoinduced electron recombination time of CuAgBi2I8 film is about 5.48 × 10−3 s at the highest decay slope, which is higher than those of Cu0.6AgBi2I7.6 film (2.91 × 10−3 s) and Cu0.7AgBi2I7.7 (3.16 × 10−4 s). The photovoltage of CuAgBi2I8 film is the biggest among the CuaAgm1Bim2In films, indicating of the highest photoinduced charge carrier concentration (Table S7). Charge carrier mobilities play a significant role in the performance of photovoltaic devices. We determined effective charge-carrier mobilities of CuAgBi2I8 by using Hall effect measurement. The value for the electron–hole total mobility is 2.80 cm2V−1 s−1 (Table S8), which is higher than that of double perovskite Cs2AgBiBr6 (0.8 cm2V−1 s−1) [36] and Cu2AgBiI6 (1.7 cm2V−1 s−1) [20, 25]. Charge carrier diffusion length is calculated by equation of \(L_{D} = \sqrt {\tau \mu_{e} ekT/q}\), and the estimated value is 3.8 µm, indicating a preferable charge carrier diffusion length. Therefore, the CuAgBi2I8 film with a two-layered structure has the best dynamical behavior of photogenerated carriers, suggesting that it has a great application potential as an absorption layer in solar cell devices.

Fig. 3
figure 3

a UV–vis absorption spectra of CuaAgm1Bim2In films after etching. b Corresponding (αhv)2 vs energy (hv) curves of CuaAgm1Bim2In films after etching. c TRPL decay curves and d TSPV curves of CuaAgm1Bim2In films after etching

Full size image

3.3 Performance of CuaAgm1Bim2In Solar Cells

We assembled the CuAgBi2I8 film as an absorption layer into a planar heterojunction photovoltaic cell. Figure 4a shows the schematic cross-sectional view of the CuaAgm1Bim2In solar cells. The n-i-p type cell has a structure of FTO/c-TiO2/m-TiO2/CuaAgm1Bim2In/CuI/carbon. A layer of activated carbon is deposited on CuaAgm1Bim2In as a bifunctional film with a work function (− 5.0 eV) close to gold (− 5.1 eV), which can be effectively used for hole extraction and collection [37, 38]. In addition, carbon electrodes are very stable, processable, and inexpensive. The band structure of CuaAgm1Bim2In was calculated through their UPS spectra (Figs. S16-S18) and UV–vis absorption spectra (Fig. 3a). As shown in Fig. 4b, band energy levels of the FTO, TiO2, CuaAgm1Bim2In, and carbon layers represent the smooth electron transport from the CuaAgm1Bim2In to the TiO2 in the conduction band and hole extraction from the CuaAgm1Bim2In to the CuI and carbon electrode in the valence band. Figure 4c shows the J-V curve of CuAgBi2I8/CuI based solar cell device. All the CuaAgm1Bim2In solar cells with different ratio of Cu/Ag/Bi were also fabricated, as shown in Fig. S19. The corresponding photovoltaic parameters of the CuaAgm1Bim2In solar cell devices with an active area of 0.04 cm2, including the short-circuit density (Jsc), open-circuit voltage (Voc), fill factor (FF), and PCE, are presented in Fig. 4c and Table S9.

Fig. 4
figure 4

a Cross-sectional SEM image of the device with the architecture of FTO/c-TiO2/m-TiO2/CuaAgm1Bim2In/CuI/carbon. b Band alignment diagram of CuaAgm1Bim2In film. c J-V plot of champion CuAgBi2I8 solar cell. The inset shows its J-V parameters PCE, Voc, Jsc and FF. d PCE, e Voc, f Jsc and g FF statistics of 20 devices under each condition. The box lines indicate the standard deviation, and the central represents the mean value

Full size image

The PCE of CuAgBi2I8 solar cell is as high as 2.76%, which is highest value reported for this class of materials [20, 22, 24]. In detail, the CuAgBi2I8 solar cell has higher Voc (0.582 V) and Jsc (up to 10.8 mA cm−2) than the reported Cu2AgBiI6 devices (Table S9). The Jsc data were almost twice as high as the reported (5.3 mA cm−2) [24]. All these are closely related to the bandgap regulation engineering, which enhances the harvest of light owing to the reduction in the bandgap. On the other hand, the CuI layer as a natural hole transport layer is beneficial to enhancing the hole extraction. As shown in Fig. S20, our CuAgBi2I8 devices show slight J–V hysteresis, which may be due to the imbalanced charge transportation at interfaces and non-radiative recombination [39]. To demonstrate the reproducibility and statistic properties, about 20 devices were fabricated and characterized as shown in Fig. 4d, e, f, g. The PCE statistics are distributed over a small range with an average value of 2.32%. This high photovoltaic performance of CuAgBi2I8 should be due to the growing of current density (Jsc) caused by the regulation of Cu/Ag/Bi metal proportion (Fig. 4f). From the IPCE spectrum in Fig. S21, CuAgBi2I8 device has a wide spectral response in the wavelength range of 300–700 nm. An integrated current of CuAgBi2I8 is obtained from the IPCE spectra of the best cell. To furtherly evaluate the stability of CuaAgm1Bim2In solar cell, the J-V curves of CuAgBi2I8 devices were recorded after 100 days in air (Fig. S22a). The unencapsulated CuaAgm1Bim2In solar cell retains nearly 63% of its initial PCE after 100 days in air. The crystal structure of CuAgBi2I8 had not changed after 100 days in air by the determination of the XRD pattern (Fig. S22b).

4 Conclusions

In summary, we fabricated a series of CuaAgm1Bim2In films with self-generated CuI hole transport layer via the gas–solid-phase reaction of copper-silver-bismuth-iodine elements at low temperature. By tuning the sputtered Cu/Ag/Bi metal film thickness, the bandgap of the corresponding quaternary compounds could be reduced from 2.06 to 1.78 eV. Interestingly, both a continuous and dense CuaAgm1Bim2In light absorption layer and a CuI hole transport layer could form simultaneously due to the strong atomic diffusion effect of Bi, Cu, and Ag during the redox reaction process. The CuAgBi2I8 solar cell with an FTO/TiO2/CuaAgm1Bim2In/CuI/carbon structure delivers an optimum PCE of 2.76% with an improved Jsc of 10.8 mA cm−2, which is the highest reported so far. This work will pave the way on the development of fabricating a new type environmentally friendly CuaAgm1Bim2In inorganic photovoltaic materials. However, we have to admit that the current PCE is still too low in comparison to lead-based devices. At the present stage, we are not skilled in the dynamic control of this gas–solid reaction, with poor crystallinity, small grains, and many defects, leading to serious charge recombination. On the other hand, the device structure is only preliminary exploration, resulting that the interface charge transfer is not optimal. In addition, the bandgap has not reduced to an ideal range (1.4–1.6 eV). In the future, the photovoltaic properties of CuaAgm1Bim2In thin film materials will be improved by introducing the coordination system and adjusting reaction activation energy to promote the generation of large crystal grains with little defects. By further adjusting the ratio of Cu, Ag, Bi, and I to tune the ratio of tetrahedral/octahedral lattice structure, the bandgap of CuaAgm1Bim2In will be further adjusted to the ideal range.