A. Fabrication and characterization

Substrate temperature is known to be a critical parameter in the thin film fabrication, and therefore, its optimization is necessary. ^{Reference Ohtake, Nukaga, Kirino and Futamoto17,Reference Sathyamoorthy, Narayandaas and Mangalaraj18} Initially, Ca₃Co₄O₉ thin film was fabricated on the LAO substrate at three different substrate temperatures, i.e., 550, 600, and 750 °C. All the parameters except the substrate temperature were kept constant. For the thin film deposited at 550 °C, the crystallinity was found to be quite poor. This can be attributed to the PLD kinetics, where high substrate temperature can provide sufficient surface diffusion to allow surface atoms to migrate toward thermodynamically stable sites and crystallize. ^{Reference Sun, Hng, Yan and Ma10} Therefore, increasing the substrate temperature to 600 °C improved the crystallinity of the film, as seen in the XRD pattern [Fig. 1(a)], which shows diffraction peaks corresponding to only (001) Ca₃Co₄O₉ film reflections or reflections from the substrate. The XRD pattern therefore shows a dominant c-axis alignment with no evidence of the secondary phase in this film. Increasing the substrate temperature further to 750 °C partly decomposed the Ca₃Co₄O₉ phase into Ca₃Co₂O₆ phase, ^{Reference Zhang, Zhang, Lu and Zhang19} with reflections marked in the XRD pattern [Fig. 1(b)].

FIG. 1.

XRD patterns of pure Ca₃Co₄O₉ (CCO) thin films deposited on LAO substrate at a substrate temperature of (a) 600 °C and (b) 750 °C.

Several kinds of growth modes and characteristic structures can be observed during the deposition of atoms on a flat substrate. ^{Reference Gilmer, Huang and Roland20} A layer-by-layer or an epitaxial growth mode is called the Frank–van der Merwe mode (FM). However, sometimes the FM structure is found to exhibit undulations in the thickness, which is an indication of the uniform film being unstable. This leads to an eventual breakup of the FM film structure into a Stranski–Krastanov mode (SK) or Volmer–Weber mode (VW). The SK growth method is observed in our Ca₃Co₄O₉ thin film. It is described as an initial 2D growth of a few monolayers with grains oriented in a (001) orientation [Fig. 2(a)], followed by 3D cluster/island growth on these monolayers [Fig. 2(b)]. In the literature, different kinds of growth modes such as VW ^{Reference Sun, Hng, Yan and Ma10} and columnar ^{Reference Cieniek and Kac21} have been achieved for the PLD deposited Ca₃Co₄O₉ system apart from the epitaxial growth. ^{Reference Hu, Si, Sutter and Li22}

FIG. 2.

SEM micrographs of the surface of the thin Ca₃Co₄O₉ film showing (a) the initial 2D growth of monolayers with grains aligned in a (001) orientation, and (b) lower magnification image of the film showing the 3D cluster growth. (c) and (d) are the EDS spectra collected from Spot 1 and Spot 2 of (a), respectively.

AFM images of the as-grown Ca₃Co₄O₉ thin film also confirm the SK growth structure. A height profile of the 2D thin film was taken over the location indicated in the AFM image in Fig. 3(a). A 1.4-μm long line profile in Fig 3(b) shows that the thickness of the 2D thin film is ∼14 nm, and the clusters are an average of ∼90 nm high. Energy dispersive spectroscopy (EDS) spectra were collected from different spots on the thin film, and it was observed that the clusters were comprised of Ca _x CoO₂ phase along with excess Ca. For example, the EDS spectra in Figs. 2(c) and 2(d) were taken from the thin film and the cluster, respectively, and clearly exhibit the difference in the Ca content between the two spots. A similar result was reported by Cieniek et al. ^{Reference Cieniek and Kac21} in their study of (Co, Ca)O PLD thin films, as they also observed Ca rich droplets on their thin films, the origin of which is not very clear. From the quantitative EDS data, the cation ratio of the 2D part of the thin film (Spot 1) was determined to be Ca:Co = 3:4.1, which shows that the stoichiometry in the thin film is maintained. It is stated in the literature ^{Reference Sun, Hng, Yan and Ma10} that during the PLD synthesis, the Ca _x CoO₂ structure forms first, particularly at the boundaries of nuclei with high defect density. The Ca to Co cationic ratio from Spot 2 (clusters) is determined to be 1.16:1. Although, the elemental composition of Ca _x CoO₂ lies between 0.26 ≤ x ≤ 0.5, ^{Reference Sun, Hng, Yan and Ma10} it is also known that excess Ca exists when Ca _x CoO₂ phase is formed on the substrate. This could explain the excess Ca in the clusters. However, if given enough ripening time during the coalescence process of the nuclei, a well-aligned CCO thin film can be formed. In a different process, Ca _x CoO₂ can also combine with ambient Ca and oxygen and transform into the CCO structure, again giving us a well-textured CCO thin film. Following the above argument, the clusters can be thought of as the initial formation of the CCO structure, and therefore, they consist of Ca _x CoO₂ phase along with excess Ca as indicated by EDS analysis.

FIG. 3.

(a) AFM images of the Ca₃Co₄O₉ thin film on LAO substrate deposited at a substrate temperature of 600 °C. (b) Height profile of the 2D thin film taken over the location indicated by the red line in the AFM image in (a). (c) Three dimensional AFM image of the thin film.

Ca₃Co₄O₉ thin film with 2 at.% Bi doping was also deposited on the LAO substrate. The SEM micrographs in Fig. 4 show the surface morphology of the Ca_2.8Bi_0.2Co₄O₉ thin film. Compared to the pure Ca₃Co₄O₉ thin film [Figs. 2(a) and 2(b)], Bi doping not only results in increased grain size [see Fig. 4(a)] but also increases the cluster density in the film [see Fig. 4(b)]. The XRD pattern (Fig. 5) shows the decrease in the peak intensity as well as the broadening of the peaks due to the Bi doping. This deterioration of film crystallinity can be a result of the high cluster density in the Ca_2.8Bi_0.2Co₄O₉ thin film. The cluster size, on the other hand, is observed to be much smaller than for the Ca₃Co₄O₉ thin film. Fig. 6 illustrates the average dimensions of the largest clusters observed in both films, along with the variation in film thickness as observed from the AFM and SEM analyses. The average diameter and the average height of the largest cluster for Ca_2.8Bi_0.2Co₄O₉ are found to be ∼70% and 22% lower than the ones observed for the Ca₃Co₄O₉ thin film, respectively. No apparent change was detected in the thickness of the 2D monolayers in both kinds of films, which was observed to be ∼14 nm. As both thin films were deposited under the exact same conditions, as well as for the same time, therefore, the difference in the morphology can be considered to be due to the presence of Bi in the system. It appears that Bi doping reduces the amplitude of the undulations and, hence, restricts the ripening of the clusters. Therefore, the merging of the smaller clusters into the larger ones is impeded, and thus, the process does not allow the average cluster size to increase. ^{Reference Gilmer, Huang and Roland20} This argument explains the large number of relatively small clusters observed in Ca_2.8Bi_0.2Co₄O₉ thin films. This kind of incomplete cluster coalescence also leads to pinhole formation ^{Reference Chang, Wu, Ramesh, Xi, Ravi, Venkatesan, Hwang, Muenchausen, Foltyn and Nogar23,Reference Hafez, Elamrawi and Elsayed-Ali24} which can be seen as the dark spots in Fig. 4. These pinholes (∼50 nm to 2 μm in size) are expected to further reduce the thermal conductivity due to increased phonon scattering.

FIG. 4.

(a) High and (b) low magnification SEM micrographs of the surface of Ca_2.8Bi_0.2Co₄O₉ thin film deposited on LAO substrate, illustrating the grain size and the cluster density in the film.

FIG. 5.

XRD patterns showing two most prominent peaks, (002) and (004), of Ca₃Co₄O₉ (CCO) and Ca_2.8Bi_0.2Co₄O₉ (CB02) thin films.

FIG. 6.

A Schematic illustration of the differences in surface morphology between the two studied thin film families.

B. TE properties of CCO and CB02 thin films

Figure 7(a) shows the electrical transport and Seebeck coefficient for the undoped and Bi-doped samples. The trend in the electrical conductivity in our films indicates semiconductor-type transport in the temperature range studied. This behavior is very different from the one observed in the literature, ^{Reference Sugiura, Ohta, Nomura, Hirano, Hosono and Koumoto12} which is metallic for the temperatures between ∼100 and 400 K. Also, our values for σ_350K are 705 and 388/Ω/m for Ca₃Co₄O₉ and Ca_2.8Bi_0.2Co₄O₉ thin films, respectively, which are lower than earlier reported values, ^{Reference Ohta, Sugiura and Koumoto25} the basic reason for which can be suggested as the island formation of the film. Now, as the effective 2D film thickness is very low in our case, therefore, a contribution of electrical conduction from the clusters is also expected and cannot be neglected. These clusters, as discussed in Sec. III. A, are considered to consist of the Ca _x CoO₂ phase, which has higher electrical resistivity than Ca₃Co₄O₉ phase, although both have layered structures. ^{Reference Sun, Ma, Yan, Huang, Wang and Hng26,Reference Kanno, Yotsuhashi and Adachi27} The clusters present in our thin films also represent areas of dislocation and incomplete nucleation and, hence, might act as carrier-scattering centers decreasing carrier mobility. We propose that the conduction in our thin films takes place through a temperature-activated hopping mechanism in the temperature range measured. During the conduction process, the holes are eventually trapped at structural defects, which are the clusters in our case, and the transport continues as a tunneling from one defect to another, causing the unusual transport phenomena. ^{Reference Hughes28}

FIG. 7.

Temperature dependence of (a) electrical conductivity (σ) and (b) Seebeck coefficient (α) for Ca₃Co₄O₉ and Ca_2.8Bi_0.2Co₄O₉ thin films.

Another point to note is that unlike the earlier reported work, Bi doped film is more resistive than the pure Ca₃Co₄O₉ thin film [Fig 7(a)]. It is known that with Bi doping, the effective carrier concentration decreases due to the Bi⁺³ substitution for divalent Ca⁺² in the (Ca₂CoO₃) layers, which inject electrons into the conductive (CoO₂) layer. ^{Reference Sun, Hng, Yan and Ma15} Therefore, the increase in conductivity due to Bi doping has been reported to take place due to the improvement in the charge carrier mobility and not through increased carrier concentration. ^{Reference Xu, Funahashi, Shikano, Matsubara and Zhou3} However, this does not seem to be happening in our case. This can be related to the increase in the cluster density as a result of Bi doping. It was shown that the trapping rate of the holes is directly proportional to the concentration of trapping sites (in our case, clusters) and the radius of trapping sites. ^{Reference Hughes28} As the increase in cluster density with Bi doping is much higher than the reduction in the cluster radius, therefore, we can assume that the increase in resistivity is mainly caused by the larger number of clusters, which consequently increases the number of defect sites and dislocations in the structure of the film.

The room temperature Seebeck coefficient [Fig. 7(b)] for our Ca₃Co₄O₉ and Ca_2.8Bi_0.2Co₄O₉ thin films is found to be 136 and 163 μ V/K, respectively, which is among the largest α_300K reported for Ca₃Co₄O₉ thin films. ^{Reference Zhou, Matsubara, Shin, Izu and Murayama16} Such a high Seebeck coefficient compensates for the lower electrical conductivity and results in a reasonable power factor of α²σ_300K ≈ 0.13 × 10⁻⁴ W/m/K² for CCO and α²σ_300K ≈ 0.1 × 10⁻⁴ W/m/K² for CB02 thin film. These values fit well among those reported previously for this system. ^{Reference Sugiura, Ohta, Nomura, Hirano, Hosono and Koumoto12,Reference Sun, Hng, Yan and Ma15} Furthermore, the Seebeck coefficient of CB02 thin film increases almost linearly with increasing temperature. Now, similar to the Drude picture, the Seebeck coefficient of the Ca₃Co₄O₉ system can be expressed as α ≈ c _e/n, ^{Reference Sun, Hng, Yan and Ma15} where c _e is the electronic specific heat and n is the carrier concentration. As n is known to decrease with Bi doping, the increase in α due to the Bi doping is expected. Also, a large thermopower in our thin films can be explained on the basis of the fact that the Seebeck coefficient for a given carrier density n increases as the scattering parameter increases. ^{Reference Heremans, Thrush and Morelli29} The presence of resistive clusters surrounded by defects and dislocations, which are considered as trap sites for holes, can be considered a source of efficient sites for filtering the charge carriers having lowest energies, resulting in a large Seebeck coefficient. Further power factor enhancements could be achieved by tuning the thickness of the thin film. Inhomogeneous thin film consisting of conducting and nonconducting constituents has been observed to exhibit an abrupt increase in electrical conductivity within a narrow thickness range, which is explained by percolation theory by some authors. ^{Reference Rogacheva, Nashchekina, Vekhov, Dresselhaus and Cronin30} A sharp growth in carrier mobility and electrical conductivity is expected during the transition from island mode to continuous thin film. A detailed work constituting the dependence of TE properties on varying thickness of the Ca₃Co₄O₉ ultrathin film is ongoing.

The obtained values of thermal conductivity (κ) are measured as a whole parameter (Fig. 8), i.e., the thin film plus the substrate. We see that the major κ contribution in the CCO and CB02 thin films comes from the LAO substrate with κ_350K ≈ 11.4 W/m/K. The carrier contribution κ_carr toward total κ was calculated using the relation ^{Reference Yang, Yu, Jinguang, Xianjie and Wenhui2} :

where, L ₀ is the Lorenz number, whose value is 2.44 × 10⁻⁸ W Ω/K², T is the temperature in Kelvin, and ρ is the electrical resistivity in Ω-m. The values of carrier thermal conductivity κ_carr are calculated to be very low, i.e., less than ∼0.005 W/m/K and ∼0.003 W/m/K for CCO and CB02, respectively. Even though the κ_carr for CB02 is ∼30% lower than for CCO, we still observe a lower κ_total of ∼8.5 W/m/K for CB02/LAO compared with the κ_total of ∼11.2 W/m/K for CCO/LAO. These figures indicate that the CB02/LAO sample has a significantly larger number of phonon scattering centers as compared with the CCO/LAO thin film. This again can be easily related to the presence of a higher cluster density as well as the presence of pinholes in the CB02 thin film. These figures are quite figurative and to determine the real thermal conductivity of the individual films, a special measurement technique is required which can separate the thermal conductivity contribution of the substrate from that of the actual thin film. This additional information would provide important information on the intrinsic thermoelectrical properties of Ca-349 thin films having a SK formation.

FIG. 8.

Temperature dependence of thermal conductivity of the CCO and CBO2 thin films on LAO substrate along with the thermal conductivity of the substrate for comparison.