Resumen de Theoretical studies of chemical processes for biomass conversion on metal catalysts
The world endeavor towards sustainability began in 1987 with the Brundtland report, which defines sustainable development as balancing our economic, social, and environmental needs without compromising the ability of future generations to meet their owns.[1] In a world threatened by global warming and the depletion of fossil fuels, the development of new technologies is crucial to reach such a future. Therefore, a lot of effort is being made in developing the basis of a green chemical industry, which is based on the use of water as solvent, catalysts, renewable resources as raw materials, and other principles.2 Among the renewable resource of chemical compounds, the most abundant by far is the non-edible fraction of biomass procedent from agriculture. Biomass-derived molecules are rich in functional groups containing oxygen atoms. A particular focus is done on alcohols, as they can be obtained in large quantities from carbohydrates and be transformed into many platform chemicals by the use of heterogeneous catalysis. Indeed, many alcohols are among the “top 14” biomass-derived compound.[3] The large-scale obtention of platform chemicals from raw materials is normaly done in industrial processes that use heterogeneous catalysis, as the separation process is much simpler than their homogeneous counterpart. Currently, 75% of industrial chemical processes depend on heterogeneous catalysts.[4, 5] Most of the economic development in the 20th century is due to the heterogeneous catalysys. Still, there are many challenges ahead for the years to come. In the beginning, new catalysts were designed by trial-and-error. This practice is grossly inefficient as it consumes lots of material and human resources. However, the development of analytical techniques allowed scientists to relate catalytic properties with composition and structure. More recently, the exponential increase of computational power6 and the development of theoretical methods such as density functional theory, DFT,[7, 8] enabled the prediction of the structure and energetics of many chemical systems.[9] Indeed, the use of theoretical simulations at several levels can give valuable insights into the activity, selectivity, and stability of most catalyst.[10] Nowadays, the combination of simulations and experiments is the most rational way to design and characterize new catalysts. In this approach, industrial and experimental conditions impose the conditions to numerical simulations, and large-scale simulations imposes contraints to the short-scale ones. Conversely, atomic-scale techniques provide design parameters for larger scale simulations. Additional tools to fill the gaps are scaling relationships[9] and open databases containing the structure and energetics of relevant systems.[11, 12] I concentrated most of my research in developing scaling relationships for thermodynamic and kinetic parameters, as well as the modelling of hydrogen production and glycerol valorization. Hydrogen can be used as energy vector in fuel cells, as well as reactant in many hydrogenation reactions, like the pre-treatment of raw biomass. It can be produced from biomass by decomposition and reforming of alcohols, under the conditions described in Fig. 1.2. To this end, the adsorption and reactivity of poly-alcohols and their fragments is treated in Chapters 3-5, and accurate scaling relationships were developed and validated against experimental data. Then, as water is ubiquitous in biomass, the metal-water interface is modelled in Chapter 6 and the effect of solvation is addressed in Chapter 7. Glycerol is an important commodity chemical3 that is being overproduced in the manufacture of biodiesel and soap. As the world market cannot absorb the surplus, the retail price of crude glycerol reached zero in 2010.[13] Its reforming into hydrogen is studied in Section 5.4.1 and its upgrade into lactic acid is treated in Chapter 8. Lactic acid is also an important platform chemical[3] whose market is expanding 5–8% per year.[14] Finally, to speed-up the discovery of new catalysts, all structures and energetics were made public in ioChem-BD.[12] The development of new technologies for biomass reforming and glycerol upgrade will result in environmental-friendly and profitable processes. This, in turn, will have a positive social impact not to depending on scarce raw materials, thus paving the way towards a sustainable future. In particular, I identified the following needs: * Despite many scalings for the adsorption of molecules exist in literature, any of them have been developed for large molecules such as those derived from biomass, while the effect of hydrogen bonding on adsorption have been overlooked or oversimplified. * A general model for adsorption of molecular fragments derived from alcohols are missing. In particular, there is a need for reliable methods to predict the thermochemistry of large set of molecular fragments on metal surfaces from a small sample of them.
* To the best of our knowledge, there are not comparative theoretical studies based on DFT and microkinetic modeling on the activity, selectivity, and stability of Cu, Ru, Pd, and Pt for hydrogen production through autothermal, steam, and aqueous-phase reforming of alcohols. Also, the reliability of linear scaling relationships for transition states has not been tested to hold on different metals.
* There are limited knowledge about the nature of the interface between metals and liquid water. In particular, previous theoretical studies do not detail which are the minimum box dimensions to reproduce the behaviour of liquid water.
* As implicit solvation models for periodic-boundary conditions are too recent, there are limited data about their reliability when describing reaction 26networks of oxygenates, such as the first steps of methanol decomposition. Also, the effect of coadsorbed methanol molecules in the reaction path has not been quantified. Whether linear scaling relationships hold if the reactions occur under solvation.
* There are not mechanistic studies on the oxidation of acetol to pyruvaldehyde, which is one of the routes to convert glycerol into lactic acid.
REFERENCES [1] Brundtland, G. H.; Khalid, M. Our common future, New York 1987, .
[2] Anastas, P.; Eghbali, N. Green chemistry: principles and practice, Chem. Soc. Rev. 2010, 39, 301–312.
[3] Bozell, J. J.; Petersen, G. R. Technology development for the production of biobased products from biorefinery carbohydratesthe US Department of Energys “top 10” revisited, Green Chem. 2010, 12, 539–554.
[4] Moulijn, J. A.; van Leeuwen, P. W. N. M.; van Santen, R. A. Catalysis: an integrated approach to homogeneous, heterogeneous and industrial catalysis; volume 79 Elsevier: 1993.
[5] Bhaduri, S.; Mukesh, D. Homogeneous catalysis: mechanisms and industrial applications; Wiley: 2000.
[6] Schaller, R. R. Moore’s law: past, present and future, IEEE Spectr. 1997, 34, 52–59.
[7] Hohenberg, P.; Kohn, W. Inhomogeneous electron gas, Phys. Rev. 1964, 136, B864.
[8] Kohn, W.; Sham, L. J. Self-consistent equations including exchange and correlation effects, Phys. Rev. 1965, 140, A1133.
[9] Nørskov, J. K.; Bligaard, T.; Rossmeisl, J.; Christensen, C. H. Towards the computational design of solid catalysts, Nat. Chem. 2009, 1, 37–46.
[10] López, N.; Almora-Barrios, N.; Carchini, G.; Blonski, P.; Bellarosa, L.; García-Muelas, R.; Novell-Leruth, G.; García-Mota, M. State-of-the-art and challenges in theoretical simulations of heterogeneous catalysis at the microscopic level, Catal. Sci. Tech. 2012, 2, 2405–2417.
[11] Wilkinson, M. D. et al. The FAIR guiding principles for scientific data management and stewardship, Sci. data 2016, 3.
[12] Alvarez Moreno, M.; De Graaf, C.; López, N.; Maseras, F.; Poblet, J. M.; Bo, C. Managing the computational chemistry big data problem: The ioChem-BD platform, J. Chem. Inf. Model. 2014, 55, 95–103.
[13] Ciriminna, R.; Pina, C. D.; Rossi, M.; Pagliaro, M. Understanding the glycerol market, Eur. J. Lipid Sci. Technol. 2014, 116, 1432–1439.
[14] Dusselier, M.; Van Wouwe, P.; Dewaele, A.; Makshina, E.; Sels, B. F. Lactic acid as a platform chemical in the biobased economy: the role of chemocatalysis, Energy Environ. Sci. 2013, 6, 1415–1442.
MAIN CONCLUSIONS The adsorption and reactivity of oxygenated molecules derived from biomass was studied by state-of-the-art theoretical methods based on density functional theory, including novel approaches like dispersion and solvation models. The first research Chapters (3–4) focused on the adsorption of large alcohols and molecular fragments, following with the reactivity of C 1 –C 3 alcohols in Chapter 5. The interfaces between liquid water and metal surfaces were studied in Chapter 6, paving the way to study the effect of solvation in reactivity, Chapter 7. As surfaces under working conditions may contain oxides and suboxides, three models were applied for silver on Chapter 8 to account for different oxidation states. In general, the methods applied in this dissertation reproduced accurately experimental values and trends. All the structures are available in our repository, ioChem-BD, thus satisfying the FAIR data principles: findability, accessibility, interoperability, and reusability. The main conclusions are summarized below.
* The ground-state conformation for a poly-alcohol in gas phase maximize the number of intramolecular hydrogen bonds and minimizes the strain.
* The main energy contributions for the adsorption of polyalcohols on Pd or Pt(111) comes from its hydroxyl groups and the weakening of its intramolecular hydrogen bonds.
* The inclusion of van der Waals interactions is fundamental to reproduce the experimental adsorption energies of alcohols.
* The carbon tail contributes to the adsorption energy via van der Waals interactions, being –0.07 and –0.06 eV/CH x for Pd and Pt respectively.
* For large molecules, structural complexity may hinder the simultaneous interaction of all the hydroxyl groups with the metal, thus reducing the total adsorption energy.
* The adsorption energy of a C 1 -C 7 poly-alcohol can be obtained from its collective descriptor, which contains (i) the number of hydroxyl groups that interact with the surface, (ii) the number of carbon atoms in the tail, and (iii) the energy penalty of loosening and intramolecular hydrogen bonds.
* The model can be generalized to include molecular fragments and other functional groups, thus paving the way to develop new catalytic routes to transform large oxygenated molecules into chemicals and fuels.
* The energetics for the 71 intermediates from the C 1 -C 2 decomposition network on 12 metal surfaces can be written as a function of two descriptors within an accuracy of 0.08 eV (mae).
* The two metal descriptors can be traced back to the d-band center and the reduction potential.
* With the formation energies of three adsorbates on a given metal, namely O*, OH*, and CCHOH*, the thermochemistry of the remaining intermediates can be predicted within an accuracy of 0.12 eV (mae), following a robust methodology based on principal component analysis and regressions.
* The adsorption energies predicted from the methodology based on principal component analysis and regressions are in good agreement with experimental and previous DFT studies * The transition states for the O–H bond breakings resemble their corresponding initial states. Also, the energies for transition and initial states scale with each other as an offset.
* The transition states for the C–H, C–C, C–O, and C–OH cleavages resemble their final states, and the potential energies for their transition and final states also scale as an offset.
* The scaling of Ru, Pd, and Pt can be lumped into one equation while maintaining sufficient accuracy to predict the H 2 production rate.
* During the reforming of C 2 alcohols, several relevant stationary states can be found in their microkinetic models.
* After three hours of reaction, all C 2 reforming processes reached a relevant stationary state within a relative variation threshold of 0.01%/s for all surface concentrations and reaction rates.
* During the autothermal reforming on Ru and Pd, and all reforming technologies on Cu, the surfaces are covered by O* and OH*, compromising the validity of the metal-only model.
* Undesirable deviations in the activity towards H2 production may arise when linear-scaling relationships are used to predict a metal not represented in the training set (MK-L1O).
* The surface reactivity has a large impact in the characteristics of the metal-water interface.
* Irregular 5-, 6-, and 7-membered rings found in experiments on Pd and Pt were reproduced by Born-Oppenheimer molecular dynamics.
* On Ru, 40% to 50% of the water molecules remained dissociated, in good agreement with the experimental measures.
* On Ru, the wetting layers form a reminiscent of the electric double layer.
* On Ru, the dissociation products (H3O+ , OH– , and others) remains bounded to the surface, so acidification does not occur, as found in experiments.
* The behavior of liquid water can be reproduced if at least 1.4 nm in each direction is provided to the set of molecules, and avoiding confinement.
* When pure methanol is present on the surfaces, the methanol molecules aggregate and the preferred decomposition path starts by the O–H breaking.
* For aqueous methanol solutions, which resembles the aqueous-phase reforming conditions, the methanol molecules on the surface are surrounded by water. The solvent changes the decomposition to start by the C–H breaking.
* The model containing two explicit water molecules and implicit solvent (S2W+) is the most suitable to describe experiments. It is also less computationally demanding than the use of a bulk liquid explicit model.
* The linear scaling relationships described on Chapter 5 hold when different solvation models are used, thus validating the MK-LSR results for aqueous-phase reforming.
* Experimental observations found that the O 2 :GLY ratio in the feed influences the nature of the active phase. For sub-stoichiometric ratios, 0.0 "menor que" O 2 :GLY " 0.5, metallic silver predominates. For oxygen-rich environments, 0.5 "menor que" O 2 :GLY, the bulk oxide Ag 2 O is formed. For stoichiometric ratios, O2:GLY=0.5, a surface oxide of silver is formed.
* The oxidation potential for each metal also influences the reactivity of the metal.
* When the reactor is fed with a mixture with an sub stoichiometric O2:GLY ratio, the surface is oxygen-poor and the acetol dehydrogenation is thermodynamically limited, thus lowering the production rate of pyruvaldehyde.
* In oxygen-rich environments, O2:GLY ratio larger than 0.5, the surface adopts a structure similar to a bulk oxide, Ag2O. This hinders the catalyst regeneration through H2O formation and desorption.
* The most reactive surface was the surface oxide on metallic silver,AgOx/Ag, which dominates in O2:GLY ratios 0.5. There, both the dehydrogenation of acetol to pyruvaldehyde and the catalyst regeneration are promoted.
