Thorough examination of a Wittig-Horner reaction using reaction calorimetry (RC-1), LabMax, and ReactIR

Article abstract of DOI:10.1021/op020075i

The kinetics and thermodynamics of a Wittig-Horner reaction was investigated, together with side reactions such as the hydrolysis of benzyi phosphonate. It was found that the water content in the reaction mass has a crucial effect on the rates both of the unwanted hydrolysis of the phosphonate and of the main reaction. The reaction was found to be fourth-order. That means that the reaction is very fast at the start but tails off towards the end. On the basis of the kinetic and thermodynamic parameters, a simulation of the whole process was made, including the expected concentrations of the product under the required conditions (temperature, reactor characteristics, etc.). The experimental values of the concentrations were obtained from spectra collected via a rapid reaction mode in the ReactIR (a spectrum was collected every ～3 s). Exact agreement between experimental and theoretical values confirms the correctness of the mathematical model and enables right planning of the industrial process.

Full text of DOI:10.1021/op020075i

Organic Process Research & Development 2003, 7, 237−243
Articles
Thorough Examination of a Wittig-Horner Reaction Using Reaction Calorimetry
(RC-1), LabMax, and ReactIR
Michael Grabarnick and Sharona Zamir*
Makhteshim Chemical Works Ltd., P.O. Box 60, Beer SheVa 84100, Israel
Abstract:
The kinetics and thermodynamics of a Wittig-Horner reaction
was investigated, together with side reactions such as the
hydrolysis of benzyl phosphonate. It was found that the water
content in the reaction mass has a crucial effect on the rates
both of the unwanted hydrolysis of the phosphonate and of the
main reaction. The reaction was found to be fourth-order. That
means that the reaction is very fast at the start but tails off
towards the end. On the basis of the kinetic and thermodynamic
parameters, a simulation of the whole process was made,
including the expected concentrations of the product under the
required conditions (temperature, reactor characteristics, etc.).
Figure 1. Heat flow during stilbene synthesis with solid KOH
The experimental values of the concentrations were obtained
additions (the third portion of KOH was suspended in solvent
after high-shear mixing).
from spectra collected via a rapid reaction mode in the ReactIR
(a spectrum was collected every ∼3 s). Exact agreement between
experimental and theoretical values confirms the correctness
of the mathematical model and enables right planning of the
industrial process.
Introduction
From early 1950, when Wittig and co-workers published
a new method for olefin synthesis from aldehydes or ketones
and phosphonium ylides, the Wittig reaction with some
modifications (for example, as shown below) has become
widely used and discussed in numerous reviews.1
Figure 2. Heat flow during stilbene synthesis with aqueous
KOH solution.
The mechanism of the Wittig reaction is commonly
presented showing formation of a four-membered cycle of
oxaphosphetane that is irreversibly decomposed to the
reaction products. In this work we consider kinetic peculiari-
ties, differences in the reaction mechanism, process optimi-
zation, and industrial usage of the Wittig-Horner reaction.
Figure 3. Heat flow during stilbene synthesis with solid KOH
suspended in solvent.
RC-1 Experiments and Process Optimization by DOE.
Preliminary experiments that were carried out in the RC-1
reactor showed that the stilbene-producing reaction is very
(1) For example: Maryanoff B. E.; Reitz A. B. Chem. ReV. 1989, 89, 863-
927.
10.1021/op020075i CCC: $25.00 © 2003 American Chemical Society
Published on Web 03/19/2003
Vol. 7, No. 3, 2003 / Organic Process Research & Development
•
237

Figure 4. Interaction plot of the three factors.
fast and that the yield is sensitive to the way the base is
added to the reaction media. The lowest yield was obtained
when solid KOH was added in four portions (Figure 1). It
can be seen that when the KOH was added suspended in
the solvent using a high-shear mixer (i.e. high availability
of the base), the heat flow was much higher. KOH addition
in aqueous solution (Figure 2) or suspended in the solvent
(Figure 3) gave better results.
Since preliminary experiments had shown that the base
addition time is the important factor to the reaction, we
included it as a parameter (parameter (c) 10 and 20 min) in
the process development by design of experiment (DOE).
Two additional factors that were examined were (a) starting
temperature: T0 (5 and 30 °C) and (b) final temperature:
T1 (40 and 80 °C). To all of these factors repeated center-
point values were added (17.5 °C, 60 °C, and 15 min,
respectively). All the experiments were performed with KOH
suspended in solvent.
Calculation of results was carried out using the STAT-
GRAPHICS Plus program.4 Figure 4 shows the interaction
plot of the three factors. It can be seen that the influence of
both the starting temperature and the final temperature on
the yield are in opposition when interacting with the addition
time (factor c). It means that a high yield is obtained both
when the starting temperature is low and the addition time
is long, and when the starting temperature is high and the
addition time is short. The same tendency was obtained with
the final temperature.
The following results were obtained from the ANOVA
(analysis of variance) Table of the DOE from the STAT-
GRAPHICS program. It can be seen that all the parameters
are significant from the statistical point of view (p values
are less than 0.05), and that the model is adequate to the
real process (p value of lack-of-fit test is greater than 0.05).
Reaction Kinetics. The prediction ability of this statistical
model is restricted usually both by the number of parameters
238
•
Vol. 7, No. 3, 2003 / Organic Process Research & Development

analysis of the samples. We used not less than 30-times
excess of KOH and BA, and to avoid quantitative analysis
of the reaction mass samples we used an internal standard.
The constants (kobs) of the reaction were calculated by linear
regression in coordinates -ln(SBP/SIS) - τ. Here SBP is the
peak area of benzyl phosphonate, SIS is the peak area of the
internal standard, and τ is the reaction time. The straight
lines that were obtained in these coordinates demonstrate
that the Wittig-Horner reaction is first-order with respect
to BP.
To find the reaction order with respect to KOH and BA
we measured kobs at different concentrations of these reagents
in the reaction mass. We obtained first-order with respect to
KOH, but surprisingly the reaction order with respect to BA
was two! In other words, the order of the Wittig-Horner
reaction under the considered conditions is four.
We checked also the influence of water on the reaction
rate at different temperatures. Changing the water content
in the reaction mass was done by the dilution of KOH. It is
evident that increasing the water concentration in the reaction
mass (reducing the KOH concentration in the water solution)
decreases the reaction rate (Figure 5).
At the same time the activation energy is not changed,
which means that this reaction series is isoenthalpic and that
the activation entropy determines the change in the reaction
rate. It was found that a linear relationship exists between
ln k0 and water concentration in the reaction mass, where k0
is the preexponential factor of the reaction constant of the
fourth order.
Our knowledge of the reaction order and the influence of
temperature and water content on the reaction rate make it
possible to calculate the real reaction constant and the real
rate of the reaction in wide intervals of the temperature and
the reagent concentrations. Comparison of the pseudo-first-
order constants that were calculated and measured in the new
Figure 5. Activation parameters as a function of water content.
and by the limits of the changes of the parameters during
DOE performance. Although the results of the DOE predic-
tion were of high quality, we must carry out at least one
additional set of DOE to obtain the optimal parameters of
the process.
Another method of process optimization is based on the
kinetic investigation of the desired reaction and of the main
side reactions. Kinetic data enable us to understand the
mechanism of the chemical reaction, to define the limiting
reaction step, to create a mathematical model of the process,
and as a result, to find the optimal parameters. As a rule,
this method needs more experiments and more calculations
but allows more precise and less limited predictions to be
made.
Kinetic measurements of the main process were carried
out under pseudo-first-order conditions by means of HPLC
Scheme 1
Vol. 7, No. 3, 2003 / Organic Process Research & Development
•
239

Table 1. Calculated and measured pseudo-first-order
constants
BA
KOH water
concn, concn, concn, temperature,
mol/L mol/L mol/L
°C
k
obs, s-1
k
calc, s-1
0.766 0.894 1.742
0.766 0.894 1.742
10
15
3.55 × 10-2 3.66 × 10-2
4.78 × 10-2 4.96 × 10-2
conditions does validate the mathematical model (Table 1).
From the theoretical viewpoint it is important to under-
stand, what is the mechanism of the Wittig-Horner reaction,
what is the rate-limiting step of the reaction, why do we
have a second-order kinetics of the reaction with respect to
BAswhat is the role of the second molecule of BA?
It seems trivial that the first stage of the reaction is a rapid
equilibrium between CH-acid (BP) and its potassium salt
(Scheme 1). Because of the high nucleophility of the anion,
synthesis of the activated complex 1 may also be very rapid.
The next reaction step must be oxygen transfer from carbonyl
carbon to phosphor and decomposition of the activated
complex to the reaction products. The problem is that an
oxygen transfer by a four-membered cyclic transition state
is very difficult! For this transfer, we probably need a second
molecule of BA. It enables the synthesis of the six-membered
cyclic activated complex 2. Producing this complex or its
decomposition to the reaction products with release of a
molecule of BA may be the rate-limiting step of the reaction.
We believe that the second case is more probable, because
the activation energy is too low for typical nucleophilic
reactions (about 41 kJ/mol). Apart from it, activation energy
is not changed with changing solvating ability of the solvent
(increasing water concentration in the solution). An increase
of the water content in the reaction mass due to the hydration
of complex 2 tends to increase the activation entropy (about
20 J/(mol‚K) more negative).2
Figure 6. Activation parameters as a function of water content.
Scheme 2
hydration effect. At the same time activation enthalpy is
increased. According to our calculations, at a temperature
of about 100 °C (isokinetic temperature) the influence of
water concentration is negligible, and at higher temperatures
the influence should be the opposite.
We have found that in the investigated intervals, linear
relationships exist both between water concentration and the
preexponential factor of the second-order reaction rate
constant, and between water concentration and the activation
energy of hydrolysis. These facts make it possible to calculate
the real reaction constant and the real rate of BP hydrolysis
at different temperature and reagent concentrations. In other
words, we now have a full mathematical model of the BP
hydrolysis.
We investigated the hydrolysis of BP by a similar method
without BA addition. Again, the straight lines that were
obtained in coordinates -ln(SBP/SIS) - τ demonstrates first-
order kinetics of this reaction with respect to BP. Measure-
ments of kobs at the different concentrations of KOH lead us
to a conclusion that the reaction is also first-order with
respect to KOH.
As in the case of the Wittig-Horner reaction, we
investigated the hydrolysis of BP at different temperatures
and in solutions at different water concentrations. It is
interesting to note that the activation energy of the BP
hydrolysis is essentially higher than in the Wittig-Horner
reaction and is changed, depending on the water content,
from 63 up to 82 kJ/mol (Figure 6).
In these experiments we have an ideal isokinetic relation-
ship.3 This means that the reaction mechanism is the same
in all reaction series. Water addition leads to reduction of
the reaction rate and activation entropy owing to the
Second-order kinetics (first with respect to BP and to
KOH) and activation parameters allow us to conclude that
the hydrolysis of BP is carried out according to the classical
(addition - elimination) (A - E) mechanism and that
probably the rate-limiting step is nucleophilic attack by the
hydroxide ion (Scheme 2). In this case a higher concentration
of water leads to higher hydration of the hydroxide ion.
Respectively, on one hand, activation enthalpy of the
nucleophilic attack rises up, and on the other hand, the
difference between hydration entropy of the hydroxide ion
and of the activated complex decreases (activation entropy
becomes less negative).
(2) Gordon A. J.; Ford R. A. The Chemist’s Companion: A Handbook of
Practical Data, Techniques, and References; John Wiley: New York, 1972.
(3) Palm, V. A. The Base of the QuantitatiVe Theory of Organic Reactions;
Khimiya: Leningrad (SU), 1977.
(4) SGWIN Software: STATGRAPHICS Plus for windows; Manugistics Inc.:
Rockville, MD.
Comparison of the kinetic and thermodynamic parameters
of the main process of the stilbene synthesis (Wittig-Horner
reaction) and the side reaction (hydrolysis of BP) permits
us to note some important conclusions about the possibilities
240
•
Vol. 7, No. 3, 2003 / Organic Process Research & Development

(4) Real decisions about process realization in industrial
conditions should be made on the basis of process simulation,
after a compulsory test of the mathematical model in pilot
reactors.
Measurement of the stilbene synthesis reaction heat was
carried out under the recommended conditions (Figure 7).
It can be seen that the reaction heat flow corresponds with
the kinetics of the high-order reactionsit is very high at the
beginning and remains low for a long time.
Process Simulation. Process simulations were carried out
by numeric integration in Excel. The integration constant of
time was 1 s. At every integration step we calculated the
temperature of the reaction mass, water content, and reagent
concentrations at the beginning of the second. On the basis
of these data, constants of the main and side processes were
determined, and changes of all reagents and all product
concentrations up to end of the second step were calculated.
After evaluating the change of the product concentration we
calculated the heat of reaction during the second step. Using
the reactor parameters and temperature of the cooling agent
we defined the heat flow to the jacket during the time interval
and the amount of heat accumulated in the reaction mass.
From the Cp of the reaction mass we calculated the
temperature change during the second step.
Figure 7. Heat flow and reactor temperature in the stilbene
synthesis.
to improve the selectivity of the stilbene synthesis.
(1) Because the activation energy of the BP hydrolysis is
higher than that of the Wittig-Horner reaction, the stilbene
synthesis should be performed at the lowest temperature
as possible.
(2) Sensitivity of the Wittig-Horner reaction to water
addition (reduction of the reaction rate constant) is higher
than that of the BP hydrolysis. Consequently, stilbene
synthesis should be carried out with the water content as
low as possible.
(3) Second-order reaction kinetics of the Wittig-Horner
reaction with respect to BA makes the rate of this reaction
very sensitive to the BA concentration. Therefore, stilbene
synthesis should be done with the concentration and excess
of BA as high as possible.
Now we have all the information to go to the next step
and to calculate the temperature of the reaction mass, water
content, and the reagent concentrations at the beginning of
the new interval. This procedure was repeated for the reaction
Figure 8. Three-dimensional plot of the reactant (BA) at 1691 cm-1 during 600 s.
Figure 9. Three-dimensional plot of the product formation (stilbene) at 942 cm-1 during 600 s.
Vol. 7, No. 3, 2003 / Organic Process Research & Development
•
241

Figure 10. Reactor temperature, theoretical - from simula-
tion; experimental - from LabMax.
Figure 12. Reactant concentration, theoretical - from simula-
tion; experimental - from ReactIR profile.
the IR profiles were fitted to the concentration by multiplica-
tion of these values by a constant (extinction coefficient)
that was the same for all the experimental values that were
obtained (but different for the reactant and for the product).
The yield that was obtained in this experiment was 94%,
whilst the calculated yield was 95%.
The same accordance was obtained in other experiments
at different temperatures and reagent concentrations.
The model was also tested on a large-scale production
reactor (8 m3). It was calculated that performing the process
in this reactor under adiabatic conditions would lead to a
temperature rise of up to 70 °C and thus to a decrease of the
yield to 80%. Exactly these results were obtained during
preparation of the first industrial samples of the stilbene. To
achieve a higher yield of the product, an industrial reactor
will be equipped by a more effective heat exchanger.•
Figure 11. Product concentration, theoretical - from simula-
tion; experimental - from ReactIR profile.
time of 1 h; after that, the yield of the desired stilbene was
calculated.
Conclusions
• In this study the mechanism of a Wittig-Horner reaction
Validation of the Simulation Process with the ReactIR
and the LabMax. To validate the simulation results we
carried out few experiments in the LabMax 1-L reactor
equipped with the ReactIR Dicomp probe. We have followed
the reaction temperature and the profiles of the reactant and
the product during the reaction. These values were compared
to the simulated values.
The reactants were dissolved in the reaction solvent at
50 °C, without the KOH solution, and then we cooled the
jacket temperature to ∼-10 °C as rapidly as possible. When
the reactor temperature reached the desired value (30 or 40
°C), we added the KOH solution all at once. When the reactor
temperature reached the maximal value, we set the temper-
ature to the starting value (in the reactor). At that time we
collected spectra in the rapid mode (eight scans at a resolution
of 8 cm-1, 2.8 s between the spectra). At the end of the
reaction, a sample for the final stilbene concentration was
analyzed for calculation of the yield. Figures 8 and 9 show
the 3D plots of the reactant and the product, respectively.
The following figures (10-12) show the degree of fit of
the simulated values to the experimental values of a sample
experiment (at 30 °C). The values that were obtained from
was found after a thorough kinetic and thermodynamic
research.
• The optimal conditions for the reaction were obtained.
• A mathematical model that enables the planning of the
industrial equipment was prepared.
Experimental Section
Materials. Solvents and reagents were obtained from
commercial sources and used without further purification.
Analytics. HPLC analysis was performed using Varian
9012 with Autosampler 9300 and UV-vis detector 9050.
Kinetic Measurements of the Stilbene Synthesis. The
solvent (between 200 and 500 mL) was added to a triple-
jacket reactor of 1 L (with thermostatic unit CC 250 of
Huber). The following materials were then added: benz-
aldehyde (from 0.2 up to 0.4 mol), benzyl phosphonate (about
0.005-0.01 mol), water (if needed), and internal standard;
at that time the mixture was then heated to 40 °C to dissolve
the components. After a clear solution was obtained, the
reaction mass was cooled to the desired temperature, and
with a high agitation rate (turbine agitator at about 1000 rpm)
an excess of the KOH was added. Sampling was begun
242
•
Vol. 7, No. 3, 2003 / Organic Process Research & Development

immediately, and for these, the reaction was stopped by a
dilution with a mixture of a 1:1 ratio of acetonitrile:phosphate
buffer, pH ) 7.
was immersed in the LabMax reactor during all the experi-
ments. The solvent of the reaction was used as the blank for
the IR spectra.
Kinetic Measurements of the Benzyl Phosphonate
Hydrolysis. In this case the same procedure was used, but
without adding the benzaldehyde.
RC-1: Automatic Calorimetric Reactor (2 L) of Mettler-
Toledo.
LabMax: Automatic Reactor (1 L) of Mettler-Toledo.
ReactIR 1000: For collecting FTIR spectra in-situ
through Dicomp probe (diamond wafer used as a multiple
reflection attenuated total reflectance (ATR), C-276 super
alloy body, MCT detector) of Applied Systems. The probe
Acknowledgment
We thank all the people from Makhteshim Chemical
Works who made the efforts to that study: Amir Etsion the
senior engineer of the project, Dali Gelen and Alex Altshuler
who performed the extensive kinetic work, and Yossi Balilty,
Ilan Fima for the LabMax and RC-1 experiments.
Received for review August 27, 2002.
OP020075I
Vol. 7, No. 3, 2003 / Organic Process Research & Development
•
243