2025 Volume 66 Issue 9 Pages 1114-1120
In this study, we developed a method for reproducibly fabricating high-performance nano-grained bulk Si-Ge thermoelectric materials free from severe oxidization. In our previous work, the oxidization of Si-Ge during mechanical alloying and sintering processes had led to poor reproducibility of the value of electrical resistivity. We found that co-sintering with Ti, which is more easily oxidized than Si and Ge near the sintering temperature, effectively reduces the oxygen concentration in the nano-grained bulk Si-Ge samples. The oxygen concentration in the sample co-sintered with Ti was found to be less than 2.4 at%, and electrical resistivity was found to be less than 3.9 mΩ cm at 922 K with good reproducibility. High Seebeck coefficient (more than 400 µV K−1) and low thermal conductivity (less than 1 Wm−1K−1) were simultaneously achieved by constructive electronic structure modification via iron doping and nano-crystallization, respectively. As a consequence, we succeeded in obtaining a surprisingly large value of dimensionless figure of merit, ZT = 4 at 922 K, and the temperature range of ZT exceeding 1 extended at high temperatures above 700 K.
This Paper was Originally Published in Japanese in J. Thermoelec. Soc. Jpn. 21 (2025) 141–146.
Si-Ge thermoelectric materials are attracting considerable interest because of their moderately good thermoelectric performance together with relatively inexpensive, abundant, and non-toxic nature. These materials are capable of possessing a high ZT of approximately unity at high temperatures of around 800 K. In the 1970s, NASA employed Si-Ge thermoelectric materials as the power-generating components in radioisotope thermoelectric generators (RTGs) space exploration probes. At that time, due to high thermal conductivity, the maximum ZT of Si-Ge thermoelectric materials was less than 0.5 for p-type and less than 0.9 for n-type [1]. Recently, nano crystallization [2] and composite effects [3–5] have succeeded in increasing the ZT of p-type and n-type Si-Ge thermoelectric materials to 0.95 [6] and 1.3 [7], respectively. However, these values are still insufficient for practical applications, and further performance enhancements have been strongly required for these materials to be widely used in our modern life as one of the key technologies leading to the low-energy-consuming, sustainable society.
Recently, we developed nano-grained bulk Si-Ge thermoelectric materials containing small amount of 3d transition metal impurity. The use of nano-grained bulk works especially at high temperatures because the nano-grained structure naturally increases the electrical resistivity at low temperatures by reducing the mean free path of electrons, but does not increase the electrical resistivity at high temperatures where the electron mean free path has become shorter than the grain size due to the strong phonon scatterings. The 3d transition metal impurity is used to constructively modify the electronic structure near the band edge for realizing a high-power factor. This concept is explained as “electron glass & phonon glass with a sharp peak in the electronic density of states near the band edge” [8].
To fabricate the nano-grained Si-Ge alloys having a significantly reduced lattice thermal conductivity κlat, we employed (a) nano-crystallization using a high-energy planetary ball mill and (b) “low-temperature & high-pressure sintering” to prevent the grain growth during the sintering process. The electronic structure was constructively modified by use of a small amount of Fe, for which DFT calculations predicted to produce an intense, narrow peak in the density of states near the edge of conduction band [8]. As a result, we successfully developed an n-type material that possesses ZT = 1.88 at 873 K.
Initially, it was considered that the working temperature of our nano-grained Si-Ge alloys was limited by the sintering temperature, because the sample was supersaturated solid solution. However, the behavior of increasing ZT with increasing temperature near 873 K promoted us to explore the thermoelectric properties of the nano-grained Si-Ge alloy at temperatures higher than 873 K. As a consequence, we discovered a trend that ZT reaches ZT = 3.6 at 1073 K [9].
After possessing ZT = 3.6 at 1073 K, the nano-grained Si-Ge alloy showed grain growth and a precipitation of crystalline phase containing Fe and P by the higher temperature annealing at above 1073 K. As a result, the absolute value of the Seebeck coefficient was significantly reduced and the lattice thermal conductivity was increased. Both behaviors naturally led to an irreversible decrease in ZT. In such a condition, the measured thermoelectric properties became nearly identical to those of the samples sintered with normal pulsed current sintering (PCS) process with 50 MPa at 1273 K. Based on these experimental findings, we conclude that the nano-grained bulk Si-Ge thermoelectric materials produced by “low-temperature & high-pressure sintering” stays in a metastable state, but their stability is sufficiently good to be used in practical applications working below 1000 K.
We realized, however, that severe oxidization often occurs to increase the electrical resistivity and drastically decrease the magnitude of ZT. The oxidization would occur most likely during the nano-crystallization process and/or sintering process. A thin oxide layer is supposed to form on the surface of the Si-Ge nanoparticles to increase the electrical resistivity of the samples, while it has less impact on Seebeck coefficient because of the small volume fraction. The oxygen concentration varied widely with the sample, and consequently, we observed significant variations both in electrical resistivity and ZT [10]. The poor reproducibility in thermoelectric performance have prevented us from utilizing the nano-grained Si-Ge thermoelectric materials in practical applications.
To solve this problem, we considered, more recently, that the use of oxygen getter during the sintering process may effectively reduce the oxygen concentration in nano-grained Si-Ge samples. We selected elements of their formation energy of oxide negatively high. During the sintering, we considered that the oxygen getter element is supposed to effectively eliminate the oxygen from the surface of the Si-Ge nanoparticles to reduce the oxygen concentration in the samples.
One of the most plausible oxygen getters is Titanium, because it shows inert nature to silicon and germanium at lower temperature. It also works as electrode provided that the metallic nature is kept remained after oxidation. Therefore, in this study, we selected Ti as the oxygen getter element.
To our best of knowledge, the suppression of oxidation by co-sintering the sample with Ti in PCS process has never been reported previously. Therefore, this study represents a novel, potential approach for effectively reducing oxygen concentration in variety of nano-particle sintering processes. We report in this paper that the electrical resistivity was significantly decreased to 3.9 mΩ cm to realize ZT = 4 at 922 K, that is the highest value ever reported for single-phase bulk materials.
We selected the sample composition (Si55Ge35P10)99Fe1 in the same manner as that in our previous studies in which ZT values of 1.88 (900 K) [8] and 3.6 (1073 K) [9] were successfully observed. High-purity Si (purity 99.99%), Ge (purity 99.99%), and P (purity 99.99%) purchased from Kojundo Chemical Lab. Co., Ltd. were well mixed together using a set of mortar and pestle in a glove box filled with an Ar atmosphere after evacuating to less than 5 × 10−3 Pa, and then milled under an atmosphere of 70 vol% Ar + 30 vol% H2 using a high-energy planetary ball mill (PULVERISETTE 7 classic line by FRITSCH). A stainless steel container was used along with 10 g each of stainless steel balls with diameters of 5 mm and 3 mm. Delime-Codrin et al. [8] reported, when they succeeded in preparing Si-Ge samples possessing ZT = 1.88, that approximately 1 at% Fe contamination from the stainless steel balls and container into the samples took place due to the friction between the samples to constructively modify the electronic structure near the edge of conduction band. Therefore, by using the same technique as that of Delime-Cdrin et al., we did not add Fe to the initially mixed powders.
The milling process has two stages: the first one for alloying the pure elements with strong milling-power and the other for homogeneous nano-crystallization using rather mild condition: the first stage was conducted at 600 rpm for 60 minutes with a pause time of 1 minutes, repeated 10 times, while the second stage at 300 rpm for 30 minutes with a pause time of 15 minutes, repeated 20 times. The milled powder samples were loaded into the die and punches inside the argon-filled glovebox.
Relatively low sintering temperature (848 K) with relatively high pressure (350 MPa) were used with a very long sintering duration (180 min.) for maintaining the nano-grains in the densified bulk. To avoid the serious oxidation during the sintering, Ti, which has a higher absolute value of the standard Gibbs free energy of oxidation than Si [11], was placed above and below the Si-Ge nano-powder in the form of compressed powder. The Ti powder (purity 99.9%) of 45 µm in average particle diameter purchased from Kojundo Chemical Lab. Co., Ltd. were co-sintered with the Si-Ge sample using pulsed electric current sintering (PCS) under low vacuum conditions of approximately 10−3 Pa. In order to eliminate the residual oxygen in the sintering atmosphere, the sintering chamber was flushed five times with pure N2 gas before the sintering process. We have to emphasize here, before passing, that the titanium used for the co-sintering can also serve as an electrode in the thermoelectric device. Therefore, it is considered that the co-sintering of Si-Ge with Ti is one of the promising approaches to prepare nano-grained, bulk, high-performance, thermoelectric Si-Ge due to two important effects: preventing oxidization and potentially serving the electrode.
The crystal structure of the synthesized sample was analyzed by conventional powder X-ray diffraction (XRD) using a Bragg-Brentano X-ray diffractometer and Cu-Kα radiation (40 mA, 40 kV) in Bruker D8 ADVANCE. The average crystallite size of the bulk sample was estimated using the Scherrer formula and subsequently confirmed directly using a transmission electron microscope (TEM). The sample density d was measured using the Archimedes method.
The chemical composition was investigated by energy-dispersive X-ray spectrometry (EDX) equipped with a scanning electron microscope (SEM). The measurement equipment was a Hitachi SU6600 scanning electron microscope, and the acceleration voltage was 15 kV.
The temperature dependence (300–1000 K) of the constant pressure specific heat Cp(T) and the thermal diffusivity α(T) of the sample were measured by using the laser flash method (NETZSCH LFA457). The thermal conductivity κ(T) was calculated from the obtained values using the following equation, where the density d, which was determined by the Archimedes method with pure ethanol, is used to convert the specific heat per unit mass (Cp) into the specific heat per unit volume.
| \begin{equation} \kappa(T) = \alpha (T) \times C_{p}(T) \times d \end{equation} | (1) |
The temperature dependence (300–1000 K) of the electrical resistivity ρ(T) and the Seebeck coefficient S(T) were measured using a standard four-probe technique and steady-state method, respectively, under a vacuum atmosphere better than 5 × 10−3 Pa. These measurements were carried out using in-house-built vacuum apparatuses in which dV-dT fitting on at least 60 dT points within dT < 5 K was performed at every setting temperature. The reliability of this measurement system has been confirmed on various materials by comparison of the data with the well-known system ULVAC ZEM3.
The XRD patterns of the milled (Si55Ge35P10)99Fe1 sample in the powder form and the bulk sample co-sintered with Ti are shown in Fig. 1(a) together with calculated XRD patterns of Si, Ge, and Ti. The XRD pattern of the bulk sample sintered without Ti is also shown in Fig. 1(b). The peaks from the unalloyed Si and Ge are absent from the XRD pattern of the powder sample after ball milling in both samples. It means that the employed ball milling process led to homogeneous alloying. The increase in the background intensity of the milled samples at around 2θ = 50° presumably suggests the presence of an amorphous phase. This fact has already been confirmed by TEM in our previous study [8, 9].
The XRD pattern of the sintered samples showed a reduction both in peak width and the amorphous background, suggesting a grain growth and a reduction in the volume fraction of the amorphous phase, respectively. The average particle sizes calculated from Scherrer’s equation for the sintered bulk samples with and without Ti were 9.0 nm and 17.2 nm, respectively. The difference in average particle size is presumably due to a displacement in the position of the thermocouple used to monitor the temperature during sintering, which resulted in a difference in the actual sample temperature. The increase in particle size leads to a decrease in electrical resistivity and an increase in thermal conductivity; thus, the impact of the particle size difference will be discussed in terms of the measured physical properties. Despite the large difference in the averaged grain size, the density of both samples showed almost the same value of 3.26 g cm−3 (93% of the theoretical density). This result confirmed that the nano-grained bulk Si-Ge sample could be well densified through “the low-temperature & high-pressure sintering technique”.
Note here that the peak near 2θ = 40° in the nano-grained Si-Ge sample co-sintered with Ti is not produced by the sintered nano-grained bulk Si-Ge sample itself, but attributed to the small fraction of co-sintered Ti. It was intentionally kept remained on the sample surface for confirming the oxygen absorption in the Ti.
The secondary electron image and EDX mapping images of the fabricated nano-bulk samples are shown in Fig. 2. No segregation of the constituent elements was observed in both the samples sintered w/wo Ti. The EDX mapping images and the secondary electron SEM image shown in Fig. 2 clearly prove that the sintered sample is dense, homogeneous, and therefore high quality. The oxygen concentrations were 4.2 at% for the sample without the oxygen getter (Ti) while 2.4 at% for the sample with Ti. Although a small amount of oxygen was kept retained in the sample co-sintered with Ti, it is strongly emphasized that Ti effectively worked as an oxygen getter during the PCS process. The results of the composition of (Si55Ge35P10)99Fe1 samples determined by EDX are summarized in Table 1.
Secondary electron image and EDX mapping of the (Si55Ge35P10)99Fe1 sintered with titanium (#11 in Table 2).
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/MT-E2025001tab01.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
Figure 3 shows EDX mapping images around the area where a tiny fraction of oxygen getter (Ti) was intentionally kept remaining on the sample surface. The higher intensity of the oxygen signal is observable in the area of the oxygen getter (Ti). This fact lends great support to the validity of the use of Ti as an effective oxygen getter. As a result, a significant reduction in electrical resistivity is took place, as it will be described in the next subsection.
EDX mapping of the oxygen-getter (Ti) section (#11 in Table 2).
It should be mentioned, however, that the sample still contained 2.4 at% of oxygen even with the oxygen getter during the PCS process. In order to eliminate potential barriers that prevent the electrons passing through the grain boundaries without the inelastic scatterings, we need to further reduce the oxygen concentration by optimizing parameters in this co-sintering method. We already invented a new, optimized method, and that will be reported in the future.
3.2 Thermoelectric propertiesFigure 4 shows the electrical resistivity of the nano-grained bulk Si-Ge samples at room temperature plotted as a function of oxygen concentration and average crystallite size. The sample preparation conditions, oxygen concentrations, particle sizes, and electrical resistivities are summarized in Table 2. It was clearly observed that lower electrical resistivity is obtainable for the samples with lower oxygen concentrations and larger average crystallite sizes. This tendency is naturally interpreted as the increase in particle size leading to a longer mean free path, and the lower oxygen concentration causes reduced potential barriers at the grain boundaries.
Relation between grain size, oxygen concentration, and electrical resistivity of the sample.
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/MT-E2025001tab02.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
Figure 5 shows the oxygen concentration dependence of the electrical resistivity and Seebeck coefficient at 922 K. By use of co-sintering with Ti, the samples with low oxygen concentration were successfully fabricated with good reproducibility, resulting in a low electrical resistivity. In sharp contrast to the significant oxygen concentration dependence of electrical resistivity, the Seebeck coefficient showed less obvious effect of oxidization. Since the thermal conductivities of the nano-grained Si-Ge and SiO2 glass are comparable at around 1 Wm−1K−1, it is naturally understood that the temperature gradient and the subsequent thermoelectric motive force are not greatly affected by the tiny volume fraction of SiO2 at the grain boundaries. As a result, thermal conductivity and Seebeck coefficient become less sensitive to the formation of thin SiO2 layers at the grain boundaries.
The dependence of (a) electrical resistivity and (b) Seebeck coefficient on oxygen concentration at 922 K.
Figure 6 shows the temperature dependence of electrical resistivity, Seebeck coefficient, thermal conductivity, power factor (PF = S2ρ−1), and ZT observed for the samples of high oxygen concentration (sample #10 in Table 2) and low oxygen concentration (sample #11 in Table 2). The electrical resistivity decreases with temperature to reach 3.9 mΩ cm at 922 K. The negative temperature coefficient of electrical resistivity (TCR) is caused by the delocalization of conduction electrons with increasing temperature. The scenario of carrier excitation as the source of electrical resistivity reduction is safely rejected because the absolute value of Seebeck coefficient is slightly increased with increasing temperature in the low temperatures where the electrical resistivity is most drastically decreased. The localization of carriers would be caused by the heavily introduced chemical disordering induced by the coexistence of Si, Ge, and P in the diamond lattice and the structural disorder induced by high-energy ball milling. Note here that a quantum interference effect known as the weak localization is tend to be observed with a small Fermi surface such as that in the previous samples, and the Anderson localization is also known to take place near the edge of valence band and conduction band where electrical density of states is kept small. The decrease in electrical resistivity with increasing temperature is caused by the intensified hopping conduction (delocalization). One of the several different mechanisms of hopping conduction, such as nearest hopping, variable range hopping, and variable range hopping with Coulomb gap, should take place in our present sample, but we cannot identify it only from the temperature dependence of electrical resistivity in the range above 300 K.
Thermoelectric properties of samples with and without titanium (a) electrical resistivity, (b) Seebeck coefficient, (c) thermal conductivity, (d) power factor, (e) ZT.
The power factor calculated from the thermoelectric properties of the sample co-sintered with Ti reached PF = 4.2 Wm−1K−2 at 922 K. At the same temperature, the dimensionless figure of merit achieved a maximum value of ZT = 4, which is the highest ever reported for bulk thermoelectric materials. Besides, ZT > 1 was maintained over a wide temperature range from 700 K to over 900 K. Since our previous studies observed ZT = 3.6 at temperatures higher than 922 K, further improvement in ZT would be expected by increasing the measurement temperature range. We intentionally avoided it in the previous study because it would cause irreversible reactions both in grain growth and precipitations of the secondary phase. To identify conditions that yield even higher ZT values, we have to precisely determine the temperature at which the second phase precipitates and grain growth occurs. It is now being conducted and the results will be reported elsewhere in near future. Future work will be focused also on the thermal and chemical stability under the condition of practical use.
In the conventional PCS process for Si-Ge alloys using high-temperatures at around 1200 K, a low-pressure of 50 MPa condition is employed with carbon dies and punches. The high temperature in the carbon die and punches is supposed to cause a reducing atmosphere. Consequently, oxygen within the sample reacts with the surrounding carbon and is expelled as CO2, thereby minimizing oxidation issues [12]. Therefore, it is considered that the severe oxidation was not considered in such a PCS process. In sharp contrast to the ordinary PCS process, “the low-temperature & high-pressure sintering process”, which plays the most important role in preparing densified nano-grained bulk Si-Ge samples, does not produce a reduction atmosphere in the die to cause severe oxidization at the surface of nano-grains. The formation of thin oxide films at the surface of nano-grains strongly affects the electrical resistivity by making barriers for electrical conduction, while their thin thickness has tiny impact on the Seebeck coefficient or thermal conductivity. In previous studies, the degree of oxidation was not fully controlled, and therefore led to a variation in electrical resistivity and difficulties in reproducibly obtaining high ZT values.
In this study. “the low-temperature & high-pressure sintering process” with “the oxygen getter co-sintering” was simultaneously employed for nano-powder samples. The amount of Ti was determined so as to absorb 10 at% of oxygen. All the residual oxygen was supposed to be absorbed in Ti to form TiO2−x. However, the residual oxygen remained within the samples, despite its concentration being effectively reduced. It means that Si-Ge nano-grained bulk samples possessing a much smaller electrical resistivity and hence the larger value of ZT could be obtainable with selecting a more appropriate oxygen getter element and optimizing its amount.
The reduction process of Si-Ge is still under investigation. For effectively eliminating the oxygen concentration in the sample by using the oxygen getter during the PCS process, the adsorbed oxygen at the surface of Si-Ge nanoparticles should be forced to move towards the Ti layer. A much larger magnitude of negative formation energy of TiO2 than that of SiO2 and Ge-O, could provide us with the effective driving force of oxygen absorption in Ti. However, the dynamics of oxygen moving from Si-Ge bulk towards Ti at the surface at the low sintering temperature of 848 K has been still unclear. It should be noted, in addition, that Ti may prevent oxygen in the vacuum chamber from coming into the sample during the sintering process. The good vacuum level of the sintering chamber with the 5 times nitrogen flushing process suggests that this scenario is less plausible interpretation.
Another method for reducing oxygen is the instant sintering at a temperature close to a reducing atmosphere, which allows for the removal of oxygen at the surface of nano-grains.
We have to also mention that the nano-crystallization process is another point of improvement. The sample preparation processes, such as storage of the raw materials, weighing, and packing in to the die, should be further improved by using in a better inert atmosphere in a glove box. The use of better oxygen getters such as Zr or Mg, which have a larger absolute value of the Gibbs free energy of oxide formation is also being considered for further studies.
In this study, we developed a novel method for synthesizing high-performance nano-grained bulk Si-Ge thermoelectric materials through co-sintering of Ti with the samples to minimize oxidation. By employing co-sintering of the Si-Ge nano-crystalline powder with Ti powder, the Ti effectively eliminated oxygen from the samples. The reduction in oxygen concentration resulted in a significant decrease in electrical resistivity to 3.9 mΩ cm at 922 K. As a consequence, the maximum value of ZT reached 4, that is the largest value ever reported for single-phase bulk material. The temperature range where ZT exceeds unity was widely extended from 700 K to 900 K.
