I. E. Grey*a,
P. Bordetb and
N. C. Wilsona
aCSIRO Mineral Resources, Private Bag 10, Clayton South, Victoria 3169, Australia. E-mail: ian.grey@csiro.au
bUniversité Grenoble Alpes, CNRS, Institut Néel, Grenoble, 38000, France
First published on 24th February 2021
Amorphous titania samples prepared by ammonia solution neutralization of titanyl sulphate have been characterized by chemical and thermal analyses, and with reciprocal-space and real-space fitting of wide-angle synchrotron X-ray scattering data. A model that fits both the chemical and structural data comprises small segments of lepidocrocite-type layer that are offset by corner-sharing as in the monoclinic titanic acids H2TinO2n+1·mH2O. The amorphous phase composition that best fits the combined chemical and scattering data is [(NH4)3H21Ti20O52]·14H2O, where the formula within the brackets is the cluster composition and the H2O outside the brackets is physically adsorbed. The NH4+ cations are an integral part of the clusters and are bonded to layer anions at the corners of the offset layers, as occurs in the alkali metal stepped-layer titanates. The stepped-layer model is shown to give a consistent mechanism for the reaction of aqueous ammonia with solid hydrated titanyl sulphate, in which the amorphous product retains the exact size and shape of the reacting titanyl sulphate crystals.
The important role of amorphous titania in photocatalysis applications has been highlighted in a recent review,10 which also noted the comparative lack of published information on the structure and properties of amorphous titania relative to the common nanocrystalline forms, anatase, brookite and rutile. Published structure analyses using the Fourier transform of the total powder diffraction pattern (pair distribution function, PDF) coupled with reverse Monte Carlo (RMC) simulations have variously described the amorphous phase structure as resembling that of brookite,11,12 lepidocrocite13 or anatase.14 Molecular dynamics (MD) and Monte Carlo (MC) simulations have also been used to generate amorphous titania particles with particle sizes from 1 to 10 nm and containing up to 3000 atoms.15–19 The MD and MC studies do not identify particular known structure types that occur locally within the particles, but instead report partial radial distribution functions (Ti–Ti, Ti–O and O–O) together with coordination number and bond angle distributions. These are broadly consistent with a network of short staggered chains of edge- and corner-shared polyhedra, of which TiO6 octahedra dominate (∼70%) together with minor TiO5 and TiO7 polyhedra.
The nature of the amorphous titania precursor phase is of importance in helping to understand how non-metallic dopants are incorporated into nanocrystalline titania. Nitrogen is the most effective non-metallic dopant for enhancing photocatalytic activity of TiO2, but there is still considerable uncertainty about the nature of the N complex species that contribute to visible-light absorption of the doped titania.3 In their review of N-doped TiO2 photocatalysts these authors3 have noted that the formation of the N complex species most likely depends on the synthesis procedure used. Understanding how the N species are incorporated into the amorphous precursor phase is an important step.
In published studies on syntheses of nanocrystalline TiO2 photocatalysts via an amorphous phase, the relationships between the amorphous phase precursor and the crystalline phase-type formed from it have often been hampered by a lack of chemical characterization of the precursor phase. Most studies on amorphous titania do not provide chemical or thermal analyses on the samples that are being characterized by diffraction studies. A notable exception is a recent study on the characterization of amorphous titania formed by reaction of ammonia solution with solid titanyl sulphates20 for which chemical analyses, thermogravimetric and differential thermal analysis (TGA/DTA), and infrared (IR) and Raman spectroscopy results were given. These authors20 reported the interesting result that reaction of aqueous ammonia solution with solid titanyl sulphate crystals at 0 °C results in complete extraction of sulphate ions and their replacement with hydroxyl groups, forming amorphous products that preserve the size and shape of the original crystals. The chemical analyses showed that a significant amount of nitrogen, as NH4+, is bound in the amorphous phase.
In this study we report the structural characterization of the amorphous N-doped precursor phase formed during neutralization of titanyl sulphate with ammonia solution. Synchrotron wide-angle X-ray scattering data was used, with checking of models using the Debye function to fit the powder X-ray diffraction (PXRD) data directly and using the PDF to screen different structural models. The diffraction data was supported by chemical and thermal analyses to help determine how the N was incorporated into the amorphous phase.
The TGA/DTA/MS results are summarised in Table 2. When heated in helium the DTA shows an endotherm centred at 145 °C corresponding to water loss and an exothermic peak at 417 °C corresponding to crystallization of anatase. When the TGA experiment was conducted in a flow of oxygen, a second exothermic peak was observed in the DTA trace, at 288 °C. From MS on the evolved gases it was confirmed that this exotherm corresponded to the oxidation of the ammonium ion to nitrogen oxides. The TGA/DTA/MS plots are reported as ESI data, Fig. S1(a)–(d).†
Helium run | Oxygen run | |
---|---|---|
DTA peaks (°C) | ||
Endotherm | 145 | 170 |
Exotherm | 417 | 288 and 400 |
TGA mass losses (%) | ||
On evacuation | 11.9 | 13.0 |
20 to 130 °C | 3.2 | 1.9 |
130 to 1000 °C | 11.2 | 12.3 |
Evolved gases temp. (°) | ||
Water | 40–600 | 40–600 |
NH3 | 100–600 | 100–280 |
CO2 | 40–200 | 40–200 |
NOx | — | 200–450 |
Thermal analyses conducted by Palkovská et al.24 on the Klementova sample D at different heating rates showed that at the slowest heating rate, of 2 °C min−1, the water release could be resolved into two steps, with losses below 130 °C corresponding to physically adsorbed waster and losses above 130 °C corresponding to more strongly bonded water. Our results are consistent with this observation, with 80% or more of the mass loss of our sample at temperatures below 130 °C occurring during evacuation of the sample. We corrected the mass loss at greater than 130 °C for loss of CO2 and NH3 to obtain the structurally bound water/hydroxyl content and combined this with the chemical analyses in Table 1 to obtain a formula for the amorphous phase. The derived formula is [(NH4)0.16H1.06TiO2.61]·(H2O)0.70(CO2)0.025, where the components outside the square brackets are physically adsorbed species. It is worth noting that the precursor phase is not a crystalline phase with a well-established periodic lattice, and so variations in the formula are expected both within the hydrolysis product and between products prepared in separate syntheses.
The formula inside the square brackets indicates that the titanate structure has a small negative charge, balanced by NH4+ cations. The titanate formula is close to H2Ti2O5, a member of the series of layered polytitanic acids, H2TinO2n+1.25 The titanate formula can also be written as ∼ TiO2·0.5H2O. It is interesting to note that Barringer and Bowen,26 who prepared amorphous titania subjected to PDF analysis by Wang11 gave the same formula TiO2·0.5H2O to account for the structurally bound water content.
Fig. 1 (a) Synchrotron PXRD pattern for amorphous precursor phase (λ = 0.4430 Å). d-Spacings shown. (b) Atomic PDF function for amorphous phase. d-Spacings are shown. |
The positions and relative intensities of the PDF peaks for our amorphous titania correspond closely to those reported by Gateshki,13 for amorphous titania prepared by reaction of hexamethylenetetramine with TiCl3 solution at 90 °C. These authors compared their experimental PDF with calculated PDFs for numerous phases; specifically rutile, brookite, anatase, two high-pressure forms of TiO2 (α-PbO2-type and baddeleyite-type structures), the reduced titanates Ti3O5 and Ti4O7, the titanic acid H2Ti3O7 and lepidocrocite (using the CsxTi2−x/4O4 structure28). They noted that the experimental PDF was sensitive enough to discriminate between the different models tested and they concluded that the one based on the lepidocrocite layer structure gave the best approximation to the precursor phase structure.
Protonated Ti-lepidocrocite, HxTi2−x/4O4, has a layer structure built from edge- and corner-shared octahedra as shown in Fig. 2(a), with a separation between the layers of ∼9 Å.29 Gateshki13 proposed that the precursor phase was composed of randomly orientated short segments of layers. They did not, however, check the validity of their model directly against the PXRD pattern.
Fig. 2 (a) 24-Ti atom segment of lepidocrocite structure used for Debye-function modelling of PXRD. (b) The segment rotated to show cubic-stacking of (111)rocksalt-like layers. |
Our fit to the experimental PDF using the lepidocrocite model is shown in Fig. 3(a).
Fig. 3 (a) Experimental PDF (blue points) and calculated PDF (red line) for the lepidocrocite structure. (b) Amorphous titania PXRD (blue points) and Debye-function calculated pattern (red line) based on the lepidocrocite-type cluster shown in Fig. 2(a). |
The fitting parameters that were refined were the scale, unit cell parameters and overall isotropic displacement parameters for Ti and O. Visually the match closely resembles that previously reported13 (their Fig. 5). In Fig. 3(b) the PXRD calculated using the Debye equation is compared with the experimental pattern. For the calculated pattern the segment of a lepidocrocite layer (24 Ti atoms) shown in Fig. 2(a) was used. The agreement is only fair; the position of the first calculated peak is displaced to a higher d-spacing of 3.35 Å compared to experiment (3.1 Å), and there is a relatively strong calculated peak at d = 2.35 Å that is not present in the experimental pattern. This peak can be explained if we consider that the lepidocrocite layer corresponds to a 2-octahedra-wide (110) slice of the face-centred cubic (fcc) rocksalt structure and the peak at d = 2.35 Å corresponds to diffraction from the cubic-stacked layers (=(111)rocksalt). In Fig. 2(b) the lepidocrocite layer shown in Fig. 2(a) has been rotated to show the cubic-stacked layers. We used the Debye equation to generate PXRD patterns for several lepidocrocite segments of varying size and shape but in all cases the 2.35 Å peak was a major feature and we conclude that lepidocrocite is not a good model for our amorphous titania phase. Debye function plots for different lepidocrocite segment sizes are given as ESI data, Fig. S2(a)–(d).†
In searching for alternative structural models to explain the diffraction data for amorphous titania we considered the possibility of titanium-oxo clusters. A variety of such clusters exist in solution depending on the pH and associated solution species and many clusters have been induced to crystallize and have had their structures determined.30,31 As described by Zhang,32 titanium-oxo clusters are “the trapped snapshots of intermediate structures in the sol–gel growth of TiO2 nanocrystals”. To date, however, there has been no direct confirmation that titanium-oxo clusters exist in amorphous precursor phases. For example Kozma33 conducted a laboratory study of the industrial sulphate process for titania pigment production and demonstrated that hydrated titanium oxysulphate clusters containing 18 Ti atoms per cluster exist in the strong sulphuric acid solution and can be crystallised, but when the solution was diluted and aged under ambient conditions, 2 nm particles of anatase precipitated without any indication of an amorphous phase precursor. On the other hand, Gautier-Luneau34 crystallized a hexanuclear titanium acetate complex by evaporation of an ethanolic solution of titanium tetraethoxide and acetic acid and noted that if the evaporation rate was increased an amorphous gel formed and a radial distribution function plot for the gel showed the same interatomic distances as for the crystalline cluster, leading them to postulate that the hexamer cluster exists in the gel.
We checked numerous published structures of titanium-oxo clusters for Ti–Ti distances consistent with the amorphous titania PDF and selected four clusters for PDF calculations, corresponding to widely different sizes and shapes of the clusters. The four clusters are shown in Fig. S3 of the ESI data.† Ti18O45 is a Keggin complex, Ti13O40, capped by five TiO5 square antiprisms;35 Ti12O32 has a hexameric belt of edge-shared square antiprisms, capped above and below by Ti3O13 edge-shared trimers;36 Ti6O28 is the Galy-cluster comprising two trimeric units of edge-and corner-connected octahedra34 and Ti15O30 is an optimised cluster obtained by simulated annealing and MC simulations.37 PDF calculations showed that none of the as-published cluster models fitted the experimental PDF as well as the lepidocrocite model. A cautionary lesson from the calculations was that unconstrained PDF refinements in which the Ti and O atom positions were allowed to move from the published values gave very good fits to the experimental PDF but resulted in unrealistically short Ti–O and O–O distances. Separate models as disparate as the planar Ti6O28 cluster and the spherical hedgehog-like Ti18O45 cluster gave equally good fits, emphasising the need to have careful control and appropriate constraints during PDF refinements. For the calculated PDFs reported in this study, the atoms were kept at the positions given for the published structures.
With the lepidocrocite model giving the closest match to the PDF of the models considered so far, we next considered derivates of this structure type. A family of related structures that has received considerable attention as precursors for controlled formation of different titania polymorphs and different nano-morphologies (nanotubes, nanobelts, nanowires) are the layered titanic acids of general formula H2TinO2n+1·mH2O, n = 2 to 5.4,25,38,39 The n = 2 member has an orthorhombic lepidocrocite-type structure whereas the other members have monoclinic symmetry and stepped-layer structures that comprise lepidocrocite strips that are n-octahedra wide and offset by corner-sharing. The series has been extended by monoclinic members with n = 6 (ref. 38) and n = 8 (ref. 40) in which 3- and 4-octahedra wide stepped layers are fused by corner-sharing. We tested the possibility of stepped layers being present in our amorphous titania by generating PDFs and Debye-function PXRDs.
The PDF for the member with n = 3 was calculated using the structural model reported for D2Ti3O7,41 shown in Fig. 4(a). The comparison with the experimental PDF after refining the same parameters used for the lepidocrocite-type PDF refinement is shown in Fig. 5(a).
Fig. 4 (a) [010] projection of the structure for H2Ti3O7. (b) Cluster of 20 Ti-centred octahedra from H2Ti3O7 stepped-layer used for Debye-function calculation of the PXRD pattern. |
It represents a significant improvement on the fit obtained with lepidocrocite-type structure (Fig. 3(a)), with the reduced chi-squared for the fit decreased from 0.031 to 0.019. A 20 Ti atom portion of the stepped layer was cut from the structure, Fig. 4(b), and used in a Debye-function calculation of the PXRD. The calculated pattern is compared to the experimental pattern in Fig. 5(b). There is a close match between the two, not only for the three strongest peaks to d = 1.5 Å, but also for the weaker peaks at higher scattering angle, suggesting that small segments of stepped-lepidocrocite layers occur in our amorphous titania sample.
We checked the effect of changing the size of the cluster on the calculated PXRD pattern, by increasing the cluster size from 20 to 30 and 40 Ti atoms per cluster. The clusters and the corresponding Debye-calculated PXRD patterns are given in the ESI data as Fig. S4.† The calculated PXRDs show increasing structure in the peaks relative to the experimental pattern and relative to the calculated pattern for the 20 Ti atom cluster, consistent with the interatomic Ti–Ti distances in these clusters extending beyond the ∼1 nm correlation length evident in the experimental PDF. The calculations suggest that the optimum size of aggregates in the amorphous phase corresponds to 20 Ti atoms clusters.
Fig. 6 (a) Location of A cations in monoclinic A2TinO2n+1. (b) Bonding of A1 cations (NH4+) to 20-Ti atom stepped layer. |
Based on these studies the NH4+ cations in our amorphous titania sample are considered to be located at the A1 sites and coordinated to oxygen atoms of octahedra involved in corner-linking of the lepidocrocite segments. The bonding of the cations in the 20 Ti atom stepped layer cluster is shown in Fig. 6(b). The ratio of available A1 sites to Ti atoms is 3 to 20 = 0.15, which is close to the value of 0.16 obtained from the chemical analyses. The Debye-calculated PXRD pattern for the 20 Ti atom cluster with 3 NH4+ cations incorporated as in Fig. 6(b) is shown as the green line in Fig. 5(b). It gives a good match with the experimental PXRD and a PDF calculated with NH4+ added resulted in a small decrease in chi-squared (to 0.017) relative to the model without NH4+ (0.019).
The composition of the 20 Ti atom cluster, Fig. 4(b), with all Ti atoms octahedrally coordinated, is Ti20X63, X = anions O, OH, H2O. For comparison, chemical and thermal analyses give a formula [(NH4)0.16H1.06TiO2.61]·(H2O)0.70(CO2)0.025, where physically adsorbed species are shown outside the square brackets. Scaling the formula within the square brackets to 20 Ti and rounding to integers gives (NH4)3H21Ti20O52. To be compatible with this formula the 20 Ti atom cluster requires 11 fewer anions. This corresponds to 55% of the Ti atoms being 5-coordinated, giving a mean coordination number for Ti of 5.45.
Comparing a 30 Ti atom cluster, Ti30X87, to the scaled chemical formula of (NH4)5H31Ti30O78, indicates that this structural cluster requires 9 fewer anions, equivalent to 30% of the Ti atoms being 5-coordinated and giving a mean coordination number for Ti of 5.70. For comparison, Petkov12 reports a Ti–O first coordination number of 5.6 and Zhang14 reports a mean coordination number for Ti of 5.3 for their sol–gel derived amorphous titania samples. The 20-atom cluster, which gives the best fit to our scattering data, is in good agreement with both reported values. Accordingly, we can describe the amorphous titania prepared by ammonia neutralization of titanyl sulphate as being composed of small clusters of stepped-layer structure, with the average cluster size corresponding to a composition (NH4)3H21Ti20O52. The NH4+ cations are coordinated to the clusters at the corner-linkages between offset lepidocrocite-type segments. Aggregates of individual clusters are most likely held together through H-bonding and ∼14 molecules of H2O per cluster are physically adsorbed within the pore space of the nanoparticles.
In Fig. 7(a) is shown the structure of titanyl sulphate monohydrate. It has orthorhombic symmetry, with a = 9.818, b = 5.133, c = 8.614 Å.46 The structure comprises crankshaft-shaped ribbons of cis-corner-connected octahedra along [010] which are interlinked along [100] and [001] by corner-sharing with sulphate tetrahedra. The octahedral corner-linkages involve alternating very short (1.70 Å) and long (2.05 Å) Ti–O bonds and so the structure can be described in terms of chains of titanyl groups, [..OTi–O..]n decorated with SO4 and H2O, running parallel to [010]. When TiOSO4·H2O is reacted with ammonia solution the sulphate groups are completely extracted giving TiO2·H2O with loss of SO3. Fig. 7(b) shows one of the resulting titanyl chains, in which each TiO5(H2O) octahedron has been altered to four-coordinated TiO3(H2O). Ti4+, a d0 transition element cation, is most stable when octahedrally coordinated. The octahedral coordination can be restored by folding of the polymeric chain so that the anions can be shared by two or three Ti atoms. In this folding the OTi–O backbone is retained, but the resulting shrinkage associated with the folding will result in the long chains being broken into shorter length chains. Fig. 7(c) shows an example of the folding of a 6-member chain to give a cluster of edge- and corner-shared octahedra. The location of the corner-sharing is expected to be influenced by the proximity of NH4+ cations. We have confirmed, using polyhedral models, that the titanyl backbone remains intact in the folding operation and that this is the structure-directing element.
In Fig. 8(a) the titanyl chains are viewed end-on along [010] after extraction of the sulphate groups. With the removal of interconnecting sulphate groups the chains are free to rotate and displace as shown by the arrow in Fig. 8(b). Condensation of adjacent chain segments, with loss of H2O gives the lepidocrocite-layer segment shown in Fig. 8(c). Thus the combination of sulphate extraction, folding and rupture of the titanyl chains and condensation of adjacent chains applied to the reaction of solid titanyl sulphate with aqueous ammonia, leads to a model for the amorphous phase reported by Klementová20 that matches the model we derived from X-ray scattering experiments on amorphous titania prepared by neutralization of titanyl sulphate solution with ammonia.
The stepped-layer model for the amorphous phase is consistent with a mechanism that we developed to explain the reaction of ammonia solution with solid titanyl sulphate hydrate as reported by Klementová.20 The ammonia solution completely extracts the sulphate groups, leaving titanyl chains. A model involving rupture, folding and condensation of the leached titanyl chains leads directly to the formation of small segments of stepped-layer structure. The operations involve only small local rotations and displacements of the chains and are thus consistent with the observation that the amorphous product retains the size and shape of the original titanyl sulphate crystals.
The stepped-layer model is different from a brookite-based model used by Petkov12 to describe the structure of amorphous titania prepared by controlled hydrolysis and condensation of tetra-iso-propoxytitanate in ethanol solution, and from a model based on a strained anatase core developed by Zhang14 to describe amorphous titania obtained by hydrolysis of titanium tetraethoxide in water at 0 °C with acetic acid added. It should be noted however that the PDFs presented by both groups differ significantly from our PDF in having lower ratios of the 3.0 Å peak (Ti–Ti edge-sharing) to the 3.6 Å peak (Ti–Ti corner-sharing). Evidently the local structure of amorphous titania precursor phase can differ depending on the synthesis conditions and this difference can result in different titania polymorphs being formed on crystallization of the precursor.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0ra08886b |
This journal is © The Royal Society of Chemistry 2021 |