- Email: [email protected]
Thermoelectric properties of Zn doped Cu2SnSe3
Materials Chemistry and Physics 147 (2014) 1022e1028
Contents lists available at ScienceDirect
Materials Chemistry and Physics journal homepage: www.elsevier.com/locate/matchemphys
Thermoelectric properties of Zn doped Cu2SnSe3 Ch Raju a, M. Falmbigl b, P. Rogl b, P. Heinrich c, E. Royanian c, E. Bauer c, Ramesh Chandra Mallik a, * a
Department of Physics, Indian Institute of Science, Bangalore 560012, India €hringerstrasse 42, A-1090 Wien, Austria Institute of Physical Chemistry, University of Vienna, Wa c Institute of Solid State Physics, Vienna University of Technology, Wiedner Hauptstrasse 8-10, A-1040 Wien, Austria b
h i g h l i g h t s In this investigation, the adverse effect of Zn doping on thermoelectric properties of Cu2SnSe3 is explained. The existence of non stoichiometry could influence the transport properties is explained. Doping could not influence the reduction of thermal conductivity, instead it is increased. Maximum thermoelectric figure of merit zT ¼ 0.48 at 773 K is obtained for the sample Cu2SnSe3.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 27 September 2013 Received in revised form 5 March 2014 Accepted 20 June 2014 Available online 17 July 2014
Zn doped ternary compounds Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075) were prepared by solid state synthesis. The undoped compound showed a monoclinic crystal structure as a major phase, while the doped compounds showed a cubic crystal structure confirmed by powder XRD (X-Ray Diffraction). The surface morphology and elemental composition analysis for all the samples were studied by SEM (Scanning Electron Microscopy) and EPMA (Electron Probe Micro Analyzer), respectively. SEM micrographs of the hot pressed samples showed the presence of continuous and homogeneous grains confirming sufficient densification. Elemental composition of all the samples revealed an off-stoichiometry, which was determined by EPMA. Transport properties were measured between 324 K and 773 K. The electrical resistivity decreased up to the samples with Zn content x ¼ 0.05 in Cu2ZnxSn1xSe3, and slightly increased in the sample Cu2Zn0.075Sn0.925Se3. This behavior is consistent with the changes in the carrier concentration confirmed by room temperature Hall coefficient data. Temperature dependent electrical resistivity of all samples showed heavily doped semiconductor behavior. All the samples exhibit positive Seebeck coefficient (S) and Hall coefficient indicating that the majority of the carriers are holes. A linear increase in Seebeck coefficient with increase in temperature indicates the degenerate semiconductor behavior. The total thermal conductivity of the doped samples increased with a higher amount of doping, due to the increase in the carrier contribution. The total and lattice thermal conductivity of all samples showed 1/T dependence, which points toward the dominance of phonon scattering at high temperatures. The maximum 1/TZT ¼ 0.48 at 773 K was obtained for the sample Cu2SnSe3 due to a low thermal conductivity compared to the doped samples. © 2014 Elsevier B.V. All rights reserved.
Keywords: Powder diffraction Rietveld analysis Thermoelectric effects Hall effect
1. Introduction Ternary Cu based semiconductor materials belonging to the family of I2IVVI3 (e.g. Cu2SnSe3) are promising materials for acousto-optic applications due to their low melting point, mass
* Corresponding author. Tel.: þ91 80 2293 2450; fax: þ91 80 2360 2602. E-mail addresses: [email protected], [email protected] (R.C. Mallik). http://dx.doi.org/10.1016/j.matchemphys.2014.06.054 0254-0584/© 2014 Elsevier B.V. All rights reserved.
density, high mean atomic weight and index of refraction [1e3]. These compounds are derivatives of binary compounds IIeVI, which were derived from group IV elemental semiconductors. Goodman (1958) defined these ternary semiconductors using the concept of cross substitution of IIeVI compounds by maintaining the ratio of number of valance electrons to number of atoms present in a compound to be constant [4]. Thus for example, Cu2SnSe3 can be derived from the binary compound ZnSe taking three formula units into account (Zn3Se3) and then substituting two monovalent Cu and one tetravalent Sn on three divalent Zn sites.
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
Since the multi-cation (Cu, Sn) substitution on the Zn site leads to a complex crystal structure and a large unit cell compared to the binary compound, it causes a decrease in thermal conductivity: ~27 mW cm1-K1 [5] for Cu2SnSe3 compared to 190 mW cm1K1 for ZnSe [6] at room temperature. Recently, these ternary compounds were found to be of great interest for studying thermoelectric (TE) properties due to their good TE performance, which depends on the TE figure of merit of a material, defined as ZT ¼ (S2/ r$l) T where S, r, T and l represent the Seebeck coefficient, electrical resistivity, absolute temperature and total thermal conductivity, which is composed of carrier (lE) and lattice contribution (lL), respectively. The promising values of ZT in these compounds are based on their low thermal conductivity and are further improved by optimizing the power factor via doping. Slack's PGEC (Phonon Glass Electron Crystal) [7] concept is applicable for these compounds, like for clathrate and skutterudite materials [8]. For example, the enhancement of ZT in Cu2SnSe3, is achieved by doping with In on the Sn site by Shi et al. [5] using the PGEC concept. In this compound, CueSe forms a charge carrying network that contributes to electrical conduction, and the Sn atom acts like a filler atom contributing to a reduction in lattice thermal conductivity. In a similar manner, by doping the Cu2SnSe3 with Mn and Ga, a maximum ZT of 0.41 at 716 K for Cu2Sn0.99Mn0.01Se3 [9] and 0.5 at 750 K for Cu2Ga0.075Sn0.925Se3 [10] is obtained, respectively. Skoug et al. [11] reported a minimum of lattice thermal conductivity (3.7 mW cm1-K1 at 760 K) through substitution of Se by S in the compound Cu2Sn0.925In0.075(Se0.7S0.3)3 by taking one of the optimum doping levels from Ref.5 and achieved a ZT of 0.62 at 750 K. Cho et al. [12] studied TE properties of another class of compounds, Cu2GeSe3 doped with Ga, and obtained a ZT of 0.43 at 700 K for Cu2Ga0.075Ge0.925Se3 through reduction in lattice thermal conductivity owing to large anharmonicity. Recently, a new colloidal synthetic route is used to produce nano-crystalline undoped Cu2SnSe3,3 which resulted in a reduction of thermal conductivity, and delivered a maximum ZT of 0.3 at 730 K [13], 0.34 at 598 K [14] and ~0.27 at 723 K for Cu2GeSe3 [15], respectively. In this work, authors aimed towards optimization of carrier concentration to improve the power factor (S2/r) of Cu2SnSe3 by doping Zn on the Sn site. Also the doping of Zn on the Sn site is expected to decrease thermal conductivity due to the mass fluctuation scattering. Thus, systematically the effect of Zn doping on thermoelectric properties along with structural and phase characterization of the ternary compound Cu2SnSe3 is studied. However, by doping with Zn the carrier contribution of the thermal conductivity increased significantly and thus led to a decrease in ZT. 2. Experimental details The stoichiometric compounds Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075) were prepared by solid state synthesis. The starting element ingots (Cu 99.9999% e Alfa Aesar, Zn 99.99% e Sigma Aldrich, Sn 99.99% e Sigma Aldrich, Se 99.999% e Alfa Aesar) were taken in stoichiometric ratio and sealed under high vacuum (~104 Torr) in quartz ampoules. Samples were heated up to 1173 K with a step of 0.5 K min1 and kept for 12 h at that temperature, followed by slow cooling to room temperature at 0.5 K min1. The slowly cooled samples were heated up to 773 K at 1 K min1 and annealed for 72 h. The prepared ingots were powdered using mortar and pestle. The powdered samples were ball milled by a Planetary Micro Mill Fritsch Pulverisette 7, and then pressed using an in-house built uniaxial hot press. The temperature and pressure conditions were chosen from the literature as applied for spark plasma sintering [5]. The samples were pressed uniaxially in a graphite die at 863 K with a pressure of 40 MPa for 5 min in argon atmosphere. Densities of the samples from 90% to 95% were
1023
estimated using Archimedes principle. Cylindrical samples were cut into discs of 6.0 mm diameter and a thickness of 0.5 mm for thermal conductivity measurements, and rectangular pieces of dimension (8.0 3.0 3.0) mm for Seebeck and electrical resistivity measurements. All the transport properties were measured in parallel to the pressing direction. The powder X-Ray Diffraction (XRD) patterns of hot pressed (HP) samples were collected by a Bruker D8 Advance diffractometer using Cu-Ka radiation. Rietveld refinement was carried out for the crystallographic phase identification using the FullProf software [16]. The surface morphology of hot pressed samples was inspected by an Environmental Scanning Electron Microscope (Quanta 200, ESEM), and the elemental compositional analysis of hot pressed samples was done by a JEOL JXA-8530F Electron Probe Micro analyzer (EPMA) using Wave length Dispersive Spectrometry (WDS). Electrical resistivities and Seebeck coefficient were measured between 324 K and 775 K by an ULVAC-RIKO ZEM-3 system. Thermal conductivity was collected by an ANTER Flashline 3000 unit in the temperature range between 423 K and 773 K. The measurement errors for the electrical resistivity and the Seebeck coefficient are 5% and 10% for the thermal conductivity. The room temperature (RT) Hall Effect measurements were performed on homemade equipment based on the van der Pauw method. Arbitrarily shaped samples with a height of ~0.5 mm were used and a magnetic field of 0.65 T was applied during the measurements. 3. Results and discussions 3.1. Structural and phase characterization Rietveld powder XRD refinement of the undoped hot pressed sample is shown in Fig. 1(a). It is confirmed the main phase as Cu2SnSe3 with a monoclinic crystal structure of space group Cc as a major contribution (~81% volume fraction) and a minor contribution by cubic crystal structure of F43m (~18% of volume fraction obtained from Rietveld analysis). The low angular portion of the XRD pattern for the undoped sample is shown in Fig. 2 and additional reflections confirming the monoclinic crystal structure as reported earlier [17] are clearly observed. The XRD patterns of all the doped samples showed a cubic crystal structure of space group F43m, confirmed through the Rietveld analysis shown in Fig. 1(bed). A similar change from monoclinic to cubic crystal symmetry by doping of Cu2SnSe3 is already reported [5,10]. In all the samples a small amount of a secondary SnSe phase (0.55 Vol. % for x ¼ 0, 1.05 Vol. % for x ¼ 0.025, 0.88 for x ¼ 0.05 and 1.40 Vol. % for x ¼ 0.075 estimated from Rietveld refinement) is observed. An additional phase, CuSe2 (2.80 Vol. %), is observed in the sample with highest Zn-content, may be due to the excess of doping, which disrupts the formation of the stable compound. This leads to the formation of secondary phases, which depend on the compositional changes and the temperatures involved. Recently,J.-M. Song et al. [14] observed CuSe as secondary phase which may be due to the high temperatures used for processing the material. Therefore, this may be also the possible explanation for the presence of the secondary phase CuSe2. Cu2SnSe3 exists in many modifications such as cubic [18], monoclinic [19] and monoclinic with a superstructure [20]. Fig. 3 shows ESEM secondary electron images with X10000 magnification of the hot pressed samples. A continuous and homogeneous growth of large grains with less porosity indicates the compaction of the hot pressed samples Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075) (see Fig. 3). The surface morphology of all the samples is almost similar with doping and no impurity phases are observed in all the samples. Further, EPMA measurements are carried out for finding the phase purity and exact composition in all
1024
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
Fig. 1. (aed) Rietveld refinement for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
the hot pressed samples. Fig. 4 shows back-scattered electron micrographs of hot pressed samples and confirms a homogeneous single phase region in all the samples. The elemental composition of an undoped sample deviating from ideal stoichiometry may be due to the formation of point defects (vacancies, interstitials, antisites) created by deficiency of Cu and excess of Sn (see Table 1) as it was reported earlier [21]. Also, the doped samples showed a small deviation from its nominal composition, possibly due to the introduction of extrinsic point defects (substitutional defects) created through the doping by Zn on the Sn site. The presence of non-stoichiometry and the sample preparation through defects
Fig. 2. Low angular part of the XRD pattern of Cu2SnSe3.
may affect the carrier concentration, which will influence the transport properties of these materials. 3.2. Transport properties Fig. 5 shows temperature dependent electrical resistivity measured between 324 K and 775 K for all samples. Electrical resistivity of all the samples increased with an increase in temperature, indicating a heavily doped semiconductor behavior [5,9,10]. The electrical resistivity decreased with increase in doping content of Zn up to x ¼ 0.05 and increased slightly above x ¼ 0.05 at RT. As the Zn content increases, it creates more holes because of the substitution of þ2 valent Zn on the þ4 valent Sn site. This causes an increase in carrier concentration resulting in the reduction of electrical resistivity. This consistent trend shows that the presence of the impurity phase SnSe (<2 Vol. %) estimated from powder XRD analysis in all the samples (at least up to x ¼ 0.05) might not influence the resistivity. However, for the sample with x ¼ 0.075, the resistivity slightly increased compared to x ¼ 0.05 most likely due to the presence of a higher content of impurity phases SnSe (1.40 Vol. %) and CuSe2 (2.80 Vol. %). The electrical resistivity values at 324 K decreased almost by 35% for the samples x ¼ 0 to x ¼ 0.025 and 70% for the sample x ¼ 0 to x ¼ 0.05, 0.075. A similar trend of decrease in r through doping is also found for Cu2Sn1xInxSe3 (x ¼ 0,0.05, 0.075, 0.10) [5] and Cu2Sn1xMnxSe3 (x ¼ 0.005, 0.01, 0.02, 0.05, 0.1) [9] at 300 K. Electrical resistivity decreased from 7.93 mU-cm (for x ¼ 0) to 1.07 mU-cm (for x ¼ 0.1) in Cu2Sn1xInxSe3 and 80 mU-cm (for x ¼ 0) to 0.275 mU-cm (for x ¼ 0.1) in Cu2Sn1xMnxSe3, respectively [5,9]. Previously, authors found a similar trend of electrical resistivity behavior for the doped quaternary compound Cu2þxZnSn1xSe4 (0 x 0.15) [22]. The
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
1025
Fig. 3. (aed) secondary electron SEM images for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
electrical resistivity of the undoped sample (1.134 mU-cm at 324 K) is smaller than the RT values reported in the literature [5,9]. This deviation in resistivity between the present work and literature values may be due to the presence of non-stoichiometry in the compound (see Table 1) and may also be due to differences in the sample preparation. Electrical resistivity could not follow the trend
in an expected manner due to the changes in observed chemical composition as compared with nominal composition. Therefore, in order to get a rough estimation of the doping level in the substitution of Sn by Zn in Cu2Sn1xZnxSe3 as found in EPMA results (see Table 1) and explained earlier by the authors for the sample Cu2.1Zn0.9Sn1xInxSe4 (0 x 0.1) [23], it is assumed that Cu2SnSe3
Fig. 4. (aed) Back scattered electron micrographs of EPMA for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
1026
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
Table 1 Nominal composition, EPMA composition, r, S, l, zT, n and mH for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075). Nominal composition Cu2SnSe3 Cu2Zn0.025Sn0.975Se3 Cu2Zn0.05Sn0.95Se3 Cu2Zn0.075Sn0.925Se3 a b c d
EPMA, compositiona Cu1.97Sn1.03Se3.00 Cu2.00Zn0.03Sn0.97Se3.00 Cu2.04Zn0.09Sn0.94Se2.93 Cu1.98Zn0.07Sn0.94Se3.01
r423 K (mU cm)
S423 K (mV K1)
l423 K
(mW cm1 K1)
zT (423 K)
nb (1021 cm3)
nc (1021 cm3)
nd (1021 cm3)
mH
1.533 0.885 0.409 0.411
62.5 59.4 33.3 40.0
17.6 25.9 32.3 34.0
0.061 0.041 0.033 0.032
1.17 1.91 2.66 2.33
e 0.32 0.98 0.76
0.72 1.05 2.91 2.23
7.1 7.8 6.4 8.2
(cm2 V1s1)
EPMA data normalized to 6 atoms/unit cell. System internal pure element standards were used for calibration. n value estimated from S(T) fit data. n value calculated from simple valence counting. n value from the Hall measurement.
is a charge balanced compound and the substitution of Zn on Sn (Zn has two valence electrons less than Sn) leads to an increase of the p-type doping. The observed Zn contents from the EPMA measurements reveal the doping quantities of 0.03 for Cu2.00Zn0.03Sn0.97Se3.00, 0.09 for Cu2.04Zn0.09Sn0.94Se2.93 and 0.07 for Cu1.98Zn0.07Sn0.94Se3.01 per formula unit, respectively. This could explain the trend of electrical resistivity for the doped samples. Fig. 6 shows the Seebeck coefficient as a function of temperature measured in the range 324e775 K. All the samples show positive Seebeck coefficient indicating that the dominant carriers are holes. All the samples showed an almost linear temperature dependence of Seebeck coefficient indicating that the degenerate semiconductor behavior follows the same trend as reported for the doped Cu2SnSe3 [5,9,10]. The Seebeck coefficient decreases with increase of doping content up to x ¼ 0.05 and increases for x ¼ 0.075, which may be due to a variation in the charge carrier density (n). This can be roughly estimated by a linear slope of S (T) using Mott's formula for the diffusion thermo power of a free electron gas at temperatures well above the Debye temperature (qD):
Sd ðT > qD Þ ¼
p2 k2B 2m* 2=3 T eZ 3np2
(1)
where m* being the carrier mass, e the carrier charge, kB the Boltzmann constant, ħ the Planck's constant, T the temperature, and n the charge carrier density, respectively [24]. A similar trend of Seebeck coefficient for the doped samples in quaternary compounds was observed by the authors in their previous report [22]. For the estimation of n, S(T) data linearly fitted well up to 700 K for all the samples by considering m* ¼ me, as explained by authors in a recent report [23]. This results in n ¼ 1.17 1021 cm3 for Cu2SnSe3, n ¼ 1.91 1021 cm3 for Cu2Zn0.025Sn0.975Se3, n ¼ 2.66 1021 cm3 for Cu2Zn0.05Sn0.95Se3 and n ¼ 2.33 1021 cm3 for Cu2Zn0.075Sn0.925Se3. A simple valence electron counting is applied taking the unit cell dimensions obtained from XRD for the doped samples by assuming that, only doping amounts of Zn contribute to the charge carrier density (Cu2SnSe3 as charge balanced), which results in n ¼ 0.32 1021 cm3 for Cu2Zn0.025Sn0.975Se3, n ¼ 0.98 1021 cm3 for Cu2Zn0.05Sn0.95Se3 and n ¼ 0.76 1021 cm3 for Cu2Zn0.075Sn0.925Se3, respectively. It is pointed out that the charge carrier densities follow a similar trend for all doped samples with small variation in magnitude, although both methods give rough estimations. To find the exact charge carrier densities, Hall measurements were carried out for all the samples at RT. Carrier concentrations (n ¼ 1/(e RH)) were calculated using the room temperature Hall coefficient (RH) data, and are plotted together in Fig. 7 with the estimated carrier concentrations using both methods discussed above as a function of Zn content. The positive RH values for all the samples indicate p-type conductivity, which is in agreement with the Seebeck data. The general trend of measured carrier concentration with doping is similar,
with a small variation in the magnitude compared to the values found from simple estimation methods. It is observed that the n value for the undoped sample is 7.25 1020 cm3, which is at least one/two orders of magnitude larger compared to the reported data 0.66 1020 cm3 [5] and (0.04 1020 cm3) [9], respectively. This difference could be explained by the existence of nonstoichiometry in the compound observed by EPMA, because the transport properties are sensitive to the exact composition of the sample. Using the carrier concentration (n) and electrical resistivity (r), Hall mobility (mH ¼ 1/ner) was calculated for all the samples (see Table 1). Hall mobility of the undoped sample 7.1 cm2 V1 s1 is slightly smaller than the literature value of 11.8 cm2 V1 s1 [5]. This may be due to the presence of high carrier concentration compared to the literature [5], and it leads to the scattering of more carriers resulting in less mobility. Fig. 8 shows a Pisarenko plot, Seebeck coefficient (S) as function of carrier concentration (n) for Cu2ZnxSn1xSe3, including undoped Cu2InxSn1xSe3 at RT from the literature. There are several reports on various thermoelectric materials discussing the Pisarenko plot [25e27]. The effective mass (m*) is calculated using the RT values of S and n via Eq. (1). The solid line in Fig. 8 represents the calculated S values for m* ¼ 0.72 me as a function of n. It is observed that S values for all the samples in the present work as well as reported as a function of n are almost in agreement with the values predicted for a single parabolic band behavior of m* ¼ 0.72 me. The effective masses increase with increase in doping content and are in the range of 0.50 me to 0.95 me for Cu2ZnxSn1xSe3 (0 x 0.075) and a similar trend is observed with the doping in the literature [5] (calculated m* using the S, n values at RT from Eq. (1)). The power factor (S2/r) is evaluated from the electrical resistivity and Seebeck coefficient values in the entire temperature
Fig. 5. Temperature dependent electrical resistivity for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
1027
Fig. 6. Seebeck coefficient as a function of temperature for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
range. The maximum power factor 8 mW cm1-K2 at 775 K has been achieved for the sample Cu2Zn0.05Sn0.95Se3. This value is somewhat smaller in magnitude compared with the power factor ~12 mW cm1-K2 around 800 K for Cu2Sn0.9In0.1Se3, and 9.5 mW cm1-K2 at 670 K Cu2Sn0.99Mn0.01Se3, respectively [5,9]. The variation of the power factor is mainly due to the lower values of the Seebeck coefficient, which is affected by the high charge carrier density. The temperature dependent total thermal conductivity measured between 423 K and 773 K for all samples is plotted in Fig. 9. The total thermal conductivity of all the samples decreased with an increase in temperature, which is similar to the reports in the literature [5,9] explaining the dominance of phonon scattering at high temperatures. The l(T) for the undoped sample is almost close to the reported data. Thermal conductivity of all the doped samples is expected to decrease with increase in doping content due to mass fluctuation scattering, but it increased with increase of doping content due to the increase in carrier contribution which is related to the electrical resistivity. The smallest total thermal conductivity value 17.6 mW cm1-K1 is obtained for the undoped sample at 423 K, which compares well to the literature value (~19 mW cm1-K1 at 425 K) [5]. These values are moderate, and about an order of magnitude smaller compared to the binary
Fig. 7. Hole concentration as a function of the Zn-content for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075), Hall measurement, estimated from S(T) data and simple valence counting (for details see text).
Fig. 8. Pisarenko plot, Seebeck coefficient (S) as a function of charge carrier density (n) for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075) including reported values for Cu2InxSn1xSe3 (x ¼ 0, 0.05, 0.1) at room temperature.
compound ZnSe (190 mW cm1-K1 at 300 K) [6], which may be due to the substitution of multi-cations I and IV (Cu and Sn) on the II (Zn) site leading to a distortion in the crystal structure. Also, the lower values of thermal conductivity for all the samples may be due to the presence of porosity, which is one possible factor for reduction in thermal conductivity. The thermal conductivity of carrier contribution was calculated using the WiedemanneFranz relation lE ¼ LT/r, where L ¼ 2.45 108 WUK2, r is electrical resistivity and T is absolute temperature. The lattice part of thermal conductivity calculated by subtracting the carrier part from the total thermal conductivity is shown in Fig. 9. The trend in the temperature dependence of lattice thermal conductivity for all the samples is similar and the almost 1/T dependence indicates a dominant Umklapp phononephonon scattering at high temperature. Fig. 10 shows the thermoelectric figure of merit (ZT) as a function of temperature. ZT was evaluated from the power factor (S2/r) and total thermal conductivity (l). The figure of merit increases with increase in temperature for all the samples. Although the maximum power factor 8 mW cm1-K2 was found for the sample Cu2Zn0.05Sn0.95Se3 by doping, a significant drop in ZT is obtained
Fig. 9. Temperature dependent total (solid symbols) and Lattice (open symbols) thermal conductivity for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
1028
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
to Cu2Zn0.05Sn0.95Se3. So the creation of more carriers by doping causes suppression in the Seebeck coefficient as well as a significant increase in the carrier contribution of the thermal conductivity, leading to an overall decrease in the thermoelectric figure of merit. Acknowledgments The authors would like to thank Prof. Arun M Umarji for providing the hot press facility and the Department of Science & Technology (DST), India for financial support through Grant No INT/ AUA/BMWF/P-13/2011 and the OEAD for support via the WTZ Austria-India under grant IN 09/2011. References
Fig. 10. Thermoelectric figure of merit (ZT) as function of temperature for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
due to the rise of thermal conductivity by the large carrier contribution. The maximum thermoelectric figure of merit 0.48 at 773 K was found for the undoped sample due to its low thermal conductivity. The adverse effect on ZT for the doped samples is due to their low power factor and high thermal conductivity, affected by the high charge carrier concentration leading to lowering the Seebeck coefficient and increasing the carrier thermal conductivity. 4. Conclusions The ternary diamond like semiconductors Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075) were prepared by melting and subsequent annealing. The XRD pattern revealed a monoclinic crystal structure as a major phase for the undoped sample and a cubic crystal structure for the doped samples. The surface morphology and elemental composition of the hot pressed samples was studied by SEM and EPMA, respectively. The electrical resistivity decreased with increase in Zn content up to x ¼ 0.05 and increased for x ¼ 0.075 at RT. This is due to the change in carrier concentration through doping, confirmed by the Hall data at RT. The positive Seebeck and Hall coefficient of all the samples indicate that the majority of the carriers are holes. The Seebeck coefficient of all the samples followed a similar trend as electrical resistivity. The total and lattice thermal conductivity decreased with increase in temperature, due to the dominance of phonon scattering at elevated temperatures. The lowest total thermal conductivity was found for the undoped sample, whereas it increased with increase in doping. This may be due to the increase in the carrier contribution to l, with doping. Although the maximum power factor of 8 mW cm1-K2 was found for the sample Cu2Zn0.05Sn0.95Se3, the maximum thermoelectric figure of merit (ZT) 0.48 at 773 K was obtained for the sample Cu2SnSe3 due to its low thermal conductivity as compared
[1] G. Marcano, C. Rincon, L.M. de Chalbaud, D.B. Bracho, G.S. Perez, J. Appl. Phys. 90 (2001) 1847e1853. [2] G.S. Babu, Y.B.K. Kumar, Y.B.K. Reddy, V.S. Raja, Mater. Chem. Phys. 96 (2006) 442e446. [3] G. Hema Chandra, O. Lakshmana Kumar, R. Prasada Rao, S. Uthanna, J. Mater. Sci. 46 (2011) 6952e6959. [4] C.H.L. Goodman, J. Phys. Chem. Solids 6 (1958) 305e314. [5] X. Shi, L. Xi, J. Fan, W. Zhang, L. Chen, Chem. Mater. 22 (2010) 6029e6031. [6] G.A. Slack, Phys. Rev. B 6 (1972) 3791e3800. [7] G.A. Slack, New materials and performance limits for thermoelectric cooling, in: D.M. Rowe (Ed.), CRC Handbook of Thermoelectrics, CRC, Boca Raton, 1995, p. 407. [8] G.S. Nolas, X. Lin, J. Martin, M. Beekman, H. Wang, J. Electron. Mater. 38 (2009) 1052e1055. [9] X. Lu, D. Morelli, J. Electron. Mater. 41 (2012) 1554e1558. [10] J. Fan, H. Liu, X. Shi, S. Bai, X. Shi, L. Chen, Acta. Mater. 61 (2013) 4297e4304. [11] E. Skoug, J. Cain, D. Morelli, J. Electron. Mater. 41 (2012) 1232e1236. [12] J.Y. Cho, X. Shi, J.R. Salvador, G.P. Meisner, J. Yang, H. Wang, A.A. Wereszczak, X. Zhou, C. Uher, Phys. Rev. B 84 (2011) 085207. [13] M. Ibanez, D. Cadavid, U. Anselmi-Tamburini, R. Zamani, S. Gorsse, W. Li, A.M. Lopez, J.R. Morante, J. Arbiol, A. Cabot, J. Mater. Chem. A 1 (2013) 1421e1426. [14] J.-M. Song, Y. Liu, H.-L. Niu, C.-J. Mao, L.-J. Cheng, S.-Y. Zhang, Y.-H. Shen, J. Alloys.Compd. 581 (2013) 646e652. n ~ ez, R. Zamani, W. Li, D. Cadavid, S. Gorsse, N.A. Katcho, A. Shavel, [15] M. Iba pez, J.R. Morante, J. Arbiol, A. Cabot, Chem. Mater. 24 (2012) A.M. Lo 4615e4622. [16] T. Roisnel, J. Rodriguez-Carvajal, Mater. Sci. Forum 118 (2001) 378e381. n, Mater. Res. Bull. 38 (2003) [17] G.E. Delgado, A.J. Mora, G. Marcano, C. Rinco 1949e1955. [18] B.B. Sharma, R. Ayyar, H. Singh, Phys.Status. Solidi A 40 (1977) 691e696. n, G. S rez, Mater. Lett. 53 [19] G. Marcano, L.M. de Chalbaud, C. Rinco anchez Pe (2002) 151e154. [20] L.D. Gulay, M. Daszkiewicz, T.A. Ostapyuk, O.S. Klymovych, O.F. Zmiy, Acta. Crystallogr. C 66 (2010) i58ei60. [21] D.-H. Kuo, W.-D. Haung, Y.-S. Huang, J.-D. Wu, Y.-J. Lin, Thin Solid Films 518 (2010) 7218e7221. [22] C. Raju, M. Falmbigl, P. Rogl, X. Yan, E. Bauer, J. Horky, M. Zehetbauer, R.C. Mallik, AIP Adv. 3 (2013) 032106e032112. [23] R. Chetty, M. Falmbigl, P. Rogl, P. Heinrich, E. Royanian, E. Bauer, S. Suwas, R.C. Mallik, Phys. Status Solidi A 210 (2013) 2471e2478. [24] F.J. Blatt, Physics of Electronic Conduction in Solids, McGraw-Hill, New York, 1968. [25] Y. Pei, N.A. Heinz, A. LaLonde, G.J. Snyder, Energy Environ. Sci. 4 (2011) 3640e3645. €fer, G.J. Snyder, [26] W.G. Zeier, A. LaLonde, Z.M. Gibbs, C.P. Heinrich, M. Pantho W. Tremel, J. Am. Chem. Soc. 134 (2012) 7147e7154. [27] M. Falmbigl, N. Nasir, A. Grytsiv, P. Rogl, S. Seichter, A. Zavarsky, E. Royanian, E. Bauer, J. Phys. D Appl. Phys. 45 (2012) 215308.
Contents lists available at ScienceDirect
Materials Chemistry and Physics journal homepage: www.elsevier.com/locate/matchemphys
Thermoelectric properties of Zn doped Cu2SnSe3 Ch Raju a, M. Falmbigl b, P. Rogl b, P. Heinrich c, E. Royanian c, E. Bauer c, Ramesh Chandra Mallik a, * a
Department of Physics, Indian Institute of Science, Bangalore 560012, India €hringerstrasse 42, A-1090 Wien, Austria Institute of Physical Chemistry, University of Vienna, Wa c Institute of Solid State Physics, Vienna University of Technology, Wiedner Hauptstrasse 8-10, A-1040 Wien, Austria b
h i g h l i g h t s In this investigation, the adverse effect of Zn doping on thermoelectric properties of Cu2SnSe3 is explained. The existence of non stoichiometry could influence the transport properties is explained. Doping could not influence the reduction of thermal conductivity, instead it is increased. Maximum thermoelectric figure of merit zT ¼ 0.48 at 773 K is obtained for the sample Cu2SnSe3.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 27 September 2013 Received in revised form 5 March 2014 Accepted 20 June 2014 Available online 17 July 2014
Zn doped ternary compounds Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075) were prepared by solid state synthesis. The undoped compound showed a monoclinic crystal structure as a major phase, while the doped compounds showed a cubic crystal structure confirmed by powder XRD (X-Ray Diffraction). The surface morphology and elemental composition analysis for all the samples were studied by SEM (Scanning Electron Microscopy) and EPMA (Electron Probe Micro Analyzer), respectively. SEM micrographs of the hot pressed samples showed the presence of continuous and homogeneous grains confirming sufficient densification. Elemental composition of all the samples revealed an off-stoichiometry, which was determined by EPMA. Transport properties were measured between 324 K and 773 K. The electrical resistivity decreased up to the samples with Zn content x ¼ 0.05 in Cu2ZnxSn1xSe3, and slightly increased in the sample Cu2Zn0.075Sn0.925Se3. This behavior is consistent with the changes in the carrier concentration confirmed by room temperature Hall coefficient data. Temperature dependent electrical resistivity of all samples showed heavily doped semiconductor behavior. All the samples exhibit positive Seebeck coefficient (S) and Hall coefficient indicating that the majority of the carriers are holes. A linear increase in Seebeck coefficient with increase in temperature indicates the degenerate semiconductor behavior. The total thermal conductivity of the doped samples increased with a higher amount of doping, due to the increase in the carrier contribution. The total and lattice thermal conductivity of all samples showed 1/T dependence, which points toward the dominance of phonon scattering at high temperatures. The maximum 1/TZT ¼ 0.48 at 773 K was obtained for the sample Cu2SnSe3 due to a low thermal conductivity compared to the doped samples. © 2014 Elsevier B.V. All rights reserved.
Keywords: Powder diffraction Rietveld analysis Thermoelectric effects Hall effect
1. Introduction Ternary Cu based semiconductor materials belonging to the family of I2IVVI3 (e.g. Cu2SnSe3) are promising materials for acousto-optic applications due to their low melting point, mass
* Corresponding author. Tel.: þ91 80 2293 2450; fax: þ91 80 2360 2602. E-mail addresses: [email protected], [email protected] (R.C. Mallik). http://dx.doi.org/10.1016/j.matchemphys.2014.06.054 0254-0584/© 2014 Elsevier B.V. All rights reserved.
density, high mean atomic weight and index of refraction [1e3]. These compounds are derivatives of binary compounds IIeVI, which were derived from group IV elemental semiconductors. Goodman (1958) defined these ternary semiconductors using the concept of cross substitution of IIeVI compounds by maintaining the ratio of number of valance electrons to number of atoms present in a compound to be constant [4]. Thus for example, Cu2SnSe3 can be derived from the binary compound ZnSe taking three formula units into account (Zn3Se3) and then substituting two monovalent Cu and one tetravalent Sn on three divalent Zn sites.
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
Since the multi-cation (Cu, Sn) substitution on the Zn site leads to a complex crystal structure and a large unit cell compared to the binary compound, it causes a decrease in thermal conductivity: ~27 mW cm1-K1 [5] for Cu2SnSe3 compared to 190 mW cm1K1 for ZnSe [6] at room temperature. Recently, these ternary compounds were found to be of great interest for studying thermoelectric (TE) properties due to their good TE performance, which depends on the TE figure of merit of a material, defined as ZT ¼ (S2/ r$l) T where S, r, T and l represent the Seebeck coefficient, electrical resistivity, absolute temperature and total thermal conductivity, which is composed of carrier (lE) and lattice contribution (lL), respectively. The promising values of ZT in these compounds are based on their low thermal conductivity and are further improved by optimizing the power factor via doping. Slack's PGEC (Phonon Glass Electron Crystal) [7] concept is applicable for these compounds, like for clathrate and skutterudite materials [8]. For example, the enhancement of ZT in Cu2SnSe3, is achieved by doping with In on the Sn site by Shi et al. [5] using the PGEC concept. In this compound, CueSe forms a charge carrying network that contributes to electrical conduction, and the Sn atom acts like a filler atom contributing to a reduction in lattice thermal conductivity. In a similar manner, by doping the Cu2SnSe3 with Mn and Ga, a maximum ZT of 0.41 at 716 K for Cu2Sn0.99Mn0.01Se3 [9] and 0.5 at 750 K for Cu2Ga0.075Sn0.925Se3 [10] is obtained, respectively. Skoug et al. [11] reported a minimum of lattice thermal conductivity (3.7 mW cm1-K1 at 760 K) through substitution of Se by S in the compound Cu2Sn0.925In0.075(Se0.7S0.3)3 by taking one of the optimum doping levels from Ref.5 and achieved a ZT of 0.62 at 750 K. Cho et al. [12] studied TE properties of another class of compounds, Cu2GeSe3 doped with Ga, and obtained a ZT of 0.43 at 700 K for Cu2Ga0.075Ge0.925Se3 through reduction in lattice thermal conductivity owing to large anharmonicity. Recently, a new colloidal synthetic route is used to produce nano-crystalline undoped Cu2SnSe3,3 which resulted in a reduction of thermal conductivity, and delivered a maximum ZT of 0.3 at 730 K [13], 0.34 at 598 K [14] and ~0.27 at 723 K for Cu2GeSe3 [15], respectively. In this work, authors aimed towards optimization of carrier concentration to improve the power factor (S2/r) of Cu2SnSe3 by doping Zn on the Sn site. Also the doping of Zn on the Sn site is expected to decrease thermal conductivity due to the mass fluctuation scattering. Thus, systematically the effect of Zn doping on thermoelectric properties along with structural and phase characterization of the ternary compound Cu2SnSe3 is studied. However, by doping with Zn the carrier contribution of the thermal conductivity increased significantly and thus led to a decrease in ZT. 2. Experimental details The stoichiometric compounds Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075) were prepared by solid state synthesis. The starting element ingots (Cu 99.9999% e Alfa Aesar, Zn 99.99% e Sigma Aldrich, Sn 99.99% e Sigma Aldrich, Se 99.999% e Alfa Aesar) were taken in stoichiometric ratio and sealed under high vacuum (~104 Torr) in quartz ampoules. Samples were heated up to 1173 K with a step of 0.5 K min1 and kept for 12 h at that temperature, followed by slow cooling to room temperature at 0.5 K min1. The slowly cooled samples were heated up to 773 K at 1 K min1 and annealed for 72 h. The prepared ingots were powdered using mortar and pestle. The powdered samples were ball milled by a Planetary Micro Mill Fritsch Pulverisette 7, and then pressed using an in-house built uniaxial hot press. The temperature and pressure conditions were chosen from the literature as applied for spark plasma sintering [5]. The samples were pressed uniaxially in a graphite die at 863 K with a pressure of 40 MPa for 5 min in argon atmosphere. Densities of the samples from 90% to 95% were
1023
estimated using Archimedes principle. Cylindrical samples were cut into discs of 6.0 mm diameter and a thickness of 0.5 mm for thermal conductivity measurements, and rectangular pieces of dimension (8.0 3.0 3.0) mm for Seebeck and electrical resistivity measurements. All the transport properties were measured in parallel to the pressing direction. The powder X-Ray Diffraction (XRD) patterns of hot pressed (HP) samples were collected by a Bruker D8 Advance diffractometer using Cu-Ka radiation. Rietveld refinement was carried out for the crystallographic phase identification using the FullProf software [16]. The surface morphology of hot pressed samples was inspected by an Environmental Scanning Electron Microscope (Quanta 200, ESEM), and the elemental compositional analysis of hot pressed samples was done by a JEOL JXA-8530F Electron Probe Micro analyzer (EPMA) using Wave length Dispersive Spectrometry (WDS). Electrical resistivities and Seebeck coefficient were measured between 324 K and 775 K by an ULVAC-RIKO ZEM-3 system. Thermal conductivity was collected by an ANTER Flashline 3000 unit in the temperature range between 423 K and 773 K. The measurement errors for the electrical resistivity and the Seebeck coefficient are 5% and 10% for the thermal conductivity. The room temperature (RT) Hall Effect measurements were performed on homemade equipment based on the van der Pauw method. Arbitrarily shaped samples with a height of ~0.5 mm were used and a magnetic field of 0.65 T was applied during the measurements. 3. Results and discussions 3.1. Structural and phase characterization Rietveld powder XRD refinement of the undoped hot pressed sample is shown in Fig. 1(a). It is confirmed the main phase as Cu2SnSe3 with a monoclinic crystal structure of space group Cc as a major contribution (~81% volume fraction) and a minor contribution by cubic crystal structure of F43m (~18% of volume fraction obtained from Rietveld analysis). The low angular portion of the XRD pattern for the undoped sample is shown in Fig. 2 and additional reflections confirming the monoclinic crystal structure as reported earlier [17] are clearly observed. The XRD patterns of all the doped samples showed a cubic crystal structure of space group F43m, confirmed through the Rietveld analysis shown in Fig. 1(bed). A similar change from monoclinic to cubic crystal symmetry by doping of Cu2SnSe3 is already reported [5,10]. In all the samples a small amount of a secondary SnSe phase (0.55 Vol. % for x ¼ 0, 1.05 Vol. % for x ¼ 0.025, 0.88 for x ¼ 0.05 and 1.40 Vol. % for x ¼ 0.075 estimated from Rietveld refinement) is observed. An additional phase, CuSe2 (2.80 Vol. %), is observed in the sample with highest Zn-content, may be due to the excess of doping, which disrupts the formation of the stable compound. This leads to the formation of secondary phases, which depend on the compositional changes and the temperatures involved. Recently,J.-M. Song et al. [14] observed CuSe as secondary phase which may be due to the high temperatures used for processing the material. Therefore, this may be also the possible explanation for the presence of the secondary phase CuSe2. Cu2SnSe3 exists in many modifications such as cubic [18], monoclinic [19] and monoclinic with a superstructure [20]. Fig. 3 shows ESEM secondary electron images with X10000 magnification of the hot pressed samples. A continuous and homogeneous growth of large grains with less porosity indicates the compaction of the hot pressed samples Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075) (see Fig. 3). The surface morphology of all the samples is almost similar with doping and no impurity phases are observed in all the samples. Further, EPMA measurements are carried out for finding the phase purity and exact composition in all
1024
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
Fig. 1. (aed) Rietveld refinement for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
the hot pressed samples. Fig. 4 shows back-scattered electron micrographs of hot pressed samples and confirms a homogeneous single phase region in all the samples. The elemental composition of an undoped sample deviating from ideal stoichiometry may be due to the formation of point defects (vacancies, interstitials, antisites) created by deficiency of Cu and excess of Sn (see Table 1) as it was reported earlier [21]. Also, the doped samples showed a small deviation from its nominal composition, possibly due to the introduction of extrinsic point defects (substitutional defects) created through the doping by Zn on the Sn site. The presence of non-stoichiometry and the sample preparation through defects
Fig. 2. Low angular part of the XRD pattern of Cu2SnSe3.
may affect the carrier concentration, which will influence the transport properties of these materials. 3.2. Transport properties Fig. 5 shows temperature dependent electrical resistivity measured between 324 K and 775 K for all samples. Electrical resistivity of all the samples increased with an increase in temperature, indicating a heavily doped semiconductor behavior [5,9,10]. The electrical resistivity decreased with increase in doping content of Zn up to x ¼ 0.05 and increased slightly above x ¼ 0.05 at RT. As the Zn content increases, it creates more holes because of the substitution of þ2 valent Zn on the þ4 valent Sn site. This causes an increase in carrier concentration resulting in the reduction of electrical resistivity. This consistent trend shows that the presence of the impurity phase SnSe (<2 Vol. %) estimated from powder XRD analysis in all the samples (at least up to x ¼ 0.05) might not influence the resistivity. However, for the sample with x ¼ 0.075, the resistivity slightly increased compared to x ¼ 0.05 most likely due to the presence of a higher content of impurity phases SnSe (1.40 Vol. %) and CuSe2 (2.80 Vol. %). The electrical resistivity values at 324 K decreased almost by 35% for the samples x ¼ 0 to x ¼ 0.025 and 70% for the sample x ¼ 0 to x ¼ 0.05, 0.075. A similar trend of decrease in r through doping is also found for Cu2Sn1xInxSe3 (x ¼ 0,0.05, 0.075, 0.10) [5] and Cu2Sn1xMnxSe3 (x ¼ 0.005, 0.01, 0.02, 0.05, 0.1) [9] at 300 K. Electrical resistivity decreased from 7.93 mU-cm (for x ¼ 0) to 1.07 mU-cm (for x ¼ 0.1) in Cu2Sn1xInxSe3 and 80 mU-cm (for x ¼ 0) to 0.275 mU-cm (for x ¼ 0.1) in Cu2Sn1xMnxSe3, respectively [5,9]. Previously, authors found a similar trend of electrical resistivity behavior for the doped quaternary compound Cu2þxZnSn1xSe4 (0 x 0.15) [22]. The
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
1025
Fig. 3. (aed) secondary electron SEM images for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
electrical resistivity of the undoped sample (1.134 mU-cm at 324 K) is smaller than the RT values reported in the literature [5,9]. This deviation in resistivity between the present work and literature values may be due to the presence of non-stoichiometry in the compound (see Table 1) and may also be due to differences in the sample preparation. Electrical resistivity could not follow the trend
in an expected manner due to the changes in observed chemical composition as compared with nominal composition. Therefore, in order to get a rough estimation of the doping level in the substitution of Sn by Zn in Cu2Sn1xZnxSe3 as found in EPMA results (see Table 1) and explained earlier by the authors for the sample Cu2.1Zn0.9Sn1xInxSe4 (0 x 0.1) [23], it is assumed that Cu2SnSe3
Fig. 4. (aed) Back scattered electron micrographs of EPMA for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
1026
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
Table 1 Nominal composition, EPMA composition, r, S, l, zT, n and mH for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075). Nominal composition Cu2SnSe3 Cu2Zn0.025Sn0.975Se3 Cu2Zn0.05Sn0.95Se3 Cu2Zn0.075Sn0.925Se3 a b c d
EPMA, compositiona Cu1.97Sn1.03Se3.00 Cu2.00Zn0.03Sn0.97Se3.00 Cu2.04Zn0.09Sn0.94Se2.93 Cu1.98Zn0.07Sn0.94Se3.01
r423 K (mU cm)
S423 K (mV K1)
l423 K
(mW cm1 K1)
zT (423 K)
nb (1021 cm3)
nc (1021 cm3)
nd (1021 cm3)
mH
1.533 0.885 0.409 0.411
62.5 59.4 33.3 40.0
17.6 25.9 32.3 34.0
0.061 0.041 0.033 0.032
1.17 1.91 2.66 2.33
e 0.32 0.98 0.76
0.72 1.05 2.91 2.23
7.1 7.8 6.4 8.2
(cm2 V1s1)
EPMA data normalized to 6 atoms/unit cell. System internal pure element standards were used for calibration. n value estimated from S(T) fit data. n value calculated from simple valence counting. n value from the Hall measurement.
is a charge balanced compound and the substitution of Zn on Sn (Zn has two valence electrons less than Sn) leads to an increase of the p-type doping. The observed Zn contents from the EPMA measurements reveal the doping quantities of 0.03 for Cu2.00Zn0.03Sn0.97Se3.00, 0.09 for Cu2.04Zn0.09Sn0.94Se2.93 and 0.07 for Cu1.98Zn0.07Sn0.94Se3.01 per formula unit, respectively. This could explain the trend of electrical resistivity for the doped samples. Fig. 6 shows the Seebeck coefficient as a function of temperature measured in the range 324e775 K. All the samples show positive Seebeck coefficient indicating that the dominant carriers are holes. All the samples showed an almost linear temperature dependence of Seebeck coefficient indicating that the degenerate semiconductor behavior follows the same trend as reported for the doped Cu2SnSe3 [5,9,10]. The Seebeck coefficient decreases with increase of doping content up to x ¼ 0.05 and increases for x ¼ 0.075, which may be due to a variation in the charge carrier density (n). This can be roughly estimated by a linear slope of S (T) using Mott's formula for the diffusion thermo power of a free electron gas at temperatures well above the Debye temperature (qD):
Sd ðT > qD Þ ¼
p2 k2B 2m* 2=3 T eZ 3np2
(1)
where m* being the carrier mass, e the carrier charge, kB the Boltzmann constant, ħ the Planck's constant, T the temperature, and n the charge carrier density, respectively [24]. A similar trend of Seebeck coefficient for the doped samples in quaternary compounds was observed by the authors in their previous report [22]. For the estimation of n, S(T) data linearly fitted well up to 700 K for all the samples by considering m* ¼ me, as explained by authors in a recent report [23]. This results in n ¼ 1.17 1021 cm3 for Cu2SnSe3, n ¼ 1.91 1021 cm3 for Cu2Zn0.025Sn0.975Se3, n ¼ 2.66 1021 cm3 for Cu2Zn0.05Sn0.95Se3 and n ¼ 2.33 1021 cm3 for Cu2Zn0.075Sn0.925Se3. A simple valence electron counting is applied taking the unit cell dimensions obtained from XRD for the doped samples by assuming that, only doping amounts of Zn contribute to the charge carrier density (Cu2SnSe3 as charge balanced), which results in n ¼ 0.32 1021 cm3 for Cu2Zn0.025Sn0.975Se3, n ¼ 0.98 1021 cm3 for Cu2Zn0.05Sn0.95Se3 and n ¼ 0.76 1021 cm3 for Cu2Zn0.075Sn0.925Se3, respectively. It is pointed out that the charge carrier densities follow a similar trend for all doped samples with small variation in magnitude, although both methods give rough estimations. To find the exact charge carrier densities, Hall measurements were carried out for all the samples at RT. Carrier concentrations (n ¼ 1/(e RH)) were calculated using the room temperature Hall coefficient (RH) data, and are plotted together in Fig. 7 with the estimated carrier concentrations using both methods discussed above as a function of Zn content. The positive RH values for all the samples indicate p-type conductivity, which is in agreement with the Seebeck data. The general trend of measured carrier concentration with doping is similar,
with a small variation in the magnitude compared to the values found from simple estimation methods. It is observed that the n value for the undoped sample is 7.25 1020 cm3, which is at least one/two orders of magnitude larger compared to the reported data 0.66 1020 cm3 [5] and (0.04 1020 cm3) [9], respectively. This difference could be explained by the existence of nonstoichiometry in the compound observed by EPMA, because the transport properties are sensitive to the exact composition of the sample. Using the carrier concentration (n) and electrical resistivity (r), Hall mobility (mH ¼ 1/ner) was calculated for all the samples (see Table 1). Hall mobility of the undoped sample 7.1 cm2 V1 s1 is slightly smaller than the literature value of 11.8 cm2 V1 s1 [5]. This may be due to the presence of high carrier concentration compared to the literature [5], and it leads to the scattering of more carriers resulting in less mobility. Fig. 8 shows a Pisarenko plot, Seebeck coefficient (S) as function of carrier concentration (n) for Cu2ZnxSn1xSe3, including undoped Cu2InxSn1xSe3 at RT from the literature. There are several reports on various thermoelectric materials discussing the Pisarenko plot [25e27]. The effective mass (m*) is calculated using the RT values of S and n via Eq. (1). The solid line in Fig. 8 represents the calculated S values for m* ¼ 0.72 me as a function of n. It is observed that S values for all the samples in the present work as well as reported as a function of n are almost in agreement with the values predicted for a single parabolic band behavior of m* ¼ 0.72 me. The effective masses increase with increase in doping content and are in the range of 0.50 me to 0.95 me for Cu2ZnxSn1xSe3 (0 x 0.075) and a similar trend is observed with the doping in the literature [5] (calculated m* using the S, n values at RT from Eq. (1)). The power factor (S2/r) is evaluated from the electrical resistivity and Seebeck coefficient values in the entire temperature
Fig. 5. Temperature dependent electrical resistivity for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
1027
Fig. 6. Seebeck coefficient as a function of temperature for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
range. The maximum power factor 8 mW cm1-K2 at 775 K has been achieved for the sample Cu2Zn0.05Sn0.95Se3. This value is somewhat smaller in magnitude compared with the power factor ~12 mW cm1-K2 around 800 K for Cu2Sn0.9In0.1Se3, and 9.5 mW cm1-K2 at 670 K Cu2Sn0.99Mn0.01Se3, respectively [5,9]. The variation of the power factor is mainly due to the lower values of the Seebeck coefficient, which is affected by the high charge carrier density. The temperature dependent total thermal conductivity measured between 423 K and 773 K for all samples is plotted in Fig. 9. The total thermal conductivity of all the samples decreased with an increase in temperature, which is similar to the reports in the literature [5,9] explaining the dominance of phonon scattering at high temperatures. The l(T) for the undoped sample is almost close to the reported data. Thermal conductivity of all the doped samples is expected to decrease with increase in doping content due to mass fluctuation scattering, but it increased with increase of doping content due to the increase in carrier contribution which is related to the electrical resistivity. The smallest total thermal conductivity value 17.6 mW cm1-K1 is obtained for the undoped sample at 423 K, which compares well to the literature value (~19 mW cm1-K1 at 425 K) [5]. These values are moderate, and about an order of magnitude smaller compared to the binary
Fig. 7. Hole concentration as a function of the Zn-content for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075), Hall measurement, estimated from S(T) data and simple valence counting (for details see text).
Fig. 8. Pisarenko plot, Seebeck coefficient (S) as a function of charge carrier density (n) for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075) including reported values for Cu2InxSn1xSe3 (x ¼ 0, 0.05, 0.1) at room temperature.
compound ZnSe (190 mW cm1-K1 at 300 K) [6], which may be due to the substitution of multi-cations I and IV (Cu and Sn) on the II (Zn) site leading to a distortion in the crystal structure. Also, the lower values of thermal conductivity for all the samples may be due to the presence of porosity, which is one possible factor for reduction in thermal conductivity. The thermal conductivity of carrier contribution was calculated using the WiedemanneFranz relation lE ¼ LT/r, where L ¼ 2.45 108 WUK2, r is electrical resistivity and T is absolute temperature. The lattice part of thermal conductivity calculated by subtracting the carrier part from the total thermal conductivity is shown in Fig. 9. The trend in the temperature dependence of lattice thermal conductivity for all the samples is similar and the almost 1/T dependence indicates a dominant Umklapp phononephonon scattering at high temperature. Fig. 10 shows the thermoelectric figure of merit (ZT) as a function of temperature. ZT was evaluated from the power factor (S2/r) and total thermal conductivity (l). The figure of merit increases with increase in temperature for all the samples. Although the maximum power factor 8 mW cm1-K2 was found for the sample Cu2Zn0.05Sn0.95Se3 by doping, a significant drop in ZT is obtained
Fig. 9. Temperature dependent total (solid symbols) and Lattice (open symbols) thermal conductivity for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
1028
C. Raju et al. / Materials Chemistry and Physics 147 (2014) 1022e1028
to Cu2Zn0.05Sn0.95Se3. So the creation of more carriers by doping causes suppression in the Seebeck coefficient as well as a significant increase in the carrier contribution of the thermal conductivity, leading to an overall decrease in the thermoelectric figure of merit. Acknowledgments The authors would like to thank Prof. Arun M Umarji for providing the hot press facility and the Department of Science & Technology (DST), India for financial support through Grant No INT/ AUA/BMWF/P-13/2011 and the OEAD for support via the WTZ Austria-India under grant IN 09/2011. References
Fig. 10. Thermoelectric figure of merit (ZT) as function of temperature for Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075).
due to the rise of thermal conductivity by the large carrier contribution. The maximum thermoelectric figure of merit 0.48 at 773 K was found for the undoped sample due to its low thermal conductivity. The adverse effect on ZT for the doped samples is due to their low power factor and high thermal conductivity, affected by the high charge carrier concentration leading to lowering the Seebeck coefficient and increasing the carrier thermal conductivity. 4. Conclusions The ternary diamond like semiconductors Cu2ZnxSn1xSe3 (x ¼ 0, 0.025, 0.05, 0.075) were prepared by melting and subsequent annealing. The XRD pattern revealed a monoclinic crystal structure as a major phase for the undoped sample and a cubic crystal structure for the doped samples. The surface morphology and elemental composition of the hot pressed samples was studied by SEM and EPMA, respectively. The electrical resistivity decreased with increase in Zn content up to x ¼ 0.05 and increased for x ¼ 0.075 at RT. This is due to the change in carrier concentration through doping, confirmed by the Hall data at RT. The positive Seebeck and Hall coefficient of all the samples indicate that the majority of the carriers are holes. The Seebeck coefficient of all the samples followed a similar trend as electrical resistivity. The total and lattice thermal conductivity decreased with increase in temperature, due to the dominance of phonon scattering at elevated temperatures. The lowest total thermal conductivity was found for the undoped sample, whereas it increased with increase in doping. This may be due to the increase in the carrier contribution to l, with doping. Although the maximum power factor of 8 mW cm1-K2 was found for the sample Cu2Zn0.05Sn0.95Se3, the maximum thermoelectric figure of merit (ZT) 0.48 at 773 K was obtained for the sample Cu2SnSe3 due to its low thermal conductivity as compared
[1] G. Marcano, C. Rincon, L.M. de Chalbaud, D.B. Bracho, G.S. Perez, J. Appl. Phys. 90 (2001) 1847e1853. [2] G.S. Babu, Y.B.K. Kumar, Y.B.K. Reddy, V.S. Raja, Mater. Chem. Phys. 96 (2006) 442e446. [3] G. Hema Chandra, O. Lakshmana Kumar, R. Prasada Rao, S. Uthanna, J. Mater. Sci. 46 (2011) 6952e6959. [4] C.H.L. Goodman, J. Phys. Chem. Solids 6 (1958) 305e314. [5] X. Shi, L. Xi, J. Fan, W. Zhang, L. Chen, Chem. Mater. 22 (2010) 6029e6031. [6] G.A. Slack, Phys. Rev. B 6 (1972) 3791e3800. [7] G.A. Slack, New materials and performance limits for thermoelectric cooling, in: D.M. Rowe (Ed.), CRC Handbook of Thermoelectrics, CRC, Boca Raton, 1995, p. 407. [8] G.S. Nolas, X. Lin, J. Martin, M. Beekman, H. Wang, J. Electron. Mater. 38 (2009) 1052e1055. [9] X. Lu, D. Morelli, J. Electron. Mater. 41 (2012) 1554e1558. [10] J. Fan, H. Liu, X. Shi, S. Bai, X. Shi, L. Chen, Acta. Mater. 61 (2013) 4297e4304. [11] E. Skoug, J. Cain, D. Morelli, J. Electron. Mater. 41 (2012) 1232e1236. [12] J.Y. Cho, X. Shi, J.R. Salvador, G.P. Meisner, J. Yang, H. Wang, A.A. Wereszczak, X. Zhou, C. Uher, Phys. Rev. B 84 (2011) 085207. [13] M. Ibanez, D. Cadavid, U. Anselmi-Tamburini, R. Zamani, S. Gorsse, W. Li, A.M. Lopez, J.R. Morante, J. Arbiol, A. Cabot, J. Mater. Chem. A 1 (2013) 1421e1426. [14] J.-M. Song, Y. Liu, H.-L. Niu, C.-J. Mao, L.-J. Cheng, S.-Y. Zhang, Y.-H. Shen, J. Alloys.Compd. 581 (2013) 646e652. n ~ ez, R. Zamani, W. Li, D. Cadavid, S. Gorsse, N.A. Katcho, A. Shavel, [15] M. Iba pez, J.R. Morante, J. Arbiol, A. Cabot, Chem. Mater. 24 (2012) A.M. Lo 4615e4622. [16] T. Roisnel, J. Rodriguez-Carvajal, Mater. Sci. Forum 118 (2001) 378e381. n, Mater. Res. Bull. 38 (2003) [17] G.E. Delgado, A.J. Mora, G. Marcano, C. Rinco 1949e1955. [18] B.B. Sharma, R. Ayyar, H. Singh, Phys.Status. Solidi A 40 (1977) 691e696. n, G. S rez, Mater. Lett. 53 [19] G. Marcano, L.M. de Chalbaud, C. Rinco anchez Pe (2002) 151e154. [20] L.D. Gulay, M. Daszkiewicz, T.A. Ostapyuk, O.S. Klymovych, O.F. Zmiy, Acta. Crystallogr. C 66 (2010) i58ei60. [21] D.-H. Kuo, W.-D. Haung, Y.-S. Huang, J.-D. Wu, Y.-J. Lin, Thin Solid Films 518 (2010) 7218e7221. [22] C. Raju, M. Falmbigl, P. Rogl, X. Yan, E. Bauer, J. Horky, M. Zehetbauer, R.C. Mallik, AIP Adv. 3 (2013) 032106e032112. [23] R. Chetty, M. Falmbigl, P. Rogl, P. Heinrich, E. Royanian, E. Bauer, S. Suwas, R.C. Mallik, Phys. Status Solidi A 210 (2013) 2471e2478. [24] F.J. Blatt, Physics of Electronic Conduction in Solids, McGraw-Hill, New York, 1968. [25] Y. Pei, N.A. Heinz, A. LaLonde, G.J. Snyder, Energy Environ. Sci. 4 (2011) 3640e3645. €fer, G.J. Snyder, [26] W.G. Zeier, A. LaLonde, Z.M. Gibbs, C.P. Heinrich, M. Pantho W. Tremel, J. Am. Chem. Soc. 134 (2012) 7147e7154. [27] M. Falmbigl, N. Nasir, A. Grytsiv, P. Rogl, S. Seichter, A. Zavarsky, E. Royanian, E. Bauer, J. Phys. D Appl. Phys. 45 (2012) 215308.