Professional Documents
Culture Documents
Received 8 November 2002; received in revised form 18 February 2003; accepted 19 February 2003
Abstract
The simulation and safe layout of a reactive distillation process requires the knowledge of the chemical reaction and separate binary phase
equilibrium data. The latter are difficult to obtain experimentally since the products of reaction affect the binary equilibrium as soon as they
are formed. With the apparatus constructed and the modeling method described, the binary vaporliquid equilibria (VLE) are obtained for
systems with reaction rates showing half-life times of 30 min and below. Results for the binary VLE are presented for compounds of methanol
and ethanol with formic acid at 313333 K and isopropanol with formic acid at 313353 K. The proposed method can be applied to obtain
binary interaction parameters of reactive compounds for commonly used gE -models, if the half-life time of the non-catalyzed reaction at the
desired temperature is above 10 min.
2003 Elsevier B.V. All rights reserved.
Keywords: Fast reacting systems; Binary vaporliquid equilibria; Chemical esterification reaction
0255-2701/$ see front matter 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0255-2701(03)00115-6
358 M. Tischmeyer, W. Arlt / Chemical Engineering and Processing 43 (2004) 357367
In this work, our method of determining the binary phase that this method is appropriate for a certain class of reaction
equilibria of reactive systems is described. The esterifica- systems used in the reactive distillation.
tion reactions of formic acid with methanol, ethanol and iso-
propanol are examined. These reactions have half-life times
of some minutes up to several hours, depending on the tem- 3. Theoretical background
perature and the amount of formic acid present in the system.
3.1. Phase equilibrium calculations
2. Overview of existing approaches For description of the VLE, a model was chosen:
xi i f0iL = yi iV P (1)
Ppken et al. [2,3] are using a computer operated static
apparatus. They measured isothermal P, x-data of the binary The activity coefficient i is calculated from a model for
systems required for the methyl acetate production process. the Gibbs excess enthalpy gE . The Wilson model [12] was
Presumption for these measurements is that the VLE is es- chosen according to the strongly non-ideal nature of the
tablished much faster than the chemical equilibrium which system:
yields operating temperatures up to 323.15 K at max.
Alsmeyer et al. [4,5] have set-up a flow-through apparatus xk
ki
for VLE measurements in reactive systems which also allows ln i = ln xj
ij + 1 (2)
j k j xj
kj
the determination of reaction kinetics in the liquid phase.
The reactive mixture is forced into a steady state which
with
does not represent its chemical equilibrium but reaches phase
equilibrium between vapor and liquid. Both phases are then vL0j ij
ij = exp (3)
analyzed in situ by mid-infrared spectroscopy. While the gas vL0i T
phase is measured in transmission, attenuated total reflection
is used for the liquid phase. where vL0i and vL0j are the molar volume of liquids i and j as
Another flow-type apparatus has been operated until 1998 given in [1] and ij the binary interaction parameters of the
by Reichl et al. [6]. It was set-up as non-recycle flow still and model.
due to the short residence time was used to determine iso- To calculate the liquid fugacity it is also necessary to in-
baric VLE data of thermally unstable components and of re- clude the standard state fugacity f0iL , which is the fugacity
active mixtures. Two esterification systems (methyl formate of the pure liquid component i at system pressure and tem-
and ethyl acetate) and one etherification system (tert-amyl perature. It can be expressed by
methyl ether) were investigated.
possible to use the experimental set-up as a flow type cell. saturated vapor pressure P0iLV , and exp(vL (P P LV )/RT) is
0i 0i
The compostion of the phases were obtained with a Fourier the Poynting correction, which accounts for the expansion
transforminfrared spectrometer. The reaction takes place or compression of the liquid from saturated vapor pressure
only by use of catalyst and was neglected during the VLE to system pressure.
measurements. The fugacity coefficient iV is calculated with the virial
Lee and Kuo [8], Lee and Liang [9], and Lee and Lin [10] equation of state by the method of Hayden and OConnell
investigated different esterification reactions which were cat- [13] to get the second virial coefficient. They assume that
alyzed to reach the chemical equilibrium. For measuring the the vapor is ideal except for association between molecules,
VLE an Othmer-type circulation still was used where the va- usually due to hydrogen bonding. The method incorporates
por phase is condensed before sampling. The pressure was the chemical theory of dimerization. Therefore, it accounts
kept constant, liquid and vapor concentrations were deter- for strong association and solvation effects, including such
mined at different constant temperatures, a gas chromato- as found in systems containing organic acids like in the
graph was used for analysis. present system.
Patel und Young [11] used a modified adiabatic calorime-
ter for a study of the ethylene oxide + nonyl phenol system. 3.2. Influence of reaction on quaternary VLE
During the reaction, the pressure was continuously measured
and stored. A linear extrapolation was used to obtain the The possible phase equilibrium conditions of a quaternary
binary VLE data. mixture can best be depicted in a tetrahedron. All points
The measuring method applied in this work follows the within the tetrahedron are possible phase equilibrium com-
proposal by Patel and Young [11] for calorimetry. It is shown positions for a system without reaction. When one or more
M. Tischmeyer, W. Arlt / Chemical Engineering and Processing 43 (2004) 357367 359
reactions occur, the degree of freedom FG is reduced by the ric lines are parallel. Their endpoints represent the chemical
number of independent reversible chemical equilibrium re- equilibrium concentrations corresponding to the initial com-
actions, see Gibbs phase rule (5): position of the components A and B. All reactions which
are started by mixing of pure A and B yield stoichiometric
FG = 2 + C Ph R (5)
lines within the triangle outlined by A, B, (C/D = 1/1).
where C is the number of components, Ph the number of The backward reaction (not depicted here for clarity in
phases and R the number of independent chemical equilib- reasons) would, if started by mixing of pure C and D, pro-
rium reactions. duce stoichiometric lines directing from the CD-border to
In a quaternary system with two phases and one reversible the chemical equilibrium surface.
reaction where the temperature is given, the remaining de-
gree of freedom is two. This number is identical to the di-
mension of the concentration space in which the equilibrium 4. Experimental
conditions are fulfilled (where either temperature or pres-
sure are kept constant). This part of the concentration space 4.1. Chemicals
is called reaction space [14]. The geometrical shape of the
reaction space varies with the type of reaction, the position Isopropanol, methanol and ethanol used for experimental
of the chemical equilibrium (i.e. the quantity of the equilib- procedure were of analytical grade (purity 99.9%, Merck).
rium constant Ka ) and the presence of inert components. Formic acid was given with 98% purity (Merck). For cal-
The stoichiometric reaction mechanism is: ibration of the gas chromatograph, methyl formate, ethyl
formate and isopropyl formate (purity >98%, Merck) and
A A + B B C C + D D, (6)
distilled water were used. The chemicals were processed
where A and B are the reactants (alcohol and organic acid), without further purification except for degassing.
C and D the products (ester and water) and i the stoichio-
metric coefficients. The latter are unity for the esterification 4.2. Equipment and experimental procedure
reactions considered here. If the equilibrium constant Ka
(see Eq. (7)) has a quantity about unity and an ideal mix- The experiments were carried out in a 550 cm3 static equi-
ture behavior (i.e. also Kx = 1), the reaction space forms a librium cell which is located in a thermostatic oil bath at a
saddle area as depicted in Fig. 1. given temperature with an accuracy of 0.1 K. The set-up
a C aD is given in Fig. 2 and explained below.
Ka = (7) Stainless steel 1.4435, Teflon and glass have been used
aA aB
for construction. The temperature inside the cell was
where ai are the activities measured by a platinum resistance thermometer PT-100
ai = xi i (8) (Conatex) with an accuracy of 0.01 K. The pressure was
measured using a calibrated high accuracy sensor head
The saddle area is stretched between the non-reacting (MKS Baratron Type 121A connected to the signal condi-
binary systems AC, AD, BC and BD. The arrows rep- tioner Type PDR-D-1) attached to the top of the cell. The
resent the stoichiometric lines which show the change in sensor head and the connecting line from the cell to the
concentration caused by the forward reaction. Since the re- sensor were thermostated at 398.15 0.1 K. The sensor
action does not change the number of moles, all stoichiomet- was calibrated by the ambient pressure (measured with a
mercury manometer), high vacuum (104 Pa) and the vapor
pressure of pure water, methanol, ethanol and formic acid
at a temperature range of 313.15353.15 K. According to
these measurements, the experimental pressure is assessed
to be reliable within 0.5% of the measured value.
Phase equilibrium is reached quickly by mixing the two
phases intensively using a stirrer head with magnetic cou-
pling (bmd 075, Bchi GmbH). The measurement starts
by injection of the necessary amount of alcohol in order to
obtain a desired mole fraction of the mixture. The alcohol
is degassed and filled into the injection cylinder (manufac-
tured by the Laboratory for Thermophysical Properties of
the University of Oldenburg). When the equilibrium tem-
perature of the oil bath is established, the vapor pressure of
pure organic acid is measured. At the time t0 , the predefined
Fig. 1. Reaction space of an equimolar reaction A + B C + D at amount of alcohol is injected into the cell within <2 s by
constant pressure. means of the pressure pump, powered with 1 MPa from a
360 M. Tischmeyer, W. Arlt / Chemical Engineering and Processing 43 (2004) 357367
Fig. 2. Experimental set-up for measuring VLE data of fast reacting systems: (1) static cell; (2) injection cylinder; (3) pneumatic power supply for
injection; (4) liquid sample loop; (5) thermostatic oil bath; (6) alcohol reservoir and degassing device; (7) vacuum pump; (8) gas chromatograph; (9)
personal computers.
nitrogen vessel. The precision of the injection is controlled For the non-reactive experiment, the injection cylinder was
by a linear transducer (Haidenhain, type ND221) with an filled with methanol, thermostated and injected step by step
accuracy of 10 m which is calibrated to determine the into the cell, which contained pure water at the beginning.
injected mass of alcohol at a given temperature. The triangles show the experimental data, the circles are
With the addition of a second component, the equi- the P, x-data calculated at the actually measured cell tem-
librium is disturbed due to the heat of mixing and heat peratures after each injection. Hence, they incorporate the
of reaction. We checked with non-reacting mixtures like temperature disturbance through mixing effects and differ
acetic acid+water and methanol+water that the cell reaches slightly from the values calculated at averaged temperature
phase equilibrium after 60 s of mixing. The VLE diagram (see continuous line). The mean absolute pressure deviation
of methanol + water at T = 315.4 K is given in Fig. 3. here is 0.29% and the mean relative deviation is 1.27%,
Fig. 3. Experimental and calculated phase equilibrium of methanol + water at 315.4 K: (triangle) experimental data; (circle) values calculated at
cell-temperature; (continuous and stroked lines) P, x- and P, y-curves calculated at averaged temperature, model parameters taken from Gmehling and
Onken [1].
M. Tischmeyer, W. Arlt / Chemical Engineering and Processing 43 (2004) 357367 361
Fig. 6. Comparison of Wilson regression to experimental LLE data (experimental data by Marsh [17]).
experimental data, a sufficient result could be achieved, but To obtain the desired binary interaction parameters for
then the mixing gap disappeared by the VLE parameters the organic acidalcohol pair, only these two are fitted to
obtained for ethyl formate + water. Since the fitting quality the P, T, xi -data of the quaternary mixture while all other
still was superior by using Wilson, the models which are ca- parameters remain unchanged. Since the VLE data were
pable of LLE calculation were not used. The same behavior taken close to the binary conditions, the resulting parameters
was found in the methanol and isopropoanol systems. are most suitable for calculating the binary VLE.
For a compilation of all used Wilson interaction param- The vapor phase association was taken into account by
eters see Tables 13, the ideal gas constant R is given in the chemical theory. The degree of association is represented
cal/(mol K) according to the format of the DECHEMA data by the ij parameter which is taken from Prausnitz [16]
base. for binary i and j phases (see Table 4). If no association
Table 1
Wilson parameters used in formic acidmethanol system
[23] [24] [25] [26] [27]
Table 2
Wilson parameters used in formic acidethanol system
[28] [29] [26] [17]
Table 3
Wilson parameters used in formic acidisopropanol system
[30] [31] [30] [26] [32]
Table 4
Binary interaction parameters for chemical theory [13]
Isopropanolethanolmethanol Formic acid Isopropyl Water
formateethyl
formatemethyl
formate
ij
Watera 1.55 2.5 1.3 1.7
Isopropyl formateb
Ethyl formatea
Methyl formatea 1.3 2 0.53
Formic acida 2.5 4.5
ii
Isopropanol 1.32
Ethanol 1.40
Methanol 1.63
a Prausnitz [16].
b No value available, estimated by analogy to ethyl formate and different acetates.
P, T, xi -data were measured during the reactions of: (at 313, 333 and 353 K).
Fig. 7. P(t) diagrams for the methanolformic acid system at 313 K: (circle) experimental; (dash) calculated.
M. Tischmeyer, W. Arlt / Chemical Engineering and Processing 43 (2004) 357367 365
Table 7
Regressed binary interaction parameters for the Wilson model
Components i, j
Temperature ( C) 40 60 40 60 40 60 80
Wilson ij 1322.2905 2202.9381 489.8492 156.5872 949.8210 1246.1398 1446.7360
Wilson ji 972.8247 1651.3409 57.8520 325.6978 562.4923 1600.1542 2102.0758
Pmean over all data (%) 2.2 6.6 12.3 4.4 2.1 3.7 2.6
Pmax over all data (%) 10.4 23.1 31.0 12.0 7.2 7.9 7.2
Pmean and
Pmax are the relative mean and maximum deviations of the calculated pressure and measured data calculated by |Pexp Pcalc |/Pexp (bar).
The ethanolformic acid reaction at 313 K was the first one examined. The apparatus (especially the accuracy and speed of liquid sampling) was advanced
and improved after completing the measurement of this phase equilibrium. Since each experiment consists of approximately nine reaction runs with about
12 liquid samples each, the amount of data points to be fitted by the model is immense. Deviations larger as compared to standard VLE measurements
were inevitable. The methanol + formic acid system at 333 K shows the fastest reaction rate. At high formic acid concentrations, the half-life time is
about 23 min. With this system, the maximum reaction rate to be examined by our measuring method is clearly reached.
A complete set of all measured data is given by Tis- use of Eq. (1) with the Wilson gE -model for liquid phase
chmeyer [21]. The respective liquid concentrations x1 x4 at and the chemical theory for vapor phase. The relative devi-
these time points and the measured temperatures are used to ation of all experimental to calculated pressure data is then
calculate the system pressure and vapor concentrations by minimized during the model fitting procedure by variation
Fig. 10. Calculated binary VLE diagram of the isopropanolformic acid system.
of the binary interaction parameters of the reacting pair. Bi- Appendix A. Nomenclature
nary parameters for the Wilson model were obtained at 313
and 333 K for all reacting systems and at 353 K for the iso- a activity
propanol system. In case of the ethanolformic acid binary a12 , a21 binary interaction parameters for gE -models
phase, experimental hE -data measured by Zhao et al. [22] Ai , Bi , Ci Antoine coefficients, P (bar) and T (K)
were also included in the regression. The resulting binary dm dipole moment (Debye)
parameters are given in Table 7. Vapor phase data are not EA,ij energies of activation for reaction ij
reproduced here due to problems in the determination of the f fugacity (Pa)
compositions. g Gibbs energy per mole (J/mol)
In Figs. 810, the resulting binary phase equilibria of the gij energies of interaction between an ij pair
reacting acidalcohol pairs can be seen. All show a min- of molecules (J/mol)
imum pressure azeotrope at low to medium alcohol con- hmix enthalpy per mole (J/mol)
centrations which shifts to lower alcohol concentrations at k12 reaction rate constants of the
higher temperatures. esterification (mol/s)
k21 reaction rate constants of the hydrolysis
(mol/s)
6. Conclusions Ka chemical equilibrium constant in terms
of activities
Quaternary P, T, xi -data have been measured for the Kf chemical equilibrium constant in terms
reacting systems isopropanol + formic acid (313353 K), of fugacities
ethanol + formic acid (313333 K) and methanol + Kx chemical equilibrium constant in terms
formic acid (313333 K). A static cell with the possibility of mole fractions
of adding thermostated, degassed components and imme- Mw molecular weight (g/mol)
diate mixing as well as a quaternary regression method n number of components
have been used. Wilson gE -model parameters for the re- ni number of moles of component i
acting binary systems were obtained. Calculated values P pressure (Pa)
and experimental data are in very good agreement in five PLV vapor pressure (Pa)
out of seven cases. The new regressed parameters for the r reaction rate (mol/s)
chosen gE -model can be recommended for accurate com- rg radius of gyration (nm)
putation of the equilibrium conditions, e.g. in simulation R universal gas constant (J/mol K)
processes. t time (s)
The results show that the applied method can be rec- t0 time of mixing, start of the reaction (s)
ommended to determine the binary interaction param- T temperature (K)
eters of reacting systems with half-life times down to v molar volume (m3 /mol)
10 min. Temperature deviations due to mixing effects x liquid mole fraction
(exothermic reactions) are covered by the regression X transformed liquid coordinate
method since each data point is evaluated at its own y vapor mole fraction
temperature. Y transformed vapor coordinate
M. Tischmeyer, W. Arlt / Chemical Engineering and Processing 43 (2004) 357367 367
ij Wilson coefficient as defined in Eq. (2) lation plant), Ph.D. Thesis, ETH Zrich, 1999.
ij Wilson binary interaction parameters [8] L.-S. Lee, M.-Z. Kuo, Phase and reaction equilibria of the acetic
(cal/mol) acidisopropanolisopropyl acetatewater system at 760 mm Hg,
Fluid Phase Equilib. 123 (1-2) (1996) 147165.
i stoichiometric coefficient of component i
[9] L.-S. Lee, S.-J. Liang, Phase and reaction equilibria of acetic
fugacity coefficient acid1-pentanolwatern-amyl acetate system at 760 mm Hg, Fluid
activity coefficient Phase Equilib. 149 (1-2) (1998) 5774.
Hayden OConnell association parameter [10] L.-S. Lee, R.-G. Lin, Reaction and phase equilibria of esterification
of isoamyl alcohol and acetic acid at 760 mm Hg, Fluid Phase
Equilib. 165 (2) (1999) 261278.
Subscripts
[11] N.C. Patel, M.A. Young, Measurement and prediction of vaporliquid
0 pure component equilibria for a reactive system: application to ethylene oxide + nonyl
after time phenol, Fluid Phase Equilib. 83 (1993) 7992.
AD components AD [12] G.M. Wilson, Vaporliquid equilibrium. Part XI. A new expression
c critical for the excess free energy of mixing, J. Am. Chem. Soc. 86 (1964)
127130.
i component i
[13] J.G. Hayden, J.P. OConnell, A generalised method for predicting
i0 initial composition of component i second virial coefficients, J. Ind. Eng. Chem. Process Des. Dev.
j component j 14 (3) (1975) 209216.
max maximum [14] B. Bessling, G. Schembecker, K.H. Simmrock, Design of processes
mean arithmetic mean with reactive distillation line diagrams, Ind. Eng. Chem. Res. 36
(1997) 30323042.
mix mixture
[15] U. Weidlich, Experimentelle und Theoretische Untersuchungen zur
Erweiterung der Gruppenbeitrags methode, UNIFAC, Ph.D. Thesis,
Superscripts Dortmund, 1985.
E excess [16] J.M. Prausnitz, T.F. Anderson, E.A. Grens, Computer Calculations for
L liquid Multicomponent VaporLiquid and LiquidLiquid Equilibria, Pren-
tice Hall Inc., Englewood Cliffs, New Jersey, 1980.
V vapor
[17] K.N. Marsh, J. Chem. Eng. Data 41 (1996) 1219.
VLE in phase equilibrium [18] VDI-Wrmeatlas: 8, Auflage, VDI-Gesellschaft Verfahrenstechnik
VLE + CE in phase and chemical equilibrium und Chemieingenieurwesen, SpringerVerlag, Berlin, Heidelberg,
+ standard state conditions 1997.
[19] A.L. Lydersen, Estimation of Critical Properties of Organic Com-
pounds, University Wisconsin Coll. Eng., Eng. Exp. Stn. Rep. 3,
Madison, Wisconsin, 1955.
References [20] R.C. Reid, J.M. Prausnitz, B.E. Poling, The Properties of Gases and
Liquids, fourth ed., Mc-Graw Hill, New York, 1987.
[1] J. Gmehling, U. Onken, VaporLiquid Equilibrium Data Collection, [21] M. Tischmeyer, Ph.D. Thesis, Technical University Berlin, Germany,
DECHEMA Chemistry Data Series, vol. I, Part 1, Frankfurt, 1977. 2004, in press.
[2] T. Ppken, R. Geisler, L. Gtze, A. Brehm, P. Moritz, J. Gmehling, [22] J. Zhao, J. Bao, Y. Hu, J. Chem. Thermodyn. 21 (1989) 811.
Reaktions kinetik in der reaktivrektifikationzur bertragbarkeit von [23] N. Kozub, H. Schuberth, E. Leibnitz, J. Prakt. Chem. 17 (1962)
kinetischen daten aus einer rhrzelle auf einen rieselbettreaktor, 282.
Chem. Ing. Technol. 71 (1/2) (1999) 96100. [24] (a) G.A. Ratcliff, K.C. Chao, Can. J. Chem. Eng. 47 (1969) 148;
[3] T. Ppken, L. Gtze, J. Gmehling, Reaction kinetics and chemical (b) M. Broul, K. Hlavaty, J. Linek, Collect. Czech. Chem. Commun.
equilibrium of homogeneously and heterogeneously catalyzed acetic 34 (1969) 3428.
acid esterification with methanol and methyl acetate hydrolysis, Ind. [25] V.T. Zharov, O.K. Pervukhin, Zh. Fiz. Khim. 46 (1972) 1970.
Eng. Chem. Res. 39 (2000) 26012611. [26] (a) V.V. Udovenko, L.P. Aleksandrova, Zh. Fiz. Khim. 34 (1960)
[4] F. Alsmeyer, W. Marquardt, H.-J. Ko, M. Roth, H.W. Siesler, D. 1366;
Mller, G. Olf, G. Ronge, Measurement of Phase Equilibria in Re- (b) M.S. Vrevsky, A.A. Glagoleva, Zh. Russ. Fiz. Khim. Obshch.
acting Systems, vol. 5.-7.4.2000, In: Proceedings of the Lecture at 59 (1927) 608.
GVC-Fachausschu on Thermische Zerlegung von Gas- und Fls- [27] A. Reichl, U. Daiminger, A. Schmidt, M. Davies, U. Hoffmann, C.
sigkeitsgemischen, Wernigerode, 2000. Brinkmeier, M. Reder, W. Marquardt, Fluid Phase Equilib. 153/1
[5] F. Alsmeyer, W. Marquardt, G. Olf, A new method for phase equi- (1998) 113.
librium measurements in reacting mixtures, Fluid Phase Equilib. 203 [28] I. Nagata, T. Ohta, M. Ogura, S. Yasuda, J. Chem. Eng. Data 21
(2002) 3151. (1976) 310.
[6] A. Reichl, U. Daiminger, A. Schmidt, M. Davies, U. Hoffmann, C. [29] I. Mertl, Collect. Czech. Commun. 37 (1972) 366.
Brinkmeier, C. Reder, W. Marquardt, A non-recycle flow still for [30] E.K. Tunik, V.T. Zharov, Viniti 1242 (1980).
the experimental determination of vaporliquid equilibria in reactive [31] E. Sada, T. Morisue, J. Chem. Eng. Jpn. 8 (1975) 191.
systems, Fluid Phase Equilib. 153 (1) (1998) 113134. [32] R. Stephenson, J. Stuart, J. Chem. Eng. Data 31 (1986) 5670.