I n d . E n g . C h e m . Res. 1987,26, 155-159
155
Kinetic Aspects in the Oxidation of Hydrogenated 2-Ethyltetrahydroanthraquinone E. Santacesaria* C.N.P.M., Centro Nazionale Propulsione ed Energetica, 20133 Milano, Italy
R. Ferro and S. Ricci Montefluos S.p.A. Company, 20133 Milano, Italy
S. CarrP Dipartimento di Chimica Fisica Applicata del Politecnico, 20133 Milano, Italy
In this paper the kinetic aspects of the oxidation of hydrogenated 2-ethyltetrahydroanthraquinone have been examined. This reaction occurs with molecular oxygen in gas-liquid reactors and is a key reaction in the industrial production of hydrogen peroxide. I t has been observed that reaction rates are strongly affected by mass transfer of oxygen across the liquid film. The mass-transfer coefficient of oxygen has been experimentally evaluated in absorption-desorption runs by employing an amperometric membrane electrode. Kinetic runs have been carried out in three types of reactors: a Levenspiel reactor with very low interfacial area, a CSTR reactor, and a well-stirred batch reactor with very high interfacial area. The kinetic regime has been determined and corresponds to a slow to moderately fast reaction with a Hatta number in the range 0.2-0.4 and enhancement factor of about 1. A second-order kinetic law has been proven to be reliable for interpreting the kinetic data.
In the process &-TETRA (Powell, 1968; Ullmann, 1969) for the industrial production of hydogen peroxide, 2ethyltetrahydroanthraquinone (THEAQ), is subjected to cyclic reduction and oxidation as in the following scheme: 0
OH (1)
*H2-
0
OH
0
OH
*02OH
W
H
2
0
2
(2)
0
A mixture of 70% THEAQ and 30% 2-ethylanthraquinone (EAQ) dissolved in a suitable solvent is usually worked up in industry, as reported by Ullmann (1969). Also EAQ is reduced by hydrogen to EAQH2, but it readily reacts, as described by Berglin and Shoon (1983), to give EAQHz + THEAQ THEAQHz + EAQ (3) This equilibrium is almost completely shifted to the right. However, in order to avoid secondary reactions, only 60-70% of the mixture of anthraquinone is reduced in the hydrogenation step. Therefore, only THEAQHz is involved in the successive oxidation step. No paper has been published, in the past, dealing with the kinetic aspects of the above described oxidation reaction. This reaction can be classified as an autoxidation in agreement with James and Weissberger (1938). However, with respect to other well-studied autoxidations, in this case, reaction rates depend on oxygen concentration. This behavior seems to be common to other autoxidable substances that form stable radicals as,for example, phenol (Denisov, 1980). The oxidation of THEAQHz with molecular oxygen is usually performed, on a large scale, in gas-liquid reactors in which gas-liquid interfaces are maximized. This fact clearly denotes that oxygen mass transfer plays an important role in limiting the reaction *To whom correspondence must be addressed. 0SSS-5SS5/S7/2626-0155$01.50/0
rates. Therefore, in this paper mass-transfer effects have been carefully considered. In particular, the mass-transfer rates of oxygen in the working solution have been experimentally measured, in transient conditions, for determining the related coefficient. Kinetics have been studied by employing two different types of CSTR reactors and have been verified in a batch reactor.
Experimental Section (A) Method, Apparatus, and Reagents. In order to study the kinetics of the oxidation of THEAQHz,we have employed three types of gas-liquid reactors: a slightly modified Levenspiel reactor (Levenspiel and Godfrey, 1974), a CSTR reactor, and a well-stirred batch reactor. The use of the first type of reactor, characterized by a very small interfacial area, had the scope to individuate the operative kinetic regime and to estimate the mass-transfer coefficient of molecular oxygen in the presence of the reaction. The second type of reactor has been employed for determining the chemical kinetic constant, while the third type has been employed for testing the reliability of both the kinetic model and the parameters. Mass-transfer rates of oxygen in the solution have been independently measured in transient conditions by the classic amperometric membrane electrode suggested by Clark (1959). The employed electrode was from Ingold Co. The response of this electrode to the oxygen concentration was fast enough to allow accurate measurements of the oxygen mass-transfer coefficient also by neglecting the time constant of the electrode, very small in respect to the duration of the run. Two types of runs were performed that are characterized by absorption or desorption of oxygen in the solvent of anthraquinones, that is, a mixture of aromatics. The liquid was gently stirred in order to avoid increase of the interfacial area by cavitation. The current signal, proportional to the oxygen concentration, was recorded during the time. In the absorption runs, oxygen was fed upon the surface of the liquid purged with nitrogen. On the contrary, in the desorption runs, nitrogen was fed upon the surface of the liquid saturated with oxygen. The presence of anthraquinone, in the oxidized form, has been tested and does not affect significantly mass-transfer rates and oxygen solubility. The measurements, therefore, were performed 0 1987 American Chemical Society
156 Ind. Eng. Chem. Res. Vol. 26, No. 1, 1987
-
y
to the gasvolumetric burette
o-\\
in
0
1
U
out
0 Figure 1. Scheme of the slightly modified Levenspiel reactor. Floatings have been employed in order to avoid formation of foams. (1)and (3) correspond to the inlet and the outlet, respectively, of the thermostating fluid, while (2) and (4) are the inlet and outlet of the working solution. ( 5 ) is connected to a gas volumetric buret.
in the solvent of anthraquinones. The solubility of oxygen in the working solution has been evaluated by the method suggested by Nitta et al. (1983). All kinetic runs have been performed, at 50 "C, on a mixture of EAQ-THEAQ not fully reduced (60-70%) or in the presence of fully reduced THEAQ. The composition of the mixture was about 30% EAQ and 70% THEAQ in moles, while the concentration of THEAQ was kept below 0.3 M. The employed Levenspiel reactor is represented in Figure 1. In this reactor the interfacial area was changed by using floatings in order to avoid the formation of foams. The reactor, having a total volume of 3 L, was filled with 1650 cm3 of liquid and was operated as a continuous reactor for the liquid and as a batch reactor for the gas. Pure oxygen and mixtures of nitrogen with oxygen were employed in the kinetic runs. The CSTR reactor, used in the kinetic runs, is represented in Figure 2. It had a capacity of about 1 L and was filled with 300 cm3 of liquid. Only pure oxygen has been employed in these runs. A stirrer of particular shape, in Figure 2, was employed. This stirrer operates in such a way that both phases can be considered completely mixed and very high interfacial areas, related to the rotating speed, can be obtained. The effect of stirring on gas-liquid mass-transfer rates can be determined with the sulfite method described by Charpentier (1981) and Linek and Vacek (1981). This can be done by comparing the absorption rates of oxygen, in aqueous solutions of sodium sulfite, of opportune concentration, containing cobalt ions, as catalyst, stirred at different values of rpm, with the one obtained in a solution of the same type, well stirred, but without producing bubbles. This last condition has been considered by us, as a fluid dynamic reference situation, in which interface area is obviously known. In the same fluid dynamic conditions, the mass-transfer coefficient of oxygen, in the working organic solution, has been determined, with the amperometric electrode. Increasing rotating speed, with
Figure 2. Scheme of the employed continuous gas-liquid reactor. In the same scheme is also represented in detail the stirrer used in the kinetic runs. (1)and (3) are the inlet and outlet, respectively, of the thermostating fluid, while (2) and (4) are the inlet and outlet of the working solution. ( 5 ) is connected with a gas volumetric buret.
30
t
2ol J
/
10 0
0
500
1mo
1500
ZOW
2500 rpm
Figure 3. Effect of stirring, with the turbine represented in Figure 2, on the apparent interfacial area.
respect to the reference described conditions, increases the mass-transfer rates of oxygen, in the sulfite solution. This effect is due mainly to the increase in interface area, but the mass-transfer coefficient also increases with stirring. Since these two parameters appear to be always coupled in the mass-transfer relationships, it is convenient to define an "apparent interface area" in which both the stirring effects on the true interface area and on the mass-transfer coefficient are contained. The mass-transfer coefficient is consequently considered constant, with stirring, and equal to the one measured in the described fluid dynamic reference conditions. The obtained values of apparent interface area, at different stirring speeds, are reported in Figure 3. It must be pointed out that the trend of Figure 3 probably is different in the case of solvents other than water. On the other hand, in our case, viscosity and density of the employed organic solvent were comparable with respect to water, and the observed values of gas holdup were also similar, despite the different surface tensions. On the basis of these facts, we have assumed, as a reasonable approximation, the apparent interface area
Ind. Eng. Chem. Res. Vol. 26, No. 1, 1987 157 equal, in the two mentioned systems, at the same rpm. The possible error in this procedure would be, according to us, of the same order of magnitude of the experimental error in the kinetic measurements. The stirred reactor, operating in batch conditions for both the liquid and gas phases, is quite similar to the one just described. Since the oxidation reaction is very fast at high levels of interfacial area, care must be taken, in batch runs to avoid the dead time in mixing reagents. If this aspect is neglected, erroneous conclusions could be reached about the presence or not of autocatalytic effects in the reaction. Kinetic runs in batch conditions have been performed on fully hydrogenated THEAQ and EAQ, respectively,and on partially hydrogenated mixtures. Runs have been performed by introducing the liquid, kept a t 50 "C and under nitrogen, a t constant feed rate into the reactor, which has been previously fiied with pure oxygen. In practice, oxygen mass transfer and reaction rates become significant only when the level of the liquid reaches the impeller. Reaction rates were measured, in all types of reactors, by measuring the oxygen consumption by a gas volumetric buret. This buret measures and replaces reacted oxygen without requiring any manual intervention. All the reagents have been supplied by Montefluos S.p.A.
Results and Discussion Oxidation of THEAQHz by molecular oxygen occurs through the following steps: (a) diffusion of oxygen from the gas phase to the gas-liquid interphase and (b) diffusion of oxygen from the liquid interphase into the bulk of the solution with chemical reaction. By operation in the presence of pure oxygen, the first step can be ignored. The other two steps can occur simultaneously. Therefore, in order to evaluate the extent of the reaction occurring in the liquid film, the enhancing factor or Hatta number must be determined. This has been done by evaluating oxygen mass-transfer rates both in the absence and in the presence of the reaction. Determination of the oxygen mass-transfer rates across the gas-liquid interphase in the absence of the reaction has been performed with the Clark electrode. These rates can be expressed, in the case of oxygen adsorption, as (4)
Integrating relation 4 from 0 to any value of Co, until saturation, after the introduction of the proportionality factor between oxygen concentration and current intensity, we obtain 1
1
1
1
where F(i) is a function of the electrode current measured. The proportionality factor b can be obtained by dividing the oxygen concentration a t saturation by the corresponding current intensity, and it resulted in b = 2.62 X IO-* mol/(cm3 nA). As expected, the trend of F ( i ) as a function of time was, in all cases, a straight line whose slope gives the mass-transfer coefficient. In Figure 4 mean values of i obtained in absorption runs, performed at 20 "C, are reported. Data have been fitted by eq 4 with an cm/s (ai= 0.16 cm-l). average value of kL = 5.5 X Table I shows the mass-transfer coefficients and Henry's law constants obtained, respectively, at 20 and 50 "C. Kinetic runs have been performed in the Levenspiel reactor with various feed rates, interfacial areas, and ox-
I InAl 400 r
300
I
200
100
0
1000
2000
3000
4000
Time I S I
Figure 4. Fitting of the evolution of current as a function of time in an absorption run performed at 20 "C. Table I. Kinetic and Thermodynamic Parameters Determined for the Oxidation of THEAQHz Henry const, atm cm3/mol H,,oc = 112 100 Hb, o c = 109 000 kL(20 "C) = 5.5 X mass-transfer coeff, cm/s kL(50 "C) = 6.6 X kc = 3830 f 300 kinetic const, cm3/(mol s) Table 11. Rates of Oxygen Consumption Obtained in the Levensoiel Reactor. in Different Ooerative Conditions" interface liq PO,' rates, 103kL, area, cm2 flow rate atm mol/s x lo6 cm/s 89.62 0.744 1 2.568 3.12 0.372 1 2.713 89.62 3.30 0.372 1 89.62 3.93 3.230 0.372 1 89.62 3.53 2.903 0.372 1 89.62 4.13 3.400 0.372 1 34.23 4.80 1.507 0.744 1 34.23 4.44 1.395 0.372 1 34.23 4.44 1.395 0.372 0.4 1.12 89.62 3.41 1.30 0.372 0.5 89.62 3.16 1.69 0.372 0.6 89.62 3.43 "Initial concentration of the organic reagent was 2.9 X lo4 mol/cm3. The rpm of the stirrer was 300 in all cases, and the temperature was 50 "C.
ygen concentrations. The results are summarized in Table 11, along with the operating conditions. In the adopted conditions, very low conversions were obtained, that is, about 3% in a hour, although residence times were rather long. Moreover, a change in the interface area produces a change in the reaction rate. Therefore, it is possible to conclude that the reaction is sufficiently fast to lower the concentration of oxygen close to zero in the bulk liquid, according to the classification of the gas-liquid reactions reported by Charpentier, consequently:
In Table I1 are reported the values obtained for KL in the different runs. The comparison of these values with the value of kL obtained in the absorption-desorption runs should give directly the enhancing factor if, effectively,Co, E 0. Since the values obtained in the comparison are in the range 0.5-0.75, it can be concluded that Co, cannot be considered equal to zero, and consequently, the enhancing factor would be near to the unity. Other kinetics runs have been performed in the CSTR reactor of Figure 2, which is characterized by a much greater interphase area. Runs have been performed at 50 OC and at different feed rates and concentrations of the
158 Ind. Eng. Chem. Res. Vol. 26, No. 1, 1987 Oxygen retained
Table 111. Kinetic Results Obtained for the Oxidation of THEAQH2 in the Continuous CSTR Reactor of Figure 2" liq flow rate 1.561 1.427 1.111 0.783 0.694 0.555 0.833 1.111 1.111
104coR, mol/cm3 3.38 3.38 3.38 3.03 2.49 2.83 3.38 2.26 0.99
rpm 1900 1900 1900 2000 2000 2000 1900 1900 1900
100(conversion) 77.8 81.0 86.4 90.7 92.4 94.6 89.0 85.0 88.0
("1
kc, cm3/(mol s) 4360 4630 4422 3440 3734 4246 3467 3100 3090
800
700
1 ~
600 SO0 -
"The temperature was kept in all cases a t 50 "C and the liquid volume a t 300 cm3.
organic reagent. The speed of the stirrer has been kept a t 1900-2000 rpm, which corresponds to 41-43 cm2/cm3 of the apparent interface area, as it can be observed in Figure 3. The results together with the operative conditions are reported in Table 111. These runs have been interpreted by assuming a second-order kinetic law and employing the relation
that is, eliminating Co, the rate of oxygen consumption becomes Po,
kLai has been calculated as the product of kL, determined with the electrode measurements, per the apparent interface area reported in Figure 3. This product must be equal to the product of the true parameters. This procedure is convenient because it allows us to separate the influence of fluid dynamic effects on mass transfer from the influence of other factors. The values obtained for kc, in the different runs, are reported in Table 111, while the mean value is reported in Table I. As it can be seen, the values obtained for kc in the different runs are reasonably scattered in the range of the experimental error for a foaming gas-liquid system as the one considered. With the introduction of the mean value of k c in the Hatta number
Ha =
(Do,k&n)"2 kL
(9)
it is possible to evaluate that in the runs reported in Table 111, the Hatta number is about 0.2-0.4, that is, a range to which corresponds an enhancement factor of 1 (Charpentier, 1981). The conclusion is that oxidation of THEAQH, is a moderately slow reaction occurring mainly in the bulk liquid, whose rate is affected by the mass transfer across the liquid film.Therefore, both interfacial area and liquid holdup should be maximized in the industrial reactor. The second-order kinetic relation has been chosen on the basis of the following observations: (a) If a zero order is attributed to oxygen, the low conversions obtained in the Levenspiel reactor cannot be justified. (b) Changes in THEAQHz concentrations in the runs of Table I11 are well interpreted with the second-order law. (c) The oxidation of phenol proceeds with the same kinetic law. (d) Also the runs performed in batch conditions can be well interpreted
Time i s )
Figure 5. Kinetic runs performed in batch conditions a t 50 "C (rpm = 1800). Curve 1 is related to a 212-cm3THEAQH, solution with concentration of 7.5 g/L; curve 2 to 270 cm3 of solution with concentration of 15 g/L; curve 3 to 290 cm3 of solution with concentration of 30 g/L. Curve 4 is related to 300 cm3 containing 30 g/L of mixture of quinones hydrogenated for about 80%. Curve 5 is related to a solution of 10 g/L of fully hydrogenated EAQ. Except for curve 5, experimental points have been fitted by employing the described kinetic model with the parameters of Table I. Continuous lines in curves 1-4 are calculated, while the points are experimental.
with the second-order kinetic law. In fact, in Figure 5 is reported the uptake of oxygen, as a function of time, for three runs performed in batch conditions a t different concentrations of fully hydrogenated THEAQ. The temperature was always kept at 50 "C. The runs have been fitted by introducing the parameters in the equations
-- -
02' '
dr
rtransf
- rreact
VMVL
d VL - - filling rate
(12)
= constant (13) dr and by integrating. Here r = t - tdelay, where tdelay is the time necessary for filling the reactor with the solution a t the level of the impeller. V0,' is the volume of the transferred oxygen, while Vo,'* is the volume of reacted oxygen. By solving the above-mentioned equations, it appears that reaction rates are only deceleratory, in accordance with a second-order kinetic law. No autocatalytic effect, therefore, is present in contrast with the behavior of hydrogenated durohydroquinone, reported by James and Weissberger (1938) and by James et al. (1938). In Figure 5 are also reported, for comparison, kinetic runs performed, respectively,on fully hydrogenated EAQ and on partially hydrogenated mixture of EAQ and THEAQ. As it can be seen, the kinetic behavior of the mixture is equal to the behavior of pure hydrogenated THEAQ, while the behavior of EAQ is completely different, because the reaction rate seems to be insensitive to the organic reagent concentration and equal to the mass-transfer rate of oxygen. That is, oxidation of EAQHz
Ind. Eng. Chem. Res. 1987,26, 159-161
is faster than oxidation of THEAQH2. The absence of autocatalytic effects could be explained in a classical autoxidation by assuming that steady-state conditions for the involved radicals are readily achieved. On the contrary, it is difficult to explain, on the basis of the corresponding radical mechanisms, the second-order kinetic law. It is reasonable to suppose that the hydrogen bond in this reaction can be transferred directly to the oxygen molecule as already suggested by Etienne and Fellion (1954), for explaining the mechanism of dihydrophenazine autoxidation. Similarly, it is possible that oxygen forms a complex with the hydroquinone followed by the formation of two radicals that rapidly react, giving a molecule of hydroperoxide as in the scheme OH
OH
159
Nomenclature ai = specific interfacial area referred to the liquid volume, cm2/cm3 CO, = concentration of oxygen in the liquid, mol/cm3 CR = concentration of the organic reagent, mol/cm3 Do, = molecular diffusivity of oxygen in the liquid, cm2/s E = enhancement factor F = feed rate of liquid, cm3/h H = Henry’s law constant, atm cm3/mol Ha = Hatta number kL = oxygen mass-transfer coefficient, cm/s kc = second-orderkinetic constant, cm3/(mols) i = current intensity, nA P O , = partial pressure of oxygen, atm Ro, = rate of oxygen consumption, molls S = exposed interfacial area, cm2 t = time, s V , = volume of liquid, cm3 V , = molar volume Registry No. EAQ, 84-51-5; THEAQ, 15547-17-8; EAQH2, 839-73-6; THEAQH2, 68279-54-9; H202, 7722-84-1; 02,7782-44-7.
Literature Cited HO
OW
The formation of radicals would be the rate-determining step, and consequently reaction rates can be expressed as
r = kcl[complex] = kclKe[CR][Oz] = kc [CR][02]
(15)
where k , is the equilibrium constant for the formation of the complex and kc1 is the true kinetic constant. Runs performed in batch conditions at different temperatures showed an activation energy of about 15 kcal/mol for the apparent kinetic constant kc. This fact confirms that the mechanism 14 is more reliable than classic autoxidation mechanism. However, both of the mechanisms explain well the greater reactivity of EAQHz due to the greater aromaticity of this molecule with respect to THEAQHz.
Acknowledgment
Berglin, T.; Schoon, N. H. Ind. Eng. Chem. Process Des. Deu. 1983, 22, 150. Charpentier, J. C. In Aduances in Chemical Engineering; Academic: New York, 1981; Vol. 11. Clark, L. C. US Patent 2913 386, 1959. Denisov, E. T. In Comprehensive Chemical Kinetic; Banford, C. H., Tipper, C. F. G., Eds.; Elsevier: New York, 1980; Vol. 16, p 125. Etienne, A.; Fellion, Y. C. R.Hebd. Seances. Acad. Sci., Ser. C 1954, 238, 1429. James, T. H., Weissberger, A. J. Am. Chem. SOC.1938, 60, 88. James, T. H.; Snell, J. M.; Weissberger, A. J . Am. Chem. SOC.1938, 60, 2984. Levenspiel, 0.; Godfrey, J. H. Chem. Eng. Sci. 1974, 29, 1723. Linek, V.; Vacek, V. Chem. Eng. Sci. 1981,36(11), 1747. Nitta, T.; Akimoto, T.; Matsui, A.; Katoyama, T. J . Chem. Eng. Jpn. 1983, 16(5), 352. Powell, R. Hydrogen Peroxide; Noyes Development Co.: New York, 1968. Ulmann, T. Encyclopedie der Technischen Chemie; Wiley: New York, 1969; Vol. 17.
Thanks are due to the Montefluos S.p.A. Company for financial support.
Received for review October 22, 1985 Accepted June 16, 1986
Vapor-Liquid Equilibria by UNIFAC Group Contribution. 4. Revision and Extension Detlef Tiegs and Jiirgen Gmehling* Lehrstuhl Technische Chemie B, University of Dortmund, 46 Dortmund 50, West Germany
Peter Rasmussen and Aage Fredenslund Instituttet for Kemiteknik, T h e Technical University of Denmark, DK-2800 Lyngby, Denmark
Revised UNIFAC interaction parameters are presented for a number of groups, especially the aniline group. Parameters are given for 52 pairs of groups where the parameters have not previously been available. N-Methylpyrrolidone has been introduced as a new group, and interaction parameters with five important groups are given. The range of applicability of the UNIFAC group-contribution method is dependent on the availability of group group-surface areas (QJ, and group-intervolumes (Rk), action parameters (ann and ann). Extensive tables with parameters for 41 commonly applicable groups have been OS88-5SS5/8?/2626-0159$01.50/0
presented by Gmehling et al. (1982) and supplemented by Macedo et al. (1983). UNIFAC parameters have been published for various specific group interactions. Recent examples of such publications are shown in Table I. The parameters 1987 American Chemical Society
