Organic Process Research & Development
Article
conversion to 3 (as measured by HPLC relative to an internal
standard) and increasing strength of the acid additive (see
Table 1).
the desired product in reasonable yield, that we turned our
focus to increased reaction productivity or throughput, central
to this would be a move to a flow photochemistry platform.
Development of the Photoredox Minisci Reaction in
Continuous Flow. The initial results in a batch photo-
chemical reactor were encouraging, but we were acutely aware
that the scale-up of photocatalytic reactions presents unique
Table 1. Effect of Acid pK on Conversion of 6 to 3 Using
a
18,31,32
challenges.
light penetration into the reaction solution is limited by the
necessarily) high molar extinction coefficient of the photo-
catalyst. Therefore, penetration beyond the first few milli-
Because of the Beer−Lambert−Bouguer law,
(
18
meters of the reaction mixture is often not possible. For this
reason, chemists have generally turned to flow reactor design
to provide the required short pathlength for irradiation of the
bulk of the solution, while maintaining the ability to process
pK of acid (in
product/internal standard (HPLC
area counts)
a
2
8,29
17−19,32−41
acid additive
water)
large quantities of the material.
trifluoroacetic
0.23
1.35
To render the current photo-Minisci reaction viable for the
manufacture of ceralasertib on a required scale of greater than
acid
oxalic acid
pyruvic acid
malonic acid
1.23
2.39
2.83
1.35
1.61
1.93
1
00 kg, we estimated that a minimum productivity of 5 kg/day
of pyrimidine 3 would be required. Thus, although our initial
focus was on reaction yield, we now focused on increasing the
reaction rate to attain shorter residence times and thus higher
productivity for a continuous flow process. We investigated our
highest yielding batch conditions using DIPEA as an additive
and 4CzIPN as a photocatalyst in DMSO using a Vapourtec
UV-150 photochemical reactor with blue LEDs (450 nm). We
were delighted to observe a 50% solution yield of product 3
with a residence time of 50 min which enabled complete
consumption of redox active ester 6. For comparison, the use
of malonic acid as an additive required a longer residence time
of 100 min to achieve complete consumption of 6 and a similar
solution yield. Unfortunately, we also observed very significant
darkening of the reaction mixture. We believed that this
darkening would lead to diminishing returns for increases in
residence time and to test this hypothesis, and to begin to
improve our understanding of the reaction, we carried out
some LED-NMR experiments.
Next, we conducted an experiment with triethylamine
hydrochloride as an acidic additive, showing that the current
Minisci reaction could proceed with a very weak acid.
Remarkably, the use of the ammonium salt produced a higher
solution yield using the EvoluChem photochemical reactor
(Table 2). Indeed, we had doubted that the role of acid in our
Table 2. Investigation of Alternative Additives
LED-NMR Reaction Profiling. LED-NMR is a data-rich
technique for the analysis of photocatalytic reactions in situ.
a
42
entry
additive
NMR solution yield (%)
1
2
3
4
malonic acid (2.0 equiv)
triethylamine hydrochloride (2.0 equiv)
no additive
23
40
33
50
This technique enabled us to gain detailed information on the
reaction rate and profile (Figure 1), as well as highlight the
formation of reaction byproducts over the course of the
reaction.
Our reaction profiling confirmed the increased rate of the
reaction when DIPEA (0.2 equiv) is used as an additive
N,N-diisopropylethylamine (0.2 equiv)
a
Yields based on proton NMR of the crude reaction mixture,
quantified against a known amount of an external standard (TCNB).
(
Figure 1). We were able to improve yield using an excess (3
reaction was to protonate the heterocycle as proposed in the
general mechanism of Sherwood, as our substrate, 2,4-
dichloropyrimidine, is a very weak base indeed, with a
equiv) of the cheap and readily available reagent 2,4-
dichloropyrimidine. However, it also became apparent that
the product 3 is oxidized to sulfoxide 4 over extended time
periods. Notably, oxidation to sulfoxide 4 was only observed
when the redox active ester 6 was almost or fully consumed.
This would be consistent with an alternative photoredox
catalytic cycle becoming operative only when the redox active
ester was consumed. In theory, the formation of sulfoxide 4
could be controlled by identifying the optimal flow rate, but we
were interested in further improving the rate of the reaction to
improve throughput.
22,30
calculated pK of −2.84.
We therefore performed an
a
experiment with no acid additive present and surprisingly
found that the desired product was formed in higher yield than
with any of the carboxylic acid additives present. For this
system, it appears that we have a combination of a particularly
nucleophilic radical and an electrophilic arene, allowing for
radical addition to occur in the absence of activation.
Realizing that an acid additive was not required, we decided
to investigate other additives, primarily DIPEA. Interestingly,
we observed an improved solution yield of 50% after 21 h
using catalytic DIPEA as an additive (Table 2). The
mechanistic effects of DIPEA were unknown at this point
but we decided to proceed to scale-up. It was at this point in
development, with reaction conditions in hand that produced
We next carried out a series of “same-excess”-type
experiments in an attempt to increase our understanding of
the reaction under our best conditions using DIPEA (0.2
4
3
equiv) as an additive. The “same-excess” experiment
demonstrated that the rate of the reaction at 50% consumption
of 6 is not equivalent to the rate of the reaction when the
C
Org. Process Res. Dev. XXXX, XXX, XXX−XXX