Ind. Eng. Chem. Res.

1997, 36, 1507-1518 1507

Biomass to Hydrogen via Fast Pyrolysis and Catalytic Steam

Reforming of the Pyrolysis Oil or Its Fractions
D. Wang, S. Czernik, D. Montané,† M. Mann, and E. Chornet*,‡
National Renewable Energy Laboratory, 1617 Cole Boulevard, Golden, Colorado 80401-3393

Pyrolysis of lignocellulosic biomass and reforming of the pyroligneous oils are being studied as
a strategy for producing hydrogen. A process of this nature has the potential to be cost
competitive with conventional means of producing hydrogen. We propose a regionalized system
of hydrogen production, where small- and medium-sized pyrolysis units (<500 Mg/day) provide
bio-oil to a central reforming unit to be catalytically converted to H2 and CO2. Thermodynamic
modeling of the major constituents of the bio-oil has shown that reforming is possible within a
wide range of temperatures and steam-to-carbon ratios. In addition, screening tests aimed at
catalytic reforming of model compounds to hydrogen using Ni-based catalysts have achieved
essentially complete conversion to H2. Existing data on the catalytic reforming of oxygenates
have been studied to guide catalyst selection. A process diagram for the pyrolysis and reforming
operations is discussed, as are initial production cost estimates. A window of opportunity clearly
exists if the bio-oil is first refined to yield valuable oxygenates so that only a residual fraction
is used for hydrogen production.

Introduction (poplar, CH1.47O0.67). The organic fraction of the biocrude

Steam reforming of hydrocarbons, partial oxidation
of heavy oil residues, and gasification of solid fuels to
yield syngas (a mixture of H2 and CO), followed by
water-gas shift conversion to produce H2 and CO2, are
well-established processes (Patil, 1987). Although re-
newable lignocellulosic biomass has been considered as
a potential feedstock for gasification to produce syngas,
the economics of current processes favor the use of
hydrocarbons (natural gas, C2-C5, and naphtha) and
inexpensive coal. An alternative approach to the pro-
duction of H2 from biomass begins with fast pyrolysis
of biomass.
Recent advances in the understanding of pyrolysis
mechanisms and in the technology required to achieve
high heat-transfer rates to particles, while controlling
secondary cracking reactions responsible for gas and
char production, have led to the development of fast
pyrolysis. Thus, conversion of biomass materials in high
yields (∼75 wt % of dry biomass) into a “pyrolytic oil”,
also known as “bio-oil” or “biocrude”, has been achieved.
The bio-oil contains numerous oxygenated organic
compounds (Elliott, 1988; Fagernas, 1995): acids, alco-
hols, aldehydes, ketones, furans, substituted phenolics,
and complex oxygenates derived from biomass carbo-
hydrates and lignin. About 15 wt % of the biocrude is
water from dehydration reactions. The water is very
well dispersed in the organic fraction of the bio-oil owing
to the hydrophilicity of the carbohydrate-derived com-
pounds that constitute the major fraction of the bio-oil.
Very little ash and char (a few ppm) are present in the
biocrude when appropriate filtration technology is used
in the pyrolysis process (Diebold et al., 1995).
Fluid-bed fast pyrolysis (Radlein et al., 1991) has been
used to obtain a 76 wt % yield of biocrude from wood
* Author to whom all correspondence should be addressed.
Phone: (303) 384-6240. Fax: (303) 384-6103. E-mail: chornete@
tcplink.nrel.gov.
† On leave from Department d’Enginyeria Quı́mica, Uni-
versitat Rovira i Virgili, Carretera de Salou, s/n, 43006
Tarragona, Spain.
‡ Also affiliated with Université de Sherbrooke, Sherbrooke,
Québec, Canada J1K-2R1.
S0888-5885(96)00396-X CCC: $14.00
represents 85 wt %; its elemental composition is
CH1.33O0.53. Table 1 shows the expected yields of
hydrogen from two thermochemical routes (pyrolysis
and gasification), compared to the stoichiometric maxi-
mum yield of the reaction between wood (poplar) and
steam using externally supplied heat.
Calculations made using current yields for nonopti-
mized fast pyrolysis (Radlein et al., 1991) and gasifica-
tion (Probstein and Hicks, 1982) processes show that
the H2 production potential for these two routes is
similar. However, the pyrolysis process is less severe
and does not require an oxygen supply as do some
gasifiers. Both options require steam for reforming (fast
pyrolysis) or shift conversion (gasification).
The objectives of this paper are as follows: (1)
determination of suitable ranges of temperature and
steam-to-carbon ratios that can reform the pyrolytic oils
to H2 and CO2 by thermodynamic simulations; (2)
assessment of the scientific feasibility of catalytically
converting the oxygen-rich pyrolytic oil compounds and
identification of suitable catalysts and conditions to be
used; and (3) generation of a tentative process flow
diagram and analysis of its economic feasibility.
Fast Pyrolysis: Bio-oil Yields and Composition
Pyrolysis of carbonaceous solid materials is a thermal
decomposition process occurring in an inert atmosphere.
For centuries, slow pyrolysis of biomass has produced
charcoal, and a large variety of reactor configurations
have been developed, which are well documented in
generic handbooks (Kirk-Othmer, 1991; Uhlman, 1989).
The thermal decomposition of carbohydrates and
lignin takes place through a sequence of dehydration
and depolymerization reactions (Antal, 1982; Shafiza-
deh, 1982; Richards, 1987; Elliott, 1988). A variety of
oligomeric and low molecular weight reactive intermedi-
ates are formed during thermal decomposition. In
terms of reaction kinetics, the heat- and mass-transfer
rates compete with diffusional and reaction rates.
Therefore, reaction intermediates can be driven either
toward charcoal or, after limited further cracking,
toward lower molecular weight products. These latter
© 1997 American Chemical Society
1508 Ind. Eng. Chem. Res., Vol. 36, No. 5, 1997

Table 1. Comparison of H2 Yields from Three Biomass Process Routes

yield of H2 energy content (HHV) in H2/
process route (% wt biomass) energy content of biomass (%)
pyrolysis + catalytic reforming 12.6 91
gasification + shift conversion 11.5 83
biomass + steam + external heat 17.1 124
(theoretical maximum)

Table 2. Composition of Pyrolytic Oils Derived from

products can be separated as volatile vapors from a Different Feedstocks by Two Processes (Yields in wt %
limited charcoal (also known as char) residue (Antal, of Dry Biomass)
1982). High yields of these volatiles can be achieved
fluidized bed Vortex
by a proper choice of reactor configuration, particle size, (University of Waterloo)a (NREL)b
fluid dynamics, and heat-transfer rates, as shown below.
poplar maple spruce oak
Advances in petroleum cracking technology, particu- product (504 °C) (508 °C) (500 °C) (∼500 °C)
larly the cracking of heavy residues, indicate that acetic acid 5.4 5.8 3.9 5.0
similar approaches could be applied to the pyrolysis of formic acid 3.1 6.4 7.2 3.3
biomass materials in order to maximize the production hydroxyacetaldehyde 10.0 7.6 7.7 4.3
of low molecular weight volatile intermediates. Such glyoxal 2.2 1.8 2.5 3.0
methylglyoxal c 0.65 c c
approaches are known as “fast pyrolysis”. Several formaldehyde c 1.2 c 2.2
technologies at the pilot and demonstration levels acetol 1.4 1.2 1.2 1.8
appear to be capable of achieving high yields (75-80 ethylene glycol 1.1 0.6 0.9 c
wt %) of bio-oil at appropriate temperatures, residence levoglucosan 3.0 2.8 4.0 3.8
1,6-anhydroglucofuranose 2.4 c c c
times, and high heat-transfer rates. They include fluid fructose 1.3 1.5 2.3 c
beds (Scott et al., 1985; Piskorz et al., 1988), transport xylose c c c 0.9
reactors (Graham et al., 1988; Kovac et al., 1988), and glucose 0.4 0.6 1.0 c
cyclonic reactors (Diebold and Scahill, 1988; Czernik et cellobiosan 1.3 1.6 2.5 c
oligosaccharides 0.7 c c c
al., 1995). The common feature of all these technologies pyrolytic lignind 16.2 20.9 20.6 24.9
is rapid heating of biomass particles to temperatures unidentified 11.9 17.1 12.9 5.8
between 450 and 550 °C and short residence times of oile 65.8 67.9 66.5 55.3
the volatile vapors in the reaction chamber (a fraction watere 12.2 9.8 11.6 10.4
char 7.7 13.7 12.2 12.4
of a second to a few seconds depending on fluid dynam- gas 10.8 9.8 7.8 12.2
ics). A fourth technology, vacuum pyrolysis (Roy et al., mass closure 96.5 101.2 97.7 90.3
1990), has also produced high biocrude yields using slow a From Piskorz et al., 1988. b Private communication. c Not
heating rates under vacuum conditions. This is possible
found. d Material precipitated by addition of water. e Oil + water
because the low molecular weight intermediates are ) bio-oil.
rapidly removed from the reaction zone before they
undergo further recondensation reactions. The organic Oxygenated bio-oil is an appropriate feedstock for
fraction produced by this process contains 50-60 wt % further catalytic steam reforming to H2 and CO2 using
of moisture-free biomass and 13-16 wt % of water. a conversion strategy paralleling that of catalytic steam
Reliable values for oil, water, char, and gas yields are reforming of alcohols and oxygenated aromatics (see
difficult to obtain because little published work closes later discussions). Optimization of pyrolysis conditions
or reports material balances. The product yields and to obtain high yields of bio-oil having a significant
composition depend on the feedstock as well as on the fraction of low molecular weight aldehydes and alcohols,
process type and conditions (temperature and vapor which are easier to reform than the lignin-derived
residence time). The quantitative composition of bio- aromatic phenolics, is one of the approaches discussed
oil has been reported from fluidized-bed studies con- in this paper.
ducted at the University of Waterloo (Scott et al., 1985,
Piskorz et al., 1988), among others. It is shown in Table
Thermodynamics
2 for several feedstocks, as is the composition of the bio-
oil derived from the cyclonic (vortex type) reactor located Thermodynamic simulations of the reforming reaction
at NREL. have determined equilibrium yields as a function of
Variations in the composition of pyrolysis oil as shown temperature and steam-to-carbon ratios (S/C). These
in Table 2 should be expected as a function of raw simulations were performed on ASPEN Plus using
material, pyrolysis treatment severity (temperature, compounds representative of bio-oil. Figure 1 shows the
residence time, and heating rate profiles), and presence approach to maximum stoichiometric yields of hydrogen
or absence of catalysts during the pyrolytic step. How- as a function of S/C at a temperature of 700 °C, which
ever, the information presented in Table 2 indicates that is in the lower end of temperatures used in current
the bio-oil is essentially a mixture of carboxylic acids hydrocarbon reformers. As expected, the water-gas
(mainly acetic and formic), numerous aldehydes and shift reaction plays a major role in the overall steam
alcohols, and a significant fraction of lignin (denoted as reforming reaction of oxygenates; hydrogen yield im-
pyrolytic lignin), present as a low to medium molecular proves significantly on increasing S/C. From such
weight material that precipitates by adding water. information we can establish, for a typical bio-oil,
Based on the mechanistic works of Antal (1982), Shafiza- theoretical maximum H2 yields under various operating
deh (1982), and Richards (1987), the unidentified com- conditions (Table 3). This table suggests that a wide
pounds contain a large number of carbohydrate-derived range of combinations of the two key operating variables
components. Monomeric lignin-derived products having can be chosen to approach stoichiometric yields.
alcohol characteristics may also be present (Elliott, The stoichiometric and equilibrium yields of hydrogen
1988). from steam reforming of several biomass types and
 Ind. Eng. Chem. Res., Vol. 36, No. 5, 1997 1509
Table 4. Hydrogen Yields from Steam Reforming
Reactions
stoichiometric equilibrium
H2 yield H2 yield
% % %
feedstock formula molesa (by wt)b (of st.)c (+WGS)c,d
methane CH4 4.00 50.3 85.6 95.0
methanol CH4O 3.00 18.9 87.1 96.8
ethanol C 2H 6O 3.00 26.3 85.9 96.2
acetone C3H6O 2.67 27.8 85.2 96.2
dimethylfuran C6H8O 2.50 31.5 84.5 96.2
anisole, cresol C7H8O 2.43 31.7 84.3 96.2
methylfuran C5H6O 2.40 29.5 84.5 96.3
ADPe C11H14O3 2.36 27.0 84.7 96.4
phenol C6H6O 2.33 30.0 84.3 96.3
guaiacol C7H8O2 2.29 26.0 84.6 96.5
furan C4H4O 2.25 23.5 84.5 96.5
syringol C8H10O3 2.25 23.5 84.8 96.6
lignin C7H9O3 2.21 22.1 84.9 96.7
furfuryl alcohol C5H6O2 2.20 22.6 84.9 96.7
Figure 1. Thermodynamic equilibrium analysis showing the catechol C6H6O2 2.17 23.8 84.6 96.7
effect of steam-to-carbon atom molar ratio on the yield of hydrogen poplar oile C6H7.98O3.18 2.14 19.7 85.2 96.9
at selected reforming and shift conversion conditions. Feed: aspen C6H8.79O3.62 2.13 18.5 85.4 97.0
syringol (O), HAA (4), and furfural (0). poplar C6H8.82O4.02 2.07 17.2 85.5 97.1
oakwood oile C6H9.14O4.53 2.01 15.8 85.6 97.3
Table 3. Percent of Stoichiometric Yield of H2 at furfural C5H4O2 2.00 21.0 84.6 96.9
Equilibrium from Steam Reforming of a Typical Bio-oila 5-HMFe C 6H 6O 3 2.00 19.2 84.9 97.0
xylan C5H8O4 2.00 15.3 85.8 97.3
% yield of H2 cellulose [C6H10O5]n 2.00 14.9 85.8 97.3
operating
temp (°C) S/Cb )5 S/Cb ) 10 glucose C6H12O6 2.00 13.4 86.3 97.5
HAc, HAAe C 2H 4O 2 2.00 13.4 86.3 97.5
500 57.0 76.3 formic acid CH2O2 1.00 4.4 87.7 98.9
600 76.9 90.4
a Moles of H produced per mole of carbon in the reactant being
700 84.6 92.8 2
800 84.7 91.7 reformed. b Amount of H2 formed divided by the sample molecular
900 82.8 89.9 weight. c The equilibrium moles of H2 predicted at 750 °C and S/C
) 5 divided by the stoichiometric yield. d With additional moles
a Using water-free poplar oil with the formula of CH
1.33O0.53. of CO present under 750 °C and S/C ) 5 to be shifted in a
b Steam-to-carbon ratio. downstream water-gas shift (WGS) reactor. e ADP: 4-allyl-2,6-
dimethoxyphenol. 5-HMF: 5-(hydroxymethyl)furfural. HAc: ace-
related materials and model compounds are listed in tic acid. HAA: hydroxyacetaldehyde. Water-free bio-oils (from
Table 4. Also shown are percentage yields by weight poplar and oakwood) are used.
(grams of hydrogen produced per gram of the oxygenate
feed being reformed). In general, lignin-derived phe- Catalytic Reforming of Oxygenates
nolics produce more hydrogen than the acids, alcohols, General Concepts. Steam reforming C1-C5 hydro-
aldehydes, and ketones that originate from the carbo- carbons, naphtha, gas oils, and simple aromatics are
hydrates in biomass. This is because, for an oxygenated commercially practiced, well-known processes. When
compound with a chemical formula of CnHmOk, the the objective is to maximize the production of H2, the
stoichiometric yield of hydrogen is 2 + m/2n - k/n moles stoichiometry describing the overall process is
per mole of carbon in the feed (eq 1). The term k/n is
CnHm + 2nH2O ) nCO2 + (2n + m/2)H2 (2)
CnHmOk + (2n - k)H2O ) nCO2 + (2n + m/2 - k)H2
(1) The simplicity of eq 2 hides the fact that, in a
hydrocarbon reformer, the following reactions take place
usually well below 1 for the aromatic phenolics from concurrently:
lignin, while k/n is close to 1 for most carbohydrate-
derived products such as sugars. CnHm + nH2O ) nCO + (n + m/2)H2 (3)
Thermodynamic calculations indicate that the equi-
librium yield of hydrogen (as a percentage of the CH4 + H2O ) CO + 3H2 (4)
stoichiometric yield) under typical steam reforming
conditions (750 °C and S/C ) 5) does not vary much for CO + H2O ) CO2 + H2 (5)
these oxygenates and averages around 85%. The small
variation in this yield reflects the true ratio (5 + k/n) of At normal operating conditions, reforming of higher
steam to carbon in the catalytic reformer: external hydrocarbons (eq 3) is irreversible (Tottrup and Nielsen,
steam supplied (S/C ) 5) plus that derived from the feed 1982), whereas the methane reforming reaction (eq 4)
(S/C ) k/n). A further water-gas shift operation to and the shift conversion reaction (eq 5) approach equi-
convert all the CO formed in the reformer would librium. A large molar ratio of steam to hydrocarbon
improve the overall yield to about 97%, with the rest will ensure that the equilibrium for reactions (4) and
hidden mostly in “carbon” (here assumed to be graphite, (5) is shifted toward H2 production.
the best alternative available in the absence of any Choice of Catalyst. The main problem in the steam
thermodynamic data on actual carbonaceous materials reforming of hydrocarbons is carbon formation. Three
formed). If the formation of carbon could be avoided, types of carbon may be formed: whiskers, encapsulating
the conversion of oxygenates to hydrogen would be deposits (“polymers”), and pyrolytic carbon (Rostrup-
almost quantitative through the combination of re- Nielsen and Tottrup, 1979; Bartholomew, 1982). Car-
forming and water-gas shift. bon formation is related to the kinetics, and hence the
1510 Ind. Eng. Chem. Res., Vol. 36, No. 5, 1997
selectivity, of the catalyst. In steam reforming, adsorp-
tion of hydrocarbons on the metal sites (Ni is commonly
used) is followed by dehydrogenation and cracking. This
results in smaller CxHy adsorbed fragments which can
polymerize to unwanted intermediates and form coke.
These reactions compete with reactions with adsorbed
water species. The catalyst must supply enough ad-
sorbed H2O-derived species, mainly OH and H, to
overcome the coke formation reactions. This implies
that the adsorbed OH and H must have sufficient
surface mobility to reach the adsorbed hydrocarbon-
derived species and react with them (Rostrup-Nielsen,
1984). The steam adsorption capability can be signifi-
cantly improved by promoting the classical Ni-alumina
or Ni-silica/alumina formulations with Ca and/or K. An
alternative is the use of magnesia as a magnesium-
aluminum spinel which enhances steam adsorption and
can be prepared as high-surface-area refractory ceramic
pellets.
The reforming operation requires relatively high
steam-to-carbon molar ratios (S/C ) 3.5-5.0 for meth-
ane), adequate temperature profiles in the tubular
reactor (low inlet temperature increased along the tube
length to a maximum of 750-775 °C owing to the
structural constraints of the Ni-Cr alloys used), and a
Ni-based catalyst (20-30 wt % as NiO) on a refractory
support having appropriate steam adsorption charac-
teristics as previously discussed. The reforming opera-
tion is not very sensitive to pressure, which is essen-
tially dictated by the applications of the product H2
(Patil, 1987). The NiO is reduced to its metal form prior
to use. Space velocities (GC1HSV, defined as C1-
equivalent volume of feed per hour per unit volume of
catalyst) used in tubular reformers typically lie between
1500 and 2000 h-1.
Sulfur is the most severe catalyst poison (Ridler and
Twigg, 1989). Biomass-derived feedstocks, i.e., bio-oils,
have only traces of sulfur. For example, bio-oils made
by fast pyrolysis of hybrid poplar have a sulfur content
in the range of 0.02-0.04% (Scahill et al., 1996), similar
to that in the hybrid poplar itself (0.03%). The sulfur
content in biomass, in general, ranges from less than
0.01% (woody types) to about 0.2% (herbaceous types).
For comparison, some naphthas contain as much as
1.5% sulfur. Therefore, a desulfurization unit prior to
the reformer may not be necessary for bio-oil feedstocks.
However, because of the oxygenated nature of the
pyrolysis oil, the catalyst to be used in the reforming
process must be able to handle high percentages of
oxygen-containing functional groups.
Efforts to modify steam reforming catalysts and
render them less susceptible to poisoning by heteroat-
oms, mainly sulfur, have led to the development of novel
preparations. Thus, sulfur-resistant rare-earth-pro-
moted catalysts for gas and oil feeds have been available
since the mid-1980s (Patil, 1987). The unreduced
catalyst form is a mixture of NiO, rare-earth oxides, and
Zr2O3. The applicability of this family of catalysts to
the reforming of oxygenates is not known, but one would
expect that sulfur-resistant catalysts should also toler-
ate oxygen. Thus, such preparations are attractive for
the reforming of oxygen-containing molecules.
Aznar et al. (1993) have demonstrated that steam re-
forming of syngas obtained from the gasification of
lignocellulosics is possible using commercially available
Ni-based catalysts (Topsoe R-67 and RKS-1). However,
the catalyst rapidly deactivates after only a few hours.
Tars produced during gasification contain a high per-
centage of polycyclic aromatics, while bio-oil is es-
sentially a mixture of carbohydrate-derived fragments
and lignin-derived phenolics (Elliott, 1988). In order to
develop a successful process for steam reforming bio-
oil, we must have a catalyst that can maintain a high
activity much longer than just several hours.
Oxygenate Reforming Literature. Literature on
the steam reforming of oxygenates falls within two
categories: reforming simple alcohols, such as methanol
and ethanol, and reforming oxygenated aromatics, i.e.,
cresols. Methanol reforming is of interest as a process
for the production of H2 in numerous applications, with
the hydrogenation of organic compounds and fuel cells
being the most obvious. A large body of information has
thus been accumulated on the catalytic reforming of
methanol. The comparison with methane is thus ap-
propriate. Methane requires high temperatures (>500
°C) to be effectively steam reformed. The following
mechanism has been proposed (Ross et al., 1978): (1)
methane, or any other hydrocarbon, is dissociatively
adsorbed on the metal sites; (2) H2O is also dissocia-
tively adsorbed on the Al sites, hydroxylating the
surfaces; (3) metal-catalyzed dehydrogenation takes
place, creating adsorbed hydrocarbon-derived frag-
ments; (4) the OH surface groups migrate to the metal
sites, activated by the temperature, and they eventually
form intermediates leading to carbon oxides.
Methanol can be steam reformed at much lower
temperatures (<300 °C) than methane. A variety of
catalysts are capable of carrying out the conversion
including reduced NiO-based preparations (Mizuno et
al., 1986), reduced CuO/ZnO shift preparations (exten-
sive literature cited by Amphlett et al. (1985) and Su
and Rei (1991)), Cu/SiO2 and Pd/SiO2 (Takahashi et al.,
1982), Cu/ZrO2 (Takezawa et al., 1987), and Pd/ZnO
(Iwasa et al., 1993). The product slate is markedly
affected by the type of catalyst and the experimental
conditions used. Two main mechanisms have been
proposed (Jiang et al., 1993a,b): (1) the decomposition-
shift mechanism where the chemisorbed methanol is
totally dehydrogenated on the metal surface prior to
forming CO (the shift reaction then proceeds with the
adsorbed water-derived species); (2) the formate mech-
anism initially proposed by Takahashi et al. (1982) and
based on the formation of methyl formate ester on the
catalyst surface between two adsorbed species, H3CO
and HCO. The ester is hydrolyzed to methanol and
formic acid. The latter decomposes directly to CO2 and
H2 with 100% selectivity, i.e., no CO is formed.
Ethanol reforming has been studied to understand the
formation of acetic acid and H2. Iwasa and Takezawa
(1991) have shown that, in the formation of acetic acid,
acetaldehyde is the key intermediate when Cu-based
catalysts are used. They found also that the role of the
support is important. Cu alone shows very high selec-
tivity toward acetaldehyde (88% at 250 °C), whereas the
presence of a support (Al2O3, SiO2, ZrO2, or ZnO)
markedly affects this selectivity and produces several
other intermediates at significant concentrations. Gar-
cia and Laborde (1991) conducted a thermodynamic
study on the reforming of ethanol to H2. They concluded
that H2 production requires temperatures higher than
those needed for the reforming of methanol and that
steam-to-ethanol ratios higher than 10 (equivalent
steam-to-carbon ratios higher than 5) were required to
prevent carbon formation.
Steam reforming of oxygenated aromatics has been
directed toward the dealkylation of cresol. The reaction

proceeds at temperatures of 400-500 °C and steam-to-
carbon ratios of 6 using Pd, Rh or Ni on alumina, silica,
or chromia-alumina. The predominant reaction is
degradation to H2 and CO (Daly, 1980; Duprez, 1992).
Assafi and Duprez (1988) proposed a mechanism in
which the OH group attached to the aromatic ring
participates in the overall reaction in the same manner
as the OH group derived from the dissociation of water.
Most importantly, under the conditions used, the aro-
matic ring is “cracked” and thus total reforming may
indeed be feasible.
During steam reforming of oxygenated aromatics,
dehydroxylation proceeds at a significant rate to produce
aromatic species on the catalyst surface that are sub-
sequently steam reformed. Duprez (1992) discusses the
reforming of mixtures of aliphatic and aromatic hydro-
carbons using a combination of Ni and V at 850 °C.
Work conducted at much lower temperatures, 440-480
°C, and at steam-to-compound ratios lower than 10 (or
S/C close to 1 for complex aromatics), on Rh-V on
alumina (Kim, 1978) and on Rh, Pt, and Ni on alumina
and silica (Delahay et al., 1985) showed that with
methylnaphthalene as the feed, partial degradation to
CO and H2 takes place in competition with ring opening
and dealkylation. Delahay and Duprez (1985) con-
cluded that intermediate hydrogenation favors, but is
not a prerequisite for, ring opening.
The bio-oil compositions shown in Table 2 make it
clear that challenges in catalytic steam reforming will
be linked to the ability to handle the “pyrolytic lignin”
fraction and the carbohydrate-derived fraction: acetic
acid, aldehydes, alcohols (sugars and anhydrosugars),
and oligomers. It is likely that the oligomeric frag-
ments, either lignin- or carbohydrate-derived, once
condensed as bio-oil cannot be revolatilized without
significant char formation. Thus, these options are
possible. We separate the pyrolytic lignin from the
condensed bio-oil and steam reform the aqueous frac-
tion. Pyrolytic lignin has a potential market of its own
as a phenolics substitute for the production of adhesives
(Chum et al., 1993). The aqueous fractionswhile still
containing some substituted, but monomeric, phenolicss
will be constituted predominantly of the carbohydrate-
derived fractions. We may carry out a fractional
condensation of the pyrolytic vapors to separate the
oligomers and thus directly recover a “light” fraction
composed essentially of mixed monomers, which is then
steam reformed, or we may try to reform all uncon-
densed vapors by coupling the reformer to the pyrolysis
unit.
Given the heterogeneity of the aqueous fraction, Ni-
based catalysts operating in the 700-800 °C range,
according to the literature reviewed above, ought to be
able to carry out the steam reforming conversion. A
temperature profile through the reactor that controls
coke formation will have to be developed experimentally.
In the next section we present and discuss our screening
results using model compounds characteristic of both
the carbohydrates and the lignin, and pyrolysis vapors
of biomass and its three major components (cellulose,
xylan, and lignin) to determine whether available
catalysts can effectively carry out the reforming opera-
tion.
Catalyst Screening Tests
Experimental Procedures. The catalyst screening
work was carried out in a vertical, dual-bed quartz
reactor interfaced with a molecular beam mass spec-
Ind. Eng. Chem. Res., Vol. 36, No. 5, 1997 1511
trometer (MBMS), as shown in Figure 2, for real-time
analysis of products (Evans and Milne, 1987). This
system has the advantage of simultaneously detecting
unconverted reactants, intermediates, and products at
the exit of the microreactor by rapidly sampling through
an orifice. A molecular beam is formed as the result of
the supersonic, free-jet expansion through the orifice,
which quenches all reaction products. These product
molecules are then ionized and analyzed by a quadru-
pole mass spectrometer.
The microreactor was housed in a tubular furnace
with four independently controlled temperature zones
(Wang et al., 1996). The dual-bed configuration of this
reactor enabled us to perform comparative experiments
investigating thermolysis vs catalysis and catalyst
performances. An outer flow was allowed for calibration
purposes as well as for the dilution of samples with
helium gas to obtain an adequate molecular beam and
high signal-to-noise ratios. Steam was generated in situ
in the reactor by vaporizing water injected with syringe
pumps through the bottom inlet or a special injection
inlet in the side arm. Similarly, liquid model com-
pounds were introduced into the reactor. The vapors
were swept by the preheated carrier gas. Solid samples
were fed in batches through the side inlets and pyro-
lyzed in a preheated mixture of helium and steam. All
compounds were obtained from commercial sources and
were used as supplied without further purification. In
particular, they included cellulose (microcrystalline,
from Sigma), xylan (extracted from larch wood, from
Aldrich), lignin (NREL sample made from steam explo-
sion of yellow poplar wood), levoglucosan (Senn Chemi-
cals), and aspen wood (ASTM standard materials).
All catalysts (Table 5) were ground and sieved before
testing and particles in the size range of 250-355 µm
(+60/-45 mesh) were used. All nickel-based catalysts
were reduced at 600 °C and the Cu/Zn shift catalyst was
reduced at 350 °C, all with 20% hydrogen in helium for
an extended period (more than 24 h). The desired
amount of the reduced catalyst was weighed out and
either used directly or mixed with quartz chips of the
same particle size if the amount of catalyst was low (less
than 0.25 g). Quartz wool was placed at both ends of
the bed to hold the catalyst in the tube. A K-type
thermocouple was placed just inside the catalyst top of
the bed (Figure 2) for monitoring catalyst temperature.
The reactor was then heated to about 20 °C above the
desired operating temperature, and the catalyst was
reduced again with 50% hydrogen in helium for 2 h. The
pressure drop through the catalyst bed was typically
3-21 kPa at the highest reactant flow rates and catalyst
temperatures. Thermal cracking reactions were carried
out in either a blank quartz tube or one packed with
quartz chips.
Mass spectra in the range of m/z 1-350 were
obtained for pyrolysis and catalysis products with the
MBMS instrument under 25 eV electron impact ioniza-
tion conditions. The detection limit of the MBMS
corresponded to conversions of approximately 99.95%
for most of the samples, i.e., from the sample signal
without catalyst at more than 1.0 × 106 counts/s (cps)
to less than the instrument’s lower limit of 500 cps
(noise ) (250 cps) with catalyst.
Although quantitative measurements using MBMS
are relatively new, we have succeeded in identifying all
significant products and quantifying them in several
cases (Wang et al., 1994; Elam et al., 1994; Wang et
al., 1996). Most of the present work belongs to such
1512 Ind. Eng. Chem. Res., Vol. 36, No. 5, 1997

Figure 2. Schematic of the microreactor and molecular beam mass spectrometer (MBMS) setup used for catalyst screening studies.

Table 5. Catalysts Used in the Screening Tests

manufacturer
ID ref name (catalyst shape and size in mm)
A S.-U. Sherbrooke Univ.
(pellets; o.d. ) 2,
length ) 2-5)
B UCI G-90C United Catalysts Inc.
(rings; o.d. ) 16,
i.d. ) 8, length ) 8)
C ICI 25-4M Imperial Chemical Industries
(4-hole shape; o.d. ) 11.5,
length ) 13.8,
hole diam. ) 3.0)
D ICI 46-1 Imperial Chemical Industries
(ring pellets; o.d. ) 16,
i.d. ) 8, length ) 16)
E UCI G-90B United Catalysts Inc.
(tablets; 6.4 × 6.4)
F UCI C18HC United Catalysts Inc.
(tablets; 4.8 × 2.4)
G BASF G1-25S BASF (rings; o.d. ) 16,
i.d. ) 9.5, length ) 9.5)
H ICI 46-4 Imperial Chemical Industries
(rings; o.d. ) 16, i.d. ) 8,
length ) 16)
I UCI G-91 United Catalysts Inc.
(rings; o.d. ) 16, i.d. ) 8,
length ) 8)
J Ni-Mg Univ. of Zaragoza
(fluidizable particles)
a NA ) not available.
composition
NiO, Al2O3, Cr2O3, La2O3, MgO
15% Ni, ceramic carrier
(70-76% Al2O3 5-8% CaO)
15% NiO, 70% Al2O3, 13% CaO,
1.5% K2O promoted
10-25% NiO refractory carrier,
1.5% K2O promoted
11% Ni, ceramic carrier (76-82%
Al2O3, 6-9% CaO)
42% CuO, 47% ZnO, 10% Al2O3
15% Ni, ceramic carrier
NiO, CaO, Al2O3, ZrO2
11% Ni, ceramic carrier (76-82%
Al2O3, 6-9% CaO), K2O promoted
NiO, MgO, Al2O3
commercial usea
NA (research catalyst for steam
reforming high MW aromatics)
steam reforming in difficult service
conditions (low S/C ratios, C3 + C4
hydrocarbons etc.)
steam reforming natural gas and
other low MW hydrocarbons
steam reforming high MW naphtha
(bp up to 235 °C, containing up to
25% aromatics)
steam reforming in difficult service
conditions (low S/C ratios, C3 + C4
hydrocarbons etc.)
low-temperature (200-260 °C) shift
conversion catalyst
steam reforming natural gas (high
activity & mechanical strength,
easy reduction)
steam reforming: used with ICI 46-1
for removing C2/C3 intermediates
steam reforming in difficult service
(low S/C ratios, C3 + C4 hydrocarbons
etc.)
NA (research catalyst for
pyrogasification of biomass)

cases. The main products formed in complete steam re- of incomplete conversions, the main products are ac-
forming reactions studied in this work are easily identi- companied by unconverted starting material as well as
fied: H2, CH4, CO, CO2, and unreacted H2O. In cases other products resulting from thermal and/or catalytic

Table 6. Product Yields in Catalytic Steam Reforming of Model Compounds, Biomass, and Related Materialsa
sample GC1HSV (h-1)
methanol 340
acetic acid 180
HAA (20 wt % in H2O) 180
phenol (20 wt % in MeOH) 290
syringol (20 wt % in MeOH) 290
ADP (41 wt % in MeOH)c 1680
levoglucosand 670
cellulosed 670
xyland 670
lignind 1350
aspend 1350
a Averaged results from triplicates. Reaction conditions were 600 °C, t ) 0.1 s, S/C ) 10-13. Catalyst used: UCI G-90C. b Limited by
the detection capability of the MBMS instrument. Also see discussion in text. c S/C ) 4.5. d Samples pyrolyzed in batches of 5-10 mg at
600 °C; the residence time before reaching the catalyst bed was about 0.5 s.
decomposition. However, because other molecules
present in the steam reforming reactions interfere very
little with the molecular ion of hydrogen at m/z 2,
hydrogen can be effectively quantified.
Hydrogen was calibrated at several known levels of
flow rates to cover the predicted range of hydrogen
production. Pure hydrogen gas was metered by a mass
flow controller (Tylan, Model FC280, 0-100 mLSTP‚min-1)
into the outer flow of the reactor under operating
conditions. The calibration was repeated as often as
necessary throughout the day; at least one level of
calibration was repeated when the reactor temperature
was changed. A constant flow (10 mLSTP‚min-1) of
argon (m/z 40) was used as an internal standard to
correct for small changes in the total flow rate of gases
exiting the reactor. The linear regression line fit of
intensity at m/z 2 versus flow rate was used to predict
the flow rate of hydrogen produced by steam reforming
of the model oxygenates. The data were fitted with
straight lines through the origin with correlation coef-
ficients better than 0.9999. For steady-state experi-
ments using continuous feeding, the calculated hydro-
gen flow rate was first multiplied by the correction
factor obtained from the internal argon standard, and
the resulting value was used to calculate hydrogen yield
by dividing it with the stoichiometric amount. There-
fore, all yield results for hydrogen are reported here as
the percentage of theoretical maximum according to the
stoichiometry of the reaction. They are estimated to be
accurate to within (3%. Hydrogen yields measured
using methanol as the standard feed under complete
conversion conditions with excessively high S/C ratios
(>15) were consistent with this estimation. For batch-
feeding experiments where a quartz boat containing a
known amount of the solid sample was placed in a
sample holder and inserted to the pyrolysis zone of the
vertical reactor (Figure 2), the averaged signal during
the whole evolution curve and the averaged sample feed
rate were used to calculate the yield of hydrogen. Errors
associated with the latter results are estimated to be
(5%. Larger errors ((10%) are expected in cases
involving incomplete conversions.
We did not measure the yields of other major products
of steam reforming reactions, CH4 (m/z 15 and 16), CO
(m/z 28), and CO2 (m/z 44) at the microreactor scale.
However, they were estimated by assuming a complete
conversion of methanol to H2 and CO2, and a comparison
on relative yields can still be made. It should be noted
that CO•+ (m/z 28) is also a fragment ion in the EI mass
spectrum of CO2, with an intensity of 3-6% relative to
that of m/z 44, depending on the mass spectrometer
tuning. For cases involving incomplete conversions,
qualitative product mass spectra were obtained by
Ind. Eng. Chem. Res., Vol. 36, No. 5, 1997 1513
convb (%) H2 yield (%) CO2 yield (%) Im/z 28/Im/z 44 (%)
>99.95 95 93 8
>99.95 86 90 5
>99.95 96 103 5
>99.95 99 98 6
>99.95 100 100 6
99.7 69 47 12
>99.95 100 87 3
>99.95 102 98 3
>99.95 100 93 5
>99.95 41 35 3
>99.95 62 60 3
subtracting the mass spectrum of the reactant (assumed
to be unconverted following vaporization conditions and
rapid passage through the bed at low temperatures)
from that of the reactant converted by pyrolysis or re-
forming reactions. Product assignments are based on
a user library of mass spectral data collected for pure
compounds in separate experiments.
Screening of Feedstocks and Catalysts. We
reformed a series of oxygen-containing model com-
pounds, of in situ pyrolysis vapors of biomass and its
major components (lignin, cellulose, and hemicellulose),
and of vaporized pyrolysis oils of biomass produced in
a large-scale pyrolysis reactor (Diebold et al., 1995).
Methanol steam reforming was used as a standard test
for checking the performance of the system and catalyst
deactivation. Acids, ketones, alcohols, and aldehydes
were represented by acetic acid, acetone, ethylenediol,
glycerol, and hydroxyacetaldehyde. The phenolic series
included phenol, anisole, cresols, resorcinol, 2,6-dimeth-
ylphenol, guaiacol, syringol, and 4-allyl-2,6-dimethoxy-
phenol (ADP). The furan family of model compounds
consisted of furan, 2-methylfuran, 2,5-dimethylfuran,
2-furfuraldehyde, furfuryl alcohol, 5-methylfurfural, and
5-(hydroxymethyl)furfural.
Parameters for the steam reforming operation in-
cluded catalyst temperature, molar steam-to-carbon
ratio (S/C), gas hourly space velocity (GC1HSV), and
residence time (t, calculated from the void volume of the
catalyst bed divided by the total flow rate of gases at
the inlet of the reactor; void fraction ) 0.4).
Table 6 summarizes quantitative results obtained
experimentally for some of these oxygenates. These
tests were performed with the same catalyst (UCI
G-90C), as described below, under conditions of 600 °C,
S/C ) 10-13, GC1HSV ) 180-1680 h-1, and t ) 0.1 s.
Vapor residence time was similar in all experiments
because the feed flow rate was small compared to that
of the helium carrier gas and steam. Essentially
complete conversion of all oxygenates was achieved, and,
in most cases, the yields of hydrogen were close to
stoichiometry. However, ADP, lignin, and aspen feed-
stocks produced hydrogen yields far below those pre-
dicted by stoichiometry and equilibrium calculations,
because nonvolatile deposits (ash and char) formed
during vaporization or pyrolysis of the feed prior to
entering the catalyst bed. The other feedstocks formed
little or no such non-volatile deposits during vaporiza-
tion or pyrolysis. Better yields can be obtained if this
side effect is eliminated. The abundance of m/z 28
relative to that of m/z 44 (Table 6) can be used to infer
the amount of CO generated during steam reforming.
Thus, ADP produced the greatest amount of CO; other
feedstocks produced much less. No clear explanation
1514 Ind. Eng. Chem. Res., Vol. 36, No. 5, 1997

Table 7. Conversions (%) of Model Compounds by

Catalytic Steam Reforminga
ADP/MeOH
catalyst MeOH,
ID MeOH HAc HAA ADP MeOH repeat
A >99.95 99.85 >99.95 >99.95 99.7 >99.95
E >99.95 >99.95 99.8 99.6 99.7 99.92
G 99.91 99.95 >99.95 99.7 99.95 99.8
D, #1 >99.95 >99.95 >99.95 98.0 99.4 >99.95
D, #2 >99.95 >99.95 99.95 95.9 98.3 99.88
C, #1 >99.95 >99.95 >99.95 >99.95 >99.95 >99.95
C, #2 99.5 99.9 99.7 99.3 99.5 99.5
B, #1 >99.95 >99.95 >99.95 >99.95 >99.95 >99.95
B, #2 >99.95 >99.95 99.90 99.8 >99.95 >99.95
B, #3 >99.95 >99.95
a Experimental conditions: 700 °C, S/C ) 5, G HSV ) 6725
C1
h-1, and t ) 0.01 s. Conversions >99.95% correspond to the
instrument detection limit.
Figure 4. Yields of hydrogen for catalytic steam reforming of
methanol (0/9), acetic acid (O/b), hydroxyacetaldehyde (4/2), and
4-allyl-2,6-dimethoxyphenol (41 wt % in methanol) ([) at varying
temperatures. Open symbols denote conditions of GC1HSV ) 336
h-1, S/C ) 10, and t ) 0.1 s, and solid symbols, conditions of GC1-
HSV ) 1680 h-1, S/C ) 4.5, and t ) 0.08 s. The thick solid and
dashed lines represent equilibrium yields for C2H4O2 (acetic acid
and hydroxyacetaldehyde) from thermodynamic calculations, while
the thin solid and dashed lines are only visual aids.

converted HAc completely to form acetone and CO2 and

Figure 3. H2 yield for six catalysts and four model compounds
under the same steam reforming conditions (700 °C, S/C ) 5, GC1-
HSV ) 6725 h-1, and t ) 0.01 s).
can be advanced given the limited data. However, the
results indicate that reforming kinetics are competitive
for the different species and that aromatic compounds
(ADP) are harder to reform than other constituents
found in bio-oil.
We also screened various Ni-based catalysts using
four model compounds under the following conditions:
700 °C, S/C ) 5, GC1HSV ) 6725 h-1, and t ) 0.01 s.
The shift catalyst was tested at lower temperatures of
350-400 °C. Table 5 lists the physical and chemical
properties of all catalysts screened in this study. The
Ni-based catalysts included two research formulations
and samples obtained from several commercial catalyst
suppliers, with variations in loading, supports, and
promoters. The four model compounds used for the
catalyst screening were methanol (MeOH), acetic acid
(HAc), hydroxyacetaldehyde (HAA), and 4-allyl-2,6-
dimethoxyphenol (ADP). Catalyst performance was
measured by compound conversion (shown in Table 7),
the yield of hydrogen (relative to stoichiometric yield,
shown in Figure 3), and resistance to deactivation. It
should be noted that these screening tests did not
discriminate between the high activity catalysts.
The low-temperature shift catalyst (UCI C18HC,
containing CuO/ZnO/Al2O3/C) worked only for methanol.
Methanol was completely converted at typical operating
temperatures (<300 °C). The catalyst showed similar
methanol reforming activity at higher temperatures.
However, it failed to show appreciable steam reforming
activity for any of the other model compounds studied
at temperatures <300 °C. At 400 °C, this catalyst
minor amounts of ketene, but no hydrogen. Some
hydrogen formation (less than 5% of stoichiometry) was
produced from HAA at 350 and 400 °C. Other more
abundant products included a series of aldehydes and
dihydroxybenzene compounds.
Table 7 lists the percent conversions for these screen-
ing tests with six catalysts and four model compounds.
High conversions (>99%) were observed with all cata-
lysts. In the steam reforming of the ADP/MeOH
mixture, the conversion of ADP was never better than
that of methanol. The ICI 46-1 catalyst had the lowest
conversion for ADP, reaching only about 97%. The H2
yields for all catalysts and model compounds were high,
averaging 89.5% ((5.3%). Within our experimental
error limit, no one catalyst was clearly better than the
others. The yields of catalysts C and D reproduced very
well (within (2-3%), but yields of catalyst B reproduced
with a larger variance ((5.5%). There seem to be some
differences in hydrogen yields among the different
model compounds, but we do not have any explanation
for it.
Effects of Operating Conditions. These studies
were carried out by varying the following parameters
in a systematic manner: temperature, residence time,
steam-to-carbon ratio, and space velocity. Their effects
on the catalyst performance were measured by conver-
sion, yields of H2, CO2, CO, and CH4. Temperature had
the most profound effect on steam reforming reactions
(Figure 4). The three small molecules, methanol, acetic
acid, and HAA, were all reformed in high conversions
(>99.95%) at 421 °C and above. ADP was more difficult
to reform with steam, and higher temperatures (>600
°C) were required for complete conversion. At lower
temperatures of 421, 500, and 600 °C, the conversions
of ADP were 21%, 81%, and 99.7%, respectively. Cata-
lyst deactivation was indicated by decreased yields of
hydrogen obtained from repeated experiments with the
used catalyst. This deactivation probably resulted from
steam reforming operations at the low temperatures
used in these experiments, temperatures that were
much lower than the optimal values. More meaningful
studies on catalyst deactivation, lifetimes, and regen-

Figure 5. Effect of (a) residence time and (b) steam-to-carbon
ratio on hydrogen yields from catalytic steam reforming of
methanol (0), acetic acid (O), hydroxyacetaldehyde (4), and 4-allyl-
2,6-dimethoxyphenol (41 wt % in methanol) (]) at 600 °C, GC1-
HSV ) 1680 h-1, and (a) S/C ) 4.5 and (b) t ) 0.08 s. Lines are
only visual aids.
eration are underway using both model compounds and
real bio-oils in a larger bench-scale fixed-bed reactor.
Within experimental error limits, varying residence
time from 0.04 to 0.15 s and increasing S/C from 4.5 to
7.5 did not significantly affect (Figure 5) the yield of
hydrogen under the conditions of 600 °C and GC1HSV
) 1680 h-1 (UCI G-90C catalyst). These conditions
resulted in almost complete conversion of all four model
compounds (methanol, acetic acid, HAA, and ADP).
Interestingly, except for the case of HAA, methane
formation did show some dependence on residence time
and S/C (Figure 6). In the case of HAA, there was
almost no change in the amount of methane formed
upon varying residence time and S/C. However, as the
residence time increased from 0.04 to 0.15 s, acetic acid
produced less methane but both methanol and ADP/
MeOH produced significantly more (Figure 6a). As
expected, less steam also favored the formation of
methane, especially from methanol (Figure 6b). At 700
°C, there was no significant change in the yield of
hydrogen.
The microreactor-MBMS system continuously moni-
tored in real time the intermediate products present in
the gas phase and also allowed the study of the reaction
mechanism in low-conversion experiments. More de-
tailed studies on the various types of model compounds
mentioned above are underway; we have recently
published the results of thermal decomposition and
catalytic steam reforming reactions of HAc and HAA
(Wang et al., 1996). These results indicate that both
thermal decomposition and catalytic reforming reactions
take place during the steam reforming of oxygenates
found in pyrolysis oils. The decomposition involves both
thermal cracking reactions prior to entering the catalyst
bed and the acid-catalyzed reactions at the acidic sites
of the catalyst support. Only very few oxygen-contain-
ing compounds are stable enough to reach the catalyst
bed without thermal cracking. These competing reac-
tions may form carbonaceous materials (coke), which
Ind. Eng. Chem. Res., Vol. 36, No. 5, 1997 1515
Figure 6. Effect of (a) residence time and (b) steam-to-carbon
ratio (S/C) on the relative amount of methane per moles of carbon
in the feed (max. ) 100) formed from catalytic steam reforming of
methanol (9), acetic acid (b), hydroxyacetaldehyde (2), and 4-allyl-
2,6-dimethoxyphenol (41 wt % in methanol) ([) at 600 °C, GC1-
HSV ) 1680 h-1, and (a) S/C ) 4.5 and (b) t ) 0.08 s. Lines are
only visual aids.
could plug the reactor, significantly lower the yield of
hydrogen, and even deactivate the catalyst. However,
under suitable operating conditions (700-750 °C, GC1-
HSV ) 2000 h-1, S/C ) 5, and t ) 0.1 s) a highly active
catalyst such as the UCI G-90C catalyst used here can
effectively steam reform all compounds, including the
secondary cracking products, into H2 and CO2. The key
to the steam reforming process for pyrolysis oils is hence
to properly inject the oil into the reactor in order to
minimize coke buildup at the entrance of the catalyst
bed.
Process Design and Preliminary Economics
Two ways to produce hydrogen from biomass pyrolysis
oil can be envisioned:
(1) A regionalized system of distributed small- and
medium-sized pyrolysis units provides condensed bio-
oil to a central reforming unit. At the reforming plant,
the separation of the lignin-derived oligomers will
generate a “clean” aqueous stream containing simple
carbohydrate-derived soluble organics that will be fed
to the reformer.
(2) An integrated system in which a larger pyrolysis
unit processes biomass transported to it. Uncondensed
vapors will be fed directly to an integrated reforming
unit located in the same plant. The heavier fraction of
the pyrolysis vapors can be recovered by fractional
condensation and the remaining uncondensed vapors
fed to the reformer.
The first way can use cheaper feedstocks, which
include biomass residues. The integrated system, how-
ever, can avoid the costs of entirely condensing the
vapors to produce bio-oil and transporting the bio-oil to
the hydrogen facility. A conceptual process design to
make hydrogen from pyrolysis oil in a regionalized
system is shown in Figure 7. Because of the low sulfur
content of biomass and bio-oil, a sulfur removal system
is not likely to be required. Also, according to thermo-
dynamic simulations and the screening results, a tem-
1516 Ind. Eng. Chem. Res., Vol. 36, No. 5, 1997
Figure 7. Process flow diagram for the fast pyrolysis and
reforming of biomass.
perature reformer ramping up to 700-750 °C (which is
lower than 825-900 °C required for reforming natural
gas) will be needed. The ratio of steam to carbon will
be determined by experimental results and economic
optimization; it will be in the 5-7 range, based on the
literature and experimental data already obtained.
In the process being evaluated, bio-oil generated from
fluid-bed pyrolysis of biomass will be refined through a
separation step (using water and ethyl acetate) to
recover an oxyaromatic coproduct which will be used
as a phenolic substitute in resin formulations. The
remaining aqueous fraction will be catalytically steam
reformed to produce hydrogen, using a process based
on that used for natural gas reforming. Laboratory
experiments will provide the basis for the choice of the
most suitable catalyst and reactor configuration; the
base case will use a fixed-bed catalytic reactor. A
pressure-swing adsorption unit will be used to purify
the H2 produced.
A feasibility analysis was performed on this process
to determine if the process could have economic viability
and to specify areas where research will help to lower
the production cost. Both laboratory data and standard
process data, where applicable, were used. Although
this analysis is not of design quality, it does provide
useful information on this research project before scaleup
and commercialization. The capital investment of the
pyrolysis plant was taken from Beckman and Graham
(1993). Biomass was considered available at a cost of
$25/dry tonne. A 15% internal rate of return was
assumed for both the pyrolysis and reforming facilities.
The phenolics substitute coproduct was assumed to be
sold for $0.44/kg, a fraction of the selling price of phenol.
Steam is produced through heat integration and is sold
as a byproduct. The design assumes that a number of
small pyrolysis operations, using biomass where it is
available, will supply bio-oil to a central extraction and
reforming facility. Because biomass pyrolysis oil is
similar to petroleum crude oil in that many fuels and
chemicals can be derived from it, further coproduct
options can also be considered as a function of current
economic and market conditions.
The current selling price of H2 in industry is generally
comprised between $5/GJ and $14/GJ, depending on the
size of the production facility. This range is for hydro-
gen as produced by the plant: purified but not com-
pressed or stored. For our conceptual process, the cost
of hydrogen has been estimated to be $7.70/GJ for the
base case (production capacity: 35.5 tonne of hydrogen
per day). Several parameters (for instance, a lower cost
for biomass) can lower this price to $3-5/GJ (Mann et
al., 1996). The process can also sustain large changes
in coproduct selling price, capital cost, and hydrogen
production capacity before the hydrogen becomes more
expensive than current markets will allow.
Conclusions
Fast pyrolysis of biomass is an advanced technology
for producing a bio-oil in high yields (70-75 wt % of
anhydrous biomass). This bio-oil is a complex aqueous
mixture of simple aldehydes, alcohols, and acids to-
gether with more complex carbohydrate- and lignin-
derived oligomeric materials. Fractional condensation
of the pyrolytic vapors could separate the simple mon-
omeric materials, which compose about one-third to half
of the biocrude, from the complex oligomeric fraction.
Steam reforming of the simple monomeric materials is
thermodynamically and chemically feasible. Steam
reforming of the entire bio-oil is also thermodynamically
feasible. Reforming the complex oxygenates seems
chemically possible, but it may require high steam-to-
carbon ratios because the oxygenates rapidly dehy-
droxylate and aromatics are formed on the surface of
the catalyst. Carbon-carbon bonds are then ruptured
and only a large supply of OH and H species will
suppress (or at least minimize) the formation of coke.
Our screening tests have shown that such catalysis is
possible using commercially available Ni-based cata-
lysts. Improvements in their formulation may be
needed to optimize the activity, selectivity, and time-
on-stream relationships when reforming the biomass-
derived oxygenates.
The preferred implementation strategy consists of
small- to medium-sized regional fast pyrolysis units
which will produce bio-oil from either dedicated crops
or plantations or from waste lignocellulosics. The bio-
oil produced will be transported to a central reforming
unit serving a given region where H2 will be produced.
An alternative strategy consists of transporting the
biomass to a central conversion site where fast pyrolysis
and reforming operations can be conducted jointly.
Fractional condensation of the vapors can be carried out
in order to recover, for instance, the depolymerized
lignin fraction which has value as a mixture of phenolics
to be used in resin formulations for adhesives.
The process concept is simple: a desulfurization unit
is not needed, and the bio-oil can be atomized in a flow
of steam which is then processed through a reforming
unit. By a proper choice of conditions the near-equi-
librium design can be driven to maximize H2 production.
A pressure-swing adsorption unit will purify the gas
stream. The preliminary economic analysis of the
process indicates that the necessary selling price of
hydrogen is well within current market values. This
analysis is based on the coproduction of hydrogen and
a phenolic substitute for resin formulation. The process
studied considers pyrolysis oil production at distributed
locations followed by extraction and steam reforming
at a centralized facility.
Acknowledgment
The authors are indebted to the U.S. DOE Hydrogen
Program (Mr. Neil Rosemeissel and Ms. Cathy Grégoire-
Padró, managers) for financial support of this research
project under Contract DE AC 36-83CH10093.

Literature Cited
Amphlett, J. C.; Evans, M. J.; Mann, R. F.; Weir, R. D. Hydrogen
Production by the Catalytic Steam Reforming of Methanol. Part
2: Kinetics of Methanol Decomposition Using Girdler G66B
Catalyst. Can. J. Chem. Eng. 1985, 63, 605-611.
Antal, M. J. Biomass Pyrolysis: a Review of the Literature. Part
1sCarbohydrate Pyrolysis. In Advances in Solar Energy; Boer,
K. W., Duffield, J. A., Eds.; Solar Energy Society: New York,
1982; pp 61-111.
Assafi, M.; Duprez, D. Steam Dealkylation of m-Cresol on Alumina
Supported Rhodium, Nickel and Palladium Catalysts. Bull. Soc.
Chim. Fr. 1988, 106-112.
Aznar, M. P.; Corella, J.; Delgado, J.; Lahoz, J. Improved Steam
Gasification of Lignocellulosic Residues in a Fluidized Bed with
Commercial Steam Reforming Catalysts. Ind. Eng. Chem. Res.
1993, 32, 1-10.
Bartholomew, C. H. Carbon Deposition in Steam Reforming and
Methanation. Catal. Rev.sSci. Eng. 1982, 24, 67-112.
Beckman, D.; Graham, R. Economic Assessment of a Wood Fast
Pyrolysis Plant. In Advances in Thermochemical Biomass
Conversion; Bridgwater, A. V., Ed.; Blackie Academic & Profes-
sional: London, 1993; Vol. 2, pp 1314-1324.
Chum, H.; Diebold, J.; Black, S.; Kreibich, R. Resole Resin
Products Derived from Fractionated Organic and Aqueous
Condensates Made by Fast Pyrolysis of Biomass Materials. U.S.
Patent 5223601, 1993.
Czernik, S.; Scahill, J.; Diebold, J. The Production of Liquid Fuel
by Fast Pyrolysis of Biomass. J. Sol. Energy Eng. 1995, 117,
2-6.
Daly, F. Catalytic Steam Dealkylation of Alkyl Phenols. J. Catal.
1980, 61, 528-532.
Delahay, G.; Duprez, D. Catalytic Steam Reforming of Heavy
Aromatic Hydrocarbons. II. Conversion of a Variety of Bicyclic
Molecules. Comparison with the Steam Reforming of 1-Meth-
ylnaphthalene. Bull. Soc. Chim. Fr. 1985, 1245-1249.
Delahay, G.; Bousquet, J.; Duprez, D. Catalytic Steam Reforming
of Heavy Aromatic Hydrocarbons. I. Steam Reforming of 1-
Methylnaphthalene on Rhodium, Nickel and Palladium Cata-
lysts. Bull. Soc. Chim. Fr. 1985, 1237-1244.
Diebold, J.; Scahill, J. Biomass to Gasoline: Upgrading Pyrolysis
Vapors to Aromatic Gasoline with Zeolite Catalysis at Atmo-
spheric Pressure. In Pyrolysis Oils from Biomass: Producing,
Analyzing and Upgrading; Soltes, E. J., Milne, T. A., Eds.; ACS
Symposium Series 376; American Chemical Society: Washing-
ton, DC, 1988; pp 31-40.
Diebold, J.; Czernik, S.; Scahill, J.; Philip, S. D.; Feik, C. J. Hot
Gas Filtration to Remove Char from Pyrolysis Vapors Produced
in the Vortex Reactor at NREL. Proceedings of Biomass Py-
rolysis Oil Properties and Combustion Meeting, Estes Park, CO,
Sept 26-28, National Renewable Energy Laboratory: Golden,
CO, 1994; pp 90-108.
Duprez, D. Selective Steam Reforming of Aromatic Compounds
on Metal Catalysts. Appl. Catal. A 1992, 82, 111-157.
Elam, C. C.; Meglen, R. R.; Tatsumoto, K.; Evans, R. J.; Chum,
H. L. Application of Molecular Beam Mass Spectrometry to the
Problem of Plastics Recycling Through Intelligent Chemical
Processing. In Proceedings of the Specialists Workshop on
Applications of Free-Jet, Molecular Beam Mass Spectrometric
Sampling; Estes Park, CO, Oct 11-14, 1994; Milne, T. A., Ed.;
NREL/CP-433-7748; National Renewable Energy Laboratory:
Golden, CO, 1994; pp 211-215.
Elliott, D. C. Relation of Reaction Time and Temperature to
Chemical Composition of Pyrolysis Oil. In Pyrolysis Oils from
Biomass: Producing, Analyzing and Upgrading; Soltes, E. J.,
Milne, T. A., Eds.; ACS Symposium Series 376; American
Chemical Society: Washington, DC, 1988; pp 55-65.
Evans, R. J.; Milne, T. A. Molecular Characterization of the
Pyrolysis of Biomass. 1. Fundamentals. Energy Fuels 1987, 1,
123-137.
Fagernas, L. Chemical and physical characterization of biomass-
based pyrolysis oils: Literature review. VTT Research Notes
1706; Technical Research Center of Finland; Espoo, Finland,
1995; pp 117.
Garcia, E. Y.; Laborde, M. A. Hydrogen Production by the Steam
Reforming of Ethanol: Thermodynamic Analysis. Int. J. Hy-
drogen Energy 1991, 16, 307-312.
Ind. Eng. Chem. Res., Vol. 36, No. 5, 1997 1517
Graham, R. G.; Freel, B. A.; Bergougnou, M. A. The Production of
Pyrolysis Liquids, Gas, and Char from Wood and Cellulose by
Fast Pyrolysis. In Research in Thermochemical Biomass Con-
version; Bridgwater, A. V., Kuester, J. L., Eds.; Elsevier Applied
Science: London, 1988; pp 629-641.
Iwasa, N.; Takezawa, N. Reforming of EthanolsDehydrogenation
to Ethyl Acetate and Steam Reforming of Acetic Acid over
Copper-Based Catalysts. Bull. Chem. Soc. Jpn. 1991, 64, 2619-
2623.
Iwasa, N.; Kudo, S.; Takahashi, H.; Masuda, S.; Takezawa, N.
Highly Selective Supported Pd Catalysts for Steam Reforming
of Methanol. Catal. Lett. 1993, 19, 211-216.
Jiang, C. J.; Trimm, D. L.; Wainright, M. S.; Cant, N. W. Kinetic
Study of Steam Reforming of Methanol over Copper-Based
Catalysts. Appl. Catal. A 1993a, 93, 245-255.
Jiang, C. J.; Trimm, D. L.; Wainright, M. S.; Cant, N. W. Kinetic
Mechanism for the Reaction between Methanol and Water over
a Cu-ZnO-Al2O3 Catalyst. Appl. Catal. A 1993b, 97, 145-158.
Kim, C. J. Noble Metal-Catalyzed Water-Hydrocarbon Reaction
Paths. J. Catal. 1978, 52, 169-175.
Kirk-Othmer Encyclopedia of Chemical Technology, 4th ed.; 1991;
Vol. 13, pp 838-894.
Kovac, R. J.; Gorton, C. W.; O’Neil, D. J. Entrained Flow Pyrolysis
of Biomass. Proceedings of Thermochemical Conversion Program
Annual Meeting; SERI/CP-231-2355; Solar Energy Research
Institute: Golden, CO, 1988; pp 5-20.
Mann, M. K.; Spath, P. L.; Kadam, K. Technoeconomic Analysis
of Renewable Hydrogen Production, Storage, and Detection
Systems. In Proceedings of the 1996 U. S. DOE Hydrogen
Program Review, Miami, FL, May 1-2, 1996, NREL/CP-430-
21968; National Renewable Energy Laboratory: Golden, CO,
1996; pp 249-272.
Mizuno, K.; Yoshikawa, K.; Wakejima, N.; Takeouchi, Y.; Wa-
tanabe, A. Production of Synthesis Gas with Various Composi-
tions of H2, CO, and CO2 from Methanol and Water on a Ni-
K/Al2O3 Catalyst. Chem. Lett., Chem. Soc. Jpn. 1986, 1969-
1972.
Patil, K. Z. Hydrogen. Chem. Age India 1987, 38, 519-527.
Piskorz, J.; Scott, D. S.; Radlein, D. Composition of Oils Obtained
by Fast Pyrolysis of Different Woods. In Pyrolysis Oils from
Biomass: Producing, Analyzing and Upgrading; Soltes, E. J.,
Milne, T. A., Eds.; ACS Symposium Series 376; American
Chemical Society: Washington, DC, 1988; pp 167-178.
Probstein, R.; Hicks, R. Synthetic Fuels; McGraw Hill: New York,
1982; p 235.
Radlein, D.; Piskorz, J.; Scott, D. S. Fast Pyrolysis of Natural
Polysaccharides as a Potential Industrial Process. J. Anal. Appl.
Pyrolysis 1991, 19, 41-63.
Richards, G. N. Glycoaldehyde from Pyrolysis of Cellulose. J. Anal.
Appl. Pyrolysis 1987, 10, 251-255.
Ridler, D. E.; Twigg, M. V. Steam Reforming. In Catalyst Hand-
book; Twigg, M. V., Ed.; Wolfe: London, 1989; pp 225-280.
Ross, J. R. H.; Steel, M. C. F.; Zeini-Isfahani, A. Evidence for the
Participation of Surface Nickel in the Steam Reforming of
Methane over Nickel/Alumina Catalysts. J. Catal. 1978, 52,
280-290.
Rostrupp-Nielsen, J. R. Catalytic Steam Reforming. In Catalysis
Science and Technology; Enderson, J. R., Boudart, M., Eds.;
Springer: Berlin, 1984; Vol. 5, Chapter 1.
Rostrup-Nielsen, J. R.; Tottrup, P. B. In Proceedings of the
Symposium on Science of Catalysis and its Applications in
Industry; Feb 22-24, 1979; FPDIL: Sindri, 1979, p 379ff.
Roy, C.; de Caumia, B.; Plante, P. Performance Study of a 30 Kg/
hr Vacuum Pyrolysis Process Development Unit. In 5th Euro-
pean Conference on Biomass for Energy and Industry; Grassi,
G., Goose, G., dos Santos, G., Eds.; Elsevier Applied Science:
London, 1990; p 2595ff.
Scahill, J.; Diebold, J. P.; Feik, C. Removal of Residual Char Fines
from Pyrolysis Vapors by Hot Gas Filtration. Proceedings of the
Developments in Thermochemical Biomass Conversion; Banff,
Canada, May 20-24, 1996; Blackie A&P: London, 1997; pp
253-266.
Scott, D. S.; Piskorz, J.; Radlein, D. Liquid Product from the
Continuous Flash Pyrolysis of Biomass. Ind. Eng. Chem. Process
Des. Dev. 1985, 24, 581-586.
Shafizadeh, F. Introduction to Pyrolysis of Biomass. J. Anal. Appl.
Pyrolysis 1982, 3, 283-305.
Su, T.-B.; Rei, M.-H. Steam Reforming Reactions of Methanol over
Nickel and Copper Catalysts. J. Chinese Chem. Soc. 1991, 38,
535-541.
1518 Ind. Eng. Chem. Res., Vol. 36, No. 5, 1997
Takahashi, K.; Takezawa, N.; Kobayashi, H. The Mechanism of
Steam Reforming of Methanol over a Copper-Silica Catalyst.
Appl. Catal. 1982, 2, 363-366.
Takezawa, N.; Shimokawabe, M.; Hiramatsu, H.; Sugiura, H.;
Asakawa, T.; Kobayashi, H. Steam Reforming of Methanol over
Cu/ZnO2. Role of ZrO2 Support. React. Kinet. Catal. Lett. 1987,
33 (1), 191-196.
Tottrup, P. B.; Nielsen, B. Higher Hydrocarbon Reforming.
Hydrocarbon Process. 1982 (March), 89-91.
Uhlman’s Encyclopedia of Industrial Chemistry; VCH Verlags-
gesellschaft mbH: Weinheim, Germany, 1989; Vol. A13, pp
297-442.
Wang, D.; Meglen, R.; Evans, R. J. Quantitative sampling with a
Free-Jet, Molecular Beam Mass Spectrometer. In Proceedings
of the Specialists Workshop on Applications of Free-Jet, Molec-
ular Beam Mass Spectrometric Sampling; Estes Park, CO,
October 11-14, 1994; Milne, T. A., Ed.; NREL/CP-433-7748;
National Renewable Energy Laboratory: Golden, CO, 1994; pp
275-288.
Wang, D.; Montané, D.; Chornet, E. Catalytic Steam Reforming
of Biomass-Derived Oxygenates: Acetic Acid and Hydroxyac-
etaldehyde. J. Appl. Catal. A 1996, 143, 245-270.
Received for review July 8, 1996
Revised manuscript received January 1, 1997
Accepted January 8, 1997X
IE960396G
X
Abstract published in Advance ACS Abstracts, February
15, 1997.