Characterization of the Nano-Rod Arrays of Pyrite Thin Films Prepared by Aqueous Chemical Growth and a Subsequent Sulfurization

Pyrite is an earth-abundant and low-cost material with a specific collection of properties including a low band gap and high absorption coefficient of solar light. These properties make pyrite a good choice in a wide variety of applications such as catalysts, batteries, and photovoltaic devices. A thin film composed of vertically aligned pyrite nano-rods was processed via a hydration-condensation method followed by subsequent aging and sulfurization. In this process, no ionic salt was used which resulted in a lower cost process with a lower level of impurities. Field emission scanning electron microscopy, X-ray diffraction, and Raman spectroscopy analyses were used to characterize the thin films in different steps of the process. The major impurity of the final thin films was the marcasite phase according to the Raman analysis which could be minimized by lowering sulfurizing time to about 60 min. In addition, after structural, electrical, and optical characterization of thin films, these layers’ performances in a photovoltaic device were also examined. After deposition of a thin aluminum layer, Schottky-type solar cells of pyrite formed which were then illuminated to measure their current-voltage characteristics. The results show that a combination of low-cost materials and a low-cost preparation method is applicable for building future solar cells.


Introduction
From industrial and economical points of view, metal sulfides have gained increasing importance due to their remarkable physical, electrical, and chemical characteristics. Pyrite (FeS 2 ), a polymorph of iron sulfide, is a good candidate for use in solar cells since it has excellent solid-state characteristics [1]. The crystal structure of the mineral pyrite (β-FeS 2 ) is face-centered cubic and pyrite belongs to the Pa3 space group [1]. The unit cell of pyrite contains four FeS 2 formula units [2].The pyrite unit cell can be defined completely by a 0 (the lattice constant) and u (the S coefficient) that indicate the position of the Pyrite atom in the unit cell. In 1914, Bragg [3] introduced the complete crystal structure of a pyrite unit cell for the first time, and now the well-accepted values of a 0 and u are 5.416Å and 0.385Å, respectively [4,5]. Natural pyrites can be n-type or p-type (usually with arsenic), depending on the impurity type that they contain. The measured conductivity of natural pyrite is in the range of 0.02 − 562 (Ωcm) −1 [6] and its measured band gap is about 0.9 eV [2].
Nanocrystals of pyrite were successfully prepared two decades ago for the first time [7]. Recently, large [8] and small [9] clusters of pyrite nanocrystals have been reported.
Nevertheless, so far, no efficient pyrite-based solar cells have been reported. It has been claimed that the main drawback is pyrite's low illuminated open-circuit voltage (lower than 0.2 V) which is the consequence of its inherent low unilluminated short-circuit current, due to the existence of impurities and the high density of surface states of pyrite [10].

Experimental Procedures
The precursors FeCl 3 ·6H 2 O, pyrite, and HCl were purchased from Merck and used as received. Fluorine-doped tin oxide (FTO)-coated glass with 8 ohms electrical resistance was purchased from Solaronix. To prepare β − FeOOH, the ACG method was established. The process started with 0.1 M iron chloride via mixing 5.4 g of FeCl 3 ·6H 2 O in 200 mL distilled water, followed by a pH reduction to 1.5 by adding HCl.
A reflux system was prepared to run ACG at 80°C. FTO-and non-coated glass were used inside this system as two different substrates. In all samples, an orange layer was formed on both sides of the substrates, and an orange-colored sediment was deposited in the flask simultaneously. The substrates were gently removed from the flask and were quickly washed with water. Meanwhile, the sediments were dried in an oven at 50°C to obtain powder samples.
A muffle furnace was used for the calcination of the layered and powdered samples. Samples were tempered for 5 h at 400°C, and then they were cooled down to room temperature inside the furnace.
Then, the tempered samples were placed in a sealed quartz capsule with 1 g sulfur and were evacuated to 0.1 Torr, as depicted in Figure 1. Afterward, these capsules were placed in the furnace at 450°C for different times, and they were cooled to room temperature inside the furnace. Then, the samples were removed from capsules. Table 1 summerizes sample coding system used in this research. The initial letter of the sample codes shows the sample type which means that glass, FTO-coated glass substrates, and powder sample codes start with G, F, and P, respectively. The number after this letter shows the process which was applied to the sample of that type. The second number shows the process variable (usually the processing time). G2 is the G17 sample after calcination. For more information about the process, please check the Supplementary Materials section. way to prepare a solar cell. It is noteworthy to mention that, in this research, for the first time, no ionic salt was used for the ACG process. The ACG process electrolyte pH was merely controlled by HCl. The results pave the way for future low-cost solar cells.

Experimental Procedures
The precursors • 6 , pyrite, and HCl were purchased from Merck and used as received. Fluorine-doped tin oxide (FTO)-coated glass with 8 ohms electrical resistance was purchased from Solaronix. To prepare , the ACG method was established. The process started with 0.1 M iron chloride via mixing 5.4 g of • 6 in 200 mL distilled water, followed by a pH reduction to 1.5 by adding HCl.
A reflux system was prepared to run ACG at 80 ℃. FTO-and non-coated glass were used inside this system as two different substrates. In all samples, an orange layer was formed on both sides of the substrates, and an orange-colored sediment was deposited in the flask simultaneously. The substrates were gently removed from the flask and were quickly washed with water. Meanwhile, the sediments were dried in an oven at 50 ℃ to obtain powder samples.
A muffle furnace was used for the calcination of the layered and powdered samples. Samples were tempered for 5 h at 400 ℃, and then they were cooled down to room temperature inside the furnace.
Then, the tempered samples were placed in a sealed quartz capsule with 1 g sulfur and were evacuated to 0.1 Torr, as depicted in Figure 1. Afterward, these capsules were placed in the furnace at 450 ℃ for different times, and they were cooled to room temperature inside the furnace. Then, the samples were removed from capsules. Table 1 summerizes sample coding system used in this research. The initial letter of the sample codes shows the sample type which means that glass, FTO-coated glass substrates, and powder sample codes start with G, F, and P, respectively. The number after this letter shows the process which was applied to the sample of that type. The second number shows the process variable (usually the processing time). G2 is the G17 sample after calcination. For more information about the process, please check the Supplementary Materials section.  The structural, electrical, and optical properties of samples from different steps of the preparation were examined. Samples G12, G14, G17, G3, and F31 were analyzed using a field emission scanning electron microscope (FESEM). The electrical resistance of G3 was  way to prepare a solar cell. It is noteworthy to mention that, in this research, for the first time, no ionic salt was used for the ACG process. The ACG process electrolyte pH was merely controlled by HCl. The results pave the way for future low-cost solar cells.

Experimental Procedures
The precursors • 6 , pyrite, and HCl were purchased from Merck and used as received. Fluorine-doped tin oxide (FTO)-coated glass with 8 ohms electrical resistance was purchased from Solaronix. To prepare , the ACG method was established. The process started with 0.1 M iron chloride via mixing 5.4 g of • 6 in 200 mL distilled water, followed by a pH reduction to 1.5 by adding HCl.
A reflux system was prepared to run ACG at 80 ℃. FTO-and non-coated glass were used inside this system as two different substrates. In all samples, an orange layer was formed on both sides of the substrates, and an orange-colored sediment was deposited in the flask simultaneously. The substrates were gently removed from the flask and were quickly washed with water. Meanwhile, the sediments were dried in an oven at 50 ℃ to obtain powder samples.
A muffle furnace was used for the calcination of the layered and powdered samples. Samples were tempered for 5 h at 400 ℃, and then they were cooled down to room temperature inside the furnace.
Then, the tempered samples were placed in a sealed quartz capsule with 1 g sulfur and were evacuated to 0.1 Torr, as depicted in Figure 1. Afterward, these capsules were placed in the furnace at 450 ℃ for different times, and they were cooled to room temperature inside the furnace. Then, the samples were removed from capsules. Table 1 summerizes sample coding system used in this research. The initial letter of the sample codes shows the sample type which means that glass, FTO-coated glass substrates, and powder sample codes start with G, F, and P, respectively. The number after this letter shows the process which was applied to the sample of that type. The second number shows the process variable (usually the processing time). G2 is the G17 sample after calcination. For more information about the process, please check the Supplementary Materials section.  The structural, electrical, and optical properties of samples from different steps of the preparation were examined. Samples G12, G14, G17, G3, and F31 were analyzed using a field emission scanning electron microscope (FESEM The structural, electrical, and optical properties of samples from different steps of the preparation were examined. Samples G12, G14, G17, G3, and F31 were analyzed using a field emission scanning electron microscope (FESEM). The electrical resistance of G3 was measured using a four-probe surface resistance scanner. X-ray diffraction analyses were carried out for P1, P2, and P3 powders and also for Samples F1, F2, F36, F31, F3.30, F3.10, G1, G2, and G3. The X-ray spectrum was emitted from Cu − K α with 40 kV voltage and 30 mA current. Samples G3, F36, F31, F3.30, and F3.10 were used for Raman analysis with a 10 MW laser with 785 nm wavelength to analyze the thin films. A UV-Vis analysis was also carried out for G3 with a light wavelength range of 190-1100 nm.
Furthermore, the performance of the prepared pyrite layers in a photovoltaic device was also examined. To examine the photovoltaic response, a 100 nm thick aluminum layer was deposited on the samples with FTO-coated glass substrate by using thermal evaporation. The current-voltage tests were carried out for final Schottky cells under airmass 1.5 with a 100 mW/cm 2 illumination of a solar simulator.

Results and Discussion
After the ACG process, an orange semi-transparent layer was formed on the substrates. The FESEM images of Sample G17 show a large number of flower-like nano-rods which are formed on the substrate, as shown in Figure 2. The mean diameter and length of the rods were 36 nm and 235 nm, respectively, measured from the FESEM images, and therefore, the mean aspect ratio of the nano-rods was 6.5. measured using a four-probe surface resistance scanner. X-ray diffraction analyses were carried out for P1, P2, and P3 powders and also for Samples F1, F2, F36, F31, F3.30, F3.10, G1, G2, and G3. The X-ray spectrum was emitted from with 40 kV voltage and 30 mA current. Samples G3, F36, F31, F3.30, and F3.10 were used for Raman analysis with a 10 MW laser with 785 nm wavelength to analyze the thin films. A UV-Vis analysis was also carried out for G3 with a light wavelength range of 190-1100 nm.
Furthermore, the performance of the prepared pyrite layers in a photovoltaic device was also examined. To examine the photovoltaic response, a 100 nm thick aluminum layer was deposited on the samples with FTO-coated glass substrate by using thermal evaporation. The current-voltage tests were carried out for final Schottky cells under airmass 1.5 with a 100 mW/cm illumination of a solar simulator.

Results and Discussion
After the ACG process, an orange semi-transparent layer was formed on the substrates. The FESEM images of Sample G17 show a large number of flower-like nano-rods which are formed on the substrate, as shown in Figure 2. The mean diameter and length of the rods were 36 nm and 235 nm, respectively, measured from the FESEM images, and therefore, the mean aspect ratio of the nano-rods was 6.5. The XRD pattern of Sample F1 is shown in Figure 3. The characteristic peaks of fluorinedoped tin oxide (FTO) and iron oxy-hydroxide (akaganéite) are visible in the pattern. This polymorph of FeOOH forms when a small amount of Cl is present in the crystal [30] which, in our solution, was supplied by HCl. The high intensity of FTO peaks may be due to the high thickness of this layer in contrast with the iron oxy-hydroxide layer. The XRD pattern of the powder sample, Sample P1, was used to examine the crystallite size and preferential orientation of the nano-rods ( Figure 4). The crystallographic direction of each peak is labeled in Figure  4 according to the reference pattern. The average crystallite size, which can be estimated from the empirical Scherrer equation, is 37 3 nm. By compareing this average crystallite size with the particle size measured from microscopic images, it can be said that each nano-rod contains six crystallographic grains, on average. The Lotgering factor, ℎ , was also used to measure the preferential orientation of the nano-rods [31]. The thin film's orientation is totally parallel to the ℎ orientation if ℎ equals one, and is random if ℎ equals zero. The calculated Lotgering factor was compared for the major crystallographic directions of Sample F1 in  The XRD pattern of Sample F1 is shown in Figure 3. The characteristic peaks of fluorinedoped tin oxide (FTO) and iron oxy-hydroxide (akaganéite) are visible in the pattern. This polymorph of FeOOH forms when a small amount of Cl is present in the crystal [30] which, in our solution, was supplied by HCl. The high intensity of FTO peaks may be due to the high thickness of this layer in contrast with the iron oxy-hydroxide layer. The XRD pattern of the powder sample, Sample P1, was used to examine the crystallite size and preferential orientation of the nano-rods ( Figure 4). The crystallographic direction of each peak is labeled in Figure 4 according to the reference pattern. The average crystallite size, which can be estimated from the empirical Scherrer equation, is 37 ± 3 nm. By compareing this average crystallite size with the particle size measured from microscopic images, it can be said that each nano-rod contains six crystallographic grains, on average. The Lotgering factor, L(hkl), was also used to measure the preferential orientation of the nano-rods [31]. The thin film's orientation is totally parallel to the hkl orientation if L(hkl) equals one, and is random if L(hkl) equals zero. The calculated Lotgering factor was compared for the major crystallographic directions of Sample F1 in Figure 5 by using the XRD patterns of the powder sample (P1) and the thin film sample (F1). The L factors for the 310 and 211 directions are higher than the other directions. The 310 direction could be omitted because of its overlap with the FTO peak. A higher L factor in some directions means that the akaganéite nano-rods are preferentially oriented. However, in the literature, it has been reported that the orientation is parallel to the c axis [32,33]. This contradiction may be due to the lack of use of ionic salts in the ACG process. Ionic salts may adhere to some crystal facets and slow down their growth rate.
could be omitted because of its overlap with the FTO peak. A higher L factor in some directions means that the akaganéite nano-rods are preferentially oriented. However, in the literature, it has been reported that the orientation is parallel to the c axis [32,33]. This contradiction may be due to the lack of use of ionic salts in the ACG process. Ionic salts may adhere to some crystal facets and slow down their growth rate.  ture, it has been reported that the orientation is parallel to the c axis [32,33]. This contradiction may be due to the lack of use of ionic salts in the ACG process. Ionic salts may adhere to some crystal facets and slow down their growth rate.  The effect of the ACG process time on morphology was also studied. Samples with 2 (G12), 4 (G14), and 7 (G17) hours of process time are compared in Figure 6. It can be concluded that the ACG process time can affect the diameter distribution of the nano-rods, based on the demonstrated nano-rod thickness histograms in Figure 6b-f. The skewness factor was calculated for different samples, as shown in Figure 6g. The skewness factor is a measure of the asymmetry of a distribution. The sample with a 7 h ACG process time has a lower skewness factor than other samples, which means that the thickness distribution of the nano-rods in Sample G17 is more uniform. The effect of the ACG process time on morphology was also studied. Samples with 2 (G12), 4 (G14), and 7 (G17) hours of process time are compared in Figure 6. It can be concluded that the ACG process time can affect the diameter distribution of the nanorods, based on the demonstrated nano-rod thickness histograms in Figures 6b-f. The skewness factor was calculated for different samples, as shown in Figure 6g. The skewness factor is a measure of the asymmetry of a distribution. The sample with a 7 h ACG process time has a lower skewness factor than other samples, which means that the thickness distribution of the nano-rods in Sample G17 is more uniform.  The effect of the ACG process time on morphology was also studied. Samples with 2 (G12), 4 (G14), and 7 (G17) hours of process time are compared in Figure 6. It can be concluded that the ACG process time can affect the diameter distribution of the nanorods, based on the demonstrated nano-rod thickness histograms in Figures 6b-f. The skewness factor was calculated for different samples, as shown in Figure 6g. The skewness factor is a measure of the asymmetry of a distribution. The sample with a 7 h ACG process time has a lower skewness factor than other samples, which means that the thickness distribution of the nano-rods in Sample G17 is more uniform. The XRD pattern of the thin film with an FTO sublayer after calcination (F2) revealed two different phases: FTO and hematite ( ), as shown in Figure 7. During calcination, water molecules abandon the akaganéite phase and the crystal structure transforms from tetragonal (I4/m) to trigonal (R3̅ c), and finally the hematite phase starts to form. The hematite crystallite size, according to the Scherrer equation, is 42 3 nm, which suggests a little grain growth during calcination. The calculated Lotgering factors for different crystal directions of the thin film sample (F2), according to the XRD pattern of the powder sample (P2), are shown in Figure 8. The preferential crystallographic direction of hematite nano-rods is [012].
The color of all the powder and thin film samples turned black after sulfurization. The FESEM image of Sample G3 from the top view of the thin film sample showed the nano-rods, as depicted in Figure 9. The diameter of the nano-rods was measured using an image software (ImageJ) and was 74 nm on average, which was two times larger than the thickness of akaganéite nano-rods. The thickness of this layer was between 400 and 700 nm according to its crosssection microscopic image shown in Figure 10. Moreover, some impurities were deposited on the thin film which could be the sulfur condensed residuals, shown by a black arrow in Figure  10. The XRD analysis of the powder sample (P3) showed an incomplete transformation of hematite to pyrite (FeS2) after sulfurization ( Figure 11). The calculated crystallite size of pyrite crystals was 37 5 nm which was about 5 nm smaller than that of Sample P2 hematite crystallite size. Smaller pyrite crystals mean that not all of the hematite crystal was transformed to pyrite, and thus, the sulfurization time was insufficient. It has been reported that the main factor which affects the kinetics of the process is the diffusion rate of sulfur atoms into the hematite crystals and oxygen atoms out of it [18]. The color of all the powder and thin film samples turned black after sulfurization. The FESEM image of Sample G3 from the top view of the thin film sample showed the nano-rods, as depicted in Figure 9. The diameter of the nano-rods was measured using an image software (ImageJ) and was 74 nm on average, which was two times larger than the thickness of akaganéite nano-rods. The thickness of this layer was between 400 and 700 nm according to its cross-section microscopic image shown in Figure 10. Moreover, some impurities were deposited on the thin film which could be the sulfur condensed residuals, shown by a black arrow in Figure 10. The XRD analysis of the powder sample (P3) showed an incomplete transformation of hematite to pyrite (FeS 2 ) after sulfurization ( Figure 11). The calculated crystallite size of pyrite crystals was 37 ± 5 nm which was about 5 nm smaller than that of Sample P2 hematite crystallite size. Smaller pyrite crystals mean that not all of the hematite crystal was transformed to pyrite, and thus, the sulfurization time was insufficient. It has been reported that the main factor which affects the kinetics of the process is the diffusion rate of sulfur atoms into the hematite crystals and oxygen atoms out of it [18].
Marcasite is an iron sulfide phase with an orthorhombic crystal structure. A Raman spectroscopy analysis is the only way to distinguish between marcasite and pyrite phases, and it was used in this report to characterize the thin film Sample G3, as shown in Figure 12. It has been reported that the active Raman peaks of pyrite and marcasite are (353,387 cm −1 ) and (324,387 cm −1 ), respectively [34]. Accordingly, the first peak of the sample could be assigned to marcasite, the second one to pyrite, and the third one to both of them which was overlapped. Therefore, the sample layer consisted of both pyrite and marcasite phases. Pyrite was the dominant phase due to the relative Raman peak intensities. The formation of the marcasite phase in this process could be justified by the uncontrollable partial pressure of gaseous pyrite in the silicon tube which could result in the stabilization of two marcasite and pyrite phases alongside.  Marcasite is an iron sulfide phase with an orthorhombic crystal structure. A Raman spectroscopy analysis is the only way to distinguish between marcasite and pyrite phases, and it was used in this report to characterize the thin film Sample G3, as shown in Figure 12. It has been reported that the active Raman peaks of pyrite and marcasite are (353, 387 cm −1 ) and (324, 387 cm −1 ), respectively [34]. Accordingly, the first peak of the sample could be assigned to marcasite, the second one to pyrite, and the third one to both of them which was overlapped. Therefore, the sample layer consisted of both pyrite and marcasite phases. Pyrite was the dominant phase due to the relative Raman peak intensities. The formation of the marcasite phase in this process could be justified by the uncontrollable partial pressure of gaseous pyrite in the silicon tube which could result in the stabilization of two marcasite and pyrite phases alongside.  Marcasite is an iron sulfide phase with an orthorhombic crystal structure. A Raman spectroscopy analysis is the only way to distinguish between marcasite and pyrite phases, and it was used in this report to characterize the thin film Sample G3, as shown in Figure 12. It has been reported that the active Raman peaks of pyrite and marcasite are (353, 387 cm −1 ) and (324, 387 cm −1 ), respectively [34]. Accordingly, the first peak of the sample could be assigned to marcasite, the second one to pyrite, and the third one to both of them which was overlapped. Therefore, the sample layer consisted of both pyrite and marcasite phases. Pyrite was the dominant phase due to the relative Raman peak intensities. The formation of the marcasite phase in this process could be justified by the uncontrollable partial pressure of gaseous pyrite in the silicon tube which could result in the stabilization of two marcasite and pyrite phases alongside.        A UV-Vis analysis was carried out for Sample G3 to investigate the optical properties. The high absorption obtained in the UV to the visible range of the light wavelength emphasizes that pyrite is a good choice for the absorption of sunlight (Figure 13). In addition, no absorption peaks were observed in this range (313 to 1090 nm). In other words, the band gap of pyrite thin film is, at most, 1.127 eV. A UV-Vis analysis was carried out for Sample G3 to investigate the optical properties. The high absorption obtained in the UV to the visible range of the light wavelength emphasizes that pyrite is a good choice for the absorption of sunlight ( Figure 13). In addition, no absorption peaks were observed in this range (313 to 1090 nm). In other words, the band gap of pyrite thin film is, at most, 1.127 eV. A UV-Vis analysis was carried out for Sample G3 to investigate the optical properties. The high absorption obtained in the UV to the visible range of the light wavelength emphasizes that pyrite is a good choice for the absorption of sunlight ( Figure 13). In addition, no absorption peaks were observed in this range (313 to 1090 nm). In other words, the band gap of pyrite thin film is, at most, 1.127 eV. The sulfurization process time varied and the Raman shifts of each sample were obtained ( Figure 14). The samples all show Raman active peaks of pyrite and marcasite but there are also some unrecognized peaks visible in the curves. To estimate the relative contents of pyrite and marcasite, the ratios of the first active Raman peaks of pyrite and marcasite for different samples were calculated and compared in the inset in Figure 14. This ratio reveals that the sample with 1 h sulfurization time (F31) has the most pyrite to marcasite relative concentration.  The sulfurization process time varied and the Raman shifts of each sample were obtained ( Figure 14). The samples all show Raman active peaks of pyrite and marcasite but there are also some unrecognized peaks visible in the curves. To estimate the relative contents of pyrite and marcasite, the ratios of the first active Raman peaks of pyrite and marcasite for different samples were calculated and compared in the inset in Figure 14. This ratio reveals that the sample with 1 h sulfurization time (F31) has the most pyrite to marcasite relative concentration. A schematic representation of the cross-section of the final solar cell is shown in Figure 15a. The structure consists of a thin film of pyrite which is sandwiched between an FTO-coated glass and a physical vapor-deposited aluminum layer. This is a Schottky-type solar cell in which a semiconductor (pyrite) is in contact with metal (aluminum). The aluminum and FTO layers also have the role of positive and negative electrodes, respectively. A schematic representation of the cross-section of the final solar cell is shown in Figure 15a. The structure consists of a thin film of pyrite which is sandwiched between an FTO-coated glass and a physical vapor-deposited aluminum layer. This is a Schottky-type solar cell in which a semiconductor (pyrite) is in contact with metal (aluminum). The aluminum and FTO layers also have the role of positive and negative electrodes, respectively. The surface electrical resistance of pyrite, , was measured as 1. 6 10 Ω/ , using a four-point probe ohmmeter. Furthermore, the calculated work function ( ) of aluminum is 4.2 [35] and the electron affinity ( ) of pyrite is 3.28 0.14 [36]. The The surface electrical resistance of pyrite, R sh , was measured as 1.6 × 10 5 Ω/sq, using a four-point probe ohmmeter. Furthermore, the calculated work function (φ m ) of aluminum is 4.2 eV [35] and the electron affinity (χ sc ) of pyrite is 3.28 ± 0.14 eV [36]. The estimated band diagram of the contact of pyrite and aluminum is drawn by considering the above parameters in Figure 15b. The resulted band bending in the valence and conduction bands of pyrite in contact with the aluminum guide photoexcited electrons and holes toward the FTO and aluminum electrodes, respectively, before recombination occurs. This process completes the external circuit and produces electrical current. The maximum voltage that this cell can make is eV bi , as shown in Figure 15b.
The current-voltage behavior of each sample was measured in both dark and illuminated conditions. The resulting J-V curves are shown in Figure 16. Samples F3.10 ( Figure 16a) and F36 (Figure 16d) did not show any photovoltaic response, and in particular, no Schottky contacts were detected. The electrical shunts might be responsible for this, as the film formation process is not fully controlled, and thus, the pyrite thin film may not cover all the substrate. Notice that the pyrite thin film thickness is as low as 500 nm and it is not possible to be increased by this type of ACG process.
Furthermore, unknown Raman peaks, as can be observed in Figure 14, can result from unknown chemical bonds that are formed during sulfurization. These conditions can cause new surface states or band defects which, in consequence, can lower the photoelectric response.
Samples F31 and F3.30 had photovoltaic responses according to Figure 16b,c. The calculated energy conversion efficiencies (η) of Samples F31 and F3.30 were 0.0044 and 0.00025%, respectively, which were lower than the η values reported by Lin [9] and Ennaoui [18] and were consistent with those reported by Steinhagen [26] and Bi [25].
Nevertheless, the η value of the sample with 1 h sulfurization time (F31) is higher than those of the other samples. This can be related to the higher pyrite to marcasite ratio of this sample. It can be stated that the main reason for the low open-circuit voltage of these solar cells is the presence of marcasite impurity, as reported previously [37]. Moreover, the formation of new surface states in pyrite can cause the recombination of excited electrons and holes before being collected in the external circuit [2,38]. nated conditions. The resulting J-V curves are shown in Figure 16. Samples F3.10 ( Figure  16a) and F36 (Figure 16d) did not show any photovoltaic response, and in particular, no Schottky contacts were detected. The electrical shunts might be responsible for this, as the film formation process is not fully controlled, and thus, the pyrite thin film may not cover all the substrate. Notice that the pyrite thin film thickness is as low as 500 nm and it is not possible to be increased by this type of ACG process. Furthermore, unknown Raman peaks, as can be observed in Figure 14, can result from unknown chemical bonds that are formed during sulfurization. These conditions can cause new surface states or band defects which, in consequence, can lower the photoelectric response.
Samples F31 and F3.30 had photovoltaic responses according to Figure 16b,c. The calculated energy conversion efficiencies ( ) of Samples F31 and F3.30 were 0.0044 and 0.00025%, respectively, which were lower than the values reported by Lin [9] and Ennaoui [18] and were consistent with those reported by Steinhagen [26] and Bi [25].

Conclusions
In this study, an ionic salt-free ACG process followed by calcination and sulfurization was developed for preparing FeS 2 thin films. The method is inexpensive and also enters lower impurity to the product. In the ACG step, the akaganéite nano-rod array thin film was formed. Akaganéite changed to hematite, and then pyrite, respectively, in the calcination and sulfurization steps. These transitions were detected by XRD; however, further studies by Raman spectroscopy proved the presence of the marcasite phase in the final layer. This phase is not desirable and is considered to be an impurity because it reduces the photovoltaic response of the layer. The pyritization time can affect the pyrite to marcasite ratio, which can be minimized for 60 min; therefore, the photovoltaic response of the Schottky-type solar cells made from pyrite thin film was improved by optimizing sulfurization times. The low-cost material and preparation method is a suitable combination for future solar cells. However, more investigations on the detailed photovoltaic mechanisms and characterizations are suggested before using these devices in real applications. The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Conflicts of Interest:
The authors have no relevant financial or non-financial interests to disclose.