Reaction progress analysis: Powerful tool for understanding suzuki-miyaura reaction and control of polychlorobiphenyl impurity

Article abstract of DOI:10.1021/op800205x

Cross coupling of unsaturated aryl or vinyl triflates/halides with aryl boronic acids using Pd as catalyst (Suzuki coupling) have become increasingly attractive for making the heterocoupled product (Ar-Ar′). However, most Pd cycle reactions produce some h

Full text of DOI:10.1021/op800205x

Organic Process Research & Development 2009, 13, 420–428
Reaction Progress Analysis: Powerful Tool for Understanding Suzuki-Miyaura
Reaction and Control of Polychlorobiphenyl Impurity
Sandeep B. Kedia* and Mark B. Mitchell
GlaxoSmithKline Research and DeVelopment, FiVe Moore DriVe, Research Triangle Park, North Carolina 27709, U.S.A.
Abstract:
Scheme 1. Suzuki-Miyaura coupling
Cross coupling of unsaturated aryl or vinyl triflates/halides with
aryl boronic acids using Pd as catalyst (Suzuki coupling) have
become increasingly attractive for making the heterocoupled
product (Ar-Ar′). However, most Pd cycle reactions produce some
homocoupled impurity. This is a major concern when the impuri-
ties are polychlorobiphenyls (PCBs). To understand and control
the impurity formation was the most important challenge. This
contribution demonstrates how reaction progress analysis along
with mathematical modeling can be used to gain fundamental
understanding and postulate reaction mechanism, providing an
insight on how to reduce the amount of byproduct generated in
the catalytic cycle with a limited number of experiments (typically
three). Fundamentals, such as stability of catalyst, catalyst poison-
ing, inhibition, and reaction mechanism, can also be answered with
these limited number of experiments. Characterization of the
catalytic cycle led to a semi-batch addition regime of boronic acid,
which effectively eliminated PCB generation by forcing the
catalytic cycle to partition between the oxidative addition inter-
mediate (I) and transmetallated intermediate (II).
ematical modeling to gain an understanding of key features of
the reaction mechanism. In this example PCB formation
stemmed solely from homocoupling of the aryl boronic acid
(
Scheme 1), and various mechanisms are proposed in the
3-5
literature for this reaction.
The principles of reaction progress analysis developed by
Blackmond proved to be invaluable in understanding the
catalytic cycle.
6,7
Questions To Consider When Performing Pd Cycle
Chemistry.
a. Is the catalyst stable?
b. Is the catalyst being poisoned during the course of
reaction?
c. Is there substrate/product inhibition observed by catalyst?
d. At what stage in the catalytic cycle is the impurity
formed?
e. Does the kinetic profile fit the mechanism; do we
understand the mechanism?
Introduction
The palladium-mediated cross coupling of aryl or vinyl
halides/triflates with aryl boronic acid (Suzuki-Miyaura reac-
tion) has been extensively applied in recent years and is the
method of choice for the synthesis of unsymmetrical biaryl
1
systems. Invariably the Suzuki reaction also furnishes low
We demonstrate here that by following the methodology
described by Blackmond, one can answer those questions in a
limited number of experiments (typically less than 5).
levels of undesired symmetrical biaryls resulting from homo-
2
3
coupling of either the aryl halide or the aryl boronic acid, and
while this is generally of little concern, this is not the case when
one of the byproducts is a polychlorinated aromatic. The
resulting homocoupled biaryls are then, by definition, poly-
chlorinated biphenyls (PCBs) which are strictly controlled in
most countries due to both their toxicicity and environmental
impact. For example, in the United States most processing plants
must control the level of PCBs to <50 ppm, unless the facility
is a registered producer of PCBs. For these cases, controlling
the impurity formation is a major challenge, requiring a detailed
mechanistic understanding of the process.
Reaction
A catalyst-base-solvent screen had identified the optimal
reagents to effect the Suzuki-Miyaura coupling of vinyl triflate
[
A] with aryl boronic acid [B] (Scheme 1) while minimizing
8
the level of the undesired PCB. However, despite this the
amount of PCB generated was still 2-5% compared to product.
This limited the manufacturing of the current product to less
than 5 kg of product/year in order to comply with the state
environmental regulations. The main goal of the kinetic study
was to understand the characteristics of the normal reaction,
since knowledge of the desired catalytic cycle and resting states
This contribution describes the utilization of in situ collected
reaction kinetic profiles in conjunction with appropriate math-
*
Author to whom correspondence may be sent. E-mail:
sandeep.b.kedia@gsk.com.
(4) Adamo, C.; Amatore, C.; Ciofini, I.; Jutand, A.; Lakmini, H. J. Am.
Chem. Soc. 1979, 103, 7547.
(
1) (a) Miyaura, N.; Suzuki, A. Chem. ReV. 1995, 95, 2457. (b) Bellina,
F.; Carpita, A.; Rossi, R. Synthesis 2004, 2419.
(5) Smith, G.; Dezeny, G. J. Org. Chem. 1994, 59, 8151.
(6) Blackmond, D. Angew. Chem., Int. Ed. 2005, 44, 4302.
(7) Blackmond, D. J. Org. Chem. 2006, 71, 4711.
(
(
2) Tsou, T.; Kochi, J. J. Am. Chem. Soc. 1979, 103, 7547.
3) Moreno-Manas, M.; Perez, M.; Pleixats, R. J. Org. Chem. 1996, 61,
2
346.
(8) Deschamps, N.; Elitzin, V.; Tabet, E. GSK unpublished data.
4
20
•
Vol. 13, No. 3, 2009 / Organic Process Research & Development
10.1021/op800205x CCC: $40.75  2009 American Chemical Society
Published on Web 01/20/2009

of intermediate species is central to understanding where
impurity generation occurs, and hence its subsequent control.
Rate Expression for Suzuki-Miyaura Coupling. Litera-
[AB*] ) K [A*][B]
(2)
(3)
2
[
A*] ) K [A][*]
1 ActiVe
1
ture precedent indicates a catalytic cycle with two intermediate
species is most likely for the Suzuki-Miyaura reaction, as
depicted in Scheme 2. The right-hand cycle depicts the proposed
Substitution of eqs 2 and 3 into the initial rate expression (eq
) gives the following rate equation:
1
Scheme 2. Postulated mechanism for the catalytic cycle for
desired product formation (Ar-R)
rate[AB] ) K K k [A][B][*]
Active
(4)
1
2 3
From mole balance for catalyst (Pd):
*]Total ) [*]Active + [A*] + [AB*]
[
(5)
Substitution of eqs 2 and 3 into eq 5 and rearrangement yields:
[
*]Total
[*]Active
)
(6)
1
+ K [A] + K K [A][B]
1 1 2
Substituting eq 6 into eq 4 gives the desired overall rate
expression for the catalytic cycle as:
K K k [A][B][*]
Total
1
2 3
rate[AB] )
(7)
1
+ K [A] + K K [A][B]
1 1 2
3
competing catalytic process for PCB generation. The rate
9
where K ) k/k
-i
equation for a two-intermediate catalytic process can be derived
as follows from consideration of the fundamental reaction steps:
. Activation of Pd-sites with triphenyl phosphine ligand [L],
assumed to be very fast reaction:
i
i
1
Experimental Procedures
In the current example, all kinetic studies were performed
10
using an iChemExplorer equipped for heating and stirring.
Use of iChemExplorer allowed reactions to be performed
directly in the HPLC vial, with the convenience of rapid direct
injection HPLC sampling and automated profile generation.
Thus, by using a small quantity of substrate (∼100 mg), one
can perform profiling very rapidly, thereby gaining insight into
the reaction mechanism.
In a 2-mL HPLC vial, 1 mL of toluene was added along
with vinyl triflate (1 equiv), boronic acid (1 equiv + excess
defined in Table 1), Hunig’s base (1.6 equiv), water (15 equiv),
triphenyl phosphine (0.3 or 0.6 equiv, see Table 1) and
palladium acetate (0.1 or 0.2 equiv, see Table 1). A magnetic
[
Pd(0)] + 2[L] f [*]
2
3
4
. Oxidative addition of Ar-triflate [A] with active Pd*:
k1
[
*] + [A] { } [A*]
k-1
. Transmetalation; insertion of boronic acid into Pd-catalyst:
k2
[
A * ] + [B] { } [AB*]
k-2
Table 1. Same excess and different excess experiments
. Reductive elimination; product formation and regeneration
of active Pd-catalyst:
boronic acid
(mol/L)
triflate
(mol/L)
Pd(OAc)
(mol/L)
2
excess
(mol/L)
exp
1
2
3
4
5
6
0.059
0.059
0.052
0.118
0.047
0.044
0.030
0.030
0.022
0.030
0.018
0.030
0.006
0.003
0.003
0.003
0.006
0.006
0.030
0.030
0.030
0.089
0.030
0.015
k3
[
AB*] 98 [AB] + [*]
The rate of formation of the desired product is given by:
rate[AB] ) k [AB*]
(1)
stirrer was inserted, and the vial was crimped to minimize
solvent loss from sampling. The vial was then inserted into the
iChemExplorer block which was then heated to 70 °C and
stirred. Samples were collected at 3-min intervals. The overall
3
The quasi-equilibrium assumption is used to derive an expres-
sion for the concentration of the bound species:
(
9) Rawlings, J.; Ekerdt, J. Chemical Reactor Analysis and Design
Fundamentals; Nob Hill Publishing Co.: Madison, WI, 2002.
(10) IChemExplorer; Reaction Analytics Inc.: Wilmington, DE, 2008; http://
www.ichemexplorer.com.
Vol. 13, No. 3, 2009 / Organic Process Research & Development
•
421

rate is calculated from first principles using eq 7 (refer to
Determining the Rate-Limiting Step section below for details).
The initial goal of the experiments was to confirm a first-
order dependence upon catalyst (eq 7) and to explore if the
catalyst activity was affected by reaction progression, allowing
us to rule in or out complicating factors such as substrate
inhibition, product inhibition, and reaction induction. Once
demonstrated, a handle on the catalytic cycle and specifically
the resting states of the catalytic species could then be
determined. Overall, six experiments were conducted (Table
1), comprising some at varying catalyst load (to probe order in
catalyst), some at the same reagent excess (to probe catalyst
activity), and some at different reagent excess (to probe the
catalytic cycle).
Two sets of experiments (10 mol % and 20 mol %) were
run to identify if there was any saturation of Pd sites with R-OTf
during the onset of these reactions and identify the resting state
in each set of experiments. The calculated concentration of vinyl
triflate was within 5% of measured value at 95% confidence
level.
Normalizing the Reaction Rate: Use of the Excess
Concept. Excess [e] is the difference in molar concentration
between the two substrates for a bimolecular reaction. The
excess for a given experiment is constant, and does not change
as the reaction progresses.
Figure 1. Rate vs concentration of vinyl triflate for same excess
experiments 1 and 5.
tion (since [catalyst] varied across the experiments) and plot it
versus [A], we now see overlay for all of the same excess
experiments 1, 2, 3, and 5 (Figure 2).
[
e] ) [B] - [A] ) [B] - [A] w [B] ) [e] + [A] (8)
i i f f t t
[
A] ) vinyl triflate; [B] ) boronic acid; excess ) [e] )
B] - [A];
i ) initial condition, f ) final condition;
t ) at time point t
[
Use of the catalytic rate expression (eq 7) to describe the
kinetics for a given reaction assumes that catalyst activity is
not affected by variation in reagent concentrations. In practice
it is not uncommon for the catalyst to be activated or deactivated
by components in the reaction mixture, e.g. substrate inhibition,
product inhibition, poisoning, and catalyst induction. The first
goal of the kinetic analysis is to determine if such issues are at
play, and this is achieved by comparing two or more same
excess experiments with differing starting substrate concentrations.
If the reaction rate is now plotted against substrate concen-
tration (either boronic acid or vinyl triflate) for two same excess
experiments differing only in initial substrate concentrations,
then the plots should exactly overlay if catalyst activity is
invariant between the reactions. This has been done for
experiments 1 and 5 (Figure 1), the exact overlay showing that
catalyst activity is indeed identical for the two experiments.
Accordingly, the total concentration of active catalyst sites
is the same in both cases and is unaffected by the catalyst
history. For instance, as experiment 1 proceeds, the reactant
concentrations will eventually wane to the initial concentrations
for experiment 5. The graphical overlay plot indicates that the
catalyst activity for the two reactions remains the same (as
expected from eq 7) even though for the case of reaction 1 the
catalyst has performed many turnovers and is now in the
presence of a significant concentration of product. Similarly,
when we normalize the rate by dividing by catalyst concentra-
Figure 2. Normalized rate {rate/[cat]} versus vinyl triflate for
all six experiments.
In both cases (Figure 1 and Figure 2) we see that
different excess experiments 4 and 6 do not overlay; we
will discuss more about this in graphical methodology
below.
Determining the Rate-Limiting Step. Blackmond has
1
1
applied spreadsheet calculators to understand the kinetics
of the catalytic cycle through derivation of the underlying
reaction constants. To determine the rate-limiting step in
our reaction, rate and equilibrium constants in eq 7 were
required, and this seemed the most logical approach to
obtain these values. We accordingly applied this approach,
utilizing a spreadsheet modified from a version supplied
1
1
to us by Blackmond.
Measurements can be either differential data which is
directly related to the reaction rate (e.g., calorimetric
power) or integral data which is directly related to
concentration (e.g., IR or UV absorption). In our case we
(
11) GSK in-house training course: Blackmond, D. Kinetics of Organic
Catalytic Reactions; GSK Chemical Development: Philadelphia, PA,
U.S.A., 2007.
4
22
•
Vol. 13, No. 3, 2009 / Organic Process Research & Development

have integral data since we are measuring the area under
curve of an HPLC UV response, and since we know the
initial concentration of reactants and that reaction proceeds
to completion (with only low levels of byproduct forma-
tion) it is both simple and valid to simply scale the UV
response to concentration units for the reactant we are
tracking the disappearance of. Accordingly, the normalized
concentration of one of the substrates [A]measured, i.e.,
R-OTf is measured with respect to time using this
approach. From this, the concentration of [B] is determined
at each time point from the known excess in the experi-
ment ) [A] + [e]. Using these concentrations, the rate of
reaction is calculated at each point by using eq 7 (initially
calculated concentrations at these points will be skewed by
the pure error in the measured values, resulting in the calculated
concentrations mirroring the error in the measurements. This
is of little consequence since it is the rate and equilibrium
constant information that is of prime importance, and as
described earlier, the synchronous fitting tends to minimize the
overall error in these values.
choosing some arbitrary values for constants K
). The rate of reaction for consecutive data points (rAB,1
at t and rAB at t ) is calculated using eq 7, then [A]calc
is derived at t by eq 9, which is simply the concentration
at t minus the product of the sampling interval and the
1 2
, K , and
k
3
1
,
2
2
2
1
average rate over the sampling interval. This works well
for a high sampling frequency (as employed in these
experiments) and clearly [A]calc at t is set to the known
0
initial concentration of A.
Figure 3. Finding rate constants using LMS method by
minimizing error between measured concentration and calcu-
lated concentration of [A] using eqs 7 and 9 simultaneously.
[A]calc,2 ) [A]calc,1 - (rAB,1 + rAB,2) * (t - t ) ⁄ 2 (9)
2 1
The goal is to minimize the error between [A]measured and
A]calc (using the least mean square (LMS) method) and to fit
simultaneously the same rate and equilibrium constants (K , K
and k ) across all six experiments (same and different excess
and differing catalyst loading). For our reactions the values
obtained for the constants are listed in Table 2.
To shed some more light on the resting state of the catalyst
and intermediate species, graphical methodology was used in
conjunction with the above data.
The difference between the calculated concentration of vinyl
triflate and the measured value was within (5% at 95%
confidence level. The mean difference was 1.3% with standard
deviation of 1.9%.
[
1
2
3
Table 2. Calculated rate and equilibrium constants
Introducing the Graphical Rate Equation. Inspection of
eq 7 suggests there may be some merit in plotting the rate of
reaction divided by concentration of one of the reactants against
the concentration of the other reactant. This is an example of a
graphical rate equation, and looking for the overlay of plots
for experiments of known excess is central to the kinetic analysis
-
2
rate constants
units
errors (10 )
K
K
1
5.24
979.
0.22
L/mol
L/mol
(1.04
(20.3
(0.34
2
-
1
k
3
min
See Figure 3 for measured versus predicted concentration
of [A]. Once the three reaction constants (K , K , k ) have been
6
methodology. We will later apply this equation for reactions
1
2
3
with different excess; however, it is also instructive to apply it
now for the same excess experiments (1, 2, 3, and 5). In Figure
synchronously fitted across all six experiments (using LMS
regression), then one can confidently use eq 7 to calculate the
rate of reaction under any of the experimental conditions. It
should be noted that the biggest source of pure error is in the
measurement of the reactant concentration, with some discon-
tinuities evident in Figure 3 clearly stemming from this. It should
be remembered, however, that the reaction constants are
synchronously fitted across all the experiments, which has the
effect of minimizing the overall pure error in the calculated
reaction constants (since the residuals should, on average, sum
close to zero over multiple measurements). All plots related to
rate should accordingly be smooth (as seen); however, the fitted
concentration profile in Figure 3 mirrors the discontinuities
arising from experimental error in the measurements. Although
this may look surprising, it is quite expected when you
remember that [A]calc is determined at each point using the mean
calculated rate between that point and the immediately preceding
point (eq 9). This is in turn evaluated with eq 7 and the
measured values of [A] at the consecutive points. Hence, the
4, the rate of reaction divided by the concentration of boronic
Figure 4. Graphical rate equation for same excess experi-
ments.
Vol. 13, No. 3, 2009 / Organic Process Research & Development
•
423

acid ([B]) is plotted against the concentration of vinyl triflate
[A]), where it can be seen that the plots for experiments 1 and
(
5
exactly overlay, while those for experiments 2 and 3 do not.
Note, however, that the catalyst concentration for experi-
ments 2 and 3 differs from that for experiments 1 and 5. Recall
that one of the initial objectives was to prove the first-order
dependence upon catalyst given in eq 7, and we have not yet
accounted for this in the graphical rate equation depicted in
Figure 4. Instead, if we plot the rate of reaction divided by the
product of boronic acid concentration ([B]) and total catalyst
concentration ([C]), we should obtain an exact overlay of all
the curves if the catalyst dependence is first order. This is
depicted in Figure 5, where the first-order dependence on
catalyst is clearly demonstrated by the exact overlay of the plots.
Figure 6. First-order dependence upon [A] and [B] demon-
strated at low [A] concentration.
[A], it can be seen that the rate doubles with doubling the [A],
indicative of first order in [A] at concentrations below 0.005
M.
A Closer Look at the Graphical Rate Equation (for high
[A] concentration). We have so far demonstrated a robust
catalyst system with a first-order dependence upon aryl boronic
acid and vinyl triflate at low [A] concentration, but what is the
dependence on [A] above 0.005 M? To answer this question
we need to plot the rate of reaction divided by [C] against [A]
or [B] {see Figure 7}.
Figure 5. First-order dependence in catalyst is demonstrated.
It could also be concluded a first-order dependence on [B]
based upon the overlap of these plots; however, there is an issue
here. We somewhat arbitrarily chose to plot rate/[B] vs [A],
and from eq 8 we know [B] and [A] are related by the reagent
excess. Because of this relationship, if we had plotted Rate/[A]
vs [B] it would have also directly overlaid; by considering only
same excess experiments we cannot distinguish between these
scenariossthis is somewhat akin to trying to solve simultaneous
equations in three unknowns with only two equations. To solve
the problem we need a third equation, and that equation is a
different excess experiment.
Figure 7. Normalized rate/[C] vs [A].
Probing the Catalytic Cycle: the Different Excess Experi-
ment. From the overlay of the same excess experiments we
can confidently conclude that the catalyst is robust with the
kinetics following the expected rate equation (eq 7). We have
also seen that there is a first-order dependence upon [A] or [B]
but that we cannot distinguish between these cases. In Figure
From Figure 7, it can be seen that for high12 [A] (greater
than 0.02 M), rate divided by [C] is nearly constant, indicative
of zero order in both [A]- and [B]-saturated regions. Whereas,
(
12) At high [A], the [B] is necessarily high since it is in excess.
(13) For LC-MS analysis, the catalyst was forced to sit entirely at (I)
by omitting aryl boronic acid from the reaction mixture (as
described in the text). HPLC analysis shows complete conversion
to (I) which when analyzed by ESI LC-MS furnished a peak with
6
the rate of reaction divided by the product of [B] and [C] is
plotted against [A] for the same excess experiments (as per
Figure 5) and also for the different excess experiments (experi-
ments 4 and 6).
The exact overlay of the plots immediately tells us the
reaction is first order in [B] {at low [A]} and [C] since the rate
was normalized by the product of [B] and [C]. Recall we had
already proved the first-order dependence on [C] from the same
excess experiments; the additional overlay with a different
excess experiment provides additional confirmation. Also at low
+
MW ) 884, corresponding to [M-OTf] for the complex
3 2
RPd(PPh ) OTf. The MS also displayed the characteristic isoto-
pomer pattern for a Pd complex. The characterization of an
analogous complex by ESI LC-MS during Pd-catalyzed reductive
carbonylation has recently been reported, the intermediate also
ionizing as [M-X] in this instance (ref 16). Such complexes clearly
have significant aqueous stability and survive ESI LC-MS condi-
tions. This is perhaps unsurprising, since for the case of Suzuki
reaction, water often accelerates the reaction (as is the case for
our reaction), and use of water as the primary solvent is often
employed in the literature for this type of coupling (ref 17).
4
24
•
Vol. 13, No. 3, 2009 / Organic Process Research & Development

at low [A] {concentration less than 0.005 M}, the normalized
rate doubles with doubling the concentration of [A] or [B],
indicative of first order in [A] and [B] for low [A], under-
saturated. The region between 0.005 and 0.02 M is where the
mechanism transitions from the resting state of the catalyst being
at [II] to being partitioned between [I] and [III], see Figures 8
and 9. Although not illustrated, the same conclusion can be
drawn for B when plotting rate/[C] vs [B].
2
rests at both the Pd(0)L stage (III) and at (I). This is
crucial information for understanding control of homo-
coupling, since we can now postulate the mechanism for
formation of PCB where the intermediate (III) is common
to both the normal reaction and impurity generation in
the catalytic cycles (see Scheme 3). In this report no
attempt has been made to investigate the kinetics of PCB
formation as this is out of the scope of this work. Here
we demonstrate that, with the kinetic profile for the main
reaction and knowing the various resting states, one can
easily delineate the plausible mechanism for PCB forma-
tion and avoid the catalyst resting at (III) to minimize
the impurity.
At the start of the reaction the concentrations of vinyl
triflate and aryl boronic acid are high, and the reaction
proceeds under saturation kinetics, the resting state of the
catalytic cycle sitting at (II). As the concentrations of
reactants fall, the catalyst becomes under-saturated, and
the catalytic cycle now partitions between both intermedi-
ates (I) and (III), providing an opportunity for effective
competing homocoupling of aryl boronic acid to occur.
This arises since a tangible concentration of intermediate
This is a demonstration of saturation kinetics, at high [A]
and [B] the catalyst is fully saturated in both A and B and the
reaction rate is constantszero order in [A] and [B]. At low
[A] and [B] the reaction is second order overallsfirst order in
both [A] and [B]. The implications for the catalytic cycle are
depicted in Figures 8 and 9.
(
III) now exists in solution to undergo the much slower
oxidative addition with the aryl boronic acid. In contrast,
when the catalyst was saturated with reactants, intermedi-
ate (III) did not reach any significant concentration since
it is consumed as fast as it is formed by rapid oxidative
addition into the vinyl triflate.
Figure 8. Catalytic cycle for saturation kinetics. Catalyst fully
saturated in vinyl triflate and aryl boronic acid, resting state is
(
II), and rate-determining step is reductive elimination step.
The increasing rate of formation of symmetrical biaryl
PCB can be seen in Figure 10 (blue curve) for batch
Suzuki-Miyaura coupling in direct response to an in-
creasing partition towards (III) as the reaction progresses.
Interestingly, real-time HPLC profiling was able to track
an intermediate species which was confirmed as the
1
3
oxidative addition intermediate (I) by LC-MS. As the
reaction approaches completion, this intermediate rapidly
disappears, commensurate with consumption of all the
vinyl triflate.
To minimize PCB formation we clearly need to force
the catalytic cycle to rest at either (I) or (II) for the
duration of the whole reaction. At low [B] we know that
transmetalation with aryl boronic acid is rate limiting, so
by removing aryl boronic acid from the reaction mixture
completely, we can force the catalytic cycle to sit at
intermediate (I), that is [I] ) [*]Total, all added palladium
exists as (I). Addition of aryl boronic acid is this manner
is atypical for Suzuki-Miyaura-type reactions, as almost
all reactions are done in batch fashion for these types of
cross-coupling reactions. Aryl boronic acid should now
be added semi-batch in a “feed controlled” manner; that
is to say as aryl boronic acid is added, it is immediately
consumed and does not accumulate in solution to any
appreciable level. This is synonymous with a rate of
addition slower than the instantaneous rate of vinyl triflate
oxidative addition. If true, this will force the catalytic cycle
to partition between (I) and (II), and we will have
effectively slowed down the transmetalation step by
limiting the reagent concentration.
Figure 9. Catalytic cycle during the end of reaction. Catalyst
is not saturated, resting states are partitioned between (III) and
(
I). Rate-determining step is addition of the boronic acid.
The above could have also been obtained by using the rate
constants. It could be easily be shown that the product of
[A][B] . 1 + K [A] in the denominator of eq 7, for
concentration greater than 0.02 M of [A]. Also the term ‘1‘
dominates (1 . K [A] + K [A][B]) in the denominator for
K K
1 2
1
1
1 2
K
concentration of [A] less than 0.005 M, hence showing how
the rate-determining step shifts during the catalytic cycle.
Impurity Reduction/Elimination. Kinetic analysis of the
normal reaction allows us to delineate the resting states
of the “on-cycle” intermediates; under saturation kinetics
the transmetallated intermediate (II) is the resting state
(
Figure 8), while under unsaturated conditions the catalyst
Vol. 13, No. 3, 2009 / Organic Process Research & Development
•
425

Scheme 3. Postulated mechanism for catalytic cycle for desired product formation on the left (Ar-R), and the impurity (PCB)
on the right
PCB Control by Semi-Batch Addition. Accordingly,
control of PCB formation was demonstrated through semi-
batch addition of aryl boronic acid, which was added as
a solution to the reaction mixture over 3-6 h. Various
temperatures were explored (60, 70, 80, and 90 °C), under
which the reaction went to completion in all cases without
significant PCB formation. Reaction profiles for the semi-
batch additions under the various temperatures can be seen
in Figures 11-14, each resulting in negligible PCB
formation.
minimal effect upon the reaction. Comparing the two
reactions side by side, one performed open to air, the other
employing rigorous degassing and nitrogen purging,
indeed furnished comparable results. Both afforded prod-
uct with <50 ppm PCB content (Figure 15).
A reaction temperature of 70 °C was chosen for
subsequent scale-up at 1 L whence product was obtained
as expected, again with negligible PCB formation (<10
ppm).
For semi-batch addition of aryl boronic acid, no aryl
boronic acid was detected by real time HPLC profiling,
confirming “feed controlled” addition rate slower than the
overall reaction rate. HPLC also detected two intermedi-
ates, the red line corresponding to the oxidative addition
intermediate (I) confirmed by LC-MS. The second
intermediate (blue line) was of fleeting existence and could
4
Some literature reports suggest oxidation of Pd(0) to
Pd(II) is required for homocoupling of the aryl boronic
acid. However, we were confident from our kinetic
analysis that the catalyst would be partitioning between
the oxidative addition intermediate (I) and transmetallated
intermediate (II), and this implied that oxygen should have
Figure 10. Batch Suzuki coupling in the presence of excess aryl boronic acid.
4
26
•
Vol. 13, No. 3, 2009 / Organic Process Research & Development

Figure 11. Semi-batch addition of boronic acid at 60 °C.
Figure 14. Aryl boronic acid addition at 90 °C.
Figure 15. HPLC profile for scale-up run on a 1-L scale.
was observed. No PCB was seen by HPLC (level of
detection is ∼100 ppm). This behavior appears to be borne
out experimentally for all the cases studied (Figures 11,
Figure 12. Aryl boronic acid addition at 70 °C.
1
2, 13, and 14).
PCB level was found to be 7.4 ppm in reaction solution.
PCB concentration (shown above with arrow) in not
detected on HPLC column, for a concentrated reaction
mixture, indicating <100 ppm PCB present in reaction
mix (LOD g 100 ppm).
Conclusions
Reaction profile analysis and application of the graphi-
cal rate equation has facilitated control of PCB generation
through understanding of the catalytic cycle. Specifically
we have:
i demonstrated the catalyst is stable and rugged from
the same excess experiments;
ii demonstrated the dependence upon aryl boronic acid
and vinyl triflate from the different excess experiments;
iii demonstrated catalyst sits at transmetallated intermedi-
ate {(II)} at high concentrations of reactants (saturation
kinetics) and rests between oxidation addition inter-
mediate {(I)} and active Pd(0) {(III)} otherwise;
iv delineated cycle must rest at oxidation additive inter-
mediate {(I)} or transmetallated intermediate {(II)} to
minimize PCB formation;
Figure 13. Aryl boronic acid addition at 80 °C.
not be confirmed by off-line LC-MS, but it is postulated
this is the transmetallated intermediate (II). This is
consistent with feed-controlled enforced partition between
(
I) and (II), and depending upon when the HPLC pulls a
sample for direct inject, one or other of these intermediates
may be predominating. As the reaction proceed, the RDS
will naturally shift towards transmetalation (as in the batch
reaction) so we would expect to mostly observe intermedi-
ate (I) in the latter stages of the addition which is what
v employed semi-batch addition of aryl boronic acid to
enforce partition between oxidation additive and trans-
metallated intermediates {(I) and (II), respectively};
Vol. 13, No. 3, 2009 / Organic Process Research & Development
•
427

vi successfully reduced PCB levels to <50 ppm;
with their support in the kinetics work. We also thank Robert
Yule for his support to the kinetics team; John Grimes, David
Black, and Kim Harvey for their valuable support in identifying
the intermediate structures and for analytical support throughout
the course of this work.
1
4,15
vii demonstrated that oxygen sensitivity
was not the
cause of PCB formation in this example.
Acknowledgment
We thank the project leader, Matthew J. Sharp; process
chemists, Vassil Elitzin, Elie Tabet, and Nicole Deschamps,
for their valuable contributions that aided in understanding this
catalytic system, for their postulation of the mechanism, along
Note Added after ASAP: The errors in Table 1 have
been corrected for the version published on the Internet April
, 2009.
7
(
(
(
(
14) Coifini, I. J. Am. Chem. Soc. 2006, 128, 6829
15) Smith, M. AdV. Synth. Catal. 2005, 347, 647
16) Beller, M. J. Am. Chem. Soc. 2008, 130, 15549
17) See for example: Moore, L.; Shaughnessy, K. Org. Lett. 2004, 6, 225
.
.
Received for review August 22, 2008.
OP800205X
.
.
4
28
•
Vol. 13, No. 3, 2009 / Organic Process Research & Development