Optoelectronic Properties of CuSbS2 and Cu12Sb4S13 Thin Films for Thermoelectric Applications

This work presents a two-step procedure to obtain thin films with a combination of CuSbS2 and Cu12Sb4S13 phases for study in thermoelectric applications. The procedure consisted of the physical evaporation of sulfides layers (Sb2S3 and CuS) on glass substrates and the subsequent annealing of the samples in a N2 atmosphere. The characterizations by Raman spectroscopy and XRD revealed that the samples presented a varied percentage of Cu12Sb4S13 and CuSbS2. The results indicated that the percentage of phases depended on the initial thickness of the sulfide layers and the annealing temperature. The lower initial ratio between sulfide thicknesses and annealing temperature above 300 °C favored the formation of Cu12Sb4S13. However, the thermoelectric properties were improved when the phases coexisted in the thin film compared to samples with high percentages of Cu12Sb4S13. In this way, a sample with a power factor of 2.30 μW /cm∙ K2 at 60 ºC was identified.


Introduction
Thermoelectric materials are those that can take advantage of the Seebeck, Peltier and Thompson effects to convert heat into electrical energy or take advantage of a current passing through them to heat or cool.These materials are small, reliable, environmentally friendly; also, they do not produce noise or vibrations, can operate in wide temperature range, and have a long period of life [1][2][3] .However, for a material to be a good thermoelectric, it must have a good efficiency, which is determined by the figure of merit ZT, which is given by the equation: where S is the Seebeck coefficient, sometimes denoted by ∝, σ is the electrical conductivity and K is the thermal conductivity and T is the working temperature 2 .The higher the ZT value, the more efficient the material.To improve the ZT value, the thermal conductivity can be reduced with the help of nanoprecipitates, nanoinclusions to disperse the phonons or with material of complex structure that reduce the mean free path of the phonon.The other option is the improve of the power factor (PF), 2 S σ .The PF value has been improved thorough the regulation the convergence of electronic band structure, resonance levels and invisible dopants 3 .
Currently, the materials that have been studied for thermoelectric applications are varied, and most of them, are based on compounds derived from lead, selenium, or tellurium.However, some of these materials can be toxic and dangerous for the environment, especially those based on lead, so new alternatives to these materials have been studied.Alternatives include bismuth, copper, and sulfur compounds.Recent investigations have even focused on the study of essential aspects, and little studied, such as the effects of corrosion, the mechanical characteristics and the electrochemical responses in compounds based on bismuth telluride, given the good thermoelectric properties of this material [4][5][6] .However, some copper and sulfur compounds, as well as ternary compounds with antimony and selenium, have been little studied and have potential use in photovoltaic and thermoelectric applications.In the case of ternary compounds of Cu-Sb-S, CuSbS 2 has been studied for thermoelectric applications 7 .CuSbS 2 (chalcostebite) is a material with advantages, the elements that compose it are abundant on the earth, they are low-toxic and they are economics 8 .CuSbS 2 compound is a p-type semiconductor, with a direct optical band gap of 1.4 to 1.6 eV, with a hole density of 10 15 to 10 18 cm -3 [8][9][10][11] .Although there are various procedures to obtain thin films by thermal evaporation, either by individual elements 12 , few investigations have focused on obtaining thin films through the sequential evaporation of two sulfides 8,13 .Materials Research However, the most studied applications are for photoelectric purposes and as a thermoelectric material, there are not many reports, being the solid-state reaction technique used to obtain pellets and be studied at high pressure to know their optical properties and structural; as well as determine a possible use in thermoelectric devices 7 .Although the main challenge in the preparation of CuSbS 2 films is the presence of secondary phases such as Sb 2 S 3 (stibnite), CuS (covelite), Cu 1.8 S (digenite), Cu 3 SbS 3 (skinnerita), Cu 3 SbS 4 (famatinite) and Cu 12 Sb 4 S 13 (tetrahedrite), which are present regardless of the process or technique used 8,12,14,15 .Of these, it is tetrahedrite (Cu 12 Sb 4 S 13 ), which is also interesting to be used as a thermoelectric.In other works, the tetrahedrite has easily formed in a Cu-rich environment 10,16,17 .This phase has been obtained in thin film by the electron beam evaporation technique, with a reported p-type conductivity, a direct optical band gap of 1.8 eV and a power factor greater than 1.0, as well as a charge carrier concentration of up to (~ 10 20 cm -3 ) 17 .In some cases, Seebeck coefficient values of 80 μV/K or more have been reported for a temperature of 300 K in powders and 56.69 μV/K to 340 K in thin films, with PF values of 2.30 μW/cm•K 2 at 495 K 3,17 .The obtaining of CuSbS 2 and Cu 12 Sb 4 S 13 as thin films by the Physical Vapor Deposition (PVD) technique is of interest in this work.This kind of deposit has been used to obtain these phases by means of sequential evaporation of sulfides and using a subsequent annealing.Also, it has been reported that both phases can be found when the concentration of CuS is varied 8,16 .
This work describes the process to obtaining thin films of CuSbS 2 and Cu 12 S 4 Sb 13 for thermoelectric applications; this through a process similar described by Medina-Montes et.al. and Trejo-Zamudio et.al.The process consisted of a sequential evaporation of antimony and copper sulfides on glass substrates.The proportion of copper sulfide was varied accordingly with stoichiometry of the compound and the annealing temperature.The samples obtained were characterized by Raman spectroscopy, Ultraviolet-Visible Spectroscopy, EDS, Hall effect, X-ray Diffraction, Scanning Electron Microscope and X-Ray Photoelectron Spectrometry.Thin films of combined phases with a high Seebeck coefficient and power factor, for use as p-type material in thermoelectric modules, were obtained.

Preparation of Cu-Sb-S films
Ternary chalcogenide films were obtained by a procedure similar to that described by Medina-Montes et al. 8 and Trejo-Zamudio et al. 16 .The procedure consisted of two main stages: the sequential deposit of sulfide films and the thermally annealed of the films in a N 2 atmosphere.
Glass substrates of 7.5 cm x 2.5 cm were cleaned with soap, rinsed with distilled water several times, and treated with a chromic solution acid for 24 hours.Subsequently, the substrates were rinsed with deionized water and treated with a 3:1 mixture of water and nitric acid for additional 3 hours close to the boiling point solution.The substrates were rinsed with deionized water several times and stored in an amber glass recipient with deionized water and ethanol.
In the first stage, by means of the physical evaporation deposit technique, a layer of antimony (III) sulfide (Sb 2 S 3 ) followed by a layer of cooper (II) sulfide (CuS) were deposited in a sandwich structure on the glass substrate.Pure Sigma-Aldrich powders were used in this work.The growth of the films was at room temperature.The vacuum pressure to start the deposit was 6 x 10 -5 mbar and the distance between the source to the substrate was kept constant at 16 cm.The applied current for the antimony and copper sulfide was 120 and 210 A respectively.The deposit was in two stages, first the antimony sulfide, the vacuum is removed, and the copper sulfide is placed and deposited under the same conditions described.Finally, the sample obtained has a structure: glass substrate/Sb 2 S 3 /CuS.
The thicknesses of the Sb 2 S 3 were adjusted at 320 nm and CuS film was varied by approximately 160, 180, 200 and 220 nm.The thickness ratio between the Sb 2 S 3 and CuS were 2.00, 1.78, 1.60 and 1.45.Each sample was labeled as CuSbS 2.00, CuSbS 1.78, CuSbS 1.60 and CuSbS 1.45.The thicknesses of the samples were measured by profilometry.The second stage consisted of a thermal annealing at 250, 300, 350 and 400 °C in a quartz tubular furnace of Lindberg Blue M.An atmosphere of N 2 gas was employed for 2 h.Table 1 summarizes the conditions in which the samples were obtained and the name of the samples.

Characterization techniques
The Cu-Sb-S chalcogenide samples were characterized for study as thermoelectric material.The thicknesses of the samples were measured with an Alpha-Step D-100 KLA Tencor.Raman spectra were acquired with a Thermo Scientific DXR2 system equipped with a 633 nm laser as excitation source.X-ray diffractograms, to determine the structure of the materials, were obtained with an Empyrean diffractometer using a Cu Kα radiation (λ=1.5406Å).The transmittance of films was measured using Genesys 10 S UV-VIS, Thermo Scientific.The electric properties were measured using a Hall Effect Measurement System (ECOPIA HMS-300).Elemental composition of the samples was determined by energy dispersive spectroscopy, EDS analysis, in a Hitachi SU1510.The chemical states of the elements present in the material deposited were analyzed by XPS, in a XPS Escalab 250 Xi (Thermofisher).The Seebeck coefficient of the samples

Raman characterization
Figure 1 (a-d) shows the Raman spectra of the samples, where two signals are observed that correspond to vibrational modes with centers marked at 335 y 356 cm -1 , attributable to CuSbS 2 and Cu 12 Sb 4 S 13 compounds respectively 7,8,16,18,19 .The signal at 335 cm -1 is attributable to stretching vibrational mode form the Sb-S bond in CuSbS 2 and is close to that reported at 329 cm -1 18 .The Raman band at 356 cm -1 has been reported in the literature 8,19 and might be related with the stretching vibrations of Sb-S bond in tetrahedrite structure 8,19,20 .The Raman spectra show that the two phases, CuSbS 2 and Cu 12 Sb 4 S 13 , coexist in the samples, in some with a higher proportion than the others and depends on the temperature and the initial sulfide ratio.The samples CuSbS 2.00 at 350 ºC and CuSbS 2.00 at 400 ºC samples present the CuSbS 2 phase in greater quantity than the other samples.Not so with the samples CuSbS 2.00 at 250ºC and 300ºC, which have the CuSbS 2 phase to a lesser extent.Therefore, it can be predicted that a greater amount of initial CuS, a greater amount of Cu, favors the formation of tetrahedrite phase, when the annealing temperatures exceed 350 ºC.This confirms previous studies where it has been reported that an excess of CuS (excess of Cu) with respect to Sb 2 S 3 favors the formation of Cu 12 Sb 4 S 13 8,12,15,16   .The other samples present only a signal related with the Cu 12 Sb 4 S 13 at 356 cm -1 , although a low intensity signal can be seen at 335 cm -1 , which corresponds to chalcostebite.Some samples show a shift to the right, such as CuSbS 1.45 at 250 ºC, CuSbS 2.00 at 300 ºC and CuSbS 1.45 at 350 ºC.This could indicate stress in the lattice of material since no other signals are observed that indicate the presence of secondary phases.The full width at half maximum (FWHM) of the signal also indicates that the crystalline quality of the material is lower with respect to other samples.In the sample CuSbS 1.78 at 300 ºC and 350 ºC, a band with center at 326 cm -1 is observed, which may be due to the presence of Cu 3 SbS 4, and it has even reported that a vibrational mode at 330 cm -1 may be due to presence of Cu 3 SbS 4 15,17 ; although it could also be due to stoichiometry deficiency or presence of the CuSbS 2 phase.No bands corresponding to secondary binary phases were found, such as CuS at 470 cm -1 as explained Medina-Montes et.al, which might indicate that there are no remains of the precursors in the samples 8,21 .
According to Medina-Montes and collaborators, the Raman data in Figure 1 can be used to quantify the percentage of phases contained in the sample, for this is necessary to exclude the part of Sb 2 S 3 that did not react.Based on this, the following equation was employed: ( ) % 100 Where Cu Sb S Area − − is the area of Raman band at either 335 cm -1 (

)
CuSbS Area or 356 cm -1 ( ) Cu Sb S Area 8 .Figure 2 shows the results of the percentage of estimated phases found in each sample.It was necessary to deconvolve the signals at 335 and 356 cm -1 .According to the results, it can be observed that when the CuS ratio increase, there is a greater amount of Cu, which favors the formation of Cu 12 Sb 4 S 13 .This can be seen in Figure 2a, 2b, and 2d.Where it is observed that the percentage of Cu 12 Sb 4 S 13 increases and the percentage of CuSbS 2 decreases when the initial concentration of Cu increases, and the annealing temperature is constant.However, the percentages are very similar to those obtained in another investigation, where at 250 ºC and a sulfide ratio of 2.5, close to 20% of CuSbS 2 was obtained 8 .
Figure 2c shows that by increasing the amount of CuS, the percentage of Cu 12 Sb 4 S 13 obtained increase with respect to the other temperatures, such as the sample CuSbS 1.45 annealed at 250, 350 and 400 ºC, as well as in the samples CuSbS 1.78 annealed at 300 and 350 ºC and for the CuSbS 1.60 samples the results are similar.However, with the sample CuSbS 2.00, the opposite occurs when the temperature increase.This could be due to the fact that a percentage of Cu 12 Sb 4 S 13 , that has been formed, reacts with Sb 2 S 3 to form CuSbS 2

8
. This is likely since the samples contain a higher proportion of Sb 2 S 3 than the other samples.Furthermore, in Figure 2c, it is observed that the maximum percentage of Cu 12 Sb 4 S 13 obtained was 97% very close to the 96% obtained with the same initial sulfide ratio at 300 ºC.However, as the proportion of sulfurs decreases (CuS increases), the percentage of CuSbS 2 increases again, maybe part of the tetrahedrite could be decompose into CuSbS 2 and Sb 2 S 3 .Although the latter could be in smaller quantity and a signal in Raman is no appreciated, in addition, only the areas of two signals were used in the calculations.Figure 3a) shows that as the initial thickness of copper sulfide film increases, the tetrahedrite phase is favored if the annealing is carried out at the same temperature.The same is observed in the other diffractograms, Figure 3b and Figure 3d, where the samples with a lower ratio between thickness are those in which the Cu 12 Sb 4 S 13 phase is favored.The samples CuSbS 1.78 at 300 and 350 ºC show the tetrahedrite phase and the (111) plane in not displayed.The CuSbS 2.00 samples present both phases coexist at different temperatures, which coincides with Medina-Montes and collaborators 8 .

XRD characterization
The other samples show the presence of both phases coexisting as CuSbS 1.60 and CuSbS 1.45.It is also observed that there is a slight shift to the right of the diffraction patterns of the samples with respect to the cards, indicating stress within the samples.The samples showing the tetrahedrite phase match with the planes (200) at 2θ 17.170º, (220) at 2θ ≈ 24.366º, (222) at 2θ ≈ 29.950º, (321) at 2θ ≈ 32.411º, (400) at 2θ ≈ 34.714º, (330) at 2θ ≈ 36.899º,(440) at 2θ ≈ 49.902º and (622) at 2θ ≈ 59.303º of the PDF cards.In the samples in which the two phases are observed, some of the planes listed above can be observed, however, the signals corresponding to the planes (200) at 2θ ≈ 12.102º, (111) at 2θ ≈ 28.447º, (410) at 2θ ≈ 28.728º, (301) at 2θ ≈ 29.909º and (212) at 2θ ≈ 52.033º are also appreciated with less intensity.These last planes belong to the CuSbS2 phase of the card PDF#44-1417.
In general, the XRD results coincide with those obtained in Raman spectroscopy, since it is shown that by increasing the amount of CuS (greater amount of Cu), the formation of tetrahedrite is favored.Likewise, the CuSbS 1.78 samples annealed at 300 and 350 ºC are those that presented the highest percentage of tetrahedrite.With the diffractograms of Figure 3, the average size of the crystals (D) was calculated.For this, the Debye-Scherrer equation was employed: Where D is the average of the crystallite size, λ = 1.5406Å is the Cu K ∝ radiation, θ is Bragg angle and β is the fullwidth at half maximum of diffraction peak 22 .The results are presented in the Table 2.The results are very similar with respect to other studies, in this case the crystallite size is smaller compared to the study by Medina-Montes et.al.This difference could be due to difference in the thickness of Sb 2 S 3 used.In this study the thickness used was 320 nm, less than 400 nm, used in the other study.This could cause the films to have a lower crystallization quality and cause a small crystallite size, because they have less material and could evaporate in the annealing process, affecting the crystalline quality of the samples.Also, having less material, the crystal does not grow as when a greater amount of Sb 2 S 3 is available.

EDS Analysis
Table 2 shows the elemental composition of the samples determined by EDS.All samples were measured at two different points to determine their homogeneity.Figure 4 shows the EDS analysis for some samples, the figure includes the two measurements in different regions of the thin film.The results shown in Table 2 represent the average of the two measurements and the standard deviation (SD).The SD has small values, which indicates homogeneous distribution of the material in the thin films obtained.
As can be seen, all the samples show an excess of Cu with respect to the ideal composition of chalcostebite and tetrahedrite.Remembering that the samples that present a high percentage of Cu 12 Sb 4 S 13 are the CuSbS 1.78 samples that were annealed at 300 and 350 ºC, these samples present more than a 10% excess of Cu with respect to the ideal stoichiometry.This can affect the electrical properties of the material.Tetrahedrite is a p-type semiconductor and holes are provided by sulfur atoms, by reducing its percentage, the electrical conductivity is affected, since it has a lower charge carrier concentration with respect to the ideal stoichiometry 3,16 .
In these same samples antimony does not have many changes, but sulfur has deficiencies, this due to the annealing process, by which sulfur evaporates first, followed by antimony and copper.The other samples show an excess of Cu as well, which is explained by the annealing process and the evaporation of sulfur.It should also be considered that the samples show both phases coexisting.In the case of the CuSbS 2.00 samples annealed at 350 and 400 ºC, they are those that have an atomic percentage close to ideal but are the samples that present the highest percentage of chalcostebite.In a balance of the two phases, the amount of Cu is lower and therefore close to the ideal, but, both phases present an excess of Cu.One likely way to find an ideal stoichiometry sample using this method would be to reduce anneal times to avoid sulfur loss.As well as an annealing using high pressure with an inert gas to avoid the loss of antimony and sulfur or in a closed system.Also, annealing could be carried out in a sulfur atmosphere.

Optical absorption
Figure 5 shows the spectral transmittance curves of the CuSbS samples prepared with different Sb 2 S 3 /CuS ratios and annealed at different temperatures.The CuSbS 1.78/350 sample presents the highest transmittance in the region from 800 to 1000 nm and the CuSbS 2.00/400 sample presents the lowest transmittance.The first sample only presents the 3% of CuSb 2 and the second the 54% of CuSbS 2 .According with the literature the fundamental absorption edge of CuSbS 2 is at 900 nm 23,24 , although it could be exhibited since the 750 nm.For the Cu 12 Sb 4 S 13 is at ~600 nm 17 .The samples CuSbS 2.00 and 1.45 annealed at 250 ºC show a similar transmittance curve, with an absorption edge close at 600-650 nm and an another close at 700 nm.These samples contain a similar percentage of CuSbS 2 but in Raman spectroscopy the second showed a shift to the right, so the crystalline quality of the sample could be lower or could indicate stress in the crystal lattice or could be another phase (although in XRD no other was found).This difference can cause the edge to shift to the right and show higher transmittance with respect at the CuSbS 2.00 sample, the higher wave in its spectrum, could confirm a good crystallinity and homogeneous grain growth of the films 15 .Therefore, the effect observed in Raman may be due to stress in the lattice.The samples annealed at 300 ºC have a different transmittance curve.The CuSbS 1.78 sample shows an absorption edge at ~550-600 nm which could be attributable to the presence of tetrahedrite and an interference wave close at 700 nm attributable to the thickness 25 .On the other hand, the CuSbS 2.00 sample presents absorption edge after 600 nm and another close at 650 nm, since it contains a 45% of chalcostebite.This last sample has a transmittance spectrum like samples annealed at 250 ºC that contain around ).The CuSbS 1.78 sample annealed at 350 ºC has a similar transmittance spectrum as the CuSbS 1.78 sample annealed at 300 ºC.This sample also contains a low percentage of the secondary phase (4%) and has a lower thickness which represents a higher percentage of transmittance.Also, the second sample has a larger crystal size and higher wave in its spectrum, confirming a good crystallinity and homogeneous grain growth of the films 15 .
The CuSbS 2.00, 1.60 and 1.45 samples annealed at 350 ºC have an absorption edge close at 600 nm and to a lesser intensity at ~650-750 nm, since these contain chalcostebite in a higher percentage.The samples CuSbS 2.00 and 1.45 are very similar because the percentage of chalcostebite phase in these sample is higher than the CuSbS 1.60 sample.Also, the wave is higher in the curve of the CuSbS 2.00 sample because it has a larger crystal size.Finally, the Figure 5d shows the transmittance spectra of the samples annealed at 400 ºC.The sample CuSbS 1.60 has a spectrum very similar to that of the samples CuSbS 1.78/300 and CuSbS 1.78/350, although the absorption edge is close to 500 nm.This sample has 21% of CuSbS 2 , according to results in Raman spectroscopy, but in XRD analysis it showed tetrahedrite as the main phase and a larger crystal size, which explain a higher transmittance with respect to the rest of the samples annealed at 400 ºC.The spectra of the CuSbS 2.00 and 1.45 samples are similar, both contain chalcostebite in a percentage greater than 10%.It is the reason why these samples present an absorption edge above 600 nm.The CuSbS 2.00 sample has a lower intensity edge near at 700 nm, which confirm the presence of chalcostebite.Its spectrum is like that of other samples with the same CuS/Sb 2 S 3 ratio and annealed at other temperature but with lower transmittance.This may be because it has a smaller crystal size and may be a less homogenous sample in the growth of its grains.
Figure 6 displays the direct optical band gap for the CuSbS samples.For this, the Beer-Lambert law was used to calculate the absorbance and the absorption coefficient α .Subsequently, the band gap was estimated using the Tauc parabolic bands model.Figure 6 shows the extrapolation of the linear region of (αhν) 2 versus hν to the x-axis.According with Figure 6, the band gap of the CuSbS samples is in the range of 1.85 to 2.25 eV; values very close to those reported in the literature 17,19,26 .According to reports the band gap of Cu 12 Sb 4 S 13 can be influenced by many factors 17 .The band gap can vary due to the effect of decreasing particle size and it can increase 26 .In other cases, it has been reported that the band gap increases when the samples are rich in Cu, in these cases the highest value reported has been of 1.9 eV 27 .Prem Kumar et.al. reported values of 1.84 eV for samples with mixes of phases (Cu 12 Sb 4 S 13 and Cu 3 SbS 4 ) and rich in Cu (~55% Cu and ~37% S) and a value of 1.93 eV for a sample rich in S (~42%) 17.In this work, the samples that contain the highest percentage of tetrahedrite are the CuSbS 1.78/300 and CuSbS 1.78/350, and these samples are rich in Cu (~59%) and present values of band gap of 2.00 eV and 1.92 eV respectively.These values agree with the reported for similar samples.Also, the sample with the lowest band gap is CuSbS 2.00/400, but this sample is rich in S with respect to the other samples.However, it is a sample that contain a high percentage of CuSbS 2 and as the band gap of this phase is smaller, that is why the band gap of the sample decreases.The values are similar for the other samples with the same initial ratio of CuS/Sb 2 S 3 .The sample with the highest band gap is CuSbS 1.60/400, its value is 2.25 eV.This sample contains 79% of tetrahedrite, is rich in Cu and has a small crystal size, this in combination could explain the increased of band gap.This in contrast to the samples that have the same initial sulfides ratio (CuSbS 1.45/350 and CuSbS 1.45/250); these samples have a smaller band gap but contain a higher percentage of chalcostebite.Finally, the samples CuSbS 1.60/350 and CuSbS 1.45/400 have band gap values of 1.94 eV and 1.89 eV respectively, are similar values.The samples have similar percentage of Cu 12 Sb 4 S 13 , Cu and S.However, the CuSbS 1.60/350 sample have a larger crystal size; this could explain the higher value of band gap.

Electrical characterization
Table 3 shows the results of the electrical measurements carried out using the Hall effect.Samples of 1x1 cm were employed in the van der Pauw configuration and the thickness of the samples used in the characterization are showed in Table 2.According to the results shown in the table, it can be confirmed that the samples present the two phases coexisting.This is because the charge carrier concentration is greater than the CuSbS 2 phase and less than the Cu 12 Sb 4 S 13 phase .The charge carrier concentration reported for CuSbS 2 is 10 15 to 10 18 cm -3 8-11 and for Cu 12 Sb 4 S 13 is up to 10 20 cm -3 .When there is a secondary phase, this affects the charge carrier concentration of the main phase 28 .In this case, this parameter is affected by the percentage of the phases and by the amount of Cu and S. In some samples the charge carrier concentration drops to 10 20 cm -3 (CuSbS 1.78/350, 1.45/250, 1.45/350, 1.45/400) which are samples that contain high amounts of Cu and low amounts of S; all three samples also have a crystal size less than 20 nm.It agrees with another study where it is explained that theoretically decreasing sulfur element decrease the charge carrier concentration 3 .These same samples have a lower conductivity with respect to the others.The CuSbS 2.00/300 sample, also has lower charge carrier concentration but also has a high concentration of Cu, this effect is not observed in the CuSbS 2.00/250, CuSbS 1.78/300 and CuSbS 1.60/400.Samples with a higher amount of S y lower percentage de Cu have a higher charge carrier concentration, since the compound is a p-type semiconductor, the holes are provided by the sulfur atoms.
The mobility of the samples is in accordance with the charge carrier concentration, at greater amount of this, the mobility decreases since there are interactions between the material lattice and the charge carriers themselves.A high charge carrier concentration leads to low electrical resistivity and high conductivity, like that reported for tetrahedrite.
The highest electrical conductivity was 4.378x10 2 (1/Ωcm) for the CuSbS 2.00/300 sample which contains a considerable percentage of chalcostebite.However, the charge carrier concentration is similar for all the samples as wells the conductivities, which allows to be in range where the highest value of ZT for a semiconductor can be reached (10 19 to 10 21 cm -3 ) 2,29 .

Morphology of the samples (SEM analysis)
The morphology of the samples with the best values of the PF (CuSbS 2.00/300 and CuSbS 1.78/300) and of the samples with the highest values of Seebeck coefficient (CuSbS 1.78/350 and CuSbS 1.45/350) was studied by scanning electron microscope (SEM).The images obtained are shown in the Figure 7 (a-d).The images shown a continues and compact surface for all the samples with non-uniformly distributed grains on this.This type of morphology has been reported in other similar Cu-Sb-S films obtained by PVD 8 , E-beam evaporation 17 and chemical bath deposition 30 .The length of the diameters of some grains was measured used ImageJ software and they were labeled in Figure 7.The CuSbS 1.78/300 sample presents a grain size in the range of 20 to 144 nm, with formation of aggregates.The aggregates have sizes of 526 to 628 nm approximately.For the CuSbS 2.00/300 sample the surface morphology is more continue than the first sample.The grain size has dimensions of 28 to 86 nm approximately without aggregates.The sample CuSbS 1.45/350 has dimensions of the grain size of 33 to 115 nm and presents the formation of aggregates with sizes of 395 to  580 nm.Finally, the CuSbS 1.78/350 sample presents grain sizes of 48 to 116 nm without aggregates.The grain size is similar to that obtained by E-beam evaporation 17 ; but lower than those obtained by evaporation and diffusion processes in layers 8,15,31,32 .This could be due to the thinner thicknesses of Sb 2 S 3 used in this investigation, which could favor the loss of material during annealing and therefore smaller grain sizes.In addition, the phase found in the highest proportion is tetrahedrite and, as can be seen in the works with which it is compared, they have a greater amount of chalcostebite, and this has greater grain growth than the other phase mentioned.
The variation in the dimensions of the grain size indicates that the grain growth was not uniform throughout the film.But the grain sizes are similar for the four samples, which is consistent with XRD, with this analysis similar crystal sizes were obtained for all the samples.No significant effects of sulfide layer ratio or temperature are observed in the surface morphology of the films.In some cases, the effect of temperature on morphology has been described, where an increased in temperature favors the growth of larger grains; this due to coalescence 8,33 .In this case, it can be explained that the smaller grains combine with others and small empty spaces appears.This can be seen in Figure 7d, the small grains start to combine and empty spaces can be seen, in relation with Figure 7a, these samples have the same ratio of sulfides but different annealing temperature.Although the presence of tetrahedrite with smaller grains has also related to obtaining thin films by evaporation 8,17 ; and all the samples present a high content of tetrahedrite.
Figure 8 shows the cross-sectional images of the a) CuSbS 2.00/300, b) CuSbS 1.45/350 and c) CuSbS 1.78/350 samples.The three samples only present growth in a single layer.There are no two layers that could be due to the precursors, this could indicate a complete diffusion between the sulfides used.The CuSbS 2.00/300 sample exhibits a compact morphology and a good adhesion with the substrate.The thickness film is of 440 nm approximately and is a value very close to that obtained by profilometry (420 nm).The difference may be due to equipment calibration (measurement uncertainty).For the CuSbS 1.45/350 sample, the morphology observed is compact in an only layer and with a good adhesion.The thickness of the film is of 430 nm approximately and is very close to the obtained by profilometry (450 nm).Finally, the CuSbS 1.78/350 sample presents a compact morphology in an only layer and a good adhesion to the substrate.The thickness estimated is of 390 nm approximately and is close to that obtained by profilometry (410 nm).

XPS analysis
For the XPS analysis, all signals were adjusted to C1s which is associated with the adventitious hydrocarbon this peak es centered at 284.6 eV 34 .Additionally, Shirley's method was used to subtract the background of the signals.
For the measurements, an Al K alpha source gun type was employed, an energy step size of 0.100 eV, a pass energy of 20.0 eV and a spot size of 650 μm. Figure 9a shows the high resolution (HR-XPS) signal for Cu 2p orbit.On the other side, one the most important aspects to the deconvolution method in the XPS analysis is the calculus of the difference in the binding energy (ΔBE element ).In that sense, the figure shows two signals at 951.56 and 932.06 eV that correspond at spin-orbit splitting of Cu 2p 1/2 and Cu 2p 3/2 , respectively.So, the ΔBE Cu = 19.5 eV this value agrees to the value that was reported in a previous work 8 .The signals correspond to the Cu + valence state 35 .Figure 9b shows the XPS signals for Sb 3d species.The spectra show two signals with maximum intensity at 530.37 and 539.68 eV that correspond to the signals of Sb 3d 5/2 and Sb 3d 3/2 8,36,37 and is associated to the ΔBE Sb = 9.31 eV close to the value that was reported by Medina-Montes et al. 8 .This first pair of peaks correspond to Sb 5+ and the other pair, with maximum intensity at 538.23 and 528.92 eV, correspond to Sb 3+ , with a ΔBE Sb = 9.31 Ev 37 .Finally, the Figure 9c shows the spectra for the S 2p species.In this case, two signals at 162.43 and 161.17 eV are observed, obtaining the ΔBE S = 1.26 eV in agreement with previous work of our group 38 .These signals correspond at S 2p 1/2 and S 2p 3/2 respectively and correspond to S 2 37 .According with the results, the valance of sates is very similar with another research, with Cu + , Sb 3+ , Sb 5+ and S 2-in the formula Cu Sb Sb S + + + − 37 ; although our sample presents mixture of phases with chalcostebite and sulfur deficiencies.

Seebeck coefficient and power factor
Table 3 shows the Seebeck coefficients and power factor (PF) of all samples.All samples have values of Seebeck coefficient greater than 50 μV/K and greater than those reported for thin films at 340 K 17 .The sample CuSbS 1.78/350 reached a maximum value of 139 μV/K, which agrees with the literature, at higher charge carrier concentration the value of S decrease 2,29 .The smallest value was for the CuSbS 2.00/400 sample that has a charge carrier concentration of 2.029x10 22 .
Table 3 also presents the PF values for all samples.The greatest value was found in the CuSbS 2.00/300 sample (2.30 μW/cm K 2 ) with a charge carrier concentration of 1.849x10 21 .This is a similar value to that reported by another technique; however, the temperature differs, and a higher value could be reached.Also, in this work it was found that the sample contained mixtures of phases (Cu 12 Sb 4 S 13 and Cu 3 Sb 4 S 4 ) 17 .On the other hand, the value of Seebeck coefficient is much lower in film than in bulk, where values of ~13.0 μW/cm K 2 at 495 K have been reported.However, in this same work it is indicated that a direct comparison cannot be made and is necessary studies about the growth method and dynamics dependent stoichiometry 17 .
For the other samples, the value of PF decreases which coincides with the literature, a maximum value can be reached with a charge carrier concentration of 10 20 -10 21 29 .

Conclusions
Thin films of CuSbS 2 /Cu 12 Sb 4 S 13 phases were obtained by a two-stage process, the sequential deposition of the Sb 2 S 3 and CuS layers on glass substrates, followed by annealing in a N 2 atmosphere.It was found that the obtaining of the CuSbS 2 or Cu 12 Sb 4 S 13 phases depends on the thickness of the layers of the precursors and the annealing temperature.If the thickness of the CuS layer is increased and the Sb 2 S 3 layer is kept constant, the formation of Cu 12 Sb 4 S 13 is favored by increasing the annealing temperature.This due to a high Cu content in the samples, favored by the evaporation of S at high temperatures, which allows obtaining the tetrahedrite.However, the thermoelectric properties improved when both phases coexisted due to the combination of properties, such as the charge carries concentration and the electrical conductivity.The best value, in relation to the power factor, was 2.30 μW/cm K 2 at 60 °C for a sample with a combination of CuSbS 2 (34%) and Cu 12 Sb 4 S 13 (66%).According to the results, it can be observed that a competitive thin film was obtained in comparison with others that have been obtained by other techniques.The techniques included spin coating and e-beam evaporation; with the advantage that the technique used in this work requires less time and does not require the manufacture of targets.The purity and stoichiometry of the Cu 12 Sb 4 S 13 phase could be improved, using this method, by varying the thicknesses of Sb 2 S 3 and CuS layers; as well as modifying the annealing temperature and reducing anneal times to avoid sulfur loss.Also, the annealing process could be carried out in high pressure with an inert gas to avoid the loss of antimony and sulfur or in a closed system.Another option could be to carry out the annealing process in a sulfur atmosphere to enrich the samples with this element.However, according to the results, obtaining the pure phase does not guarantee an improve in the thermoelectric properties of the thin film compared to when both phases coexist.Finally, quantification of thermal conductivity is required to determine the ZT values of the samples.

Figure 3 (
Figure3 (a-d) shows the XRD patterns of the annealed samples.The XRD patterns of the samples were compared with the corresponding cards, PDF#44-1417 and PDF#24-1318 for CuSbS 2 (orthorhombic) and Cu 12 Sb 4 S 13 (cubic) respectively.These phases were the only detected by XRD and Raman.Figure3a) shows that as the initial thickness of copper sulfide film increases, the tetrahedrite phase is favored if the annealing is carried out at the same temperature.The same is observed in the other diffractograms, Figure3band Figure3d, where the samples with a lower ratio between thickness are those in which the Cu 12 Sb 4 S 13 phase is favored.The samples CuSbS 1.78 at 300 and 350 ºC show the tetrahedrite phase and the (111) plane in not displayed.The CuSbS 2.00 samples present both phases coexist at different temperatures, which coincides with Medina-Montes and collaborators8 .The other samples show the presence of both phases coexisting as CuSbS 1.60 and CuSbS 1.45.It is also observed that there is a slight shift to the right of the diffraction patterns of the samples with respect to the cards, indicating stress within the samples.The samples showing the tetrahedrite phase match with the planes (200) at 2θ 17.170º, (220) at 2θ ≈ 24.366º, (222) at 2θ ≈ 29.950º, (321) at 2θ ≈ 32.411º, (400) at 2θ ≈ 34.714º, (330) at 2θ ≈ 36.899º,(440) at 2θ ≈ 49.902º and (622) at 2θ ≈ 59.303º of the PDF cards.In the samples in which the two phases are observed, some of the planes listed above can be observed, however, the signals corresponding to the planes (200) at 2θ ≈ 12.102º, (111) at 2θ ≈ 28.447º, (410) at 2θ ≈ 28.728º, (301) at 2θ ≈ 29.909º and (212) at 2θ ≈ 52.033º are also appreciated with less intensity.These last planes belong to the CuSbS2 phase of the card PDF#44-1417.In general, the XRD results coincide with those obtained in Raman spectroscopy, since it is shown that by increasing the amount of CuS (greater amount of Cu), the formation of tetrahedrite is favored.Likewise, the CuSbS 1.78 samples annealed at 300 and 350 ºC are those that presented the highest percentage of tetrahedrite.

Table 1 .
Conditions of deposit CuSbS samples.

Table 2 .
Elemental composition, crystallite size and thickness of CuSbS samples.