Effect of B4C addition on mechanical and electrical properties of Ca3Co4O9

B4C added Ca3Co4O9 in different proportions (0, 0.1, 0.25, 0.5, and 0.75 wt.%) have been fabricated using solid-state method. Powder XRD patterns have shown that only Ca3Co4O9 phase can be identied in all the samples. Microstructural observations have allowed determining that B4C compound has been supercially oxidized, producing liquid B2O3 phase during sintering, which has reacted with the Ca3Co4O9 grains to produce bridges between them. In spite of the increase of porosity, these bridges led to an important raise (more than two times) of mechanical properties. On the other hand, while B4C addition has not signicantly inuenced S values, it has decreased electrical resistivity, thermal conductivity, and thermal expansion. Consequently, ZT values have been also increased, reaching 0.24 in 0.25 wt.% B4C containing samples, which is very close to the best values reported in the literature for Ca3Co4O9 compounds, and two times higher than the obtained in pure materials in this work.


Introduction
Thermoelectric (TE) technology is considered as a very promising one to enhance energy transforming devices e ciency by harvesting wasted heat [1]. This characteristic is associated to the Seebeck effect inherent to the TE materials [2]. On the other hand, these materials should have appropriate thermoelectric properties, evaluated using the equation [3]: where ZT is the dimensionless Figure-of-Merit, and S, T, σ, and κ are Seebeck coe cient, absolute temperature, electrical, and thermal conductivity, respectively. TE materials are considered adequate for practical applications when their ZT values are higher than 1. Nowadays, only intermetallic materials display these values at low-medium temperatures [4,5], with the drawback of their usual heavy elements content. Moreover, they cannot be used at high temperatures due to oxidation processes, unless they are encapsulated to avoid oxidant environments. On the other hand, oxide materials possess high chemical and thermal stability, and can work at high temperatures. Moreover, they are composed of abundant, cheaper, and environmentally friendly elements [6]. However, their TE performances are lower than those determined in intermetallic materials. Consequently, most of the works published in this eld are dealing with the improvement of their TE performances through many different ways [7][8][9][10][11][12][13].
Nevertheless, TE properties of materials are not the only parameters determining the e ciency of a TE module. One of them is the so-called manufacturing factor (MF) which is a function of the ideal resistance in the module (R id , taking into account the resistance of legs), and the real internal resistance (R int , where the contact resistance is also present), as MF = R id /R int [14]. This MF is usually determined in the asfabricated modules, but it can be decreased during the working life of these modules due to the differential expansion of the different components (p, and n thermoelectric legs, metallic parts, and insulating ceramic) which can damage the contacts among them, increasing the internal resistance.
Consequently, before building a TE module, it is very convenient to choose materials with very close thermal expansion coe cients to avoid internal stresses when working at high temperature. Besides, high mechanical properties are also desired in these TE legs in order to maintain the module integrity to increase its useful life.
In this work, Ca 3 Co 4 O 9 material will be prepared, with the addition of small amounts of B 4 C (0, 0.10, 0.25, 0.50, and 0.75 wt.%) via the classical solid-state route. The effect of these additions on the structure and microstructure of Ca 3 Co 4 O 9 sintered pellets will be studied, and related to the modi cations of linear expansion coe cient, mechanical, and thermoelectric properties. Experimental Ca 3 Co 4 O 9 + x wt.% B 4 C, with x = 0.0, 0.10, 0.25, 0.50, and 0.75, have been prepared using the classical solid-state route from CaCO 3 (Panreac, 98 + %), and Co 2 O 3 (Aldrich, 98 + %) commercial powders. They were weighed in the adequate amounts, mixed and ball milled under water media for 30 minutes at 300 rpm to produce a suspension, which was dried under infrared radiation. The resulting dry mixture was manually milled to produce a very ne powder, subsequently thermally treated at 750 and 800 ºC for 12 h under air, with an intermediate manual milling. This thermal process is adequate to decompose CaCO 3 and to produce intermediate by-products, in agreement with previously reported data [15]. After milling these calcined powders, B 4 C (Alfa Aesar, 99 + %) has been added in the stoichiometric proportions and the mixture has been ball milled 30 minutes at 300 rpm to homogenize the mixture. After this process, the different materials were cold-pressed in form of pellets under 400 MPa applied pressure. Finally, the compacts were sintered at 900 ºC for 24 h under air atmosphere, followed by a nal furnace cooling.
Powder X-ray diffraction (XRD) patterns were determined in a theta-theta PANalyticalX'Pert Pro diffractometer (CuKα radiation, λ = 1.54059 Å) between 10 and 40 degrees, where the main peaks of Ca 3 Co 4 O 9 phase, and the most intense of B 4 C, appear. Density measurements were performed using Archimedes' method on several samples for each B 4 C addition. The relative density values were calculated with respect to the theoretical one, assuming a dense and perfect mixture of Ca 3 Co 4 O 9 (4.677 g/cm 3 [16]) and B 4 C (2.50 g/cm 3 [17]).
Microstructural observations have been performed on the samples surfaces and fractured sections in a Field Emission Scanning Electron Microscope (FESEM, Carl Zeiss Merlin) combined with an energydispersive spectrometry (EDS) system. Flexural strength has been determined in several samples for each composition using the three point bending test in an Instron 5565. Samples were measured with 30 µm/s punch displacement and 10 mm span. Seebeck coe cient and electrical resistivity were simultaneously determined through the four-point contact using a LSR-3 (Linseis GmbH) under He atmosphere in the 50-800 ºC temperature range. Thermal diffusivity (α) has been measured in a laser-ash system (Linseis LFA 1000). Thermal conductivity (κ) has been calculated as κ = α·C p ·ρ, being C p speci c heat and ρ, sample density. C p has been determined through Dulong-Petit law. In order to stablish the samples TE performances, ZT was determined from Seebeck coe cient, electrical resistivity, and thermal conductivity data. These data were used to evaluate the properties evolution as a function of the dopant content, and were also compared with previously reported values in the literature for similar compounds. Finally, dilatometric behaviour of samples has been studied in the 25-800 ºC temperature range in a L79 HCS dilatometer (Linseis GmbH) to determine their linear expansion coe cient evolution as a function of B 4 C content.
Results And Discussion Figure 1 illustrates the XRD patterns obtained in Ca 3 Co 4 O 9 + x wt.% B 4 C powdered samples. As it can be easily observed in the gure, all samples present the same patterns, independently of the B 4 C addition.
Moreover, all peaks can be associated to the diffraction planes (indicated in the graph) of Ca 3 Co 4 O 9 phase with monoclinic symmetry [18,19]. Furthermore, the most intense diffractions correspond to its ab-planes, which could be associated to preferential grain orientation. However, it should be highlighted that XRD patterns were obtained on powdered samples and, consequently, this grain orientation is produced by the samples preparation, as observed in previous works [20]. On the other hand, no B 4 C has been identi ed with this technique, probably due to its small amount, and the fact that its most intense peak appears at 39.4 degrees [21], together with a possible partial decomposition during sintering, and it can be masked by the (020) diffraction of Ca 3 Co 4 O 9 phase. images show that all samples are composed of ake-like particles with no preferential orientation, which is the characteristic situation in this system, as reported in previous works [22,23]. Moreover, all grains composition has been determined, through EDS, as Ca 3 Co 4 O 9 phase, con rming the XRD results previously presented. On the other hand, the Ca 3 Co 4 O 9 grain sizes are decreased, when the amount of B 4 C in the samples is increased, while porosity seems to be increased. In order to con rm the raise in porosity when the B 4 C content is larger, the samples density was evaluated and the mean results, together with their standard error, are displayed in Table 1. As it is clear from these data, samples density is slightly decreased when the amount of B 4 C is increased, corroborating the SEM observations. In any case, the relative density values are in the order of the typically reported in materials prepared through the classical solid-state method [23][24][25]. This low densi cation can be explained when considering the phase equilibria diagram [26], which shows that Ca 3 Co 4 O 9 phase is stable up to 926 ºC. Consequently, sintering procedure has to be performed below this temperature, which is, in turn, much lower than the eutectic temperature (1350 ºC), drastically limiting the samples densi cation. properties are drastically increased, when compared to the pure samples. It is increased in around 50% for 0.1% B 4 C addition, and more than 100% for higher contents. These results seem to be contradictory and, consequently, the fractured sections of samples were microstructurally studied. Figure 4 shows a representative micrograph at high magni cation of 0.75 B 4 C sample. As it can be observed in the picture, besides the grey contrast (Ca 3 Co 4 O 9 phase), a new contrast (dark grey) is appearing. EDS analysis has shown that this contrast contains B, C, O, Ca, and Co, and it is connecting thermoelectric grains. This behaviour can be explained when considering that B 4 C starts to be oxidized under air at around 450 ºC, producing CO 2 and B 2 O 3 [27]. This effect could lead to the total B 4 C oxidation, however, the formed B 2 O 3 has melting point of 450 ºC [28] and, between 600 and 1000 ºC, forms a protective layer on the particles surface avoiding further oxidation [29]. On the other hand, the presence of this liquid phase, making close contact with the thermoelectric grains, lead to Ca-B-O [30], and Co-B-O [31] compounds formation on the B 4 C particles surface. Consequently, these compounds form bridges between the thermoelectric grains, enhancing the samples mechanical properties.
Electrical resistivity variation with temperature, as a function of B 4 C content is presented in Fig. 5. As shown in the plot, all samples possess very similar behaviour in the whole measured temperature range.
The samples display a minimum at around 450 ºC, which corresponds to the change between semiconducting (dρ/dT < 0) and metallic (dρ/dT > 0) behaviour. This temperature indicates the change from a hole hopping from Co 4+ to Co 3+ [32] to a charge carriers transport in the valence or conduction band [33]. This is a very common behaviour observed in Ca 3 Co 4 O 9 sintered materials [23,34]. On the other hand, B 4 C addition up to 0.25 wt.% decreases electrical resistivity, while further addition drastically increases it, when compared to samples without additions. This behaviour with B 4 C is similar to the one observed in thin lms, where a slightly lower Co content, with respect to the stoichiometric one, leads to lower electrical resistivity [35]. Moreover, this effect also induces a lower grain size, which agrees with the previous observations. On the other hand, the decrease of electrical resistivity of B 4 C containing samples, when compared to the pure ones, is lower than the observed in [35], which can be associated to a lower reduction of Co content. Furthermore, the formation of B-containing phases in these samples can also affect the electrical transport properties. Consequently, the lowest electrical resistivity values have been obtained in samples with 0.25 wt.% B 4 C. The minimum resistivity values measured in these samples at 800 ºC (14 mΩ cm) are slightly lower than the best values reported for Ca 3 Co 4 O 9 samples sintered or textured through spark plasma sintering (15-18 mΩ cm) [36] and around 20% lower than in pure samples. However, they are higher than the measured in highly dense materials prepared using alternative methods (10 mΩ cm) [37]. The highest values at 800 ºC (195 µV/K) are higher than the reported in pure Ca 3 Co 4 O 9 sintered or textured through spark plasma sintering (170-175 µV/K) [36], but lower than the reported in very dense materials obtained by alternative methods (205 µV/K) [37].
Total κ consists in the addition of two components, κ = κ l + κ e , where κ l is the lattice thermal conductivity contribution and κ e is the electronic counterpart. κ e can be estimated from the Wiedemann-Franz's law [38], which is expressed as κ e = L σ T, where L, and σ are Lorenz number (2.45 × 10 − 8 V 2 /K 2 ), and electrical conductivity, respectively. The variation with temperature and B 4 C content of the electronic and total thermal conductivity are presented in Fig. 7. When comparing both graphs, it is clear that electronic thermal conductivity is increased when the temperature rises. In contrast, the total thermal conductivity decreases from room temperature to around 600-650 ºC induced by the increase of lattice vibrations which enhance phonon scattering. At higher temperatures, the total thermal conductivity increases due to the larger in uence of electronic thermal conductivity. Moreover, B 4 C addition leads to a clear decrease of total thermal conductivity, when compared to the pure samples, especially at high temperatures. These results are in agreement with previous works in Ca 3 − x B x Co 4 O 9 which showed that thermal conductivity is decreased with B-content up to 0.5 [39]. At 800 ºC, the minimum total thermal conductivity value determined in 0.25 wt.% B 4 C samples (1.2 W/K m) is about 20% higher than the lowest measured in these samples (1.0 W/K m at 650 ºC). Moreover, these values are comparable to the reported in classically sintered materials (0.9 W/K m with ∼ 60% theoretical density) which show higher amount of porosity [40] than the samples in this work. On the other hand, it is lower than the best values obtained by spark plasma sintering (2.1 W/K m) [41], and much lower than the measured in textured materials by hot-uniaxial pressing (2.2, 3, and 4.7 W/K m) [13,40,42].
With all the previously discussed data, ZT has been calculated and plotted in Fig. 8. As displayed in the graph, ZT values are increased with temperature in the whole measured temperature range, and also with B 4 C content up to 0.25 wt.%, decreasing for higher additions. The highest value has been measured at 800 ºC in 0.25 wt.% B 4 C samples (0.24), which is more than two times higher than the calculated for pure ones. Moreover, it is higher than the measured in classically (0.08 [40]), or spark plasma sintered samples (0.15 [41]), and even than the determined in textured samples (0.18, and 0.16 [13,40]). However, this value is lower than the reported one for materials prepared through non-conventional methods (0.35 [15]) or in Ca 3 − x Sr x Co 4 O 9 textured materials (0.29 [42]).
Thermal expansion coe cient has been determined in all samples and presented in Table 2. As it can be observed in this table, thermal expansion of pure samples (10.32 ppm/K) is in the range of the reported value for Ca 3 Co 4 O 9 materials (10.6 ppm/K [25]) and lower than the measured in Bi-substituted materials (12.8 ppm/K [25]). On the other hand, B 4 C addition, up to 0.25 wt.%, decreases thermal expansion coe cient, increasing with further additions. The minimum value (9.3 ppm/K) measured in this work corresponds to a decrease of only around 10% but closer to the one reported for Al 2 O 3 (7.5 ppm/K [43]). This reduction, even if it does not seem to be signi cant, may reduce the differential thermal expansion between two of the main components of thermoelectric modules, decreasing the internal stresses at working temperatures, and increasing the life span of these devices.

Conclusions
Ca 3 Co 4 O 9 + x wt.% B 4 C (x = 0, 0.1, 0.25, 0.5, and 0.75) polycrystalline materials have been prepared through the classical ceramic method. XRD analysis has only identi ed Ca 3 Co 4 O 9 phase in all samples, independently of the B 4 C content. SEM observations have shown that B 4 C has reacted with oxygen producing liquid B 2 O 3 on its surface, slightly reacting with Ca 3 Co 4 O 9 grains and forming bridges between the grains. These microstructural modi cations have been re ected in the mechanical properties, which were enhanced when compared with the pure samples, despite the slight increase of porosity. Seebeck coe cient has not been affected by B 4 C addition, while electrical resistivity, thermal conductivity and thermal expansion have been decreased, leading to maximum ZT values close to the best reported in the literature, and about two times higher than the measured in pure samples in this work. It is worth to mention that these results have been obtained through a simple and scalable process which can be easily transferred to industry.      Seebeck coe cient variation with temperature, as a function of B4C content, in Ca3Co4O9 + x wt.% B4C samples.

Figure 7
Evolution of a) Electronic thermal conductivity; and b) Total thermal conductivity, with temperature, as a function of B4C content, in Ca3Co4O9 + x wt.% B4C samples.