Thermoelectric properties of chalcogenide based Cu 2 + xZnSn 1-xSe 4

Quaternary chalcogenide compounds Cu 2+xZnSn1−xSe4 (0 ≤ x ≤ 0.15) were prepared by solid state synthesis. Rietveld powder X-ray diffraction (XRD) refinements combined with Electron Probe Micro Analyses (EPMA, WDS-Wavelength Dispersive Spectroscopy) and Raman spectra of all samples confirmed the stannite structure (Cu 2FeSnS4-type) as the main phase. In addition to the main phase, small amounts of secondary phases like ZnSe, CuSe and SnSe were observed. Transport properties of all samples were measured as a function of temperature in the range from 300 K to 720 K. The electrical resistivity of all samples decreases with an increase in Cu content except for Cu 2.1ZnSn0.9Se4, most likely due to a higher content of the ZnSe. All samples showed positive Seebeck coefficients indicating that holes are the majority charge carriers. The thermal conductivity of doped samples was high compared to Cu 2ZnSnSe4 and this may be due to the larger electronic contribution and the presence of the ZnSe phase in the doped samples. The maximum zT = 0.3 at 720 K occurs for Cu 2.05ZnSn0.95Se4 for which a high-pressure torsion treatment resulted in an enhancement of zT by 30% at 625 K.


I. INTRODUCTION
Quaternary compounds (I 2 -II-IV-VI 4 ) are promising materials for solar cell applications due to their suitable direct band gap E g (E g = 1.44 eV for Cu 2 ZnSnSe 4 and E g = 0.98 eV for Cu 2 CdSnSe 4 ) and high absorption coefficient for wave numbers around 10 5 cm −1 . 1 Additionally, these compounds can also be used in thermoelectric devices, which can convert thermal energy into electrical energy and vice-versa.The efficiency of a thermoelectric (TE) device depends on a material parameter defined as a dimensionless thermoelectric figure of merit zT = (S 2 σ /λ) T, where S, σ , T and λ represent the Seebeck coefficient, electrical conductivity, absolute temperature and total thermal conductivity, respectively.The total thermal conductivity is composed of an electronic (λ el ) and a phonon contribution (λ ph ).There have been many attempts to improve zT of skutterudite, clathrate materials etc. 2 by applying Slack's PGEC (Phonon Glass Electron Crystal) concept 3 to their complex crystal structures.These PGEC materials possess high electrical conductivity and at the same time low thermal conductivity like a glass.Recently there has been considerable interest in studying the thermoelectric properties of wide band gap quaternary compounds, because conventional TE materials exhibit a narrow band gap in which a bipolar effect may reduce the thermoelectric efficiency.Also, these compounds (e.g.Cu 2 ZnSnSe 4 ) have complex crystal structures, which follow the concept of two structural/functional units: 4 a Cu 2 Se 4 tetrahedral array acting as an electrically conducting unit and the other ZnSnSe 4 tetrahedral array acting as an insulating path.These compounds are like PGEC materials such as filled skutterudites, because the electrically insulating ZnSnSe 4 layer (which plays the role of filler atoms in skutterudites) responsible for low thermal conductivity is inserted between two electrically conducting Cu 2 Se 4 layers (plays a role as the charge carrying network in skutterudites) accountable for high electrical conductivity. 5Furthermore, these compounds (I 2 -II-IV-VI 4 ) have been derived from binary II 4 -VI 4 compounds with four unit cells using the concept of cross substitution (replacement of II 4 by two I, one II and one IV group element, respectively) and maintaining the ratio of the number of valence electrons to the number of atoms present in the compound as four. 6This exhibits a naturally distorted structure and may lead to a decrease in thermal conductivity. 7The improvement of zT can be obtained by enhancing electrical conductivity through doping /partial substitution and a high Seebeck coefficient because of their wide band gap and low thermal conductivity due to a naturally distorted crystal structure. 4,7 9][10][11] The maximum value of zT observed for bulk materials such as Cu 2.1 Zn 0.9 SnSe 4, Cu 2.1 Cd 0.9 SnSe 4 , Cu 2 ZnSn 0.9 In 0.1 Se 4 and Cu 2.075 Zn 0.925 GeSe 4 reached 0.91 at 860 K, 4 0.65 at 700 K, 5 0.95 at 850 K 7 and 0.45 at 670 K, 8 respectively.The nanostructured compounds Cu 2 CdSnSe 4 , Cu 2.15 Cd 0.85 SnSe 3.9 , Cu 2.15 Zn 0.85 GeSe 3.9 and non-stoichiometric Cu 2 ZnSnSe 4 when using a chemical synthesis route showed peak zT values 0.65 at 723 K, 9 0.71 at 685 K, 10 0.55 at 723 K 11 and 0.44 at 723 K, 12 respectively.Though the thermoelectric properties of the quaternary chalcogenides have been studied as above mentioned, there is no literature on phase purity and micro structural information on these compounds.
In this investigation, the authors synthesized the compounds Cu 2+x ZnSn 1−x Se 4 (with x = 0, 0.025, 0.050, 0.750, 0.100, 0.125, 0.150) through partial substitution of Cu for Sn, which is supposed to create more charge carriers and conducting pathways to improve the electrical conductivity by using the concept described above.A systematic study of the structural, phase purity and compositional analysis followed by thermoelectric properties of the prepared compounds Cu 2+x ZnSn 1−x Se 4 (with x = 0, 0.025, 0.050, 0.750, 0.100, 0.125, 0.150) is presented.
Two of the resulting samples were treated by a high-pressure torsion (HPT) process introducing additional structural defects and grain boundaries via severe plastic deformation.The influence of the HPT-treatment on the thermoelectric properties was studied.

II. EXPERIMENTAL DETAILS
The stoichiometric compounds Cu 2+x ZnSn 1−x Se 4 (x = 0, 0.025, 0.05, 0.75, 0.1, 0.125, 0.15) were prepared (about 10 g each) by a standard solid state synthesis method from ingots of starting materials (Cu 99.9999% Alfa Aesar, Zn 99.99% Sigma Aldrich, Sn 99.99% Sigma Aldrich, Se 99.999% Alfa Aesar), followed by sealing in quartz ampoules under a high vacuum of 10 −2 Pa.All the samples were slowly heated up to 1170 K and kept at this temperature for 6 hrs.Then the samples were allowed to cool down to 773 K within 24 hrs, followed by subsequent annealing at 773 K for 172 hrs.
The ingots were crushed to a particle size below 500 μm in an argon-filled glove box (O 2 -, H 2 O-content < 15 ppm) and transferred to WC-ball mill containers with WC-balls.The ball milling procedure was carried out in a high-energy planetary mill (Vario Pulverisette 4).The resulting powders were transferred into graphite dies and hot pressed in an argon atmosphere at 773 K at a pressure of 56 MPa in a HP W 200/250-2200-200-KS (FCT Systeme GmbH).The high-pressure torsion process was carried out using a device from W. Klement, Austria, equipped with an inductive heating coil and an infrared pyrometer.Sample discs with a diameter of 10 mm and a height of ∼1 mm were exposed to a temperature of 573 K and one revolution at a pressure of 4 GPa.The densities of the hot-pressed as well as the HPT-deformed samples were determined using Archimedes' principle.After ball milling and hot pressing, all the measured densities were at least >97% of the x-ray density.The HPT-treatment did not result in any detectable change of the sample density.All the samples were characterized by powder X-ray diffraction (XRD) after annealing and hot pressing by employing a Guinier-Huber Imaging plate system with CuK α1 -radiation as well as by electron probe microanalysis (EPMA) (Zeiss Supra 55 VP, 20 kV, EDX-detector (Oxford Instr.) and (JEOL JXA-8530F wavelength dispersive spectrometer (WDS)).Lattice parameters were determined by XRD using Si (a = 0.5431065 nm) as an internal standard.Crystallographic phase characterization was carried out using Rietveld refinement. 13High temperature XRD patterns of Cu 2+x ZnSn 1−x Se 4 (0 ≤ x ≤ 0.15) samples were collected by a Bruker D8 Advance X-ray diffractometer using Cu K α radiation and a Ta strip for heating.
Unpolarized Raman spectra of Cu 2+x ZnSn 1−x Se 4 (0 ≤ x ≤ 0.125) samples were recorded in the wave number range from 50 to 500 cm −1 at room temperature using a HORIBA Jobin Yvon LabRAM HR800 spectrometer equipped with an argon laser (514.5 nm).
Measurements of electrical resistivities and Seebeck-coefficients from 300 K to 700 K were carried out on an ULVAC ZEM 3-system.Thermal conductivity was evaluated from thermal diffusivity and heat capacity data collected with an ANTER Flashline 3000 unit in the same temperature range.The measurement errors for the electrical resistivity and the Seebeck coefficient are 5% and 10% for the thermal conductivity.

A. Structural characterization and phase composition
XRD-patterns of the annealed as well as the hot pressed samples revealed the presence of the stannite phase for all the samples of Cu 2+x ZnSn 1−x Se 4 (0 ≤ x ≤ 0.15) crystallizing in a tetragonal crystal structure isotypic with the structure of Cu 2 FeSnS 4 .In all the samples the presence of ZnSe as a secondary phase was confirmed mainly by WDS measurements and partially by X-ray patterns.In addition to the ZnSe phase, a CuSe phase was detected in the case of the two samples Cu 2+x ZnSn 1−x Se 4 (x = 0.125,0.15)with the highest nominal Cu-content indicating the exceeding of the solubility limit, whereas the undoped sample revealed SnSe as an additional secondary phase (see Table I).The presence of ZnSe, SnSe in the sample with nominal composition Cu 2 ZnSnSe 4 may be due to the deviation from its stoichiometry confirmed byWDS.The doping of Sn on Cu site in Cu 2 ZnSnSe 4 leads to the off stoichiometry in the nominal composition of the compounds, which further may lead to the instability of the crystal structure.The existence of non-stoichiometry in these compounds may cause an increase in ZnSe phase with increase in Cu content.The Cu 2 ZnSnSe 4 stannite structure is a derivative of the ZnSe-structure (sphalerite-type).In the chalcopyrite structure (CuFeS 2 -type) the alternate ordering of Fe and Cu atoms on the cationic sublattice causes a doubling of the unit cell in c-direction and therefore a reduction to tetragonal symmetry.Although similar lattice geometry exists, in the stannite structure the three different elements  occupying the cationic sublattice are arranged in a different manner.The ratio of the lattice parameters a/c is very close to 0.5 and thus the splitting of several reflections is rather weak and hardly resolved.Moreover overlapping with the peaks of the ZnSe-phase (with very close lattice parameter a = 0.5664(1) nm) 14 is observed in most patterns and does not allow a proper determination of the ZnSe-content from XRD.In order to disentangle the contributions of both phases a complementary technique, Raman spectroscopy, was employed.Raman spectra of samples Cu 2+x ZnSn 1−x Se 4 (0 ≤ x ≤ 0.125) including a Lorentzian fitting for the Raman spectrum of Cu 2 ZnSnSe 4 are shown in Figure 1.
The results from Raman spectroscopy essentially confirm the information extracted by XRD and WDS as for all samples the corresponding peaks of the Cu 2 ZnSnSe 4 phase were observed [see Table II], confirming the stannite structure to be the main phase. 15In addition, a secondary phase peak related to ZnSe in the sample Cu 2 ZnSnSe 4 and a peak of the CuSe phase in Cu 2+x ZnSn 1−x Se 4 (x = 0.025, 0.1, 0.125) were observed [see Table II].The amount of the ZnSe phase was estimated only from the backscattered electron micrographs.In all the cases except for the sample with nominal composition Cu 2.1 ZnSn 0.9 Se 4 (ZnSe >10%) the amount of the ZnSe-phase is below 3%, and SnSe as well as CuSe impurities are less than 2% as documented by XRD as well as WDS data [see Table I].In general, there is a unit cell volume decrease with increasing Cu-content, as both a and c slightly decrease (Fig. 2). Figure 3 shows the back-scattered electron micrographs of Cu 2 ZnSnSe 4 and Cu 2.1 ZnSn 0.9 Se 4 in which the dark regions correspond to the secondary ZnSe phase.The sample compositions derived from WDS-data are listed in Table I.Due to the very small changes in the nominal composition no clear conclusion on the substitution of Sn by Cu in Cu 2+x ZnSn 1−x Se 4 can be extracted from XRD, Raman-Spectroscopy, and WDS.High temperature XRD measurements for all samples were carried out at temperatures of 300 K, 723 K, 748 K, and 773 K respectively.No structural changes were observed (see Fig. 4 for the representative sample Cu 2.05 ZnSn 0.95 Se 4 ).The thermal expansion coefficients calculated from the lattice parameters a and c range at 11.3×10 −6 K −1 in the temperature interval from RT to 773 K [see Table III].The only observable change in the powder patterns was the appearance of two small peaks at 26.28 • and 43.56 • in 2θ , which were successfully indexed with the high temperature modification of Cu 2 Se (Pearson symbol: cF44, transition at 130 • C). 16 Here the main interest is to give the possible explanation for the impact of secondary phases on transport properties of the main phase.The ZnSe secondary phase has an adverse effect on the transport properties as compared with other secondary phases because of the presence of the high ZnSe content in all the samples.

B. Transport properties
The temperature dependent electrical resistivity of all the samples above 300 K is plotted in Fig. 5.The electrical resistivity decreases with an increasing amount of Cu-doping, except for the sample Cu 2.1 ZnSn 0.9 Se 4 (ρ 300K = 2664 × 10 −6 cm), which is attributed to the significantly higher content of the secondary ZnSe-phase (ρ 300K = 4.16 -30.1 cm for bulk) 17 and (ρ = 175 − 5 cm in between 300 K and 500 K for thin film) 18 and also may be due to higher scattering of holes with the ZnSe phase.The effect of SnSe and CuSe on the electrical resistivity is almost negligible because of their low resistivities (i.e, 0.169-0.176cm between 300 K and 780 K for SnSe, 19 0.2 − 0.1 cm between 300 K and 350 K for CuSe) 20 as compared with the electrical resistivity of ZnSe.All the doped samples exhibit a maximum in ρ(T) between 300 K and 480 K followed by an increase of electrical resistivity with temperature.This maximum is not related to a structural change of the main phase as confirmed by high-TXRD.Only the undoped sample (Cu nom = 2), exhibiting a much higher absolute electrical resistivity, shows a reverse trend.The room temperature values of electrical resistivity are decreasing dramatically for the samples Cu 2+x ZnSnSe 4 with 0 ≤ x nom ≤ 0.05 by almost 94%.The samples with a higher nominal Cu-content show a comparatively smaller change by a further decrease of 30%.Comparing the room temperature values of the electrical resistivity and Seebeck coefficient of Cu 2 ZnSnSe 4 (nominal composition) with the literature data also demonstrates the sensitivity of the transport properties to the exact composition: 33 × 10 −4 cm and 75 μV/K 4 , 22×10 −3 cm and 130 μV/K 7 and 40×10 −3 cm and 240 μV/K in this work.The difference between the transport properties of reported 4,7 and presently observed values can be explained on the basis of the non stoichiometry of the sample confirmed by the WDS.
Figure 6 displays the Seebeck coefficient S, as a function of temperature for all samples.The positive S-values evidence holes as the predominant charge carriers.S of all the samples increased with increasing temperature.The Cu-doped samples reveal an almost linear temperature dependence typically observed for metallic compounds.The Seebeck-coefficient at room temperature is systematically decreased with increasing nominal Cu-content, which may be due to the increase in carrier concentration by doping and it follows the Equation (1).The Equation (1) expresses the relationship between Seebeck coefficient and carrier concentration as which describes the diffusion thermopower in terms of a free electron model, 21 with m e being the electron mass, and e the carrier charge.As negative S values are observed for ZnSe (−178 to −370 μV/K between 300 K and 500 K for ZnSe thin film), 18 ZnSe has an adverse effect on the total S value because of negative charge carriers.Thus the total S value deceased because the presence of the electron dominant ZnSe phase in the hole dominant phase may suppress the thermoelectric voltage values due to their opposite charge carriers in the temperature gradient.
Figure 7 displays the temperature dependent thermal conductivity, λ (T), of all the samples.The electron contribution was calculated from the electrical resistivity applying the Wiedemann-Franz relation with a Lorentz number L = 2.44×10 −8 W K −2 .The remaining phonon part, λ ph , is plotted in Fig. 7.The effect of ZnSe on thermal conductivity is rather due to the high thermal conductivity value of ZnSe (λ 300K = 19 W m −1 K −1 ). 17The sample Cu 2 ZnSnSe 4 exhibits by far the smallest total as well as phonon thermal conductivity.The increase of λ for all the other samples can be partially explained by the higher electronic contribution.The micrographs clearly show that the presence of ZnSe impurities (λ 300K = 19 W m −1 K −1 ) 17 in these samples is significantly higher than in Cu 2 ZnSnSe 4 , and thus a larger contribution of this phase to the overall thermal conductivity is expected.As Umklapp-processes dominate the scattering at high temperatures, a 1/T-dependence of λ ph is found.
The thermoelectric figure of merit (zT) was evaluated from the power factor (PF), S 2 /ρ, and the thermal conductivity (λ).The results for zT are plotted in Fig. 8 and show an increase as a function of temperature for all samples.Although the doping increased the power factor for all the samples with higher Cu-content, due to the significant rise of thermal conductivity, only the sample Cu for Cu 2.05 ZnSn 0.95 Se 4 .The observed values are slightly smaller than in the literature reported for the uncoated bulk materials Cu 2.1 Zn 0.9 SnSe 4 (zT = 0.45 at 700 K), 4 and Cu 2 ZnSn 0.9 In 0.1 Se 4 (zT = 0.37 at 700 K). 7 The presence of the ZnSe secondary phase, which is believed to show an adverse effect on transport properties, prevents a better thermoelectric performance.

C. Transport properties after high-pressure torsion (HPT)
As demonstrated for skutterudite compounds, severe plastic deformation created by HPTtreatment can lead to a significant improvement in the thermoelectric figure of merit.By lowering the average grain size, enhancement of the dislocation density and introducing additional defects in the crystal structure the thermal conductivity is decreased.As the unfavorable effect on the power factor is overcompensated by this reduction, an increased zT was observed. 22he samples Cu 2 ZnSnSe 4 and Cu 2.05 ZnSn 0.95 Se 4 were exposed to the HPT process described in details in the experimental section.As expected, both samples show an increase of the electrical resistivity (Fig. 9 and 10), but in the case of Cu 2 ZnSnSe 4 a decrease of the Seebeck coefficient was observed (Fig. 9), whereas S slightly increases for Cu 2.05 ZnSn 0.95 Se 4 (Fig. 10).As the HPT was carried out at 573 K small variations in composition are most likely the reasons for the change of the Seebeck-coefficients.In Fig. 11 the power factors (PF) of the two samples before and after high-pressure torsion are displayed.In both cases PF(T) after HPT is smaller, mainly due to the increase in electrical resistivity, but for Cu 2.05 ZnSn 0.95 Se 4 the marginal higher Seebeck coefficient is almost compensating the cutback of ρ.
The thermal conductivity of the HPT-treated samples at 420 K is decreased by ∼30% in both cases, but with increasing temperature the difference to the ball milled and hot pressed samples is shrinking and expected to reach the same values around 700 K-750 K (Fig. 12).This observation leads to the conclusion that the introduced defects are already healing at these temperatures.Due to the reduced power factor accompanied by the moderate difference of λ no improvement of the thermoelectric figure of merit zT for Cu 2 ZnSnSe 4 is observed.In contrast to this, for Cu 2.05 ZnSn 0.95 Se 4 an improved thermoelectric performance up to 625 K is observed, where zT is increased by more than 30% mainly due to the decreased thermal conductivity (Fig. 13).

IV. CONCLUSIONS
The quaternary compounds Cu 2+x ZnSn 1−x Se 4 (x = 0, 0.025, 0.05, 0.75, 0.1, 0.125, 0.15) were synthesized by melting and annealing.The stannite phase with ZnSe as a secondary phase in all the samples was confirmed by Rietveld analysis of XRD patterns, WDS as well as Raman spectroscopy.In addition to these phases, the undoped and Cu 2+x ZnSn 1−x Se 4 (x = 0.125, 0.15) samples showed SnSe or CuSe as additional phases.
The temperature dependence of the electrical resistivity, the Seebeck coefficient and the thermal conductivity of samples Cu 2+x ZnSn 1−x Se 4 with 0 ≤ x ≤ 0.15 were measured.The electrical resistivity of all samples decreases with an increase of the doping level except for the sample Cu 2.1 ZnSn 0.9 Se 4 due to the presence of a higher content of ZnSe as a secondary phase (4.16 -30.1 cm at room temperature). 17The positive Seebeck coefficient in all the samples documents that holes are the majority carriers.The smallest total as well as phonon thermal conductivity was observed for Cu 2 ZnSnSe 4 with respect to the remaining doped samples.This may be attributed to the high content of the ZnSe phase (λ 300K = 19 W m −1 K −1 ). 17A dominant Umklapp scattering mechanism was found by the 1/T dependence of phonon thermal conductivity at elevated temperatures.The maximum zT = 0.3 was obtained for Cu 2.05 ZnSn 0.95 Se 4.
High-pressure torsion treatment resulted in an increase of the electrical resistivity and at the same time a significant reduction of the thermal conductivity.In case of Cu 2.05 ZnSn 0.95 Se 4 an improvement of the thermoelectric figure of merit at 625 K by more than 30% is observed.

FIG. 10 .
FIG. 10.Comparison of the electrical resistivity (full symbols) and the Seebeck-coefficient (Open symbols) as a function of temperature for Cu 2.05 ZnSn 0.95 Se 4 before and after HPT.

TABLE II .
Nominal composition, presence of stannite phase Raman modes including secondary phase modes such as ZnSe and CuSe for Cu 2+x ZnSn 1−x Se 4 (0 ≤ x≤ 0.125).