Introduction

The inability to capture photon energy lower than the band gap of the light absorber poses a fundamental limit to the theoretical efficiency of photovoltaic (PV) cells1. Although solar cells consisting of multiple absorber layers with different band gaps can mitigate this problem and such strategy is indeed implemented in some III–V based solar cells2, the increased complexity inevitably increases the cost of the cells. Alternatively, solar cells with an intermediate band (IB) promise to extend the absorption spectrum while at the same time preserve the output voltage. The theoretical efficiency (η) of intermediate band cells can reach 63%, far beyond the single-junction silicon solar cells (η ≤ 31%)3. Band diagram of the IB cells is shown in Fig. 1. The partially-filled intermediate band can provide empty states to receive the electrons pumped by photon “3” and further pump the electrons from there by photon “2”. The solar response is significantly extended to lower energy photons, especially in near-infrared (NIR) region. As well known, the solar spectrum (on earth) is a blackbody radiation with strong visible and NIR emissions. In consequence, the multi-photon absorption of two sub-band gaps (EH and EL) efficaciously enhances the photocurrent without photo-voltage degradation.

Figure 1
figure 1

A proposed scheme of intermediate band in Sn-doped Chalcopyrites.

(a) Band diagram of an intermediate band solar cell, showing the bandgap (Eg), the sub-bandgaps (EH, EL) and the CB, IB and VB. “(1)”, “(2)” and “(3)” represent photon absorptions. (b) Band structure of Sn-doped CuMS2 (M = In, Ga).

Full size image

The intermediate band material of N-doped GaAs was estimated to achieve η ~ 40% in a multiband cell4, but the III–V thin films are fabricated by expensive metal organic chemical vapor deposition, which is not affordable for civilian uses. Cu-based chalcopyrite semiconductors are nowadays used as light absorbers in the highest efficiency thin-film solar cells (Cu(In,Ga)(Se,S)2, η = 20.3%)5,6. The solar cells possess an extraordinary defect tolerance and the possibility for economic applications. Due to the band gaps deviated from 2.4 eV for CuGaS2 (CGS) to 1.45 eV for CuInS2 (CIS), they only absorb partial solar spectrum. Thus, intermediate bands introduced to these chalcopyrites should enable them to more efficiently utilize solar energy.

Many predictions have been theoretically placed on the wide-spectrum absorption of chalcopyrites by introducing various dopants (Ti, V, Cr, Ni, etc.)7,8,9,10,11, but direct experimental evidences on the existence of intermediate band were still lack12,13. For example, the 3d-metal atoms commonly doped in CuGaS2 form IBs in the forbidden band gap from these 3d-metal orbitals7,8,9. Since the 3d states of transition metal are rather localized, the optical transitions related to IB may not be favorable due to indirect transition or low light absorption coefficients. Furthermore, the low carrier mobility on the 3d orbitals of transition metal cannot satisfy the requirement of photovoltaic cells. Thus, the transition metal might not be the best candidate to achieve high efficiently IB absorption in chalcopyrite.

In this paper, intermediate bands were introduced into CuMS2(M = In, Ga) by Sn doping (Fig. 1) to greatly enhance solar energy absorption and conversion. Nanostrcutured Sn-doped CuMS2 powder was synthesized by simple mechanically ball-milling of Cu2S and M2S3 and the corresponding thin films were prepared by a non-vacuum solution-based approach. The intermediate band in Sn-doped CuMS2 was directly evidenced by diffuse reflection spectra and photoluminescence spectra. With the aid of wide-spectrum solar response, especially from the NIR region, the more efficient solar energy conversions were yielded in photocatalytic reactions and photoelectrochemical cells.

Results

In order to evaluate the ball-milling method, the harvested CuMS2 (M = In, Ga) powders were investigated by transmission electron microscopy (TEM) (Fig. 2a). The morphologies of CuIn1−xGaxS2 are fairly uniform in size of 100–200 nm. STEM elemental mapping analysis was further examined the distribution of Sn in the doped samples. The elemental mappings of Ga, Sn and S, as shown in Fig. 2b, have a homogeneous spatial distribution. The uniform elemental distribution of CuIn1−xSnxS2 was also confirmed (see in Fig. S1 of Supporting Information). The concentration of Sn was investigated by quantitative analyses of energy-dispersive X-ray spectroscopy (EDS). The EDS measurements are in good agreement with the nominal values, as summarized in Table S1 of Supporting Information.

Figure 2
figure 2

Analyses of as-prepared CuGa0.96Sn0.04S2 particles.

(a) TEM, (b) STEM and elemental mapping of Cu. (c) XRD powder patterns of CuIn1−xSnxS2 (CIS) and CuGa1−xSnxS2 (CGS); the right: (112) peak shift. (d) Linear dependence of cell constants (a, c) on x for CuM1−xSnxS2.

Full size image

The phase purity of Sn-doped CuMS2 (M = In, Ga) was investigated by X-ray diffraction (XRD). The diffraction peaks were all indexed to a chalcopyrite structure (Fig. 2c)14 and no impurities were observed. Linear dependence of lattice constants (a, c) on the Sn-doping level (x) for CuM1−xSnxS2 reveals the typical feature of solid solution, as shown in Fig. 2d. The successful doping of Sn is also evidenced by a right shift of XRD peaks in CuIn1−xSnxS2 and a left shift in CuGa1−xSnxS2 due to the atom radius of Sn (1.41 Å) smaller than In (1.44 Å) and larger than Ga (1.26 Å).

From the above results and discussion, the ball milling is demonstrated to be an excellent method to synthesize uniform nanostructured CuMS2 powder. Quantitative composition analysis and crystal structure evolution reveal that the as-prepared samples possess both the homogeneity of element distributions and the compatibility of lattice structure. High quality of samples is due to selecting Cu2S and M2S3 as “metastable” starting materials15. The mechanic energy from ball-milling initializes the exothermic reaction of Cu2S and M2S3 to rapidly form CuMS2. Similarly, ceramic CuInSe2 was sintered by us from the reaction of Cu2Se and In2Se3 at 550°C16, whose sintering temperature is lower than the reported samples (~700°C)17. The polycrystalline CuIn0.7Ga0.3Se2 films were also prepared from Se-coated Cu(Ga) and In layers at 550°C in a few seconds18. The proposed method can avoids introducing some potential impurities, which is difficult for solution process methods19,20. This newly-developed method is applicable to synthesize the other chalcopyrites.

As shown in Fig. 3a, the color of the pressed CuGa1−xSnxS2 powder varies from orange (x = 0), dark red (x = 0.02), to dark brown (x = 0.04). The light absorptions of these powders were verified by UV-Vis-NIR diffuse reflection spectra. The plots of (αhν)2 versus hν are shown in Fig. 3b and S2a, based on the modified Kubelka–Munk function of F(R) = (1−R)2/(2R) = α/S, where R is the reflectance and α and S are the absorption and scattering coefficients. With increasing x, the onset of the absorption coefficient shifts significantly to lower photon energy, which is indexed by the direct band gap (Eg) from the extrapolated intercept of the linear portion of the plot of (αhν)2 against hν (Fig. 3b and S2b). The corresponding band gaps of CuGa1−xSnxS2 are 2.35 eV (x = 0), 1.98 eV (x = 0.02) and 1.80 eV (x = 0.04); the Eg of CuIn1−xSnxS2 are 1.42 eV (x = 0), 1.14 eV (x = 0.02) and 1.03 eV (x = 0.04). It indicates that Sn doping in CuMS2 successfully extends their solar response spectra (Fig. 3d). In order to show the detail of the absorption spectra, the low absorption coefficient part of CuGa1−xSnxS2 are magnified and shown in Fig. 3c. The extra absorption peaks at 0.7 eV and 1.7 eV are related to the absorption extra photons of hν2 (IB → CBM, >0.7 eV) and hν3 (VBM → IB, >1.7 eV). The similar extra absorption spectra were predicted or observed in other intermediate band related materials (In2S3:V, ZnTe:O etc.)30,31.

Figure 3
figure 3

IBs-induced Wide-spectrum absorption of Sn-doped Chalcopyrites.

(a) Photographs of CuGa1−xSnxS2 powders. (b) The whole and (c) the partial UV-Vis-NIR diffuse reflectance spectra of CuGa1−xSnxS2 powders; inset: the derived band gaps. (d) UV-Vis-NIR diffuse reflectance spectra of CuIn1−xSnxS2 films.

Full size image

In order to meet the requirements for the future applications on thin film photovoltaic cells, the CuIn1−xSnxS2 films were prepared by a non-vacuum method and annealed in 5 mol% H2S-contained Ar atmosphere at 600°C for 15 min. The film is polycrystalline and the average grain size is about 1 μm (Fig. S5), comparable to the CuInGaSe2 (CIGS) film in CIGS solar cells21. The CuInS2 film shows a direct optical band gap of 1.45 eV in Fig. 3d, consistent with the reported data22. With increasing x in the CuIn1−xSnxS2 film, the onset of the absorption coefficient also shifts significantly to NIR range (i.e., ~1.0 eV @ x = 0.04), which is similar to the powdered samples. As well known, CuInS2 thin films have been used as light absorber in CuInS2 solar cells with η = 11.4%23. The drawback is that CuInS2 only absorbs the visible and ultraviolet lights (λ < 825 nm). The introduction of intermediate band in the CuIn0.96Sn0.04S2 film shows a larger absorption in the solar spectrum (AM 1.5, shown in yellow background in Fig. 3d). Herein, it is promising for the CuIn1−xSnxS2 films to combine wide spectrum absorption and high open-circuit voltage in further PV applications.

The underlying reason for Sn doping to extend their absorption spectra needs further investigation. The Sn atoms in the M3+ sites have to possess the oxidation state of +3 to balance the charge, similar to Sn2S324. The coexisting Sn2+ and Sn4+ were evidenced by X-ray photoemission spectroscopy (XPS) (Fig. 4a). Two peak structures in the binding energy range from 485 eV to 500 eV observed in CuGa1−xSnxS2 are assignable to be Sn 3d5/2 and Sn 3d3/2. The asymmetric Sn 3d5/2 peak was fitted by two peaks with binding energy at 488.3 eV and 487.2 eV, assigned to Sn4+−S and Sn2+−S bonds, respectively24,25. Similarly, Sn 3d3/2 peak was also fitted by two peaks at 496.7 eV and 495.7 eV. Therefore, the partially filled Sn 5s states should be located in-between the valance band maximum (VBM) and the conduction band minimum (CBM) of CuMS2. In order to monitor the energy states of Sn, the photoluminescence (PL) spectra of the CuIn1−xSnxS2 films were performed, as shown in Fig. 4b. The green laser light (λ = 514.5 nm) was employed to excite the films at room temperature. The obvious PL peak centered at ~1.4 eV was observed in all the CuIn1−xSnxS2 films, which is identical to the optical absorption edge of CuInS2 (Eg = 1.45 eV), due to the near-bandgap emission. An extra PL peak in the NIR region (~0.78 eV) was also revealed in the Sn-doped film (e.g., x = 0.04). The recombination model of light-excited electron−hole pairs at ~0.78 eV was proposed to occur from IB to VBM, as shown in the inset of Fig. 4b. This is a solid evidence to observe the existence of an intermediate band in CuM1−xSnxS2 (x ≠ 0). The CuIn0.98Sn0.02S2 also shows an additional PL peak (~0.75 eV) at nearly same position as x = 0.04. The lower peak intensity and narrower full width at half maximum of the peak were observed, which indicate that the distribution of the IB electron states is depended on the doping level of Sn.

Figure 4
figure 4

Analyses of IBs location and band structure.

(a) Sn 3d XPS spectra of CuGa0.98Sn0.02S2. (b) PL spectra of CuIn1−xSnxS2 films; inset: schematic band structure. (c) Band structure of CuGa1−xSnxS2; the intermediate band in red. (d) Total and partial DOS of CuGa1−xSnxS2.

Full size image

Therefore, the Eg reduction in CuM1−xSnxS2, compared with CuMS2, is due to the intermediate band. According to the theory of quantum leap, the leap probability of electrons from VBM to IB and from IB to CBM is proportional to the density of states (DOS) in IB. Thus, electrons in Sn-doped CuMS2 are able to be excited from VBM to IB or from IB to CBM and some photons with hν < Eg can be absorbed. However, large Sn doping (>4 mol%) favors solar absorption, but a secondary phase (SnS2) may occurs.

Discussion

In order to further confirm the intermediate band, the electronic structures of CuM1−xSnxS2 (M = In, Ga) were calculated by first-principles theoretical calculations. The direct transition gaps at Γ (0, 0, 0) are 2.6 eV for CGS and 1.6 eV for CIS (Fig. 4c and S4a), which are consistent with the experimental data (Eg = 2.35 eV for CGS, 1.45 eV for CIS). The direct transition (VBM → CBM) usually leads a high light absorption coefficient due to needing no extra phonon momentum. Surprisingly, the band structure of Sn-doped CuMS2 is nearly same to the undoped one, except that an intermediate band appears between VBM and CBM. The intermediate band was confirmed by analyzing the DOS of the undoped and the Sn-doped CuMS2 (Fig. 4d and S4b), which is consistent with the previous calculations26. The VBM of CuMS2 consists of the Cu-3d states hybridized with S-3p and the CBM consists of M-ns (n = 4 for Ga, 5 for In) and S-3p states. The partially filled IB mainly contains Sn-5s and S-3p states, as desired for the IB scheme (Fig. 1b). Hence, CuM1−xSnxS2 has three direct bandgap transitions (VBM → CBM, VBM → IB, IB → CBM) enabling to absorb more photons.

Experimental and theoretical results confirm the existence of intermediate band and much wider spectrum solar response. Such IB materials have not been studied in photocatalysis. Hereby, CuGa1−xSnxS2 is selected to verify its IB light absorption and enhanced photocatalysis, as schematized in Fig. 5a. The photocatalytic performance is evaluated by degrading methylene blue (MB) under a xenon light source chopped by two optical high-pass filters (420 nm, 650 nm). The Sn-doped samples show much higher photocatalytic activities than the undoped one under the irradiation of visible-light and NIR spectrum (>420 nm) (Fig. 5b). The respective MB removals over CuGaS2, CuGa0.98Sn0.02S2 and CuGa0.96Sn0.04S2 are 62%, 78% and 81% after the 90 min illumination. The enhanced photocatalytic activities in the IB samples are apparently due to the wide-spectrum solar response, as derived from the UV-Vis-NIR diffuse reflectance spectra in Fig. 3b. CuGaS2 only absorbs hν1 photons (VBM → CBM, >2.35 eV) from the xenon lamp to degrade MB. The IB samples can absorb the extra photons of hν2 (IB → CBM, >0.7 eV) and hν3 (VBM → IB, >1.7 eV). In order to verify the IB absorption, a 650 nm (1.9 eV) optical high-pass filter was used to cutoff the >Eg photons and the NIR-light degradations are shown in Fig. 5c. Again, the IB samples are better by comparing the MB removals in 120 min over CuGa1−xSnxS2, 10% (x = 0), 24% (x = 0.02) and 34% (x = 0.04), which is attributed to the cascade absorption of hν3 and hν2 photons in the Sn-doped CuGaS2. The cycling degradation improvement of CuGa0.96Sn0.04S2 is nearly same in five cycles (Fig. 5d).

Figure 5
figure 5

IBs-induced photocatalysis enhancement.

(a) Schematic photocatalysis of CuGa1−xSnxS2. Photocatalytic methylene blue degradation under (b) λ > 420 nm and (c) λ > 650 nm. (d) Cycling photocatalytic degradations.

Full size image

Wide-spectrum solar response materials were further verified to improve photovoltaic conversion. The CuIn1−xSnxS2 thin films on Mo/SiO2 substrates were assembled into photoelectrochemical (PEC) cells (Fig. 6a). The electrochemical analysis was conducted in a three-electrode configuration using 0.25 M Na2S aqueous solution electrolyte. The Mott-Schottky (M-S) plot obtained at the frequency of 1,000 Hz in dark has a negative slope to indicate the p-type conduction of CuInS2 (Fig. 6c). The flat-band potential estimated from the extrapolation of the M-S plot is about 0.67 eV relative to Ag/AgCl27. The alignment of band edge of the CuInS2 film and water redox potential (the insert of Fig. 6c) indicates that the CuInS2 film is available as photocathode for PEC application. Fig. 6b shows linear sweep voltammetry under illumination (λ > 420 nm) and the photocurrent densities of Sn-doped film are higher than CuInS2 film. At 0 VAg/AgCl, the photocurrent densities were 2.83 mA/cm2 for CuInS2 and 3.52 mA/cm2 for CuIn0.96Sn0.04S2. The improvement of photocurrent densities is contributed by the wide-spectrum solar response from the intermediate band effect. In order to get an obvious evidence of this effect, a 900 nm (1.38 eV) optical high-pass filter was used to cutoff the most >Eg photons. Under the NIR light, CuIn0.96Sn0.04S2 has a photocurrent density of is 1.4 mA/cm2 at 0 VAg/AgCl, much higher than CuInS2 (0.2 mA/cm2). It was further confirmed by ON/OFF light cycling with the potential swept at 5 mV s−1 toward negative potential (Fig. 6d). The remarkable improvement in PEC properties verifies that the intermediate band increases the photocurrent densities by the cascade absorption of hν3 (VBM → IB) and hν2 (IB → CBM) photons.

Figure 6
figure 6

IBs-induced PEC enhancement.

(a) Photoelectrochemical cell with thin-film CuIn1−xSnxS2 photoelectrode. (b) Current-potential curves. (c) Mott-Schottky plot of CuInS2 capacitance against potential; inset: CBM and VBM positions. (d) Current-potential curves of CuIn1−xSnxS2 under intermittent irradiation.

Full size image

In summary, the intermediate band was demonstrated in Sn-doped CuMS2 (M = In, Ga). The uniform nanostructured CuMS2 particles and polycrystalline thin films were synthesized by newly-developed methods. Significant red shifts found in the absorption spectra of CuM1−xSnxS2 are caused by the triple absorption bands (VBM → CBM, VBM → IB, IB → CBM), further confirmed by photoluminescence and band structure calculations. Due to wide spectrum response, the enhanced photoelectric conversion revealed the better performance in photocatalysis and photocurrent generation. These results may introduce Sn-doped CuMS2 to be applied in future full-spectrum solar cells.

Methods

Samples preparation

A non-toxic solvent-based process was first developed to prepare nanostructured particles and films of CuMS2 (M = In, Ga). The raw powder of Cu2S, In2S3 and Ga2S3 were synthesized in the solid state reactions from the stoichiometric elements of Cu (99.999%, SinoReag), S (99.999%, SinoReag), In (99.999%, SinoReag) and Ga (99.999%, SinoReag) in a sealed evacuated fused silica tube. The nanostructured particles of CuMS2 were harvested from ball-milling the mixtured Cu2S and M2S3 powder with the atomic ratio of Cu:M:S = 1:1:2 for 24 h. The as-prepared powder were further dispersed in ethanol and further milled for 6 h to obtain the slurry. The precursor thin films of CuInS2 were spin-coated on both SiO2 and Mo/SiO2 substrates by using the slurry. All the as-coated films were annealed in H2S/Ar atmosphere (5 mol% H2S, 5000 Pa) at 600°C for 15 min. The Sn-doped samples were prepared only by substituting M with Sn from the same procedure described above.

Characterizations

CuM1−xSnxS2 powder samples were transferred to a TEM grid (Quantifoil Cu 200 mesh) for these investigations. TEM and STEM elemental mapping were conducted using a JEOL 2100F microscope, operating at 200 kV. Patterns of XRD of the samples were obtained (Bruker D8 FOCUS, at 40 KV and 40 mA) using Cu Kα radiation (λ = 0.15418 nm). The optical absorption spectra of samples were obtained at room temperature by the UV-Vis-NIR spectrometer (Hitachi U4100) equipped with an integrating sphere. The photoluminescence spectra of samples were obtained at room temperature, excited by the green laser light (λ = 514.5 nm) and detected by visible light and NIR (InSb) detectors. XPS experiments were carried out on a RBD upgraded PHI-5000C ESCA system (Perkin Elmer) with Mg Kα radiation (hν = 1253.6 eV). Binding energies were calibrated by using the containment carbon (C 1 s = 284.6 eV).

Calculation details

Band structure calculations were performed using the plane-wave projector augmented-wave (PAW) method applying the Heyd-Scuseria-Ernzerhof (HSE) hybrid function28,29. We constructed a supercell which contained 32 atoms and then substituted one M atom by Sn corresponding to the doping content of 12.5%. The cutoff energy of plane wave was chosen at 300 eV.

Photocatalytic and PEC test

The photocatalytic degradation of methylene blue was carried out in a reaction container equipped with an optical system including a 300 W xenon lamp and optical high-pass filters with the different cutoff wavelengths. The transparency spectra of the filters are shown in Fig. S3 of Supporting Information. The volume of the initial 10 mg/L MB solution is 200 ml and the catalyst powder are 100 mg. Light illumination is conducted after the suspension is stirred in the dark for 30 min to reach the adsorption–desorption equilibrium of organic dye on catalyst surface. During irradiation, about 3 ml suspension is continually taken from the reaction cell at given time intervals for subsequent MB concentration analysis after centrifuging. PEC measurements were performed in a typical three-electrode potentiostat system (Parstat 2773), in which the CuIn1−xSnxS2 film on Mo/SiO2 substrate, a Pt wire and an Ag/AgCl electrode were used as the working, counter and reference electrodes, respectively. 0.25 M Na2S aqueous solution was used as the supporting electrolyte to maintain the stability of films. A solar simulator (AM 1.5) with a power of 100 mW/cm2 was used as the illumination source. Photocurrent ON/OFF cycles were measured using the same electrochemical workstation coupled with a mechanical chopper.