Shiyang
He‡
ab,
Amin
Bahrami‡
*a,
Pingjun
Ying
a,
Lars
Giebeler
a,
Xiang
Zhang
c,
Kornelius
Nielsch
ab and
Ran
He
*a
aLeibniz Institute of Solid State and Materials Science, 01069 Dresden, Germany. E-mail: a.bahrami@ifw-dresden.de; r.he@ifw-dresden.de
bInstitute of Materials Science, Technische Universität Dresden, 01062 Dresden, Germany
cNational Center for International Joint Research of Micro-Nano Molding Technology, School of Mechanics and Safety Engineering, Zhengzhou University, 450001 Zhengzhou, China
First published on 31st May 2022
The complexity of crystal structures plays an intriguing role in manipulating properties in thermoelectrics, spintronics, and batteries. In comparison to the widely studied ternary half-Heusler thermoelectric compounds, quaternary double half-Heusler compounds are promising due to their intrinsically low lattice thermal conductivities (κL). However, they have been much less investigated due to the limited material availability. In this study, we report a new double half-Heusler compound based on ZrNi(In,Sb). Upon tuning the ratio of In/Sb from 0.5/0.5 to 0.4/0.6 and reducing the nominal concentrations of Zr and Ni by 10%, we greatly reduce the intensities of the impurity-phase peaks in the diffraction patterns. An even better phase purity, in combination with an optimized power factor, is realized by substituting Co at the Ni sites. Further alloying Hf at the Zr sites enhances the point defect scattering of phonons, which yielded a minimum κL of ∼1.8 W m−1 K−1 and a maximum zT of ∼0.5 for Zr0.7Hf0.2Ni0.65Co0.25In0.4Sb0.6 at 973 K. Our work thus confirms the intrinsically low κL of ZrNi(In,Sb) double half-Heusler compounds and indicates their promising applications upon further improving their electrical transport properties.
Thermoelectric (TE) technology is a promising sustainable solution to alleviate the energy crisis and environmental pollution.4 The performance of TE materials is determined by the coupled electrical and thermal transport, as manifested in the expression of the figure-of-merit, zT, given by zT = S2ρ−1(κele + κL)−1T, where S, ρ, κele, κL, and T are the Seebeck coefficient, the resistivity, the electrical thermal conductivity, the lattice thermal conductivity and the absolute temperature, respectively.5 Thus, good TE materials require high power factors (PF = S2ρ−1) as well as low total thermal conductivity to realize high zT.6 However, optimizing the zT is challenging due to the intercorrelation of the transport parameters (S, ρ, and κele), such as via the carrier concentration.7 To realize effective decoupling, tremendous efforts are devoted, which can generally be categorized into two approaches.8 One is to maximize the PF by optimizing the carrier concentration via doping or band engineering,9 while another one targets seeking materials or processes to minimize the thermal conductivity.10
As a prominent TE material candidate in the moderate temperature region, half-Heusler (HH) compounds have attracted significant attention due to their low cost, high thermal stability, and mechanical robustness, as well as their excellent electrical properties.11 The bandgap of HH compounds varies from 0.1 to 3.7 eV based on the composition.11 The valence electron count (VEC) per unit cell has been confirmed to play a critical role in the transport properties.12 A slight variation from the 18 VEC (i.e., through doping) will optimize the PF, and the high PF is marked as the unique advantage of HH compounds.13 For instance, an ultrahigh PF value of ∼10.6 mW m−1 K−2 has been achieved in Ti-doped NbFeSb, and a record output power density of ∼22 W cm−2 is obtained in a single-leg TE device.14 In another example, Zhu et al. reported an outstanding zT of ∼1.5 at 923 K for p-type HH in a Sn-doped ZrCoBi alloy with a high PF (3.7 mW m−1 K−2 at 823 K).15 Despite the excellent electronic performance, in comparison with other classic thermoelectric materials (e.g. Bi2Te3-, PbTe- and Cu2Se-based TE materials),13,16,17 the intrinsically large κL of HH compounds impedes their further zT improvement. Because of their innately low thermal conductivity, the nominal 19-electron half-Heusler compounds,18 or defective half-Heusler compounds,19 have recently been identified as attractive TE material options for HH compounds. As a result, the primary objective of discovery in the TE discipline is to create novel HH alloys with low κL yet strong electrical characteristics.
The complexity of a primitive unit cell is a key factor to reduce the κL of HH compounds. Anand et al.20 compared the phonon frequency dependence of calculated cumulative κL of ternary (TiCoSb) and quaternary (Ti2FeNiSb2) HH compounds. A κL reduction of ∼50% as found for Ti2FeNiSb2 occurred in the high energy range of the phonons. In comparison to TiCoSb, a lower group velocity and a higher scattering rate of Ti2FeNiSb2 have been found, which indicates that quaternary HH compounds with more complex crystal structures would in general have lower κL than the traditional ternary HH compounds. Hence, HH systems are promising candidates for designing complex crystal structures to suppress the κL. Anand et al. predicted a total number of 713 possible ternary, defect-free HH compounds, of which 487 have been fully explored.20 However, quaternary double HH compounds, based on the same calculations, have been rarely investigated, with only TiFe1−xNixSb being successfully synthesized.21 The lowest κL of 2.5 W m−1 K−1 at 973 K has been achieved in TiFe0.6Ni0.4Sb while still reaching a high PF (∼1.5 mW m−1 K2), indicating that enhancing the complexity of a primitive unit cell for HH alloys can dramatically suppress the κL.21 Thus, quaternary HH compounds have huge potential to enable the discovery of further new highly performing TE materials due to the huge unexplored phase space of ∼7719 quaternary HH compounds.20
In this study, a new n-type double HH compound based on ZrNi(In,Sb) is realized (Fig. 1(a)). Starting from ZrNiIn0.5Sb0.5, we find obvious impurity peaks besides the main HH phase by X-ray powder diffraction (XRD). The intensity of the impurity peaks is substantially reduced after tuning the composition by changing the nominal In/Sb ratio to 0.4/0.6. Lowering the nominal amount of Zr and Ni by 10% improves this impurity phase reduction. Further substituting Co at the Ni sites not only tunes the carrier concentrations to realize an almost doubled peak PF from 1.08 to 2.03 mW m−1 K2 at 973 K in Zr0.9Ni0.65Co0.25In0.4Sb0.6 but also further removes the impurity phase completely according to the detection limit of the diffractometer. Additionally, we further suppressed the κL from 2.10 to 1.81 W m−1 K−1 at 973 K by alloying a small amount of Hf to partially substitute Zr on the respective sites, which is comparable with other HH materials that have shown minimized κL (Fig. 1(b)). As a result of simultaneously optimizing the electrical properties and suppressing the κL, the final zT values were enhanced by 336% from 0.11 to 0.48 at 973 K in a compound with the composition of Zr0.7Hf0.2Ni0.65Co0.25In0.4Sb0.6 (Fig. 1(c)), demonstrating the great potential of ZrNi(In,Sb) HH compounds for intermediate- and high-temperature energy conversion.
![]() | ||
Fig. 1 (a) Unit cell of the crystal structure of ZrNi(In,Sb) double HH compounds (space group F![]() |
Subsequently, to evaluate the lattice parameters and phase contents, Rietveld analyses based on the XRD results were performed for four specimens with the nominal compositions of ZrNiIn0.5Sb0.5, ZrNiIn0.4Sb0.6, Zr0.9Ni0.9In0.4Sb0.6, Zr0.9Ni0.65Co0.25In0.4Sb0.6, and Zr0.7Hf0.2Ni0.65Co0.25In0.4Sb0.6 as shown in Fig. 2, and the detailed information can be found in Table S1.†
For the Rietveld analyses, the structural model of Romaka et al.24 was adopted for the previously discussed elements. The resulting structural model fits the X-ray powder diffraction patterns to a great extent and with a phase content > 93 wt% to almost 99 wt% depending on the total amount of minor phases. Here, the specimen with the nominal composition ZrNiIn0.5Sb0.5 has the highest percentage of ∼7 wt% and ∼1 to 2 wt% for all other compounds. The evaluated sizes of the lattice parameter a result in the following order of the compounds: ZrNiIn0.4Sb0.6 > ZrNiIn0.5Sb0.5 > Zr0.9Ni0.9In0.4Sb0.6 > Zr0.9Ni0.65Co0.25In0.4Sb0.6 > Zr0.7Hf0.2Ni0.65Co0.25In0.4Sb0.6 reflecting the different concentrations of substituent elements and their respective radii. As Sb has a smaller radius than Zr, Ni, and Co, reduced lattice parameters are expected, and the lattice parameters of the individual compounds therewith follow their composition obtained by the ICP-OES analyses shown in Table S2.† We find that while the concentrations of other elements agree well with the nominal one, In always displays certain deficiencies that may originate from the losses during the arc melting. Additionally, increased Sb and Ni/Co concentrations are found, whereas the In concentration decreased. Both may lead to the resulting lattice parameter order. The occupancy has not been included in this consideration as these results need further investigation to explain them plausibly. Fig. 3 illustrates the fracture surface obtained by SEM (scanning electron microscopy) for the compounds ZrNiInxSb1−x (x = 0.4–0.6). The grain boundaries on this scale become less defined with decreasing In content, as displayed in Fig. 3(b) and (c). Additionally, the average grain size values of ZrNiInxSb1−x are similar based on statistics but still show a slightly increasing trend with lower In content. This behavior may be attributed to the distinct response to the current-assisted sintering process, as will be discussed soon since the Sb-rich compound is more conductive than the In-rich one. It is worth noting that the increasing trend of grain size is more visible, while the electrical resistivities increased in Co-doped samples. As shown in Fig. S2,† the clear grain boundaries can be observed, and the grain size significantly increased with increased Co doping content (Fig. S2(h)†). Additionally, grain sizes above 1 μm appeared and were distributed uniformly – marked with the red circles – when the Co-doped content reached 0.3 (Fig. S2†).
![]() | ||
Fig. 3 The fracture surfaces obtained by SEM for (a) ZrNiIn0.6Sb0.4, (b) ZrNiIn0.5Sb0.5 and (c) ZrNiIn0.4Sb0.6, and the distribution of the (d) grain size for the above composition. |
![]() | ||
Fig. 4 Temperature dependence of S (a and d), ρ (b and e) and PF (c and f) for ZrNi(In,Sb) double HH compounds. |
Considering the low electrical resistivity and the Seebeck coefficient, it is desirable to reduce the carrier concentration to optimize the power factor, which is realized by substituting Co for Ni. As illustrated in Fig. 4(d) and (e), both the electrical resistivity and the Seebeck coefficient of Zr0.9Ni0.9−yCoyIn0.4Sb0.6 (y = 0–0.3) are linearly dependent on the temperature in the degenerated nature. The value of ρ increases monotonically from 2.35 to 12.57 μΩ m with the doping level of Co at 973 K (Fig. 4(e)). Correspondingly, the S also increases three times from −51.3 to −151.1 μV K−1 at 973 K (Fig. 4(d)), implying that the Co doping could effectively tune the electrical properties. Benefiting from the huge increase of S, the PF displays ∼100% enhancement from 1.07 to 2.04 mW m−1 K2 for the composition of y = 0.25 (Fig. 4(f)).
The variation in the electrical properties can be understood from the results of the carrier concentration (n) and the Hall mobility (μ). Fig. 5(a) and (b) show the Hall measurement results in the range of 10–300 K. The n values of all the samples are found in the range of 1021 cm−3 and are weakly dependent on the temperature. As shown in Fig. 5(a), the ZrNiIn0.5Sb0.5 sample has the lowest Hall concentration (nH) among the ZrNiIn1−xSbx samples in the temperature range of 10–300 K. The Hall concentration was found to increase by either increasing the In/Sb ratio to 0.6/0.4 or decreasing it to 0.4/0.6. Whereas the increased nH stems from an increased VEC in the Sb-rich composition (nominally the VECs are 18 for ZrNiIn0.5Sb0.5, and 18.2 for ZrNiIn0.4Sb0.6), it has more complex origins in the In-rich composition because of the much higher concentration of impurities (Fig. S1(a)†). Interestingly, the carrier concentration changes little by decreasing the nominal concentrations of both Zr and Ni by 10% from ZrNiIn0.4Sb0.6 to Zr0.9Ni0.9In0.4Sb0.6. The Co-substituted composition, Zr0.9Ni0.65Co0.25In0.4Sb0.6, displays a further reduced n to ∼2 × 1021 cm−3, thus explaining its higher power factor in Fig. 4(f). Additionally, Fig. 5(a) shows that the substitution of Hf on the Zr site doesn't influence the nH significantly due to the isovalent nature of Zr and Hf, as well as the lanthanide-contraction of Hf as has been discussed by Liu et al.25 Accordingly, the electrical properties do not change significantly upon Hf substitution, as shown in Fig. S3.†
![]() | ||
Fig. 5 (a) Temperature dependence of n; (b) μ and (c) n dependence of |S| at 300 K ; (d) optimum PF at 300 K for the ZrNi(In,Sb) double HH compounds. |
On the other hand, the μ values for all the Co-free samples are similar, typically ranging from 14 to 10 cm2 V−1 s−1. The substitution of Co in Zr0.9Ni0.65Co0.25In0.4Sb0.6 yields a reduction in the mobility by nearly ∼40%, probably due to the enhanced electron scattering through alloying. Based on the Hall measurement and electrical measurement results, we plotted the carrier-concentration as a function of the absolute Seebeck coefficient in Fig. 5(c) together with the calculations to evaluate the DOS effective mass m*, using the single parabolic band (SPB) model,26 which is estimated to be 2.8me and similar to the ZrNiSn-based compounds.27–29 This approach is reasonable since the band edges mainly constitute the d-orbitals from the transition metals (Zr and Ni), so the modification at the most electronegative sites (In/Sb or Sn) is not expected to change the effective mass drastically. Fig. 5(d) shows the n-dependent optimum PF of ZrNi(In,Sb) samples at 300 K following an SPB model,26 together with our experimental results. Clearly, there is great potential to further improve the power factor by further reducing the carrier concentration. However, this idea is not realized by increasing the Co concentration to y = 0.3, as shown in Fig. 4, which might originate from an enhanced alloy scattering when the Co amount exceeds a certain limit. Therefore, other approaches are sought to further enhance the thermoelectric performance, which will be introduced later.
![]() | ||
Fig. 6 Temperature dependence of (a) κtot and κL (b) for ZrNiInxSb1−x, and (c) the comparison of κL of the calculated/experimental double HH alloys at 300 K.20 |
Subsequently, we investigated the dependence of the thermal transport properties on the Co and Hf substitutions. In parallel, the temperature-dependent κtot and κL of Zr0.9Ni0.9−yCoyIn0.4Sb0.6 and Zr0.9−zHfzNi0.65Co0.25In0.4Sb0.6 compounds are presented in Fig. 7. For Zr0.9Ni0.9−yCoyIn0.4Sb0.6, while κtot is suppressed by increasing the Co content (Fig. 7(a)), κL displays an inverse trend (Fig. 7(b)). Thus, the reduction of κe is very important to overcompensate the rebound of κL. Furthermore, the increased κL with respect to the increased Co content is counter-intuitive, since heavier alloying usually scatters phonons more strongly. The Rietveld analysis (Table S1†) suggested that the half-Heusler phase has a phase content of ∼98.6 mass% for the Zr0.9Ni0.65Co0.25In0.4Sb0.6 sample; therefore, the impurity phases should contribute negligibly to the varied lattice thermal conductivity. Furthermore, we observe a trend of grain enlargement with more Co following the SEM results (Fig. S2†), which could be partially accountable for the enhanced κL. The mechanism of such a grain-enlargement is not well understood, but it has been widely reported for half-Heusler materials that, under the same preparation conditions, a lower electrical resistivity usually yields a small grain size.33,34 This is probably due to less Joule heating produced during sintering. In addition, our group recently revealed that a higher doping level will induce, spontaneously, a stronger charge-compensation effect in the half-Heusler compounds by producing charged point defects to counteract the extrinsic doping.34 This effect is widely appreciated in tuning the electronic transport, but it can also significantly intensify the phononic scattering. The substitution of Co at the Ni sites reduces the carrier concentration/doping level, thus reducing the formation of charged point defects that are against the extrinsic doping. Admittedly, considering the complex multiple-elemental nature and the formation of impurities, it is very challenging to quantify the point defect through refinement based on a Mo anode. Diffraction patterns with higher quality are necessary such as those obtained by synchrotron or even neutron radiation. These will be subject to future studies.
![]() | ||
Fig. 7 Temperature dependence of κtot (a and c) and κL (b and d) for the Co-doped and Hf-doped ZrNiInxSb1−x based HH alloy samples. |
For Zr0.9−zHfzNi0.65Co0.25In0.4Sb0.6 samples, the κtot gradually decreases as the Hf content increases, while there is only a negligible difference of κe (Fig. S4(c)†). Hence, the decrease of κtot is related to a stronger mass fluctuation induced by higher Hf contents. In particular, the lowest κL of Zr0.7Hf0.2Ni0.65Co0.25In0.4Sb0.6 is ∼4.1 W m−1 K−1 at room temperature and ∼1.81 W m−1 K−1 at 973 K as given in Fig. 7(d). Finally, the zT values increase from ∼0.1 to ∼0.4 at 973 K with a tuned carrier concentration by Co-doping and are further improved to ∼0.5 at 973 K with suppressed lattice thermal conductivity realized in the compound with the nominal stoichiometry Zr0.7Hf0.2Ni0.65Co0.25In0.4Sb0.6 as shown in Fig. 1(c). Additionally, the optimum zT of the newly discovered ZrNi(In,Sb) compounds is comparable with that of other double HH systems21,35–37 (Fig. S5†), demonstrating that ZrNi(In,Sb) compounds enrich the HH families and can potentially be applied in electricity generation.
Footnotes |
† Electronic supplementary information (ESI) available. See https://doi.org/10.1039/d2ta02413f |
‡ These authors contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2022 |