Hydride Reduction of BaTiO3 - Oxyhydride Versus O Vacancy Formation.

We investigated the hydride reduction of tetragonal BaTiO3 using the metal hydrides CaH2, NaH, MgH2, NaBH4, and NaAlH4. The reactions employed molar BaTiO3/H ratios of up to 1.8 and temperatures near 600 °C. The air-stable reduced products were characterized by powder X-ray diffraction (PXRD), transmission electron microscopy, thermogravimetric analysis (TGA), and 1H magic angle spinning (MAS) NMR spectroscopy. PXRD showed the formation of cubic products-indicative of the formation of BaTiO3-x H x -except for NaH. Lattice parameters were in a range between 4.005 Å (for NaBH4-reduced samples) and 4.033 Å (for MgH2-reduced samples). With increasing H/BaTiO3 ratio, CaH2-, NaAlH4-, and MgH2-reduced samples were afforded as two-phase mixtures. TGA in air flow showed significant weight increases of up to 3.5% for reduced BaTiO3, suggesting that metal hydride reduction yielded oxyhydrides BaTiO3-x H x with x values larger than 0.5. 1H MAS NMR spectroscopy, however, revealed rather low concentrations of H and thus a simultaneous presence of O vacancies in reduced BaTiO3. It has to be concluded that hydride reduction of BaTiO3 yields complex disordered materials BaTiO3-x H y □(x-y) with x up to 0.6 and y in a range 0.04-0.25, rather than homogeneous solid solutions BaTiO3-x H x . Resonances of (hydridic) H substituting O in the cubic perovskite structure appear in the -2 to -60 ppm spectral region. The large range of negative chemical shifts and breadth of the signals signifies metallic conductivity and structural disorder in BaTiO3-x H y □(x-y). Sintering of BaTiO3-x H y □(x-y) in a gaseous H2 atmosphere resulted in more ordered materials, as indicated by considerably sharper 1H resonances.

Introduction

Hydride reduction has developed into a versatile method for modifying transition metal oxides, yielding highly reduced products with unusual coordination environments and exotic electronic and magnetic properties.[1−4] Especially fruitful precursors are oxides that relate to perovskite and Ruddlesden–Popper phases. Specific prominent examples are the LaNiO3 to LaNiO2, SrFeO3 to SrFeO2, and YBaCo2O5 to YBaCo2O4.5 reductions.[5−7] In a typical hydride reduction, a ternary transition metal oxide is reacted with CaH2, NaH, or LiH at relatively mild, low temperature (“chimie douce”) conditions (150–600 °C). The reduction is accompanied with O removal (i.e., O vacancy formation), and frequently, there is a strict topotactic relationship between the original oxide and its reduced form. In rare cases, transition metal reduction and hydride ion insertion occur simultaneously, leading to oxyhydrides.[8] This scenario is thought to be highly unusual because of the incompatibility of O2– and strongly reducing H– in forming a common anion substructure.[9]

It then came as a surprise when in 2012 Kobayashi et al. reported that the reaction of CaH2 with the archetypical perovskite BaTiO3 affords BaTiO3–H with large amounts (x < 0.6) of hydrogen incorporated.[10] The cubic perovskite oxyhydride BaTiO3–H is remarkable in several respects. It represents a defect-/vacancy-free solid solution of O2– and H– ions, which commonly form the octahedral environment around Ti that is now in a mixed IV/III oxidation state (Figure a). BaTiO3–H is stable in air and water. Further, it is stable at elevated temperatures of up to approximately 400 °C above which hydrogen is released. When present, oxygen is scavenged and BaTiO3 is retained. In inert gas atmospheres containing D2, a hydride exchange H/D occurs at hydrogen release temperatures. Also, oxynitrides BaTiO3−δN may be prepared by heating BaTiO3–H under N2 flow at 400–600 °C.[11] These observations led to the conclusion that the hydride species in BaTiO3–H is labile and that the material represents a versatile precursor toward new mixed-anion compounds.[12] Sakaguchi et al.[13] and Yamamoto et al.[14] showed that perovskite oxyhydrides are also obtained with CaTiO3, SrTiO3, and EuTiO3 although with smaller hydride contents x (<0.3). A high degree of O2–/H– exchange was further observed for cubic SrVO3 for which topochemical hydride reduction with CaH2 yields tetragonal SrVO2H.[15] Remarkably, the structure of SrVO2H is fully anion-ordered, displaying two-dimensional VO2 sheets that are connected by H– ligands (cf. Figure b).

Figure 1

(a) Structure of BaTiO2.75H0.25 represented as the 2 × 2 × 2 supercell of cubic BaTiO3 in which two O atoms were randomly replaced by two H atoms. The right-hand side shows the local coordination of H in BaTiO3–H. The Ti is in a mixed IV/III oxidation state. (b) Anion-ordered tetragonal structure of SrVO2H.[15] The alkaline earth metal, O, and H atoms are depicted as gray, red, and green spheres, respectively.

Mechanisms or processes behind metal hydride reductions are not well investigated and far from understood. They involve intermediates, which in turn depend on the particular system and conditions applied. Also, the active reducing species may be H– or H2 or a combination of both. Hayward et al. could show that the reduction of LiNiO3 with NaH corresponds to a solid-state diffusion-controlled reaction that proceeds via the intermediate LaNiO2+.[5] For other systems, there may be significant H2 gas-phase contributions to reduction.[16] Conditions can be varied by the choice of reducing agents (i.e., metal hydride), reaction temperature and time, and the activity/concentration of H–. Hernden et al. performed a comprehensive investigation of the hydride reduction of Sr2MnO4 and identified a two-electron transfer H– → H+ (as opposed to one-electron transfer H– → 1/2H2) during reduction.[17] For oxyhydride formation, additional complexity arises. Are reduction and exchange mechanisms coupled, or does oxyhydride formation occur via an O vacancy-rich intermediate? Bridges et al. studied the formation pathway of the first reported transition metal oxyhydride LaSrCoO3H0.7 from LaSrCoO4.[18] From in situ X-ray diffraction studies, they could identify oxygen-deficient LaSrCo3O3.38– as the reduced intermediate, in which oxide was subsequently substituted by H–. At the same time, the presence of gas-phase H2 appeared to be important for achieving the final composition LaSrCo3H0.7. The proposed three-step mechanism is rather complex, but it is clear that direct hydride substitution for oxide, coupled with reduction, is excluded.

An important question is what factors govern oxyhydride formation as opposed to exclusive vacancy formation during hydride reduction. What mechanism could apply for BaTiO3 reduction? Whereas defined reduced phases can be obtained from LaSrCoO4 (i.e., LaSrCoO3.50 and LaSrCoO3.38),[18] BaTiO3– with significant O deficiency is not known. We report a systematic study of the hydride reduction of BaTiO3, especially investigating the influence of reducing agents. In contrast to previous studies, we find that hydride reduction leads to phases BaTiO3–H□( with comparatively low H content and, accordingly, large O vacancy concentrations.

Experimental Methods

Synthesis

As starting materials, we used BaTiO3 (500 nm particle size, 99.9% purity, ABCR GmbH) and powders of the metal hydrides CaH2 (99.99%, Sigma-Aldrich), NaH (95%, Sigma-Aldrich), MgH2 (96.5%, Sigma-Aldrich), NaBH4 (98%, ABCR GmbH), and NaAlH4 (93%, Sigma-Aldrich). Prior to use, BaTiO3 was dried overnight in an oven at 200 °C. All steps of sample preparation for the synthesis reactions were performed in an Ar-filled glovebox. For a typical synthesis, ∼1.4 g (6 mmol) of BaTiO3 was intimately mixed with metal hydride by grinding the materials together in an agate mortar for 15 min. We considered the molar proportions BaTiO3 + nH, with n = 0.2, 0.6, 1.2, and 1.8 and with H = NaH, 0.5 CaH2, 0.5 MgH2, 0.25 NaAlH4, and 0.25 NaBH4. Samples are in the following termed “n-H-MH”, to indicate the hydride concentration and hydride source (M = Na, Ca, Mg, “NaAl”, “NaB”) in the reaction. The BaTiO3/MH mixture was subsequently pressed into a pellet with a diameter of 8 mm. The pellet was then sealed inside a stainless steel ampule (with dimensions 10 mm ID and 80 mm length), which in turn was placed in a fused silica jacket.

Following removal from the glovebox, the silica jacket was placed in a vertical tube furnace and connected to a vacuum line. The empty space of the furnace was filled with silica insulation wool, and a K-type thermocouple was introduced parallel to the silica jacket to monitor the temperature close to the location of the metal ampule. The silica jacket was then evacuated, and ampules were heated for times between 1 and 7 days. Reaction temperatures were varied from 500 to 700 °C. Most reactions reported in this work were carried out at 600 °C in 2 days’ duration. A deuterated sample 1.2-D-CaD2 was prepared by reacting a mixture of BaTiO3 and CaD2 at 600 °C for 2 days. CaD2 was synthesized prior by reacting dendritic pieces of Ca (99.99% trace metals purity, Sigma-Aldrich) with D2 (99.9% isotope purity, AGA) in a stainless-steel autoclave pressurized to 30 bar at 400 °C for 12 h.

Products were washed with 0.1 M CH3COOH (HAc) to remove excess metal hydride and metal oxide formed during hydride reduction and dried at 120 °C under dynamic vacuum (<10–5 bar). In order to study the effect of the acidic washing agents, we also used a 0.1 M NH4Cl/methanol (MeOH) mixture and 0.1 M HCl. For washing, the pellets were crushed and sonicated with 50 mL acidic washing agent for 15 min and then centrifuged. The procedure was repeated three times. As a last step, samples were treated with 20 mL of pure methanol.

Selected reduced BaTiO3 samples were subjected to a posttreatment in which the pellets were heated in a corundum crucible in a stainless steel autoclave at 600 °C under a pressure of 30–50 bar of hydrogen gas (H2 AGA 99.99% purity) for 24 h.

Powder X-ray Diffraction (PXRD) Analysis

Ambient-temperature PXRD patterns were collected on a PANalytical X’Pert PRO diffractometer operated with Cu Kα1 radiation in θ–2θ diffraction geometry. Powder samples were mounted on a Si wafer zero-background holder, and diffraction patterns were measured in a 2θ range of 10–90° with 0.013° step size. In situ high-temperature PXRD studies were performed on a PANalytical X’Pert PRO instrument with Cu Kα radiation in θ–2θ diffraction geometry. Samples were heated in air to 900 °C using an Anton Paar XRK 900 high-temperature chamber equipped with Be windows and connected to a temperature controller. Powder samples were mounted on a Au wafer. Data in a 2θ range of 20–60° were collected at room temperature and then in steps of 100 °C with 15 min acquisition time and 5 °C/min heating rate between the steps. Au reflections were used for correction of the 2θ-scale of data. The contribution of Kα2 radiation to the PXRD patterns was removed using the PANalytical X’Pert HighScore Plus software. The Rietveld method as implemented in the FullProf program was used for structure and phase analysis.[19] A six-coefficient polynomial function was applied for the background. The peak shape was described by a pseudo-Voigt function. Patterns with pronounced peak shape asymmetries were refined as mixtures of two cubic phases. Site occupancies for the O atoms could not be refined reliably and were constrained to 1.

Transmission Electron Microscopy (TEM) Investigations

The TEM images were recorded using either a JEOL JEM-2100F (Cs = 0.5 mm and point resolution of 1.9 Å) or a JEOL JEM-2100 microscope (Cs = 1.4 mm and point resolution 2.5 Å). Both microscopes were operated at room temperature with an accelerating voltage of 200 kV. The TEM specimens were prepared by crushing the dry powders in an agate mortar, followed by dispersion in ethanol. One droplet of this suspension was transferred onto a copper grid coated with a holey-carbon film.

Thermogravimetric Analysis (TGA)

TGA experiments were carried out using a TA instruments Discovery system. The samples (∼15 mg powders) were heated in a platinum crucible from room temperature to 900 °C with a heating rate of 5 °C/min. A dry air gas flow of 20 mL/min was applied.

Magic Angle Spinning (MAS) NMR Spectroscopy

The 1H MAS NMR experiments were performed at a magnetic field of 14.1 T (600.12 MHz 1H Larmor frequency) and a MAS frequency of 60 kHz on a Bruker AVANCE-III spectrometer equipped with a 1.3 mm MAS HX probe. Spectra were acquired using a rotor-synchronized, double-adiabatic spin-echo sequence with a 1.2 μs 90° excitation pulse length, followed by a pair of 50 μs tanh/tan short high-power adiabatic pulses (SHAPs) with 5 MHz frequency sweep.[20−22] All pulses were applied at a nutation frequency of 208 kHz. Signal transients (4096) with a 5 s recycle delay were accumulated for each sample. The shifts were referenced with respect to tetramethylsilane (TMS) at 0 ppm. The 2H MAS NMR spectrum was recorded on a 9.4 T (61.41 MHz 2H Larmor frequency) Bruker AVANCE-III spectrometer with a 2.5 mm MAS HX probe at a spinning frequency of 30 kHz. The same pulse sequence as for protons was used but with a 3.0 μs 90° excitation pulse and 66.67 μs SHAPs at a nutation frequency of 83 kHz. Scans (4096) with a 20 s recycle delay were collected. The shifts were referenced to deuterated TMS (TMS-d12) at 0 ppm.

Results and Discussion

Hydride Reduction of BaTiO3

Kobayashi et al. prepared BaTiO3–H by reacting BaTiO3 in the form of 170 nm sized particles with CaH2 at temperatures between 500 and 580 °C for 4–7 days in pyrex or fused silica tubes.[10,13] CaH2 was employed with a rather large excess, 3 M, corresponding to 6 H according to our nomenclature. The reacted samples were washed with 0.1 M NH4Cl/MeOH in air to remove excess CaH2 as well as the side product CaO. Very weakly acidic NH4Cl/MeOH is commonly used as a washing agent for concluding hydride reductions.[1−4] The dark blue-colored cubic products were identified as BaTiO3–H and described as a defect-free O2–/H– solid solution containing xTi(III) and 1–xTi(IV) with x up to 0.6.

In the following, we outline possible processes that occur during the hydride reduction of BaTiO3. The H– species as a reductant may transfer one or two electrons toward oxygen vacancy formation (“one-” and “two”-electron processes, respectively). In addition, the metal hydride may decompose at elevated temperatures into hydrogen gas and metal, which can both act as reductants. H2 gas-phase contribution to hydride reduction reactions has been demonstrated by Kobayashi et al.[16] For oxyhydride formation, H– formally acts by simultaneously reducing Ti and replacing O2– in the perovskite structure. However, the reaction may proceed via initial vacancy formation and subsequent oxidation by gaseous H2. Processes involving hydrogen are summarized in Table . Additionally, the explicit hydride reduction reactions for BaTiO3 are formulated below for CaH2 (a) and for a generic metal hydride MH (b).

Table 1

Processes Involving Hydrogen Species during Hydride Reduction

process	oxidation	side products
vacancy, 1e	H– → H• + e	0.5(O2– + H2)
vacany, 2e	H– → H+ + 2e	OH–
H2, gas phase	H• → H+ + e	0.5H2O
oxyhydride formation	2H– → H– + H• + e	0.5(O2– + H2)

Vacancy formation, one-electron process

Vacancy formation, two-electron process

Vacancy formation, H2

Oxyhydride formation, direct substitution

Oxyhydride formation, via a defect intermediate

Note that if hydroxide is formed (two-electron process), it may most likely not be observed in the ex situ analysis of products because hydroxide will react with excess hydride to yield oxide and H2 (OH– + H– → O2– + H2).

Reduction with CaH2

Our reactions were performed at 600 °C in welded stainless steel ampules. The average particle size of the BaTiO3 starting material was 500 nm, which is slightly larger than the material used by Kobayashi et al. Products were washed with 0.1 M HAc, which expedited the procedure considerably compared to the much less acidic NH4Cl/MeOH. From PXRD, TGA, and 1H NMR measurements, we concluded that the products after washing with 0.1 M HAc and 0.1 M NH4Cl/MeOH were identical (see the Supporting Information, part III). Figure shows the evolution of products with increasing CaH2 concentration during 2 day experiments at 600 °C, and Table presents the results from the refinement of the PXRD patterns. The product obtained with 0.2 H (“0.2-H-CaH2”) had a pale blue color and remained tetragonal. With the higher concentrations, 0.6, 1.2, and 1.8 H cubic products were obtained. A higher degree of reduction with increasing H concentration was recognizable by a deepening of the color to dark blue, almost black, in agreement with the observation of Kobayashi et al.[10,13] The unit cell volume of the reduced products is very similar to that of the starting material, increasing slightly with increasing H concentration. A closer inspection of the PXRD patterns of 1.2-H-CaH2 and 1.8-H-CaH2 revealed that reflections have a pronounced shoulder at lower angles, suggesting phase heterogeneity. These patterns were refined as a mixture of two cubic phases. For both products, the minority phase had a weight fraction of about 11% and a unit cell parameter that was larger than that of the majority phase by about 0.014 Å. We note that the lattice parameter reported by Kobayashi et al. for BaTiO2.38H0.62, 4.0236 Å,[10] is similar to the minority phase of the 1.2-H-CaH2 and 1.8-H-CaH2 samples. Higher concentrations than 1.8 H led to a drastic broadening of reflections and a diminished crystallinity.

Figure 2

(a) PXRD patterns of products obtained from the hydride reduction of BaTiO3 at 600 °C during 2 days, using various concentrations of CaH2. (b) Rietveld plots for the PXRD patterns of the samples obtained with 0.6, 1.2, and 1.8 H concentrations of CaH2 showing the evolution of two-phase mixture with increasing H concentration.

Table 2

Synthesis Products from Hydride Reduction with CaH2 during 2 Day Experiments at 600 °C

sample	product/fraction (wt %)	lattice parameters (Å)	volume (Å3)	x from TGa
0-H	tetragonal	a = 3.9964(1), c = 4.0310(1)	64.379(2)	0
0.2-H	tetragonal	a = 3.9971(1), c = 4.0260(1)	64.324(3)	0.03
0.6-H	cubic	4.0051(6)	64.247(2)	0.10
1.2-H	cubic-I/89(1)	4.0096(2)	64.461(2)	0.24
	cubic-II/11(1)	4.0219	65.095
1.8-H	cubic-I/89(1)	4.0138(1)	64.662(2)	0.34
	cubic-II/11(1)	4.0288	65.391

x refers to a reaction BaTiO3–H + 0.75xO2 → BaTiO3 + 0.5xH2O.

The presence of two cubic phases in the 1.2-H-CaH2 and 1.8-H-CaH2 samples, as evidenced from PXRD, could not be reconciled from TEM investigations. Figure a,b shows TEM images of the starting material and 1.2-H-CaH2, respectively. The morphology of the crystalline particles is not influenced by the hydride reduction reaction. The electron diffraction pattern of investigated 1.2-H-CaH2 crystallites, shown in Figure c, is cubic without signs of superstructuring or diffuse scattering. However, the thickness of the crystallites hampered detailed diffraction studies.

Figure 3

(a) TEM images of the BaTiO3 starting material (with an average particle size of 500 nm according to the specification of the supplier). (b) TEM image of a 1.2 H CaH2-reduced sample. (c) Electron diffraction pattern along [001] shows cubic symmetry.

Kobayashi et al. suggested that TGA represents a convenient way to assess the H content of BaTiO3–H.[10,13] TGA under the flowing air will monitor the reactionaccording to which a substantial weight increase occurs, and thus, a high accuracy can be associated with the determination of x. The product after TGA (i.e., after heating and cooling to 900 °C) is white and corresponds to tetragonal BaTiO3. Figure collects the TGA traces for the CaH2-reduced samples. All samples show initially a small weight loss (0.1–0.15%) which we attribute to the loss of surface water/hydroxyls. For the 0.2-H-CaH2 sample, the weight loss continued up to 600 °C, for 0.6-H-CaH2 up to 450 °C, and for 1.2-H-CaH2 and 1.8-H-CaH2 up to 350 °C. The initial weight loss relates well to the TG behavior of the starting material, which shows a continuous weight loss amounting to 0.15% up to 700 °C (as shown later). The subsequent weight increase should be due to oxidation, which for all samples is completed above 700 °C. The associated x values according to eq are contained in Table . The value x = 0.34 for the 1.8-H-CaH2 sample is clearly below the maximum value reported by Kobayashi et al. (x ≈ 0.6).[10,13]

Figure 4

TGA traces for products obtained from the hydride reduction of BaTiO3 at 600 °C during 2 days, using various concentrations of CaH2.

Figure shows the lattice parameter variation as a function of temperature as obtained from a multi-temperature PXRD experiment in which the 1.2-H-CaH2 sample was heated in air to 900 °C and subsequently cooled. In agreement with the TGA experiment, oxidation occurs between 500 and 600 °C. Above 600 °C, the lattice parameters during heating and cooling coincide. Upon cooling, the phase transition into tetragonal BaTiO3 occurs below 200 °C.[23] The lattice parameter of the cubic high-temperature form of BaTiO3 is clearly smaller than that of the reduced samples [by about 0.07 Å for 1.2-H-CaH2 and by about 0.015 Å for 1.2-H-MgH2 (Supporting Information, Figure S1)].

Figure 5

Evolution of the lattice parameters during heating and cooling of 1.2 H CaH2-reduced BaTiO3 in air. Standard deviations are less than the size of the symbols.

The influence of reaction time is depicted in Figure for the 1.2 H reductions, and results from the refinement of the PXRD data are compiled in Table . All products represent mixtures of two cubic phases. The 1- and 2 day experiments produced virtually identical products, which are also very similar to the 4 day experiment, whereas the product of the 7 day experiment is clearly different. The PXRD pattern shows significantly broader reflections and also the presence of an impurity phase (Ti3O), which is indicative of the onset of decomposition of BaTiO3. The unit cell parameters are increased with respect to the products obtained after shorter times. Peculiar is the TGA trace, exhibiting significant weight loss (∼0.4%) between room temperature and 350 °C, after which a weight increase in excess of 3% occurs. We also mention briefly the influence of temperature. Below 550 °C, the reactivity of CaH2 is low and a conversion into a cubic product was not observed after 5 days. Temperatures above 650 °C represented harsh conditions, equivalent to long reaction times and/or high concentrations of CaH2, H/BaTiO3 > 4.

Figure 6

PXRD patterns (a) and TGA traces (b) of products obtained from the hydride reduction of BaTiO3 with 1.2 H CaH2 at 600 °C during 1, 2, 4, and 7 day experiments. The arrow marks a reflection from Ti3O in the 7 day pattern.

Table 3

Synthesis Products from Hydride Reduction with 0.6 M CaH2 at 600 °C at Varying Reaction Times

1.2-H-CaH2 reaction time	product/fraction (wt %)	lattice parameters (Å)	volume (Å3)	x from TG
1 d	cubic-I/91(2)	4.0093(2)	64.446(4)	0.24
	cubic-II/9(2)	4.0219(1)	65.056
2 d	cubic-I/87(2)	4.0079(1)	64.380(4)	0.24
	cubic-II/13(2)	4.0205(1)	64.987
4 d	cubic-I/89(2)	4.0094(1)	64.451(4)	0.26
	cubic-II/11(2)	4.0221(1)	65.068
7 d	cubic-I/70(1)	4.0173(1)	64.834(2)	0.51
	cubic-II/28(1)	4.0275(1)	65.330
	Ti3O/2(1)

To summarize, we could reproduce the hydride reduction of BaTiO3 with CaH2 as reported by Kobayashi et al.[10,13] At the same time, we noticed profound differences, such as heterogeneous products, which indicates that the hydride reduction of BaTiO3 is sensitive to the experimental conditions.

Reduction with NaH, MgH2, NaBH4, and NaAlH4

Hydride reduction of BaTiO3 has hitherto only been reported with CaH2. In the following, we describe the reaction with the reductants NaH (MH), MgH2 (MH2), and NaBH4 and NaAlH4 (MH4). Originally, we also included LiH, LiBH4, and LiAlH4 in this investigation. However, Li-containing metal hydrides exhibited a different behavior, which not only involved H but also Li as the reactive species toward the modification of BaTiO3. A detailed investigation of the Li-containing metal hydride reactions will be reported elsewhere. The decomposition temperatures of NaH, MgH2, NaBH4, and NaAlH4, referring to an equilibrium pressure of 1 bar, are in the range 300–400 °C, which is substantially lower than that of CaH2 (∼600 °C).[24] Consequently, one may expect a substantial H2 gas-phase contribution to hydride reduction. The results are compiled in Table , in Figure (TGA traces), and in the Supporting Information (Figure S2, PXRD patterns).

Table 4

Synthesis Products from Hydride Reduction with NaH, MgH2, NaBH4, and NaAlH4 during 2 Day Experiments at 600 °C

sample	product/fraction (wt %)	lattice parameters (Å)	volume	x from TGa
0-H	tetragonal	a = 3.9964(1), c = 4.0310(1)	64.379(2)	0
NaH
0.2-H	tetragonal	a = 3.9989(1), c = 4.0197(1)	64.280(3)	N/A
0.6-H	tetragonal	a = 4.0001(1), c = 4.0171(1)	64.276(2)	N/A
1.2-H	tetragonal	a = 4.0002(1), c = 4.0164(1)	64.269(2)	N/A
1.8-H	tetragonal	a = 4.0007(1), c = 4.0151(1)	64.263(3)	N/A
MgH2
0.2-H	cubic	4.0046(1)	64.222(1)	0.04
0.6-H	cubic	4.0091(1)	64.438(2)	0.21
1.2-H	cubic	4.0198(1)	64.954(2)	0.41
1.8-H	cubic-I/48(1)	4.0331(1)	64.96(1)	0.55
	cubic-II/52(1)	4.0199(1)	64.566(9)
NaBH4
0.2-H	tetragonal	a = 4.0017(1), c = 4.0132(1)	64.265(3)	0.04
0.6-H	cubic	4.0046(1)	64.220(1)	0.18
1.2-H	cubic	4.0046(1)	64.216(2)	0.36
1.8-H	cubic	4.0045(1)	64.217(2)	0.60
NaAlH4
0.2-H	cubic	4.0045(1)	64.22(2)	0.05
0.6-H	cubic-I/83(1)	4.0065(1)	64.312(1)	0.19
	cubic-II/17(1)	4.0155(1)	64.747(1)
1.2-H	cubic-I/44(1)	4.0107(1)	64.514(1)	0.39
	cubic-II/56(1)	4.0202(1)	64.973(1)
1.8-H	cubic-I/30(3)	4.0220(1)	64.78(1)	0.56
	cubic-II/70(3)	4.0163(3)	65.064(9)

x refers to a reaction BaTiO3–H + 0.75xO2 → BaTiO3 + 0.5xH2O.

Figure 7

TGA traces of products obtained from the hydride reduction of BaTiO3 at 600 °C during 2 day experiments using NaH (a), MgH2 (b), NaAlH4 (c), and NaBH4 (d). (a) also contains the TGA trace of the starting material.

All samples from NaH reductions displayed a pale blue color, and the PXRD patterns (Figure S2a) show that conversion to a cubic product was not achieved. The trend in the lattice parameters indicates decreased tetragonality with increasing NaH concentration (Table ). The TGA traces (Figure a) reveal a small weight loss up to 350 °C (0.25–0.45%), which is followed by a slightly irregular behavior up to 700 °C, above which weights are constant. In Figure a included is the TG trace of the starting material. BaTiO3 exhibited a small (0.15%) weight loss between room temperature and 740 °C. The continuous nature of this weight loss indicates gradual release of water, initially physically adsorbed water on surface hydroxyls and then water from the condensation of surface hydroxyls. Typically, surface hydroxyls have various bonding energies and are desorbed over a wide temperature range.[25,26] We infer that this small weight loss is also present in metal hydride-reduced samples. During TGA, this process overlaps with the weight increase because of oxidation, thus leading to a minor underestimation of the weight increase. The weight loss of the NaH-reduced samples significantly exceeds that of the starting material. This suggests that reduction primarily affected and modified the surface of BaTiO3 particles, which, after washing, resulted in a higher concentration of hydroxyl and absorbed water. Note that the TGA weight loss behavior (up to 350 °C) of the NaH-reduced samples and the 1.2-H-CaH2-7d sample appears to be very similar (cf. Figure b).

Using MgH2 as the reducing agent yielded the cubic phase already with 0.2 H (Figure S2b). The samples produced with 0.2 H, 0.6 H, and 1.2 H represented single-phase products, with the lattice parameter increasing from 4.005 to 4.020 Å (Table ). With 1.8 H, a mixture of two phases—with lattice parameters 4.020 and 4.033 Å and roughly equal proportions—was obtained. Generally, the lattice parameters of the MgH2-reduced samples are larger compared to that of the CaH2-reduced ones. Also, and perhaps related, the TGA weight increase (Figure b) due to oxidation is larger compared to that of CaH2-reduced ones. The x value attained for the 1.8-H-MgH2 sample approaches the maximum value 0.6 proposed by Kobayashi et al.[10] The results for NaAlH4 are shown in Figures c and S2c. Similar to MgH2 reduction, a single-phase cubic product is obtained already with 0.2 H. The samples 0.6-H-NaAlH4, 1.2-H-NaAlH4, and 1.8-H-NaAlH4, however, represented two phase mixtures. The TGA traces of NaAlH4- and MgH2-reduced samples compare well.

The results for the NaBH4 reductions are shown in Figures d and S2d. The sample 0.2-H-NaBH4 corresponded to the tetragonal phase, whereas higher synthesis H concentrations afforded single-phase cubic products. Remarkably, the lattice parameter of these products shows virtually no dependence on synthesis H concentration and attains a comparatively small value, 4.005 Å (Table ). The TGA traces of the NaBH4-reduced samples (Figure d) are different from those of the CaH2-, MgH2-, and NaAlH4-reduced ones, in that the weight increase is already observed between 200 and 400 °C, that is, at temperatures about 200 °C lower. In addition, a peculiar kink near 500 °C is noticeable, which suggests that the oxidation occurs in two steps. Above 650 °C, the weights are constant and the total weight increases compare well with the corresponding MgH2- and NaAlH4-reduced samples.

In summary, NaH acts only as a weak reducing agent toward BaTiO3, which should relate to the low decomposition temperature of NaH, possibly making gaseous H2 the reducing species. We conjecture that reduction with NaH primarily affects and modifies the surface of BaTiO3 particles. The other metal hydrides afford cubic products. The TGA weight increase for NaBH4-, MgH2-, and NaAlH4-reduced samples is very similar, with x values for 1.8 H products being near to 0.6, which was previously reported as the maximum value for BaTiO3–H.[10] Judging from the TGA weight increase (x values) of the reduced samples, the reductive reactivity of CaH2 appears to be lower compared to that of MgH2, NaAlH4, and NaBH4. From the PXRD, one can notice differences between the cubic products obtained with the various reductants. Products from NaBH4 reduction are distinguished in that they are single phase even at the highest H concentrations applied, whereas the other reductants yield mixtures of two cubic phases with increasing H concentration. MgH2-reduced samples attain the largest lattice parameters. With respect to unit cell parameters, one may discern three main scenarios, 4.005 Å—NaBH4, 4.01 Å—CaH2, and 4.02 Å—MgH2. We note that 0.1 HAc as the washing agent removed effectively the oxides MgO and “NaAlO2”, which is rather difficult when using NH4Cl/MeOH. Using NaBH4 as the reducing agent yielded an amorphous side product that could not be removed by washing with a weak acid. The presence of amorphous “NaBO2” only became apparent when analyzing the samples after TGA. The PXRD pattern showed extra reflections from BaTi(BO3)2, presumably formed according to 2 NaBO2 + BaTiO3 → BaTi(BO3)2 + Na2O. Perhaps this reaction also influences the shape of the TG curve and may explain the peculiar kink feature near 500 °C.

Solid-State 1H MAS NMR Investigations

1H MAS NMR has not yet been considered for the detailed analysis of new BaTiO3–H. Kobayashi et al., in their initially report on BaTiO3–H, showed a spectrum with a single sharp signal at 4.4 ppm (the sample composition was not stated).[10] They concluded that all H species are in the same chemical state and environment, in agreement with a defect-free solid solution in which O2– is randomly replaced by H– in the perovskite structure.

Figure shows 1H MAS NMR spectra recorded of the starting material before and after TGA, that is, before and after heating to 900 °C in air. Three distinct resonances are observed at 1.1, 4.8, and 6.5 ppm, together with two shoulders at ∼2 and ∼4 ppm, respectively. All of these proton signals are attributed to the surface OH species; the presence of structural hydroxyl in the starting material is excluded because of the great similarity of the spectra before and after TGA. Unlike surface OH, structural OH will not reform upon exposure to ambient atmosphere after the TGA treatment. Also, the occurrence of structural OH is more typical for the hydrothermally synthesized BaTiO3, which are afforded as nanoparticles with cubic symmetry.[25−28] These are clearly not the characteristics of our starting material (cf. Figure a). Because the distinct chemical shifts of 1H resonances stem from the specific surface environments, the bonding energies associated with the respective proton sites are also likely to differ, which is reflected by the continuous weight loss seen in the TG experiment (cf. Figure a).

Figure 8

1H MAS NMR spectrum of the BaTiO3 starting material before (black line) and after (red line) heating to 900 °C in air flow (during a TGA experiment).

Figure a shows the 1H MAS NMR spectra recorded from the 0.2-H-CaH2 and 1.2-H-CaH2 samples. The most noticeable difference is a broad resonance centered at −18 ppm, with an additional shoulder at −60 ppm, which appears in the spectrum of the 1.2-H-CaH2 sample and which is absent in the spectrum of 0.2-H-CaH2. We attribute this signal to hydridic H on the O position in the perovskite structure. This is unambiguously corroborated by the 2H MAS NMR spectrum of the analogous sample 1.2-D-CaD2, which was prepared in the same way as the 1.2-H-CaH2 specimen (i.e., 2 day synthesis at 600 °C), but with a deuterated reducing agent. Obviously, the hydroxyl resonances are absent in the 2H spectrum, which shows a single signal that exhibits the same chemical shift as the broad resonance in the 1H spectrum (Figure b). The 1H MAS NMR signals of 0.2-H-CaH2 and 1.2-H-CaH2 in the (positive) 0–10 ppm range have a similar pattern (seen in Figure c). When compared to the untreated BaTiO3, the CaH2-reduced samples exhibit additional signals in the 1–7 ppm region and also altered intensities. This indicates that upon reduction (and subsequent washing), additional surface OH sites are created, along with a population redistribution of the pre-existing ones.

Figure 9

(a) 1H MAS NMR spectra of 0.2 and 1.2 H CaH2-reduced samples (600 °C, 2 day synthesis). (b) 2H MAS NMR spectrum of a 1.2 D CaD2-reduced sample (600 °C, 2 day synthesis). The dotted line highlights the identical location of the maxima of the broad resonances in the 1H and 2H spectra. (c) Close-up of the resonances in the positive parts per million range. The inset shows integrated proton signal intensity with respect to the starting material (BaTiO3). The signal of the 1.2-H sample is deconvoluted into a protic (positive parts per million) and hydridic (negative parts per million) contribution.

As a next step, we attempted to quantify the H content of the BaTiO3 starting material by relating its 1H NMR signal intensity to that of adamantane (C10H16) in the same rotor volume and under identical experimental conditions. The BaTiO3 sample (before TGA) exhibits a total 1H NMR signal integral of 0.94% compared to the adamantane sample. From the densities and molecular weights of adamantane (1.08 g/cm3, 136.23 g/mol) and BaTiO3 (6.02 g/cm3, 233.2 g/mol), a molar ratio of H/BaTiO3 ≈ 0.039 can then be derived. This ratio relates well to the TGA weight loss of the BaTiO3 starting material (0.15%, cf. Figure b): when attributing this weight loss to H2O, one obtains a molar ratio H2O/BaTiO3 ≈ 0.02 (i.e., H/BaTiO3 = 0.04). In order to quantify the H content of the reduced samples, we related their 1H NMR signal integrals to this value. This is demonstrated in the inset of Figure c. In addition, the 1H signals can be deconvoluted into a protic (positive parts per million) and hydridic (negative parts per million) part.

Surprisingly, the total 1H NMR signal integral for 1.2-H-CaH2 is only approximately 5 times larger than that of the 0.2-H-CaH2 sample (which in turn is only slightly higher than that of the starting material). From deconvolution of the total 1H NMR signal integral for 1.2-H-CaH2, one can estimate the molar ratio between hydridic H and BaTiO3 to be approximately 0.16. The large discrepancy with the TGA experiment (cf. Figure ), which delivers an x value of 0.24 with respect to BaTiO3–H, has to be attributed to the simultaneous presence of O vacancies. As a matter of fact, the combined 1H MAS NMR and TGA results suggest that hydridic H and O vacancies are present in roughly equal concentrations. Thus, we reformulate 1.2-H-CaH2 as BaTiO3–H□( with x ≈ 0.24 and y ≈ 0.16 and conclude that 1H MAS NMR spectroscopy represents a reliable and convenient method to analyze H environments and assess H contents in reduced BaTiO3 samples.

The observed negative shift for the hydridic 1H can be attributed to the contact hyperfine interaction between the 3d electrons of Ti(III) and the 1H nucleus. This shift mechanism is operative in both metallic systems and paramagnetic insulators.[29,30] In metallic systems, the electrons are delocalized in a conduction band and the hyperfine interaction results in a Knight shift that depends on the density of states at the Fermi level.[31] The magnitude of the Knight shift is known to increase with both the unit cell volume[32] and the concentration of charge carriers.[33] In paramagnetic insulators, the shift mechanism is dominated by the Fermi-contact hyperfine interaction of 1H with the 3d electron of the directly bonded Ti(III) ion, which gives a Fermi-contact shift.[29] In the latter case, the electronic conductivity is hypothesized to arise from the electron polaron formation between localized states formed in the band gap. Both mechanisms may give negative shifts, if the hyperfine interaction is dominated by a spin polarization mechanism. The identity of the shift mechanism in our reduced BaTiO3 samples is addressed in the following discussion.

The 1H MAS NMR spectra recorded from the series of 0.2 H-reduced samples are shown in Figure . Although surface hydroxyl sites and their respective populations are altered, one recognizes a general similarity with the starting material. More interestingly, the spectra of 0.2-H-MgH2 and 0.2-H-NaAlH4 specimens have an additional resonance at around 0 ppm. As discussed previously, these samples are cubic, whereas the other 0.2 H-reduced materials remained tetragonal. We therefore interpret this additional resonance at around 0 ppm as being from hydridic H. The total 1H signal of the 0.2-H-CaH2 and 0.2-H-NaBH4 samples is similar to that of the starting material (H/BaTiO3 ≈ 0.04), whereas that of the remaining samples is by 70–80% higher. For 0.2-H-NaH, this is clearly due to an increased concentration of surface OH. For 0.2-H-MgH2 and 0.2-H-NaAlH4, the higher 1H concentration is attributed to the hydridic contribution. This contribution, y, can then be estimated to be ∼0.03, which is slightly lower than the x values obtained from TGA, 0.04–0.05 (cf. Table ). The relatively small hydride shifts observed for 0.2-H-MgH2 and 0.2-H-NaAlH4 compared to 1.2-H-CaH2 may then be consistent with a Knight shift for all the three samples: 1.2-H-CaH2 has a considerably higher concentration of hydride and also vacancies and therefore a higher concentration of charge carriers, which in turn results in a more negative Knight shift.

Figure 10

1H MAS NMR spectra of the 0.2-H sample series. The blue dotted line indicates an additional resonance in the 0.2-H-MgH2 and 0.2-H-NaAlH4 spectra at ∼0 ppm, which is attributed to hydridic H. The inset shows integrated 1H signal intensity with respect to the starting material (BaTiO3). The violet color for 0.2-H-MgH2 and 0.2-H-NaAlH4 illustrates a mixture of protic and hydridic H.

Figure shows the 1H MAS NMR spectra recorded for the series of 1.2 H-reduced samples. First, 1.2-H-NaH, which is tetragonal, does not exhibit a hydridic component, as expected given the results discussed previously. However, the spectrum does reveal a high concentration of surface hydroxyl, with an especially intense resonance at around 1 ppm. This supports the conjecture that reduction with NaH primarily affects the surface of the particles: that is, reduction leads to a high concentration of vacancies at the surface, which subsequently terminate/react into hydroxyl during the washing process. The spectrum of 1.2-H-NaBH4 has a distinct signal with a small negative shift of −1.8 ppm, which is assigned to hydridic H. Note that the comparatively low shift is similar to that observed for 0.2-H-MgH2 and 0.2-H-NaAlH4. As already discussed, 1.2-H-CaH2 exhibits a broad resonance with a maximum at around −18 ppm and with a shoulder at around −60 ppm. The 1H NMR signals at negative shifts for the 1.2-H-NaAlH4 and 1.2-H-MgH2 samples are even broader with their peak maxima at around −30 and −60 ppm, respectively. Clearly, among the 1.2-H reduced samples, 1.2-H-NaBH4 is distinguished because of its relatively sharp 1H MAS signal, which indicates a homogeneous coordination environment for H, and low shift. As will be discussed in the next section, this sample contains a low concentration of hydride, y ≈ 0.04, which is comparable to that of 0.2-H-MgH2 and 0.2-H-NaAlH4.

Figure 11

(a) 1H MAS NMR spectra of the 1.2-H sample series. (b) Close-up of the resonances in the positive parts per million range attributed to protic (surface) H species. Note that the close-up also contains the signal from hydridic H for 1.2-H-NaBH4, which is centered at −1.8 ppm.

We now remark on a possible correlation between the shifts and line widths of the hydridic 1H resonances, which broaden according to NaBH4 < CaH2 < NaAlH4 < MgH2 and concurrently shift toward more negative parts per million values. The shoulder at −60 ppm in the spectrum for 1.2-H-CaH2 and a pronounced asymmetry of the negative parts per million signal in the spectrum for 1.2-H-NaAlH4 probably relate to the two-phase nature of these samples. We recall that both samples constituted two cubic phases with lattice parameters ∼4.01 and ∼4.02 Å and one is tempted to correlate the shift to the size of the lattice parameter: 4.005 Å—–2 ppm (NaBH4, single phase); 4.01 Å—–20 to −30 ppm (CaH2, NaAlH4, two-phase); 4.02 Å—–60 ppm (MgH2, single phase). This is corroborated by the 1H MAS NMR spectra from the samples obtained by 1.2 H CaH2 reduction during 1, 2, 4, and 7 days (Supporting Information, Figure S3). The similarity of the 1, 2, and 4 day samples established from PXRD and TGA (cf. Figure ) is also reflected in their virtually identical 1H MAS NMR responses. The spectrum of the 7 day sample, however, reveals a largely asymmetric resonance with the peak maximum at around −50 ppm. Concomitant with the more negative shift of the NMR signal of hydridic H is an increase of the lattice parameter of the majority phase from ∼4.01 to ∼4.02 Å (cf. Table ).

This gradual shifting of the resonance to more negative shifts is consistent with the behavior previously observed for metallic hydrides such as NbH, where an increase in the magnitude of the Knight shift correlates with an increasing unit cell volume.[32] This observation suggests that our reduced 1.2-H samples correspond to metals with delocalized electrons in a conduction band, as opposed to electron polaron formation, leading to localized states in the band gap. Thus, the observed 1H shifts are Knight shifts. Values are consistent with the ranges observed for other metallic hydride materials, which are typically 0 to −250 ppm.[34] As mentioned above, the magnitude of the Knight shift also depends on the density of states at the Fermi level.[31] This in turn should correlate with the charge carrier concentration, and one would also expect that the charge carrier concentration in our samples increases with increasing Knight shift. However, the overall trends are complicated to predict. Kageyama et al. reported measurable electrical conductivities for their materials BaTiO2.4H0.6 and BaTiO2.7H0.3, while at the same time their 1H MAS NMR spectrum indicates the absence of the Knight effect.[10] This may be explained by assuming a different electronic structure and conduction mechanism, that is, thermally activated polaron hopping. However, we note that the 1H MAS NMR spectrum reported by Kageyama et al. showing a single sharp resonance at 4.4 ppm is slightly inconceivable because a more complex spectrum is expected from the simultaneous presence of unavoidable surface hydroxyl and surface water.

Oxyhydride versus O Vacancy Formation

It appears that the hydride reduction of BaTiO3 is strongly influenced by the reducing agent (metal hydride) and that reduced BaTiO3 actually represents complex heterogeneous materials because of the simultaneous presence of vacancies and H in the anion substructure. The overall concentration of hydridic H is comparatively rather low.

Figure presents the results of the 1H MAS NMR spectral deconvolution into positive(protic)- and negative(hydridic)-parts per million shift contributions to the total proton signal of 1.2 H-reduced BaTiO3 samples. It is seen that 1.2-H-MgH2 and 1.2-H-NaAlH4 have a very similar hydridic H content as 1.2-H-CaH2. Surprising is the high H content of the 1.2-H-NaH sample, which, however, refers to protic (surface) hydroxyl, and the low hydridic H content (on the order of 0.04) of 1.2-H-NaBH4. This implies that O vacancies dominate in the NaBH4-reduced sample. Thus, when using the formula BaTiO3–H□( for hydride-reduced BaTiO3, x as estimated from TGA is ∼0.35 (cf. Table ) and y is ∼0.04. Accordingly, the H–/O vacancy ratio is around 1:10. Analogous reasoning for the MgH2- and NaAlH4-reduced samples yields y values similar to the CaH2-reduced sample (0.14–0.15); however, because of the significantly larger x (around 0.4), O vacancy concentrations for these samples exceed the concentration of hydridic H by 2–3 times. The compositions of 1.2-H samples derived from the combined TGA and NMR analysis are compiled in Table .

Figure 12

Relative strengths and shift contributions to the 1H MAS NMR signal integral of the 1.2-H series of reduced BaTiO3 samples.

Table 5

Compositions of 1.2-H Samples from Hydride Reductions during 2 Days at 600 °C

sample	xH from TGa	x□ from TGb	y from NMR	formula BaTiO3–xHyx–y
CaH2	0.24	0.23	0.16	BaTiO3.76H0.16□0.08
MgH2	0.41	0.39	0.14	BaTiO3.60H0.14□0.26
NaAlH4	0.39	0.37	0.15	BaTiO3.62H0.15□0.23
NaBH4	0.36	0.34	0.04	BaTiO3.65H0.04□0.31

xH refers to a reaction BaTiO3–H + 0.75xO2 → BaTiO3 + 0.5xH2O.

x□ refers to a reaction BaTiO3– + 0.5xO2 → BaTiO3.

The formation of O-deficient phases BaTiO3–H□( is surprising and at variance with the results of Kobayashi et al.[10,13] The question arises whether these phases represent intermediates toward stoichiometric oxyhydride formation. Bridge et al. pointed out the significance of gaseous H2 present in the hydride reduction of LaSrCo3O4 for obtaining LaSrCo3H0.7.[18] In addition, perhaps Kobayashi et al. attained during their synthesis of BaTiO3–H—for unknown reasons—a pressurized H2 atmosphere. Following this conjecture, we subjected the CaH2-, MgH2-, and NaAlH4-reduced samples to pressurized H2 atmospheres (30–50 bar) at 600 °C, which is the temperature applied in the preceding hydride reduction. The results are summarized in Table S1 (Supporting Information) and Figure . Interestingly, during annealing under pressurized H2, the two-phase samples essentially became single phase, with a lattice parameter closely corresponding to the previous majority phase (Table S1, cf. Table ). The 1H NMR spectra (Figure ) revealed that the annealing procedure did not lead to an increased hydride content. However, the signal narrowed significantly, which indicates that more ordered phases BaTiO3–H□( were obtained (cf. Figure a and Supporting Information, part V, where spectra of post- and prehydrogenated samples are directly compared). Apparent again is a correlation of the size of the lattice parameter and the chemical shift of the 1H resonance: the larger the parameter, the more negative the shift is. The TGA weight increases are by about 20% lower after the annealing procedure. This coincides with a lower H content of the posthydrogenated samples. The diminished hydridic H content may be due to a slight oxidation of the sample from traces of oxygen in the autoclave setup.

Figure 13

(a) 1H MAS NMR spectra of posthydrogenated samples. (b) Comparison of relative strengths and shift contributions to the 1H MAS NMR signal integral for regular and posthydrogenated 1.2-H samples.

In summary, it remains unclear why our hydride reduction experiments with BaTiO3 resulted in a different product as described by Kobayashi et al. Yet, highly O-deficient BaTiO3–H□( is an interesting material. Reduced BaTiO3–, with applications in electroceramics, electrocatalysis, and electronics, has been described and studied earlier.[35] There is general agreement that the crystal symmetry changes from tetragonal to cubic for small values of x, x < 0.02, before transforming to the hexagonal 6H-perovskite for x > 0.02.[36,37] The hexagonal form can maintain a maximum x of about 0.15.[38] Characteristic for all forms O-deficient BaTiO3– is their blue/black color. Cubic BaTiO3– is synthesized at temperatures 900–1250 °C in a flowing mixture of 5% H2/95% N2. Reduction at higher temperatures results in the hexagonal form.[37] Mixtures of tetragonal and cubic BaTiO3– are also obtained when sintering BaTiO3 in the reducing environment of an SPS apparatus at 1200 °C,[39,40] and the hexagonal form can be synthesized when heating tetragonal BaTiO3 in a graphite-vacuum furnace to 1300–1500 °C.[38] Against this background, the existence of cubic phases BaTiO3–H□( with x values exceeding 0.5 is surprising. Clearly, the incorporation of H plays an important role in stabilizing these highly O-deficient variants of BaTiO3, which in contrast to previously reported forms are accessible at comparatively low temperatures (∼600 °C). The two-phase behavior observed for most reductions may correlate with a segregation of BaTiO3–H□( into a vacancy and hydride-rich composition. Interestingly, our defect samples show a similar reactivity toward the formation of mixed-anion derivatives as that reported for BaTiO2.5H0.5.[12] For example, BaTiO3–H□( can be easily converted into green-colored nitridized BaTiO3 (BaTiO3−δN) under a stream of N2 gas at temperatures above 450 °C.

Conclusions

Following previous reports on the reaction of tetragonal BaTiO3 with CaH2 yielding cubic oxyhydrides BaTiO3–H,[10,13] we investigated hydride reduction of BaTiO3 with various metal hydrides as reducing agents. At the applied temperature of 600 °C, we find that NaH acts only as a weak reducing agent. A conversion into a cubic product is not obtained. This is attributed to the low decomposition temperature of NaH, instead of making H– gaseous H2 the reducing species. Apart from the originally employed CaH2, also MgH2 and the ternary hydrides NaBH4 and NaAlH4 reduce BaTiO3 to a blue-/black-colored cubic product. However, at variance with the original reports, we find that only a small concentration of H replaces O in the anionic substructure of BaTiO3. Instead, highly O-deficient disordered cubic phases BaTiO3–H□( with x up to 0.6 and y in a range 0.04–0.25 are obtained. The existence of such phases is surprising, and the incorporation of H supposedly plays an important role in stabilizing these highly O-deficient variants of BaTiO3. These defective samples show high reactivity toward the formation of mixed-anion derivatives, such as oxynitrides. Finally, we emphasize the important role of NMR spectroscopy for characterizing both H species and H contents in reduced BaTiO3 samples. The negative shift of hydridic H is attributed to the metallic nature of the phases BaTiO3–H□(.