Appendix D Maxwell-Stefan Equations

For Multicomponent Transport

Solution of simultaneous mass and energy transfer equations:
For a system with n components, the number of unknowns in the problem are 2n mole fractions at
the interface (xi
I
and yi
I
, n molar fluxes (incorporating !i
"
#!i
$
#!i
$"
and an interface temperature (%
I
&
Since energy equilibrium at the interface is also considered, it acts as a bootstrap to relate the total flux to
the absolute quantities without needing any further assumptions (of inert flux being 'ero or stoichiometric
fluxes etc&& %hus the total number of (ariables are )n*+& %he number of equations relating these (ariables
is
a n,+ rate equations for the liquid phase gi(en by
N J x J x
q
i
L
i
L
i k k
L
k
n
i
x
+

+
+

(-& +
where ( ) [ ]( ) J c k x x
L
t
L
L
I L

(in matrix form, (n,+x(n,+ (-& 2

and
k
#(
k
,
n
.
x
, q#h
&
"
(%
I
,%
"
,h
&
$
(%
/
,%
I
(-& )
(
i i
V
i
L
H H and

x i i
i
n
x

+
,

y i i
i
n
y

+
(-& 0
b n,+ rate equations for the (apor(gas phase gi(en by
N J y J y
q
i
V
i
V
i k k
V
k
n
i
y
+

+
+

(-& 1
where ( ) [ ]( ) J c k y y
V
t
V
V
V I

(in matrix form, (n,+x(n,+ (-& 2

c n equilibrium relations y K x
i
I
i i
I
(-& 3
d Interfacial energy balance
h T T N H T h T T N H T
L I L i
L
i
L
L
i
n
V G I i
V
i
V
G
i
n
. .
( ) ( ) ( ) ( ) + +


+ +
(-& 4
e %wo more equations are necessary to complete the set and are gi(en by mole fraction sum for both gas
and liquid compositions at the interface&
x
i
I

+
(-& 5
y
i
I

+
(-& +6
%his completes the set of equations required to sol(e for all the fluxes, compositions and
temperatures& %he mass and heat transfer coefficients are the corrected (alues at the fluxes and can be
re(ised based on this flux calculation until con(erged (alues are obtained& -etails of iteration for finite flux
transfer coefficients can be found in %aylor and 7rishna (+55)&
A.) The Bootstrap matrix for interphase transport
+& 8ootstrap matrix for gas transport is e(aluated based on interphase energy flux for the gas ,liquid
interface
9 :
,

i k G ik i k
y
(-& ++
where

ik k nc y
( .
(-& +2
and

y i i
i
y

,
i i i
V
G i
L
L
y H T H T ( (; (;
(-& +)
For the gas,solid interface similar equations are used except the enthalpy is now of the liquid in the solid
(liquid at solid temperature&
2& 8ootstrap matrix for the liquid phase is also based on the energy flux term and gi(en by
9 :
,

i k L ik i k
x
(-& +0
where

ik k nc x
( .
(-& +1
and

x i i
i
x

,
i i i
V
G i
L
L
x H T H T ( (; (;
(-& +2
)& <atalyst le(el equations ha(e unsymmetric boundary conditions due to gas and liquid bulk flow
streams on either side of partially wetted pellets and suitable boundary conditions ha(e to be
incorporated into the bootstrap matrix to ensure that mass balance is satisfied in the pellets& Se(eral
bootstraps are possible in this case such as stoichiometric fluxes, 'ero flux for inert component, and net
'ero flux for 'ero net mole production& 8ut these are not deemed suitable for the present problem&
=ence the most likely condition would be that net (olumetric flux entering on the wetted side is equal
to net mass flux lea(ing on the dry side& %his can be incorporated directly in the fully implicit beta
e(aluation as


S / "i i i S " "i i i
. > ! ( . > ! (
(-& +3
or in general since beta is not e(aluated at each intra,catalyst location, it can be represented as
9 :
,

i k CP ik ci k
x
(-& +4
where

k k nc mx
M M ( .
(-& +5
and

mx i i
i
x M

(-& 26
B) The transport matrices for interphase mass and energy flux
+& For the gas phase the low flux mass transfer coefficient matrix can be calculated as
9 : 9 : 9 : k B
G

+
(-& 2+
where
9 :
,
, ,
B
y
k
y
k
i j
i
i nc
k
i k k
+
(for i # ? (-& 22
9 : (
,
, ,
B y
k k
i j i
i j i nc

+ +
(for i ? (-& 2)
=ere
9 :
,
,
k
D
l
i j
i j

where the diffusi(ity coefficients are calculated from infinite dilution diffusi(ities
(from kinetic theory using mole fraction weighting factors& %he 9: matrix is the non,ideality coefficient
matrix, which is considered to be identity for the gas side&
2& %he liquid phase mass transfer coefficient matrix are calculated similarly as
9 : 9 : 9 : k B
L

+
(-& 20
where
9 :
,
, ,
B
y
k
y
k
i j
i
i nc
k
i k k
+
(for i # ? (-& 21
9 : (
,
, ,
B y
k k
i j i
i j i nc

+ +
(for i ? (-& 22
=ere
9 :
,
,
k
D
l
i j
i j

where the diffusi(ity coefficients are calculated from infinite dilution diffusi(ities
(using the @ilke,<hang correlation as gi(en in Appendix B using mole fraction weighting factors& %he
9: matrix is the non,ideality coefficient matrix, which is considered to be identity for the gas side&
C. High flux correction
+& =igh flux correction is important for the gas side mass and heat transfer coefficients and has been
incorporated as below&
-etermine 9: matrix based on the calculated fluxes

ii
i
t in
k
t ik
N
c k
N
c k
+
(-& 23

ii i
t ij t in
N
c k c k
(
+ +
(-& 24
correct the low flux mass transfer coefficient matrix by the correction as
9 : 9 : 9 :9exp(9 : 9 :: k k I
hih!l"x l#$!l"x



+
(-& 25
where exp9: is e(aluated by a Cade approximation as
exp9 :
9 : 9 : .
9 : 9 : .

I
I
2
2
(-& )6
Similarly, heat transfer coefficient is also corrected by

ht
i %i
N C
h

(-& )+
h h
hih!l"x l#$!l"x ht ht
. ((exp( +
(-& )2
Appendix E Evaluation Of arameters
For !nstead" State Model
"iquid phase infinite dilution diffusi(ities are determined using the @ilke and <hang (+511 correlation as
2 & 6
i ?
2 . +
? ? +2 6
i?
$
% > (
+6 x 0 & 3 -

(B& +
/as phase infinite dilution diffusi(ities are determined as gi(en by Deid et al& (+543
2 ) . +
i
) . +
i
31 & + 2 . +
? i ? ? 2 6
i?
$ $ ( C
% > > . > > ((
+6 x 6+) & + -
+
+

(B& 2
%hese are corrected for concentration effect using the $ignes (+522 correlation as recommended by taylor
and 7rishna (+55)
x ( 6
?i
x + ( 6
i? i?
i i
- ( - ( -

(B& )
%he film thickness for gas and liquid phase is determined using the Sherwood correlation by using the
particle diameter as the characteristic dimension as
00 & 6 4) & 6 C
Sc De 62) & 6
d

(B& 0
%he Schmidt number used in the abo(e correlation is different for each species and can gi(e multiple (alues
of film thickness according to the species used in the calculation& %his dilemma is resol(ed using either an
a(erage (alue of diffusi(ity or the diffusi(ity of the gas phase key reactant&
=eat transfer coefficients for gas and liquid phase are determined using either the !usselt (B&1 number or
the <hilton,<olburn analogy (B&2 as recommended by %aylor and 7rishna (+55)
55 & 6 De +0 & 2 ( Cr
k
d h
2 . + ) . + C
+

(B& 1
) . 2 mt
C t

Sc
Cr
( k < c h



(B& 2
%he interfacial areas required for interphase transport (gas,liquid, liquid,solid, and gas,solid are e(aluated
using the following correlations& ,
2
D
C 0 & 6
"
C 8
" 0
/"

-
d
( De
d
+
( + ( +6 x 5 & ) a

(B& 3
<B
E
C
8 "S

S
$
( + ( a
(B& 4
+ (
S
$
( + ( a
<B
E
C
8 /S

(B& 5
%he modeling of liquid phase non,ideality effects can be done by using any of the se(eral
approaches a(ailable (such as simple 2 parameter models such as the >argules, $an,
"aar, or the @ilson equation or complex rigorous methods such as the F!IGFA< or
F!IFA<& In the present study, the @ilson equation was used to obtain the equilibrium
constants as well as the acti(ity matrix (9: as gi(en by the equations below&
For the @ilson equation, the excess energy G is gi(en by

n
+ i
i i
S ln( x G
(B& +6
where

n
+ ?
i? ? i
x S
(B& ++
D% . ( exp(
$
$
ii i?
i
?
i?

(B& +2
First and second deri(ati(es of G w&r&t& compositions i and ? are then gi(en by

n
+ k
k ki k i i
S . x S ln( G
(B& +)

+
n
+ k
2
k k? ki k ? ?i i i? i?
S . x S . S . G
(B& +0
%he acti(ity coefficients needed for calculating the equilibrium constants are then gi(en directly by
G exp(
i i
+ (B& +1
%he thermodynamic factors (i? required for the nonideality matrix are gi(en by
G G ( x
in i? i i? i?
+
(B& +2
Appendix F. Experimental Data From
Stead" And !nstead" Experiments
Appendix #. Simulation of Flow usin$
%FD&'(
A. ro)lem Definition
Credicting the fluid dynamics of trickle bed reactors is important for their proper scale,up to
industrial scale& Cre(ious studies ha(e resorted primarily to prediction of o(erall phase holdup and pressure
drop based on an empirical or phenomenological approach& Decent ad(ances in simulation of multiphase
flow and de(elopment of robust codes that can handle two and three dimensional simulations ha(e made
flow simulation feasible in complex flows such as those obser(ed in trickle beds& <DB" has access to the
<F-"I8 multiphase flow codes de(eloped by the "os Alamos !ational "aboratory which are used in this
study&
(. *esearc+ O),ectives
%he ob?ecti(e of this study is to simulate the effect of single point and multi,point inlet flow
distribution on the flow distribution inside the reactor and to expand the test database beyond the
preliminary simulations presented by 7umar (+551,+552& %his in(ol(es modification of con(entional drag
and interfacial exchange terms implemented in <F-"I8 using drag formulations de(eloped at <DB" as
well as those a(ailable in trickle bed literature& Another ob?ecti(e is to study the influence of surface
tension on the spreading of liquid in single and multi,point inlet conditions& %his will ser(e as a benchmark
for comparison with experimental (elocities and phase holdup data which ha(e not yet been reported in the
open literature&
%. *esearc+ Accomplis+ments
%-. Modelin$ 'nterp+ase Exc+an$e and 'nterfacial Tension Terms
%he underlying equations for the <F-"I8 code ha(e been discussed in detail in earlier reports by
7umar (+552 and can be found in 7ashiwa et al& (+550& %he special case of one fixed phase (the catalyst
bed has also been incorporated in the code for single and two phase flow simulation& %he important terms
in simulating trickle bed reactors is the drag equation and the influence of phasic pressure difference due to
interfacial tension& Chenomenological models de(eloped at <DB" by =olub (+556 are incorporated in
simulating the drag between the stationary solid phase and each of the flowing phases& %he code models the
drag force in terms of phase fractions and relati(e (elocity gi(en for any combination of phases k and l as
& ' " "
D k l k l kl k l (
(

(+
where the Ekl is modeled by the modified Brgun equation (=olub, +556, Sae' and <arbonell, +541 with
Brgun constants either determined by single phase experiments or using uni(ersal (alues&
'
(
G)
(
G)

"
l *
*
L
L
L
L
L
L
L* *
(
( De De
H H(

_
,
+

1
]
1

+
+
)
+ 2
2

(2
'
(
G)
(
G)

"
G *
*
G
G
G
G
G
G
G* *
(
( De De
H H(

_
,
+

1
]
1

+
+
)
+ 2
2

()
For gas,liquid drag either no interaction is assumed or interaction based on a drag coefficient is used as
'
C "
+
G L
L D GL
L %
(
& H H

6 31

(0
For modeling interfacial tension the famous "e(erettIs J function (-ankworth et al&, +556 is used to yield
the difference between the gas and liquid pressure calculated in the code in terms of the interfacial tension
( , bed permeability (k and phase fractions as
% %
k
L G
* * L
L

_
,
+

'

+
6 04 6 6)2
+
+ 2

.
& & &ln(
(1
%he bed permeability (k is related to Brguns constant B+ and equi(alent particle diameter (de as
(( .
(
(
.
+
+
+ 2
+

*
*
* ,
k
(
+
(2
%he simulations are conducted both by discounting and incorporating the abo(e equation and results
presented in the next section&
%.. Simulation of Test %ases/ *esults and Discussion
<ase I: (Deactor -imensions 26x++ cm, 8#6&0, dp# +&1 mm,
Fs"#6&6)2 cm.s, Fs/#)&2) cm.s
(a P#int *#".c, Inl,t $ith /ni!#.m B,+ P#.#0ity
(b P#int *#".c, Inl,t1 Incl"+in *".!)c, T,n0i#n (!!,ct0
For the test case of point source liquid inlet when the surface tension forces are ignored, Figure +
shows that the liquid stays mainly in the central core without much spreading and with almost negligible
wall flow& For the case where interfacial tension was included and modeled using a capillary pressure gi(en
by the "e(erett Function, significant liquid spreading was obser(ed at a depth of o(er +6 cm from the inlet&
Figures 2 and ) show the contours of the liquid holdup and de(elopment of the (elocity and holdup profile
down the reactor to a constant and fairly uniform flow profile at the reactor exit& In this case some wall
flow is indeed obser(ed after a depth of 24 cm as shown in Figure 2& %his is corroborated by (isual and
photography experiments done at <DB" on a 2- bed (Jiang, +553 and by the tomography results of "utran
et al&(+55+&
<ase II: Fniform "iquid and /as Inlet : !o Surface %ension Bffects, Fniform Corosity e(erywhere except
at the wall and exit section (Deactor -imensions 24&4x3&2 cm, 8#6&0 in the core and 8#6&1 at the wall, dp#
) mm, Fs"#6&+ cm.s, Fs/#1&6 cm.s
(a N# G)0 &l#$ (/n0)t".)t,+ Liq"i+ &l#$)
(b L#$ G)0 &l#$ (L#$ Int,.)cti#n)
(c Hih G)0 &l#$ (M#+,.)t, Int,.)cti#n)
For the uniform distribution tests, it was assumed that the wall 'one extends three particle
diameters from the wall and the exit of the reactor for which a higher porosity was assigned, and to the rest
of the central core a uniform lower porosity was assigned& It is much more difficult to discern the flow
distribution profiles in this case as compared to <ase I, primarily due to local non,uniformity in the phase
holdups and (elocity& %he only ob(ious effect is that of higher wall flow of gas at the wall as compared to
the central core as depicted in Figure 0 for the moderate gas flow rate (case IIc& For the low and 'ero gas
flow rate such clear gas wall flow was not obser(ed& "arger reactor si'e tests for this case are underway and
results will be reported in future reports&
<ase III: M"lti%#int Inl,t #! Liq"i+ (Including Surface %ension Bffects (Deactor -imensions 26x2) cm,
8#6&0, dp# +&1 mm, Fs"#6&612 cm.s, Fs/#)&03 cm.s
%his case is shown to illustrate conditions similar to industrial distributors with multiple points of
a large distributor each of which is similar to that in <ase I& %he holdup contour plot shown in Figure 1
shows underutili'ation of the top part of the bed indicating the required pitch between distributor inlet
points should be smaller than 1 cm that was used here to achie(e better distribution& %his mal,distribution
may result in 'ones where hotspots may de(elop such as in the gas phase ad?acent to the boundary of the
gas,liquid 'one (where both gas and liquid reactants are abundantly supplied to the catalyst for the case of
non,(olatile liquid reactant and in the gas rich 'ones in the case of (olatile liquid reactant between the well
irrigated 'ones& Such simulations can yield direct information as to the extent and consequence of
maldistribution due to well separated multi,point inlets typical in industrial reactors&
D. Future 0or1
%he potential for <F-"I8 to predict flow distribution in trickle bed reactors is shown in this
report for se(eral test cases& Further work will encompass simulation of more complex reactor geometry
and flow situations and extension of flow simulations from cold,flow modeling to reacti(e cases where
accurate prediction of product distribution and hot,spot formation is crucial to optimal and safe operation of
pilot and large scale reactors&
E. 2omenclature
<- # -rag <oefficient
dp # Carticle -iameter
de # Carticle Bqui(alent -iameter
B+,B2 # Brguns <onstants
F-(kl) # -rag Force between Chases k and l&
g # /ra(itational Acceleration
/a

# Chase /alileo !umber

k # 8ed Cermeability
p

# Chase Cressure
De

# Chase Deynolds !umber

uk # Interstitial $elocity of Chase k.
Ekl # Interphase Bxchange <oefficient between phases k and l&
G.,,k *ym2#l0
k # Chase Fraction of Chase k.

# Chase -ensity
# Interfacial %ension
F. *eferences
+& -ankworth, -& <&, 7e(rekidis, I&/&, and Sundaresan, S&, %ime -ependent =ydrodynamics in
>ultiphase Deactors, <hem& Bng& Sci&, $ol& 01, !o& 4, pp& 22)5,2202 (+556&
2& =olub, D& A&, =ydrodynamics of %rickle 8ed Deactors& Ch&-& %hesis, @ashington Fni(ersity in St&
"ouis, >K (+556&
)& Jiang& L&, Fnpublished Desults on 2,- %rickle 8ed Deactors using Coint and Fniform "iquid Inlet
-istributors (+553&
0& 7ashiwa, 8& A&, Cadial, !& %&, Dauen'ahn, D& >& and @& 8& $ander=eyden, A <ell centered I<B
>ethod for >ultiphase Flow Simulations, AS>B Symposium on !umerical >ethods for >ultiphase
Flows, "ake %ahoe, !e(ada (+550
1& 7umar, S& 8&, Simulation of >ultiphase Flow Systems using <F-"I8 code <DB" Annual >eeting
@orkshop (+551&
2& 7umar, S& 8&, !umerical Simulation of Flow in 8ubble <olumns, <DB" Annual Deport (+552&
3& "utran, C& /&, !g, 7& >& and -elikat, B& C&, "iquid distribution in trickle beds: An experimental study
using <omputer,Assisted %omography& Ind& Bngng& <hem& Des& )6 (+55+
4& Sae', A& /& and <arbonell, D& /&, =ydrodynamic Carameters for /as,"iquid <ocurrent Flow in Cacked
8eds, AI<hB J& )+, 12 (+541
Title:
Creator: TECPLOT
CreationDate:
Fi$ure -/ &iquid 3oldup %ontours for Sin$le oint Source 'nlet 4%ase 'a)
Title:
Creator: TECPLOT
CreationDate:
Fi$ure ./ &iquid 3oldup %ontours for Sin$le oint Source 'nlet 'ncludin$ 'nterfacial Tension Effects
4%ase '))
6
6&6+
6&62
6&6)
6&60
6&61
6&62
6&63
6&64
6 1 +6
E(cm
"
i
q
u
i
d

=
o
l
d
u
p

(
e
"

L#16 cm from Inlet

L#)6 cm from Inlet
L#+6 cm from Inlet
6&66B*66
2&66B,6+
0&66B,6+
2&66B,6+
4&66B,6+
+&66B*66
+&26B*66
+&06B*66
+&26B*66
+&46B*66
2&66B*66
6 1 +6
E(cm
"
i
q
u
i
d

$
e
l
o
c
i
t
y

(
$
i
"

L#16 cm from Inlet

L#)6 cm from Inlet
L#+6 cm from Inlet
Fi$ure 5/ &iquid 3oldup and 6elocit" rofiles for Sin$le oint Source &iquid 'nlet 4wit+ 'nterfacial
Tension Effects)4%ase '))
6&)1
6&0
6&01
6&1
6&11
6 +&4 )&2 1&0 3&2
E(cm
/
a
s

=
o
l
d
u
p

(
e
/

L#23 cm from Inlet

L#+4 cm from Inlet
L#5 cm from Inlet
)
)&1
0
0&1
1
1&1
2
6 +&4 )&2 1&0 3&2
E(cm
/
a
s

F
l
o
w

(
e
/
M
$
i
/

L#23 cm from Inlet

L#+4 cm from Inlet
L#5 cm from Inlet
Fi$ure 7/ #as 3oldup and Flow rofiles for %ase ''c 4Moderate #as Flow)
Title:
Creator: TECPLOT
CreationDate:
Title:
Creator: TECPLOT
CreationDate:
Fi$ure 8/ &iquid 3oldup and 6elocit" for Multipoint 'nlet Distri)utor 4%ase ''').