Mechanism of isopropanol dehydration on -Al2O3:

experiments and DFT for a predictive kinetic model

Kim Larmier,1,2 Cline Chizallet,1,* Andr Nicolle, 3 Sylvie Maury,1 Nicolas Cadran,1 Johnny Abboud,2
Anne-Flicie Lamic-Humblot,2 Eric Marceau,2 Hlne Lauron-Pernot2
1
IFP Energies nouvelles, Catalysis and Separation Division, Rond-Point de lchangeur de Solaize, BP3, 69360 Solaize, France
2
Sorbonne Universits, UPMC Univ Paris 06, UMR CNRS 7197, Laboratoire de Ractivit de Surface, 4 place Jussieu, F-75005
Paris, France
3
IFP Energies nouvelles, Powertrain and Vehicule Division, 1-4 avenue de Bois-Prau, 92852 Rueil-Malmaison Cedex, France
*Corresponding author: celine.chizallet@ifpen.fr

Keywords: -alumina, alcohol dehydration, isopropanol, DFT, kinetics, kinetic modeling

1. Introduction calculations (DFT+D2) were performed using the

Alcohols are important platform molecules obtained
from the conversion of lignocellulosic biomass. It is
well known that dehydration reactions of alcohols
toward both olefins and ethers are successfully
catalyzed by common and cost-effective acidic
oxides such as -alumina. Both types of products,
currently supplied by the petrochemicals industry,
are industrially significant, as monomers for
polymer production (olefins) or as fuel additives
(ethers).
However, the key parameters governing each
dehydration reaction, in terms of active sites and
mechanisms, are still under debate, as shown by
several experimental[1,2] or computational[3]
studies dealing with ethanol dehydration. We
address here the selectivity issue stemming from the
two competing dehydration paths offered to
isopropanol on -Al2O3 and related solids, and to
rationalize the evolution of the olefin-to-ether ratio
over the whole conversion range of the alcohol. To
this end, we have set up a combined experimental
and theoretical approach, in which experimental
measurements are compared with the results of DFT
calculations (modeling at the molecular scale) and
microkinetic simulations (modeling at the
macroscopic scale).[4]
VASP software (PBE+D2, PAW, Ecutoff = 400 eV),
and transition state determinations were performed
using the NEB method. Vibrational analyses were
performed to deduce free energy profiles. We
employed the models for the -alumina surface
proposed by Digne et al.[5] Reactor modeling was
performed with the Chemkin software, by
implementing the kinetic parameters of the
elementary steps calculated ab initio.
3. Results and discussion
Isopropanol conversion was measured at 200C over
three related aluminas. While the specific activity
varied with the catalyst (-Al2O3 being the most
active one), the selectivity towards diisopropylether
was found to be identical at a given conversion (Fig.
1, marks), with an increase to a maximum for an
isopropanol conversion of 35%, and a subsequent
decrease to the benefit of propene.

2. Experimental Part and Theoretical Details

Experimental measurements were carried out on a
gas-phase catalytic test using a plug-flow reactor.
Before reaction, aluminas (-alumina, Puralox
TH100/150, Sasol ; Na-poisoned - alumina; -
alumina) were activated at 450 C (5C.min-1) under
nitrogen flow (20 mL.min-1) for 3 h. Isopropanol
(Sigma-Aldrich, 99%) was introduced with a partial
pressure of 1.5 kPa in a N2 flow (5 to 60 cc.min-1).
The composition of the effluent was determined by
gas chromatography (Carbowax column). Ab initio
Figure 1. Evolution of partial pressures with an
equivalent of contact time on -alumina, experimental
(dots) versus simulated (lines) with our final model
(reaction temperature 200 C, initial P(iPOH) = 1.5kPa,
mcat = 30 mg).
This leads to the conclusion that the reaction paths
giving propene and diisopropylether should not be
treated as parallel and independent reactions in
modeling, and that the active sites should be similar
regardless of alumina, though more or less abundant.
In order to identify plausible active sites, the
reaction pathways leading to the formation of
propene or diisopropylether from isopropanol were
calculated by DFT on the most abundant surfaces of
-alumina, taken in their relevant hydration state
under reaction conditions: (100) facets are
dehydrated while (110) terminations are partially
hydrated (around 9.0 OH.nm-2). The most favorable
pathways for both reactions have been found on the
(100) surface, with an identical intermediate, an
isopropanol molecule adsorbed on a
pentacoordinated AlV Lewis site (Fig. 2).
Figure 2. (a) Enthalpic diagram for the formation of
propene from isopropanol on the AlVa site of the -
Al2O3(100) surface. First transition state structures for
the: (b) E1 pathway, (c) E1cb partway, (d) E2 pathway.
Grey: Al, Red: O, Blue: C, yellow: H.
Computed activation enthalpies for the formation of
propene and diisopropylether are close (125 kJ.mol-
1
/E2 mechanism, and 112 kJ.mol-1/SN2 mechanism,
respectively); they match very well the experimental
values (128 5 and 118 5 kJ.mol-1, respectively).
Calculations show that the selectivity toward one
product or the other is thus more entropy- than
enthalpy- driven (rS = 8 and 36 J.K-1.mol-1,
respectively).
A microkinetic model was established according to
the molecular modeling results, which also included
the ether decomposition back to propene and
isopropanol, occurring at higher temperatures and
also favored on the same active site, as well as water
inhibition as proposed by DeWilde et al.[1] (Fig. 3).
The simulated evolution of partial pressures with
contact time and of selectivity with conversion (Fig.
1, solid line) successfully reproduces the
experimental data on the three materials.
Figure 3. (a) Macro site on the (100) surface of -alumina
and its schematic representation. Gray: aluminum; red:
oxygen. (b) Simplified network representation of the
kinetic model for isopropanol dehydration on alumina,
including the effect of the inhibition of the reaction by
water affecting only the formation of propene.
4. Conclusions
Comparison between experimental data and multi-
scale modeling shows that dehydration reactions to
olefin and ether may involve the same active site, an
AlV Lewis acid site located on the lateral (100)
surface of the alumina grains. As a consequence, the
various paths leading to olefin and ether during
alcohol dehydration cannot be treated as
independent and parallel reactions, but interact
following a complex interplay on a unique site.
Selectivity is thus mainly governed by experimental
reaction conditions, and not by the action of distinct
sites on the catalyst.
References
[1] J. F. DeWilde, H. Chiang, D. A. Hickman, C. R. Ho, A.
Bhan ACS Catalysis 2013, 3, 798.
[2] T. K. Phung, A. Lagazzo, M. . Rivero Crespo, V. Snchez
Escribano, G. Busca J. Catal. 2014, 311, 102.
[3] J. H. Kwak, R. Rousseau, D. Mei, C. H. F. Peden, J. Szanyi
ChemCatChem 2011, 3, 1557; M. A. Christiansen, G.
Mpourmpakis, D. G. Vlachos ACS Catalysis 2013, 3, 1965.
[4] K. Larmier, A. Nicolle, C. Chizallet, N. Cadran, S. Maury,
A.-F. Lamic-Humblot, E. Marceau, H. Lauron-Pernot ACS
Catalysis 2016, 6, 1905; K. Larmier, C. Chizallet, N. Cadran, S.
Maury, J. Abboud, A.-F. Lamic-Humblot, E. Marceau, H.
Lauron-Pernot ACS Catalysis 2015, 5, 4423.
[5] M. Digne, P. Sautet, P. Raybaud, P. Euzen, H. Toulhoat J.
Catal. 2004, 226, 54.