Formation and growth mechanism for niobium oxide nanoparticles: atomistic insight from in situ X-ray total scattering

Understanding the mechanisms for nanoparticle nucleation and growth is crucial for the development of tailormade nanomaterials. Here, we use X-ray total scattering and Pair Distribution Function analysis to follow the formation and growth of niobium oxide nanoparticles. We study the solvothermal synthesis from niobium chloride in benzyl alcohol, and through investigations of the influence of reaction temperature, a formation pathway can be suggested. Upon dissolution of niobium chloride in benzyl alcohol, octahedral [NbCl6−xOx] complexes form through exchange of chloride ligands. Heating of the solution results in polymerization, where larger clusters built from multiple edge-sharing [NbCl6−xOx] octahedra assemble. This leads to the formation of a nucleation cluster with the ReO3 type structure, which grows to form nanoparticles of the Wadsley–Roth type H-Nb2O5 structure, which in the bulk phase usually only forms at high temperature. Upon further growth, structural defects appear, and the presence of shear-planes in the structure appears highly dependent on nanoparticle size.


SUPPORTING INFORMATION for
Formation and growth mechanism for niobium oxide nanoparticles: Atomistic insight from in situ X-ray total scattering Olivia Aalling-Frederiksen, Mikkel Juelsholt, Andy S. Anker, and Kirsten M. Ø. Jensen* Department of Chemistry and Nano-Science Center, University of Copenhagen *kirsten@chem.ku.dk

Background scattering signal and subtraction
The in situ experiments were initiated upon applied heating. Figure S1 shows how the scattering signal from the solvent used in the experiment changes upon heating. The rapid change in the scattering pattern of the benzyl alcohol followed by a stabilization reflect that after 10 s of heating, the temperature stabilizes.

Fig. S1
Heating of benzyl alcohol to 300 °C. Electronic Supplementary Material (ESI) for Nanoscale. This journal is © The Royal Society of Chemistry 2021 Figure S2 shows the scattering pattern obtained from the reaction solution (benzyl alcohol and NbCl5) together with the background, i.e. the scattering pattern measured from the pure solvent (benzyl alcohol) in the fused silica capillary at the appropriate temperature and pressure. The scattering patterns measured from the background and precursor were first normalized to have the same intensity at the Q-max value of 14.5 Å -1 , after which a background scaling factor was determined by identifying the scaling value, where no peaks from the solvent and glass could be identified in the final PDF after background subtraction and Fourier transformation. Figure S2 shows the signal and background data obtained for frames at room temperature and after heating to 300 °C along with the final background subtracted data.

Fig. S2
Background subtraction for the room temperature data and the data collected after 24 min of heating at 300 °C.

Fig. S4
Scattering pattern and PDF collected after 24 min of reaction at 300 °C compared with calculated scattering patterns of selected niobium oxide structures. It shows that many of the niobium oxides have the same structural motif, however, none of the reported structures can fully describe the data collected.

PDF refinements
One phase sequential refinement, H-Nb2O5 model Table S2-S4 show the refinement values for datasets collected at 300 °C, 200 °C and 160 °C using the H-Nb2O5 as the structural starting model. The first frame in the sequential refinement (1 min of reaction) and the last frame were refined using all data points in the PDF. The sequential refinement of the rest of the frames was performed on the Nyquist data sampling. The scale factor, lattice parameters a, b, c and β, Uiso for oxygen and the sp-diameter were refined in the sequential refinement. The atomic positions of oxygen were kept fixed at the initial values obtained from the cif-file. Uiso for niobium was furthermore fixed to a value of 0.005 Å 2 . d2 was kept at a value of 3 Å 2 which was determined from a refinement including the very low r-range. Fixed parameters were chosen in order to be able to perform a stabilized sequential refinement and keep parameters at physically reasonable values. The atomic positions of niobium were refined for the first frame in the sequential refinement (1 min of reaction) but kept fixed throughout the sequential refinement.

Table S2
Refined values for the data collected 1 min and 24 min into the reaction at 300 °C using the H-Nb2O5 model (Fits shown in Figure 3A + B).  Figure 6D).   Table S4 Refined values for the data collected 1 min and 13 min into the reaction at 1600 °C using the H-Nb2O5 model (Fit shown in Figure 6C).

One-phase sequential refinement, Nb12O29 and Nb22O54 model
Sequential refinements using other niobium oxides show how other structures with the same local motifs give highly similar results compared to the refinements using the H-Nb2O5 model. In the refinement using Nb12O29 as the starting model a scale factor, lattice parameters a, b, c and β, Uiso for oxygen and niobium and the sp-diameter were refined in the sequential refinement. The niobium positions were refined for the first frame (1 min of reaction) and kept fixed in the sequential refinement.
In the refinement using Nb22O29 as the structural starting model a scale factor, lattice parameters a, b, c and β, Uiso for oxygen and niobium and the sp-diameter were refined in the sequential refinement.

Fig. S8
Sequential refinement of data collected at 300 °C using Nb12O29 and Nb22O54 respectively as the structure model. The changes in both Rw and sp-diameter with time look highly similar to those observed in the refinement using the H-Nb2O5 model Table S5 Refined values for data collected 1 min and 24 min of reaction at 300 °C using Nb12O29 as the structural starting model ( Figure S8).

Table S6
Refined values for data collected 1 min and 24 min of reaction at 300 °C using Nb22O54 as the structural starting model ( Figure S8).

Nb22O54
Initial  Table S7 Refined values for data collected 24 min into the reaction at 300 °C using ReO3 as the structural model ( Figure  5A).

Table S8
Refined values with data collected 24 min into the reaction at 300 °C using ReO3 and R-Nb2O5 as the structural models ( Figure 5B).

Two-phase sequential refinement with R-Nb2O5 and H-Nb2O5
The data collected at 300 °C were refined using a two-phase refinement with H-Nb2O5 and R-Nb2O5. Both Uiso for niobium and oxygen along with the sp-diameter were constrained to the same values for both phases throughout the sequential refinement.

Fig. S9
Selected fits from the sequential refinement of the data collected at 300 °C using H-and R-Nb2O5 as the structural starting models.

Fig. S10
Lattice parameters as a function of time from the two-phase refinement of the data collected at 300 °C using Hand R-Nb2O5 as models.

Fig. S11
Uiso for niobium and oxygen as a function of time from the two-phase refinement of the data collected at 300 °C using H-and R-Nb2O5 as models.

One-phase sequential refinement of 160 °C and 200 °C with H-Nb2O5
Fig. S12 From sequential refinement of the data collected at 160 °C and 200 °C with H-Nb2O5 we observe no significant changes in the scale factors.

Cluster refinement of precursor PDFs
For the refinement of the precursor PDFs we apply Diffpy-CMI 2 and perform a least-square optimization between a theoretical calculated PDF (calculated with the Debye equation) and the experimental PDF. The cluster models used for the calculated PDF were all obtained as cutouts from the Cs2[Nb3O5Cl7] crystal structure shown in Figure S12. Clusters with different sizes and different Cl/O ratios were extracted in order to get the best structural starting models for the refinements. Figures S12 also shows such a cut-out. In all refinements, the scale factor and Nb atomic positions were refined. The isotropic ADP values of Cl, Nb and O were all fixed to 0.03 Å 2 . For the refinement with single octahedra, a wavefunction was implemented in order to describe the solvent-cluster interaction. 3 Figure S13 shows the refinement of the precursor data for the 200 o C experiment both with a without implementing the wave function.   Table S10 Refined values of the precursor collected for the experiment conducted at 300 °C ( Figure 7B). Here, a chain of four [NbCl6-xOx] was used in the structural refinement.

Data analysis of experiment conducted at 100 °C with ambient pressure
In order to slow down the reaction an experiment at 100 °C and ambient pressure was conducted.

Table S13
Refined values for data collected after 13 min of reaction at 100 °C using ReO3 as the structural starting model ( Figure 8D).

Table S14
Refined values for data collected after 13 min of reaction at 100 °C using HNbO3 as the structural starting model ( Figure S15).

Sp-diameter (Å)
16.50 Table S15 Refined values for data collected after 13 min of reaction at 100 °C using H-Nb2O5 as the structural starting model ( Figure S16 A-C).

Ex situ experiment: Synthesis of niobium oxide nanoparticles ambient pressure and 100 °C
A synthesis similar to the in situ experiment was done in our home laboratory to produce particles for TEM and SAXS characterization, as well as ex situ PDF analysis.
Total scattering data of the as-prepared solution was measured for 50 hours on a Panalytical Empyrean Series 3 with an Ag-source (wavelength of 0.56 Å), equipped with a Galipix detector. Background scattering was collected in the same way with a capillary filled with pure benzyl alcohol. The total scattering data was Fourier Transformed to obtain the PDF using PDFgetX3. 4 The following parameters were used for the data reduction: : Qmin=1.3 Å -1 , Qmax=13.5 Å -1 , Qmaxinst=18.4 Å -1 and rpoly=0.9. TEM images were collected on a Tecnai T20 G2 200 kV TEM at the National Center for Micro-and Nanofabrication at the Technical University of Denmark. SAXS and WAXS data were measured on a SAXSLab instrument (JJ-X-ray, Denmark) at the Niels Bohr Institute, University of Copenhagen. The instrument is equipped with a 100XL + microfocus sealed X-ray tube from Rigaku that produces a photon beam with a wavelength of 1.54 Å and a 2D 300 K Pilatus detector from Dectris. The 2D scattering patterns were azimuthally averaged, normalised for sample transmission, primary beam intensity and exposure time and corrected for detector inhomogeneities using Saxsgui. Scattering patterns from an empty quartz capillary and a capillary with benzyl alcohol were measured as backgrounds. The background subtraction was done in a Python script. The scattering angles 2θ were converted to a Q-scale: Simulated SAXS formfactor models were calculated in Diffpy-CMI, which uses SASVIEW functions, with a monodisperse sphere model.

Fig. S18
Top: Total scattering data collected after 13 min of reaction at 100 °C and ambient pressure at the Beamline P.01. DESY (in-situ) compared an experiment performed in our home laboratory. Middle: PDFs obtained for the two datasets. Bottom: Structural refinement on PDF collected for the ex situ experiment using ReO3 as the structural starting model.

Table S16
Refined values obtained for data collected on Panalytical Empyrean using ReO3 as the structural starting model ( Figure S17).

Fig. S20
SAXS data obtained for the as-synthesized niobium oxide nanoparticles. A calculated SAXS form factor assuming a spherical shape and a particle radius of 0.8 nm gives a good description of the data in the high Q-regime of the data. A power law can describe the aggregation of particles into a two-dimensional network, seen in the low Qregime.