1 Introduction

Nowadays, the global energetic situation, and the Glasgow agreement in 2021, point out to a reduction of fossil fuels consumption. In this scenario, thermoelectric (TE) systems are very promising to decrease fuels consumption, by increasing the efficiency of classical energy transforming devices. By means of the Seebeck effect, these systems allow the harvesting of the wasted heat [1], accompanied by a decrease of CO2, and greenhouse gases emissions [2]. The adequate materials for use in TE generation systems should display values above 1 of the dimensionless figure-of-merit (ZT), defined as [3]:

$$ZT=\frac{{S}^{2}T\sigma }{\kappa }$$

where S, T, σ, and κ are the Seebeck coefficient, absolute temperature, electrical conductivity, and thermal conductivity, respectively.

On the other hand, only intermetallic materials show ZT > 1 values when working at moderate temperatures [4, 5]. However, they are usually composed by a large amount of heavy elements which can be oxidized or volatilized at high temperatures under air. When considering TE systems working at high temperatures, oxide materials can be a very good option as, in spite of their relatively low ZT values, they show high chemical and thermal stability under air. Moreover, these oxides are mostly abundant, cheap, and environmentally friendly [6]. As a result, most of the studies which are published about TE oxides focus on raising their performances by different approaches [7,8,9,10,11,12,13,14].

On the other hand, it should be taken into account that TE properties of materials are very important to have an efficient TE module, but they are not the only parameters, as the manufacturing factor (MF), defined as the ideal internal resistance of the module, Rid, and the real internal resistance, Rint, in the form MF = Rid/Rint [15]. The main difference between them is that the Rid only considers the resistance of the TE legs, while the Rint also takes into account the electrical contact resistances. The MF factor determined in the modules after fabrication is not constant during their working life, as it can decrease due to the different thermal expansion of the components of these modules, leading to internal stresses which can increase the real internal resistance. As a consequence, choosing materials with similar thermal expansion coefficients is important to minimize these induced thermal stresses to maintain constant the performances of the modules during their lifespan. Moreover, the TE materials should possess high mechanical properties to assure the module integrity. All these characteristics are also very important to extend the modules lifecycle.

In this work, Ca3Co4O9 material has been prepared through the classical ceramic route, with the addition of small amounts (0, 0.25, 0.50, and 0.75 wt%) of B4C, AlN, TiC, TiB2, or TiN. These materials have been selected as additives as they show very high hardness and moduli [16], and some of them have been previously reported highlighting their ability to modify electrical properties [17,18,19]. The effect produced by these compounds on the structural and microstructural characteristics of the sintered specimens has been evaluated, and correlated to the changes on the linear expansion coefficient, and thermoelectric properties.

In the next section, all the preparation steps and the characterization equipments and conditions, are described; Sect. 3 shows the obtained results and their discussion, trying to link them. Finally, in Sect. 4 we present the main results and conclusions of this work, indicating some future recommendations and directions for deeper knowledge of these systems.

2 Experimental

Ca3Co4O9 + x wt% B4C, AlN, TiC, TiB2, or TiN, with x = 0.0, 0.25, 0.50, and 0.75, were made using the conventional solid-state route from CaCO3 (Panreac, 98 + %), and Co2O3 (Aldrich, 98 + %) powders. They were weighed in stoichiometric proportions, mixed and ball milled 30 min at 300 rpm in double distilled water. The resulting suspension was totally dried using infrared radiation, and the powder was manually milled. The mixture was then calcined in two steps, at 750 and 800 ºC for 12 h under air, with a manual milling after each thermal treatment. After calcination, the powder is composed of intermediate products and is carbonate-free, as previously reported [20]. The calcined powders were then mixed with B4C (Alfa Aesar, 99 + %), AlN (Aldrich, 99.8%), TiC (Aldrich, 99.8%), TiB2 (Alfa Aesar, 99.5%), or TiN (Alfa Aesar, 99.7%) in the adequate proportions and ball milled at 300 rpm for 30 min. The mixtures were cold uniaxially pressed under 400 MPa in form of pellets (3 × 3 × 15 mm3), which were sintered using the optimal conditions previously determined [21], and consisting in one step heating at 900 ºC for 24 h under air atmosphere, with a final furnace cooling.

Powder X-ray diffraction (XRD) patterns were performed on all samples using a theta-theta PANalyticalX'Pert Pro diffractometer (CuKα radiation, λ = 1.54059 Å) between 10° and 50° (step 0.013°, and 3 s recording time), which is the range where most of the peaks of Ca3Co4O9 phase, and the most intense of the different added compounds, appear.

Microstructure was studied on surfaces of samples, using backscattered electrons to easily identify the different contrasts, in a Field Emission Scanning Electron Microscope (FESEM, Carl Zeiss Merlin) attached to an energy-dispersive spectrometry (EDS) system. Density was determined in all samples through the Archimedes’ method [22] using a minimum of three samples for each composition. Relative densities [22] were determined using the theoretical values for each compound, and considering a homogeneous mixture of Ca3Co4O9 (4.677 g/cm3 [23]), B4C (2.50 g/cm3 [24]), AlN (3.255 g/cm3 [24]), TiC (4.93 g/cm3 [24]), TiB2 (4.38 g/cm3 [24]), and TiN (5.21 g/cm3 [24]). Simultaneous Seebeck coefficient (S), and electrical resistivity (ρ), measurements were performed, in four-point contact configuration [3], in a LSR-3 system (Linseis GmbH) between 50 and 800 ºC under He atmosphere. The performances of samples were determined through the power factor, PF, defined as PF = S2/ρ [3]. These data were used to determine the evolution of performances with the type and content of the different added compounds. Moreover, they were also compared to the reported values in the literature for related compounds. Dilatometric studies were made on all samples between 25 and 800 ºC, under air, in a L79 HCS dilatometer (Linseis GmbH) to evaluate the modification of the linear expansion coefficient with the type, and amount, of the added compounds.

3 Results and discussion

Figure 1 shows representative powder XRD patterns of undoped Ca3Co4O9 samples and those with the highest added compounds. In the figure, it can be seen that all samples are composed by nearly pure Ca3Co4O9 phase [25,26,27], but small amounts of Ca3Co2O6 can be distinguished [25]. On the other hand, it should be noted that the fact that the most intense peaks are associated to the ab-planes, does not reflect a real grain orientation in the bulk sample. This effect is only associated to the grains shape which favours their orientation when preparing the samples for the XRD characterization, as demonstrated in previous reports [28]. Furthermore, no B4C, AlN, TiC, TiB2, or TiN reflections have been observed in their corresponding samples, even if their most intense peaks should be in the observed 2θ range [29,30,31,32,33], around 38, 33, 42 (and second most intense one at 36°), 44.5°, and 41°, respectively, which is probably due to the small amount added to the samples.

Fig. 1
figure 1

Powder X-ray diffraction patterns of pristine and 0.75 wt% B4C, AlN, TiC, TiB2, and TiN Ca3Co4O9 samples. The diffraction planes indicate the reflections of Ca3Co4O9 phase, while * identify those associated to the Ca3Co2O6 one

Figure 2 illustrates representative SEM micrographs of all compositions, taken on their surfaces. As it can be easily observed in the micrographs, all samples are formed by randomly oriented plate-like grains, as observed in previous reports [34,35,36]. Table 1 summarizes the composition, determined through EDS, of the different contrasts observed in the micrographs. These analysis have shown that major phase is the Ca3Co4O9 one, independently of the amount and nature of the added compounds, accompanied by small amounts of the Ca3Co2O6 one (indicated by #1 in Fig. 2a), in agreement with the XRD data previously discussed. On the other hand, the additives behave differently under the sintering conditions and produce different effects; the B4C, TiB2, and TiN compounds greatly decrease the grain sizes (see Figs. 2b, e, and f), when these are compared to the pristine sample (see Fig. 2a), while the AlN, and TiC present a softer effect (see Figs. 2c, and d). Moreover, these compounds also show different reactivity with the Ca3Co4O9 phase during the sintering procedure, as determined by EDS analysis; B4C and TiB2 addition leads to the formation of a boron rich compound, with the release of CO2 in the former case, and the formation of CaTiO3 phase in the last one (grains indicated by #2, and #5 in Fig. 2). Both effects seem to be very similar and can be explained by considering the case of B4C compound reported in previous works [37]. In this case, during sintering the B4C oxidizes at about 450 ºC in air atmosphere, decomposing into CO2 and B2O3 [38], with 450 ºC melting point [39], which acts as a protective layer on the surface of these particles and avoids further oxidation [40]. Moreover, this liquid phase also dissolves cations from the Ca3Co4O9 one, producing the B-rich phase on the B4C surface. In the case of TiB2, some similar behaviour can be expected, forming the B-rich phase, and CaTiO3 phase on the TiB2 surface. In the case of AlN, no reaction has been observed and only AlN agglomerates have been found in the samples. Finally, in the case of TiC, and TiN, only CaTiO3 phase has been found in the surface of these particles, probably with a release of CO2, and N2, respectively.

Fig. 2
figure 2

Representative SEM micrographs obtained on the surfaces of the different Ca3Co4O9 samples, for a Pristine; and 0.75 wt% b B4C; c AlN; d TiC; e TiB2; and f TiN ones. #1 indicates the Ca3Co2O6 phase; #2 a boron-rich compound; #3 AlN; #4 CaTiO3; #5 CaTiO3 + boron-rich compound; and #6 CaTiO3

Table 1 Composition, determined through EDS on points with the diferent contrasts observed in the Ca3Co4O9 samples

In Table 2 the mean relative density is presented for all compositions. As it can be easily observed, they are relatively low, but they are in the range of the reported values for sintered materials prepared by the ceramic method [36, 41,42,43]. These characteristics can be easily understood when observing the phase equilibria diagram of the CaO–CoO system previously published, where it can be seen that the Ca3Co4O9 phase is only stable below 926 ºC [44]. Moreover, the eutectic point is observed at 1350 ºC, which is much higher than the maximum stability of Ca3Co4O9 phase, or the sintering temperature used in this work (900 ºC) and, as a consequence, densification of samples is highly limited, explaining the low values of density in sintered samples. On the other hand, the larger densification has been obtained with nitride additions, which could be due to their greater resistance to oxidation, compared to the other compounds. The lower densification obtained with TiC and TiB2 can be associated to oxidation processes, with the release of CO2 in the first case, and the formation of liquid B2O3 in the second one, even if they could be only produced in the particles surfaces. As a consequence, for the B4C, both of these processes are produced, resulting in a decrease of samples densities. On the other hand, taking into account the reactions previously described, other factor that can influence density values is the increase in oxygen content in these samples due to the oxidation processes of these additives, which increases the mass of the particles even if it is only produced in their surfaces. However, this trend would be in competition with the formation of lower density species than those of the additives, as B2O3 (2.55 g/cm3), CaTiO3 (3.98 g/cm3), or, even if not detected, Al2O3 (3.97 g/cm3) [24].

Table 2 Relative density (% of theoretical one) determined through Archimedes’ method for all samples

Electrical resistivity variation with temperature is presented in Fig. 3 for the best samples of each composition, and the pristine one as reference. When comparing the different curves, it can be seen that all samples behave very similarly within the measured temperature range, reaching a minimum value at about 450 ºC. This point marks a behaviour change, from semiconducting (dρ/dT < 0) below this point, to metallic one (dρ/dT > 0) at higher temperatures. Physically, it can be explained by a hole hopping transport mechanism from Co4+ to Co3+ [45] in the semiconducting region, while a charge carriers transport mechanism in the valence or conduction band is produced in the metallic zone [46]. This behaviour is in agreement with the one typically reported on sintered materials [36, 47,48,49]. When considering the electrical resistivity values, it is clear that all type added compounds contribute to their decrease, generally for low contents. This decrease can be associated to two effects; firstly, the added compounds can lead to small chemical modifications on the surfaces of the Ca3Co4O9 phase, and secondly, the largest decrease in some of the samples is due to the important oxidation processes in TiB2 and TiC (which is even enhanced in B4C), probably decreasing the oxygen content. The minimum value at 800 ºC has been determined in 0.25 wt% B4C added samples, 14 mΩ cm, which is about 20% lower than the measured in the pristine sample in this work. Moreover, it is also lower than those reported in sintered or spark plasma sintered (SPS) Ca3Co4O9 samples, between 15 and 18 mΩ cm [50]. On the other hand, it is still higher than the values determined in materials with high density, between 8 and 12.5 mΩ cm [13, 51, 52].

Fig. 3
figure 3

Electrical resistivity variation with temperature for the best samples of each composition, using the pristine one as a reference

Figure 4 illustrates the Seebeck coefficient evolution with temperature for the best samples of each composition, together with the pristine one, as reference. In the plot, it can be seen that S is positive in all cases, which is associated to dominating hole conduction mechanism. Furthermore, despite slight differences found at high temperatures for some samples, the values seem similar, independently of the nature of the added compound. This behaviour can be explained taking into account that reactions between the added compound and the Ca3Co4O9 phase is very limited and could only take place along the grain boundaries, as previously described. As a consequence, these very similar S results for all samples clearly show that nearly no compositional modifications are produced on the thermoelectric phase. The maximum S value at 800 ºC has been reached by most of the samples, 195 μV/K, which is higher than those reported in sintered or SPS Ca3Co4O9 samples, between 170 and 180 μV/K [50, 52]. On the other hand, it is slightly lower than the value determined for materials with high density, 205 μV/K [13, 51].

Fig. 4
figure 4

Seebeck coefficient variation with temperature for the best samples of each composition, using the pristine one as a reference

Using the electrical resistivity and Seebeck coefficient data, PF has been calculated, and plotted as a function of temperature, in Fig. 5 for the best samples of each composition, and the pristine one as reference. As it can be observed in the graph, pristine sample displays the lowest values among all samples, in the whole measured temperature range. The difference on the PF values are only due to the different values of electrical resistivity, as all samples display the same S values. The highest PF values have been determined in 0.25 wt% B4C samples, in the measured temperature range, reaching its maximum at 800 ºC, around 0.27 mW/K2m, which is lower than the reported for highly densified materials, between 0.30 and 1.15 mW/K2m [13, 50,51,52,53]. However, it is higher than the values obtained in conventionally sintered materials, between 0.09 and 0.19 mW/K2m [21, 47, 48].

Fig. 5
figure 5

Power factor variation with temperature for the best samples of each composition, using the pristine one as a reference

Finally, linear thermal expansion coefficient, determined in all samples, is shown in Table 3. The thermal expansion coefficient of pure samples (10.32 1/K) is the highest one, revealing that all additions led to a decrease of this parameter. This effect could be related with the reactivity of added particles with the thermoelectric phase, producing larger decrease of thermal expansion when the reactivity is higher in the sintering conditions. On the other hand, the value determined in the pristine samples is very similar to that reported in Ca3Co4O9 materials (10.6 1/K [36]), and clearly lower than the obtained in Co-site doped samples (12.0–13.1 1/K [36]), or Ca-site doped ones (10.6–12.9 1/K [42, 49]. However, it is higher than the reported for hot-pressed Sr-doped Ca3Co4O9 (between 9.4 and 9.7 1/K) [54]. Other interesting observation is that the lowest thermal expansion coefficients of each composition have been measured on the samples with the lowest electrical resistivity, clearly confirming the higher reactivity of these compounds. These characteristics could be explained by the formation of relatively strong grains connections which can enhance charge carrier mobility and decrease grains expansion. The lowest value has been obtained in 0.25 wt% B4C samples, 9.3 1/K, about 10% lower than the measured in the pristine sample, and only around 20% higher than that of the Al2O3 (7.5 1/K [55]). The decrease on thermal expansion, even if still far from that of Al2O3, is important to reduce the internal stresses in TE modules by diminishing their differential thermal expansion at high temperatures, allowing increasing the lifespan of these systems.

Table 3 Thermal expansion coefficient (α, 1/K) determined for all samples

4 Conclusions

Ca3Co4O9 + x wt% B4C, AlN, TiC, TiB2, or TiN (x = 0, 0.25, 0.5, and 0.75) polycrystalline ceramic materials were prepared using the classical solid-state route. Powder XRD analysis has found the Ca3Co4O9 phase in all samples, independently of the added compound and proportion. SEM observations showed that some new phases appear in the surface of the added compounds by reacting with air and the Ca3Co4O9 grains. Moreover, grain sizes decreased with these additives, while no drastic changes in density were noticed. Seebeck coefficient was kept nearly constant, pointing to a negligible modification of Ca3Co4O9 phase by the added compounds. On the other hand, electrical resistivity and linear thermal expansion decreased, independently of the added compound and proportion, when compared to the pristine sample. These changes could be explained by the formation of relatively strong grain boundaries which could enhance carrier mobility and slightly decrease the thermal expansion. Finally, power factor has been improved in all cases, when compared to the pure sample, reaching the maximum value at 800 ºC in 0.25 B4C samples (0.27 mW/K2m), which is much higher than the reported for conventionally sintered materials.

On the other hand, from the results obtained in this work, it is clear that deeper studies regarding the behaviour of the different ceramic additives under air at sintering temperature are necessary to clarify the chemical reactions produced in these conditions. Moreover, it should be also very important to know their effect on the thermal conductivity of these bulk materials to exactly evaluate their contribution to the improvement of thermoelectric performances.