Tuning Thermoelectric Properties of Bi2Ca2Co2Oy through K doping and Laser Floating Zone processing

In the present study, thermoelectric Bi 2 Ca 2-x K x Co 2 O y ceramic materials (x=0.0, 0.05, 0.075, 0.10, and 0.125) in different forms (called bulk, as-grown and annealed fibers) have been manufactured via a classical solid-state method and textured using the laser floating zone (LFZ) technique. The identification and characteristics of undoped and doped samples were determined by X-ray diffraction (XRD) and scanning electron microscopy (SEM). The XRD patterns of all samples have shown great similarity, and the major peaks can be assigned to the Bi 2 Sr 2 Co 2 O y thermoelectric phase, independently of the processing technique and K-doping. SEM-EDS have indicated the randomly oriented plate like grains of different sizes in bulk sample, evolving to longer and well-oriented grain structure through K-doping and LFZ. Because of the incongruent melting properties of compound, the high number of secondary phases formed in the as-grown samples. In order to reduce it, an annealing and K-doping process have been applied. The microstructural evolution is reflected on the electrical properties, and the lowest resistivity values are found in the annealed K-doped fibers. Seebeck coefficient is positive in all cases, pointing out to p-type conduction mechanism. These modifications led to PF values up to 0.162 mW/K 2 m, obtained in 0.10 K-doped annealed fibers at


Introduction
Thermoelectric (TE) materials are quite popular owing their capability of converting wasted heat energy to useful electric power using Seebeck effect [1]. The efficiency of such conversion is evaluated through the dimensionless figure of merit, ZT, defined as TS 2 /ρκ, where T is the absolute temperature, S the Seebeck coefficient, κ the thermal conductivity, and ρ the electrical resistivity. It seems that promising TE materials must possess high Seebeck coefficient, low thermal conductivity and electrical resistivity, and high working temperatures. TE materials comprise a large number of families involving binary phases of Bi-Te, Co-Sb, Pb-Te, silicides, as Mg-Si, or Mn-Si, and oxides [2]. Among these types, ceramic CoO-based TE materials are quite attractive owing to their abundance in the earth crust, high working temperatures, and relatively lower toxicity and costs than the other families [3,4]. The first discovered CoO-based material, NaxCoO2, exhibited large thermoelectric power breaking the general belief that oxides had poor thermoelectric properties [5]. This study led to the discovery of new layered cobaltites such as Ca-Co-O [6], Bi-Ca-Co-O [7], and Bi-Sr-Co-O [8]. From various crystallographic studies, it has been found that the crystal structures of CoO-based (TE) materials can be described through a monoclinic structure, which is composed of an alternate stacking of two different layers, namely CdI2type CoO2 conductive layer and rock salt (RS) Bi2X2O4 (X=Ca, Sr and Ba) insulating one.
These two layers have common a-and c-axis lattice parameters with different b-axis length, which causes a misfit along the b-direction [7,9,10]. Layered cobaltites are known to have a large crystallographic anisotropy, which is reflected in an anisotropic behavior of electrical properties. For example, Seebeck coefficient, S, may be tuned up through variations of misfit factor or the oxidation state of cations in the RS substructure [11]. These variations can be carried out through different material preparation procedures, as the alignment of plate-like grains by texturing techniques [12][13][14], or chemical processes as substituting different elements into the matrix [15][16][17][18][19].
The aim of the present study is to determine the effect of microstructure modification on the thermoelectric properties of Bi2Ca2Co2Oy ceramic materials prepared by solid state method and processed using different techniques. In addition, the influence of partial Ca by K substitution on the different processed materials will be evaluated.

Experimental Procedure
The Bi2Ca2-xKxCo2Oy (x = 0.0, 0.05, 0.075, 0.10 and 0.125) ceramic precursors were obtained from commercial Bi2O3 (Panreac, 98 + %), CaCO3 (98.5 %, Panreac), K2CO3 (Panreac, 98 + %), and Co2O3 (Aldrich, 98 + %) powders. They were weighed in stoichiometric proportions, and milled in a ball-mill for 30 minutes at 300 rpm in acetone medium. Then, the mixture was dried using infrared irradiation and manually re-grinded to produce fine powders, which were subsequently calcined twice at 750 and 800 ºC for 12 hours to decompose the metallic carbonates. Part of these powders has been pressed into pellets under 400 MPa applied pressure, and sintered at 810 ºC for 24 h with a final furnace cooling. The obtained materials after this process will be called bulk samples throughout the text. The other part of the powders were pressed into cylinders (around 3 mm diameter and 100 mm length) by means of isostatic pressure under 200 MPa, followed by texturing in a LFZ system, using previously described conditions [20]. The samples produced through the LFZ treatment will be called asgrown fibers. Moreover, due to the incongruent melting shown by this compound some of the as-grown fibers were annealed in an oven for 24 hours at 800°C with a final furnace cooling in order to increase the amount of thermoelectric phase [21]. The samples obtained after this process will be named as annealed fibers throughout the text. Moreover, these data have been used to calculate the thermoelectric power factor (PF, defined as S 2 /).

Results and discussion
XRD patterns of Bi2Ca2-xKxCo2Oy bulk, as-grown, and annealed fibers are given in Fig. 1.
From these figures, it is clear that most of the observed peaks correspond to Bi2Ca2Co2Oy thermoelectric phase, independently of the processing technique and K-substitution in agreement with previous studies [22]. Further, there is no significant peaks shift with K substitution, while the absence of K-based secondary phases could indicate its successful incorporation into the main matrix in all cases. On the other hand, some secondary phases are observed in all as-grown samples and in undoped bulk sample and annealed fibers. The larger content of secondary phases in as-grown samples, when compared to the other two types of samples prepared in this work, is related to the incongruent melting resulting from LFZ process and observed in similar systems [20,23]. Consequently, it can be easily deduced that annealing process at 800 °C for 12 h, drastically decreases the amount of these secondary phases as observed in the XRD patterns of the annealed fibers. In addition, secondary phases content is further decreased in bulk samples and annealed fibers with K-substitution.
In order to evaluate grain orientation in the different samples, Lotgering factor (LF), defined as LF = (P-P0)/(1-P0), where P is the sum of the peak intensities of preferred orientation, and P0 is the P of samples with randomly-oriented grains [24][25], has been calculated for all samples, and presented in Table 1 When comparing undoped specimens to the doped ones, it is evident that K-substitution increases density in bulk samples, eliminating the typical porosity produced through the solid state method [26], while slightly increases grain orientation in as-grown and annealed fibers.
When considering the secondary phases content, as-grown fibers show the highest content, as previously mentioned, due to the incongruent melting observed in these compounds.
However, annealing procedure drastically decreases black contrast (Co oxide, #3) and, in lower content, white one (Bi-Ca oxides, #2). In any case, the sizes of these Bi-Ca oxide phases are much larger than the observed in the bulk samples, but they display some orientation along the growth direction. All these observations fit well with the previously discussed XRD data. In addition, K + incorporation into the system further increases grain orientation most probably due to the decrease of melting point of the system associated to the formation of a Bi2O3-K2CO3 eutectic which decreases the thermal radial gradient during the LFZ processing, as observed in similar systems [27]. while for the bulk sample it is semiconductor-like up to 500 ºC, and metallic-like one above this temperature. In all cases, electrical resistivity values are decreased by K-doping, regardless of the processing technique. Since K + substitution for Ca 2+ reduces the total charge in the rock-salt layers and induces the promotion of Co 3+ to Co 4+ in the conduction layer, increasing the charge carrier concentration which leads to lower resistivity values [19]. On the other hand, the high resistivity observed in as-grown fibers is due to the high amount of secondary phases, together with a large content of oxygen vacancies, as previously reported [28]. Furthermore, the drastic decrease of electrical resistivity in annealed fibers, when compared to the as-grown ones, is due to the formation of thermoelectric phase from the secondary ones, the enhancement of grain orientation, and the decrease of oxygen vacancies which increase charge carrier concentration. This evolution can be easily observed in Table 1, where the resistivity values of all samples at 650 ºC are displayed. The minimum values at 650 ºC have been determined in 0.075K annealed fibers (26.4 m cm), which are same order of the Pb-substituted textured materials with much higher cationic substitution and prepared through solution method ( 30 m cm) [14], and much lower than the reported in sintered materials ( 70 m cm) [7]. On the other hand, the values are still far from the reported in single crystals at room temperature (6 m cm) [29].
The electrical conduction mechanism in misfit-layered Bi2Ca2-xKxCo2Oy is based on the small polaron hopping model [30][31][32] where resistivity is given as: where is the pre-exponential term related to the scattering mechanism, n the carrier concentration, e the elementary charge, the intersite distance of the hopping, the Boltzman constant, T the absolute temperature, and the activation energy. The linear fits of ln(ρ/T) versus 1000/T presented in Fig. 4 (for clarity for x=0 and 0.075K are presented) indicate that small polaron hopping transport model is a good description of the conduction mechanism of these samples. In order to calculate , the slopes of fits in Fig. 4 are used, and the results are presented in Table 1. According to the table, it can be observed that values for undoped samples are higher than the calculated for K-substituted samples, independently of the processing method. The decrease in the values is another explanation for the observed decrease in the electrical resistivity, as discussed above [31].
Seebeck coefficient variation with respect to temperature for all Bi2Ca2-xKxCo2Oy samples is given in Fig. 5. All Seebeck coefficients are positive in the whole measured temperature range which is an indication of p-type conduction mechanism. When comparing the values determined in the different samples, it is clear that as-grown fibers display the highest Seebeck coefficient at any temperature and composition. This effect is associated to a higher amount of oxygen vacancies in these fibers, as reported in previous works [33]. The maximum value of this coefficient at room temperature has been determined in 0.05K asgrown fibers (around 225 μV/K), much higher than the reported in single crystals ( 150 μV/K) [34]. A theoretical model of thermopower (or Seebeck coefficient) in cobalt oxides was proposed by Koshibae et al. [35], as follows: where kB is the Boltzmann constant, e the charge of electron, and x the concentration of Co +4 ions in the conduction band. By using this formula and the Seebeck values measured at 650 o C (see Table-1), the fraction of Co +4 in each sample has been calculated and presented in Table   1. From these values, it can be observed that all doped samples have higher Co +4 concentration, in agreement with the previous discussion in the electrical resistivity section.
According to Koshibae's model, there is no S dependence with temperature when measuring it at high temperatures. However, this model is not realistic enough as it is ignoring the peculiar splitting of the t2g levels in the CoO2 layer [36]. Hence, it is well known that, in real conditions, Seebeck coefficient values are influenced by temperature. Consequently, the highest S values for each sample have been determined at 650 ºC, independently of processing method. The maximum values in annealed samples at 650 ºC (about 225 μV/K) are lower than the measured in as-grown samples at the same temperature ( 285 μV/K). In any case, they are higher than the reported in sintered materials (∼ 150 μV/K) [37], or produced through solution methods ( 215 μV/K) [38], but lower than the measured in hot-pressed ceramics (∼ 250 μV/K) [39].
The estimation of thermoelectric performances of all samples has been made using power factor (PF) values, obtained from Seebeck coefficient and resistivity data, and represented in  Table-1, has been determined in 0.10K annealed fibers (0.162 mW/K 2 m), which is around eight times higher than the reported in sintered materials (∼ 0.02 mW/K 2 m) [37], slightly higher than the obtained in samples prepared through solution methods (∼ 0.09 mW/K 2 m) [38], but still lower than the reported in hot-pressed materials (∼ 0.25 mW/K 2 m) [39]. It should be highlighted that these last samples combine high grain orientation, high density, low amount of secondary phases, and probably optimal oxygen content in the thermoelectric phase.

Conclusions
In this study, Bi2Ca2-xKxCo2Oy (x=0, 0.05, 0.075, 0.10, and 0.125) bulk samples, as-grown and annealed fibers have been prepared by solid-state reaction, LFZ processing, and LFZ processing followed by annealing, respectively. XRD graphs showed that most of the observed peaks corresponded to the Bi2Ca2Co2Oy thermoelectric phase, showing the independent character of the processing technique and K-substitution. From SEM-EDS investigations, it has been determined that K-substitution reduced the amount of porosity and Co oxide secondary phase in bulk samples, and improved grain alignment in annealed fibers.
These structural and microstructural modifications were reflected in the electrical resistivity values, which were lower in K-substituted specimens than in undoped ones, regardless of the processing route. On the other hand, Seebeck coefficient values agreed with the electrical resistivity ones, being higher when the measured electrical resistivity was higher due to the decrease of Co 4+ concentration in undoped samples, when compared to the K-substituted ones. The best thermoelectric performances, evaluated using PF, were obtained in annealed fibers, which show the best combination of density, grain orientation, secondary phases content, and charge carrier concentration. However, these high PF values were still lower than the obtained in hot-pressed materials, probably due to the difficult oxygen diffusion through the fibers during annealing process due to their high density, resulting in higher oxygen vacancies than usual in sintered or hot-pressed materials.           ln(ρ/T) versus 1000/T graph of representative BiCa2-xKxCo2Oy samples, for x=0 and 0.075.