Gas Phase Glycerol Valorization over Ceria Nanostructures with Well-Defined Morphologies

Glycerol solutions were vaporized and reacted over ceria catalysts with different morphologies to investigate the relationship of product distribution to the surface facets exposed, particularly, the yield of bio-renewable methanol. Ceria was prepared with cubic, rodlike, and polyhedral morphologies via hydrothermal synthesis by altering the concentration of the precipitating agent or synthesis temperature. Glycerol conversion was found to be low over the ceria with a cubic morphology, and this was ascribed to both a low surface area and relatively high acidity. Density functional theory calculations also showed that the (100) surface is likely to be hydroxylated under reaction conditions which could limit the availability of basic sites. Methanol space-time-yields over the polyhedral ceria samples were more than four times that for the cubic material at 400 °C, where 201 g of methanol was produced per hour per kilogram of the catalyst. Under comparable glycerol conversions, we show that the rodlike and polyhedral catalysts produce a major intermediate to methanol, hydroxyacetone (HA), with a selectivity of ca. 45%, but that over the cubic sample, this was found to be 15%. This equates to a 13-fold increase in the space-time-yield of HA over the polyhedral samples compared to the cubes at 320 °C. The implications of this difference are discussed with respect to the reaction mechanism, suggesting that a different mechanism dominates over the cubic catalysts to that for rodlike and polyhedral catalysts. The strong association between exposed surface facets of ceria to high methanol yields is an important consideration for future catalyst design in this area.


Page
Description S3 Figure S1. The effect of cell volume on the calculated lattice energy for a series of k-point sampling grids. S4 Table S1. Structural parameters for the cubic cell of bulk ceria refined using the Murnaghan approach. S5 Table S2. Calculated surface energies from slab model simulations. Thermochemistry calculation details and Estimation of partial pressures S16 Fig S11 & fig S12. CO2 and NH3 TPD profiles S18 Estimation of CO2 desorption energy using the Redhead equation. S19 Table S3. GC retention times for full product list S20 Table S4. Qualitative LC-MS analysis conditions S2 S21 Table S5a. Full product distribution at a space velocity of 3600 h -1 and reaction temperature of 320 °C S22 Table S5b. Full product distribution at a space velocity of 3600 h -1 and reaction temperature of 360 °C S23 Table S5c. Full product distribution at a space velocity of 3600 h -1 and reaction temperature of 400 °C S24 Table S6a. Full product distribution at isoconversion at Cgly ≈ 15% and a reaction temperature of 320 °C. Space velocities: Ce-C = 3600 h -1 , Ce-R = 11250 h -1 , Ce-P = 9000 h -1 . S25 Table S6b. Full product distribution at isoconversion at Cgly ≈ 99% and a reaction temperature of 320 °C. Space velocities: Ce-C = 1800 h -1 , Ce-R = 3600 h -1 , Ce-P = 3600 h -1 . S26 Table S7. Full carbon based product distribution at iso-conversion at Cgly ≈ 99% and a reaction temperature of 320 °C. Space velocities: Ce-C = 1800 h -1 , Ce-R = 3600 h -1 , Ce-P = 3600 h -1 Figure S13. Post reaction TGA profiles S27 Table S8. Glycerol conversion and main product selectivies as a function of contact time at various temperatures S28 Table S9. Glycerol conversion to MeOH over various catalysts S29 References S3

S1 DFT Calculation Parameter Choices
The cell vector for each k-point sampling grid was obtained by optimisation of atom co-ordinates at a series of cell volumes by systematically expanding and contracting the cell vector around the experimentally reported value. The resulting energy vs cell volume plots for expansion factors up to ±1% in 0.1% steps are shown in Figure S8.
From the region close to the minimum (indicated by the lines of best fit on figure S8) an initial estimate for the bulk modulus can be made based on a quadratic fit: where V0 the cell volume at the minimum cell energy, E(V0) and a is the coefficient for the V 2 term in the quadratic fit.
The cell energy as a function of cell volume, E(V), can be fitted over a broader range using the Murnaghan equation of state: where the additional parameter, 0 ′ is the derivative of the bulk modulus, K0, with respect to pressure: Cell Volume / Å 3 Energy / eV Figure S1: The effect of cell volume on the calculated lattice energy for a series of k-point sampling grids. Plot shows result of k-point sampling of 3x3x3 (diamonds), 5x5x5 (squares), and 13x13x13 (triangles).
The data from the cell expansion calculations in VASP was fitted to the Murnaghan equation of state using a simple minimisation of the sum of squared deviations between data and values calculated from the equation of state with V0, E0, K0 and K0' as variable parameters using the Excel spreadsheet solver function. The initial values of the first 3 parameters were taken from the quadratic fit results and the initial value of K0' was set to 4. The final values obtained by the fitting procedure are given in Table S1.    Raman shift (cm -1 ) Intensity (a.u.) S12 Figure S8. Temperature programmed reduction profiles using H2 as reductant for Ce-C (pink), Ce-R (orange) and Ce-P (blue). All samples were heated at a rate of 10 K min -1 under 30 mL min -1 10 % H2/Ar.

S3 Thermochemistry calculations
To calculate the free energy of hydroxylation for the surfaces the VASP code was used to evaluate the vibrational modes of the relaxed clean surface, the surface with one monolayer coverage and an isolated water molecule. For vibrational calculations of slabs a single oxide layer and all adsorbate atoms were included in the degrees of freedom used to form the second derivative matrix. The enthalpy, H, entropy, S, and free energy, G, at a particular temperature, T, and pressure, P, are then calculated using the formulae: where Uelec is the PBE electronic energy of the system calculated by the VASP optimisation. The conditions was estimated as described in the following section.
For each system the calculations of the enthalpy and entropy was achieved using modules from the Atomic Simulation Environment python library. 6 As part of this work we have implemented python scripts to read the required data from VASP output files and carry out the set of calculations required to give the enthalpy, H, entropy, S, and free energy, G, change for the formation of the monolayer of water from the clean slab and isolated water molecules. The script makes additional checks such as ensuring that the number of degrees of freedom in reactant and product states is correctly matched.

S4 Estimation of the partial pressure of water under the experimental conditions
The partial pressure of water under experimental conditions (reaction temperature of 400 °C) was estimated using Dalton's law and the molar flow rates of glycerol, water, and argon carrier gas. The ideal gas law was used to calculate the mole fraction of water due to the large excess of inert carrier gas present.

S5 Estimation of CO2 desorption energy using the Redhead equation.
To estimate the desorption energy for CO2 from the temperature of a TPD peak maximum, Tm, we employ the Redhead equation: 7 This equation is derived from the Polanyi-Wigner equation assuming coverage independent desorption and that the CO2 adsorption/desorption process is first order. In equation (S9), R is the molar gas constant and 1 is the frequency factor in the Arrhenius expression for the desorption rate constant, which is taken to be 1 = 10 13 s -1 . The parameter, , is the heating rate of the TPD experiment, in our measurements = 0.167 K s -1 throughout ( Figure S10). To obtain the value of the activation energy, , the right hand and left hand expressions in equation (S9) were evaluated for an initial guess value using an excel spreadsheet. The solver function was then used to minimise the square of the difference between the two sides of the equation by varying , , for the values reported the difference between the left and right side was always less then 10 -30 K -1 . S18

S6 Analysis of Reaction Products
Glycerol

Glycerol to MeOH: Qualitative analysis
Some additional qualitative analysis of the post reaction effluent was conducted by liquid chromatography-mass spectrometry (LCMS). This was conducted on a Bruker Amazon SL ion trap mass spectrometer which was operated in positive electrospray ion mode and coupled to a Thermo Ultimate HPLC system. The HPLC was equipped with a C-18 column (maintained at 40 ºC) and utilized a gradient elution consisting of 0.1% formic acid in H2O (A) and 0.1% formic acid in acetonitrile. 10 μL of sample was injected and the gradient elution was performed as illustrated in Table S3.   Ce-C Ce-R Ce-P a) corresponds to 0.85 mL/min glycerol flow; b) not reported -assumed; c) corresponds to 6.6 mL/min glycerol flow; d) not reported