An optimized process to 10-bromo-1-decanol

Article abstract of DOI:10.1021/op100143t

A multivariate design and optimization study for the synthesis of the bromoalkanol 10-bromo-1-decanol using decane-1,10-diol as substrate is reported. The bromination process was supported by the phase transfer catalyst tetrabutylammonium bromide with aqueous HBr (48%) as the brominating reagent. The optimized batch protocol provided a yield of 64% of 10-bromo-1-decanol 2 TM with a conversion of 80%, and 10% of the dibrominated alkane 1,10-dibromodecane 3, a characteristic byproduct, was formed.

Full text of DOI:10.1021/op100143t

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Organic Process Research & Development 2010, 14, 1215–1220
An Optimized Process to 10-Bromo-1-decanol
Raffaele Spaccini,†,⊥ Anna Tsoukala,† Lucia Liguori,†,§ Carlo Punta,‡ and Hans-Rene´ Bjørsvik*,†,§
Department of Chemistry, UniVersity of Bergen, Alle´gaten 41, N-5007 Bergen, Norway, Politecnico di Milano, Dipartimento di
Chimica, Materiali e Ingegneria Chimica, ”Giulio Natta”, Via Mancinelli 7, I-20131 Milano (MI), Italy, and Fluens Synthesis,
Thormøhlensgate 55, N-5008 Bergen, Norway
Chart 1. Idebenol [2-(10-hydroxydecyl)-5,6-dimethoxy-3-
methylcyclohexa-2,5-diene-1,4-dione]
Abstract:
A multivariate design and optimization study for the synthesis of the
bromoalkanol 10-bromo-1-decanol using decane-1,10-diol as substrate
is reported. The bromination process was supported by the phase
transfer catalyst tetrabutylammonium bromide with aqueous HBr
(48%) as the brominating reagent. The optimized batch protocol
provided a yield of 64% of 10-bromo-1-decanol 2 TM with a
conversion of 80%, and 10% of the dibrominated alkane 1,10-
dibromodecane 3, a characteristic byproduct, was formed.
Scheme 1. Process to 10-bromo-1-decanol 2
Introduction
For a project dedicated to a novel total synthesis of idebenol
(Chart 1) and derivates thereof we needed access to 10-bromo-
1-decanol 2 on a multigram scale with a prospective future
requirement of multikilogram scale.
Alcohols can be converted to the corresponding alkyl halides
via a multitude of reagents, and this transformation has thus in
the past been a subject for extensive studies that have resulted
in a series of sophisticated protocols. Unfortunately, harsh
reaction conditions are often mandatory in several of these
procedures. Reagents that have been reported for the alcohol
to halide transformation embrace a wide range of methods.1-10
The advantage of several of these reagents is the relatively low
cost, but the drawbacks include toxicity, handling problems
(especially on large scale), and low selectivity in some of them.
After evaluating the various synthetic routes to alkyl
bromides from alcohols,11 a pathway that appeared attractive
to us involved decane-1,10-diol 1 as substrate that is readily
available at low cost. Moreover, such a synthetic protocol
involves only a simple bromination step at one of the two
available hydroxyl centers.11a However, drawbacks exist in such
a process, namely the requirements of a highly chemoselective
process since the desired product 10-bromo-1-decanol 2 can
readily undergo a second bromination to produce 1,10-dibro-
modecane 3, Scheme 1, an unwanted byproduct that reduces
effective yield of the desired product 2.
Methods and Results
Design and Analysis of the Experimental Investigation.
We commenced this study by a combined synthetic route
discovery and pre-experimental design study, that brought us
to a synthetic protocol disclosed in this journal.10 With the basis
in that protocol we elaborated an Ishikawa cause-effect (ICE)
diagram,12 Figure 1, that also encompasses prospective process
variables that must be taken into account for an eventually future
scale-up project.
However, for the present experimental optimization study
we decided to perform an investigation of the variables related
to the synthetic protocol, which are the branches enumerated
1, 2, 5, 6, 7, and 8 in the ICE diagram (Figure 1). We believed
that the reaction temperature, the reaction time, the quantities
of the brominating reagent, the phase transfer catalyst, and the
solvent volume were the variables that influenced the perfor-
mance in terms of the yield and selectivity. The solvent, other
than in the bromination reagent (HBr in water), was not used,
that is the diol 1 was added neat to the reaction flask.
* Author to whom correspondence may be sent. E-mail: hans.bjorsvik@
kj.uib.no. Telephone: +47 55 58 34 52. Fax: +47 55 58 94 90.
† University of Bergen.
‡ Politecnico di Milano.
§ Fluens Synthesis.
⊥ Current address is Fluens Synthesis AS, Bergen, Norway.
(1) Martinez, A. G.; Ruiz, M. O. Synthesis 1983, 663.
(2) (a) Weiss, R. G.; Snyder, E. I. J. Chem. Soc., Chem. Commun. 1968,
1358. (b) Weiss, R. G.; Snyder, E. I. J. Org. Chem. 1972, 36, 403.
(3) Fujisawa, T.; Iida, S.; Sato, T. Chem. Lett. 1977, 1173.
(4) Benazza, T.; Uzan, R.; Beaupe`re, D.; Demailly, G. Tetrahedron Lett.
1992, 33, 3129.
(5) Benazza, T.; Uzan, R.; Beaupe`re, D.; Demailly, G. Tetrahedron Lett.
1992, 33, 4901.
(6) (a) Tortajada, A.; Mesters, R.; Iglesias- Arteaga, M. A. Synth. Commun.
2003, 1809. (b) Chong, J. M.; Heuft, A. M.; Rabbat, P. J. Org Chem.
2000, 65, 5837.
(7) Ranu, B. C.; Jana, R. Eur. J. Org. Chem. 2005, 755.
(8) Abad, J.-L.; Rodriguez, S.; Camps, F.; Fabrias, G. J. Org. Chem. 2006,
71, 7558.
(9) Hooz, J.; Gilani, S. S. H. Can. J. Chem. 1968, 46, 86.
(10) Kad, G. L.; Kaur, I.; Bhandari, M.; Singh, J.; Kaur, J. Org. Process
Res. DeV. 2003, 7, 339.
(11) See for example: (a) March, J. AdVanced Organic Chemistry, 4th ed.;
Wiley: New York, 1992; pp 431-433. (b) Larock, R. C. Compre-
hensiVe Organic Transformations, 2nd ed.; Wiley: New York., 1999;
pp 693-695.
(12) Ishikawa, K. Guide to Quality Control, 2nd ed.; Asian Productivity
Organization: Minato-Ku, Japan, 1990; pp 18-29.
10.1021/op100143t  2010 American Chemical Society
Published on Web 07/20/2010
Vol. 14, No. 5, 2010 / Organic Process Research & Development
•
1215

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Figure 1. Ishikawa diagram for a potential bromination of decane-1,10-diol 1 to achieve 10-bromo-1-decanol 2 and minimize the
production of the byproduct 1,10-dibromodecane 3.
Table 1. Statistical experimental design for the investigation
investigated are described in the footnotes of Table 1. Each of
the experimental variables (xi, i ) 1, ..., 3) was scaled according
to eq 115 in order to facilitate the calculation of the regression
coefficients (the numerical values of the ꢀ coefficients) of the
predictive empirical mathematical model provided in eq 2.
of the bromination of decane-1,10-diol 1
experimental variablesa
response variablesb
#
x1
x2
x3
c120
s120
y120
1
-1
+1
-1
+1
-1
+1
+1
-1
+1
+1
0
-1
-1
+1
+1
-1
-1
-1
+1
+1
+1
0
-1
-1
-1
-1
+1
+1
+1
+1
+1
+1
0
36
35
54
53
38
38
44
53
69
49
46
64
23
55
63
25
44
79
89
89
81
72
89
87
82
77
72
84
80
78
94
77
75
92
84
78
32
31
44
38
34
33
36
41
50
41
37
50
22
42
47
23
37
62
2
3
1
2
1
2
zi - zi,L
+
× (zi,H - zi,L)
4
{
}
5
xi )
,i ) 1, ..., 3
6a
6b
7
zi,H - zi,L
+
× (zi,H - zi,L)
{
}
(1)
8a
8b
9
10
11
12
13
14
15
16
3
2
3
0
0
0
y ) f(x1, x2, x3) ) ꢀ0 +
ꢀ x +
ꢀ x x +
∑
∑ ∑
-2
+2
0
0
0
0
1
1
ij 1 j
i)1
i)1 j)2
0
3
-2
+2
0
0
ꢀ x2 (2)
0
0
∑
ii
i
0
-2
i)1
0
0
+2
a Procedure: decane-1,10-diol 1 (2.87 mmol, 0.500 g) was used as substrate in
all of the experiments. Experimental variables: xk (definition) [levels: -2, -1, 0,
+1, +2]: x1 (reaction temperature) [85 °C, 90 °C, 95 °C, 100 °C, 105 °C,]; x2
(quantity of hydrobromic acid) [0.15, 0.30, 0.45, 0.60, 0.75 equiv]; x3 (quantity of
tetrabutylammonium bromide) [0.01, 0.10, 0.20, 0.30, 0.40 equiv]. b Responses
measured by GC: ct ) conversion at time t ) 120 min, st ) selectivity at time t )
120 min, and yt ) yield at time t ) 120 min.
xi of eqs 1 and 2 is the experimental variable i given in scaled
units, zi is the experimental variable i given in real units, zi,L
and zi,H are the selected low and high experimental values in
real units, of the experimental variable i.15
Prior to carrying out in the laboratory the initial experimental
design, that is objects #1-10 of Table 1, the extreme points,
namely, objects #1, #8, and #9 were conducted in order to assess
whether the selected experimental levels represented a sufficient
span to provide significant variations in the responses. Moreover,
these experiments could also provide information whether
quadratic terms intervened in the prospective models. Figure 2
shows the reaction profile for the three extreme point experi-
ments, object #1, 8, and 9.
The experimental study was planned to be implemented by
means of statistical experimental design13 and multivariate
modeling14 to simultaneously optimize the two responses, yield
and selectivity, respectively. Such an approach will furthermore
furnish a picture of the process robustness and at the same time
provide an indication on the cheapest performance.
The original experimental design D, entries # 1-10, Table
1, was a full factorial design (2k) of k ) 3 independent
experimental variables (x1, x2, x3) at two experimental levels
(denoted as -1 and +1) with c ) 2 experiments in the center
of the experimental domain, that provides in total 2k + c ) 23
+ 2 ) 10 experiments. The selectivity (st t ) 120 min) and the
yield (yt t ) 120 min) of each experiment were measured by
means of gas chromatography and utilized as the responses.
The numerical values of the responses are provided in the right-
hand columns adjacent to the statistical experimental design of
Table 1. The experimental variables and levels that were
(13) (a) 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. (b) Montgomery, D. C. Design
and Analysis of Experiments, 3rd ed.; Wiley, New York, 1991. (c)
Box, G. E. P.; Draper, N. R. Empirical Model-Building an Response
Surfaces; Wiley: New York, 1987. (d) Myers, R. H.; Montgomery,
D. C. Response Surface Methodology. Process and Product Optimiza-
tion Using Designed Experiments; Wiley: New York, 1995.
(14) (a) Montgomery, D. C.; Peck, E. A. Introduction to Linear Regression
Analysis; Wiley: New York, 1982. (b) Draper, N. A.; Smith, H. Applied
Regression Analysis, 2nd ed.; Wiley: New York, 1981.
(15) When scaling is performed according to eq 1, the scaled low value is
set at -1, and the scaled high value becomes set at +1.
1216
•
Vol. 14, No. 5, 2010 / Organic Process Research & Development

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Figure 2. Reaction profiles for (a) yield, (b) conversion, and (c) selectivity for the three experiments: #1 (X) where (x1 x2 x3) ) [-1
-1 -1], #8 (•) mean of two runs where (x1 x2 x3) ) [+1 +1 +1], and #9 (∇) mean of two runs where (x1 x2 x3) ) [0 0 0].
y120 ) f(x1, x2, x3) ) 41.799 + 2.034 × x1 +
The experimental span appeared to be sufficient with a ymax
3.231 × x3 + 0.897 × x1 × x2 + 0.897 × x2 × x3 -
2.874 × x21 - 2.124 × x22 + 1.501 × x23 (3)
- ymin ) 18, at the reaction time t ) 120 min. Furthermore, it
appeared that the reaction system is influenced by some
nonlinear terms; experiment #1 and #9 provide almost quite
similar results (∆y ≈ 5) at the maximum value at reaction time
t ) 120 min, while experiment #3 provides substantially higher
yield throughout the whole reaction profile, an experiment that
moreover provides the maximum yield at t ) 60 min. On this
basis we decided to eventually also take into account quadratic
terms by expanding the original design (objects #1-10) to
including a centered star design (through the center point) which
in total constitute a central composite design, see object #1-16
of Table 1. The experiments were conducted in the laboratory,
and the responses were carefully measured on GC.
A model matrix X16 lines×10 columns ) [1 x1 x2 x3 x1 × x2 x1 ×
x3 x2 × x3 x12 x22 x32] was created on the basis of the design
matrix D. The model matrix X with the adjacent yield and
selectivity values of target molecule 2 was submitted for
multivariate regression using the partial least-squares regression
method (PLSR).16 Rough evaluations of the predictive models
were performed using cumulative normal probability plots and
the adjacent bar plots, shown in a and b of Figure 3. The final
variable pruned empirical mathematical models, that describe
the variation of the yield y120 min ) f(x1,x2,x3) and the selectivity
s120 min ) g(x1,x2,x3) when the reaction time is t ) 120 min, are
given in eqs 3 and 4.
s120 ) g(x1, x2, x3) ) 79.889 - 3.429 × x1 -
0.636 × x3 - 1.542 × x1 × x2 + 3.024 × x1 × x3 +
0.942 × x21 + 1.429 × x22 (4)
Optimization Experiments. The two models shown in eqs
3 and 4 were used for the production of the isocontour maps in
Figure 4. The isocontour maps describe the selectivity and the
yield as a function of the three experimental variables (x1, x2,
x3). The isocontour projections of the response surfaces were
used for the purpose of prediction of a few of the optimized
protocols (Table 2) for the bromination of decane-1,10-diol 1.
Prediction of Optimized Conditions. The conditions for
the first optimization experiment (O1, X) were selected as the
following: T ) 100 °C, 0.45 equiv of HBr, and 0.50 equiv of
TBAB, that represent an extrapolation of experimental variable
x3 (TBAB ) +3 in coded unit, eq 1). The obtained result of
this experiment proved to be very promising, see Figure 5, and
according to predicted values (ypred ) 65%, spred ) 85). Spurred
by those results, we performed another two optimization
experiments (O2 (•) and O3 (∇) of Table 2). Both of those
experiments represented “extremes” to medium long-range
extrapolations involving the two experimental variables quantity
of HBr (x2) and quantity of TBAB (x3). Experiment O2 of Table
2 provided a reasonable result with a yield of 66% of TM with
a selectivity of 79% and conversion of 84%. The only observed
side product was the dibrominated alkane 3. In the third
optimization experiment (O3) we attempted an extreme ex-
trapolation in the experiment variable x3 (the quantity of
tetrabutylammonium bromide). In this experiment a quantitative
The model for the yield: y120 min ) f(x1,x2,x3) was described
by a ) 5 PLS components, a model that accounts for ≈96.8%
[) 87.307(a)1) + 2.164(a)2) + 5.017(a)3) + 2.190(a)4)
+
0.102(a)5)] of the variance in the data. The model for the
selectivity: s120 min ) g(x1,x2,x3) was described by a ) 3 PLS-
components, a model that account for ≈99.5% [) 84.117(a)1)
+ 13.936(a)2) + 1.442(a)3)] of the data.
Despite the high number of the explained variance in the
two models, the product statistics17,18 of the models apparently
indicate a somewhat weak fitting of the data: the model for the
yield shows (with a ) 5 PLS components) R2y ) 0.424 and
RMSEPy ) 7.19; the model for the selectivity shows (with a
) 3 PLS components): Rs2 ) 0.451 and RMSEPs ) 5.86. The
models should prove to have highly predictive power.
(17) Product statistics are calculated according to the following equation:
R2 ) 1 - SSresidual/(SSmodel + SSresidual)]. Root means squares error of
prediction (RMSEP) is an estimate of the prediction error expressed
in real units, in this case, % yield and % selectivity:
I
2
yPred - y
i
(
)
∑
i
i)1
RMSEP )
ꢀ
I
(16) Malinowski, E. R. Factor Analysis in Chemistry, 3rd ed.; Wiley: New
(18) Malinowski, E. R. Factor Analysis in Chemistry, 3rd ed.; Wiley: New
York, 2002.
York, 2002.
Vol. 14, No. 5, 2010 / Organic Process Research & Development
•
1217

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Figure 3. (a) First model including all of the model parameters with yield of product at time 120 min as response variable. (b) First
model including all of the model parameters with selectivity of product at time 120 min as response variable.
conversion of the substrate was observed, but the yield of TM
was slightly reduced compared to the yields of the optimization
experiments O1 and O2. In the O3 experiment, a subsequent
reaction had taken place under the formation of 10-bromodec-
1-ene 4 in a quantity of ≈20% (measured by means of GC/
MS). The remaining part of the converted substrate 1 was
transformed to the byproduct 1,10-dibromodecane 3. A pro-
longed reaction time (t > 120 min) resulted in a significantly
augmentation of the quantity of the side product 3 and a minor
increase in the quantity of the byproduct 4.
The Robustness of the Process. Even though the product
statistic17 parameter RMSEP shows values in the range 6-7,
the long-distance (outside the experimental domain) model
predictions, the models shows surprisingly small deviation
between the predicted and achieved value of the optimization
experiment. The RMSEP values can be explained by the steep
ascent from the saddle point and towards the expected optimum
area. The steep ascent of the area in close proximity to the
expected optimal operational range for the brominating process
can be a drawback, since the process is much affected by even
small changes in the control variables (x1, x2, x3). As such, the
process is not robust in the sense that it will not provide high
yield and selectivity without precise control of the reaction
temperature and exactly measured amounts of the substrate, the
reagent, and the phase transfer catalyst.
Scaled-Up Optimized Experiment. The optimized proce-
dure was attempted scaled-up using 50 g of the substrate 1.
The reaction was conducted in a cylindrical flat-bottomed glass
reactor with no baffles (D 78 mm, H 140 mm, V ≈ 500 mL).
The reactor was equipped with a radial turbine stirrer (flat
blades, 50 mm, and stirrer rate ) 200 turns × min-1). The
experiment was run for 2 h 30 min to provide a yield of ≈60%
of desired product 2, that is slightly lower than in the original
scale (64%) used during the optimization study.
Conclusion
With the presence of two identical reactive sites in substrate
molecule such as decane-1,10-diol 1, it is well-known that a
high selectivity can be achieved by means of a strictly restricted
conversion of the substrate but often low yield of desired
product. However, in this work we have concurrently optimized
the selectivity and the yield of the brominating process by means
of multivariate experimental design and modeling using re-
sponse surface methodology. Multivariate empirical model were
derived on the basis of experimental data that subsequently was
used to predict optimized protocols for the production of target
1218
•
Vol. 14, No. 5, 2010 / Organic Process Research & Development