Mg2Si-Based Materials for the Thermoelectric Energy Conversion
- 511 Downloads
Thermoelectric materials are capable of converting a temperature gradient into electricity (thermoelectric power generation) and vice versa (Peltier cooling). The thermoelectric power generation has been used for decades in spacecraft, where radioactive decay provides the heat source. Additional applications under consideration are based on the utilization of waste heat, for example in automotives or the manufacturing industries. Commercial thermoelectric materials are normally based on Bi2Te3, PbTe, or possibly in the future on the so-called filled skutterudites, such as YbxCo4Sb12. The downside of these materials is that some of their major constituent elements are toxic, namely tellurium, lead, and antimony, and in part rare and expensive (ytterbium, tellurium). Mg2Si on the other hand is composed of abundant, environmentally benign elements, and thus offers a huge advantage for commercial applications. Here, we provide a review of Mg2Si-based materials for thermoelectric energy conversion, discussing how competitive these materials have become in comparison to the above-mentioned more traditional materials.
KeywordsThermoelectrics energy magnesium silicon tin bismuth
Thermoelectric Energy Conversion
Thermoelectric (TE) materials can convert a temperature gradient into electricity, utilizing the Seebeck effect, or use electricity to generate a temperature gradient, known as Peltier cooling. Peltier cooling is employed in portable coolers and laptops, as well as possibly for temperature control in car seats.1 More than half of the energy generated by mankind is lost as waste heat, but TE materials may tap into this waste heat, otherwise lost, and convert it in part into useful energy. For several decades, space missions from Voyager to Curiosity have used the Seebeck effect2,3 and nowadays thermoelectric generators are being tested on Earth for waste heat recovery in stationary applications as well as automotives.4,5 The relatively low conversion efficiency, as well as the price and toxicity of these materials, still mitigate against a wider use of this technology.6,7
There TH denotes the hot temperature and TC the cold temperature. Larger ZT values thus lead to larger conversion efficiency. Advanced thermoelectrics are semiconductors,8,9 so that the energy gap can prevent the co-existence of p- and n-type carriers. To obtain the best compromise of having both a high Seebeck coefficient α and a high electrical conductivity σ, these semiconductors are typically doped to a charge carrier concentration of the order of 1019 to 1020 cm−3.10 Furthermore, most of the advanced thermoelectrics are comprised of heavy elements, which provide a lower phonon contribution to the thermal conductivity, in turn reducing the total thermal conductivity. The latter consists of two parts, namely the lattice, κL, and the electronic thermal conductivity, κe, arising from the charge carriers, according to κ = κL + κe. Highly complex structures may also provide an advantage in reducing thermal conductivity, but this may also cause a lowered electrical conductivity through lowering the mobility of the charge carriers.11
Advanced Thermoelectric Materials
Traditionally, variants of Bi2Te3, a semiconductor with a narrow gap of 0.16 eV.12 have been employed for applications at temperatures below 500 K. The crystal structure of Bi2Te3 is a layered defect variant of the NaCl type, which in turn is adopted by PbTe, a well-known high-temperature thermoelectric material with a gap of 0.32 eV.13 Alloying with Sb afforded a maximum zT = 1.0 at 300 K for p-type Bi2Te3, and nanostructuring led to a performance increase as evidenced in zT = 1.4 at 300 K.14 Nanowires can exhibit even better performance15,16 for example in case of Bi2Te3 with zT values > 2 at 300 K.17 Bulk n-type Bi2Te3 may also exhibit zT values in excess of unity, e.g. zT = 1.2 at 423 K.18 Modifications of PbTe may exceed these zT values, for example in the p-type Tl0.02Pb0.98Te with zT = 1.5 at 773 K,19 in Na0.02Pb0.98Se0.15Te0.85 with zT = 1.8 at 850 K20 in SALT materials (sodium-antimony-lead-tellurium), e.g. Na0.95Pb20SbTe22 with zT = 1.7 at 650 K,21 and in the LASTT materials (lead-antimony-silver-tellurium-tin) Ag(Pb1−ySny)mSbTem+2 with nanodomains, achieving a high zT = 1.5 at the lower temperature of 627 K.22 On the n-type side, the LAST materials (lead-antimony–silver–tellurium), e.g. AgPb18SbTe20, stand out with zT = 2.1 at 800 K.23 Finally, a p-type composite of PbTe doped with 2% Na and embedded 4% of SrTe was reported to achieve zT = 2.2 at 915 K.24
More complex materials exhibiting comparable performance, in part because of their low thermal conductivity, are skutterudites and clathrates, including the n-type triple-filled skutterudite Ba0.08La0.05Yb0.04Co4Sb12 with zT = 1.7 at 850 K25 and the n-type clathrate Ba8Ni0.31Zn0.52Ga13.06Ge32.2 with zT = 1.2 at 1000 K26 and the best Si-based clathrate Ba8Ga16Si30 with zT = 0.87 at 873 K.27 Zintl materials28 often adopt even more complex structures with lower thermal conductivity, with examples such as the p-types Ni0.06Mo3Sb5.4Te1.6 with zT = 0.93 at 1050 K,29 Sr3Ga1–xZnxSb3 with zT = 0.97 at 1000 K,30 La0.4Yb13.6MnSb11 with zT = 1.15 at 1150 K,31 Yb14Mn0.4Al0.6Sb11 with zT = 1.32 at 1275 K32 and β-Zn4Sb3 with zT = 1.3 at 670 K.33
Among the materials with low lattice thermal conductivity of κL < 1 W m−1 K−1, thallium tellurides stand out with values often below 0.5 W m−1 K−1.34 This has led to several Tl tellurides exhibiting zT values above unity (at intermediate temperatures) as well, including values of 1.1 for Tl9Bi0.98Te6 at 500 K,35 1.2 for TlAg9Te5 at 700 K,36 1.2 for zone-refined Tl9BiTe6 at 500 K37 and finally 1.3 for Tl8.05Sn1.95Te6 and 1.5 for Tl8.10Pb1.90Te6 at 680 K.38
Most of the materials discussed above are antimonides and tellurides, and therefore toxicity is a major concern. On top of that, several metals occur in these materials that are of comparable or even higher toxicity, most notably thallium and lead. Affordability is another important issue for commercial applications of these materials, as some of these elements are rare and/or expensive, namely tellurium, ytterbium and silver. While cations from Tl to Bi and anions from Sb to Te are natural choices to form heavy materials exhibiting low thermal conductivity, the performance of Mg2Si-based materials demonstrates that lighter elements may also be employed as advanced thermoelectrics.39 Details of their thermoelectric properties are presented in the current review in comparison with established state-of-the-art materials.
RESULTS AND DISCUSSION
Synthesis of Mg2Si-Based Materials
The synthesis of Mg2Si is challenging for a number of reasons, namely the high affinity of magnesium for oxygen, the high vapor pressure of Mg, the high melting temperature of silicon at 1687 K, and the almost identical melting point of Mg2Si at 1358 K and boiling point of Mg at 1363 K. As a consequence, many Mg2Si samples exhibit MgO impurities.40 Therefore, several different methods in addition to classical ceramic methods have been devised, including the vertical Bridgman method using molten Mg2Si,41 crystal growth from molten Mg2Si1−xGex,42 B2O3 flux method,43 microwave assisted fast synthesis after ball-milling,44 vacuum plasma spray,45 a solid-state metathesis reaction of NaSi, MgCl2 and Na,46 and liquid encapsulated vertical gradient freezing using a KCl/MgCl2 eutectic mixture as encapsulant.47 Recently, an “ultrafast self-propagating high-temperature synthesis” was developed in which a cylinder of Mg, Si and Sb reacted after ignition within seconds in air to yield Sb-doped Mg2Si.48 The same authors also prepared other thermoelectrics via this method.49 In several of these methods, an excess of magnesium was used to compensate for its vapor loss.
Crystal Structure and Semiconducting Properties of Mg2Si
With such a band gap, the intrinsic carrier concentration of Mg2Si is very low, resulting in a low electrical conductivity of undoped Mg2Si of only σ = 3 Ω−1 cm−1 around room temperature (RT).56 Because of the light constituent elements and the small high symmetry unit cell, the lattice thermal conductivity is of the order of κ = 10 W m−1 K−1 at RT, which needs to be decreased, in addition to increasing the charge carrier concentration, before Mg2Si can be considered for the thermoelectric energy conversion.
Thermoelectric Properties of Bulk Mg2Si-Based Materials
Substituting Si in part with Ge or Sn atoms aids in reducing the thermal conductivity, because of the enhanced mass fluctuation combined with an overall increased molar mass.57 For example, using 30% Ge on the Si site, we obtained less than 3 W m−1 K−1 at RT58 compared to about 10 W m−1 K−156 for Mg2Si. A problem with using tin stems from the reported miscibility gap, in contrast to the complete solid solution of Mg2Si1−xGex. Depending on the synthesis method,59 the miscibility gap may occur between 40% and 60% Sn on the Si site.60 This problem may be overcome by optimizing the synthesis conditions, e.g. by using a B2O3 flux (preparing Sb-doped Mg2Si0.5Sn0.5).43 In 2015, a low thermal conductivity of κ = 1.9 W m−1 K−1 at 300 K was reported for a sample of nominal composition Mg2Si0.5Sn0.5 prepared from Mg2Si and Mg2Sn, but this sample inadvertently contained MgO nanoparticles as well as traces of unreacted Si and Sn that may have contributed to this low value.61 With values between 2 and 3 W m−1 K−1 at RT, the lattice thermal conductivity begins to be competitive with other thermoelectrics such as PbTe, while still being higher than that of Bi2Te3, the filled skutterudites or various clathrates.
The most widely used n-type doping elements here are antimony and bismuth, shown to substitute for the Si atoms.62 These elements belong to group 15 of the periodic table, and thus contribute one extra electron for each replaced Si atom from group 14. Because of the larger size of the Sb and Bi atoms, compared to Si, the amount of Sb and Bi is limited to less than 3% on the Si site.63 This amount may be increased in case of the alloyed samples Mg2Si1−xGex and Mg2Si1−xSnx, likely because of their unit cell expansions stemming from the larger sizes of Ge and Sn in comparison to Si atoms.64 Each percent of Sb or Bi corresponds to an expected increase in the carrier concentration of 1.6 × 1020 cm−3, but most often smaller increases were found, possibly because of the formation of Mg3Sb2 or Mg3Bi2 in the grain boundaries. For example, we found n = 3.5 × 1019 electrons per cm3 and 9.5 × 1019 electrons per cm3 for the hot-pressed materials of nominal composition Mg2Si0.98Sb0.02 and Mg2Si0.98Bi0.02, respectively, instead of the 3.1 × 1020 cm−3 calculated based on the formulae.62 Similarly, Zhang et al.65 reported n = 8.6 × 1019 electrons per cm3 for Mg2Si0.4Sn0.6Bi0.02 and n = 2.9 × 1020 electrons per cm3 for Mg2Si0.4Sn0.6Bi0.03, prepared by melt-spinning and consolidated by spark-plasma-sintering (MS-SPS). Depending on the synthesis technique, the density and doping level, electrical conductivity values of Sb- and Bi-doped samples typically range from σ = 400 Ω−1 cm−158 to 2000 Ω−1 cm−1 in the case of Mg2.16(Si0.4Sn0.6)0.97Bi0.03 with a carrier concentration of 2.4 × 1020 cm−3 at RT,66 and then decrease with increasing temperature.
The two samples prepared via MS-SPS (labeled MS in Fig. 2) and pressureless SPS (labeled Porous) exhibit significantly lower thermal and electrical conductivity below 600 K. An upturn is observed in the thermal conductivity in both of these samples beginning between 500 K and 600 K, likely caused by the bipolar effect. The other three samples, synthesized by solid-state reactions and consolidated by SPS, now show a clear bipolar effect until 700 K; at least until then, both the thermal and the electrical conductivity decrease steadily with increasing temperature. Around RT, κ ranges from 2.56 W m−1 K−1 to 2.77 W m−1 K−1 for the regular solid-state samples, compared to 1.53 W m−1 K−1 for the porous sample, and possibly less (extrapolated) for the MS sample. Similarly, the electrical conductivity of the solid state samples varies from σ = 1470 Ω−1 cm−1 to 2040 Ω−1 cm−1, compared to 715 Ω−1 cm−1 for the porous sample around RT. The highest conductivity occurs in the sample with the highest dopant concentration of the three, namely 3% Bi versus 1% and 1.5% Sb on the Si/Sn site. Utilizing the Wiedemann–Franz law, κe = LσT with L representing the Lorenz number, one can estimate the electronic contribution to the thermal conductivity, and then derive the lattice part via κL = κ − κe. The three solid-state materials exhibit very comparable lattice thermal conductivity, namely from κL = 1.65 W m−1 K−1 to 1.92 W m−1 K−1 at RT, while the porous sample is significantly below that, with κL = 1.16 W m−1 K−1. These values compare nicely with Bi2Te3.
Therefore, the Sn-containing samples are the ones with the highest power factor, P.F. = α2σ. Because α increases with increasing temperature, and P.F. is proportional to α2, P.F. continues to increase with increasing temperature despite the parallel decrease of σ. In the case of Mg2.16(Si0.4Sn0.6)0.97Bi0.03, the power factor changes from 30 μW cm−1 K−2 at 300 K to 42 μW cm−1 K−2 at 800 K. These numbers are highly competitive, as evident from a comparison with established thermoelectrics. The above-mentioned filled skutterudites Ba0.08La0.05Yb0.04Co4Sb12 (zTmax = 1.7) exhibits a maximum power factor of 52 μW cm−1 K−225 and the clathrate Ba8Ni0.31Zn0.52Ga13.06Ge32.2 (zTmax = 1.2) a maximum P.F. = 15 μW cm−1 K−2.26 Similarly, Tl-doped PbTe with zTmax = 1.5 has a maximum P.F. = 19 μW cm−1 K−2.19
The figure-of-merit values zT, depicted in Fig. 3b, are too low for applications at ambient temperatures, all being below 0.4 around RT. On the other hand, the steady increase of zT until 600–800 K or beyond, lead to competitive performance at those temperatures, with several examples surpassing zT of unity. The first examples with zTmax = 1.4 at 800 K were published in this decade, beginning with Mg2.16(Si0.4Sn0.6)0.97Bi0.03 as shown in Fig. 3b,66 and also including the Ge-containing Mg2Si0.53Sn0.4Ge0.05Bi0.02.71 Thus far, the highest value, zTmax = 1.63 at 615 K, was achieved by the porous Sb-doped Mg2Si0.5Sn0.5 consolidated via pressureless SPS, also shown in Fig. 3b.69 However, a problem with its porosity of 37% may lie in its applicability, i.e. the formation of a robust device.
No comparable p-type Mg2Si materials have been identified to date. The best examples exhibit figure-of-merit values between 0.3 and 0.7, including Mg2.10(Si0.3Sn0.7)0.95Ga0.05 with zTmax = 0.35 at 650 K,72 Mg2Si0.6Ge0.4:Ga(0.8%) with zTmax = 0.36 at 625 K,73 and Mg1.86Li0.14Si0.3Sn0.7 with zTmax = 0.5 at 750 K, where Li atoms were substituted onto the Mg site.74 Most recently, Mg1.86Li0.14Si0.4Sn0.6 was prepared via melt-spinning and spark-plasma-sintering within 1 h, resulting in zTmax = 0.58 at 760 K,75 and Mg2Li0.025Si0.4Sn0.6 was synthesized under a layer of B2O3, yielding zTmax = 0.7 at 675 K.76 Comparable or even better performance may be observed in p-type manganese silicides, namely in MnSi1.75 and variants thereof. This strongly depends on the preparation methods, as a comparison of the binary Mn silicides reveals that MnSi1.72, prepared via furnace melting and SPS, exhibits zTmax = 0.63 at 723 K,77 MnSi1.75 (Mn4Si7), synthesized via induction melting and SPS, shows zTmax = 0.62 at 800 K,78 MnSi1.84, prepared via ball-milling and pulse discharge sintering, a zTmax = 0.83 at 818 K,79 and MnSi1.85, obtained via ball-milling and SPS, a zTmax = 0.67 at 873 K.80
Thermoelectric Properties of Bulk Mg2Si-Based Nanocomposites
Forming nanocomposites is a proven strategy to further enhance thermoelectric materials. Ideally, this occurs with a reduction of the thermal conductivity by enhanced scattering of mid- to long-wavelength phonons, a less affected electrical conductivity and an increased Seebeck coefficient via energy filtering of the charge carriers.81, 82, 83 Since 2012, composites of Bi-doped Mg2Si with single-walled carbon nanohorns (SWCNH)84 and with Si nanoparticles,85 Bi-doped Mg2(Si,Ge) with multi-walled carbon nanotubes (MWCNT),86 and As- and Sb-co-doped Mg2(Si,Sn) with TiO287 have been investigated. While adding these nanoparticles to Mg2Si reduced the thermal conductivity, e.g. down to 6 W m−1 K−1 at RT in the case of the Si additions, these reductions cannot match the above-discussed Mg2(Si,Ge) and Mg2(Si,Sn) solid solutions.
The fact that the additions of MWCNT to Mg2Si0.877Ge0.1Bi0.023 and of TiO2 to Mg2(Si0.392Sn0.6)0.9925As0.008Sb0.0075 both caused an increase in zTmax by approximately 20%, demonstrates that this concept may also be applied to other solid solutions of Mg2Si with Mg2Ge and Mg2Sn, most interestingly to the materials with the best performance to date.
Within this review, we have demonstrated that several methods exist to render Mg2Si competitive with the best thermoelectrics at temperatures between 500 K and 800 K. Most importantly, alloying with Sn and doping with Sb and Bi, both on the Si site, followed by spark-plasma sintering, yielded several different bulk materials with naturally occurring nanoparticles, that exhibit zT values between 1.0 and 1.4 around 800 K. In addition, pressureless spark-plasma sintering afforded a porous material with enhanced Seebeck coefficient and reduced thermal conductivity, providing for the record zT = 1.6 at 615 K to date in this system. These performances are only surpassed by few triple-filled skutterudites and PbTe modifications with nanodomains or nano-additions, exhibiting bulk zT values between 1.7 and 2.2 at similar temperatures as presented in the “Introduction”.
In addition, the Mg2Si materials may further be enhanced via nanocomposites: first attempts yielded improvements of the order of 20% in the figure-of-merit, but were not performed on the best materials in this family. Considering all these findings occurred recently in the current decade, further improvements can be anticipated. This is particularly exciting because of the sustainability of these materials in contrast to the toxic antimonides and tellurides. Therefore, variants of environmentally benign Mg2Si may become the next generation materials.
Financial support from the Natural Sciences and Engineering Research Council is highly appreciated.
- 2.R.R. Furlong and E.J. Wahlquist, Nucl. News 42, 26 (1999).Google Scholar
- 5.M. Matsumoto, M. Mori, T. Haraguchi, M. Ohtani, T. Kubo, K. Matsumoto, and H. Matsuda, SAE Int. J. Eng. 8, 1815 (2015).Google Scholar
- 6.D.M. Rowe, Thermoelectrics Handbook: Macro to Nano (Boca Raton: CRC Press, Taylor & Francis Group, 2006).Google Scholar
- 54.O. Madelung, U. Rössler, M. Schulz (eds.), Non-Tetrahedrally Bonded Elements and Binary Compounds I, Landolt-Börnstein—Group III Condensed Matter, Vol. 41C (Berlin: Springer, 1998)Google Scholar
- 55.U. Winkler, Helv. Phys. Acta 28, 633 (1955).Google Scholar
- 60.E.N. Nikitin, V.G. Bazanov, and V.I. Tarasov, Sov. Phys. Solid State 3, 2648 (1961).Google Scholar