Heat Transfer Limits JPS
Heat Transfer Limits JPS
Heat Transfer Limits JPS
= (8)
) 33 . 4
8 . 4577
exp(
3
=
T
K (9)
The majority of the PrOx studies used the power law expression for CO oxidation
due to its simplicity [9,10,11]. This form is simplified from a Langmuir-Hinshelwood (L-H)
expression and not suitable for small CO concentrations [5]. Therefore we adopted the full
L-H expression instead of the power law expression for CO oxidation (Equation 4). The H
2
oxidation was previously modeled using empirical power law rate expressions by others
[4]. However, in the presence of CO, the rate-limiting CO desorption strongly inhibits H
2
and O
2
adsorption and hence H
2
oxidation in PrOx [1]. Thus the incorporation of P
CO
in
the H
2
oxidation rate expression is necessary (Equation 5). The kinetics of r-WGS reaction
was well studied previously [15], in which an empirical reversible rate expression [16] is
attractive due to its relative simplicity and its appropriateness in PrOx kinetic studies as
demonstrated previously [4]. With the rates of CO oxidation, H
2
oxidation and r-WGS
8
reaction obtained, the net rates of individual reactant species (i.e.,
'
CO
r ,
'
2
CO
r ,
'
2
H
r ,
'
2
O H
r
'
2
O
r )
are calculated as in Equations 10 12.
'
3
'
1
' '
2
r r r r
CO CO
= = (10)
'
3
'
2
' '
2 2
r r r r
O H H
+ = = (11)
) ( 5 . 0
'
2
'
1
'
2
r r r
O
+ = (12)
A multiple nonlinear regression analysis with the Marquardt method [17] was
selected to evaluate the independent variables (A
1
, A
2
, A
3
, E
1
, E
2
, E
3
,
1
,
2
,
3
,
0
1
K ,
ads
H )
in the kinetic expressions. Following the initial input of the independent variables, the
dependent variables (
CO
P ,
2
O
P ,
2
CO
P ,
2
H
P ,
O H
P
2
) were calculated with the 4
th
order Runge-
Kutta method [17]. Data fitting of the calculated dependent variables to experimental data
was then carried out to derive corrected values of the independent variables with the
Marquardt method [17]. Afterwards, these corrected values were fed to the kinetic
expressions for the next iteration. The final values of the dependent variables were reached
by minimizing the weighted sum of squared residuals for dependent variables. The above
algorithm was realized with MATLAB.
2.3. Quasi-3D Non-Isothermal Reactor Model
The first part of the non-isothermal reactor model was the construction of mole
balance and energy balance equations. The kinetic rate expressions of different reactant
species in Equations 10 12 were used as the mole balance equations. The pressure drop
across the reactor was neglected in the modeling for m-PBRs since the Ergun equation [18]
9
suggested acceptably negligible pressure drop for the flow conditions and reactor
characterizations used in the modeling. The first law of thermodynamics was used for the
derivation of the energy balance equation (Equation 13) [18]. After dissecting the heat flow
(
Q d ) into radial and axial terms, a partial differential equation (Equation 14) is obtained
as the energy balance equation in the model. The first bracketed term is the heat removal in
axial and radial directions and the second is the heat generation from the three reactions in
PrOx.
=
5
1
0 ] ) ( [
i
R i
dV
dF
T H
dV
Q d
o
(13)
0 ] ) ( [
1
5
1
2
2
2
2
=
)
`
)
`
|
|
.
|
\
|
= i
i
R i
dV
dF
T H
l
T
r
T
r
T
r
k
o
(14)
The next part describes the quasi-3D structural model of the reactors. Cylindrical
geometry was used to approximate both the microchannel and the packed bed. Thus the 3D
reactor structure can be represented by the quasi-3D finite difference grid in radial and
axial directions, as shown in Figure 2. The volume of a differential 3D element is therefore
calculated as 2[(j-1)dr] dr dl (Figure 2).
As the third step, derivatives were transformed to finite-difference forms by a well-
established approach [19,20]. The resulting linear equations are shown in Equations 15
17.
) 2 (
) (
1
) (
) , 1 (
) (
) , (
) (
) , 1 ( 2 2
) 1 (
) , (
2
k
i j
k
i j
k
i j
k
i j
T T T
dr r
T
+
+
+ =
(15)
10
) 2 (
) (
1
) (
) 1 , (
) (
) , (
) (
) 1 , (
2 2
) 1 (
) , (
2
k
i j
k
i j
k
i j
k
i j
T T T
dl l
T
+
+
+ =
(16)
) (
2
1
) (
) , 1 (
) (
) , 1 (
) 1 (
) , ( k
i j
k
i j
k
i j
T T
dr r
T
+
+
=
(17)
Finally, the Gauss-Seidel iteration method [20] was used to solve the mole balance
and energy balance equations simultaneously. The sum of squared residuals was
minimized as the criterion to end iteration.
3. Results and Discussion
3.1. Qualitative Analysis of Heat Transport Limitations in Reactors
Our addtional PrOx studies [1] discussed the heat transport resistance with Mears
criterion, which suggested that interparticle heat transport dominates in both the
microreactor and m-PBRs compared to intraparticle and interphase heat transport. The
Damkhler number (Da) was then used for qualitative comparison of heat transport
limitations for the two types of reactors, in which the heat conduction distance (t
cat
) is the
determining factor. The t
cat
of the thin-film catalyst (510
-6
m) is orders of magnitude
smaller than that for the packed-bed catalyst (~210
-3
m). Due to the quadratic dependence
of Da on the t
cat
, Da of the m-PBR with 210
-3
m radius (2-mm m-PBR) is more than 5
orders of magnitude of Da for the microreactor. The Mears criterion then suggested severe
heat transport limitations for the m-PBRs for the highly exothermic PrOx reaction.
11
3.2. Temperature Gradients for the microreactor and the 2-mm m-PBR
In this section, in order to derive the correlations between reactor size and reaction
performance, we discuss the temperature distribution and PrOx performance for both types
of reactor based on identical operating conditions (reactor wall temperature and WHSV).
The modeling results of the microreactor showed essentially isothermal
temperature distribution in the thin-film catalyst even at the highest operating temperature
(300C), supporting our qualitative analysis with Mears criterion. Also as predicted, the
results for the 2-mm m-PBR showed significant temperature gradients and effect on PrOx
performance, as shown in Figures 3 and 4. Figure 3 plots the conversions (O
2
, CO, H
2
) and
selectivity (CO
2
) along the reactor length for three representative wall temperatures (T
w
=
120C, 180C, 220C). Figure 4 shows the 3D figures of the temperature distribution also
at these wall temperatures.
As described in Section 2.4 and Equation 14, the overall product of the reaction
rates of the PrOx reactions (Equations 1 3) and their reaction enthalpies determine the
local heat generation in the reactor. Since the reaction enthalpies for the exothermic
oxidations of CO and H
2
are similar (Equations 1 and 2) and the endothermic r-WGS
reaction has reaction enthalpy almost one order of magnitude smaller (Equation 3), the
amount of local heat generation is controlled by the oxidation reactions and roughly
proportional to the local O
2
reaction activity. This argument allows the correlation of the
modeling results of reaction and temperature distribution as follows.
Due to the low O
2
reaction activity at T
w
= 120C (Figure 3a), the temperature
gradient is negligible in both axial and radial directions (< 1C), as shown in Figure 4a. At
12
180C, O
2
reaction activity is mild up to l = 210
-2
m, which leads to small heat
accumulation and temperature gradients in this region. However, the sudden increase of
both CO and H
2
oxidation activity between l = 210
-2
~2.510
-2
m causes a dramatic
increase of net O
2
reaction activity, leading to temperatures much higher internally than at
the wall (Figures 3b and 4b). Figures 3c and 4c show the results of reaction and
temperature distribution at T
w
= 220C; more severe temperature gradients are developed
close to the entrance of the reactor since full O
2
conversion is reached at a length of only
10
-3
m.
In order to gain further insight into the effect of wall temperature on hot spot
formation, the axial temperature distribution in the center (r = 0.0 m) and the radial
temperature distribution at l = l
Tmax
are plotted in Figures 5 and 6, respectively. Figure 5
shows that the temperature gradients become more dramatic and move upstream with
higher wall temperatures. With the wall temperature increases, full conversion of O
2
(with
CO and H
2
) takes place in shorter reactor lengths. Due to the similar reaction heat of CO
and H
2
oxidations, the total heat generated by full O
2
conversion is almost constant even as
selectivity varies. A shorter length for full O
2
conversion thus implies a higher density of
heat accumulation and consequently larger temperature gradients in the region close to
l
Tmax
. Correspondingly, the radial temperature gradients also increase with the wall
temperature, as shown in Figure 6.
3.3. Operating Temperature Window of PrOx
13
Figure 7 plots the modeling results of CO conversion at different wall temperatures
for the microreactor and m-PBRs with 210
-3
m and 410
-3
m radii (2-mm and 4-mm m-
PBRs), all with the same WHSV (1500 hr
-1
) and isothermal wall temperatures. The CO
conversion curve for the microreactor essentially coincides with the result for ideal
isothermal operation. On the contrary, the CO light-off curves for the m-PBRs shift to
lower temperatures and CO conversion drops significantly at higher wall temperatures. The
light-off shift to lower temperature is due to an increase in CO oxidation rate at local hot
spots. With further temperature increase, however, these temperature non-uniformities
activate the r-WGS reaction, leading to a drop of net CO conversion. These phenomena are
seen to intensify as the radial thermal resistance increases (4-mm vs. 2-mm). The inability
of the 4-mm m-PBR to reach 100% conversion is indicative of severe hot spots even at
relatively low wall temperatures. The CO conversion curves from literature PrOx studies
with m-PBRs [2,3,4] agree qualitatively with these modeling results and thus can be
understood with the above discussion. However, due to the lack of detailed kinetic data of
other PrOx catalysts in the literature, further effort to specifically predict PrOx behavior for
these PrOx reactors is impractical.
Figure 8 then shows the average reaction rates of CO oxidation, H
2
oxidation and r-
WGS reaction for the microreactor, the 2-mm and 4-mm m-PBRs at the same reaction
conditions used in Figure 7, which were calculated by integrating and then averaging the
reaction rates over the entire reactor volume. It clearly indicates that the m-PBRs has
increasing r-WGS reaction rate with wall temperatures higher than 220C. Consequently,
these phenomena are more significant for the 4-mm than the 2-mm m-PBR. In contrast, the
14
isothermality within the microreactor effectively minimizes the extent of the r-WGS side
reaction at these temperatures.
Additional results from our model also clarified other factors that influence the
reactor temperature profile and PrOx reaction performance. For example, perfect wall
insulation leading to adiabatic operation results in heat removal only through the axial
direction and as expected, eliminates radial temperature gradients. Heat accumulation is
more severe than in the isothermal wall condition, leading to an even higher local
temperature and adverse PrOx performance. Similarly, a higher catalyst active metal
density causes large O
2
conversion in shorter reactor lengths and higher local temperatures.
Thus thermal management through the control of reactor radius, heat exchange conditions
and catalyst loading density is crucial in the PrOx reactor design.
4. Conclusions
Divergence of the PrOx reaction behavior between a microreactor with thin-film
catalyst and m-PBRs were discovered and delineated here. A quasi-3D finite difference
non-isothermal reactor model was developed to gain insight into the origins of the
differences. The microreactor has negligible temperature gradients within the entire
experimental temperature range due to its small t
cat
. On the contrary, due to its greater
effective t
cat
, the m-PBRs develop significant temperature gradients in both radial and
axial directions under the conditions considered here experimentally and through
simulation. For wall temperatures before and during light-off, the slight local temperature
gradients in the m-PBRs cause the light-off curve to shift to lower temperatures. At
15
moderately higher temperatures, severe temperature gradients rapidly develop in the m-
PBRs which degrade the PrOx performance by activating the r-WGS reaction and
decreasing the net CO conversion.
Nomenclature and Units
Greek letter
i
reaction order for rate expression of H
2
oxidation (i: O
2
, H
2
, CO)
letters
A
i
pre-exponential factor in kinetic expressions
D
a
Damkhlers number for heat transport
E
i
activation energy for kinetic expressions (J mol
-1
)
F
tot
total reactant flow (m
3
s
-1
at 20C and 101330 Pa)
H
ads
adsorption energy (J mol
-1
)
k
i
specific reaction rate for
'
1
r ,
'
2
r and
'
3
r (i = 1, 2 or 3)
K
1,
0
1
K adsorption equilibrium constant for CO
K
3
equilibrium constant of the r-WGS reaction
l reactor axial position, l = 0 for the entrance of the reactor (m)
l
Tmax
the axial position where the temperature (at r = 0) is the highest (m)
L total reactor length (m)
P
i
partial pressure of reactant species i (i: CO, O
2
, CO
2
, H
2
, H
2
O) (Pa)
r reactor radial position, r = 0 for the center of the reactor (m)
16
'
1
r reaction rate of CO in CO oxidation, moles of reactant per gram of active
metal of catalyst per second (mol kg
-1
s
-1
)
'
2
r reaction rate of H
2
in H
2
oxidation, moles of reactant per gram of active
metal of catalyst per second (mol kg
-1
s
-1
)
'
3
r reaction rate of CO
2
in r-WGS oxidation, moles of reactant per gram of
active metal of catalyst per second (mol kg
-1
s
-1
)
R reactor radius (m)
S
CO2
CO
2
selectivity (defined as the ratio of O
2
reacted with CO and total O
2
reacted)
cat
t effective interparticle heat conduction distance: catalyst thickness for the
microreactor or reactor inner radius for m-PBR (m)
T local temperature in the reactor (C)
T
w,
T
b
reactor inner wall temperature (C)
WHSV weight hourly space velocity, defined as mole of total reactant flow per
mole of precious metal in the catalyst per hour (hr
-1
)*
X
i
conversion of species i (i: CO, O
2
, H
2
)
* Non-SI unit used by convention
17
References:
[1] X. Ouyang, L. Bednarova, P. Ho, R.S. Besser, manuscript submitted to AIChE J.
(2004).
[2] M.J. Kahlich, H.A. Gasteiger, R.J. Behm, J. Catal., 171 (1997) 93105.
[3] S.H. Oh, R.M. Sinkevitch, J. Catal., 142 (1993) 254-262.
[4] Y. Choi, H.G. Stenger, J. Power Sources, 129 (2004) 246-254.
[5] R.H. Venderbosch, W. Prins, W.P.M. van Swaajj, Chem. Eng. Sci., 53 (1998) 3355-
3366.
[6] G.W. Roberts, P. Chin, X.L. Sun, J.J. Spivey, Appl. Catal. B: Env., 46 (2003) 601-611.
[7] H. Surangalikar, X. Ouyang, R.S. Besser., Experimental Study of Hydrocarbon
Hydrogenation and Dehydrogenation Reactions in Silicon Microfabricated Reactors of
Two Different Geometries, Chem. Eng. J. 93 (2003) 217-224.
[8] R.S. Besser and W.C. Shin, Deep Reactive Ion Etching Characteristics of a
Micromachined Chemical Reactor, J. Vac. Sci. Technol., B 21 2 (2003) 912-915.
[9] S. Zhao, R.S. Besser, in Proceedings of the 6
th
International Conference on
Microreaction Technology, New Orleans, LA (2002) 289-296.
[10] A. Manasilp, E. Gulari, Appl. Catal. B: Env., 37 (2002) 1725.
[11] M.J. Kahlich, H.A. Gasteiger, R.J. Behm, J. Catal., 171, (1997) 93105.
[12] R.H. Nibbelke, M.A.J. Campman, J.H.B.J. Hoebink, G.B. Marin, J. Catal., 171 (1997)
358-373.
[13] D.H. Kim, M.S. Lim, Appl. Catal. A: Gen., 224 (2002) 27-38.
18
[14] X. Ouyang, P. Ho, H. Chen, W.C. Shin, L. Bednarova, W. Lee, R.S. Besser, S. Pau,
C.S. Pai, J.A. Taylor, W.M. Mansfield, in AIChE 2003 Annual Meeting Proceedings
(2003).
[15] Y. Choi, H.G. Stenger, J. Power Sources, 124 (2003) 432-439.
[16] J. M. Moe, Design of Water-Gas Shift Reactors, Chem. Eng. Pro., 58 (1962), 33-36.
[17] A. Constantinides, N. Mostoufi, Numerical Methods for Chemical Engineers with
MATLAB Applications, Prentice Hall Inc., Upper Saddle River, NJ, 1999, p. 598.
[18] Steady State Nonisothermal Reactor Design, in: H.S. Fogler, Elements of Chemical
Reaction Engineering, third ed., Prentice Hall, Inc., Upper Saddle River, NJ, 2000, pp.
339-425.
[19] Partial Differential Equations, in: S.C. Chapra, R.P. Canale, Numerical Methods for
Engineers, third ed., McGraw-Hill. Inc., 1998, pp. 812-848.
[20] J.D. Hoffman, Numerical Methods for Engineers and Scientists, second ed., Part 3,
McGraw-Hill, Inc., 1992.
Figure Captions
Figure 1. CO conversion vs. temperature: Comparison of the results of microreactor to
results of other PrOx studies in the literature. WHSV: our work 1500 hr
-1
, Kahlich et al.
[2] 1250 hr
-1
, Oh et al. [3] 75 hr
-1
, Choi et al. [4] 200 hr
-1
.
Figure 2. The finite difference grid in radial (r) and axial (l) directions. Each grid point is
denoted as (j, i) with R as the reactor radius and L as the reactor length.
Figure 3. O
2
, CO, H
2
conversion and CO
2
selectivity along the reactor length of 2-mm m-
PBR with T
w
= 120C, 180C, 220C. WHSV = 1500 hr
-1
.
Figure 4. 3D Plots of reactor temperature distribution for the 2-mm m-PBR with T
w
=
120C, 180C, 220C. WHSV = 1500 hr
-1
.
Figure 5. Temperature distribution in the axial direction at r = 0.0 m for the 2-mm m-PBR
at different wall temperatures. WHSV: 1500 hr
-1
.
Figure 6. Temperature distribution in the radial direction at l = l
Tmax
for the 2-mm m-PBR
at different wall temperatures. WHSV: 1500 hr
-1
.
Figure 7. CO conversion vs. different reactor wall temperatures for the microreactor, 2-mm
and 4-mm m-PBRs. WHSV: 1500 hr
-1
.
Figure 8. The average reaction rates for CO oxidation, H
2
oxidation and the r-WGS
reaction at T
w
= 120C, 180C, 220C for the microreactor, 2-mm and 4-mm m-PBRs. All
vertical axes have the same scales, with full scale (1.0) corresponding to 3.9 mol/kg-s.