The performance of Mg2Si-based

thermoelectric materials prepared from MgH2

Y. Z. Zhang 1, Y. H. Han 2 and Q. S. Meng*1
In this paper, thermoelectric materials of pure Mg2Si, Mg2Si1-xSnx and 1% Bi-doped Mg2Si were
prepared by a low-temperature solid-state reaction using MgH2, Sn and Si, followed by a field-
activated and pressure-assisted sintering process for consolidation. The reaction between
silicon, tin and magnesium hydride, instead of pure magnesium, minimises the oxidation and
volatilisation of magnesium. After consolidation, the prepared thermoelectric materials, which
had a grain size of 500 nm, were evaluated for their thermoelectric properties. Furthermore, the
electrical and thermal conductivities and the Seebeck coefficient were measured over the
temperature range of 300–750 K. A sample of 1% Bi-doped Mg2Si showed a maximum figure of
merit, ZT, of 0·53 at 750 K. Additionally, the Mg2Si1-xSnx samples were seen to possess an extra
low thermal conductivity. The results are discussed in comparison with previously reported
values and a prediction of modelling studies.
Keywords: MgH2, Mg2Si1-xSnx, Field-activated and pressure-assisted sintering process, Thermoelectric material, Thermoelectric conductivity, Electric
conductivity, Carrier concentration

Introduction and Mg2Si1-xSnx are particularly suitable for operating

in the mid-temperature range, 500–800 K. Additionally,
The current concern over future energy prospects and the they have the advantages of non-toxicity, low density
impact of fossil fuels on the environment has prompted a and the relative abundance of their elements in nature.
concerted effort to seek alternative energy sources. The silicide of Mg2Si and Mg2Si1-xSnx can be prepared
Among the many potential approaches considered for by various methods, including induction melting, spark
alternative energy, the use of thermoelectric (TE) plasma reactive sintering, solid-state reaction, vertical
materials has gained considerable attention in recent Bridgman growth and mechanical alloying.6,7 However,
years. Such materials, which can convert heat directly to melting methods suffer from numerous difficulties,
electrical energy, have already been used in relatively making the preparation of the desired pure material diffi-
small-scale applications.1,2 The efficiency of TE materials cult, if not impossible. The difficulties arise from the vola-
depends on their property and structure and is deter- tility of Mg, its oxidation tendency and the significant
mined by the value of the dimensionless figure of merit, difference between the melting points of Mg (650°C),
ZT, defined by Sn (232°C) and Si (1414°C). For this reason, alternative

2  methods have been investigated, as indicated above.
s σ In this paper, MgH2 powder was used instead of Mg
ZT = ·T (1)
κ powder, and it reacted with Si and Sn powder at a relatively
low temperature.8 In doing so, the problems of volatilisation
and oxidation (of Mg) are avoided by adjusting the relative
where S is the Seebeck coefficient, σ is the electrical amounts of MgH2 and Si, and it was possible to obtain
conductivity, κ is the thermal conductivity and T is Si nanodots-doped Mg2Si TE materials. Pure Mg2Si,
temperature. Mg2Si1-xSnx and 1 at% Bi-doped Mg2Si were prepared
Among the many material systems identified as poten- by this method and consolidated by the field-activated
tially attractive for such applications, the compounds and pressure-assisted sintering process (FAPAS).
Mg2Si and their solid solutions have received considerable
attention in recent years.3,4 These materials have a rela- The experimental procedure
tively high electrical conductivity, high Seebeck coefficient The following powders were used as the starting materials
and low thermal conductivity.5 As TE materials, Mg2Si to prepare pure Mg2Si, Mg2Si1-xSnx and Bi-doped Mg2Si:
MgH2 (−325 mesh, 98% pure, Alfa Aesar (Tianjin) Co.,
1
College of Material Science and Engineering, Taiyuan University of Ltd.); Si (−200 mesh, 99·99% pure, Shanghai Crystal
Technology, Taiyuan 030024 China Pure Industrial Co., Ltd.); Sn(−325 mesh, 99·99% pure,
2
Shanxi Jinghui Solar Energy Co., Taiyuan 030024, China Alfa Aesar (Tianjin) Co., Ltd.) and Bi (−200 mesh,
*Corresponding author, email [email protected] 99·99% pure, Crystal Pure Industrial Co., Ltd.). All the

© W. S. Maney & Son Ltd 2015

Received 29 September 2014; accepted 4 December 2014
S1-264 DOI 10.1179/1432891715Z.0000000001483 Materials Research Innovations 2015 VOL 19 SUPPL 1
 Zhang et al. The performance of Mg 2 Si-based thermoelectric materials prepared from MgH 2

raw materials and processes were handled in an argon- Results and discussion
filled glove box. The Si powders were initially ball-
milled for 8 hours with 15 minutes intervals in a high- Figure 1 shows the XRD patterns of the prepared pure
speed vibration ball mill (QM-3B, Nanjing Nanda Mg2Si, the 1% Bi-doped Mg2Si powders and Mg2Si1-
Instrument Factory) to obtain a grain size of less than xSnx samples. All the diffraction peaks can be indexed

100 nm. Furthermore, the ball-to-powder mass ratio to the face-centred cubic (fcc) antifluorite structure with
was 10:1. To avoid contamination during milling, soft the space group of Fm3m. Within the limit of detection
Teflon vials and hard zirconia balls were employed. For of XRD, there was no evidence of the presence of
the Bi-doped sample, the Si and Bi powders, with a MgO. Furthermore, the dopant atoms (Bi) are believed
mole ratio of 99:1, were milled in advance to enhance to enter substitutionally into the lattice of Mg2Si, which
the mixing. Then, this mixture and the MgH2 powders reside on the Si sublattice sites.10,11 The grain size of
were weighed in a molar ratio of 1:1·95 in a glove box, Mg2Si calculated from the XRD patterns was greater
and ball-milled for 30 minutes. The molar ratio of than 100 nm.
1:1·95 was used to avoiding excess of MgH2 for generat- The effect of milling the Si powder on the microstruc-
ing MgO, which acts against the TE performance of ture of the resulting Mg2Si is shown in the micrographs
Mg2Si. The ball-to-powder mass ratio was 5:1 for the of Fig. 2. With the unmilled Si (45 μm), the grain size
whole process. The use of different ball-to-powder mass of the resulting silicide is relatively large, about 2 μm. In
ratios for pure Si and the mixture was in order to contrast, the grain size of Mg2Si prepared with the
obtain a fine grain Si in a short time, and to only mix milled Si is relatively small, about 500 nm. This obser-
the powders for the second case. After milling, the vation is significant, because improving the ZT value
mixture was poured into an alumina boat and placed through a reduction of thermal conductivity has been
inside a quartz tube. The tube was then sealed with a stop- reported for TE materials with a nanostructure.12 The
cock adapter on each end, and placed in a tube furnace. grain size of the Mg2Si obtained in this work, 500 nm,
The reaction was carried out at 623 K for 20 hours is much smaller than that reported in other literature,
under a flow of Ar gas. The powder product was placed typically in the tens of micrometres. The density was
in a 20 mm diameter graphite die and sintered by the measured by the Archimedes method, and the theoretical
FAPAS apparatus. Details of the apparatus have been density of Mg2Si is 1·98 g cm−3, whereas the density of
reported in previous publications.9 the samples had an average value of 2·074 g cm−3,
An X-ray powder diffraction (XRD, A-6000, Cu, which is a little larger than the theoretical one. For the
λ=1·54060 Å) and scanning electron microscopy (SEM, pure Mg2Si, the ratio of MgH2 and Si (1·95:1) turned
Philips FEIXL30-SFEG) were utilised to determine the the excess Si nanodots into an impurity for the sample
phase composition and microstructure of the Mg2Si and introduced lattice distortion, which increased the
powders after the solid-state reaction. The grain size of density of the samples. For the Bi-doped sample, the
the resulting pure and doped Mg2Si samples was approxi- dopant atoms (Bi) enter substitutionally into the lattice
mately 500 nm, as determined from the SEM images. of Mg2Si, which reside on the Si sublattice sites,10 result-
The TE properties were determined from various ing in an increase in density.
measurements. The Seebeck coefficient and electrical Figure 3 shows the temperature dependence of the elec-
conductivity were measured by the Seebeck coefficient/ trical conductivity of the pure Mg2Si, Mg2Si1-xSnx and 1%
electric resistance measuring system (ZEM-1, ULVAC Bi-doped Mg2Si samples prepared in this work, and the
Inc., Japan). Furthermore, a temperature difference of corresponding values reported elsewhere in the literature
about 4°C between the cool and hot ends of the sample for materials prepared from Mg and Si.9,13 The references
was used for the Seebeck coefficient measurement. The are used for the comparison of all the TE properties in this
thermal conductivity was calculated by k = αCpρ, where report. Figure 3 shows a gradual increase of the electrical
α is the thermal diffusivity, ρ is the sample density and conductivity, with increasing temperatures over the range
Cp is the specific heat capacity. The parameters α and of 300–750 K for samples prepared in this study. The
Cp were measured on an NETSCH LFA457/ DSC404, values reported in the literature are nearly constant with
and ρ was measured by the Archimedes method. these temperatures. Furthermore, an increase in the temp-
erature is seen to lead to an increase in electrical

1 The XRD patterns of pure Mg2Si, MgSi1-xSnx and 1% Bi-doped Mg2Si

Materials Research Innovations 2015 VOL 19 SUPPL 1 S1-265

Zhang et al. The performance of Mg 2 Si-based thermoelectric materials prepared from MgH 2

2 The SEM images of Mg2Si powders: a Mg2Si prepared with unmilled Si powders; b Mg2Si prepared with milled Si powders

3 The temperature dependence of the electrical conductivity of Mg2Si, Mg2Si1-xSnx and 1% Bi-doped Mg2Si compared with the
values reported in the literature4,9,14

conductivity, which is consistent with the behaviour of a and n the carrier concentration. From Table 1, it can be
non-degenerate semiconductor, as seen in Fig. 3a. When seen that the carrier concentration is increased by
the temperature reaches 700 K, the exhaustion region is doping the 1% Bi, thus the Seebeck coefficient of the
reached and the electrical conductivity for the doped 1% Bi-doped sample is lower than the pure sample.
samples also reaches its maximum value. From Figures 3b and 4b, it can be seen that no obvious
The results of the Seebeck coefficient measurements of the advantages can be found in the MgSi1-xSnx samples pre-
samples prepared in this study are shown in Fig. 4 along with pared from the MgH2 powder when compared with other
the corresponding results from published studies.4,9,14 The samples prepared from the Mg powder. The decrease in
negative values of the coefficient indicate n-type conduction. MgO impurity will increase the carrier concentration,
For the undoped Mg2Si sample, the absolute value of the and the reduction in the grain size will decrease the mobi-
Seebeck coefficient first increases to a maximum absolute lity. The increase in the carrier concentration is beneficial
value of 475 μV K−1 at 500 K with an increasing tempera- to electrical conductivity, but harmful to the Seebeck
ture, and then decreases. This trend is consistent with that coefficient. Alternatively, the decrease in mobility is ben-
reported by others for undoped Mg2Si13 (not shown in eficial to the Seebeck coefficient, but harmful to electrical
Fig. 4). For the 1% Bi-doped sample, the absolute value of conductivity. Therefore, owing to the double effect, there
the Seebeck coefficient increases with the increasing temp- is no big difference in the electrical properties between the
erature, with values nearly identical to those in the cited samples prepared from the MgH2 powder and those pre-
references. Expectedly, the absolute values of the Seebeck pared from the Mg powder.
coefficient of the 1% Bi-doped samples are lower than that
of undoped samples. From equation (2), it can be found
that the Seebeck coefficient is inversely proportional to the
carrier concentration Table 1 The electrical properties of Bi-doped and undoped

 Mg2Si at 300 K
8π2 κ2B π
S= m ∗ T 2/3 (2) Electrical
3eh2 3n Carrier Mobility conductivity
conc. (cm−3) (cm2 V−1·s−1) (103 sm−1)

Here, kB is the Boltzmann constant, h the Planck con- Mg2Si 3·496×1017 239·9 7·12961
Mg2Si+1%Bi 4·651×1020 38·12 11·21898
stant, m* the effective mass, T the Kelvin temperature

S1-266 Materials Research Innovations 2015 VOL 19 SUPPL 1

 Zhang et al. The performance of Mg 2 Si-based thermoelectric materials prepared from MgH 2

4 The temperature dependence of the Seebeck coefficient as a function of the temperature of Mg2Si, Mg2Si1-xSnx and 1% Bi-
doped Mg2Si, compared with the values reported in the literature4,9,14

5 The temperature dependence of the thermal conductivity for Mg2Si, Mg2Si1-xSnx and 1% Bi-doped Mg2Si, compared with the
values reported in the literature 4,9,14

Figure 5 shows the temperature dependence of the can scatter more phonons, which transport heat in
thermal conductivity for Mg2Si, MgSi1-xSnx and 1% Bi- solids. Therefore, the samples prepared from MgH2
doped Mg2Si compared with the references.4,9,14 It can have an especially lower thermal conductivity.
be seen from Fig. 5 that the thermo conductivity of all The figure of merit, ZT, of the samples, calculated from
the samples prepared from the MgH2 powder is much the measured transport parameters is shown in Fig. 6 as a
better than that of the samples prepared from the Mg function of temperature. Owing to the excellent thermo
powder. In particular for the Mg2Si1-xSnx samples, the conductivity, the highest ZT value is increased by 50%
thermo conductivity reduced by 28% for Mg2Si0·4Sn0·6 for MgSi0·4Sn0·6 and 180% for MgSi0·3Sn0·7.
and 32% for Mg2Si0·3Sn0·7. The reason is first that the Furthermore, the value of ZT increased with the increas-
decrease in MgO impurity is helpful at lowering the ing temperature. The higher ZT value for the undoped
thermal conductivity; and second, the nanosize grain Mg2Si at low temperatures (relative to the doped

6 The temperature dependence of the figure of merit, ZT, for Mg2Si and 1% Bi-doped Mg2Si, compared with the values reported
in the literature 4,9,14

Materials Research Innovations 2015 VOL 19 SUPPL 1 S1-267

Zhang et al. The performance of Mg 2 Si-based thermoelectric materials prepared from MgH 2

samples) is the consequence of a higher Seebeck coeffi- References

cient. However, there is a contribution from the carrier
1. L. E. Bell: ‘Cooling, heating, generating power, and recovering waste
concentration and hence conductivity becomes dominant heat with thermoelectric systems’, Science, 2008, 321, 1457–1461.
with the increasing temperature. At 750 K, the value of 2. G. J. Snyder, T. Caillat and J. P. Fleurial: ‘Thermoelectric properties
ZT exceeded that of the undoped sample, reaching a of chalcogenides with the spinel structure’, Mater. Res. Innov., 2001,
maximum of 0·53. This value is slightly higher than 5, (2), 67–73.
that reported by Tani and Kido, 0·50,11 and significantly 3. H. L. Gao, T. J. Zhu, X. X. Liu, L. X. Chen and X. B. Zhao: ‘Flux
synthesis and thermoelectric properties of eco-friendly Sb doped
higher than the value (0·37) reported by Choi et al.13 A Mg2Si0·5Sn0·5 solid solutions for energy harvesting’, J. Mater.
modelling study of the TE properties of Mg2Si predicted Chem., 2011, 21, (16), 5933–5937.
that the highest ZT value for n-type Mg2Si at 850 K is 4. Q. Zhang, H. Yin, X. B. Zhao, J. He, X. H. Ji, T. J. Zhu and T. M. Tritt:
0·75, with a carrier concentration of 5·3×1019 cm−3. ‘Thermoelectric properties of n-type Mg2Si0·6-ySbySn0·4 compounds’,
Phys. Status Solidi A-Appl. Mater. Sci., 2008, 205, 1657–1661.
According to this modelling study, a maximum ZT 5. N. Satyala, J. S. Krasinski and D. Vashaee: ‘Simultaneous enhance-
value of 0·3 at 850 K would be expected for n≈5×- ment of mechanical and thermoelectric properties of polycrystalline
1020 cm−3 (approximately the value obtained in this magnesium silicide with conductive glass inclusion’, Acta Mater.,
work). The highest temperature for the ZT value obtained 2014, 74, 141–150.
in this work, 0·53, is 750 K. When the ZT curve in Fig. 6 6. T. Jun-ichi and K. Hiroyasu: ‘Thermoelectric properties of Sb-doped
Mg2Si semiconductors’, Intermetallics, 2007, 15, 1202–1207.
is extrapolated to 850 K, a value of 0·8 is obtained. This is 7. Q. Zhang, X. B. Zhao, H. Yin and T. J. Zhu: ‘Thermoelectric perform-
higher than the predicted value is likely to be a conse- ance of Mg2-xCaxSi compounds’, J. Alloys Compounds, 2008, 464, 9–12.
quence of the effect of the microstructure, the small 8. J. S. Marcus Schmidt, Y. Grin, A. Boehm, B. Kieback, T. Scholl, T.
grain size and its effect on the thermal conductivity.15 Weissgaerber and M. Zumdick, International Patent WO02083561,
2002.
9. Q. S. Meng, W. H. Fan, R. X. Chen and Z. A. Munir:
‘Thermoelectric properties of Sc- and Y-doped Mg2Si prepared by
field-activated and pressure-assisted reactive sintering’, J. Alloys
Compounds, 2011, 509, 7922–7926.
Conclusions 10. M. Akasaka, T. Iida, A. Matsumoto, K. Yamanaka, Y. Takanashi,
T. Imai and N. Hamada: ‘The thermoelectric properties of bulk crys-
In this paper, high purity, nanoscale (500 nm) Mg2Si- talline n- and p-type Mg2Si prepared by the vertical Bridgman
based and Mg2Si1-xSnx thermoelectric materials were pre- method’, J. Appl. Phys., 2008, 104, 013703.
pared from MgH2, Si and Sn, and then sintered by the 11. S. Fiameni, S. Battiston, S. Boldrini, A. Famengo, F. Agresti, S.
field-activated and pressure-assisted sintering process Barison and M. Fabrizio: ‘Synthesis and characterization of Bi-
doped Mg2Si thermoelectric materials’, J. Solid State Chem., 2012,
method. This synthesis was judged to be preferable to 193, 142–146.
that of the direct reaction between Si, Sn and Mg, as it 12. S. W. You and I. H. Kim: ‘Solid-state synthesis and thermoelectric
minimises the deleterious effects associated with the properties of Bi-doped Mg2Si compounds’, Curr. Appl. Phys.,
latter method, namely the oxidation and vaporisation of 2011, 11, S392–S395.
13. M. Ioannou, G. S. Polymeris, E. Hatzikraniotis, K. M.
Mg. All prepared samples got an extra low thermal con-
Paraskevopoulos and T. Kyratsi: ‘Effect of Bi-doping and Mg-
ductivity in the temperature range of this study. The ZT excess on the thermoelectric properties of Mg2Si materials’,
value of 1% Bi-doped Mg2Si at 750 K was 0·53. Using J. Phys. Chem. Solids, 2014, 75, (8), 984–991.
MgH2 instead of Mg (references) leads to a better ZT at 14. W. Liu, K. Yin, X. Su, H. Li, Y. Yan, X. Tang and C. Uher: ‘Enhanced
high temperatures. Furthermore, the ZT value of the hole concentration through Ga doping and excess of Mg and
thermoelectric properties of p-type Mg2(1+z)(Si0·3Sn0·7)1- yGay’,
Mg2Si0·4Sn0·6 and Mg2Si0·3Sn0·7 samples are respectively Intermetallics, 2013, 32, 352–361.
1·5 and 2·8 times as great as the values of the samples pre- 15. J. Tani and H. Kido: ‘Thermoelectric properties of Bi-doped Mg2Si
pared from the Mg powder. semiconductors’, Physica B, Condensed Matter, 2005, 364, 218–224.

S1-268 Materials Research Innovations 2015 VOL 19 SUPPL 1