Article Cite This: Inorg. Chem. XXXX, XXX, XXX−XXX
pubs.acs.org/IC
CsCu5S3: Promising Thermoelectric Material with Enhanced Phase Transition Temperature Ni Ma, Fei Jia, Lin Xiong, Ling Chen,* Yan-Yan Li, and Li-Ming Wu* Beijing Key Laboratory of Energy Conversion and Storage Materials, College of Chemistry, Beijing Normal University, Beijing 100875, People’s Republic of China
Downloaded via WESTERN SYDNEY UNIV on January 9, 2019 at 13:23:31 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
S Supporting Information *
ABSTRACT: Cu2S, featuring low cost, nontoxicity, and earth abundance, has been recently recognized as a high efficiency thermoelectric (TE) material. However, before reaching the maximum of the figure of merit (ZT), Cu2S undergoes three phase transformations starting at 370 K, which give rise to severe problems, such as possible decomposition and low reliability. Herein, we discover CsCu5S3 with phase transformation at 823 K, which is significantly higher than the 370 K value of Cu2S. Single crystal diffraction data reveal that its two phases are constructed by the same Cu4S4 columnar building unit via propagating either at the opposite sides into a layered o-CsCu5S3, or at the four apexes into a 3D t-CsCu5S3, respectively. Interestingly, the oto-t transformation is quick, but the reverse one is relatively slow. Theoretical studies reveal that the Cu4S4 column exhibits not only the most condensed atomic aggregation (Dcolumn) but also the lightest effective mass (m*), along which higher σ is realized. More interestingly, both phases exhibit remarkable ZT enhancements, 0.46 at 800 K for o-CsCu5S3, and 0.56 at 875 K for tCsCu5S3, which are 170% and 175% that of Cu2S at the same temperature.
■
independent Wyckoff sites;6,7 (2) between 370 and 700 K, β phase, P63/mmm (No. 194), hP16, in which 4 Cu atoms are disordered over 18 possible sites in the unit cell;8 and (3) above 700 K, α phase with Fm3̅m (No. 225), cF208, in which the 8 Cu ions are distributing over 204 possible sites in the unit cell. Such a strikingly high degree of disorder indicates a low threshold for the copper ion motion that gives rise to a liquidlike behavior, through which Cu ions strongly scatter phonons and reduce specific heat by suppression of transverse phonon modes. Such a phonon-liquid electron-crystal concept explains well the extraordinarily low κ and high σ of Cu2S.5 These phase transitions cause acute changes in not only structure but also density and carrier concentration and, therefore, give rise to significant questions regarding device reliability and long-term stability. As reported, the ZT peak value may not be repeatable on the same Cu2S sample.5 In particular, when Cu2S is subjected to high electronic currents and large temperature gradients, metallic elemental Cu is deposited on the surface of a Cu2S sample.9 From a chemistry viewpoint, we consider that the two following approaches are worth trying, one is to reduce the number of phase transition in the working temperature range; the other is to increase the phase transition temperature so as to enlarge the service temperature range. Herein, we report CsCu5S3 with an enhanced phase transition temperature at 823 K (significantly higher than the 370 K of Cu2S). Note that the nonradioactive cesium compounds are mildly toxic, although
INTRODUCTION The worldwide severe energy shortage with increasing energy demand is one of the current hottest topics, which inspires considerable interest in thermoelectric (TE) material, a promising green energy material that is able to convert heat into electricity directly by the Seebeck effect, or, vice versa, to generate a temperature gradient by Peltier effect.1−4 High performance TE materials thus provide a promising and attractive solution for the energy crisis; for example, a TE solidstate refrigerator provides cooling without moving parts or the release of greenhouse gases, or a TE power generator converts the low grade industrial waste heat or automobile exhaust heat into electricity. The TE performance is determined by the dimensionless figure of merit, ZT, with the definition ZT = S2σT/κ (S, Seebeck coefficient; σ, electrical conductivity; κ, thermal conductivity; T, absolute temperature; σS2, power factor). From the viewpoint of industrial applications, some considerations of a high performance TE material, such as low cost, nontoxicity, and environmental friendliness, are especially crucial. However, current state-of-the-art TE materials always contain expensive or toxic elements Pb, Te, Bi, and Sb, etc. Thus, cheap and low toxicity Cu2S has attracted much attention; with suitable Cu-deficiency, Cu1.97S can realize a very high ZT of 1.7 at 1000 K.5 Despite these, Cu2S faces severe problems that are mainly caused by its multiple phase transitions as follows: (1) γ phase below 370 K, with space group P21/c (No. 14) and Pearson symbol of mP144, a very complicated monoclinic structure in which the 96 Cu atoms in the unit cell are distributed over 24 crystallographically © XXXX American Chemical Society
Received: October 14, 2018
A
DOI: 10.1021/acs.inorgchem.8b02919 Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
room temperature to 923 K with a rate of ±20 K/min under the constant nitrogen atmosphere (Figure S2). The densified ingots of o-CsCu5S3 and t-CsCu5S3 were obtained by SPS at 723 and 773 K, respectively, with axial pressure of 55 MPa for 5 min. The density (∼95% of the theoretical value) was calculated with the measured dimensions and mass (Table S3). Electrical conductivity and Seebeck coefficient were measured simultaneously on a ZEM-3 (Ulvac Riko, Inc.) under the helium atmosphere. The rectangular specimens were cut from the abovementioned ingots parallel or perpendicular to the pressing direction (denoted as sample-para, or sample-perp, hereafter, Figure 4f, inset). A thin layer of boron nitride was coated on the pellet before measurement. The total thermal conductivity (κ) was calculated from the formula κ = D × Cp × d. The thermal diffusivity (D) was measured on a Netzsch LFA-457 instrument under a nitrogen atmosphere. Also, the specific heat (Cp) was indirectly measured with the pyroceram 9606 as reference. The uncertainty for the measurement is within 15%. Theoretical Calculations. The Vienna ab initio Package (VASP)14 based on density functional theory (DFT) methodology using the generalized gradient approximation of Perdew−Burke− Ernzerhof15 with projector augmented wave potentials was carried out to calculate the electronic structures. Energy cutoff was set as 500 eV, and Monkhorst−Pack k-point meshes were set as 5 × 13 × 5 or 4 × 4 × 6, for o-CsCu5S3 and t-CsCu5S3, respectively. The Cs 5s25p66s2, Cu 3d103p1, and S 3s23p4 were considered in the calculations (Figures S4−6). The lattice constants were optimized to be a = 9.6766 Å, b = 3.9300 Å, and c = 9.0962 Å for o-CsCu5S3, and a = b = 13.2732 Å, c = 7.8556 Å, for t-CsCu5S3, respectively. These agreed well with the single crystal refinement results.
excess cesium can lead to hypokalemia, arrhythmia, etc., but ordinarily such amounts would not be encountered in nature, and Cs2CO3 exhibits less toxicity than cesium. Interestingly, the high temperature phase (t-CsCu5S3) is relatively stable and slowly transforms. Single crystal diffraction data reveal that both o- and t-CsCu5S3 are constructed by a Cu4S4 column building unit with different extending patterns. Theoretical studies reveal that the Cu4S4 column has not only the most condensed atomic aggregation (Dcolumn), but also the lightest effective mass (m*), along which higher σ is realized. More interestingly, both phases exhibit remarkable ZT enhancements, 0.46 at 800 K for o-CsCu5S3, and 0.56 at 875 K for tCsCu5S3, which are 170% and 175% that of Cu2S at the same temperature.
■
EXPERIMENTAL SECTION
Syntheses. Bronger first synthesized CsCu5S3 in 2001 and reported that this compound has orthorhombic (o-CsCu5S3) and tetragonal (t-CsCu5S3) modifications.10 Here, we utilized totally new routes including neat solid-state reactions as well as an extremely fast spark plasma sintering (SPS) method to prepare them at large scales. The pure and polycrystalline CsCu4S3 was synthesized according to the literature (Figure S8).11 The pure o-CsCu5S3 was prepared by loading CsCu4S3 and Cu powder (Cu, Alfa, 99.999%, used as purchased) with a 1:1 mol ratio (1 g in total weight) into a silica tubing under vacuum and heated to 973 K in 3 h, maintained there for 3 h, and then cooled to 300 K in 13 h (Figure 1a). In the product, black single crystals with metallic luster are suitable for the single crystal diffraction measurements.
■
RESULTS AND DISCUSSION Large-Scale Syntheses and Thermal Stability. The syntheses of o-CsCu5S3 were first reported in 2001 by reactions of Cs2S, Cu, and S in a flux of cesium thiocyanate, and the single crystal structure was well-established.10 The solvothermal synthesis of o-CsCu5S3 was recently reported.13 However, these two methods use toxic cesium thiocyanate reactants,10 and the yield is extremely low; in addition, the remaining organic ligands are hard to remove.13 In comparison, our synthesis method is superior with regard to its low cost, nontoxicity, and high yield with a nearly stoichiometric reaction that is able to scale up to 10 g per reaction. Interestingly, the formation of the o- or t-CsCu5S3 phase is highly sensitive to the reaction temperature, pressure, as well as cooling rate. For example, XRD studies reveal that slow cooling (i.e., radiant cooling process) produces solely o-CsCu5S3, whereas extremely fast cooling (i.e., quenching in liquid N2) yields merely t-CsCu5S3 (Figure 1). Utilizing the SPS technique, t-CsCu5S3 is prepared at a 10 g per reaction scale as a pure phase with no other detectable impurity (Figure 1b and Table S1). The thermal stability of t-CsCu5S3 was checked by a batch of experiments, in which samples were annealed for 2 h at 773, 823, and 873 K, respectively, and then quenched into liquid nitrogen. As shown in Figure 2a, all these samples show no phase transformation. Further, another batch of samples was annealed at 873 K for 12 and 24 h, as shown in Figure 2b; merely, the t-CsCu5S3 phase was obtained. According to the TG/DSC curves of o-CsCu5S3 (Figure S2), an endothermic peak at 823 K indicates a phase transition, which is consistent with that reported earlier.10 However, on the cooling curve, no obvious exothermic peak is observed, which means the expected reversible t- to-o-CsCu5S3 transformation is not seen. Since the cooling of the DSC measurement is completed within 1 h, perhaps it is too short
Figure 1. Purity of (a) o-CsCu5S3 and (b) t-CsCu5S3 samples was confirmed after the SPS densifying (quick process within 5 min) and property measurements. Utilizing SPS-211LX (Fuji Electronic Industrial Co., Ltd.), pure phase t-CsCu5S3 was prepared by spark plasma sintering a 10 g sample of o-CsCu5S3 powder at 773 K under pressure of 55 MPa for 5 min (Figure 1b). To get single crystals, a 1 g t-CsCu5S3 sample was sealed into silica tubing, annealed at 973 K for 12 h and, then, quenched in liquid nitrogen (extremely fast quenching); many black shining single crystals were also found in the product. X-ray Diffraction. The room temperature (RT) PXRD pattern was taken on a Bruker D8 ADVANCE instrument, with Cu Kα radiation λ = 1.5406 Å in the 2θ ranges of 8° to 80° with step of 0.02°. The single crystals with dimensions of about 0.6 × 0.9 × 1.0 mm3 were used to collect the diffraction data with the aid of a Bruker APEX-II CCD instrument (Mo Kα λ = 0.71073 Å) at room temperature. The structures were solved and refined utilizing the SHELXTL program package.12 The Pmma (No. 51) and P4̅21c (No. 114) symmetries and R1/wR2 values of 6.55/13.80% and 4.80/ 11.08%, for o-CsCu5S3 and t-CsCu5S3, respectively, agree well with the previous reports.10,13 Crystallographic data are listed in Table S1. Property Measurements. UV−vis diffuse reflectance spectra were collected at room temperature with the aid of a SHIMADZU UV-2600 spectrometer in the range 850−250 nm. TG and DSC curves were measured on finely ground samples on a NETZSCH STA 449 F3 thermal analyzer in an Al2O3 crucible from B
DOI: 10.1021/acs.inorgchem.8b02919 Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
fully localized on the corresponding Wyckoff sites with 100% occupancy (Table S2). Both o- and t-CsCu5S3 structures are built from the same Cu4S4 columnar structure building unit, which is constructed by 3-fold coordinated Cu1 and Cu2 atoms with Cu1−S bonds of 2.26−2.40 or 2.32−2.41 Å, and Cu2−S, 2.35 Å or 2.29− 2.41 Å in o-, or t-CsCu5S3, respectively (Figure 3). Because the c parameter of t-CsCu5S3 (7.88 Å) is nearly 2 times the b value of o-CsCu5S3 (b = 3.96 Å), the Cu4S4 columns in two structures are thus propagating with similar periodicity along different directions, [001] in the former vs [010] in the latter. However, the structure differences are also obvious. In oCsCu5S3, the Cu4S4 column is extending at two opposite sides via the 2-fold coordinated Cu3 (Cu3−S = 2.21 Å) atoms into a wavy layer. Further, these layers are stacking along the [001] direction; in between, an array of Cs+ cations is accommodated (Figure 3a). The structural anisotropy is roughly represented by the Cu/S atomic aggregation density (D) as follows: Db (column) > Da (intralayer) > Dc (interlayer). Interestingly, such a decreasing trend agrees well with the increasing effective mass trend: my* (0.17) < mx* (5.47) < mz* (12.21). These suggest that the lightest m* along the b axis leads to the highest mobility where the D is the highest. In addition, along the z direction that is perpendicular to the layers, the mz* is as high as 12.21, significantly higher than my* and mx*, indicating a strong structural anisotropy that is consistent with the crystallographic structure feature. Dissimilarly, in the t-CsCu5S3 structure, the Cu4S4 column is extending simultaneously at four apexes via Cu3 (3-fold coordinated) and Cu4,5 (2-fold coordinated) (blue ball, Figure 3b) into a 3D network that defines large channels running parallel to the [001] (Figure 3b). The structural anisotropy roughly varies as follows: Dc (column) > Da (intralayer) = Db (intralayer). This is consistent with the increased effective mass trend of mz* (0.19) < mx* = my* (3.52). Note that mz* (0.19) is similar to my* (0.17) in o-CsCu5S3, because different extending directions should not vary noticeably the D value of the Cu4S4 column itself. However, the mx=y* (3.52) in the isotropic (001) layer in t-CsCu5S3 decreases significantly in comparison to that (mx* = 5.47) in o-CsCu5S3 owing to the separation of the Cs+ cations (Figure 3). In addition, the divergence of m* in t-CsCu5S3 is much less than that in oCsCu5S3, indicating a smaller degree of structural anisotropy. Such structural anisotropy is also reflected by the TE properties as discussed below.
Figure 2. Thermal stability of t-CsCu5S3 was checked by an XRD technique. (a) Samples were annealed at various temperatures for 2 h. (b) A batch of samples were annealed at 873 K for 24 or 12 h, respectively, and then quenched into liquid nitrogen. No phase transformation was detected. (c) A batch of samples were heated at 873 K and annealed for 12 h, and then slowly cooled to RT over 8, 10, or 12 h, respectively. The o-CsCu5S3 phase appears when cooling was over 10 h as indicated by dark red asterisks (*); when the cooling is over 12 h, the t-CsCu5S3 phase transforms completely into the oCsCu5S3 phase.
to allow the expected phase transformation, which also agrees with the previously report.10 In order to probe how slow such a transformation reaction is, some experiments were carried out: a batch of t-CsCu5S3 samples were annealed at 873 K for 12 h and then cooled to RT at different cooling rates (Figure 2c). The results reveal that, upon cooling over 8 h (i.e., cooling rate of 71 K/h), no phase change is observed, whereas when cooling over 10 h (57 K/h), o-CsCu5S3 appears as a minor phase (about 35%). If the cooling extends to 12 h (48 K/h), t-CsCu5S3 completely transforms into o-CsCu5S3. As a consequence, the t- to-oCsCu5S3 transformation requires a relatively long time. In the TE property measurements, for example, the ZEM-3 measurement, the cooling from 873 K to RT is usually finished within 2 h, which ensures the sample purity. Crystal Structure. Single crystal diffraction data reveal that o-CsCu5S3 crystallizes in Pmma (No. 51) with a = 9.6343(14) Å, b = 3.9590(7) Å, c = 8.9592(13) Å, and Z = 2, and tCsCu5S3 crystallizes in P4̅21c (No. 114) with a = b = 13.0550(17) Å, c = 7.8836(10) Å, Z = 8; these are in good agreement with literature values.10,13 Markedly different from cubic Cu2S with a high degree of disorder,5−7 in which 8 Cu atoms are distributed over 204 Wyckoff sites showing a liquidlike behavior, all Cu atoms in both structures of CsCu5S3 are
Figure 3. Single crystal structures of (a) o-CsCu5S3 and (b) t-CsCu5S3 that are constructed by a common building unit, the Cu4S4 column, which is extending at two opposite sides into a waved layer in part a, or simultaneously at four apexes into a 3D network in part b. Black, Cs; yellow, S; red, Cu1, Cu2; blue, Cu3 in part a and Cu3−5 in part b. C
DOI: 10.1021/acs.inorgchem.8b02919 Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
Figure 4. Thermal properties of o-CsCu5S3 and t-CsCu5S3 along the perpendicular and parallel directions of the SPS pressing, respectively. Temperature dependence of (a) electric conductivity, (b) Seebeck coefficient, (c) power factor, (d) total thermal conductivity, (e) lattice thermal conductivity, and (f) figure of merit ZT. Those of Cu2S are listed as reference.5
In good agreement with the reported value 1.41 eV,13 the experimental band gaps are assessed to be 1.59 and 1.68 eV, for o- and t-CsCu5S3 (Figure S1), respectively, which are larger than the 1.2 eV value of Cu2S5 owing to the involvement of the less electronegative Cs cations, whose lower lying states broaden the band gap. Similarly, CsAg5Te3 with Cs+ cations also shows an enlarged band gap of 0.67 eV19 that is wider than the 0.59 eV of Ag5Te3.21 Thermoelectric Properties. The σ is measured in the temperature range 323−873 K for t-CsCu5S3 and 323−810 K for o-CsCu5S3 considering its phase transition at 823 K. As shown in Figure 4a, the σ increases with temperature featuring an intrinsic semiconducting behavior and is significantly higher than that of Cu2S,5 about 3−15 times higher in the entire temperature range. For instance, t-CsCu5S3 reaches the highest σ of 42 S/cm at 773 K that is roughly 5 times higher than the 8 S/cm value of Cu2S. Both o- and t-CsCu5S3 exhibit anisotropic features in σ (Figure 4a). The σ values of t-CsCu5S3-perp are systematically larger than those of t-CsCu5S3-para. Such an anisotropy associates with the structural anisotropy. As shown in Figure S3b, the XRD pattern of t-CsCu5S3-perp differs significantly from those of t-CsCu5S3-para, with profound (hk0) diffraction peaks, i.e., (110), (320), (400), and (600), indicating that the perp-sample contains more polycrystalline particles that are aligned along the c direction. Consequently, t-CsCu5S3-perp exhibits a higher σ (Figure 4a). In comparison, the orientation preference of o-CsCu5S3 sample is weaker (Figure S3a); we consider this may be related to the different densifying temperature (o- vs tCsCu5S3: 723 vs 773 K). Nevertheless, the difference between perp- and para-slices of o-CsCu5S3 are still seen. The oCsCu5S3-para exhibits a stronger (001) diffraction peak indicating that more particles in such a pellet are c-orientated. Since the c direction is the stacking direction of layers in oCsCu5S3, which is poor in conduction, o-CsCu5S3-para shows poor σ (Figure 4a). On the other hand, o-CsCu5S3-perp shows profound (201) and (104) diffraction peaks (Figure S3a) indicating that more b-orientated particles are presented. The b value is the extending direction of the Cu4S4 columns that ensures a higher σ in o-CsCu5S3-perp (Figure 4a).
In summary, both o- and t-CsCu5S3 samples reveal that where the Cu4S4 column extends, the σ is higher, where the m* is the smallest. These two facts are consistent, because the lightest effective mass always associates with the highest mobility.22 The positive S value indicates the p-type feature with major carriers being holes. The S data monotonously increases with temperature which markedly differs from that of Cu2S (Figure 4b). The S falls in the range 120−273 μV/K, which is significantly smaller than that for Cu2S (300−450 μV/K)5 but is comparable with those of Sn1+xTe (50−150 μV/K),16 Cu2Se (60−300 μV/K),17 and GeTe (20−180 μV/K).18 As a result, CsCu5S3 exhibits moderate power factor (PF = S2σ) with the maximum value of 2.76 μW cm−1 K−2 at 823 K for t-CsCu5S3, about twice that of Cu2S (1.09 μW cm−1 K−2 at 823 K). Above 600 K, thePFs for both o- and t-CsCu5S3 are higher than that of Cu2S (Figure 4c). The total κ (κtot) values are very low (
