C. Williamson, A Gardiner & E Pilbrow
HYDROGEN FROM ETHANOL VIA STEAM REFORMING IN A
PLASMA REACTOR
1
2
2
C. J. Williamson , A Gardiner and E Pilbrow
1
University of Canterbury,
Christchurch, New Zealand.
[email protected]
2
Industrial Research Ltd, IRL,
Christchurch, New Zealand
ABSTRACT
Industrial Research Limited (IRL) and the University of Canterbury are interested in the
production of hydrogen from ethanol, primarily for small scale applications in stationary
fuel cells. An example is for use as a complementary stored fuel source for an IRL
concept called Hylink that proposes the use of hydrogen as an energy store and carrier in
remote area power system applications.
Plasma reformation of mixtures of ethanol and steam shows promise as capable of
achieving high hydrogen selectivity without some of the problems associated with
catalytic processes. These include the requirement for expensive metals in the catalyst,
lack of selectivity, coking and deactivation.
A laboratory scale reactor for ethanol steam reformation in a non-thermal plasma was
designed, built and tested in the University of Canterbury and IRL laboratories. An
ethanol/water mixture was vaporised at close to atmospheric pressure and passed
through a plasma created by applying a voltage of 7 kV across carbon electrodes. The
mixture breaks down into hydrogen, carbon dioxide, carbon monoxide and small
amounts of hydrocarbons (mainly methane and ethylene). Initial experimental results
show gas mixtures of 60-70 mol % hydrogen are formed from feeds varying between 2.5
and 10 mol % ethanol. The results also show good selectivity to hydrogen at up to 4 mol
hydrogen/mol ethanol fed, comparable to or better than catalytic systems. Fuel cell grade
hydrogen can subsequently be formed from the predominantly hydrogen/carbon oxide
mixtures using a hydrogen permeable membrane.
INTRODUCTION
Industrial Research Ltd (IRL) have a concept for remote area power applications
(Hylink). Many small rural communities in New Zealand require reliable power supply
but running overhead power lines from the national grid can cost upward of
$30,000/km. These small communities are often located in or near areas that would be
suited to the installation of a wind turbine for power generation, although once again the
cost of installing transmission lines can be prohibitive. The Hylink proposal aims to
overcome this cost barrier by using hydrogen as the energy transport medium. Hydrogen
can be transported several kilometres in a polymer pipeline at a fraction of the cost of
transmission lines. The Hylink system consists of a wind turbine (or other power
1
C. Williamson, A Gardiner & E Pilbrow
source) connected to a water electrolyser, the electrolyser converts the electrical power
to hydrogen which is then transported to the community in a pipeline. The hydrogen is
then either converted to electricity, if required, or used directly for heating and cooking.
Ethanol reformation is proposed to complement water electrolysis as a source of
renewable hydrogen within the Hylink system. Ethanol is arguably the most
environmentally friendly alcohol as the majority of it is produced by fermenting
renewable biomass sugars producing low concentration ethanol/water solutions.
However, current systems for producing hydrogen from ethanol are constrained by
limitations within conventional catalytic reactors. This is in contrast to methanol which
is easily steam reformed to close to stoichiometric mixtures of hydrogen and carbon
dioxide via the reforming reaction:
CH3OH + H2O → CO2 + 3H2
(1)
The CO2 and hydrogen can then be separated by a membrane to make fuel cell grade
hydrogen available. Methanol, although easily reformed, is predominantly produced
from natural gas and can therefore be regarded as a non-renewable fuel. The
corresponding ethanol steam reforming reaction is;
C2H5OH + 3H2O → 6H2 + 2CO2
(2)
The reaction demonstrates a second key advantage of ethanol as a source of hydrogen in
comparison to methanol, 6 moles of hydrogen are released per mole of ethanol as
opposed to 3 moles/mole methanol. On a volumetric basis, a litre of ethanol contains
potentially 40% more hydrogen than a litre of methanol. However, there is at present no
catalytically based system that can reform ethanol at low temperatures (around 300 oC)
and atmospheric pressure with high conversion and high selectivity to hydrogen via the
steam reforming reaction. The successful ethanol reforming reactors rely on high
temperatures (>450 oC) and expensive catalysts to achieve results. Less expensive
catalysts and lower temperatures tend to favour side reactions as well as the required
breakdown of ethanol into H2 and CO2. All catalysts, expensive or not, are also subject
to the usual deactivation and coking problems.
An alternative is a plasma reactor where the ethanol/water mixture is passed through a
region of ionised gas generated between electrodes by a high voltage and low current
field. Recent studies have shown that this approach holds some promise for ethanol
steam reforming at low temperatures (Aubry et al, 2005).
PLASMA REACTOR DESIGN
The plasma state can be either thermal (high temperature) or non-thermal (low
temperature and non-equilibrium). For example, the sun represents a thermal plasma
while neon lights are non-thermal. The plasmas of interest for ethanol reforming are
2
C. Williamson, A Gardiner & E Pilbrow
non-thermal as these have been shown to assist the reaction at low inlet gas mixture
temperatures while requiring less energy for generation than the thermal kind.
The University of Canterbury and IRL combined to design a plasma reactor for ethanol
steam reforming (the reactor is shown in Figure 1). The reactor housing is a QVF glass
chamber flanged at each end to secure tufnel end caps. The glass body can withstand
pressures and temperatures envisaged for the reactor while preventing arc flare from the
electrodes to the reactor walls. Tufnel is a resilient and insulating material and the end
caps form sections that reduce the reactor volume, increasing the proportion of the feed
mixture that passes through the plasma, while holding the graphite electrodes in place.
Figure. 1 University of Canterbury and IRL Plasma reactor with ignited arc.
A voltage of approximately 7 kV is applied across a small gap (around 3 mm) between
the electrodes, generating a string shaped plasma region. The plasma region has
characteristics of a “glow discharge” illustrating the non-thermal nature of the plasma. A
vaporised ethanol/water mixture enters the reactor above the plasma and leaves at the
base. Downward flow through the reactor allows any liquid formed by vapour
condensation to drain from the reactor.
The saturated vapour feed stream is generated using a Russell Hobbs Espresso coffee
machine modified to produce a vapour flow (monobloc steam generator in Fig. 2), a
convenient and inexpensive source of heat and low flow pumping capacity. The
unreacted components in the product gas (ethanol and water) are condensed and
3
C. Williamson, A Gardiner & E Pilbrow
separated from the product gas mixture. This product gas can then be sampled and its
composition analysed using an Agilent Micro GC.
A schematic diagram of the experimental setup is shown in Figure 2.
Figure. 2 Experimental setup for plasma steam reforming of ethanol
EXPERIMENTAL METHODOLOGY
Ethanol used in the experiments was High Grade Absolute Alcohol (> 99.85 % v/v
ethanol) supplied by Anchor Ethanol Ltd. Water was treated using a Millipore Elix 5
laboratory purification system to meet type 2 quality standards. Ethanol/water solutions
were prepared by adding measured volumes of the components to a feed storage
container. Since the density of the components was known accurately volumetric
quantities were chosen to provide the required mass/molar concentrations in the feed
mixtures.
The monobloc steam generator was set up to provide a constant vapour flowrate. The
flowrate was confirmed for the 5 mol % ethanol solution by a number of runs where the
change in liquid volume in the feed supply reservoir and the time taken were both
measured. The flowrate was calculated as 17.7 g min-1 as solution density and
volumetric flow were both known.
A series of experiments were conducted with a constant volumetric flow of feed leaving
the feed reservoir but with ethanol concentrations ranging from 2.5 to 10 mol %. Two or
more runs were made at each ethanol concentration to assess the repeatability of the
results. The product gas composition was measured using the Agilent Micro GC for
4
C. Williamson, A Gardiner & E Pilbrow
each run. A single run was also made with pure water in order to establish whether there
was appreciable breakdown of water occurring in the plasma.
Plasma reactions are not well understood or modelled. However, the constituent
compounds that appeared in gas analyses suggested that the reacting system could be
characterised by a relatively simple scheme that involved two major and three minor
reactions:
Major reactions:
C2H5OH + 3H2O → 6H2 + 2CO2
(3)
C2H5OH + H2O → 4H2 + 2CO
(4)
Minor reactions:
C2H5OH → CO + CH4 + H2
(5)
C2H5OH → C2H4 + H2O
(6)
C2H4 + H2 → C2H6
(7)
Assuming that the system was described by these reactions a carbon balance could be
made from the gas analyses to establish the amount of ethanol required to produce a
mole of the dry product gas. The accuracy of the calculation was checked by
comparing the expected hydrogen content of the gas with the measured hydrogen
composition. The calculated ethanol amount allowed a hydrogen selectivity in terms of
moles hydrogen produced/mole ethanol consumed to be calculated. As the ethanol
consumption during the experimental runs was small and the concentration in the liquid
changed only slightly this was the most accurate method available for finding the
selectivity to hydrogen and therefore gauging how the plasma reactor compared with
literature values for catalyst based systems operating at similar conditions.
A material balance was possible for the 5 mol % ethanol feed because the gas flow (N4
in Figure 3) was measured for this case.
RESULTS
Measurements were made of the change in liquid inventory in the steam generator with
time to establish that the flow of ethanol/water solution at 5 mol % was 17.7 g min-1. As
the volumetric flowrates were constant for other feed concentrations of ethanol the mass
flowrates will vary slightly. A block diagram for the 5 mol % ethanol case and an
associated stream table is shown in Figure 3 and Table 1.
N4
Steam
Generator
N1
Plasma
Reactor
N2
N3
Condenser
Separator
N5
Figure. 3 Block diagram of the plasma reactor experiment.
5
C. Williamson, A Gardiner & E Pilbrow
Stream
N1
N2
N3
Ethanol
11.9
10.2
10.2
10.3
Water
88.1
87.3
87.3
89.7
Hydrogen
0.3
0.3
11
Methane
0.1
0.1
4
Carbon
Monoxide
1.0
1.0
41
Carbon
Dioxide
0.9
0.9
36
Ethylene
0.2
0.2
7.5
Ethane
0.0
0.0
0.5
17.7
17.7
0.45
Total Flow
-1
g min
17.7
N4
N5
17.25
Table. 1 Stream Table for the 5 mol % ethanol case showing stream composition in
mass % and total flow in g min-1
The results indicate that, for the 5 mol % ethanol case, single pass ethanol conversion
through the plasma reactor is 0.3 g min-1 or 14.6 % of the ethanol in the feed.
The plasma reactor was fed mixtures with ethanol content varying from 0 to 10 mol %
and the variation in gas composition for both major components (H2, CO and CO2) and
minor components (CH4, C2H4, C2H6) is shown in Figures 4 and 5.
In one experiment, the reactor was fed with pure water and product gas analysis (result
shown in Figure 4) showed the presence of both hydrogen and oxygen indicative that
some breakdown of the water had occurred in the plasma. However, in this experiment
the gas flow was so small it could not be accurately measured.
The molar ratio of hydrogen produced/ethanol consumed was also calculated for the
different feed ethanol concentrations and this is shown in Table 2.
Ethanol mol % in the feed
Selectivity
[Mol H2/Mol C2H5OH reacted]
2.5
3.93
5.0
3.46
10.0
3.06
Table. 2 Selectivity to hydrogen (mol H2/mol ethanol reacted) with feed concentration.
6
C. Williamson, A Gardiner & E Pilbrow
Figure 4. Major component composition (mol %) in the product gas vs. feed ethanol
content
Figure 5 Minor component composition in the product gas (mol %) vs. feed ethanol
content
These results are all taken from the final year research project work of Kyle (2009).
7
C. Williamson, A Gardiner & E Pilbrow
DISCUSSION & CONCLUSIONS
The results are encouraging with product gas mixture compositions of 60-70 mol %
hydrogen and selectivities of up to 3.93 mols H2/mole C2H5OH at a feed ethanol
concentration of 2.5 mol %. Total hydrocarbon byproducts in the gas are low ranging
between 5 and 8 mol %.
The results achieved with the plasma reactor have improved hydrogen selectivity when
compared with published results for similar scale catalytic systems operating at low
temperatures (Cussins, 2005, Mariño et al, 2001 and Sun et al, 2005 shown in Table 3).
This indicates significantly more ethanol reformation is favoured in this case rather than
byproduct formation.
EtOH Conversion
Catalyst Type
(max)
Hydrogen Yield
(max)
Feed
concentration
at maximum
Reference
(mol %)
Moles H2/mole ethanol
fed
Ni(6)/γ-Al2O3
89.4
0.02
29.9
Cussins (2005)
Ni(10)/γ-Al2O3
97.3
1.2
36.2
“
Ni(20)/γ-Al2O3
99.8
1.8
29.9
“
Ni(16)/γ-Al2O3
86.4
-
Cu(6)/Ni(20)/γ-Al2O3
99.5
1.3
Cu(6)/Ni(6)/γ-Al2O3
90
0.9
(EtOH wt%)
Sun et al (2005)
29.9
“
Marino et al
(2001)
Table. 3 Ethanol conversion and hydrogen selectivity in low temperature catalytic
studies (300 oC) of ethanol steam reforming (Cussins, 2005)
The main byproduct reaction produces carbon monoxide and hydrogen as opposed to
the ideal carbon dioxide/hydrogen mixture. However, hydrogen composition could be
increased relatively easily by reacting carbon monoxide with water via the water gas
shift reaction. This mixture is then easily separated using existing membranes.
The ethanol conversion in the plasma case is much lower than comparable catalytic
systems. The most likely reason for the low conversion is that the plasma region
occupies only a fraction of the volume of the reactor and therefore much of the feed
ethanol/steam mixture can bypass the plasma and remain unreacted.
In comparison to a published plasma reactor study (Aubry et al, 2005) the results show
similar gas compositions in terms of the major components hydrogen, carbon dioxide
and carbon monoxide content. The minor impurities show a higher ethylene content
with similar methane and ethane compositions. The gas composition trend with
increasing ethanol concentration in the feed, reducing hydrogen concentration,
8
C. Williamson, A Gardiner & E Pilbrow
increasing carbon monoxide and reducing carbon dioxide also reflects what has
previously been found. Total hydrocarbon content (methane, ethylene and ethane) also
increases in a similar fashion with ethanol concentration. This suggests that the plasma
region in this work has similar properties to that formed by Aubry et al. (2005) even
though the reactor designs are significantly different.
In contrast to previous work, experiments with pure water as the feed have shown that
some breakdown of water occurs in the plasma. The most likely explanation is that a
small amount of water electolysis occurs within the plasma, as well as the reactions
involving ethanol.
The results are encouraging in terms of hydrogen selectivity indicating an improved
reaction mix compared with catalyst based systems and future work will focus on
improving reactor design to increase the ethanol conversion. In a review paper on nonthermal plasma reforming (Petitpas et al, 2007) there are descriptions of a number of
plasma reactor designs where the plasma occupies a much larger fraction of reactor
volume than the design in this study. Modifications to change this aspect of current
design would allow more of the feed to pass through the plasma region and should
increase ethanol conversion while retaining the improved hydrogen selectivity.
REFERENCES
1. Aubry O., Met C., Khacef A., Cormier J. M., (2005), On the Use of a nonthermal plasma reactor ethanol steam reforming, Chemical Engineering Journal,
106, 241-247.
2. Cussins, M. J., (2005) Clean Distributed Hydrogen Production : Small scale
Alcohol Reformation Technology for Renewable Alcohols, ME thesis, University
of Canterbury.
3. Kyle, W. J.,(2009) Ethanol Steam reforming Using a Non-Thermal Plasma
reactor, BE research project report, University of Canterbury.
4. Mariño, F., Boveri, M., Baronetti, G. and Laborde, M. (2001). Hydrogen
production from steam reforming of bioethanol using Cu/Ni/K/γ-Al2O3
catalysts. Effect of Ni. International Journal of Hydrogen Energy, 26(7): 665668.
5. Petitpas, G., Rolliera, J.-D., Darmonb, A. Gonzalez-Aguilara, J.,
Metkememeijera, R., Fulcheria, L. (2007), A comparative study of non-thermal
plasma assisted reforming technologies, International Journal of Hydrogen
Energy, 32, 2848-2867.
6. Sun, J., Qiu, X.-P., Wu, F. and Zhu, W.-T. (2005). H2 from steam reforming of
ethanol at low temperature over Ni/Y2O3, Ni/La2O3 and Ni/Al2O3 catalysts for
fuel-cell application., International Journal of Hydrogen Energy, 30(4): 437-445.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the efforts of the final year Chemical and
Process Engineering students, Daniel Sheely, Alex Watson and Will Kyle, who carried
out the experimentation for this project. We would also like to acknowledge the work of
9
C. Williamson, A Gardiner & E Pilbrow
Steven Broome who made a major contribution to the development of the plasma
reactor system while he worked for IRL.
BIOGRAPHY OF PRESENTER
Chris Williamson is a Senior Lecturer in Chemical and Process Engineering at the
University of Canterbury. After graduation Chris spent eight years working as an
industrial chemical engineer before returning to university to complete a PhD. After
completing his doctorate he joined the academic staff at Canterbury. During his time
there he has lectured on thermodynamics, process control and process engineering
design. Since 2004 Chris has been involved in renewable energy research projects
including gasification of biomass and the production of fuel cell grade hydrogen from
alcohols.
10