In Situ Microfocus Chemical Computed Tomography of the Composition of a Single Catalyst Particle During Hydrogenation of Nitrobenzene in the Liquid Phase

Heterogeneous catalysis performed in the liquid phase is an important type of catalytic process which is rarely studied in situ. Using microfocus X-ray fluorescence and X-ray diffraction computed tomography (μ-XRF-CT, μ-XRD-CT) in combination with X-ray absorption near-edge spectroscopy (XANES), we have determined the active state of a Mo-promoted Pt/C catalyst (NanoSelect) for the liquid-phase hydrogenation of nitrobenzene under standard operating conditions. First, μ-XRF-CT and μ-XRD-CT reveal the active state of Pt catalyst to be reduced, noncrystalline, and evenly dispersed across the support surface. Second, imaging of the Pt and Mo distribution reveals they are highly stable on the support and not prone to leaching during the reaction. This study demonstrates the ability of chemical computed tomography to image the nature and spatial distribution of catalysts under reaction conditions.

XRF mapping of particles in situ (SI Fig 1) and during operation (Fig 1) reveals more inhomogeneity in the element distributions, in particular with Mo, than in previously mapped particles [1] . The Mo signal from the core of the particle measured in situ is far stronger than that for the particle measured under operating conditions, and also far higher than that measured for the particles ex situ [1] , indicating there is a greater degree of variation in the concentration of Mo deposited than for Pt. It may be that in this instance the lower surface coverage of the Pt colloid allowed for a far greater amount of Mo to reach the inner pore structure of the carbon. The greater Mo variation observed may be a result of the doubled loadings of catalyst and promoter, chosen to improve the signal to noise ratio.
The intensities of each elemental signal in the XRF map (Fig 1) are higher than those in the reconstructed sinograms (Fig 2 and SI Fig 2). However, the intensities of each pixel of the XRF map are the sum of all the counts through the particle at the given projection whilst the sinograms are spatially resolved. The intensities of each element vary slightly between sinograms, as can be seen from the map (Fig 1), this is the result of small variations in the deposited concentration throughout the support.  ; experimental data (black), fit (red), residual (blue) and Pt/C nanoparticle reference (green). The R-factor of the linear combination fit in Athena is 0.0016 (measured between 0 and 1 where 0 equates to a perfect fit and 1 failure to fit), and the majority of the residual is attributed to experimental noise. Every other reference spectra fitted with a contribution of less than 1% with an error at least double the contribution. The Mo XANES under operating conditions differs from that ex situ (which corresponds to the heptamolybdate complex), and is likely a result of both coordination environment and oxidation state changing (SI Fig 5). The spectral shape over the absorption edge changes, and it is likely that this is a reduction of the molybdate [2] from [Mo 7 O 24 ] 6to Mo 4 O 11 . The switch from all octahedral sites to a combination of tetrahedral and octahedral sites in Mo 4 O 11 would explain both the reduction in the intensity of the pre-edge feature as well as the increase in the rising edge intensity [3] .
The molybdate may have been partially reduced by the hydrogen in the reaction, although the temperature of the reactor is far lower than would normally be used and as such partial reduction of the sample by the X-ray beam must be considered. This was observed at the Pt L 3 edge for the sample ex situ [1] , and a small degree of beam interaction is also observed under operating conditions (SI Fig  6). The very low volume of ethanol caused complications during the data collection as localised heating occurred due to high flux of the X-ray beam, leading to micro-bubbles forming in the capillary and necessitating reducing the exposure time of the solution to the X-ray beam. Whilst the movement of these bubbles within the capillary may have helped to circulate the reaction solution (since stirring was not possible), the extent of this is likely to have been limited; thus it is unlikely that the final solution in the capillary was thoroughly mixed, and therefore quantification of the turnover of nitrobenzene to aniline is not possible.
SI Figure 6: Reduction by beam Pt XANESas particles are already very small may be a result of decrease in particle size (or beam damage). First scan (black), second scan (red).
The degree of reduction is far less than that observed for the sample ex situ as the oxidising colloid shell has already been removed. Both the increased absorption edge intensity and the change in the immediate post edge feature between 11570-11575 eV are indicative of an increase in Pt nanoparticle size. The X-ray beam may be increasing the mobility of the Pt clusters on the surface of the carbon, leading to localised aggregation of the clusters that, whilst still too small to be observed by XRD, is enough to cause a change in the XANES due to an increase in the degree of multiple scattering of the photoelectron.  Each row in the sinogram is a projection through the particle; a diffraction peak occurring at a certain point in the row is indicative of crystalline material being present at some point along the path of the X-ray beam through the sample. Normally when collecting XRD-CT on (poly)crystalline samples, a feature can be tracked through all rotations, forming the distinctive wave through the sinogram. However if only a few crystallites are present, then diffraction will only occur at a discrete number of rotations. The algorithms underlying the tomographic reconstruction rely on a continuous "wave" of data through the sinogram, and so fail to cleanly reconstruct the limited amount of information. The streak artefacts in SI Fig 8 are an example of reconstruction with very low information content in the sinogram.
SI Figure 10: 785 nm Raman spectra of 22.3M reference solutions of aniline and nitrobenzene in ethanol, with mixtures of 25, 50 and 75% nitrobenzene with aniline (also in ethanol). The 75% aniline/25% nitrobenzene ratio best reproduces the peak shape observed for the reaction mixture. All spectra were measured in identical capillaries to that used for the tomography measurements. 100% Aniline 75% Aniline 25% Nitrobenzene 50% Aniline 50% Nitrobenzene 25% Aniline 75% Nitrobenzene 100% Nitrobenzene Reaction mix SI Figure 11: Microreactor for in situ tomography. (a) gas inlet, (b) single catalyst particle mounted in 0.4 mm diameter glass capillary, (c) capillary sealed with two o-rings, (d) nano-positioning motors for centring capillary, (e) base to connect to beamline rotation stage.