Investigation of Optimum Mg Doping Content and Annealing Parameters of Cu2MgxZn1−xSnS4 Thin Films for Solar Cells

Cu2MgxZn1−xSnS4 (0 ≤ x ≤0.6) thin films were prepared by a simple, low-temperature (300 °C) and low-cost sol–gel spin coating method followed by post-annealing at optimum conditions. We optimized the annealing conditions and investigated the effect of Mg content on the crystalline quality, electrical and optical performances of the Cu2MgxZn1−xSnS4 thin films. It was found that the Cu2MgxZn1−xSnS4 film annealed at 580 °C for 60 min contained large grain, less grain boundaries and high carrier concentration. Pure phase kesterite Cu2MgxZn1−xSnS4 (0 ≤ x ≤ 0.6) thin films were obtained by using optimal annealing conditions; notably, the smaller Zn2+ ions in the Cu2ZnSnS4 lattice were replaced by larger Mg2+ ions. With an increase in x from 0 to 0.6, the band gap energy of the films decreased from 1.43 to 1.29 eV. When the ratio of Mg/Mg + Zn is 0.2 (x = 0.2), the grain size of Cu2MgxZn1−xSnS4 reaches a maximum value of 1.5 μm and the surface morphology is smooth and dense. Simultaneously, the electrical performance of Cu2MgxZn1−xSnS4 thin film is optimized at x = 0.2, the carrier concentration reaches a maximum value of 3.29 × 1018 cm−3.


Introduction
In recent years, the semiconductor Cu 2 ZnSnS 4 (CZTS) has attracted enormous attention as an ideal absorber material for low-cost thin film solar cells. For thin film CuInGaSe 2 (CIGS) and CdTe solar cells, reliable efficiencies of more than 20% have been achieved [1,2]. However, the limited resources and extremely high costs of In and Ga, and toxicity of Se and Cd significantly limit further development of CIGS and CdTe solar cells. CZTS is regarded as a substitute for CIGS, wherein the high-cost and rare In and Ga, and toxic Se are replaced by low-cost and earth-abundant Zn, Sn and S, respectively. In addition to being composed of abundant and non-toxic elements, CZTS exhibits remarkable photoelectric properties as an absorbing layer, including a high absorption coefficient (>10 4 cm −1 ) and a suitable band gap (1.40-1.50 eV) [3]. To date the best efficiency of pure CZTS has broken through 11% [4,5], but it is still far below than that of CIGS (21.7%) [6]. In order to realize the industrialization of low-cost and environmental protection CZTS solar cells, it is necessary to further improve the efficiency of CZTS based thin film solar cells. The low efficiencies of CZTS solar cells are attributed to factors such as low crystallinity, large open circuit voltage (V oc ) deficit as well as poor band alignment at the CdS/CZTS interface [7][8][9]. As we all know, V oc is linearly related to the band gap of CZTS. Band gap engineering has emerged as an effective method to adjust the band alignment at

Results and Discussion
It has been widely reported that annealing conditions significantly affect the properties of CZTS films. To optimize the annealing conditions for Cu 2 Mg x Zn 1−x SnS 4 films, they were annealed under different conditions. Figure 1a-f show the SEM images of the Cu 2 Mg x Zn 1−x SnS 4 (x = 0.2) films annealed at different conditions. Samples A1, A2 and A3 were annealed for 60 min under a sulfur atmosphere at 540, 580 and 600 • C, respectively. Figure 1a-c shows the SEM images of samples A1, A2 and A3, respectively. As shown in the surface SEM image in Figure 1a, sample A1 contained small nanoparticles (30-100 nm); in addition, a small hole was observed on the surface of the film. Figure 1b shows the SEM image of sample A2; as observed, with an increase in the annealing temperature, the crystalline quality of the Cu 2 Mg x Zn 1−x SnS 4 film significantly improved; grain size increased up to 1.4 µm; and the surface became smooth, dense and crack free. However, with a further increase in the annealing temperature to 600 • C, the grain size decreased to 500-900 nm, and more voids and nanoparticles were observed on the surface, as shown in Figure 1c. As shown in Figure 1a-c, sample A2 exhibited optimal crystalline quality, indicating that the optimum annealing temperature was 580 • C. Samples B1, B2 and B3 were annealed at 580 • C under a sulfur atmosphere for 30, 45 and 75 min respectively. Figure 1d-f show the SEM images of samples B1, B2 and B3. Compared to that of sample A2 annealed at 580 • C for 60 min, the crystalline quality of samples B1, B2 and B3 was inferior. Moreover, the surfaces of B1, B2 and B3 were uneven and porous, as shown in Figure 1d-f. As shown in Figure 1b,d,e, with an increase in the annealing time from 30 min and 45 min, the grain size increased from 70-200 nm to 100-500 nm, and then, with a further increase in the annealing time to 60 min, the grain size increased to 100-1400 nm. However, as the film was annealed for a longer time (75 min), the grain size of the Cu 2 Mg x Zn 1−x SnS 4 film decreased to 200-700 nm. The surface morphological examination indicated that the Cu 2 Mg x Zn 1−x SnS 4 grain growth gradually occurred, and an optimal crystallization quality was achieved by annealing at 580 • C for 60 min.  Table 1 lists the electrical transport parameters of the Cu2MgxZn1-xSnS4 (x = 0.2) film annealed at different annealing conditions. As observed, the film invariably exhibited p-type conductivity. With an increase in the annealing temperature from 540 °C to 600 °C, the carrier concentration first sharply increased from 4.12 × 10 15 cm −3 (sample A1) to 3.29 × 10 18 cm −3 (sample A2), and then decreased to 3.79 × 10 17 cm −3 (sample A3); notably, the resistivity decreased from 9.43 × 10 0 Ωcm to 1.16 × 10 −1 Ωcm, and then increased to 1.53 × 10 0 Ωcm. The mobility decreased from 3.70 × 10 0 cm 2 V −1 S −1 to 1.01 × 10 −1 cm 2 V −1 S −1 and then increased to 7.87 × 10 −1 cm 2 V −1 S −1 . Similarly, with an increase in the annealing time from 30 min to 75 min, the carrier concentration first increased from 3.21 × 10 14 cm −3 (sample B1) to 3.79 × 10 15 cm −3 (sample B2), reached the maximum value of 3.29 × 10 18 cm −3 (sample A2) and finally reduced to 4.62 × 10 16 cm −3 (sample B3). Simultaneously, the mobility decreased from 6.02 × 10 0 cm 2 V −1 S −1 for sample B1 to 2.02 × 10 0 cm 2 V −1 S −1 for sample B2, and then reached the minimum 1.01 × 10 −1 cm 2 V −1 S −1 for A2 and finally slightly elevated to 9.32 × 10 −1 cm 2 V −1 S −1 for sample B3. According to the SEM and Hall results, the carrier concentration gradually increases with the annealing time changes from 30 to 60 min, but it starts to decrease when the annealing time increases from 60 to 75 min. It can be explained that when the annealing time increases from 30 to 60 min, the crystallinity of Cu2MgxZn1-xSnS4 films is improved, the defects at the grain boundaries are passivated, resulting in the increase of carrier concentration. When the annealing time increases from 60 to 75 min, the crystallization quality is slightly deteriorated, as shown in the previous SEM results, therefore, the carrier concentration decrease. The change of mobility is opposite to that of the carrier concentration. When the annealing time changes from 30 to 60 min, the mobility decreases with the increasing of the carrier concentration, and when the annealing time increases from 60 to 75 min, the mobility starts to increase with the decreasing of the carrier concentration. Finally, it was found when the film was annealed at 580 °C for 60 min, the Cu2MgxZn1-xSnS4 film has the optimum crystallization quality and the best electrical performance with the carrier concentration of 3.29 × 10 18 cm −3 and the mobility of 1.01 × 10 −1 cm 2 V −1 s −1 . It is concluded that the change of the electrical properties may have great relevance to the defects passivated in the grain boundaries by improving the crystallinity properties.  Table 1 lists the electrical transport parameters of the Cu 2 Mg x Zn 1−x SnS 4 (x = 0.2) film annealed at different annealing conditions. As observed, the film invariably exhibited p-type conductivity. With an increase in the annealing temperature from 540 • C to 600 • C, the carrier concentration first sharply increased from 4.12 × 10 15 cm −3 (sample A1) to 3.29 × 10 18 cm −3 (sample A2), and then decreased to 3.79 × 10 17 cm −3 (sample A3); notably, the resistivity decreased from 9.43 × 10 0 Ωcm to 1.16 × 10 −1 Ωcm, and then increased to 1.53 × 10 0 Ωcm. The mobility decreased from 3.70 × 10 0 cm 2 V −1 S −1 to 1.01 × 10 −1 cm 2 V −1 S −1 and then increased to 7.87 × 10 −1 cm 2 V −1 S −1 . Similarly, with an increase in the annealing time from 30 min to 75 min, the carrier concentration first increased from 3.21 × 10 14 cm −3 (sample B1) to 3.79 × 10 15 cm −3 (sample B2), reached the maximum value of 3.29 × 10 18 cm −3 (sample A2) and finally reduced to 4.62 × 10 16 cm −3 (sample B3). Simultaneously, the mobility decreased from 6.02 × 10 0 cm 2 V −1 S −1 for sample B1 to 2.02 × 10 0 cm 2 V −1 S −1 for sample B2, and then reached the minimum 1.01 × 10 −1 cm 2 V −1 S −1 for A2 and finally slightly elevated to 9.32 × 10 −1 cm 2 V −1 S −1 for sample B3. According to the SEM and Hall results, the carrier concentration gradually increases with the annealing time changes from 30 to 60 min, but it starts to decrease when the annealing time increases from 60 to 75 min. It can be explained that when the annealing time increases from 30 to 60 min, the crystallinity of Cu 2 Mg x Zn 1−x SnS 4 films is improved, the defects at the grain boundaries are passivated, resulting in the increase of carrier concentration. When the annealing time increases from 60 to 75 min, the crystallization quality is slightly deteriorated, as shown in the previous SEM results, therefore, the carrier concentration decrease. The change of mobility is opposite to that of the carrier concentration. When the annealing time changes from 30 to 60 min, the mobility decreases with the increasing of the carrier concentration, and when the annealing time increases from 60 to 75 min, the mobility starts to increase with the decreasing of the carrier concentration. Finally, it was found when the film was annealed at 580 • C for 60 min, the Cu 2 Mg x Zn 1-x SnS 4 film has the optimum crystallization quality and the best electrical performance with the carrier concentration of 3.29 × 10 18 cm −3 and the mobility of 1.01 × 10 −1 cm 2 V −1 s −1 . It is concluded that the change of the electrical properties may have great relevance to the defects passivated in the grain boundaries by improving the crystallinity properties. To evaluate the crystalline quality and investigate the existence of impurity phases, the films were subjected to XRD analysis. Figure 2 illustrates the XRD patterns of the Cu 2 Mg x Zn 1−x SnS 4 (0 ≤ x ≤ 0.6) thin films. As shown in Figure 2a, strong diffraction peaks at 2θ = 28.53, 32.99, 47.33 and 56.17 • were observed for all films, which were assigned to the (112), (200), (220) and (312) diffraction planes of kesterite CZTS (JCPDS card no. 26-0575) [26,27]. In addition, two weak peaks were observed at 2θ = 69.27 • and 76.44 • , which were ascribed to the (008) and (332) planes of kesterite CZTS [28], suggesting that the crystalline quality of the Cu 2 Mg x Zn 1−x SnS 4 films was satisfactory. Apart from the diffraction peaks of CZTS, no secondary phase peaks were detected, indicating that Mg doping did not affect the crystal structure of the CZTS film. As observed, with an increase in x from 0 to 0.1, the intensity of the (112) peak slightly increased, and then with a further increase in x to 0.2, the peak intensity reached the maximum, implying that the crystalline quality of the Cu 2 Mg x Zn 1−x SnS 4 thin film with x = 0.2 is the best. However, with increasing x from 0.2 to 0.6, the intensity of the (112) peak gradually decreased and became the lowest at x = 0.6; this gradual deterioration in the crystallinity of the Cu 2 Mg x Zn 1−x SnS 4 thin films with increasing x was attributed to excessive Mg doping. Figure 2b shows the enlarged view of the (112) peaks. As observed, with an increasing Mg content, the (112) peak unidirectionally shifted to smaller 2θ values, suggesting an increase in the lattice constant of Cu 2 Mg x Zn 1−x SnS 4 . It is well known that the change of the ion radius in CZTS usually results in the change of lattice parameters [29][30][31]. The occupation of Zn 2+ sites in the CZTS host lattice by Mg 2+ ions results in an increase in the Cu 2 Mg x Zn 1−x SnS 4 lattice parameters, because the covalent radius of Mg 2+ (1.36 Å) is larger than that of Zn 2+ (1.25 Å). Thus, the XRD results indicated that with Mg doping, the phase structure of CZTS did not change, and the Zn 2+ sites in the CZTS host lattice were occupied by Mg 2+ .
The formation of pure kesterite Cu 2 Mg x Zn 1−x SnS 4 cannot be properly confirmed by XRD, because the lattice parameters of CZTS and some possible impurity phases such as tetragonal Cu 2 SnS 3 , cubic ZnS and Cu x S are similar [32,33]. Therefore, to confirm the formation of pure kesterite Cu 2 Mg x Zn 1−x SnS 4 , the samples were subjected to Raman spectroscopy analysis. Figure 3 shows the Raman spectra of Cu 2 Mg x Zn 1−x SnS 4 (0 ≤ x ≤ 0.6) films. As shown, the spectra contained the dominant characteristic peak at 333 cm −1 and two relatively weak peaks at 288 cm −1 and 375 cm −1 . These Raman peaks were attributed to the A1, A2 and E vibration modes of the S atom in kesterite CZTS, respectively; these results agreed well with those previously reported [34,35]. Notably, no other ternary or binary phase (Cu 2 SnS 3 , SnS 2 , SnS, ZnS) peaks were observed in the Raman spectra. In addition, as shown in Figure 3, with an increase in x from 0 to 0.6, the Raman peak, particularly for the peak of A 1 vibration mode, was slightly red shift systematically. Figure 3 displays the A1 mode peak position variation as a function of the Mg content; as observed, the peak shifted from 336.79 cm −1 to 332.13 cm −1 with an increase in the Mg content. Combining with the XRD results, the change in the A1 peak position could be ascribed to lattice expansion due to the substitution of the smaller Zn ions by the larger Mg ions in Cu 2 Mg x Zn 1−x SnS 4 . The redshift in the lattice vibrations were attributed to the lower bonding force of Mg-S than that of Zn-S, resulting from the larger covalent radius of Mg than that of Zn. A similar Raman peak shift caused by ion replacement has been reported in previous studies [36]. Combined with XRD results to analyze the result of Raman spectra, it was found that no other impurity compounds were detected in Cu 2 Mg x Zn 1−x SnS 4 films when the x was in the range of 0 to 0.6. The pure kesterite Cu 2 Mg x Zn 1−x SnS 4 thin films were successfully prepared. attributed to excessive Mg doping. Figure 2b shows the enlarged view of the (112) peaks. As observed, with an increasing Mg content, the (112) peak unidirectionally shifted to smaller 2θ values, suggesting an increase in the lattice constant of Cu2MgxZn1-xSnS4. It is well known that the change of the ion radius in CZTS usually results in the change of lattice parameters [29][30][31]. The occupation of Zn 2+ sites in the CZTS host lattice by Mg 2+ ions results in an increase in the Cu2MgxZn1-xSnS4 lattice parameters, because the covalent radius of Mg 2+ (1.36 Å) is larger than that of Zn 2+ (1.25 Å). Thus, the XRD results indicated that with Mg doping, the phase structure of CZTS did not change, and the Zn 2+ sites in the CZTS host lattice were occupied by Mg 2+ .
The formation of pure kesterite Cu2MgxZn1-xSnS4 cannot be properly confirmed by XRD, because the lattice parameters of CZTS and some possible impurity phases such as tetragonal Cu2SnS3, cubic ZnS and CuxS are similar [32,33]. Therefore, to confirm the formation of pure kesterite Cu2MgxZn1-xSnS4, the samples were subjected to Raman spectroscopy analysis.  Figure 3 shows the Raman spectra of Cu2MgxZn1-xSnS4 (0 ≤ x ≤ 0.6) films. As shown, the spectra contained the dominant characteristic peak at 333 cm −1 and two relatively weak peaks at 288 cm −1 and 375 cm −1 . These Raman peaks were attributed to the A1, A2 and E vibration modes of the S atom in kesterite CZTS, respectively; these results agreed well with those previously reported [34,35]. Notably, no other ternary or binary phase (Cu2SnS3, SnS2, SnS, ZnS) peaks were observed in the Raman spectra. In addition, as shown in Figure 3, with an increase in x from 0 to 0.6, the Raman peak, particularly for the peak of A1 vibration mode, was slightly red shift systematically. Figure 3 displays the A1 mode peak position variation as a function of the Mg content; as observed, the peak shifted from 336.79 cm −1 to 332.13 cm −1 with an increase in the Mg content. Combining with the XRD results, the change in the A1 peak position could be ascribed to lattice expansion due to the substitution of the smaller Zn ions by the larger Mg ions in Cu2MgxZn1-xSnS4. The redshift in the lattice vibrations were attributed to the lower bonding force of Mg-S than that of Zn-S, resulting from the larger covalent radius of Mg than that of Zn. A similar Raman peak shift caused by ion replacement has been reported in previous studies [36]. Combined with XRD results to analyze the result of Raman spectra, it was found that no other impurity compounds were detected in Cu2MgxZn1-xSnS4 films when the x was in the range of 0 to 0.6. The pure kesterite Cu2MgxZn1-xSnS4 thin films were successfully prepared. Notably, the chemical composition of the Cu2MgxZn1-xSnS4 films and the chemical bonding states of the constituents significantly affect the solar cell performance. Hence, the Cu2MgxZn1-xSnS4 films were characterized by XPS. Figure 4a-d show the XPS profiles of the constituent metals (Cu, Zn, Sn and Mg) of the representative Cu2MgxZn1-xSnS4 (x = 0.2) sample. Figure 4a displays the Cu 2p XPS profile. The two peaks at 952.4 eV and 931.7 eV were attributed to Cu 2p1/2 and Cu 2p3/2. In addition, the peak separation value agreed well with the standard value of 20.7 eV, indicating that Cu was present in the +1 combined-state [37]. Figure 4b illustrates the XPS spectrum of Zn 2p. The two peaks located at 1044.6 eV and 1022.1 eV were attributed to Zn 2p1/2 and Zn 2p3/2, respectively, the splitting energy was 22.5 eV. The splitting value is consistent with the standard value of 22.97 eV, which confirms that Zn exists in a +1 state [38]. The Sn 3d XPS profile is displayed in Figure 4c. As observed, two peaks of Sn 3d3/2 and Sn 3d5/2, situated at 494.3 and 485.9 eV were detected; the peak separation value was 8.4 eV, which agreed with the standard value, implying  Figure 4a displays the Cu 2p XPS profile. The two peaks at 952.4 eV and 931.7 eV were attributed to Cu 2p1/2 and Cu 2p3/2. In addition, the peak separation value agreed well with the standard value of 20.7 eV, indicating that Cu was present in the +1 combined-state [37]. Figure 4b illustrates the XPS spectrum of Zn 2p. The two peaks located at 1044.6 eV and 1022.1 eV were attributed to Zn 2p1/2 and Zn 2p3/2, respectively, the splitting energy was 22.5 eV. The splitting value is consistent with the standard value of 22.97 eV, which confirms that Zn exists in a +1 state [38]. The Sn 3d XPS profile is displayed in Figure 4c. As observed, two peaks of Sn 3d 3/2 and Sn 3d 5/2 , situated at 494.3 and 485.9 eV were detected; the peak separation value was 8.4 eV, which agreed with the standard value, implying that Sn was in the Sn 4+ oxidation state [39]. Figure 4d presents the Mg 1s XPS profile; the peak at 1303.7 eV was assigned to the Mg 1s core level, indicating the presence of divalent Mg 2+ [24]. According to the results of XPS, the valence states of Cu, Zn, Mg and Sn were +1, +2, +4 and +2 respectively. This further confirmed the substitution of Zn in CZTS by Mg, agreeing well with the XRD and Raman results. that Sn was in the Sn 4+ oxidation state [39]. Figure 4d presents the Mg 1s XPS profile; the peak at 1303.7 eV was assigned to the Mg 1s core level, indicating the presence of divalent Mg 2+ [24]. According to the results of XPS, the valence states of Cu, Zn, Mg and Sn were +1, +2, +4 and +2 respectively. This further confirmed the substitution of Zn in CZTS by Mg, agreeing well with the XRD and Raman results. The atomic contents of Cu, Zn, Sn, S and Mg in the Cu2MgxZn1-xSnS4 (0 ≤ x ≤ 0.6) films are listed in Table 2. When the percentages of Mg/(Mg + Zn) for the precursor solution of Cu2MgxZn1-xSnS4 (0 ≤ x ≤ 0.6) were 0, 10, 20, 40 and 60, the percentages of Mg/(Mg + Zn) in Cu2MgxZn1-xSnS4 films were 0, 7.79, 14.22, 34.72 and 54.61, respectively. Notably, the elemental loss during annealing and the preparation process cannot be neglected; nonetheless, the Mg/(Mg + Zn) ratio in the Cu2MgxZn1-xSnS4 films increased with an increase in the Mg content in the precursor solution. Figure 5 summarizes the atomic contents of the constituent elements of the Cu2MgxZn1-xSnS4 (0 ≤ x ≤ 0.6) films, according to the energy dispersive X-ray spectroscopy (EDS) results presented in Table 2. As observed, the atomic content of Mg gradually increased with a decrease in the atomic content of Zn from 17.95 to 7.39; moreover, the Mg/(Mg + Zn) ratio also increased. This indicated that Mg was incorporated into the CZTS lattice, replacing Zn. Furthermore, the changes in the atomic contents of other elements in the Cu2MgxZn1-xSnS4 films were negligible. The result is in good agreement with the conclusion that Mg will substitute the site of Zn obtained from the analysis result of XRD and Raman.  Figure 5 summarizes the atomic contents of the constituent elements of the Cu 2 Mg x Zn 1−x SnS 4 (0 ≤ x ≤ 0.6) films, according to the energy dispersive X-ray spectroscopy (EDS) results presented in Table 2. As observed, the atomic content of Mg gradually increased with a decrease in the atomic content of Zn from 17.95 to 7.39; moreover, the Mg/(Mg + Zn) ratio also increased. This indicated that Mg was incorporated into the CZTS lattice, replacing Zn. Furthermore, the changes in the atomic contents of other elements in the Cu 2 Mg x Zn 1−x SnS 4 films were negligible. The result is in good agreement with the conclusion that Mg will substitute the site of Zn obtained from the analysis result of XRD and Raman.   To determine the effect of the Mg content on the crystalline quality of the Cu2MgxZn1-xSnS4 (0 ≤ x ≤ 0.6) films, the films were detected by SEM as shown in Figure 6a-e. Figure 6a displays the surface SEM images of the Cu2MgxZn1-xSnS4 film with x = 0. As observed, the film consisted of irregular nanoscale grains (40-500 nm). Moreover, the surface of the film was relatively rough, but compact. Obviously, the irregular grain boundaries and small particles are not conducive to the improvement of the efficiency for the CZTS solar cells. As shown in Figure 6b, with an increase in x to 0.1, the film crystallinity enhanced and the grain size increased to 400-1200 nm. Furthermore, the surface morphology was significantly improved and become smooth and compact. With a further increase in the value of x to 0.2, the film surface became very flat and dense, as displayed in Figure  6c; in addition, the grain size further increased to 0.7-1.5 μm, which was conducive to achieving high efficiencies for CZTS solar cells. Figure 6d shows the SEM image of the Cu2MgxZn1-xSnS4 film with x = 0.4. As observed, the grain size sharply decreased to 300-900 nm, but the grains were larger than those of Cu2MgxZn1-xSnS4 with x = 0, and densely stacked. The crystalline quality of the Cu2MgxZn1-xSnS4 film continued to deteriorate with further increase in x to 0.6. As shown in Figure  6e, the grain size of Cu2MgxZn1-xSnS4 film reduced to 200-400 nm, occasionally, a few larger grains were observed on the film surface. As seen, the surface morphology of the film with x = 0.6 was uneven and irregular. The bar chart in Figure 6 shows the average particle diameter as a function of the Mg content. As seen, the average size gradually increased with an increase in the value of x from 0 to 0.2 and reached the maximum at x = 0.2, then with further increase in x from 0.2 to 0.6, the size sharply decreased. It is well known that the good grain growth and smooth surface is of great significance to the fabrication of high power conversion efficiency (PCE) CZTS solar cells. Because the absorption layer with larger particle size can reduce the grain boundaries area, which is conducive to decrease the recombination of photon-generated carrier and increase the efficiency of CZTS solar cells. Based on the results of SEM, when the value of x was 0.2, it was concluded that the To determine the effect of the Mg content on the crystalline quality of the Cu 2 Mg x Zn 1−x SnS 4 (0 ≤ x ≤ 0.6) films, the films were detected by SEM as shown in Figure 6a-e. Figure 6a displays the surface SEM images of the Cu 2 Mg x Zn 1−x SnS 4 film with x = 0. As observed, the film consisted of irregular nanoscale grains (40-500 nm). Moreover, the surface of the film was relatively rough, but compact. Obviously, the irregular grain boundaries and small particles are not conducive to the improvement of the efficiency for the CZTS solar cells. As shown in Figure 6b, with an increase in x to 0.1, the film crystallinity enhanced and the grain size increased to 400-1200 nm. Furthermore, the surface morphology was significantly improved and become smooth and compact. With a further increase in the value of x to 0.2, the film surface became very flat and dense, as displayed in Figure 6c; in addition, the grain size further increased to 0.7-1.5 µm, which was conducive to achieving high efficiencies for CZTS solar cells. Figure 6d shows the SEM image of the Cu 2 Mg x Zn 1−x SnS 4 film with x = 0.4. As observed, the grain size sharply decreased to 300-900 nm, but the grains were larger than those of Cu 2 Mg x Zn 1−x SnS 4 with x = 0, and densely stacked. The crystalline quality of the Cu 2 Mg x Zn 1−x SnS 4 film continued to deteriorate with further increase in x to 0.6. As shown in Figure 6e, the grain size of Cu 2 Mg x Zn 1−x SnS 4 film reduced to 200-400 nm, occasionally, a few larger grains were observed on the film surface. As seen, the surface morphology of the film with x = 0.6 was uneven and irregular. The bar chart in Figure 6 shows the average particle diameter as a function of the Mg content. As seen, the average size gradually increased with an increase in the value of x from 0 to 0.2 and reached the maximum at x = 0.2, then with further increase in x from 0.2 to 0.6, the size sharply decreased. It is well known that the good grain growth and smooth surface is of great significance to the fabrication of high power conversion efficiency (PCE) CZTS solar cells. Because the absorption layer with larger particle size can reduce the grain boundaries area, which is conducive to decrease the recombination of photon-generated carrier and increase the efficiency of CZTS solar cells. Based on the results of SEM, when the value of x was 0.2, it was concluded that the crystallization quality of Cu 2 Mg x Zn 1−x SnS 4 films achieved the best results, the grain size was the largest and the surface was the smoothest and denser, which is most suitable for the absorber layer of the CZTS solar cells. Nanomaterials 2019, 9, x FOR PEER REVIEW 9 of 13 crystallization quality of Cu2MgxZn1-xSnS4 films achieved the best results, the grain size was the largest and the surface was the smoothest and denser, which is most suitable for the absorber layer of the CZTS solar cells. UV-Vis-NIR spectroscopy was carried out to investigate the influence of Mg content on the band gap (Eg) values of the Cu2MgxZn1-xSnS4 (0 ≤ x ≤ 0.6) thin films. Figure 7a illustrates the plots of (αhυ) 2 against hυ for the films, where α and hυ are the absorption coefficient and photon energy, respectively. The Eg values for the Cu2MgxZn1-xSnS4 (0 ≤ x ≤ 0.6) films can be obtained by optical absorption measurements, according to Tauc's relation [40]: where A is a constant, n = 1/2, 3/2, 2 and 3 for the allowed direct, forbidden direct, allowed indirect and forbidden indirect transitions, respectively [41]. In general, Cu2MgxZn1-xSnS4 is regarded as a direct band gap semiconductor, therefore, n = 1/2. The values of Eg for the Cu2MgxZn1-xSnS4 thin films with x = 0, 0.1, 0.2, 0.4 and 0.6 calculated according to Tauc's relation were 1.43, 1.36, 1.35, 1.33 and 1.29 eV, respectively. The inset of Figure 7a shows the UV-vis absorption spectra of the representative Cu2MgxZn1-xSnS4 with x = 0.2. It was found that the Cu2MgxZn1-xSnS4 film had a stronger absorption intensity in the short wavelength range, which is suitable as the absorber layer UV-Vis-NIR spectroscopy was carried out to investigate the influence of Mg content on the band gap (E g ) values of the Cu 2 Mg x Zn 1−x SnS 4 (0 ≤ x ≤ 0.6) thin films. Figure 7a illustrates the plots of (αhυ) 2 against hυ for the films, where α and hυ are the absorption coefficient and photon energy, respectively. The E g values for the Cu 2 Mg x Zn 1−x SnS 4 (0 ≤ x ≤ 0.6) films can be obtained by optical absorption measurements, according to Tauc's relation [40]: where A is a constant, n = 1/2, 3/2, 2 and 3 for the allowed direct, forbidden direct, allowed indirect and forbidden indirect transitions, respectively [41]. In general, Cu 2 Mg x Zn 1−x SnS 4 is regarded as a direct band gap semiconductor, therefore, n = 1/2.  Figure 7a shows the UV-vis absorption spectra of the representative Cu 2 Mg x Zn 1−x SnS 4 with x = 0.2. It was found that the Cu 2 Mg x Zn 1−x SnS 4 film had a stronger absorption intensity in the short wavelength range, which is suitable as the absorber layer of the CZTS solar cells. Figure 7b shows the variation in the band gap energy as a function of the Mg content. As observed, the E g value reduced from 1.43 to 1.29 eV with an increase in x from 0 to 0.6, which can be ascribed to the change in the lattice parameter, resulting from the occupation of Zn sites by Mg. According to the first principles calculation results for the CZTS semiconductor, the minimum of the conduction band depends on the Sn 3d and S 3p antibonding orbitals and the maximum of the valence band is primarily related to p-d hybridization between Cu and S [42][43][44]. In the present work, the Mg element will take the site of Zn in CZTS, which will not affect the band gap of CZTS based on the theoretical analysis mentioned above. However, the band gap of Cu 2 Mg x Zn 1−x SnS 4 linearly varied as x increased from 0 to 0.6. Similar phenomena that the band gap of CZTS changes regularly because Zn is replaced by other elements (Cd, Ge) have been mentioned in previous studies [33,44], they ascribed the change to an increase in the unit cell volume, which led to a reduction of the CZTS solar cells. Figure 7b shows the variation in the band gap energy as a function of the Mg content. As observed, the Eg value reduced from 1.43 to 1.29 eV with an increase in x from 0 to 0.6, which can be ascribed to the change in the lattice parameter, resulting from the occupation of Zn sites by Mg. According to the first principles calculation results for the CZTS semiconductor, the minimum of the conduction band depends on the Sn 3d and S 3p antibonding orbitals and the maximum of the valence band is primarily related to p-d hybridization between Cu and S [42][43][44].
In the present work, the Mg element will take the site of Zn in CZTS, which will not affect the band gap of CZTS based on the theoretical analysis mentioned above. However, the band gap of Cu2MgxZn1-xSnS4 linearly varied as x increased from 0 to 0.6. Similar phenomena that the band gap of CZTS changes regularly because Zn is replaced by other elements (Cd, Ge) have been mentioned in previous studies [33,44], they ascribed the change to an increase in the unit cell volume, which led to a reduction in the antibonding component of the s-p and s-s hybridization between S 2− and Sn 4+ , resulting in a decrease in the minimum of the conduction band. In the present work, the substitution of Zn by Mg increased the volume of the unit cell and reduced the antibonding component of s-p and s-s hybridization between S 2− and Sn 4+ . Hence, the minimum of the conduction band and the band gap of Cu2MgxZn1-xSnS4 gradually decreased with increasing Mg content. The conductivity (ρ), carrier concentration (n) and mobility (μ) for the Cu2MgxZn1-xSnS4 (0 ≤ x ≤ 0.6) thin films were determined by the van der Pauw method at room temperature and the results are presented in Table 3. Tests were repeated on the same sample to ensure precision and reliability of the electrical performances of the Cu2MgxZn1-xSnS4 films. All samples with different Mg contents exhibited p-type semiconductor characteristics. In addition, as the value of x increased from 0 to 0.2, the carrier concentration of the Cu2MgxZn1-xSnS4 films increased from 6.95 × 10 16 cm −3 to 3.29 × 10 18 cm −3 . However, with a further increase in x from 0.2 to 0.6, the carrier concentration gradually decreased to 2.02 × 10 17 cm −3 . Simultaneously, the mobility decreased from 2.63 × 10 0 cm 2 V −1 s −1 to 1.01 × 10 −1 cm 2 V −1 s −1 with an increase in x from 0 to 0.2, and then, the mobility increased to 1.43 × 10 0 cm 2 V −1 s −1 as x increased to 0.6. The increase in the carrier concentration with an increase in x from 0 to 0.2 was attributed to the passivation of grain boundary defects, resulting from an improvement in the crystalline quality of the Cu2MgxZn1-xSnS4 thin film. As previously reported for K-doped and The conductivity (ρ), carrier concentration (n) and mobility (µ) for the Cu 2 Mg x Zn 1−x SnS 4 (0 ≤ x ≤ 0.6) thin films were determined by the van der Pauw method at room temperature and the results are presented in Table 3. Tests were repeated on the same sample to ensure precision and reliability of the electrical performances of the Cu 2 Mg x Zn 1−x SnS 4 films. All samples with different Mg contents exhibited p-type semiconductor characteristics. In addition, as the value of x increased from 0 to 0.2, the carrier concentration of the Cu 2 Mg x Zn 1−x SnS 4 films increased from 6.95 × 10 16 cm −3 to 3.29 × 10 18 cm −3 . However, with a further increase in x from 0.2 to 0.6, the carrier concentration gradually decreased to 2.02 × 10 17 cm −3 . Simultaneously, the mobility decreased from 2.63 × 10 0 cm 2 V −1 s −1 to 1.01 × 10 −1 cm 2 V −1 s −1 with an increase in x from 0 to 0.2, and then, the mobility increased to 1.43 × 10 0 cm 2 V −1 s −1 as x increased to 0.6. The increase in the carrier concentration with an increase in x from 0 to 0.2 was attributed to the passivation of grain boundary defects, resulting from an improvement in the crystalline quality of the Cu 2 Mg x Zn 1−x SnS 4 thin film. As previously reported for K-doped and Na-doped CZTSSe and CIGS solar cells [45,46], the carrier concentration markedly improved because of passivation of grain boundary defects, resulting from an improvement in the crystallization properties. Moreover, the decrease in the carrier concentration observed in this study with an increase in x from 0.2 to 0.6 was attributed to the deterioration of crystalline quality due to the small grain size, multi-hole and irregular surface morphology. Notably, the Cu 2 Mg x Zn 1−x SnS 4 film with x = 0.2 exhibited an optimal electrical conductivity.

Conclusions
In conclusion, we have prepared Cu 2 Mg x Zn 1−x SnS 4 films with different Mg contents and investigated the influence of the annealing temperature and time on the performance of the films. The optimal annealing temperature and time was found to be 580 • C and 60 min, respectively. Moreover, under the optimal annealing conditions, we investigated the effect of Mg content on the performance of the Cu 2 Mg x Zn 1−x SnS 4 films in detail. It was found that the Cu 2 Mg x Zn 1−x SnS 4 films were ideal for use as absorption layers in solar cells because of their continuous tunable band gaps, favorable photoelectric performance and high crystallinity. With an increase in x from 0 to 0.6, the band gap increased from 1.43 to 1.29 eV. Notably, the continuous tunable band gap could facilitate the tuning of band alignment at the Cu 2 Mg x Zn 1−x SnS 4 /CdS heterojunction by changing the Mg content. Furthermore, the Cu 2 Mg x Zn 1−x SnS 4 film with x = 0.2 exhibited superior crystallinity and surface morphology compared to other Cu 2 Mg x Zn 1−x SnS 4 films. Meanwhile, at x = 0.2, the electrical conductivity of Cu 2 Mg x Zn 1−x SnS 4 film reached the optimal level, with a carrier concentration of 3.29 × 10 18 cm −3 and a mobility of 1.01 × 10 −1 cm 2 V −1 s −1 . Notably, after being annealed at 580 • C for 60 min, the Cu 2 Mg x Zn 1−x SnS 4 film with the optimal Mg/(Mg + Zn) ratio of 0.2 exhibited favorable photoelectric performance and enhanced crystalline quality, making it a promising candidate for the preparation of high-efficiency solar cells with tunable band gap absorption layers.