Synthesis and structural characterization of Ca12Ga14O33

Ca12Ga14O33 was successfully synthesized using a wet chemistry technique to promote the homogenous mixing of the Ca and Ga cations. Rietveld refinements on X-ray and neutron powder diffraction data confirm that the compound is isostructural to Ca12Al14O33, however, with a significantly larger lattice parameter allowing for the cages that result from the framework arrangement to expand. In naturally occurring Ca12Al14O33, the mineral mayenite, these cages are occupied by O2− anions, however, experimental studies exchanging the O2− anions with other anions has led to a host of applications, depending on the caged anion. The functional nature of the structure, where framework distortions coupled with cage occupants, are correlated to electronic band structure and modifications to the framework could lead to interesting physical properties. The phase evolution was tracked using thermogravimetric analysis and high temperature X-ray diffraction and showed a lower formation temperature for the Ca12Ga14O33 analogue compared to Ca12Al14O33 synthesized using the same wet chemistry technique. Analyzing both X-ray and neutron powder diffraction using the Rietveld method with two different starting models results in one structural model, with one Ca position and the caged O on a 24d special position, being preferred.

The atomic structure of Ca 12 Al 14 O 33 , also known as C12A7, consists of a positively charged cage-like framework where occluded anions occupy a fraction of the cages to balance the positively charged framework. The diffusion of anions from cage to cage allows the material high ion mobility and the potential for ion storage 1 . This functionality leads to applications for Ca 12 Al 14 O 33 as a CO 2 absorber 2 , a catalyst 3 , and as an inorganic electride 4 . The highly functional nature of the material raises questions about isostructural compounds and their potential properties. Framework distortion and cage occupancy are correlated to electronic band structure changes responsible for electrical insulator to conductor transition. Further modifications to the framework through alteration of cations and charge balancing anions in the structure should lead to changes in physical properties; the alteration of material properties with atomic substitutions has been experimentally and theoretically explored and are summarized in 5 . The ionic radius of Ga 3+ in fourfold coordination with O is 47 pm compared to the 39 pm radius of Al 3+6 suggesting that this substitution could result in a cage with a larger diameter. Partial site replacement of tri-valent Al cations has been performed with similar tri-valent Ga cations. However, results led to the conclusion that the single phase (Ca 12 Al 14-x Ga x O 32 )O formed when x = 1 and when x = 2 both Ca 5 Ga 6 O 14 and Ca 3 Al 2 O 6 were observed in addition to Ca 12 Al 12 Ga 2 O 33 7 . When Ca 12 Al 13 GaO 33 was subjected to reducing conditions at 1350 °C for 6 h, buried in graphite powder, the compound decomposed to Ca 3 Al 2 O 6 and Ca 12 Al 14−x Ga x O 33 7 . The cause of decomposition is not known, but Ca 12 Al 14 O 33 structure in general demonstrates complicated thermodynamic instability/stability trends under dry reducing conditions above 1100 °C 8,9 . The influence of Ga on the cage structure and localized electron behavior could alter the physical properties including the mobility of occluded species, either atomic, molecular, or electrons, and the electrical conductivity.
The existence of Ca 12 Ga 14 O 33 has been theorized due to the similarity of the CaO-Al 2 O 3 and CaO-Ga 2 O 3 binary phase diagrams and existence of isostructural compounds from the two systems including Ca 5 23 synthesized Ca 5 Ga 6 O 14 using solid state synthesis techniques and characterized the compound using X-ray diffraction, infrared and Raman spectroscopy, differential scanning calorimetry, thermogravimetric analysis, density measurements, and dilatometry. The latter was studied in both dry and moist atmospheres and after the sample was heated under moist conditions X-ray diffraction revealed CaGa 2 O 4 , Ga 2 O 3 , and CaCO 3 , thought to have transformed from CaO during cooling in an environment where CO 2 was present.
It has previously been identified that the Ca 5 Al 6 O 14 phase forms through decomposition kinetics of Ca 12 Al 14 O 33 , indicating that Ca 12 Ga 14 O 33 may be stable at low temperatures 8,24 . Similar to Ca 12 Al 14 O 33 it is expected the thermodynamic stability of the Ca 12 Ga 14 O 33 (ρ = 3.54 g/cm 3 ) will be dependent on availability of anions that are able to stabilize cage structure over the layered and more dense CaGa 4 O 7 (ρ = 4.46 g/cm 3 ), Ca 3 Ga 4 O 9 (ρ = 4.21 g/cm 3 ), and Ca 5 Ga 6 O 14 (ρ = 4.12 g/cm 3 ) phases.
Here we report on the successful synthesis of Ca 12 Ga 14 O 33 through a solution-based route, provide a crystal structure characterization from X-ray and neutron powder diffraction data, evaluate the phase formation through thermogravimetric analysis (TGA) and high temperature X-ray diffraction (HTXRD) studies, and compare and contrast the Ca 12 Al 14 O 33 and Ca 12 Ga 14 O 33 compounds. Solution synthesis allows for homogenous mixing of the metal atoms, shortening diffusion pathways and allowing for kinetically favorable, but thermodynamically unfavorable, phases to form.

Experimental and methods
The polymer assisted sol-gel synthesis method was employed utilizing poly vinyl alcohol (PVA) 25 . The 88% hydrolyzed PVA, with molecular weight between 20,000 and 30,000 g/mol, was dissolved in deionized water and allowed to stir for 1 h and Ca(NO 3 ) 2 ·4H 2 O (ACS Grade, Fisher Chemical) and GaCl 3 (99.99 +%, Acros Organics) were dissolved separately into deionized water. The stoichiometry of the solutions was characterized through gravimetric titration. For synthesis, stoichiometric amounts of the Ca(NO 3 ) 2 and GaCl 3 solutions were added to the PVA solution and allowed to stir for 1 h prior to heating. A 4:1 ratio of the number of metal cations to PVA monomer units was used as previously identified as the ideal amount for Ca 12 Al 14 O 33 8 . The solution was heated on a 300 °C hotplate until most of solvent evaporated and the solution became viscous. The viscous liquid was placed in a 120 °C drying oven for 12 h and dried to a light foam; foaming occurs due to the evaporation of nitrate and chloride species. The foam was ground to a fine powder and divided for the various characterization studies.
Thermogravimetric Analysis (TGA) was performed on the dried non calcined polymer powder using a TA Instruments Q50 TGA. The sample was heated at a constant rate of 10 °C/min from 25 to 650 °C. Mass loss was recorded as a function of temperature.
A fraction of the powder was then calcined to 600 °C, based on the TGA data, to decompose the organics to allow for clean processing in the XRD environmental chamber, and immediately quenched. The resulting calcined powder was pressed into a 13 mm pellet. The remaining organic/inorganic PVA powder was retained for further thermal processing. Both room temperature and HTXRD data were collected using a Malvern PANalytical Empyrean diffractometer with a Cu radiation source operating with an accelerating voltage of 45 kV and current of 40 mA. A PIXcel 3D area detector with 255 active channels with ~ 3°2θ of coverage was used for rapid non-ambient data collection. For non-ambient data collection the sample was heated at 5 °C/min up to 800 °C, then the rate was slowed to 1.5 °C/min as the sample was heated to 1000 °C. Data were collected in the range from 25 to 36.5°2θ, this region contains major peaks related to the Ca 12 Ga 14 O 33 phase and possible secondary phases including Ga 2 O 3 , Ca 5 Ga 6 O 14 , CaGa 2 O 4 , CaGa 4 O 7 , and Ca 3 Ga 4 O, and CaO. Using a 0.0131° step size and 13.77 s counting time each scan was approximately 1.5 min allowing for phase transitions to be detected while they occur. A longer data collection range, from 15-80°2θ, was used at 1000 °C to determine the full diffraction pattern and evaluate the evolved phases. The non-calcined powder was pressed into a pellet and was fired at 800 °C for 1 h. Subsequently ambient temperature data were collected from 15-120°2θ in high spatial resolution mode to verify the structure of the material and allow for refinement on the structural details including the lattice parameter, atomic positions, and atomic displacement parameters.
Time-of-flight powder neutron diffraction data were collected on the Nanoscale Ordered Materials Diffractometer (NOMAD) beamline 26 of the Spallation Neutron Source (SNS) at Oak Ridge National Laboratory (ORNL). About 1.5 g of near single-phase sample was contained in a 6 mm V sample canister. The partially calcined powder was fired at 800 °C for 1 h prior to loading the sample canister. NOMAD's detectors were calibrated using diamond, and silicon was used to generate the starting instrumental parameters. S(Q) was produced through normalizing the sample scattering via a solid vanadium rod and subtracting the background collected for an empty 6 mm V sample canister. Four individual data sets were collected and merged together for the data analysis.
HighScore Plus software package 27 interfaced with the ICDD PDF4 + database 28 was used for phase identification and the GSAS II software package 29 was used for the analyzing both the XRD and neutron diffraction data using the Rietveld method. For the data collected the lowest possible R obtainable for the data or wR min , based in part on the number of data points and calculated in GSAS II, was 10.39% for the XRD data and 0.43% for the NPD data.  25 . In this region the decomposition follows a relatively constant slope, indicating that decomposition is uniform as the temperature increases. This is ideal for preventing segregation of species within the sample during calcination. The sharp drop in mass above 425 °C is likely caused by the oxidation of carbonaceous residue left by polymer decomposition 25 . The sample experiences no further weight loss above 450 °C indicating that the remaining components in the sample are inorganic. The TGA results suggest that between 500 and 600 °C is the optimal temperature for calcination to ensure that all organics decompose. The sample experienced a total weight loss of 35% making this a high yield solution-based synthesis method.

XRD characterization.
In-situ high temperature X-ray diffraction was used to observe phase transformations during heating in the data range from 25 to 36.5°2θ. The intensity heat map of the scans is shown in Fig. 2 and has five distinct phase regions. Representative patterns from each temperature region are shown in Fig. 3. Data collected on the 600 °C pre-calcined pellet shows that at temperatures below approximately 300 °C the  www.nature.com/scientificreports/ sample is amorphous (Fig. 3a). After calcination most of the organics have been removed and a structureless mix of metal and oxygen atoms remains, with no diffraction peaks present in the data range. When heated above approximately 300 °C, two peaks appear at 28.9 and 31.2°2θ (Fig. 3b). In an attempt to identify the phase(s) that form between 300 and 650 °C data were collected in a larger °2θ range, between 10-80°2θ, and is shown in Fig. 4, revealing additional peaks associated with the phase(s). Despite attempts using the HighScore Plus software and the ICDD PDF4 + database, with and without chemical constraints, the peaks remain unidentified and could be representing either a single phase or a multiphase mixture. The sample had previously been heated to 600 °C during calcination to remove the organics, so this phase transformation is likely an atmospheric response of the amorphous Ca, Ga, and O species. At approximately 650 °C the unknown phase(s) disappear(s) and a multiphase mixture forms with peaks suggesting a compound isostructural to the Ca 12 Al 14 O 33 , with Ga substituted  (Fig. 3c). The peaks belonging to the Ca 12 Ga 14 O 33 compound get sharper and more intense as the temperature increases, consistent with phase and crystallite growth. CaGa 4 O 7 is present between 650 and 750 °C disappearing above 750 °C. The presence of this Ga rich phase could indicate a non-equilibrium assemblage that forms due to the high heating rate or potentially during the solution-based synthesis where nonstoichiometric areas of the precursor form but get quickly removed when the diffusion increases with higher temperature. Above 750 °C (Fig. 3d) single phase Ca 12 Ga 14 O 33 is present until a new peak at approximately 31.5°2θ appears at 975 °C (Fig. 3e). To identify the additional phase that is represented by the new peak, data were collection on a larger °2θ range (10-80°2θ) at 1000 °C. It was confirmed that the peak belongs to the CaO and using the Rietveld technique results in 19.8(3) wt% of CaO with all peaks in the pattern accounted for by either Ca 12 Ga 14 O 33 or CaO. A pellet of the amorphous sol-gel reactants was fired at 800 °C to form Ca 12 Ga 14 O 33 and high spatial resolution X-ray diffraction data were collected on the powdered sample at room temperature to verify the phase purity and characterize the crystal structure. To verify the crystal structure Rietveld refinements using two different models, based on the Ca 12 Al 14 O 33 structure, were attempted. The two slightly different atomic structures were reported by Bartl and Scheller 30 , determined from single crystal X-ray diffraction data, and Boysen et al. 31 (ICSD 241000) determined from neutron powder diffraction data. Both structures crystallized in the cubic crystal system, space group I43d (space group number 220) and with a unit cell edge close to 12 Å. The structure by Boysen et al. 31 differed from the structure by Bartl and Scheller 30 by having two unique partially occupied Ca positions on 24d sites as opposed to one fully occupied Ca 24d site. Another difference between the two reported structures is that Bartl and Scheller 30 placed the caged O on a 24d site (x, 0, ¼, specifically 0.337, 0, ¼) with a corresponding site occupancy of 0.083 for filling approximately two of the 12 cages/unit cell. Boysen et al. 31 working with neutron powder diffraction data, which is more sensitive to the O atoms (the bound coherent neutron scattering lengths for Ca, Al, and O are 4.70, 3.449, and 5.803 fm, respectively 32 ) moved the caged O from 0.337 to 0.375 or the 12a site ( 3 8 , 0, ¼) with a correspondingly higher site occupancy for filling approximately two of the 12 cages/unit cell. Additional studies have been carried out using synchrotron radiation to understand the details of the cage distortions and cage occupancy 33,34 . The intent of the characterization studies presented here is not to provide crystallographic details but to confirm the isostructural nature of Ca 12 Ga 14 O 33 to Ca 12 Al 14 O 33 and the model that fits best.
When using the structure with the two unique partially occupied Ca sites and caged O on the 12a site 31 and changing the Al atoms to Ga atoms, it was impossible to refine on the atomic displacement parameter (adp) for the O atom in the cage center without the isotropic atomic displacement parameter (U iso ) going negative. If the adps for all three oxygens in the structure were constrained together, a positive U iso was obtained but it was impossible to refine on the site occupancy factor of the caged oxygen and obtain a physically reasonable result. When the site occupancy factor was fixed at 1 6 and the U iso of the three oxygens constrained together, plausible results were obtained, however, there were intensity mismatches, shown in Fig. 5, for several of the peaks including the (211), (400), (420), and (422). The resulting agreement factor for the refinement is wR = 18.54% for a goodness of fit (GOF) (wR/wR min ) of 1.79. When using the structure with only one Ca position and the caged O on a 24d site 30 and changing the Al atoms to Ga atoms it was possible to refine on the adps of the oxygen atoms separately and the site occupancy factor of the caged oxygen. The intensity differences, between the calculated XRD pattern and the observed data, for the reflections discussed above were smaller, as shown in Fig. 6, and a better overall structural model is supported by the resulting smaller wR = 14.94% and the GOF = 1.44. Upon close www.nature.com/scientificreports/ examination there were a few additional weak peaks but these were accounted for by adding a small amount of CaGa 4 O 7 35 as a secondary phase, resulting in a wR = 13.75% and GOF = 1.33 with 1.9(2) wt% CaGa 4 O 7 . Experimental and refinement details, crystal data, and refined fractional coordinates, site occupancy factors, and isotropic thermal parameters are given in Tables 1, 2, and 3, respectively. The above suggests the better structural model for Ca 12 Ga 14 O 33 is the model with only one Ca position and the caged O in the 24d site 30 . However, given the low occupancy of any atomic or molecular species occluded in the cage, coupled with the occluded species having fewer electrons than the framework elements, Ca and Ga, the best structural model will depend on the elusive occluded atomic or molecular species.       31,32 . Both crystallographic models 30, 31 discussed above were attempted to determine which structural model provided a better fit to the neutron diffraction data. The refinement agreement factors using the model with one Ca position and caged O in the 24d site 30 were almost identical (wR = 4.84% and GOF = 16.61 refining on 57 variables) to those obtained using the crystallographic structural model with two unique Ca sites and the caged O on the 12a site 31 (wR = 4.83% and GOF = 16.59 refining on 59 variables). However, using the model with the two unique Ca positions and the caged O on the 12a site when the site occupancy of the caged O was refined it resulted in a value > 1 and poor background fitting using the same function and number of coefficients. The refined lattice parameter obtained from the neutron powder diffraction data, a = 12.316(3) Å (lattice parameter esds reported as 3σ), is significantly larger than the lattice parameter, a = 12.2993(3) Å, refined based on the X-ray diffraction data. Figure 7 compares the calculated patterns, generated from the refined variables, to the observed data in each of the four detector banks. The neutron powder diffraction data show a few unexplained peaks, marked in Fig. 7  www.nature.com/scientificreports/ mining secondary phases in this study could be due to the ability to definitively determine amount of cage occupant, to charge balance 1 6 of the cages are occupied by a O 2− anion. When using the structural model with one Ca position and the caged O on the 12a position 30 the site occupancy fraction (sof) of the caged oxygen refined to 0.610 (9), only slightly lower than that determined by the X-ray powder diffraction study (sof = 0.62(1)), and still high compared to sof = 0.167 that should result if one out of every six cages are occupied. Figure 8 illustrates how manually changing the site occupancy of the caged oxygen atom has a significant impact on several of the peaks in the neutron powder diffraction pattern. Future work using total scattering or pair distribution function (pdf) analysis on time-of-flight neutron powder diffraction data will reveal more information for a better understanding of the nanoscale structural features of the compound and specifically about the atom(s) present in the Ca 12 Ga 14 O 33 cages. The results presented here are only to provide a second characterization technique that supports the synthesis of Ca 12 Ga 14 O 33 .
Comparison of Ca 12 Ga 14 o 33 to ca 12 Al 14 o 33 and conclusions Ca 12 Ga 14 O 33 was successfully synthesized using the polymer-assisted steric entrapment method 25 . X-ray and neutron diffraction data indicate that the new compound is isostructural to Ca 12 Al 14 O 33 with a full exchanged of Al to Ga. HTXRD showed a formation temperature of 650 °C, which is significantly lower than that observed for Ca 12 Al 14 O 33 synthesized in a similar manner 8 . The best refinement, based on both the laboratory XRD and time-of-flight neutron powder diffraction data, resulted with the model that only has one Ca position and the caged O on the 24d site. The refined lattice parameter for the new Ca 12 Ga 14 O 33 compound is a = 12.2993(3) Å (XRD), approximately 2.6% larger than the 11.9794 Å of Ca 12 Al 14 O 33 31 . This expansion in lattice parameter leads to a 1.6% expansion of the cages as shown in Fig. 9. The larger framework and cages potentially open the possibility of occluding larger molecular species within the cage. Future work is needed to elucidate the current cage occupants, better understand the local structure of the compound, determine potential processing conditions that converts the compound into an electride structure, and evaluate the changes in electrical properties based on cage occupants.