COMMUNICATION
Maximising opportunities in supercritical chemistry: the continuous
conversion of levulinic acid to c-valerolactone in CO2
Richard A. Bourne, James G. Stevens, Jie Ke and Martyn Poliakoff*
Received (in Cambridge, UK) 11th June 2007, Accepted 5th October 2007
First published as an Advance Article on the web 16th October 2007
DOI: 10.1039/b708754c
be achieved with a tubular reactor at 200 uC and 20 MPa with a
20% excess of H2 over a 5% Ru on Al2O3 catalyst.11
Phase behaviour is manipulated during the hydrogenation of
aqueous levulinic acid in supercritical CO2 to separate almost
pure c-valerolactone from water and unreacted acid with
reduced energy requirements compared to conventional
processing.
We have found that the co-solvent can be replaced by water;
a concentrated mixture of LA and H2O (75% w/w LA, ca.
1 : 2 mol/mol) is an easily pumpable liquid. Surprisingly, this
aqueous solution of LA can be converted into GVL, even though
the second stage of the reaction involves elimination of H2O. We
used an automated reactor12 to identify the optimal conditions and
at 200 uC; almost quantitative conversion can be obtained, see
Table 1.
Compared to the patent,11 our reaction involves lower pressures
(10 MPa vs. 20 MPa) and a higher concentration of LA (LA : CO2
ca. 1 : 10 vs. 1 : 28). However we require a larger excess of H2
(H2 : LA = 3 : 1), possibly because of reduced mass transport
across the gas/liquid interface. Since GVL is miscible with H2O, it
is obtained in aqueous solution at the end of the reaction, (ca. 3 : 1
H2O : GVL, mol/mol) and separation is required. It is here that
scCO2 offers a real advantage.
Several research groups have now demonstrated that supercritical
CO2 (scCO2), is a highly effective medium for continuous catalytic
reactions,1 including alkylation,2 etherification,3 hydroformyla-
tion,4 oxidation5 and particularly hydrogenation.6 These reactions
have been very successful in terms of process intensification and
scale-up. Indeed, conditions have sometimes been optimised to
give products sufficiently pure to eliminate completely the need for
downstream purification.6 However, there are definite limitations.
In particular, the relatively poor solvent power of scCO2 and the
need to pump the substrate into the reactor means that it is often
necessary to use a co-solvent, which has then to be separated from
the product, considerably compromising the efficiency of the
process and entailing increased energy costs.
The advantage derives from the work of Lazzaroni et al. who
showed that adding a moderate, sub-critical pressure of CO2 to
aqueous THF causes liquid/liquid separation into THF-rich and
H2O-rich phases.13 This is quite distinct from supercritical
extraction since the THF remains in a separate, ‘‘gas-expanded’’
liquid phase (THF + CO2) which is immiscible with H2O but does
not dissolve substantially in the gaseous CO2 phase at these
pressures.14 We therefore investigated whether a similar phase
separation occurs when aqueous GVL is pressurised with CO2.
Our results, Fig. 1, show that apart from a slightly unusual phase
inversion, GVL behaves just like THF. Since GVL and H2O
separate at a pressure of CO2 similar to that inside our reactor, we
have the basis for a process which combines reaction and
separation. The phase inversion is a bonus which simplifies the
equipment needed but is not inherent to the principle.
In this communication, we describe a new approach that
combines the use of water as a co-solvent with phase manipulation
using scCO2 to integrate reaction and separation into a single
process with reduced energy requirements compared to conven-
tional distillation. We illustrate our approach with the conversion
of levulinic acid (LA) to c-valerolactone (GVL) (Scheme 1). LA
can be obtained from renewable biomass by the acid catalysed
dehydration of hexose sugars.7 Its conversion to GVL involves
hydrogenation followed by intramolecular cyclisation, with the
loss of water.7,8 GVL has been proposed as a sustainable liquid
and as a precursor to a biomass-derived acrylic monomer.9,10
A recent patent11 describes the continuous hydrogenation of LA
to GVL in scCO2. GVL is a liquid but LA is a solid (mp 30 uC)
which had to be pumped into the reactor dissolved in 1,4-dioxane.
At the end of the reaction the GVL had to be separated from the
dioxane solvent, the H2O produced in the reaction and any
unreacted LA, processes which are potentially energy intensive. A
variety of conditions were reported but ca. 98% conversion could
Therefore, we reconfigured our apparatus to incorporate a
simple liquid separator between the reactor and back pressure
regulator, see Fig. 2. The modified system was tested first by
Table 1 Reaction optimisation of LA to GVL in scCO2
Expt. no.
T/uC
H2 equiv.
Yield of GVL (%)
1
2
3
4
5
a
180
190
200
200
200
4.5
4.5
4.5
3.0
1.5
73
74
.99
.99
52
Scheme 1 Conversion of levulinic acid (LA) to c-valerolactone (GVL).
All reactions performed at 10 MPa system pressure with 2.76 g of
5% Ru on SiO2 (Degussa H 3036 XH/D). Flow rates 1.0 ml min21
liquid CO2, 0.3 ml min21 LA–H2O (75% w/w LA).
School of Chemistry, University of Nottingham, Nottingham, UK.
E-mail: Martyn.Poliakoff@nottingham.ac.uk; Fax: +44 115 951 3508;
Tel: +44 115 951 3520
4632 | Chem. Commun., 2007, 4632–4634
This journal is ß The Royal Society of Chemistry 2007