Optimized thermoelectric performance of the n-type half-Heusler material TiNiSn by substitution and addition of Mn

Energy consumption and efficiency is increasing continuously, but in many industrial processes and consumer applications, a significant amount of energy is lost as waste heat.¹ The interest in recycling waste heat into electricity makes thermoelectric generators one of the intensive research fields. In contrast to small-scale devices, for waste-heat recovery applications abundant, cheap and environmentally-friendly materials are required. Thermoelectric materials have been explored and developed over time to reach technological requirements for various use, depending on the operating temperature scale.² The efficiency of a single thermoelectric material, the dimensionless figure of merit, ZT, is defined by the interdependent parameters σ (electrical conductivity), S (Seebeck coefficient), κ (thermal conductivity) and T (absolute temperature).

Among various materials, narrow band-gap half-Heusler compounds with 18 valence electrons became promising thermoelectric materials³ e.g., with n-type conduction MNiSn, (M = Ti, Zr, Hf) or with p-type conduction MCoSb (M =Ti, Zr, Hf). They exhibit high electrical conductivity and Seebeck coefficient, have good thermal stability and consists of low-cost environmentally friendly constituents.^1,4–6 Half-Heusler compounds have the general formula XYZ, and crystallize in a face-centered cubic crystal structure (MgAgAs structure type) with the Wyckoff positions 4a (0, 0, 0), 4b (1/2, 1/2, 1/2), and 4c (1/4, 1/4, 1/4) occupied.⁷ The main obstacle for thermoelectric applications is their too high thermal conductivity originating from the high lattice thermal conductivity.^8–10 

Many strategies have been applied to enhance the performance of half-Heusler compounds, such as adding nanoparticles to bulk material,¹¹ nano-structuring by ball milling followed by spark plasma sintering or hot pressing,^12–16 phase separation induced through isoelectronic alloying to have increased point defect, boundary phonon scattering^17–19 and systematic doping studies^20–22 for yielding an optimized power factor. Especially phase separation reduces the thermal conductivity at the cost of a minimal effect on power factor, which results in a better figure of merit.

It is reported that the attempt of substituting Ti for Mn in Co₂MnSn leads to the formation of two compounds, Co₂MnSn and Co₂TiSn, and the grain boundaries formed thereby are beneficial to increase phonon scattering.²³ In the present work it is demonstrated that phase separation in TiNiSn alloys can also be induced by the substitution of Ti by Mn or the addition of Mn. Both alloying strategies result in a significant reduction of the thermal conductivity and an increase of the electrical conductivity.

The two series of compounds Ti_1-xMn_xNiSn and TiMn_yNiSn (x or y = 0, 0.01-0.05, 0.10, 0.15, 0.20) were prepared by arc melting of stoichiometric amounts of Ti, Mn, Ni and Sn (all with purities >99.9%) in argon atmosphere. Samples were flipped at least three times with intermediate crushing to ensure homogeneity. The as-cast samples were annealed at 900°C for 3 days or at 800°C for 7 days in evacuated quartz ampoules (details are given in the supplementary material). For higher Mn concentrations (x or y = 0.15, 0.20) tantalum ampoules were used to avoid Mn reacting with the quartz ampoules. Following the annealing, the samples were quenched in ice-water. The obtained ingots were cut into platelets and bars for transport measurements.

The crystal structure was characterized by powder x-ray diffraction (XRD, Huber imaging plate, Guiner camera G670) with Cu Kα₁ radiation and Ge (111) monochromator at room temperature. Microstructure and composition was analyzed using inductively coupled plasma optical emission spectrometry (ICP-OES 5100 SVDV, Agilent Technologies) and scanning electron microscopy (SEM; JEOL 7800F, field emission gun cathode) with energy dispersive x-ray spectroscopy (EDXS; Bruker, silicon drift detector).

The pycnometric density was measured using an AccuPyc 1330 device in helium gas atmosphere. The chamber was flushed with helium gas several times before measurement. The values given are averages of 20 measurements on each sample. Specific heat capacity was determined by a Netzsch 404C differential scanning calorimeter (DSC) using Ar atmosphere and with a heating rate of 20°C/min up to 500°C.

Thermal diffusivity was determined using the laser-flash method (Netzsch LFA 457). The Seebeck coefficient (S) and the electrical conductivity (σ) were measured simultaneously (ZEM 3, ULVAC) under helium gas atmosphere up to 900K.

The XRD patterns of all samples show the C1_b structure as main phase. On substituting Mn on the Ti site, Mn-containing impurity phases appear from 3% Mn concentration, which is in agreement with literature.²⁴ LeBail fitting using the Jana2006 software²⁵ gives the lattice parameters 5.929Å, 5.921Å and 5.927Å for the 0%, 1%, 2% nominal substitution samples, respectively. Taking into account that Mn has a smaller atomic radius than Ti, lattice parameters get smaller upon successful substitution. The larger lattice parameter observed at 2% than at 1% attributes to its analyzed Mn concentration of 0.2% revealed by ICP-OES analysis (see supplementary material). The powder XRD patterns are also given in the supplementary material. Phase analysis is performed using XRD data as well as SEM-EDXS and ICP-OES analyses. It is seen that there is no clear correlation between Mn adding or substitution concentration (x or y) and the appearance of impurity phases. Small amounts of β-Sn were present in the majority of the samples. The highest β-Sn phase amount of 2.6 % is found in the Ti_0.85Mn_0.15NiSn sample.

A LeBail fit of the powder XRD data of TiMn_0.02NiSn (i.e. y = 0.02; see supplementary material) indicates that besides the main phase the sample contains Sn and MnSn₂ impurity phases which are also identified in the backscattered electron SEM images (cf. Fig. 1a).

In Fig. 1a, based on EDXS analyses, the gray colored region corresponds to the TiNiSn phase, whereas darker grains are Ti-rich phase and the light colored regions are MnSn₂ or Sn impurities. There are small black dots in the whole region which are identified as titanium oxide precipitations in element-specific EDXS mappings (Fig. 1b). These spots have diameters in the range of 1 μm. However, no oxygen content could be detected in regular EDXS analyses.

In the TiMn_yNiSn series, increasing Mn concentration enhances the phase separation (Fig. 2d–f). However, in the Ti_1-xMn_xNiSn series, phase separation does not correspond linearly to Mn concentration (Fig.2a–c). The sample Ti_0.85Mn_0.15NiSn has regular sized grains compared to Ti_0.8Mn_0.2NiSn. This feature leads to higher electrical conductivity and lower Seebeck coefficient than for Ti_0.8Mn_0.2NiSn (Fig. 3). However, the two samples have similar thermal conductivity (Fig. 4a1). Unlike other samples, Ti_0.98Mn_0.02NiSn shows uniformly dispersed Sn-rich dendritic phase (Fig. 2a). Its morphological effect is discussed below.

The electrical conductivity, Seebeck coefficient and power factor for the samples of the two series are plotted as function of temperature in Fig. 3. All samples show semiconducting behavior, i.e. the conductivity increases linearly as temperature increases. The most dramatic change occurs for 2% Mn substitution (x = 0.02), where the electrical conductivity increases two fold compared to TiNiSn at T = 800 K. However, this unusual increase is attributed to its morphology which differs quite strongly from the rest of the series. The present dendritically formed Sn-rich metallic phase favors carrier conductivity effectively, which results in higher electrical conductivity and lower Seebeck coefficient. This morphological effect to the transport properties can be explained via general effective media (GEM) approach.²⁶ 

Also adding high amounts of Mn (y = 0.10, y = 0.15 and y = 0.20) leads to a less temperature-dependent behavior of the electrical conductivity. Similar results are reported when adding 10 or 15 % Ni to TiNiSn.²⁷ In all these cases this is probably due to the increased amount of impurity phases that contribute to electrical conductivity. The element-specific EDXS mappings and microstructure images of TiMn_0.15NiSn are given in the supplementary material.

All samples show negative Seebeck coefficient values in the whole investigated temperature range, which indicates that the majority of the charge carriers are electrons (n-type). Substitution with Mn changes the Seebeck coefficient significantly, while adding Mn does not change it much until y = 0.10. For this level of adding the electrical conductivity increases, but the Seebeck coefficient decreases due to inverse relations between S (Seebeck coefficient) and n (electron carrier concentration),² 

where m* is an effective mass. The highest Seebeck coefficient reached is -249 μV K^-1 for Ti_0.97Mn_0.03NiSn and -242 μV K^-1 for Ti_0.95Mn_0.05NiSn at 566 K. In both series, the maximum power factor occurs in the temperature range 700-800 K. The highest value of 3.6 mW K^-2m^-1 (Fig. 3c1) is attained in Ti_0.95Mn_0.05NiSn at 712 K. The spurious peak in power factor is ascribed to decomposition of a minor Mn₃Sn₂ impurity phase. It is reported that, Mn₃Sn₂ lies in narrow homogeneity range and decomposes peritectoidally into Mn_2-xSn and MnSn₂ upon heating at 753K.²⁸ 

The thermal conductivity of the samples (Fig. 4a) is calculated using κ = DC_pd, where C_p is the specific heat capacity, d the density of the samples, and D the thermal diffusivity. In all samples, Mn addition as well as substitution decreases the thermal conductivity roughly according to Mn concentration. This implies enhanced phonon scattering in the phase and at the grain boundaries. The lowest thermal conductivities are found for 10% Mn (x = 0.1 or y = 0.1) in both series. The values are 4.4 W K^-1m^-1 in TiMn_0.1NiSn and 4.7 W K^-1 m^-1 in Ti_0.9Mn_0.1NiSn which are significantly lower than 8.0 W K^-1m^-1 in TiNiSn at room temperature. Some samples show small cusp in thermal conductivity at 500K implying the sign of β-Sn impurity phase.

The figures of merit ZT are presented in Fig. 4b. The highest figure of merit ZT = 0.42 is found in Ti_0.95Mn_0.05NiSn at 712 K and ZT = 0.43 in TiMn_0.2NiSn at 813 K. These values are 94% and 96% improvements, respectively, compared to TiNiSn. TiMn_0.02NiSn shows an enhanced figure of merit ZT = 0.38 in the temperature range of 711-837 K due to its improved Seebeck coefficient and reduced thermal conductivity.

In conclusion, modifying the half-Heusler compound TiNiSn by introduction of Mn is proven to be an effective way to reduce the lattice thermal conductivity and enhance the electronic properties for thermoelectric applications. As a result, the best materials yield almost twice higher figure of merit than the unmodified material. This work illustrates the possibility to enhance the figure of merit through phase separation induced in an alloy of cheap, abundant, and overall quite environmentally-friendly materials.