You are on page 1of 20

~ APT

CAPA
LE
IY
LDSS
I
A: GENERAL
ELSEVIER Applied Catalysis A: General 161 (1997) 59-78

Review

Design, synthesis, and use of cobalt-based


Fischer-Tropsch synthesis catalysts
Enrique Iglesia*
Department of Chemical Engineering, Universi~, ~ff"California. Berkeley. CA 94720, USA

Received 24 September 1996: accepted 13 May 1997

Abstract

Catalyst productivity and selectivity to C5+ hydrocarbons are critical design criteria in the choice of Fischer-Tropsch
synthesis (FTS) catalysts and reactors. Cobalt-based catalysts appear to provide the best compromise between performance
and cost for the synthesis of hydrocarbons from CO/H2 mixtures. Optimum catalysts with high cobalt concentration and site
density can be prepared by controlled reduction of nitrate precursors introduced via melt or aqueous impregnation methods.
FTS turnover rates are independent of Co dispersion and support identity over the accessible dispersion range (0.01-0.12) at
typical FFS conditions. At low reactant pressures or conversions, water increases FI'S reaction rates and the selectivity to
olefins and to C5+ hydrocarbons. These water effects depend on the identity of the support and lead to support effects on
turnover rates at low CO conversions. Turnover rates increase when small amounts of Ru (Ru/Co < 0.008 at.) are added to Co
catalysts. C5+ selectivity increases with increasing Co site density because diffusion-enhanced readsorption of (~-olefins
reverses, ,3-hydrogen abstraction steps and inhibits chain termination. Severe diffusional restrictions, however, can also
deplete CO within catalyst pellets and decrease chain growth probabilities. Therefore, optimum C5+ selectivities are obtained
on catalysts with moderate diffusional restrictions. Diffusional constraints depend on pellet size and porosity and on the
density and radial location of Co sites within catalyst pellets. Slurry bubble column reactors and the use of eggshell catalyst
pellets in packed-bed reactors introduce design flexibility by decoupling the characteristic diffusion distance in catalyst pellets
from pressure drop and other reactor constraints. ~ 1997 Elsevier Science B.V.

Keywords: Fischer-Tropsch synthesis; CO hydrogenation; Reaction-transport coupling; Cobalt catalysts; Synthesis gas
conversion; Cobalt-ruthenium alloys

I. Introduction chain length and functionality. Chain growth occurs


by addition of surface methylene species to adsorbed
The Fischer-Tropsch synthesis (FFS) is a surface- alkyl groups; these alkyl groups can undergo/3-hydro-
catalyzed polymerization process that uses CHx mono- gen abstraction to form linear ct-olefins or hydrogen
mers, formed by hydrogenation of adsorbed CO, in addition to form n-paraffins on Co catalysts [1,2]. f~-
order to produce hydrocarbons with a broad range of hydrogen abstraction is a reversible chain termination
step at typical Fischer-Tropsch synthesis conditions
*Corresponding author. Fax: (+1-510) 6424778; E-mail: and the readsorption of c~-olefins leads to the re-
iglesia(o)cchem.berkeley.edu initiation of alkyl surface chains and to the formation

0926-860X/97/$17.00 1997 Elsevier Science B.V. All rights reserved.


P l l S 0 9 2 6 - 8 6 0 X ( 9 7 ) 0 0 1 86-5
60 E. lglesia/Applied Catalysis A: General 161 (1997) 59-78

(Cn) paraffins

kp l kh
kp
(Cn'l*) ~'- "~ I (Cn*)I-" " (Cn+l*)

kc ks
(Cm,Cn.m) ~ ..... (Cn) olefins .... ~ (Cn)
paraffins , paraffins
,ko
I
T
(Cn+ 1 OH)
alcohols secondary
reactions
Scheme 1. Chain growth pathways and possible secondary reactions of olefins in the Fischer-Tropsch synthesis.

of larger hydrocarbons [3-8]. The proposed mechan- the logarithm of the molar hydrocarbon size distribu-
ism for chain growth in FFS and some potential tion is proportional to chain size. Reported chain
secondary reactions (olefin hydrogenation, hydroge- growth kinetics, however, often depend on chain size
nolysis, and hydroformylation) are shown in Scheme and lead to non-Flory carbon number distribution on
1. These reaction pathways are well established for Co Co, Ru, and Fe catalysts [3-6,9-11].
and Ru FTS catalysts. On these catalysts, light hydro- In this review, we describe the effects of the size of
carbons consist mostly of linear c~-olefins and heavier metal crystallites, the support identity, and the pre-
products are predominantly paraffinic, unless hydro- sence of an alloying element on the turnover rate and
genation or isomerization sites are present. Fe-based selectivity on supported Co catalysts. These effects
catalysts, which are usually operated at higher tem- arise at the level of a catalytic site and reflect the
peratures, show higher selectivity to oxygenates and influence of site structure, composition, and electronic
branched hydrocarbons, suggesting more complex properties on the rate of the primary chain growth
termination pathways and secondary reactions, but catalytic steps. In addition, we explore the subtle
chain growth pathways remain similar to those effects of secondary reactions, such as c~-olefin re-
described in Scheme 1[5]. In particular, CO insertion adsorption and hydrogenation, and of diffusional
termination pathways that form alcohols and ",/-hydro- restrictions on chain growth pathways; these processes
gen abstraction steps leading to internal olefins influence FFS selectivity and introduce catalyst design
become much more prevalent on Fe catalysts because parameters unavailable in kinetic-limited catalysts.
of the higher reaction temperature and multicompo- These effects occur at the level of a catalyst pellet
nent surfaces typical of Fe-based Fischer-Tropsch and our research attempts to exploit intrapellet con-
synthesis. FTS products follow Flory-type carbon centration gradients and radially non-uniform catalyst
number distributions when chain growth and termina- pellets in order to control the selectivity of chain
tion rates are independent of chain size. In such cases, growth reactions. Water, a reaction product in FTS,
E. lglesia/Applied Catalysis A: General 161 (1997) 59-78 61

influences chain growth selectivity and on some cata- size [20,21], and where the number of Co atoms in
lysts, also the rate of the primary CO hydrogenation contact with the support is a very small fraction of the
reaction. total number of metal atoms in a crystal. Later studies
In the characteristic length scale of a catalytic have shown that these observed effects of crystallite
reactor, we discuss the use of slurry bubble column size may be indirect. They appear to reflect the
reactors in order to decouple the chemical design of incomplete reduction of CoO, precursors or the re-
catalyst pellets for optimum selectivity from the pellet oxidation of Co metal as water is formed in FTS
size restrictions imposed by pressure drop considera- reactions [ 18], both of which depend on Co crystallite
tions within packed-bed reactors. Recent advances in size. Similarly, the formation of TiO~ overlayers,
the design of slurry reactors [12-14] have increased which increase methanation turnover rates by favoring
the economic incentives for their application in indus- the adsorption of hydrogen over CO, is rapidly
trial FTS practice. These reactors have also restored reversed by the significant water concentrations pre-
the pre-eminent importance of surface reactions in sent at useful FTS reactant pressures and conversions
FFS by lessening the significant heat and mass transfer [3,8]. As we discuss below, the reversal of these TiO~
disguises common in packed-bed reactors. decoration processes may also be involved in the rate
The details of the experimental and simulation enhancements induced by the addition of water to low
methods used in the results reported in this review conversion FTS reactors and discussed in detail later
have been reported in detail elsewhere [3-8,15,16]. in this review.
Results mentioned without references were obtained Turnover rates (site-time yields) are not influenced
in our laboratory. by dispersion or support effects at conditions favoring
chain growth (C5+ selectivity > 80%). At 2000 kPa,
473 K, and CO conversion levels above 30-40%, we
2. Design of supported cobalt catalysts for the observe that cobalt-time yields (mole CO converted/g-
Fischer-Tropsch synthesis atom total Co-s) increase linearly with increasing Co
dispersion, irrespective of the chemical identity of the
Fischer-Tropsch synthesis catalysts with high volu- metal oxide support (Fig. 1), over the entire experi-
metric productivity decrease reactor volume require- mental dispersion range (0-0.12). Turnover rates vary
ments and improve significantly process economics. between 1.6 x 10-2s i and 3.0 x 10 2 S-I for a
In general, volumetric productivity is controlled by series of Co catalysts supported on A1203, SiO2, TiO2,
varying the active site density and the turnover rates SiO2-modified TiO2, and MgCr204 (Fig. 2).
characteristic of such sites. Supported Co catalysts These data show that FTS reactions do not detect
with high specific rates require the synthesis of small the presence of any structural surface features that
metal crystallites at high local surface densities on vary as dispersion changes within our experimental
support surfaces and the use of supports or alloy range or any electronic perturbations introduced by
elements that increase the rate per surface Co atom contact with a specific metal oxide support. These
(turnover rate). results suggest that Fischer-Tropsch synthesis is a
At reaction conditions leading to the predominant structure-insensitive reaction by the definition of Bou-
formation of short paraffins (100 kPa, > 493 K), cobalt dart [20], although the experimental dispersion range
atoms at the surface of small crystallites appear to be is not sufficient to reach an unequivocal conclusion,
less active for CO hydrogenation reactions than those because crystallite surface structure and orientation
in larger crystals [17,18]. Also, TiO2 supports appear (and thus turnover rates) depend only weakly on
to lead to higher methanation turnover rates on several crystallite diameter over this dispersion range
metals [19] than other supports, apparently because [20,21]. Recent studies have shown that methanation
TiO, overlayers that form during reduction pretreat- turnover rates are also independent of Co dispersion
ments influence CO and H2 adsorption thermo- on supported catalysts [ 18,22] and of surface orienta-
dynamics. These crystallite size and support effects tion on Co single crystals and films [18,23].
are found in a range of crystallite sizes in which An interesting type of support effect appears at low
surface structure should not be very sensitive to crystal synthesis gas pressures (560 kPa), and at higher pres-
62 E. Iglesia/Applied Catalysis A: General 161 (1997) 59-78

0=o11 •4 SlO= I , ~J=O= ~ Other C


25
r l o.,o. ,co-.u..;
== =o r ,v. I_.=-a-z-,
_ = . __o °

"~' 15 ~'~-'/ ~ loi-


/ , "/

8 1 v

~ I I ~ I I I I I I I
0 0.02 0.04 0.06 0.08 0.1 0.12 o 0.02 0.04 o.oe o.o8 o.1 o.12 o.14
Co Dispersion
Cobalt Fractional Dispersion
Fig. 1. Effect of cobalt dispersion(ratio of surface Co to total Co
atoms), support, and alloying on FTS Co-time yields (moles CO Fig. 2. Effect of cobalt dispersion(ratio of surface Co to total Co
converted/total g-atom Co-s) [tubularpacked-bed reactor; reaction atoms), support, and alloying on FTS turnover rates (moles CO
conditions: 473 K, 2000 kPa, H2/CO=2.05, 55-65% CO conver- converted/surface g-atom Co-s) [tubular packed-bed reactor;
sion, C5+ selectivity>80%; 0.17 mm pellet size]. reaction conditions:473 K, 2000 kPa, H2/CO=2.05, 55-65% CO
conversion, C5+ selectivity>80%; 0.17 mm pellet size].

sures (2000kPa) when CO conversions are low


(<30%). At these conditions, turnover rates on Table 1
TiO2-supported Co and on Co catalysts supported Effect of added water on FTS rate and selectivity
on large-pore, low surface area silicas (80 m 2 g - I ) Avg. H20 Site-time Methane Propene/propane
are significantly lower than on other supported Co pressure yield selectivity ratio
catalysts (Fig. 3), but turnover rates reach values (kPa) (102 S 1) (%)
similar to those reported in Fig. 2 when water is added 51 0.7 9.3 2.3
with the H2/CO reactants or when conversion is raised 115 1.3 7.5 2.4
185 1.6 6.1 2.8
above 30-40% (Fig. 4). This autocatalytic effect of 228 1.9 5.0 2.9
water leads to an increase in reaction rate as CO 350 2.3 4.2 3.0
conversion increases, because of the concomitant 585 2.5 3.8 3.1
increase in water concentration within the reactor 940 2.1 3.5 3.3
(Fig. 4). At CO conversions above 30%, or when 12.1% Co/TiO2, 0.058 dispersion,conditionsas in Fig. 1, 8% CO
water is added to the H2/CO feed at lower conversions, conversion, no H20 added for 51 kPa point.
turnover rate differences among samples disappear, as
shown in Figs. 1 and 2.
Indigenous or added water initially increase FT Thus, it appears that at conditions typical of practical
synthesis rates and olefin and C5+ selectivity on FTS operations (high pressure and CO conversion)
Co/TiO2 catalysts (Table 1, Fig. 4); at high enough support effects disappear, water concentration effects
water concentrations, such effects appear to 'saturate' become 'saturated', and catalyst productivity can be
and rates and selectivities are much more weakly predicted directly from Co dispersion values measured
affected by further increases in water concentration. by hydrogen chemisorption.
E. Iglesia/Applied Catalysis A." General 161 (1997) 59-78 63

3
._. 1001
2.5 o
-i m .... • ....... •/
-i--
r- . . . .
o

1.5
10 - I.U
"~,
. Z~_/~ . . . . . A_A . . . . . . . . .
E / / • 8102(amall pore)
[] SlOe(large pore)

O9
I

~
~--- o.6
09
t
1

O/k
AI=O a
TiO 2

Lzx~ o ~ u,~,o, k 560 kPe


0 I I I
1 I I I I I I I 0 20 40 60 80
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
CO CONVERSION (%)
Cobalt Fractional Dispersion Fig. 4, Autocatalytic water effects on Co supported on TiO2,
A1203, and SIO2. CO conversion effects on FTS turnover rates
Fig. 3. Effect of reactant pressure on F r S site-time yields (moles
CO convened/surface g-atom Co-s) [Reaction conditions: 473 K, (site-time yields) [Conditions as in Fig. 1, conversion varied by
changing bed residence time].
H2/CO=2.05, 55-65% CO conversion, C5+ selectivity >80%:
0.17 mm pellet size[.

The addition of water to H2/CO reactants increases behavior were similar. Rate enhancements caused by
C5+ selectivity and reduces C H 4 yields on Co/SiO2 water were fully reversible upon changes in water
[24], C o ] T i O 2 [25], and Co-Mg-Th-SiO2 [26] cata- concentration.
lysts. The latter study also reports an increase in FFS The addition of water also increases the olefin
rate and in olefin selectivity when water is added to the content and C5+ selectivity of FTS products (Table 1).
synthesis gas feed [25]. C5+ selectivity increased and Water selectively inhibits irreversible chain termina-
t h e C H 4 selectivity decreased as the concentration of tion by hydrogen addition and secondary hydrogena-
indigenous or added water increases in packed-bed tion pathways. Surface modifications by chemisorbed
FTS reactors [25]. water are unlikely to be influenced by the identity of
The underlying processes responsible for these the support for the large Co crystallites of this study
water effects remain unclear. The autocatalytic effects (>10 nm). Thus, it is difficult to explain how such
initially observed on TiO2 supports could be attributed chemical modifications of the surface could be influ-
to the reversal of decoration processes that occur enced by support effects. An alternate explanation,
during catalyst reduction and block Co surface atoms. which is discussed in Section 5, is that small-pore
Water effects, however, were also observed on some supports, such as A1203 and small-pore SiO2, contain
silica samples and even on Co powders, on which the a separate intrapellet water phase, which facilitates
inhibition of chemisorption uptakes with increasing CO and H2 transport within the porous structure and
reduction temperatures, typical of decorated metal increases the accessibility of isolated transport-limited
crystals, was much weaker or did not occur. The regions within porous pellets. This condensed water
removal of a catalyst impurity by water was also phase would form only at much higher water partial
considered, but the concentration of trace elements pressures in larger-pore materials, such as TiO2, large-
on silica samples showing normal and autocatalytic pore SiO2, and low surface area Co powders.
64 E. Iglesia/Applied Catalysis A: General 161 (1997) 59-78

3. Synthesis of supported cobalt catalysts metal dispersions for a given precursor-support pair.
It appears that the exothermic decomposition of nitrate
The limited range of cobalt dispersion shown in the precursors and oxidation of CoO to CO304 during
data of Figs. 1 and 2 reflects the significant synthetic oxidative treatments and the high H20 concentrations
hurdles that exist in the preparation of small Co metal that develop during rapid reduction of oxides and
crystallites on oxide supports. The synthesis of highly nitrates by H2 lead to agglomeration of Co particles
dispersed Co catalysts requires the initial formation of and to low metal dispersion. The elimination of 'cal-
very small CoO or Co304 crystallites. The formation cination' steps, by reducing nitrate precursors directly
of these small oxide clusters in turn requires strong in flowing H2, increases Co dispersion from 0.030-
interactions between the support and the Co precursor; 0.057 (Table 2). In addition, the use of high H2 flow
such interactions tend to interfere with the low-tem- rates and slow temperature ramping protocols lead to
perature reduction of such precursors to Co metal [27]. additional improvements in cobalt dispersion to values
On strongly interacting supports, high reduction tem- above 0.10 (Table 2), even at high Co concentrations,
peratures are required and they lead to extensive such as those required in eggshell catalyst configura-
agglomeration and to the formation of large Co metal tions described later [16]. Subsequent oxidation-
particles. Cobalt precursors can be readily reduced on reduction cycles after initial reduction do not lead
weakly interacting supports, but such supports are to detectable agglomeration of Co crystallites.
unable to stabilize very small precursor crystallites Cobalt dispersions above 0.15-0.20 (5-6 nm crys-
during impregnation and drying. Thus, optimum tallite diameters) are difficult to achieve during synth-
cobalt dispersions are favored by support-precursor esis; they are also difficult to maintain during periodic
combinations with intermediate interaction strength oxidative regenerations at high temperatures, which
[27]. are ultimately required to remove deactivating depos-
For example, alumina-cobalt nitrate and silica- its during long-term reactor operation. Also, small Co
cobalt carboxylate support-precursor pairs lead to very metal crystallites (<5-6 nm diameter) appear to re-
small CoOx particles, which reduce completely only oxidize and deactivate rapidly in the presence of water
above 800 K and form sintered metal particles. Silica- reaction products at typical FFS conditions. Thus, it is
cobalt nitrate and MgCrzO4-cobalt nitrate pairs lead to unlikely that marked improvements in volumetric
CoOx particles of intermediate size during nitrate productivities can be achieved by increasing Co metal
decomposition; these crystallites reduce at 573- dispersions above 0.15. Instead, catalyst productivities
673 K without significant agglomeration, and lead can be increased significantly by the synthesis of
to cobalt dispersions of 0.08-0.11. Co dispersions materials with Co dispersions of 0.10-0.15 at higher
above 0.10 can also be achieved by using carbonyl cobalt concentrations (40-50% wt.).
precursors or by impregnation with concentrated The use of support-precursor pairs with intermedi-
nitrate solutions, followed by the direct reduction of ate interaction strengths and the slow and controlled
nitrate precursors using slow temperature ramping reduction of impregnated precursors appear to be the
protocols (Table 2) [16]. most promising route to the synthesis of supported Co
The procedures used for the decomposition and catalysts with high Co concentrations and modest
reduction of catalyst precursors also influence Co dispersions (0.10-0.15). These methods have been
used in the synthesis of eggshell catalysts, in which
very high concentrations of Co are placed locally
Table 2
within the external regions of large catalyst pellets
Pretreatment effects on Co dispersion [13% wt. Co]SiO2, impreg-
nated to incipient wetness with Co nitrate solution] using cobalt nitrate melt precursors [16]. Co disper-
sions of 0.08-0.12 have been obtained at local Co
Treatment Co dispersion from
concentrations above 50% wt. in these eggshell cata-
hydrogen chemisorption
lysts by using nitrate melts and their slow direct
Air (673 K); H2 (to 623 K at 4 K min - l ) 0.030 reduction in flowing He. These synthesis techniques
H2 (to 623 K at 4 K min ~) 0.057
lead to very high volumetric Co site densities and FTS
H2 (to 623 K at 0.2 K min -1) 0.118
reaction rates. As we discuss in a later section, high
E. Iglesia/Applied Catalysis A: General 161 (1997) 59-78 65

volumetric site densities also lead to high selectivity to of Co surface atoms in monometallic catalysts. The
desired C5+ products, because they favor the reversal addition of very small amounts of Ru (Ru/Co =
of chain termination steps via diffusion-limited re- 0.0067 at.) increases FTS tumover rates on Co/SiO2
adsorption of c~-olefins. and CoFFiO2 catalysts (Fig. 1) [15]. On Co/
TiO2.(I 1.7% wt., Co; 0.022 dispersion), the addition
of Ru increases turnover rates increase from 1.7 ×
4. Alloy effects in cobalt-ruthenium and cobalt- 10 3s i to 5.6 x 10 3s 1 without any apparent
rhenium bimetallic catalysts change in the cobalt dispersion measured on fresh
catalysts or in the rate parameters (activation energy
Catalyst productivity in FTS reactions can also be and reaction orders) observed on monometallic Co
increased by combining Co with another metal in catalysts (Fig. 1). This apparent chemical promotion
order to increase the number of exposed Co sites is enhanced by oxidative pretreatments above 573 K
(structural promotion) or the FTS rate per exposed before catalytic reaction tests. In situ X-ray absorption
Co metal atom (chemical promotion). A second metal measurements have shown that oxidation above 573 K
component in intimate contact with Co can increase increases the number of Co neighbors near Ru atoms
the number of exposed Co atoms by modifying either in reduced crystallites, suggesting that the enhanced
the strength of CoOx-support interactions or the redu- synergistic effects caused by oxidative treatments
cibility of CoO~ precursors. For example, the addition reflect better mixing of Co and Ru components during
of Ru to Co/A1203, Co/SiO2, and Cofrio2 catalysts oxidation [15]. The reduction temperature of bime-
decreases the temperature at which CoOx precursors tallic oxide precursors decreases as the temperature of
reduce to Co metal [15]. This structural promotion the oxidation treatment increase, also suggesting more
requires intimate contact between Co and Ru compo- intimate contact between Co and Ru after oxidation
nents; such contact appears to require oxidation treat- treatments [ 15]
ments that form mobile RuOx species and The higher turnover rates observed on well-mixed
thermodynamically stable Co-Ru mixed oxides. On Co-Ru crystallites cannot be explained by an additive
A1203, the presence of Ru leads to higher apparent Co contribution of surface Ru atoms to the FTS activity of
dispersions, because CoO and Co304 crystallites the exposed surface Co atoms, even if all the added Ru
strongly interacting with the alumina support are more atoms were exposed at the surface. First, turnover rates
easily reduced when Ru is present, as also reported by on Ru catalysts are similar or even slightly lower than
others [28]. These effects of Ru on CoO~ reduction are on Co catalysts [3,7,8], suggesting that the replace-
enhanced by reduction-oxidation cycles that promote ment of a surface Co atom by Ru would make
bimetallic mixing [ 15,28] negligible contributions to the observed reaction rate.
Rhenium is also a structural promoter of supported Also, an additive contribution should not require
Co catalysts, but it acts by preventing agglomeration intimate mixing between the two components. Finally,
of CoO~ particles during calcination treatments and the addition of Ru does not affect FTS reaction rate
oxidative regenerations [29-31]; Re leads to higher Co orders or activation energies, even though these
metal dispersions on TiO2, without influencing FTS kinetic properties are measurably different on mono-
turnover rates. The presence of Re (0.8% wt.) in Co/ metallic Ru and Co catalysts [8].
TiO2 leads to an increase in dispersion from 0.022- The presence of Ru appears to preserve a larger
0.053 (Fig. 1, open squares). This increase in disper- fraction of the surface Co atoms available for catalytic
sion disappears when catalysts are prepared by direct reactions during FTS. The reported turnover rates are
reduction of cobalt nitrate precursors without inter- based on the initial dispersion of Co catalysts, because
vening oxidation at high temperatures. Re oxide of the obvious difficulty in obtaining such measure-
species appear to anchor CoOx clusters or to decrease ments on liquid-filled catalysts pellets after reaction.
their 'volatility' and thus inhibit their migration and The constant turnover rates shown in Fig. 2 for mono-
sintering in oxidizing environments. metallic Co catalysts suggest that the fraction of sur-
A second element can also act as a chemical face Co atoms that remains available for reaction at
promoter and increase the turnover rate characteristic steady-state is not influenced by support or dispersion
66 E. lglesia/Applied Catalysis A: General 161 (1997) 59-78

effects. The higher turnover rates observed on Co-Ru Figs. 5 and 6 also illustrate two unusual but com-
catalysts may reflect a protective effect of the Ru monly observed features of chain growth selectivity in
component on the accessibility of reduced surface FTS reactions. The carbon number distributions do not
Co atoms, perhaps by inhibiting their oxidation or obey Flory chain growth kinetics, which would lead to
the deposition of carbon during FFS reactions. straight lines in the semi-logarithmic distribution plots
These inhibited deactivation effects are consistent of Fig. 5. Instead, the apparent chain growth prob-
with thermogravimetric data and X-ray photoelectron ability increases with increasing chain size, leading to
spectra, which show that carbon deposition from H2/ the curved semi-logarithmic plots of Fig. 5. Also,
CO mixtures occurs at higher temperatures when Ru is Fig. 6 shows an increase in C5÷ selectivity as CO
present and well-mixed with Co on Co]TiO2 catalysts conversion is increased by increasing reactor resi-
and Co foils [15]. The presence of Ru may also dence time, suggesting that either secondary reactions
facilitate the removal of surface oxygen during FFS contribute to chain growth or that one of the reaction
reactions. CO dissociation elementary steps during products (e.g., H20) increases chain growth probabil-
FrS lead to the formation of surface oxygen and ities and product molecular weight. As CO conversion
hydroxyl species that must be removed by reaction increases, C H 4 selectivity decreases, the olefin content
with H2 to form the water product of the synthesis within each carbon number group decreases, but the
reaction. The presence of Ru is likely to increase the chain growth probability for Cls+ chains is only
rate of this surface reduction reaction and to inhibit the weakly influenced by bed residence time.
surface and bulk oxidation of Co crystallites. These Termination probabilities for chains with n carbon
inhibited oxidation effects may increase the stability atoms (13T,,,) can be obtained from complete carbon
of cobalt catalysts during FTS, especially for small number distributions using the formalism described in
crystallites, at the high water concentrations typical of Scheme 2 and first proposed by Herrington [32]. The
backmixed slurry reactors at high synthesis gas con- termination probability of C2-C15 chains decreases
version levels. with increasing bed residence time. Chain termination
appears to slow down, or possibly be reversed, as bed
residence time increases; termination also becomes
5. Chain growth selectivity on supported cobalt less likely or more easily reversed for larger surface
catalysts chains at a given bed residence time (Fig. 7). This
latter effect leads to the non-Flory molecular weight
Chain growth selectivity depends on the size and distributions widely reported in the FTS literature and
composition of cobalt crystallites and on the chemical discussed in detail elsewhere [3,7].
identity of the metal oxide support [3], an unexpected
result in view of the constant site activity observed on
these samples. For example, carbon number distribu- 6. Chain growth reaction pathways and Fischer-
tions on Co/A1203 (19.5% wt. Co, 0.036 dispersion), Tropsch synthesis selectivity
Co/SiO2 (24.8% wt,, 0.042 dispersion), and Co/TiO2
(11.8% wt., 0.022 dispersion) are significantly different Cobalt crystallites on non-acidic supports lead to
(Fig. 5). Also, increasing the dispersion of Co crystal- the selective formation of c~-olefins and n-paraffins at
lites on TiO2 from 0.022-0.058 increases C5+ selec- the high CO pressures and conversions typical of
tivity from 85%-90.5% (Fig. 6) at about 50% CO industrial practice [3,7]. High CO and H20 partial
conversion. A similar increase in C5+ selectivity is pressure inhibit secondary hydrogenation and isomer-
observed when Ru is added (Fig. 6). C5+ selectivities ization reactions on supported cobalt catalysts.
on Ru catalysts are very similar to those measured on
Co. Thus, this increase in C5+ selectivity cannot be 6.1. Chain termination and readsorption of ~-olefins
explained by the presence of isolated Ru sites in Co-
Ru bimetallic catalysts. The addition of Ru increases The readsorption of c~-olefins merely reverses the
apparent turnover rates by a factor of three, without predominant chain termination pathways of small
influencing the initial Co metal dispersion (Fig. 1). surface chains (~-hydrogen abstraction) by re-initiat-
E. Iglesia/Applied Catalysis A: General 161 (1997) 59-78 67

10
24.8% Co/SiO 2
0
5

11.8% C o / T i O 2
©
2

19.5% Co/AI203
~" 1.0 •
I•o-

t-
o 0.2
.o
t~
u
0.1

0.05

0.02

0.01 i i J i i
10 20 30 40 50
C a r b o n N u m b e r (n)
Fig. 5. Support effects on carbon number distribution in supported cobalt catalysts. [0.17 mm pellet size; reaction conditions as in Fig. l].

ing surface chains that continue to grow and ultimately and olefins as bed residence time increases. Clearly,
desorb as larger hydrocarbons (Scheme 1). In effect, readsorption effects become more influential as olefin
termination of alkyl chains via/3-hydrogen abstraction concentrations increase with increasing bed residence
becomes increasingly reversible as olefin concentra- time and CO conversion.
tion or residence time increase; this readsorption The effects of bed residence time and of olefin
reaction also becomes faster as reactive olefins are addition to H2/CO feeds confirm the significant role
retained longer within liquid-filled catalyst pellets. of olefin readsorption on chain termination pathways.
This well-known secondary reaction ([3] and refer- Increasing bed residence time leads to a decrease in
ences therein) leads to an increase in the selectivity to the selectivity to 1-butene, without a corresponding
C5+ hydrocarbons and to lower selectivities to CH4 increase in the selectivity to n-butane or 2-butenes
68 E. lglesia/Applied Catalysis A: General 161 (1997) 59-78

'61 0.3 I I l I I !

0.25

°-
i

ii ""': ~ o.ls

O
% i 0.1

_
0.05

B
r~ ! 1 . ! I I I ,I I I I
0 20 40 60 80 I00 5 10 15 20 25 30 35
CO Conversion (%) Carbon Number (n)

Fig. 7. Bed residence time and carbon number effects on total


Fig. 6. Site density and alloy effects on C5+ selectivity (Co/TiOz)
[A: 11.7% Co, 0.022 dispersion (d), 10-6 g-atom surface Co m 2 chain termination probability. A: 9.5% CO conversion, 2 s bed
residence time; B: 72% CO conversion, 12 s bed residence time
(0co); B: 12.1% Co, d=0.058, 0co=3.3 x 10-6; C: 11.7% Co, (Ru/
[Co/TiO2, 11.7% Co, 0.015 dispersion, 0.17 mm pellet size; 473 K,
CO)at =0.0067, d=0.024, 0Co----1.1x 10-6] [0.17 nun pellets].
2000 kPa, H2/CO=2. I]

13r = I = ¢. / Z®.+I ¢i content in reaction products increases with increasing


bed residence time, as FTS products become domi-
nated by 'dead-end' n-paraffins, which cannot
= 13o + undergo surface reattachment and chain initiation
because of unfavorable thermodynamics. Secondary
hydrogenation reactions appear to be strongly inhib-
On = molar rate of formation of chains ited by water and CO concentrations typical of FFS
with n carbons reactions [3], but these reactions can influence product
selectivity at the low pressure and conversion condi-
Scheme 2. Chain growth parameters in Fischer-Tropsch synthesis.
tions typical of many previous literature reports. As
expected from the reaction pathways (Scheme 1) and
(Fig. 8). Thus, the depletion of o~-olefins, such as 1- the definition of chain termination probability
butene, from FTS products is not caused by secondary (Scheme 2), higher extents of readsorption for larger
hydrogenation or double-bond migration reactions, chains and at long residence time appear to be related
but instead by the selective disappearance of 1-butene, to the low chain termination probabilities observed for
caused by its re-attachment to the surface and sub- these products and reaction conditions.
sequent chain growth. The decrease in olefin/paraffin These bed residence time studies show that c~-
ratios observed with increasing bed residence is also olefins initiate surface chains with very high selectiv-
caused by selective readsorption steps (Fig. 9). Re- ity, even though c~-olefins added to H2/CO feeds
adsorption pathways remain available as long as chain undergo significant hydrogenation (Table 3). Added
growth is terminated by ~3-hydrogen abstraction, but olefins hydrogenate predominantly at the dry condi-
they cannot occur again after chains terminate irre- tions of the reactor inlet, but the water product of FTS
versibly as unreactive paraffins. Thus, the paraffin reactions strongly inhibits the hydrogenation of olefins
E. lglesia/Applied Catalysis A: General 161 (1997) 59-78 69

4 I I I u I I

~an ~ Total C4

m
I1~ n~l--~
A

o~
v

>

o
E
2 - •~ • 1-Butene
(I)

(J
N-Butane
-- V ""--" • V "--- V "----
~ •,-,. V v
1-

2-Butenes
0 -..--,I,-O qp._....., o.----- ,o
0 2 4 6 8 10 12 14
Bed Residence Time (s)
Fig. 8. Bed residence time effects on olefin and paraffin selectivity within C 4 products on Co/TiO2 (catalyst: 11.7% wt. Co, 0.015 dispersion,
9.5-72% CO conversion; reaction conditions as in Fig. 1; selectivity reported on a carbon basis as the percentage of the converted CO
appearing as a given product).

Table 3 dies of the role of readsorption pathways were carried


Hydrogenation and chain initiation reactions of ethylene during out at low pressures and 'dry' conditions. As a result,
FTS [11.7% Co/TiO2, 0.017 dispersion, 2070kPa, <15% CO
conversion, 8% mol C2H4 in feed, HJCO=2.1] [9]
these studies have significantly underestimated the
critical role of these chain initiation pathways in the
Ethylene source Hydrogenation Chain control of carbon number distributions in FTS re-
(%) initiation (%)
actions.
Added (at inlet) with H2/CO 70.5 29.5 The total chain termination probabilities obtained
Formed from H2/CO during FTS 15.1 84.9 from the approach described in Scheme 2 (/~T,n, Fig. 7)
Added (below inlet) with H J C O 31.5 68.5
can be separated into individual termination probabil-
Added (at inlet) with 15% 22.0 78.0
HzO-H2/CO ities for the two predominant termination steps: hydro-
gen addition (¢3h,.) and hydrogen abstraction (/3o, .)
(Figs. 10 and 11). In effect, the contributions of the
formed (along with water) by FTS reactions. The two termination pathways to the formation of hydro-
presence of water in the H2/CO feed or the external carbons of a given length can be obtained from the
introduction of olefins below the 'dry' reactor inlet complete carbon number distribution and the olefin/
markedly increase the chain initiation selectivity of paraffin content for each chain length (Scheme 2). A
added olefins (Table 3). Previous olefin addition stu- similar separation is not possible with the normally
70 E. Iglesia/Applied Catalysis A: General 161 (1997) 59-78

6 I i I I I I t i t 'u i

0
I m
4
Ig

i m

O.
3

%0
\"
2
2s, 5 % CO c o n v e r s i o n _
0
f.
i m

_.¢
0 1 m

• ~0 • •
0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

12s, 60*/0 CO conversion

I I I I I I I I I I I I
0 2 4 6 8 10 12 14 16 18 20 22 24
Carbon N u m b e r (n)
Fig. 9. Bed residence time and carbon number effects on c~-olefin to n-paraffin ratio on Co/TiO2 [Catalyst and reaction conditions as in Fig. 8]

reported values of chain growth probability (rp/rt), paraffins, as shown by a comparison of the chain
because the additive chain termination rates appear in termination probabilities at short (Fig. 10; 2 s) and
the denominator. long (Fig. 11; 12 s)residence times.
The net probability of chain termination to olefins It appears unlikely that the forward rate of chain
decreases with increasing chain length, but the prob- termination to olefins (/3-hydrogen abstraction) would
ability of termination to paraffins is independent of depend strongly on chain length, in view of the chain
chain size (Fig. 10). Therefore, the decrease in total size independent termination rate for the irreversible
chain termination probability with chain size (Fig. 7) hydrogen addition step and of the observed effect of
and the non-Flory carbon number distributions bed residence time on the calculated probability of
(Fig. 5) reflect a selective decrease in the net rate termination to olefins. The results reported here
of chain termination to olefins, without any apparent strongly suggest that the observed decrease in the
influence of chain size on the irreversible H-addition net olefin termination probability reflects the higher
termination step that forms paraffins (Fig. 10). The probability of reversing the t3-hydrogen abstraction
higher C5+ selectivity and paraffin content obtained at step by readsorption as the size of hydrocarbon chains
long bed residence times is caused by a decrease in the increase. This higher probability in turn reflects the
net rate of chain termination to olefins, without sig- slower diffusivity and the generally higher reactivity
nificant changes in the rate of chain termination to of larger olefins, both of which lead to more severe
E. lglesia/Applied Catalysis A: General 161 (1997)59-78 71

0.2 I I

0.15

(X - olefins
0=
¢:
0.1

n- paraffins
O0 "
0.05 • 0(

0
0 10 20 30
Carbon Number (n)
Fig. 10. Chain size effects on termination probabilities to olefins and paraffins [9.5% CO conversion, 2 s bed residence time; CofFiOz, 11.7%
Co, 0.015 dispersion, 0.17 mm pellet size; 473 K, 2000 kPa, H2/CO=2.1]

diffusional restrictions within liquid-filled pores as agreement with experimental measurements of chain
chain length increases [3]. This proposal, however, termination probabilities [3,7]. For example, the
must also be able to account for the otherwise un- curves shown along with the experimental data in
explained effects of alloying, dispersion, and support Fig. 10 were obtained from simulations using these
identity on FTS selectivity (Figs. 5 and 6), Indeed, the reaction-transport models [3,7] and a chain length
severity of diffusional restrictions is typically con- dependence only for the diffusivity and the readsorp-
trolled by varying the volumetric density of active tion rate constant of olefins with n carbon atoms:
sites and the diameter and pore structure of support
(D,,/kr.,,) = (D6/kr.6)exp[-'~(n - 6)]. (l)
pellets.
This exponential dependence includes chain size
6.2. Diffusion-enhanced chain growth and site effects on olefin diffusivity (D,) and reactivity (kr.,)
densiO, effects on selectivity from independently obtained values for 1-hexene and
a value of "~(0.29) adjusted to match the experimental
Rigorous mathematical descriptions of diffusional results in Fig. 4.
transport within catalyst pellets and of chain growth Dimensional analysis of the boundary value pro-
and termination surface kinetics have led to excellent blem posed by a reaction-diffusion mole balance for
72 E. lglesia/Applied Catalysis A: General 161 (1997) 59-78

0.2 I I I 1 I I

0.15

-g /
_./(X - olefins
i 0.1

..~.AA~ AA AA AAAA AAAA~'AAA~AAAA A~


0.05 \ n- paraffins

0
I ,! I I I I
0 5 10 15 20 25 30 35
Carbon Number (n)
Fig. 11. Chain size effects on termination probabilities to olefins and paraffins [72% CO conversion, 12 s bed residence time; Co/TiO2, 11.7%
Co, 0.015 dispersion, 0.17 mm pellet size; 473 K, 2000 kPa, H2/CO=2.1]

each hydrocarbon chain within a catalyst pellet shows a catalyst pellet. This term reflects predominantly the
that diffusional restrictions depend on a dimensionless molecular properties of olefins with n carbon atoms
parameter ~,, which reflects the ratio of maximum and leads to the observed effect of chain size on
diffusion rates to maximum reaction rates for an termination probability and olefin content (Figs. 3
c~-olefin with n carbon atoms: and 4). This ~'n term is only weakly affected by the
support pore structure within the narrow range of void
ff~n = ( k r , n / O n ) x (RZoeOco/rp) = (zb,) × (X) (2)
fraction and tortuosity typical of the type of supports
The severity of transport restrictions, and the extent to used in this study.
which these restrictions enhance the probability of The second term (X) in Eq. (2) contains the average
olefin readsorption, increase as the value of ~bn radius of catalyst pellets (Ro) the density of Co sites
increases. per unit area (0Co), and the void fraction (e) and
The first term (~b,,) in Eq. (2) contains a first-order average pore radius (rp) of the metal oxide support.
rate constant for olefin readsorption reactions (k~,,,) The last three catalyst properties appear as a combined
and the effective diffusivity (D,,) of an c~-olefin within term (eOco/rp), which is proportional to the volumetric
E. lglesia/Applied Catalysis A: General 161 (1997) 59-78 73

density of surface Co atoms (Co surface atoms per unit a 100


pellet volume). This structural term (~) contains
structural catalyst properties that can be independently
measured and it appears in the ~/,,,term for all values of 95-
n. The structural parameter ;~ term depends only on the
number of sites available for chain desorption and 90-
readsorption and on the average distance that mole-
cules must traverse by diffusion within the intrapellet
liquid phase before they are removed by convective 85-
i@
gas flow within intrapellet bed interstices. ¢ • •
The effects of catalyst structure and site density es • • •0
+.,, 80-
suggested by the parameter ~ give rise to an increase o I
qe
in the extent of readsorption and in C5+ selectivity as
Co site density is increased either by increasing Co 75-
loading or dispersion in monometallic catalysts or by
maintaining a larger fraction of the exposed Co sur-
70 I ! t i
face atoms available during steadystate FTS in Co-Ru
10 100 1000 10,000
catalysts (Figs. 5 and 6). Chain termination probabil-
X (10"16 m)
ities decrease with increasing site density (at least on
small pellets) because termination to olefins is more
likely to be reversed when diffusional restrictions b 15
increase intrapellet olefin concentrations and lengthen
CH4
intrapellet residence times. The resulting changes in
C~+ selectivity occur without any changes in the
asymptotic chain growth probability, which reflects
A
the intrinsic rate of the irreversible hydrogen addition 10
termination step. Therefore, the kinetics of intrinsic
chain growth and termination do not need to be
influenced by changes in Co dispersion or support
structure in order to account for their observed effects
on olefin content, carbon number distribution, and C5+
selectivity. o 5
Carbon number and olefin selectivity changes
caused by dispersion, support, and alloy effects can
also be described by the effects of site density and sup-
port structure on the rate of diffusion-enhanced olefin
readsorption reactions. A rigorous comparison of the 0 I I I I
10 100 1000 10,000
predicted and observed selectivity changes induced by
variations in the parameter ~ is shown in Fig. 12(a) X (10-16 m)
and (b) for C5+ and CH,, selectivities, respectively.
Fig. 12. The effect of structural parameters (%) on Fischer-Tropsch
For small values of ~, the selectivity to C5+ pro- synthesis selectivity: diffusion-enhanced readsorption ( . . . . ) and
ducts increases (Fig. 12(a)) and methane selectivity diffusion-inhibited chain growth ( . . . . ) simulations and experi-
decreases (Fig. 12(b)) as ~ increases, because diffu- mental data ( © , dispersion/support effects; A , pellet size
sional restrictions decrease chain termination prob- variations; O , eggshell thickness variations: (a) C5~ selectivity;
(b) CH4 selectivity [473 K, 2000 kPa, H2/CO=2.1, 55-65% CO
abilities and increase average chain size. Fig. 12
conversion] (~ values from Eq. (2), with r v and R,. in m and Oc. in
includes Co, Co-Ru, and Co-Re catalysts of varying surface Co atoms m 2) [3,16].
dispersion supported on A1203, TiO2, and SiO2. The
broken curves in Fig. 12 were obtained using identical
74 E. lglesia/Applied Catalysis A: General 161 (1997) 59-78

values of chain growth kinetics, desorption rate con- as long as these phases co-exist in thermodynamic
stants, and olefin diffusivities for all catalysts. equilibrium. In the absence of transport limitations,
Selectivity modifications caused by alloying, sup- which would prevent attainment of thermodynamic
port, or dispersion changes reflect the indirect effect of equilibrium between phases, the rate of a chemical
these properties on diffusion-enhanced c~-olefin read- reaction cannot depend on the identity of the phase in
sorption. Dispersion, support, and bimetallic effects which the catalyst resides or on the solubility of the
do not influence the intrinsic chain growth properties reactants in this phase. Also, in three-phase reactors,
of Co surfaces, at least in the experimentally acces- the residence time of any products removed predomi-
sible dispersion range; this conclusion is consistent nantly in the gas phase (C1-C2o) is controlled exclu-
with the observed structure insensitivity of CO hydro- sively by gas phase residence time and the reactor gas
genation reactions (Figs. 1 and 2). Excellent agree- hold-up as long as vapor-liquid equilibrium is main-
ment between experiments and simulations can be tained. Thus, a liquid phase holdup of higher olefins in
obtained without varying the nature of the surface the C2-C2o range cannot explain the observed carbon
chain growth kinetics or introducing a carbon number number effects in the absence of transport limitations.
dependence on intrinsic chain termination rates. Thus, Olefin readsorption rates depend only on the olefin
intrinsic chain termination kinetics need not depend thermodynamic activity (not its concentration), which
on carbon number, surface structure, or support mod- at vapor-liquid equilibrium is identical in the liquid
ifications in order to explain non-Flory carbon number and gas phases and independent of solubility at steady-
distributions (Fig. 5) or the observed effect of site state. The rate of readsorption depends only on the gas
density and pore structure on carbon number and phase olefin fugacity, unless olefin partitioning
olefin selectivities (Fig. 6). Clearly, diffusion- between the gas and liquid phases is restricted by
enhanced c~-olefin readsorption processes cannot slow diffusion rates. In our model, a transport restric-
explain the decrease in product molecular weight that tion in the removal of olefins leads to intrapellet olefin
ultimately occurs at values of X above 200 × fugacities higher than in the interpellet gas phase;
1016 m -1. These simulations predict that C5+ selec- these fugacity gradients increase with increasing chain
tivities would reach an asymptotic value of about 92% size because transport restrictions become more
as all olefins disappear from the effluent stream at very severe. In effect, a chemical reaction cannot 'detect'
high values of X. In the next section, we show that the presence of a liquid phase unless the liquid phase
diffusion-limited arrival of CO monomer precursors introduces either a transport restriction or a change in
decrease chain growth probabilities and product mole- the reactive properties of the catalytic surface. Solu-
cular weight in large catalyst pellets and lead to the bility-enhanced olefin readsorption, even if it were
low Cs+ selectivities observed on catalysts with high consistent with thermodynamics, cannot account for
X values. the observed effects of site density and pellet size on
Recently, non-Flory distributions have been incor- selectivity, because neither catalyst property can influ-
rectly attributed to the higher solubility of larger c~- ence in any way vapor-liquid equilibrium.
olefins within intrapellet and interpellet liquids in At very high conversions or on catalysts with sig-
three phase reactors [33-35]; these authors proposed nificant numbers of secondary hydrogenation or dou-
that this higher solubility leads to a selective increase ble-bond isomerization sites, reaction products
in the concentration of heavier olefins, in their reactor approach Flory carbon number distributions. These
residence time, and in the rate and extent of their secondary reactions convert c~-olefins into species
secondary hydrogenation [33-35]. These proposals with much lower readsorption rates and thus minimize
correctly assume that vapor-liquid equilibrium favors readsorption events that cause changes in growth
the presence of larger olefins in FTS liquids, but probability with chain size.
incorrectly conclude that such higher concentrations
lead to faster readsorption rates. Chemical reactions, 6.3. Diffusion-inhibited chain growth
whether catalytic or not, are driven by the chemical
potential of reactants. For a given reactant species, its The selectivity trends caused by diffusion-enhanced
chemical potential is identical in two or more phases, olefin readsorption are ultimately reversed as transport
E. Iglesia/Applied Catalysis A: General 161 (1997) 59-78 75

restrictions become more severe (Fig. 12), leading to a reactant transport restrictions that can inhibit chain
decrease in product molecular weight as X increases growth. CO transport restrictions decrease reaction
above 200 x 1 0 " m -1 . Large pellets and high Co site rates, after an initial increase for modest transport
densities inhibit chain growth, because they lead to restrictions (because of the negative order CO depen-
significant depletion of monomer precursors (CO) dence of the reaction rate at high CO concentrations),
within intrapellet liquids and on catalytic surfaces. by limiting the accessibility of sites near the center of
Intrapellet gradients in the thermodynamic activity of large pellets. The lower intrapellet CO fugacities
CO arise when diffusion rates cannot satisfy the decrease chain growth probability and favor second-
kinetic rate of CO hydrogenation reactions, either ary hydrogenation of c~-olefins and preferential termi-
because of the distance CO must traverse or the nation by hydrogen addition. The presence of water
number of sites that must be supplied with reactants has directly opposite effects; it increases reaction rates
per unit pellet volume. Chain termination probabilities (at least on some catalysts), C5+ and olefin selectivity,
(rt/rp) increase as X increases above a threshold value, and chain growth probabilities of light hydrocarbon
because diffusional restrictions inhibit chain growth chains. Thus, we consider the possibility that the
steps by reducing the supply of monomer species presence of water enhances the rate of transport of
required for chain growth. A dimensional analysis CO and H 2 reactants and increases the accessibility of
of the reaction-diffusion equations for the CO hydro- sites located within the least accessible porous regions
genation steps shows that the severity of CO diffu- of catalyst pellets.
sional restrictions depends on a dimensionless CO and H2 diffusion coefficients are about three
parameter epco that accounts for the relative rates of times larger in water than in typical FT hydrocarbon
CO diffusion and hydrogenation: liquids [36,37]. Capillary effects can lead to the con-
2 densation of a separate water phase within catalyst
(]'CO ---- (~/'CO) × (RocOco/r p) = ('@co) x (X), (3)
pores, or to the formation of a hydrocarbon-water
where (~co is a function of the CO hydrogenation rate emulsion stabilized by minor alcohol components
constant and the intrapellet CO diffusivity and X is among FTS products, at concentrations well below
identical to the structural parameter derived from the the saturation vapor pressure of water at reaction
readsorption-transport equations for c~-olefins, The conditions. Also, diffusional restrictions in the
solid curves in Fig. 12(a) and (b) show the result of removal of the water formed in FT synthesis can lead
simulations using CO hydrogenation rate expressions to significantly higher water fugacities near the center
obtained from kinetic-limited measurements and inde- of catalysts pellets than in the interpellet gas phase.
pendently measured values for the H2 and CO diffu- Transport restrictions lead to intrapellet water gradi-
sivities and solubilities in FTS liquid products ents in the same range of ~ as required for CO
[3,7,16]. These simulations are in excellent agreement gradients, because one water molecule is formed for
with experimental results for large values of X. For- each CO consumed and diffusion coefficients are
tunately, the onset of CO transport restrictions occurs similar for these two species. Preliminary simulation
for X values significantly above those required in order results suggest that the presence of a water or water-
to enhance olefin readsorption rates, because CO rich phase within intrapellet voids increases FT synth-
diffuses through intrapellet liquids much more rapidly esis rates and C5+ and olefin selectivities [38]. Also,
than large olefins ('¢'co<<(,',,). Very few olefins remain the slight model overprediction of C H 4 selectivity at
in the product stream for )( values leading to maximum high values of X (Fig. 12) may reflect the failure to
C5+ selectivities, suggesting that the full benefit of include water concentration gradients in the simula-
olefin readsorption can be achieved before CO deple- tions. Thus, one possible explanation for the autoca-
tion begins to inhibit chain growth. talytic and selectivity effects of water is that it
At this point, we reconsider the observed effects of enhances the rate of CO and H2 diffusion within
water on FTS rate and selectivity in the context of the intrapellet liquids and removes deleterious diffusional
reaction-transport models just discussed. In particular, effects affecting a minority of Co sites, which, how-
we observe that the effects of water on rate and ever, form methane and light paraffins with high
selectivity mimic those predicted by the removal of selectivity. If correct, this proposal would predict that
76 E. lglesia/Applied Catalysis A: General 161 (1997) 59-78

water can act as an indigenous moderator of CO composition on supported Co catalysts. Thus, trans-
diffusional restrictions in PT synthesis. port restrictions introduce catalyst design variables
This simple proposal, however, seems to contradict that are not available in kinetic-limited catalysts.
the observation that water effects appear to be stronger For example, light olefins are favored at very low X
on large-pore materials, in which capillary condensa- values, while light paraffins are formed preferentially
tion is least likely to occur. This apparent contra- at very high values of X- Intermediate X values lead to
diction can be resolved by considering that small- heavy paraffins and minimize the formation of light
pore materials may already contain a separate water products.
phase or a water-hydrocarbon emulsion, even at low Reactor and economic constraints determine the
water concentrations. Thus, these materials do not range of volumetric site densities (productivity) and
require high water concentrations in order to form a pellet size (pressure drop in packed beds). These
water-rich intrapellet liquid phase that makes internal practical considerations may limit the range of X
pellet regions more accessible to reactants. Pellets values that can be achieved using uniformly impreg-
with larger pores would form a water-rich phase only nated pellets in the packed-bed reactors used through-
at higher water partial pressures, because of the out most of experimental studies described in this
weaker capillary effects expected within such larger review. For example, pressure drop constraints may
pores. Independent evidence for this proposal or alter- require the use of large catalyst pellets (1-2 mm
nate explanations for these water effects are still diameter), for which optimum values of X can only
needed. This two-phase model of intrapellet liquids, be obtained at very low site densities (and volumetric
however, appears to provide the most reasonable productivities).
explanation for the observed role of support pore size In such instances, the use of pellets with Co sites
on the autocatalytic and selectivity effects of water preferentially located near the outer pellet surface
during FTS reactions at this time. An alternate expla- allows the characteristic diffusion length to be varied
nation involving reversible decoration of Co surfaces independently of pellet diameter [16,39,40]. The rele-
by reducible support species cannot account for the vant structural factor for such eggshell pellets
water effects observed on Co/SiO2 catalysts, but may becomes:
account for the effect of water o n C o / T i O 2 catalysts. In
X = (r2ogOco/rp), (4)
this case, indigenous or added water can reverse the
blocking of Co surface atoms by re-oxidizing TiOx where ~5ois the thickness of the eggshell layer [3,16].
species responsible for these decoration phenomena. The open symbols included in Fig. 12 show the
experimental results obtained using eggshell pellets
of varying eggshell thickness. These data illustrate the
7. Optimum Cs+ selectivity and catalyst and remarkable flexibility introduced by eggshell config-
reactor design options in the Fischer-Tropsch urations in the design of FTS catalysts. The open
synthesis symbol to the extreme right corresponds to a uni-
formly impregnated large SiO2 pellet (2.2 mm dia-
The experimental and simulation results shown in meter); the right to left sequence of open symbols
Fig. 12 suggest that optimum C5+ selectivities shows how decreasing the eggshell thickness within
(>90%) are obtained on PTS catalysts with intermedi- such large pellets markedly increases C5+ selectivity
ate values of the structural parameter X. The physical and minimizes the formation of CH4 [3,16], without a
structure and the Co site density in FTS catalysts can significant decrease in volumetric catalyst productiv-
be designed and constructed in a way that leads to ity. High productivity can be achieved, in spite of the
desired carbon number distributions and olefin content much smaller support volume and surface area onto
in products. None of these design variables influences which Co sites must be introduced, by using precur-
FTS selectivity on kinetic-limited catalyst pellets, sor-support pairs with intermediate interaction
because intrinsic chain growth probabilities and strength and by decomposing Co nitrate precursors
hydrogen addition and abstraction rates appear to slowly in flowing Hz. Slow reduction of nitrate pre-
be independent of surface structure and of bimetallic cursors on SiO2 leads to Co dispersions of 0.08--0. l 2
E. Iglesia/Applied Catalysis A: General 161 (1997) 59-78 77

even at local Co concentrations of about 50% wt. [6] R.J. Madon, S.C. Reyes, E. Iglesia, J. Phys. Chem. 95 (1991)
7795.
within the eggshell regions [16].
[7] E. Iglesia, S.C. Reyes, S.L. Soled, in: E.R. Becker, C.J.
Eggshell catalysts are useful in packed-bed reac- Pereira (Eds.), Computer-Aided Design of Catalysts and
tors, in which volumetric productivity and pressure Reactors, Marcel Dekker, New York, 1993, p. 199.
drop constraints require the use of pellets with high X [8] E. Iglesia, S.L. Soled, R.A. Fiato, J. Catal. 137 (1992)
values. Slurry bubble column reactors contain much 212.
[9] H. Pichler, H. Schulz, M. Elsmer, Brennst. Chem. 48 (1967)
smaller catalyst pellets and typical X values tend to be
78,
lower than optimum values. Bubble column reactors [10] H. Schulz, K. Beck, E. Erich, Proc. 9th Int. Cnngr. Catal. 2
require pellets that are small enough to ensure com- (1988) 829.
plete fluidization at economical linear velocities, but [11] D. Vanhove, Z. Zhuyong, L. Makambo, M. Blachard, Appl.
large enough to be easily separated from liquid reac- Catal. 9 (1984) 327.
[12] H. Kolbel, M. Ralek, Catal. Rev. 21 (1980) 225.
tion products. This imposes a fairly narrow range of
[13] B. Jager, R.C. Kelfkens, A.E Steynberg, Stud. Surf. Sci.
pellet diameters, usually between 0.01 and 0.1 mm, Catal. 81 (1994) 419.
for practical operation of bubble column reactors [41 ]. [14] B.L. Bhatt, R. Frame, A. Hoek, K. Kinnari. V.U.S. Rao, EL.
In this pellet size range, optimum values of X will Tungate, Topics in Catal. 2 (1995) 235.
require the selective deposition of high Co site den- [15] E. lglesia, S.L. Soled, R.A+ Fiato, G.H. Via. J. Catal. 143
sities near the pellet center or a significant increase in (1993) 345.
[16] E. Iglesia, S.L. Soled, J.E. Baumgartner, S.C. Reyes, J. Catal.
the site density and volumetric productivity of uni- 153 (1995) 00.
formly impregnated pellets. The latter approach is [17] C.H. Bartholomew, R.C. Reuel, J. Catal. 85 (1984) 78.
preferred because it also increases reactor productiv- [18] B.G. Johnson, C.H. Bartholomew. D.W. Goodman, J. Catal.
ity; it requires, however, the synthesis of high-loading 128 (1991) 231.
supported Co catalysts with high Co dispersion [19] M.A. Vannice, Catal. Rev. 14 11976) 153
[20] M. Boudart, Adv. Catal. Rel. Subj. 20 (1969) 85.
(>10%). Slurry liquids that inhibit olefin diffusion [21] R. van Hardeveld, E Hartog, Adv. Catal. Rel. Subj. 22 (1972)
also increase values of G, Eq. (2) and can be used 75.
to compensate for the low values of X typical of small [22] S.W. Ho, M. Houalla, D.M. Hercules. J. Phys. Chem. 94
catalyst particles in slurry bubble column reactors. (1990) 6396.
[23] J.J. Geerlings, M.C. Zonnevylle, C.P. de Groot, Surf. Sci. 241
(1991) 302, 315.
[24] Eur. Pat. Appls. 201 557 (1983), assigned to Shell.
Acknowledgements [25] C.J. Kim, Eur, Pat. Appls. 339 923 (1988) and 355 218
(1990), assigned to Exxon.
The results described in this review were obtained [26] H. Schulz, E. van Steen, M. Claeys, in: H.E. Curry-Hyde, R.E
by the author in collaborations with Stuart L. Soled, Howe (Eds.), Natural Gas Conversion II, Elsevier, Amester-
dam, 1994, p. 455.
Sebastian C. Reyes, Joseph E. Baumgartner, Rostam J.
[27] S.L. Soled, J.E. Baumgartner, S.C. Reyes, E. Iglesia, Proc.
Madon, and Rocco A. Fiato at the Corporate Research Mater. Res. Soc. Symp. 368 (1995) 113.
Laboratory of Exxon Research and Engineering Co. [28] H. Beuther, C.L. Kibby, T.P. Kobylinski, R.B. Pannell, U.S.
Pat. 4 413 064 (1983); 4 493 905 (1985); 4 585 798 (1986),
assigned to Gulf.
[29] C.H. Mauldin, K.L. Riley, Eur. Pat. Appls. 453 674 (1991),
References assigned to Exxon.
[30] S. Eri, J.G. Goodwin, G. Marcelin, T. Riis, U.S. Pat. 4 801
[1] E Biloen, J.N. Helle, WM.H. Sachtler, J. Catal. 58 (1979) 573, (1989).
58. [31 ] S. Vada, A. Hoff, E. Adnanes, D. Schanke, A. Holmen. Top.
[2] R.C. Brady, R. Pettit, J. Am. Chem. Soc. 103 (1981) 1287. Catal. 2 (1995) 155.
[3] E. Iglesia, S.C. Reyes, R.J. Madon, S.L. Soled, in: D.D. Eley, [32] E.EG. Herrington, Chem. Ind. (N.Y.) 65 (1946) 346.
H. Pines, EB. Weisz (Eds.), Adv. Catal. Rel. Subj., Academic [33] H, Schulz, K. Beck, E. Erich, Stud. Surf. Sci. Catal. 36 (1988)
Press, New York, vol. 39, 1993, p. 239. 457.
[4] E. Iglesia, S.C. Reyes, R.J. Madon, J. Catal. 129 (1991) [34] L.M. Tau, H.A. Dabbagh, B.H. Davis, Catal. Lett. 7 (1990)
238. 127.
[5] R.J. Marion, S.C. Reyes, E. Iglesia, Am. Chem. Soc. Symp. [35] E.W. Kuipers, I.H. Vinkenburg, H.J. Oosterbeek, J. Catal. 152
Ser. 517 (19921 383. (1995) 137.
78 E. Iglesia/Applied Catalysis A: General 161 (1997) 59-78

[36] R.C. Reid, J.M. Prausnitz, T.K. Sherwood, The Properties of [39] W.A. van Erp, J.M. Nanne, M.E Post, U.S. Pat. 4 637 993
Gases and Liquids, McGraw-Hill, New York, 1977. (1987), assigned to Shell.
[37] R.S. Albal, Y.T. Shah, N.L. Carr, A.T. Bell, Chem. Eng. Sci. [40] C.H. Mauldin, K.L. Riley, U.S. Pat. 4 977 126 (1990),
39 (1984) 905; M.A. Mathews, A. Akgerman, AICHE J. 33 assigned to Exxon.
(1987) 881. [41] E. Herbolzheimer, E. Iglesia, U.S. Pat. 5 348 982 (1994); Eur.
[38] S.C. Reyes, E. Iglesia, unpublished results. Pat. Appls. 302 710 (1991),