GUO Ki HUO Ho-Jie LUO Jun,

a (School of Materials Science and Engineering, Shanghai University, Shanghai 200444, China)

b (Materials Genome Institute, Shanghai University, Shanghai 200444, China)

ABSTRACT The layered 122 Zintl compounds have become an intriguing class of thermoelectric materials due to the promising electronic transport properties and inherently low thermal conductivity, showing the typical characteristics of “phonon-glass electron-crystal”. Owing to the unprecedented performance tunability, the thermoelectric properties of the layered-structure compounds are completive with some traditional thermoelectric materials. Point defects involving vacancy, aliovalent doping and equivalent alloying atoms have been introduced to further enhance the thermoelectric properties. This review emphasizes the effects of various point defects on the thermoelectric parameters, and provides perspective on the strategies for increasing the thermoelectric figure of merit zT, which are believed to be applicable for improving the thermoelectric properties of many other compounds.

Keywords: layered structure, 122 Zintl phase, point defect, orbital alignment, phonon scattering,

1 INTRODUCTION

Thermoelectric materials can realize direct conversion between heat and electrical energy, showing promising appli- cations in electricity generation and portable refrigeration[1]. In the last decades, a lot of efforts have been focused on increasing the conversion efficiency of thermoelectric materials in order to achieve commercial competitiveness of thermoelectric tech- nology. In principle, the conversion efficiency of a thermo- electric material is assessed by the dimensionless figure of meritzT=S2σΤ/(κe+κL), whereSis the Seebeck coefficient,σis the electrical conductivity,κeandκLrepresent respectively the electronic and lattice contribution to the thermal conductivity, andTis the working temperature[2]. Accordingly, the ideal thermoelectric materials demand large Seebeck coefficients, high electrical conductivities, and low thermal conductivities. However, the strong coupling betweenS,σandκethrough carrier concentrationnmakes it hard to essentially increase the thermoelectric figure of meritzT. Slack proposed the concept of “phonon-glass electron-crystal”, implying that the electronic and thermal transport properties could be decoupled via constructing functional units to design high-performance thermoelectric materials[3]. The filled-skutterudite is a typical representative in which CoSb3frame-work and filled atoms behave as electronic transport media and phonon scattering centers, respectively[4]. As a result,zT= 1.7 at 850 K has been achieved via tuning the composition, concentration and valence state of multiple-filled atoms[5].

Recently, intensive attention has been paid to the layered 122 phase Zintl compounds AB2X2, which adopts the CaAl2Si2-type structure with space groupP-3m1 (Fig. 1a)[6-13]. Herein, A is a divalent cation, e.g., alkaline earth metals (Mg, Ca, Sr, Ba) and rare earth metals (Yb, Eu, Sm), B stands for Mg, Mn, Zn, and Cd, and X represents As, Sb, and Bi. The structure of CaAl2Si2consists of two-dimensional [Al2Si2]2-slabs and single Ca2+atomic slabs. The [Al2Si2]2-slabs are formed by AlSi4tetrahedra which share three of their edges with neighboring tetrahedra. Within the Zintl-Klemm concept, Ca donates electrons to polyanions [Al2Si2]2-to satisfy the charge balance, assuming complete charge transfer from the cation to the polyanions. The layered 122 Zintl phase AB2X2is a kind of “phonon-glass electron-crystal” compound, where polyanionic slabs are good charge transport media and cation layers are responsible for phonon scattering. Fig. 1b displays the band structure of representative YbMg2Sb2. The valence band is composed of nondegeneratepz(Nv= 1) and doubly degeneratepxy(Nv= 2) band orbitals, which dominate the charge transport properties of YbMg2Sb2due to itsp-type conduction. This compound has an indirect band gap of 0.7 eV, within the appropriate range of excellent thermoelectric material[14]. Furthermore, the rich solid-state chemistry of Zintl phases enables various possibilities for chemical substitutions and/or structural modification, and thus high thermoelectric performance has been explored[15,16].

Fig. 1. (a) Crystal structure of trigonal AB2X2. (b) Band structure of the representative Zintl compound YbMg2Sb2

Especially, point defects engineering proves to be an effective strategy leading to enhanced thermoelectric pro- perties of layered Zintl compounds AB2X2. However, the various point defects involved have highly differentiated influence on the charge and thermal transport properties. Herein, we summarize the researches on the point defects including vacancy, aliovalent doping and equivalent alloying atoms for improving the thermoelectric performance. The corresponding mechanisms on manipulation of thermoelectric parameters have been analyzed and disclosed, which is instructive for future optimization of thermoelectric per- formance for many other compounds.

2 MANIPULATION STRATEGIES

2. 1 Vacancy engineering for increasing the power factor

Strictly, Zintl phases are defined as valence-precise com- pounds and are thus expected to exhibit intrinsic semicon- ducting behavior[15]. However, the thermodynamic effect usually causes the formation of defects, resulting in the deviation of the final product from the ideal stoichiometric ratio (Fig. 2a). In 2015, Snyder and Bux reported that cation Yb vacancies are responsible for the high hole concentrations (1019～1020cm-3) in Yb1-δZn2Sb2[17]according to the following equation:

The occurrence of one hole is actually accompanied by two cation vacancies due to the limitation of doping efficiency. Thus, through simple manipulation of vacancy, the hole concentration can be controlled within a relatively wider single-phase region by valence imbalance (Fig. 2b), which has been realized to increase the power factor PF. Similar vacancy manipulation has been adopted in Eu1-xZn2Sb2system[18]. It is worth noting that the temperature dependence of vacancy concentration usually results in the unstable electronic transport behaviors as well as the deviation from the optimal carrier concentration in AB2X2[19].

Fig. 2. (a) Calculated ZnSb-YbZn2Sb2 phase diagram demonstrating the large stable concentration of Yb vacancies tolerated by YbxZn2Sb2. (b) Hall carrier concentration of YbxZn2Sb2 samples decreasing linearly for x < 1.00 and remaining constant when x > 1.00. Reproduced from [17] with permission from the American Chemical Society

One should not ignore the effects of cation vacancy on thermal conductivity. On one hand, the increased hole con- centration results in the augment of electronic contribution to the total thermal conductivity. On the other hand, the cation vacancies can also contribute to the phonon scattering for reducing the lattice thermal conductivity. Significant im- provements in the average and peakzTof Yb1-δZn2Sb2and Eu1-xZn2Sb2were achieved by the vacancy engineering[17,18].

2. 2 Aliovalent doping for optimizing the hole concentration

Despite that manipulation of carrier concentration is a conventional strategy for performance optimization of almost all thermoelectric materials, it is essential because it does realize the compromise betweenσandS, both of which are dependent on the carrier concentrationn. Although the idea seems to be simple, selecting a proper dopant is not easy. The layered 122 phase Zintl compounds AB2X2arep-type semiconductors, indicating that the hole is the major carrier as the result of the intrinsic cation vacancy. Thus, it is reasonable to use the alkali metals (Li, Na, K) to substitute A site or Cu/Ag to substitute the B site, which is so calledp-type (acceptor) doping because the alkali metals (Cu/Ag) possess less valence electrons than the corresponding matrix elements[20-23]. The substitution limit and doping efficiency of the dopants should be addressed in some cases when the optimum carrier concentration cannot be achieved, since that the higher TE performances at the optimal carrier con- centration predicted by the theoretical calculations have not been realized in experiments. Fig. 3a displays the hole concentration as a function of the composition for BaCd2-xAgxSb2and Ba1-yNayCd2Sb2at room temperature(20]. As the content of the dopant increases, the hole concentration increases linearly, evidencing the hole concentration can be tuned within a wide range. The slopes for these two straight lines are distinctively different, which reflects the difference in the doping efficient of Na and Ag.

Fig. 3. (a) Hall carrier concentration (p) depending on the composition for BaCd2-xAgxSb2 and Ba1-yNayCd2Sb2 at room temperature[20]. (b) Hall carrier concentration-dependent Seebeck coefficient for BaCd2-xAgxSb2 and Ba1-yNayCd2Sb2 and Ba0.8-yYb0.2NayCd2Sb2 at 300 K[20]

As mentioned above, thepzorbital separates from thepxandpyorbitals due to the spin orbital coupling in layered materials. Thus, the increase in hole concentration caused byp-type doping would push the Fermi level to enter into the deep valence band, which opens the possibility of charge transport through all three orbitalspzandpxy. Fig. 3b shows the Seebeck coefficient dependent on the hole concentration for BaCd2-xAgxSb2, Ba1-yNayCd2Sb2and Ba0.8-yYb0.2NayCd2Sb2at 300 K[20]. The hole begins to get heavier at the concentrationpHof 1019cm-3and changes to be a constant again whenpH> 1020cm-3, strongly indicating that orbitalspzandpxycontribute to the electrical transport properties. Thus, the enhanced effective mass is beneficial for the enhancement of Seebeck coefficients and power factors.

2. 3 Equivalent alloying for realizing the orbital alignment

In addition to the carrier optimization for maximizing the power factor, the increase of Seebeck coefficient proves to be another effective way to enhance the charge transport pro- perties by enlarging the carrier effective massm* via band convergence or orbital alignment[24]. In principle, the density- of-states effective massm* is given bym* =Nv2/3mb*, whereNvinvolves the orbital degeneracy, andmb* is the mass of a single valley. One can raise the orbital degeneracy or/and the mass of a single valley to increasem*. However, flat valley in the electronic structure implying a largemb* would result in essential reduction in mobilityμ, which in turn suppresses the electrical conductivity.The ideal solution is to increase the orbital degeneracyNvrather thanmb*.Taken 122 phaseYbMg2Sb2as an example, the valence band maximum is dominated by theporbital characteristics of Sb (Fig. 1b). The effect of spin orbital coupling in layered materials causes the separation ofpzorbital from thepxandpyorbitals. In this case, the crystal field splitting energy between these two bands defined as ΔE = E(Γ(pz) - Γ(pxy)) is about 0.2 eV. Interestingly, negative crystal field splitting energies have been identified for many other 122 layered compounds such as YbZn2Sb2and EuCd2Sb2, which promotes a significant research topic on realization of the orbital alignment (ΔE = 0) by simple equivalent alloying (Fig. 4a)[25-28]. The systems including YbCd2-xZnxSb2and EuCd2-xZnxSb2alloys demonstrate the success of zero-ΔE rule, while the recent results on YbMg2-xZnxSb2alloy evidence insignificant orbital alignment effect[14]. Therefore, it is necessary to reveal this unknown mechanism in the near future.

Fig. 4. (a) Diagrammatic sketching for the composition of EuCd2-xZnxSb2 illustraing the orbital alignment by changing the Zn content. (b) Lattice thermal conductivity κL of EuCd2-xZnxSb2 as a function of the composition at various temperature[26]

Equivalent alloying also shows significant influence on the lattice thermal conductivityκL, as a result of the introduction of point-defect phonon scattering. Although the layered 122 Zintl compounds normally exhibit intrinsically low lattice thermal conductivity originated from the strong anhar- monicity, the point defect would further lead to the reduction inκL. Fig. 4b presents the lattice thermal conductivityκLdepending on the composition for EuCd2-xZnxSb2alloy at various temperature. The dashed line denotes the predicted data using the Debye-Callaway model[29]according to equation 2:

Wherexis defined by the equation ofx=?ω/kBT(ωis the phonon frequency), andυ,kB,?andτare the phonon velocity, Boltzman constant, Planck constant, Debye temperature, and phonon relaxation time, respectively. Under the assumption that phonon-phonon scattering and point-defect scattering dominate the phonon transport, the total phonon relaxation time can be expressed as

The minimum lattice thermal conductivity is usually observed for alloys with the intermediate composition. This fact demonstrates the strongest point defect scattering from the fluctuation of mass and stress (Fig. 4b). Combined with the zero-ΔE rule, one can expect that the perfect alloy system should be the one with the orbital alignment for the intermediate composition. In this situation, high power factor and low lattice thermal conductivity can be achieved simul- taneously, which is beneficial for achieving excellent ther- moelectric performance in the layered 122 Zintl compounds.

3 CONCLUSION

The layered 122 Zintl compounds are appealing ther- moelectric materials as a result of the tunability of electronic transport properties and inherently low thermal conductivity. The point defects have been engineered in order to increase the thermoelectric figure of meritzT. The intrinsic vacancy can rise the hole concentration and reduce the lattice thermal conductivity. The aliovalent dopants are usually utilized to optimize the hole concentration for the compromise of the electrical conductivity and Seebeck coefficient. Equivalent alloying in AB2X2can realize the orbital alignment for enhancing the hole effective mass without damaging the mobility. Thus, the point defect engineering makes their performance competitive with other thermoelectric materials.