Nano- and Microstructure Engineering: An ... - ACS Publications

Nov 24, 2016 - Nano- and Microstructure Engineering: An Effective Method for ... Department of Chemistry and Waterloo Institute for Nanotechnology, ...
0 downloads 0 Views 2MB Size
Research Article www.acsami.org

Nano- and Microstructure Engineering: An Effective Method for Creating High Efficiency Magnesium Silicide Based Thermoelectrics Nader Farahi,† Sagar Prabhudev,‡ Gianluigi A. Botton,‡ James R. Salvador,§ and Holger Kleinke*,† †

Department of Chemistry and Waterloo Institute for Nanotechnology, University of Waterloo, Waterloo, ON N2L 3G1, Canada Materials Science and Engineering Department, McMaster University, Hamilton, ON L8S 4L8, Canada § General Motors Research & Development Center, Warren, Michigan 48154, United States ‡

S Supporting Information *

ABSTRACT: Considering the effect of CO2 emission together with the depletion of fossil fuel resources on future generations, industries in particular the transportation sector are in deep need of a viable solution to follow the environmental regulation to limit the CO2 emission. Thermoelectrics may be a practical choice for recovering the waste heat, provided their conversion energy can be improved. Here, the high temperature thermoelectric properties of high purity Bi doped Mg2(Si,Sn) are presented. The samples Mg2Si1−x−ySnxBiy with x(Sn) ≥ 0.6 and y(Bi) ≥ 0.03 exhibited electrical conductivities and Seebeck coefficients of approximately 1000 Ω−1 cm−1 and −200 μV K−1 at 773 K, respectively, attributable to a combination of band convergence and microstructure engineering through ball mill processing. In addition to the high electrical conductivity and Seebeck coefficient, the thermal conductivity of the solid solutions reached values below 2.5 W m−1 K−1 due to highly efficient phonon scattering from mass fluctuation and grain boundary effects. These properties combined for zT values of 1.4 at 773 K with an average zT of 0.9 between 400 and 773 K. The transport properties were both highly reproducible across several measurement systems and were stable with thermal cycling. KEYWORDS: magnesium silicide, thermoelectric, high efficiency, nanostructure, microstructure

worth understanding the origin of its low efficiency. Mg2Si has a fairly wide band gap of 0.77 eV6 with a low room temperature intrinsic carrier concentration of 1014 cm−3,6 which leads to a low electrical conductivity of around 0.001 Ω−1 cm−1 at 300 K. The highly symmetric antifluorite cubic structure with Fm3̅m space group, and also the absence of heavy elements together with the highly covalent bonds facilitates phonon transport, which results in high thermal conductivity of approximately 10 W m−1 K−1 at 300 K.7 To increase the carrier concentration, doping with group 15 elements such as phosphorus (P),8 antimony (Sb),9−16 and bismuth (Bi)17−23 which substitute for silicon (Si), act as electron donors and have been shown to be profoundly effective dopants. Among all the dopants substituting at the Si site, Sb and Bi seemed to be the most effective for improving the electrical conductivity through increasing the carrier concentration by 6 orders of magnitude to >1020 cm−3 at 300 K.20 As mentioned above, a significant amount of heat is carried through lattice vibrations in Mg2Si materials; therefore to reduce the thermal conductivity the best approach is to focus on reducing the lattice thermal conductivity (κl). Based on the kinetic theory, lattice thermal conductivity can be estimated as κl = 1/3 l ν Cv where l, ν, and Cv designate phonon mean free path, phonon group velocity, and specific heat at constant volume, respectively.24 Introducing germanium (Ge) or tin (Sn) into the Mg2Si structure has been shown to reduce both the phonon group velocity and mean free path, due to their heavier mass and larger size compare

1. INTRODUCTION With respect to the visible impact of fossil fuel consumption and greenhouse gas emissions on the climate and environment, we are now, more than ever, in critical need of advanced technologies to reduce their effects in the near future.1 The automotive sector, due to its global ubiquity and large fuel energy usage based on the current engines, has a very high potential to benefit from these emerging technologies to improve overall vehicle efficiency. In addition to enhancing the engine efficiency, developing an efficient and costeffective waste energy recovery mechanism seems to be a reasonable way to proceed.2,3 For the recovery of waste heat, thermoelectric (TE) materials were demonstrated to be a reliable and irreplaceable candidate.4 To be suitable for industrial scale applications, the materials need to be relatively cheap, made from elements that are abundant in nature, and also nontoxic to meet the environmental regulations. Last but not least, the materials should have sustained high efficiency to enhance energy recovery as much as possible. The efficiency of a TE material depends on the figure of merit, zT, via zT = TS2σ (κe + κl)−1, where T, S, σ, κe, and κl represent absolute temperature, Seebeck coefficient, electrical conductivity, and electronic and lattice thermal conductivity, respectively. An ideal, practical TE material should exhibit not only low thermal conductivity to maintain the temperature gradient with low heat flux, but also high electrical conductivity and Seebeck coefficient to facilitate charge transport to attain larger power output.5 Aside from its low efficiency, magnesium silicide fulfills almost all the aforementioned criteria for large scale applications. To find the most feasible strategy to improve the efficiency of this material, it is © 2016 American Chemical Society

Received: September 28, 2016 Accepted: November 24, 2016 Published: November 24, 2016 34431

DOI: 10.1021/acsami.6b12297 ACS Appl. Mater. Interfaces 2016, 8, 34431−34437

Research Article

ACS Applied Materials & Interfaces

Figure 1. SEM/EDX mapping of Mg2Si0.3Sn0.665Bi0.035 and Mg2Si0.365Sn0.6Bi0.035 samples. to Si. This alloying drastically reduced the lattice thermal conductivity of Mg2Si from ∼10 to ∼2.5 W m−1 K−1 for Mg2Si0.3Sn0.7.7,12 The localized atomic scale distortion generated by alloying is more effective for scattering short-wavelength phonons. To scatter a broader range of phonons other techniques such as nanostructuring, nanophase inclusion and grain boundary engineering could be combined with alloying to decrease lattice thermal conductivity beyond the alloy limit. For instance, inclusion of multiwalled carbon nanotubes (MWCNT) into Mg2Si0.877Ge0.1Bi0.023 reduced the low temperature thermal conductivity (T < 500 K) by 15%,25 while adding SiC had a similar effect, but only minor changes to the figure of merit.26 It did, however, enhance the mechanical properties of Sb-doped Mg2(Si,Sn).27 Nanostructuring in Si80Ge20 through ball mill processing decreased the thermal conductivity by around 40%,28,29 and Te coated grain boundaries in Bi0.5Sb1.5Te3.0 reduced κl by 50% as compared to the noncoated sample.30 The best performing alloys in the Mg2(Si,Ge,Sn) system include: Mg 2. 1 5 Si 0. 2 8 Sn 0. 7 1 Sb 0 . 0 06 , 3 1 Mg 2 . 16 (Si 0 .4 Sn 0 . 6 ) 0 . 97 Bi 0 . 03 , 2 1 and Mg2Si0.53Ge0.05Sn0.4Bi0.02,19 all demonstrating figures of merit between 1.2 and 1.4 near 800 K.32 The Si/Sn ratios of these phases are around the reported miscibility gap in the Mg2Si − Mg2Sn system, which may lead to inhomogeneous products and irreproducibility of properties. Previously, this gap was believed to occur between 0.4 Sn and 0.6 Sn, but it seemed to fluctuate based on the synthesis conditions, and was reported to vary between 0.2 Sn and 0.45 Sn.33,34 In an attempt to overcome the potential kinetic barriers to a completely homogeneous solid solution in the materials we report here, we utilize a two stage ball milling assisted synthesis method developed to produce Mg2Si1−xSnx solid solutions with stable microstructure, that are oxide free, within the postulated forbidden composition range. Although the powder X-ray pattern of all samples were single phase solid solutions, the nanostructure and the composition of grain boundaries were examined to shed more light on possible nanoscale inhomogeneities of the composition that is supposed to be in the so-called miscibility gap. Finally, the reliability and the reproducibility of the presented data were assured through consecutive measurements.

under vacuum. The tubes were heated in a resistance furnace at 923 K for 1 week and thereafter for 1223 K for another week. The heat treated samples were ground into powders and dry ball milled for 2 h under inert atmosphere by using Fritzsch Pulverisette 7 Premium planetary mill. To consolidate the powders to near full density for physical property measurements, an Oxy-Gon hot press was used to densify the mixtures in an 95% Ar - 5% H2 atmosphere at a maximum temperature of 973 K under 56 MPa. The applied pressure was released during cooling, to mitigate stress and strain on the pellets. The pressed pellets were 2 mm thick and 12.7 mm in diameter, and were 98% of the theoretical densities as determined via the Archimedes method. An Inel powder X-ray diffractometer with Cu Kα radiation and a position sensitive detector was used to examine the purity of the pressed samples; the diffraction patterns (Supporting Information, Figure S1) revealed no traces of MgO, which is a typical side product in Mg2Si based compounds that is challenging to avoid. The four samples with x(Sn) = 0.4, 0.6, 0.665, and 0.67 and y = 0.03−0.035 Bi per formula unit were prepared phase-pure within detection limits of the instrument. Thermal diffusivity (α) was measured, between 300 and 800 K, under flowing argon using the Anter Flashline FL3000 thermal properties analyzer. The diffusivity values were then multiplied by the density (ρ) of the pellets, and the specific heat (Cp) of the compounds, as calculated from the Dulong-Petit approximation, to obtain thermal conductivity (κ), κ = α ρ Cp. The validity of using the Dulong−Petit approximation for the high temperature (above 400 K) specific heat of Mg2Si based materials was examined previously,22,35 and it was established to be a reliable approximation for the thermal conductivity calculation in this class of materials. For measuring the electrical conductivity (σ) and Seebeck coefficient (S), the consolidated pellets used in the thermal diffusivity measurements were subsequently cut into rectangular bars with the dimensions of roughly 12 × 2 × 2 mm. The measurements were performed under a helium atmosphere between 300 and 800 K by using the ULVAC-RIKO ZEM-3 apparatus. To confirm the results, a second set of measurements was performed using a Linseis LRS-3 system. Estimated experimental errors are 3% for the Seebeck coefficient, and 5% both for the electrical and thermal conductivity, which results in an error of about 10% for the figure-of-merit.36 Hall effect measurements and four probe resistivity measurements were made with a cryostat equipped with a 5.0 T magnet and a Linear Research AC resistance bridge. Hall data was collected with both a positive and negative field (−3 to +3 T) to account for probe misalignment. Estimated errors are again 5% for the conductivity and

2. EXPERIMENTAL SECTION Bismuth doped Mg2Si1−xSnx samples were synthesized by mixing the stoichiometric ratios of elements in an argon filled glovebox. Mg chips (99.98%, Sigma-Aldrich, 4−30 mesh), Si powder (99.9%, Alfa Aesar, −100 mesh), Sn granules (99.9%, Alfa Aesar, ≤ 2 cm), and Bi granules (99.99%, Sigma-Aldrich) were used for the synthesis. The elements were put in tantalum crucibles, which were sealed under argon in an arc melter and then placed into silica tubes that were then flame-sealed 34432

DOI: 10.1021/acsami.6b12297 ACS Appl. Mater. Interfaces 2016, 8, 34431−34437

Research Article

ACS Applied Materials & Interfaces 3% for the carrier concentration, corresponding to an error of 6% for the mobility. To investigate the homogeneity of the samples at the micron scale, scanning electron microscopic (SEM) analysis was performed on consolidated pieces using a Zeiss ULTRA electron microscope equipped with an EDAX Pegasus 1200E. The transmission electron microscopy (TEM) sample was prepared by focused ion beam (FIB) using a Zeiss NVision 40 instrument. A FIB liftout was prepared and milled to a final thickness of ∼20 nm. The sample was cleaned from hydrocarbon using a hydrogen−oxygen plasma for 2 min (Gatan Solarus plasma cleaner). The sample was imaged in a FEI Titan 80− 300 cubed microscope, equipped with a CEOS image and probe corrector. The point resolution of the microscope is