3,4,5-tri-dodecyloxybenzoic acid: Optimisation and scale-up of the synthesis

Article abstract of DOI:10.1021/op000066e

The synthesis of tris-0-dodecyl-gallic acid (3,4,5-tri-dodecyloxybenzoic acid) - a versatile building block for organic liquid crystalline materials - has been selected for fine chemical scale-up. A large-scale procedure of the alkylation of methyl gallat

Full text of DOI:10.1021/op000066e

Organic Process Research & Development 2001, 5, 54−60
3,4,5-Tri-dodecyloxybenzoic Acid: Optimisation and Scale-Up of the Synthesis
M. C. Hersmis,‡ A. J. H. Spiering, R. J. M. Waterval, J. Meuldijk, J. A. J. M. Vekemans, and L. A. Hulshof*
EindhoVen UniVersity of Technology, Laboratory of Macromolecular and Organic Chemistry and Process DeVelopment
Group, Department of Chemical Engineering and Chemistry, P.O. Box 513, 5600 MB EindhoVen, The Netherlands
Scheme 1
Abstract:
The synthesis of tris-O-dodecyl-gallic acid (3,4,5-tri-dodecyl-
oxybenzoic acid)sa versatile building block for organic liquid
crystalline materialsshas been selected for fine chemical scale-
up. A large-scale procedure of the alkylation of methyl gallate
was optimised with experimental design techniques. Apart from
the solvent effect, also the temperature, phase-transfer catalyst,
stirring speed, and amount of base were found to be most
significant for the reaction rate. Reaction calorimetry revealed
no excessive exothermic reaction steps in the process. Reaction
kinetics on the alkylation reaction was studied as a function of
particle size distribution of the base, potassium carbonate, and
formation of carbon dioxide. Combination of all experimental
results has debouched into a master recipe for kilogram-scale
synthesis in a 10 dm3 fully automated (semi)batchwise operated
reactor.
wise operated reactor.7 The newly developed process (master
recipe) should be: (1) selective (affording highly pure
material in high yield), (2) rapid (high conversion rates), (3)
cheap, (4) safe, and (5) environmentally acceptable.
Introduction
Optimisation Studies
Process research and development for the production of
fine chemicals asks for a completely different approach than
that for bulk chemicals. Bulk chemicals are produced in
large-scale continuous processes, based on detailed, quantita-
tive insights into the chemical and physical aspects of the
process and on rigorous process development for one specific
product.
The described tri-alkylation3 of methyl 3,4,5-trihydroxy-
benzoate or methyl gallate (1) with 1-bromododecane,
depicted in Scheme 1 uses 5 mol equiv of potassium
carbonate in cyclohexanone as solvent at a 0.1 M concentra-
tion of 1. The reaction times are approximately 40 h, and
the resulting products are purified by column chromatogra-
phy.3
In contrast, fine chemicals are usually produced in
multipurpose (semi)batchwise operated equipment. New fine
chemicals ask for a short time to market and have a relatively
short lifetime in the market as compared to that for bulk
chemicals. This makes batch-process design quite challenging
due to lack of design tools and generic methodologies when
compared to continuous processing.
As part of our program to define design tools and a
methodology for fine chemical scale-up, tris-O-dodecyl-gallic
acid (3,4,5-tri-dodecyloxybenzoic acid) was selected as
target. Tris-O-dodecyl-gallic acid (TDGA) is a versatile
building block for the flexible part of discotic liquid
crystalline materials.1-4 Its versatility and broad scope are
reflected in diverse applications such as helical tobacco
mosaic virus models,5 intra- and intermolecularly hydrogen-
bonded supramolecular polymers,6 and C3-symmetrical super
helices.3
The conversion rate, throughput, and purification as
reported by Palmans et al.3 are not attractive for large-scale
(1) Discovered in the nineteenth century by Lehman and Reinitzer2, liquid
crystals constitute a class of molecules sharing mobility and orientational
order. Molecular anisotropy or a dichotomy of the structure may provide
this ordering. The dichotomy relates to different structural properties (e.g.
rigid and flexible)3 or different chemical properties (e.g. hydrophobic or
hydrophilic).4 Liquid crystalline materials consist of two domains: the rigid
part of the molecule, tending to aggregate into large stacks, and the flexible
part, which can be derived from TDGA. Liquid crystals exhibit both fluid
and solid properties over a certain temperature range. Liquid crystal display
(LCD) technology has found new applications in a diverse range of consumer
goods, therefore the global demand for new applications of LCD technology
is increasing.
(2) (a) Reinitzer, F. Monath. Chem. 1888, 9, 421. (b) Collins, P. J. Liquid
Crystals, IOP publishing Ltd, Princeton University Press: New Jersey, 1990,
24-34.
(3) Palmans, A. R. A.; Vekemans, J. A. J. M.; Fischer, H.; Hikmet, R. A.;
Meijer, E. W. Chem. Eur. J. 1997, 3, 300.
(4) Schenning, A. P. H. J.; Elissen-Roma´n, C.; Weener, J.-W.; Baars, M. W.
P. L.; van der Gaast, S. J.; Meijer, E. W. J. Am. Chem. Soc. 1998, 120,
8199.
An extensive study was initiated to design the optimal
parameters affording a robust process to produce TDGA on
a kilogram scale in a 10 dm3 fully automated (semi)batch
(5) Percec, V.; Heck, J.; Lee, M.; Ungar, G.; Alvarez-Castillo, A. J. Mater.
Chem. 1992, 2, 1033.
(6) Sijbesma, R. P.; Beijer, F. H.; Brunsveld, L.; Folmer, B. J. B.; Hirschberg,
K.; Lange, R. F. M.; Lowe, J. K. L.; Meijer, E. W. Science 1997, 278,
1601.
* Author for correspondence: E-mail: L.A.Hulshof@tue.nl.
‡ E-mail: mchersmis@yahoo.com.
(7) The description of this reactor has been published: Soldaat, A.
Chemisch2Weekblad. 1999, 8, 18.
54
•
Vol. 5, No. 1, 2001 / Organic Process Research & Development
10.1021/op000066e CCC: $20.00 © 2001 American Chemical Society and The Royal Society of Chemistry
Published on Web 12/06/2000

Table 1. Results prior to factorial designa
exp
solvent
catalyst
99% conversion (h)
1
2
3
4
5
6
7
cyclohexanone
MIBK
MEK
MIBK
MEK
MIBK
MIBK
-
-
-
5
>48
>48
2
TBAIb
TBAI
4
TBABc
Aliquat 336e
3d
1.5
a All experiments were performed under reflux conditions with the same
batches of 1, dodecyl bromide, and base. b Tetrabutylammonium iodide. c Tet-
rabutylammonium bromide. d Poor agitation. e Methyltrioctylammonium chloride.
synthesis. Therefore, the reaction was studied on lab-scale
with respect to the five criteria mentioned above. Initially,
other solvents than those described in the literature,8 that is,
acetone, DMF, and cyclohexanone, were investigated. Apart
from aldol-condensation sensitive acetone and toxic DMF,
other polar aprotic solvents were studied with special
emphasis on the reduction of reaction time. Ketones with
increasing boiling points, such as methyl ethyl ketone, methyl
isobutyl ketone, and cyclohexanone were selected. A higher
reaction rate was expected, provided that the solubility of
the reactants and intermediates remains sufficiently high. To
further accelerate the alkylation of 1, the influence of a phase-
transfer catalyst (PTC) on the conversion rate was studied.
A set of experiments was designed to screen the solvent
and phase-transfer effects rapidly. The experiments, collected
in Table 1, were all performed under reflux conditions with
10 mol equiv of potassium carbonate and 10 mol % of PTC,
based on the amount of methyl gallate. The reaction was
Figure 1. Course of three consecutive alkylations of methyl
gallate versus time. Reaction conditions: 0.4 M methyl gallate,
6.1 mol eq K2CO3, 0.05 mol eq TBAB and 3.2 mol eq
1-bromododecane in MIBK were heated to reflux (t ) 0 min.)
and maintained at T ) 115 °C for 3 h. The moment at which
reflux temperature is reached is t ) 0.
Scheme 2
1
monitored by thin-layer chromatography and H NMR. A
dramatic effect of the phase-transfer catalyst on the conver-
sion rate was observed in this preliminary study as displayed
in Table 1.
The combination of MIBK with TBAB was chosen as
the basis for experimental design. Although Aliquat 366
appeared to be more difficult to remove in the work-up, it
was included in the experimental design to study phase-
transfer catalytic effects.
The pathway followed by the three consecutive alkylations
was studied with 1H NMR spectroscopy. The first O-
alkylation of 1 did almost exclusively occur at the para
position. As a consequence, only asymmetric bis-alkylated
methyl gallate was observed, see Scheme 2. An example of
the progress of alkylation with time is displayed in Figure
1. The reaction is started at room temperature, and the
reaction mixture is heated to reflux. Only mono-alkylation
occurs at lower temperatures. The subsequent reactions start
upon reaching reflux temperature. Tri-alkylation was com-
pleted after 2.5 hours.11
The first three entries show that the alkylation is only
approximately 10 times faster in cyclohexanone at 155 °C
than either in refluxing methyl isobutyl (MIBK) or refluxing
methyl ethyl ketone.9 The data in Table 1 point out that
cyclohexanone is a poor solvent for the alkylation of 1,
probably due to the poor solubility of either base or (mono-
alkylated) intermediate salts of 1 and the limited polarity.
However, entries 4-7 clearly show that phase-transfer
catalysis is beneficial. Although Aliquat 336 was the most
active PTCs2 times as active as TBAI and 3 times as active
as TBABsthe latter was selected as PTC. TBAB is easily
available in large quantities and lower-priced per mol than
TBAI, and above all, TBAB is more easily removed in down-
stream-processing than Aliquat 336, which was difficult to
separate from compound 2, according to 1H NMR analysis.10
Fractional Factorial Design
On the basis of the exploratory results collected in Table
1, the experimental design was set up for the alkylation
(8) (a) Allen, C. F. H.; Gates, J. W. Organic Syntheses 1952, 3, 140. (b)
Maltheˆte, J.; Tinh, N. H.; Levelut, A. M. J. Chem. Soc., Chem. Commun.
1986, 1548. (c) Johansson, G.; Percec, V.; Ungar, G.; Abramic, D. J. Chem.
Soc.; Perkin Trans. 1. 1994, 447.
(9) Since the temperature difference is more than 40 °C, one would expect a
larger difference in reaction rate. Based on an activation energy of
approximately 90 kJ/mol per mol 1-bromododecane in a temperature range
of 0-100 °C a considerably shorter reaction time was expected in
cyclohexanone. See also Morrison, R. T.; Boyd, R. N. Organic Chemistry,
Allyn and Bacon, Inc., Boston, 1987, 55-58.
(10) To enhance reactivity and catalyst separation see (a) Halpern, M. E.;
Grinstein, R. Spec. Publ. -R. Soc. Chem. 1999, 236, 30. (b) Halpern, M.
E.; Grinstein, R. Spec. Chem. 1999, 19, 204. (c) Halpern, M. E. ACS Symp.
Ser. 1997, 659, 97.
(11) The reaction constants for each of the three consecutive alkylations are not
yet determined in view of the complexity of the system.
Vol. 5, No. 1, 2001 / Organic Process Research & Development
•
55

Table 2. Variables and minimum and maximum levels used
in the factorial designa
Xi
variable
(-)
0
(+)
X1 conc methyl gallate (mol/L)
X2 temperature (°C)
0.2
95
0.3
105
0.4
115
X3 type of PTCb
TBAB 0.5/ Aliquat
0.5c
6.0
0.075
3.2
336
8.0
0.10
3.3
X4 amount of K2CO3 (mol equiv)
X5 amout of PTC (mol equiv)
X6 amount of C12H25Br (mol equiv)
X7 stirring speed (rpm)d
4.0
0.05
3.1
0
Figure 2. Results of fractional factorial design after 1.5 h.
300
600
Table 4. Results factorial design: estimated effects for the
a All experiments were performed with the same batches of chemicals. b Phase-
transfer catalyst: tetrabutylammonium bromide or methyltrioctylammonium
chloride. c Mixture of TBAB and Aliquat 336 (mol equiv) d Magnetic stirrer
equipped with 20 mm stirrer bar.
experimental variablesa
effect (%)b
Xi
variable
1.5 h
std error
3 h
std error
X2 temperature
53.5
21.3
26.0
10.3
5.5
5.5
5.5
5.5
5.5
5.5
5.5
2.5
48.6
24.9
18.6
13.4
-12.1
-4.6
4.6
68.7
0.89
0.72
12.3
6.3
6.3
6.3
6.3
6.3
6.3
6.3
2.9
Table 3. Fractional factorial design: experimental matrix
and results
X7 stirring speed
X4 amount of K2CO3
X5 amout of PTC
X6 amount of C12H25Br -13.3
X3 type of PTC
response
variables
yield of 2 (%)
-5.8
3.5
54.2
0.93
0.79
11.0
X1 conc methyl gallate
exp X1 X2 X3 X4 X5 X6 X7
1.5 h
3 h
mean value â0
R2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
-
+
-
+
-
+
-
+
-
+
-
+
-
+
-
+
0
-
-
+
+
-
-
+
+
-
-
+
+
-
-
+
+
0
-
-
-
-
+
+
+
+
-
-
-
-
+
+
+
+
0
+
-
-
+
+
-
-
+
-
+
+
-
-
+
+
-
0
+
-
+
-
-
+
-
+
-
+
-
+
+
-
+
-
0
+
+
-
-
-
-
+
+
-
-
+
+
+
+
-
-
0
-
+
+
-
+
-
-
+
-
+
+
-
+
-
-
+
0
15
17
93
98
38
13
30
99
7
60
96
54
31
22
93
68
63
71
62
30
37
100
99
57
29
57
99
10
91
100
77
53
24
100
88
86
86
82
Q2
RSD
a Number of experiments ) 16 (excluding 3 centre points). b Estimated
standard error of the regression coefficient (scaled and centered).
considering previous experiments and the five constraints
mentioned in the Introduction.
The results are depicted in Figure 2, and the statistical
analysis and significant influences of this factorial designs
using the Modde program14sare shown in Table 4. The effect
of each variable Xi has been calculated using a polynomial
function of the seven experimental variables.
The process parameters in Table 4 are ranked according
to the estimated and calculated effects. Temperature, stirring
speed, and the amount of base were found to be the most
important parameters. The concentration of methyl gallate
and the type of catalyst are less important parameters.
Conclusions could be drawn with respect to the relevance
of the parameters. Moreover, due to the fold-over approach,
all effects are free from confoundings of first-generation
interactions.15
0
0
0
0
0
0
0
0
0
0
0
0
0
0
reactions of 1. The influence of seven critical process
parameters was studied with 16 experiments, a fold-over,
and three centre point experiments,12 see Table 2.
The conversion and selectivity of each experiment were
determined by 1H NMR analysis at 1.5 and 3 h time intervals,
enabling a seven-factor set up with two responses, in which
the reaction time is used to indicate the time elapsed when
the response of product yield is measured. The experiments
have been performed in a random order, and the experimental
matrix and responses are depicted in Table 3.13 The selection
of the levels for the different factors has been carried out
A high concentration of methyl gallate would result in
more space-time-yield, and the TBAB phase-transfer catalyst
can be separated from the product more easily than Aliquat
336 as described above.
(13) The experimental design was generated according to the lines for a fold-
over design indicated by R. Carlson, Design and Optimization in Organic
Reactions, Elsevier: Amsterdam, 1990, 141-149.
(12) Box, G. E. P.; Draper, N. R. Empirical model-building and response
surfaces, John Wiley & Sons: New York, 1987.
(14) Modde 4.0, Umetri AB, Umeå, Sweden.
56
•
Vol. 5, No. 1, 2001 / Organic Process Research & Development

Table 5. Experiments based on the results of the factorial
design
variable
exp A
115
600
8
exp B
115
600
4
exp C
115
600
6
0.05
3.1
TBAB
0.4
1.5
temperature (°C)
stirring speed (rpm)a
amount of K2CO3 (mol equiv)
amout of PTC (mol equiv)
amount of C12H25Br (mol equiv)
type of PTC
concd methyl gallate (mol/L)
reaction time (h)
0.10
0.05
3.1
TBAB
0.4
3.1
TBAB
0.4
1
6
a Magnetic stirrer equipped with 20 mm stirrer bar.
Table 6. Predicted and obtained results for the optimising
experiments
response
Figure 3. The response surfaces X2 (temperature from 95 to
115 °C) vs X7 (stirring speed from 0 to 600 rpm.) for Y1.5 (1).
Other variables Xi ) 0.
yield predicted (%)
yield observed (%)
expa
1.5 h
3 h
1.5 h
3 h
exp A
exp B
exp C
100b
85
98
100b
100
100
100c
73
100c
89
Table 4 also provides information on the optimal reaction
conditions. Operating at higher temperatures, stirring at maxi-
mal speed, using a large amount of base and catalyst, and
working with a small excess of 1-bromododecane should
result in a high production rate of 2. The regression analysis
of this factorial design resulted into two models that would
fit the experimental data as given in the following equations:15
100d
100d
a Variables in Table 5. b Actual prediction was above 100%. Reaction was
c
d
completed within 1 h. Reaction was completed within 1.5 h.
per mol 1, respectively, were rather high. Particularly, a large
amount of base is unfavourable for scale-up. Experiment B
was set up to investigate whether smaller amounts of base
and catalyst, 4 and 0.05 equiv, respectively, could be used
without indulging too low reaction rates. The time needed
for complete conversion in this experiment was 6 h. The
reaction rate dropped dramatically, so a final experiment C
with 6 equiv of base and 0.05 equiv of catalyst was carried
out. The alkylation proceeded to completion within 1.5 h,
and the crude yield based on 1 was nearly quantitative, that
is, 99% before purification.
The methyl 3,4,5-tridodecyloxy-benzoate (2) was saponi-
fied with either sodium or potassium hydroxide in boiling
ethanol within 2 h, acidification with hydrochloric acid,
followed by cooling gave an off-white lumpy crystalline
material, TDGA (3) in good yield (86-90%).
Y1.5 ) 54.2 + 26.8X1 + 1.8X2 - 6.6X3 + 10.6X4 + 13.0X5 +
5.1X6 - 2.9X7 (1)
Y3 ) 68.7 + 24.3X1 + 2.3X2 - 6.1X3 + 12.4X4 + 9.3X5 +
6.7X6 - 2.3X7 (2)
wherein:
Y1.5, Y3 ) predicted space-time-yield at 1.5 and 3 h, respectively
Xi ) variables described in Tables 1 and 4
See Figure 3 for an example of the response surfaces of
variables X2 and X7. On the basis of the results described in
the parts Optimisation Studies and Fractional Factorial
Design, three additional experiments were carried out to
verify the model eqs 1 and 2 and to fine-tune the reaction
conditions. The experimental conditions of these three
experiments are presented in Tables 5 and 6.
The experimental conditions in Tables 5 and 6 were the
desired ones according to the model eq 1 and in view of
further scale-up (heat effects and productivity), leading to
1.5 h reaction time for complete alkylation. A synthesis was
performed, and the reaction time turned out to be 1 h as
was predicted by the model eq 1. This result confirmed that
the required reaction conditions were found. However, the
amounts of base and catalyst used, 8 and 0.10 mol equiv
Thermochemical Properties
The selected reaction conditions for the synthesis of
TDGA on lab-scale were determined as described above. A
100 g scale experiment was carried out applying these
reaction conditions in a 1 dm3 reactor, and surprisingly the
reaction rate was much lower than in the smaller-scale
experiment. Only 25% trialkylated 2 was formed, in 5 h
reaction time. On the 1 dm3 (100 g) scale the same batches
of starting materials were used as in the optimisation studies
except for the potassium carbonate. Down-scaling to the
initial lab-scale with identical batches of starting materials
gave the same poor results. Therefore, it was concluded that
the quality of K2CO3 was the only different factor between
these experiments and the optimisation studies. Mass transfer
of potassium carbonate turned out to be the most critical
factor for the reaction rate. Grinding the base before use gave
(15) Supplementary information on the statistics is available. Modeling including
an interaction term of X2 X7 gives some less of fit for the Y 3.0 response
as well as a model excluding X1 and X3. The best fit was obtained when
X2 was included as a quadratic term (R2 ) 0.94, Q2 ) 0.85 for both
responses).
Vol. 5, No. 1, 2001 / Organic Process Research & Development
•
57

Table 7. RC1 results: heat of reaction and adiabatic
temperature rise
-∆Hr‚Conc
∆Tadia
)
Cp
concentration )
0.26 mol/kg, Cp ) 2 kJ/kg‚K
∆Hr (kJ/mol)
∆Tadia (K)
trialkylation
-216.6
38.4
-98.4
-92.1
28
-5
13
CO2 formation
solvatation NaOH
neutralization HCl
12
Scheme 3
Figure 4. The temperature profile during the alkylation
reaction. Reaction conditions: 0.4 M methyl gallate, 6.1 mol
eq K2CO3, 0.05 mol eq TBAB and 3.2 mol eq 1-bromododecane
in MIBK were heated to reflux and reflux was maintained for
3 h.
Disproportionation of KHCO3 yields water besides carbon
dioxide. This water forms an azeotrope with MIBK,17 which
is clearly illustrated in Figure 4. The reaction started at 115
°C; during the reaction the boiling point of the azeotrope
dropped to 112 °C.
the same satisfactory results as before, that is, complete
conversion within 2 h, on both 10 and 100 g scale. A particle
size study before and after grinding demonstrated that an
average particle size of approximately 125 µm or smaller is
needed for complete reaction to TDGA within 1.5 h.
Both the tri-alkylation and the saponification/acidification
steps are exothermic reactions. To enable further scale-up
of the synthesis of TDGA, insight into the heats of reaction
and an estimation of the adiabatic temperature rise are
required. The thermochemical data for the different reactions
was determined with reactor calorimetry. Instead of perform-
ing the reaction under reflux conditions, the alkylation was
studied at 110 °C to improve the accuracy of the RC1
measurements. The results of the alkylation reaction and
saponification/acidification are listed in Table 7.
Although the tri-alkylation step has a reasonably high heat
of reaction (∆Hr ) -216.6 kJ/mol per mol methyl gallate),
the corresponding adiabatic temperature rise of approximately
30 °C is not problematic during scale-up. The first alkylation
step is rather slow at a relatively low concentration since
both methyl gallate and K2CO3 must solubilise from their
solid state.
Finally, it was concluded that neither release of CO2 nor
the azeotrope formation was significant with respect to the
tri-alkylation rate, yield, and scale-up.
One Kilogram Scale
The insights and results described above were translated
into a master recipe for a 1 kg product-scale. The starting
materials (methyl gallate, K2CO3, TBAB, MIBK, and
1-bromododecane) were added to a fully automated 10 dm3
1
reactor. The reaction was monitored by H NMR, temper-
ature, heat flow, stirring speed, and energy dissipation due
to stirring. The alkylation was complete within 2 h. Extrac-
tion and solvent evaporation yielded a brown oil, which was
saponified with potassium hydroxide in ethanol within 2 h.
Neutralisation and cooling to 45 °C afforded an off-white
lumpy precipitate, TDGA (3). Unfortunately, crude TDGA
contained salts, and a purification step was needed. The
product was dissolved in dichloromethane, and the insoluble
salts were filtered off. Evaporation of the filtrate resulted in
pure and white TDGA in 92% yield based on 1 and 99%
purity.
The insoluble salts in the residue were identified as mainly
consisting of potassium chloride. KCl has a poor solubility
in 92% ethanol, that has been used to wash the product
during filtration. A few small-scale experiments were carried
out to investigate the saponification with lithium hydroxide.
Although the reaction times for complete hydrolysis varied
from 3 to 4 h, no insoluble salts were formed. LiCl is very
soluble in 92% ethanol and hence did not accumulate in the
The heat of dissolution of the solid NaOH in ethanol in
the second step can be used to heat the reaction mixture.
The saponification is performed at 78 °C under reflux
conditions. The subsequent neutralisation can be controlled
without difficulties by dosing the hydrochloric acid solution.
It was concluded that these three exothermic reaction steps
could be managed safely on larger scale.
The reaction calorimetry experiments also showed an
endothermic effect after completion of the tri-alkylation, due
to gas formation. The released gas was unambiguously
identified as carbon dioxide (see Scheme 3). CO2 formation
in time was studied by volumetric techniques. It was
relatively constant in time and temperature dependence.16
(16) The release of CO2 was very small in a temperature range from 25 to 80
°C and increased to 2.0 dm3/h at 110 °C and atmospheric pressure on 100
g scale. The total amount of CO2 evolved during the alkylation was 10-15
mol % of the KHCO3 formed in the alkylation. This amount appeared to
be constant in time. CO2 is presumably formed by disproportionation of
KHCO3 as depicted in Scheme 3.
(17) Horsley, L. H. Azeotropic data, American Chemical Society: Washington,
D. C., 1952; Azeotrope MIBK/H2O, Wt. % H2O ) 24.3.
58
•
Vol. 5, No. 1, 2001 / Organic Process Research & Development

Scheme 4. Block diagram of master recipe for the
alkylation of methyl gallate
The insights and knowledge acquired will be used to
expand the current scale-up methodology to industrial scale.
Experimental Section
All reagents and solvents were used without further
purification. The water content in the solvent MIBK (4-
methyl-2-pentanone, Aldrich) was analysed by a Karl Fisher
titration and was within specification. All proton NMR
spectra were recorded on a Varian 300 or 400 MHz
spectrometer with TMS as internal standard. Mass spectra
were measured on a Perseptive Voyager DE-PRO Maldi-
TOF. Reaction calorimetry was performed with a Mettler-
Toledo RC1e Reaction Calorimeter equipped with a 2 dm3
SV01 glass reactor equipped with a pitch blade impeller.
Schlummberg volumetric gas meter was used for the carbon
dioxide measurements. Large-scale reaction was carried out
in a 10 dm3 fully automated (semi)batch-wise operated
reactor.7 This Belatec reactor is able to perform under a
variety of conditions: a temperature range from -50 to 200
°C, solvent distillation, different agitator types, a pressure
range from 0.03 to 1.1 bar. The reactor is controlled by a
PLC, a special computer, which monitors physical data (batch
history) and secures optimal process and safety conditions.
Representative Procedure for Methyl 3,4,5-tri-dode-
cyloxy-benzoate (2) for Optimisation Studies. Methyl
gallate (1) (Fluka, 2.95 g, 16 mmol), K2CO3 (13.27 g, 96
mmol), tetrabutylammonium bromide (TBAB, Fluka, 0.26
g, 0.8 mmol), MIBK (40 mL) and 1-bromododecane (Acros,
12.37 g, 50 mmol), were added to a 100 mL three-necked
flask. Subsequently, the reaction mixture was heated to reflux
and stirred for 2 h. The conversion of the reaction was
monitored by TLC (eluent hexane:ethyl acetate, 24:1) and
1H NMR. Upon completion the brown mixture was cooled
below 100 °C, and water (40 mL) was added. The aqueous
layer was separated, and the organic layer was washed with
water (40 mL), diluted HCl solution (40 mL 1.0 M), and
water (40 mL) again. Solvent evaporation resulted in a yellow
oil (11.7 g), which crystallises at approximately T ) 40 °C
into a light brown solid (2). Purification by column chro-
matography (flash SiO2; eluent hexane:EtOAc (96:4)) yielded
product. The TDGA obtained was 99% pure, a small
impurity was tentatively identified as 4,4′,5,5′,6,6′-hexakis-
(dodecyloxy)-biphenyl-2,2′-dicarboxylic acid.18
The master recipe was adapted, and the reaction was
performed on 1 kg scale once more. The block diagram of
the alkylation is shown in Scheme 4.
The second campaign was successfully performed. Using
lithium hydroxide in the hydrolysis step resulted in the
desired product of a perfect quality. The saponification was
completed within 5 h and afforded TDGA without salt
inclusion in 96% yield.
Conclusions
In this report, the scale-up and optimisation of the
synthesis of 3,4,5-tri-dodecyloxybenzoic acid has been
scrutinised. The results obtained from the optimisation study,
reaction calorimetry, saponification, and work-up formed a
firm basis for the master recipe. The 1 kg scale synthesis of
TDGA has been performed successfully (96% yield, 99%
purity). The newly developed process meets all the con-
straints of nowadays; it is selective, highly productive, cheap,
safe, and environmentally acceptable. The salt production
has been reduced to a minimum.
The selected approach for process development and scale-
up demonstrates some important steps, such as:
1. Definition of the important constraints for the TDGA
synthesis.
2. Identification and range definition of critical process
parameters.
1
a white powder (10.2 g): mp 43.2-43.8 °C. H NMR
(acetone-d6, 300 MHz): δ ) 7.27 (s, 2H, ortho-H), 4.11-
3.98 (m, 6H, OCH2), 3.85 (s, 3H, OCH3), 1.90-1.70 (m,
6H, OCH2CH2), 1.55-1.20 (m, 54H, (CH2)8), 0.86 (t, 9H,
CH3). Anal. Calcd for C44H80O5 (MW 689.12): C 76.69, H
11.70. Found: C 77.3, H 11.8.
Purification of Methyl 3,4,5-tri-dodecyloxy-benzoate
(2). Crude yellow 2 (27.4 g) was heated above its melting
point and poured gently into 1 L methanol (tech) under
stirring. The initial yellow clusters were dispersed, and a
white precipitate was formed within an hour. Filtration
yielded 97% pure (based on NMR analysis) material, yield:
25.1 g ) 91%.
General Procedure to Synthesise 3,4,5-Tri-dodecyloxy-
benzoic acid (3). Crude yellow 2 (11.7 g) was dissolved in
ethanol (96%, 40 mL) at 45 °C. Sodium hydroxide pellets
(Merck, 0.77 g, 19.2 mmol) were added, and the mixture
was heated to reflux. After two h of reflux, the reaction
mixture was cooled and acidified with 12 M HCl solution
3. Establishment of the heat and mass transfer limitations
on lab-scale.
4. Verification and fine-tuning on 1 kg scale.
(18) Methyl gallate contains about 1% 4, 4′,5, 5′,6, 6′-hexakis(hydroxy)-biphenyl-
2, 2′-dicarboxylic acid.
Vol. 5, No. 1, 2001 / Organic Process Research & Development
•
59

(1.7 mL, 20.4 mmol). An off-white, lumpy product precipi-
tated from the solution at T ) 45 °C. The precipitate was
filtered and washed twice with water (40 mL) to obtain 3
(9.82 g). Crude compound 3 was taken up in CH2Cl2, and
the insoluble inorganic salts were removed by filtration.
Evaporation of the filtrate yielded the pure white solid 3 (9.5
g, 88% yield based on 1): mp 57.5-58 °C. 1H NMR (CDCl3,
400 MHz): δ ) 7.31 (s, 2H, ortho-H), 4.02 (m, 6H, OCH2),
1.80 (m, 6H, OCH2CH2), 1.50-1.20 (m, 54H, (CH2)8), 0.88
(t, 9H, CH3). Anal. Calcd. for C43H78O5 (MW 675.09): C
76.50, H 11.65. Found: C 76.51, H 11.67. Mass calcd
674.56. Found 675.59 (M + H), 697.57 (M + Na).
Master Recipe to Synthesise 3,4,5-Tri-dodecyloxyben-
zoic Acid (3). Methyl gallate (1) (Fluka, 368.3 g, 2.0 mol),
ground K2CO3 (1660 g, 12.0 mol), tetrabutylammonium
bromide (TBAB, Fluka, 32.0 g, 0.10 mol), MIBK (5.0 L),
and 1-bromododecane (Acros, 1560 g, 6.26 mol) were
introduced in a 10 dm3 automated reactor. Subsequently, the
reaction mixture was heated (in 20 min) to reflux and stirred
(300 rpm) for 2 h. The conversion of the reaction was
monitored by 1H NMR. Upon completion the brown mixture
was cooled to 90 °C, and water (3 L) was added. The
aqueous layer was separated, and the organic layer was
washed with water (3 L), with diluted HCl solution (1.5 L
1.0 M), and twice with water (2 × 1 L) again. Solvent
evaporation of the organic layer resulted in a yellow-brown
oil (2). This was kept in the reactor overnight at 45 °C under
stirring (150 rpm). The oil (2) was dissolved in ethanol (96%,
4.8 L) and saponified with lithium hydroxide (Merck, 97.0
g, 2.3 mol) within 5 h under reflux conditions. The
conversion of the reaction was monitored by 1H NMR. Upon
completion, the reaction mixture was cooled to 60 °C and
acidified by dosing a solution of HCl (260 mL 37% HCl
(Merck) diluted in 250 mL 96% ethanol) in 15 min. An off-
white, lumpy product precipitated from the solution at T )
45 °C. The slurry was filtered and washed three times with
92% ethanol to remove residual salts. White TDGA (3) was
obtained after drying overnight in 96% yield (1300 g) and
99% purity: mp 57.5-58 °C. 1H NMR (CDCl3, 400 MHz):
δ ) 7.31 (s, 2H, ortho-H), 4.02 (m, 6H, OCH2), 1.80 (m,
6H, OCH2CH2), 1.50-1.20 (m, 54H, (CH2)8), 0.88 (t, 9H,
CH3). Anal. Calcd for C43H78O5 (MW 675.10): C 76.50, H
11.65. Found: C 76.47, H 11.46. Mass calcd. 674.56. Found
675.59 (M + H), 697.57 (M + Na).
Acknowledgment
This research is supported by Professor E. W. Meijer
(Laboratory of Macromolecular and Organic Chemistry,
TUE, Netherlands) and DSM Fine Chemicals, Netherlands.
The authors gratefully acknowledge R. Hoekstra of DSM
for performing the RC1 measurements and J. P. A. Custers,
M. Fransen, and B. Smeets for their support in the synthesis
of 2. H. C. C. K. van Bakel of DSM Fine Chemicals is
acknowledged for his assistance with the statistics. J. G. M.
Moerel, H. G. Fiedler, P. H. van Eeten, and A. J. Bombeeck
are acknowledged for all of their efforts and support in
constructing the 10 dm3 Belatec reactor.
Received for review June 13, 2000.
OP000066E
60
•
Vol. 5, No. 1, 2001 / Organic Process Research & Development