Layered Cobaltites and Natural Chalcogenides for Thermoelectrics

Abstract

We have systematically investigated thermoelectric properties by a series of doping in layered cobaltites Bi2Sr2Co2Oy, verifying the contribution of narrow band. In particular, Sommerfeld coefficient is dependent on charge carriers’ density and as function of density of states (DOS) at Fermi level, which is responsible for the persistent enhancement of large thermoelectric power. Especially for Bi2Sr1.9Ca0.1Co2Oy, it may provide an excellent platform to be a promising candidate of thermoelectric materials. On the other hand, high‐performance thermoelectric materials require elaborate doping and synthesis procedures, particularly the essential thermoelectric mechanism still remains extremely challenging to resolve. In this chapter, we show evidence that thermoelectricity can be directly generated by a natural chalcopyrite mineral Cu1+xFe1−xS2 from a deep‐sea hydrothermal vent, wherein the resistivity displays an excellent semiconducting character, while the large thermoelectric power and high power factor emerge in the low x region where the electron‐magnon scattering and large effective mass manifest, indicative of the strong coupling between doped carriers and localized antiferromagnetic spins, adding a new dimension to realizing the charge dynamics. The present findings advance our understanding of basic behaviors of exotic states and demonstrate that low‐cost thermoelectric energy generation and electron/hole carrier modulation in naturally abundant materials is feasible.

1. Introduction

Layered cobaltites with CdI₂‐type CoO₂ block provide an excellent platform for investigating thermoelectric properties. A key to unveil mysterious thermoelectric properties lies in the two‐dimensional (2D) conducting CoO₂ layer. For layered Bi‐A‐Co‐O (A = Ca, Sr, and Ba), it also contains analogous conducting CoO₂ layer [1]. In particular, layered Bi₂Sr₂Co₂O_y (BSC) exhibits a rather large thermoelectric power S (∼100 μV/K) at room temperature, which makes Bi₂Sr₂Co₂O_y one of promising thermoelectric materials from the viewpoint of potential applications, analogous to other misfit‐layered cobaltites, such as NaCo₂O₄ and Ca₃Co₄O₉ [2–5]. However, most studies of Bi₂Sr₂Co₂O_y system are mainly focused on the thermoelectric improvement [2, 3, 6]. The transport mechanism based on resistivity ρ and thermoelectric power S has not been clarified. Moreover, large S is totally different from conventional value (<10 μV/K) based on a broad band model [7]. In this chapter, we will show evidence on a narrow band contribution in doped Bi₂Sr₂Co₂O_y [8]. And what's more, exotic enhancement of large S is related to local density of states (DOS) near Fermi level (E_F) [9]. It could be effectively modulated thermoelectric performance by utilizing different doping. It is plausible to distinguish, which thermoelectric materials in doped Bi₂Sr₂Co₂O_y could be regarded as potential candidates.

On the other hand, ternary chalcogenides serve as an ideal platform for investigating intricate physical and chemical characteristics controlling the efficiency of thermoelectric materials, and also are promising materials for potential applications in photovoltaics, luminescence, as well as thermoelectric and spintronic devices [10–13]. Ternary chalcopyrite‐structured chalcogenides, such as CuFeS₂, have attracted particular attention owing to their unique optical, electrical, magnetic, and thermal properties [14–28]. Studies on chalcopyrite (CuFeS₂) have primarily focused on its electronic states [14, 15, 29–31]. However, the microscopic mechanism of electronic structure and thermoelectric character in CuFeS₂, which presumably arises from some scenarios such as delocalization of the Fe 3d electrons, charge‐transfer‐driven hybridization between Fe 3d and S 3p orbitals, or density of the conduction band electron states, still remains highly controversial [17, 30, 32]. The intrinsic mechanism of good thermoelectric properties is still a vital question which needs to be clarified. Another important issue is that the fabrication of artificial chalcopyrite itself requires expensively complex synthesis procedures and relatively high cost of constituent precursors, thereby potentially limiting the large‐scale applications in the thermoelectric field.

In this chapter, we confirm that an unexpected thermoelectricity can directly be generated in a natural chalcopyrite mineral Cu_1+xFe_1−xS₂ from a deep‐sea hydrothermal vent, and demonstrate that doped carriers have strong coupling with localized antiferromagnetic (AFM) spins, which greatly enhance the thermoelectric power S and power factor, revealing the significance of electron‐magnon scattering and large effective mass [33]. This will open up another useful avenue in manipulating low‐cost thermoelectricity or even electron/hole carriers via the natural energy materials abundantly deposited in the earth.

2. Thermoelectric properties and narrow band contribution of Bi₂Sr_1.9M_0.1Co₂O_y and Bi₂Sr₂Co_1.9X_0.1O_y

2.1. Crystal structure and valence states of Co ions

The crystal structure of Bi₂Sr₂Co₂O_y is shown in the inset in Figure 1, where conducting CoO₂ layer with triangular lattice and insulating rocksalt Bi₂Sr₂O₄ block layer are alternatively stacked along c‐axis, similar to the case of high‐temperature superconductors like Bi₂Sr₂CaCu₂O_y. Scanning electron microscopy (SEM) characterization of Bi₂Sr₂Co₂O_y indicates surface morphology of plate‐like grains. Figure 1 shows X‐ray diffraction (XRD) patterns of selected samples Bi₂Sr₂Co₂O_y, Bi₂Sr_1.9Ca_0.1Co₂O_y, and Bi₂Sr₂Co_1.9Mo_0.1O_y with single phase, in agreement with XRD result of Bi_1.4Pb_0.6Sr₂Co₂O_y [34]. The average Co valence was determined based on energy dispersive spectroscopy (EDS) measurement for all samples. For Bi₂Sr₂Co₂O_y, average Co valence is +3.330. For Bi₂Sr_1.9M_0.1Co₂O_y (M = Ag, Ca, and Y), average Co valence is +3.380, +3.330, and +3.280, respectively. For Bi₂Sr₂Co_1.9X_0.1O_y (X = Zr, Al, and Mo), average Co valence is +3.295, +3.347, and +3.189, respectively. X‐ray photoemission spectroscopy (XPS) spectra (see Figure 4a) also show the valence states of Co 2p_3/2 and 2p_1/2 for selected Bi₂Sr_1.9Ca_0.1Co₂O_y sample. Photon energy of Co 2p_3/2 and 2p_1/2 is 779.4 and 794.2 eV, respectively, demonstrating mixed Co valence between +3 and +4.

Figure 1.

Powder XRD patterns for Bi₂Sr₂Co₂O_y, Bi₂Sr_1.9Ca_0.1Co₂O_y, and Bi₂Sr₂Co_1.9Mo_0.1O_y samples at room temperature. Inset: crystal structure and SEM image of Bi₂Sr₂Co₂O_y.

2.2. Resistivity and transport mechanism

Figure 2a and d shows temperature dependence of resistivity ρ(T) of all samples. For parent Bi₂Sr₂Co₂O_y sample, an upturning point at T_p (∼75 K) is observed. Metallic behavior above T_p appears, demonstrating existence of itinerant charge carriers. Compared with Bi₂Sr₂Co₂O_y, ρ(T) of all doped samples (except Bi₂Sr_1.9Ca_0.1Co₂O_y) display total increase in view of the disorder effect. Furthermore, enhanced random Coulomb potential because of the doping induces the obvious shift of T_p toward higher temperature. On the other hand, ρ(T) of Bi₂Sr_1.9Ca_0.1Co₂O_y presents an overall decrease due to introduction of hole–type charge carriers into conducting CoO₂ layers.

Figure 2.

(a) Temperature dependence of resistivity ρ(T) and inset: magnification plot of ρ(T) for Bi₂Sr₂Co₂O_y (BSC) and Bi₂Sr_1.9M_0.1Co₂O_y (M = Ag, Ca, and Y) samples. (b) Plot of ln ρ against T⁻¹ for Bi₂Sr₂Co₂O_y and Bi₂Sr_1.9M_0.1Co₂O_y samples. Solid lines stand for TAC fitting. Dashed curves express VRH fitting. (c) Bi₂Sr₂Co₂O_y and Bi₂Sr_1.9M_0.1Co₂O_y dependence of activation energy $\Delta E$, onset temperature T_p of TAC, and onset temperature T_hopping of VRH. The shadow in bold is guide to the eyes. (d)–(f) are similar to (a)–(c) but for Bi₂Sr₂Co_1.9X_0.1O_y (X = Zr, Al, and Mo) samples.

To get insight into the conduction mechanism below T_p, dependences of ln ρ on T⁻¹ are plotted in Figure 2b and e. At the beginning, it is found that thermally activated conduction (TAC) law matches ρ(T) data well below T_p, namely [35], $\rho (T)={\rho }_0\text{exp}(\Delta E/k_{B}T)$, where $\Delta E$ is activation energy. Interestingly, ρ(T) apparently deviates from the TAC behavior with decreasing temperature further, and it follows Mott's variable‐range-hopping (VRH) model described by equation [35]: $\rho (T)={\rho }_0\text{exp}[{\left(T_0/T\right)}^n]$. As seen from Figure 2c and f, obtained values of $\Delta E$ and onset temperature T_hopping of Bi₂Sr_1.9Ca_0.1Co₂O_y (0.66 meV and 15.3 K) are the respective minimum, even smaller, than those of parent Bi₂Sr₂Co₂O_y (0.70 meV and 16.2 K), while $\Delta E$ and T_hopping of Bi₂Sr₂Co_1.9Zr_0.1O_y (2.75 meV and 63.6 K) are both maximum among all samples.

2.3. Thermoelectric power and narrow band model

Figure 3a and b shows temperature dependence of thermoelectric power S(T) for all samples. Positive values of S reflect electrical transport feature dominated by holes. Values of S at room temperature for all doped samples produce a substantial increase, especially for Bi₂Sr₂Co_1.9Mo_0.1O_y (∼117 μV/K), compared with pristine Bi₂Sr₂Co₂O_y (∼92 μV/K). Particularly, with decreasing the temperature until below T_hopping, S(T) behavior follows with VRH model [36]: S_VRH(T) ∼ aT^1/2, where a is factor determined by density of localized states at Fermi level N(E_F). The inset in Figure 3b reveals Anderson localization of Bi₂Sr₂Co_1.9Mo_0.1O_y, in correspondence with low‐temperature resistivity.

Figure 3.

Temperature dependence of thermoelectric power S(T) for Bi₂Sr₂Co₂O_y, (a) Bi₂Sr_1.9M_0.1Co₂O_y (M = Ag, Ca, and Y), and (b) Bi₂Sr₂Co_1.9X_0.1O_y (X = Zr, Al, and Mo) samples. Inset: calculated and fitted results of (a) Boltzmann formula and (b) VRH model for Bi₂Sr₂Co_1.9Mo_0.1O_y sample, respectively. Schematic diagram of density of states in a narrow band with Anderson localization at (c) high temperatures (metallic or TAC region) and (d) low temperatures (VRH region).

In general, S is extremely small (<10 μV/K) and presents a metallic behavior in a broad band [7]. Taking into account the huge difference, large S at high temperatures (above 200 K) in a narrow band matches Heikes model [37]: $S=k_B/e\{\text{ln}[d/(1-d)]\}$, where d is concentration of Co⁴⁺. The enhanced S at high temperatures is attributed to the competition between d and spin entropy. It is noted that S(T) is also described by narrow band model at intermediate temperatures. S(T) follows with Boltzmann formula [38]: $S(T)=1/eT\{{{\displaystyle \int }}^{}(E-E_{\text{F}})E^2\text{d}E/{[e^{\left(E-E_{\text{F}}\right)/2k_{\text{B}}T}+e^{-(E-E_{\text{F}})/2k_{\text{B}}T}]}^2\}/\{{{\displaystyle \int }}^{}{E}^2\text{d}E/{[e^{\left(E-E_{\text{F}}\right)/2k_{\text{B}}T}+e^{-(E-E_{\text{F}})/2k_{\text{B}}T}]}^2\}$. Calculated S(T) indicates monotonous increase with increasing T, as well as experimental result as plotted in the inset in Figure 3a, revealing the validity of narrow band model.

Actually, activation energy $\Delta E$ is equal to E_F−E_C, where E_C is the upper mobility edge. As k_BT/2 >$\Delta E$, conduction mainly determined by contribution of excited holes in itinerant states as specified in Figure 3c. At high temperatures, the majority of acceptor‐like states are fully ionized, that is, occurs complete excitation of holes, that resulting in metallic behavior of ρ(T)and diffused S(T) (Heikes formula). As k_BT/2 is near to $\Delta E$, TAC conduction forms (Boltzmann dispersion). As k_BT/2 <$\Delta E$, VRH conduction dominates the transport mechanism as shown in Figure 3d.

2.4. X‐ray photoemission spectroscopy and thermal conductivity

In order to further verify the narrow band model, we carried out XPS spectra for Bi₂Sr_1.9Ca_0.1Co₂O_y. As shown in Figure 4b, XPS spectra present an intense peak at ∼ 0.95 eV, in line with large S and metallic behavior. Between E_F and ∼2.0 eV, Co 3d and O 2p orbitals play an important role, similar to pristine Bi₂Sr₂Co₂O_y [39]. Moreover, strong hybridization between Co 3d and O 2p forms [39, 40]. Namely, antibonding t_2g narrow bands contribute to intense peak at ∼0.95 eV, while bonding e_g broad bands are responsible to peak within 3–8 eV. In addition, calculated S(T) is also consistent with experimental value based on magnitude and temperature dependence [39]. Therefore, the narrow band model is very suitable for explaining all experimental and theoretical results.

Figure 4.

(a) Co 2p XPS spectra and (b) XPS spectra in wide binding‐energy range for selected Bi₂Sr_1.9Ca_0.1Co₂O_y sample at room temperature.

Temperature dependence of total thermal conductivity κ(T) for all samples are shown in Figure 5a and d. κ(T) can be expressed by the sum of phononic component κ_ph(T) and mobile charge carriers’ component κ_e(T) as κ(T) = κ_ph(T) + κ_e(T). Value of κ_e(T) can be estimated from the Wiedemann‐Franz law, κ_e(T) = L₀T/ρ, where L₀ ∼ 2.44 × 10⁻⁸ V²/K² stands for Lorenz number. In Figure 5b and e, κ_ph(T) dominates the thermal conductivity because CoO₂ layer and Bi‐Sr‐O block layer induces the interface scattering. Dimension less figure of merit ZT = S²T/ρκ reflects total thermoelectric performance (see Figure 5c and f). For pristine Bi₂Sr₂Co₂O_y, ZT value reaches ∼ 0.007 at 300 K, while ZT value reaches 0.19 at 973 K, indicative of promising thermoelectric material for Bi₂Sr₂Co₂O_y at high temperatures [2]. Especially for Bi₂Sr_1.9Ca_0.1 Co₂O_y, ZT value reaches maximum ∼ 0.012 at 137 K. Therefore, it is reasonable to predict that Bi₂Sr_1.9Ca_0.1Co₂O_y could be considered as one of potential ultra‐high temperature thermoelectric materials, as well as pristine Bi₂Sr₂Co₂O_y.

Figure 5.

Temperature dependence of (a) total thermal conductivity κ(T), (b) phononic component κ_ph(T), and (c) dimensionless figure of merit ZT for BSC and Bi₂Sr_1.9M_0.1Co₂O_y (M = Ag, Ca, and Y) samples. (d)–(f) are similar to (a)–(c), but for Bi₂Sr₂Co_1.9X_0.1O_y (X = Zr, Al, and Mo) samples.

3. Exotic reinforcement of thermoelectric power in layered Bi₂Sr_2−xCa_xCo₂O_y

3.1. XRD patterns and electrical transport properties

The crystal structure of Bi₂Sr₂Co₂O_y is shown in Figure 6a. Figure 6b shows XRD patterns of all Ca-doping samples with single phase in Bi₂Sr_2−xCa_xCo₂O_y (0.0 ≤ x ≤ 2.0). With increasing Ca content, diffraction peak along [003] direction distinctly shifts to higher angle as shown in the inset in Figure 6b, confirming the smaller ionic radius of Ca²⁺, than that of Sr²⁺. SEM characterization indicates surface morphology of plate‐like grains and regular grain sizes for selected samples with x = 0.0 and 1.0, respectively.

Figure 6.

(a) Crystal structure of Bi₂Sr₂Co₂O_y. (b) Powder XRD patterns for Bi₂Sr_2‐xCa_xCo₂O_y (0.0 ≤ x ≤ 2.0) samples at room temperature. Insets: magnified powder's XRD patterns along [003] direction for all samples and SEM images for selected samples with x = 0.0 and 1.0, respectively.

Figure 7a and b shows resistivity ρ(T) of all samples in Bi₂Sr_2−xCa_xCo₂O_y. For the present x = 0.0 polycrystalline sample, upturning point at T_p (∼150 K) appears. Metallic behavior above T_p is observed, demonstrating the existence of itinerant charge carriers. In comparison, for x = 0.0 single crystal [41], in‐plane resistivity ρ_ab also shows metallic behavior around room temperature, while it arises minimum near 80 K and diverges with further decreasing the temperature. Resistivity ρ_ab value of single crystal for x = 0.0 at room temperature is ∼4 mOhm×cm and is smaller than that of our polycrystalline sample (∼15 mOhm×cm). On the other hand, compared with x = 0.0, ρ(T) of all Ca‐doped samples produce total increase due to disorder effect. For the samples with x ≤ 0.5, enhanced random Coulomb potential because of Ca doping induces the shift of T_p toward higher temperature. Interestingly, for the samples with x ≥ 1.0, the signature of transition at T_p completely vanishes and ρ(T) only presents an insulating‐like behavior.

Figure 7.

(a) and (b) Temperature dependence of resistivity ρ(T). Insets: magnification plot of ρ(T) for Bi₂Sr_2‐xCa_xCo₂O_y samples. (c) and (d) Plot of lnρ against 1/T. Solid lines present TAC fitting. Dashed curves stand for VRH fitting. (e) Ca concentration x dependence of activation energy $\Delta E$, onset temperature T_p of TAC, and onset temperature T_hopping of VRH. (f) Ca concentration x dependence of resistivityρ_{300 K} at room temperature.

To discern conduction mechanism below T_p, relationship of lnρ against 1/T is plotted in Figure 7c and d. As for x ≤ 0.5, at the beginning, it is found that TAC law matches ρ(T) data well below T_p, namely [35], $\rho (T)={\rho }_0\text{exp}(\Delta E/k_{\text{B}}T)$, where $\Delta E$ is activation energy. But ρ(T) apparently deviates from TAC behavior with decreasing the temperature further, and it follows Mott's VRH model described by equation [32]: $\rho (T)={\rho }_0\text{exp}[{\left(T_0/T\right)}^n]$. However, as for x ≥ 1.0, ρ(T) meets VRH model only, in agreement with the insulating feature of x = 2.0 single crystal [1, 42, 43]. Obtained values of $\Delta E$ and onset temperature T_hopping are plotted in Figure 7e. Basically, $\Delta E$ increases with Ca content, as well as T_p for x ≤ 0.5. In comparison, the present value of $\Delta E$ based on sintering temperature 800°C is larger than the previous one of x = 0.0 at 900°C [8], revealing the difference of grain size effect. It is worth noting that values of T_hopping and ρ_300K at room temperature first increase and then decrease in whole Ca‐doped range (see Figure 7e and f).

3.2. Enhancement of thermoelectric power driven by Ca doping

Figure 8a shows thermoelectric power S(T) for all samples. Positive values of S demonstrate that majority of charge carriers are hole type. In addition, S exhibits a nearly T‐independent behavior above 200 K, while S strongly depends on T peculiarly below 150 K. Ca doping obviously boosts S_300K at room temperature especially for heavy Ca contents (see Figure 8b). Large S_300K value monotonously increases from 105 μV/K(x = 0.0) to 157 μV/K (x = 2.0). In general, the change of S should be related to variation of n. For x = 0 single crystal [38], Hall coefficient (R_H) is positive and strongly dependent on the temperature in the range from 300 to 0 K. Increase of R_H toward the lowest temperature is not simple due to the decrease of n, but rather due to anomalous Hall effect. It is noted that variation of R_H with Pb doping is also similar to that of ρ_ab. Pb doping slightly reduces the magnitude of R_H, but the increase in number of charge carriers is much smaller than expected from chemical composition [41, 44].

Figure 8.

(a) Temperature dependence of thermoelectric power S(T) for Bi₂Sr_2‐xCa_xCo₂O_y samples. (b) Ca concentration x dependence of S and charge carriers’ density n at room temperature, respectively. (c) Ca concentration x dependence of Co⁴⁺ ion (deduced from charge carriers’ density n) and corresponding S_Heikes (originating from Heikes formula) at room temperature, respectively. (d) Ca concentration x dependence of electronic coefficient γ deriving from specific heat C(T). Inset: temperature dependence of C(T) plotted as C/T versus T² based on fitting lines for x = 0.0, 0.5, 1.5, and 2.0, respectively.

As we know, S is rather low (<10 μV/K) with a metallic behavior in a broad band [7]. Taking into account the tremendous discrepancy, large S of Bi₂Sr_2−xCa_xCo₂O_y with a nearly T-independence at high temperatures in a narrow band should follow the so‐called Heikes formula [37]: $S=k_B/e\mathrm{\{}\text{ln}[(g_3/g_4)d/(1-d)]\mathrm{\}}$, where d is concentration of Co⁴⁺, and $g_3$ and $g_4$ are spin orbital degeneracies for Co³⁺ and Co⁴⁺ ions, respectively. Concentration d at room temperature can be deduced from charge carriers’ density n. As visible in Figure 8c, as for x< 1.5, d decreases, while S_Heikes (deriving from Heikes formula) increases, which is consistent with the trend of S_300K. But for x ≥ 1.5, reduced S_Heikes is reverse to persistent enhancement of S_300K. Thus, we have to consider other possible reason of enhanced S for heavily doped samples.

3.3. Specific heat and Sommerfeld coefficient

Next we will check whether the enhanced S originates from the increased effective masses through electronic correlation. To test this point, we performed measurement of specific heat C(T), which is plotted as C/T versus T² (see the inset in Figure 8d) for selected samples with x = 0.0, 0.5, 1.5, and 2.0. C(T) at low temperatures can be described as C(T) = γT + βT³ [45], where γT and βT³ denote electronic and lattice contribution to C(T), respectively. We can get electronic coefficient γ by the linear fitting according to C/T = γ + βT² [45]. Here, we need to explicitly interpret Sommerfeld coefficient γ. For the present system, unit formula should involve two cobalt atoms. For our polycrystalline sample with x = 0.0, a conventional way to get γ by extrapolating high‐temperature linear part of C/T versus T = 0 gives very large value of ∼ 135 mJ mol⁻¹ K⁻² (see Figure 8d), comparable with that of x = 0.0 single crystal (∼140 mJ mol⁻¹ K⁻²) [41]. However, it is observed that γ rapidly decreases with increasing Ca doping. For our sample with x = 2.0, value of γ is ∼85 mJ mol⁻¹ K⁻². Differently, it is noted that value of γ is only 50 mJ mol⁻¹ K⁻² for Bi‐Ca‐Co‐O system, while such a unit formula merely includes one cobalt atom [45].

Now we discuss the underlying implications of enhanced S with Ca doping. As mentioned above, as for x < 1.5, decreased d based on Heikes formula should be responsible for the enhanced S. But for x ≥ 1.5, local modification of DOS and band structure near E_F could play crucial role. S(T) can be defined by Mott formula [39]: $S(T)=({\pi }^2{k}_{\text{B}}T)/\left(3e\right){[\text{dln}\sigma (E)/\text{d}E]}_{E=E_{\text{F}}}$, where σ(E) is electrical conductivity with σ(E) = n(E)eυ(E), υ(E) is mobility, n(E) is charge carriers’ density with n(E) = D(E)f(E), D(E) is DOS, and f(E) is Fermi function. Apparently, in terms of Mott formula, the enhancement of S for x ≥ 1.5 should be attributed to the increase of local DOS near E_F. In details, with decreasing A‐site ionic radius (i.e., with increasing Ca content), tolerance factor decreases (not shown here), which leads to changes of lattice distortion in CoO₂ layer and local band structure near E_F, reminiscent of layered perovskite cobaltite SrLnCoO₄ (Ln stands for different rare earth elements) [46]. Ultimately, value of S for x ≥ 1.5 would be enhanced. Based on all of above results, one should emphasize that Sommerfeld coefficient γ is dependent on n, and also as function of DOS at E_F, which leads to continuous enhancement of large S.

4. Thermoelectricity generation and electron‐magnon scattering in a natural chalcopyrite mineral

4.1. Crystal structure and SEM characterization

A series of natural chalcopyrite minerals, Cu_1+xFe_1−xS₂ (x = 0.17, 0.08, and 0.02), were obtained from a hydrothermal vent site named Snow Chimney in the Mariner field of Lau Basin [47]. Basically, mineral composition obtained from intact natural sulfide chimneys has no variation. Subsamples with x = 0.02 and 0.08 were obtained from the most interior chimney part, whereas subsample with x = 0.17 was obtained from the middle chimney wall region. The highly fluctuated and variable physicochemical conditions lead to obvious differences in mineral composition [48]. Figure 9 shows sketches of its crystal structure and atomic planes, in which chalcopyrite crystallizes in a tetragonal lattice with space group of I‐42d and produces honeycomb structure characteristic [49]. Each Fe and Cu atom is encircled by tetrahedron of S atom. The highlighted planes indicateatomic zig‐zag pattern, which is likely responsible to phonon scattering. XRD Rietveld refinement of power pattern indicates that three natural samples are single phase with standard chalcopyrite structure. For x = 0.08, refined lattice parameters a and c are 5.278 and 10.402 Å, respectively (see Figure 10).

Figure 9.

Crystal structure of Cu_1+xFe_1‐xS₂. Ball‐and‐stick model of the crystal structure (left) viewed along a‐axis with black lines indicating unit cell. Stick model (right) showing characteristic honeycomb structure of chalcopyrite. Identical atomic arrangement is highlighted in gray in both structures, but projection is along different axes.

Figure 10.

Powder XRD patterns with Rietveld refinement for natural sample of Cu_1+xFe_1‐xS₂ (x = 0.08). Red line indicates experimentally observed data, and black line overlapping them refers to calculated data. Vertical tick is related to the Bragg angles positions in space group I‐42d. The lowest profile shows the difference between observed and calculated patterns. Rietveld refinement indicates that it is standard chalcopyrite structure.

To probe the microstructures of natural Cu_1+xFe_1−xS₂, we performed SEM characterization (Figure 11). SEM analysis revealed that natural chalcopyrite with x = 0.08 had layered structure. Three examined natural samples were found to contain morphological diversity, which is characteristic of chalcopyrite minerals, and suggest different physical and chemical behaviors of various microstructures. The SEM observation may provide important insights of the relevance between physical and chemical functions and behaviors of chalcopyrite minerals.

Figure 11.

Surface morphology of natural sample of Cu_1+xFe_1‐xS₂ (x = 0.08) showing characteristic layered structure. (a) Areas showing cracked layered structure in natural sample Cu_1+xFe_1−xS₂ (x = 0.08), scale bar: 10 μm. (b) Densely layered structure, scale bar: 5 μm. (c) Triangular pattern surrounded by layered structure, scale bar: 1 μm.

4.2. Thermoelectricity generation and electronic states

To examine the functional properties of natural Cu_1+xFe_1−xS₂ samples, we first measured resistivity (ρ) as function of temperature (T). Three examined natural samples exhibited excellent conductive behavior with semiconductive characteristics (Figure 12a). With the reduction of x, the overall resistivity decreased due to the emergence of doped charge carriers. Value of ρ_300K for x = 0.17, 0.08, and 0.02 was 4.97, 0.11, and 1.01 Ohm×cm, respectively. Compared with x = 0.08, the increase of resistivity for x = 0.02 stems from the enhanced random Coulomb potential owing to the natural doping.

Figure 12.

Formation of thermoelectricity by Cu_1+xFe_1‐xS₂. (a) Temperature dependence of resistivity ρ in three natural samples of Cu_1+xFe_1‐xS₂. (b) Temperature dependence of thermoelectric power S for three samples.

In order to track the evolution of electronic states, we carried out thermoelectric power (S) measurement (Figure 12b), where the sign of S changes. For x = 0.17, the sign of S switches from negative to positive around 235 K with decreasing temperature (Figure 12b). It is amazing to observe two unusual peaks: a broad peak (T_m; 32 μV/K, 186 K) and a sharper peak (T_p; 215 μV/K, 11 K), indicating the majority of hole carriers (p-type). It is of particular interest that, for x = 0.08 and 0.02, T_p peak utterly disappears, while T_m peak becomes wider and rapidly shifts to a lower temperature, where S presents very large negative values, demonstrating the majority of electron carriers (n-type), in line with negative Hall coefficient R_H (Figure 13). Large S_300K reached a remarkable value of −713 and −457 μV/K for x = 0.08 and 0.02, respectively. Namely, more electrons are activated at room temperature with increasing Fe concentration. For x = 0.08, charge carriers’ mobility μ_300K and density n_300K are 1.8 cm² V⁻¹ s⁻¹ and 3.5 × 10¹⁹ cm⁻³, obtained from R_H = 1/ne and μ= 1/neρ. In addition, Fe magnetic moment may also play an key role to induce large S, indicative of strong coupling between magnetic ions and doped charge carriers because synthetic CuFeS₂ presents AFM ordering at 823 K [15].

Figure 13.

Hall effect of natural sample of Cu_1+xFe_1‐xS₂ (x = 0.08). (a) Temperature dependence of Hall coefficient R_H. (b) Temperature dependence of charge carriers’ density n. Value of R_H (cm³ C⁻¹) is determined by n (cm⁻³) and electron charge e, that is, R_H = 1/ne, where e = 1.602176 × 10⁻¹⁹ C. The shadow in bold is guide to the eyes.

4.3. Electron‐magnon scattering and large effective mass

The matter of imperative concern is how to understand the origin of T_m peak and conduction mechanism. According to Mott's formula, S can be qualitatively expressed as $S=-{\pi }^2{k}_{\text{B}}^{2}T/3e[\text{σ’}(E_{\text{F}})/\sigma (E_{\text{F}})]$, where k_B is Boltzmann constant, $\sigma (E_{\text{F}})$ is electrical conductivity at Fermi level E_F, and $\sigma \text{’}$ denotes d[$\sigma (E)$]/dE [35]. If one assumes $\sigma \text{’}$ is a constant accompanied by isotropic electrical transport properties, namely, $\sigma$⁻¹= ρ, then ΔS/S₀ Δρ/ρ₀ can be derived. Plot of ΔS/S₀ versus Δρ/ρ₀ for x = 0.17 (Figure 14) shows that all experimental data near T_m at T₀ from 155 to 300 K deviate from the theoretical calculation, the linearity. These results verify that exotic mechanism of S(T) in natural sample is beyond the framework of conventional thermoelectric picture [50].

Figure 14.

Correlation between thermoelectric power S(T) and resistivity ρ(T). Relative changes of ΔS/S₀ versus Δρ/ρ₀ in natural sample with x = 0.17 at various temperatures (T₀ = 155, 185, 200, 215, 230, 240, 250, 270, and 300 K). The present experimental data substantially deviates from the linear relationship predicted by Mott's formula, which is indicated by dotted line.

To better discern intrinsic transport mechanism of Cu_1+xFe_1−xS₂, we incorporate spin-wave theory to analyze temperature dependence of S. For x = 0.08 and 0.02, field‐cooling magnetization and loop hysteresis indicate the localized ferromagnetism (FM) at low temperatures because of additional Fe moments (Figure 15). However, strong AFM interaction at high temperatures dominates for three natural samples. Generally speaking, spin waves can scatter electrons for AFM or FM materials, resulting in magnon‐drag effect [12]. To check this issue, we developed magnon‐drag model, S=S₀+S_3/2T^3/2+S₄T⁴, where S₀ is value of S at T = 0, S_3/2T^3/2 term stems from electron‐magnon scattering, and S₄T⁴ term is related to spin‐wave fluctuation in AFM phase. Using this model of magnon drag, the predicted values for three samples closely matched S(T) data (Figure 16a and b). As the absolute value of S_3/2 is nearly six orders of magnitude larger than that of S₄ (Table 1), electron‐magnon scattering dominates S(T) curve. Thus, T_m peak is predicted to originate from magnon drag due to the strong electron‐magnon interaction.

Figure 15.

Magnetic properties of natural Cu_1+xFe_1−xS₂. (a, b) Temperature dependence of field‐cooling (FC) magnetization, M, in three natural samples of Cu_1+xFe_1−xS₂, measured in applied magnetic field of H = 0.1 T (a) and H = 1 T (b). (c) Magnetic field dependence of magnetization, M, for three samples, measured at 40 K.

Figure 16.

Temperature dependence of S for Cu_1+xFe_1−xS₂ samples with x = 0.17 (a) and x = 0.08 and 0.02 (b). Symbols represent experimental data and solid lines correspond to theoretical simulation based on the model of magnon drag, S = S₀ + S_3/2T^3/2 + S₄T⁴. (c) Obtained parameters S_3/2 and ΔE are plotted as function of Fe content, where S_3/2 represents the electron‐magnon scattering process and ΔE is activation energy. (d) Temperature dependence of power factor, S²/ρ, for three samples.

Parameter	Tm(K)	S0(μVK^−1)	S3/2(μVK^−5/2)	S4 (μVK^−5)	ΔE(meV)
x = 0.17	186	−6.21	0.03	−3.84×10^−8	60.1
x = 0.08	68	−75.45	−0.08	−5.47×10^−8	4.9
x = 0.02	38	−10.61	−0.04	−3.95×10^−8	11.8

Table 1.

Obtained parameters based on theoretical simulation.

The parameter T_m represents the peak of magnon drag, which stems from the experimental S(T) curve. The parameters S₀, S_3/2, and S₄ stem from the model of magnon drag, S= S₀ + S_3/2T^3/2 + S₄T⁴. The parameter ΔE is the activation energy, which stems from the TAC model, ρ(T) = ρ₀ exp (ΔE/k_BT).

To gain more insight into the correlation between magnon drag, doped carriers, and S, we plotted parameters S₀, S_3/2, and S₄ as a function of x (Table 1). S₀, S_3/2, and S₄ for x = 0.08 has largest absolute values among three natural samples, in agreement with the largest S, smallest ρ, and highest power factor. Unlike S₀ and S₄, dependence of S_3/2 is quite unique (Figure 16c). The sign of S_3/2 varies from positive to negative with increasing Fe concentration, suggesting the alternation of p‐type and n‐type charge carriers and orbital degree of freedom of Fe 3d band with AFM ordering. Additionally, electron‐magnon scattering occupies thermoelectric properties, indicating strong coupling between doped charge carriers and AFM spins. Furthermore, ρ(T) follows TAC model $\rho (T)={\rho }_0\text{exp}(\Delta E/k_{B}T)$, where $\Delta E$ is activation energy [35]. Notably, the fitted energy gap of ΔE (60.1, 4.9, and 11.8 meV for x = 0.17, 0.08, and 0.02, respectively), which verifies the existence of localized Fe spins, is markedly smaller than that of artificial chalcopyrite [21, 29–31]. It is noted that experimental S(T) result is well described by electron‐magnon scattering up to ∼200 K, while it deviates from theoretical lines for higher temperatures. In particular, power factor S²/ρ shows an abrupt enhancement above 200 K for x = 0.08 (Figure 16d), in agreement with that of R_H and n (Figure 13). Above 200 K, large effective mass (m^*) leads to high power factor and large S due to low μ and high n. For x = 0.08, it exhibits the largest m^* value (1.6 m₀) at room temperature, where m₀ is free electron mass. Therefore, we can conclude that robust electron‐magnon scattering and large m^* induce unexpected thermoelectricity generation in natural chalcopyrite mineral.

In terms of thermal conductivity κ, phononic component κ_ph dominates for three natural samples owing to negligible electronic component κ_e (Figure 17). For the optimal sample with x = 0.08, value of ZT can reach 0.03 at room temperature (Figure 17), thus indicating that natural chalcopyrite semiconductor is a promising candidate for thermoelectric energy materials. It is quite striking that the spontaneous doping process during deep‐sea hydrothermal vent mineral precipitations led to natural thermoelectric improvement, which is similar to natural mineral tetrahedrites [51].

Figure 17.

Thermal conductivity and phonon scattering of natural Cu_1+xFe_1−xS₂. (a) Temperature dependence of total thermal conductivity κ. (b) Temperature dependence of electronic component κ_e. (c) Temperature dependence of phononic component κ_ph. (d) Temperature dependence of dimensionless figure of merit ZT.

5. Conclusions

Our results of layered cobaltites Bi₂Sr₂Co₂O_y system based on narrow band model are not only helpful to understand large S and transport mechanism but also differentiate other systems based on a broad band model. In particular, we give the experimental evidence by Hall effect and C(T) measurements, demonstrating that Sommerfeld coefficient γ is dependent on charge carriers’ density n, and also as a function of DOS at E_F, which induces exotic enhancement of large S in Bi₂Sr_2−xCa_xCo₂O_y. Especially for Bi₂Sr_1.9Ca_0.1Co₂O_y, it may provide an excellent platform to be regarded as potential candidates for thermoelectric materials.

In addition, we demonstrated direct thermoelectricity generation in natural chalcogenides, Cu_1+xFe_1−xS₂, which was shown to have large S value and high power factor in the low x region, in which electron‐magnon scattering and large m^* values were detected. Since doped charge carriers exist in strong coupling with localized spins, the unusual alternation of p‐ and n‐type carriers should be of paramount importance in understanding charge dynamics arising from 3d orbital degrees of freedom. Such a finding of exotic thermoelectric properties in natural but not synthetic chalcopyrite opens a novel research field for manipulating low‐cost thermoelectricity or even electron/hole carriers, providing therefore a new perspective on technical feasibility for designing and pinpointing the surface‐morphology‐engineered devices via the naturally abundant materials.

Acknowledgments

The author gratefully thanks L. H. Yin, W. H. Song, Y. P. Sun, A. U. Khan, N. Tsujii, K. Takai, R. Nakamura, and T. Mori for their fruitful collaboration in the study of layered cobaltites and natural chalcogenides for thermoelectrics. This work was supported by the National Natural Science Foundation of China under Contract No. 10904151, the Fund of Chinese Academy of Sciences for Excellent Graduates, and the NIMS Open Innovation Center (NOIC) of Japan. The author thanks the Sichuan University Talent Introduction Research Funding (grant No. YJ201537) and Sichuan University Outstanding Young Scholars Research Funding (grant No. 2015SCU04A20) of China for financial support.