A low-cost platinum-free counter electrode is undoubtedly important for developing economic dye-sensitized solar cells (DSSCs). In recent years, many kinds of materials such as carbonaceous materials [1-5], conductive polymers [6-9], and transition metal compounds [10-18], have been investigated to use as counter electrodes for DSSCs. Among the investigated alternatives, transition metal chalcogenides show excellent intrinsic electrocatalytic ability for electrode reaction on counter electrodes of DSSCs [19, 20]. Moreover, transition metal chalcogenides are able to form transparent counter electrodes with balance between electrical conductivity and optical transparency. Especially, molybdenum dichalcogenides i.e. MoS₂ [19] and MoSe₂ [20] have graphite-like layered structure that is helpful to improve active sites for reduction of triiodine ions on the counter electrodes.

Preparation of non-platinum counter electrodes usually employ two or more steps. Active materials are firstly prepared by chemical synthesis, and then the obtained materials are filmed on conductive substrates by spin-coating or doctor-blade coating, etc. to fabricate counter electrodes [21-24]. In these routes, besides the active materials, the film process usually has directly effect on the final electrode performance. Some in situ methods such as chemical bath deposition [25-27], chemical vapor deposition [28], and electrochemical deposition [29-32] are also developed to prepare transition metal chalcogenide counter electrodes. To compare with the conventional routes, these in situ methods are usually able to make better interface contact between active film and conductive substrate, reducing the interface resistance and improving electrochemical performance. Lee et al. [33] prepared the MoSe₂ electrode by the surface selenization of Mo-coated soda-lime glass with lower resistance. Chen et al. [34] obtained the MoSe₂/Mo electrode via in situ selenization of Mo films, in which the bottom Mo layer can act as conductive collector for electron extraction. Compared with these methods, magnetron sputtering method has advantages of simple equipment, easy control, large coating area and strong adhesion [4].

In this work, an in situ magnetron sputtering method is developed to prepare MoSe₂ counter electrodes. In the electrodes, MoSe₂ shows nanoplate-like morphology and forms a layer of active film with an average thickness of 1.3 μm. As counter electrode for a dye-sensitized solar cell, MoSe₂ electrodes prepared at 400 ᵒC has relatively low charge transfer resistance (R_ct), and the corresponding DSSCs provide an energy conversion efficiency of 6.83% which is comparable with that of the reference DSSC with platinum as counter electrode (6.51%).

The preparation process of MoSe₂ counter electrode is shown in Fig. 1. Transparent conductive glass (fluorine-doped tin oxide, FTO, 15Ω/square, Nippon Sheet Glass) was used as substrate. Before used, FTO glass was cleaned up with deionized water, ethanol, and deionized water in turn. 0.5 g of Se powder (99.99%, Aladdin) was mixed with 5% aqueous solution of sodium carboxymethyl cellulose (CMC; 0.05 mL) and ground to a sticky mixture. A layer of Se film was made by doctor-blade method on the cleaned FTO substrate with a thickness of 50 m (thickness of a layer of 3M tape).The as-prepared Se film was dried in the air, metal molybdenum was radio-frequency sputtered on the Se film when it drying, forming the Se-Mo film. In the process of sputtering, High purity argon (Ar) was used as the bombardment gas, and the sputter power is 80 W, radiofrequency sputtering 30 min in the vacuum pressure is 5 × 10^‒3 torr. Finally, the sample was calcined in a tubular furnace for 1 h at 300, 400, 450, 500, 550 or 600 ℃ separately with a heating rate of 5 ℃ min^‒1 under flowing Ar atmosphere (80 ml min^‒1). For the sake of comparison, pure Mo electrode was prepared in the same way. The preparation process is as shown in Fig. 1, and there is no process "a" when prepared the pure Mo electrode.

As a reference sample, the sputtering Pt counter electrode was obtained by purchase from Wuhan Lattice.

The cleaned FTO conductive glass was immersed into 40 mmol L^-1 TiCl₄ aqueous solution at 70 ℃ for 30 min, and then washed with water and ethanol. A commercial TiO₂ sol (20 nm, Wuhan Lattice) was coated on the FTO glass to a certain thickness by doctor-blade coating. The film was then heated in a muffle furnace at 130 ℃ for 8 min. After cooling to room temperature, doctor-blade coating the TiO₂ sol (200 nm, Wuhan Lattice) on the first layer, heated in a muffle furnace at 500 ℃ for 1 h. The obtained film was treated with TiCl₄ solution again, and then calcined at 500 ℃ for 0.5 h. Finally, the film was soaked in 0.3 mmol L^‒1 ethanol solution of N719 (Solaronix SA) for 24 h.

The DSSCs were assembled by clipping the TiO₂ photoanode and counter electrode together, and injecting the electrolyte into the aperture between the two electrodes. Surlyn 1702 was used as the spacer between the two electrodes. The liquid electrolyte contained 0.6 mol/L 1, 2-dimethyl-3- propylimidazoliumiodide (DMPII), 0.5 M 4-tert-butyl pyridine (4tBP), 0.05 mol/L I₂ and 0.1 mol/L LiI, with acetonitrile as the solvent. Solid paraffin was used as a sealant to prevent the electrolyte solution from leaking. The effective cell area was 0.20 cm².

Photocurrent density-voltage (J-V) characteristic curves, electrochemical impedance spectra (EIS) and cyclic voltammetry (CV) measured by Zahner IM6ex electrochemical workstation. The J-V characteristic of the DSSCs was illuminated under AM1.5 solar irradiations. The EIS measurements were carried out in a frequency range of 100 KHz to 100 mHz, with an AC modulation signal of 10 mV, and a bias DC voltage is -0.6 V. CV test was finished in the three-electrode system, metal Pt used as the counter electrode, an Ag/Ag⁺ electrode as the reference electrode, and counter electrode of DSSCs as working electrode. The electrolyte was 5 mmol/L LiI, 0.5 mmol/L I₂, 0.5 mmol/L LiClO₄ dissolved in acetonitrile solution. And the scan rate was 50 mV s^‒1 when tested.

The structure and morphology of the as-prepared materials were evaluated by X-ray diffraction (XRD, Rigaku MiniFlex Ⅱ), scanning electron microscopy (SEM, SUPRA 55VP), and transmission electron microscopy (TEM, JEM-2010, JEOL).

Fig. 2 shows the XRD patterns of FTO, Se-Mo-300, Se-Mo-400, Se-Mo-450, Se-Mo-500, Se-Mo-550, and the numbers represent the heat treatment temperature of the samples. All the patterns of the prepared samples contain the peaks of fluorine and tin oxide on account of FTO conductive substrates. In the case of Se-Mo-300, the diffraction peaks can be indexed, respectively, to crystalline selenium (JCPDS No. 1-848) and metallic molybdenum (JCPDS No. 44-1120). Clearly, this means that no reactions occur between the coating selenium layer and the sputtered molybdenum layer at 300 ℃. With increasing calcination temperature, Se-Mo-400 shows two diffraction peaks at 2θ = 13.5ᵒ and 31.6ᵒ, which can be indexed to (002) and (100) lattice planes of MoSe₂ (JCPDS No. 87-2419), respectively. The results reveals that a solid-solid chemical reaction occurs between selenium and metallic molybdenum, and MoSe₂ layer is in situ generated on the FTO substrate. Meanwhile, the very weak diffraction peak of (101) plane of selenium at 2θ = 29.5ᵒ imply existence of a small amount of selenium in the Se-Mo-400. When the calcination temperature increases to 450 ℃, all the diffraction peaks can be indexed to MoSe₂, and no selenium or molybdenum is observed. With the calcination temperature rises to 550 ℃, the diffraction peaks of MoSe₂ become gradually strong, suggesting improved crystallinity. As a layered chalcogenide like MoS₂, MoSe₂ similarly has (002) dominated plane as basal plane, and (100) plane in the vertical direction as edge plane. In details, the peak intensity ratios of (100) plane to (002) plane are different due to the different calcination temperatures. The ratio of (100)/(002) is 0.73, 1.1, 0.54 and 0.69 at 400, 450, 500, and 550 ℃, respectively. According to the previous results, the greater ratio means more electrocatalytically active sites for reduction of triiodide ions [19].

SEM and TEM of the Se-Mo-450 as the representative sample are performed to analyze morphology and structure details on the FTO substrate. Fig. 3(a) shows that Se-Mo-450 presents irregular particles with a size of 0.5‒2 μm. The irregular particles are formed by aggregation of MoSe₂ nanoplates, which is a classic morphology for layered metal dichalcogenides. Thickness of the MoSe₂ nanoplates is about 20 nm, and size of the plates is around 100‒200 nm (Fig. 3(b)). Fig. 3(c) shows the cross section of the Se-Mo-450 electrode. It is clear that the MoSe₂ layer is around 1.3 μm of thickness, and the nanoplates present relatively incompact but interwoven. This morphology is helpful for improving electrocatalytically active surface and infiltration of electrolyte. In addition, there is good contact between the MoSe₂ layer and FTO layer, indicating a low interface resistance. The TEM image shows the edge structure of a plate of MoSe₂ from Se-Mo-450 (Fig. 3(d)). The interplanar spacing is 0.63 nm in the edge, corresponding to the (002) plane of MoSe₂, and the other lattice fringe spacing of 0.28 nm corresponds to the (100) plane of MoSe₂.

Photocurrent density-voltage curves of DSSCs using Pt, Se-Mo-300, Se-Mo-400, Se-Mo-450, Se-Mo-500, and Se-Mo-550 as counter electrode are shown in Fig. 4, and the detailed photovoltaic performance parameters are listed in Table 1. In the case of Se-Mo-300, though the DSSC shows a relatively high V_oc, 0.791 V, the photoelectric conversion efficiency (PCE) is only 2.36% due to the poor fill factor (FF), 0.29. This means that the metallic molybdenum and selenium in Se-Mo-300, in which no MoSe₂ is generated as shown in the XRD result, have poor electrocatalytic activity for reduction of triiodide ions [21]. The low electrocatalytic activity causes low charge transfer process on the counter electrode. Therefore, the electrons from the external circuit cannot be transferred to triiodine ions promptly, and stay on the counter electrode to result in increase of potential, which causes the high open voltage. Meanwhile, the low electrocatalytic activity means large charge transfer resistance that leads to the poor fill factor. The DSSCs using the MoSe₂ counter electrodes obtained at 400, 450 and 500 ℃ show the photoelectric conversion efficiency (PCE) of 6.83%, 6.68% and 6.55%, respectively. The results are comparable with that of the DSSC with Pt counter electrode (6.51%). The device with Se-Mo-550 electrode yields a reduced efficiency of 5.98%, and the one using Se-Mo-300 shows only the value of 2.36%. For Se-Mo-550, the corresponding low J_sc (11.78 mA cm^‒1) is the main factor for the low efficiency. As known well, FTO substrate can be partially influenced when calcined at 550 ℃.

EIS of the symmetric cells with two identical counter electrodes of Pt, Se-Mo-300, Se-Mo-400, Se-Mo-450, Se-Mo-500 and Se-Mo-550, respectively, is carried out to investigate kinetics of the electrode processes. The measurement and fitting results are shown in Fig. 5, and the detailed fitting data are listed in Table 2. All the Nyquist plots of the different electrodes obtained at the various calcination temperatures are composed of two semicircles. The semicircle in high frequency region represents charge transfer impedance (R_ct), reflecting electrocatalytic activity of the electrodes for reduction of triiodide ions, and the other one in low frequency region represents Nernst diffusion impedance (Z_w), reflecting diffusion rate of triiodide ions on the electrode surface [35-37]. Among the measurement results, both R_ct and Z_w of Se-Mo-400 are the smallest values, indicating relatively high electrocatalytic activity and fast ion diffusion rate. These reveal why Se-Mo-400 has the maximum FF and the highest power conversion efficiency in the J-V curve measurement. R_ct of Se-Mo-450 is 2.42 Ω, lower than the value of the Pt electrode (4.06 Ω), suggesting that its electrocatalytic activity is comparable with that of Pt electrode. Meanwhile, Z_w of Se-Mo-450 is also 3.84 Ω, close to Z_w of the Pt electrode (4.25 Ω), indicating similar ion diffusion rate for the two electrodes. This leads to the very close power conversion efficiency for the DSSCs with Se-Mo-450 and the Pt counter electrodes. To compare Se-Mo-400 with Se-Mo-450, the lower R_ct of the former should be attributed to a higher active surface area, considering the fact that the sample is heated at the relatively low temperature in spite of existence of the small amount of Se impurity. R_ct of Se-Mo-500 is 4.73 Ω, slightly higher than the value of the Pt electrode, namely, the electrocatalytic activity is slightly lower than that of the Pt electrode. However, Z_w of Se-Mo-500 is lower than that of the Pt electrode, suggesting a relatively fast ion diffusion rate. As mentioned above, the incompact but interwoven structure for MoSe₂ electrodes is helpful for mass transfer process. As the combination result of the two parameters, the power conversion efficiency of the DSSCs with Se-Mo-500 and Pt counter electrodes is very close. In the case of Se-Mo-550, R_ct is 1.73 Ω, and Z_w is 5.28. This means that Se-Mo-550 has high electrocatalytic activity for reduction of triiodide ions, but relatively slow ion diffusion rate due to the high calcination temperature. For the Se-Mo-300 electrode with different composition, R_ct is 403 Ω, much higher than those of the other electrodes, suggesting that its electrocatalytic activity is very poor. And its Z_w is also very high, resulting in the low power conversion efficiency. For series resistance (R_s), the relative gap among the MoSe₂ electrodes is small, so no further discussion here.

Electrocatalytic activity of the electrodes to iodide/triiodide ions are characterize by CV measurements. As shown in Fig. 6, in the four peaks for each, the oxidation peaks (O_x^‒1, O_x^‒2) is located in the upper portion of the curve, and the reduction peaks (Red^-1, Red^-2) is at the bottom. The corresponding redox reactions are as follow:

(1)

(2)

In the CV curves, peak current density and peak to peak separation (between the Red^‒1 and O_x^‒1, E_pp) are related to electrochemical rate constant of redox reaction, namely, the larger the current density and the thinner the E_pp, the greater the rate constant, and the stronger the electrocatalytic activity. In Fig. 6, peak current density decreases in the order of Se-Mo-450 > Se-Mo-400 > Se-Mo-550 > Se-Mo-500 > Pt, and E_pp of Se-Mo-400, Se-Mo-450 and Pt are lower than those of Se-Mo-500 and Se-Mo-550. This suggests that the reaction rate constant of Se-Mo-400 and Se-Mo-450 for reduction of triiodide ions is larger than those of Se-Mo-500 and Se-Mo-550 electrodes. The results are consistent with the above analysis from the EIS measurements.

Molybdenum selenide electrodes are prepared by in situ magnetron sputtering method. The obtained MoSe₂ presents nanoplate-like morphology with a thickness of around20 nm and a diameter of about 100 nm, and the nanoplates interweave into a layer of MoSe₂ film with about an average thickness of 1.3 μm. The MoSe₂ electrodes obtained at various temperatures from 300 and 550 ᵒC are used as counter electrode for a dye-sensitized solar cell. The MoSe₂ electrodes prepared at 400 ᵒC has the optimized performance, and the corresponding DSSCs provide an energy conversion efficiency of 6.83% which is comparable than that of the reference DSSC with Pt as counter electrode (6.51%). The electrocatalytic activity of the MoSe₂ electrodes for reduction of triiodide ions increases with rising of the calcination temperature from 400 to 500 ᵒC. However, the MoSe₂ electrode prepared at 550 ᵒC has obviously reduced electrocatalytic activity. Briefly, this work provides a simple and effective method to prepare low-cost MoSe₂ counter electrodes for dye-sensitized solar cells.