Xinke
Wang
a,
Igor
Veremchuk
a,
Ulrich
Burkhardt
a,
Matej
Bobnar
a,
Harald
Böttner
b,
Chang-Yang
Kuo
ac,
Chien-Te
Chen
c,
Chun-Fu
Chang
a,
Jing-Tai
Zhao
de and
Yuri
Grin
*a
aMax-Planck-Institut für Chemische Physik fester Stoffe, Dresden 01187, Germany. E-mail: grin@cpfs.mpg.de
bRetired from Fraunhofer-Institut für Physikalische Messtechnik, Freiburg 79110, Germany
cNational Synchrotron Radiation Research Center, 101 Hsin-Ann Road, Hsinchu 30076, Taiwan
dSchool of Materials Science and Engineering, Shanghai University, Shanghai 200444, China
eNanophotonics Center, Texas Tech University, Lubbock, Texas 79409, USA
First published on 6th August 2018
As one family of the most investigated thermoelectrics (TE), PbTe-based materials have been developed into state-of-the-art p-type and n-type TE materials. However, there are quite a few studies focusing on the reproducibility of TE properties and microstructure evolution during different heat treatments. In this work, Pb0.98−xNa0.02EuxTe (x = 0–0.030) samples were systematically examined after three different kinds of heat treatments: spark plasma sintering (SPS), laser flash measurement (LFA), and long-term annealing. The maximal solubility of Eu (ca. 1.0 atom%) in Pb0.98−xNa0.02EuxTe was established at 873 K. The most inhomogeneous samples (samples after SPS) show highest values of figure-of-merit, ZTmax, of up to 2.1 at 760 K, due to a large number of micrometer-scale sodium- and europium-rich aggregations in them. After additional heat treatment (LFA measurement or long-term annealing), the ZTmax value reduces to 1.6. The distribution of Eu and Na in the samples becomes much more homogeneous, accompanied by increased lattice parameters and decreased carrier concentrations. The long-term annealed samples have the best stable TE properties and good mechanical stability in the cyclic measurements. Surface protection needs to be considered for the temperatures above 773 K in order to avoid material decomposition.
Among the thermoelectrics suitable for applications, PbTe-based materials reveal outstanding performance. Sodium and iodine are well documented p- and n-type substituents used for enhancement of the thermoelectric properties of pristine PbTe, respectively.33–36 Significant progress has recently been made in obtaining high ZT in PbTe-based bulk materials using various substitutions.3,18,20,37–46 Lattice thermal conductivity (κL) of PbTe can be minimized by strong phonon scattering through defect formation and nanostructuring.18,41,47 The increase of degenerate valleys of transport bands by reduction of the energy offset between the light (at the L-point of the Brillouin zone) and the heavy (along the Σ-line of the Brillouin zone) sub-valence bands in PbTe leads to an effective band convergence within a few kBT, which improves the Seebeck coefficient without reducing electric conductivity.48–50 Recently reported high ZT p-type materials based on Na-substituted PbTe are summarized in Table 1. Among the 4f rare-earth elements, substitution by Eu has been proven to significantly influence the band structure of PbTe.51,52 Thin films of Pb1−xEuxTe containing quantum wells have been predicted theoretically and confirmed experimentally as a route towards enhanced thermoelectric properties at room temperature (RT).11,53–55 However, no significant influence of the Eu substitution on the thermoelectric figure-of-merit was observed in stoichiometric bulk materials,56 which is in contrast to the behavior of thin films reported in the literature.11,53–55
Reported systems | Solubility limit of substituents | Temperature range with ZT > 1, along with x values | ZT max, temperature (K) | σ | S | κ L | Ref. |
---|---|---|---|---|---|---|---|
a ↓: Decreased with respect to Na-substituted PbTe. ↑: Increased with respect to Na-substituted PbTe. ×: No effect or not mentioned. | |||||||
Pb0.98Na0.02Te–xSrTe (x ≤ 0.12) | x = 0.05 | 515–923 K | 2.5, 923 | ↓a | ↑a | ↓a | 42 |
x = 0.08 | |||||||
Pb0.98Na0.02Te–xMgTe (x ≤ 0.08) | x = 0.04 | 540–923 K | 2.0, 823 | ↓ | ↑ | ↓ | 20 |
x = 0.06 | |||||||
Pb1−xYbxTe:Na (x ≤ 0.10) | x > 0.10 | 550–850 K | 1.7, 850 | ↓ | ↑ | × | 41 |
x = 0.01 | |||||||
NaxPb0.97−xCd0.03Te (x ≤ 0.02) | No data on the Na solubility limit | 550–800 K | 1.7, 750 | ↓ | × | ↓ | 38 |
x = 0.012 | |||||||
PbTe−0.01Na2Te–xHgTe (x ≤ 0.03) | x < 0.02 | 520–800 K | 1.6, 770 | ↓ | ↑ | ↓ | 39 |
x = 0.02 | |||||||
Pb1−xMnxTe:Na (x ≤ 0.15) | x < 0.10 | 500–750 K | 1.6, 700 | ↓ | ↑ | ↓ | 44 |
x = 0.04 | |||||||
PbTe–0.01Na2Te–xCaTe (x ≤ 0.08) | No data on the CaTe solubility limit | 550–800 K | 1.5, 765 | ↑ | ↓ | ↓ | 37 |
x = 0.06 | |||||||
Pb1−x−yEuxNayTe (x ≤ 0.05, y ≤ 0.05) | x > 0.05; no data on the Na solubility limit | 500–850 K | 2.2, 850 | ↓ | ↑ | ↓ | 63 |
x = 0.03; y = 0.025 |
All lanthanide (Y, La–Lu) monotellurides adopt the NaCl-type of crystal structure.57,58 It can be expected that all of them may form the respective solid solutions with isostructural PbTe. Almost all of them are three-valent metals and will serve as donors for PbTe.46,58,59 However, in monotellurides, Eu and Sm may be divalent and therefore the electronic balance in a solid solution with PbTe should not change, as was proven experimentally.57,60,61 However, adding a monovalent element like Na to the system may change the electronic configuration of potentially mixed-valence f metal. By introducing Na into the Pb1−xEuxSe system, a part of Eu atoms may change the electron configuration to 4f6.62 Moreover, the study of Pb1−x−yEuxNayTe suggests that the solubility of Na in PbTe increases with increasing EuTe content. With nanometer-scale precipitates and high density of dislocations, a ZTmax of 2.2 at 850 K was reported.63
Nevertheless, the additive role of the Eu- and Na-substitution for ZT enhancement of the PbTe is still unclear, especially from the chemical point of view. The solubility of Na in PbTe was recently shown to be more complex than expected: Na forms two different local arrangements in the crystal structure of PbTe.36 Although there are several previous publications on p-type PbTe materials with high values of ZT (Table 1), and also Pb1−x−yEuxNayTe have been reported,63 there are only a few investigations about the stability (chemical and physical properties) of the studied materials.20,64–66 Moreover, different compositions reported in different publications reveal problems in the reproducibility of the results. That prompts us to carefully consider possible chemical issues for manufacturing quaternary PbTe-based TE materials.
Here we present a systematic investigation of structural and chemical features, carrier transport, and thermoelectric properties in the Eu–Na–PbTe system. We discuss the possibility of using these materials for the thermoelectric applications, based on thermal stability studies of variously heat-treated samples.
Phase identification was performed using the X-ray Guinier diffraction technique (Huber G670 camera, Cu Kα1 radiation, λ = 1.54056 Å, Δ2θ = 0.005°, 2θ range 3.0–100°, exposure time 6 × 15 min). The reflection positions, obtained by profile deconvolution, were corrected using an internal standard LaB6 (a = 4.15689(8) Å). Lattice parameter refinement was performed with the program package WinCSD.67
X-ray absorption spectroscopy (XAS) experiments were performed at the Dragon beamline of the National Synchrotron Radiation Research Center (NSRRC) in Taiwan. The Eu M4,5 spectra were recorded at 300 K using the total electron yield (TEY) mode with a photon energy resolution of ∼0.6 eV. Clean sample surfaces were obtained by cleaving the specimens in situ in a vacuum of 1 × 10−9mbar. By making weighted sums with the two reference samples EuO and Eu2O3 with the Eu2+ and Eu3+ ions in an octahedron, respectively,68 the relative amounts of Eu2+ and Eu3+ ions were extracted using the “NMimimize” function of the Mathematica software69 to obtain the best fit to the experimental spectrum of each sample.
For the metallographic study, the samples were embedded in conductive resin, and subsequently polished, finally using 0.1 μm diamond powder in a slurry. The material homogeneity was examined by optical microscopy (Zeiss Axioplan 2) in bright-field and polarized light. Elemental mappings were produced on an Energy-dispersive X-ray spectroscopy (EDX) system with a silicon-drift detector (SDD: XFlash 6|30; Bruker Quantax 400 system), attached to a Scanning Electron Microscopy (SEM JEOL 7800F) system. Areas of 500 × 500 μm2 were analyzed by stitching 3 × 4 images. A magnification of 200× was chosen to analyze the elemental distribution on a larger scale. All experiments were performed at an acceleration voltage of 10 kV with an exposure time of 60 min per image. Background-subtracted intensities of Pb Mα, Te Lα, Eu Lα, and Na Kα lines are used to visualize the concentration of the elements. Elemental chemical analysis of Pb, Te, Na, and Eu was performed using inductively-coupled plasma optical-emission spectrometry (ICP-OES, Agilent 5100 SVDV setup).
Electrical resistivity (ρ) and Seebeck coefficient (S) were measured simultaneously with the ZEM-3 setup (Ulvac-Riko) in the temperature range from 300 K to 760 K. Thermal diffusivity (D) was established with the Netzsch LFA 457 equipment. The Hall effect (RH) was measured with a standard four-point ac technique in a physical property measurement system (PPMS, Quantum Design), by sweeping magnetic fields up to 9 T. The Hall carrier concentration was calculated as 1/(RH·e), where RH is the Hall coefficient, and e is the charge of an electron. Magnetic susceptibility measurements were performed using a magnetometer system MPMS XL-7 (Quantum Design) in the temperature range 50–400 K and in magnetic fields up to 7 T. The effective magnetic moment per Eu atom was calculated using the Curie–Weiss fitting. Thermal diffusivity (D) was established with Netzsch LFA 457 equipment. The heat capacity per atom (Cp) was estimated from the relation Cp/kB = 3.07 + 0.00047(T − 300).70 Thermal conductivity was calculated as κ = dCpD, where d is the density obtained using the mass and geometric volume of the specimen disk. Lattice thermal conductivity (κL) was calculated by subtracting the electronic part κe = LσT from the total conductivity (the Wiedemann–Franz equation); the Lorenz number is evaluated as L = 1.5 + exp[−(|S|)/116], which is accurate within 20% for PbTe.71 The uncertainty of the Seebeck coefficient and electrical conductivity measurements is 5%, the uncertainty of the thermal conductivity is estimated to be within 8%. The combined uncertainty for the experimental determination of ZT is ∼20%.72
Fig. 2 (a) Location of the solid solution of Na and Eu in the phase diagram of the Pb–Eu–Na–Te system. (b) Lattice parameters of Pb0.98−xEuxNa0.02Te before SPS (black), after SPS (red) and LFA (blue), after 900 hours annealing at 873 K (green) along with pristine PbTe (orange square) as a reference.56 |
The concentration of 1.0 atom% Na was chosen, because the respective ternary sample revealed the highest charge carrier concentration.36 With increasing Eu content, the lattice parameter increases until x = 0.02 for all studied series of samples, which is unlike the results from ref. 63, where a monotonic increase of the lattice parameter is observed towards x = 0.05. The thermal history of our materials is different in ref. 63. At the same time, additional heat-treatment (SPS, LFA measurements, long-term annealing) increased the lattice parameter, regardless of the sample series (Fig. 2b). The lattice parameters of the samples after SPS and LFA increase linearly; for other series this trend is not so obvious (Fig. 2b). The similar lattice change was observed in the Pb1−xNaxTe series, which was explained by the redistribution of some sodium during the heat treatment.36 Introducing sodium into PbTe leads to either non-balanced Pb-by-Na substitution (rNa < rPb) or aggregation formation of defects in the Te sublattice. In both scenarios, this leads to the reduction of the lattice parameter of the majority phase. The minor amounts of the remaining Te are distributed in the product mixture, and due to the minimal amount cannot be detected easily. The thermal treatment leads to partial removal of sodium from the matrix (Table S1, ESI†) and equilibrium of the structure, which in turn increases the lattice parameter (cf.ref. 36). Introducing europium as the substituent of Pb increases the lattice parameter due to the size difference (rEu > rPb). The latter was already clearly shown in the ternary system Pb–Eu–Te. Based on these and early published data,36 it can be deduced that additional heat treatments increase the homogeneity of the quaternary materials. The analysis of the full width at half maximum (FWHM) of the XRD-reflections indirectly confirms this suggestion (Fig. 3). Only for the samples after long-term annealing, the FWHM values of the respective reflections are the smallest for all series. The inhomogeneity of the samples may originate in the heterogeneous distribution of the sodium and europium (Fig. S1 and S2, ESI†). The metallographic studies for the samples after SPS (believed to be most inhomogeneous) and after long-term annealing (believed to be most homogeneous) were performed. The sample after SPS is strongly inhomogeneous (Fig. 4a). There are a large number of micrometer-sized sodium- and europium-rich aggregations (Fig. 4b and c). The most homogeneous sample (the sample after annealing) looks differently (Fig. 5a). The convex areas are Na-reach zones (Fig. 5c, red circles), the elongated areas are Eu-rich (Fig. 5b, blue circles). Different distributions of Eu and Na, even after long-term annealing, may correspond to the immiscibility of the elements in the liquid state, which can be deduced from the results obtained from investigations of the rare-earth metals–lithium–gallium system.75
Fig. 3 Full width at half maximum (FWHM) of reflections in powder XRD patterns of Pb0.98−xEuxNa0.02Te: (a) before SPS, (b) after SPS, (c) after SPS and LFA, and (d) after annealing. |
In ternary Pb1−xEuxTe solution (x = 0.01 and 0.02) Eu remains in the 4f7 (Eu2+) state, consistent with reported data.60,76 This excludes the possible presence of europium(III) oxide. The effective magnetic moment of Eu is in good agreement with the theoretical value for the S = 7/2 Eu state. In the composition of the ternary samples which is located on the line, Pb1−xEuxTe1+0.5x, the effective magnetic moment is reduced (red point in Fig. S3, ESI†), suggesting the presence of Eu in the 4f6 state. The addition of sodium depending on the substitution scenario may force oxidation of europium towards Eu3+ in order to compensate the charge disbalance in the system. The effective magnetic moment of Eu in the quaternary samples is smaller (Fig. S3, ESI†). Since the magnetic susceptibility measurements cannot provide a precise evaluation of the 4f7/4f6 ratio, we performed XAS measurements for a series of samples after SPS and LFA, for which the lattice parameters obey Vegard's rule. Eu atoms were found to be in a mixed-valence state (Fig. 6). The amount of Eu in the 4f6 state increases with the Eu concentration until the solubility limit, the non-monotonicity of the Eu3+ ratio could originate from different deviations of the sample composition from the blue line in Fig. 2a towards the Te-rich region (Table S2, ESI†). For understanding this effect, we assume that adding Na moves the Fermi level towards the Eu 4f states, which leads to instability of the 4f7 configuration and drives the 4f7–4f6 transition.62
Fig. 7 Thermoelectric properties of Pb0.98−xEuxNa0.02Te after SPS: (a) resistivity, (b) Seebeck coefficient, (c) power factor, and (d) thermoelectric figure-of-merit ZT. |
Fig. 8 Thermoelectric properties of Pb0.98−xEuxNa0.02Te after SPS and LFA: (a) resistivity, (b) Seebeck coefficient, (c) power factor, and (d) thermoelectric figure-of-merit ZT. |
Fig. 9 Thermoelectric properties of Pb0.98−xEuxNa0.02Te after 900 h annealing at 873 K: (a) resistivity, (b) Seebeck coefficient, (c) power factor, and (d) thermoelectric figure-of-merit ZT. |
Fig. 10 (a) Hall carrier concentration p and (b) carrier mobility μ of Pb0.98−xEuxNa0.02Te after SPS (red), after SPS and LFA (blue), and after annealing at 873 K (green). Dashed lines are a guide to the eye. (c) Room temperature Seebeck coefficient S and (d) room temperature Hall mobility μ versus Hall carrier concentration p for Pb0.98−xEuxNa0.02Te, PbTe: Na,15,35,41 and PbTe: Tl.4,44,77 |
Highest values of the charge carrier concentration were observed for the series of samples after SPS. The carrier concentration practically does not change within the Eu solubility range (Fig. 10a, red symbols, Fig. S5, ESI†). Additional heat treatment reduces the carrier concentration drastically, which is due to redistribution of sodium during the heat treatment,36 consistent with the lattice parameter changes, and to the reduction of metal-rich aggregations into the microstructure (Fig. 4 and 5). The carrier concentration after long-term annealing (Fig. 10a, green symbols) is reduced by 40–50%, compared to the samples just only after SPS (Fig. 10a, red symbols). For samples after LFA and after annealing, the carrier concentration decreases slightly with the increasing Eu content, which may be attributed also to the increase of Eu3+ concentrations after additional heat-treatment. At the same time, samples after SPS are characterized as the most inhomogeneous, showing the lowest values of carrier mobility (Fig. 10b). Due to homogenization by heat treatment, the mobility increases for other series of samples. Comparing these values with PbTe: Na,15,35,41 Eu-substituted samples reveal higher Seebeck coefficients (Fig. 10c) and lower carrier mobilities (Fig. 10d). In the case of PbTe: Tl,4,44,77 the Seebeck coefficient is lower and the mobility is significantly higher (Fig. 10c and d). This indicates the effect of the Eu substitution on the band structure of pristine PbTe. As shown in Fig. 4 and 5, different thermal treatments have changed the distribution of Eu in samples, which will influence the electronic band structure of PbTe (as investigated in ref. 63). Therefore, a large dispersion of the Seebeck coefficient dependent on thermal treatments (Fig. 10d) was found. Thermal conductivity for all series decreases with increasing Eu concentration (Fig. 11a and b). The structural disorder, introduced by Eu substitution, and additional inhomogeneity (Fig. S4, ESI†) enhance the phonon scattering, which directly affects the lattice thermal conductivity (Fig. 11c and d). Interestingly, the samples just after SPS show higher values of thermal conductivity (with respect to the consecutive heat treatments), which is mainly due to higher contribution from electric thermal conductivity, since the samples after SPS have the highest carrier concentrations. The high lattice thermal conductivity of the samples after SPS is most probably caused by metal-rich aggregations remaining in the microstructure (Fig. 4). They are strongly reduced after enhanced thermal treatment (Fig. 5), and the thermal conductivity of the material has more intrinsic character, i.e. shows lower values. Similar to the Pb1−xNaxTe series,36 a jump-like decrease is observed in the vicinity of 650 K in the first heating cycle (Fig. 12). We used the first heating LFA data to calculate the total and lattice thermal conductivity for the samples after SPS. Since at high temperature (>650 K), the first heating LFA data are almost the same compared to the following cyclic LFA data, so the total thermal conductivity of after SPS samples and after LFA samples is almost the same (Fig. 11 and 12). However, the after-SPS samples have higher values of electric thermal conductivity at high temperature due to lower resistivity. According to the Wiedemann–Franz equation, the calculated lattice thermal conductivity show lower values in the high temperature range compared to after LFA samples (Fig. 11).
Fig. 12 Cyclic measurement of total thermal conductivity (κ) of Pb0.98−xEuxNa0.02Te: (a) x = 0.005, (b) x = 0.010, (c) x = 0.015, and (d) x = 0.020. |
The most inhomogeneous samples (after SPS) reveal the highest values of ZTmax up to 2.1 at 760 K. However, metallographic studies show that there are a large number of micrometer-scale sodium- and europium-rich aggregations. After additional heat treatment (LFA measurement or long-term annealing), the ZTmax value reduces to 1.6. The distribution of Eu and Na within the samples becomes much more homogeneous, while the lattice parameters increase and the carrier concentrations decrease.
The structural disorder introduced by Na and Eu substitutions and the non-homogeneity of materials increase the phonon scattering, which decrease the total and lattice thermal conductivities. The cyclic measurements of thermal conductivity show a significant difference between the first heating cycle and the subsequent cycles.
The cyclic TE properties were investigated for all three sample series. The long-term annealed samples show the best reproducible TE properties and good mechanical stability. In summary, these materials can be used in thermoelectric modules, just only after long-term annealing. In order to avoid material decomposition, surface protection need to be considered for working temperatures above 773 K.
This study clearly demonstrates that out-of-equilibrium materials may have promising high ZT values under certain circumstances, but these ZT values will generally approach lower values after heat treatment similar to the “working conditions”.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c8tc03142h |
This journal is © The Royal Society of Chemistry 2018 |