Development of Na0.5CoO2 Thick Film Prepared by Screen-Printing Process

The Na0.5Co0.9Cu0.1O2 thick film with the same thermoelectric performance as a Na0.5CoO2 bulk was formed on an alumina substrate by the screen-printing process. The power factor exceeded 0.3 mW/K2m, with the resistivity of 3.8 mΩcm and the thermopower of 108 μV/K. The thick film without any cracks strongly adhered to the substrate. The high-quality thick film had been realized through the carefully designed and improved process, mixing NaCl to promote the anisotropic sintering of Na0.5Co0.9Cu0.1O2, inserting a CuO interlayer to adhere the film and substrate, and Co–Cu substituting Cu for Co to control the sintering temperature.


Introduction
Thermoelectric (TE) materials can convert heat energy directly into electrical energy without any moving parts or emitting environmental pollutants. This clean power generation is expected to be a technology that can effectively utilize the waste heat from other power generations as well as various devices, because TE can generate power even from a small amount of thermal energy that cannot turn the turbine. In order to enhance the conversion efficiency, various inorganic materials, such as the Bi-Te system [1][2][3], SnSe [4,5], and Heusler compounds [6][7][8], have been studied steadily for a long time. Recently, flexible and/or wearable devices contributing to IoT have attracted attention, even in the field of thermoelectricity [9], and organic thermoelectric materials such as PEDOT:PSS (Poly (3,4-ethylenedioxythiophene):Poly(styrenesulfonate)) are also being actively studied [10].
Harvesting diluted waste heat requires a module design that can concentrate a heat flow into TE materials, and a planar TE device is one of the ways [11,12]. Such a TE device could potentially replace the battery in IoT devices which need a power of several mW, otherwise the battery should be replaced every month. Furthermore, the planar-shaped device realized by the printing process could be freely designed in size and shape, and would be able to propose an appropriate design for various heat sources. A screen-printing process is the most popular and traditional of the various printing processes, because the low process cost and high productivity are suitable for industrial applications. The process has long been used for conductive circuit patterning, and many pastes for metal thick films are commercially available. Even today, the processes and pastes are investigated to realize fine line patterning for the product miniaturization and integration [13,14], and the latest research has achieved printing with a width of 30 µm or less using silver paste [15]. The development of carbon-based conductive circuit printing for flexible devices is also active [16,17], but there are few reports on ceramic material printing, with a serious problem in thick film sintering such as a cracking due to the sintering shrinkage [18,19].
We chose oxides as the thermoelectric material, favoring the high stability and low toxicity. The high thermoelectric performance of the metallic layered transition-metal oxide, Na 0.5 CoO 2 , as a p-type TE material had been reported by Terasaki et al. [20]. The low resistivity originates in its high carrier density, whereas the high thermopower is maintained by the orbital and spin entropies derived from the Mott insulator. The thermoelectric power factor S 2 ρ −1 , where ρ is resistivity and S is thermopower, of a Na 0.5 CoO 2 single crystal reaches 5 mW/K 2 m at 300 K in the in-plane direction [20] and exceeds 4 mW/K 2 m of Bi 2 Te 3 [21]. Since Na 0.5 CoO 2 shows anisotropic electrical conductivity and the resistivity in the out-of-plane direction is approximately 50 times higher than that in the in-plane direction, the power factor of Na 0.5 CoO 2 polycrystalline sample is 0.25 mW/K 2 m at 300 K. However, Na 0.5 CoO 2 demonstrates superior performance compared to the other oxide materials, and is a promising thermoelectric material.
In this paper, we report the thermoelectric properties and structure of developed Na 0.5 CoO 2 thick film and discuss the significance of our improved process. In particular, we have introduced a sintering additive into the thick film and an interlayer between the thick film and a substrate, and further adjusted the composition of the thick film.

Experimental
The Na 0.5 Co 0.9 Cu 0.1 O 2 powder was prepared by a solid-state reaction. Na 2 CO 3 , Co 3 O 4 , and CuO were mixed and calcined at 860°C for 12 h in air. The calcined products were ground, pressed into a pellet, and sintered at 920°C for 12 h in air. The Na 0.5 Co 0.9 Cu 0.1 O 2 powder was obtained via a mechanical grinding and ball milling of the pellet. The Na 0.5 CoO 2 powder was prepared by the same process from Na 2 CO 3 and Co 3 O 4 .
The paste for the main Na 0.5 Co 0.9 Cu 0.1 O 2 layer was comprised of a mixture of synthesized Na 0.5 Co 0.9 Cu 0.1 O 2 powder and ground NaCl powder, a vehicle of Terpineol and Ethyl cellulose. The volume fraction of NaCl in the mixture was optimized to 20 vol%. The paste for the CuO interlayer consisted of CuO powder and the same vehicle as that of the main layer. The pastes of Na 0.5 CoO 2 without NaCl, and Na 0.5 CoO 2 with 20 vol% NaCl, were also prepared by mixing the same vehicle.
The CuO paste was printed on an alumina substrate (10 mm × 20 mm × 1 mm) in a rectangular shape (10 mm × 2 mm) using a screen-printing process. The printed film was sintered at 950°C for 12 h in air after drying the vehicle at 200°C for 5 min. Subsequently, the NaCl-mixed Na 0.5 Co 0.9 Cu 0.1 O 2 paste was printed on the sintered CuO interlayer in the same shape. The vehicle was dried in the same manner as mentioned. The screen printing of both layers was performed at the print speed of 50 mm/s using a stainless steel mesh screen with 250-mesh/inch by a semi-auto screen printing machine (SERIA CORPORATION, SSA-PC150E-IP, Tokyo, Japan). In order to achieve 100 µm of the thickness after drying, the Na 0.5 Co 0.9 Cu 0.1 O 2 paste was laminated 10 times under the same printing condition. The sintering of the dried film was carried out at 920°C for 1 h in air.
In this paper, two other kinds of thick films were also prepared for comparison. One is a Na 0.5 CoO 2 thick film prepared by a simple screen-printing process. Here, only the Na 0.5 CoO 2 paste was printed and laminated 10 times on an alumina substrate in a rectangular shape (10 mm × 5 mm). Another thick film was prepared by a partially improved process. The NaCl-mixed Na 0.5 CoO 2 paste was printed and laminated 10 times on a CuO interlayer, which was printed and sintered on an alumina substrate in a rectangular shape (10 mm × 2 mm). The printing, drying, and sintering conditions of both of the main layers (the Na 0.5 CoO 2 layer and the NaCl-mixed Na 0.5 CoO 2 layer) and the CuO interlayer were the same with those of the NaCl-mixed Na 0.5 Co 0.9 Cu 0.1 O 2 layer and the CuO interlayer as mentioned above, respectively. X-ray diffraction (XRD) patterns of the thick film were taken with CuKα radiation using a standard diffractometer with a monochromator (Rigaku SmartLab, Tokyo, Japan). The XRD measurement was performed in 2θ/θ-scan mode against the sample surface, i.e., a plane parallel to the substrate surface.
The resistivity and thermopower were measured by the four-point probe method and EMFs under temperature gradients using Pt thermocouples, respectively, in air from 50°C to 550°C (Ozawa Science RZ2001S, Nagoya, Japan). The morphology and element mapping of the thick film were observed using a scanning electron microscope (SEM; JEOL, JSM-5600, Tokyo, Japan) and energy-dispersive X-ray spectrometry (EDX; JEOL, EX-54145JMU, Tokyo, Japan). The thermal behaviors of the Na 0.5 CoO 2 and Na 0.5 Co 0.9 Cu 0.1 O 2 powders were analyzed by a thermogravimetry and differential thermal analyzer (TG/DTA; BRUKER, MTC1000SA, Billerica, MA, USA) from 15°C to 950°C in air.

Problems with a Thick Film Prepared by a Simple Printing Process
As a first approach to forming a Na 0.5 CoO 2 thick film, we simply screen-printed and sintered the Na 0.5 CoO 2 paste on an alumina substrate. Figure 1a,b show a cross-sectional schematic of a printed thick film, and photographs of a dried and a sintered thick film prepared by the simple screen-printing process, respectively. The thick film was cracked severely after sintering and easily peeled from the substrate. In addition, Na 0.5 CoO 2 has decomposed to Co 3 O 4 due to the evaporation of Na during sintering, as shown in Figure 1c. ℃ to 550 ℃ (Ozawa Science RZ2001S, Nagoya, Japan). The morphology and element mapping of the thick film were observed using a scanning electron microscope (SEM; JEOL, JSM-5600, Tokyo, Japan) and energy-dispersive X-ray spectrometry (EDX; JEOL, EX-54145JMU, Tokyo, Japan). The thermal behaviors of the Na0.5CoO2 and Na0.5Co0.9Cu0.1O2 powders were analyzed by a thermogravimetry and differential thermal analyzer (TG/DTA; BRUKER, MTC1000SA, Billerica, MA, USA) from 15 ℃ to 950 ℃ in air.

Problems with a Thick Film Prepared by a Simple Printing Process
As a first approach to forming a Na0.5CoO2 thick film, we simply screen-printed and sintered the Na0.5CoO2 paste on an alumina substrate. Figure 1a,b show a cross-sectional schematic of a printed thick film, and photographs of a dried and a sintered thick film prepared by the simple screenprinting process, respectively. The thick film was cracked severely after sintering and easily peeled from the substrate. In addition, Na0.5CoO2 has decomposed to Co3O4 due to the evaporation of Na during sintering, as shown in Figure 1c.   In order to suppress the evaporation of Na, we employed flux sintering using NaCl. Furthermore, the interface between the film and substrate was modified by an insertion of a CuO interlayer in order to improve the adhesion strength. The cross-sectional schematic and the photograph of a partially improved Na 0.5 CoO 2 thick film with NaCl mixing and a CuO interlayer introduction are shown in Figure 2a,b, respectively. NaCl mixing drastically changes the appearance of the film to be free of cracks, and more importantly, the decomposition was totally prevented. As expected, the CuO interlayer introduction improved the adhesion strength and prevented the peeling of the thick film.
to improve the adhesion strength. The cross-sectional schematic and the photograph of a partially improved Na0.5CoO2 thick film with NaCl mixing and a CuO interlayer introduction are shown in Figure 2a,b, respectively. NaCl mixing drastically changes the appearance of the film to be free of cracks, and more importantly, the decomposition was totally prevented. As expected, the CuO interlayer introduction improved the adhesion strength and prevented the peeling of the thick film.
However, bending and arching of the film on the CuO interlayer were found due to the sintering shrinkage. Through several attempts, we found that Cu substitution into Co site prevent the delamination. TG/DTA analysis revealed the difference in the stability between Na0.5CoO2 and Na0.5Co0.9Cu0.1O2 as shown in Figure 3. The weight loss was significantly reduced by the substitution. In addition to that, a faint endothermic peak at 890 ℃ shifted to 830 ℃. K. Park et al. has reported that this Co-Cu substitution reduces the melting temperature of Na0.5CoO2 [22]. Although the observed peaks are much lower than the melting point, the shift indicates the reduced temperature stability by the substitution as well. In other words, Na0.5Co0.9Cu0.1O2 is more reactive than Na0.5CoO2, which is beneficial to improve sintering. As a result, well-adhered thick film without any cracks was attained as shown later.   Figure 4a shows the cross-sectional schematic of the Na0.5Co0.9Cu0.1O2 thick film prepared by the improved process. Compared to the partially improved thick film shown in Figure 2a, the composition of the main layer had been changed from Na0.5CoO2 to Na0.5Co0.9Cu0.1O2. Figure 4 However, bending and arching of the film on the CuO interlayer were found due to the sintering shrinkage. Through several attempts, we found that Cu substitution into Co site prevent the delamination. TG/DTA analysis revealed the difference in the stability between Na 0.5 CoO 2 and Na 0.5 Co 0.9 Cu 0.1 O 2 as shown in Figure 3. The weight loss was significantly reduced by the substitution. In addition to that, a faint endothermic peak at 890°C shifted to 830°C. K. Park et al. has reported that this Co-Cu substitution reduces the melting temperature of Na 0.5 CoO 2 [22]. Although the observed peaks are much lower than the melting point, the shift indicates the reduced temperature stability by the substitution as well. In other words, Na 0.5 Co 0.9 Cu 0.1 O 2 is more reactive than Na 0.5 CoO 2 , which is beneficial to improve sintering. As a result, well-adhered thick film without any cracks was attained as shown later.

Microstructure of Crack-Free Thick Film
Materials 2020, 13, x FOR PEER REVIEW 4 of 11 In order to suppress the evaporation of Na, we employed flux sintering using NaCl. Furthermore, the interface between the film and substrate was modified by an insertion of a CuO interlayer in order to improve the adhesion strength. The cross-sectional schematic and the photograph of a partially improved Na0.5CoO2 thick film with NaCl mixing and a CuO interlayer introduction are shown in Figure 2a,b, respectively. NaCl mixing drastically changes the appearance of the film to be free of cracks, and more importantly, the decomposition was totally prevented. As expected, the CuO interlayer introduction improved the adhesion strength and prevented the peeling of the thick film.
However, bending and arching of the film on the CuO interlayer were found due to the sintering shrinkage. Through several attempts, we found that Cu substitution into Co site prevent the delamination. TG/DTA analysis revealed the difference in the stability between Na0.5CoO2 and Na0.5Co0.9Cu0.1O2 as shown in Figure 3. The weight loss was significantly reduced by the substitution. In addition to that, a faint endothermic peak at 890 ℃ shifted to 830 ℃. K. Park et al. has reported that this Co-Cu substitution reduces the melting temperature of Na0.5CoO2 [22]. Although the observed peaks are much lower than the melting point, the shift indicates the reduced temperature stability by the substitution as well. In other words, Na0.5Co0.9Cu0.1O2 is more reactive than Na0.5CoO2, which is beneficial to improve sintering. As a result, well-adhered thick film without any cracks was attained as shown later.   Figure 4a shows the cross-sectional schematic of the Na0.5Co0.9Cu0.1O2 thick film prepared by the improved process. Compared to the partially improved thick film shown in Figure 2a, the composition of the main layer had been changed from Na0.5CoO2 to Na0.5Co0.9Cu0.1O2. Figure 4(b1) shows a photograph of the entire Na0.5Co0.9Cu0.1O2 thick film, and Figure 4 Figure 4a shows the cross-sectional schematic of the Na 0.5 Co 0.9 Cu 0.1 O 2 thick film prepared by the improved process. Compared to the partially improved thick film shown in Figure 2a, the composition of the main layer had been changed from Na 0.5 CoO 2 to Na 0.5 Co 0.9 Cu 0.1 O 2 . Figure 4(b1) shows a photograph of the entire Na 0.5 Co 0.9 Cu 0.1 O 2 thick film, and Figure 4(b2) shows an enlarged portion of the area surrounded by the red broken line in Figure 4(b1). The thick film has a glossy surface, indicating a finely sintered dense film. No cracks can be seen with an optical microscope. In addition, the thick film is strongly adhered to an alumina substrate and passed the adhesive tape peeling test. This appearance fully satisfies the requirements for a thick film structure and is expected to perform well.

Microstructure of Crack-Free Thick Film
correspond to the Na0.5CoO2 phase. Co3O4, which is thought to have been formed by the decomposition of a small amount of Na0.5Co0.9Cu0.1O2 by the evaporation of Na, is confirmed as an impurity. It should be noted that the Na0.5CoO2 phase, which was completely degraded in the trial sample, remained mostly in the improved thick film. Since the sintering temperature and atmosphere of the films were exactly the same, the decomposition of the Na0.5CoO2 phase was suppressed by the Na-rich surroundings with mixed NaCl. The peaks corresponding to the (00n) planes of the Na0.5CoO2 phase are stronger than the other peaks compared to reported powder diffraction patterns. Namely, the c-axis of the Na0.5CoO2 phase is oriented perpendicular to the substrate, at least near the film surface where the X-ray can penetrate. The c-axis length, calculated from the 00n peaks, are 10.95 Å , and it corresponds to the Na-rich composition of Na0.68CoO2 in comparison with data reported by Shu et al. [23]. The excess Na, induced from NaCl, has been intercalated into Na0.5Co0.9Cu0.1O2. Although the Co-Cu substitution should change the axis length, the effect on the c-axis should be slight since the ionic radii of Co and Cu are similar. Certainly, the 00n peak shift with the Co-Cu substitution has been invisible in the previous report [22,24].    Figure 4c shows the XRD pattern of the Na 0.5 Co 0.9 Cu 0.1 O 2 thick film. The indexed peaks correspond to the Na 0.5 CoO 2 phase. Co 3 O 4 , which is thought to have been formed by the decomposition of a small amount of Na 0.5 Co 0.9 Cu 0.1 O 2 by the evaporation of Na, is confirmed as an impurity. It should be noted that the Na 0.5 CoO 2 phase, which was completely degraded in the trial sample, remained mostly in the improved thick film. Since the sintering temperature and atmosphere of the films were exactly the same, the decomposition of the Na 0.5 CoO 2 phase was suppressed by the Na-rich surroundings with mixed NaCl. The peaks corresponding to the (00n) planes of the Na 0.5 CoO 2 phase are stronger than the other peaks compared to reported powder diffraction patterns. Namely, the c-axis of the Na 0.5 CoO 2 phase is oriented perpendicular to the substrate, at least near the film surface where the X-ray can penetrate. The c-axis length, calculated from the 00n peaks, are 10.95 Å, and it corresponds to the Na-rich composition of Na 0.68 CoO 2 in comparison with data reported by Shu et al. [23]. The excess Na, induced from NaCl, has been intercalated into Na 0.5 Co 0.9 Cu 0.1 O 2 . Although the Co-Cu substitution should change the axis length, the effect on the c-axis should be slight since the ionic radii of Co and Cu are similar. Certainly, the 00n peak shift with the Co-Cu substitution has been invisible in the previous report [22,24].
The cross-sectional SEM observation image of the Na 0.5 Co 0.9 Cu 0.1 O 2 thick film is shown in Figure 5(a1). The well-sintered CuO interlayer with a thickness of approximately 15 µm is clearly observed on the alumina substrate, on which the Na 0.5 Co 0.9 Cu 0.1 O 2 layer with a thickness of 50 µm is laminated. No cracks can be observed, and all layers are firmly adhered. It is noteworthy that the Na 0.5 Co 0.9 Cu 0.1 O 2 layer clearly separates a 30 µm-thick plate-crystal layer from the surface and a 20 µm-thick isotropic-crystal layer near the CuO interlayer. Figure 5(a2) shows an enlarged image of the interface between the layers. So far, preferential growth of Na 0.5 CoO 2 has been reported in the templated grain growth process using NaCl flux [25] as well as hot-forged sintering [26], leading to the well-aligned microstructure along the ab-plane. In our Na 0.5 Co 0.9 Cu 0.1 O 2 thick film, the mixed NaCl would have acted as a flux that promotes the crystal growth of Na 0.5 Co 0.9 Cu 0.1 O 2 . This leads to the growth of plate-crystals, reflecting its layered crystal structure and crystal orientation along with the interface. Typically, cracks in printed thick films are caused by the mismatch between the sintering shrinkage of the film and in-plane constraint from the substrate. Then, brittle ceramic materials are hard to sinter into crack-free thick films (compared to metallic materials that can deform). On the contrary, in our flux sintered thick film, the crystals grew in the in-plane direction of the film, supporting the anisotropic sintering of the film in such a way that effectively suppressed crack formation. Notably, with both the layered crystal structure of Na 0.5 Co 0.9 Cu 0.1 O 2 and the crystal growth with NaCl flux, the isotropic crystal growth of Na 0.5 Co 0.9 Cu 0.1 O 2 just above the CuO interlayer is rather unnatural.
Materials 2020, 13, x FOR PEER REVIEW 6 of 11 The cross-sectional SEM observation image of the Na0.5Co0.9Cu0.1O2 thick film is shown in Figure  5(a1). The well-sintered CuO interlayer with a thickness of approximately 15 μm is clearly observed on the alumina substrate, on which the Na0.5Co0.9Cu0.1O2 layer with a thickness of 50 μm is laminated. No cracks can be observed, and all layers are firmly adhered. It is noteworthy that the Na0.5Co0.9Cu0.1O2 layer clearly separates a 30 μm-thick plate-crystal layer from the surface and a 20 μm-thick isotropic-crystal layer near the CuO interlayer. Figure 5(a2) shows an enlarged image of the interface between the layers. So far, preferential growth of Na0.5CoO2 has been reported in the templated grain growth process using NaCl flux [25] as well as hot-forged sintering [26], leading to the well-aligned microstructure along the ab-plane. In our Na0.5Co0.9Cu0.1O2 thick film, the mixed NaCl would have acted as a flux that promotes the crystal growth of Na0.5Co0.9Cu0.1O2. This leads to the growth of plate-crystals, reflecting its layered crystal structure and crystal orientation along with the interface. Typically, cracks in printed thick films are caused by the mismatch between the sintering shrinkage of the film and in-plane constraint from the substrate. Then, brittle ceramic materials are hard to sinter into crack-free thick films (compared to metallic materials that can deform). On the contrary, in our flux sintered thick film, the crystals grew in the in-plane direction of the film, supporting the anisotropic sintering of the film in such a way that effectively suppressed crack formation. Notably, with both the layered crystal structure of Na0.5Co0.9Cu0.1O2 and the crystal growth with NaCl flux, the isotropic crystal growth of Na0.5Co0.9Cu0.1O2 just above the CuO interlayer is rather unnatural.  Figure 5(b1,b2) show the cross-sectional element mapping images of the Na 0.5 Co 0.9 Cu 0.1 O 2 thick film for Co and Cu, respectively. In the plate-crystal layer, Co and Cu coexist, and both are uniformly distributed throughout the layer, indicating that Cu substitutes to Co site in Na 0.5 CoO 2 . On the other hand, in the isotropic-crystal layer, Co is uniformly distributed throughout the layer just as with the plate-crystal layer, but the Cu distribution has a distinct gradation. The high brightness of Cu concentration in the isotropic-crystal layer is approximately the same with the CuO interlayer, suggesting the presence of Cu-rich impurities. Co is also found in the CuO interlayer. Co can substitute Cu site in CuO, thus Co ions would diffuse into the CuO interlayer during the sintering, and NaCl flux would promote the diffusion.
In order to identify the compounds contained in each layer, the thick film was polished from the surface, and each exposed layer was measured by XRD. Figure 6 shows XRD patterns of each layer. The indexed peaks and the unassigned peaks correspond to the Na 0.5 CoO 2 phase and alumina substrate, respectively. The peaks of the Na 0.5 CoO 2 phase are located at the same angle in all layers. There are peaks of CuO in addition to the peaks of the Na 0.5 CoO 2 phase in the isotropic-crystal layer, and Co 3 O 4 observed in the plate-crystal layer. Hence, the Cu-rich impurity observed in the elemental mapping image is CuO, and the isotropic-crystal layer is a composite of Na 0.5 Co 0.9 Cu 0.1 O 2 , Co 3 O 4 , and CuO. CuO forms large grains because Cu-rich grains are relatively large in the mapping image. Such CuO with a C12/c1 space group of monoclinic should form isotropic grains and may act as a grain growth inhibitor for Na 0.5 Co 0.9 Cu 0.1 O 2 , resulting in the isotropic Na 0.5 Co 0.9 Cu 0.1 O 2 crystals. The origin of Cu-source that form CuO grains should be the CuO interlayer exuded as a liquid phase or dissolved in NaCl flux, since the Cu concentration in Na 0.5 Co 0.9 Cu 0.1 O 2 is much lower than that of CuO. Co in the CuO interlayer formed Co 3 O 4 .
Materials 2020, 13, x FOR PEER REVIEW 7 of 11 Figure 5(b1,b2) show the cross-sectional element mapping images of the Na0.5Co0.9Cu0.1O2 thick film for Co and Cu, respectively. In the plate-crystal layer, Co and Cu coexist, and both are uniformly distributed throughout the layer, indicating that Cu substitutes to Co site in Na0.5CoO2. On the other hand, in the isotropic-crystal layer, Co is uniformly distributed throughout the layer just as with the plate-crystal layer, but the Cu distribution has a distinct gradation. The high brightness of Cu concentration in the isotropic-crystal layer is approximately the same with the CuO interlayer, suggesting the presence of Cu-rich impurities. Co is also found in the CuO interlayer. Co can substitute Cu site in CuO, thus Co ions would diffuse into the CuO interlayer during the sintering, and NaCl flux would promote the diffusion.
In order to identify the compounds contained in each layer, the thick film was polished from the surface, and each exposed layer was measured by XRD. Figure 6 shows XRD patterns of each layer. The indexed peaks and the unassigned peaks correspond to the Na0.5CoO2 phase and alumina substrate, respectively. The peaks of the Na0.5CoO2 phase are located at the same angle in all layers. There are peaks of CuO in addition to the peaks of the Na0.5CoO2 phase in the isotropic-crystal layer, and Co3O4 observed in the plate-crystal layer. Hence, the Cu-rich impurity observed in the elemental mapping image is CuO, and the isotropic-crystal layer is a composite of Na0.5Co0.9Cu0.1O2, Co3O4, and CuO. CuO forms large grains because Cu-rich grains are relatively large in the mapping image. Such CuO with a C12/c1 space group of monoclinic should form isotropic grains and may act as a grain growth inhibitor for Na0.5Co0.9Cu0.1O2, resulting in the isotropic Na0.5Co0.9Cu0.1O2 crystals. The origin of Cu-source that form CuO grains should be the CuO interlayer exuded as a liquid phase or dissolved in NaCl flux, since the Cu concentration in Na0.5Co0.9Cu0.1O2 is much lower than that of CuO. Co in the CuO interlayer formed Co3O4.
CuO has been sintered in a liquid phase at an interlayer sintering temperature of 950 ℃, and penetrated into a Na0.5Co0.9Cu0.1O2 layer with NaCl flux. Therefore, the CuO interlayer adheres to both of the alumina substrate and the Na0.5Co0.9Cu0.1O2 thick film by the anchor effect, and adheres the thick film to the substrate strongly. In addition, CuO is compatible with alumina substrates from the viewpoint of thermal expansion [27,28]. . XRD (CuKα) patterns of the plate-crystal layer, the isotropic crystal layer, the CuO interlayer, and the alumina substrate in the Na0.5Co0.9Cu0.1O2 thick film. Each layer was exposed by polishing from the surface of the film. Figure 7 shows the temperature (T), dependence of resistivity (ρ), thermopower (S), and power factor (S 2 ρ −1 ) of a Na0.5CoO2 bulk, a Na0.5Co0.9Cu0.1O2 bulk, and the Na0.5Co0.9Cu0.1O2 thick film. The   Figure 6. XRD (CuKα) patterns of the plate-crystal layer, the isotropic crystal layer, the CuO interlayer, and the alumina substrate in the Na 0.5 Co 0.9 Cu 0.1 O 2 thick film. Each layer was exposed by polishing from the surface of the film.

Thermoelectric Properties of Na0.5Co0.9Cu0.1O2 Thick Film
CuO has been sintered in a liquid phase at an interlayer sintering temperature of 950°C, and penetrated into a Na 0.5 Co 0.9 Cu 0.1 O 2 layer with NaCl flux. Therefore, the CuO interlayer adheres to both of the alumina substrate and the Na 0.5 Co 0.9 Cu 0.1 O 2 thick film by the anchor effect, and adheres the thick film to the substrate strongly. In addition, CuO is compatible with alumina substrates from the viewpoint of thermal expansion [27,28].
3.3. Thermoelectric Properties of Na 0.5 Co 0.9 Cu 0.1 O 2 Thick Film Figure 7 shows the temperature (T), dependence of resistivity (ρ), thermopower (S), and power factor (S 2 ρ −1 ) of a Na 0.5 CoO 2 bulk, a Na 0.5 Co 0.9 Cu 0.1 O 2 bulk, and the Na 0.5 Co 0.9 Cu 0.1 O 2 thick film. The Na 0.5 CoO 2 bulk showed similar thermoelectric properties with various reported polycrystalline samples. On the other hand, although the thermopower of our Na 0.5 Co 0.9 Cu 0.1 O 2 bulk was a similar value to the Na 0.5 Co 0.9 Cu 0.1 O 2 bulk reported by Terasaki et al. [29], the resistivity was more than twice the reported values [22,29]. The cause of the high resistivity of the bulk should be its low density due to non-optimized process. The Na 0.5 Co 0.9 Cu 0.1 O 2 thick film shows lower resistivity than our Na 0.5 CoO 2 bulk and the reported Na 0.5 Co 0.9 Cu 0.1 O 2 bulk at all temperatures. It has been reported that the resistivity in the in-plane direction is one order of magnitude lower than the resistivity in the out-of-plane direction [20]. Co-Cu substitution and Na content in Na 0.5 CoO 2 also have an influence on resistivity, because both change the carrier density through reducing the amount, or changing the valence, of Co. However, the change in resistivity caused by the factor related to composition is smaller than that caused by the anisotropy. Hence the preferred orientation of ab-plane is responsible for the low resistivity. Certainly, the textured Na 0.5 CoO 2 ceramics, with the well aligned ab-plane prepared by the reactive templated grain growth (RTGG) method [30], show lowest resistivity in Figure 7a. Na0.5CoO2 bulk showed similar thermoelectric properties with various reported polycrystalline samples. On the other hand, although the thermopower of our Na0.5Co0.9Cu0.1O2 bulk was a similar value to the Na0.5Co0.9Cu0.1O2 bulk reported by Terasaki et al. [29], the resistivity was more than twice the reported values [22,29]. The cause of the high resistivity of the bulk should be its low density due to non-optimized process. The Na0.5Co0.9Cu0.1O2 thick film shows lower resistivity than our Na0.5CoO2 bulk and the reported Na0.5Co0.9Cu0.1O2 bulk at all temperatures. It has been reported that the resistivity in the in-plane direction is one order of magnitude lower than the resistivity in the out-ofplane direction [20]. Co-Cu substitution and Na content in Na0.5CoO2 also have an influence on resistivity, because both change the carrier density through reducing the amount, or changing the valence, of Co. However, the change in resistivity caused by the factor related to composition is smaller than that caused by the anisotropy. Hence the preferred orientation of ab-plane is responsible for the low resistivity. Certainly, the textured Na0.5CoO2 ceramics, with the well aligned ab-plane prepared by the reactive templated grain growth (RTGG) method [30], show lowest resistivity in Figure 7a.
The thermopower of the Na0.5Co0.9Cu0.1O2 thick film was lower than that of the bulk samples. The thermopower has been reported to be enhanced by a Co-Cu substitution in Na0.5CoO2, and increased with increasing Na content in a previous study [29,31]. As shown in Figure 7a, Na0.5Co0.9Cu0.1O2 bulk shows higher thermopower than the Na0.5CoO2 bulk. The thermopower of the Na0.5Co0.9Cu0.1O2 bulk reported by Park et al. has exceeded 400 µ V/K at 600 ℃ [22], and it is much higher than the other reports and the theoretical value calculated from the Heikes formula with the valence ratio of Co 3+ :Co 4+ = 1:1 [32,33], but the reason for this is not clear. Nevertheless, the Na0.5Co0.9Cu0.1O2 thick film, which is expected to be a Na-rich composition with Co-Cu substitution, shows lower thermopower than the Na0.5CoO2 bulk.  Temperature dependence of (a) resistivity, thermopower, and (b) power factor for a Na 0.5 CoO 2 bulk, a Na 0.5 Co 0.9 Cu 0.1 O 2 bulk, and the Na 0.5 Co 0.9 Cu 0.1 O 2 thick film measured in air.
The thermopower of the Na 0.5 Co 0.9 Cu 0.1 O 2 thick film was lower than that of the bulk samples. The thermopower has been reported to be enhanced by a Co-Cu substitution in Na 0.5 CoO 2 , and increased with increasing Na content in a previous study [29,31]. As shown in Figure 7a, Na 0.5 Co 0.9 Cu 0.1 O 2 bulk shows higher thermopower than the Na 0.5 CoO 2 bulk. The thermopower of the Na 0.5 Co 0.9 Cu 0.1 O 2 bulk reported by Park et al. has exceeded 400 µV/K at 600°C [22], and it is much higher than the other reports and the theoretical value calculated from the Heikes formula