Ind. Eng. Chem. Prod. Res. Dev. 1986, 25,424-430
424
(Muraki et al., 1985): the activity of Pd catalyst is increased under the cycling feed due to the periodic operation effect (Muraki et al., 1986).
Grill, C. M.; Gonzalez, R. D. J. Phys. Chem. 1980, 84, 878. Hecker, W. C.;Bell, A. T. J. Catal. 1983, 84, 200. Kobylinski. T. P.; Taylor, B. W. J. Catal. 1974, 33, 376. Kummer, J. T. Prog. Energy Combust. Sci. 1980, 6, 177. Lorimer, D.; Bell, A. T. J. Catal. 1979, 59, 223. Muraki, H.; Shinjoh, H.; Sobukawa, H.; Yokota, K.; Fujitani, Y. Ind. Eng. Chem. Prod. Res. Dev. 1985, 24, 43. Muraki, H.; Fujitani, Y. Ind. Eng. Chem. Prod. Res. Dev. 1988, preceding paper in this issue. Schlatter. J. C.; Taylor, K. C. J . Catal. 1977, 49,42. Shelef, M.; Gandhi, H. Ind. €ng. Chem. Prod. Res. Dew. 1972, 7 1 . 393. Taylor, K. C.;Schlatter, J. C. J. Catal. 1980, 63, 53. Yokota, K.; Muraki, H.: Fujitani. Y. Presented at the Society of Automotive Engineers Congress, Detroit, MI, March 1985; paper 850129.
Acknowledgment We are grateful to Professor Y. Murakami and Associate Professor T. Hattori, of Nagoya University, for their helpful suggestions. Registry No. NO, 10102-43-9; CO, 630-08-0; Pd, 7440-05-3. Literature Cited Butler, J. D.: Davis, D. R. J. Chem. SOC., Dalton Trans. 1978, 27, 2249. Campbell, C. T.: White, J. M. Appl. Surf. Sci. 1978, 7 , 347. Dalla Betta, R. A.; Shelef, M. J. Mol. Catal. 1976, 1 , 431. Dubois, L. H.; Hansma, P. K.: Somorjai, G. A. J. Catal. 1980, 65, 318.
Received for review September 9, 1985 Revised manuscript received February 10, 1986 Accepted March 12, 1986
Comparison of Some Solid Catalysts for the Production of Ethanolamines from Ammonia and Ethylene Oxide in the Liquid Phase Lennart Vamling' and Lennart Cider Department of Chemical Reaction Englneering, Chalmers University of Technology, S-4 12 96 Goteborg, Sweden
The ability of different forms of zeolites, such as 13X, 4A, Y, and AW500, to catalyze the formation of mono-, di-, and triethanolamines from ethylene oxide and ammonia in the liquid phase has been examined and compared with that of an organic ion-exchange resin. The ion-exchange resin, Amberlite 200, gives the highest yield of monoethanolamine (MEA), while the zeolite 13X has the highest capacity of producing MEA. Furthermore, it is demonstrated that any contrlbutlons from homogeneous reactions to the total reaction rate are negligible. A fast and highly reproducible method for the direct GC analysis of ethanolamines using fused-silica capillary columns is also presented.
RpNHB-p + HA
Introduction Mono-, di-, and triethanolamines (MEA, DEA, and TEA) are produced by reacting ethylene oxide (EO) with ammonia (NH,) according to NH3 + C2H40 RNH2 NH, EO MEA
CH CHz
CHzCHz
-
'd
RNH2 + C2H40 R2NH MEA EO DEA RzNH C2H40 R,N (111) DEA EO TEA where R represents the group HOCH2CH2-. As already pointed out by Knorr (1899), no reaction occurs by just mixing NH, and EO. Traditionally, water has been added to promote the reactions. The kinetics for this alternative have been studied by Potter and McLaughlin (1947) and by Miki et al. (1966)among others. The water must be separated from the products, an energy-consuming disadvantage. This has led to an increasing interest in anhydrous methods using heterogeneous catalysts. Weibull et al. (1957, 1973) have shown that it is possible to use ion-exchange resins. One plausible reaction mechanism explaining the catalytic effect of these resins is the following, given by Weibull (1970). First, an NH,, MEA, DEA, or TEA molecule (represented by R,NH,, where p = Ck3 and R is the same group as above) is adsorbed a t an acidic site, in this case a sulfonic acid group.
-
0196-4321/86/1225-0424$01.50/0
RPH4-,N**A
reversible adsorption step Second, the oxygen atom in the EO molecule is attracted by the adsorbed molecule.
-+
+
2
+
RPH4.,N..A
C
/7
O*.H4-,RpN..A
reversible adsorption step
This weakens the C-0 bond in the EO molecule and makes it possible for the nitrogen atom in another NH,, MEA, or DEA molecule (R,NH,, where q = 0-2) to make a successful nucleophilic attack, which causes the ring to break. CH CH,
R,NH3-,
+ $7 **H,-,R,N.*A
R,+lH3-,N*.A
+
RpNH3-,
reaction step
This led us to the idea that other sources of acidity could also be used, perhaps ones having higher activity or higher yield of MEA. A potential disadvantage with organic ion-exchange resins is their lack of stability at high temperatures. The requirements of having a high-temperature stability and of containing acidic sites are met by many zeolites. We therefore decided to investigate four different types, one with small cages (4A), one with medium-sized cages (AW500), and two with large cages (13X and Y). For 0
1986 American Chemical Society
Ind. Eng. Chem. Prod. Res. Dev., Vol. 25, No. 3, 1986 EO
I
Table I. Conditions Used for t h e GC Analysis of t h e Ethanolamines GC column carrier gas column inlet pressure solvent internal standard injection volume split ratio detector injector temp detector temp
Figure 1. Flow scheme.
comparison, measurements with an acidic organic macroporous ion-exchanger (Amberlite 200) were also made. The idea of using zeolites and other inorganic catalysts is also found in a recent patent by Johnson (1984). Stoyanov et al. (1983) have studied the kinetics for a strongly acidic gel type ion exchanger (Wofatit KPS). They compared its activity and selectivity with that of water and proposed a reaction mechanism similar to the one described above. The main purpose of this work is to predict the product distribution and the capacity of a full-scale adiabatic packed-bed reactor for the different catalysts. Consequently, we used a well-insulated packed-bed reactor without internal cooling or heating for our laboratory experiments. We also restricted our operating region to that of practical interest, using nearly complete conversions of EO and a large excess of NH3.
Experimental Section Equipment. A flow scheme of the equipment is presented in Figure 1. A metering pump with two heads was used to obtain the desired mixture and flow. The feeds consisted of 99.6% pure EO and 99.96% pure NH3 as delivered by AGA, Sweden. To avoid formation of gas bubbles in the pump heads during the suction phase, we pressurized the storage cylinders. This was done by filling the EO cylinder with He to about 0.5 MPa and by heating the NH, cylinders to about 313 K. To o b h n a nearly pulsation-free flow, a diaphragm-type pulse damper was mounted directly after the pump. Before entering the reactor, the mixture was preheated by leading it through a tube coil, which in turn was heated with oil from a thermostat bath. The reactor consisted of a 1-m-long Type-316 stainless steel tube with i.d. of 24.3 mm and 0.d. of 33.4 mm. It was insulated with 50-60 mm of ceramic wool. Type K thermocouples were used to measure the temperature at 12 different locations. The first 83 mm of the reactor was fiied with glass beads of about the same size as the catalyst particles, followed by 830 mm packed with catalyst. The size and form of the catalyst pellets are given in Table 11. The rest of the reactor was filled with glass beads up to the outlet. After the reactor was a regulating valve that, together with a pressure transmitter and a PI controller, was used to keep the reactor pressure at the desired level. A pressure of about 13 MPa was used t o avoid vaporization of the reaction mixture. Finally, most of the remaining NH3 was stripped off in a distillation column. More details of the experimental setup can be found in Vamling (1986). Measurements. The EO storage cylinder was placed on a balance. The mass flow was obtained by measuring
425
temp, "C programming sample concn, g/L internal standard concn, g/L
Varian 3700 with autoinjector SGE BPI 25 m fused silica, i d . 0.22 mm H2 11 psi ethanol N-methylpiperazin 0.8 PL adjusted to maximum column capacity FID 300 "C 300 "C procedure 1 Drocedure 2 70 25 190 115 25 190 1 min min-' 2 min 1 min mi+ 2 min 6.4 80 1.6 20
the time required for a 50-g weight change. This gave an average value over a period of 15-30 min. The uncertainty was estimated to be about 4%, a figure based on an overestimate of the minimum detectable weight change. The total mass flow was obtained by directing the reactor output to a steel cylinder placed on another balance. Here also the time for a 50-g weight change was measured, giving a measurement period of 1-3 min. In this case the uncertainty was estimated to be about 1% , but we observed differences between successive readings as high as 10% due to flow irregularities. The reaction products, after the remaining NH, was stripped off, were sampled in 25-cm3E flasks, for a period of 10-30 min and then weighed. The product distribution was obtained by direct GC analysis as described below. GC Analysis. For the determination of the amounts of the ethanolamines, we sought an accurate, simple, fast method and found direct GC analysis to be the most attractive alternative. A problem with such methods is that ethanolamines have both electron-donor and H-donor properties, which causes excessive tailing on many packings. In the literature we found three methods for this analysis. First we used a column packed with Tenax, a porous polymer, as suggested by Saha et al. (1977). This method gave an excellent separation of the ethanolamines with sharp and narrow peaks, but the reproducibility was unsatisfactory. The relative standard deviation for successive injections of the same sample was over 4%. A somewhat better reproducibility was achieved when 15% Carbowax 20M on 80-100 mesh Porapak Q was used, as suggested by Boneva and Dimov (1979). We had, however, some problems with the detection of small amounts of TEA since its peak was quite broad. The column also had a relatively short lifetime since it had to be operated at its temperature limit. The paper of Boneva and Dimov (1981) drew our attention to the possibility of using capillary columns. We therefore tested three fused-silica columns with bonded phases having different polarities. We could not use the highly polar stationary phase (poly(ethy1ene glycol)) since TEA was not desorbed when the column was operated below the temperature limit. Both the nonpolar (methylsilicon) and the medium polar (cyanopropylsilicon) phases were found to have acceptable properties. We chose the nonpolar because it gave a somewhat shorter analysis time. The procedures used for our samples are presented in Table I. The reproducibility with procedure 1was very good, with relative standard deviations less than 1% . The
Ind. Eng. Chem. Prod. Res. Dev., Vol. 25, No. 3, 1986
420
Table 11. Catalysts Used symbol A200 13X 4A AW500 Y
OJ4O
name
delivd or manufd by Serva EKA AB EKA AB Union Carbide Linde
Amberlite 200 Ekasorb 13X Ekasorb 4A Chabasite LZ-Y62
t
*
0.035
form beads 20-50 mesh H beads Na beads Na extr l/g in. mixed extr in. H
radius/” mean 0.27 0.83 0.66 1.56 0.79
length/” mean
SD 0.09 0.14 0.05 0.06 0.01
4.0 3.2
SD
2.3 10
directly based on the mechanism presented in the Introduction; the main reason we have chosen them is that they fit our results well. Vamling (1986) has shown that using more advanced models containing additional parameters does not lead to any significant improvements. Note also that this model assumes the three reactions have the same activation energy, which is plausible considering they all include a breakage of the EO ring. Reactor Modeling. Neglecting dispersion, the reactor can be modeled as
+
0 0101 0.015
0.020
0.025
0.030
0035
0.040 0.045
0.050
0.055 YE,-%
Figure 2. Yield of MEA for the various catalysts: A200 (+), 13X (A),4A ( O ) , AW500 ( O ) ,and Y ( 0 ) . Ob
response factors showed a slow drift, so frequent calibrations were necessary. When the column had been used for a while, the detection limit for TEA increased. An alternative is procedure 2, showing a slightly lower reproducibility but having a lower detection limit for TEA (between 0.5% and 1% weight) and showing a somewhat slower drift. The calibration curves were linear but not directly proportional so two or three calibration points had to be used. Catalysts. The catalysts used are presented in Table 11. Amberlite 200 is a macroporous strongly acidic ion exchanger made of poly(styrene-divinylbenzene) with sulfonic acid groups, and the others are different forms of zeolites. Before being loaded into the reactor, the zeolites 13X, AW500, and 4A were pretreated by washing them with a 10% solution of ammonium nitrate at 343 K for a period of 8 h. This ion exchange was not needed for the other catalysts since they were delivered in suitable form.
Theoretical Section Reaction Modeling. It is quite possible to derive a reaction rate model from the reaction mechanism presented in the Introduction (Vamling, 1986), but it will contain a large number of parameters, and they are not easy to determine separately. Since our goal was not to perform a detailed kinetic investigation, we used a model as simple as possible, with as few parameters as possible:
ki = ki(Tref)exp(
r , = klcNcE
(1)
r2 =
~ ~ C M C E
(2)
r3 =
~BCDCE
(3)
L( %)) RTref
1-
i
= 1-3
(4)
The symbols used are explained in Nomenclature. These rate expressions have previously been used for the corresponding homogeneous reactions. Note that they are not
=
(8)
Tb/Tref
with the following boundary conditions at z* = 0 YE
=
3)Ef
YM
=
YD =
Ob
=
Tbf/Tref
In a laboratory reactor, wall effects have greater influence than in a large-scale reactor. Since the intention is to use the reactor model for scale-up, we have to consider such effects. We have therefore included the following equation for the wall temperature:
Since the flanges contribute about half of the reactor’s outer area, we must take this into consideration when formulating the boundary conditions for the wall temperature:
&AW d8, = hfl’Afl(Ow- 8,) at z* = 1 (11) L dz* To calculate the observed reaction rates, rob,, we would normally have to solve equations describing the mass transfer in the pellets for each grid point used in the solution of the reactor equations. Since our main interest is to obtain a model suitable for scale-up, using the same pellet sizes as in the laboratory experiments, we can model the observed reaction rates instead of the “true” surface reaction rates. We tested this model of observed reaction rates in the followingway. First we simulated a reactor with significant mass-transfer resistance for a set of inlet temperatures and concentrations, using the same form of rate expressions as described above. Thereafter we performed a regression on the obtained simulation results using our model. We found that the model was able to describe more than 99.9% of the variations in our simulated data. The reason for this success is mainly that the dominating first reaction is pseudo first order with respect to EO because of the large excess of NH,.
Ind. Eng. Chem. Prod. Res. Dev., Vol. 25, No. 3, 1986
Solution Methods. The reactor equations were solved by using COLSYS, a subroutine package for numerical solution of multipoint boundary problems described by Ascher et al. (1981). The pellet equations were solved simultaneously by using orthogonal collocation as described in Villadsen and Michelsen (1978) for each set of bulk concentrations and temperature. The required observed reaction rates were found by Gaussian quadrature. Results The measured data are given in Table 111. From these data, the values in Table IV were calculated; a possible small EO leakage in the pumps was neglected. When the MEA production values were calculated, an estimate of the amount of MEA leaving the stripping column together with the NH3 flow is included. These corrections are about 1%. The runs presented in the tables were selected from a larger set. Runs showing instabilities in flow or temperature have been excluded. The rate constants and the activation energy for the different catalysts have been determined by nonlinear regression using the following stepwise procedure. If the observed reaction rates have the same EO concentration dependence, as assumed in our model, we can start by determining the values of k 2 / k l and k 3 / k l according to the following procedure. By combining eq 1-3 with eq 5, we obtain dYM = rl - r2 = -1 + k2YM (12) dYN -r1 klYN k2YM k?YD -dY =D- = - r-2 - r 3 (13) dYN -r1 klYN klYN From these equations we see that we can determine k z / k l and k 3 / k l using only product distributions as dependent variables. The values of the ratios determined in this way are given in Table V. We have also determined these ratios using data from other sources. For heterogeneous catalysts they are included in Table V, and for water they are given in Table VI. In the patent by Johnson (1984) only the area percentage from a GC analysis is presented. By assuming that this area is proportional to the number of carbon atoms in the molecule, we were able to make a rough estimate of the product distribution and of the desired ratios. After having fixed the rate constant ratios, we can proceed to determine the remaining parameters in our model, using temperatures and MEA production as dependent variables. The results are given in Table VII. To do this we also need estimates of the physical properties of the mixture and their temperature and composition dependence. Since there are no data available for the mixture, we have tried to estimate them from known data of the pure substances. However, use of these estimates for heat capacity and for the reaction enthalpy gave an energy balance incompatible with our temperature measurements. Because of this, we had to include an equation with adjustable parameters for AHlC, in our regression. The parameters obtained for this equation are also given in Table VII. While there therefore may be some doubt concerning the absolute values of the activation energies given, they are still useful for comparisons between our different catalysts. More details about the estimates used are given by Vamling (1986).
+-
Discussion Yield. A convenient way of comparing the yield of MEA for the different catalysts is to plot the MEA production
427
vs. the EO consumption. This is done in Figure 2, where all of our measurements are included. From this plot we can see directly that A200 has the highest yield, followed by AW500,4A, Y, and 13X in decreasing order. Another way is to look at the rate constant ratio k 2 / k l . We can do this since k 3 / k l has only a minor influence in our operating region. The lower k z / k l is, the higher is the yield of MEA. One question that has to be answered is to what extent the differences in yield between the catalysts are due to differences in mass transfer. To check this, we simulated a reactor using the simple reaction rate model, but including mass transfer. When calculating the true reaction rate, we used k z / k l = 4.65, k , / k l = 6, and cN,fklRq2/DNveff = 10 000, which means a severe mass-transfer resistance. Using the same procedure as described above, we then obtained k z / k l = 6.9 and k 3 / k l = 12.6. In these calculations we have assumed that the pellets can be considered as isothermal. This assumption can be checked by calculating the maximum temperature rise within a pellet according to AT- = DE,e~E,s(-AH)/Xefi, which in our case is estimated to be less than 0.5 K, which means that our assumption is not unreasonable. Since the ratios used for the true reaction rate are the highest possible and since the results change very little with increasing mass-transfer resistance, we therefore conclude that differences in mass transfer cannot be the only cause of the observed differences in yield. In these calculations, we have assumed that the ratios of the diffusion coefficients are the same as for bulk diffusion, and this might not be true in the narrow pores of the zeolites. Other possible reasons include differences in acidity of the reaction sites, and differences in steric hindrances to formation of activated complexes. Capacity. The capacities of the catalysts can be compared in a number of ways. For isothermal operation, we can simply compare the rate constant k l at the desired operating temperature, since the capacity is directly proportional to the rate constant. As an example, the values at 375 K are given in Table VII. Here, 13X is clearly the most active catalyst, followed Y, A200,4A, and AW500 in decreasing order. For adiabatic operation, the influence of the different activation energies has to be included. In this case, we chose to compare the amount that can be produced or consumed for a certain time and reactor volume, when the EO conversion is restricted to 99.9% and the outlet temperature to 400 K. In Figure 3 we have compared the capacity of consumption of EO, and in Figure 4 the capacity of production of MEA. These plots show that we can divide our catalysts into three groups: high capacity (13X), medium capacity (A200 and Y), and low capacity (4A and AW500). The required inlet temperatures to reach 400 K at the outlet are given in Figure 5. Stability. In the course of our investigations, the activity of the A200 has been fairly constant. This is also in accordance with industrial experience. The Y zeolite has lost about half of its activity during our testing period. This, of course, influences the quality of our estimates of its rate constants and activation energies. For the other zeolites, we have not performed enough experiments to make any definitive conclusions regarding their stability. From this we conclude that the organic ion-exchanger A200 is at least as stable as the zeolites, at least up to about 400 K. Working with higher temperatures means entering the supercritical region, so that we must work with even higher pressures if our mixture is to remain dense. Contribution from Homogeneous Reactions. To test whether there were any contributions from homogeneous reactions to the total reaction rate, we replaced part of the
428
Ind. Eng. Chem. Prod. Res. Dev., Vol. 25, No. 3, 1986
f
-s c
a c
Ind. Eng. Chem. Prod. Res. Dev., Vol. 25, No. 3, 1986 429
Table IV. Calculated Results for Selected Runs production, @molls catalyst A200 A200 A200 A200 A200 A200 13X 13X 13X 13X 13X 4A 4A 4A 4A AW500 AW500 AW500 AW500 AW500 AW500
ID
CN,f/CE,f
525 602 604 608 610 658 3 6 8
50.5 33.5 33.3 24.4 36.3 29.9 25.5 31.0 25.7 20.7 24.0 28.2 22.9 23.9 23.9 21.0 20.5 21.0 29.8 30.0 37.3 29.5 29.8 30.2 38.0 36.6 37.4 42.2
11
14 3 4 14 18 3 5 7 10 12 14 2 3 5 6 7 14 16
Y Y Y Y Y Y Y
convn of EO 1.035 0.965 0.967 0.985 0.940 0.935 0.991 1.011 1.015 1.082 1.002 0.902 0.876 0.989 0.916 0.775 0.805 0.741 0.839 0.861 0.724 0.957 0.972 0.991 0.995 0.952 0.840 0.962
Table V. Ratios of Rate Constants for Some Solid Catalystsa catalvst source k2lh kdki A200 4.34 f 0.49 this work 13X 9.4 f 8.2 13.5 1.1 4A 8.5 i 1.6 8.7 2.3 AW500 6.03 f 0.23 7.8 i 4.9 AW500 5.4 6.0 Johnson (1984) 10.7 0.72 15.9 f 10.0 this work Y 6.4 12.1 Si-A1 Johnson (1984) 18.9 12.6 Wofatit KPS Stoyanov et al. (1983) 4 Dowex 50WX8 5 Weibull et al. (1973)
*
'The limits given are based on a confidence level of approximately 95%. Capacity kg/(s m3)
18.9 15.4 28.1 29.0 12.3 10.9 14.0 5.6 4.7 6.4 6.0 3.2 4.6 4.8 3.2 15.4 21.5 20.9 11.5 11.2 7.7 7.1
temp rise, K 19.4 34.3 34.7 45.9 32.3 37.1 45.0 37.3 50.8 64.1 54.2 34.6 44.0 38.7 41.6 38.4 40.5 36.1 25.4 23.8 17.8 41.2 38.4 38.8 31.4 31.5 28.4 24.9
kg/(s m3)
0.35
1
020
t
O.O5I
--- - - - - - -- - - - ---- - --0.060
0.070
0.080
0.090
0.100
0.110
Fraction EO w / w
_----
0.25 /
Figure 4. Capacity of production of MEA under adiabatic conditions, with an outlet temperature of 400 K and an EO conversion of 99.9%, for the various catalysts: A200 (-1, 13X (--), 4A (--), AW500 and Y (---). (-e-),
, ,,
Tin/K
0 00 0.050
TEA
Capacity
0.050
t
0.05
DEA 27.0 39.5 43.3 46.9 45.3 50.8 161.2 159.7 230.7 229.6 168.4 107.4 107.3 82.6 73.3 66.8 62.5 47.6 41.8 39.6 26.2 127.0 142.1 142.3 80.5 80.2 56.7 52.9
*
*
0.40
MEA 662.5 637.2 613.7 599.4 707.3 718.1 691.7 745.3 1008.5 878.5 763.7 769.6 675.4 608.2 571.4 655.1 606.2 503.9 580.6 561.1 473.9 1007.1 1004.3 990.5 620.1 631.0 467.9 480.1
380
1
360
-
355
-
350
-
._ ._ ._._._._._._. _. _. _. _. _. . . . . . . . 0.060
0.070
0.080
0.090
0.100
0.110
Fraction EO w / w
Figure 3. Capacity of consumption of EO under adiabatic conditions, with an outlet temperature of 400 K and an EO conversion of 99.9%, for the various catalysts: A200 (-1, 13X (--), 4A (--), AW500 and Y (---).
345 -
(-e-),
catalyst in the reactor with glass beads. The temperature profile in the reactor in this experiment is given in Figure 6. From this it is obvious that the effect of any such contribution is negligible compared with the contribution
0.050
0.060
0.070
0.080
0.090
0,100
0.110
Fraction EO w/w
Figure 5. Required inlet temperatures to achieve 400 K at outlet under adiabatic conditions.
Ind. Eng. Chem. Prod. Res. Dev., Vol. 25, No. 3, 1986
430
Table VI. Ratios of Rate Constants as a Function of H,O Concentration for the Homogeneously Catalyzed Reactions concn of HzO, kz/k, k3/k, source kmol/m3 6 4 Potter and McLaughlin (1947) 18-30 40 5.5 7 Miki et al. (1966) 4 7 15 0.693 40 14 Stoyanov et al. (1983) 0.411 93 81 0.215 105 100 0 14 40 Johnson (1984)
-
Table VII. Rate Constant k,and Activation Energiesn k , X lo7 activation m3/(mol s) energy, catalyst a t 375 K kJ/mol A200 2.6 0.2 95 f 8 13X 6.8 f 0.6 67 f 4 4A 1.43 f 0.08 50 f 10 AW500 0.76 f 0.04 55 f 6 Y 3.4 f 0.2 60 f 9 64 Wofatit KPS (Stoyanov et al., 1983) 61 HzO (Stoyanov et al., 1983) 46 H,O (Miki et al., 1966) 69 H,O (Potter and McLaughlin, 1947) Expression for C, Value exuression
units
kgjmol K K/Pa
Cp,,(13 MPa) = 424.6 f 7.4 dCp,cdP= 5.3 X lo4 f 3.8 X lo4
"The limits given are based on a confidence level of approximately 95%.
Acknowledgment We are grateful to the Swedish Board of Technical Development (STU) and to Berol AB for their financial support of part of this work. Nomenclature ci = concentration of substance i, mol/m3 ni = concentration of substance i, mol/kg(fluid) y i = concentration of substance i, dimensionless, defined by eq 6 T = temperature, K Tref= reference temperature, 375 K L = length of reactor bed, m z* = reactor coordinate, dimensionless (=t/L) G = mass flux in reactor, kg/(m2 s) R i= radius, m R = gas constant, J/(mol K) E = activation energy, J/mol rj = reaction rate for jth reaction, mol/(s m3) k i = ith rate constant, m3/(mol s) kfl' = heat-transfer coefficient between flanges and room, W/(m2 K) U = heat-transfer coefficient between bed and steel wall, W/(m2 K) A , = cross-sectional area of steel wall, m2 Afl = outer area of one flange, m2 C = heat capacity of mixture, J/(kg K) Akj = reaction enthalpy of jth reaction, J/mol Di,eff= effective diffusivity for substance i, m2/s Subscripts N = ammonia E = ethylene oxide
M = monoethanolamine D = diethanolamine T = triethanolamine ins = insulation w = steel wall b = packed bed fl = flange r = room f = feed s = catalyst surface
T/K 41c
tribution. However, there is a need for better knowledge of the thermodynamic properties of the reaction mixture and for investigations of the long-term activity trend with the zeolites.
t
Greek Letters Glass beads
Catalyst
0 = temperature, dimensionless, (=T / Tref) X = heat conductivity, W/(m K) vi, = stoichiometric coefficient for substance i in reaction j
Catalyst
360 0.00
L
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90 Z/m
Figure 6. Temperature profile in reactor with catalyst Y partly replaced with glass beads.
from the heterogeneously catalyzed reactions. Scale-up. We have had the opportunity to compare our model for the A200 catalyst with production data from a full-scale adiabatic reactor a t Berol AB, Stenungsund, Sweden. The predicted yield of MEA was found to be about 4% lower than the observed value, and the capacity was about equally well predicted. Conclusions If the yield of MEA is the major concern, then the strongly acidic ion-exchanger Amberlite 200 is clearly the most attractive alternative. But if capacity is more important, then the zeolite 13X is a strong alternative. We believe that our results can be used for a preliminary reactor design, especially for predicting the product dis-
Registry No. MEA, 141-43-5; DEA, 111-42-2; TEA, 102-71-6;
NH3, 7664-41-7; Amberlite 200, 12626-25-4.
Literature Cited Ascher, U.; Christiansen, J.; Russel, R. D. Assoc. Comput. Mach. Trans. Math. Software 1981, 7(2), 209. Boneva, S . ; Dimov, N. Chromatographia 1979, 12(4), 204. Boneva, S.; Dimov, N. Chromatographla lB81, 14(10), 601. Johnson, F. L., Jr. (Texaco) U.S. Patent 438281, 1984. Knorr. L. Ber. Dtsch. Chem. Ges. 1899, 32, 729. Miki, M.; Ito, T.; Hatta, M.; Okabe, T. Yukagaku 1988, 75(5), 215. Potter, C.; McLaughlin, R. R. Can. J . Res., Sect. B : 1947, 25, 405. Saha, N. C.; Jain, S. K.; Dua, R. K. J. Chromatogr. 1977, 10(7), 368. Stoyanov, A. D.; Boeva, R. S.;Kotov, S . V. Zh. Prikl. Khim. (Leningrad) 1983, 56, 1966. Vamling, L. Ph.D. Dissertation, Chalmers University of Technology, Goteborg, Sweden, 1986. Villadsen, J.; Mlchelsen, L. M. Solution of Differential Equation Models by Polynomial Approximation; Prentlce Hall: Englewood Cliffs, NJ, 1978. Weibull, B. J. G. Swedish Patent 158 167, 1957. Weibull, 8. J. G., presented at the Swedish Symposium on Catalysis, Lund. 1970. Weibull, B. J. G.; Thorsell, L. U. F.; Lindstrom, S.0. Swedish Patent 355 570, 1973.
Received for review October 9, 1985 Accepted February 24, 1986
