Fuel 81 (2002) 1583±1591

www.fuel®rst.com

Fischer±Tropsch synthesis using Co/SiO2 catalysts prepared

from mixed precursors and addition effect of noble metals q
S. Sun a, K. Fujimoto b, Y. Yoneyama c, N. Tsubaki c,*
a
Department of Applied Chemistry, School of Engineering, The University of Tokyo, Hongo 7-3-1, Bunkyo-ku, Tokyo 113-8565, Japan
b
Department of Chemical Processes and Environments, The University of Kitakyushu, Hibikino 1-1, Wakamatsu-ku, Kitakyushu, Fukuoka 808-0135, Japan
c
Department of Material System and Life Science, School of Engineering, Toyama University, Gofuku 3190, Toyama 930-8555, Japan
Received 8 May 2001; revised 25 October 2001; accepted 25 October 2001; available online 10 April 2002

Abstract
Fischer±Tropsch synthesis in Co/SiO2 catalysts, which were prepared by mixed impregnation of cobalt (II) nitrate and cobalt (II) acetate,
was studied under mild reaction conditions (Total pressure ˆ 1 MPa, H2/CO ˆ 2, T ˆ 513 K). X-ray diffraction indicated that highly
dispersed cobalt metal was the main active sites on the catalyst prepared by the same method. It was considered that the metallic crystallines,
which were readily reduced from cobalt nitrate, promoted the reduction of Co 21 to metallic a state in cobalt acetate by H2 spillover
mechanism during the catalyst reduction process. The reduced cobalt, from cobalt acetate, was highly dispersed one and remarkably
enhanced the catalytic activity. The addition of a small amount of Ru to this type of catalyst remarkably increased the catalytic activity
and the reduction degree. Its turn over frequency (TOF) increased but the selectivity of CH4 was unchanged. However, when Pt or Pd were
added into catalysts, they exhibited a higher selectivity of CH4. Although Pt and Pd hardly exerted an effect on cobalt reduction degree, they
promoted cobalt dispersion and decreased the value of TOF. Characterization of these bimetallic catalysts suggested that a different contact
between Co and Ru, Pt or Pd existed. Ru was enriched on the metallic cobalt surface but, Pt or Pd dispersed well in the form of Pt±Co or
Pd±Co alloy. q 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Co/SiO2 catalysts; Fischer±Tropsch synthesis; Noble metal; CO hydrogenation; Reduction degree

1. Introduction strong interaction between supports and cobalt precursors,

but in turn such interactions generally lower the reduction
Cobalt-based catalysts are known to be effective in CO degree of such precursors to Co metal. Support-precursor
hydrogenation [1,2], which is very important for the produc- pairs of Silica±cobalt carboxylate lead to very small CoOx
tion of liquid synthetic fuel. The activity of the Fischer± particles, which are reduced completely only above 1000 K
Tropsch synthesis (FTS) catalyst can be increased at high and form sintered metal particles [6]. Silica±cobalt nitrate
temperature, but selectivities of unfavorable products such pairs form CoOx particles in an intermediate size during
as CH4, CO2 are also enhanced sharply. Thus, the high nitrate decomposition; then these crystallites can be reduced
activity of catalysts at a low temperature is ideally expected. at 673 K in ¯owing H2.
Recent studies indicated that the FTS activity of Co cata- A number of investigations have been focused so far on
lysts solely depends on the number of active-site located on the noble metal as the promoter for cobalt supported cata-
the surface of crystalline metal formed by the reduction lysts. Iglesia et al. [7] showed that the presence of Ru in the
process [3]. The number of active-sites was determined by supported cobalt catalysts not only enhanced the reduction
the Co particle size, loading amount and the reduction of cobalt oxide but also inhibited carbon deposition during
degree [4,5]. FTS. Schanke et al. [8] studied Co±Pt catalysts with low Pt/
Synthesis of highly dispersed Co catalysts requires the Co ratio and found that the selectivity was not in¯uenced
initial formation of a small size of CoO or Co3O4 crystal- by the presence of Pt, while the cobalt oxide reducibility
lites. The formation of highly dispersed catalysts needs and dispersion were increased signi®cantly for Co/Al2O3
catalysts. However, on SiO2-supported Co catalysts, the
presence of Pt did not in¯uence the ®nal extent of reduction,
* Corresponding author. Tel./fax: 181-76-445-6846.
E-mail address: [email protected] (N. Tsubaki). but increased the dispersion slightly.
q
Published ®rst on the web via Fuel®rst.comÐhttp://www.fuel®rst.com In this paper, it is ®rstly reported that the mixed cobalt
0016-2361/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0016-236 1(02)00090-X
1584 S. Sun et al. / Fuel 81 (2002) 1583±1591

salts, as a precursor, had a great effect on the activity and Hydrocarbons dissolved in the solvent and cooled in the trap
selectivity of cobalt catalysts. Furthermore, trace amount of were analyzed on a Silicone SE-30 column with FID.
noble metals, such as Ru, Pd, or Pt were added into the
Co/SiO2 catalyst prepared from the mixed cobalt precursors, 2.3. Catalyst characterization
the performance and structure of these catalysts were
An X-ray diffractometer (XRD, Rigaku) with monochro-
investigated.
matized Cu Ka radiation, and a scan rate of 2.08 min 21, was
used. The catalyst samples for XRD measurement were at
the oxide state. The average particle sizes of Co3O4 were
2. Experimental
calculated using the Scherrer equation. The diameter of a
2.1. Catalyst preparations and its nomenclature given Co3O4 particle could be used to calculate the diameter
of metallic Co crystallite by the formula below [9]:
The nomenclature for catalysts in this study consists of dv …Co0 † ˆ 0:75 d …Co3 O4 † …1†
two parts: [xN/yA] or [xN 1 yA]. N indicates cobalt (II)
nitrate and A indicates cobalt (II) acetate; x refers to the The active surface area for the metal of the catalyst was
weight percent of cobalt metal from N and y refers to the measured by chemisorption in a conventional static volu-
weight percent of cobalt metal from A. The second part metric adsorption system. Before adsorption of H2 or CO,
gives the order of impregnation of metal salts: [N/A] the catalysts, previously reduced by H2 and passivated, were
means that N was impregnated into the support previous treated in H2 at 673 K for 1 h, then followed by evacuation.
to A, and [N 1 A] means that N and A were co-impreg- H2 adsorption isotherms were measured at 373 K, as
nated. For the noble metals added into catalysts, the molar described in detail elsewhere [10]. The reduction level
ratio of noble metal to Co was between 0.61 and 1.16%. The was determined after the reduction of the catalyst followed
noble metal salts are Ru(NO3)3, Pd(NH3)2(NO2)2 and by O2 titration at 673 K. The reduction percentage was
Pt(NH3)2(NO2)2 (Tanaka Co.), respectively. calculated by assuming stoichiometric conversion of metal-
The supported cobalt catalysts were prepared by incipi- lic Co to Co3O4 [11].
ent-wetness impregnating cobalt (II) salts on a commer- FTIR spectra were recorded in transmission mode (Nicolet
cially available silica gel (Fuji Davison ID gel, speci®c Magna 550) with a 2 cm 21 resolution using a MCT detector. A
surface area of 270 m 2 g 21). The samples were dried, in 30±40 mg sample was pressed into a 10 mm B self-support-
an oven, at 393 K for 12 h. Furthermore, they were calcined ing disk and put into a slit of the holder in an IR cell. The
in air by raising the temperature to 723 K with a ramping catalyst in the disk was reduced with hydrogen in the cell at
rate of 3 K min 21, keeping that temperature for 2 h. After 673 K for 1 h. CO of 26 kPa was adsorbed on the catalysts at
calcination, catalysts were activated in ¯owing hydrogen at room temperature for 10 min and then evacuated before the
673 K for 10 h; the temperature ramping rate was also FTIR spectra of the CO adsorption were measured.
3 K min 21 also. The catalysts were passivated by 1% Transmission electron microscope (TEM) images were
oxygen in nitrogen. For the sequentially impregnated cata- taken on JEM-2010F (JEOL) operated at 200 kV. A catalyst
lysts, after the ®rst impregnation step, the catalyst precur- pellet was embedded into resin and cut by microtome
sors were dried in an oven at 393 K for 12 h and the samples method (Leica Ultracut UCT). For elemental analysis of
were calcined in air at 723 K for 2 h. The procedures were individual metal particles, a suitable sampling area was
repeated for the second salt impregnation. identi®ed and photographed in the transmission mode.
The microscope was then switched over to scanning trans-
mission mode for energy dispersive spectroscopy
P 3 P (EDS).
2.2. Catalytic measurements Average particle size is estimated by ni di = ni di2 : A
Catalysts were tested in a ¯ow-type semi-batch autoclave 0.5 nm electron probe was focused on individual metal
slurry-phase reactor with an inner volume of 80 ml. Twenty particles and X-ray counts were acquired for 100 s.
milliliters of n-C16H34 were used as the liquid medium. A X-ray photoelectron spectroscopy (XPS) spectra were
cooling trap was set between the reactor exit and the pres- recorded using a VG ESCALAB 220 spectrometer with a
sure regulator to collect the water and liquid hydrocarbons. Mg Ka radiation source, where reduced sample could be
The standard reaction conditions were P (total) ˆ 1.0 MPa, prepared in situ without exposure to air. The C1s photo-
CO/H2 ˆ 1/2, W/F (CO 1 H2) ˆ 5 g h mol 21, T ˆ 513 K. emission line was taken as a reference to correct electro-
Argon was employed as an internal standard with a concen- static charging.
tration of 3% in the feed gas.
Gaseous products were analyzed by on-line gas chroma- 3. Results and discussion
tographs. CO and CO2 were analyzed by using an active
charcoal column equipped with a thermal conductivity 3.1. Catalytic activity of various Co-based catalysts
detector (TCD). Light hydrocarbons were analyzed on a
Porapak-Q column with a ¯ame ionization detector (FID). Table 1 shows the FTS activity and selectivity of the
 S. Sun et al. / Fuel 81 (2002) 1583±1591 1585

Table 1
Catalytic behaviors of various Co(10 wt%)/SiO2 catalysts for CO hydro-
genation (Reaction conditions: 513 K, 1.0 MPa, H2/CO ˆ 2, steady state,
W/F ˆ 5 g cat h mol 21)

Catalyst CO conv. (%) CH4 sel. (%) CO2 sel. (%) a

10N 29.8 6.6 1.6 0.85

5N/5A 42.5 6.3 1.8 0.86
5A/5N 31.3 6.7 1.2 0.85
5N 1 5A 33.5 8.8 1.3 0.84
10A 6.2 7.6 1.6

catalysts with different preparation procedures. 5N/5A

showed the highest activity. For the catalyst 10A, whose
total loading was 10 wt%, the reduction degree and the
particle sizes determined from H2 adsorption varied with
the method of preparation. Table 2 clearly shows that the
catalysts prepared from mixed salts had smaller particle size
than that of the catalyst prepared from cobalt nitrate only. It
should be noted that in Table 2, the metal particle size Fig. 1. The relation of CO conversion, reduction degree from O2 titration
and particle size determined from H2 chemisorption with cobalt nitrate
calculated from XRD data is generally in accordance with percentage in N/A catalyst series. Total cobalt loading: 10 wt%. R: reduc-
the value calculated from H2 chemisorption. In the case of tion degree; P: particle size; C: CO conversion. Standard reaction condi-
the catalyst where cobalt acetate was the precursor, metallic tions.
cobalt peak, in XRD, was not detected. The catalyst from
nitrate only had a high reduction degree but a low degree of
dispersion. The catalyst from acetate solely had a high obtained from O2 titration. It seems that the H2 spillover
degree of dispersion but very low degree of reduction. took place in the catalysts prepared from mixed salts
The interaction between the cobalt oxide, from cobalt acet- system, similar to the facts that noble metals promoted the
ate, and the silica surface was very strong. reduction of Co 21 to cobalt metals [12,13]. Thus, during the
Assuming the theoretical reduction degree of the catalysts catalyst preparation, the CoOx derived from cobalt nitrate
prepared from mixed salts can be calculated from the equa- was reduced at ®rst, which activated gaseous hydrogen and
tion promoted the CoOx reducibility derived from the cobalt
acetate through H2 spillover, and then highly dispersed
R ˆ NRN 2 ARA …2† supported cobalt catalyst was prepared.
For the catalysts 10A, 3N/7A, 5N/5A, 7N/3A and 10N,
where R means the theoretical reduction degree of the cata- whose total loading were 10 wt%, the reduction degree and
lysts prepared from mixed salts; N and A mean molar the particle size determined from H2 adsorption increased
percentage of cobalt nitrate and cobalt acetate, respectively; with increasing N/A ratio, as shown in Fig. 1. But the CO
RN and RA are the reduction degrees of catalysts 10N and conversion had the largest value at N/A ˆ 1. This phenom-
10A obtained from O2 titration at 673 K. The results calcu- enon can be attributed to the fact that the catalyst activity of
lated from Eq. (2) are listed in Table 3. The results calcu- Co depends on the number of active sites. As shown in
lated from Eq. (2) were obviously smaller than those Table 2, cobalt oxide species, from cobalt acetate, had

Table 2
The characterization results of Co(10 wt%)/SiO2 catalyst (N.D.: not detectable)

Catalyst a H2 chemisorption XRD Reduction degree O2 TOF b ( £ 10 2 s 21)

titration at 673 K (%)
H2 uptake c (mmol H2/g) Average dp d of Co 0 (nm) Co D (%) Average dv of Co 0 (nm)

10N 35 20.5 4.7 15.5 88 7.9

5N/5A 31 15.0 6.4 10.1 57 12.7
5A/5N 36 12.7 7.6 12.1 56 8.0
5N 1 5A 34 12.0 8.0 6.2 50 9.1
10A 7 2.0 63.5 N.D. 1.3 8.2
a
After reduction at 673 K.
b
TOF was based on the amount of total H2 uptake of fresh catalysts.
c
Adsorption temperature was 373 K.
d
Calculated by dp (nm) ˆ 96/D (%).
1586 S. Sun et al. / Fuel 81 (2002) 1583±1591

Table 3 Table 4
The H2 spillover effect on various catalysts Catalytic behaviors of various Co(20 wt%)/SiO2 catalysts for CO hydro-
genation (Reaction conditions: 513 K, 1.0 MPa, H2/CO ˆ 2, steady state,
Catalyst Measured Calculated W/F ˆ 5 g cat h mol 21)
reduction degree reduction degree
(%) (%) Catalyst CO conv. (%) CH4 sel. (%) CO2 sel. (%) a

10N 88 88 20N 40.2 6.2 1.5 0.86

5N/5A 57 45 10N/10A 59.8 6.5 2.4 0.87
5A/5N 50 45 10A/10N 64.7 6.4 2.4 0.86
5N 1 5A 50 45 10N 1 10A 66.4 6.8 2.6 0.87
10A 1.3 1.3 20A 9.3 7.3 1.8
20N 88 88
10N/10A 63 45
10A/10N 75 45 with the increase of Co loading. It should be noted that only
10N 1 10A 52 45
for the A/N and N 1 A series, cobalt particle size and reduc-
20A 1.4 1.4
tion degree were nearly kept constant when the cobalt load-
ing increased, consequently CO conversion was doubled
when the cobalt loading was doubled. Because cobalt
high dispersion but were dif®cult to be reduced. On the other
carboxylate from cobalt acetate had a strong interaction
hand, cobalt oxide species, from cobalt nitrate, were easy to
with silica [3], it might have adsorbed onto the silica surface
be reduced but had low dispersion. The compromise effect
much faster than the cobalt nitrate and formed some
of the reduction degree and dispersion determined that 5N/
`islands' during the A/N and N 1 A catalyst preparation.
5A catalyst exhibited best performances.
So, when cobalt nitrate was adsorbed on the silica surface
The reaction performances of various Co (20 wt%)/SiO2
covered by these `islands', it did not have enough space to
catalysts were shown in Table 4. Although the CO conver-
form large particles. From this point, using mixing precur-
sion of 5N 1 5A was lower than that of catalyst 5N/5A,
sors appears to be a promising route to prepare an active
nearly the same as that of catalyst 5A/5N, the catalyst
cobalt catalyst with high dispersion and high reduction
10N 1 10A was slightly more active than 10N/10A and
degrees simultaneously.
10A/10N. The catalysts of N 1 A had only single calcina-
tion step, whereas the sequentially impregnated catalysts 3.2. The reaction performance of Co-based catalysts
were calcined twice. The catalyst of 20N was calcined promoted by noble metals
once, but the CO conversion of 10N 1 10A was 1.5 times
as high as that of the catalyst 20N. Fig. 2 shows the CO conversion, CH4 and CO2 selectivity
It is interesting that the activity of the N and N/A series as a function of Ru loadings. It was found that the catalytic
catalysts increased 1.4 times when the Co loading was activity was apparently dependent on the amount of Ru
doubled. It was doubled for the A/N and N 1 A series cata- added in the system. The catalytic activity showed an initial
lysts when the Co loading was double. The turn over sharp increase with Ru loadings, reaching maximum at
frequency (TOF) only exhibited slight change for all cata- around 0.2% (absolute weight percent, Ru/(Ru 1
lysts. For the 10N and 20N catalysts, the particle size of 20N Co) ˆ 1.16 mol%). CH4 selectivity did not show a mean-
was larger than that of 10N, as shown in Tables 2 and 5. As a ingful change. CO2 selectivity increased slightly with Ru
result, even if the cobalt loading was doubled, the active site loadings.
number only increased 1.4 times. For the N/A series cata- The reaction performance of various cobalt-based cata-
lysts, the cobalt particle size and reduction degree increased lysts prepared from a mixed precursor promoted by Ru is
Table 5
The characterization results of Co(20 wt%)/SiO2 catalyst (N.D.: not detectable)

Catalyst a H2 chemisorption XRD Reduction degree O2 TOF b

titration at 673 K (%) ( £ 10 2 s 21)
H2 uptake c mmol H2/g Average dp d of Co 0 (nm) Co D (%) Average dv of Co 0 (nm)

20N 50 28.7 3.4 24.6 88 7.5

10N/10A 56 18.3 5.2 16.8 63 9.9
10A/10N 95 12.9 7.5 15.1 75 6.3
10N 1 10A 70 12.1 7.9 15.0 52 8.9
20A 10 2.3 42.1 N.D. 1.4 8.6
a
After reduction at 673 K.
b
TOF was based on the amount of total H2 uptake of fresh catalysts.
c
Adsorption temperature was 373 K.
d
Calculated by dp (nm) ˆ 96/D (%).
 S. Sun et al. / Fuel 81 (2002) 1583±1591 1587

Table 7
Catalytic performance of various Co(10 wt%)/SiO2 catalysts for CO hydro-
genation (Reaction condition: 513 K, 1.0 MPa, CO/H2 ˆ 1/2, W/F ˆ 5 g
cat h mol 21, total cobalt loading: 10 wt%)

Catalyst CO conv. (%) CH4 sel. (%) CO2 sel. (%) a

a
Co 33.5 8.8 1.3 0.83
RuCo a 72.3 8.8 3.3 0.76
PtCo a 49.5 13.8 9.5 0.75
PdCo a 57.9 17.8 4.8 0.75
Ru 2.5 4.8 0.1
RuCo b 56.7 6.3 2.2 0.88
Co c 29.8 6.6 1.6 0.86
a
Catalyst precursor was the mixture of cobalt nitrate and cobalt acetate,
molar ratio of nitrate to acetate was 1; noble metal loadings were 0.2 wt%.
b
10 wt% Co (from cobalt nitrate) and 0.2 wt% Ru by coimpregnation.
c
10 wt% Co (from cobalt nitrate).
Fig. 2. Dependence of CO conversion, CH4 and CO2 selectivity on the
Ru loadings. Reaction conditions: 513 K, 1.0 MPa, W/F ˆ 5 g cat h mol 21,
H2/CO ˆ 2.
Ru exhibited high C51 selectivity and large chain growth
probability (a ) [7]. The same phenomenon was observed
on the catalyst RuCo b as seen in Table 7. The catalyst
shown in Table 6. It clearly shows that the catalytic activity, prepared from the mixture of cobalt nitrate and cobalt
CH4 and CO2 selectivity did not have a meaningful differ- acetate, and promoted by Ru, Pt or Pd, however, showed a
ence. According to the results, it was estimated that the Ru smaller chain growth probability. It was even smaller
promoting effect on cobalt±silica catalysts, prepared from than that of the catalyst derived from a mixture of cobalt
the mixed precursor with different procedures, was nearly salts. As explained below, PdCo and PtCo had small
the same. The difference of reduction or metallic cobalt metallic particles, which were not favorable for chain
particle size among N/A, A/N, and N 1 A was weakened growth reaction. RuCo had a ruthenium-rich metallic
by the strong effect of the added Ru. Based on this fact, all of surface, although its metallic particle size was large,
catalysts prepared by co-impregnation were investigated as where hydrogen could be activated effectively. The hydro-
seen below. gen-rich atmosphere, on the surface of RuCo, would
The results of the catalytic activity of Ru, Pt and Pd increase the chain-termination probability and stop the
promoting catalysts are summarized in Table 7. Although chain growth easily.
the CO conversion of Co catalyst was 33.5%, the value of
the RuCo catalyst was 72.3%. However, if the catalyst 3.3. Dispersion and particle size of cobalt
precursor was only cobalt nitrate and also promoted by Table 8 shows the results of the catalyst characterization
0.2 wt% Ru, the CO conversion was only 56.6%. According regarding the dispersion and particle size of cobalt. It is
to Table 7, the activity order was RuCo . PtCo . PdCo . interesting that the amount of H2 adsorption on the catalyst
Co . 0.2 Ru. The CO conversion of catalyst 0.2 Ru itself promoted by Pt or Pd, was much larger than that of the other
was only 2.5% because of low loading. It seems that ruthe- two catalysts. The total H2 uptake by catalysts was in the
nium itself in the RuCo catalyst did not contribute directly following order: PtCo . PdCo . RuCo . Co. Based on the
to the activity of RuCo. amount of O2 consumption at 673 K, the fraction of metallic
The catalyst promoted by Pd and Pt displayed higher cobalt was calculated using the stoichiometric formula
methane selectivity than that promoted by Ru or Co described below:
only. The methane selectivity decreased in the order of
PdCo . PtCo . RuCo, Co. Iglesia et al. reported that the 3Co0 1 2O2 ! Co3 O4
catalyst prepared only from cobalt nitrate and promoted by
Pt0 1 1=2O2 ! PtO
Table 6
The reaction performance of Ru promoted cobalt-based catalysts (Reaction Pd0 1 1=2O2 ! PdO
condition: 513 K, 1.0 MPa, H2/CO ˆ 2, W/F ˆ 5 g cat h mol 21, total cobalt
loading: 10 wt%)
2Ru0 1 3=2O2 ! Ru2 O3
Catalyst CO conv. (%) CH4 sel. (%) CO2 sel. (%)
Assuming that all reduced Co and noble metal species
Ru 1 5N 1 5A 72.3 8.8 3.3 can be oxidized at 673 K, and all noble metal atoms are
Ru 1 5N/5A 72.5 7.5 3.9
completely reduced at the pretreatment conditions.
Ru 1 5A/5N 70.9 8.0 3.2
Based on this assumption, the reduction degrees of Co in
1588 S. Sun et al. / Fuel 81 (2002) 1583±1591

Table 8
The characterization of various catalysts (B: before reaction; A: after reduction; R: after reduction; O: before reduction)

Catalyst a Uptake (mmol g 21) Reduction degree b (%) Dispersion c (%) Particle size (nm) TOF d ( £ 10 2 s 21)

Total H2 e Total CO f Irre. CO f H2(ad) g R XRD

B A R O

Co 36.6 37.3 46.3 32.6 50 8.63 11.12 5.94 7.64 8.2

RuCo 58.6 56.9 93.6 88.9 99.8 6.91 13.09 10.28 6.61 11.1
PtCo 108.9 104.4 186.7 141.9 56 22.67 4.15 3.12 6.49 4.1
PdCo 90.4 85.6 145.2 102.1 63 16.62 5.63 4.38 7.38 5.8
a
After reduction at 673 K.
b
Assuming the noble metals are completely reduced at 673 K.
c
Assuming the stoichiometry Had/Cos or Ms ˆ 1 (M ˆ Ru, Pt and Pd).
d
TOF was based on the amount of total H2 uptake.
e
Adsorption temperature was 373 K.
f
Adsorption temperature was room temperature.
g
Calculated by ds (nm) ˆ 96/D (%).

monometallic and bimetallic catalysts were calculated, as This calculation method was used by Cheng et al. [14]. For
shown in Table 8. The reducibility of cobalt oxide increased the Ru-added catalyst, the dispersion decreased. The
from 50 to 56±99.8% by the addition of noble metals. This decrease in dispersion by the addition of Ru might result
is a major reason why the (M 1 Co) catalyst exhibited a from the high higher reduction degree of the reduced Co and
higher activity than Co. The dispersion of the catalyst is Ru. But for the Pt- and Pd-added catalysts, the dispersions
de®ned by the ratio of …Cos 1 Ms †=…Co0 1 M0 †; where Cos increased remarkably. It is obvious from Table 8 that the
or Ms represents the number of surface Co, or noble metal, reducibility of Pt and Pd-added catalysts only increased
obtained from H2 adsorption, assuming H/Cos or H/Ms to be slightly but adsorbed large amounts of H2. Thus, it is clear
1. Co0 or M0 represents the metallic state Co or noble that Ru increased the cobalt reducibility remarkably,
metals. It should be noted that oxidized cobalt species however Pt and Pd had little contribution to the cobalt
were not included in the calculation of the metal dispersion. reduction. On the other hand, Ru had little effect on the
dispersion, while Pt and Pd had signi®cantly in¯uences on
the cobalt metal dispersion. Furthermore, the Ru-added
catalyst exhibited large TOF; but the TOF values of Pt-
and Pd-added catalysts were even smaller than that of
catalyst Co, as in Table 8.
H2 chemisorption of the used catalyst, after reaction,
showed that the amount of H2 adsorption changed little.
This indicates that most of the surface cobalt kept its metal-
lic state during reaction. It seems that at the reaction stage
the molecules H2O and CO2 did not oxidize metallic cobalt
sites on the surface.
The XRD spectra exhibited only the cobalt peaks, as
shown in Fig. 3. No peaks attributed to Ru, Pt or Pd were
observed. The particle sizes calculated from the XRD spec-
tra by Scherrer's equation are also shown in Table 8. It is
interesting that all of the catalysts had nearly the same
particle size at oxide state. However, after reduction, size
of metallic cobalt particles changed signi®cantly. The XRD
pattern of the metallic cobalt exhibited narrow and intensive
peaks only on the Ru-added catalyst, which also provided
the evidence that Ru-added catalyst had a higher reduction
degree and formed large metallic cobalt particle after reduc-
tion. For the other three reduced catalysts, only very small
peaks of unreduced CoO were observed. It should be
mentioned that the reduced catalyst samples used here
Fig. 3. XRD patterns of Co, RuCo, PtCo and PdCo catalysts. (A) after were in passivated form, and the reduced bulk metallic
reduction; (B) before reduction. cobalt was covered by a tailor-made oxide membrane before
 S. Sun et al. / Fuel 81 (2002) 1583±1591 1589

Fig. 5. FTIR spectra of CO adsorbed on Co, RuCo, PtCo and PdCo

catalysts.

it was crushed for the XRD measurement. Some cobalt

oxides might exist in the membrane besides the originally
unreducible cobalt oxide.
Individual metallic particles in the bimetallic Ru±Co, Pt±
Co and Pd±Co catalysts were analyzed by EDS. Represen-
tative EDS spectra are shown in Fig. 4. The compositions of
some bimetallic particles resulting from EDS are listed in
Table 9. The results give clear evidence that the noble
metals had an intimate interaction with Co, and alloys of
RuCo, PtCo and PdCo formed with different composition. It
was found that Ru was very rich in alloy RuCo, compared to
its original composition (1.16 mol%). Composition, of
alloys such as PtCo and PdCo, was near to the original
preparation ratio. It is suggested that Ru only enriches on
the surface of large Co crystallites. On the surface of small
Co crystallites, it is still Co rich. The difference of the
composition of alloys may lead to the change in the reduc-
tion degree and the size of metallic crystalline during cata-
Fig. 4. Energy dispersive spectrum obtained from individual metal particle lyst preparation. The Ru was enriched on the metallic cobalt
in RuCo, PtCo and PdCo catalysts. Electron beam size: 0.5 nm.
surface but Pt or Pd formed alloy with cobalt.

Table 9 3.4. FTIR measurement

Metal composition in the alloy as determined by EDS
The transmission FTIR spectra of the adsorbed CO of the
Catalyst Particle no. Metal composition (at.%) reduced catalysts are compared in Fig. 5. For the Co cata-
Noble metal Co
lyst, an intense peak at 2014 cm 21 with two shoulder peaks
at 2065, 1943 cm 21 and weak peaks at 2181, 1814 cm 21
RuCo 1 93.3 6.7 were observed. The peak at 2014 cm 21 was attributed to
2 77.7 22.3 CO adsorbed on cobalt metal in a linear geometrical
3 85.4 14.6
PtCo 1 5.2 94.8
progression [15±18]. The shoulder peak at 2065 cm 21 was
2 6.4 93.6 attributed to the surface carbonyl species, Co(CO)x (where
3 5.1 94.9 x . 1), which readily happened to corner sites on the cobalt
PdCo 1 13.2 86.8 metal, and the shoulder peak at 1943 cm 21 was due to the
2 7.9 92.1 bridge-type CO on Co sites [17,19]. The weak peak at
3 7.4 92.6
1814 cm 21 was attributed to an adsorbed CO onto multifold
1590 S. Sun et al. / Fuel 81 (2002) 1583±1591

surface. The shift of linearly adsorbed CO from 2014 to

2006 cm 21 was mainly due to the increase of the Co particle
size of the RuCo catalyst.
From the view point of particle size effect, the wave
number of linearly adsorbed CO on PtCo or PdCo should
be higher than 2014 cm 21, the analogy on the Co catalyst, as
the metallic particle sizes of PtCo and PdCo were smaller
than that of the Co catalyst. It is estimated that due to the
electronic shift from Pt or Pd to CO, the wave numbers of
linear CO on PtCo and PdCo were 1997 and 2001 cm 21,
respectively, which were lower than that of the Co catalyst.
This electronic effect might also be the reason why bridge
CO on PdCo and PtCo had a stronger peak than that on Co
catalyst.

3.5. XPS analysis

Four catalysts in the reduced state were investigated by

XPS. The spectra recorded are shown in Fig. 6. In catalysts
Co, RuCo, PtCo and PdCo, the binding energy (BE) of the
Fig. 6. XPS spectra of cobalt in the noble metal promoted Co catalysts. Co2p3/2 located at the peaks of 780.1, 779.4, 778.6, and
778.5, respectively. The Co2p3/2 peaks of catalysts PtCo
and PdCo obviously appeared at lower position than that
site [20]. The peak at 2181 cm 21 was an attributed to the CO
of catalyst Co, indicating that the electron shifted from Pt
adsorbed on Co n1 (n ˆ 2,3) species [20], as surface cobalt
or Pd to Co. This result was in accordance with the EDS and
oxide species derived from supported cobalt acetate were
FTIR very well. For RuCo, the peak of Co2p3/2 was sharp.
dif®cult to be reduced.
This feature was attributed to the fact that Co existed at
For the catalyst PtCo and PdCo, the band of CO adsorbed
metallic state. The line shape of the other three catalysts
on cobalt metal in a linear model shifted from 2014 cm 21 on
were broad, providing the evidence that the peaks consist
Co catalyst to 1997, 2001 cm 21, respectively. Similarly, the
of metallic Co and Co(II) [23]. From O2 titration data in
peak at 2065 cm 21 derived from surface carbonyl species
Table 8, the catalyst RuCo has the highest reduction degree.
Co(CO)x and the peak at 1950 cm 21 was from CO in bridge
The BE value of Co2p3/2 clearly shows that Pt and Pd have the
adsorption state. For the catalyst RuCo, the peak of CO
intimate interaction with cobalt species in these catalysts.
adsorbed in linear form shifted to 2006 cm 21, and the
peak of Co(CO)x at 2065 cm 21 shifted to 2054 cm 21. At 3.6. Relationship between structure and reaction behavior
peak 1950 cm 21, the bridge adsorbed CO, was very strong for noble-metal-added catalysts
in RuCo. It is believed that the bridge-type CO was easily
formed on large Co particles here. As previous pointed out, The enrichment of Ru metal, at the cobalt surface, acti-
the bridge-type CO was much more active than the linear- vated gaseous hydrogen activated more easily on Ru to
type CO [21]. The high activity of the RuCo catalyst could reduce the cobalt oxide effectively during the catalyst reduc-
be attributed to the increase in bridge-type adsorbed CO, tion [12]. Consequently, the reduction degree of cobalt was
which was easily dissociated to carbon and oxygen. From very high as shown in Table 8. It is highly signi®cantly that
EDS, Pt or Pd dispersed very well in cobalt to form an alloy the high reduction degree of cobalt makes the metallic
of uniform composition. It is considered that an electronic cobalt crystal growth effectively, as no cobalt oxide phase
effect, a kind of legend effect, appeared in PtCo and PdCo could stop the sintering of the cobalt metal phase. On the
catalysts since linear CO peaks of these catalysts shifted contrary, Pd or Pt highly dispersed in the alloy could not
remarkably to the lower side, even if the Co particle size form large bulk noble metal crystalline and concentration of
decreased. From IR spectra of PtCo and PdCo, the new peak the noble metals at the surface of cobalt is lower than 1%.
derived from adsorbed CO with the structure between the More importantly, Pd or Pt atoms in the form of an alloy of
linear and the bridges mode might form, resulting in a this kind of composition lost their original metallic property.
shoulder-peak at 1950 cm 21. It suggested that this electronic As indicated in XPS results in Fig. 6, due to the electron
effect determines the lower TOF and higher methane selec- shift from noble metal to cobalt, the noble metal exists in a
tivity [22]. For RuCo, Ru became very rich on cobalt cation-like state, especially for Pd and Pt. This kind of
surfaces from EDS and bulk cobalt contained little Ru. valency changes make Pt and Pd less effective to hydrogen
This might lead to large supported cobalt particles and activation, to some extent, both in the catalyst reduction as
more bridge-type adsorbed CO. Legend effect was not as well as the preparation process and the FTS reaction itself.
obvious in RuCo as in Ru enriched at the Co particle Based on the assemble effect as well as the legend effect of
 S. Sun et al. / Fuel 81 (2002) 1583±1591 1591

the PdCo and PtCo alloys, reduction degree is low and was enriched at Co surface. This kind of structure deter-
cobalt particle size is small. mined the different reaction behaviors of PtCo, PdCo and
From the IR results, bridge-adsorbed CO appeared in a RuCo catalysts. (4) The Ru-rich spots on the Co surface
large extent on RuCo which is generally considered as determined a high reduction degree, as well as a large parti-
more reactive adsorbed CO. Due to the large crystalline cle of RuCo. This structure formed during the reduction
size of cobalt in the RuCo catalyst, more ¯at cobalt process of the catalyst. Most of the Pd and Pt were
surfaces could be provided to form bridge-type CO [24]. embedded in the bulk phase of Co or CoOx, not contributing
This feature determined the high activity of RuCo. From to the reduction of supported cobalt oxides.
Table 7, the chain growth probability of RuCo was slightly
larger than those of PtCo and PdCo, which agreed well
with the observation of Jacobs and Nijs [24]. Large
Acknowledgements
supported metal favored carbon chain growth, as indicated
also by Bell and Kellner [25] and Tuel et al. [26]. As this
This work was partly supported by NEDO (contract
reaction was conducted in a semi-batch ¯ow-type reactor,
number: 97E10005) and Sasakawa Scienti®c Research
hydrocarbons formed especially heavier one readily stayed
Grant from The Japan Science Society.
inside the reactor to receive secondary hydrocracking. It
was not easy to obtain an obvious difference in the chain
growth probability among these three catalysts. If a reaction
was carried out in a gas-phase reactor, a more distinct References
difference in the chain growth probability would have
appeared [27]. [1] East±West Center Energy Advisory. Economic conversion of natural
gas to liquid synthetic fuelsÐthe next megatrend, Honolulu; January
Compared to other catalysts in Table 7, the noble-metal-
1997.
added catalysts showed lower chain growth probability. One [2] Schulz H, van Steen E, Claeys M. Stud Surf Sci Catal 1994;81:204.
reason is the strong ability of hydrogen activation of the [3] Iglesia E. Appl Catal A 1997;161:59.
noble metal, which introduced much hydrogen to the cata- [4] Iglesia E, Soled SL, Fiato RA. J Catal 1992;137:212.
lyst surface, enhancing the methane formation and suppres- [5] Johnson BG, Bartholomew CH, Goodman DW. J Catal 1991;128:231.
[6] Matsuzaki T, Takeuchi K, Arakawa H, Hanaoka T, Sugi Y. In:
sing the carbon-chain growth process. The other reason was
Yoshida S, Takezawa N, Ono T, editors. Catalytic science and tech-
the existence of acetate-derived cobalt crystalline with a nology, vol. 1. Tokyo/Weinheim: Kodansha/VCH, 1991. p. 249.
very small size [28]. For either the ruthenium-added cata- [7] Iglesia E, Soled SL, Fiato RA, Via GH. J Catal 1993;143:345.
lysts or the non-addition catalysts, the chain growth prob- [8] Schanke D, Vada S, Blekkan EA, Hilmen AM, Hoff A, Holmen A. J
ability of the catalyst from the mixture of acetate and nitrate Catal 1995;156:85.
[9] Van't Blik HFL, Koningsberger DC, Prins R. J Catal 1986;97:210.
was lower than that of the catalyst from nitrate only. As the
[10] Zowtiak JM, Bartholomew CH. J Catal 1983;83:107.
cobalt metallic crystalline reduced from acetate was very [11] Zsoldos Z, Hoffer T, Guczi L. J Phys Chem 1991;95:795.
small, the observed overall carbon-chain growth probability [12] Matsuzaki T, Takeuchi K, Hanaoka T, Arakawa H, Sugi Y. Catal
of the catalyst with dual cobalt precursors is low. It seems Today 1996;28:251.
that a part of the acetate-derived cobalt existed separately, [13] Matsuzaki T, Takeuchi K, Hanaoka T, Arakawa H, Sugi Y. Appl
Catal 1993;105:159.
without sintering or combination with nitrate-derived large
[14] Cheng YG, Tomishige K, Yokoyama K, Fujimoto K. Appl Catal
cobalt metals. 1997;165:335.
[15] Sato K, Inoue Y, Kojima I, Miyazaki E, Yasumori I. J Chem Soc
Faraday Trans 1984;80(1):841.
4. Conclusions [16] Mohana KR, Spoto G, Zecchina A. J Catal 1988;113:466.
[17] Dees MJ, Dchidi T, Iwazawa Y, Ponec V. J Catal 1990;124:530.
(1) Using a mixture of cobalt nitrate and cobalt acetate as [18] Lapidus A, Krylova A, Kazanskii V, Borovkov V, Zaitsev A,
precursors for impregnation resulted in highly active cobalt Rathousky J, Zukal A, Jancalkova M. Appl Catal 1978;73:65.
silica catalyst. The addition of small amounts of noble metal [19] Heal MJ, Leisegang EC, Torrington RG. J Catal 1978;51:314.
[20] Ischi S, Ohno Y, Viswanathan B. Surf Sci 1985;161:349.
increased the catalyst activity effectively. (2) The added [21] Fujimoto K, Kameyama M, Kunugi T. J Catal 1980;61:7.
noble metal had a signi®cant effect on the methane selectiv- [22] Vannice MA. Catalysis, science and technology, vol. 3. Berlin:
ity. Pt and Pd added catalysts showed higher a methane Springer:1982.
selectivity, which was attributed to the small metallic cobalt [23] Chuang TJ, Brundle CR, Rice DW. Surf Sci 1976;59:413.
particle and the H2 spillover effect. (3) The EDS and XPS [24] Nijs HH, Jacobs PA. J Catal 1980;65:328.
[25] Kellner CS, Bell AT. J Catal 1982;75:251.
results con®rmed that noble metals and cobalt had an inti- [26] Barbier A, Tuel A, Arcon I, Kodre A, Martin A. J Catal 2001;200:106.
mated interaction. Pd or Pt formed well-dispersed alloys [27] Fan L, Yokota K, Fujimoto K. Top Catal 1995;2:267.
with Co, while Ru formed two distinct phases where Ru [28] Sun S, Fan L, Fujimoto K. Chem Lett 1999:343.