Structures and Role of the Intermediate Phases on the Crystallization of BaTiO3 from an Aqueous Synthesis Route

Carbonate formation is a prevailing challenge in synthesis of BaTiO3, especially through wet chemical synthesis routes. In this work, we report the phase evolution during thermal annealing of an aqueous BaTiO3 precursor solution, with a particular focus on the structures and role of intermediate phases forming prior to BaTiO3 nucleation. In situ infrared spectroscopy, in situ X-ray total scattering, and transmission electron microscopy were used to reveal the decomposition, pyrolysis, and crystallization reactions occurring during thermal processing. Our results show that the intermediate phases consist of nanosized calcite-like BaCO3 and BaTi4O9 phases and that the intimate mixing of these along with their metastability ensures complete decomposition to form BaTiO3 above 600 °C. We demonstrate that the stability of the intermediate phases is dependent on the processing atmosphere, where especially enhanced CO2 levels is detrimental for the formation of phase pure BaTiO3.


INTRODUCTION
The thermodynamic stability of BaCO 3 poses a common synthesis challenge for producing BaTiO 3 , which is a ferroelectric material widely used in capacitors. 1 The stability of BaCO 3 relative to BaTiO 3 increases with a high partial pressure of CO 2 in the atmosphere and analogously with the activity (concentration) of CO 2 in water. 2 Moreover, the solubility of BaCO 3 in water is limited, 3 which combined with the thermodynamic stability makes carbonate secondary phases prevalent both in solid-state reactions and aqueous processing of BaTiO 3 materials. The affinity for carbonate formation arises from the basicity of BaO combined with the abundance of CO 2 in the atmosphere but also dissolved in water. Similar synthesis challenges are observed in other Bacontaining oxides or ceramics consisting of basic oxides. 4 Wet chemical synthesis studies of BaTiO 3 -based powders and thin films generally report BaCO 3 compounds as intermediate and secondary phases. 1,5 Using sol−gel-related methods based on organic solvents, the precursors tend to decompose to form aragonite-type carbonate (BaCO 3 (A)) and TiO 2 , where BaTiO 3 nucleates through the solid-state reactions of these above 600°C. 6−10 However, in Pechinibased synthesis methods and from aqueous processing, a socalled oxycarbonate phase with the global proposed stoichiometry Ba 2 Ti 2 O 5 CO 3 is observed. 10−20 Recently, aqueous chemical solution deposition (CSD) synthesis routes for BaTiO 3 -based thin films have been reported, 21−23 where this intermediate oxycarbonate phase was observed to form prior to the perovskite, dependent on the thermal processing. The pyrolysis reactions and the formation of the oxycarbonate were reported to play an integral part in the texture formation, phase purity, and quality of the films. 22, 23 The oxycarbonate seems to inhibit the formation of BaCO 3 (A) and therefore also shifts the BaTiO 3 formation from the solid-state reaction of TiO 2 and BaCO 3 (A) to proceed through decomposition of the oxycarbonate. The formation of the intermediate oxycarbonate phase has been linked to the presence of a "carbonate-like" linkage in the barium carboxylates 24 or to the formation of a mixed metal citric acid complex in the precursor solution. 12−17 Several local structures for the oxycarbonate phase have been proposed based on the overall global structure of Ba 2 Ti 2 O 5 CO 3 . 15−18 However, Ischenko et al. observed that the intermediate oxycarbonate locally consisted of Ba-and Ti-rich nanosized regions, where there was no clear single crystalline structure in the Ti-rich area but instead a range of Ba y TiO 2+y phases were observed. 19 The Ba-rich areas had a structure close to the high-temperature calcite-type polymorph of BaCO 3 (R3̅ mH, no. 166), 19,20 where substitution of CO 3 2− with O 2 − stabilized the calcite structure giving BaO x (CO 3 ) 1−x . 19,20 The proposed general reaction for the transformation pathway of BaTiO 3 The BaO x (CO 3 ) 1−x phase was observed to form preferably in the presence of titanium. 19,20 Hence, a second stabilizing mechanism was also proposed by Ischenko et al.: topotaxial formation of structural domains of calcite-type carbonate (BaCO 3 (C)) by templating with oxygen-deficient Ti-rich BaTiO 3 -like structures. 20 A comparison of the aragonite and calcite modifications of BaCO 3 can be found in Table 1. Although the models for the intermediate phases have been proposed, it is not known how these phases are affected by processing conditions, such as the heating rate, annealing temperature, and atmosphere, and the intermediate phases influence the formation of BaTiO 3 .
In situ characterization is a rapidly growing field as studying materials under real-time conditions is necessary for further development, especially in battery research 25,26 and hydrogen technology. 27 Moreover, as new synthesis methods are implemented for complex material systems, much insight into the reaction pathway is required, where in situ techniques are becoming more prominent. 28−31 In situ studies by X-ray diffraction on BaTiO 3 are reported both during deposition 32,33 and the annealing 22 of thin films.
Here, we report the thermal decomposition and phase evolution for BaTiO 3 powders from an oxycarbonate forming aqueous synthesis route. In situ infrared (IR) spectroscopy and synchrotron X-ray total scattering were used to study the decomposition of the precursor, formation of intermediate phases, and nucleation of BaTiO 3 from the intermediate phases. Moreover, the effect of partial pressure of CO 2 during the decomposition was investigated. Rietveld and pair distribution function (PDF) refinements revealed the global and local structures of the intermediate phases present in the powders before BaTiO 3 nucleation, supported by electron microscopy. The intermediate phases were found to consist of BaCO 3 (C) and a BaTi 4 O 9 phase; however, the formation of these depends heavily on the processing conditions, especially the heating rate, and partial pressure of CO 2 .

RESULTS
2.1. In Situ Characterization during Thermal Annealing of BaTiO 3 Precursor Powders. Figure 1 shows the in situ IR spectra of the BaTiO 3 precursor powder during annealing in synthetic air with a hold step at 520°C to facilitate formation of the intermediate phases. (The full spectra are included in Figure S2.) The IR spectrum of the amorphous as-prepared precursor contained a split asymmetric stretching (as) mode (1340 and 1400 cm −1 ) of NO 3 − , showing a perturbation of the nitrate ion by cation interactions. 34,35 The symmetric stretching (ss, 1240−1450 cm −1 ) and as-(1590− 1750 cm −1 ) modes for the carboxylic acid groups originating from EDTA, citric acid, and their derivatives are indicated as wide bands, reflecting bonding to different metal ions. 34,35 The frequency range for the as-band of the nitrate and carboxylic acid groups overlap, so the intensity of the wide band also has a contribution from the nitrate. The characteristic C−OH outof-plane (oop) bending mode for carboxylic acid groups was observed at 930 cm −1 . The C−N stretching mode (1020− 1250 cm −1 ) from the EDTA derivatives was identified as a wide band. 34,35 As the temperature is increased above 200°C, the intensity of the bands decreases due to the decomposition  of the nitrate. Between 300 and 400°C, the bands assigned to the different functional groups in the precursor merge to a wide band (1000−1800 cm −1 ). In the temperature range of 470−520°C, this wide band narrows to the asymmetric stretching band of the carbonate ion. Carbonate species remain in the powder until 610°C where there is only BaTiO 3 , as seen from the absence of other bands in the in situ spectra and from the ex situ spectrum taken from the same sample after cooling (marked "RT after"). These observations correspond well with the IR spectra reported for calcined powders from the same precursor solution, where the carbonate absorption band was observed in the temperature range of 450−650°C. 21 The IR spectra from Figure 15−20 Broad bands developed at the start of the hold period became sharper with prolonged annealing, and the frequencies shifted toward that of calcite. Isothermal formation of the intermediate phases and heating with a low heating rate ( Figure S3) were also investigated and showed the same trends for the carbonate band development. The wide carbonate asband was also accompanied by shoulders at 1580 and at 1280 cm −1 in some cases (summarized in Table S1). It is proposed that the shoulders could be assigned to a splitting of the asymmetric stretching band of CO 3 2− bonded to Ti 4+ as the presence of titanium is necessary for the BaCO 3 (C) formation ( Figure S4).
To ease the comparison of the development of the bands assigned to the perovskite and the carbonate, a BaTiO 3 precursor powder was heated without the dome of the reaction chamber and no instrument vacuum. The in situ IR spectra of the precursor in the temperature region of the formation of intermediates and decomposition are shown in Figure 2d−f. The carbonate oop-band ( Figure 2e) and as-band (not shown) were visible above 400°C, similar to the sample with a hold step (Figure 2b). The oop-band was seen as a wide feature at the aragonite frequency (859 cm −1 ) but shifted toward that of calcite (871 cm −1 ) as the temperature increased. No shoulder accompanying the as-band was observed for this heating program. The signature of the perovskite band associated with the Ti−O octahedra (Figure 2f) was observed at 500°C and became more pronounced as the temperature increased. The formation of the perovskite band (Figure 2f) was accompanied with the formation of BaCO 3 (C) (Figure  2e), and as the perovskite band became more pronounced, the carbonate oop-band decreased in intensity, showing how the BaCO 3 (C) decomposes and BaTiO 3 is formed.
The total scattering patterns and converted PDFs of the BaTiO 3 precursor powder heated in situ with a heating rate of 0.17°C/s are shown in Figure 3 (and Figure S5). The diffraction patterns (Figure 3a) demonstrate that the sample is amorphous until diffraction lines corresponding to intermediate phases appear at 630°C, quickly followed by BaTiO 3 crystallization at 690°C. The crystallinity of the BaTiO 3 phase is poor but increases during a hold period of 70 min at 734°C. The PDF (Figure 3b) further demonstrates that the sample is amorphous at low temperatures but transforms into the   Figure 4 for BaTiO 3 precursor powders during different stages of the thermal annealing. The refinement of crystalline BaTiO 3 (Figure 4a,d) shows a reasonable fit using a rhombohedral local structure for BaTiO 3 (R3m, nr. 160 36−38 ), although there are deviations indicating a form of local disorder that could not be refined with periodic structures. The refined PDF for the crystallized powder heated with 0.17°C/s (Figure 4a) also has a contribution from the intermediate phases. The refined PDF for the amorphous powder (prior to any crystallization) is shown in Figure 4b (0.17°C/s), where the nearest atomic pair distances found in the BaCO 3 (A) and BaTi 4 O 9 structures correspond well with the experimental data, but these phases do not give the correct long-range structure. Using 1°C/s, the PDFs of the amorphous powder ( Figure 4e, the full temperature range is shown in Figures S6 and S7) show sharp and narrow peaks. Both amorphous powders ( Figure  4b,e) have peaks that can be described by short-range distances found in BaCO 3 (A) and BaTi 4 O 9 . The intermediate phases only formed using 0.17°C/s (Figure 4c), which had a contribution to the PDF from the amorphous phase compared to the ex situ annealed powder (Figure 4f), impacting the scale factor and crystallite sizes. However, both PDFs were fitted with a BaCO 3 (C) structure (R3m, nr. 160 38,39 ) and a BaTi 4 O 9 phase (Pmmn, nr. 59 37 ), and the ratio between these was locked to 75:25 during refinements, keeping with the global stoichiometry. The CO 3 2− in the calcite structure was allowed full rotational freedom. However, the carbonate groups were observed to only exhibit slight vibrations around their equilibrium position, which eliminate the mirror plane compared to the calcite structure reported by Ischenko et al. 19,20 There is an uncertainty associated with the refined structure and composition of the BaTi 4 O 9 phase as the PDF pattern of the intermediates is largely dominated by the signal from the carbonate ( Figure S8). Furthermore, the obtained fit of the intermediate phases indicates that a periodic structure cannot fully represent the local structure; hence, local disorder is expected. Refined lattice parameters for the PDFs displayed in Figure 4 are listed in Table S2.
The intermediate phases were also investigated by TEM imaging of a BaTiO 3 precursor powder that was preannealed at 530°C for 1 h. The TEM bright-field (BF) images and selected area diffraction patterns (SADPs) of selected particles are shown in Figure 5, demonstrating that the crystallinity of the intermediate phases was limited. However, both nanocrystalline and poorly crystalline BaTiO 3 particles (Figure 5a,b, respectively) were observed in the powder even if the XRD patterns 21 do not show BaTiO 3 at this temperature (530°C). Separate BaCO 3 (C) (Figure 5c) and BaTi 4 O 9 (Figure 5d,e) crystallites were also identified based on the SADP, although the crystallinity of these particles was also low. The diffraction from the presence of BaTi 4 O 9 in Figure 5d also includes a contribution from BaCO 3 (A) and possibly also BaCO 3 (C). The particle in Figure 5f showed diffraction both from the    (110) planes, is slightly larger in the mixed particle (Figure 5f) than the diffraction spots seen in the BaTi 4 O 9 particle (Figure 5e), illustrating the change in unit cell with increasing crystallinity. A weak diffraction ring with a d-spacing of 4.065 Å is observed in the particle in Figure  5f, which could correspond to the (101) diffraction line of BaCO 3 (C), but it could also be the emerging (100) diffraction of BaTiO 3 . The identified d-spacings of the particles in Figure  5 are summarized in Table S3. 2.3. Variable CO 2 Partial Pressure during Annealing of BaTiO 3 Precursors. Since CO 2 is released during the decomposition and formation of BaTiO 3 according to the proposed reaction in eq 1, the partial pressure of CO 2 is an important processing parameter. The phases present in each recorded diffractogram during in situ HT-XRD as a function of temperature and CO 2 partial pressure are summarized in Figure 6 (all measured XRD patterns are shown in Figure S9).
The results for the 0% CO 2 (synthetic air) correspond well with previous ex situ powder XRD results 21 and in situ XRD of BaTiO 3 films from the same precursor solution. 22 The formation of the intermediate phases was unaffected by increasing CO 2 partial pressure as they formed above 505°C in all the gas mixtures, except in pure CO 2 . However, for the powders heated in a CO 2 -rich atmosphere, BaCO 3 (A) formed as a thermodynamically stable phase alongside the intermediates. The temperatures used during the in situ powder HT-XRD measurements were not sufficient to remove BaCO 3 (A), so once formed the aragonite remained, even after formation of BaTiO 3 . However, due to stoichiometric concerns, there should also be Ti-rich phases present, but since these are amorphous, they do not appear in the XRD patterns. The diffractogram of the sample heated to 546°C in synthetic air (Figure 6b) demonstrates the presence of only the intermediate phases (BaCO 3 (C) and BaTi 4 O 9 ), while at 585°C in 50% CO 2 (Figure 6c), the powder contains about 15 wt % BaCO 3 (A) in addition to the intermediate phases.
The refined cell parameters and crystallite sizes from the Rietveld refinements are listed in Table S4, where the poor crystallinity in this temperature region resulted in small refined crystallite sizes for both BaCO 3 (C) and BaTi 4 O 9 .
The stability of intermediate phases over BaTiO 3 nucleation exhibited a strong CO 2 dependence even though the formation temperature for the intermediates was unaffected by the CO 2 partial pressure. In synthetic air, the intermediates decomposed below 600°C, while in a CO 2 -rich atmosphere, they were still present at 675°C (Figure 6a), which likely is caused by a stabilization of the BaCO 3 (C) phase by CO 2 . The increased carbonate stability also affected the BaTiO 3 nucleation temperature, which was below 565°C in synthetic air, but increased to above 650°C in the CO 2 -rich atmosphere. There were no significant differences in the phase evolution of the samples heated in 25−75% CO 2 .

Transformation Pathway: Decomposition and
Pyrolysis. The basicity of BaO and high stability of BaCO 3 make carbonate formation almost inevitable during the thermal processing of the BaTiO 3 precursor powder from an aqueous synthesis route. However, the phase evolution of the powders studied in this work followed the "oxycarbonate forming route" commonly reported for Pechini and sol−gel-based synthesis, instead of the solid-state reaction route. Therefore, it is probable that a mixed metal citric acid complex similar to those reported in refs 12, 14 formed in the precursor solution. A proposed transformation pathway is illustrated in Figure 7, and the decomposition and pyrolysis reactions can be divided into the following steps:  . At the start of this period, the nitrate decomposes, ammonia evaporates, and the decomposition/combustion of the organics initiates (step 1 in Figure 7). However, the wide RCOO − stretching bands are still present, as seen from the in situ IR spectra of the powder precursors (Figure 1), originating from a variety of organic groups in the sample. The symmetric and asymmetric stretching bands of the RCOO − groups become narrower and shift toward the asymmetric carbonate band as temperature is increased, demonstrating the preference for carbonate formation in this system. The organic removal (pyrolysis) can be expressed through the following proposed reaction (adapted from Ischenko et al. 19 which occurs at the initiation of this temperature region, leaving only BaCO 3 -like and BaTi 4 O 9 phases in the system, with no long-range order. This is in accordance with the thermal analysis of the precursor powder, where a decrease in the mass fraction over a narrow temperature range is reported at 550°C 21 corresponding to removal of the remaining organic compounds.

Carbonate Formation (480−570°C).
In this temperature range, the IR spectra only showed bands corresponding to carbonate (Figures 1 and 2), which were initially broad features of an amorphous carbonate species with a frequency corresponding to BaCO 3 (A) but quickly shifted toward the calcite frequency as the temperature was increased, according to the following proposed reaction The carbonate frequency shift is accompanied by formation of a Ti-carbonate-like phase as indicated in the IR spectra (Figures 1 and 2 The nucleation of BaTiO 3 occurs homogeneously throughout the powder as TEM showed that the carbonate and Ti-rich phases are intimately mixed, and the metastable nature of the intermediate phases ensures complete decomposition. 3.2. Structures of the Intermediate Phases. The intermediate phases were found to consist locally of poorly crystalline and nanosized BaCO 3 (C) and BaTi 4 O 9 with limited long-range order. Since BaCO 3 (A) is the thermodynamically stable structure in the temperature range of carbonate formation (eq 3), 38 the formation of BaCO 3 (C) is likely enabled due to one or several stabilizing mechanisms. The presence of titanium was observed to be a requirement for the calcite carbonate formation ( Figure S4) and bands in the IR-spectra assigned to a Ti-carbonate-like group (Figures 1 and  2) formed alongside BaCO 3 (C) bands. Both observations demonstrate an interaction between the carbonate and Ti-rich intermediate phases, possibly at the interface between small domains of each phase as illustrated in Figure 7. The proposed model is further supported by the Rietveld and PDF refinements, which showed that the intermediate phases could be described by nanosized poorly crystalline BaCO 3 (C) and BaTi 4 O 9 with limited long-range order. TEM further demonstrated limited crystallinity and small crystallites, but diffraction from both BaCO 3 (C) and BaTi 4 O 9 was observed from individual (Figure 5c,e, respectively) and mixed particles (Figure 5f).
A similar model for the intermediate phases with BaCO 3 (C) stabilized by topotaxial templating on oxygen-deficient Ti−O interface layers was suggested by Ischenko et al. 20 Although the composition of the Ti-rich phase determined in this work, BaTi 4 O 9 , fits with the reported range given by Ischenko et al.,19,20 the structure of BaTi 4 O 9 does not seem to comply with the suggested BaTiO 3 -like building blocks for the topotaxial templating model. 20 Moreover, the refined BaCO 3 (C) in this work has a compressed unit cell compared to the structure reported in the literature, 19,20,38 with 0.4 % expansion of the aparameter but 6.9 % compression of the c-parameter. This reduction in unit cell volume is due to only slight vibrations of the carbonate groups instead of full rotations. O 2− -substitution, as suggested by Ischenko et al. 19,20 to act as a second stabilizing mechanism for BaCO 3 (C), was not investigated, but it seems likely given the deviating structure that the degree of substitution is different in this work compared to previous studies. However, since the intermediate phases are intimately mixed nanosized domains, with a certain degree of interaction between them (Figure 7), the structure could still follow the topotaxial templating model, even if the precursor chemistry and crystal structure of the intermediate phases deviate slightly. Stabilization of the calcite modification of BaCO 3 has been reported in mixed alkaline earth carbonates 41 or by quenching. 42 Intermediate carbonates with the calcite structure are also reported in YBa 2 Cu 3 O 7−x, 43 Ba 1−x SrTiO 3 , 11 and Ba 1−x Ca x Zr y Ti 1−y O 3 23 systems. Controlling the carbonate formation was reported to be crucial for phase purity in Ba 0.85 Ca .15 Zr 0.1 Ti 0.9 O 3 thin films. 23 Barium carbonate and Tirich intermediate phases are therefore probable intermediates during wet chemical processing of BaTiO 3 -based materials, although the exact nature of these would depend on the precursor chemistry.
3.3. Influence of the Heating Rate and CO 2 on the Transformation Pathway for BaTiO 3 . Slower heating generally decreased the temperature regions for the reaction occurring during the thermal decomposition of the BaTiO 3 precursor (Section 4.1), for certain heating rates. For fast heating (>1°C/s), the transformation pathway is altered, and the intermediate phases are inhibited due to kinetics, giving direct BaTiO 3 nucleation from the amorphous network. This is in line with previous results on BaTiO 3 thin films from a similar precursor solution during fast heating (>1°C/s). 22 Moreover, the in situ IR spectra of the precursor powders ( Figure S3) show that carbonate formation was less pronounced for the slow heating (0.05°C/s), which could relate to the metastable nature of the intermediate phases and the decomposition kinetics.
Increased partial pressure of CO 2 in the atmosphere stabilizes the intermediate phases over BaTiO 3 nucleation ACS Omega http://pubs.acs.org/journal/acsodf Article ( Figure 6) by drastically limiting the decomposition reaction in eq 4. For high CO 2 partial pressures (>25 vol % CO 2 ), the perovskite nucleation temperature increased more than 100°C compared to that in synthetic air (0 vol % CO 2 ). Moreover, the temperature region for the coexistence of intermediates and perovskite increased for high CO 2 partial pressures, which means that even if the decomposition reaction in eq 4 can take place the high partial pressure of CO 2 serves as a kinetic limitation. A second effect of a high CO 2 partial pressure (>25 vol % CO 2 ) is the formation of the thermodynamically stable BaCO 3 (A), which once formed require temperatures above 700°C to decompose. No secondary Ti-rich phases were observed by XRD alongside the perovskite once aragonite-type BaCO 3 formed; hence, titanium remains as unreacted amorphous BaTi 4 O 9 . BaCO 3 (A) formed at the same temperature as the intermediate phases; therefore, CO 2 stabilizes a second reaction during the organic removal step described by this modified version of eq 3 leading to the formation of BaCO 3 (A) alongside the BaCO 3 (C) phase. The formation of a broad aragonite band in the IR spectra (Figure 2a−c) before the shift toward calcite formation when the precursor powders were annealed in air (eq 5a) might be caused by locally enhanced CO 2 partial pressure from decomposing organics. However, as the organics are further decomposed/combusted during heating, the enhanced CO 2 partial pressure decrease before the aragonite-type BaCO 3 can fully crystallize resulting in formation of BaCO 3 (C). It is also likely that the apparent stability of the intermediate phases observed in the total scattering data ( Figure 3) is due to locally enhanced CO 2 partial pressure inside the capillaries during the experiments, which causes these phases to remain even during the prolonged annealing at 740°C. Enhanced CO 2 levels would shift the reactions along the x-axis in Figure 6a, which also fits with the increased reaction temperatures observed.
Although it is important to note that direct comparison between the different experiments carried out in this work cannot be done due to different reaction volumes used for the different techniques. The volume used for annealing will affect the kinetics of the reactions, which is why thin films 22 were observed to crystallize at a lower temperature than powders. 21 However, the trends for the decomposition of the precursor and crystallization can still be discussed, independent of reaction volume. Carbonate formation is a well-known prevailing synthesis challenge in BaTiO 3 -based materials and for Ba oxides, especially during wet chemical processing if the dissolved and atmospheric CO 2 levels are not controlled. The structure of the carbonate forming depends on the precursor chemistry, where the BaCO 3 (C) type is highly sensitive toward the processing conditions. However, the BaCO 3 (C) type could be preferable over the formation of BaCO 3 (A) due to the metastable nature of BaCO 3 (C), which under the right processing conditions results in phase-pure BaTiO 3 in the temperature range of 550−600°C, as reported in this work.

CONCLUSIONS
The decomposition, pyrolysis, and crystallization reactions during synthesis of BaTiO 3 by an aqueous-based synthesis route were characterized by in situ IR and synchrotron X-ray total scattering. The in situ analysis revealed the transformation pathway for BaTiO 3 crystallization and the structure and composition of the intermediate metastable calcite-type BaCO 3 and BaTi 4 O 9 phases that formed prior to BaTiO 3 nucleation. The crystallinity of the nanosized intermediate phases is poor as there is limited long-range order. BaTiO 3 nucleates both directly from an amorphous network but also through the diffusion-controlled reaction of the intermediate phases. Intimate mixing of the intermediate phases and their metastable nature ensure full decomposition in controlled atmospheres. However, the stability of the intermediates over BaTiO 3 formation is governed by the CO 2 partial pressure, where enhanced CO 2 levels stabilize calcite-type BaCO 3 but also leads to the formation of the thermodynamically stable aragonite-type BaCO 3 . Therefore, control of the processing atmosphere is crucial when fabricating phase-pure BaTiO 3 through this aqueous synthesis route. Carbonate intermediate phases by similar formation mechanisms can also be expected for BaTiO 3 -based materials and basic oxides in general.

EXPERIMENTAL SECTION
5.1. Synthesis. The preparation of the aqueous precursor solution has been reported previously. 21 BaTiO 3 precursor powder was prepared by drying the precursor solution at 150°C for 24 h in air, resulting in a sponge-like brown material, which was crushed in an agar mortar to yield the precursor power. The BaTiO 3 precursor samples for in situ IR measurements were prepared by dispersing the precursor powder in deionized water and depositing droplets directly onto platinized silicon substrates (Pt/Si, SINTEF, Oslo, Norway). The droplets were flattened by draining most of the liquid of the substrate edge, leaving a wet precursor powder layer, which were dried at ambient temperature for 30−60 min. Precursor powders for ex situ Xray total scattering and electron microscopy investigation of the intermediate phases were annealed at 530°C for 1 h with a heating/cooling rate of 0.056°C/s in air. The IR spectra of the prepared powders are included in Figure S1. 5 scans were averaged. The IR spectra of the samples after heating were measured at ambient temperature in vacuum and without the dome of the reaction chamber, labeled "RT after".
In situ high-temperature X-ray diffraction (HT-XRD) on the precursor powder under a controlled CO 2 atmosphere was performed on a D8 Advance Diffractometer (Bruker, Billerica, MA, USA) with Cu Kα radiation (λ = 1.54 Å) equipped with a Vantec-1 SuperSpeed detector. The samples were prepared in a radiant heater sample holder of alumina. The partial pressure of CO 2 in synthetic air was varied in the range of 0−100%. The sample chamber was closed and purged with the desired gas mixture for 1 h before measurements were started. The diffractograms were recorded with a step size of 0.033°and 0.5 s scan time per step during a hold step at selected temperatures, and the heating rate in between hold steps was 0.2°C/s. Rietveld refinements of the powder XRD patterns were done with the TOPAS software (v5, Bruker, Billerica, MA, USA).
X-ray total scattering data for pair distribution function (PDF) analysis of the BaTiO 3 precursor powders was collected on BL08W 44 at SPring-8 (Japan) using a flat 2D panel detector and a wavelength of 0.10765 Å. The time resolution of the recorded data was 5 s. For the ex situ measurement, a powder preannealed at 530°C in air was filled in a Kapton tube (OD 1.05 mm, Goodfellow, England), and for the in situ measurements, the precursor powders were loaded in quartz capillaries (OD 1.5 mm, CharlesSupper Company, Westborough, USA) with glass wool on each side. The capillaries had continuous air flow of 0.12 L/min and were measured in transmission. A hot air blower was used to heat the samples continuously (0.17−1°C/s) and then held at the maximum annealing temperature for 30−70 min. The 2D images were masked and integrated with the pyFAI python package, 45 where 5 patterns were averaged. Empty capillaries were used for background subtractions, which was done with the pdfgetx3 (v2.0.0) software using periodic structures. 46 The Q-range for the samples was 16−22 Å −1 , and the PDFs were refined with PDFGui (v1.1.2). 47 For the transmission electron microscopy (TEM) investigation, a powder preannealed at 530°C in air was dispersed in isopropanol in an ultrasound bath before depositing the powder particles on a holey carbon Cu-grid. TEM was performed on a JEOL JEM 2100 equipped with a LaB 6 electron gun. Selected area diffraction patterns (SADPs) were obtained using a circular aperture covering an area in real space with a diameter of approximately 750 nm.
Additional infrared spectra, in situ total scattering data, all XRD patterns, and refined values from PDF and Rietveld refinements are listed (PDF)