Applied Catalysis A: General 167 (1998) 161±172

Hydrogen production via the direct cracking of

methane over silica-supported nickel catalysts
Tiejun Zhang, Michael D. Amiridis*
Department of Chemical Engineering, University of South Carolina, Columbia, SC 29208, USA

Received 6 January 1997; received in revised form 4 April 1997; accepted 23 December 1997

Abstract

We have investigated the catalytic cracking of methane as an alternative route for the production of hydrogen from natural
gas. Nickel supported on silica was found to be active for this reaction producing stoichiometric amounts of hydrogen and
carbon. SEM and TEM characterization of the spent catalyst indicated that carbon deposits on the catalyst in the form of
hollow cylindrical ®laments, with a nickel particle located on the tip of each ®lament. As a result, deactivation due to carbon
deposition does not occur until the carbon ®laments begin to interfere with each other and the silica surface, due to the spatial
limitations of the pore structure of the catalyst. The deactivated catalyst can be fully regenerated by either oxidation in air or
steam gasi®cation of the deposited carbon. Additional hydrogen is produced during the steam regeneration process. # 1998
Elsevier Science B.V.

Keywords: Hydrogen production; Filamentous carbon; Methane cracking; Nickel catalysts

1. Introduction vehicles can then be propelled by electricity produced

from an on-board fuel cell rather than by an internal
Signi®cant progress made in fuel cell technologies combustion engine [2,3].
during the past decade has prompted exploration of The current proton-exchange membrane (PEM)
replacing traditional central large power plants with fuel cells utilize hydrogen as the energy source and
so-called distributed power generators, consisting of a require essential elimination (ideally below 20 ppm)
hydrogen generator and a membrane fuel cell [1]. This of carbon monoxide from the hydrogen stream to
latter technology generates electricity at locations prevent poisoning of the electrocatalyst. Hydrogen
where it is aimed to be used, and therefore, eliminates is typically produced through steam reforming of
the loss of electricity during its transmission. In methanol for vehicles [2±4] and through steam
addition, a fuel cell process does not emit any en- reforming, partial oxidation or autothermal reforming
vironmental pollutants including NOx, SOx and other of natural gas for stationary uses [5,6]. In all these
hydrocarbons. As a result, such a process becomes cases, however, carbon monoxide is a co-product,
attractive for the automobile industry as well, since which has to be converted into carbon dioxide in
subsequent steps.
*Corresponding author. Fax: +1 803 777 8265; e-mail: An alternative route is to directly crack the hydro-
amiridis@sun.che.sc.edu carbon fuel into hydrogen and carbon. In this case,

0926-860X/98/$19.00 # 1998 Elsevier Science B.V. All rights reserved.

PII S0926-860X(97)00143-9
162 T. Zhang, M.D. Amiridis / Applied Catalysis A: General 167 (1998) 161±172

formation of carbon oxides is avoided and the need for impregnation of powder silica (Davison-Syloid 74),
downstream reactions such as water-gas shift and which was pre-calcined overnight in a muf¯e furnace
selective oxidation is eliminated. Surprisingly, this at 773 K. An aqueous solution of nickel nitrate
approach has not been extensively studied. While (Aldrich, purity 99.999%) was utilized as the nickel
commercial processes exist that utilize thermal crack- precursor. The impregnated sample was dried in a
ing of methane at extremely high temperatures for the vacuum oven at 423 K for 2 h, and subsequently
production of acetylene and carbon black [7,8], hydro- calcined in a muf¯e furnace at 773 K for 4 h. The
gen production via the catalytic cracking of methane metal oxide loading was estimated by the weight
has been only brie¯y considered in the past [9]. difference between the blank silica and the ®nal
Recently, Muradov [10] studied the use of iron catalyst and was veri®ed by elemental Ni analysis
oxide as a catalyst for the cracking of methane and conducted by ¯ame atomic absorption (AA) spectro-
reported that equilibrium conversions were achieved scopy. The catalyst powder thus obtained was pressed
at temperatures above 1073 K. The iron oxide, also into pellets, which were then crushed and sieved to
appeared to maintain its activity for several hours, in obtain a granulometric fraction in the desired size
contrast to a Pt/Al2O3 catalyst which deactivated within range (25±35 mesh). The catalyst was ®nally dried
minutes under similar conditions. Furthermore, Ishihara immediately prior to its use, to remove any moisture
et al. [11,12] reported that methane cracking takes place that might have accumulated during its storage, and
at low temperatures over a 10% Ni/SiO2 catalyst, reduced in a 1 : 2 H2 : He mixture for 2 h at 723 K.
which does not deactivate even after approximately The reduced catalyst was found to have a BET surface
200 carbon per nickel atoms have been deposited on it. area of 250 m2/g; and a Ni dispersion of 3.5% (via
The prolonged life of the iron and nickel catalysts, hydrogen chemisorption) and thus, an average Ni
as observed by the above investigators, suggests that particle size of 30 nm.
catalyst deactivation is not caused by blocking of the
active sites by carbon. If this were the case, the 2.2. Reactor apparatus
catalysts would have rapidly lost their activity as
carbon deposited on their surface. On the other hand, A conventional ®xed-bed plug-¯ow reactor was
catalytic formation of carbon ®laments has been fre- constructed with a quartz tube (10.5 mm i.d.). The
quently reported during decomposition of carbon reactor tube was placed within a Lindberg tubular
monoxide and several hydrocarbons over supported furnace controlled by an Omega temperature control-
iron, cobalt and nickel particles, and extensive studies ler. This combination could reach temperatures of up
have been conducted regarding the growth of these to 1300 K with an accuracy of 1 K. The feed stream
®laments and its possible mechanism [13±22]. In this was preheated in the furnace to the reaction tempera-
paper we examine the cracking of methane as a ture (measured by a K-type thermocouple) before it
potential route for hydrogen production over silica- came into contact with the catalyst. The catalyst bed
supported nickel catalysts. Activity studies are com- temperature was also measured with a K-type thermo-
plemented by TEM and SEM characterization of the couple.
deactivated catalyst samples, and are utilized in High purity methane (National Welders, 99.99%),
elucidating the mechanism of catalyst deactivation frequently diluted in helium (National Welders,
due to carbon deposition. We further examine the 99.999%), was used as the reactant. Compressed
catalyst regeneration via either oxidation in air or hydrogen (National Welders, 99.999%) diluted in
steam gasi®cation of the deposited carbon. helium was utilized for in situ reduction of the cata-
lyst. Compressed air (National Welders, zero grade)
2. Experimental was utilized as the source of oxygen during the
catalyst regeneration. The ¯ow rate of each gas stream
2.1. Catalyst preparation was measured by a precalibrated rotameter. During
steam regeneration of the catalyst, liquid deionized
The 16.4 wt% Ni/SiO2 catalyst utilized in this work water was pumped (Sage syringe pump, model 352) at
was prepared in the laboratory by incipient wetness a desired rate into the preheating zone of the reactor,
 T. Zhang, M.D. Amiridis / Applied Catalysis A: General 167 (1998) 161±172 163

where it was quickly evaporated and carried into the was coupled with the SEM for elemental identi®ca-
catalyst bed by helium. tion. In preparation for TEM experiments, a few
A Perkin-Elmer 3920 gas chromatograph equipped particles of each sample were dispersed in distilled
with a thermal conductivity detector (TCD) was used water and subjected to an ultrasonic treatment. Powder
for gas analysis on an Alltech HayeSep DB packed X-ray diffraction (XRD) patterns were recorded with a
column (30 ft1/8 in. S.S., 100/120 mesh) at an oven Rigaku D/max-B diffractometer using nickel-®ltered
temperature of 323 K. Feed and product streams, as CuK radiation (ˆ1.5406 A Ê ).
well as the standard calibration gas, were selected via
a Valco 4-position valve (4SF4P) and on-line sampled
via a Valco external gas sample injector (6UWE, 3. Results and discussion
sample loop 250 ml). Water vapor in the product
stream, if any, was removed by using a condensation 3.1. Activity measurements
trap and a desiccant drier in series. The GC was
calibrated over a wide range of compositions, with Activity measurements for the methane cracking
the calibration routinely checked with a standard reaction were conducted with the Ni/SiO2 catalyst in a
mixture of H2, N2, O2, CO, CO2 and CH4 (Alltech 20% CH4 in He stream at 823 K and a gas hourly space
Mix 34). The sensitivity in mole fraction measure- velocity (GHSV) of 30 000 hÿ1. Under these condi-
ments was estimated to be 0.1% for hydrogen and tions the catalyst exhibited a high initial activity for
0.01% for other gases. the cracking of methane (approximately 35% CH4
Preliminary activity tests were conducted with the conversion). Hydrogen was the only gaseous product
blank silica support. No measurable methane conver- detected. In addition, the rates of methane conversion
sion was observed at temperatures below 1073 K even and hydrogen formation were found to be in a ratio of
at a low space velocity of 6000 hÿ1. Subsequent 1 : 2, thus, verifying the reaction stoichiometry for
experiments were therefore performed with 0.2 g of methane cracking. Finally, the amounts of carbon
the Ni/SiO2 catalyst (25±35 mesh) diluted in 0.3 g of deposited on the spent catalyst and methane reacted
silica (25±35 mesh), to reduce temperature gradients indicated a good closure of the carbon balance
along the catalyst bed. In a typical run, the catalyst was (1005%).
initially reduced in situ for 2 h at 723 K in a 1 : 2 H2/
He stream. Following the reduction, residual hydrogen 3.2. Catalyst deactivation
was ¯ushed with helium and the reactor was heated to
the desired temperature in a helium stream. Methane It is anticipated that a catalyst used for methane
was then switched into the stream to start the reaction. cracking will eventually deactivate as a result of
The reactor was maintained at atmospheric pressure carbon deposition. Carbon may deposit on the surface
and the pressure drop across the catalyst bed was to cover the active sites (site-blocking) or accumulate
negligible. Usually, a steady state product composition at the entrance of the pores to block further access of
was detected after 3±5 GC samplings. Deactivation the reactants to the interior (pore-mouth plugging). We
experiments were continued until no hydrogen was have estimated that in both cases catalyst deactivation
detected in the products. At that time, heating and the would occur within a short period of time. Even if 10
¯ow of methane were interrupted, and the reactor was carbon atoms are needed to block each surface Ni
¯ushed with helium until it reached room temperature. atom, for example, 11 mg of carbon deposition would
be enough to completely deactivate 1 g of the 16.4%
2.3. Catalyst characterization Ni/SiO2 catalyst. Furthermore, if pore-mouth plug-
ging was the main deactivation mechanism, approxi-
Spent catalysts recovered from deactivation and mately 250 mg of carbon would be suf®cient to clog
regeneration experiments were studied using scanning the external 10% of the pores, in 1 g of the Ni/SiO2
and transmission electron microscopies (SEM, Hita- catalyst sample.
chi S-2500 Delta; TEM, Hitachi H-8000). Energy- We have experimentally observed a signi®cantly
dispersive X-ray spectroscopy (EDS, Kevex 8000) higher amount of carbon deposition on the Ni/SiO2
164 T. Zhang, M.D. Amiridis / Applied Catalysis A: General 167 (1998) 161±172

Fig. 1. Deactivation of the Ni/SiO2 catalyst at 823 K and GHSVˆ30 000 hÿ1 (feed: 20% CH4, balance He).

catalysts before deactivation occurs. At 823 K for face. These ®laments appear to grow out of the silica
example, we observed a very slow deactivation of support surface, with their length increasing with
the Ni/SiO2 catalyst for the ®rst 2 h (Fig. 1) followed time-on-stream. It is also noticed in these micrographs
by a more rapid loss of activity during the third hour. that each ®lament has a bright tip, identi®ed by the use
By the time the catalyst was completely deactivated of SEM/EDS to be a nickel particle. Spent catalyst
(200 min), approximately 0.59 g of carbon had accu- samples were further studied by the use of X-Ray
mulated on the 0.2 g of the Ni/SiO2 catalyst sample. diffraction (XRD). As compared to the XRD pattern of
This amount is in very good agreement with the the fresh catalyst (Fig. 3(a)), the presence of carbon on
amount of carbon calculated based on the integration the spent samples after exposure to methane cracking
of the methane conversion (0.61 g), and corresponds at 823 K is clearly indicated by the strong peak at
to approximately 2700 carbon atoms accumulated on 2ˆ268, as well as the weaker ones at 2ˆ42.68, 53.88
the catalyst per surface nickel atom. and 77.68 (Fig. 3(b)±(d)). The intensity of each peak is
It becomes therefore apparent, that the capability of linearly proportional to the amount of carbon present
this catalyst to accommodate carbon is signi®cantly on the sample. Although these peaks appear to corre-
higher than those predicted by either the site-blocking spond to a material of graphitic nature, they cannot be
or pore-mouth plugging models. It also contradicts directly assigned to a single form of graphite when
previous reports by Kuijpers et al. [23] (50% Ni/SiO2 compared to standard XRD patterns, suggesting that
catalyst at 623 K) and Poirier et al. [24] (10% Ni/ graphitic carbon constituents with different degrees of
Al2O3 catalyst at 1173 K) who reported a fast catalyst defect or distortion are present in the samples. In
deactivation due to active site blocking. Our results contrast, the metallic nickel peaks at 2ˆ44.38,
suggest a completely different mechanism of catalyst 51.78 and 76.38 were not signi®cantly affected follow-
deactivation under the reaction conditions studied. ing the use of the catalyst for 60 min.
SEM and TEM analyses of the spent catalysts were Formation of carbon ®laments as a result of hydro-
utilized in the elucidation of this mechanism. In fact, carbon cracking has been extensively reported in the
the SEM micrographs in this case (Fig. 2) indicate the literature with higher molecular weight hydrocarbons
formation of ®lamentous carbon on the catalyst sur- over supported nickel, iron, cobalt and several alloy
 T. Zhang, M.D. Amiridis / Applied Catalysis A: General 167 (1998) 161±172 165

Fig. 2. SEM micrographs of the Ni/SiO2 catalyst after exposure to methane cracking for: (a) 0.5, and (b) 1 h (same experimental conditions as
in Fig. 1).

catalysts (e.g. Refs. [13±21] and references therein). ®laments over Ni catalysts at a selected temperature
In addition, de Bokx et al. [22] reported the formation range (625±1050 K) [22]. Finally, selective carbon
of ®lamentous carbon on a 50% Ni/SiO2 catalyst, ®lament formation may also be the reason for the
following exposure to methane at 750 K. A commonly differences observed in the deactivation behavior of
accepted mechanism of ®lament growth in these cases supported iron and platinum catalysts for methane
consists of carbon species formation on the front cracking [10].
surface of the metal, dissolution in the metal, diffusion Nevertheless, as the experimental data show in
through the metal, and precipitation from the metal at Fig. 1, the nickel catalyst eventually experienced
the rear surface for the formation of the ®lament body. some gradual deactivation due to the carbon deposi-
It has been also established that the rate-determining tion which after 2 h accelerated and led to complete
step is the bulk diffusion of the carbon through the loss of activity after approximately 3.5 h. A TEM
metal particle, while excess carbon migrates along the micrograph of the fully deactivated sample (Fig. 4)
metal surface to form the skin component of the shows that the growth of the carbon is terminated as a
®laments. result of spatial limitations. As can be seen in the
This mechanism may also provide an explanation ®gure, the modes of ®lament termination include the
for the disparity between the present observations and nickel particle's restriction by the silica surface, the
those by Kuijpers et al. [23] and Poirier et al. [24] who arm and the tip of another carbon ®lament.
observed a fast catalyst deactivation (and therefore, no
formation of carbon ®laments) during methane crack- 3.3. Catalyst regeneration
ing at lower (623 K) or higher (1173 K) temperatures.
Carbon ®lament formation at low temperatures is The reversibility of ®lament growth has been pre-
expected to be hindered by the low solubility of carbon viously studied by others [25], but no results have been
in nickel, while at high temperatures methane crack- reported on whether the catalyst activity can be
ing is much faster than the bulk diffusion and surface restored. In fact, some workers have suggested that
migration. In both cases, the nickel surface quickly catalyst regeneration may be futile in view of the
becomes fully covered with carbon, leading to the changes caused to the catalyst support structure as a
interruption of the catalytic cycle. Similar arguments result of the ®lament growth process. Conventional
have been used to rationalize the formation of carbon oxidation in air (C‡O2ˆCO2) and steam gasi®cation
166 T. Zhang, M.D. Amiridis / Applied Catalysis A: General 167 (1998) 161±172

Fig. 3. XRD patterns of the Ni/SiO2 catalyst after exposure to methane cracking for: (a) 0, (b) 0.5, (c) 1, and (d) 3.5 h (same experimental
conditions as in Fig. 1).

(C‡xH2OˆCOx‡xH2) were examined in this work as fully restore the activity of the catalyst, as shown in
potential routes for the regeneration of the catalyst at Fig. 5 where the methane conversion is plotted against
823 K. Surprisingly, both methods appear to be able to time-on-stream for the fresh and the regenerated
 T. Zhang, M.D. Amiridis / Applied Catalysis A: General 167 (1998) 161±172 167

Fig. 4. TEM micrograph of the fully deactivated Ni/SiO2 catalyst (same experimental conditions as in Fig. 1).

Fig. 5. Methane conversions obtained over the fresh (*, ^) and regenerated (* in air, & in steam) Ni/SiO2 catalyst at 823 K and two
different space velocities.
168 T. Zhang, M.D. Amiridis / Applied Catalysis A: General 167 (1998) 161±172

Fig. 6. XRD patterns of the Ni/SiO2 catalyst after: (a) complete deactivation; (b) regeneration in steam and (c) regeneration in air.

catalysts. The oxidation process was faster than the powder. The XRD patterns (Fig. 6(c)) suggest that
steam gasi®cation, but caused a high temperature the oxidation process completely removed the depos-
front. This front gradually moved through the catalyst ited carbon and converted the metallic nickel (peaks at
bed, causing the collapse of the sample to a ®ne 2ˆ44.38, 51.78 and 76.38) into nickel oxide (peaks at
 T. Zhang, M.D. Amiridis / Applied Catalysis A: General 167 (1998) 161±172 169

2ˆ37.18, 43.18 and 62.78) which had to be reduced and steam regeneration steps. Integration of the areas
before the next reaction cycle shown in Fig. 5. On the under the curves indicates that 2 moles of hydrogen
contrary, the catalyst bed maintained a uniform tem- were obtained from each mole of methane during the
perature pro®le during the steam regeneration process cracking step and an additional 1.4 moles were pro-
and the catalyst preserved its metallic nickel form at duced during the subsequent steam gasi®cation of the
the end of the process (Fig. 6(b)). deposited carbon, leading to an overall hydrogen yield
Another difference between the two regeneration of 3.4 moles for each mole of methane.
methods is that the steam gasi®cation leads to the Formation of carbon monoxide and methane were
production of additional hydrogen. This is at the also detected during the initial stage of the steam
expense of external thermal energy, since the oxida- regeneration, where, as shown in Fig. 8, the carbon
tion of carbon releases a large amount of heat. Never- dioxide and hydrogen selectivities are less than 100%
theless, the additional production of hydrogen may be (two selectivities are de®ned in this case in terms of
of signi®cant practical importance. Theoretical yields carbon and hydrogen containing products; the carbon
of four and three moles of hydrogen per mole of dioxide selectivity is de®ned as the amount of carbon
methane can be achieved, respectively, with steam in the carbon dioxide produced, divided by the total
reforming and partial oxidation of methane, provided amount of carbon in all products ± i.e. carbon dioxide,
that all carbon monoxide is converted to carbon carbon monoxide and methane; similarly, the hydro-
dioxide by the water-gas shift reaction. In comparison, gen selectivity is de®ned as the amount of hydrogen in
methane cracking produces less hydrogen (two moles) molecular hydrogen produced, divided by the total
from each mole of methane. The hydrogen yield, amount of hydrogen in all products ± i.e. hydrogen and
however, can be dramatically improved if the hydro- methane). Increased selectivities of carbon dioxide
gen produced during the regeneration step is also and hydrogen, obtained when the regeneration was
accounted for. Fig. 7 shows the rate of hydrogen conducted at a higher space velocity, appear to suggest
production at 823 K during both the methane cracking that carbon monoxide and methane may be secondary

Fig. 7. Hydrogen production over the Ni/SiO2 catalyst during methane cracking (*) and subsequent steam regeneration (*) at 823 K and
GHSVˆ15 000 hÿ1.
170 T. Zhang, M.D. Amiridis / Applied Catalysis A: General 167 (1998) 161±172

Fig. 8. Selectivities of hydrogen (* at GHSV ˆ15 000 hÿ1, ^ at GHSVˆ37 500 hÿ1) and carbon dioxide (* at GHSV ˆ15 000 hÿ1, & at
GHSVˆ37 500 hÿ1) during steam regeneration of the deactivated Ni/SiO2 catalyst at 823 K.

Fig. 9. TEM micrograph of a carbon filament present on the Ni/SiO2 catalyst following steam regeneration at 823 K and GHSVˆ15 000 hÿ1.
 T. Zhang, M.D. Amiridis / Applied Catalysis A: General 167 (1998) 161±172 171

products, formed due to the reverse water-gas shift oxidation step will be probably required, to comple-
(H2‡CO2ˆCO‡H2O) and methanation (CO‡3H2ˆ tely remove the remaining ``skins''.
CH4‡H2O) reactions. No other species were detected
within the GC/TCD limit.
TEM micrographs of the steam regenerated sample 4. Conclusions
(Fig. 9) show that some remains of the large ®laments
are still present on this catalyst after the regeneration As a potential alternative to steam reforming and
has been completed. This is consistent with the XRD partial oxidation, catalytic cracking of methane may
pattern of this sample (Fig. 6(b)), which shows that, provide hydrogen without carbon monoxide contam-
although substantially reduced, carbon peaks are still ination, for niche applications, such as use in fuel
present. In addition, a change is observed in the ratio cells. Silica-supported nickel catalysts are active for
of the intensities of the two carbon peaks at 2ˆ268 this reaction at relatively low temperatures (823 K).
and 42.68, between the deactivated and regenerated These catalysts maintain their activity for a sustained
samples. The TEM micrograph of the regenerated period of time because of their capability to accumu-
(Fig. 9) sample shows that the carbon walls of the late large amounts of carbon on their surface in the
remaining ®laments have become much thinner, sug- form of long cylindrical hollow ®laments, with a
gesting that the carbon forming the inner portion of the nickel particle located at the front tip of each indivi-
®laments is more susceptible to the steam gasi®cation dual ®lament. Deactivation eventually occurs due to
process. These results suggest that the ®lament wall is spatial limitations imposed on the ®lament growth,
formed by two different types of graphitic carbon, which lead to the loss of exposed nickel surface. The
with the external ``skin'' being more resistant to steam deactivated catalyst can be fully regenerated by either
gasi®cation than the inner portion. oxidation in air or steam gasi®cation of the deposited
Fig. 9 also shows that the nickel particle on the tip carbon. Additional hydrogen is produced during the
of the ®lament has been removed, leaving the end of steam regeneration process. In fact, one can view the
the ®lament open. Small holes and small Ni particles methane cracking-steam regeneration combination as
are also visible at other locations on the ®lament. a breakdown of the conventional steam reforming
Carbon ®laments on the catalyst after its complete process in two distinct steps. This breakdown allows
deactivation (Fig. 4) appeared to be in contact with for a partial separation of the products, since only
each other, with some nickel particles being ``seeded'' hydrogen is produced during the cracking step, and
into the arms of other ®laments. These results suggest probably, a better control of the selectivity during the
that the steam gasi®cation starts at the nickel±carbon steam gasi®cation step. Such a breakdown is made
interface and is catalyzed by nickel. Such an explana- practically possible because of the ability of the nickel
tion is consistent with the high initial rate of hydrogen particles to form carbon ®laments, and thus accumu-
production during the regeneration as shown in Fig. 7. late signi®cant amounts of carbon on the catalyst
Eventually, the nickel particles lose contact with the before deactivation occurs.
carbon and are removed from the ®laments, providing
easier access for water vapor to the interior. Conse-
quently, the non-catalyzed steam gasi®cation of the Acknowledgements
inner less-graphitic carbon wall is largely responsible
for the continuing production of hydrogen. The exter- The funds provided by Niagara Mohawk Power
nal graphitic skin, which accounts for 25±30% of the Corporation (NMPC) for this study is gratefully
total deposited carbon according to our results, acknowledged, as are the helpful discussions with
appears to resist steam gasi®cation under these con- David Bloom®eld (Analytical Power) and Cicero
ditions. The presence of these ``skins'' however, does (Bernie) Bernales (NMPC). The experimental assis-
not appear to have a negative effect on catalytic tance of Charles Wilson and Heather Agler in catalyst
activity which is fully restored following the steam preparation and evaluation, as well as that of Platon
regeneration step (Fig. 5). Following a few cycles of Tsakiroglou in the XRD characterization studies, is
deactivation and steam regeneration, however, an air also acknowledged.
172 T. Zhang, M.D. Amiridis / Applied Catalysis A: General 167 (1998) 161±172

References [13] R.T.K. Baker, P.S. Harris, in: P.L. Walker, Jr., P.A. Thrower
(Eds.), Chemistry and Physics of Carbon, vol. 14, p. 83
Dekker, New York, 1978.
[1] K. Foger, in: H.E. Curry-Hyde, R.F. Howe (Eds.), Natural Gas [14] R.T.K. Baker, Catal. Rev. Sci. Eng. 19 (1979) 161.
Conversion II, Elsevier, Amsterdam, 1994, p. 115 . [15] M. Audier, M. Coulon, Carbon 23 (1985) 317.
[2] J. Haggin, C and EN, 8, July 24 (1995). [16] I. Alstrup, J. Catal. 109 (1988) 241.
[3] R.A. Lemons, J. Power Sources 29 (1990) 251. [17] M.S. Dresselhaus, G. Dresselhaus, K. Sugihara, I.L. Spain,
[4] J.C. Amphlett, A.M. Creber, J.M. Davis, R.F. Mann, H.A. Goldberg, Graphite Fibers and Filaments, Springer,
B.A. Peppley, D.M. Stokes, Int. J. Hydrogen Energy 19 Berlin, 1988.
(1994) 131. [18] R.T.K. Baker, Carbon 27 (1989) 315.
[5] N. Dave, G.A. Foulds, Ind. Eng. Chem. Res. 34 (1995) 1037. [19] R.T. Yang, J.P. Chen, J. Catal. 115 (1989) 52.
[6] T.S. Cristensen, I.I. Primdahl, Hydrocarbon Process 73(3) [20] N.M. Rodriguez, J. Materials Research 8 (1993) 3233.
(1994) 39. [21] W.L. Holstein, J. Catal. 152 (1995) 42.
[7] Kirk-Othmer Encyclopedia of Chemical Technology, 4th ed., [22] P.K. De Bokx, A.J.H.M. Kock, E. Boellaard, W. Klop, J.W.
vol. 4/13, Wiley/Interscience, New York, 1991. Geus, J. Catal. 96 (1985) 454.
[8] Ullmann's Encyclopedia of Industrial Chemistry, 5th revised [23] E.G.M. Kuijpers, J.W. Jansen, A.J. van Dillen, J.W. Geus, J.
ed., vol. A1/A5, VCH Verlagsgesellschaft mbh, Weinheim, Catal. 72 (1981) 75.
1985. [24] M.G. Poirier, C. Perreault, L. Couture, C. Sapundzhiev, in: O.
[9] Chem. Eng. 69, 90 (1962). Savadogo, P.R. Roberge, T.N. Veziroglu (Eds.), Proceedings
[10] N.Z. Muradov, Int. J. Hydrogen Energy 18 (1993) 211. of the first International Symposium on New Materials for
[11] T. Ishihara, T. Fujita, Y. Miyashita, Y. Takita, Shokubai 35 Fuel Cell Systems, Montreal, Que., Canada, 9±13 July 1995,
(1993) 324. p. 258 .
[12] T. Ishihara, Y. Miyashita, H. Iseda, Y. Takita, Chem. Lett. [25] J.L. Figueiredo, C.A. Bernardo, J.J. Chludzinski, Jr, R.T.K.
(1995) 93. Baker, J. Catal. 110 (1988) 127.