Formation mechanism and thermoelectric properties of CaMnO3 thin films synthesized by annealing of Ca0.5Mn0.5O films

A two-step synthesis approach was utilized to grow CaMnO3 on M-, R- and C-plane sapphire substrates. Radio-frequency reactive magnetron sputtering was used to grow rock-salt-structured (Ca, Mn)O followed by a 3-h annealing step at 800 °C in oxygen flow to form the distorted perovskite phase CaMnO3. The effect of temperature in the post-annealing step was investigated using x-ray diffraction. The phase transformation to CaMnO3 started at 450 °C and was completed at 550 °C. Films grown on R- and C-plane sapphire showed similar structure with a mixed orientation, whereas the film grown on M-plane sapphire was epitaxially grown with an out-of-plane orientation in the [202] direction. The thermoelectric characterization showed that the film grown on M-plane sapphire has about 3.5 times lower resistivity compared to the other films with a resistivity of 0.077 Ωcm at 500 °C. The difference in resistivity is a result from difference in crystal structure, single orientation for M-plane sapphire compared to mixed for R- and C-plane sapphire. The highest absolute Seebeck coefficient value is − 350 µV K−1 for all films and is decreasing with temperature.


Introduction
Materials for thermoelectric generation are widely researched due to their capability to convert waste heat into electricity [1]. They have no moving parts and do not require any chemical reactions to operate which make them ideal for waste heat recovery. However, the most utilized thermoelectric modules are based on toxic and rare materials such as tellurium [2]. In addition, there are issues with chemical stability and low oxidation resistance at elevated temperatures. Oxides are therefore promising materials due to their high chemical stability and good oxidation resistance, and an additional advantage is that they are largely based on abundant and low-tomoderate cost materials. Ca 3 Co 4 O 9 is a good candidate as a p-type thermoelectric material with a reported dimensionless figure of merit ZT = rS 2 T J -1 around unity near 1000 K [3,4], where r, S, T and J are the electrical conductivity, Seebeck coefficient, absolute temperature and thermal conductivity, respectively. To realize full oxide modules, the n-type oxides need improvement to achieve similar ZT as p-type oxides. The most investigated n-type thermoelectric materials are CaMnO 3 , ZnO, SrTiO 3 and In 2 O 3 , although these materials have so far shown only modest ZT values [5][6][7]. The perovskite CaMnO 3 suffers from low electric conductivity and relatively high thermal conductivity, yielding low ZT. In the CaMnO 3 system, the best ZT value reported is 0.32 for Nb-doped CaMnO 3 [8]. However, CaMnO 3 can be a promising n-type candidate if the power factor (rS 2 ) is enhanced by optimal doping, since for high output power from a thermoelectric device the power factor is more important than the efficiency [9,10], especially, in thin films for low-power applications.
For thin film growth of oxide materials, both chemical, e.g., chemical vapor deposition (CVD) [11,12], chemical solution deposition (CSD) [13][14][15][16], and physical methods, e.g., physical vapor deposition (PVD) [17][18][19], have been reported. CVD exhibits a high growth rate with a high film uniformity, but high processing temperatures combined with byproducts that are generally toxic, explosive and corrosive in nature limit the technique. CSD is a fast and cheap method for growing films. Magnetron sputtering is a versatile PVD method and can be suitable for industrial upscaling. However, growing oxide thin films is a challenge which can be handled in several different ways. A few options are oxide targets sputtered in RF mode, metallic target using reactive gas operated in poisoned mode using RF or pulsed DC or metallic targets using reactive gas operated in metallic mode. Using oxide targets or running metal targets in poisoned mode results in slow growth rates. Running the system in metallic mode usually results in under stoichiometry of oxygen in the grown film. The optimal process parameters, where high deposition rates combined with desired stoichiometry are achieved, are often in a narrow window between metallic and poison mode. Careful process control is needed to remain in this narrow window [20].
In this study, we report a growth method which circumvents the above-mentioned problems for CaMnO 3 thin films by a modified sputtering technique using a two-step sputtering/annealing technique [21][22][23] to maintain high deposition rates keeping the stoichiometry of the films. First, singlephase cubic (Ca, Mn)O thin films are deposited by cosputtering from elemental metallic targets of Ca and Mn in an oxygen-containing environment. Followed by annealing in air with additional oxygen to form the final phase of perovskite CaMnO 3 . This has the benefit of a high deposition rate, since the sputtering yield from metallic targets is always higher than that from the respective oxide targets. An additional advantage is that it allows perfect control of the composition of the films by controlling the target power. The phase transformation mechanism, from initial cubic (Ca,Mn)O phase to the final perovskite CaMnO 3 phase, has been studied. Evaluation of the thermoelectric properties of the films is performed in terms of their power factor.

Experimental details
(Ca, Mn)O thin films were deposited onto Al 2 O 3 (0001, C-plane), (1 102, R-plane) and (1 100, M-plane) substrates from elemental Ca (99.5% purity) and elemental Mn (99.9% purity) targets. Prior to deposition, the substrates were cleaned in an ultrasonic bath in acetone followed by isopropanol for 10 min each. The vacuum chamber at 4 9 10 -8 torr base pressure is described elsewhere [24]. Oxygen and argon were used as sputtering gases at a flow ratio of O 2 (O 2 ? Ar) -1 = 1.5% resulting in a total sputtering J Mater Sci (2019) 54:8482-8491 gas pressure of 0.27 Pa. The substrate was kept for 10 min at 400°C prior a 30-min deposition resulting in film thicknesses around 250 nm. The two magnetrons used unbalanced magnetic field configurations that were coupling to each other, and both were operated in RF mode at a power of 60 W. The substrate table was kept electrically isolated from ground. The Ca/Mn ratio was determined using energy-dispersive spectroscopy (EDS) in a scanning electron microscope (SEM, LEO Gemini 1550, Zeiss) at different regions of each sample.
The post-deposition ex situ annealing study was performed in a tube furnace in an air with O 2 gas flowing atmosphere. The study was performed in steps where prior to sample insertion, the temperature of the furnace was stabilized at set temperature. The sample was removed from the furnace when it reached the intended annealing temperature. Prior to x-ray diffraction (XRD) measurements, the sample cooled down to room temperature. This process was repeated for increasingly higher temperature from 400°C to 1000°C. The post-deposition annealing for the samples investigated more in depth was based on the ex situ annealing study, and 800°C for 3 h was used at the same gas condition as in the ex situ annealing study.
XRD h-2h measurements were performed using a Philips PW1710 powder diffractometer with a copper anode source (Cu K a , k = 1.54 Å ), operated at 40 kV and 40 mA. A Ni filter was used directly after the X-ray source to remove Cu K b . A 0.5°divergence slit was positioned after the Ni filter. In the diffracted beam path, a 2-mm antiscatter slit, a 0.5°receiving slit, a curved Ge-crystal monochromator and a proportional Xe-gas-filled detector to detect the intensity were used.
The temperature-dependent in-plane electrical resistivity and Seebeck coefficient were simultaneously measured using an ULVAC-RIKO ZEM3 system in a low-pressure helium atmosphere. The instrumental error is less than 7%.
Samples for transmission electron microscopy (TEM) were prepared by ion beam milling. For crosssectional TEM, two pieces of the sample were glued together face-to-face and clamped with a Ti grid and then polished down to 50 lm thickness. Finally, the polished sample was ion-milled in a Gatan precision ion polishing system at an Ar ? energy of 5 kV and a gun angle of 5°, with a final polishing step at 2 kV Ar ? energy. The TEM analysis was performed using a FEI Tecnai G2 TF20 UT instrument with a field emission gun operated at 200 kV and with a point resolution of 1.9 Å .
Thermal conductivity of the films was obtained using modulated thermoreflectance microscopy (MTRM). In brief, a pump beam at 532 nm delivered by a Cobolt MLD laser, intensity modulated by an acousto-optical modulator at a frequency f, is focused on the surface of the sample with an objective lens (N.A. = 0.5). In order to prevent effects from possible changes in the optical properties versus the Mn concentration, the layers were covered by a 250-nm gold film, this top layer ensuring a heat source located at the surface. Then, thermal waves are excited in the sample and monitored by the reflectivity surface change recorded around the pump location by another focused laser beam. We use a 488-nm Oxxius laser to maximize the probe sensitivity to the thermal field in the gold cap layer. A photodiode and a lockin amplifier record the AC reflectivity component, in a frequency range between 1 kHz and 1 MHz. Finally, the amplitude and phase experimental profiles were fitted according to a standard Fourier diffusion law to extract the thermal conductivity of the CaMnO 3 films [25][26][27][28]. More details on the procedure of fit and extraction of the thermal conductivity of the film with a low thermal conductivity layer in between the gold and sapphire substrate can be found elsewhere [29].

Results and discussion
As-deposited films  Fig. 2c, d. Figure 3b shows the TEM images of the as-deposited Ca 0.5 Mn 0.5 O films grown on C-plane sapphire. The films grown on C-plane sapphire are polycrystalline, which is also the case for the films grown on R-plane sapphire (not shown here). Figure 3c-e shows the SAED patterns of the C-plane sapphire where (d), corresponding to the film, shows elongated spots (almost forming rings) confirming the polycrystalline nature of the film.
In summary, the XRD and TEM results show that the as-deposited films grown on M-plane sapphire are phase-pure containing epitaxial domains with a well-defined microstructure. Films grown on C-and R-plane sapphire show similar structural properties where both are polycrystalline with less well-defined microstructure than those grown on M-plane sapphire.

Phase transformation
To study the phase transformation mechanism from initial Ca 0.5 Mn 0.5 O film to the final phase of perovskite CaMnO 3 , ex situ x-ray annealing was performed in air with O 2 flowing at increasing temperature from 450°C to 1000°C. Between each annealing step, the sample was cooled down to room temperature and measured using XRD. The general trend is similar for both samples, and Fig. 4a, b shows the h-2h x-ray diffractograms from the annealing study for the films grown on M-and R-plane sapphire, respectively. Above 500°C, the phase transformation from Fm-3m to Pnma is almost complete, as the peak from rock-salt structure disappears and 101, 202, 121 and 242 peaks of CaMnO 3 gradually appear at 2h angles 23.83°, 34.07°, 48.90°a nd 71.70°, respectively. With further increase in temperature, the peaks of CaMnO 3 become more intense with gradual decrease in full width at half maximum (FWHM), which indicates the improvement of crystal quality with temperature. Pole figure analyses of the films annealed at 800°C confirm the texture quality of the film (see figure S1 and S2 in supplementary information). Above 800°C, the diffraction peaks shift to higher 2h values, which implies macroscopic stress in the sample. Appearance of low intense peaks is also clear from Fig. 4a, b, which is attributed to the partial presence of secondary orientation in both the films on M-and R-plane sapphire substrates. The overall trend on films grown on R-plane sapphire is similar for films grown on C-plane sapphire (not shown here).
From the above study, it is clear that the annealing temperatures in the range 500-800°C facilitate the formation of single-phase and singly oriented CaMnO 3 films, and annealing temperature of 800°C is favorable to produce films with a well-defined microstructure.    Fig. 6a-d, respectively. The film grown on M-plane sapphire no longer shows the well-defined diffraction spots of the as-deposited film, but it is still highly textured with an out-of-plane (101) orientation. The film on M-plane sapphire has a single-phased structure with a relatively small amount of grains with secondary orientations forming throughout the films as seen in Fig. 6a. In contrast, the CaMnO 3 film grown on R-plane sapphire is polycrystalline as seen in the SAED pattern, which is also typical for the films grown on C-plane sapphire, as expected given their starting microstructure before annealing.

Thermoelectric properties
The thermal conductivity measured at room temperature is tabulated in Table 1 and is in the same range as previously reported bulk values, 3.5 W (mK) -1 , 4.1 W (mK) -1 and 2.8 W (mK) -1 [32][33][34]. Figure 7 shows the resistivity (q) in (a), ln(rT) (b), Seebeck coefficient (S) in (c) and power factor (rS 2 ) in Figure 5 h-2h x-ray diffractogram of annealed, at 800°C for three hours, CaMnO 3 films grown on M-, R-and C-plane sapphire, corresponding to a-c, respectively. Peaks marked with a star are from the substrate (ICDD file 00-046-1212).  (d) as a function of temperature for films grown on M-, R-and C-plane sapphire. The electrical resistivity ranges from 1.75 Xcm at room temperature to 0.077 Xcm * 500°C for the film grown on M-plane sapphire, 23.40 Xcm at room temperature to 0.25 Xcm at * 500°C for R-plane sapphire and 11.33 Xcm to 0.21 Xcm for C-plane sapphire. The Seebeck coefficient is the highest for the film grown on R-plane sapphire substrate ranging from -350 lV K -1 at room temperature to -250 lV K -1 at 500°C. The Seebeck values for the films grown on C-and M-plane sapphire are similar and range from -350 lV K -1 at room temperature to -200 lV K -1 at 500°C. The decrease in the absolute value of the Seebeck coefficient, |S|, with temperature is attributed to the excitation of minority charge carriers [35]. In Fig. 7b, ln(rT) vs T -1 is plotted to determine the activation energy E a for the electrical conductivity following the Arrhenius formula, as shown in Eq. (1) and (2) [36].
Here, r, A, T and k b are the electrical conductivity, a constant, the temperature and Boltzmann's constant, respectively. This equation describes a small polaron hopping model which is commonly used to describe electrical conductivity for CaMnO 3 [37]. The Arrhenius plot in Fig. 7b is linear in the whole temperature range for films grown on M-plane sapphire and nonlinear above 200°C for films grown on R-and C-plane sapphire. The activation energy for film grown on M-plane sapphire is evaluated to 0.19 ± 0.01 eV from the slope in Fig. 7b. The nonlinear behavior in films grown on R-and C-plane sapphire has been observed in other studies [38]. The reason for the nonlinear behavior is unknown; however, Zhou et al. report an anomaly in the octahedral tilt angles in this temperature range which could explain the observed change in activation energy [39]. The power factor (rS 2 ) is the highest for the film grown on M-plane sapphire reaching 45 lW m -1-K -2 (500°C), which is comparable to Xu et al.  [41]). The highest power factor for our films is the film on M-plane sapphire which is due to its low electrical resistivity.
Some reported values in the literature for resistivity, Seebeck coefficient and power factor at room temperature and activation energy are tabulated in Table 2. All references in Table 2 use solid-state reaction (SSR) for synthesizing the CaMnO 3 bulk samples. Huang et al. [33] report a resistivity value that is in the range of 10 Xcm. A trend is clearly visible. Higher resistivity results in both higher absolute Seebeck coefficient and activation energy. However, the reported values are rather different between studies and differ almost 100 times in the extreme case for resistivity. A possible explanation for these differences can be found by taking oxygen deficiency into consideration. Goldyreva et al. [42,43] report that oxygen vacancies act as electron donors by creating Mn 3? from Mn 4? . This change reduces, similarly too alloying with rare earth metals, resistivity, Seebeck coefficient and activation energy. Based on this, our films grown on M-plane sapphire are most likely deficient in oxygen as they have lower values in both resistivity and Seebeck compared to three of the listed references in Table 2. As the crystallinity is poorer for films grown on R-and C-plane sapphire, one should be careful to draw the same conclusion on these samples because the thermoelectric properties are likely heavily influenced by defects. From Table 2 and our discussion above, it is clear that oxygen vacancies are favorable for the performance of CaMnO 3 as a thermoelectric material.

Conclusion
We have synthesized CaMnO 3 thin films on sapphire using a two-step synthesis method. First, (Ca,Mn)O is deposited on sapphire substrates using RF reactive magnetron co-sputtering from elemental targets. Secondly, the as-deposited films are annealed to transform the films to orthorhombic (Pnma) structured CaMnO 3 . An annealing study showed that CaMnO 3 is formed from 500°C and 800°C is favorable for the formation of a well-defined microstructure. The structural and thermoelectric properties on different sapphire substrates (M-, R-and C-plane) were evaluated. The effect of post-annealing temperature is also investigated.
Films grown on M-plane sapphire are epitaxially related to the substrate without any secondary orientation in the as-deposited state. After the annealing process, the film remains highly textured, in contrast to polycrystalline films grown on C-or R-plane sapphire. A film on M-plane sapphire exhibits a power factor of 45 lW m -1 K -2 at 500°C.