High selectivity in the oxidation of mandelic acid derivatives and in 0-methylation of protocatechualdehyde: New processes for synthesis of vanillin, iso-vanillin, and heliotropin

Article abstract of DOI:10.1021/op0000529

New synthetic procedures for vanillin, iso-vanillin, heliotropin, and protocatechualdehyde starting from catechol are described. The utilisation of statistical experimental design and multivariate modelling and the mechanistic interpretation of the acid and base catalysis in the condensation of catechol derivatives with glyoxylic acid and in the regiocontrolled methylation of protocatechualdehyde and of the Cu(II) salt catalysis in the oxidative decarboxylation of mandelic acid derivatives have allowed the development of new highly selective processes.

Full text of DOI:10.1021/op0000529

Organic Process Research & Development 2000, 4, 534−543
High Selectivity in the Oxidation of Mandelic Acid Derivatives and in
O-Methylation of Protocatechualdehyde: New Processes for Synthesis of
Vanillin, iso-Vanillin, and Heliotropin
,
†
‡
,‡
Hans-Ren e´ Bjørsvik,* Lucia Liguori, and Francesco Minisci*
Department of Chemistry, UniVersity of Bergen, All e´ gaten 41, N-5007 Bergen, Norway, and Dipartimento di Chimica del
Politecnico di Milano, Via Mancinelli 7, I-20131 Milano, Italy
Abstract:
Scheme 1
New synthetic procedures for vanillin, iso-vanillin, heliotropin,
and protocatechualdehyde starting from catechol are described.
The utilisation of statistical experimental design and multivari-
ate modelling and the mechanistic interpretation of the acid
and base catalysis in the condensation of catechol derivatives
with glyoxylic acid and in the regiocontrolled methylation of
protocatechualdehyde and of the Cu(II) salt catalysis in the
oxidative decarboxylation of mandelic acid derivatives have
allowed the development of new highly selective processes.
Introduction
Vanillin (3-methoxy-4-hydroxy benzaldehyde) 1, ethyl-
vanillin (3-ethoxy-4-hydroxy benzaldehyde) 2, iso-vanillin
(
(
3-hydroxy-4-methoxy benzaldehyde) 3, and heliotropin
benzo[1,3]dioxol-5-carbaldehyde) 4, are important products
of fine chemical industry. Vanillin 1 is used as intermediate
in pharmaceuticals, and it is one of the most important
flavouring agents in confectionery, beverage, food, and in
perfumery. iso-Vanillin 3 is an important building block used
Scheme 2
in the synthesis of the PDE4 inhibitors,1 Ariflo (SB207499),
,2
2
targeted for treatment of bronchial asthma, and (R)-Rolip-
1
,3
ram that is approved as an antidepressant. Since the
olfactory profile of iso-vanillin 3 varies with the temperature,
it is also suitable as a perfumery active ingredient for the
cosmetic and perfumery industry.
Heliotropin (or piperonal) 4 is also used in perfumery,
cherry and vanilla flavors, and as an intermediate for fine
chemicals. Vanillin is available from two commercial
processes: from catechol 5 (Scheme 1) and by catalytic
oxidation of lignosulfonate. Ethyl-vanillin 2 is produced by
a process similar to that of Scheme 1.
vanillin 1. The selectivity of this dealkylation was improved7
starting from 3-alkoxy-4-hydroxy benzaldehyde, in which
the alkoxy group is different from the methoxy and can
undergo a faster hydrolysis; the only starting material
commercially available on large scale for this purpose is
ethylvanillin 2. Thus, the overall process involves five steps
starting from catechol 5: (i) monomethylation or ethylation
of catechol 5, giving guaiachol 7 and guethol 8, respectively;
(ii) reaction of guaiachol 7 or guethol 8 with glyoxylic acid
9; (iii) catalytic oxidation of the mandelic acid derivatives,
4
Two substantial processes (Scheme 2) have been reported
for the synthesis of iso-vanillin 3: regiocontrolled demeth-
5
6
ylation by sulphuric or Lewis acids of veratraldehyde (3,4-
dimethoxy benzaldehyde) 6, obtained by methylation of
*
Authors to whom correspondence should be addressed. E-mail:
Hans.Bjorsvik@kj.uib.no and Francesco.Minisci@polimi.it.
†
University of Bergen.
Politecnico di Milano.
‡
(
1) DalPiaz, V.; Giovannoni, M. P.; Castellana, C.; Palacios, J. M.; Beleta, J.;
Domenech, T.; Segarra, V. J. Med. Chem. 1997, 40, 1417.
2) Torphy, T. J.; Barnette, M. S.; Underwood, D. C.; Griswold, D. E.;
Christensen, S. B.; Murdoch, R. D.; Nieman, R. B.; Compton, C. H. Pulm.
Pharmacol. Ther. 1999, 12, 131.
1
0 or 11; (iv) methylation of vanillin 1 or ethylvanillin 2;
(
(5) Brossi, A.; Gurien, H.; Rachlin, A. I.; Teitel, S. J. Org. Chem. 1967, 32,
1269.
(
(
3) Braun, M.; Opdenbusch, K.; Unger, C. Synlett 1995, 1174.
4) Bjørsvik, H.-R.; Minisci, F. Org. Process Res. DeV. 1999, 3, 330.
(6) Prager, R. H.; Tan, Y. T. Tetrahedron Lett. 1967, 3661.
(7) Maliverney, C. (Rhone-Polenc Chimie). EP 0 709 361 A1, 1996.
5
34
•
Vol. 4, No. 6, 2000 / Organic Process Research & Development
10.1021/op0000529 CCC: $19.00 © 2000 American Chemical Society and The Royal Society of Chemistry
Published on Web 10/28/2000

Scheme 3
Scheme 4
our attempts to reproduce the reported procedure led to poor
yields (<20%). The synthesis is based on the reaction of
catechol with glyoxylic acid in aqueous basic medium
(
NaOH) and in the presence of Al
similar to the industrial procedure for the synthesis of vanillin
Scheme 1). We have performed a development of this
2 3
O ; it is substantially
(
12
reaction by using statistical experimental design and
1
3
multivariate modelling, since, as mentioned above, the
availability of protocatechualdehyde 13 is a restricting factor
for the success of our new process. An assessment of the
existing synthetic procedure gave that the following variables
could be considered in an optimisation study: the amount
(v) regiocontrolled dealkylation of compound 6 or 12. The
complete process is shown by Schemes 1 and 2. Another
method reported for synthesis of iso-vanillin 3 involves the
of catechol (f
), the amount of glyoxylic acid (x
aluminium oxide (x ), the reaction temperature (x
amount of NaOH (x ). Initially, the variables indicated as
-x were selected as independent variables for the
experimental design, whereas the other variables, f -f were
1
), the volume of water (f
2
), the reaction time
), the amount of
), and the
8
(
f
3
1
mono O-methylation of protocatechualdehyde 13 by using
sodiumhydride and methyliodide in dimethyl sulfoxide, but
is too expensive from industrial point of view. The synthesis
of heliotropin 4 involves the chromic oxidation of piperonyl
2
3
4
x
1
4
1
3
9
alcohol, the condensation of 4-bromo benzo[1,3]dioxole by
selected at convenient fixed levels. A full factorial design at
N-methylformanilide and ozonolysis,10 or chromic(VI)salt
oxidation of iso-safrole. We have designed a new strategy
for the synthesis of vanillin 1, iso-vanillin 3, and heliotropin
4
two levels with two “center” experiments (2 + 2 f 18 run)
was set up. Some introductory experiments were performed
to obtain a preliminary overview of the experimental
variables’ effects on the response parameters: the amount
of recovered catechol (r [mmol]), the amount of 3,4-
dihydroxy mandelic acid (o [mmol]), the selectivity percent
to 3,4-dihydroxy mandelic acid (s [%]), and finally the yield
percent of 3,4-dihydroxy mandelic acid calculated from
catechol (y [%]). These introductory experiments showed that
the isolated product was found to be much more impure when
an excess of NaOH was used in the reaction mixture (implies
that both catechol and glyoxylic acid appear as Na salts)
compared to the experiments performed under less basic
4
from catechol-based general Scheme 3.
We succeeded in the development of an efficient synthesis
of protocatechualdehyde 13 and the selective mono O-
methylation by dimethylsulphate to vanillin 1 or iso-vanillin
3, depending on the reaction conditions. However, attempts
to obtain heliotropin 4 from protocatechualdehyde 13 ac-
cording to the route c of Scheme 3 gave only moderate
results, and we realized that the best way to heliotropin 4
from catechol was to reverse the oxidation and methylation
steps according to Scheme 4.
In any case all of the syntheses in Schemes 3 and 4
involve only three steps, starting from catechol.
conditions. Thus, only the variables x
1
-x
a full factorial design (2 + 1 f 9 experiments), and the
variable “amount of sodium hydroxide” (x ) turned out to
be the fourth variable used at a fixed experimental level (x
f f ). The statistical experimental design with the variables
-x with their adjacent measured and calculated responses
η ) [r, o, s, y] are given in Table 1. The response s,
%]-selectivity, and the variables x -x with their interactions
3
were explored in
3
4
Methods and Results
4
Synthesis of Protocatechualdehyde 13. The synthesis
of 3,4-dihydroxymandelic acid 14 from catechol 5 and
glyoxylic acid 9 (Scheme 3) was previously11 reported, but
4
x
1
3
[
1
3
(8) Kessar, S.; Gupta, Y. P.; Mohammad, T.; Goyal, M.; Sawal, K. K. J. Chem.
Soc., Chem. Commun. 1983, 7, 400.
(
9) Holum, J. R. J. Org. Chem. 1961, 26, 4815.
(12) Box, G. E. P.; Hunter, W. G.; Hunter, J. S. Statistics for Experimenters, An
Introduction to Design, Data Analysis, and Model Building; Wiley: New
York, 1978.
(
(
10) Feugeas C. Bull. Soc. Chim. 1964, 1892.
11) Umemura, S.; Takamitsu, N.; Enomiya, T.; Shirashi, H.; Nakamura, T. (Ube
Industries Ltd.). German Patent DE 28 04 063 B2, 1980.
(13) Martens, H.; Næs, T. MultiVariate Calibration; Wiley: Chichester, 1989.
Vol. 4, No. 6, 2000 / Organic Process Research & Development
•
535

Scheme 5
main direction of the experimental domain. Due to an
aberrant response in Table 1, the inlying values of the CND
plot are fitted by two straight lines in parallel, and not by
1
5
only one straight line. See Box et al.
The obtained model s ) f (x , x , x ) eq 1 holds out
1
2 3
expectations of high selectivity for the reaction 5 + 9 f 14
when one simultaneously (i) increases the quantity of
glyoxylic acid 9 to slightly more than 1 equiv of catechol 5,
Figure 1. CND plot of the model for 3,4-dihydroxy mandelic
acid.
(
ii) adjusts the quantity fraction catechol:aluminiumoxide
Table 1. Experimental conditions with corresponding
measured and calculated responses for the screening and
introductory optimization study of the reaction 5 + 9 f 14
within the range 2.17-2.28, and (iii) adjusts the temperature
to be within the range 50-60 °C. Even though the model
shows that it may be beneficial with a low quantity of
aluminium oxide, experiments have shown that without
aluminium oxide present, the reaction did not proceed at all.
Increased quantity of glyoxylic acid implies double alkylation
on the aromatic ring, with a consequence of a lower
selectivity and higher cost of the process. The model eq 1 is
graphically represented as a response surface projection in
Figure 2, from which the conditions for an optimising
experiment were predicted. The model predicted a selectivity
of 100% for 3,4-dihydroxymandelic acid 14; the selectivity
of the optimised experiment was 90.5% with a conversion
of 78.4%. The unconverted catechol 5 was easily recovered
and recycled by extraction under suitable acidity; the aqueous
phase containing the mandelic acid derivative 14, was
directly utilised for the oxidative decarboxylation to obtain
protocatechualdehyde 13.
(3,4-dihydroxy mandelic acid)
experimental variablesa
responsesb
no.
x
1
x
2
x
3
r
o
s
y
1
2
3
4
5
6
7
8
9
-1
+1
-1
+1
-1
+1
-1
+1
0
-1
-1
+1
+1
-1
-1
+1
+1
0
-1
-1
-1
-1
+1
+1
+1
+1
0
10.34 24.78 70.66 54.57
11.00 18.53 53.81 40.81
15.08 20.67 68.15 45.52
9.22 25.75 71.15 56.71
12.80 20.11 61.67 44.29
10.04 27.99 79.13 61.64
12.15 16.88 50.75 37.17
6.87 22.57 58.56 49.70
12.09 19.76 59.30 43.51
a
In each of the experiments: 45.41 mmol of catechol, 80.37 mmol of NaOH,
5 mL of water, and a reaction time of 24h was used. The other experimental
variables were varied: x : amount of glyoxylic acid /[mmol] [34.96, 45.41],
: amount of aluminiumoxide /[mmol] [22.70, 31.79], x : reaction temperature
5
1
x
2
3
b
/
[°C] [25, 50]. r ) mmol recovered catechol, o ) mmol of 3,4-dihydroxy
mandelic acid, s ) percent selectivity to 3,4-dihydroxy mandelic acid, and y )
yield percent of 3,4-dihydroxy mandelic acid calculated from catechol.
Preliminary attempts of oxidation by oxygen and Cu(II)
catalysis, which is effective for vanillin (Scheme 1) and
heliotropin 4 (Scheme 4), led to poor yields of protocate-
chualdehyde 13 due to the formation of a large amount of a
by-product, which from NMR analysis appears to arise from
oxidative dimerisation of protocatechualdehyde 13 to give
were multivariate-correlated using the multiple linear regres-
14
sion (MLR) method to obtain a regression model as given
in eq 1
1
7 (Scheme 5).
We have therefore developed a different procedure
s ) 63.68 + 1.428 × x - 2.082 × x - 1.708 × x +
1
2
3
1
.275 × x × x + 4.890 × x × x -
1
2
1
3
involving a two-liquid system (water:ethyl acetate) and the
use of a 3-fold excess of CuCl and HCl under nitrogen.
The large excess of the copper salt is used to ensure a
complete conversion of the mandelic acid derivative 14.
Under these conditions the oxidation take place with high
yields (>95%); the organic phase, containg 13, is separated,
and CuCl
air and recycled so that the overall process is catalytic in
CuCl and the actual oxidant is air.
5
.790 × x × x - 3.688 × x × x × x (1)
2
3
1
2
3
2
2
with R of 0.968. Even though cumulative normal probability
15
plot (CND-plot) is a rough method for assessment of model
parameters, it is a suitable method for assessment of such a
simple model as eq 1. The CND plot given in Figure 1 shows
that the â13, â23, and possibly â123 also are the significant
parameters of this model and gives information about the
2
is regenerated in the aqueous phase by bubbling
2
Selective O-Methylation of Protocatechualdehyde 13.
The O-methylation of protocatechualdehyde 13 by dimeth-
ylsulphate in basic medium can lead to three different
products: vanillin 1, veratraldehyde 6, and iso-vanillin 3
(Scheme 6).
(
14) Montgomery, D. C.; Peck, E. A. Introduction to Linear Regression Analysis;
Wiley: New York, 1982; Draper, N. A.; Smith, H. Applied Regression
Analysis, 2nd ed.; Wiley: New York, 1981.
15) Daniel, C. Technometrics 1959, 1, 311. Box, G. E. P.; Draper, N. R.
Empirical Model-Building and Response Surfaces; Wiley: New York, 1987;
pp 128-134.
(
536
•
Vol. 4, No. 6, 2000 / Organic Process Research & Development

Figure 2. Contour projections of the response surface where the response is the selectivity to 3,4-dihydroxymandelic acid 12.
Scheme 6
Table 2. Experimental design with adjacent measured
responses, the quantities of vanillin 1, veratraldehyde,
iso-vanillin 3, and protocatechualdehyde
experimental variablesa
responsesb
no.
x
1
x
2
x
3
x
4
x
5
x
6
x
7
x
8
y
1
y
2
y
3
y
4
1
2
3
4
5
6
7
8
9
-1 -1 -1 -1 -1 -1 -1 -1 6.46 2.09 19.06 63.73
+1 -1 -1 -1 +1 -1 +1 +1 6.15 1.60 31.23 59.75
-1 +1 -1 -1 +1 +1 -1 +1 1.49 0.94 34.43 34.88
+1 +1 -1 -1 -1 +1 +1 -1 3.02 0.63 14.35 77.23
-1 -1 +1 -1 +1 +1 +1 -1 6.86 5.92 21.57 56.70
+1 -1 +1 -1 -1 +1 -1 +1 4.70 4.70 13.85 42.35
-1 +1 +1 -1 -1 -1 +1 +1 4.01 2.76 11.93 74.67
+1 +1 +1 -1 +1 -1 -1 -1 6.74 8.38 26.74 49.95
-1 -1 -1 +1 -1 +1 +1 +1 4.28 2.09 14.52 68.61
+1 -1 -1 +1 +1 +1 -1 -1 6.85 4.07 38.01 40.81
-1 +1 -1 +1 +1 -1 +1 -1 6.35 3.48 32.13 51.39
1
0
1
1
Clearly the formation of veratraldehyde 6 is related to
12 +1 +1 -1 +1 -1 -1 -1 +1 5.58 3.25 26.47 63.89
13 -1 -1 +1 +1 +1 -1 -1 +1 8.10 10.01 23.42 51.31
the conversions; with an excess of dimethylsulphate and
NaOH and complete conversion veratraldehyde 6 is the only
reaction product. Thus partial conversions are necessary in
order to obtain vanillin 1 and iso-vanillin 3. On the basis of
14
15
16
17
18
+1 -1 +1 +1 -1 -1 +1 -1 2.12 2.71 10.80 65.34
-1 +1 +1 +1 -1 +1 -1 -1 7.95 10.01 19.06 56.57
+1 +1 +1 +1 +1 +1 +1 +1 8.32 10.37 25.84 47.04
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0 5.92 4.31 29.78 46.88
0 6.20 4.78 27.88 52.16
the differences in the pK
3 (pKa(3-OH) ≈ 11.6, and pKa(4-OH) ≈ 7.5) we tried to develop
direct O-methylation procedures to iso-vanillin 3 or vanillin
. We carried out some O-methylation preliminary experi-
a
values of protocatechualdehyde
1
a
Generators for the experimental design: x
; x ) x . The present experimental design makes it possible to estimate
the effects of the single variables and the values for the following confounded
two-factor interactions: â12 + â35 + â48 + â67, â13 + â25 + â47 + â68, â23
+ â + â , â + â + â + â , â + â + â + â , â + â + â
5 1 2 3 6 2 3 4 7
) x x x ; x ) x x x ; x )
x
1
x
3
x
4
8
1 2 4
x x
1
+
26
ments on protocatechualdehyde 13, that showed that the
product distribution, the quantity of vanillin 1, iso-vanillin
â
15
+
46
78
14
28
37
56
24
18
36
57
34
17
â
58, â16 + â27 + â38 + â45. Experimental variables [-1, +1] /unit x
time [45, 75]/min., x : reaction time [45, 75]/min., x : amount of NaOH added
during the reaction [0.70, 0.85]/g, x : start amount of NaOH [0.18, 0.23]/g, x
: amount of dimethyl sulphate [2.3, 2.7]/g,
: volum of methylene chloride [14, 20]/mL, x : volum of water [14, 20]/mL.
Confounding patterns for the present design are: y : yield mol % of vanillin
, y yield mol % veratraldehyde 6, y : yield mol % of iso-vanillin 3, y
1
: addition
2
3
3
, and veratraldehyde 6 varied, depending on several of the
experimental variables; the addition time of reagents (x ),
the reaction time (x ), the amount of NaOH added during
the reaction (x ), the start amount of NaOH (x ), the reaction
temperature (x ), the amount of dimethyl sulphate (x ), the
volume of methylene chloride (which revealed better than
ethyl acetate) (x ), and the volume of water (x ). To study
4
5
:
reaction temperature [35, 55]/°C, x
6
1
x
7
8
b
1
2
1
2
:
3
4
3
4
unconverted mol % protocathecuic aldehyde 13.
5
6
7
8
3
vanillin 1 and y ) mol % yield of iso-vanillin 3) to the
these variables’ influence on the selectivity and the yield, a
fractional factorial design of the type 28 , including two
experiments in the centre of the experimental domain was
used. The experimental levels and design with adjacent
measured responses from the experimental runs are reported
in Table 2.
experimental variables are two linear models to which cross-
product terms have been added. However, due to the
resolution (resolution ) IV) of the experimental design that
has been used, all of cross-terms in the two models have a
certain confounding pattern with the other cross-terms. The
regression coefficients were estimated by application of the
-4
1
3
The estimated models [y
1
) f(x
1
-x
8
) and y
3
) h(x
1
-x
8
)]
partial least-squares regression (PLSR) method, and their
significances were determined by using CND plots and
which relate the observed responses (y
1
) mol % yield of
Vol. 4, No. 6, 2000 / Organic Process Research & Development
•
537

Table 3. Estimated model parameter values for eq 2 and eq 3
model parameters for y a
(iso-vanillin)
model parameters for the y b
(vanillin)
3
1
coef./value
std. err.
P
conf. int. (
coef./value
std. err.
P
conf. int. (
-
9
-7
R
R
R
R
R
R
R
R
R
R
0
23.018
0.700
1.150
-3.562
1.062
0.64
0.66
0.66
0.66
0.66
0.66
0.66
0.66
0.66
0.66
3.34 × 10
0.324
0.125
1.514
1.56
1.56
1.56
1.56
1.56
1.56
1.56
1.56
1.56
â
â
â
â
â
â
â
â
â
â
â
0
5.582
0.249
0.256
0.256
0.256
0.256
0.256
0.256
0.256
0.256
0.256
0.256
5.09 × 10
0.611
0.611
0.081
0.047
0.022
0.054
0.092
0.022
0.008
0.158
0.608
0.627
0.627
0.627
0.627
0.627
0.627
0.627
0.627
0.627
0.627
1
1
-0.138
-0.138
0.537
0.638
0.788
0.612
-0.513
0.787
2
2
-
-
3
5
3
1.01 × 10
3
4
0.151
4
5
6.437
2.52 × 10
0.008
0.109
0.193
0.575
5
7
-2.425
-1.212
0.950
12
13
15
24
45
12
24
45
-0.387
0.987
0.413
a
2
2
Statistical products for the model y
3
describing the formation of iso-vanillin (PLS components ) 4, N ) 17, confidence level ) 95%) R ) 0.956, Q ) 0.734,
b
2
RMSEP ) 2.64. Statistical products for the model y
Q ) 0.523, RMSEP ) 1.0247.
1
describing the formation of vanillin (PLS components ) 4, N ) 17, Confidence level ) 95%): R ) 0.905,
2
14
Scheme 7
statistical analysis. The final model- and statistical param-
eters are given in Table 3.
Model for iso-Vanillin 3. The experimental variables,
x
1
, x
determined to have influence on the formation of iso-vanillin
. Three two-factor interaction terms were also determined
to have influence in the model; x , x , and x . The final
2 3 4 5 7
, x , x , x , and x were the variables that were
3
x
1 2
x
2 4
4 5
x
of vanillin with 44.6% conversion of protocatechualdehyde
multivariate predictive model for iso-vanillin 3 is shown in
eq 2, numerical values are given in Table 3. The statistical
products, (root mean squares-error of prediction) RMSEP
1
3 was obtained.
For both of the two procedures, for vanillin 1 and for
iso-vanillin 3, the reaction was carried out in a two-phase
system (water:ethyl acetate), and the unreacted protocate-
chualdehyde 13 is easily recovered and recycled by taking
advantage of the higher acidity of protocatechualdehyde 13
compared to that of vanillin 1 and iso-vanillin 3.
2
2
)
2.64, R ) 0.956, and Q ) 0.734 indicate a quite good
explanatory power of the model. The model eq 2 was
estimated using the PLS method.13
y ) R + R x + R x + R x + R x + R x + R x +
3
0
1
1
2
2
3
3
4
4
5
5
7 7
Synthesis of Heliotropin 4. (A) The reaction of proto-
R x x + R x x + R x x (2)
1
2
1
2
24
2
4
45
4
5
catechualdehyde 13 with CH
Scheme 7 was known.
2
Cl
On the basis of the reported
synthetic procedure an experimental study concerning the
2
to obtain heliotropin 4,
1
6-18
16
The model in eq 2 was used to produce the response surface
given in Figure 3. By using such a response surface, a
selectivity of 94.6% was obtained for the para O-methylation
of 3,4-dihydroxybenzaldehyde. The conversion is kept low
12
ring-closure step using statistical experimental design with
1
3
subsequently multivariate modelling was performed.
From this study we have derived a mathematical model,
which describes the yield of heliotropin 4 under the influence
of the experimental variables: the amount of NaOH, reaction
temperature, reaction time, and amount of phase-transfer
(42.5%) to limit formation of veratraldehyde 2.
Model for Vanillin 1. The experimental variables x
1 2
, x ,
x
3
, x
model. In this model some more two-factor interactions were
determined to have influence; x , x , x , x , and x
4 5
, and x were found to have significant influence in the
+
-
4
catalyst (tert-Bu N Br ). Optimised experiments based on
this model have led to selectivity in the range 61-64%,
which are not sufficient for an industrial development. On
the other hand attempts to reproduce the results reported in
1
x
2
1
x
3
1
x
5
2
x
4
4 5
x .
The final multivariate predictive model for vanillin 1 is
shown in Equation 3. Also for this model, the statistical
1
6
2
2
the literature gave much lower yields (<20%).
B) The process according to Scheme 4 appeared more
products, RMSEP ) 1.025, R ) 0.905, and Q ) 0.523
indicate a fairly good explanatory power of the model. The
estimation of the model eq 3 was also here carried out using
(
convenient for the synthesis of heliotropin 4. In this case
the basic medium is not suitable for the condensation of
glyoxylic acid with benzo[1,3]dioxol 15, as in the reaction
with catechol for obtaining protocatechualdehyde 13 (Scheme
1
3
the PLS method.
y ) â + â x + â x + â x + â x + â x + â x x +
1
0
1
1
2
2
3
3
4
4
5
5
12 1 2
3
). We have developed an acid catalysis, which has allowed
â x x + â x x + â x x + â x x (3)
1
3
1
3
15
1
5
24
2
4
45 4 5
(16) Kirby, J. A. (Eli Lilly & Co.). U.S. Patent 4,082,774, 1978.
As for iso-vanillin, the derived model for vanillin in eq 3
was used to produce a response surface as given in Figure
(17) Maggioni, P. (Brichima SpA). U.S. Patent 4,183,861, 1980.
(
18) Gensler W. J.; Samour, C. M. J. Org. Chem. 1953, 18, 13; Tampta, M.;
Aoyopi, Y. Chem. Pharm. Bull. 1968, 16, 523; Bonthrone W.; Cornorth, J.
W. J. Chem. Soc. (C) 1969, 1202.
4. By using such a response surface, a selectivity of 86.8%
538
•
Vol. 4, No. 6, 2000 / Organic Process Research & Development

Figure 3. Contour projections of the response surface where the response is the selectivity to iso-vanillin 3.
the synthesis of 3,4-dioxymethylene mandelic acid in high
yields (in the range 86-90%). The oxidative decarboxylation
was carried out in a two-phase system (water:toluene) by
(pKa(4-OH) ≈ 7.5) because of the influence of the carbonyl
group in the para position. To overcome this inconvenience
we have developed an oxidation process in two steps and a
two-phase system (water:ethyl acetate). The first step was
2
oxygen and CuCl catalysis with quantitative yield of
heliotropin 4. The oxidative decarboxylation was carried out
also with high yields by Na S O and catalysis by Ag(I) salt.
2 2 8
carried out under nitrogen atmosphere by an excess of CuCl
2
and HCl so that route b of Scheme 8 was avoided and only
route a took place. The oxidation occurs in the aqueous
phase, and the organic solvent continuously extracts the
formed aldehyde; the aqueous phase is separated, and CuCl
is oxidised by bubbling air and recycled. In this way the
process is always catalytic in CuCl , and the selectivity is
2
very high (>95%) with complete conversion.
For the synthesis of vanillin 1 and iso-vanillin 3 from
protocatechualdehyde 13 the critical point was the regio-
controlled O-methylation. We have taken advantage of the
different acidity of the two hydroxyl groups, as mentioned
before. In the presence of an excess of NaOH both the
hydroxyl groups are salified, but the phenoxide ion in
position 3 is more nucleophilic, and it is preferentially
attacked by dimethylsulphate leading to vanillin 1, due to
the conjugation of the phenoxide ion in position 4 with the
carbonyl group (Scheme 9).
Discussion
The good yields and the simple and convenient procedure
for the synthesis of protocatechualdehyde 13 by statistical
experimental design and multivariate modelling for the
synthesis of 3,4-dihydroxy mandelic acid 14 and the devel-
opment of a highly selective catalytic process for its oxidation
suggested that protocatechualdehyde 13 could be an interest-
ing intermediate for the synthesis of vanillin 1, iso-vanillin
3, and heliotropin 4. A critical point for the oxidation of
3,4-dihydroxy mandelic acid 14 was the presence of oxygen,
which leads to large amounts of the by-product 17, Scheme
. We explain this behaviour by the mechanism of the
Scheme 8.
The hydrogen abstraction by the peroxyl radical B is faster
for the hydroxyl group in position 3, which is less acidic
pKa(3-OH) ≈ 11.6) than the hydroxyl group in position 4
5
(
Vol. 4, No. 6, 2000 / Organic Process Research & Development
•
539

Figure 4. Contour projections of the response surface where the response is the selectivity to vanillin 1.
For the same reason when we use one mole or less of
NaOH per mole of protocatechualdehyde 13, the hydroxy-
lgroup in position 4 is more acidic, is preferentially salified,
and is attacked by dimethylsulphate leading to iso-vanillin
basic mechanism; we have further developed catalysis by
sulphuric acid which leads under high selectivity to the
mandelic acid derivative 16 (Scheme 11).
The basic catalysis on the contrary involves the nucleo-
philic addition of phenoxide ion to the carbonyl group
(Scheme 12).
3
(Scheme 10).
In both cases partial conversion (40-50%) is necessary
to avoid dimethylation leading to veratraldehyde 6, but the
separation and the recycle of unreacted protocatechualdehyde
The oxidative decarboxylation of the mandelic acid
derivative 16 to heliotropin 4 can be carried out in high yield
with catalytic amount of Cu(II) salt in the presence of oxygen
because the absence of hydroxyl groups prevents the oxida-
tive dimerisation of heliotropin 4 by a process similar to that
described in Scheme 8. The oxidative decarboxylation was
1
1
3 is quite simple and inexpensive; protocatechualdehyde
3 remains in the water phase while the reaction products
(1, 3, 6) are soluble in dichloromethane. The starting
aldehyde 13 can then be recovered by acidifying the water
phase and extracting with ethyl acetate. The recovered
substrate 13 is practically pure enough to be reused, but an
eventually further purification can be performed by using
the different acidity of the compounds. Attempts to utilise
Scheme 3 for the synthesis of heliotropin 4 led to 64%
maximum selectivity, probably because of the instability of
protocatechualdehyde 13 under basic experimental condi-
tions. Thus, Scheme 4 appeared to be more convenient in
this case. Contrary to the behaviour of guaiachol 7 (Scheme
carried out also by Ag(I)-catalysed Na
mechanism is shown in Scheme 13.
2 2 8
S O oxidation. The
Conclusions
A new strategy of industrial interest for the synthesis of
protocatechualdehyde 13, vanillin 1, iso-vanillin 3, and
heliotropin 4 was designed and optimised. The condensation
of catechol or benzo[1,3]dioxole with glyoxylic acid, the
oxidative decarboxylation of the mandelic acid derivatives,
and the regiocontrolled O-methylation of protocatechualde-
hyde 13 were the key points in all of these syntheses and
1) or catechol 5 (Scheme 3), the condensation of benzo[1,3]-
dioxole 15 with glyoxylic acid 9 cannot be catalysed by a
540
•
Vol. 4, No. 6, 2000 / Organic Process Research & Development

Scheme 8
Scheme 9
Scheme 10
Scheme 11
Experimental Section
General Methods. Mass spectra were performed on a
GLC-MS Finnigan TSQ 70 instrument, using a Varian 3700
gas chromatograph equipped with SBP-1 fused silica column
(
L 30 m × 0.2 mm i.d., 0.2 µm film thickness) and He as
carrier gas.
GLC analyses were performed on a capillary gas chro-
matograph equipped with SBP-5 fused silica column (L 25
m × 0.25 mm i.d., 1 µm film thickness) at a hydrogen flow
3
-1
rate of 8 cm min , PTV injector, and flame ionisation
detector.
they were developed by a combination of statistical experi-
mental design, multivariate modelling, and reaction mech-
anisms. A flexible procedure that allows to perform the
O-methylation of the hydroxy groups either in position 3 or
HPLC analyses were performed on a HPLC instrument
equipped with a RP-18 (5 µm packing material, L 250 mm,
4
mm i.d.) and UV detector (λ ) 281 nm).
H NMR spectra were recorded on a NMR spectrometer
4
1
of protocatechualdehyde 13 to obtain respectively vanillin
and iso-vanillin 3 was developed; it was based on the
1
operating at 400 MHz. Chemical shifts are referenced to
internal TMS.
Starting materials and reagents were purchased com-
mercially and used without further purification. All of the
different acidity of the two hydroxyl groups. For the synthesis
of heliotropin 4 it appeared more convenient to start from
benzo[1,3]dioxole 15 than from protocatechualdehyde 13.
Vol. 4, No. 6, 2000 / Organic Process Research & Development
•
541

Scheme 12
analysis revealed a complete conversion of the mandelic acid
derivative and the yield of protocatechualdehyde of 96%.
The copper salt aqueous solution/suspension was recycled
by oxidising Cu(I) to Cu(II) by air after the removal of the
organic phase; the results were substantially unchanged.
Vanillin 1. Protocatechualdehyde 13 ( 7.2 mmol) and
NaOH (18 mmol) were dissolved in 25 mL of water; 20 mL
2 2
of CH Cl was added, and 5 mmol dimethyl sulphate was
added dropwise under vigorous stirring during 1 h at a
temperature of 55-60 °C. The reaction was run for a further
4
h at the same temperature. The two phases were separated,
and the water phase was further extracted with 2 × 25 mL
of CH Cl . The organic layers were combined and dried with
2
2
sodium sulphate, and the solvent was removed under vacuum
to give the desired product. The aqueous phase was acidified
by 35% HCl to pH 1-2 and extracted by ethyl acetate (2 ×
5
0 mL) to recover the unreacted protocatechualdehyde 13.
HPLC analyses revealed a recovery of 55.4% of unreacted
protocatechualehyde 13 (conversion 44.6%) with a selectivity
of 77.5% of vanillin, 8% of iso-vanillin, and 3.4% of
veratraldehyde.
Scheme 13
iso-Vanillin 3. (A) Protocatechualdehyde 13 (7.2 mmol)
and NaOH (2.1 mmol) were mixed in 3.5 mL of water and
3
2 2
.3 mL of CH Cl ; 6 mmol of dimethyl sulphate and 3.7
mmol of NaOH in 1.5 mL of water were simultaneously
added dropwise with stirring over a period of 45 min. After
the complete addition of the reagents, the reaction mixture
was stirred for a further 50 min at a temperature of 55-60
°
C. The reaction mixture was allowed to cool to room
temperature, the two phases were separated, and the water
Cl . The
phase was further extracted with 2 × 25 mL of CH
2
2
organic layers were combined and dried with sodium
sulphate, and the solvent was removed under vacuum to give
the desired product. The aqueous phase was acidified by 35%
HCl to pH 1-2 and extracted by ethyl acetate (2 × 50 mL)
to recover the unreacted protocatechualdehyde 13. HPLC
analyses revealed a conversion of 18.3% with a selectivity
of 93.5% of iso-vanillin, 5.8% of vanillin, and 0.5% of
veratraldehyde.
reaction products were known and were analysed by GC and
GC-MS and by comparison with authentic samples.
3,4-Dihydroxy Mandelic Acid 1 (Optimised Proce-
dure). Catechol (5.00 g, 45.41 mmol) was dissolved in
aqueous NaOH (3.21 g, 80.3 mmol in 55.0 mL of water)
2 3
followed by addition of Al O (2.04 g, 20 mmol). After 5
(
B) The reaction was carried out as in (A) at 55 °C by
min glyoxylic acid (7.10 g of 50% aqueous solution, 48.0
mmol) was added to the reaction mixture, and the mixture
was heated at 60 °C for 24 h under vigorous stirring. The
reaction mixture was then allowed to precipitate for 10 min.
using 21.7 mmol of protocatechualdehyde and 4.5 mmol of
NaOH and by simultaneously adding dropwise 21.4 mmol
of dimethyl sulphate and 17.5 mmol of NaOH. The conver-
sion was 70.8%, and the selectivity 93.2% in iso-vanillin,
2 3
and filtered to remove Al O . The obtained filter cake was
4
.0% of vanillin, and 2.8% of veratraldehyde.
,4-Dioxymethylene Mandelic Acid 16. Benzo[1,3]-
washed with 1 M NaOH (20 mL). The basic washing water
was combined with the water solution, and this was acidified
to pH 3-4 with 6.0 mL of 37% HCl and extracted with ethyl
acetate to recover the unreacted catechol (1.2 g). The aqueous
solution was further acidified to pH 1 by 2 mL of concen-
trated HCl and extracted with ethyl acetate to isolate the
mandelic acid derivative (5.1 g, 28.08 mmol). Conversion
3
dioxole 15 (8.20 mmol) was added dropwise over a period
of 30 min to a mixture of 8.41 mmol of glyoxylic acid, 14.59
mmol of sulphuric acid, and 11.34 mmol of water at 0-5
°C. The reaction was run for 20 h at 0 °C and then diluted
with 20 mL of water. The 3,4-dioxymethylene mandelic acid
precipitated and was then filtered off. HPLC analysis
revealed a yield of 86% with selectivity >95%.
77.5%, selectivity 90.5%.
Protocatechualdehyde 13. 3,4-Dihydroxy mandelic acid
(
2 g, 10.86 mmol) was dissolved in 140 mL of ethyl acetate,
and 11.11 g of CuCl ‚2H O was dissolved in 30 mL of water.
The two-phase system was vigorously stirred and heated at
0 °C for 5 h under nitrogen atmosphere. The organic phase
was separated, and the solvent was removed. The HPLC
Heliotropin 4. (A) 3,4-Dioxymethylene mandelic acid (5
mmol), Na
mL of water and 30 mL of CH
The CH Cl phase was removed and analysed by HPLC; a
yield of 96% of heliotropin was obtained.
S
2 2
O
8
(6 mmol), and AgNO
3
(0.05 mmol) in 30
2
2
2
2
Cl were refluxed for 5 h.
6
2
2
542
•
Vol. 4, No. 6, 2000 / Organic Process Research & Development

(
B) 3,4-Dioxymethylene mandelic acid (5 mmol), CuCl
2
extracted by ethyl acetate to recover the unreacted proto-
catechualdehyde. The conversion was 58.6% and selectivity
in heliotropin 64%.
(
3
0.5 mmol), and 4.05 g of 37% HCl in 30 mL of water and
0 mL of toluene was heated at 80 °C and bubbling O
2
for
22 h. The organic phase was separated and analysed by
HPLC; a yield of 98% of heliotropin was obtained. The
aqueous solution is ready to be used for further oxidation.
Acknowledgment
Borregaard Synthesis (Norway) is gratefully acknowl-
edged for financial support to this project.
(
C) Protocatechualdehyde (13.1 g, 0.1 mol), NaOH (12
g), and tetrabutylammonium bromide (3.2 g) in 100 mL of
water and 100 mL of CH Cl were heated for 4 h at 80 °C
2
2
Received for review May 15, 2000.
OP0000529
in an autoclave. The aqueous phase was then separated and
analysed by HPLC. The aqueous solution was acidified and
Vol. 4, No. 6, 2000 / Organic Process Research & Development
•
543