Sunteți pe pagina 1din 10

Ethoxylation 1635

Gisele M. Amaral Research Article


Reinaldo Giudici
Universidade de São Paulo, Kinetics and Modeling of Fatty Alcohol
Escola Politécnica, Department
of Chemical Engineering, Ethoxylation in an Industrial Spray Loop
São Paulo, SP, Brasil.
Reactor
The kinetics of the ethoxylation of fatty alcohols catalyzed by potassium hydro-
xide was studied to obtain the rate constants for modeling of the industrial pro-
cess. Experimental data obtained in a lab-scale semibatch autoclave reactor were
used to evaluate kinetic and equilibrium parameters. The kinetic model was em-
ployed to model the performance of an industrial-scale spray tower reactor for
fatty alcohol ethoxylation. The reactor model considers that mass transfer and re-
action occur independently in two distinct zones of the reactor. Good agreement
between the model predictions and real data was found. These findings confirm
the reliability of the kinetic and reactor model for simulating fatty alcohol ethoxy-
lation processes under industrial conditions.

Keywords: Ethoxylation, Fatty alcohol, Kinetic model, Mathematical model, Spray loop reactor
Received: May 03, 2011; revised: May 31, 2011; accepted: July 04, 2011
DOI: 10.1002/ceat.201100215

1 Introduction by the group of Santacesaria et al. [2–6]. They studied the


kinetics of fatty alcohol (1-dodecanol) ethoxylation using
The nonionic surfactants, represented mostly by linear potassium hydroxide as catalyst [3] or barium dodecanoate as
ethoxylated alcohols, have experienced a growing demand in catalyst [4], the products presenting different chain length
the last decade. The reasons for this growth are their excellent distributions depending on the catalyst used. Other related
detergency properties, along with their rapid biodegradability studies are the kinetics of nonylphenol ethoxylation catalyzed
and low toxicity. The polyoxyethylene surfactants are by potassium hydroxide [2], ethoxylation and propoxylation
commercially produced by the oligomerization reaction of of 1-octanol and 2-octanol catalyzed by potassium hydroxide
ethylene oxide (EO) and/or propylene oxide with an active [5], and ethoxylation and propoxylation of ethylene glycol
hydrogen-containing compound (RXH). The starter is catalyzed by potassium hydroxide [6]. The ethoxylation
composed of hydrophobic molecules with a polar group in kinetics is described as an ionic polymerization with
their termination. The final product can be represented by the equilibrium between the protonated oligomers. The catalyst
general formula RX(CH2CH2O)nH, where R can be hydrogen concentration affects the protonated oligomer concentrations
(for polyglycols) or a hydrophobic group (usually C1 to C12) and the acidity ratio between the starter and the ethoxylated
and X represents a heteroatom such as O, S, or N [1]. The final oligomers, thus affecting the chain size distribution of the
product is formed by molecules with different numbers (n) of oligomers. The industrial catalysts can be either acid or basic;
added EO units that can vary from 0 to 20, depending on the the most used basic catalysts are potassium hydroxide and
product desired. The chain distribution is important because sodium hydroxide.
it influences the product application. Short-chain products are The process is carried out industrially in spray tower loop
used for oil removal whereas long-chain length products are reactors [6–11] or in jet loop (buss) reactors [6, 12, 13]. In a
used for liquid household detergents. The reaction is usually spray tower loop reactor, presented schematically in Fig. 1, the
performed in batch reactors. reacting liquid containing the catalyst and fatty alcohols is
Until 1990, only a few studies on ethoxylation kinetics had sprayed into an atmosphere of gaseous EO. Mass transfer of
been published. Extensive studies on this topic were performed EO occurs to the liquid phase, where the ethoxylation reaction
occurs. The heat released by the exothermic reaction is
counterbalanced by an external heat exchanger through which
– the liquid is circulated prior to injection in the spray nozzle. In
Correspondence: Prof. R. Giudici (rgiudici@usp.br), Universidade de
a typical industrial operation, the catalyst is added to the
São Paulo, Escola Politécnica, Department of Chemical Engineering, Av.
Prof. Luciano Gualberto, travessa 3, No. 380, 05508-010, São Paulo, SP,
starter and the mixture is activated by the production of a
Brasil. potassium salt of the starter and water in an equilibrium

Chem. Eng. Technol. 2011, 34, No. 10, 1635–1644 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1636 R. Giudici, G. M. Amaral

reaction. The water must be removed to achieve high For droplets with diameters larger than 4 mm, the model
conversion to the potassium salt and to minimize the proposed by Angelo et al. [16] seems to be better suited.
formation of polyglycols. The usual reaction temperature In this work, a mathematical model for the production of
varies from 140 °C to 180 °C. An increment in temperature ethoxylated fatty alcohols in an industrial spray tower loop
increases the reaction rate constant and decreases the EO reactor is developed. The model considers the reaction kinetics
solubility. To keep the process within adequate safety levels, and the mass transfer in the industrial reactor, for which the
reactor pressure and temperature should be controlled. most appropriate mass transfer coefficient model was
Reactor pressure is directly affected by the EO concentration determined. For the reaction kinetics, experimental data for
and reactor temperature. The higher the amount of EO fed to fatty alcohol ethoxylation carried out in a controlled well-
the reactor, the higher the pressure. Consequently, by mixed semibatch laboratory reactor were analyzed to estimate
manipulating the EO flow rate, it is possible to control the the kinetic parameters.
reactor pressure [7]. The reactor pressure usually varies
between 4 and 5 bar (absolute) and is defined by the reactor
productivity and limited by the reactor safety system. 2 Reaction Mechanism
The general mechanism of the EO ring-opening reaction has
been discussed in details by Edwards [1] and is formed by the
following steps:
Activation (or catalyst formation):

RXH ‡ M‡ OH → RX M‡ ‡ H2 O

Initiation:

RX M‡ ‡ EO → RX…EO† M‡

Propagation: (n = 1, 2, 3,...)

RX…EO†n M‡ ‡ EO → RX…EO†n‡1 M‡

Chain transfer (or proton transfer): (n, m = 1, 2, 3,... with


n ≠ m)

RX M‡ ‡ RX…EO†m H > RXH ‡ RX…EO†m M‡

Figure 1. Scheme of the spray tower loop reactor [3]. RX…EO†n M‡ ‡ RX…EO†m H > RX…EO†n H ‡ RX…EO†m M‡

The initiation step involves reaction of the deprotonated


Previous works on the modeling of spray tower loop starter (RX–) with ethylene oxide via a nucleophilic
reactors were presented by Hall and Agrawal [8], Santacesaria substitution mechanism (SN2). The deprotonated starter
et al. [9], Khuu et al. [7, 10], and Di Serio et al. [6]. Typically, concentration is equal to the catalyst concentration because
in the reaction zone, the liquid pool is described as a plug-flow the activation step is extremely fast. The chain transfer reaction
reactor, with variation of the EO concentration and the is the fastest step of the overall reaction because it involves an
temperature along the liquid height. In the mass transfer zone, acid-base proton transfer. Since this is an equilibrium reaction,
the liquid formed droplets that absorb EO can be considered the concentration of the active species in solution will be
to be either stagnated (rigid) spheres or droplets with internal determined by the relative acidities of RXH and RX(EO)H.
circulation. There is a wide variety of sprays with different This dramatically affects the outcome of the reaction [1].
characteristics of the droplets in terms of mass transfer The propagation of the polyoxyethylene chain under basic
coefficients, which can affect the performance of the conditions proceeds by kinetically nearly identical steps, since
ethoxylation process. Therefore, predictive models for the the structure of the propagating polyoxyethylene anions are
performance of ethoxylation reactors should adequately essentially the same. The acidities of the polyoxyethylene
account for the particular features of the droplets. Moreover, alcohols RX(EO)nH are nearly identical when n > 3 and
most of fundamental works on sprays are based on laboratory independent of the chain length. Consequently, proton
sprays with small droplets (100–300 lm), while the industrial exchange between species should occur statistically with no
process droplets are greater than 1 mm [14]. Amokrane et al. species being favored over another. At low degrees of
[15] studied the absorption and desorption of sulfur dioxide polymerization, there are small differences between the
in water spray and observed that, for droplets of ∼1.1 mm nucleophilicity and acidity of the adducts that lead to small
diameter, oscillations in the drop surface can be observed with differences in the reaction kinetics. As the degree of
a frequency of 200–30 Hz, from the smaller to the greater size. polymerization increases, the effect of the initiator on the

www.cet-journal.com © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2011, 34, No. 10, 1635–1644
Ethoxylation 1637

polyoxyethylene chain length rapidly diminishes, with the The overall catalyst concentration can be defined as:
chain length distribution controlled by a statistical addition of
EO to the chain. X

B0 ˆ ‰RO…EO†i M‡ Š (7)
iˆ0

3 Kinetic Model Using the approximations m0 ≅ ‰ROHŠ and mi ≅ ‰RO…EO†i HŠ


in Eq. (6) and considering the adducts from 1 to ∞, it follows:
For the case of fatty alcohols as initiator and potassium or
sodium hydroxide as catalyst, the previous reaction model can m0 B0
‰RO M‡ Š ˆ (8)
be simplified by considering the same rate constant for all P∞
propagation steps (independent of the adduct size i) and the m0 ‡ mi Ke0i
iˆ1
proton transfer equilibrium [3, 5]:
Consequently:
k
RO…EO†i 1 M‡ ‡ EO → →RO…EO†i M‡ …i ˆ 1; :::; ∞† 0 1
mi B B0 m0 C
‰RO…EO†i M‡ Š ˆ K B∞ C (9)
Keij m0 e0i @P A
‡
RO…EO†i M ‡ RO…EO†j H > RO…EO†i H ‡ RO…EO†j M ‡ mi Ke0i ‡ m0
iˆ1

…i; j ˆ 1; :::; ∞ with i ≠ j† This model, proposed by Santacesaria et al. [3], allows for
evaluation of the concentration of each ionic couple. Previous
Chain growth normally continues until all of the EO has studies indicated that reaction parameters such as
been consumed, thus polyoxyethylation can be considered to temperature, EO partial pressure, and catalyst concentration
be a living polymerization. There have been reports that chain (for small concentrations) do not significantly affect the chain
termination occurs at very high degrees of polymerization length distribution [1]. Also, a unique value of the equilibrium
because of a hydrogen abstraction reaction, which explains the constant can be used for all proton transfer equilibrium
difficulty of obtaining polyethylene glycols with molecular reactions, i.e., Ke = Ke0i [3].
weights greater than 1 million using basic catalysts [1]. Under In order to study the ethoxylation kinetics, high mass
these conditions, the mass balance equations for the total transfer rates should be attained so that the EO concentration
alcohol (including deprotonated alcohol), m0, and for the in the liquid phase can be considered to be in equilibrium; the
oligomer (adduct) of size i, mi, in a batch reactor can be solubility of EO in the reaction medium must be known. The
written as [3]: EO concentration in the liquid phase varies during the batch,
as the liquid properties change from hydrophobic to
dm0
ˆ k‰RO M‡ Š‰EOŠ (1) hydrophilic. Santacesaria et al. [5] and Di Serio et al. [17]
dt presented data for the vapor-liquid equilibrium of EO in
different starters (nonylphenol, dodecanol, and dodecanoic
dmi  acid, in both nonethoxylated and ethoxylated form with
ˆ k‰EOŠ ‰RO…EO†i 1 M‡ Š ‰RO…EO†i M‡ Š …i ˆ 1; :::; ∞† different number of EO units). The experimental data were
dt
fitted using the UNIFAC, NRTL, and Wilson models. The
(2)
Wilson model was found to give the best results. For accurate
where kinetic evaluations, it is important to describe the change of
EO solubility with the extent of reaction [3]. The
m0 ˆ ‰ROHŠ ‡ ‰RO M‡ Š (3) concentration of EO in the liquid phase [EO] was calculated
by assuming that the reactor content is well mixed and the
liquid and vapor phase are in equilibrium:
mi ˆ ‰RO…EO†i HŠ ‡ ‰RO…EO†i M‡ Š …i ˆ 1; :::; ∞† (4†
~yEO P ˆ cEO Pv;EO ~xEO (10)
The equilibrium constant can be defined as:
The vapor phase consisted of only EO (~yEO = 1) because the
‰RO…EO†j HŠ‰RO…EO†i M‡ Š reactor was under vacuum before the EO charging started. The
Keji ˆ (5)
‰RO…EO†i HŠ‰RO…EO†j M‡ Š vapor pressure of EO was calculated by using the Antoine
equation [17]:
Considering just the oligomer equilibrium with the alcohol,  
2568
Eq. (5) can be written as: Pv;EO ˆ exp 16:74 760 (11)
T 29:01
‰ROHŠ‰RO…EO†i M‡ Š
Ke0i ˆ (6) where Pv,EO is given in (atm) and T in (K). The activity
‰RO…EO†i HŠ‰RO M‡ Š coefficient of EO in the liquid phase was calculated using

Chem. Eng. Technol. 2011, 34, No. 10, 1635–1644 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1638 R. Giudici, G. M. Amaral

Wilson’s thermodynamic equilibrium model (index 1 pressure is increased slowly to the reaction pressure. Once the
corresponds to EO and index 2 to the ethoxylated product). reactor reached the desired pressure, the EO flow rate was ma-
nipulated to keep the pressure at the set-point. The reaction
ln c1 ˆ ln…~x1 ‡ ~x2 K12 † ‡ ~x2 ‰K12 =…~x1 ‡ ~x2 K12 † temperature was also controlled and recorded at each instant.
(12)
K21 =…~x1 K21 ‡ ~x2 †Š The total amount of EO charged to the reactor was that
corresponding to the desired average molecular weight of the
product. As soon as this value was achieved, the EO feed was
ln c2 ˆ ln…~x1 K21 ‡ ~x2 † ~x2 ‰K12 =…~x1 ‡ ~x2 K12 † stopped, but there was still unreacted EO remaining in the
(13) reactor. The duration of the reaction phase was measured. The
K21 =…~x1 K21 ‡ ~x2 †Š
reactor was kept under heating for some additional minutes to
assure that all the EO was consumed or cooked down until the
where the parameters for the ethoxylation of 1-dodecanol in reactor pressure falls, indicating the consumption of EO in the
the range of 0 ≤ nEO ≤ 10 and 70 ≤ T… C† ≤ 150 were vapor phase. For product end treatment, vacuum was applied
determined by Di Serio et al. [17]: to the reactor for a time and temperature set to strip all
residual EO. The samples were analyzed for the adduct length
K12 ˆ 13 ‡ 0:9611nEO 0:01967nEO 2 (14) distribution by the following procedure: first, the samples
were reacted with N,O-bis-(trimethylsilyl)-fluoroacetamide
K21 ˆ 0:4069 ‡ 0:04714nEO 0:00134nEO 2 (15) (BSTFA) to replace the active hydrogen of the hydroxyl groups
by trimethylsilyl groups (TMS); then the separation of the
Eq. (9) was solved to obtain the mole fraction of EO in the adducts were performed by gas chromatography using a
liquid phase, ~xEO , using the function fsolve in Matlab; the column of polyphenyl-methylsiloxane (PPMS) at high
concentration of EO in the liquid phase was then calculated by temperature and a flame ionization detector. Five experiments
  were performed at different reaction temperatures. The data
qL ~xEO are presented in Tab. 1.
‰EOŠ ˆ (16)
MW 1 ~xEO Parameters k0, Ea, and Ke were obtained by adjusting the
kinetic model to the experimental data of the oligomer
where qL is the density of the liquid phase and MW is the distribution (ROH and RO(EO)iH, with i = 1...17) of the final
average molecular weight of the liquid phase. product of each run. Eqs. (1) and (2) were simultaneously
The rate constant is considered to vary with temperature solved using the function ode45 in MatLab, and the adjustment
according to the Arrhenius law: of the kinetic constants was performed using the function
  lsqnonlin in MatLab, where the least squares objective function
Ea was given by:
k ˆ k0 exp (17)
RT
minf…k0 ; Ea ; Ke † ˆ
X
N n 
run X 2 (18)
‰RO…EO†i HŠexp ‰RO…EO†i HŠcalc
4 Experimental jˆ1 iˆ0

The kinetic (k0 and Ea) and equilibrium (Ke)


VACUUM
parameters were determined from experimental
data supplied by Oxiteno [18], using the above-de- PC
scribed kinetic model. The experimental data were
obtained from experiments performed in a 2-L N2 M
ETHYLENE OXIDE
stainless-steel Parr reactor, schematically
represented in Fig. 2. The agitation was fixed at
1000 rpm, which in previous studies proved to be
enough to keep the liquid and the vapor phases in
equilibrium.
Industrial-grade lauryl alcohol (main component
1-dodecanol) was used as starter. In the TC EFL.
experiments, an aqueous 50 wt-% solution of
potassium hydroxide was added to the reactor to
act as catalyst. Water formed in this step was COOLING WATER

removed under vacuum (100 mmHg) until its


concentration was less than 0.1 wt-%. The reactor
was then heated up to the reaction temperature by
a steam jacket. The feeding of EO was then started
STEAM
and the amount of added EO measured by a
weighing cell. With the addition of EO, the reactor Figure 2. Experimental setup used in the kinetic study.

www.cet-journal.com © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2011, 34, No. 10, 1635–1644
Ethoxylation 1639

Table 1. Data used in the kinetic model.

Batch T Pressure Reaction Alcohol Catalyst Initial Final


run [°C] [MPa] time [mol] [mol] volume volume
[min] [L] [L]

1 145 0.343 39 2.49 0.0373 0.543 1.57


2 145 0.343 44 2.48 0.0371 0.540 1.56
3 130 0.343 75 2.78 0.0415 0.604 1.75
4 160 0.343 27 2.62 0.0392 0.570 1.64
5 180 0.343 17 2.76 0.0413 0.601 1.73

5 Results and Discussion


For each experiment presented in Tab. 1, the experimentally
measured oligomer distribution was fitted and the kinetic
constants for the lauryl alcohol ethoxylation were determined.
Figs. 3 and 4 compare the oligomer distributions experimen-
tally measured (symbols) with those calculated by the model
(curves). The oligomer distribution obtained is coherent with

Figure 4. Experimental and calculated oligomer distribution for


T = 130 °C, 160 °C, and 180 °C.

ed in the literature. Fig. 5 compares the values of the kinetic


constant and equilibrium constant obtained at different tem-
peratures with the values reported by Santacesaria et al. [3]
and Di Serio et al. [5]. The agreement for the kinetic constant
Figure 3. Experimental and calculated oligomer distribution for is remarkably good for lower temperatures (deviations lower
T = 145 °C, at t = 39 min and 44 min. than 1 %) and satisfactory for the whole temperature range of

Table 2. Kinetic constants for the ethoxylation of lauryl alcohol.


the literature. The model was able to represent adequately
the chain length distributions of the oligomers for the dif- Present work Santacesaria et al. [3] Di Serio et al. [5]
ferent experimental conditions.
Tab. 2 compares the parameters obtained with the ln k0 [cm3mol–1s–1] 22.27 20.3 ± 0.3 20.2 ± 0.3
corresponding values reported in the literature [3, 5]. The Ea [kJ mol–1] 71.61 55.59 ± 0.95 55.3 ± 1.3
agreement is satisfactory for the equilibrium constant, the
Ke [–] 4.35 4.8 4.1 ± 0.5
value obtained being within the range of the values report-

Chem. Eng. Technol. 2011, 34, No. 10, 1635–1644 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1640 R. Giudici, G. M. Amaral

6.1 Model for the Mass Transfer Zone

The average concentration of EO in the droplet at the time the


droplet reaches the liquid pool can be calculated by integrating
Eq. (19)
dCEO 
ˆ klm alm …CEO C EO † (19)
dt

from t = 0 (where C EO ˆ CEO B is the concentration in the

liquid at the spray nozzle) to t = tfm and will depend on the


EO solubility in the liquid phase (CEO  ), the average droplet

specific surface (alm), the average mass transfer coefficient


(klm), the droplet fall time (tfm), and the EO concentration in
the liquid at the spray nozzle (CEOB ). The EO solubility in the

Figure 5. Arrhenius plot for the rate constant and Van’t Hoff plot mixture can be calculated by using the Wilson thermodynamic
for the equilibrium constant of lauryl alcohol ethoxylation. Com- model, as previously described [5].
parison of obtained results with experimental data from [3]. The average droplet specific surface is calculated by
P
the experiments. Larger differences are observed for higher p ni di 2
i 6
temperatures, the rate constants obtained in the present work alm ˆ p P 3
ˆ (20)
nd d 32
being up to 30 % higher than those reported in the literature. 6 i i i
The larger differences observed for higher temperatures can
be explained by two factors. First, the parameters in Eqs. (14) The diameter of the droplet formed in the spray nozzle at
and (15), used to calculate the EO concentration in the liquid the reactor top was supplied by the nozzle manufacturer for
phase, were obtained over a lower temperature range (70– water. This diameter can be converted to other liquids by
150 °C), while the experiments reported here were carried out considering the relation [19]:
at higher temperatures (130–180 °C). Thus, the extrapolation
of the phase-equilibrium model to higher temperatures could d32  r 0:5  l 0:2 1; 00:3
impart some deviations in the fitted rate constants. Second, at ˆ L
(21)
d32;water 73 0:1 qL
higher temperatures, the overall process kinetics can become
limited by the mass transfer between gas and liquid phases. The average droplet falling time is calculated as function of
The model used for data treatment, however, assumes that the the initial fall rate and the average droplet path (xm), which
EO concentration in the liquid phase is in equilibrium with depends on the liquid level distance to the spray nozzle (h*),
the gas phase. If mass transfer limitations become important, the spray angle (a), and the reactor radius (r). Assuming the
the value of the apparent activation energy is lowered, because same probability for all the possible paths, the following
the mass transfer is less sensitive to temperature than the relation can be obtained [9]:
chemical reaction. Lower activation energy is noted for the q
literature data, especially in the range of higher temperatures. xm ˆ rm 2 ‡ hm 2 (22)
Indeed, Santacesaria et al. [2] observed that, at higher
temperatures, the reaction rates are strongly influenced by the where
stirring speed and reported, in that case, the necessity of taking
into account in the model the mass transfer limitation. jh* r ho j …h* r ho †
rm ˆ r ‡
2…ho ‡ r h*†
 o   
jh h*j …ho h*† h* ‡ r ho
6 Industrial Reactor Modeling r ‡ (23)
2…h* ho † 2
 
The model employed for the industrial ethoxylation reactor jh* ho j …h* ho † h* a
‡ tan
was adapted from the model presented by Santacesaria et al. 2…ho h*† 2 2
[9]. The reactor scheme is presented in Fig. 1. Falling droplets
formed in the spray nozzle make contact and absorb EO from
the gas phase. Liquid at the reactor bottom is forced by an ex- jho ‡ r
h*j …ho ‡ r h*†
hm ˆ h* ‡
ternal pump to circulate through a heat exchanger and is then 2…h* ho r†
sprayed by the nozzle placed in the reactor top. During the   
h* ‡ r ‡ ho
droplet fall time, only the occurrence of EO absorption (mass × ‡ h* (24)
2
transfer) is considered, without significant extent of reaction.
In the liquid pool, EO is consumed as the liquid flows to the
reactor bottom, according to the recirculation flow rate. The r
released reaction heat is removed by the heat exchanger in- ho ˆ a; h* ˆ h hl (25)
tan
stalled in the recirculation line. 2

www.cet-journal.com © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2011, 34, No. 10, 1635–1644
Ethoxylation 1641

The liquid level is given by It should be mentioned that the mass balance (Eq. (32)) has
been incorrectly written in several previous articles that used
V Vhead‡lines this model; the correct cross-sectional area is (pr2) instead of
hl ˆ (26)
pr2 (4pr2). The solution procedure is similar to the algorithm pre-
sented by Santacesaria et al. [9]. For each time increment, cor-
The droplet fall time can be calculated by solving the responding to the reactor residence time, Eqs. (20) and (21)
momentum balance describing the droplet movement: are numerically integrated from the top (z = 0) to the bottom
  (z = hl) of the liquid column to obtain the concentration and
dv qL qg 3CD qg v2
ˆg t ˆ 0 v ˆ v0 (27) temperature profiles, and then all variables are updated for the
dt qL qL d32 next time interval by considering the amount of EO reacted
and the related mass balances. The relevant physical properties
of the liquid phase, which vary with the ethoxylation degree of
dx
ˆv tˆ0 xˆ0 (28) the fatty alcohol, are presented in Tab. 3 [3, 18, 19, 22].
dt

where the friction factor CD can be calculated as [19]:


7 Simulation and Model Validation

24…1 ‡ Re 0:70
†=Re for 0:1 < Re < 103
CD ˆ The described model was employed to simulate the behavior
0:445 for 103 < Re < 3:5  105
of an industrial reactor used to produce ethoxylated lauryl
(29) alcohol. The industrial data measured included the evolution
of reactor pressure, temperature of the liquid in the nozzle,
The initial droplet fall rate is given by [18]:
temperature of the liquid at the reactor bottom, EO flow rate,
s amount of EO fed to the reactor, liquid recirculation flow rate,
2DPspray and initial amount of fatty alcohol. The industrial
v0 ˆ w (30)
qL measurements are reported as normalized values due to confi-
dentiality reasons.
Eqs. (27) and (28) can be integrated up to x = xm to obtain In the simulation, the temperature of the liquid in the
t = tfm, the droplet fall time. nozzle and the reactor pressure are used as inputs. The
Different models were tested to calculate the mass transfer calculation procedure was similar to that described by
coefficient, among them the model of stagnant spherical Santacesaria et al. [9, 23]. The flight time of the droplets is cal-
droplets [20], and the correlations for internal circulation culated and Eq. (19) is solved to determine the EO concentra-
droplets presented by Srinivasan and Aiken [21], Amokrane et tion at the top of the liquid column. The reactor zone is then
al. [15], and Angelo et al. [16]. The best results were obtained simulated by solving Eqs. (32) and (33) to obtain the profiles
with the correlation of Srinivas and Aiken [21], which was of EO concentration and temperature along the liquid column.
finally adopted in the model: The EO concentration at the reactor bottom is the same as the
liquid sprayed in the nozzle after recirculation, and the tem-
 1=2  2 1=2  
0:16DEO lL v qL d32 vqL d32 5=16 perature of the liquid is adjusted to the desired value by the
klm ˆ heat exchanger in the recirculating loop.
d32 qL DEO r lL
(31) Fig. 6 illustrates the variation of the reactor pressure during
a typical industrial run and Fig. 7 the corresponding variation
of the temperature. At the beginning of the batch, pressure
and temperature are increased slowly until they reach the
6.2 Model for the Reaction Zone reaction condition. Then the heat released by the reaction
causes an additional increase in the temperature, and the heat
The reaction zone, i.e., the liquid pool at the reactor bottom, exchanger in the recirculating loop is used to control the
can be described as a plug-flow reactor assuming a pseudo- temperature. Fig. 7 indicates the variation of the temperature
steady state condition for the concentration and temperature of the liquid at the nozzle, at the top, and at the bottom of the
profiles. The corresponding mass and energy balances can be liquid column as well as the calculated temperature at the
written as: reactor bottom. Quite satisfactory agreement between the
calculated and experimental temperature at the reactor bottom
 
dCEO pr2 Ea is observed.
ˆ k0 exp B0 CEO (32) The flow rate of EO calculated by the model is compared
dz ql RT
with the measured data in Fig. 8. In the industrial reactor, the
flow rate is the variable manipulated for reactor pressure
   
dT pr2 Ea A control; it is influenced by the EO transferred to the liquid and
ˆ … DH†k0 exp B0 CEO hloss l …T Tair † by the quantity remaining in the gas phase. At the beginning
dz ql qL cp RT V
of the batch, the EO flow rate increases because, as the pressure
(33)
increases, the EO solubility also increases. In addition, as the
temperature increases, the solubility decreases, but the reaction

Chem. Eng. Technol. 2011, 34, No. 10, 1635–1644 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1642 R. Giudici, G. M. Amaral

Table 3. Physical properties calculation methods.

Property Equation Source/method

Surface tension r ˆ 10 3 ‰21:619 0:0232Ti exp… 0:049nEO †Š Group contribution [22]


r in [N m–1] and Ti in [°C]
Viscosity lL ˆ 10 3 exp‰0:1423nEO 1:9354 ‡ 500=…76:15 ‡ Ti †Š Experimental data [18]
lL in [Pa s] and Ti in [°C]
Density qL ˆ 834 ‡ 34:3…nEO †0:7281 0:744Ti Data regression [3]
qL in [kg m–3] and Ti in [°C]
1=2
vMW
Gas diffusivity DEO ˆ 7:4  10 5 Ti 0:6 Wilke-Chang equation [19]
lL Va
DEO in [cm2s–1], Va in [cm3mol–1], and lL in [Pa s]
Specific heat cp ˆ 3:01 0:105nEO Experimental data [18]
cp in [kJ kg–1C–1]

8 Concluding Remarks
The kinetics of ethoxylation of fatty alcohols cata-
lyzed by potassium hydroxide was determined for
industrial-grade lauryl alcohol (1-dodecanol) in a
laboratory-scale reactor in the range of 130–
180 °C. The kinetic parameters thus determined
are in fair agreement with the literature, especially
at lower temperatures. Some deviations were ob-
served at higher temperatures, probably due to the
extrapolation of the phase equilibrium model
used. Also, differences in mass transfer limitations
could affect the temperature dependence of the
reaction kinetics. The kinetics thus obtained was
applied to model the performance of an industrial-
scale spray tower reactor for fatty alcohol ethoxyla-
tion. The model development was based on a
similar work found in [9] and assumes that mass
Figure 6. Measured evolution of the reactor pressure during the batch. transfer and reaction occur independently in two

rate (consumption of EO) increases. Fig. 9 presents the EO


concentrations in equilibrium, at the liquid interface, and at
the reactor bottom, calculated by the model. At the reactor
bottom there is some unreacted EO.
The calculated values of the average droplet flight path
length and the average droplet flight time as a function of the
batch time are displayed in Fig. 10. The droplet path length
reduces due to the increase of the level of the liquid inside the
reactor, thus reducing the mass transfer zone. The average
flight time is always lower than 1 s, thus justifying the adopted
hypothesis of a negligible reaction extent in the falling
droplets.
The consideration of a pseudo-steady state for the
calculation of the reaction zone in the liquid column is
reasonable because the total batch time corresponds to about
80 residence times. Initially, the residence time is small
(0.4 min) due to the small reactor volume and increases with
the reactor volume, but is not greater than 1.1 min. Figure 7. Reactor temperatures at the top of the liquid column Ti
(+), at the spray nozzle Ti,spray (*), and at the reactor bottom Tb
(experimental (o), calculated (x)).

www.cet-journal.com © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2011, 34, No. 10, 1635–1644
Ethoxylation 1643

distinct zones of the reactor. Good agreement between the


model predictions and the real, industrial data was found.
These findings confirm the reliability of the reactor model for
simulating fatty alcohol ethoxylation processes under
industrial conditions.

Acknowledgment
The authors thank FAPESP, CNPq, CAPES and INCT-CEPE-
MA for financial support, Oxiteno for making available the
experimental data used in this work, and Prof. Frank Quina
for revising the manuscript.

The authors have declared no conflict of interest.

Figure 8. Ethylene oxide flow rate (experimental (o), calculated Symbols used
(x)).
[EO] [mol m–3] ethylene oxide concentration
alm [m2] droplet surface area
Al [m2] heat transfer surface area
B0 [mol m–3] catalyst concentration
CD [–] friction coefficient
CEO [mol m–3] ethylene oxide concentration
B
CEO [mol m–3] ethylene oxide concentration at the
reactor bottom

CEO [mol m–3] equilibrium ethylene oxide
concentration
cp [J kg–1K–1] specific heat
d32 [m] Sauter diameter
di [m] droplet i diameter
DEO [m2s–1] gas diffusivity
Ea [J mol–1] activation energy
g [m s–2] gravity acceleration
hloss [J s–1m–2C–1] heat transfer coefficient between
reactor and external ambient
hl [m] liquid height
Figure 9. Calculated ethylene oxide concentrations in the hm [m] average droplet path height
reactor.
DH [J mol–1] heat of reaction
k [m3mol–1s–1] reaction rate coefficient
k0 [m3mol–1s–1] frequency factor
klm [m s–1] mass transfer coefficient
Ke [–] equilibrium constant for proton
xm (m)

exchange reactions
MW [kg mol–1] molecular weight
nEO [–] number of ethylene oxide units per
molecule
t (min) ni [–] number of droplets of diameter di
P [Pa] pressure
PEO [Pa] partial pressure of ethylene oxide
Pv [Pa] vapor pressure
tfm (s)

DPspray [Pa] spray pressure drop


ql [m3s–1] reactor recirculation flow rate
r [m] reactor radius
R [J mol–1K–1] ideal gas constant
Re [–] Reynolds number
t (min)
rm [m] average droplet path radius
Figure 10. Average droplet path and average droplet flight time. t [s] time

Chem. Eng. Technol. 2011, 34, No. 10, 1635–1644 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1644 R. Giudici, G. M. Amaral

tfm [s] average flight time of sprayed [5] M. Di Serio, R. Tesser, F. Felippone, E. Santacesaria, Ind. Eng.
droplets Chem. Res. 1995, 34, 4092.
T [K] temperature [6] M. Di Serio, R. Tesser, E. Santacesaria, Ind. Eng. Chem. Res.
Tair [K] ambient temperature 2005, 44, 9482.
v [m s–1] droplet velocity [7] S. M. Khuu, J. A. Rodriguez, J. A. Romagnoli, K. F. Ngian,
V [m3] reactor liquid volume Comp. Chem. Eng. 2000, 24, 863.
Va [m3mol–1] molar volume [8] C. Hall, P. Agrawal, Can. J. Chem. Eng. 1990, 68, 104.
x [m] droplet position (distance from the [9] E. Santacesaria, M. Di Serio, P. Iengo, Chem. Eng. Sci. 1999,
nozzle) 54, 1499.
xm [m] average path of sprayed droplets [10] S. M. Khuu, J. A. Romagnoli, P. A. Bahri, G. Weiss, Comp.
~xi [–] molar fraction of i in the liquid Chem. Eng. 1998, 22 Suppl., S715.
phase [11] A. Dimiccoli, M. Di Serio, E. Santacesaria, Ind. Eng. Chem.
~yi [–] molar fraction of i in the vapor Res. 2000, 39, 4082.
phase [12] L. L. van Dierendonck, J. Zahradnik, V. Linek, Ind. Eng.
Chem. Res. 1998, 37, 734.
Greek Symbols [13] P. H. M. R. Cramers, A. A. C. M. Beenackers, L. L. van Dier-
endonck, Chem. Eng. Sci. 1992, 47 (13–14), 3557.
a [rad] spray angle
[14] N. Yeh, G. Rochelle, AIChE J. 2003, 49 (9), 2363.
ci [–] activity coefficient
[15] H. Amokrane, A. Saboni, B. Caussade, AIChE J. 1994, 40
Kij [–] interaction parameter
(12), 1950.
lL [kg m–1s–1] liquid viscosity
[16] J. B. Angelo, E. N. Lightfoot, D. W. Howard, AIChE J. 1966,
qG [kg m–3] gas density
12 (4), 751.
qL [kg m–3] liquid density
[17] E. Santacesaria, M. Di Serio, R. Tesser, Catal. Today 1995, 24,
r [N m–1] surface tension
23.
v [–] solvent parameter
[18] Oxiteno, Internal Report, Influence of process variables on the
w [–] spray nozzle efficiency
consumption rate of ethylene oxide during ethoxylation of
laurylic alcohol to Unitol L80 (in Portuguese), 1996.
[19] R. H. Perry, D. W. Green, Perry’s Chemical Engineers Hand-
References
book, 7th ed., McGraw Hill, New York 1999.
[20] A. I. Johnson, A. E. Hamielec, AIChE J. 1960, 6 (1), 145.
[1] C. Edwards, in Nonionic Surfactants Organic Chemistry (Ed:
[21] V. Srinivasan, R. C. Aiken, Chem. Eng. Sci. 1988, 43 (12),
N. M. Van Os), Marcel Dekker, New York 1998.
3141.
[2] E. Santacesaria, M. Di Serio, L. Lisi, D. Gelosa, Ind. Eng.
[22] R. C. Reid, J. M. Prausnitz, B. E. Poling, The Properties of
Chem. Res. 1990, 29, 719.
Gases and Liquids, 4th ed., McGraw Hill, New York 1986.
[3] E. Santacesaria, M. Di Serio, R. Garaffa, G. Addino, Ind. Eng.
[23] E. Santacesaria, M. Di Serio, R. Tesser, Ind. Eng. Chem. Res.
Chem. Res. 1992, 31, 2413.
2005, 44, 9461.
[4] E. Santacesaria, M. Di Serio, R. Garaffa, G. Addino, Ind. Eng.
Chem. Res. 1992, 31, 2419.

www.cet-journal.com © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2011, 34, No. 10, 1635–1644

S-ar putea să vă placă și