Applied Catalysis A: General 375 (2010) 116–123

Contents lists available at ScienceDirect

Applied Catalysis A: General

journal homepage: www.elsevier.com/locate/apcata

TAP investigation of hydrogen and carbon monoxide adsorption on a

silica-supported cobalt catalyst
Jingping Hong a,b, Stanislas Pietrzyk a, Andrei Y. Khodakov a,*, Wei Chu b, Maria Olea c,1,
Veerle Balcaen c, Guy B. Marin c
a
Unité de Catalyse et de Chimie du Solide, UMR 8181 CNRS, Bât. C3, USTL-ENSCL-EC Lille, Cite Scientifique, 59655 Villeneuve d’Ascq, France
b
Department of Chemical Engineering, Sichuan University, Chengdu 610065, China
c
Laboratory for Chemical Technology, Department of Chemical Engineering, Ghent University, Krijgslaan 281 (S5), B-9000 Gent, Belgium

A R T I C L E I N F O A B S T R A C T

Article history: A temporal analysis of products (TAP) set-up was applied to investigate the initial elementary steps of
Received 20 July 2009 the Fischer–Tropsch synthesis over silica-supported cobalt catalysts, i.e., hydrogen and carbon monoxide
Received in revised form 16 December 2009 adsorption. Single-pulse TAP experiments were performed over both reduced and passivated catalysts in
Accepted 17 December 2009
order to qualitatively and quantitatively assess the mechanism and kinetics of H2 and CO adsorption.
Available online 24 December 2009
Multi-pulse experiments were performed to change the state of the catalyst in a predefined way. Only
fast and equilibrated physical hydrogen adsorption was observed over a passivated silica-supported
Keywords:
cobalt catalyst. Over the reduced catalyst, the surface sites exhibited significant heterogeneity for
Fischer–Tropsch synthesis
Clean fuels
adsorption. TAP responses were successfully described assuming two different reversible adsorption
TAP processes for hydrogen and both irreversible and reversible adsorption for carbon monoxide. Supporting
Elementary steps qualitative information about reversibility of the adsorption was also obtained through alternating pulse
Transient kinetics experiments.
Mechanism ß 2009 Elsevier B.V. All rights reserved.
Adsorption
Hydrogen
Carbon monoxide

1. Introduction and allow evaluation of the intrinsic kinetic coefficients [5–11].

Fischer–Tropsch (FT) synthesis is an essential catalytic process
of gas-to-liquids (GTL), biomass-to-liquids (BTL) and coal-to-
liquids (CTL) technologies, which manufacture clean hydrocarbon
fuels without sulfur and aromatics [1–4]. FT synthesis typically
proceeds at the conditions far from the thermodynamic equilibri-
um, thus, carbon monoxide conversion and hydrocarbon selectiv-
ities are principally governed by chemical kinetics, heat, and mass
transfer. Supported cobalt catalysts have proven to be particularly
efficient for synthesis of middle distillate and long-chain hydro-
carbons. Although FT synthesis has been an extensively studied
process, little information is available about the rates of initial
elementary steps of the reaction, i.e. hydrogen and carbon
monoxide adsorption.
Transient kinetic experiments are particularly helpful in
yielding information about reversibility of the elementary steps
* Corresponding author. Tel.: +33 3 20 33 54 37; fax: +33 3 20 43 65 61.
E-mail address: andrei.khodakov@univ-lille1.fr (A.Y. Khodakov).
1
Present address: School of Science and Technology, University of Teesside,
Middlesbrough, Tees Valley, TS1 3BA, UK.
Moreover, these experiments can be conducted in the isothermal
mode which is usually difficult to attain in steady-state regime for
highly exothermic reactions such as FT synthesis. Our previous
report [12] addressed the kinetics of the FT synthesis on silica-
supported cobalt catalysts in a conventional pulse reactor. Though
valuable information about the reaction mechanism and kinetics
has been obtained, the conventional pulse method did not provide
reliable quantitative kinetic data about the fast carbon monoxide
and hydrogen adsorption kinetics because of moderate time
resolution. In comparison with conventional pulse methods, a
temporal analysis of products (TAP) set-up is particularly suitable
for kinetic investigation of adsorption and reaction processes
which proceed on a very short time scale.
TAP kinetic studies of carbon monoxide adsorption kinetics
involving oxide and noble metal catalysts have been a subject of
several reports [13–19]. A few works have addressed the kinetic
investigation of hydrogen adsorption mostly on Rh- [20] and Pt-
catalysts [21] using TAP. Carbon monoxide and hydrogen
adsorption on cobalt catalysts has not been previously investigated
using TAP.
The objective of this paper is to investigate the mechanism and
kinetics of H2 and CO adsorption. Throughout this study a TAP

0926-860X/$ – see front matter ß 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.apcata.2009.12.026
 J. Hong et al. / Applied Catalysis A: General 375 (2010) 116–123
reactor, a very powerful transient technique combined with
transient kinetic modeling and catalyst characterization, was
employed. Single-pulse, multi-pulse and alternating pulse experi-
ments have been conducted to qualitatively and quantitatively
assess the mechanism and kinetics of H2 and CO adsorption on a
silica supported cobalt catalyst. To the best of our knowledge, this
is the first TAP paper which addresses kinetics of carbon monoxide
and hydrogen adsorption on cobalt FT catalysts.
2. Experimental
2.1. Catalyst preparation and characterization
The catalyst was prepared by incipient wetness impregnation of
Cab-o-Sil M5 silica (specific surface area of 230 m2/g) with a
solution of cobalt nitrate. After impregnation, the catalyst was
dried and calcined at 673 K in air flow. The temperature ramp from
room temperature to 673 K was 1 K/min. The catalyst was
characterized by low temperature nitrogen adsorption, X-ray
diffraction and X-ray photoelectron spectroscopy.
Cobalt dispersion was measured by X-ray diffraction; the powder
diffraction patterns of oxidized cobalt catalysts were recorded using
a Siemens D5000 diffractometer (Cu(Ka) radiation). The average
crystallite size of Co3O4 was calculated according to the Scherrer
equation [22] using a (4 4 0) peak at 2u = 65.3448.
Cobalt reducibility was evaluated from X-ray photoelectron
spectra. XPS surface analysis was performed using a VG ESCALAB
220XL spectrometer. The Al(Ka) non-monochromatized line
(1486.6 eV) was used for excitation with a 300 W-applied power.
The analyzer was operated in a constant pass energy mode
(Epass = 40 eV) and the binding energies were referenced to the Si
2p core level (103.6 eV) of the SiO2 support. In situ catalyst
reduction was carried out in pure hydrogen at 673 K for 5 h in the
reactor cell of the preparation chamber attached to the analysis
chamber of the spectrometer, while the spectroscopic measure-
ments were carried out under a vacuum better than 107 Pa.
2.2. FT catalytic test
The carbon monoxide hydrogenation was carried out in a fixed
bed microreactor (inner diameter = 9 mm) at atmospheric pressure.
First, the catalyst was crushed and sieved to obtain catalyst grains
between 0.05 and 0.2 mm in diameter. The catalyst loading was
0.5 g. Then, the catalyst was heated until 673 K and reduced at this
temperature with a flow of hydrogen. After the reduction, a flow of
premixed synthesis gas with the molar ratio of H2/CO = 2 was
introduced into the reactor at 463 K. The carbon monoxide
contained 5% of nitrogen, which was used as an internal standard
for calculating carbon monoxide conversion. The gas transfer lines
were kept at 423 K to avoid possible condensation of the reaction
products. The reaction products were analyzed on-line by gas
chromatography using a Hewlett-Packard 5890 Series II Gas
Chromatograph with a 13X molecular sieve column and a thermal
conductivity detector for the analysis of H2, N2, CO, CO2 and CH4. C1–
C18 hydrocarbons were separated on a packed 10% CP-Sil5 column
on Chromosorb WHP and analyzed by a flame-ionization detector.
The FT reaction rates are expressed as cobalt-time yields; they were
calculated from carbon monoxide conversion, inlet carbon monox-
ide partial pressure, gas-space velocity and amount of cobalt in the
reactor. Hydrocarbon selectivity was calculated on carbon basis.
2.3. TAP experiments
Before the TAP experiments, the catalyst was first reduced in a
fixed bed reactor described in Section 2.2 in hydrogen flow at 673 K
and then passivated at room temperature by injecting small pulses
117
of air to helium flow. The passivated catalyst was removed from
the fixed bed reactor and loaded into the TAP reactor. In the TAP
reactor, the catalyst was reduced by multi-pulses of hydrogen at
673 K.
The TAP reactor applied in this study is a TAP-1 instrument,
consisting of three vacuum chambers in line. The first chamber
contains a microreactor in quartz, 45 mm long, with an inner
diameter of 4.86 mm. In single-pulse experiments, the reactor
contained a pre-bed filled with quartz (length 3.37 mm), a catalyst
bed containing 0.1 g of catalyst (13.85 mm) and a post-bed filled
with quartz (13.54 mm). The catalyst and quartz particle size are in
the range of 250–500 mm. The reactor is connected via a small
entrance volume (0.0218 cm3) with two high-speed pulse valves
and two continuous-feed valves. The background pressure in the
reactor chamber varies between 104 and 105 Pa. The second
chamber is the differential chamber. This chamber also acts as a
pre-vacuum chamber for the analysis chamber. The pressure in the
differential chamber is about 106 Pa. The responses of the
reactant and products at the exit of the reactor are measured
with a UTI 100C quadrupole mass spectrometer located in the third
chamber, the analysis chamber. The pressure in the analysis
chamber is 107 Pa. This ensures that the signal detected with the
mass spectrometer is proportional to the flow rate at the outlet of
the reactor. The time resolution at which the evolution of the outlet
flow rate can be measured is far less than 1 ms. Preliminary
experiments with Ar pulses were performed over an inert bed
consisting of the quartz particles of 250–500 mm which were also
used in other TAP experiments to determine Knudsen diffusion
coefficients. Single-pulse experiments were carried out with the
catalyst sample at 423–493 K temperature range.
H2/Kr (9:1) and CO/Kr (9:1) mixtures were used for the
experiments in which Kr was present as internal standard. The
number of molecules admitted per pulse was in the range of 1014–
1015 molecules, allowing the transport in the reactor to occur
through Knudsen diffusion. The amount of surface sites in 0.1 g of
catalyst was estimated as 4  1018, therefore the number of
molecules per pulse was considerably lower that the number of
surface sites in the catalyst. In alternating-pulse experiments, the
reactors load consisted in a pre-bed of quartz (length 2.77 mm), a
mixture of 0.0496 g of catalyst and 0.0506 g of quartz (length
9.34 mm), and a post-bed of quartz of 18.54 mm. The gas mixtures
used were H2/Kr (2/8) and CO/Ar (1/9).
For mass spectroscopic quantification of the different com-
pounds leaving the TAP reactor, the following m/e values were
chosen, based on sensitivity and specificity for each component: 2
for hydrogen, 15 for methane, 18 for water, 27 for C2 and C3
hydrocarbons, 28 for carbon monoxide, 44 for carbon dioxide and
84 for krypton. The fragmentation patterns were experimentally
determined. The absolute sensitivities for the above ions have been
measured in separate experiments. Preliminary data analysis was
performed using the theory of the TAP reactor as presented in [10],
and the detailed modeling was done using COMSOL and MATLAB
software packages.
3. Results and discussion
3.1. Catalyst characterization
Characterization data for the Co/SiO2 catalyst are displayed in
Table 1. The catalyst has a BET surface area of 168.6 m2/g and pore
volume of 1.06 cm3/g. The pore size distribution in mesopore range
is relatively broad with a maximum at 200–300 Å. Co3O4 was the
major phase in the calcined catalyst as identified by XRD and XPS.
The average size of the Co3O4 crystallites calculated using the
Scherrer equation [12] was about 21 nm. The fraction of cobalt
metal phases in the catalyst reduced at 673 K was estimated as 65%
118 J. Hong et al. / Applied Catalysis A: General 375 (2010) 116–123

Table 1
Silica-supported cobalt catalyst.

Cobalt Average Fraction of cobalt Amount of FT reaction rate (s1) SCH4 (%) SC5+ (%)
content (wt.%) crystallite size metal phases after cobalt metal
of Co3O4 (nm) reduction at 673 K sites (mmol/g)

12.0 21 0.65 0.067 2.9 *104 13 70

from decomposition of XPS spectra as described in our previous The approach proposed by Gleaves et al. [11] could be
report [23]. The size of the cobalt metal particles, Co0 was particularly interesting to analyze in a greater detail adsorption of
evaluated from the size of the cobalt oxide particles, Co3O4, H2 on the reduced catalyst. For a one-zone TAP reactor, containing
according to the relative molar volumes of metallic cobalt and a solid (catalyst) layer, the exit curves have been presented [11] as
Co3O4. The resulting conversion factor for the diameter of a given a dependence of dimensionless flow FieL2/Dei (Fi = reduced flow, in
Co3O4 particle being reduced to metallic cobalt is: dCo = 0.75dCo3O4. 1/s) on dimensionless time t = tDei/(eL2). This method however is
Cobalt dispersion (DCo, %) was calculated from average metal
particle sizes (Å) assuming spherical uniform particles with site
density 14.6 atoms/nm2, by the use of formula: DCo = 960/
dCo = 6.4% [24,25]. After taking into account that only about 65%
of cobalt can be reduced to the metallic phases, the number of
cobalt metal sites in the catalyst would be 0.067 mmol per gram of
catalyst.
The catalytic activity was measured in a fixed bed differential
microreactor at 1 atm and 463 K. At carbon monoxide conversions
up to 15%, the FT reaction rate was about 2.9  104 s1 with
methane and C5+ selectivities of 13% and 70% respectively. The
obtained catalytic results are consistent with the catalytic
performance previously measured at these conditions for similar
silica-supported cobalt catalysts [22,26].

3.2. TAP studies

3.2.1. Reduction of the passivated catalyst with multi-pulses of H2

The catalyst was passivated before loading into the TAP reactor
and therefore, a series of multi-pulse experiments with H2/Kr were
performed over the catalyst at 673 K in order to reduce it.
Intermediately, the state of the catalyst was monitored by
performing single-pulse experiments with H2/Kr. No hydrogen
consumption was observed in the single-pulse experiments after
9000 pulses of H2/Kr. After introducing about 9000 pulses of H2/Kr
at 673 K, the catalyst had the maximal extent of reduction
achievable at these conditions. Most of hydrogen in the first 9000
pulses was consumed by the catalyst. The number of pulses
required for reduction of the passivated catalysts corresponds
approximately to the reduction of surface and subsurface layers of
cobalt particles in the catalyst loaded into the reactor. The catalyst
sample loaded in the TAP reactor (0.1 g) contained 1018 metallic
surface sites. TAP experiments did not show any noticeable carbon
monoxide or hydrogen adsorption on silica at 423–523 K.

3.2.2. H2 adsorption
Fig. 1a and b shows the height normalized H2 and Kr responses
on a passivated and reduced catalyst for single-pulse experiments
at 473 K. The H2 response curves are quite broad with longer tails.
This points to a reversible adsorption of hydrogen on the cobalt
catalyst. Note that a much narrower hydrogen response is
observed on the passivated catalyst, which most probably does
not contain cobalt surface sites. Thus, the broad TAP response and
thus reversible hydrogen adsorption observed on the reduced
catalyst can be attributed to the presence of cobalt metal sites.
Careful examination of the response curve of H2 on the reduced
catalyst suggests a two-site adsorption mechanism. The broaden-
ing of H2 transient response indicates the presence of sites of
reversible hydrogen adsorption. The integration of H2 and Kr
curves shows that more than 85% of initial hydrogen is recovered Fig. 1. Height normalized H2 and Kr responses from H2/Kr single-pulse at 473 K on
during the acquisition time (1 s). The shape of the peaks was 0.1 g of the passivated (a) and reduced (b) Co/SiO2 plotted versus time and plot of
1=2 1=2
slightly affected by the temperature at 423–498 K range. Flowi Mi versus t=Mi (c).
 J. Hong et al. / Applied Catalysis A: General 375 (2010) 116–123
not directly applicable to a 3-zones TAP reactor, because the
diffusion coefficients are different in the 3 zones. Since the
Knudsen effective gas diffusion coefficients in the inert, in the
catalyst and in the free volume are inversely proportional to the
square root of its molar mass, the idea of the dimensionless flow
and time could be adapted to a 3-zones TAP reactor, by multiplying
the flow and dividing the time by the square root of the molar mass
(M1/2). Fig. 1c was modified accordingly. The adsorption of
hydrogen appears to be a rather complex process, including at least
one reversible step.
3.2.3. CO adsorption
The typical single-pulse transient responses obtained in CO/Kr
experiments are shown in Fig. 2a and b. The TAP responses were
normalized relative to the amounts of krypton and carbon
monoxide in the pulses incoming into the reactor. The area
normalized carbon monoxide response curve exhibits a higher
noise to signal level which is due to the low absolute intensity. The
low intensity of carbon monoxide response results from its high
uptake which suggests its irreversible adsorption on the catalyst.
Indeed, evaluation suggests that 94% and 95% of the carbon
monoxide pulses was adsorbed on the catalyst at respectively
1=2 1=2
423 K and 473 K. In addition, in the plot of Flowi Mi vs. t=Mi (see
Section 3.2.2) the curve of the CO is situated inside the curve of the
inert gas, Kr. This also suggests an essentially irreversible
adsorption of CO on the cobalt catalyst under these conditions
[11]. The shape of the curve of the CO and its long tail could
indicate however, the existence of the sites of reversible
adsorption.
Fig. 2. Normalized CO and Kr responses from CO/Kr single-pulse at 473 K on 0.1 g of
reduced catalyst versus time (a). The responses were normalized to the amounts of
1=2 1=2
Kr and CO incoming to the reactor. Plot of Flowi Mi versus t=Mi (b).
119
Strong adsorption of CO on the silica-supported cobalt catalyst
was also confirmed by alternating pulse experiments. CO was the
pump molecule, while H2 was the probe one. Hydrogen was
injected into the TAP reactor with delays relative to a carbon
monoxide pulse (Fig. 3). Hydrogen pulses on the catalyst exposed
to carbon monoxide lead to partial desorption of carbon monoxide.
It seems that the catalyst contains adsorption sites on which CO is
strongly adsorbed. Hydrogen pulses however, force to desorb this
strongly adsorbed carbon monoxide, characterized by the lifetime
longer than 2 s (Fig. 3b). This irreversibly chemisorbed carbon
monoxide cannot be removed without assisting pulses of hydrogen
(cross and assisted desorption [27,28]). When H2 was used as
pump molecule, and CO as a probe (Fig. 4), no carbon monoxide-
assisted desorption of hydrogen was detected (Fig. 4). Indeed, this
is in agreement with TAP results showing rapid and predominantly
reversible adsorption of hydrogen on the cobalt catalyst. Thus,
because of H2 desorption, no hydrogen species are present on the
surface when carbon monoxide is injected with some delay into
the reactor.
Thus, TAP experiments suggest that hydrogen adsorption is
principally reversible while, carbon monoxide adsorption is chiefly
irreversible on the cobalt catalyst. No reaction products were
observed in TAP experiments.
3.2.4. TAP modeling
TAP experiments were exploited using COMSOL software, a
package oriented to model systems of partial differential equations
(PDE) using the finite element method. COMSOL generated models
have been used as such, or transformed into Matlab m-files and
Fig. 3. Alternating pulse experiments at T = 498 K on 0.1 g of reduced catalyst: pump
CO, probe H2—(a) 0.5 s delay; (b) 2 s delay.
120 J. Hong et al. / Applied Catalysis A: General 375 (2010) 116–123

Table 2
Invariables used for modeling TAP transient responses.

DAr (573 K) (m2/s) eb

3
Inert 2.88  10 0.53
Catalyst 2.0  103 0.53

bed in the inert zones (eb = 0.53) was determined experimentally.

In the zone packed with the catalyst, the Knudsen diffusion was
combined with adsorption/desorption for hydrogen and carbon
monoxide, while no adsorption was postulated for krypton.
The initial conditions for four zones (j = 1, 2, 3 and 4) were

Cij ðz; 0Þ ¼ 0; uij ðz; 0Þ ¼ 0 (4)

The boundary conditions between the zones for j = 1, 2, 3 were

those of continuity of gas phase concentrations and fluxes.
The boundary condition for the gas-phase species at the
entrance of the reactor was defined as
   
Ni t t
Fi0 ¼ exp  (5)
A t2 t
where t is a coefficient related to the pulse width, and its value was
taken as 1 ms. The boundary condition at the outlet of the reactor
was:

Ci ¼ 0 (6)

The boundary conditions for the adsorbed species were Fi = 0

(insulation) for all boundaries.
The effective Knudsen diffusivities in the inert zones were
determined using Ar single-pulse experiments over a bed filled
with inert quartz at 573 K. These experiments were modeled using
Eqs. (1) and (3)–(6), yielding an effective Knudsen diffusivity of
argon at T0 = 573 K of 2.88  103 m2/s (Table 2). The agreement
Fig. 4. Alternating pulse experiments at T = 498 K on 0.1 g of reduced catalyst: pump
H2, probe CO—(a) 0.5 s delay; (b) 2 s delay. between the experimental and calculated exit flows (height
normalized) was very good. The Knudsen diffusivities of other
gases in the inert zones at other temperatures T, could then be
calculated using the following expression:
completed with Matlab lines to perform optimization of para- sffiffiffiffiffiffiffiffiffiffiffi
meters. Some details are given in Appendix A. TM Ar
inert inert
The TAP reactor was modeled in 1D approximation with four Di ðTÞ ¼ DAr ðT 0 Þ ; (7)
T 0 Mi
zones: a void inlet section before the reactor bed followed by three
packed zones. For each packed zone the usual partial differential where MAr and Mi are the molecular weights of argon and of gas i,
equation (PDE) was applied to each gas phase component (reactant respectively. Thus, the effective Knudsen diffusivity of krypton in
or inert): the inert zone at 573 K is 2.00  103 m2/g.
The effective diffusivities in the catalyst zone were estimated
@C i @2 C X by fitting the krypton response curves obtained in all single-
eb ¼ Di 2 i þ ns rb nij r i; j (1)
@t @z j pulse experiments at different temperatures over the TAP
reactor packed with three zones: inert, catalyst, inert (see
(ri,j = 0 for inert gases and in the inert zones of the bed). Section 2.3). A COMSOL model corresponding to Eq. (1) was
Additional ODEs were written for each surface species as built and converted to a Matlab m-file. This m-file was used to
follows: build a Matlab function computing an objective function equal
@ui X j to the sum of squared differences between experimental and
¼ ni r j (2) computed values of the Kr signal, using experimental calibration
@t j coefficients. When estimating the diffusivity over the catalyst
The first and third zones were packed with inert material, bed, the Knudsen diffusivity over the inert beds was taken as
quartz, while the second zone was packed with the catalyst. The equal to the value mentioned above and kept constant. The
flow in the free volume was modeled as Knudsen diffusion with effective diffusivity of krypton in the catalyst at a given
diffusivities corresponding to the average diameter of this part of temperature T, obtained by the optimization of the objective
the reactor, d̄. function, was expressed using Eq. (7) as a function of the
sffiffiffiffiffiffiffiffiffiffi diffusivity of the reference gas, argon, at the reference
2d̄ 2RT temperature, T0, Dcatalyst
Ar ðT 0 Þ. The porosity value for the catalyst
DKnudsen ¼ (3) bed was taken as equal to that of the inert bed, i.e. 0.53. As
3 pM i
expected, the obtained Dcatalyst
Ar ðT 0 Þ values vary very little with
The flow in the zones packed with inert material was modeled the temperature of the experiment, and the mean value found
as Knudsen diffusion. The porosity (fractional void fraction) of the was 2.0  103 m2/s (Table 2). Some small deviations could be
 J. Hong et al. / Applied Catalysis A: General 375 (2010) 116–123 121

due to the neglecting of the participation of the internal porosity

of the catalyst.
On the basis of the above values of porosities and reference
diffusivities the responses of hydrogen and carbon monoxide were
simulated in corresponding single-pulse experiments.
At first, an attempt was made to model the experimental single-
pulse responses obtained with H2/Kr mixtures assuming reversible
dissociative adsorption and those for CO/Kr mixtures assuming
irreversible adsorption (see Section 3.2.3). It was assumed that
adsorption of hydrogen and carbon monoxide occurred on a single
type of cobalt metal sites:
H2 þ 2  , 2Had coefficients ka ; kd

CO þ  ! COad coefficient ka
Fig. 5. Experimental and calculated responses of Kr at 473 K on Co/SiO2 catalyst. The
i.e. the source (reaction) term of Eq. (1) was taken as calculated response was simulated using a three-zone TAP model.
 
2
RH2 ¼  ka C H2  kd u H ns rb (8)
The hydrogen and carbon monoxide TAP transient response
and curves were successfully fitted assuming the presence of two types
RCO ¼ ka C CO ns rb (9) of sites. For CO, the irreversible stronger adsorption was assumed
to be non dissociative and irreversible, while for H2 the stronger
for C H2 and CCO respectively. u*, surface coverage of active sites adsorption was assumed to be dissociative and reversible, as
which are available for adsorption, under the conditions of single- described by Eqs. (8) and (9). The weaker adsorption process was
pulse experiments was considered as equal to 1 and omitted from assumed to be at equilibrium for both, CO and H2, and
the expressions for adsorption or reaction rates. characterized by a Henry-type adsorption isotherm, Ci = Hiui (‘i’
The assumption that carbon monoxide adsorption is only stands for H2 or CO). Both types of weaker adsorption for each gas
irreversible, has not been consistent with the experimental were assumed as independent. It is assumed that the equilibrium
transient responses. The experimental carbon monoxide tran- between gas phase and weakly adsorbed species is immediate. In
sient responses were much broader than those obtained by the this case, the contribution of the weak adsorption to the general
model which involves only irreversible carbon monoxide
adsorption. The transient response can not be either described
assuming only reversible carbon monoxide adsorption on one
adsorption site. More than 90% of carbon monoxide pulse has
been retained by the catalyst. This indicates that carbon
monoxide adsorption is to a larger extent irreversible. In case
of reversible carbon monoxide adsorption, most of it should be
desorbed after a relatively long time between pulses of 10 s.
Obviously, carbon monoxide responses simulated assuming only
single reversible carbon monoxide adsorption, were very
different from the experimental curve both in their integral
intensity and shape. Similar behavior was observed for hydrogen
single-pulse experiments. The discrepancy between the experi-
mental and calculated transient responses can be possibly
related to either the internal porosity effects or to the presence
of two or more different sites for hydrogen and carbon
monoxide adsorption. Note that hydrogen responses over
passivated catalyst (as shown in Fig. 1a) were satisfactorily
described by the model which takes into consideration only
external Knudsen diffusion and reversible physical adsorption
without any effects arising from diffusion limitations in the
catalyst pores. As already stated, transient responses of krypton
were also fitted adequately without any assumption about the
influence of catalyst internal porosity (see Fig. 5). Indeed,
carbon monoxide and particularly hydrogen have a much higher
Knudsen diffusivity than krypton. It seems rather doubtful that
the diffusion of these gases was significantly affected by
diffusion in catalysts pores, whereas the diffusion of Kr has
not been affected. Cab-O-Sil M-5 silica support has rather larger
mesopores (d = 20–30 nm, Table 1) to create any diffusion
limitations for hydrogen and carbon monoxide. The most
plausible explanation of the discrepancies between experimen-
tal and model curves seems to be related to the presence of
different types of hydrogen and carbon monoxide adsorption
sites. Such effects have been invoked in earlier TAP studies on Fig. 6. Experimental and computed TAP responses: (a) H2, (b) CO on 0.1 g Co/SiO2
chemisorption [16,17]. reduced catalyst at 473 K. Two types of adsorption sites were considered.
122 J. Hong et al. / Applied Catalysis A: General 375 (2010) 116–123
Table 3
Optimized parameters of CO and H2 adsorption from TAP experiments. The units are
given in Appendix B.
Adsorbing gas Temperature (K) ka (m3/mol s) kd (s1)
CO 423 2.02 – 3.7
CO 473 2.20 – 4.0
H2 423 0.020 3.37  108
H2 448 0.020 4.60  108
H2 473 0.019 5.37  108
TAP equation for species ‘‘i’’ can be transferred to the accumulation
term [29]. Eq. (1) can be re-written as:
 
rb n0s @C i @2 C
eb þ ¼ Di 2 i þ R i
Hi @t @z
where Ri are given by Eq. (8) and (9) for H2 and CO respectively.
Since the weak adsorption is considered in equilibrium, the
corresponding surface coverage0 is proportional to the gas phase
@u @C
concentration. Its variations: @ti ¼ H1 @ti gives rise to an accumula-
i
tion term and is combined with the gas-phase accumulation term
eb @@Cti .
Note that the rb n0s =Hi terms are the only additional parameters
to be determined with this model. As can be seen in Fig. 6a and b,
an excellent fit of experimental and calculated response curves
was obtained, for both CO and H2. The optimized parameters (rate
and equilibrium constants) are shown in Table 3. Our kinetic data
predict low activation energy for carbon monoxide and hydrogen
adsorption on cobalt catalysts. It is well known that activation
energy of hydrogen adsorption is relatively low on cobalt metallic
particles [30,31] and molecular hydrogen adsorbs [32] dissocia-
tively on the surface of transition metals, including cobalt at
temperature higher than 80–90 K. Previous reports also suggest
that carbon monoxide chemisorption on cobalt catalysts is usually
inactivated [12,33]. Note however that no attempt has been made
in the paper to evaluate the exact values of the Arrhenius activation
energy from our kinetic data (Table 3). All the kinetic data have
been obtained at very narrow temperature range which would
significantly affect the accuracy of activation energy values.
Thus, the TAP experiments suggest the presence of two sites of
adsorption for carbon monoxide and hydrogen. The information
about the nature of carbon monoxide adsorption sites can be
possibly extracted from different methods, e.g. FTIR and TPD. On
the basis of previous works [34–36], it can be suggested that the
stronger adsorption is related to cobalt metallic sites present in the
catalysts, while weaker reversible adsorption species can be
assigned to partially reduced cobalt species, e.g. Co2+ sites.
Previous TPD experiments [37] were also indicative of several
carbon monoxide adsorption sites in cobalt catalysts.
4. Conclusion
TAP experiments provided important qualitative and quantita-
tive information about reversibility and kinetics of carbon
monoxide and hydrogen adsorption on silica-supported cobalt
catalyst. Only weak hydrogen adsorption was observed on the
passivated cobalt catalyst. In the reduced catalyst, the surface sites
exhibited significant heterogeneity for adsorption. Stronger
adsorption sites were found to coexist with weaker adsorption
sites, for both species. The adsorption/desorption rate coefficients
for CO and H2 on stronger adsorption sites along with Henry-type
constants on weaker adsorption sites were determined by
regression of the experimental results obtained in separate
single-pulse over the reduced catalyst. The built model described
with high accuracy the adsorption of CO and H2 over the reduced
Co/SiO2 catalyst. The heterogeneity of the surface could possibly
influence the overall kinetics of the Fischer–Tropsch synthesis on
cobalt supported catalysts.
n0s  rb =H FT synthesis is a complex catalytic reaction involving
adsorption of hydrogen and carbon monoxide, formation of
C1 monomeric species and subsequent polymerization and
7.7 desorption of reaction products. TAP experiments provide
7.7
important information about the kinetics of the initial elemen-
8.5
tary steps of FT synthesis over cobalt catalysts such hydrogen
and carbon monoxide adsorption/desorption. In addition to
supplementary mechanistic and kinetic data, the information
about the initial steps of FT synthesis can be included in a more
comprehensive model which would predict the performance of a
given FT catalyst at a wide range of reaction conditions. Detailed
intrinsic kinetic data about the elementary steps of FT synthesis
(10) obtained through this TAP study seem essential for elucidation
of reaction mechanisms and optimizing of yield of higher
hydrocarbons.
Acknowledgements
The financial support by IDECAT Network of Excellence is
gratefully acknowledged. J.H. thanks the PhD scholarship support
of the Embassy of France in China. The authors thank Reviewers for
helpful and interesting discussion.
Appendix A. Modeling tools
COMSOL MultiphysicsTM, version 3.2b, with Chemical Engineering
Module, was used together with Matlab1 (version R2007a). COMSOL
models were built using Chemical Engineering Module, Mass balance,
Diffusion, Transient analysis (1D).
Tested COMSOL models were saved as Matlab m-files, and
converted into Matlab programs for reading files of experimental
data, extracting the necessary output results from the ‘‘fem’’
structures produced by the solving of model PDE systems, and
computing the objective function for optimization. The objective
function was minimized using a Matlab procedure fminsearch based
on the Nelder–Mead optimization algorithm.
Experimental data from TAP saved files, containing 10–25 scans of
8000 points for each component, covering an time interval of 10 s,
were averaged and used to compute the experimental fluxes. These
were compared as described above with corresponding values
obtained by simulations. The mean-square errors for carbon
monoxide and hydrogen adsorption–desorption were respectively
less than 10 and 6 per cents of maximal flux.
Appendix B. Notations
A reactor section area (m2)
Ci concentration of molecules i in the gas phase (mol/m3)
D pore diameter (m)
dCo3 O4 particle size of Co3O4 (nm)
dCo particle size of Co (nm)
DCo cobalt metal dispersion
d̄ average diameter of the free volume of the TAP reactor (m)
D axial dispersion (m2/s)
DKnudsen Knudsen diffusivity (m2/s)
Di effective Knudsen diffusivity of gas A (m2/s)
Dinert
i effective Knudsen diffusivity of gas i in inert bed (m2/s)
Dcatalyst
i
effective Knudsen diffusivity of gas i in catalyst bed (m2/s)
Fi0 flux of i (mol/m2 s)
 J. Hong et al. / Applied Catalysis A: General 375 (2010) 116–123 123

[9] P. Phanawadee, M. Phongaksorn, N. Chaimongkol, A. Jaree, J. Limtrakul, Chem.

Hi Henry-type adsorption constant of i (mol/m3)
Eng. J. 115 (2005) 51–62.
ka adsorption rate coefficient (m3 of gas/mol s) [10] A.C. van Veen, D. Farrusseng, M. Rebeilleau, T. Decamp, A. Holzwarth, Y. Schuur-
kd desorption rate coefficient (s1) man, C. Mirodatos, J. Catal. 216 (2003) 135–143.
[11] J.T. Gleaves, G.S. Yablonskii, P. Phanawadee, Y. Schuurman, Appl. Catal. A 160
Mi molecular weight of gas i (g/mol) (1997) 55–88.
ns the amount of active sites (mol/kg) [12] A.Y. Khodakov, B. Peregryn, A.S. Lermontov, J.-S. Girardon, S. Pietrzyk, Catal. Today
106 (2005) 132–136.
n0s the amount of active sites for reversible adsorption (mol/ [13] O. Dewaele, G.F. Froment, Appl. Catal. A 185 (1999) 203–210.
kg) [14] F.H.M. Dekker, J.G. Nazloomian, A. Bliek, F. Kapteijn, J.A. Moulijn, D.R. Coulson, P.L.
Mills, J.J. Lerou, Appl. Catal. 151 (1997) 247–266.
nsM the amount of active cobalt metal sites in the catalysts
[15] E.V. Kondratenko, Y. Sakamoto, K. Okumura, H. Shinjoh, Appl. Catal. B 89 (2009)
calculated from characterization data (mol/kg) 476–483.
Ni number of moles of gas i in the pulse [16] M. Olea, M. Kunitake, T. Shido, Y. Iwasawa, Phys. Chem. Chem. Phys. 3 (2001) 627–
631.
ri,j rates of formation of i species in reaction j (s1) [17] M. Olea, M. Kunitake, T. Shido, K. Asakura, Y. Iwasawa, Bull. Chem. Soc. Jpn. 74
R gas constant (J/kmol K) (2001) 255–265.
[18] S.O. Shekhtman, A. Goguet, R. Burch, C. Hardacre, N. Maguire, J. Catal. 253 (2008)
Ri source term, associated with adsorption or reaction rate 303–311.
ðmol=m3bed sÞ; i = 0 for inert gases [19] Z. Song, H. Nishiguchi, W. Liu, Appl. Catal. A 306 (2006) 175–183.
[20] O. Dewaele, D.Z. Wang, G.F. Froment, J. Mol. Catal. A 149 (1999) 263–273.
t time (s) [21] A. Goguet, S.O. Shekhtman, R. Burch, C. Hardacre, F. Meunier, G.S. Yablonsky, J.
T temperature (K) Catal. 237 (2006) 102–110.
[22] B.D. Cullity, Elements of X-ray Diffraction, Addision-Wesley Publishing Company,
T0 reference temperature (K)
London, 1978, p. 555.
vij stoichiometric coefficient [23] J.-S. Girardon, A.S. Lermontov, L. Gengembre, P.A. Chernavskii, A. Griboval-Con-
Z catalyst bed coordinate (m) stant, A.Y. Khodakov, J. Catal. 230 (2005) 339–352.
[24] D. Schanke, S. Vada, E.A. Blekkan, A.M. Hilmen, A. Hoff, A. Holmen, J. Catal. 156
eb porosity of the catalyst bed (fraction voidage) (1995) 85–95.
Q* surface coverage by free active sites [25] S. Sun, N. Tsubaki, K. Fujimoto, Appl. Catal. A 202 (2000) 121–131.
[26] A.Y. Khodakov, R. Bechara, A. Griboval-Constant, Appl. Catal. A 254 (2003) 273–
ui surface coverage of species ‘‘i’’ 288.
u0i surface coverage in weak reversible adsorption [27] T. Yamada, R. Zhai, Y. Iwasawa, K. Tamaru, Bull. Chem. Soc. Jpn. 62 (1989) 2387–
3 2389.
rb bed density ðkg=molbed Þ [28] M. Boudart, J. Molec. Catal. A 141 (1999) 1–7.
t coefficient related to the pulse shape (s) [29] Comsol Multiphysics, Chemical Engineering Module, Model Library, Version 3.4,
October 2007, p. 636.
[30] J.M. Zowtiak, C.H. Bartholomew, J. Catal. 83 (1983) 107.
References [31] J.M. Zowtiak, G.D. Weatherbee, C.H. Bartholomew, J. Catal. 82 (1983) 230.
[32] K.H. Ernst, E. Schwarz, K. Christmann, J. Chem. Phys. 101 (1994) 5388.
[33] R.C. Reuel, C.H. Bartholomew, J. Catal. 85 (1984) 63.
[1] M.E. Dry, Catal. Today 71 (2002) 227–241.
[34] A.A. Khassin, T.M. Yurieva, V.V. Kaichev, V.I. Bukhtiyarov, A.A. Budneva, E.A.
[2] A.Y. Khodakov, W. Chu, P. Fongarland, Chem. Rev. 107 (2007) 1692–1744.
Paukshtis, V.N. Parmon, J. Mol. Catal. A 175 (2001) 189–204.
[3] A.Y. Khodakov, Catal. Today 144 (2009) 251–257.
[35] L.E.S. Rygh, O.H. Ellestad, P. Klæboe, C.J. Nielsen, Phys. Chem. Chem. Phys. 2 (2000)
[4] E. Iglesia, Appl. Catal. A 161 (1997) 59–78.
1835–1846.
[5] P. Biloen, J. Mol. Catal. 21 (1983) 17–24.
[36] A.Y. Khodakov, J. Lynch, D. Bazin, B. Rebours, N. Zanier, B. Moisson, P. Chaumette, J.
[6] G.S. Yablonsky, M. Olea, G.B. Marin, J. Catal. 216 (2003) 120–134.
Catal. 168 (1997) 16–25.
[7] R. Burch, Top. Catal. 24 (2003) 97–102.
[37] J. Rathouský, A. Zukal, A. Lapidus, A. Krylova, Appl. Catal. A 79 (1991) 167–180.
[8] J. Perez-Ramirez, E.V. Kondratenko, Catal. Today 121 (2007) 160–169.