Ind. Eng. Chem. Res. 1991, 30, 969-978 Phase Cocurrent Trickle Flows through Beds of Small Packings. Ind. Eng. Chem. Process Des. Deu. 1978,18,740-746. Kolb, W. B.; Melli, T. R.; de Santos, J. M.; Scriven, L. E. Cocurrent Downflow in Packed Beds. Flow Regimes and Their Acoustic Signatures. Ind. Eng. Chem. Res. 1990,29,2380-2389. Larkins, R. P.; White, R. R.; Jeffrey, D. Two-Phase Cocurrent Flow in Packed Beds. AIChE J. 1961,7,231-239. Levec, J.; Saez, A. E.; Carbonell, R. G. Holdup and Pressure Drop in Trickle Bed Reactors. Inst. Chem. Eng. Symp. Ser. 1986,87, 185-194. Melli, T. R. Two-Phase Cocurrent Downflow in Packed-Beds; Macroscale from Microscale. Ph.D. Thesis, University of Minnesota, Minneapolis, MN, 1989. Melli, T. R.; de Santos, J. M.; Kolb, W. B.; Scriven, L. E. Cocurrent Downflow in Networks of Passages. Microscale Roots of Macroscale Flow Regimes. Ind. Eng. Chem. Res. 1990,29,2367-2379. Rao, V. G.; Ananth, M. S.; Varma, Y. B. G. Hydrodynamics of Two-Phase Cocurrent Downflow through Packed Beds. Part I. Macroscopic Model. AIChE J. 1983a,29,467-472. Rao, V. G.; Ananth, M. S.; Varma, Y. B. G. Hydrodynamics of Two-Phase Cocurrent Downflow through Packed Beds. Part 11: Experiments and Correlations. AIChE J. 1983b,29, 473-477. Saez, A. E.; Carbonell, R. G. Hydrodynamic Parameters for GasLiquid Cocurrent Flow in Packed Beds. AZChE J. 1985, 31, 52-62. Saez, A. E.; Carbonell, R. G.; Levec, J. The Hydrodynamic of Trickling Flow in Packed Beds. Part I Conduit Models. AIChE J. 1986,32, 353-368.
969
Sherwood, T. K.; Shipley, G. H.; Holloway, G. A. L. Flooding Velocities in Packed Columns. Ind. Eng. Chem. 1938,30,765-779. Stanley, W. D.; Dougherty, G. R.; Dougherty, R. Digital Signal Processing; Reston: 1984,Reston, VA. Stauffer, D. Introduction to Percolation Theory; Taylor and Francis: London, England, 1985. Talmor, E. Two-Phase Downflow through Catalyst Beds. Part I; Flow Maps. AIChE J. 1977,23,868-873. Tosun, G. A Study of Cocurrent Downflow of Nonfoaming GasLiquid Systems in a Packed Bed. 1. Flow Regimes: Search for a Generalized Flow Map. Jnd. Eng. Chem. Process Des Dev. 1984, 23,29-34. Vasiliev, L. L.; Maiorov, V. A. An Analytical Study of Resistance, Heat Transfer and Stability in Evaporative Cooling of a Porous Heat-Producing Element. Int. J. Heat MQSSTransfer 1979,22, 301-306. Westerterp, W. P. M.; Van Swaaij, W. P. M.; Beenackers, A. A. C. M. Chemical Reactor Design and Operation; Wiley: New York, NY, 1984. White, A. M. Pressure Drop and Loading Velocities in Packed Towers. Trans. AIChE 1935,391-397. Zimmerman, S.P.; Chu, C. F.; Ng, K. M. Axial and Radial Dispersion in Trickle-Bed Reactors with Trickling Gas-Liquid Downflow. Chem. Eng. Commun. 1987,50,213-240. Received for review December 27, 1989 Revised manuscript receiued December 4, 1990 Accepted December 14, 1990
Design of Monolith Catalysts for Power Plant NO, Emission Control Jean W. Beeckman and L. Louis Hegedus* W.R. Grace & Go.-Conn., 7379 Route 32, Columbia, Maryland 21044
Research Diuision,
A mathematical model was developed to describe the reduction of NO with NH, in simulated power plant stack gases, over the internal surfaces of monolith-shaped catalysts of the vanadia-titania type. The model indicated that up to 50% improvement in the volumetric activity of these catalysts would be possible if they could be prepared with the computed optimum pore structure. Based on the guidance of the computer calculations, a new type of catalyst was developed. It consists of vanadia, titania, and silica and shows the predicted 50% improvement in activity in laboratory experiments.
Introduction The selective catalytic reduction (SCR) of nitrogen oxides in power plant stack gases is in large-scale commercial practice in Japan and West Germany. Tightening nitrogen oxide emission standards are expected to stimulate the application of SCR processes in the United States as well (Boer et al., 1990). The overall chemical reaction for NO reduction in the SCR process can be expressed as follows: 4N0 + 4NH3 + O2 4N2 + 6H20 The above reaction is equimolar with respect to NO and NH,, and control of NO removal in commercial units can easily be accomplished by injecting the required amount of NHD The SCR process employs either ceramic monoliths or catalyst-coated metal plates as catalytic elements. Typically, 1-1.5 m3 of catalyst is required for each megawatt of electric power generating capacity. A typical coal-fired power plant has a capacity of about 700 electric MW, and the United States has a total fossil-fuel-fired electric power generating capacity of about 530 000 MW. On the basis of recent German experience, the SCR process increases the cost of electricity by about 5%,while providing about 80% NO, removal (Gouker et al., 1989). Approximately 50% of the "levelized" cost (i.e., the sum of operating charges and prorated capital charges) of SCR is catalyst-related (Boer et al., 1990). Therefore, a powerful
-
economic incentive exists to improve the performance of present-day SCR catalysts. Briefly, the process employs vanadia or tungsten-vanadia as active components deposited on porous anatase-type titania which is extruded into the shape of square-channeled monoliths (Ando, 1983). Figure 1shows one such element. The catalyst can be placed either before or after the fly ash precipitators. Typical operating temperatures are in the range of 300-400 "C, and gas hourly space velocities (GHSV) are in the 2000-7000 h-'(STP) range. A typical catalyst block is 70-100 cm long with a face area of 15 cm by 15 cm. The channels are 0.3-1.0 cm wide, and the porous catalyst walls are between 0.05 and 0.15 cm thick. Several outstanding reviews of this technology are available (Ando, 1983, 1985; Bosch and Janssen, 1987). The current SCR process is the result of about 15 years of development, primarily in Japan. The objective of the present authors was to investigate what further potential may exist for catalyst improvements and to identify avenues that would lead to the indicated improvements, if any. For this, we have employed the methods of chemical reaction engineering. Specifically, we have set out to determine the intrinsic kinetics of the key SCR reactions over typical catalytic surfaces, to determine the pore structure of typical SCR catalysts, to determine the effective diffusivity of the key reactant and product species in the porous walls of SCR
0888-5885/91/2630-0969$02.50/00 1991 American Chemical Society
970 Ind. Eng. Chem. Res., Vol. 30, No. 5,1991
Figure 1. Monolith catalyst element for power plant flue gas denitrification.
Table I. Characterization of the Commercial SCR Catalyst Samples Employed for ComDarieon catalyst A
cms/cms cM,cm3/cms ,P 10%cm PM, IO* cm S,10' cm2/cms V20s,wt ?& in V2OS-TiO2 c,
0.43 0.02 140 10OOO 125 0.8
B 0.48 0.02 156 10OOO 123 0.4
catalysts, to quantify the above information in the form of a mathematical model, to use the model to compute NO and SO2 conversions under a wide variety of operating conditions, and finally to compare the computed results with our experimental data base. Once the validity of the model has thus been verified, we set out to compute catalyst performance as a function of the pore size distribution of the catalyst, leaving its intrinsic kinetics (i.e., its surface chemical composition) intact. The results would indicate what potential for catalyst improvement may exist due to pore structure optimization. Lastly, we have set out to prepare catalysts which incorporate the desired pore structures so that the computed performance improvements can be experimentallyverified.
Experimental Section In this study, we have evaluated two commercial catalysts. Their physical and chemical properties are shown in Table I. For intrinsic rate measurements, we employed a differential reactor. It was heated by a fluidized sand bath, and the catalyst temperature was kept constant within 1 "C. The experiments covered the temperature range of commercial interest (280-400 "C); in that range, the catalyst was found to be highly selective toward NO reduction with ammonia versus the competing oxidation of NH3 by 0,; i.e., ammonia was only consumed by NO. The nitrogen oxide content of the gas stream was monitored before and after the catalyst by means of a Model lOAR Thermo Electron NO Analyzer. Care was
taken in the experiments to blend the nitrogendiluted NO and NH3 streams after the air and N2 preheater to avoid background NO reduction and NH3 oxidation. The catmesh powder alyst amount (in the form of a %/BO- (U.S.) or as a slab-shaped cutout of the monolith) and gas flow rate were chosen to obtain conversions between 5 and 20% so that the NO reduction rate could be directly calculated from the difference between the reactor inlet and reactor outlet NO concentrations. Monolith experiments were conducted in an integral reactor into which cored-out sections of monoliths were inserted; the pieces were surrounded by insulation. The length of the nine-channel monoliths employed was 15 cm, while the channel opening and wall thickness were 0.60 and 0.13 cm, respectively. Five thermocouples were installed at various points between the inlet and outlet faces of the monolith; the maximum deviation among these thermocouples was less than 5 "C. Catalyst surface areas were obtained via BET nitrogen adsorption using the Model MS-13 Quantachrome instrument. Mercury pore size distributions were obtained with the Micromeritics Autopore 9200 ultra-high-pressure mercury porosimeter. Catalyst poisoning experiments were carried out in a reactor similar to the one described by Vogel et al. (1988). A small boat containing b o 3powder was mounted in the furnace section upstream of a nine-channel monolith piece. Nitrogen sweep gas was then metered to the reactor and carried the subliming A e 0 3 to the catalyst section where it deposited on the catalyst's internal surfaces. The catalyst was kept at 350 "C, while the As203 was kept at 250 "C by appropriately positioning it in the tubular furnace. After a 1-h treatment, the monolith was removed from the reactor and analyzed for the internal arsenic distribution by means of a Cameca Camebax electron microprobe.
Mathematical Model and Experimental Verification In the experiments described in this section, H20- and S02-free feedstreams were used. The objective was to establish our ability to quantitatively describe the kinetics
,
Ind. Eng. Chem. Res., Vol. 30, No. 5, 1991 971
.
" 2
4
6
8
1
0
Figure 2. Dependence of NO reduction rate on NO concentration. Differential reactor: 50/80-mesh-sized catalyst A; T = 350 O C ; C,, = 1.95 X 10" mol/cm3. 2 .
0
I
1
1
observed
0
4
/
2
3
4
5
SH, c o n c en t r a t i o n (no1 /cm3) x 1 O8
Figure 3. NO reduction rate dependence on NH3 concentration. Differential reactor: 50/80-mesh-sized catalyst A; T = 350 O C ; CNO = 1.95 X 10" mol/cm3.
and pore diffusion in this reaction system. Subsequently, we have employed the parameter values to match our data base in a "wetw(i.e., SOz-and H20-containing)feedstream; catalyst optimization studies were then conducted with those "wet" parameters. Intrinsic Kinetics. The dependence of the intrinsic NO reduction rate on the concentration of NO and NH, was determined in the differential reactor on catalyst A, Table I. Figure 2 shows that the NO reduction rate is first-order in NO, while Figure 3 reveals that it is approximately zero-order in NH, down to about 200 ppm. Accordingly, Figure 4 shows that the NO conversion was not affected by the inlet NH, concentration in the integral reactor for an NH, to NO molar inlet ratio of unity. In order to cover a broad range of commercially significant operating conditions, we have decided to use a Langmuir-Hinshelwood-type rate expression. The intrinsic NO reduction rate was then expressed as
where RNO is the intrinsic NO reduction rate expressed on a unit catalyst surface area basis. The Langmuir-Hinshelwood-type rate behavior is in agreement with NO reduction rate studies by Odenbrand et al. (1985), while Wong (1982) and Kittrell and Eldridge (1985) found simple zero-order behavior in NH3 and first-order behavior in NO, which are of course degenerate cases of Langmuir-Hinshelwood kinetics. Catalysts employed in those studies were V205/Si02,titania-supported metal oxides, and a vanadia-coated FeCr screen catalyst, respectively. The rate of oxidation of SO2to SO, in commercial SCR reactors is very slow, with only a few percent conversion at typical reactor conditions. The accurate quantification of this side reaction is important since depending on the concentrations of NH, and SO,, the formation and de-
8
NH, c o n c e n t r a t i on (mo1 /cm3) x 1 Og
NO c o n c e n t r a t i o n (mol/~n~)xl0~
"
20
I
/
0
I
100
Figure 4. Dependence of NO conversion on NH3 concentration. Integral reactor; monolith-shaped catalyst A: T = 300 O C ; GHSV at STP = 7000 hr-l; NH3/N0 = 1.0.
position of ammonium bisulfate, (NH4)HS04,can occur at the lower temperatures prevailing in the heat exchangers downstream of the SCR reactor (Bosch and Janssen, 1987). The rate equation for SO2oxidation over base metal oxides is first-order in SO2 and zero-order in O2 due to the large excess of the latter (up to about 4 vol %). The energy of activation and the surface area specific preexponential factor were determined from integral reactor measurements over the actual catalyst at hand. The SO2oxidation reaction was found to be kinetically controlled; i.e., it is not affected by pore diffusion. Therefore, its catalystvolume-specific rate is proportional to the volume-specific surface area of the catalyst. Pore Diffusion. Beeckman (1991) developed a technique to measure the diffusivity of NO in the pores of an SCR catalyst monolith. It involves the passage of gas streams inside and outside of a particular monolith channel and measuring the net flux across its walls. Such measurements, when compared with the predictions of the random pore model of Wakao and Smith (1962), indicated that the effective diffusivity can be well determined (&lo%)from inputs of Hg porosimetry measurements. The random pore model, then, allows one to calculate effective diffusivities as functions of parameters that can be measured via Hg porosimetry. Therefore, the random pore model is suitable for the design and optimization of supported catalysts, once a suitable mathematical model for catalyst performance has been constructed. The methodology has been described by Hegedus (1980) and applied, e.g., by Pereira et al. (1988). According to the random pore model, the design parameters of a porous catalyst structure are the micro- and macroporosities (expressed as volume fractions) and the volume-integral-averaged micro- and macropore radii. The molecular diffusion coefficients of NO and NH3 were estimated via standard methods. External Mass Transfer. The external mass transfer coefficient for developing laminar flow in rectangular ducts is given by the Hawthorn correlation (e.g., Hegedus (1973)):
where Rh, the hydraulic radius, is defined as twice the channel open area divided by its wetted perimeter. Re is the Reynolds number Re = 2GRh/p (3) and
is the Schmidt number SCNO = p/PDNO (4) The parameters in (2)-(4) are defied in the Nomenclature section. SCNO
972 Ind. Eng. Chem. Res., Vol. 30, No. 5, 1991
Mat hematical Model Equations. Experiments in the differential reactor were performed on crushed 50/80-mesh catalyst as well as on catalyst slabs in order to clearly show the impact of pore diffusion limitations. Effectiveness factors were calculated analytically. Experiments in the integral reactor involved monolith catalyst elements. At any axial position x in the monolith, a material balance over a slab of width dy perpendicular to the axial direction yields the following equation:
-25
I
-a - 2 9 . -a +
?.-
with boundary conditions dCNo/dy = 0 C,,
for y = 0 fory = W/2
= Cho
(5)
(6)
(7)
-31' ' 0.0016
- 'NO)
= De,NH,(ChH,
- 'NHS)
'
'
' ' 0.0018 1/T( "K
'
'
' '
0.0020
1
Figure 5. Measured and predicted NO reduction rates on powders and slabs. Differential reactor; catalyst A; NO = loo0 ppm; NHs loo0 ppm; O2= 4 % .
A similar set of equations can be written down for NH3 and yield after combination with (5) and subsequent double integration the following relationship: De,NO(chO
predicted
-
I
De,NO d2CNo/dY2 = SRNO
0.125 em s l a b 50/80 Mesh powder
A
I
&-
(8)
/A
A material balance at axial position x of the reactor over the external gas film yields
T=380°C,GHSV= 8 , 6 0 0 h - ' T=341°C,GHSV=12,000h-1 A T=281aCC,GHSV=12,000h~' -- p r e d i c t e d 0
The molecular fluxes for NH, and NO over the external gas film are equal because of the equimolarity of the SCR reaction and yield
0.50 0.75 1.00 1.25 NH,/NO m o l a r i n l e t r a t i o
Figure 6. Measured and predicted NO conversions versus NH,/NO molar inlet ratio. Integral reactor; monolith-shaped catalyst A; NO = loo0 ppm; O2 = 4 % .
Assuming plug flow, i.e., mixing-cup average properties across the bulk gas phase, a material balance in the fluid field over a differential distance dx in the monolith yields
100
r
2o
\
with boundary conditions: 0
-
A
-
Similar equations can be written down for NH3 and yield after combination with (11) and subsequent integration the following relationship:
The solution of (5)-(13) allows prediction of the NO conversion in an integral reactor; the equations were solved numerically.
Results and Discussion The mathematical model has three adjustable parameters, all related to the intrinsic rate equation, namely, the preexponential factor k,, the activation energy E , and the NH3 adsorption constant KNHs. For both cataysts A and B, the NH3 adsorption constant could not be estimated in a statistically significant way because of the lack of accurate data at very low NH3 concentrations. This is consistent with the observed zero-order behavior. The best fit of the differential- and integral-reactor data for catalyst A was obtained with a preexponential factor k , = 8.64 x lo3 cm/s and an activation energy of 19 kcal/mol. The activation energy is at the upper end of a rather wide range of values reported in the literature (Weidner and Hofmann, 1989; Bosch and Janssen, 1987; Wong, 1982). Figure 5 shows the model fit of the differential reactor data for catalyst crushed to 50/80-mesh
0
275
"
observed predicted "
300
'
"
325
'
350 T ('CC)
'
,
,
I
' i
375 4 ' 0
Figure 7. Measured and predicted NO conversions versus reactor temperature. Integral reactor; monolith-shaped catalyst A; NO = loo0 ppm; NH, = loo0 ppm; O2= 4%; GHSV at STP = 8600 h-l.
particles, as well as for catalyst slabs cut from the monolith. The model fit is quite good and shows clearly the strong pore diffusion effects on the observed rates for the slabs (Shikada et al., 1983). Figures 6 and 7 show the model predictions for the NO conversion in the integral reactor as a function of the NH3/N0 molar inlet ratio and reactor temperature, respectively. From the above, it can be observed that the proposed mathematical model adequately describes the NO reduction performance of the catalyst. This same model was then also applied to catalyst B in "wet" reactor conditions yielding a preexponential factor of 3.51 X lo2 cm/s and an activation energy of 14 kcal/mol. The above differences in the kinetic parameters for catalysts A and B are due in part to the differences in vanadia levels and due to the presence of SO2 (800 ppm) and H20 (10 vol % ) in the feedstream. The predictive powers of the model, under wet conditions, are shown in Figure 8. It is well-known that SO2and its oxidation product SO3 alter the catalyst activity over time until an equilibrium activity is established. The rate of oxidation of SO2to SO3
Ind. Eng. Chem. Res., Vol. 30, No. 5, 1991 973 100,
Table 11. Reactor Conditions Used for Catalyst Optimization NO, ppm 400 NH3/N0 molar inlet ratio 1 so29 PPm 800 0 2 , vol % 4 HzO, V O ~% 10 GHSV at STP, h-I 6700
r
a
T,O C
t
2o [ 275
300
325
350
375
400
T (“1 Figure 8. Measured and predicted NO conversions versus reactor temperature. Integral reactor; monolith-shaped catalyst B; NO = 400 ppm; NHS = 400 ppm; SO2 = lo00 ppm; O2 = 4%; H20 = 10%. -16
-
t
:-11
.6-.
Svl91 ,
0
,
,
,
,
observed
- predicted -20 0.0013
350
observed
A
0.0015
1/T
‘,
1 0.0017
(OK)
Figure 9. SO2 oxidation activity as a function of reactor temperature. Integral reactor; monolith-shaped catalyst B; NO = 400 ppm; NH3 = 400 ppm; SO2 = lo00 ppm; O2 = 4%; H20 = 10%.
was measured in *wetnconditions on catalyst B. The SO3 at the exit of the reactor was measured by standard titration methods. The conversion of SO2 at SCR conditions is very low (
