Multivariate optimization of a cyclopropanation, the key step in the synthesis of 3,3,4,4-tetraethoxybut-1-yne

Article abstract of DOI:10.1021/op5001012

3,3,4,4-Tetraethoxybut-1-yne (TEB) is a versatile synthon that can be produced in a four-step synthesis. The third step of the synthesis is a cycloproanation, which has been thoroughly investigated and optimized by means of statistical experimental design and multivariate modeling. At the outset, an exhaustively pre-experimental design was performed resulting in a copious Ishikawa cause-effect diagram. In total six of the experimental variables were assessed to be of large importance and thus selected for further investigation by fractional factorial design. The results of that screening and first step optimization formed the basis for a response surface modeling (RSM) study. The RSM investigation was completed by using a central composite design from which a response surface was graphically produced as an iso-contour projection. The derived multivariate predictive model in terms the iso-contour projection plots were ultimately utilized to establish experimental conditions that concomitantly provided excellent yield (>99%) and minimized amounts of inputs and thus obtain the desired product at the lowest production cost and minimized side-streams.

Full text of DOI:10.1021/op5001012

Article
pubs.acs.org/OPRD
Multivariate Optimization of a Cyclopropanation, the Key Step in the
Synthesis of 3,3,4,4-Tetraethoxybut-1-yne
†
Weidong Shang, Mariangela Terranova, Leiv K. Sydnes, and Hans-Rene
́
Bjørsvik*
́
Department of Chemistry, University of Bergen, Allegaten 41, N-5007 Bergen, Norway
*
S Supporting Information
ABSTRACT: 3,3,4,4-Tetraethoxybut-1-yne (TEB) is a versatile synthon that can be produced in a four-step synthesis. The third
step of the synthesis is a cycloproanation, which has been thoroughly investigated and optimized by means of statistical
experimental design and multivariate modeling. At the outset, an exhaustively pre-experimental design was performed resulting in
a copious Ishikawa cause−effect diagram. In total six of the experimental variables were assessed to be of large importance and
thus selected for further investigation by fractional factorial design. The results of that screening and first step optimization
formed the basis for a response surface modeling (RSM) study. The RSM investigation was completed by using a central
composite design from which a response surface was graphically produced as an iso-contour projection. The derived multivariate
predictive model in terms the iso-contour projection plots were ultimately utilized to establish experimental conditions that
concomitantly provided excellent yield (>99%) and minimized amounts of inputs and thus obtain the desired product at the
lowest production cost and minimized side-streams.
INTRODUCTION
produce 1,1-dibromo-2-chloro-2-diethoxymethyl-cyclopropane
under phase-transfer catalytic (PTC) conditions according to
■
1
3
A few years ago, a synthesis leading to 3,3,4,4-tetraethoxy-but-
-yne (TEB, 5) was disclosed by Sydnes and collaborators.
TEB possesses a significant potential as a synthon in organic
synthesis. The reagent has been utilized to prepare a variety of
6
1
Makosza’s method, varies significantly and has given yields in
the 20−60% range. Attempts to optimize and improve the
outcome and the robustness of this synthetic step by
7
2
evolutionary operations involving simple modifications of the
compounds such as unsaturated alcohols, deoxygenated
3
4
procedure have been implemented over a rather long period of
time. Moreover, a considerable surplus of several of the
reagents (relative to the amount of the alkene) is used to
achieve a tolerable conversion and yield.
carbohydrates, furans, and a number of other heterocyclic
5
compounds.
TEB is prepared in a four-step linear synthesis, which
encompasses two cyclopropanations steps (1 and 3) and two
ring-opening reactions (steps 2 and 4). The synthetic pathway
leading to 5 is outlined in Scheme 1.
METHODS AND RESULTS
■
On the basis of these facts, we realized the need of a thorough
and in-depth investigation of the process step 3 with the goal to
develop a robust and high yielding synthetic step 3 that
moreover demands minimized quantities of the various inputs.
Pre-experimental Design. Based on the principles and
Scheme 1. Synthesis of 3,3,4,4-tetraethoxybut-1-yne (TEB) 5
8
methodology of statistical experimental design and multi-
9,10
variate regression analysis,
an experimental study was
designed and carried through with the goal to solve the
challenges that exist in the actual reaction step of the TEB
synthesis.
At the outset, we carried through an in-depth analysis of the
existing synthetic procedure; the various experimental variables
we came across successively were incorporated into an Ishikawa
11
cause−effect (ICE) diagram, see Figure 1.
The complete set of experimental variables that we believed
to have influence on the 3 → 4 transformation was divided into
two distinct groups, namely, (1) the variables that were related
to the performance of the synthetic reaction, and (2) the
variables that concern the reaction workup. Herein we focused
on the performance of the reaction, so all experimental
During the period since the synthesis of synthon 5 was
designed and performed for the first time, the compound has
been made a substantial number of times by chemists at various
levels of expertise. The overall picture that has emerged is that
steps 1, 2, and 4 are reproducible, relatively robust, and provide
the expected reaction products (2, 3, and 5) in high, medium,
and excellent yields, respectively. On the other hand, step 3,
which is a cyclopropanation that is performed by addition of
dibromocarbene to 2-chloro-3,3-diethoxyprop-1-ene 2 to
Received: March 24, 2014
©
XXXX American Chemical Society
A
dx.doi.org/10.1021/op5001012 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
the details of which are given in the Experimental Section. The
statistical experimental design we selected for this study (Table
1
) is a fractional factorial design with a couple of experiments in
k−r
6−2
the center of the experimental domain (2 + c = 2 + 2).
The scaled experimental variables x , ..., x were used to prepare
1
6
the confounded two-variable and some of the three-variable
interactions; see the model parameters in Table 2.
Table 2. Regression coefficients with adjacent estimated
numerical values
Figure 1. Ishikawa cause−effect (ICE) diagram for the 1,1-dibromo-2-
chloro-2-(diethoxymethyl)cyclopropane (4). Experimental variables
that affect the performance of the reaction.
coefficient and alias
value
coefficient and alias
value
β0
39.278
8.875
β13 (+ β25)
7.750
−4.750
1.875
β1 reaction temperature
β14 (+ β56)
β2 quantity of bromoform
3.250
β15 (+ β23 + β46)
β3 reaction time
−1.125
−4.625
22.250
−0.500
−3.625
β16 (+ β45)
β24 (+ β36)
β34 (+ β26)
β124
0.125
variables related to the workup were excluded from the final
ICE diagram, Figure 1.
β4 addition time, NaOH
β5 quantity of NaOH
β6 amount of TEBA
β12 (+ β35)
0.875
−5.000
−5.250
−1.875
This elaborated subset of the experimental variables was
12
further assessed according to priority categories to provide
the following list of variables that were forwarded for further
investigation in a screening and first step optimization study.
This variable subset included x , reaction temperature [°C]; x ,
β134
1
2
quantity of bromoform [equiv]; x , reaction time [h]; x ,
3
4
Assessment of the Experimental Levels. Prior to the
implementation of the whole experimental plan (Table 1), a
small subset of experiments were conducted in the laboratory in
order to investigate and ensure that the selected experimental
levels delivered sufficient variation in the selected response, that
is, the yield of cyclopropane 4. The entries 1, 16, and 17 of
Table 1 were thus conducted, which revealed a significant
variation and that the selected experimental levels were
satisfactory selected. Hence, the rest of the objects of the
experimental design (2−15 and 18 that is a replicate of 17)
were successively carried out in a random order. The achieved
results of all of the experiments are reported in the right-hand
columns of Table 1.
Computation: Development of Multivariate Predic-
tive Model. The experimental variables and levels are
displayed in Table 1. Each of the variables was scaled according
to eq 1 to facilitate the estimation of the regression coefficients,
the β’s of eq 2. The x of the equations is the experimental
variable i given in scaled units, z is the identical experimental
addition time for NaOH [min]; x , quantity of NaOH [equiv];
5
and x , amount of phase transfer catalyst (TEBA) [equiv]. Each
6
of the variables x , x , ..., x were then assessed with respect to
1
2
6
ample experimental ranges,; see footnote a of Table 1. The
substrate 3 was prepared according to Scheme 1 (1 → 2 → 3),
Table 1. Statistical experimental design for screening of the
experimental variables and first step optimization of the
reaction using a 26 fractional factorial design with one
response variable
‑2
,
ab
experimental variables
entry
x1
x2
x3
x4
x5
x6
y
1
2
3
4
5
6
7
8
9
−1
1
−1
−1
1
−1
−1
−1
−1
1
−1
−1
−1
−1
−1
−1
−1
−1
1
−1
1
−1
−1
1
14
63
55
22
42
41
4
−1
1
1
1
−1
1
1
i
−1
1
−1
−1
1
1
i
1
−1
−1
1
1
variable i = 1, ..., 6, although expressed in real (noncoded) units,
zi,L and zi,H are the selected low (L) and high (H) experimental
values in real units, of the experimental variable zi.
−1
1
1
−1
−1
1
1
3
1
1
98
7
−1
1
−1
−1
1
−1
−1
−1
−1
1
−1
1
1
2
10
11
12
13
14
15
16
17
18
1
1
66
75
9
z − z + (z − z )
i
{
i,L
i,H
i,L } ,
−1
1
1
1
−1
−1
−1
−1
1
xi =
i = 1, ..., 6
1
2
zi,H − z + (z − z )
1
1
−1
1
{
i,L
i,H
i,L
}
−1
1
−1
−1
1
1
25
18
9
(1)
1
1
−1
−1
1
6
6
6
−1
1
1
1
y = β +
∑
βx + ∑ ∑ β xx
0
i
i
ij i j
1
1
1
1
56
52
51
i=1
i<j i<j
(2)
0
0
0
0
0
0
0
0
0
0
0
0
The model matrix X was created as the following X = [1 D x ×
1
x x × x x × x x × x x × x x × x x × x x × x × x x
1
2
1
3
1
4
1
5
1
6
2
4
3
4
1
2
4
a
Experimental variables: xk (description [unit]) [low level (−1),
center level (0), high value (+1) ]. x (reaction temperature [°C ])
× x × x ], included the design matrix D after a column of ones
3
4
1
that was followed by the two-variable interactions (x × x ) for
i
j
[
(
20, 25, 30]; x2 (quantity of bromoform [equiv]) [5, 8, 10]; x3
reaction time [h]) [12, 24, 36]; x (addition time for NaOH [min])
each of the six experimental variables (x , i = 1, ..., 6). The final
i
4
model matrix was used in scaled values [11 lines × 8 columns,
eq 2] with their corresponding achieved % yield values of target
molecule 1,1-dibromo-2-chloro-2-(diethoxy-methyl)-
cyclopropane 4 then submitted to multivariate regression in
[
5, 10, 15]; x (quantity of NaOH [equiv]) [8, 10, 12]; x (amount of
5
6
PTC (TEBA) [equiv]) [0.01, 0.015, 0.02]. A fixed quantity of
substrate 3 (3.29 g, 20 mmol) was used in all of the experiments 1−18.
b
Generators used in the design: x = x × x × x and x = x × x ×
5
1
2
3
6
2
3
9
x .
4
terms of the multiple linear regression (MLR) and partial
B
dx.doi.org/10.1021/op5001012 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
1
0
14
least-squares regression (PLSR) methods using the SAS and
two-variable interactions x × x and x × x are confounded,
1
3
2
5
1
5
MATLAB computer software packages.
The estimated numerical values and confounding pattern for
the various regression coefficients are provided in Table 2. The
product statistics show a reasonable good model: R = 0.973,
RMSEP = 4.324, and RSD = 4.587.
The cumulative normal probability (CND) plot and the
regression coefficient spectrum of the estimated numerical
values are displayed in Figure 2, from which it is evident that
and it is not straightforward to discern which of the two
interactions intervenes in the model. However, from a chemical
point of view, it is reasonable to believe that the quantity of
bromoform (x ) and the quantity of sodium hydroxide (x )
1
6
2
2
5
interact (that is, the two factor interaction x × x ). Thus, the
2
5
final model includes β (mean yield value) and β , β , and β .
0
1
5
25
Moreover, the two coefficients β34 (+ β ) and β are both
26
124
borderline cases. This means that these can be included in the
model (or they may be excluded).
The implementation of the screening revealed the exper-
imental variables with significantly influence and their relative
importance (due to the numerical value of each single
regressions coefficient). Hence, the experimental variables
reaction temperature (x ), quantity of bromoform (x ),
1
2
reaction time (x ), quantity of NaOH (x ), and amount of
3
5
TEBA (x ) appeared to be essential in the cyclopropanation
6
(
step 3) of Scheme 1.
Even though the design and model actually indicated an
optimized synthetic protocol (experiment 8 of Table 1), we
wanted to further investigate the cyclopropanation step, where
we also wanted concomitantly (1) to determine a maximized
yield, (2) to design and establish a robust process, and (3) to
minimize the inputs and thus the cost of the synthesis, this by
means of an detailed iso-contour projection of the response
(yield) surface.
Figure 2. CND plot and β-spectrum (regression coefficients plotted in
a bar graph).
the regression coefficients β , β , β and the two-factor
0
1
5
interaction term β13 (+ β ) are significant in the model. The
25
k
3
1
Table 3. Statistical experimental design (2 + c = 2 + 3) with responses measured by means of a 400 MHz H NMR
spectrometer
a
b
experimental variables
measured responses
y(1)
y(2)
y(3)
y [%]
yisol [g]
c
entry
χ1
χ2
χ3
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
89.52
86.48
99.91
99.78
80.24
87.48
93.02
99.97
95.28
96.22
98.76
99.14
90.96
93.38
64.59
99.94
94.58
74.94
76.69
76.02
99.97
99.98
99.55
89.41
86.76
99.95
99.69
80.05
87.46
92.88
99.92
95.48
96.67
98.76
99.36
93.03
90.98
64.75
99.96
97.43
75.02
76.56
75.72
99.99
99.97
99.62
89.58
86.60
99.93
99.85
80.35
87.60
92.78
99.97
95.32
96.30
98.35
99.02
92.62
90.11
64.86
99.94
95.07
74.87
76.58
75.64
99.98
99.96
99.56
89.5
86.6
99.9
99.8
80.2
87.5
92.9
99.9
95.4
96.4
98.6
99.2
92.2
91.5
64.7
99.9
96.0
74.9
76.6
75.8
≈100
≈100
99.6
14.58
14.11
16.27
16.26
13.06
14.25
15.13
16.27
15.54
15.70
16.06
d
10
11
12
13
14
15
16
17
18
19
20
21
22
23
0
0
0
0
0
0
−1
−2
+2
0
−1
0
−1
0
15.02
14.91
10.54
16.27
15.64
12.20
12.48
12.35
16.29
16.29
16.22
0
0
−2
+2
0
0
0
0
0
−2
+2
−2
−1
−2
−2.5
−3.0
0
0
e
e
e
e
e
−2
−2
−1
−1
−1
−1
−1
0
0
0
a
Variable explaination. χ : variable name [unit], levels [−2, −1, 0, +1, +2]. χ : Q
[g], [0.1185 0.158, 0.198, 0.237 0.2765]. χ : Q
[mL], [31.875 42.50, 53.13, 63.75 74.375]. For each experiment 8.0 g of the substrate
(as a 50%
k
1
PTC
2
NaOH
solution) [g], [22.46 29.95, 37.44, 44.93 52.42]. χ : Q
3
bromoform
2
-chloro-3,3-diethoxyprop-1-ene 3 was used, to which were added the PTC (χ ), then the bromoform (χ ), and finally the sodium hydroxide (χ ) as
1
3
2
a 50% aqueous solution over a period of 15−20 min at 0 °C. Then the reaction was stirred at 0 °C for 2 h and then at room temperature for another
b
1
c
2
2 h. The responses were measured by means of H NMR. Isolated yields after workup of the experiment. Total isolated yield achieved during the
d
e
response surface optimizing study was ∑y ≈ 325 g. Isolated yield not measured for this experiment. Optimization experiments performed in
order to test the final model.
isol
C
dx.doi.org/10.1021/op5001012 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
After the screening study was completed, the developed
process (entry 8 of Table 1) was used for a long period (several
months) to produce target cyclopropane intermediate 4 and
final synthon 5 (see details in the Experimental Section).
During this production period, a general evolution of the
synthesis process continued. These amendments were mainly
related to changes in reaction time, time of addition, and the
reaction temperature. As a final task in this exploration and
optimization study, we decided to investigate the experimental
variables related to reagent quantities, namely, the quantities of
phase transfer catalyst (χ ), sodium hydroxide (χ ), and
α , α , α , and α are significant in contributing in the
response surface model.
The model matrix M was pruned (that is, removal of the
3
13
22
33
nonsignificant variables), and a new final multivariate regression
2
model with a product statistics of R = 0.853, RMSEP = 3.638,
and RSD = 3.859 was established. The final variable-pruned
empirical model is provided in eq 3. The model, eq 3, was
subsequently utilized for the production of the iso-contour
projections of the multidimensional response surface, that is
spanned by the three experimental variables χ , ..., χ and the
1
3
response y. The two most influencing variables, χ and χ , were
1
2
2
3
bromoform (χ ), respectively. The three other experimental
investigated continuously over the whole range [−2, +2] of the
experimental space, as defined in the experimental design. The
third of the experimental variables was investigated at five
discrete values, namely, at the five set values [−2, −1, 0, +1, +2]
as used in the experimental design. From the iso-contour map,
Figure 4, it is evident how the effect of the variation of the
variables influences the outcome of the process.
3
variables we decided to keep at levels that we had identified to
be convenient and comprise an addition time for sodium
hydroxide in the range 15−20 min (at 0 °C), then continuously
stirring for 2 h at 0 °C, and then at ∼20 °C for a period of 22 h.
This information was elaborated and implemented in a central
composite design in three variables, where the experimental
variables were renamed and renumbered differently from the
screening design to provide the design matrix Δ provided in
Table 3. Based on this, a model matrix M was created as the
following M = [1 Δ χ × χ χ × χ χ × χ χ × χ χ × χ χ ×
y = f(χ , χ , χ )
1
2
3
2
2
=
α + α χ + α χ + α χ χ + α χ + α χ
2 2 3 3 13 1 3 22 2 33 3
0
1
2
1
3
2
3
1
1
2
2
3
2
χ ]. The model matrix M was multivariate, correlated to the
3
= 96.35 + 6.959χ − 4.077χ + 3.152χ χ − 3.022χ
2 3 1 3 2
response y . The estimated numerical values of the coefficients
isol
2
(α) are provided in Table 4, and presented in a CND and
− 2.236χ3
(3)
Table 4. Regression coefficients with adjacent estimated
numerical values
CONCLUSION
■
Step 3 of Scheme 1 was investigated and successfully optimized
by means of multivariate modeling to provide an excellent yield
(>99%) of 1,1-dibromo-2-chloro-2-(diethoxy-methyl)-
cyclopropane 4. The optimized process was not saddled with
production of any impurities or such impurities were only
formed in quantities that could be removed via the normal
workup procedure. Furthermore, these models operated as (1)
tools for the purpose of optimizing the yield of the process, (2)
a means to insight into how the various experimental variables
influence the performance of the process, thus mading it
possible to establish a high-yielding and robust process, and (3)
an instrument to concomitantly minimize the inputs, i.e., the
consumption of bromoform as an crucial reagent, which is
beneficial for the process economy and environmental aspects.
During the response surface modelling, 325 g of intermediate 4
was successfully produced, which subsequently was converted
to TEB. Thus, this optimization study operated also as a
reagent supply source for several TEB reagent reliant projects
in progress. For the sake of completeness a complete
experimental procedure for the steps for the total synthesis
coefficient
value
coefficient
value
α0
α1
α2
α3
α12
98.093
0.216
α13
α23
α11
α22
α33
2.981
0.996
7.044
−1.075
−3.451
−2.665
−3.992
1.116
(
1−4) of Scheme 1 is included.
Figure 3. Cumulative normal probability distribution plot (CND) and
β-spectrum (all the regression coefficients plotted as a stem graph).
EXPERIMENTAL SECTION
General Methods. H NMR spectra were obtained using a
NMR spectrometer operating at 400 MHz. Chemical shifts are
given relative to internal TMS. 2-Chloro-3,3-diethoxyprop-1-
ene 3 was synthesized as described in the literature, and
reagents were purchased commercially and used without
further purification.
Procedure for the Synthesis of 1,1-Dichloro-2-ethox-
ycyclopropane 2 from Ethoxyethene 1. A three-necked
round-bottom flask was charged with ethoxyethene (0.8 mol,
57.688 g), chloroform (3.2 mol, 381.952 g), and TBAI (3.0
mmol, 1.124 g) and placed in a water−ice bath under
vigorously stirring. To this mixture was added dropwise a
■
1
1a
regression coefficient spectrum; see Figure 3. The product
statistics for the model shows a reasonable good fit of the data,
2
R = 0.879, RMSEP = 3.289, and RSD = 3.489. The model
17
selection was also supported by using Mallows Cp statistics in
9a
combination with stepwise regression. A printout from the
regression analyses performed with the SAS program is
provided in the Supporting Information.
The CND plot and regression coefficient spectrum shows
that in addition to the mean value term α , the coefficients α ,
0
2
D
dx.doi.org/10.1021/op5001012 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 4. Response surface iso-contour projection produced by using the multivariate empirical model eq 3. The iso-contour lines (red lines) shows
the predicted %- yield of target intermediate 1,1-dibromo-2-chloro-2-(diethoxymethyl)cyclopropane (4). To read the plot: the outer frame, that is,
the horizontal row composed of the five subplots, shows the variation in the quantity of the phase transfer catalyst (x ) at five discrete experimental
1
levels. Each of the five subplots displays the iso-contour lines of the response surface when the two experimental variables quantity of sodium
hydroxide (x ) and the quantity of bromoform (x ) are continuously varied within the limits of the abscissa and ordinate axes. The second subplot
2
3
from left, when x = 0.16 was used to predict of the setting values of the variables x and x with the goal to approach an optimized procedure. The
1
2
3
blue bullet points shown in the subplot predicted for phase transfer catalyst (PTC) quantity of 0.16 g show the optimization experiments 21−23 of
Table 3, experiments that confirm the excellent predictive ability of the model eq 3 as all three experiments provided practically quantitative yields.
5
0% fresh aqueous solution of sodium hydroxide (192.01 g)
using a dropping funnel. The temperature was controlled at 0
C. The reaction mixture was left vigorously stirring at 0 °C for
h, and then at room temperature for 22 h. The reaction
after the phases were separated. The dichloromethane layers
was combined and dried with magnesium sulfate (MgSO4·
°
2
xH O) overnight, filtered, and concentrated under reduced
2
pressure on a rotary evaporator. Distillation of the crude
mixture was quenched by adding 240 mL of 6 M hydrochloric
acid. Then water (420 mL) was added, and the aqueous phase
was extracted with dichloromethane (5 times 200 mL). The
combined extracts were dried over magnesium sulfate, filtered,
and concentrated under reduced pressure to afford an almost
colorless residue (the raw product), which was distilled under
reduced pressure to give 106.4 g (86%) of the title intermediate
residue gave the title product 4.
Optimized Procedure Leading to 1,1-Dibromo-2-
chloro-2-(diethoxymethyl)-cyclopropane 4. Aqueous so-
dium hydroxide (50%, 37.44 g) was added dropwise using a
syringe to a gently stirred mixture of 2-chloro-3,3-diethox-
yprop-1-ene 3 (8.00 g, 24 mmol), bromoform (31.875 g), and
tetrabutylammonium (TBAI) (0.158 g). The resulting mixture
was stirred vigorously at ∼0 °C for 2 h and then at room
temperature for another 22 h. After that, three samples were
1
,1-dichloro-2-ethoxycyclopropane 2.
Procedure for the Synthesis of 3 from 2. A single-
1
necked flask is charged with absolute ethanol (250 mL),
pyridine (0.45 mol, 35 g), and 1,1-dichloro-2-ethoxycyclopro-
pane (0.35 mol, 54.250 g). The resulting mixture was refluxed
for 48 h and then was cooled to room temperature before was
concentrated under reduced pressure until the volume was
around 100 mL. After the residue was transferred to a
separatory funnel, water was added, and the aqueous phase
was extracted with dichloromethane. The combined organic
phases were washed with a 0.7 M aqueous solution of copper
sulfate, dried over magnesium sulfate, filtered through a plug of
aluminum oxide, concentrated, and distilled to give 37.767 g
withdrawn and analyzed by H NMR to determine the quantity
of 4. The mean value of the three samples demonstrated a
quantitative conversion to target molecule 4.
Workup. Water (comparable in volume to the organic
phase) was then added. The aqueous phase was extracted with
dichloromethane (3 × 1/3 the volume of the aqueous phase)
after the phases were separated. The dichloromethane layers
was combined and dried with magnesium sulfate (MgSO4·
xH O) overnight, filtered, and concentrated under reduced
2
pressure by on a rotary evaporator. Distillation of the crude
residue gave the title product 4.
(
66%) of 2-chloro- 3,3-diethoxyprop-1-ene.
Procedure for the Synthesis of 1,1-Dibromo-2-chloro-
Procedure for the Synthesis of 5 from 4. To a stirred
mixture of 1,1-dibrom-2-chloro-2-diethoxymethylcyclopropane
(168.00 g, 0.50 mol), ethanol (138.00 g, 3.00 mol), TBAI (1.6
g), and dichloromethane (700 mL) kept in an water−ice bath
was added dropwise fresh 50% aqueous sodium hydroxide
(160.00 g). The mixture above was stirred vigorously at bath
temperature for 24 h, before water was added. The aqueous
phase was extracted with dichloromethane. Then combined
organic extracts were dried over magnesium sulfate, filtered,
concentrated, and distilled to give TEB (110.8 g, 96%) as a
clear liquid.
2
-(diethoxymethyl-cyclopropane 4 Used in the Exper-
imental Design. Aqueous sodium hydroxide (50%, 29.95 or
4.93 g) was added dropwise using a syringe to a gently stirred
4
mixture of 2-chloro-3,3-diethoxyprop-1-ene 3 (8.00 g, 24
mmol), bromoform (122.8 g, 42.50 mL or 184.2 g, 63.75
mL), and tetrabutylammonium (TBAI) (0.158 g or 0.237g).
The resulting mixture was stirred vigorously at ∼0 °C for 2 h
and then at room temperature for another 22 h. The
experiments were performed in random order from which
1
1
three samples were withdrawn and analyzed by H NMR to
Chemical Analysis by Means of H NMR. When the
determine the quantity of 4. The mean value of the three
samples was used as the response value for the actual
experiment of the statistical experimental design.
Workup. Water (comparable in volume to the organic
phase) was then added. The aqueous phase was extracted with
dichloromethane (3 × 1/3 the volume of the aqueous phase)
reaction was complete, the reaction mixture was left in the flask
for precipitation of the solids. An aliquot of the mixture was
stored in refrigerator, and after ∼15 min three samples were
1
withdrawn for recording of the H NMR spectra.
Each of the three samples that were withdrawn from each
single experiment (from statistical experimental design and the
E
dx.doi.org/10.1021/op5001012 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
follow-up experiments) was analyzed on a 1H NMR
spectrometer in order to determine the quantity of target
intermediate 4. Relative quantification was used to calculate the
yield of each sample, using the molar ratio between the target
intermediate 4 and the unconverted substrate 3.
The unreacted substrate 3 and target intermediate 4 both
have a single proton at a tertiary carbon (red methine hydrogen
in Chart 1). Both protons appears as a singlet with chemical
shifts 4.86 and 4.57 ppm, respectively.
(8) See for example: (a) Box, G. E. P.; Hunter, J.; Hunter, W. G.
Statistics for Experimenters: Design, Innovation, and Discovery, 2nd ed.;
Wiley: New York, 2005. (b) Box, G. E. P.; Draper, N. R. Empirical
Model-Building and Response Surfaces; Wiley: New York, 1987.
(9) See for example: (a) Draper, N. R.; Smith, H. Applied Regression
Analysis, 3rd ed.; Wiley: New York, 1998. (b) Montgomery, D. C.;
Peck, E. A. Introduction to Linear Regression Analysis; Wiley: New York,
1
982.
(10) (a) Wold, S.; Sjo
̈
̈
strom, M.; Eriksson, L. Chemom. Intell. Lab.
Syst. 2001, 58, 109−130. (b) Malinowski, E. R. Factor Analysis in
Chemistry, 3rd ed.; Wiley: New York, 2002; pp 1−432. (c) Livingstone,
D. A Practical Guie to Scientific Data Analysis; Wiley: Chichester, 2009.
Chart 1
(
11) Ishikawa, K. Guide to Quality Control, 2nd ed.; Asian
Productivity Organization: Tokyo, 1982.
12) Categories used in the context of the Ishikawa diagram and
(
statistical experimental design that also can be mentioned as pre-
experimental design or prelude to experimental design are (A)
experimental variables known to influence the performance of the
reaction, (B) experimental variables suspected to influence the
performance of the reaction, (C) experimental variables suspected
not to influence the performance of the reaction, and (D)
experimental variables known not to influence the performance of
the reaction.
1
2
-Chloro-3,3-diethoxyprop-1-ene (3). H NMR (CDCl ,
3
4
1
00 MHz) δ: 1.23−1.27 (t, 6H), 3.50−3.69 (m, 4H), 4.86 (s,
H), 5.48 (d, J = 0.76, 1H), 5.68 (t, J = 1.1, 1H).
,1-Dibromo-2-chloro-2-diethoxymethylcyclopropane (4).
H NMR (CDCl , 400 MHz) δ: 1.25−1.34 (m, 6H), 2.02 (d, J
9.4, 1H), 2.12 (d, J = 9.4, 1H), 3.59−3.84 (m, 4H), 4.57 (s,
1
1
(13) When scaling is performed according to eq 1 the scaled low
3
value is set at −1, and the scaled high value becomes set at +1.
=
1
(14) The SAS program system was used for the regression analysis;
H).
see: SAS/STAT 9.1 User’s Guide; SAS Institute Inc.: Cary, NC, 2004.
15) The MATLAB program was utilized for the graphical
(
ASSOCIATED CONTENT
Supporting Information
■
representation of the results and models; see: (a) Using MATLAB
Version 6; The MathWorks, Inc., Natick, MA, 2000. (b) Using
MATLAB Graphics Version 6; The MathWorks, Inc., Natick, MA, 2000.
(16) Product statistics are calculated according to the following
equation: R2 = 1 − SSresidual/(SSmodel + SSresidual)], R2Adj is an
R2 adjusted for the number of parameters in the model relative to the
number of experiments in the experimental design. The R2Adj
provides a measure of the amount of variation about the mean
explained by the model. R2Pred = 1 − (PRESS/ SStotal). PRESS =
Predicted Residual Sum of Squares for the model. PRESS provides a
measure of how well a model fits each of the experiments of the
design. The first step is to calculate the coefficients, the β′s, for the
model without including the first experiment in the modelling. This
model is then used to predict the value of the omitted experiment, and
the residual for this point is then calculated. This procedure is
performed for each of the experiments of the design. Finally the
squared residuals are summed.
*
S
This material is available free of charge via the Internet at
http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Author
*
Phone +47 55 58 34 52. Fax: +47 55 58 94 90. E-mail: Hans.
Bjorsvik@kj.uib.no.
Present Address
†
Department of Medicinal Chemistry at University of Messina,
Italy.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
(
17) (a) Mallows, C. L. Technometrics 1973, 15 (4), 661−675.
Financial support from University of Bergen and Norsk Hydro
is gratefully acknowledged. Mariangela Terranova acknowl-
edges economical support (Erasmus study grant) from
University of Messina. W.S. acknowledges financial support
from China Scholarship Council (CSC).
(b) Hocking, R. R. Biometrics 1976, 32, 1−49. (c) Gilmour, S. G. J. R.
Stat. Soc. Ser. D 1996, 46 (1), 49−56.
REFERENCES
■
(
1
1) (a) Kvernenes, O. H.; Sydnes, L. K. Org. Synth. 2005, 83, 184−
92. (b) Sydnes, L. K.; Holmelid, B.; Kvernenes, O. H.; Sandberg, M.;
Hodne, M.; Bakstad, E. Tetrahedron 2007, 63, 4144−4148.
2) Sydnes, L. K.; Holmelid, B.; Kvernenes, O. H.; Valdersnes, S.;
Hodne, M.; Boman, K. ARKIVOC 2008, No. xiv, 242−268.
3) (a) Sydnes, L. K.; Kvernenes, O. H.; Valdersnes, S. Pure Appl.
(
(
Chem. 2005, 77, 119−130. (b) Valdersnes, S.; Sydnes, L. K. Eur. J. Org.
Chem. 2009, 5816−5831. (c) Valdersnes, S.; Apeland, I.; Flemmen, G.;
Sydnes, L. K. Helv. Chim. Acta 2012, 95, 2099−2122.
(
4) Sydnes, L. K.; Holmelid, B.; Sengee, M.; Hanstein, M. J. Org.
Chem. 2009, 74, 3430−3443.
(
(
5) Sengee, M.; Sydnes, L. K. Synthesis 2011, 3899−3907.
6) Wawrzyniewicz, M.; Makosza, M. Tetrahedron Lett. 1969, 10,
4
(
(
659−4662.
7) (a) Box, G. E. P. J. R. Stat. Soc. Ser. C 1957, 6 (2), 81−101.
b) Barnett, E. H. Ind. Eng. Chem. 1960, 51 (6), 500−503.
F
dx.doi.org/10.1021/op5001012 | Org. Process Res. Dev. XXXX, XXX, XXX−XXX