Toward a simulation approach for alkene ring-closing metathesis: Scope and limitations of a model for RCM

Article abstract of DOI:10.1021/jo201611z

A published model for revealing solvent effects on the ring-closing metathesis (RCM) reaction of diethyl diallylmalonate 7 has been evaluated over a wider range of conditions, to assess its suitability for new applications. Unfortunately, the model is too flexible and the published rate constants do not agree with experimental studies in the literature. However, by fixing the values of important rate constants and restricting the concentration ranges studied, useful conclusions can be drawn about the relative rates of RCM of different substrates, precatalyst concentration can be simulated accurately and the effect of precatalyst loading can be anticipated. Progress has also been made toward applying the model to precatalyst evaluation, but further modifications to the model are necessary to achieve much broader aims.

Full text of DOI:10.1021/jo201611z

ARTICLE
pubs.acs.org/joc
Toward a Simulation Approach for Alkene Ring-closing Metathesis:
Scope and Limitations of a Model for RCM
†
†,‡
§
,†
David J. Nelson, Davide Carboni, Ian W. Ashworth, and Jonathan M. Percy*
†
WestCHEM, Department of Pure and Applied Chemistry, University of Strathclyde, 295 Cathedral Street, Glasgow, G1 1XL,
United Kingdom
§
AstraZeneca Pharmaceutical Development, Silk Road Business Park, Charter Way, Macclesfield, SK10 2NA, United Kingdom
S Supporting Information
b
ABSTRACT: A published model for revealing solvent effects on
the ring-closing metathesis (RCM) reaction of diethyl diallylma-
lonate 7 has been evaluated over a wider range of conditions, to
assess its suitability for new applications. Unfortunately, the model
is too flexible and the published rate constants do not agree with
experimental studies in the literature. However, by fixing the values
of important rate constants and restricting the concentration
ranges studied, useful conclusions can be drawn about the relative
rates of RCM of different substrates, precatalyst concentration can
be simulated accurately and the effect of precatalyst loading can be anticipated. Progress has also been made toward applying the model to
precatalyst evaluation, but further modifications to the model are necessary to achieve much broader aims.
’
INTRODUCTION
The study showed that the quantification of reaction out-
comes required some care, as the process of product isolation can
change the reaction yield unless all the metathesis-active species
The alkene ring-closing metathesis (RCM) reaction has enabled
many complex and important molecules to be constructed in new
ways. The popularity of the RCM reaction has been driven by the
commercial availability of highly active and stable ruthenium
precatalysts such as 1ꢀ3.
6
are quenched before samples are concentrated. The increasing
concentration during solvent removal can result in secondary
2
metathesis events that may be complex in nature and demon-
strates the difficulty involved in using reaction yields from the
synthetic chemistry literature to draw firm conclusions about the
1
interplay between structure and reactivity. We have used H
NMR, a less invasive method, to quantify the outcomes of small-
7
scale metathesis reactions of simple α,ω-dienes and to follow
8
the course of their RCM reactions in kinetic studies. We have
also modeled the potential energy surfaces for these reactions in
9
some detail using the M06-L functional. It should be possible to
interpret the concentration/time data obtained from kinetic
experiments using a combination of computational insights and
quantitative data obtained from fundamental studies of precata-
lyst systems. An appropriate kinetic model that takes account of
precatalyst initiation and active catalyst decomposition would
allow the interpretation of this concentration/time data (which
Industry has seized upon the potential of the technology,
and reactions have now progressed from the realm of the dis-
covery chemist into chemical development and scale-up.
The Boehringer-Ingelheim group have prepared BILN 2061 4
on a multikilogram scale using RCM of diene 5 to form 15-
membered macrocycle 6 as a key step (Scheme 1).
1
2ꢀ4
8
very rarely correspond to a simple kinetic order ) and compar-
isons between different reaction solvents and precatalyst systems.
During the course of the campaign, extensive optimization
studies involving changes of solvent, precatalyst and substrate
structure were carried out, and cyclization efficiency was quanti-
The 2008 study published by Adjiman, Taylor and co-workers
10
5
was therefore of great interest. The study applied a very simple
kinetic model to the cyclization of diethyl diallylmalonate 7 to
form product 8 (using Grubbs second generation precatalyst 1)
and promoted acetic acid as an unexpectedly effective solvent for
fied using the method described by Ercolani et al.; plotting the
concentration of cyclic product squared versus the concentration of
2
cyclic dimer yields a linear plot with slope (EMproduct) /(EMdimer)
(eq 1), which can be used to compare different reaction condi-
tions quantitatively.
Received: August 4, 2011
Published: September 14, 2011
2
2
½
productꢁ =½cyclicdimerꢁ ¼ ðEMproductÞ =ðEMdimerÞ
ð1Þ
r 2011 American Chemical Society
8386
dx.doi.org/10.1021/jo201611z
|J. Org. Chem. 2011, 76, 8386–8393

The Journal of Organic Chemistry
ARTICLE
4
Scheme 1.
1
0
Scheme 2.
’ RESULTS AND DISCUSSION
Reaction concentration/time data were collected for the RCM
reactions of diethyl diallylmalonate 7 to form product 8, hepta-
diene 11 to form cyclopentene 12 and octadiene 13 to form
1
cyclohexene 14 (Scheme 3) catalyzed by 1 and 2, by H NMR (at
8
400 or 600 MHz) following the procedures described previously.
Dichloromethane-d (DCM-d ) was used initially as it is the
2
2
default solvent for synthetic RCM; however, we also used
chloroform-d for many experiments because it is a fraction of
the cost of DCM-d and precatalyst initiation occurs at similar
2
rates in the two solvents (vide infra). Stock solutions of substrates
and precatalysts were prepared in dried solvents (solvents were
allowed to stand over activated 4 Å molecular sieves overnight,
then degassed before use by sparging with dry oxygen-free nitrogen)
using dried volumetric glassware so that weighing and dilution
errors would be minimized. For more dilute reactions, concen-
trated stock solutions were prepared and diluted down immedi-
ately before use. Solid 1 and 2 and their solutions were handled
with care under a flow of nitrogen. Reactions were carried out in
NMR tubes fitted with pierced caps, allowing free egress of
RCM reactions on the basis of the findings; the model took
account of precatalyst initiation and active catalyst decomposi-
10
ethene. The published model (Scheme 2, eqs 2ꢀ6) was con-
tion explicitly (Scheme 2, L = H IMes). The model comprises
12
13
2
structed in both Berkeley Madonna and Micromath Scientist
reversible initiation of 1 to form 14e benzylidene 9, reversible
RCM of 7 to form 8 and irreversible decomposition of two
molecules of 9 to form 10. Numerical integration software was
used to analyze and simulate concentration/time data from
software packages (using the default RungeꢀKutta fourth-order
14
integration method for each) and reaction concentration/time
data (for dienes 7, 11 and 13 and products 8, 12 and 14) were
imported as text files or spreadsheets, respectively. Each experi-
ment was checked for mass balance: the reactions treated in this
manuscript all convert diene to cycloalkene (plus ethene) with-
1
1
reactions that were carried out under reaction conditions
similar to those used for the majority of target synthesis projects.
The meticulous exclusion of oxygen and water from reactions
using glovebox techniques is required for maximum rigor, but
such procedures bear almost no resemblance to the way RCM
reactions are carried out in the synthetic laboratory. Indeed, the
relative robustness of ruthenium-based precatalysts represents
one of the reasons for their wide appeal and ubiquity compared
to the more active but far more fragile molybdenum-based sys-
tems. While synthetic chemists would use dried (and typically
degassed) solvents, they would not usually do more, so the level
of rigor employed by Adjiman, Taylor and co-workers is entirely
appropriate and relevant. If their model could be applied to other
systems, it could be used by synthetic chemists to predict con-
ditions under which maximum reaction yield could be obtained
at optimal substrate concentration, with the most appropriate pre-
catalyst at the lowest practical loading and reaction time, with
obvious and considerable advantages to those involved in scale-
up campaigns. We therefore set out to apply the published model
to our own systems of interest and investigate its scope and
limitations; we report our findings in this manuscript.
7
out the formation of oligomers or side-products.
Initially, attempts were made to reproduce the experiment in
DCM-d reported by Adjiman and Taylor (120 mM 7, 6.7 mM 1)
2
and obtain the published rate constants through fitting.
d=dt½1ꢁ ¼ k1½1ꢁ þ k-1½9ꢁ½PCy3ꢁ
ð2Þ
ð3Þ
ð4Þ
ð5Þ
ð6Þ
d=dt½9ꢁ ¼ k1½1ꢁ þ k-1½9ꢁ½PCy ꢁ-2k3½9ꢁ
3
d=dt½PCy ꢁ ¼ k1½1ꢁ þ k-1½9ꢁ½PCy ꢁ ꢀ k3½9ꢁ
3
3
d=dt½7ꢁ ¼ k2½7ꢁ½9ꢁ þ k-2½8ꢁ½9ꢁ
d=dt½8ꢁ ¼ k2½7ꢁ½9ꢁ ꢀ k-2½8ꢁ½9ꢁ
Unfortunately, the concentration/time profile recorded indi-
cated a faster reaction than the one published (Figure 1). A
slower reaction took place in DCM-d from a freshly opened
2
ampule, although this did not match the literature concentra-
tion/time profile either. Karl Fischer titration showed that the
8
387
dx.doi.org/10.1021/jo201611z |J. Org. Chem. 2011, 76, 8386–8393

The Journal of Organic Chemistry
ARTICLE
Scheme 3
Figure 2. Testing the flexibility of k in the model with (a) k fixed to
1
3
Figure 1. Experimental concentration/time data and simulations there-
of obtained using the Adjiman and Taylor model and rate constants for
zero and with (b) k
3
allowed to change in the fitting routine.
2 2
the RCM of 7 in dry DCM-d (red b, red line) and wet DCM-d (b,
black line) at 298 K. Diene and cycloalkene concentration were both
fitted, but only cycloalkene concentration is shown for clarity.
fitting approach but it represents one way to achieve significant
simplification without recourse to unrealistic reaction conditions.
The fitting was carried out twice: once where k (catalyst de-
3
composition) was fixed to zero and once where k was allowed to
3
ꢀ
1
ꢀ1
Table 1. Rate Constants (L mol
s
) Used to Obtain a
vary in the fitting. Good fits were obtained for all values of k from
1
ꢀ
8
ꢀ4 ꢀ1
ꢀ10
ꢀ2 ꢀ1
Good Fit of the 120 mM Data Set, Plus Those Reported for the
10 to 10
s
when k was fixed, and from 10 to 10
s
3
RCM of 7 in DCM-d2
when k was allowed to change, which is very surprising given the
importance of precatalyst activation. The values of other rate
3
entry
reference
k
1
k
ꢀ1
k
2
k
ꢀ2
k
3
constants changed smoothly with changes in k (Figure 2).
1
b
ꢀ6
1
2
This work
0.000394 19.8
5.39 0.0529 3.28 ꢂ 10
It is clear these rate constants are correlated; the model will
a
tolerate quite large changes in k by readjusting the value of other
Adjiman et al. 0.0617
0.373 0.137 0.00775
0
1
a
10 b
ꢀ1
rate constants in direct proportion to the change in k , because
From reference. Units s
.
1
the rate constants in the model are all dependent on k . The
1
irrelevance of the absolute value of k is a source of considerable
1
drying procedure had been effective, with a water content of ca.
0 ppm before drying (untreated chloroform-d and DCM-d2)
concern; unconstrained fitting of even moderately complex reac-
tions is unwise because of the number of variables and the depen-
dencies between them. The absolute values of strongly correlated
rate constants cannot be obtained from fitting.
With these results in hand, concentration/time data were
recorded for the RCM reactions of diethyl diallylmalonate 7 over
a range of reaction conditions (10 to 500 mM 7, 1 to 10 mol % 1,
Table 2).
Multiple data sets were imported into Berkeley Madonna and
fitted simultaneously (batch fitted) to the Adjiman model, on the
basis that fitting a number of the data sets in Table 2 would yield a
set of rate constants that could describe a wider range of con-
ditions. Unfortunately, a fit that described all data sets at once
could not be achieved through this batch fitting approach. The
rate constants corresponding to the best fit are in Table 3.
While the best fit described the shapes of the concentration/
time profiles well for some data sets (where [7] g 250 mM),
fitting of the experimental data was moderate or poor for others.
Figure 3 shows excellent ([7] = 500 mM), moderate ([7] =
120 mM) and poor ([7] = 50 mM) fits from the batch fitting
results. We refer to the maximum difference between experi-
mental and simulated concentrations during the reaction. At this
point, it is clear that the flexibility of the model with respect to
diene concentration is a major limitation; at a give diene con-
centration, precatalyst concentration can be varied with good fits
obtaining (vide infra). Different best fits can be found for each
3
but less than 10 ppm afterward. It is clear that commercial NMR
solvents from ampules must be treated with some caution if re-
latively significant amounts of water are not to be introduced.
Straightforward drying of solvents using molecular sieves re-
duced the water content of the deuterated solvents to a level
similar to that found in DCM obtained from widely used solvent
purification systems; for example, DCM from the Innovative
Technologies PureSolv system used in-house contains water at
no more than 5 ppm. Although the kinetic data were not con-
sistent with the published literature profile, we fitted the data set;
diene and cycloalkene concentration were both fitted, but only
cycloalkene concentration is shown for clarity. An excellent fit
was obtained for both species. However, quite different values for
all rate constants were required (Table 1), prompting further
investigation with this data set.
Diethyl diallylmalonate 7 is used as a benchmark substrate to
15
assess the performance of new precatalyst systems because it is
assumed to cyclize irreversibly and via rate-limiting precatalyst
1
6,17
initiation.
The absolute values of k and k should be very
1 ꢀ1
influential on the predicted concentration/time profiles, so the
value of k was stepped in factors of 10 between fixed values from
1
ꢀ
10
ꢀ2 ꢀ1
1
0
to 10
s
in the fitting, allowing k and k to optimize.
2 ꢀ2
The k and k constants represent many events comprising the
ring-closing metathesis process; this is an inherent weakness in a
2
ꢀ2
8
388
dx.doi.org/10.1021/jo201611z |J. Org. Chem. 2011, 76, 8386–8393

The Journal of Organic Chemistry
ARTICLE
24
Table 2. Conditions Explored for the RCM Reactions of 7
Scheme 4.
data
a
a
a
set
[7]
[1]
loading of 1
1
2
3
4
5
6
505.9 mM (500 mM) 25.7 mM (25 mM)
5.1 mol % (5.0 mol %)
491.7 mM (500 mM) 12.0 mM (12.5 mM) 2.4 mol % (2.5 mol %)
507.0 mM (500 mM) 5.2 mM (5 mM)
404.5 mM (400 mM) 14.4 mM (14 mM)
1.0 mol % (1.0 mol %)
3.6 mol % (3.5 mol %)
Attention then turned to constraining the value of k in the
fitting, reflecting the importance of the precatalyst initiation
1
250.7 mM (250 mM) 25.9 mM (25 mM) 10.3 mol % (10.0 mol %)
250.5 mM (250 mM) 12.7 mM (12.5 mM) 5.1 mol % (5.0 mol %)
event. The kinetics of the initiation of precatalysts 1, 2 and 3
b
1
8ꢀ20
7
121.7 mM (120 mM) 6.7 mM (6.7 mM)
121.8 mM (120 mM) 6.7 mM (6.7 mM)
119.6 mM (120 mM) 6.7 mM (6.7 mM)
5.5 mol % (5.6 mol %)
5.5 mol % (5.6 mol %)
5.6 mol % (5.6 mol %)
1.6 mol % (1.5 mol %)
10.1 mol % (10.0 mol %)
2.5 mol % (2.5 mol %)
mol % (1.0 mol %)
have been studied in some depth
and the initiation rate of 2
b
ꢀ1 ꢀ1 20
8
in DCM at 298 K is already known (0.0263 L mol
s ).
b,c
ꢀ4
ꢀ5 ꢀ1
9
Values of 1.40 ꢂ 10 and 4.5 ꢂ 10
s
were obtained for 1
ꢀ
1
ꢀ1
1
1
1
1
0
1
2
3
76.1 mM (75 mM)
50.2 mM (50 mM)
48.2 mM (50 mM)
9.9 mM (10 mM)
1.2 mM (1.1 mM)
5.1 mM (5 mM)
1.2 mM (1.3 mM)
in DCM and chloroform, respectively, and 0.0234 L mol
s
for 2 in chloroform (see the Supporting Information for details)
by following the reaction of the precatalyst with ethyl vinyl ether
1
at 298 K by either H NMR (for 1) or UV/vis spectroscopy
0.1 mM (0.1 mM)
a
c
b
10
(for 2). These modest differences in initiation rate render chloro-
form-d a cost-effective solvent for the study of RCM.
Nominal values in parentheses. Conditions used by Adjiman et al.
Kinetic data collected using H NMR at 600 MHz with 2 scans per data point.
1
Though the 14e complex 9 is an accepted intermediate on the
energy surfaces for metathesis pathways, the barriers to alkene
ꢀ1
ꢀ1
2
1,22
Table 3. Rate Constants (L mol
Fitting 13 RCM Reaction Concentration/Time Profiles to the
Model in Scheme 2
s ) Obtained from Batch-
coordination or phosphane recoordination are very small.
The value of k 1/k (where k refers to the rate of alkene co-
ꢀ
2
2
ordination) is close to 1 from the experimental work of Sanford
1
8
a
et al., consistent with similar barriers for both pathways.
However, complex 9 undergoes barrierless forward and back-
ward reactions, so it follows that k /k should reflect this in a
k
1
k
ꢀ1
k
2
k
ꢀ2
k
3
ꢀ5
1
.64 ꢂ 10
243
64.1
0.890
0.567
1
ꢀ1
a
ꢀ1
Units s
.
very low numerical value. The absolute value of K for the initiation
ꢀ
1
therefore derives from the ca. 25 kcal mol energy for RuꢀP
ꢀ
19
bond separation; k /k = K ≈ 10 at 298 K follows from this.
1
ꢀ1
The ratios of k /k derived by Adjiman and Taylor (k /k =
ꢀ1
1
ꢀ1
1
0
.18 in dichloromethane, Table 1) are therefore inconsistent
with detailed experimental and theoretical studies on the phos-
phane dissociation event.
The values of k , which represent the rate of a bimolecular
3
decomposition of active catalyst, also varied considerably with
10
solvent. According to the work of Hong et al., phosphane-
bound methylidene 15 decomposes via 14e methylidene 16 to
23,24
yield diruthenium complex 10 (Scheme 4);
two molecules of
1
6 and one of tricyclohexylphosphane are consumed to produce
one molecule of 10, with tricyclohexylphosphane being con-
verted to tricyclohexylphosphonium chloride in the process.
In the Adjiman model, this pathway is represented by the term
2
1
k [9] , but inspection of the high-field region of the H NMR
3
spectrum of a 500 mM RCM of 7 (2.5 mol % 1) did not reveal 10,
whereas 1 and 15 are clearly visible. Formation of the diruthe-
nium hydride complex 10 has been observed in concentrated
solutions (>20 mM) of phosphane-bound methylidene 15 at
higher temperatures (322 K), or in the presence of a large excess
of ethene, so the formation of 10 (which requires two bimole-
cular steps) would be expected to be slow under the mild and
more dilute conditions used here and in most synthetic RCM
experiments. The decomposition pathway characterized by Hong
et al. has not yet been shown to account for significant quantities of
ruthenium-carbene loss in synthetic RCM experiments.
Figure 3. Testing the quality of fitting as a function of diene
concentration for (a) [7] = 500 mM (1 mol % 1) (b = data points,
solid line = simulation) and (b) [7] = 120 mM (5.5 mol % 1)
The value of k was then fixed to the experimentally deter-
1
(
2, red solid line) and [7] = 50 mM (2.5 mol % 1) (blue (, blue solid
mined value (for the relevant precatalyst/solvent combination)
for subsequent fitting. Simultaneous fitting of the 13 RCM
line).
reaction concentration/time profiles (with k fixed to zero) in
3
data set; either these are not global but local minima, or the model
simply fails to describe a range of conditions.
Table 2 yielded a new set of rate constants (Table 4). Once again,
these rate constants failed to describe a range of concentrations
8
389
dx.doi.org/10.1021/jo201611z |J. Org. Chem. 2011, 76, 8386–8393

The Journal of Organic Chemistry
ARTICLE
and only described reactions well when [7] g 250 mM. At this
point, the limitations appear to follow from the way in which the
model is constructed (that is, the differential equations used to
set it up).
Adjiman and Taylor is an artifact arising from the use of an
ꢀ2
ꢀ1
excessively large value of k (6.17 ꢂ 10
s ).
1
A second alkylidene species was observed to form slowly
during RCM reactions, with a chemical shift consistent with
The literature model was also constructed in Micromath
Scientist, which provides a number of useful statistics with which
to evaluate and compare reaction models, in order to evaluate
metathesis inactive methylidene complex 15 (δ = 17.8 ppm),
H
1
8
formed by irreversible recapture of 16 by phosphane; despite
recent and detailed low-temperature NMR studies by van der
13
27
how flexible the model was with a fixed value for k . Three data
Eide, 16 has not yet been detected by NMR. Capture of 16 by
1
sets covering a wide concentration range (entries 3, 7, and 12 in
Table 2) were fitted to the model in turn, providing good fits in
each case but requiring vastly different rate constants (entries 1,
phosphane is slow because the concentrations of 16 and phos-
phane are very low; substrates, ethene and products, all compe-
titive and reversibly binding ligands, are all present at much
higher concentrations. In the reaction profile in Figure 4, com-
plete turnover of 7 has been achieved at the expense of ca. 30% of
the original charge of 1; of this, ca. 15% appears as 15 so an equal
amount (ca. 15%) of the original charge of 1 remains unac-
counted for. The continued decrease in the concentration of
precatalyst 1 even after the reaction has finished is due to reaction
with ethene, which is known to yield a low energy metallocyclo-
2
, and 3 in Table 5).
Analysis of the fitting statistics showed that all rate constants
were highly correlated, in agreement with the previous result
Figure 2), and that 95% confidence limits for each rate constant
were very wide indeed. The confidence limits for k are so wide
(
3
(and encompass zero) as to suggest that its value has very little
bearing on the reaction, as would be expected for a bimolecular
process occurring under very dilute conditions. Attempts at fit-
ting multiple data sets using Scientist were unsuccessful; fitting
the three data sets which had been examined separately yielded a
good fit for the 500 mM reaction but underestimated the rates of
the 120 and 50 mM reactions by ca. 30 and 50%, respectively
28
butane after reaction with two molecules of ethene. This initiation
behavior was confirmed by following the decay of the precatalyst 1
in ethene-sparged solvent, which occurred at the same rate as pre-
catalyst initiation with ethyl vinyl ether.
Adjiman and Taylor present a simulation of active catalyst
concentration which shows rapid and complete precatalyst dis-
sociation followed by active catalyst decomposition at various
(
entry 4 in Table 5).
It is now useful to examine the model in more detail, under-
10
stand more fully the importance of k and identify the scope and
rates in each solvent, which is not consistent with experimental
1
utility of the model in its current form. One of the important
predictions of Adjiman, Taylor and co-workers concerns the ex-
tent to which precatalyst 1 dissociates, and the concentration/
time profile of active catalyst 9. The published paper does not
report any measurements of the concentration of 1, even though
the solution concentration would have allowed the benzylidene
observations. For each of the reactions in Table 2, the precatalyst
concentration after approximately 2000 s was measured by H
1
NMR integration and compared to the values obtained at the
same time point using simulations with rate constants from Adjiman
and Taylor, or using our best fit set of rate constants from Table 4
(Figure 5).
It can be concluded that the rate constants in Table 4 re-
present the behavior of the precatalyst more effectively, but still
do not completely account for its behavior in RCM reactions.
proton to be observed (δ = ca. 19.2 ppm in chloroform-d and
H
2
5
DCM-d2) and integrated with confidence. When spectra were
recorded over a 20 ppm chemical shift range, the change in
concentration of 1 was seen to be modest during the course of the
reaction. Most of the initial charge of 1 remains well after
complete conversion of 7 to 8, in agreement with previous
Various values of k 1, k and k can still be obtained, depending
ꢀ
2
ꢀ2
on which data set is fitted, presumably due to the large number of
2
6
literature reports, indicating clearly that the reaction conditions
are not causing extensive catalyst decomposition on the time
scale of the RCM (Figure 4).
Complete dissociation of phosphane clearly does not occur;
the simulated catalyst concentration/time profile presented by
ꢀ1
ꢀ1
Table 4. Rate Constants (L mol
s ) Obtained from Batch-
Fitting 13 RCM Reaction Concentration/Time Profiles to the
ꢀ
4
ꢀ1
Model in Scheme 2, Fixing k to 1.4 ꢂ 10
s
and k to 0 L
1
3
ꢀ
1
ꢀ1
mol
s
a,b
a
k
1
k
ꢀ1
k
2
k
ꢀ2
k
3
Figure 4. Concentration/time profiles of 8 (b), precatalyst 1 (O) and
inactive phosphane-bound methylidene 13 (red O) during the RCM of
7 (120 mM, 5.5 mol % 1; entry 9 in Table 2).
ꢀ
4
1
.4 ꢂ 10
29.8
7.69
0.110
0
a
b
ꢀ1
Value fixed. Units s
.
ꢀ
1
ꢀ1
Table 5. Rate Constants (L mol
s ) Obtained through Fitting Data Sets in Micromath Scientist; Error Bars Are 95%
Confidence Intervals
entry
data set(s)
k
ꢀ1
k
2
k
ꢀ2
k
3
1
2
3
4
3
23.4 ( 5.9
28.8 ( 11.5
166 ( 61
7.19 ( 0.525
13.9 ( 1.95
35.9 ( 5.1
0.0394 ( 0.0011
0.213 ( 0.0797
0.195 ( 0.118
0.461 ( 0.501
1.42 ꢂ 10ꢀ ( 9.41 ꢂ 10ꢀ13
0.0308 ( 2.3763
14
7
12
3, 7, 12
27.4 ( 14.5
7.99 ( 1.30
0.0452 ( 0.0271
0.621 ( 1.049
8
390
dx.doi.org/10.1021/jo201611z |J. Org. Chem. 2011, 76, 8386–8393

The Journal of Organic Chemistry
ARTICLE
Figure 5. Plot of simulated versus measured [1] after ca. 2000 s in
experiments in Table 2 (except data set 13) using (a) rate constants in
Table 1, entry 2 (b) or (b) rate constants in Table 4 (red b); the black
line represents identity between simulated and measured values (i.e.,
slope = 1).
ꢀ
1
ꢀ1
Table 6. Rate Constants (L mol
s ) Used to Describe
RCM of 7 Across Defined Concentration Ranges
entry
[7] range
k
ꢀ1
k
2
k
ꢀ2
Figure 6. Rate constants in Table 6 describe experiments at (a)
1
2
250ꢀ500 mM
50ꢀ120 mM
26.6
71.1
6.90
20.0
0.103
0.165
500 mM, using entry 1, and (b) 50 mM, using entry 2, relatively well,
even when the precatalyst concentration varies.
local minima for the fitting method. Further changes to the
model are required before a wide range of concentrations can be
described.
ꢀ
1
ꢀ1
Table 7. Rate Constants (L mol
s ) Obtained from Fitting
Concentration/Time Data for Heptadiene 11, Octadiene 13
and Diethyl Diallylmalonate 7 RCM Reactions
Simultaneously
Attempts were therefore made to fit multiple data sets within a
narrower range of concentrations and explore the concentration
range over which the approximations and simplifications of the
model hold. Rate constants were obtained (through data fitting)
that describe all experiments with 250ꢀ500 mM 7 (Table 6,
entry 1; data sets 1 to 6 in Table 2) and a second set that describe
all experiments with 50ꢀ120 mM 7 (Table 6, entry 2; data sets 7
to 12 in Table 2); k and k were fixed in the fitting. Examples of
a
diene
k
ꢀ1
k
2
k
ꢀ2
k
rel
11
359
610
166
11.6
3.32
13.1
0.589
1.000
1
7
2
2733
0.272
a
1
3
Calculated by dividing k
2
for that substrate by k
2
for octadiene RCM.
the fits obtained can be found in Figure 6.
There are clear differences between the rate constants re-
quired to fit the different concentration regimes, with higher k
ꢀ
1
and k values required to successfully model the lower concen-
2
tration reactions. The different value of k 1 is required due to the
ꢀ
incorrect modeling of the 14e species (methylidene 16 does not
capture phosphane reversibly) and the absence of ethene from
the reaction model means that [ethene] is effectively embedded
in k 2. However, the rate constants successfully fit different pre-
ꢀ
catalyst loadings and substrate concentrations within these ranges
and are therefore useful for predicting the effect of changing these
parameters, for example, to predict the lowest loading of 1 that will
effect complete reaction within a given time frame.
The extraction of rate constants for different substrates was
the original aim of this project, so attention turned to assessing
different substrates in RCM reactions quantitatively from reac-
tions conducted at the same initial diene concentration. Con-
centration/time data for the RCM of 7 (Table 2, data set 13),
heptadiene 11 to form cyclopentene (in duplicate) and octadiene
Figure 7. Fitting concentration/time data for the RCM reactions of
heptadiene (( = data points, solid black line = simulation; ), dashed
black line), octadiene (red 2, solid red line and red 4, dashed red line)
and diethyl diallylmalonate (blue b, solid blue line); the initial substrate
concentration was 10 mM (1 mol % 1).
1
2 (in duplicate) in DCM-d , all with 10 mM substrate and
The model is therefore useful for deriving quantitative con-
clusions about the relative rate of RCM of different substrates.
Interestingly, the RCM of diethyl diallylmalonate appears to
occur at half the rate of the ‘parent’ heptadiene substrate, despite
the gem-disubstitution; this order of reactivity was confirmed by a
competition RCM experiment (5 mM heptadiene, 5 mM diethyl
diallylmalonate, 0.1 mM 1) in which the heptadiene cyclized
2
0
.1 mM 1, were imported into Berkeley Madonna and fitted to
ꢀ
4
ꢀ1
the simple model in Scheme 2. The values of k (1.4 ꢂ 10
s )
1
and k (zero) were fixed and the data was fitted with a common
3
k1, k and k (substrate independent rate constants) but a
different k and k for each substrate. The rate constants in Table 7
and the fit in Figure 7 were obtained.
ꢀ1
3
2
ꢀ2
8
391
dx.doi.org/10.1021/jo201611z |J. Org. Chem. 2011, 76, 8386–8393

The Journal of Organic Chemistry
ARTICLE
ꢀ
1
ꢀ1
Table 8. Rate Constants (L mol
Fitting Data Sets
s ) Obtained from Batch
G2
GH2
1
5
2
5
6
2
6
entry
k
ꢀ1
k
k
k
ꢀ2
k
k
ꢀ2
ꢀ
6
1
2
3
984
789
N/A
224
200
1.44 ꢂ 10
405 8.28
426 0.0136
ꢀ5a
8.89 ꢂ 10
0.0234
0.0511
b
22.7
83.6 6.69
a
ꢀ1
b
Units s . Measured by following the reaction of 2 with ethyl vinyl
ether in chloroform at 298 K by UV/vis spectrometry following the
experimental procedure in ref 18 and fixed in the fitting.
more rapidly (see the Supporting Information). Small scale RCM
experiments at up to [7] = 4 M showed that despite the fact that 7
undergoes RCM slower than heptadiene 11, it is a far more
Figure 8. Experimental data points and simulated concentration/time
profiles for the RCM reactions of a 5 mM/5 mM solution of 11 and 13
with either 1 (red 2 = data points for 14; solid red line = simulation of
14, red ( = data points for 12, dashed red line = simulation of 12) or 2
29
efficient reaction (higher EMT) and does not produce oligo-
meric material even when run almost neat. This is in stark con-
trast to heptadiene which produces oligomeric material even at
(
2, solid black line and (, dashed black line); diamonds/dashed lines
represent 12 and triangles/solid lines represent 14.
ꢀ
2
7
concentrations as low as 10 M, further highlighting the need
for a quantitative understanding of the effects of substrate struc-
ture on RCM rate and efficiency; trends in RCM rate are not
GH2
d=dt½12ꢁ ¼ k1 ½2ꢁ½11ꢁ þ k2½9ꢁ½11ꢁ ꢀ k-2½9ꢁ½12ꢁ
ð10Þ
ð11Þ
ð12Þ
6
always the same as those in RCM efficiency.
Fitting of a batch of data in chloroform-d (11/13 mixtures,
each with 0.1 mM 1: 10/0, 9/1, 7.5/2.5, 5/5, 2.5/7.5, 1/9, 0/10)
GH2
d=dt½13ꢁ ¼ k1 ½2ꢁ½13ꢁ ꢀ k ½9ꢁ½13ꢁ þ k ½9ꢁ½14ꢁ
2
-2
ꢀ
5
ꢀ1
with k fixed (4.5 ꢂ 10
s ) yielded a set of rate constants for
1
heptadiene and octadiene RCM in chloroform-d (Table 8, entry 1).
Many studies have been conducted to understand the effects
of precatalyst on reaction rate and outcome,
GH2
d=dt½14ꢁ ¼ k1 ½2ꢁ½13ꢁ þ k2½9ꢁ½13ꢁ ꢀ k-2½9ꢁ½14ꢁ
17,30
so the ability of
The model was set up to allow simulation of RCM of binary
mixtures of dienes 11 and 13 with 2, and to see how the model
the model to predict the effect of changing the precatalyst was
investigated. Precatalyst 2 generates the same active species as 1
after the first turnover, so two competition reactions (in duplicate
in chloroform-d with either 0.1 mM 1 or 0.1 mM 2, plus 5 mM
heptadiene and 5 mM octadiene) were conducted and the data
were imported into Berkeley Madonna.
behaved for data obtained with 1. Values of k obtained from
1
5
initiation kinetics were fixed in the simulations. Values of k ,
2
5
6
6
k
, k , k
and k were allowed to vary (Table 8, entry 3).
ꢀ2
2
ꢀ2 ꢀ1
The fitting was quite successful for 2, though less successful for 1;
the exact shape of concentration/time profile cannot be simu-
lated (Figure 8).
5
5
6
The four experiments were fitted with common k , k , k2
2
ꢀ2
6
and k (forward and backward rates for heptadiene and octadiene
ꢀ
2
It does not appear that the same model can be used for both
precatalysts, which may suggest strongly that there are significant
mechanistic differences between the processes involving 1 and 2.
Nevertheless, these initial successes for the simulation for the
interchange mechanism are extremely promising.
respectively, the superscript refers to the product ring size)
G2
G2
ꢀ1
GH2
with k1 and k
to model the behavior of 1 and k1
to model
the initiation of 2; although the initiation of 2 has been shown to
19,20
proceed via an interchange mechanism,
a dissociative me-
chanism (the simplest model) was trialled first to see if a correctly
shaped concentration/time plot could be generated. Fitting the
’
CONCLUSIONS
data (k fixed to zero) yielded a set of rate constants similar to
3
We have found that the kinetic model published by Adjiman
those obtained for the various chloroform/1 data (Table 8, entry 1)
that approximately described the data from reactions with 1 but
more accurately described the octadiene/cyclohexene profiles
from reactions with 2 (Table 8, entry 2). Recently, the timing of
events involved in the initiation of Grubbs-Hoveyda second
generation precatalyst 2 has been re-evaluated. Vorfalt et al. have
and Taylor significantly simplifies the canonical mechanism for
RCM at some cost. Specifically, the treatment of the ruthenium
species is not detailed enough, the precatalyst initiation rates
obtained in the original paper through data fitting are incon-
sistent with literature knowledge, the prediction of active catalyst
concentration is incorrect, and ethene is not considered at all.
Consequently the model does not describe even a modest range
of substrate concentrations well. However, when constrained
with measured initiation rate constants, concentration/time profiles
can be simulated accurately, albeit over relatively narrow (ca. 100
to 300 mM) concentration ranges, and changes in precatalyst
loading can be described. Pleasingly, from kinetic experiments on
different substrates (at a common concentration, and with a
31
recently shown that precatalyst initiation is irreversible, in dis-
agreement with the more established “boomerang” mechanism
32
previously proposed.
Modifying the way in which the initiation behavior is modeled
for 2 may therefore be appropriate; an interchange mechanism,
(eqs 7ꢀ12) was therefore explored.
GH2
GH2
d=dt½2ꢁ ¼ k1 ½2ꢁ½11ꢁ ꢀ k1 ½2ꢁ½13ꢁ
ð7Þ
ð8Þ
ð9Þ
constant k /k in the fitting), relative rates can be extracted, and
1
ꢀ1
thus useful quantitative performance comparisons can be made
between RCM substrates. Unfortunately, it is not possible to
use a common model for accurate simulation with precatalysts
GH2
GH2
d=dt½9ꢁ ¼ k1 ½2ꢁ½11ꢁ þ k1 ½2ꢁ½13ꢁ
1
and 2, consistent with recent findings on the mechanism of
GH2
d=dt½11ꢁ ¼ k1
½2ꢁ½11ꢁ ꢀ k ½9ꢁ½11ꢁ þ k ½9ꢁ½12ꢁ
2
-2
initiation of 2.
8
392
dx.doi.org/10.1021/jo201611z |J. Org. Chem. 2011, 76, 8386–8393

The Journal of Organic Chemistry
ARTICLE
These studies show the need for the development of a more
robust modeling approach for RCM if kinetic data are to be
interpreted successfully and fully. We are currently working on
further development and elaboration of these models for appli-
cations over wider concentration ranges and with different pre-
catalysts.
’ REFERENCES
(1) Hoveyda, A. H.; Zhugralin, A. R. Nature 2007, 450, 243.
(2) Yee, N. K.; Farina, V.; Houpis, I. N.; Haddad, N.; Frutos, R. P.;
Gallou, F.; Wang, X.-J.; Wei, X.; Simpson, R. D.; Feng, X.; Fuchs, V.; Xu,
Y.; Tan, J.; Zhang, L.; Xu, J.; Smith-Keenan, L. L.; Vitous, J.; Ridges,
M. D.; Spinelli, E. M.; Johnson, M.; Donsbach, K.; Nicola, T.; Brenner,
M.; Winter, E.; Kreye, P.; Samstag, W. J. Org. Chem. 2006, 71, 7133.
(3) Shu, C.; Zeng, X.; Hao, M.-H.; Wei, X.; Yee, N. K.; Busacca,
’
EXPERIMENTAL SECTION
C. A.; Han, Z.; Farina, V.; Senanayake, C. H. Org. Lett. 2008, 10, 1303.
(4) Farina, V.; Shu, C.; Zeng, X.; Wei, X.; Han, Z.; Yee, N. K.;
Senanayake, C. H. Org. Process Res. Dev. 2009, 13, 250.
1
Kinetics experiments were followed by H NMR, either by observa-
tion at 400 MHz with a BBFO-z-ATMA probe or at 600 MHz with a
TBI-z probe (inverse probe). Both instruments possess temperature
control units which maintained the samples at 298 K throughout. All
kinetics experiments were conducted in NMR tubes fittedwithpiercedcaps.
Solutions for kinetics experiments were prepared using methods similar
to those reported previously. Glove box conditions were not employed,
but the precatalysts and solutions thereof were handled with care under
nitrogen. All solutions were prepared in dry volumetric glassware and
(
5) Ercolani, G.; Mandolini, L.; Mencarelli, P.; Roelens, S. J. Am.
Chem. Soc. 1993, 115, 3901.
6) Mitchell, L.; Parkinson, J. A.; Percy, J. M.; Singh, K. J. Org. Chem.
008, 73, 2389.
7) Nelson, D. J.; Ashworth, I. W.; Hillier, I. H.; Kyne, S. H.; Pandian,
(
2
(
8
S.; Parkinson, J. A.; Percy, J. M.; Rinaudo, G.; Vincent, M. A. Chem. Eur.
J. 2011, DOI: 10.1002/chem.201101662.
(8) Ashworth, I. W.; Carboni, D.; Hillier, I. H.; Nelson, D. J.; Percy,
J. M.; Rinaudo, G.; Vincent, M. A. Chem. Commun. 2010, 46, 7145.
(9) Hillier, I. H.; Pandian, S.; Percy, J. M.; Vincent, M. A. Dalton
Trans. 2011, 40, 1061.
2
carefully transferred using dry gastight syringes. DCM-d was purchased
from Goss Scientific and chloroform-d from Sigma-Aldrich; solvents
were dried on activated 4 Å molecular sieves for at least 24 h before use.
The internal standard (1,3,5-trimethoxybenzene) and precatalysts 1 and
(
10) Adjiman, C. S.; Clarke, A. J.; Cooper, G.; Taylor, P. C. Chem.
Commun. 2008, 2806.
11) Wieberg, K. B. In Techniques of Chemistry; Lewis, E. S., Ed.;
Wiley: New York, 1974.
12) Zahnley, T.; Macey, R.; Oster, G. In Berkeley Madonna, 8.3.18
ed.; University of California at Berkeley: Berkeley, CA, 2010.
13) In Micromath Scientist, 3.0 ed.; Micromath: St. Louis, MO,
011.
14) Butcher, J. C. Numerical Methods for Ordinary Differential
Equations; Wiley: Chichester, 2003.
15) Ritter, T.; Hejl, A.; Wenzel, A. G.; Funk, T. W.; Grubbs, R. H.
2
were purchased from Sigma-Aldrich and used as supplied.
(
For low concentration reactions, the substrate/internal standard
solution was used to tune, match, lock and shim the instrument. A small
ca. 5 to 20 μL) volume of a concentrated precatalyst solution was added
(
(
and the tube was shaken. For higher concentration reactions the sub-
strate/internal standard solution and precatalyst solution were prepared,
a “t
substrate/standard solution and this was used to prepare the instrument
and acquire the first spectrum. Following this, in a separate tube, x μL
precatalyst solution was added to (500 ꢀ x) μL substrate/standard
solution and the tube was shaken.
NMR spectra were then acquired periodically until after the RCM
reaction had finished; a D setting of 35 s was used (5 times the largest T
(
2
0
” sample was prepared from x μL solvent plus (500 ꢀ x) μL
(
(
Organometallics 2006, 25, 5740.
(16) Dias, E. L.; Nguyen, S. T.; Grubbs, R. H. J. Am. Chem. Soc. 1997,
119, 3887.
(17) Stewart, I. C.; Keitz, B. K.; Kuhn, K. M.; Thomas, R. M.;
Grubbs, R. H. J. Am. Chem. Soc. 2010, 132, 8534.
(18) Sanford, M. S.; Love, J. A.; Grubbs, R. H. J. Am. Chem. Soc. 2001,
1
1
measured: 7 s, see Supporting Information). Spectra were processed
using proprietary software.
33
1
23, 6543.
19) Vorfalt, T.; Wannowius, K. J.; Plenio, H. Angew. Chem., Int. Ed.
2010, 49, 5533.
20) Ashworth, I. W.; Hillier, I. H.; Nelson, D. J.; Percy, J. M.;
(
’
ASSOCIATED CONTENT
(
S
Supporting Information. Kinetic data, sample spectra,
b
Vincent, M. A. Chem. Commun. 2011, 47, 5428.
(21) Tsipis, A. C.; Orpen, A. G.; Harvey, J. N. Dalton Trans.
2005, 2849.
rate constants and simulated profiles from data-fitting. This
material is available free of charge via the Internet at http://
pubs.acs.org.
(22) Torker, S.; Merki, D.; Chen, P. J. Am. Chem. Soc. 2008,
1
1
30, 4808.
(23) Hong, S. H.; Day, M. W.; Grubbs, R. H. J. Am. Chem. Soc. 2004,
’
AUTHOR INFORMATION
26, 7414.
(
24) Hong, S. H.; Wenzel, A. G.; Salguero, T. T.; Day, M. W.;
Grubbs, R. H. J. Am. Chem. Soc. 2007, 129, 7961.
25) Scholl, M.; Ding, S.; Lee, C. W.; Grubbs, R. H. Org. Lett. 1999,
1, 953.
Corresponding Author
*
jonathan.percy@strath.ac.uk
(
Present Addresses
Laboratorio di Scienze dei Materiali e Nanotecnologie (LMNT),
‡
(26) Love, J. A.; Sanford, M. S.; Day, M. W.; Grubbs, R. H. J. Am.
Chem. Soc. 2003, 125, 10103.
D.ADU Universit ꢀa di Sassari, c/o Porto Conte, Ricerche, 07041
Alghero (SS), Sardinia, Italy
(
(
(
(
27) van der Eide, E. F.; Piers, W. E. Nat. Chem. 2010, 2, 571.
28) Romero, P. E.; Piers, W. E. J. Am. Chem. Soc. 2005, 127, 5032.
29) Kirby, A. J. Adv. Phys. Org. Chem. 1980, 17, 183.
30) Bieniek, M.; Michrowska, A.; Usanov, D. L.; Grela, K. Chem.—
’
ACKNOWLEDGMENT
Eur. J. 2008, 14, 806.
31) Vorfalt, T.; Wannowius, K. J.; Thiel, V.; Plenio, H. Chem.—Eur.
J. 2010, 16, 12312.
32) Kingsbury, J. S.; Harrity, J. P. A.; Bonitatebus, P. J.; Hoveyda,
A. H. J. Am. Chem. Soc. 1999, 121, 791.
33) Topspin, 2.1 ed.; Bruker: Billerica, MA, 2007.
(
We thank Dr. John Parkinson and Mr. Craig Irving for assistance
with NMR experiments and Mr. Gavin Bain for Karl Fischer
titrimetric measurements. We thank AstraZeneca (Industrial
CASE to D.J.N.) and the EPSRC Initiative in Physical Organic
Chemistry 2 (EP/G013160/1 and EP/G013020/1, fellowship to
D.C.) for funding.
(
(
8
393
dx.doi.org/10.1021/jo201611z |J. Org. Chem. 2011, 76, 8386–8393