You are on page 1of 91

Natural gas conversion

The Reforming and Fischer-Tropsch processes


by
Sarah Lgdberg
Hugo A. Jakobsen
TKP 4145 Reactor technology
Compendium
i
Prelude
This compendium consists of two parts. The first part deals with natural gas reforming in
general, whereas the second part treats the modelling of trickle bed reactors and the ischer!
Tropsch s"nthesis.
ii
Table of contents
Prelude.................................................................................................ii
Reforming of natural gas....................................................................1
#. $ntroduction.........................................................................................................................%
%. &ifferent natural gas reforming concepts...........................................................................'
%.# (eactions and thermod"namics)...................................................................................*
%.% Steam reforming +tubular reforming,............................................................................-
%.' Adiabatic pre!reforming..............................................................................................#'
%.. /artial o0idation..........................................................................................................#.
%.* Autothermal reforming and secondar" reforming......................................................#*
%.1 Heat e0change reforming............................................................................................#-
'. &ifferent reforming concepts for different applications...................................................%#
'.# $ntroduction.................................................................................................................%#
'.% 2ethanol s"nthesis......................................................................................................%'
'.' Ammonia s"nthesis.....................................................................................................%.
'.. ischer!Tropsch s"nthesis...........................................................................................%.
.. 3ew reforming concepts under de4elopment...................................................................%1
*. Some notes on modeling of C/56 and AT(...................................................................%-
Modeling of tricklebed reactors......................................................!1
#. i0ed!bed reactors 4s trickle!bed reactors........................................................................'%
%. &ispersion model e7uations for multiphase s"stems........................................................'.
'. Special issues in modeling of trickle!bed reactors............................................................'-
'.# $ntroduction.................................................................................................................'-
'.% Catal"st wetting and relating kapp to k.......................................................................'8
'.' Li7uid hold!up and mass transfer.................................................................................#
'.. Heat transfer..................................................................................................................
'.* 9ffecti4eness factor......................................................................................................*
'.1 :ackmi0ing..................................................................................................................1
'.1.# 5ne!parameter models..............................................................................................1
'.1.% Two!parameter model...............................................................................................8
iii
'.; Conclusion..................................................................................................................*#
The ischer!Tropsch process................................................................................................*%
..# $ntroduction.................................................................................................................*%
..% &ifferent reactor t"pes and reaction conditions..........................................................*.
..%.# HTT........................................................................................................................**
..%.% LTT .......................................................................................................................*1
..' Scope of the work.......................................................................................................*-
... 2odeling of the trickle!bed T!reactor......................................................................*8
....# <ang et al. =%>>'?....................................................................................................1#
....% Jess et al. =#888? ......................................................................................................;>
..* 2odeling of the slurr" phase reactor..........................................................................;'
..*.# &ifferent flow regimes and estimation of gas hold!up............................................;'
..*.% 2ass transfer............................................................................................................;1
..*.'. 2i0ing.....................................................................................................................;;
..*.. Heat transfer.............................................................................................................;-
..*.* The model................................................................................................................;-
..*.1 The results................................................................................................................-#
*. Conclusion@discussion.......................................................................................................-'
References.........................................................................................."5
i4
Reforming of natural gas
The different conce#ts
#
1. Introduction
The largest natural gas reser4es are located in the former ASS(, $ran and Batar. There are
also large reser4es in the 3orth Sea. About .> C of the resources are in the former ASS(.
=www.gasforeningen.se? $n #88', 3orwa" produced #.# C of the world natural gas production
=Dirk!5thmer, #88*?. 3atural gas stands for appro0imatel" %# C of the worldEs energ"
consumption =www.iea.org?.

90cept for being used as fuel, natural gas is the most common feedstock for h"drogen
production or s"ngas production for s"nthesis of base chemicals +such as methanol and
ammonia,, oil refining, and in man" other industrial applications +iron ore reduction,
h"drogenation of fats, production of o0o alcohols,. 5f the total h"drogen production b"
catal"tic reforming, *> C is used for ammonia production, #> C for methanol production and
'* C for h"drotreatment in petrochemical processes +h"drocracking, h"drodesulfuriFation
etc., =AllmanEs, #8-*?.
S"ngas ma" howe4er be produced from a range of feedstocks +naphtha, coal, biomass etc.,.
or hea4" h"drocarbons as feedstock, partial o0idation with steam and o0"gen is used for
s"ngas production. Also for solid feedstocks, such as coal and biomass, partial o0idation with
steam and o0"gen is used for the s"ngas production, but in this case it is often referred to as
gasification. <hen producing s"ngas from natural gas or light h"drocarbons, steam reforming
is often the preferred route if a high H
%
@C5 ratio is desired. Steam reforming can be
performed in the presence of o0"gen or C5
%
. A combination of steam reforming and partial
o0idation is called autothermal reforming, in which the endothermic and e0othermic reactions
are coupled =2ouliGn et al., %>>'?.
or light h"drocarbons, partial o0idation is usuall" not economical since it re7uires an
e0pensi4e cr"ogenic air separation unit. Howe4er, for special applications, a high C5@H
%
ratio
in the s"ngas is desired, which can be accomplished b" partial o0idation. /artial o0idation
ma" be carried out with or without a catal"st =2ouliGn et al., %>>'?.
$t is desirable to operate at higher pressures, if the s"ngas is to be used for production of
chemicals such as methanol, ammonia or ischer!Tropsch fuels. This is because these fuel
s"nthesis operate at ele4ated pressures, and hence a pressuriFed s"ngas remo4es the cost of
compressing. urthermore, plants operating at ele4ated pressures can ha4e higher capacities
due to their smaller gas 4olumes, which in turn means lower in4estment costs =AllmanEs,
#8-*?.
The gasification process ma" be conducted with or without a catal"st. The use of a catal"st
makes it possible to come closer to the e7uilibrium con4ersion, but catal"sts can onl" be used
for gaseous feeds or for distillate li7uid feeds.
$n this report, the state!of!the!art of reforming of natural gas into s"nthesis gas will be
described. &ifferent commercial concepts will be discussed and compared, as well as
concepts being de4eloped at present, not "et commercialised.
%
2. Different natural gas reforming concepts
Table # shows the appro0imate molar H
%
@C5 ratios that are achie4ed upon different natural
gas reforming concepts. <ilhelm et al. =%>>#? summariFed the ad4antages and disad4antages
with the maGor natural gas s"ngas production concepts in Table %.
Table #. <a"s to increase@reduce H%@C5 ratios, and appro0imate H%@C5 ratios from different concepts =<ilhelm
et al., %>>#?.
'
Table %. Ad4antages and disad4antages of se4eral natural gas reforming concepts. S2( H steam methane
reforming, AT( H autothernal reforming, /56 H partial o0idation.
a
S2( followed b" o0"gen!blown secondar"
reforming =<ilhelm et al., %>>#?.
.
2.1 Reactions and thermodynamics:
The standard enthalpies of reaction +at %8- D, are gi4en in brackets. The most important
reactions in steam reforming (SR) of methane are =2ouliGn et al., %>>'?)
#. CH
.
I H
%
5 C5 I 'H
%
+I%>1 kJ@mol,
%. C5 I H
%
5 C5
%
I H
%
+water!gas!shift, <JS, !.# kJ@mol,
'. CH
.
I C5
%
%C5 I %H
%
+dr" reforming, I%.; kJ@mol,
The last one is called Kdr" reformingL because it can be used to produce a C5!rich s"ngas
without steam as reactant.
These main reactions ma" be accompanied b" coke formation according to)
.. CH
.
C I %H
%
+decomposition of methane, I;* kJ@mol,
*. %C5 C I C5
%
+the :oudouard reaction, !#;' kJ@mol,
(eactions +., and +*, are re4ersible. At temperatures abo4e 1*> MC, also higher h"drocarbons
ma" be thermall" cracked into coke, and this reaction is irre4ersible. =&"bkGaer, #88*? The
carbon formation will be discussed more deepl" in 2.2 Steam reforming.
$n the presence of o0"gen, the following reactions ma" occur =2ouliGn et al., %>>'?)
1. CH
.
I N5
%
C5 I %H
%
+!'1 kJ@mol,
;. CH
.
I %5
%
C5
%
I %H
%
5 +!->' kJ@mol,
-. C5 I N5
%
C5
%
+!%-. kJ@mol,
8. H
%
I N5
%
H
%
5 +!%.% kJ@mol,
Since the reaction with methane and steam +#, is highl" endothermic, and the reactions
in4ol4ing 5
%
+1 O 8, are highl" e0othermic, the production of s"ngas from methane can me
allothermal or autothermal. Allothermal means that the re7uired heat for the reaction is
produced outside of the reactor. This is the case for pure steam reforming, or when onl" little
amount of 5
%
is present. Autothermal means that all heat re7uired for the steam reforming
reaction is formed inside of the reactor b" means of e0othermic reactions with 5
%
, and hence
no e0ternal heating is needed. $t is the steam@o0"gen ratio that determines the mode of
operation =2ouliGn et al., %>>'?. <hen the reaction is e0othermal it takes place in adiabatic
reactors.
&"bkGaer =#88*? remarked that also higher h"crocarbons are present in natural gas, and that
the" will react according to the following endothermal reaction)
C
n
H
m
I nH
%
5 nC5 I +nIm@%,H
%
+#>,
This reaction is irre4ersible and proceeds to full con4ersion. The o4erall heat of reaction of
+#,, +%, and +#>, ma" be positi4e, Fero or negati4e, depending on the process conditions. At
low steam@carbon +S@C, ratios, and at low catal"st e0it temperatures, the o4erall reaction is
*
onl" slightl" endothermal or e4en e0othermal, if the feed contains high concentrations of
higher h"drocarbons. This is due to that the C5 formed in reaction +#>, is con4erted into CH
.
according to the re4erse of reaction +#,. $n this case, the process ma" be carried out without
e0ternal heating, e.g. in an adiabatic pre-reformer. Howe4er, if a s"ngas with a low methane
content is desired, a high e0it temperature is re7uired and the o4erall heat of reaction will be
endothermal, and e0ternal heat is necessar" =&"bkGaer, #88*?.
As shown in igure #, at high temperatures +P ->> MC, the e7uilibrium H
%
@C5!ratio of the
s"ngas reaches a 4alue of '@# in case of pure steam reforming +#,, and %@# in case of partial
oxidation (POX) +1,, respecti4el". These 4alues are 4alid if the feed has the stoichiometric
CH
.
@H
%
5 and CH
.
@5
%
ratios for the steam reforming and partial o0idation reactions,
respecti4el". The data in igure # are obtained if respect is taken to all reactions # O ' for the
left picture +a,, and # O ' and 1 O 8 for right picture +b,. Their similar appearance is due to the
fact that the same reactions actuall" occur in both cases, e0cept for the reactions in which 5
%
are in4ol4ed which onl" occur in the partial o0idation case =2ouliGn et al., %>>'?.
igure #. 9ffect of temperature on e7uilibrium composition at # bar in steam reforming of methane +a,, and
partial o0idation of methane +b,. H%5@CH. H #@# in +a,, 5%@CH. H >.*@# in +b, =2ouliGn et al., %>>'?.
The peak in C5
%
concentration is e0plained b" that formation of C5
%
is fa4oured at low
temperatures, since C5
%
is onl" formed in e0othermal reactions. At higher temperatures, the
C5
%
is consumed in the endothermal reforming reactions, since these reactions are fa4oured
b" higher temperatures =2ouliGn et al., %>>'?.
igure % shows the pressure dependence of the e7uilibrium composition in case of pure steam
reforming with stoichiometric feed. A higher pressure shifts the steam reforming e7uilibrium
towards the reactants since the" compose fewer molecules. At '> bar, a temperature of #.>>
D is needed to reach an e7uilibrium in which onl" the products C5 and H
%
e0ist. Howe4er,
the ma0imum temperatures used in industr" for steam reforming are around #%>> D due to
reactor material constraints =2ouliGn et al., %>>'?.
1
igure %. 9ffect of temperature and pressure on e7uilibrium composition in steam reforming of methane.
H%5@CH. H #@#. +a, H% and CH., +b, C5, H%5 and C5% =2ouliGn et al., %>>'?.
Since most s"ngas applications, such as methanol and ammonia s"nthesis, re7uire high
pressures, the s"ngas production is often performed at high pressure in order to a4oid an
e0pensi4e compression and to make the reformer siFe small. A high pressure is not fa4ourable
for e7uilibrium of the s"ngas production, and hence a high temperature and steam in e0cess is
used as compensation, otherwise the methane con4ersion is too low. igures ' shows the
effect of pressure and the effect of steam@methane ratio on uncon4erted methane +the methane
slip,, respecti4el". The most economical operation is as high pressure and temperature as
possible, the tube material setting the limits. =2ouliGn et al., %>>'? $n other cases, the pressure
is set b" the re7uirements of downstream separation or purification processes, such as /SA
units, membranes or cr"ogenic units. =&"bkGaer, #88*?.
igure '. 9ffect of pressure +left, and steam@methane ratio +right, on uncon4erted methane as a function of
temperature =2ouliGn et al., %>>'?
;
2.2 Steam reforming (tubular reforming)
The o4erall steam reforming (SR) is highl" endothermic and it is carried out at high
temperature +8>> MC, and at pressures between #* and '> bar =2ouliGn et al., %>>'? o4er a
3i@Al
%
5
'
catal"st =:haradwaG and Schmidt, #88*?. The composition of the gas at the reactor
outlet reflects the e7uilibria of reaction +#, and +%, abo4e. $n some cases, it is ad4antageous to
add C5
%
at the inlet of the reformer. This is done in order to sa4e h"drocarbon feedstock and
decrease the H
%
@C5 ratio in the product gas. The C5
%
coming out from the reformer is then
rec"cled, and in some cases C5
%
is also imported. C5
%
then is supposed to react to form C5
4ia the re4erse <JS!reaction, and this is fa4ored at low S@C ratios. Howe4er, low S@C ratios
leads to high methane concentrations in the outlet. To compensate for this, a higher
temperature can be used =&"bkGaer, #88*?.
The 3i!catal"st is needed since methane is a 4er" thermod"namicall" stable molecule e4en at
high temperatures. The steam reformer consists of two sections O a con4ection section and a
radiant section +see igure .,. $n the con4ection section, methane and steam are preheated to
*>> MC =AllmanEs, #8-*? b" heat e0change with the hot flue gases from the fuel combusted in
the reformer furnace, and in the radiant section the reforming reactions take place in the
catal"st filled tubes which are hanging in rows inside of the reformer furnace =2ouliGn et al.,
%>>'?. The process gas is heated graduall" to appro0imatel" ->> C inside of the tubes
=AllmanEs, #8-*?. All tubular reformers use catal"st inside the tubes in order to reduce the
operating temperature. This is important in order to reduce the tube stresses resulting from
high pressure and high temperatures =roment and :ischoff, #88>?. H/!steam is also
produced in the con4ection Fone, which ma" be con4erted to electricit" which is used for
compression of the produced s"ngas in case of methanol or ammonia s"nthesis. Steam for the
reforming is produced from the heat in the product gas out from the reformer. 5ne furnace
can contain *>> O 1>> tubes with inner diameters of ;> O#'> mm and lengths of ; O #% m
=2ouliGn et al., %>>'?. The wall thickness of the tubes is between #> O %> mm =AllmanEs,
#8-*?. The tubes ha4e small diameters in order to achie4e the highest possible heat flu0 to the
catal"st, and hence to achie4e the highest possible capacit" for a gi4en amount of catal"st
=roment and :ischoff, #88>?.
The heat needed in the tubular reformer is the enthalp" difference between inlet and e0it gas.
The heat dut" consists of the heat of the reaction and the heat needed to bring the products to
the e0it temperature. Appro0imatel" *> C of the heat produced b" combustion in the burners
is transferred to the process gas. The other *> C e0its the s"stem in the hot flue gases from
the burners, and this heat is reco4ered in the con4ection section as described abo4e. The
o4erall thermal efficienc" of the reformer ma" approach 8* C =&"bkGaer, #88*?. $n the case
where there is no need for surplus energ", and hence production of H/!steam is not desirable,
one solution is to install smaller tubes inside the reformer tube. The catal"st is placed in the
space between the main and the smaller tubes. After being preheated, the steam and methane
enter the catal"st!filled space, where it is heated and reforming takes place as usual. At the
end of the reformer tube, the gas then enters the smaller tubes and transmits some heat to the
catal"st bed before being discharged at the top. This reduces the amount of heat transmitted
b" the reformer tube and hence the number of tubes and their surface area b" appro0imatel"
%> C =AllmanEs, #8-*?.
-
igure .. Schematic picture of a steam reformer, showing the radiation and con4ection sections =2ouliGn et al.,
%>>'?.
2ethane reforming can be described b" a first!order reaction, irrespecti4e of pressure. At high
temperatures the o4erall rate can be limited b" pore diffusion, but at low temperatures the
molecular diffusion rate is much higher than the reaction rate so that the catal"st acti4it" can
be full" used =AllmanEs, #8-*?. According to &"bkGaer =#88*?, the o4erall rate in steam
reforming is limited b" the heat transfer, at high temperatures. The 3i!catal"st is often in the
form of thick!walled (aschig rings, with #1 mm in diameter and height, and a 1 O - mm hole
in the middle. $f the heat load per unit area is too high, the limits of such catal"sts will be
reached, and hence smaller particles will be necessar" in order to make use of more of the
catal"st. Smaller particles will howe4er lead to increased pressure drop. Therefore, special
packing shapes such as spoked wheels or rings with se4eral holes ha4e been de4eloped for
these smaller particles =AllmanEs, #8-*?.
The 3i!catal"st is poisoned b" sulfur, which is present in practicall" all gaseous feedstocks,
wh" desulfuriFation is a necessar" step prior to reforming +see igure .,. The desulfuriFation
is often made with Finc!o0ide absorption beds working at '*> O .>> MC. Qinc o0ide absorbs
se4eral different sulfur species such as h"drogen sulfides +H
%
S,, carbon"l sulfide and
mercaptans. Howe4er, c"clic organic sulfur compounds such as thiophenes normall" re7uire
h"drogenation o4er Co!2o or 3i!2o catal"sts to H
%
S. The H
%
S is then adsorbed o4er the
Finc!o0ide bed. The h"drogenation and Finc!o0ide catal"sts can be combined into one 4essel.
(esidual sulfur content is R >.% mg@m
'
=AllmanEs, #8-*?.
There are four t"pes of burner configurations used in tubular reforming as shown in igure *.
Top!fired reformers can ha4e se4eral parallel rows of tubes, while wall! or side!fired
reformers onl" can ha4e one row. The homogeneit" of the heat transfer to the tubes is
determined b" burner geometr", flame length and diameter, tube!to!tube and row!to!row
spacing, fired tube length and distance from the flame to the reformer wall. The temperature
of the flame is appro0imatel" #->> C at the hottest place, and ##>> C at the coldest place.
The heat transfer in the wall!fired reformers is mainl" b" the radiant side!wall, while in top!
fired reformers, the heat is transferred through radiation from the flame and hot flue gases
=AllmanEs, #8-*?.
8
igure *. T"pical configurations of reformer furnaces =&"bkGaer, #88*?.
T"pical reformer wall temperature profiles in the top! and side!fired reformers are shown in
igure 1 +left,. The top!fired reformer is characteriFed b" a temperature peak in the top, and it
has the highest heat flu0 where the metal temperature is at its ma0imum. The side!fired
reformer allows a better temperature control, and the ma0imum temperature is at the outlet of
the tube. The highest heat flu0 is at a rather low temperature. The side!fired reformer has a
higher a4erage heat flu0 than the top fired. 2oreo4er, the short residence time in the flames in
the side fired reformer ensures 4er" low emissions of 35
0
in the flue gases =&"bkGaer, #88*?.
A central problem in steam reforming is to balance the heat input through the tube with the
heat consumption b" the endothermic reforming reaction, while at the same time limiting the
stress on the tubes b" minimising the ma0imum tube wall temperature and the ma0imum tube
wall temperature difference. The side!fired reformer is ideall" suited to sol4e this comple0
interrelation =www.topsoe.com?. Jodfre" and Shackcloth =#8;>? and Stehlik =#88#? ha4e
made some modelling of the radiation heat transfer from burner to steam reformer tube.

igure 1. Left) Tube wall temperature and heat flu0 profiles. Top! and side!fired reformers. Start of run. (ight)
Tube wall temperature profiles at start and end of run. Top! and side!fired reformers =&"bkGaer, #88*?
#>
The catal"st will deacti4ate and lose acti4it" upon use. $t is howe4er actuall" possible to
retain the producti4it" essentiall" constant b" a slight increase in temperature in the lower end
of the tube, pro4ided the re7uired heat is a4ailable. A decrease in acti4it" will howe4er result
in a large temperature increase in the top of the tube, especiall" in case of a top!fired reformer
+see igure 1 +right,,. This is since the reaction consumes less heat when the catal"st is
deacti4ated, and hence the dri4ing force for heat transfer to the process gas decreases. &ue to
these effects, the top fired reformers must be designed with a considerable margin abo4e the
ma0imum temperature at the start of the run. There is also risk of carbon formation at the hot
wall =&"bkGaer, #88*?.
$n the side!fired reformer, a decrease in catal"st acti4it" will lead to an increase in
temperature in the upper part, as well, but the temperature will still be highest in the lower
end. Therefore, in this case, the reformer does not ha4e to be designed for much higher
temperatures than at the start of the run =&"bkGaer, #88*?.
Carbon formation
As alread" mentioned abo4e, a critical parameter in steam reforming is the molar S@C molar
ratio. Carbon deposits will occur that deacti4ates the catal"st b" cooking, if this ratio is too
low. Large carbon deposits ma" also block the tubes and hence cause hot!spots, which might
destro" the tubes. A higher S@C ratio reduces the carbon formation, and a common
steam@carbon ratio lies between %.* and ..*, the higher 4alue being used for steam reforming
of naphtha. A higher S@C ratio than stoichiometric, also helps to shift the steam reforming +#,
e7uilibrium towards the products, and hence to increase the methane con4ersion +reduce the
methane slip, =2ouliGn et al., %>>'?. The carbon is formed according to =AllmanEs, #8-*?)
%C5 C I C5
%
+The :oudouard reaction,
CH
.
C I %H
%
+&ecomposition of methane,
C5 I H
%
C I H
%
5 +Heterogeneous water gas reaction,
The e7uilibrium for the :oudouard reaction and the decomposition of methane are shown in
igure ; and -. Howe4er, it has been shown that for some catal"sts the thermod"namicall"
predicted carbon formation can be suppressed at temperatures below ;>> MC. See for instance
line +f, in igure ;, which shows the regions where carbon was formed and not for a nickel!
uranium catal"st, and compare with +d, or +e, that show the true thermod"namic constant for
the :oudouard reaction. Theoreticall", carbon could form in the inlet of the steam reformer.
Howe4er, this does not occur below 1*> MC, but onl" at higher temperatures when the catal"st
has a low acti4it" so that it fails to con4ert sufficient methane. The addition of C5
%
to the feed
gas minimiFes the risk of carbon formation as shown in igure ; and - =AllmanEs, #8-*?.
##
igure ;. (eaction 7uote for :oudouard reaction +*,) %C5 C I C5% as a function of temperature. +a,
condition of rich gas from naphtha at inlet to tubular reformer, +b, e7uilibrium reached for <JS +%,, +c,
e7uilibrium reached for S( of methane +#, and <JS +%,, +d, and +e, show the e7uilibrium constant for the
:oudouard reaction according to two different references in AllmanEs =#8-*?, +f, boarder between carbon
formation and no carbon formation for a nickel!uranium catal"st, +g, boarder between carbon formation and no
carbon formation for a sulfided catal"st. =AllmanEs, #8-*?
igure -. (eaction 7uote for methane decomposition +.,) CH. C I %H% as a function of temperature. +a,
e7uilibrium reached for S( of methane +#, and <JS +%,, +b, e7uilibrium reached for <JS +%,, +c, e7uilibrium
constant for methane decomposition +.,, +d, and +e, show working lines for methane reforming in a reactor tube
with a high acti4it" catal"st and a low acti4it" catal"st, respecti4el" +figures at the lines are fractions of tube
length from reactor inlets, +f, e7uilibrium constant determined b" e0periments from a reference in AllmanEs
=#8-*?, +g, connecting line between e7uilibrium constants determined b" e0periments at .#> and *%* MC from
two references in AllmanEs =#8-*?.
#%
As described abo4e, traditionall" the S@C ratios ha4e been rather high in order to a4oid carbon
formation. Howe4er, new highl" acti4e steam reforming catal"sts and the use of adiabatic pre!
reformers make it possible to use S@C ratios below #.>. The lower S@C ratio is ad4antageous if
a C5!rich gas is wanted, and to make the o4erall reaction less endothermic =&"bkGaer, #88*?.
$n a pre!reformer +see 2.3 Adiabatic pre-reforming,, the worst carbon precursors +i.e. higher
h"drocarbons, are remo4ed up!stream the steam reformer. This is the reason for wh" the main
reformer is able to operate at a lower S@C ratio =2ouliGn et al., %>>'?.
$f C5
%
is added to the reformer feed in order to make the s"ngas richer in C5, as discussed
abo4e, less steam is actuall" needed for the suppression of carbon formation since C5
%
is an
o0idiFing agent =2ouliGn et al., %>>'?.
2. !diabatic pre-reforming
Adiabatic pre!reforming is used for reforming of natural gas to hea4" naphtha. The process is
carried out in a fi0ed!bed upstream the tubular reformer +see igure 8,. Higher h"drocarbons
are completel" con4erted into C5, H
%
and CH
.
. The reactions taking place in the pre!
reformer, e0cept for <JS, are =&"bkGaer, #88*?)
C
n
H
m
I nH
%
5 nC5 I +nIm@%,H
%
+#>,
'H
%
I C5 CH
.
I H
%
5 +re4ersed #,
$n case of natural gas as feedstock, the o4erall process is endothermic, and therefore will
result in a temperature drop since the reactor is adiabatic. or higher h"drocarbon feedstocks,
the pre!reforming is e0othermic or thermoneutral. $n the pre!reformer, the temperature is
relati4el" low. This makes the chemisorption of sulfur to the 3i!catal"st fa4orable. Therefore,
traces of sulfur from the desulfuriFation unit will be trapped in the pre!reformer =&"bkGaer,
#88*?.
igure 8. $nstallation of a pre!reformer =&"bkGaer, #88*?.
Ad4antages of installing a pre!reformer =&"bkGaer, #88*?)
#'
! All higher h"drocarbons are completel" con4erted into C5, H
%
and CH
.
. 94en for
naphtha, the natural gas steam reforming catal"st ma" be used in the prereformer.
! All traces of sulfur are remo4ed, which increases the life!time of the tubular steam
reforming catal"st.
! Since no sulfur is poisoning the top la"er of the catal"st in the steam reformer, risks of
hot spots are reduced.
! The production capacit" of the plant ma" be increased. :" installing a new pre!heat
coil or heater between the pre!reformer and the steam reformer, the load on the
reformer is reduced. This ma" be used as a capacit" increase or as a decrease in firing
with unchanged capacit".
! The sensiti4it" of the steam reformer to 4ariations in S@C ratios and feedstock
composition is essentiall" eliminated.
2." #artial o$idation
Partial oxidation (POX) is often used for gasification of hea4" oil, but all h"drocarbons are
possible as feedstocks. Since no water is added, the H
%
@C5 ratio is lower than compared with
steam reforming or autothermal reforming. /artial o0idation of natural gas is used in small
plants and in regions where natural gas is cheap. (eaction temperatures are #'*> O #1>> C
and pressures up to #*> bar. The concentrations of different compounds in the product
mi0ture are determined b" se4eral e7uilibria, which are 7uickl" tuned in at these high
temperatures. /artial o0idation can be performed with or without a catal"st. $f a catal"st is
used, the reaction temperature can be lower, the reactions still reaching e7uilibrium, since the
catal"st lowers the acti4ation energies. A lower temperature, at this high temperatures, would
not significantl" change the e7uilibrium composition =AllmanEs, #8-*?.
:haradwaG and Schmidt =#88*? claimed that the de4elopment o4er the "ears has been to
minimiFe the use of steam in reforming due to the following disad4antages)
! 9ndothermic reactions
! The product gas has a H
%
@C5 ratio of '
! Steam corrosion problems
! Costs in handling e0cess H
%
5
The trend is to mo4e from steam reforming to KwetL o0idation +autothermal reforming, to
Kdr"L o0idation. The dr" o0idation is the partial o0idation of methane +1,. This reaction
directl" gi4es the desired ratio H
%
@C5 H % for ischer!Tropsch or methanol s"nthesis, at
temperatures P 8>> MC, as shown in igure # +b, abo4e. Howe4er, the selecti4ites are also
affected b" the H
%
5 and C5
%
formed in the complete o0idation reactions +;, #>, =:haradwaG
and Schmidt, #88*?.
Since the partial o0idation reaction is slightl" e0othermic, the partial o0idation reactor would
be much more energ" efficient than the energ" intensi4e steam reformer. Since the reaction is
fast, the reactor siFe could be reduced significantl" compared to the S( reactor =:haradwaG
and Schmidt, #88*?.
The non!catal"tic partial o0idation +/56, of methane is alread" commercialiFed, for instance
in the ischer!Tropsch plant in 2ala"sia, :intulu. Howe4er, there are se4eral problems with
/56 that can be a4oided b" cataltic partial oxidation (CPOX). $n C/56, onl"
heterogeneous reactions are taking place, lower temperatures are used and no soot or
unwanted b"!products are formed. 3o burner is used. $t is mainl" the absence of
#.
homogeneous reactions that pre4ent the formation of unwanted o0idation products or flames
which can lead to soot formation. The newl" de4eloped highl" selecti4e partial o0idation
catal"sts can o4ercome the carbon problem without an" steam input. $t appears that high
selecti4it" to C5 and H
%
, as compared to C5
%
and H
%
5, can be achie4ed at short contact
times, i.e. at high JHSSs. (esults showing high methane con4ersion at short contact times
ha4e also been reported. :haradwaG and Schmidt =#88*? concluded that for partial o0idation
of methane, (h seems to be the catal"st of choice to achie4e selecti4ities to H
%
and C5 abo4e
8* C, and methane con4ersions abo4e 8> C. The maGor engineering challenge is to ensure
safe operation with the premi0ed CH
.
@5
%
mi0tures. The reaction mechanism seemed to be
direct o0idation 4ia CH
.
p"rol"sis at contact times shorter than >.# s. At longer residence
times, reforming reactions with CH
.
and H
%
5 or C5
%
and shift reactions also ma" take place.
Howe4er, the direct catal"tic partial o0idation of methane has not "et been commercialiFed
because it is difficult to stud" since it in4ol4es premi0ing of CH
.
@5
%
mi0tures which can be
flammable or e0plosi4e =:haradwaG and Schmidt, #88*?.
&ue to the high temperatures needed in order to reach high con4ersions and high selecti4ities
to H
%
and C5, the C/56!reactor needs e0tremel" tolerant materials, which are e0pensi4e.
Current research at 3T3A@S$3T9 aims at de4eloping catal"sts that can gi4e satisfactor"
high reaction rates and high selecti4ities to H
%
and C5 at lower temperatures +1*> MC,. This
would be possible if the catal"st could ensure that onl" the partial o0idation, i.e. reaction +1,,
occurs, and hence that no water is formed. (eaction +1, is e0othermic and hence fa4ored at
low temperatures, and if no water is e4er present in the reactor, a s"ngas with a H
%
@C5 ratio of
%.> could theoreticall" be formed at low temperatures in relati4el" cheap steel reactors
=:GTrgum, %>>1?.

According to &"bkGaer =#88*? and LTdeng =%>>1?, this far, no catal"st has shown selecti4it"
for C/56 of methane to C5 and H
%
at industrial conditions +i.e. at pressures abo4e %> bar,
"ielding less C5
%
than predicted b" e7uilibrium of methane steam reforming and <JS. $n
order to be economicall" attracti4e, catal"tic /56 of methane must be carried out at ele4ated
pressure and at high methane con4ersion. The composition of the product gas will alwa"s be
determined b" the e7uilibrium at the reactor e0it, irrespecti4e of whether a fi0ed or fluidiFed
bed is used =&"bkGaer, #88*?.
2.% !utothermal reforming and secondary reforming
$n a!tot"ermal reforming (A#R), partial o0idation of the fuel is conducted in order to produce
the heat re7uired for the endothermic reforming reactions of the same fuel. 90ternal steam is
added. Characteristic of the autothermal process is that the o0"gen added to generate heat is
chemicall" bound in the product gas, which results in that the H
%
@C5 ratio in the product gas
is lower than in other processes. :oth temperature and pressure is in between what is used in
steam reforming and partial o0idation. T"pical operating conditions are -*> O #>>> C and %>
O #>> bar =2ouliGn et al., %>>'?. 5ne ad4antage with autothermal reforming is that the
thermal efficienc" +i.e. ratio of heat content of reformed gas to that of the h"drocarbon feed, at
> C, is higher +--.* C, than that of steam reforming +-# C, including fuel, and than that of
partial o0idation +-'.* C,. The ma0imum temperature is not limited b" the tube material but
b" the stabilit" of the catal"st and the refractor" lining of the reactor. Autothermal reforming
is more fle0ible than tubular reforming, since the higher operating temperature can
compensate for the increase in methane slip which higher pressure would cause otherwise
=AllmanEs, #8-*?.
#*
There are two t"pes of autothermal reformers O pure catal"tic, and with a pre!combustion
unit. $n the pure catal"tic reformer, the reactants +methane, o0"gen and steam, enter the
catal"st bed directl" after mi0ing. This performance, without an" residence time in the empt"
space abo4e the catal"st bed, results in no carbon formation, e4en at low preheat temperatures
of the feedstock +methane,. Steam reforming reactions take place more 7uickl" than the
:oudouard reaction e7uilibrium is tuned in. This makes it possible to use less steam, and
hence less o0"gen, which results in a higher C5@C5
%
ratio in the product gas. The drawbacks
with this purel" catal"tic process are high thermal and mechanical loads on the catal"st in the
immediate 4icinit" of the burner. Temperature 4ariations during start!up and shut!down, and
high gas 4elocities lead to attrition and catal"st disintegration, which in turn makes it
necessar" to replace the catal"st e4er" two "ears =AllmanEs, #8-*?.
(eforming with pre!combustion in the empt" space abo4e the catal"st bed is preferred for
gases that ha4e 4er" low risk of carbon deposition, i.e. gases with high H
%
content and low
h"drocarbon content. This t"pe of reforming re7uires slightl" more o0"gen than the pure
catal"tic t"pe, and the gas 4elocities must be lower =AllmanEs, #8-*?. or natural gas
autothermal reforming, the concept with the pre!combustion Fone is preferred, and will be
described further.
AT(, as a concept, is a stand!alone process in which the entire natural gas con4ersion is
carried out b" means of internal combustion with o0"gen. Secondar reforming is a process in
which partiall" con4erted process gas from a tubular steam reformer is further con4erted b"
means of internal combustion. The secondar" reformer in an ammonia plant will be air!
blown, while that of a methanol plant will be o0"gen blown. =&"bkGaer, #88*?
The autothermal reforming is usuall" not used on its own, due to high in4estment costs for
separation of the o0"gen from air. $t is rather used downstream a steam reformer, i.e. as a
secondar" reformer, in order to reform the unreacted methane from the steam reformer. This
makes it possible to increase the pressure, without increasing the temperature, in the steam
reformer +t"pical conditions are #>;> D and '> bar,, which is not fa4ourable for the
e7uilibrium but economicall" fa4ourable if the s"ngas will be used for a high!pressure
chemical s"nthesis. $n case of ammonia production +P #>> bar,, 3
%
is needed in the s"nthesis
and hence air can be used instead of o0"gen in the autothermal reformer, which reduces the
cost for this reformer considerabl". The two!step reforming concept with a S( combined with
a secondar" reformer +AT(, is suitable also for methanol s"ngas production, since the
methanol takes place at high pressures +*> O #>> bar, =2ouliGn et al., %>>'?.
Since the concentration of combustibles are different in the gas going to the AT( and the
secondar" reformer, different designs of burner and reactor are needed, in order to ensure
suitable heat release and a4oid soot formation in each case. The AT( and the secondar"
reformer are howe4er 4er" similar in reactor design. The" consist of a refractor"!lined
pressure 4essel +and therefore stands higher pressures and temperatures than the steam
reformer, with a burner, combustion chamber and catal"tic bed. The reactor space can be
di4ided into three Fones +see igure #>, in which different reactions take place according to
Table '. =&"bkGaer, #88*?
#1
igure #>. Schematic picture of an AT( =2ouliGn et al., %>>'?.
Table '. (eaction Fones in an AT(
Reactor e$ui#ment Reaction %ones
:urner ! +pro4ide mi0ing,
Combustion
chamber
Combustion Fone
Thermal Fone
Catal"st bed Catal"tic Fone
$!rner%
The burner pro4ides the mi0ing of the feedstreams in a turbulent diffusion flame. The core of
the flame has a high temperature, often abo4e %>>> MC, wh" it is important to minimiFe the
transfer of heat back to the burner from the flame. This can be made b" recirculating the gas
from the thermal Fone back to the burner =&"bkGaer, #88*?. The burner is the ke" element for
the o0"gen!fired reformer. Careful design of the burner noFFles ensures a flow pattern with
efficient mi0ing that protects the refractor" and burner from the hot flame core
=www.topsoe.com?
Comb!stion &one%
This is the turbulent Fone in which the h"drocarbon and o0"gen are mi0ed and combusted.
5ften the principle Kmi0ed!is burntL is 4alid, since the e0othermic combustion reactions are
4er" fast. The o4erall o0"gen to h"drocarbon ratios in the combustion Fone 4ar" between >.**
and >.1, which means that the conditions are substoichiometric with respect to complete
combustion. The combustion reactions are numerous comple0 radical reactions, but for
modeling purposes, it is often enough to describe the reactions b" an o4erall one molecular
reaction. $n case of natural gas =&"bkGaer, #88*?)
CH
.
I '@% 5
%
C5 I %H
%
5 +##,
This reaction is a mi0ture of +1, and +;,. All o0"gen is consumed in the combustion Fone, and
the uncon4erted CH
.
will continue down to the thermal Fone.
#;
$n case of a secondar" reformer, also H
%
will be burnt to water in the combustion Fone
according to reaction +8,.
#"ermal &one%
$n the thermal Fone, the con4ersion of the h"drocarbon proceeds 4ia homogeneous gas phase
reactions. The main reactions are thermal methane reforming +#, and <JS +%,. Also p"rol"sis
of higher h"drocarbons takes place. =&"bkGaer, #88*?
Cataltic &one%
The catal"tic Fone is a fi0ed!bed, in which the h"drocarbons are finall" con4erted through
heterogeneous catal"tic reactions. At the e0it of the catal"tic Fone, the gas mi0ture will be in
e7uilibrium with respect to reactions +#, and +%, at the e0it temperature and pressure, and the
catal"st will destro" an" soot precursors formed in the combustion chamber. The s"ngas is
completel" free of o0"gen. The top of the catal"st bed is e0posed to gases with temperatures
of ##>> O #.>> MC. A 3i!catal"st supported on a magnesia!alumina spinel has the re7uired
acti4it" and high!temperature stabilit". The o4erall reaction rate is mainl" controlled b" the
e0ternal diffusion rate, i.e. the transport rate of the reactants through the gas film surrounding
the catal"st pellets. This means that the process can be carried out at 4er" high space
4elocities, since the catal"tic reaction is 4er" fast. $t is also known that higher space 4elocities
reduce the film thickness surrounding the pellets. The amount of catal"st needed is actuall"
determined b" optimal flow distribution and pressure drop in the reactor =&"bkGaer, #88*?. $n
s"nthesis gas production, a S@C ratio as low as >.1 is industriall" pro4en. Soot!free operation
is achie4ed through optimised burner design and b" catal"tic con4ersion of soot precursors
o4er the catal"st bed =www.topsoe.com?. The AT( or secondar" reformer is operated close to
adiabaticall", and hence the temperature is gi4en b" the adiabatic heat balance =Aasberg!
/etersen et al., %>>'?.
Applications of the concepts AT( and secondar" reforming are discussed in 3. 'ifferent
reforming concepts for different applications.
<hen steam reforming and autothermal reforming are combined, it is logical to use the heat
produced in the latter to warm the tubes in the former one. This configuration is called "eat-
exc"ange reformer, and is described below.
2.& 'eat e$change reforming
The idea with heat e0change reforming is that part of the heat is supplied to the tubes b" heat
e0changing with process gas. Heat e0change reforming eliminates the e0pensi4e fired
reformer. $n this configuration, onl" medium pressure steam can be reco4ered from the
s"ngas plant and electricit" for the s"ngas compressor must be imported. igure ## shows two
different t"pes of heat!e0change reformer. The left is the $C$ combined reforming process and
the right the 900on CA( +combined autothermal reforming, process. $n the $C$ process, ;* C
of the methane is reformed in the steam reformer at 8;> D and .> bar, and the rest is
con4erted in the autothermal reformer. $n the 900on CA( process, steam reforming and
partial o0idation is combined in a single reactor and reaction Fone. The reactor is a fluidised
bed, and o0"gen is introduced in the middle of the bed =2ouliGn et al., %>>'?.
#-
igure ##. $C$ combined reforming process +left, and 900on CA( process +right,
$n the case where two reformers are combined, the heat needed in the tubular S( reformer is
obtained from the hot product gas from the second reformer. The second reformer can be an
air! or o0"gen fired AT( or a partial o0idation unit. This concept is used for production of
h"drogen or s"ngas for the methanol s"nthesis. $n case of s"ngas production for the ammonia
s"nthesis, there is a poor correspondence between the heat re7uired and the heat a4ailable in
the air!fired secondar" reformer, and hence the concept of heat e0change reformer is not
common =&"bkGaer, #88*?.
A problem associated with the heat e0change reforming described abo4e, in which C5!rich
gases contact metals at high temperatures, is the risk of metal dusting corrosion. A certain gas
mi0ture will potentiall" form carbon 4ia the e0othermic :oudouard reaction at temperatures
below which the mi0ture satisfies the :oudouard reaction e7uilibrium. The temperature below
which a certain gas mi0ture ma" form carbon is called the :oudouard temperature. or a gas
mi0ture with a high :oudouard temperature, i.e. a C5!rich gas, the :oudouard reaction ma"
be catal"Fed b" hot metal surfaces. Therefore, it is important that a gas mi0ture with a high
:oudouard temperature does not come in contact with a metal surface of a slightl" lower
temperature. $f carbon deposition on the metal starts, there is a big risk of metal corrosion. $f
carbon is deposited on the catal"st, this will of course lead to deacti4ation. Therefore, the
catal"st outlet temperature is alwa"s higher than the :oudouard temperature. Howe4er, a
temperature drop in the reformer could make carbon deposition possible =&"bkGaer, #88*?.
Another t"pe of heat e0change reformer is the con(ection reformer. $n this reformer, the tubes
are heated mainl" b" the flue gas flowing upwards on the outside of the tubes, but also b" the
reformer gas flowing upwards inside the tube =&"bkGaer, #88*?.
The Topsoe con4ection reformer, see igure #% and #', has a single burner, which simplifies
the design and control of the unit. The burner is separated from the tube section. $t is called
con4ection reformer since it essentiall" means that the radiant tube section and the hot part of
the con4ection section are combined in a relati4el" small unit. The e0it temperature from the
reformer is appro0imatel" 1>> MC for both product gas and flue gas after heat e0change. This
reduction in temperature means that -> C of the fired dut" is utiliFed in the process,
compared to *> C in a con4entional S(. $n plants using the con4ection reformer, it is possible
#8
to balance the energ" need of the reformer b" the energ" a4ailable in the /SA off!gas, hence
a4oiding the energ" surplus in con4entional s"stems =&"bkGaer, #88*?.
igure #%. The Haldor Topsoe Con4ection (eformer, HTC(, for h"drogen production =www.topsoe.com?.
igure #'. HTC( reformer +left, with burner, principle of heat transfer in the tubes +right, =www.topsoe.com?
%>
3. Different reforming concepts for different applications
.1 (ntroduction
A t"pical AT( plant is shown in igure #.. $t comprises a feed preheat section, an AT(!
reactor, a heat reco4er" section and a gas separation unit. The AT( plant contains much less
e7uipment than the con4entional S( plant. &esulfuriFation is not needed if the feed is low!
sulfur natural gas. The schematic picture in igure #. shows the la"!out of an AT( plant for
the production of a s"ngas of H
%
@C5!ratio of %.>. Such a la"!out can be used for s"ngas
production to a methanol plant, ischer!Tropsch plant, for h"drogen production or for s"ngas
to an ammonia plant. $n the ammonia case, air has to be used in the AT( =&"bkGaer, #88*?.
igure #.. T"pical process la"!out for an AT(, for production of a s"ngas with H%@C5 ratio of %.> =&"bkGaer,
#88*?.
The concept with secondar" reforming is often used for s"ngas production to methanol or
ammonia s"nthesis. igure #* shows the la"!out of a plant for production of ammonia s"ngas.
90cept from the tubular S( reformer and the secondar" air!blown reformer, igure #* also
contains a pre!reformer. $n case of natural gas as feedstock, the pre!reformer is usuall" not
necessar" in the production of ammonia s"ngas. The concept with the pre!reformer is onl"
interesting when the steam production must be minimiFed.
%#
igure #*. State!of!the!art plant for production of ammonia s"nthesis gas =&"bkGaer, #88*?.
igure #1 shows the la"!out of the two!step production of s"ngas for methanol s"nthesis. The
conditions in the primar" S( unit are mild. $n the second reformer, o0"gen and steam are
used. =&"bkGaer, #88*?
igure #1. Two!step reforming plant for production of methanol s"nthesis gas =&"bkGaer, #88*?.
The methanol s"nthesis takes place at *> O #>> bar, and hence also for this application a
steam reformer and an autothermal reformer in series is sometimes used. Howe4er, since pure
o0"gen is needed for the methanol s"nthesis, the autothermal reformer will be rather
e0pensi4e. Therefore, more than -> C of the s"ngas production for methanol s"nthesis is
made solel" with steam reforming of methane =2ouliGn et al., %>>'?.
%%
.2 )ethanol synthesis
The ideal s"ngas composition for methanol production is H
%
@C5 H % and * 4olC of C5
%
,
which increases the acti4it". $t is actuall" the C5
%
, and not the C5, that reacts with H
%
on the
catal"st surface, Cu@Qn5@Al
%
5
'
, to gi4e methanol.
C5 I %H
%
CH
'
5H UH
(,%8-D
H !8>.- kJ@mol
C5
%
I 'H
%
CH
'
5H I H
%
5 UH
(,%8-D
H !.8.1 kJ@mol
C5 I H
%
5 C5
%
I H
%
UH
(,%8-D
H !.# kJ@mol
The newl" de4eloped methanol catal"sts ha4e selecti4ities of up to 88 C. This is essential
since methanol is thermod"namicall" less stable than other possible products, for instance
CH
.
. The methanol s"nthesis is limited b" e7uilibrium, wh" it is important to keep the
temperature low. This is obtained b" 7uenching the reaction b" cold feedgas at different
places in the bed as in the $C$!process +see igure #;,, or b" se4eral reactors with
intermediate cooling as in Haldor Topsoes process. Thus, the catal"sts must be acti4e at low
temperatures. (ec"cling is alwa"s performed in order to increase the o4erall "ield. T"pical
conditions for the methanol s"nthesis toda" are *> O #>> bar and *>> O **> D. Steam
reforming of methane is the most common wa" to produce the s"ngas, in which a H
%
@C5 ratio
of ' is obtained =2ouliGn et al., %>>'?.
igure #;. low scheme of the $C$ low!pressure methanol plant =2ouliGn et al., %>>'?.
$n order to correct the H
%
@C5@C5
%
ratio, prior to the methanol s"nthesis, either the surplus of
h"drogen is burnt or the C5 content is increased. This can be done in two wa"s)
#. $f C5
%
is a4ailable it ma" be added to the reformer or to the raw s"ngas.
%. :" installing an o0"gen!fired AT( downstream of the S(, or b" onl" using an AT( with
adGustment b" C5
%
remo4al in order to correct the carbon o0ide content.
%'
or both S( and AT(, the e0it gas mi0ture is in e7uilibrium.
H/!steam is often produced at the e0it of the AT(, or from the effluent gases from the
burners in the S(. This H/!steam can for instance be used to dri4e s"ngas compressors. <hen
S( and AT( are combined and heat e0changed, the high!temperature heat in the product gas
from the AT( is used to heat the steam reformer tubes, and hence onl" 2/!steam can be
produced in this case.
. !mmonia synthesis
The s"ngas for ammonia production re7uires a H
%
@3
%
ratio of ', which in -> C is made from
steam reforming of natural gas followed b" AT( with air.
3
%
I 'H
%
%3H
'
UH
(,%8-D
H !8#... kJ@mol
The e7uilibrium is fa4oured b" low temperature and high pressure. Howe4er, the reaction rate
is 4er" low below 1;> D. The t"pical ammonia s"nthesis conditions are 1;* D at inlet, ;%> O
;;> D at e0it, and #>> O %*> bar. The same reactor solutions as used in methanol s"nthesis are
used in the ammonia s"nthesis, hence cold!feed 7uenching +see igure #-, or se4eral reactors
with intermediate cooling. The single!pass con4ersion is low +%> O '> C, since the reaction is
limited b" e7uilibrium, and hence rec"cling is necessar" =2ouliGn et al., %>>'?.
.
igure #-. $C$ 7uench reactor and temperature!concentration profile =2ouliGn et al., %>>'?
." Fischer-Tropsch synthesis
The fundamentals in ischer!Tropsch +T, s"nthesis ha4e been described thoroughl" b"
Lgdberg =%>>1?. Here, the choice of s"ngas production concept will be discussed. igure #8
gi4es a schematic T la"!out.
The preparation of s"nthesis gas from natural gas stands for *> O ;* C of the capital cost in a
ischer!Tropsch plant. According to Aasberg!/etersen et al. =%>>'?, o0"gen!blown AT( is
considered to be the best option for large!scale, safe and economic s"ngas production. Air!
blown AT( has shown to be less efficient and more cost intensi4e considering the whole
%.
JTL!plant. The energ" consumption for the air!blown alternati4e is much higher. The energ"
consumption in the s"ngas section alone is #> C higher for the air!blown AT( compared to
the o0"gen!blown. The natural gas is often pre!reformed in an adiabatic reactor in order to
con4ert higher h"drocarbons to CH
.
, H
%
and C5, b" steam addition. The use of a pre!reformer
reduces the o4erall o0"gen consumption.
According to <ilhelm et al. =%>>#?, the s"ngas production for large!scale JTL plants in the
future will be the two!step reforming concept, or ultimatel" the AT( concept. The" claimed
that air!blown reformers will ne4er be competiti4e to o0"gen!blown ones in JTL!plants,
although the in4estment cost for the air separation unit is eliminated. The reason is that the
air!blown reformer appears much less fle0ible, has lower thermal efficienc" and re7uires high
air compression power. urthermore, the usage of an air!blown reformer makes it impossible
to recirculate T tail!gas, and the gas flows out from the reformer will be much larger which
in turn will increase the siFe of the e7uipment downstream the reformer. The T!tailgas can
usuall" be used for power production b" combustion in gas turbines. Howe4er, if it contains a
high concentration of 3
%
, it will ha4e a low heating 4alue and the power production might not
be 4iable.
igure #8. Schematic la"out of ischer!Tropsch plant =Aasberg!/etersen et al., %>>'?.
Steam to carbon ratios as low as >.1 inside the AT( ha4e been pro4en industriall". The
desired H
%
@C5 ratio out from the AT( is, in case of T, %.>. This low ratio can onl" be
achie4ed at e0tremel" low S@C ratios, see igure %>, without recirculating C5
%
=Aasberg!
/etersen et al., %>>'?.
igure %>. H%@C5 ratio as a function of S@C ratio in AT( outlet gas =Aasberg!/etersen et al., %>>'?.
%*
5peration at low S@C ratios impro4es process economics, which is shown in Table ..
Howe4er, low S@C increases the risk of carbon formation and hence catal"st deacti4ation. A
lot of research is going on in order to elucidate the causes of carbon formation, and the means
to pre4ent carbon formation or to destro" the carbon in the catal"tic bed. Carbon formation
has been found to depend upon feed gas composition, temperature, pressure and especiall"
burner design. /ilot plant demonstrations ha4e succeeded to run soot free operations at S@C
ratios as low as >.% =Aasberg!/etersen et al., %>>'?.
Table .. /rocess economics with different S@C ratios =Aasberg!/etersen et al., %>>'?.
:haradwaG and Schmidt =#88*? seemed to ha4e another opinion than Asberg!/etersen et al.
=%>>'? and <ilhelm et al. =%>>#?. The" claimed that partial o0idation is much faster, highl"
selecti4e in a single reactor, and much more energ" efficient than S( and AT(. $t could hence
reduce the capital and operating costs of s"ngas production, and would hence be the choice of
reforming concept in the future. Howe4er, the catal"tic partial o0idation +C/56, is still not
commercialiFed, and was discussed more in 2.) Partial oxidation.
According to Aasberg!/etersen et al. =%>>'?, C/56 suffers from a highl" flammable mi0ture
upstream the reactor at conditions rele4ant to JTL. The autoignition temperature is
appro0imatel" %*> MC for the mi0ture, therefore the inlet temperature of the mi0ture to the
reactor is low, t"picall" %>> MC. Compared to an AT( with a pre!reformer, the C/56 concept
consumes more natural gas and o0"gen for the same producti4it". This is due to that much of
the natural gas must be completel" o0idiFed in order to release enough energ" to raise the
temperature in the reactor to the desired e0it temperature. The e0it compositions will be
essentiall" the same in both concepts, pro4ided the amount and properties of the inlet streams
are the same. The authors meant that C/56 is not economical especiall" since it re7uires a lot
more o0"gen than the AT( concept. urthermore, the inherent safet" constraints make the
C/56 less attracti4e. Air!blown C/56 could be an alternati4e in case of making fuels for
fuel cells, howe4er not for T or methanol s"nthesis, as discussed abo4e.
4. New reforming concepts under development
3ormal steam reforming accounts for 1> O ;> C of the in4estment costs of a methanol or T!
plant based on natural gas, and the air separation unit stands for a maGor part. The
combination of steam reforming and autothermal reforming +the heat e0change reformer
concept, could to some e0tent reduce the in4estment and operating costs, and especiall" the
catal"tic partial o0idation +C/56, has a large potential to reduce the s"ngas production cost,
according to 2ouliGn et al. =%>>'?. C/56 is attracti4e since it has a high e0erg" efficienc",
which is due to the fact that no heat e0change is needed between a hot side and a cold side
since all reactions take place in the same 4essel. Also the $C$ combined reforming process
gi4es appro0imatel" '> C lower e0erg" losses compared to a con4entional steam reformer.
This is mainl" due to that the need for irre4ersible combustion reactions to heat the furnace in
the steam reformer case is eliminated in the combined reforming process. Howe4er, 2ouliGn
et al. =%>>'? claimed that Ka maGor breakthrough would be re7uired to make an" substantial
%1
sa4ings in the cost of producing s"ngasL. Some e0amples of such new ideas, still on the
de4elopment@demonstration stage are discussed below.
As discussed abo4e, the air!separation unit is a cost intensi4e part of the s"ngas preparation in
the methanol and T cases. $t has, howe4er, been shown economicall" ad4antageous to run
the AT(!burner on o0"gen compared to air, as discussed abo4e. A more promising alternati4e
to eliminate the air!separation unit would be a ceramic membrane reactor. $n this concept, an
ionic or o0"gen transporting membrane makes it possible to combine air separation and
partial o0idation in one unit. The de4elopment of such membranes is still relati4el" earl". The
membranes t"picall" work at ;*> MC or higher. 50"gen is transported through the membrane
with a selecti4it" of #>> C. 5n the inside of the membrane it reacts with the h"drocarbon. $f
the air must be compressed in order to ensure similar pressures on the two sides, the concept
might not be economical. $f ambient pressure air can be used, the membrane must ha4e a high
mechanical strength in order not to break at the high inner pressures desired for the s"ngas
production. The concept with ceramic membrane reforming seems promising, but its
feasibilit" must be pro4en =Aasberg!/etersen et al., %>>'?.
2embranes can also be used in order to remo4e the produced h"drogen from the product gas
+see igure %#,. This is called membrane reforming. $n this wa", the amount of h"drogen
produced is not limited b" the e7uilibrium and hence much lower temperatures can be used,
and no traditional C5
%
remo4al s"stem is needed. =2ouliGn et al., %>>'?
igure %#. Schematic picture of membrane reforming =www.co%captureproGect.org?.
Another techni7ue under de4elopment is the chemical looping reforming, in which chemicall"
bound o0"gen is used to o0idiFe methane. The s"stem consists of two reactors, as shown in
igure %%. $n the air!reactor, metal particles are o0idiFed to metal o0ide b" air +e0othermic
reaction,, and in the fuel reactor, methane is partiall" o0idiFed to s"ngas b" means of the
o0"gen in the metal o0ide +endothermic reaction,. The o0"gen carrier is continuousl"
circulated between the two beds, thus e0changing both o0"gen and heat
=www.co%captureproGect.org?. A possible alternati4e would be to install steam reformer tubes
in the fuel reactor, letting the complete o0idation of methane on the fluidiFed metal o0ide
particles gi4e heat to the steam reforming reactions inside of the tubes. Actuall", the heat used
for the steam reforming would be released when the o0"gen carrier is o0idiFed in the air
reactor, and transported to the fuel reactor with the metal o0ide particles. The complete
combustion of methane outside of the steam reformer tubes is actuall" endothermic and this
process is needed in order to reduce the metal o0ide back to the metal =("dVn, %>>1?. The
ad4antage with chemical looping reforming is that the separation of o0"gen from air is simple
and, in case of complete combustion, the C5
%
can be captured easil" since it is free from 3
%
.
Another ad4antage is that partial o0idation could be performed without forming the e0plosi4e
gas mi0tures of o0"gen and methane discussed in 2.) Partial oxidation.
%;
igure %%. Schematic picture of chemical looping partial o0idation =www.co%captureproGect.org?.
Another concept that implies C5
%
capture from hot flue gases, which is gaining interest due to
the possibilit" of C5
%
se7uestration, combined with h"drogen production, is the sorption
en"anced steam reforming process (S*RP). The idea is to mi0 the ordinar" steam reforming
catal"st with a C5
%
adsorbent, e.g. dolomite or lithium Firconate. C5
%
is captured, which
shifts the e7uilibriums of reactions +#, and +%, towards h"drogen, and makes it possible to
perform the steam reforming reaction at much lower temperatures +.*> O 1'> MC, than
con4entional. Hence, in4estment and operational costs ma" be significantl" reduced. $t is
possible to obtain a product gas containing 8; C H
%
on a dr" basis. After saturation, the C5
%
!
saturated adsorbent is regenerated b" heating. Dinetic limitations of the adsorbent howe4er
hinder the concept from being commercialiFed =5choa ernandVF et al., %>>*?.
5. Some notes on modeling of CP! and "#$
&e Jroote et al. =#881? made a one!dimensional heterogeneous model to simulate the
adiabatic C/56 of methane in a fi0ed!bed reactor. A plug!flow was assumed. The reactions
considered were C5
%
reforming, steam reforming, <JS, complete combustion, the
:oudouard reaction and methane cracking, for which kinetic rate e0pressions were set up.
$nternal gradients in the catal"st pellets were accounted for b" using effecti4eness factors.
These effecti4eness factors were 4er" small +between >.>* and >.>;, for all reactions e0cept
for the <JS +>.;,. $t was concluded that when air is used as o0idant, or when steam and C5
%
are added to the feed, acceptable temperatures are achie4ed in the bed. <hen o0"gen is used,
%-
a temperature ma0imum of #*>> MC was achie4ed, which would destro" the 3i!catal"st.
<hen steam is added, the amount of coke becomes negligible. C5
%
also reduces carbon
formation but increases the area in which coke is formed.
Hoang and Chan =%>>.? made a two!dimensional heterogeneous reactor model for an
adiabatic AT(. The application of the AT( was for h"drogen production to fuel cells. The"
onl" included four reactions, i.e. the complete combustion of methane, methane steam
reforming to C5 and H
%
+and to C5
%
, and <JS. The carbon deposition reactions were
neglected since the chosen S@C and air to carbon +A@C, ratios would ensure no formation of
carbon. The combustion chamber was actuall" not modeled, but the complete reactor was
assumed filled with 3i!catal"st. $t was assumed that the inlet temperature of the catal"st bed
was appro0imatel" '>> MC, which is the light!off temperature, i.e. the temperature at which
the autothermal reaction can be self!acti4ated.
:iesheu4el et al. =%>>'? made a one!dimensional AT(!reactor model for con4ersion of
methane. The" di4ided the model in two sections) an upstream o0idation section and a
downstream reforming section. $n the o0idation section, all of the o0"gen is con4erted, with
partial con4ersion of the fuel. An empirical fuel utiliFation ratio is used to 7uantif" which part
of the feed is con4erted in the o0idation section as a function of the relati4e flows of air and
steam. $n the o0idation section, the gas temperature rapidl" increases toward the top
temperature at the intersection with the reforming section. $n this section the temperature
decreases while the fuel is further con4erted with water and C5
%
as o0idant.
or the o0idation section, no kinetics were considered. The +partial, o0idation reaction in this
Fone was assumed to take place instantaneousl", and all o0"gen reacted, and hence no need
for a catal"st. 90perimental data of the top!temperature +i.e. the temperature at the start of the
catal"tic bed +the reforming section,, as a function of inlet flows led to an empirical
e0pression for the fuel utiliFation ratio, i.e. how much of the steam and CH
.
that was
con4erted in the o0idation section and the selecti4it" to H
%
and C5, and H
%
5 and C5
%
. $n
order to find this relation, the heat release and heat loss in the o0idation section and the C
p
!
4alues of the gas mi0tures must of course be taken into account. /robabl", the composition of
the gas e0iting the o0idation section was found b" iteration b" sol4ing mass and energ"
balances o4er the o0idation section, for a guessed outlet composition, until the calculated
temperature agreed with the measured. Hence the authors could correlate a measured top
temperature to a certain gas composition@flow at the inlet of the reforming section, and then
the actual modelling of the reformer section was made based on these inlet data.
igure %' and %. show some results from the simulation.
%8
igure %'. 50idation section in AT(. W H S@C ratio. Top temperature on left "!a0is. AtiliFation factor +dashed
lines, fraction of methane con4erted in the o0idation section, and thermod"namic CH. con4ersion +solid lines,
on right "!a0is. T> H .>> MC, adiabatic =:iesheu4el et al., %>>'?.
igure %.. Temperature profiles in the reforming section at different S@C ratios. The e0it temperature and the
inlet gas composition will describe the composition of the product gas. 50"gen@carbon ratio H >.*; =:iesheu4el
et al., %>>'?.
'>
Modeling of tricklebed reactors
! The &ischerTro#sch reaction
'#
1. %i&ed'(ed reactors vs tric)le'(ed reactors
$n a con4entional fi0ed!bed reactor, onl" gases and solid phases +the catal"st particles, are
present. This makes the modeling rather straight!forward. Howe4er, if a li7uid phase is added
to the fi0ed bed, the modelling becomes a bit more comple0. This concept is called trickle!
bed, and often means that the reaction takes place between a gaseous and a li7uid reactant, the
gaseous reactant ha4ing to diffuse through the li7uid phase to the catal"st surface and there
react with the other reactant alread" present in the li7uid.
The downflow cocurrent column packed with catal"st ma" operate in mainl" two different
flow regimes) Kthe trickle flowL in which the gas phase is continuous and the li7uid phase
dispersed, and Kthe bubble flowL in which the gas phase is dispersed and the li7uid phase
continuous. $ncreasing the gas flow rate will e4entuall" lead to pulsed flow. igure #
illustrates the different regimes.
igure #. low regimes in cocurrentl" operated packed bed. + is the gas load in =kg@m
%
, s? and , the li7uid load
in =kg@m
%
, s? =Jianetto and Sil4eston, #8-1?.
The flow patterns in a trickle!bed range between gas!continuous and li7uid!continuous
regimes, depending on shape and siFe of particles as well as on ph"sical properties and flow
4elocities of the gas and li7uid
Trickle!beds are widel" used in the refineries for processes such as h"drodesulfuriFation of
different petroleum fractions, h"drocracking of hea4" gasoils and atmospheric residues,
h"drotreating of lubricant oils and for h"drogenation =roment and :ischoff, #88>?. Trickle!
beds are also often used in the ischer!Tropsch s"nthesis, in which s"ngas is con4erted into
h"drocarbons. $n this case, the reactants are gaseous and the products gases and li7uids.
'%
The trickle!bed, with its stationar" catal"st bed, is preferred to a slurr" t"pe operation, in
which the catal"st is suspended in the li7uid, when the gas flow rate is relati4el" low because
it leads to a gas and li7uid flow pattern that better appro0imates the plug flow. A plug flow,
with its high reactant concentration, is fa4ourable if the total reaction order is positi4e. or
high gas flows, the slurr" operation might be better suited since it a4oids the pulsed flow
regime that can come up in a fi0ed!bed. Howe4er, Huang et al. =%>>.? reported that there
might be some ad4antages to work in the pulsing flow mode. The heat transfer is often poor in
the con4entional trickle!flow regime. This is undesirable especiall" if the reaction is
e0othermic, due to risk of run!awa", and also due to formation of undesired b"!products. 5ne
wa" to enhance the heat transfer in trickle!beds is to operate in a high gas!li7uid interaction
regime called pulsing flow b" Huang et al. =%>>.?. $n the pulsing flow mode, flows of gas!
continuous bases and li7uid!rich slugs +pulses, are alternated. The heat transfer rates in the
pulses can be ' O . times higher that those in the gas!continuous bases. The pulsing flow can
also enhance the wetting of the catal"st pellets and hence impro4e the li7uid distribution,
which in turn will minimiFe the risk for hot spot formation. $n the pulsing flow mode also the
mass transfer rates and the li7uid hold!up can be significantl" higher than in the con4entional
trickle!flow mode.
The main difference between the modeling of a fi0ed!bed and a trickle!bed is the
h"drod"namics because of two fluid phases in the latter case. urthermore, due to the e0tra
li7uid phase in the trickle!bed, more mass and heat transfer resistances ha4e to be
encountered, end hence gas!li7uid and li7uid!solid interfacial areas must be known. or a
con4entional fi0ed!bed, the mass transfer coefficient, -
g
, and the geometrical e0ternal particle
surface area, a
(
, is enough. or the mass transfer o4er the gas@li7uid interface, usuall" Henr"Es
constant, ., is used to estimate the concentration of a component at the interface.
igure % shows schematic concentration profiles of reactants and products in the different
phases e0isting in a trickle!bed.
''
igure %. $dealiFed diffusion!reaction s"stem in a three!phase trickle!bed reactor, where one reactant is in gas!
phase and the other in li7uid phase =Jianetto and Sil4eston, #8-1?.
2. Dispersion model e*uations for multip+ase s,stems
T"picall", mass and energ" balances for a #!dimensional dispersion model of a trickle!bed
reactor, with A as the reactant of the gas phase and : as that of the li7uid phase, would look
like this if the reaction onl" takes place at solid, and assuming constant superficial gas and
li7uid 4elocities)
)* for component ! in gas phase:
> , + , +
X
%
%

,
_


A,
A
( ,
A+
i+ ,
A+
eA+ ,
C
.
p
a /
d&
dC
!
d&
C d
'
where 0 and 0
,
are the 4oid fraction of packing and li7uid hold!up +=m
'
li7uid@m
'
reactor?,,
respecti4el". +0 - 0
,
, is e7ual to the gas hold!up, 0
+
,

in =m
'
gas@m
'
reactor?. !
i+
is the interstitial
4elocit" of gas in =m reactor@s? and hence the e0pression +0 ! 0
,
,Y!
i+
is e7ual to !
s+
, the
superficial gas 4elocit" in =m
'
gas@m
%
reactor, s?. $f !
s+
is not constant, it should be included in
the differential. a
(
1
is the gas!li7uid interfacial area per unit packed 4olume in =m
%
interface@m
'
reactor?. /
,
is an o4erall mass transfer coefficient from gas bulk to li7uid bulk in terms of the
li7uid concentration gradient, according to)
+ , ,
.- - /
# # #
+
where -
,
and -
+
are mass transfer coefficients from interface to li7uid bulk based on
concentration as dri4ing force, and from gas bulk to interface based on partial pressure as
dri4ing force, respecti4el", in =m
'
li7uid@m
%
interface, s? and =kmol@m
%
, bar, s?.
)* for component ! in li+uid phase:
'.
> , +
X X X
%
%

,
_

+
s
As A, ( Al A,
A
( ,
A,
i, ,
A,
eA, ,
C C a - C
.
p
a /
d&
dC
!
d&
C d
'
)* for component * in li+uid phase:
> , +
X X
%
%

s
$s $, ( $l
$,
i, ,
$,
e$, ,
C C a -
d&
dC
!
d&
C d
'
where 0
,
Y!
i,
is the same as the superficial li7uid 4elocit", !
s,
, in =m
'
li7uid@m
%
reactor, s?. -
l
is
the mass transfer coefficient between li7uid and catal"st surface in =m
'
li7uid@m
%
interface, s?,
and a
(
11
is the li7uid!solid interfacial area per unit packed 4olume in =m
%
interface@m
'
reactor?.
)* for component ! inside catalyst pellet, concentration gradients accounted for:
> , +
#
#
%
%

2R
3
3 A3 s
As
effA
r
d
dC
d
d
'


)* for component * inside catalyst pellet, concentration gradients accounted for:
> , +
#
#
%
%

2R
3
3 $3 s
$s
eff$
r
d
dC
d
d
'


where '
eff
is the effecti4e diffusi4it" coefficient inside the catal"st pellet in =m
%
@s?, 4
s
is the
catal"st densit" in =kg cat.@m
'
particle?, 2R is the number of reactions, 5
A3
is the stoichiometric
coefficient of component A in the G
th
reaction, and r
3
is the reaction rate for reaction G. The sum
of 5
A3
6r
3
is negati4e for a component that is consumed. $f mass transfer limitation inside of the
pellet can be neglected, for instance when the particle diameter is 4er" small which often is
the case in slurr" reactors, the last 2: ma" be simplified with)
)* for component ! at catalyst surface, no concentration gradients:
> , # + , +
#
X X
+

3
2R
3
A3 s
s
As A, ( Al
r C C a -
)* for component * at catalyst surface, no concentration gradients:
> , # + , +
#
X X
+

3
2R
3
$3 s
s
$s $, ( $l
r C C a -
'*
The mass balances abo4e are t"pical for an" multiphase s"stem. Simplifications are often
made such as assuming plug!flow of the gas. $n slurr" reactor models, in which the catal"st
particles are entrained with the li7uid, the slurr" phase is often modeled as a CST(, while in
trickle beds the li7uid phase often is treated as a plug flow. According to Shah and Sharma
=#8-;?, in general, the number of e7uations that must be set up to sol4e a multi!phase s"stem
e7uals the sum of all component balances for each phase plus the number of o4erall material
balances.
-* for gas phase:
> , + . , + , + , +
X
%
%
+
g 7 g7
t
g
l g ( a pg g i+ , ag ,
# # a
d
8
# # a "
d&
d#
c !
d&
# d

where 9
ag
is the a0ial conducti4it" in the gas phase in =<@m gas, D?, "
a
the heat transfer
coefficient from bulk gas to bulk li7uid in =<@m
%
interface, D?, 8
g
is an o4erall heat transfer
coefficient for heat transfer between gas and tube wall in =<@m
%
gas!wall interface, D? and a
g7
is the gas!wall interfacial area per inner tube wall area.
-* for li+uid phase:
> , + . , + , +
X X X
%
%
+ +
l 7 l7
t
l s
s l ( l l g ( a p, , i, , al ,
# # a
d
8
# # a " # # a "
d&
d#
c !
d&
# d

where 9
al
is the a0ial conducti4it" in the li7uid phase in =<@m li7uid, D?, "
l
the heat transfer
coefficient between li7uid and catal"st surface in =<@m
%
interface, D?, 8
l
is an o4erall heat
transfer coefficient for heat transfer between li7uid and tube wall in =<@m
%
li7uid!wall
interface, D? and a
l7
is the li7uid!wall interfacial area per inner tube wall area.
-* in particle, intraparticle temperature gradients accounted for:
> , +
#
%
%

,
_

3
2R
3
3 s
s e
r .
d
d#
d
d

<hile slurr" reactors are often considered as isothermal, trickle!beds are not. Therefore it is
necessar" to set up energ" balances for the different phases. Howe4er, according to Jakobsen
=%>>'?, the catal"st particles can often be considered as isothermal, and hence the last 9:
could be simplified with)
-* at particle surface, no intraparticle temperature gradients:
> , + , # + , +
#
X X
+

3
2R
3
3 s
s
s l ( l
r . # # a "
'1
Howe4er, according to Jianetto and Sil4eston =#8-1?, commercial trickle!bed reactors are
normall" considered as adiabatic, since the heat losses from the reactor are negligible
compared to the heat generated b" the reaction, and heat conduction is neglected.
urthermore, according to Shah =#8;8? and Jianetto and Sil4eston =#8-1?, all phases at an"
gi4en a0ial position in the reactor is often assumed to ha4e the same temperature. <ith these
assumptions, the 9: for a trickle bed could be simplified with the following e0pression)
-* for adiabatic tric.le-bed /ith temperature e+uality at any 0-point in the reactor:
> , + , # + , , + +
#
+

3
2R
3
3 s pg g i+ , p, , i, ,
r .
d&
d#
c ! c !
)omentum e+uation
$n the case of trickle beds, a momentum e7uation has to be set up. The standard 9rgun
e7uation was used b" <ang et al. =%>>'? in the case of a ischer!Tropsch reactor, in which
the li7uid phase was assumed Gust to be a thin film surrounding the catal"st particles.
roment and :ischoff =#88>? gi4e some alternati4e e0pressions for calculation of the pressure
drop in trickle beds, based on pressure drops measured for a certain bed with either li7uid, :
,
,
and gas, :
+
, flow onl". The e7uation below is one e0ample)
111 . > log
.#1 . >
log
%
%
+

,
_

+
+
,
+ ,

where :
2
is the two!phase frictional pressure drop. The measured pressure drop, dP
tot
;d&, is
related to :
2
according to)
, ,
tot
g
d&
dP

%
g is the acceleration of gra4it" in =m@s
%
?. The e7uations abo4e are onl" 4alid if !
s+
H !
s,
, since
the" are based on the assumption that the gas and li7uid are 7uasi!homogeneous. See roment
and :ischoff =#88>? for further e0amples. roment and :ischoff =#88>? also discussed that in
trickle!bed operation, the li7uid and gas flow rates are much lower than in for instance packed
absorbers, and hence single!flow pressure drop e7uations could be used as a first
appro0imation, with the 4oid fraction reduced to account for the li7uid hold!up.
';
3. Special issues in modeling of tric)le'(ed reactors
.1 (ntroduction
<hat make trickle!beds a bit more complicated to model than other multi!phase reactors, are
the facts that the catal"st is often not completel" wetted, insufficient li7uid hold!up,
backmi0ing etc. =Shah and Sharma, #8-;?. Jianetto and Sil4eston =#8-1? summariFed the
main differences between transport and reaction in porous catal"sts in multi!phase reactors
and con4entional gas!solid reactors with the following points)
! The catal"st pores in multiphase reactors are often filled with li7uid, and the diffusion in
li7uids is appro0imatel" . orders of magnitudes lower than that in gases. Concentration of
solutes in most cases will be lower than the concentrations of the gas species. The low
concentrations and the slow diffusion result in a low catal"st surface utiliFation, which in turn
means that the o4erall rates are much slower than if the reaction could be performed in gas!
phase.
! <hen the pores are filled, the heat conducti4it" is about one order of magnitude higher than
for gas!filled pores. This means that the catal"st particles can be assumed isothermal at each
a0ial point in the reactor, pro4ided the wetting of the catal"st is sufficient.
! $n a multiphase s"stem, the catal"st particle ma" be e0posed to non!s"mmetrical surface
conditions. /art of the catal"st surface ma" be in contact with the flowing li7uid, another part
in contact with a stagnant li7uid, and the rest of the particle might be unwetted. This means
that the solute concentration will 4ar" significantl" o4er the e0ternal catal"st surface. <hen
the wetting is poor, the e0othermic reactions can lead to e4aporation of the li7uid in some
pores. Those pores are hence dried out and completel" filled with gas. $n these cases, comple0
temperature gradients will be present inside of the catal"st.
A poor wetting is also belie4ed to be the maGor cause of poor effecti4eness. The incomplete
wetting leads to non!s"mmetrical concentration gradients in the catal"st pores and
complicates the prediction of reaction rates and the anal"sis of kinetic data from trickle!beds.
$n order to be able to e4aluate the true rate constant, -, from laborator" e0periments, it is
necessar" to be able to correlate the obser4ed rate constant, -
obs
, to the true one.
$n Table #, the ad4antages and disad4antages with trickle!beds as summariFed b" Shah =#8;8?
are gi4en.
'-
Table #. Ad4antages and disad4antages with trickle!beds =Shah, #8;8?.
.2 1atalyst /etting and relating .
app
to .
Assuming first order kinetics of the reaction inside a trickle!bed, the apparent rate constant,
-
app
, will for instance depend on the thickness of the li7uid film surrounding the catal"st
pellets, which in turn is dependent on the li7uid flow rate, and on the diffusion resistance
inside the pellets. The -
app
will of course be different from +lower than, the true rate constant,
-, as long as the reaction is limited b" mass transfer. $f the reaction is not trul" first order,
different -
app
4alues will be obtained at different con4ersions.
:ondi =#8;#?, a reference in roment and :ischoff =#88>?, deri4ed the following relation for
trickle!beds)
b
, app
,
A
- - , @ +
# #
X

where >.* R b R >.;


where A1 is a constant. The higher the ,4
,
@<, the closer -
app
@- is to # for h"drodesulfuriFation
of a hea4" gasoil, which is understandable since the film thickness of li7uid surrounding the
particles decreases with increased li7uid flow rate, and the wetting of the pellets increases.
-
app
@- would howe4er ne4er reach # as long as there are intraparticle concentration gradients.
'8
$n case of porous packing, two t"pes of wetting can be defined =Shah, #8;8?)
! $nternal wetting or pore filling. This is often complete due to capillar" action.
! 90ternal effecti4e wetting, i.e. the amount of the particlesE outside area effecti4el"
contacted b" the li7uid. Almost all mass e0change between the internal li7uid and the
flowing li7uid occurs through this area.
Little is known on how the li7uid!solid contact efficienc" depends upon the gas and li7uid
flow rates. Higher gas and li7uid loadings can increase the contact. Also a higher 4iscosit"
and lower surface tension increases the contacting efficienc". Satterfield, a reference in Shah
=#8;8?, defined the contacting effecti4eness in terms of the apparent rate constant, -
app
,
obtained from laborator" e0periments in trickle!beds, and the true rate constant, -, obtained in
a CST(, and found the correlation depictured in igure '.
igure '. Contacting effecti4eness 4s li7uid loading, as proposed b" Satterfield. +, is the superficial li7uid flow
rate per unit cross!section +!s,, =Shah, #8;8?.
According to Doros, another reference in Shah =#8;8?, there is little correlation between flow
uniformit" and contacting effecti4eness. The contacting efficienc" ma" increase e4en if the
flow uniformit" is poor. The interstitial bed geometr", catal"st shape and siFe pla" important
roles in the contacting effecti4eness. At high li7uid mass 4elocities, high contacting
effecti4eness can be achie4ed e4en for reactor ) catal"st diameter ratios as low as #.. ) #. or
some s"stems, no correlation between the contacting effecti4eness and mass 4elocit" can be
found.
2ears =#8;.?, another reference in roment and :ischoff =#88>?, meant that correlating the
-
app
to - and li7uid flow rate +or li7uid d"namic hold!up, was not correct, and instead
suggested that -
app
should be proportional to the e0ternal wetted area of the catal"st, which in
turn depends on parameters such as =
,
, >
,
, >
,?c
etc. This is understood since in con4entional
trickle!beds both components in gas and li7uid are reactants that must meet and react at the
catal"st surface. The lower the degree of wetting, the smaller the interfacial contact area
between li7uid and solid and hence the slower the mass transfer rates of : to the catal"st
acti4e sites. The transport of the gas phase reactant A will howe4er be faster the lower the
.>
degree of wetting, which might actuall" increase the global reaction rate if A is the limiting
reactant. $n the h"droprocessing of oils, a sufficientl" large li7uid hold!up is essential, since if
this is too low, the whole capacit" of the catal"st will not be used, due to incomplete wetting.
$n case of ischer!Tropsch, where all reactants are in the gas phase, the degree of wetting
should primaril" affect the mass transfer rates from gas to catal"st surface, probabl" gi4ing a
higher o4erall rate the lower the degree of wetting, pro4ided that the reaction rate is diffusion
controlled. Jianetto and Sil4eston =#8-1? concluded that in trickle!beds, essentiall" all
reaction occurs b" mass transfer of reactant from the gas!co4ered part of the catal"st surface.
. 2i+uid hold-up and mass transfer
:iswas et al. =#8--?, a reference within roment and :ischoff =#88>?, performed e0periments
that led to the conclusion that in order for the catal"st particles to be full" wetted, an efficient
li7uid distribution at the top is re7uired, and the column!to!particle diameter should e0ceed %>
O %* to a4oid li7uid b"passing along the wall. This is of course of big importance if the
trickle!bed in4ol4es reaction between components in the gas and the li7uid, in order to full"
utiliFe the capacit" of the catal"st loaded into the reactor.
$n the trickle!bed, the li7uid holdup, 0
,
, consists of li7uid inside the pores +internal li7uid
hold!up, and outside of the pores +e0ternal li7uid holdup,. The e0ternal li7uid hold!up
consists, in turn, of the d"namic hold!up, 0
,
1
, and the static hold!up. The static hold!up was
related to the 9t4s number, :
,
YgYd
p
%
@>
,
+>
,
is the surface tension of the li7uid in =3@m?,, and
it 4aries between >.>% and >.>* +Charpentier et al. =#81-?, a reference within roment and
:ischoff =#88>?,. roment and :ischoff =#88>? presented the following e0pression for the
d"namic hold!up. 3ote that it is onl" dependent on the li7uid flow rate and not on the gas
flow rate)

, +
%
% '
X
p (
,
, p
,
p ,
,
d a
g d , d
c

,
_

,
_

where , is the 4olumetric li7uid flow rate in =m


'
li7uid@s?, < the cross sectional reactor area
=m
%
reactor?, =
,
the li7uid 4iscosit" in =kg@m, s? and a
(
is the e0ternal particle surface area per
unit of reactor 4olume in =m
%
particle@m
'
reactor?. c, 5, @ and A are constants +see roment and
:ischoff =#88>?,. Although not discussed b" roment and :ischoff =#88>?, this d"namic hold!
up could probabl" find use in a so!called t7o-&one model, in which the li7uid is looked upon
as a flowing fraction and a stagnant fraction, as described below in $ac-mixing.
Henr" and Jilbert, a reference in Shah =#8;8?, claimed that a certain minimum li7uid hold!up
is re7uired for #>> C catal"st utiliFation, as illustrated in igure ..
.#
igure .. 9ffect of fluid d"namics on hold!up =Shah, #8;8?.
The li7uid hold!up was claimed to depend on +Br
,
@Re
,
,
#@'
, where Br
,
H !
s,
@4
,
%
d
p
g and Re
,
H
!
s,
d
p
@=
,
.
There are some correlations that might be used to estimate interfacial areas in the trickle!bed.
5ne e0ample is the following e0pression, originall" de4eloped for a countercurrent packed
column, for 4
,
,@< H #.* kg@m
%
, s)
#- . >
, #'* . > >. . >
X
(e >* . #

,
_

,
c ,
,
(
(
Ce
a
a

where Ce H 4
,
,
%
d
p
@<
%
>
,
, and >
,?c
is the critical surface tension abo4e which the packing cannot
be wetted.
(especting the mass transfer between gas and li7uid in trickle!beds, there are some specific
results for cocurrent operation a4ailable in literature. or low li7uid and gas flows,
Charpentier =#8;;? +a reference in roment and :ischoff =#88>?, de4eloped the following
correlation for the li7uid side mass transfer)
8
X
#> . . %
>>## . >

A,
, ( ,
'
* a - =s
!#
?
where '
A,
is the diffusi4it" in li7uid in =m
%
@s?, assuming that the li7uid 4iscosit" is not too far
from that of water, and *
,
H +DP
tot
@D&,Y!
s,
in =<@m
'
?. (eiss, a reference in Shah =#8;8?, found
a correlation for the li7uid side mass transfer at higher gas and li7uid rates +pulsing or spra"
flow,)
* . > X
#;' . >
, ( ,
* a -
igure * illustrates the abo4e discussion.
.%
igure *. Correlation between li7uid side mass transfer coefficient -,a, 4alues for cocurrent downflow in packed
beds. a, in the figure is the gas!li7uid interfacial area, i.e. a(
1
. 8E, in the figure is the superficial li7uid 4elocit",
i.e. !s, =Shah, #8;8?.
The gas!side mass transfer, -
+
a
(
1
, is also depending on the li7uid and gas flow rates which is
illustrated in igure 1.
igure 1. 9nerg" correlation for -+. 8E is the superficial li7uid 4elocit" +!s,. 0 is the bed 4oidage. F is the packing
shape coefficient, 4alues are found in Table % below =Shah, #8;8?.
.'
Table %. Salues for porosit" and packing shape coefficients used in igure 1 =Shah, #8;8?.
or the li7uid!solid mass transfer, 3
'
correlations for single!phase flow can be used as a first
appro0imation. These correlations are gi4en in roment and :ischoff =#88>?.
&ealing with mass transfer limitations inside of the catal"st pellets in a trickle!bed is done in
the same wa" as for single!flow fi0ed beds, i.e. with particle e7uations or effecti4eness
factors. Howe4er, the trickle!bed case is much more comple0, since in the case of li7uid
e4aporation +in h"drodesulfuriFation or h"drocracking, the pores will be filled with both
4apour and li7uid, and the theor" of the effecti4eness factor for such a situation still has to be
worked out.
." 'eat transfer
The trickle!bed is characteriFed b" poor heat transfer rates, compared to reactor
configurations in which the li7uid is continuous. The transition from homogeneous trickling
flow to pulsed flow correspond to an increase of se4eral hundred percent in the radial heat!
transfer rate. There are not man" studies about the heat transfer in trickle!beds. 5ne relation
describing the heat transfer in the trickle!mode was deri4ed b" <eekman and 2"ers, a
reference in Shah =#8;8?)
+ +
l
g
, ,
l l
e
-
-
- -
-
/r (e #;% . > /r (e >>#;. . >
;# . ;

,
_

+ +
where -
e
is the effecti4e conducti4it" of the bed, -
l
and -
g
the li7uid and gas conducti4ities.
Re
+
and Re
,
are the gas and li7uid (e"nolds numbers based on superficial 4elocities and tube
diameter. Pr
+
and Pr
,
are the gas and li7uid /randtl numbers. The o4erall heat!transfer
coefficient was related to the effecti4e conducti4it" as)
G
c ! c ! r
-
r
8
p+ s+ p, s, t
e
t
, + #-' . >
-8% . %

+
+ =cal h
!#
cm
!%
D
!#
?
where r
t
is the tube radius in =cm? and !
s,
and !
s+
are in =g@h, cm
%
?. c
p+
6
is the specific heat of
saturated air in =cal@g, D?. G is the a0ial distance from the inlet in =cm?. These relations were
found for an air!water cocurrent downward s"stem with packing of spheres with siFes
between '- mm and 1* mm. The radial heat transfer coefficients were found to be much
larger than those obser4ed for single phase li7uid flow. The effect of the gas on the heat
transfer was mainl" that it increased the 4elocit" of the li7uid phase. The radial 4elocit" in the
li7uid was larger in two!phase flow.
..
.% -ffecti3eness factor
The use of effecti4eness factors instead of sol4ing the particle e7uations is con4enient.
According to Jianetto and Sil4eston =#8-1?, the effecti4eness factor change much more
slowl" through the reactor than do the reaction rate or concentration. Thus, the effecti4eness
factor onl" has to be estimated at the inlet, in the middle and at the outlet of the reactor.
Jianetto and Sil4eston =#8-1? reported e0pressions for the Thiele modulus and effecti4eness
factor for 4arious kinetic reaction rate e0pressions.
Howe4er, when the wetting is not complete, con4entional intraparticle effecti4eness factors
are not applicable since the concentration of reactant at the e0ternal catal"st surface is not
uniform. According to Jianetto and Sil4eston =#8-1?, it is necessar" to consider mass transfer
from gas to li7uid, from li7uid to the portion of the particle in contact with flowing li7uid, and
direct transport from gas to the part of the particle in contact with the gas. 90pression for an
o4erall effecti4eness factor in case of spherical catal"st particles partl" in contact with
flowing li7uid, and partl" in direct contact with gas, was deri4ed b" Jianetto and Sil4eston
=#8-1? for first order kinetics. The whole idea is that the o4erall effecti4eness factor, H
o
, is
di4ided into two parts O one for the wetted catal"st surface +f, and one for the Kdr"L catal"st
surface +#!f,, where f is the wetting efficienc")
g l o
f f , # + +
where H
l
and H
g
are effecti4eness factors for entire particle surface +i.e. o4erall effecti4eness
factors that include e0ternal mass transfer resistance as well as internal, co4ered b" li7uid and
gas, respecti4el". The particle effecti4eness factor, H, i.e. the one concerning internal mass
transfer restrictions onl", is included in the e0pressions for H
l
and H
g
. See Jianetto and
Sil4eston =#8-1? for the complete e0pression. The o4erall rate, r, of formation@disappearance
of one component can then be e0pressed as)
Y
l o
-C r
where C
l
6
is the li7uid phase concentration in e7uilibrium with the bulk gas concentration.
/artial wetting should onl" affect the o4erall rate@effecti4eness factor if H
g
I H
l
. $n order to
estimate the role of wetting on the o4erall rate, it is necessar" to know the wetting efficienc"
f. A correlation useful for scale!up is)
1
1
]
1

,
_


;* . >
, % . > >* . >
'1 . # e0p #
,
c ,
l l
Ce +a f

where +a
l
H d
p
'
g@J
l
%
, and J
l
the kinematic 4iscosit" of the li7uid in =m
%
@s?.
.*
.& *ac.mi$ing
.&.1 4ne-parameter models
The #!dimensional dispersion models characteriFe the backmi0ing with Gust a single
parameter O the dispersion coefficient. The assumption that all the mi0ing processes follow a
ickEs law t"pe of diffusion becomes increasingl" dubious with increasing degree of
backmi0ing. :ut the dispersion models are still 4er" widel" used due to their simplicit", since
the" describe the backmi0ing +or (T&, b" simpl" one parameter. $n multiphase reactors, the
degree of backmi0ing is accounted for in each phase. The degree of backmi0ing is often much
higher for the slow mo4ing li7uid phase compared to the fast mo4ing gas phase, which often
leads to a plug!flow model +Fero dispersion@backmi0ing, for the latter =Shah, #8;8?. The
effecti4e a0ial dispersion coefficient, '
ea
, is e0pressed in dimensionless form as the /eclet
number +based on particle diameter,)
ea
p i
a
'
d !
Pe
where !
i
is the interstitial 4elocit" in =m reactor@s? of the phase encountered. The /eclet
number, at a (e"nolds number of #> +based on particle diameter,, is appro0imatel" >.% for the
li7uid phase in the trickle!bed, compared to appro0imatel" % for the fluid in a single!phase
operation =roment and :ischoff, #88>?.
roment and :ischoff =#88>? claimed that in laborator"!scale, the minimum reactor length for
absence of a0ial dispersion in the gas phase can be an order of magnitude higher for a trickle!
bed than for single!phase fi0ed beds. Howe4er, in industrial reactors, the a0ial dispersion in
gas phase can be neglected.
9lenko4 and Dole4 =#8;%?, a reference in roment and :ischoff =#88>?, de4eloped the
following e0pression for estimation of '
ea
in the li7uid in a trickle!bed)
'' . >
% '
%
;- . >
.
>1- . >

,
_

,
_

, (
,
, (
,
( ea
s,
a
g
a
,
a '
!

&e4iations from plug!flow in the gas!phase of trickle!bed reactors are often not concerned. 5f
course the con4entional wa" to find '
ea
is to make a tracer e0periment in the phase of interest,
and then to calculate the *(t), the residence time distribution function. The *(t) describes the
dimensionless outlet tracer concentration +C
o!t
(t);C
o!t
(t) integrated from t
E
- t

, as a function of
time, and b" setting up mass balances and boundar" conditions @ initial 4alues for the tracer
+including the dispersion term,, the numerical solution +i.e. C
o!t
(t), e0pression ma" be fitted to
the e0perimentall" found *(t) b" choosing a suitable '
ea
.
According to Shah =#8;8?, maldistribution of the flow is common in laborator"!scale trickle!
beds. The maldistribution, such as channeling, dead Fones and b"passing etc. result in
comple0 (T& e4aluation. $f such phenomena cannot be a4oided in the s"stem, micromi0ing
models +see ogler =#888? for an e0planation, ha4e to be set up to correlate such unusual
shapes of (T& cur4es. 2icromi0ing models describe how molecules of different ages
encounter one another in the reactor, unlike the macromi0ing models +i.e. the (T&, which
onl" describe the distribution of residence times.
.1
(T&s for each phase is needed in order to describe the backmi0ing. Howe4er, as mentioned
abo4e, the gas phase is often considered to be a plug flow. igures ;!#> below show an ideal
(T& and (T&s for s"stems with maldistributional flows for t"pical trickle!bed reactors.
igure ;. An ideal (T& cur4e, and a (T& cur4e with tailing =Shah, #8;8?.
igure -. (T& cur4e for a reactor with channeling =Shah, #8;8?.
igure 8. (T& cur4e for a reactor with large dead space =Shah, #8;8?.
.;
igure #>. (T& cur4e for a reactor with b"passing =Shah, #8;8?.
Channeling and b"passing are undesirable phenomena since the" reduce the utiliFation of the
catal"st. /roper design of the li7uid distributor and a calming section in large!scale trickle!
beds can essentiall" eliminate the maldistributions of flows. $t is also important to use the
same t"pe of particles +same porosit", in the large!scale as in the laborator" scale since tailing
will occur due to the static li7uid hold!up in the pores, shown in igure ;.
Shah =#8;8? presented se4eral models for describing the backmi0ing@macromi0ing in trickle!
bed reactors. The most common, although with a rather poor capabilit" of predicting the
backmi0ing, are the a0ial dispersion models, as described abo4e in 2. 'ispersion model
eK!ations for m!ltip"ase sstems. The dispersion model gi4es satisfactor" correlation to the
measured (T& data as long as radial!flow nonuniformities are minimiFed b" keeping a large
reactor diameter @ particle diameter +P %*,. 2ichell and urFer, a reference in Shah =#8;8?,
claimed that a dispersion model of the li7uid phase does not describe the true nature of
trickling flow. $n their so called Kb"pass modelL, the li7uid phase is modeled as a laminar film
flow o4er a series of packing elements with imperfect mi0ing, or b"passing, at each packing
Gunction. The li7uid film is assumed plug!flow, and the static li7uid hold!up at the Gunctions
well!mi0ed b" h"drod"namic effects, such as ripples, that cause a0ial dispersion in the
reactor. At each Gunction, a fraction K of the flow enters the perfectl" mi0ed regions, while the
rest of the flow b"passes the mi0ing. This is illustrated in igure #>.
igure #>. Schematic picture of the b"pass model =Shah, #8;8?.
.-
2ichell and urFer deri4ed the following e0pression for the mean residence time, t
m,
, of the
li7uid)
L B
m,
t K t t +
where Z is the bed height di4ided b" packing diameter, t
B
is the mean residence time of a
single film and t
L
the mean residence time in a single mi0er. 90pression for t
B
and t
L
are gi4en
below)
f
p
B
!
d
t
s,
p stat, ,
L
K!
d
t

where !
f
is the mean flow 4elocit" in the film, 0
stat,
the static li7uid hold!up and !
s,
the
superficial li7uid 4elocit" in =m
'
li7uid@m
%
reactor, s?. !
f
is related to the wetted surface area,
a
7
, according to)
' @ # ' @ %
.-
.

,
_

,
_

,
,
, 7
s,
f
g
a
!
!

The e0pression for *(t) for a &irac pulse tracer input, in case of the b"pass model e7uations,
will be)

,
_

,
_

L x
x
L
x x
t
t
t
t
x x x
K K
t K t * e0p
,[ # + ,[ # + ,[ +
, # + [
, + , # + , +
#
#

Since a
7
and !
s,
often can be estimated from e0periments, the onl" unknown parameter that
has to be found, in order to fit the e0perimental results to the model, is K. Hence, both the
dispersion and b"pass models are t"pical single!parameter models +'
ea
and K,.
.&.2 T/o-parameter model
Shah =#8;8? presented three two!parameter models and one three!parameter model for
description of the backmi0ing in multiphase reactors. :elow, one two!parameter will be
shown as an e0ample.
This model was referred to as t"e t7o-parameter modified mixing-cell model in Shah =#8;8?,
and as t"e t7o-&one model in roment and :ischoff =#88>?. According to roment and
:ischoff =#88>?, the de4iations from plug!flow in the li7uid phase of a multi!phase packed!
bed reactor are mainl" caused b" insufficient contact between the gas and li7uid. The reason
for this phenomenon ma" be preferential paths in the packing or stagnant Fones. These effects
can often not be described satisfactoril" b" effecti4e diffusion models. A more appropriate
.8
description could be made using a two!Fone model. :riefl", in such a model, onl" a fraction
of the li7uid flows through the packing, and the rest of the li7uid is considered stagnant. The
stagnant li7uid phase is assumed to be well!mi0ed and e0changes mass with the flowing
fraction of the li7uid and with the catal"st particles. 3ote that in the one!parameter b"pass
model described abo4e, no mass transfer between the film and stagnant Fones was
encountered.
The mass balance for A in the gas!phase is written as shown in 2. 'ispersion model eK!ations
for m!ltip"ase sstems. The mass balance for A in the li7uid phases can be written =roment
and :ischoff, #88>?)
)* for ! in flo/ing fraction of li+uid:
> , + , +
X X X X

,
_

+
s
As
f
A, ( l
f
A,
A
( ,
d
A,
f
A, #
f
A,
i,
f
,
C C a - C
.
p
a / C C -
d&
dC
!
where !
i,
1
is the interstitial 4elocit" of the flowing fraction of the li7uid, -
l
is the mass transfer
between the flowing fraction and the solid, -
#
the mass transfer between the flowing and
stagnant li7uid. The superscripts f and d stand for flowing fraction and stagnant fraction,
respecti4el".
)* for ! in stagnant fraction of li+uid:
, + , +
X s
As
d
A, l
f
A,
d
A, #
C C - C C -
where -
l
1
is the mass transfer coefficient between stagnant li7uid and solid.
)* for ! at catalyst:
, # + , + , +
X X X
+
s A
s
As
d
A, l
s
As
f
A, ( l
r C C - C C a -
-
#
and -
l
1
contain interfacial areas that are not "et well established.
$n order to sol4e the e7uations, the flowing fraction of li7uid, f, must be determined, so that 0
,
f
can be estimated. $t is f and -
#
that comprise the two parameters needed to fit the e0perimental
tracer data with *(t). The e0pression for *(t) can be found in Shah =#8;8?. f can be estimated
b")
m
i
t
t
f #
where t
i
and t
m
are the time when the tracer first appears in the effluent, and the a4erage
residence time of li7uid in the reactor, respecti4el". The 4ariance of the (T& is gi4en b")
m #
t - f @ %
% %

*>
t
i
, t
m
and >
2
are easil" achie4ed from the (T& cur4e, and hence both f and -
#
can be estimated
rather easil" from the (T& e0periment.
.5 1onclusion
There is still not enough knowledge in order to make a model that can predict the different
results obtained in laborator"!scale and large!scale trickle!bed reactors, and hence the design
practice for trickle!beds is still in an earl" stage of de4elopment. The high degree of
backmi0ing that can arise in multiphase s"stems is often difficult to describe with a
con4entional dispersion model, as discussed abo4e. This is due to the comple0 phenomena
that often take place, such as droplets breaking loose from the li7uid phase and getting
mi0ed@entrained with the gas flow, or in the case of slurr" reactors gas bubbles coalesce and
break continuousl". Some li7uids will also form foam in the column at certain conditions,
which is not eas" to predict. $n the multiphase s"stems the interfacial contact areas between
the different phases is important for the mass and heat transfer since the resistance towards
transfer mainl" e0ist in the interfacial KfilmsL. At certain conditions, these films will break up
and the mass and heat transfer rates ma" therefore change drasticall". =Jakobsen, %>>1?
Shah =#8;8? summariFed the problems with (T& and scale!up for trickle!beds with the
following points)
! (T& data from laborator"!scale e0periments cannot be used for large!scale units.
5ften the pre4ailing flow regimes are different in different scales.
! $n trickle!beds, maldistribution of the flow is common, which makes it difficult to
e4aluate (T& data. 2aldistribution, such as channeling, dead Fones and b"passing,
often e0ist in laborator"!scale units but not in larger scale.
*#
#+e %isc+er'#ropsc+ process
".1 (ntroduction
The ischer!Tropsch +T, s"nthesis, in which s"nthesis gas is con4erted to li7uid fuels such
as gasoline and diesel is also called JTL +Jas!To!Li7uids,. eedstocks for the s"nthesis gas
can be coal, natural gas, biomass etc. The T!process was disco4ered in the earl" #8%>Es and
has been used from time to time at places short of petroleum, such as Jerman" during <<%,
and in South Africa during the oil embargo. $n the last ten "ears, the T!process has gained
new world!wide interest due to the forecasts about diminishing petroleum reser4es, and due to
the more stringent emission legislations coming up due to the green!house effect. T!fuels are
much cleaner than petroleum!based fuels since the" are essentiall" free from sulfur and
aromatics. urthermore, the" can be C5
%
!neutral if made from biomass.
The commercial!scale T!plants e0isting in the world toda" are three coal!based plants in
South Africa +#*> >>> barrels
#
per da" +bpd,, Sasol,, one natural!gas based plant in South
Africa +%' >>> bpd, /etroSA,, and one natural!gas based plant in 2ala"sia +#* >>> bpd,
Shell,. rom this "ear and within the ne0t * "ears there will be se4eral commercial!scale
natural!gas based T!plants set up in Batar, which is a countr" with huge natural!gas
resources. The companies acti4e in this countr" are Batar /etroleum@Sasol Che4ron, Shell,
900on2obile and Conoco/hillips. Statoil recentl" erected a natural!gas based T!
demonstration unit in South Africa, which is a slurr" reactor with a capacit" of # >>> bpd. $n
Jerman", Choren@Shell are building a commercial!scale biomass!based T!plant. Sweden has
currentl" one pilot!plant for gasification of biomass in S\0G, and one for gasification of
black li7uor +a rest product from the pulp industr", in /ite]. Still, the Swedish energ" has not
decided what fuels will be made from the s"ngas produced at this pilot!plants. A feasibilit"
stud" made b" 9kbom et al. =%>>*? showed that appro0imatel" %*!'> C of the transportation
fuels in Sweden can be replaced b" C5
%
!neutral fuels +T!fuel, &29 or methanol, if all pulp
mills in Sweden replace their old reco4er" boilers b" new black li7uor gasifiers. This option
would still co4er the steam and energ" need of the pulp mills.
$n the ischer!Tropsch +T, s"nthesis, s"nthesis gas +C5 I H
%
, is con4erted o4er e or Co!
based catal"sts into h"drocarbons and water according to reaction +#, =Sie and Drishna,
#888?.
C5 I %H
%
!CH
%
! I H
%
5 UH
(,%8-D
H !#1* kJ@mol +#,
2ainl" straight!chain saturated h"drocarbons from CH
.
up to hea4" wa0es ma" be produced.
Some olefins and o0"genates are also produced. The reaction mechanism is a pol"merisation
process in which the building block, or monomer, is the CH
%
!unit. &ue to the step!wise
growth mechanism, the h"drocarbon products will alwa"s be a range of products of C
#I
,
described b" the AS +Anderson, SchulF, lor", !distribution, shown in figure #. The chain
growth probabilit", 5, is a common parameter used to characterise a T product distribution.
A higher 5 means that the product contains higher h"drocarbons, and hence less CH
.
. $t is
known that the higher the pressure, and the lower the temperature, and the lower the inlet
H
%
@C5!ratio, the higher is 5. The chain growth probabilit" is also dependent on the
characteristics of the catal"st +e.g. pellet siFe, pore siFe, acti4e site densit", promoters etc.,. $t
#
# barrel e7uals #*8 dm
'
.
*%
is commonl" accepted that the highest "ield to diesel +the most promising T!fuel due to a
4er" high cetane number realised b" the un!branched straight h"drocarbon chains, is achie4ed
b" first making wa0, which is then h"drocracked into the diesel fraction +C
#%
O C
%>
,. $n this
wa" "ou will minimise the formation of shorter b"products, especiall" CH
.
. Hence, the world!
wide T!research is toda" focused on how to prepare catal"sts that gi4e high 5, and how to
increase the dispersion of the Co in order to minimiFe the amount of e0pensi4e Co needed for
a certain producti4it". A normal 5 toda" is around >.8 for a wa0!producing T!process.
The molar usage ratio of H
%
@C5 is about % o4er a Co!based catal"st +see reaction +#,,. The
reaction is strongl" e0othermic, which makes a rapid heat remo4al necessar" from the reactor.
An efficient heat remo4al is a ke" point in the design of T reactors, due to the risk of run!
awa", and also due to the temperature sensiti4it" of the product selecti4it".
or e!based catal"sts, the water!gas!shift +<JS, reaction +reaction +%,, =Jess, #888?, takes
place simultaneousl" with reaction +#, in the reactor, and hence lowering the usage ratio,
which makes it possible to feed the T!reactor with a molar inlet ratio of H
%
@C5 of less than
%. Such h"drogen poor s"nthesis gas is obtained from gasified coal or biomass. The <JS
reaction is known for its e7uilibrium composition which is located Kin the middleL between
the reactants and products. Therefore, the <JS e7uilibrium is often tuned in in chemical
processes.
C5 I H
%
5 C5
%
I H
%
UH
(,%8-D
H !.# kJ@mol +%,
Howe4er, for the case of low!temperature T +LTT,, which is used when high wa0es are
wanted, the <JS reaction is slow and does often not reach e7uilibrium =&r", #881?. Hence,
no commercial T!process e0ists toda" that can handle, directl" in the T!reactor, such a low
H
%
@C5!ratio as #.>, which is the ratio in for instance gasified biomass. $nstead, e0ternal <JS!
units are used in order to correct the H
%
@C5!ratio in the s"ngas before entering the T!reactor.
The T!reaction is not limited b" e7uilibrium, as is the case in, for instance, the methanol
s"nthesis. 54er a Co!based catal"st, the reaction products +HCs and water, are not belie4ed to
affect the reaction rate, and hence a high con4ersion per pass is possible +abo4e 8> C possible
in a slurr" phase reactor,. 54er a e!based catal"st, water has shown to limit the reaction rate,
and hence the con4ersion per pass is rather low and water is knocked out prior to
recirculation. :elow, two generalised rate e0pressions for T o4er a e! and Co!based catal"st
are presented, respecti4el" =&r", %>>%?.
or e)
or Co)
Co is often preferred to e if the s"ngas is not poor in h"drogen. This is due to its higher
acti4it", the lack of <JS!acti4it" +which would otherwise lead to a too high H
%
@C5!ratio
inside of the reactor,, and the fact that the reaction rate is not limited b" the presence of water.
<hat limits the con4ersion per pass o4er a Co!based catal"st is actuall" the abilit" of heat
*'
remo4al in the reactor. Se4eral different reactor solutions ha4e been de4eloped for T during
the "ears. The" will be presented in the ne0t section.
".2 6ifferent reactor types and reaction conditions
$n a T!plant, the production of s"ngas, if produced from natural gas, accounts for 1> ! ;> C
of the total capital and running costs. 9conom" of scale is 4er" important for the s"ngas
production, and therefore the downstream T!reactors must also ha4e a high capacit".
Therefore, the ease of up!scaling is 4er" important when making the selection of t"pe of T!
reactor. =&r", %>>%?
The T!s"nthesis takes place in a three!phase s"stem. The gas phase contains the reactants
and water 4apour and gaseous h"drocarbon products. The higher h"drocarbons compose the
li7uid phase, and the catal"st is the solid phase. $n some cases, often high!temperature T
+HTT,, all products are 4aporised under reaction conditions and hence there are onl" two
phases present. The amounts of s"ngas and product molecules that must be transferred
between the phases are 7uite large in T compared to the h"drogen molecules transferred in
the h"drotreatment of petroleum oils in toda"Es refineries. Therefore, in T s"nthesis there is
a demand for a high interfacial mass transfer. =Sie and Drishna, #888?
A nearl" isothermal reactor, i.e. a reactor in which the temperature control is good, is possible
to operate at a slightl" higher a4erage temperature than a non!isothermal reactor in which
there are risk of hot spots and hence run!awa". A slightl" higher temperature gi4es a slightl"
higher reaction rate, and therefore a higher producti4it" =SchulF, #888?. Hence, a reactor with
a good temperature control can ha4e a higher producti4it" per gram catal"st, or put in another
wa", can host a more acti4e catal"st.
2an" of the earl" T!reactors +before and during <<%, were onl" possible for low
capacities +#* bbl@da", due to inefficient heat remo4al and re7uired 4er" large cooling
surfaces. Low pressure was used +e4en down to atmospheric, which re7uired 4er" low gas
flows in order to reach high con4ersions. The lower gas flows ga4e poor heat transfer.
=Drishna and Sie, %>>>?
5ther earl" T!reactors had e0ternal cooling, i.e. the hot outlet gas or li7uids were cooled and
then rec"cled into the reactor. This solution re7uires 4er" large rec"cle streams in order to
remo4e the generated heat in the reactor, which in turn results in large pressure drops and
high energ" consumption for the circulation. =Drishna and Sie, %>>>?
High heat e0change rates are toda" accomplished b" forcing the s"ngas at high linear
4elocities through long narrow tubes packed with catal"st particles +tubular fi0ed!bed reactor,
T:(, to achie4e turbulent flow and each tube cooled b" boiling water, or b" using a slurr"
phase reactor +S/(, in which the catal"st is dispersed in the li7uid products and in which 4er"
high heat transfer rates are achie4ed. or HTT in which ^ R >.;#, fluidised catal"tic bed
reactors are used which ha4e 4er" good heat transfer and temperature e7ualisation
characteristics. &ue to the risk of hea4" products depositing on the catal"st pellets, fluidised
beds cannot be used when high wa0es are produced. =&r", %>>%_ Sie and Drishna, #888?
There are currentl" two T operating modes) =&r", %>>%_ <ender, #881?
*.
#. .ig"-temperat!re B# (.#B#)_ '>> O '*> MC, e!based catal"sts, production of
gasoline and linear low molecular mass olefines. Significant amounts of o0"genates
are also produced. &iesel ma" be produced b" oligomeriFation of the olefins.
%. ,o7-temperat!re B# (,#B#)_ %>> O %.> MC, either e! or Co!based catal"sts,
production of high amounts of paraffins and linear products and the selecti4it" to high
molecular mass linear wa0es can be 4er" high. The primar" diesel cut and the
h"drocracking of the wa0es "ield e0ellent diesel fuels. The primar" gasoline cut needs
further treatment to obtain a high octane number.
The reactor operating conditions also depend on the reactor t"pe. The pressure is between #
and .> atm +H . 2/a,. The SS is between #>> O #>>> h
!#
. =<illiams, %>>%?
".2.1 'TFT
The Sasol T plants e0isting in #8-% all had circ!lating fl!idi&ed beds (CB$s) for their HTT
process, see igure ## +left picture,. The C: reactors are also called the Snt"ol reactors. $n
C:s there are two phases of fluidiFed catal"st O dense and lean. The catal"st particles mo4e
down the standpipe in dense phase while the" are transported up the reaction!Fone in lean
phase. $n order to a4oid that the feedgas goes up the standpipe, the differential pressure o4er
the standpipe must alwa"s e0ceed that of the reaction Fone. &uring operation at high
temperature, carbon is deposited on the catal"st particles +the e!based, which lowers the bulk
densit" of the catal"st and hence the differential pressure o4er the standpipe. Therefore it is
not possible to raise the catal"st loading in the reaction section in order to compensate for the
normal catal"st deacti4ation with time on stream. =&r", %>>%?
Sasol performed pilot scale e0periments in fixed fl!idi&ed beds +BB$s, see igure ## +right
picture,, in the #8;>s. The" disco4ered that the : reactor had a higher performance than
their commercial C: unit. #8-8 a commercial : unit was brought on!line successfull".
:etween #88* and #888, the #1 second generation C: reactors at Secunda were replaced b"
eight : reactors called Sasol Ad(anced Snt"ol (SAS). =&r", %>>%? The fluid bed T
s"nthesis at Sasol is mainl" used for production of olefins +C
%
!C
;
,. =SchulF, #888?
or the fluidised HTT reactors, the catal"st pellets can be small +R #>> um,, and hence
internal diffusion limitations are e0pected to be absent e4en though the catal"st pores are
filled with wa0. =&r", #881?
The ad4antages of :s o4er C:s are) =&r", %>>%_ <ender, #881??
! .> C lower construction costs. or the same capacit", the : reactor is much
smaller.
! Lower pressure!drop.
! &ue to the wider reaction section, more cooling coils can be installed, which increases
the capacit" of the : reactor due to the more isothermal beha4iour
! At an" moment, all of the charged catal"st participates in the reaction.
! The lowering of bulk densit" due to fouling is of less significance in the : which
implies that less on!line catal"st remo4al and replacement with fresh catal"st is
re7uired to maintain high con4ersions, which lowers the o4erall catal"st consumption.
! The : reactor re7uires less maintenance due to the lower linear 4elocities. The C:
reactors are comple0 and subGected to considerable erosion at the bends in the reactor.
**
igure ##. HTT reactors =<ender, #881?.
".2.2 2TFT
$n top!fed m!ltit!b!lar fixed-bed reactors +T:(, see igure #% +left picture,,, the wa0es
trickle down and out of the catal"st bed. $n sl!rr p"ase reactors +SPRs, see igure #% +right
picture,, the wa0es accumulate inside the reactors and so the net wa0 produced needs to be
continuousl" remo4ed from the reactor. =&r", %>>%?
$n the T plant in Sasolburg that was started #8**, fi4e multitubular AR+* reactors were
installed for wa0 production. The" are still in operation. 9ach reactor consists of %>*> tubes
+* cm i. d., #% m long,. The" operate at %.; 2/a and %'> C. The production capacit" of each
reactor is %# >>> t@"ear. An additional reactor was installed in #8-; operating at ..* 2/a.
=&r", %>>%?
$n the Shell plant at :intuli there are four large multitubular reactors with a capacit" of #%*
>>> t@"ear each. There are ca. #> >>> tubes in each reactor. $n this plant, Co!based catal"sts
are used. These are much more acti4e than the e!based catal"sts used in the Sasol plants.
This implies that the tube diameters of the Shell reactors are narrower in order to be able to
handle the higher rate of heat release. =&r", %>>%?
&ue to pressure drop constraints in packed beds, the catal"st pellet diameter used in the T:(
is often larger than # mm. <hen the catal"st pellet diameter is larger than >.* mm,
intraparticle diffusion can be rate limiting +H
%
@C5 H %, %# bar, %>> O %.> C, pore diameter H
#> O 1* nm, =Sie and Drishna, #888?. This problem can howe4er be o4ercome b" using egg!
shell impregnation, in which onl" the outmost la"er of the pellets contains the acti4e metal. $n
this wa", the diffusion distance for the reactants to the acti4e sites is not go4erned b" the siFe
of the pellets. Since Co is e0pensi4e, it is important that all of the impregnated metal is used
in the reaction. =9spinoFa et al., #888?
*1
$n the #8*>s, sl!rr p"ase reactors (SPRs) as alternati4e to fi0ed bed reactors were first
encountered. 3ot until #88' the first commercial unit based on a S/( for wa0 production was
brought on!line b" Sasol. $ts capacit" is ca. #>> >>> t@"ear, which e7uals the total capacit" of
the fi4e A(J9 reactors, and the catal"st is e!based. S/(s re7uire efficient filtration de4ices
in order to separate the produced wa0es from the catal"st slurr". =&r", %>>%?
The S/( consists of a shell in which cooling coils are present for steam generation. The
s"ngas is fed to the bottom of the reactor and it rises through the slurr". The slurr" consists of
li7uid products +mainl" wa0es, with catal"st particles suspended in it. The reactant gases
diffuse through the li7uid phase to the catal"st particles. The hea4" products form part of the
slurr" and the lighter gaseous products and water diffuse through the li7uid. The gaseous
products and unreacted reactants pass through the slurr" to the freeboard abo4e the slurr" bed
and then to the gas outlet. =9spinoFa et al., #888?
$n the S/(, the catal"st pellets can be smaller than in the fi0ed bed, since the" are dispersed in
the li7uid phase and hence there are essentiall" no pressure drop constraints. The smaller
catal"st pellets pre4ent internal diffusion limitations on the o4erall rate. Howe4er, since the
pellets are suspended in the li7uid wa0, diffusion of the reactants from the gas phase through
the wa0 to the solid catal"st ma" be rate limiting. =&r", #881?
5ne ad4antage with the multitubular fi0ed!bed reactor is that there are no problems in
separating the wa0 product from the catal"st. =9spinoFa et al., #888? The ad4antages of slurr"
phase reactors o4er multitubular reactors are) =&r", %>>%_ <ender, #881_ 9spinoFa et al.,
#888?
! The cost of the reactor is onl" %* C of the cost of a multitubular reactor.
! The pressure drop o4er the reactor is four times lower which results in lower gas
compression costs.
! Lower comple0it", easier scale!up.
! A lower catal"st loading is possible, which results in a four times lower catal"st
consumption per tonne of product.
! The slurr" bed can operate at higher temperatures and hence results in higher a4erage
con4ersions due to the more isothermal beha4iour. This makes it easier to control the
product selecti4ities. The S/( can handle catal"sts with higher acti4ities than the
multitubular fi0ed!bed reactor can.
! 5n!line remo4al@addition of catal"st allows longer reactor runs due to fre7uent
catal"st renewal, which is necessar" in order to obtain a stead" selecti4it" profile for a
single reactor.
The periodic replacement of the catal"st in the multitubular reactor is cumbersome and
maintenance and labour intensi4e, and it causes disturbances in the operation of the plant. The
cooling s"stem, with cooling onl" on the shell!side of the tubes, gi4es rise to a0ial and radial
temperature profiles in the tubes. or ma0imum reaction rates, a ma0imum a4erage
temperature is desired. This temperature is howe4er limited b" the ma0imum allowable
temperature peak. This ma0imum temperature is set in order to a4oid catal"st sintering and
fouling. The product selecti4ities are temperature dependent, wh" it is desirable to be able to
4ar" the temperature. $n tubular fi0ed bed reactors, the temperature control is howe4er
se4erel" complicated b" the need to a4oid e0ceeding the ma0imum peak temperature.
=9spinoFa et al., #888?
*;
5ne disad4antage with a fluidiFed bed +such as in the S/(, is that if a poison, like H
%
S,
should enter the reactor, the whole amount of catal"st will be poisoned. $n the case of a fi0ed
bed, onl" the la"ers of the catal"st closest to the entrance will be poisoned. =&r", %>>%? To
achie4e this, the remo4al of sulphur for slurr" bed application must be 4er" effecti4e. This is
one of the critical aspects for successful commercialisation of S/(s operating with a Co
catal"st. =9spinoFa et al., #888?
igure #%. LTT reactors =<ender, #881?.
The slurr" phase LTT process is b" man" authors regarded as the most efficient process for
T clean diesel production toda". This T technolog" is also under de4elopment b" Statoil
+3orwa", for use to con4ert associated gas at offshore oil fields into a h"drocarbon li7uid.
=SchulF, #888? Shell howe4er claims that their T:( works as well as a S/( =/resentation at
Jasuel>* in :rugge, :elgium?. i0ed!bed reactors are also more commonl" used for
laborator"!scale e0periments, due to faster and simpler operation for screening of different
catal"sts.
". Scope of the /or.
The scope of this report is to describe how to model a fi0ed!bed +also called trickle!bed, for
the LTT!s"nthesis. The laborator"!scale of these reactor t"pes are used at 3T3A and b"
Statoil for screening of catal"sts. The kinetic information from these reactors can also be used
to model slurr" phase reactors. $n the kinetic e0periments, the pellet siFe of the catal"st in the
fi0ed!bed is between *' and 8> `m, in order to a4oid mass transfer limitations on the
measured rate. The tube diameter is #> mm, and the bed height appro0imatel" *> mm. $n
larger!scale fi0ed!beds, this pellet siFe would result in high pressure drop.
or comparison, the modelling of a slurr" phase T!reactor will be discussed briefl".
2odelling of reactors is made in order to be able to scale!up a process from laborator" results,
to e0trapolate the performance of a reactor at conditions outside the conditions on which the
*-
model is built, to predict hot!spots and run!awa", to predict optimum operating conditions
respecting temperature control, rec"cling ratio etc.
"." )odeling of the tric.le-bed FT-reactor
As mentioned abo4e, high heat e0change rates in the T:( is accomplished b" forcing the
s"ngas at high linear 4elocities through long narrow tubes to achie4e turbulent flow. 9ach
tube is cooled b" boiling water. The wish is to obtain an isothermal reactor, and hence the
reaction heat should be remo4ed b" radial transport +mainl" conduction,. Howe4er, the heat
conducti4it" is rather poor, especiall" at the tube wall, compared to fluidised beds and this
often results in radial temperature gradients, see igure #'. These gradients affect the
selecti4it" and ma" also lead to increased deacti4ation of the catal"st. The gradients are
highest near the entrance of the reactor since the reaction rates are highest there, and since the
li7uid phase is rather small. urther down in the reactor, the reaction rates are slower due to
depletion of reactants and more and more li7uid wa0 is present, which reduces the radial
temperature gradients. =Drishna and Sie, %>>>_ Sie and Drishna, #888?
igure #'. Concentration and temperature profiles in a reactor of the a2itteldruck E ischerOTropsch process =Sie
and Drishna, #888?.
The radial heat transport is enhanced b" larger catal"st pellets and higher gas 4elocities. There
must be a compromise between the heat transport@pressure drop and the effecti4eness, Z, of
the catal"st. The larger the catal"st particle, the larger is the Thiele modulus, which can be
thought of as Ka reaction rateL @ Ka diffusion rateL. A large Thiele modulus will result in a low
effecti4eness factor, as shown in igure #., which means that the usage of the catal"st particle
is low, i.e. onl" the outer part of the catal"st particle is taking part in the reaction.
*8
ig. #.. Catal"st effecti4eness Z as a function of the Thiele modulus b for 4arious cobalt and iron catal"sts.
H"drogen@ carbon mono0ide molar ratio %, P H %.# 2/a, # H .;'O*#' D =Sie and Drishna, #888?.
:oth the effecti4e radial thermal conducti4it", 9
eff
, and the wall heat transfer coefficient, 5
7
,
increase with increased (e"nolds number +Re
p
, +see igure #*,. =Sie and Drishna, #888?
igure #*. 9ffecti4e radial thermal conducti4it" ceff and wall heat transfer coefficient ^w as functions of the
particle (e"nolds number Rep. dp is the particle diameter and cJ is the thermal conducti4it" of the gas =Sie and
Drishna, #888?.
The problem with the poorest heat e0change rate in the inlet of the reactor, where the need for
heat remo4al is largest, can partl" be sol4ed b" adding li7uid in order to make sure that the
complete tube operates in the trickle!flow mode. =Sie and Drishna, #888? Another wa" to
reduce the gradients at the inlet of the reactor is to lower the con4ersion per pass +to %> O '>
C, and rec"cle the gas outlet. $n this wa", linear gas 4elocities are increased. The A(J9
process mentioned abo4e, combines a low con4ersion per pass and rec"cling with higher
reaction temperature and pressure. Compared to the classical once!trough T:(, the A(J9
concept results in an increase in capacit" b" a factor %*, a reduction in cooling area b" a factor
#%, and a reduction of the amount of catal"st and steel b" a factor of ;. =Drishna and Sie,
%>>>?
As mentioned abo4e, in T s"nthesis there is a demand for a high interfacial mass transfer,
since capillar" condensation of the formed wa0es makes the pores filled with li7uid. =Sie and
Drishna, #888? Therefore, the modelling of the T fi0ed!bed reactor in4ol4es description of
the interactions between comple0 chemical kinetics and transport phenomena.
According to <ang et al. =%>>'?, onl" a few models for T fi0ed!beds are reported in
literature. Atwood and :ennett =#8;8?, a reference in <ang et al. =%>>'?, made a one!
dimensional, heterogeneous plug!flow model. :ub et al. =#8->? de4eloped a two!dimensional,
pseudo!homogeneous plug!flow model. Jess et al. =#888? made a pseudo!homogeneous two!
dimensional model. $n case of the pseudo!homogeneous models, the intraparticle diffusion
1>
was not considered. $n the heterogeneous model b" Atwood and :ennett =#8;8?, the
intraparticle gradients were encountered for b" means of an effecti4eness factor. According to
<ang et al. =%>>'?, the use of an effecti4eness factor instead of sol4ing the catal"st
pellet@particle e7uations, makes a fundamental anal"sis of intraparticle diffusion!reaction
beha4iour impossible.
".".1 7ang et al. 8299:
<ang et al. =%>>'? de4eloped a one!dimensional heterogeneous model, with assumed plug!
flow in the gas!phase. 5nl" two phases were accounted for ! gas and solid@li7uid. The catal"st
was e!based, and hence acti4e for the <JS!reaction. 2ass and energ" balances were set up
for the two phases. The concentration of the components in the solid phase was e0pressed as
concentration in li7uid, and hence the pores were assumed to be filled with li7uid wa0. Li7uid
surrounding the pellets was not taken into account, and it was assumed that the mass transfer
restriction through the gas film surrounding the pellets was negligible. Hence, gas!li7uid
e7uilibrium was assumed at the interface between the bulk gas and the li7uid filled pellet. A
modified Soa4e!(edlich!Dwong +S(D, 95S was used to correlate the e7uilibrium between
the components in the gas bulk and the same components sol4ed in the wa0 in the catal"st
pores. The description of this 95S is gi4en in <ang et al. =#888?. or the heat transfer,
resistance between solid and gas was assumed to be located in a film around the pellets and
was e0pressed as a boundar" condition for the pellet e7uation.
<ang et al. =%>>'? chose to include the reaction rate terms in the 2: +and 9:, for the gas
phase, howe4er integrated o4er the 4olume of the pellet and then di4ided b" the pellet 4olume
in order to obtain 4olume!a4erage 4alues. This was done due to the simplification that there
e0ist no mass transfer resistance between li7uid and gas phase, an assumption which means
that the e7uilibrium 4alues are attained immediatel". rom :ischoff and roment =#88>? we
ha4e learnt that when interfacial gradients are accounted for, these terms would ha4e been
replaced b" the flu0 between the phases, i.e. b" k
g
a
4
+C
Ag
! C
S
AS
, +and b" h
f
a
4
+T
s
s
O T
g
,,. <ang
et al. =%>>'? did not estimate this -
g
, but instead simpl" Gust assumed e7uilibrium
concentrations in li7uid and gas.
or the gas phase mass transfer, con4ection and flu0 from pellets was encountered. or the
energ" balance in the gas phase, con4ecti4e heat transport, heat flu0 from pellets and
conduction through tube wall were included.
or the solid phase mass transfer, diffusion and reaction was encountered. or the energ"
balance, conduction inside pellets +dispersion term, and reaction were included.
The one!dimensional model cannot predict an" radial gradients, which according to the
discussion abo4e might be of significant magnitude in the fi0ed!bed.
1#
:elow follows the model e7uations used b" <ang et al. =%>>'? with the same s"mbols as
used b" roment and :ischoff =#88>?)
*ul. gas phase:
2:)
d r
R d&
C ! d
p
R
2R
3
3 i3 s
p
i s %
>
#
'
, # +
' , +


+i H #, nb of total components in4ol4ed,
9:)
, + . , + , # +
'
%
>
#
'
g 7
t
R
2R
3
3 3 s
p
g
p g s
# #
d
8
d r .
R d&
d#
c !
p
+


Mnitial conditions%
& H >) c
i
H c
i?E
P H P
in
#
g
H #
in
2i+uid;solid phase:
2:)


2R
3
3 i3 s
is
ei
r
d
dC
d
d
'
#
%
%
, +
#


+i H #, nb of ke" components in4ol4ed,
9:)


2R
3
3 3 s
s
e
r .
d
d#
d
d
#
%
%
, + , +
#

$o!ndar conditions%
N H >)
>
d
dC
is
>
d
d#
s
N H R
p
)
i
O
i
,
i
i
x

, +
g s
e
f
s
# #
"
d
d#


where
i
and x
i
are the mole fractions of component i in gas and li7uid, respecti4el", and P
,
i
and P
O
i
are the fugacit" coefficients in li7uid and gas phase, respecti4el". This boundar"
condition means that component i is in its e7uilibrium concentrations in the wa0 and the gas.
<ang et al. =%>>'? chose to describe the pressure drop in the bed with the classical 9rgun
e7uation)
p
s g
p
d
!
f
d&
dP
%



1%
The /eng!(obinson 95S +e7uation of state, was used to calculate d
g
and hence u
s
. The ideal
gas!law can onl" be applied at low reactor pressures.
The rather comple0 kinetic rate e0pressions used in the model b" <ang et al. =%>>'? can be
found in <ang et al. =%>>#?. (ate e0pressions were set up for the products, which included
CH
.
, paraffins and olefins of different carbon numbers, and C5
%
+<JS!reaction,. All reaction
rate e0pressions included the chain growth parameter, 5, defined abo4e. The chain growth
parameter used b" <ang et al. =%>>#, %>>'? was slightl" modified compared to the original
5
ASB
as defined b" Anderson, SchulF and lor". The modified 5 is not constant, but changes
with carbon number, and also accounts for the so!called olefin readsorption factor +@
n
,. $f the
catal"st has a high @
n
, the probabilit" that a desorbed h"drocarbon chain will readsorb on the
catal"st surface and continue to grow is high. This phenomenon gi4es higher wa0es and hence
higher 5. The definition of @
n
can be found in <ang et al. =%>>#?. This factor can be
calculated knowing the rate constant for the olefin readsorption reaction, -
-Q
, and the partial
pressures of the h"drocarbon products +paraffins and olefins,. The @
n
howe4er approaches >
rather 7uickl" with increasing carbon number. /racticall", ethene, which is the shortest olefin,
has the highest readsorption abilit", and it is often an order of magnitude higher than the
readsorption abilit" of propene.
1 * #
#
%
- P - P -
P -
. CO
CO
ASB
+ +

, # +
1 * #
#
%
n . CO
CO
n
- P - P -
P -

+ +

The parameters -
R
and -
Q
are rate constants for paraffin formation and olefin desorption,
respecti4el". The description of k
#
was not gi4en in <ang et al =%>>#?, but is probabl" the rate
constant for consumption of C5.
<ang et al. =%>>#? characteriFed the wa0 b" 32( and found it to be appro0imatel" n!C
%-
H
*-
in all e0perimental runs, wh" this was also put into the model. &ispersion coefficients for H
%
,
C5 and C5
%
in this wa0, '
i
, were taken from literature and could be fitted to e0pressions as a
function of temperature. &ispersion coefficients for the other components were appro0imated
b" means of diffusi4it" correlations in infinitel" dilute solutions)
1 . >
, @ +
i CO CO i
O O ' '
where O is the mole 4olume. The effecti4e diffusi4it" of component i was calculated b"
correcting the molecular diffusi4ities for the porosit" and tortuosit" of the catal"st pellets)

i s
ei
'
'
Clearl", this model should onl" work satisfactoril" if the composition of the wa0 is close to n!
C
%-
H
*-
, and therefore should not be so accurate in predicting reactor operation under
conditions far awa" from those upon which the model is based.
1'
$n order to make the s"stem of e7uations con4erge, the surface li7uid concentrations in
e7uilibrium with the bulk gas concentrations were set as guess 4alues for the unknown
concentrations inside the catal"st pellets.
:elow, some of the results from the simulation are discussed. igure #1 shows the
concentration profiles of C5, H
%
, C5
%
and H
%
5 inside of the wa0!filled catal"st pores +> is at
center and # at pellet surface,.
igure #1. Concentration profiles of ke"
components in catal"st pellet =<ang et al., %>>#?.
The further the T!reaction proceeds, the
more water is produced, which is seen in
the increasing H
%
5 concentration towards
the middle of the particle. The more water,
the more C5 is consumed in the <JS reaction, at the same time as it is consumed in the T!
reaction, which e0plains the rapid decrease in C5 concentration between # and >.-R
p
. This
means that the reaction Fone is actuall" onl" in >.%R
p
, which is in good agreement with a
4alue of >.%.R
p
reported in literature for catal"st pellets of ' mm in diameter. The rapid
decrease in C5 concentration in the outmost particle shell is also due to the far more se4ere
diffusion limitations on C5 compared to H
%
. Howe4er, since the concentrations of C5
%
and H
%
are rather high at this point +>.-*R
p
,, the re4ersed <JS!reaction will be triggered, and hence
pro4ide some C5 for the T!reaction e4en far inside of the pellet.
Since both the reaction order with reference to P
.2
and P
CO
are positi4e, lower partial
pressures of these reactants will result in a lower reaction rate inside of the catal"st pellet
compared to at the surface. urthermore, because of the higher H
%
@C5 ratio inside of the
particle, see igure #;, the selecti4it" to
wa0es will be lower here.
igure #;. Sariation of H%@C5 ratio along the
pellet dimension =<ang et al., %>>#?.
An impro4ed diffusion of the C5 would
increase the reaction rate as well as the
selecti4it" to wa0es.
1.
The authors calculated the effecti4eness factor for se4eral different cases. The effecti4eness
factor was calculated b" di4iding the true pellet!4olume!a4eraged consumption rate of C5
with the pellet!4olume!a4eraged consumption rate of C5 if the C5 concentration would be as
high as at the surface throughout the whole pellet. igure #- and #8 show the effecti4eness
factor as a function of temperature and pellet diameter, respecti4el".
igure #-. Sariation of effecti4eness factor with operation temperatures at different pressures =<ang et al.,
%>>#?.
igure #8. Sariation of effecti4eness factor with
pellet siFe =<ang et al., %>>#?.
igure %> shows the HC distribution as a
function of temperature. igures %# and %%
show the selecti4it" to CH
.
and C5
%
, and
C
%I
and C
*I
, respecti4el", as a function of
pellet siFe.


1*
igure %>. /roduct distribution of h"drocarbons at different temperatures =<ang et al., %>>#?.
igure %#. Selecti4it" 4ariations of C5% and CH. with pellet radius =<ang et al., %>>#?.
11
igure %%. Selecti4it" 4ariations of C%
I
and C*
I
products with pellet radius =<ang et al., %>>#?.
The conclusion is that small catal"st particles fa4our the formation of the desired products, at
the same time as the reaction rate is higher. Howe4er, in industrial applications the pellet siFe
should not be below % mm, in order to a4oid a large pressure drop. The solution is to use so!
called egg!shell impregnation, in which the acti4e metal, in this case e, is situated onl" in the
outmost shell of the pellet. The siFe of the inert core will ha4e to be optimised in order to gain
a high selecti4it" to wa0, at the same time keeping the total acti4it" not so much under the
acti4it" of the uniforml" impregnated pellet. igure %' shows the effect of the inert core
radius on the selecti4it" to C
*I
for a catal"st pellet with a radius of #.* mm.
igure %'. 9ffect of inert core radius on C*
I
selecti4it" =<ang et al., %>>#?.
igure %. shows that with an inert core radius of up to ->C of R
p
, the reaction rate is actuall"
not lower compared to the uniform pellet. This indicates that the egg!shell impregnation can
impro4e the selecti4it" to wa0es, a4oiding a significant increase of the reactor 4olume to
obtain the same producti4it" as compared to uniform catal"st pellets.
1;
igure %.. 9ffect of inert core radius on effecti4eness factor =<ang et al., %>>#?.
The model could satisfactoril" predict the performance of different!scale demonstration
processes. The main conclusions from the modelling of <ang et al. =%>>'? were that a large
tube diameter was unfa4ourable for the total "ield of higher HCs, and that the inner diameter
should not e0ceed 1> mm. This is a direct effect of the increased bed temperature as a larger
tube diameter implies a poorer heat remo4al +see igure %*,.
(ec"cling resulted in thermal stabilit" of the reactor, a4oiding the hot spot in the beginning of
the reactor, and also enhanced the "ield of desired products. <ith a high rec"cle ratio the bed
was almost isothermal. The rec"cle operation means that the superficial gas 4elocit" is
increased, but the space 4elocit" of the fresh feed is unchanged. The lower o4erall
temperature decreases the C5!con4ersion, but the total s"ngas con4ersion is actuall"
increased due to the increase in H
%
@C5 usage ratio +i.e. decrease in <JS!reaction rate,.
Howe4er, from a practical point of 4iew, the power consumption should alwa"s be taken into
consideration when choosing the rec"cle ratio.
An increase in cooling temperature can increase the s"ngas con4ersion, but lowers the "ield of
higher HCs. Hence, a lower cooling temperature is recommended. An increase in pressure
fa4ours both con4ersion and "ield to desired products.
1-
ig. %*. 9ffect of process parameters on the a0ial temperature profile in the fi0ed!bed reactor. :ase conditions)
dtH'% mm, ,H;.> m, JHSSH*>> h
e#
, (c"cH'.>, #7H#g,>H*%' D, PinH%* bar. resh s"ngas composition +C,) C5)
'>.*8_ H%) *;.;*_ C5%) ;.>_ 3%) ..>-_ CH.) >.*- =<ang et al., %>>'?.
18
".".2 <ess et al. 81===:
Jess et al. =#888? de4eloped a two!dimensional pseudo!homogeneous model for T!s"nthesis
o4er a e!catal"st.
$n the model b" Jess et al. =#888?, onl" mass transfer b" con4ection +onl" in a0ial direction,
and reaction was considered in the 2:. $n the 9:, radial conduction, ad4ecti4e heat transport
in a0ial direction, and reaction was encountered. 3ote that Jess et al. considered the radial
temperature gradients and radial heat transfer, while neglecting radial concentration gradients.
:elow follows the model e7uations used b" Jess et al. =#888? with the same s"mbols as used
b" roment and :ischoff =#88>?)
2:)
3
2R
3
i3 s
i s
r
d&
C ! d


#
, # +
, +

+i H #, nb of total components in4ol4ed,
9:)

,
_

+
2R
3
3 3 s p g s er
r .
d&
d#
c !
dr
d#
r dr
# d
#
%
%
, + , # +
#

Mnitial conditions%
& H >, > r R
i
) c
i
H c
i?E
P H P
in
#
g
H #
in
$o!ndar conditions%
r H >, all &)
>
dr
d#
All &, all r)
>
dr
dC

r H R
i
, all &)
, +
, ,
,
i 7 i R
er
i 7
# #
dr
d#

r H R
o
, all &)
, +
,
,
cool o 7
7
o 7
# #
dr
d#

where R
i
is the tube inner radius, and R
o
the tube outer radius, 5
7?i
and 5
7?o
=<@m
%
,D? the inner
and outer wall heat transfer coefficient, respecti4el". 9
7
is the heat conducti4it" of the wall
=<@m,D?.
Jess et al. =#888? e0pressed the consumption rate of C5 as the sum of three rates, as shown in
igure %1.
;>
ig. %1. Dinetic data of the three main reactions during T!s"nthesis on the iron!catal"st +A(J9!Lurgi!
(uhrchemie, =Jess et al., #888?.
The rate e0pressions abo4e are not trul" kinetic. (ather, the" are onl" 4alid in the case of
mass transfer limitations inside the catal"st pellets. Hence, the rate e0pressions describe the
rate at which C5 is consumed if diffusion of reactants through the catal"st pores is the rate
limiting step, and are, for pellet diameters P %.* mm, onl" 4alid at temperatures abo4e #;> MC.
90ternal diffusion limitations could, according to Jess et al. =#888?, be neglected up to .>> MC,
but this is Gust a h"pothetic 4alue since the catal"st would rapidl" deacti4ate at temperatures
abo4e %1> MC. These rate e0pressions corresponds to an effecti4eness factor of >.% at %*> MC,
and the acti4ation energies in the e0pressions are appro0imatel" half of the true acti4ation
energies, which is t"pical in the range of pore diffusion limitations =ogler, #888?. $n their
pseudo!homogeneous model, :ub et al. =#8->? also used Knon!kineticL rate data based on
laborator" e0periments with rather large particle diameters. :" using the same particle
diameter in the laborator" e0periments as used in the pilot!scale reactor to be modeled, the
possible intraparticle mass transfer limitation on the o4erall rate was not necessar" to be
considered.
Anlike <ang et al. =%>>'?, the rate e0pressions of Jess et al. =#888? do not include the product
selecti4it". $t was not e0plained how the selecti4it" was calculated. Since the model is
pseudo!homogeneous, the diffusion of reactants through the wa0es inside the catal"st pores is
;#
baked in in the rate e0pressions, and in this wa" intraparticle gradients are somehow
accounted for, but this model should not be so accurate in predicting reactor operation under
conditions far awa" from those upon which the model is based. or instance, it can onl" be
applied if the catal"st pellets ha4e e0actl" the same dimensions as the pellets used in the
KkineticL measurements.
The obGect was to make a model that could predict the performance with a nitrogen!rich +*>
4ol!C, s"ngas. :" producing the s"ngas b" partial o0idation of natural gas with air, instead of
with pure o0"gen, the need for the capital intense air separation unit is eliminated. &ue to the
high inert concentration, this concept can +or must, in order not to build up inerts or to be
forced to bleed off a large stream, work without rec"cling of the tail gas and hence does not
need a rec"cle compressor.
The results of the modeling showed that nitrogen pla"s an important role in remo4ing the heat
released b" the T!reaction. This led to an optimum tube diameter of ;> mm for a stable and
safe operation, compared to .* mm for the nitrogen!free s"ngas. The production rate per tube
was about three times higher than in the case with a nitrogen!free s"ngas, due to the possible
high once!through con4ersion. This would decrease the in4estment cost of multitubular fi0ed!
bed reactors. Howe4er, since nitrogen is diluting the s"stem, a higher total pressure will be
needed in order to preser4e the partial reactant pressures, which is important for the kinetics.
The critical temperatures were defined as the temperatures not to be e0ceeded in order to
a4oid a temperature runawa". The estimation of these temperatures can be found in Jess et al.
=#888?. igure %; shows how the tube diameter influences the critical temperatures in the
nitrogen!free and rich cases.
ig. %;. $nfluence of the diameter of the single tubes of a T!reactor on the critical ma0imum temperature for
nitrogen!rich and nitrogen!free s"ngas +4olume rate of s"ngas according to Se H %* s, =Jess et al., #888?.
;%
".% )odeling of the slurry phase reactor
According to Drishna and Sie =%>>>?, the best reactor choice is the slurr" phase reactor +or as
the" call it Kthe bubble column slurr" reactorL, for large!scale plants with capacities around
.>,>>> bpd. $t was calculated that for such a large total production, #> T:(, #; slurr"
reactors operating in the "omogeneo!s regime or four slurr" reactors operating in the
"eterogeneo!s regime are needed. The limiting criterion was a ma0imum reactor weight of
8>> tons.
".%.1 6ifferent flo/ regimes and estimation of gas hold-up
The homogeneous regime is characteriFed b" a relati4el" low superficial gas 4elocit", !
s+
, and
b" small bubbles +d
b
H # O ; mm,. <hen the superficial gas 4elocit" reaches the 4alue 8
trans
,
coalescence of the small bubbles occurs and larger bubbles are produced. The regime with
larger bubbles +d
b
H %> O ;> mm, and higher !
s+
is called the heterogeneous regime. +igure
%- shows a schematic of the two regions., These larger bubbles ha4e a high rising speed
through the column +# O % m@s, in a plug!flow manner. Smaller bubbles are also present in the
heterogeneous regime, and since the" are entrained in the li7uid phase, the" are often assumed
to ha4e the same backmi0ing characteristics as the li7uid phase. <hen the catal"st particles
are small +R *> `m,, the" are well mi0ed with the li7uid, and the slurr" phase can be regarded
as a pseudo!homogeneous phase. This assumption, with no catal"st settling, works well at
larger reactor diameters +P >.* m, at high gas 4elocities +!
s+
P >.% m@s,.
ig. %-. Homogeneous and churnOturbulent regimes in a gasOli7uid bubble column. The figure on the right
shows that increasing the s"stem pressure dela"s the regime transition point =Drishna and Sie, %>>>?.
The 4olume fraction of catal"st in the li7uid, 0
cat?,
+0
s
in igure %8, strongl" affects the gas
hold up, 0
+
+0 in igure %8,.
;'
ig. %8. $nfluence of increased particles concentration on the total gas hold!up in columns of +a, >.# m and +b,
>.'- m diameter. Air was used as the gas phase in all e0periments. The li7uid phase was paraffin oil +densit",
4LH;8> kg m
e'
_ 4iscosit", =LH>.>>%8 /a s_ surface tension, >H>.>%- 3@m, to which solid particles in 4ar"ing
concentrations were added. The solid phase used consisted of porous silica particles +skeleton densit"H%#>> kg
m
e'
_ pore 4olumeH#.>* ml g
e#
_ particle siFe distribution, dp) #>CR%; `m_ *>CR'- `m_ 8>CR.; `m,. The solids
concentration s, is e0pressed as the 4olume fraction of solids in gas free slurr". The pore 4olume of the particles
+li7uid!filled during operation, is counted as a part of the solid phase =Drishna and Sie, %>>>?.
or modeling purposes it is important to distinguish between the gas hold up in large bubbles
and that in the small bubbles. This can be made b" instantaneousl" switching off the gas flow
to a slurr" column at time t H >. The gas hold!up will 7uickl" decrease due to the escape of
large bubbles. After this 7uick reduction in gas hold!up, the small bubbles escape and the gas
hold!up is further reduced, as illustrated in igure '>. rom this e0periment, the 4oidage of
gas in the dense phase +see igure '>,, 0
+?df
, the gas hold!up in the large bubbles, 0
+?b
, and the
superficial 4elocit" through the dense phase, !
s?df
, can be estimated. !
s?df
is often taken to be the
same as 8
trans
.
ig. '>. &"namic gas disengagement e0periments for air@paraffin oil and air@'1 4ol.C paraffin oil slurr" in the
>.'-!m!diameter column. The s"stem properties are as gi4en in the legend to igure %8 =Drishna and Sie, %>>>?.
;.
$t is clearl" seen that the higher the catal"st concentration in the slurr" phase, 0
cat?,
, the smaller
is the gas hold!up in the small bubbles. This is due to the increased coalescence of small
particles into bigger in the presence of solid particles. At 0
cat?,
P '> C, the population of small
bubbles is destro"ed. The 0
+?df
for a certain 0
cat?,
is essentiall" constant at !
s+
P >.# m@s for the
heterogeneous regime. urthermore, the 0
+?df
is essentiall" independent on reactor diameter
+'
#
,. This is 4er" useful for scale!up purposes, since a rather small reactor can be used to
determine the 0
+?df
under actual reaction conditions.
The 0
+?b
, howe4er, of course increases with increasing superficial 4elocit" through the large
bubble phase +!
s+
!!
s?df
,, but seems to be independent on 0
cat?,
in the range >.#1 R 0
cat?,
R >.'1.
Anlike the 0
+?df
, the 0
+?b
+0
b
in igure '#, is strongl" dependent of the reactor diameter. The 0
+?b
is reduced at large diameters due to the increase in rise 4elocit", which, in turn is due to the
reduced wall effects at large diameters. (elationships to find the large bubble rise 4elocit" for
different d
b
@'
#
ratios is found in Drishna end Sie =%>>>?.
ig. '#. $nfluence of column diameter on the hold!up of large bubbles b in +a, paraffin slurries and +b, Tellus oil
=Drishna and Sie, %>>>?.
These relationships must be modified to be 4alid at higher pressures, as shown in igure '%.
The modifications are presented in Drishna and Sie =%>>>?. :riefl", the e0pression for
calculation of the rise 4elocit" of large bubbles, !
b
, is corrected for the higher densit" of a gas
at higher pressure compared to atmospheric pressure. !
b
for s"ngas at '> bars, which is
common in T, is for instance onl" .'C of the !
b
for air at atmospheric pressure, which is
easil" understood since the gas densities are ; kg@m
'
and #.%8 kg@m
'
for the pressuriFed
s"ngas and the atmospheric air, respecti4el".
;*
ig. '%. The influence of ele4ated pressure on gas hold!up =Drishna and Sie, %>>>?.
<hen the !
b
is known, the 0
+?b
can be calculated through the steps below. 0
+?df
in a pressuriFed
s"stem can be estimated b")

,
_

,
_

, cat
df + ref +
+
df + df + ,
> , ,
.- . >
,
> , , ,
; . >
#


where 0
+?df?E
is 0
+?df
at >C catal"st concentration, 4
+?ref
is the densit" of gas at ambient
conditions, and 4
+
the densit" of the gas at pressuriFed conditions. The rise 4elocit" of the
small bubbles, in atmospheric or pressuriFed s"stems, can be calculated b")

,
_

+
, cat
small
small small
!
! !
,
> ,
> ,
- . >
#
where !
small?E
is the rise 4elocit" at >C catal"st concentration. This is not much affected b"
pressure. The superficial gas 4elocit" through the dense phase is estimated b")
df + small df s
! !
, ,

Since !
b
is known and defined as)
b +
df s s+
b
! !
!
,
,
, +

we can now calculate 0


+?b
. The total gas hold!up in a pressuriFed slurr" column can be
calculated according to)
, # +
, , , b + df + b + +
+
".%.2 )ass transfer
As mentioned abo4e, in slurr" reactors intraparticle diffusion limitations are often negligible,
due to the small pellets used. 2ass transfer from gas to li7uid can howe4er be rate limiting.
This is often the case with highl" acti4e catal"sts in high concentrations in slurr" reactors
operating in the heterogeneous regime.
;1
The large bubbles represent the maGor gas throughput. Con4entional calculation of mass
transfer rates predicts a 4er" low mass transfer between the gas and slurr" phase for large
bubbles. 90periments ha4e, howe4er, shown that the real mass transfer rates are * O #> times
higher than that e0pected from con4entional calculations based on hold!up and bubble siFe.
The e0planation to this is that during the characteristic time for mass transfer from the gas to
the li7uid, the bubbles e0change gas with other bubble siFe classes. $n conclusion, the large
bubbles represent the maGor gas throughput, but gas!li7uid mass transfer is mainl" determined
b" the interfacial area between the small bubbles and the slurr". $n most cases, the mass
transfer between gas and li7uid is not limiting the o4erall rate in T!s"nthesis. The 4olumetric
mass transfer +g@l, coefficient for the large bubbles, +-
,
a)
large
, can be calculated according to)
ref ,
,
b +
e l ,
'
'
a -
, ,
arg
* . >
, +

where '
,
is the diffusi4it" in the li7uid phase and '
,?ref
is a constant of % Y #>
!8
m
%
@s. The
diffusi4ities of the reactants C5 and H
%
at %.> MC are #;.% Y #>
!8
and .*.* Y #>
!8
m
%
@s,
respecti4el". or the small bubbles, the e0pression has been found to be the same as for the
large bubbles, but with a constant of #.> instead of >.*)
ref ,
,
df +
small ,
'
' a -
, ,
> . #
, +

3ote that the mass transfer coefficients -


,
a, are not the same as -
,
in :ischoff and roment
=#88>? in which this is a mass transfer coefficient from g@l interface to li7uid bulk. -
,
a used b"
2aretto and Drishna in this case is an o4erall mass transfer coefficient from gas bulk to li7uid
bulk, including the interfacial area.
".%.. )i$ing
The large bubbles concentrate in the middle of the column and carr" the li7uid upwards. At
the top of the column, the large bubbles escape from the slurr", and the li7uid is recirculated.
The larger the reactor diameter, the bigger is the li7uid 4elocit", O
,
, upwards in the middle
and downwards at the wall. Howe4er, b" normaliFing the li7uid 4elocities b" di4iding O
,
(r)
with the li7uid 4elocit" at the central line, O
,
(E), the radial distributions are similar for all
reactor diameters. Hence, the li7uid phase a0ial dispersion coefficient, '
a?,
, should be
proportional to O
,
(E). 5ne useful relationship is)
- @ #
'
% @ #
, + % . > , > +

,
_

,
# ,
g
8
g' O

where J
,
is the kinematic 4iscosit" of the li7uid phase. Drishna and Sie =%>>>? showed that
the backmi0ing of the slurr" phase is actuall" the same as for low 4iscosit" li7uids, such as
water, at the same superficial gas 4elocit". 2easurements of the '
a?,
ha4e led to the following
simple relationship)
# , , a
' O ' , > + '# . >
,

;;
9stimation of '
a?,
in a commercial scale bubble slurr" column with a reactor diameter of ; m
and a superficial gas 4elocit" of >.'* m@s "ields '
a?,
H #> m
%
@s. This high 4alue of the
dispersion coefficient suggests that the slurr" phase is well mi0ed, and hence can be modeled
as a CST(.
".%." 'eat transfer
The heat transfer is fa4ored b" high gas 4elocities and high 0
cat?,
+0
s
in igure '',? which are
t"pical for the heterogeneous regime.
ig. ''. 9stimate of the heat transfer coefficients to 4ertical cooling tubes =Drishna and Sie, %>>>?.
".%.% The model
The model, the results of which were briefl" described in Drishna and Sie =%>>>?, can be
found in 2aretto and Drishna =#888?. A schematic picture of the model is gi4en in igure '.,
in which 8 H !
s+
, and 8
df
H !
s?df
.
ig. '.. JeneraliFed two!phase model applied to a bubble column slurr" reactor operating in the churnOturbulent
regime =Drishna and Sie, %>>>?.
The slurr" was assumed to ha4e the properties of C
#1
H
'.
. The rate e0pression for the
consumption of s"ngas was taken from fates and Satterfield =#88#?)
;-
%
, # +
%
%
CO
CO .
. CO
bp
p ap
R
+

+
where a and b depend on the catal"st used, and are 4ar"ing with temperature. 3ote that
according to the T!reaction stoichiometr", R
.2
H %R
CO
. The selecti4it" was described with the
AS!distribution and onl" paraffins +C
n
H
%nI%
, were assumed, the molar fraction of which was
e0pressed as)
#
, # +


n
ASB ASB n
x
5
ASB
was set to >.8.
As mentioned abo4e, in the heterogeneous regime, which was modeled in this article, the
larger bubbles +%> O ;> mm, tra4el up the column at high 4elocities +# O % m@s, in a plug!flow
manner, resulting in a well!mi0ed slurr". The smaller bubbles are hence entrained in the
slurr".
A one!dimensional heterogeneous model was made of a slurr" reactor with a diameter of ; m
and height of '> m. The temperature was %.> MC and pressure '> bar. The large bubbles were
modeled as if being in plug!flow, with a superficial 4elocit" of !
s
!!
s?df
, and the slurr" phase
was assumed to be well!mi0ed. The temperature was assumed being constant in the reactor
and the catal"st particles, wh" no energ" balances are needed. The model assumes that
resistance between li7uid and catal"st surface is negligible, and also intraparticle gradients
were neglected. The catal"st pellets had a diameter of *> `m.
)* large bubbles:

,
_

, CO
CO
e l + CO
e l CO ,
e l + CO df s s+
C
.
C
a -
d&
C ! ! d
,
arg , ,
arg ,
arg , , ,
, +
, , ++

,
_

, .
.
e l + .
e l . ,
e l + . df s s+
C
.
C
a -
d&
C ! ! d
,
arg , ,
arg ,
arg , , ,
%
%
%
%
%
, +
, , ++
)* small bubbles:
G C
.
C
a - C C !
, CO
CO
small + CO
small CO , small + CO
i
+ CO df s

,
_


,
, ,
, , , , ,
, + , +
G C
.
C
a - C C !
, .
.
small + .
small . , small + .
i
+ . df s

,
_


,
, ,
, , , , ,
%
%
%
% % %
, + , +
where < is the cross sectional area of the reactor in =m
%
?, G is the dispersion height of the
reactor in =m?, and superscript i means inlet of reactor. $t was assumed that the concentration
of C5 in the inlet gas is e7ual in large and small bubbles and e7uals C
i
CO?+
. The same was
;8
assumed for H
%
. The superficial gas 4elocit" through small bubbles, !
s?df
, was assumed
constant in the whole reactor.
)* li+uid phase:
> , + , +
%
, ,
, ,
, ,
arg . ,
>
arg ,
+

,
_

,
_

. CO , , CO
o
s, , CO
CO
small + CO
small CO , , CO
CO
e l + CO
G
e l CO ,
R G C ! C
.
C
a - G d& C
.
C
a -
> % , + , +
% % %
%
%
% %
%
%
%
, ,
, ,
, ,
arg . ,
>
arg ,
+

,
_

,
_

. CO , , .
o
s, , .
.
small + .
small . , , .
.
e l + .
G
e l . ,
R G C ! C
.
C
a - G d& C
.
C
a -
where 0
,
is the slurr" hold!up, e7ual to #!0
+
, since in the case with a slurr" reactor the catal"st
particles are included in the slurr". The e7uations for the li7uid phase are t"pical e7uations for
a CST(, i.e. the concentrations in the slurr" are assumed constant in the reactor, and also the
concentrations in the small bubbles, which means that the 2: for the slurr" is an o4erall 2:
for that phase. Since the concentrations in the large bubbles are not constant, the mass transfer
between gas and li7uid must be integrated along the reactor in order to fit in in the o4erall 2:
for the slurr" phase. The Henr"Es constants were used as estimated solubilities for H
%
and C5
in the wa0, which was assumed to ha4e the characteristics of C
#1
H
'.
, and e7ual to %..;- and
%.81. at a temperature of %.> MC for C5 and H
%
, respecti4el".
$t was also assumed that the contraction of gas 4olume due to reaction onl" affects the large
bubbles. The contraction factor, T, for #>> C s"ngas con4ersion was estimated to O>.1.-,
assuming * C inert in the s"ngas feed. Hence, the superficial gas 4elocit" through large
bubbles decreases with con4ersion according to)
, # ,+ +
, ,
X ! ! ! !
df s
i
s+ df s s+
+
where X is the con4ersion of C5 +or H
%
,. The con4ersion of C5 e7uals the con4ersion of H
%
in
this case since the inlet molar H
%
@C5 ratio is the same as the usage ratio in the reactor, i.e. %.
The reason for wh" this relationship was set up is that the authors onl" considered mass
balances and distribution between gas and li7uid phase for the reactants +and not the
products,. Therefore, there is lack of information considering how man" moles of gas that
e0ist at different F!points in the reactor. The e0pression abo4e estimates this information.
The solution to the e7uations abo4e was found b" means of an iterati4e procedure. This kind
of solution is t"pical when differential e7uations are coupled with algebraic ones, and in this
case most probabl" means that the concentration in the well!mi0ed phases are guessed as start
4alues, and then integration of the gas!phase along the reactor will gi4e new guessed 4alues of
the concentrations in the well!mi0ed phases, closer to the true ones. The authors mentioned
that it was possible to describe the concentration of one reactant as a function of the other,
which simplified the solution of the e7uations.
Simulations were made with !
s+
H >.#% O >.. m@s and 0
cat?,
H >.%> O >.'*. Since the reactor was
assumed to be isothermal, all reaction heat must be remo4ed. This was modeled b" assuming
4ertical cooling tubes of *> mm in diameter, with coolant of %'> MC. The heat transfer
coefficient from slurr" to coolant was estimated b")
, /r +(e # . >
%
+
Br St
->
where the different dimensionless groups are e0plained in &eckwer et al. =#8-%?, a reference
in 2aretto and Drishna =#888?. The slurr" densit" was calculated from)
, cat p , cat
S-
,
, Sl , ,
#

,
_



where 4
p
is the catal"st densit" in =kg@m
'
of particle including 4oids? and 4
S-
is the catal"st
skeleton densit" in =kg@m
'
of catal"st solids without 4oids?. The slurr" 4iscosit" was estimated
according to)
, * . . # +
,, cat , Sl
+
".%.& The results
(esults from the simulation are shown in igures '* ! ';. $t can be seen that as the superficial
gas 4elocit", !
s+
+or 8 in igure '*,, increases, the con4ersion decreases, but the producti4it"
increases. That the producti4it" increases as the space 4elocit" is increased should be due to
an enhanced mass transfer. /racticall", a single pass con4ersion in a slurr" phase T!reactor
should be around 8> C so that no rec"cling is necessar". or such a high con4ersion, it is
necessar" to keep the superficial gas flow below >.' m@s.
ig. '*. ischerOTropsch reactor simulation results) s"ngas con4ersion =2aretto and Drishna, #888?.
-#
ig. '1. ischerOTropsch reactor simulation results) total reactor producti4it" +column diameter 'TH; m_
dispersion height .H'> m, =2aretto and Drishna, #888?.
$ncreasing the catal"st concentration in the slurr", 0
cat?,
+or 0
s
in igure '1,, of course increases
both con4ersion and producti4it", but the influence of 0
cat?,
is not onl" because of the
increasing number of acti4e catal"st sites, but also due to its reducing effect on the total gas
hold!up. So, increasing the 0
cat?,
leads to lower 0
+
and hence higher slurr" hold!up, 0
,
, which in
turn means that more catal"st can be loaded into the reactor. The higher the producti4it", the
higher the number of cooling tubes needed +igure ';,.
ig. ';. ischerOTropsch reactor simulation results) number of *>!mm!diameter cooling tubes =2aretto and
Drishna, #888?.
A sensiti4it" anal"sis was also performed to in4estigate the effect of the mass transfer
coefficient from gas to li7uid, and the effect of the fates!Satterfield kinetic constant a, on the
o4erall producti4it". The superficial gas 4elocit", !
s+
, was set to >.. m@s and the catal"st
concentration in the slurr" to 0
cat?,
H >.%*. A #>!fold increase or '!fold reduction of -
,
a as
compared with the base case, had a negligible effect on the producti4it". Howe4er, doubling
the kinetic constant a resulted in a 1> C increase of the producti4it". Hence, it was concluded
that the slurr" phase reactor at this base case is kineticall" controlled.
-%
5. Conclusion-discussion
$t seems common for the few models describing T in fi0ed!beds +trickle!beds, not to
consider the li7uid as a separate phase. This is probabl" since the li7uid essentiall" Gust is
e0isting as a thin film around the catal"st pellets, and hence the gas phase can be seen as the
continuous phase, and the li7uid e0ternal hold!up, 0
,
, is practicall" Fero. 5ne could of course
include the li7uid phase in the model, as would be the case for other gas@li7uid@solid s"stems
such as slurr" reactors or more con4entional trickle bed reactors +such as used in the
h"drotreatment of oils, in which the li7uid hold!up is much larger and also the li7uid flow.
The simplification with no separate li7uid phase, in the case of the model b" <ang et al.
=%>>'? who used true intrinsic kinetic rate e0pressions, is Gustified if the e0istence of the
li7uid film around the pellets does not affect the o4erall rate, i.e. if no interfacial gradients
occur between gas and gas@li7uid interface, and gas@li7uid interface and li7uid, and between
li7uid and solid. <ang et al. =%>>'? did account for temperature gradients in a gas film around
the pellets. $f mass transfer restrictions in the li7uid film around the pellets can be neglected,
there is no need to find the interfacial areas, a
4
g

and a
4
gg, which otherwise seem rather
troublesome to find especiall" if the catal"st pellets are not completel" wetted b" the li7uid.

The models described abo4e de4eloped b" <ang et al. =%>>'? and Jess et al. =#888? did not
e0plain how the product "ield was calculated. /robabl", a material balance o4er the whole
reactor is needed since no li7uid phase is accounted for, and the wa0 composition is assumed
to be constant, and hence cannot describe the proper product distribution.
$n all models presented here, the composition of the wa0 is assumed to be constant. $n order
to de4elop a model able to predict the performance of a wide range of conditions, the model
should be able to predict the composition of the wa0 phase. The composition of the wa0
determines '
ei
inside the pores and also the concentrations of components at the gas@li7uid
interface at the pore mouths. 2a"be, it would be possible to sol4e such a problem iterati4el",
b" guessing a composition of the wa0, and then integrate the gas phase o4er the reactor in
order to get a new guess 4alue for the composition. or each composition of the wa0, the
e7uilibrium 4alues for gas!phase concentration and li7uid!phase concentration of reactants
and products would be different. $n a real T!trickle!bed, the composition of the wa0 would
4ar" along the reactor, and, ma"be more important, the thickness of the li7uid film will
increase downwards the reactor. Another approach would be to assume that the wa0
composition in a F!point is gi4en b" the selecti4it" in that point. $n this case, of course the
selecti4it" is not constant in the reactor, but 4aries with for instance H
%
@C5 ratio at catal"st
surface, as suggested b" <ang et al. =%>>#?.
The fact that the catal"st pores are filled with li7uid wa0, and that diffusion through wa0 is
much slower than diffusion through gas, leads to the choice of making a heterogeneous model
to account for interfacial and intraparticle gradients. This could ha4e been made with an
effecti4eness factor instead of sol4ing the particle e7uations. Howe4er, such a model would
not be able to predict the product selecti4it" with a high accurac" outside the conditions used
for building the model, since the selecti4it" is 4er" much dependent on the H
%
@C5 ratio at the
acti4e site on the internal catal"st surface. 5nl" b" sol4ing the particle mass balance the true
H
%
@C5 ratio at the acti4e site can be predicted.
The comple0it" of the T s"nthesis regarding the broad product distribution makes it
desirable to simplif" the model e7uations most often b" choosing a #!dimensional model, and
-'
also to treat the li7uid and catal"st as a pseudo!homogeneous phase. $t also seems to be
common onl" to consider the mass balances for the reactants.
-.
References
Aasberg!/etersen, D., Christensen, T., 3ielsen, C., &"bkGaer, $., Recent de(elopments in
a!tot"ermal reforming and pre-reforming for snt"esis gas prod!ction in +#, applications,
uel. /roc. Tech., %>>', -', %*'!%1#
:haradwaG, S., Schmidt, L., Cataltic partial oxidation of nat!ral gas to sngas, uel. /roc.
Tech., #88*, .%, #>8!#%;
:iesheu4el, /., Dramer, J., #7o-section reactor model for a!tot"ermal reforming of met"ane
to snt"esis gas, A$Che Journal, %>>', .8, #-%;!#-';
:GTrgum, 9., researcher at S$3T9, Trondheim, personal communication April %>>1
:ub, J., :aerns, 2., :ussemeier, :., rohning, C., Prediction of t"e performance of cataltic
fixed bed reactors for Bisc"er-#ropsc" snt"esis, Chem. 9ng. Sci., #8->, '*, '.-!'**
&e Jroote, A., roment, J., Sim!lation of t"e cataltic partial oxidation of met"ane to
snt"esis gas, Appl. Catal. A, #881, #'-, %.*!%1.
&r", 2., Practical and t"eoretical aspects of t"e cataltic Bisc"er-#ropsc" process, Appl.
Catal. A, #881, #'-, '#8!'..
&r", 2., #"e Bisc"er-#ropsc" process% UVRE-2EEE, Catal. Toda", %>>%, ;#, %%;!%.#
&"bkGaer, $., #!b!lar reforming and a!tot"ermal reforming of nat!ral gas W an o(er(ie7 of
a(ailable processes, uel. /roc. Tech., #88*, .%, -*!#>;
9kbom, T., :erglin, 3., Lgdberg, S., $lac- ,iK!or +asification 7it" Lotor B!el Prod!ction
W $,+LB MM W A tec"no-economic feasibilit st!d of cataltic Bisc"er-#ropsc" snt"esis for
snt"etic diesel prod!ction in comparison 7it" met"anol and 'L* as transport f!els, proGect
report for contract /#'%-.!# in the Swedish 9nerg" Agenc"Es program Korsknings och
ut4ecklingsprogrammet fr alternati4a dri4medel %>>' ! %>>1 +ALT,L, %>>* +The report is
a4ailable at www.n"komb.se,
9spinoFa, (., Ste"nberg, A., Jager, :., Sosloo, A., ,o7-temperat!re Bisc"er-#ropsc"
snt"esis from a Sasol perspecti(e, Appl. Catal. A, #888, #-1, #'!%1
ogler, H. S., *lements of C"emical Reaction *ngineering, '
rd
ed., /rentice!Hall $nternational,
Anited States of America, #888
roment, J., :ischoff, D., C"emical reactor analsis and design, %
nd
ed., John <ile" h Sons,
$nc., Anited States of America, #88>
Jianetto, A., Sil4eston, /., L!ltip"ase c"emical reactors% t"eor? design? scale-!p,
Hemisphere /ublishing Corporation, Anited States of America, #8-1
-*
Jodfre", D., Shackcloth, :., 'namic modeling of a steam reformer and t"e implementation
of feedfor7ard;feebac- control, 2easurement and Control, #8;>, ', T1*!T;%
Hoang, &., Chan, S., Lodeling of a cataltic a!tot"ermal met"ane reformer for f!el cell
applications, Appl. Catal. A, %>>., %1-, %>;!%#1
Huang, 6., Sarma, A., 2cCread", 2., .eat transfer c"aracteri&ation of gas-liK!id flo7s in a
tric-le-bed, Chem. 9ng. Sci., %>>., *8, ';1;!';;1
Jakobsen, H., professor at &epartment of Chemical 9ngineering, 3orwegian Ani4ersit" of
Science and Technolog" +3T3A,, personal communication, %>>1!>'
Jakobsen, H., Bixed $ed Reactors, lecture notes in subGect) S$D%>*', (eactor Technolog",
spring %>>', &epartment of Chemical 9ngineering, 3orwegian Ani4ersit" of Science and
Technolog" +3T3A,, %>>'
Jess, A., /opp, (., Hedden, D., Bisc"er-#ropsc"-snt"esis 7it" nitrogen-ric" sngas W
f!ndamentals and reactor design aspects, Appl. Catal. A, #888, #-1, '%#!'.%
Dirk!5thmer 9nc"clopedia of Chemical Technolog", Solume #', .
th
ed., John <ile" h Sons,
ASA, #88*
Drishna, (., Sie, S., 'esign and scale-!p of t"e Bisc"er-#ropsc" b!bble col!mn sl!rr
reactor, uel /roc. Tech., %>>>, 1., ;'!#>*
LTdeng, (., researcher at S$3T9, Trondheim, personal communication April %>>1
Lgdberg, S., Lodeling of tric-le-bed reactors W #"e Bisc"er-#ropsc" reaction, e0amination
assignment in KTD/ .#.* (eaktorteknologiL at 3T3A, Trondheim, spring %>>1
2aretto, C., Drishna, (., Lodelling of a b!bble col!mn sl!rr reactor for Bisc"er-#ropsc"
snt"esis, Catal. Toda", #888, *%, %;8!%-8
2ouliGn, J., 2akkee, 2., 4an &iepen, A., C"emical Process #ec"nolog, John <ile" h Sons
Ltd, 9ngland, %>>#
5choa ernandVF, 9., (usten, H., Jakobsen, H., (Tnning, 2., Holmen, A., Chen, &., Sorption
enhanced h"drogen production b" steam methane reforming using Li
%
Qr5
'
as sorbent)
sorption kinetics and reactor simulation, Catal. Toda", %>>*, #>1, .#!.1
("dVn, 2., /h& student at Chalmers, Jothenburg, Sweden, personal communication 2arch
%>>1
SchulF, H., S"ort "istor and present trends of Bisc"er-#ropsc" snt"esis, Appl. Catal. A,
#888, #-1, '!#%
-1
Shah, f., +as-,iK!id-Solid Reactor 'esign, 2cJraw!Hill $nc., Anited States of America,
#8;8
Shah, f., Sharma, 2., +as-,iK!id-Solid Reactors, Ch. #> in C"emical reaction and reactor
engineering? 2arcel &ekker, $nc., #8-;
Sie, S., Drishna, (., B!ndamentals and selection of ad(anced Bisc"er-#ropsc" reactors,
Appl. Catal. A, #888, #-1, **!;>
Stehlik, /., Lat"ematical model of a steam-reforming radiation c"amber - O. *ffect of
selected parameters on t"e main c"aracteristics (al!es of t"e process, Chemick" /rum"sl,
#88#, .#, %'!%-
AllmanEs 9nc"clopedia of $ndustrial Chemistr", Solume A #%, *
th
ed., SCH, Jerman" #8-8
<ang, f., Li, f., :ai, L., Qhao, f., Qhang, :., Correlation for gas-liK!id eK!ilibri!m
prediction in Bisc"er-#ropsc" snt"esis, uel, #888, ;-, 8##!8#;
<ang, f., 6u, f., 6iang, H., Li, f., Qhang, :., Lodeling of catalst pellets for Bisc"er-
#ropsc" snt"esis, $nd. 9ng. Chem. (es., %>>#, .>, .'%.!.''*
<ang, f., 6u, f., Li, f., Qhao, f., Qhang, :., .eterogeneo!s modeling for fixed-bed
Bisc"er-#ropsc" snt"esis% Reactor model and its applications, Chem. 9ng. Sci., %>>', *-,
-1;!-;*
<ender, $., Reactions of snt"esis gas, uel /rocessing Technolog", #881, .-, #-8!%8;
<ilhelm, &., Simbeck, &., Darp, A., &ickenson, (., Sngas prod!ction for gas-to-liK!ids
applications% tec"nologies? iss!es and o!tloo-, uel /roc. Tech., %>>#, ;#, #'8!#.-
<illiams, Bisc"er-#ropsc" snt"esis, 5ral presentation b" (achid 5ukaci at the %
nd
Annual
Jlobal JTL Summit 90ecuti4e :riefing, 2a" %-!'>, %>>%, London, AD.
-;

You might also like