Batch versus flow photochemistry: A revealing comparison of yield and productivity

Article abstract of DOI:10.1002/chem.201404347

The use of flow photochemistry and its apparent superiority over batch has been reported by a number of groups in recent years. To rigorously determine whether flow does indeed have an advantage over batch, a broad range of synthetic photochemical transformations were optimized in both reactor modes and their yields and productivities compared. Surprisingly, yields were essentially identical in all comparative cases. Even more revealing was the observation that the productivity of flow reactors varied very little to that of their batch counterparts when the key reaction parameters were matched. Those with a single layer of fluorinated ethylene propylene (FEP) had an average productivity 20% lower than that of batch, whereas three-layer reactors were 20% more productive. Finally, the utility of flow chemistry was demonstrated in the scale-up of the ringopening reaction of a potentially explosive [1.1.1] propellane with butane-2,3-dione.

Full text of DOI:10.1002/chem.201404347

DOI: 10.1002/chem.201404347
Full Paper
&
Flow Photochemistry
Batch versus Flow Photochemistry:
A Revealing Comparison of Yield and Productivity
Luke D. Elliott,[a] Jonathan P. Knowles,[a] Paul J. Koovits,[a] Katie G. Maskill,[a] Michael J. Ralph,[a]
Guillaume Lejeune,[a] Lee J. Edwards,[b] Richard I. Robinson,[b] Ian R. Clemens,[b] Brian Cox,[b]
David D. Pascoe,[b] Guido Koch,[c] Martin Eberle,[c] Malcolm B. Berry,[d] and Kevin I. Booker-
Milburn*[a]
Abstract: The use of flow photochemistry and its apparent
superiority over batch has been reported by a number of
groups in recent years. To rigorously determine whether
flow does indeed have an advantage over batch, a broad
range of synthetic photochemical transformations were opti-
mized in both reactor modes and their yields and productivi-
ties compared. Surprisingly, yields were essentially identical
in all comparative cases. Even more revealing was the obser-
vation that the productivity of flow reactors varied very little
to that of their batch counterparts when the key reaction
parameters were matched. Those with a single layer of fluo-
rinated ethylene propylene (FEP) had an average productivi-
ty 20% lower than that of batch, whereas three-layer reac-
tors were 20% more productive. Finally, the utility of flow
chemistry was demonstrated in the scale-up of the ring-
opening reaction of a potentially explosive [1.1.1] propellane
with butane-2,3-dione.
Introduction
straints, resulting in the production of 20–500 g of photochem-
ical products in 24 h. Precise regulation of flow rate enabled
residence time to be controlled in the cases when over-irradia-
tion of evolving product is a problem.[14–16] Recently, a second-
generation reactor has enabled use of air-cooled low-pressure
36 W lamps in conjunction with FEP tubing,[16–18] with the ad-
vantage of allowing controlled irradiation at selected wave-
lengths centered around 254, 312 and 365 nm by simply ex-
changing lamps (Figure 1b).
In the last ten years, the use of flow chemistry in organic syn-
thesis has increased dramatically and is now rapidly being es-
tablished as a routine tool for mainstream synthesis.[1–8] Flow
techniques have also been applied to synthetic organic photo-
chemistry on a variety of platforms.[9,10] In 2005, Booker-Mil-
burn and Berry reported a practical flow reactor for general
lab-based synthetic photochemistry.[11–13] This employed a reac-
tor of flexible, UV-transparent fluorinated ethylene propylene
(FEP) tubing wrapped closely around a UV-emitting source
(Figure 1a). The surface area and proximity of the photolysate
to the UV source ensured effective irradiation of large volumes
of solution minimizing transmission versus distance con-
The FEP reactor concept has now been used and adapted
for a number of reactions, resulting in the synthesis of a wide
range of complex, photochemically derived products.[19–28] Per-
haps the most striking of these has been the use of an FEP
flow reactor in the scaled-up continuous synthesis of the front-
line anti-malarial drug artemisinin described by Seeberger and
co-workers.[29,30]
[a] Dr. L. D. Elliott, Dr. J. P. Knowles, Dr. P. J. Koovits, K. G. Maskill, M. J. Ralph,
Dr. G. Lejeune, Prof. K. I. Booker-Milburn
Many groups have now reported the advantages of flow
photochemistry over previously reported batch results.[9,10]
Whilst it is true that scale-up has now been clearly demonstrat-
ed in flow, and there is a real safety advantage when irradiat-
ing large volumes of flammable solvents, flow photochemistry
involves a considerable equipment investment compared to
the simplicity of a conventional batch reactor. On scale-up, the
increased path length of larger batch reactors theoretically
suffer from the constraints imposed by the Beer–Lambert law
(e.g., longer reaction times/photo-degradation), nevertheless,
huge photochemical industrial processes have been developed
in batch. For example, the Toray process for the synthesis of
cyclohexanone oxime (Nylon 6 production) is a semi-continu-
ous photochemical batch process that in 1976 had a worldwide
production capacity of 160000 tonnespa.[31]
School of Chemistry, University of Bristol
Cantock’s Close, Bristol, BS8 1TS (UK)
E-mail: k.booker-milburn@bristol.ac.uk
[b] L. J. Edwards, R. I. Robinson, I. R. Clemens, B. Cox, D. D. Pascoe
Novartis Institutes for Biomedical Research
Wimblehurst Road, Horsham, West Sussex, RH12 4AB (UK)
[c] G. Koch, M. Eberle
Novartis Pharma AG
Novartis Institutes for Biomedical Research, 4002 Basel (Switzerland)
[d] M. B. Berry
GlaxoSmithKline, Gunnels Wood Road, Stevenage, Herts, SG1 2NY (UK)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/chem.201404347. It contains a comprehensive ac-
count of all experimental details, including supplementary figures and
tables, full experimental procedures, analytical data, as well as copies of H
and 13C NMR data.
1
Chem. Eur. J. 2014, 20, 1 – 8
1
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
&
These are not the final page numbers! ÞÞ

Full Paper
tubing, gaps between the tubing and absorption of light by
the FEP. To enable fair comparison between reactors, all reac-
tions were performed keeping parameters of concentration,
lamp power and distance of photolysate from the UV source
as identical as possible. Although it is commonplace in flow
chemistry to use residence times (i.e., reactor volume/flow
rate) for comparison with batch reaction times, in photochem-
istry, this often leads to incorrect conclusions (see below) re-
garding various reaction outcomes.[9] Consequently, both
modes were optimized independently under the above-de-
scribed reaction parameters, and their yields, and critically, pro-
ductivities compared at the end-time point obtained for the
batch process.
Table 1 summarizes the extensive data collected on twelve
different cycloaddition and rearrangement reactions. Perhaps
the most strikingly obvious feature is the similarity in isolated
yields between batch and flow, which in most cases are essen-
tially identical. After extensive optimization of each reaction, it
also became clear that the productivity of the flow reactor dif-
fered to that of the batch reactor by a relatively small amount.
Similar to previous studies,[11] the number of layers of FEP had
a significant effect. For example, in entry 1, Table 1, although
the yields of 3[32] were the same, the productivity was 50%
higher for a three-layer compared to one-layer system. This
was due to the more effective capture of light in the three-
layer system, enabling a faster flow rate (3 mLminÀ1) and the
isolation of 62.7 g of 3 in a single 22 h run. Conversely, it was
clear that the one-layer system was less effective at capturing
light thus reflecting its lower productivity compared to batch
(5.70 vs. 6.56 g in 3 h). For this reason, all subsequent reactions
studied with a 400 W lamp utilized the three-layer system. The
other maleimide cycloadditions (entries 2-4)[11,32–33] proceeded
in a similar fashion, and the desired [2+2] cycloaddition ad-
ducts were obtained in near identical yields for both batch
and flow reactions but with slightly greater productivity levels
for flow.
Figure 1. Batch and flow reactors used in this study. (a) Comparison of the
footprints of a 400 W medium-pressure Hg lamp batch reactor and power
supply (left) with equivalent three-layer FEP flow reactor and peristaltic
pump (right). (b) A triad of one-layer FEP reactors customized for use with
40 cm 36 W low-pressure Hg lamps. This example depicts the use of 254 nm
(UVC) lamps, in which three-reactors have been daisy chained together for
increased productivity (see Table 1, entry 7).
Entry 5 proved to be very interesting. It was found that the
absorption of amino maleimide 10 is redshifted with respect
to the parent maleimide, similar to the methoxy maleimides
we have previously studied.[34,35] This amino maleimide chro-
mophore had a lmax centered at 350 nm, ideally suited to irra-
diation with 36 W UVA lamps (Figure 2). Under flow conditions,
this intramolecular [2+2] cycloaddition was an exceptional per-
former with a productivity of 12.69 g/100 min at 80% isolated
yield. The absorption was so strong, and the efficiency of the
reaction so high, that it could be run at a flow rate of
16 mLminÀ1 using two 36 W lamps in series, enabling the pro-
duction of 40 g of tricyclic product 11 in a 5 h run. However,
when it came to comparing the batch process by using the
same low-pressure 36 W lamp, we were faced with the prob-
lem that the lamp was too long to be fully immersed in
a 400 mL batch immersion well, and only half of the lamp
could be utilized. When this batch irradiation was carried out
in this way, it gave an optimized 77% isolated yield of the tri-
cyclic product at a productivity of 3.11 g/100 min. On face
value, this appears to be a quarter as productive as the flow
result for the same time period. However, when the four-fold
Faced with this conflicting picture, the following question
arises: is it possible that batch limitations in photochemistry
are not as great as feared and the superiority of flow is yet to
be proven unequivocally? With many readily available bespoke
and commercially available reactor choices, today’s first-time
photochemistry user may well ask the question: which is
better, batch or flow? With this in mind, we embarked on an
investigation and careful comparison of batch versus flow-reac-
tor performance for a variety of different photochemical reac-
tions. This study would prove for the first time whether photo-
chemical flow reactors are more or less productive under the
same time period of an optimized batch process when both
are exposed to the same amount of UV light.
Results and Discussion
Attempting to compare batch and flow photo-reactors is not
straightforward, because encapsulating the light efficiently in
flow is more challenging due to the non-planar surface of the
&
&
Chem. Eur. J. 2014, 20, 1 – 8
www.chemeurj.org
2
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Paper
Table 1. Batch versus flow yields and productivities for a variety of photochemical reactions.
Entry
Lamp
Mode
Conditions[a]
Yield [%]
Productivity
at batch end [g]
Batch vs.
flow[g]
400 W
400 W
400 W
batch
flow[b]
flow[c]
0.096 m, 180 min
68
68
68
6.56
8.55
5.70
1/1.30
1.15/1
1
0.096 m, 3 mLminÀ1
0.096 m, 2 mLminÀ1
400 W
400 W
batch
flow[b]
0.1 m,120 min
67
68
4.77
5.90
0.1 m, 4 mLminÀ1
2
3
1/1.24
1/1.12
400 W
400 W
batch
flow[b]
0.1 m, 80 min
0.1 m, 6 mLminÀ1
56
52
4.32
4.82
400 W
400 W
batch
flow[b]
0.1 m, 120 min
65
64
4.0
4.70
4
5
1/1.18
1/1.02
0.1 m, 4 mLminÀ1
1ꢁ36 W UVA
2ꢁ36 W UVA
batch
flow[c]
0.05 m, 100 min
0.055 m, 16 mLminÀ1
77
80
3.11 [12.44][f]
12.69
1ꢁ36 W UVA
1ꢁ36 W UVA
batch
flow[c]
0.1 m, 105 min
0.1 m, 7.5 mLminÀ1
80
77
7.09 [13.10][f]
10.70
6
1.22/1
1ꢁ36 W UVC
1ꢁ36 W UVC
3ꢁ36 W UVC
batch
flow[c]
flow[c]
0.02 m, 7 h
84
75
72
2.10 [4.20][f]
3.54
1.19/1
1.25/1
0.02 m, 2 mLminÀ1
0.03 m, 4 mLminÀ1
7
8
9
10.06
400 W
400 W
batch
flow[b]
0.3 m, 3 h
3 mLminÀ1
67
72
14.05
20.52
1/1.46
400 W
400 W
400 W
400 W
batch
flow[b]
batch
flow[b]
0.1 m, 4 h
63
63
76
77
4.88
5.87
1.18
1.43
1/1.20
1/1.21
0.1 m, 2 mLminÀ1
0.02 m, 2 h
0.02 m, 4 mLminÀ1
400 W
400 W
batch
flow[b]
0.1 m, 100 min
62
60
5.82
5.90
0.1 m, 4.2 mLminÀ1
10
1/1.02
1/1.10
400 W
400 W
batch
flow[b]
0.6 m, 135 min
0.6 m, 3.3 mLminÀ1
90
89
30.70
33.83
11[d]
1ꢁ36 W UVC
1ꢁ36 W UVC
batch
flow[c]
0.02 m, 310 min
0.02 m, 3.1 mLminÀ1
49
46
0.80 [2.0][f]
1.86
12
1.08/1
[a] All reactions run in degassed solvents. [b] Three layers of FEP, flow direction outer layer to inner. [c] One layer of FEP. [d] Acetophenone sensitizer
(50 mol%). [f] Parentheses indicate power corrected value for length of the lamp effectively covered (see the Supporting Information for detail). [g] Ratio
of productivities; all products are racemic.
power difference is factored in the results between batch and
flow are remarkably similar at the 100 min batch end point.
With the 36 W low-pressure lamps, a three-layer version was
not investigated, because the longer length of the lamps
(40 cm) would require a large amount of FEP, leading to an ex-
cessive reactor volume (ca. 300 mL).
Chem. Eur. J. 2014, 20, 1 – 8
www.chemeurj.org
3
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
&
These are not the final page numbers! ÞÞ

Full Paper
azepines (entries 9 and 10).[42,43] With dimethylmaleimide 18,
the reactions were run at two different concentrations, the
lower concentrations enabling a higher conversion to be ob-
tained in a shorter reaction time. For this reaction, it was re-
markable to observe identical yields for batch and flow, even
with a ten-fold difference in concentration. However, in both
cases, flow conditions proved to be approximately 20% more
productive than batch.
The [5+2] cycloaddition of dichloromaleimide 20 proved to
be a very interesting example (entry 10). Previously in batch,
this reaction had proved to be very sensitive using a 125 W
lamp at 0.02m, and the window between maximum conver-
sion of starting material and over-irradiation of the sensitive
product was very narrow.[9,14] Consequently, we had previously
never been able to perform this reaction on scales much great-
er than 0.5 g by using batch apparatus. So it was to our sur-
prise that when this reaction was performed at the much
higher concentration of 0.1m in batch (400 mL) with a 400 W
lamp, we were able to follow the reaction over time without
observing any noticeable product degradation (Figure 3).
Figure 2. UV absorption of amino maleimide 10 overlaid with emission of
36 W UVA low-pressure lamp.
Entry 6 describes the known[36–38] intramolecular cycloaddi-
tion of dieneone 12 to caged diketone 13 with a single 36 W
UVA lamp in flow. The batch reaction was conducted by using
approximately 50% immersion of the lamp. Optimized yields
were again very similar, and it was found that the batch reac-
tion was approximately 20% more productive once the power
difference had been accounted for. Interestingly, unlike the
flow reactor, for this example, the batch reactor suffered from
foul-up over-time and required cleaning (base bath) between
runs. Recently, Aillet et al. described a chemical engineering
analysis of this reaction in a microcapillary flow reactor and
also concluded that in this case, productivity was higher in
batch.[39]
The intramolecular [2+2] cycloaddition of pyrrole 14
(entry 7, Table 1) provided an example of the use of a 36 W
UVC lamp at 254 nm, which was also conducted with 50%
lamp immersion.[40] Once power correction had been applied,
this also displayed the same trend as entry 6, Table 1. Although
the yield and power corrected productivity in batch was
higher, by using a three-lamp system in flow, we were able to
produce 29.2 g of 15 in a single 21 h run (cf. 10.06 g/7 h). To
process the same 5 L solution of substrate 14 in batch would
be impractical. The cycloaddition was also carried out with
a 125 W medium-pressure lamp through a quartz filter (see
Figure S14 in the Supporting Information; 69%, 0.57 g/2 h).
Over the course of 7 h, this lamp could only produce about
2 g of 15, the same as half a 36 W UVC lamp. This clearly dem-
onstrates the superior efficiency of the low-pressure UVC emis-
sion over that of a medium-pressure lamp.
Figure 3. Batch [5+2] photocycloaddition of dichloromaleimide 20 moni-
tored over time by using H NMR technique with 1,3,5-trimethoxy benzene
as internal standard.
1
In batch, azepine formation progressed rapidly over the first
100 min or so to give an optimal conversion in the region of
60–70% (by NMR spectroscopy). Cessation of irradiation at this
point gave a 62% isolated yield of pure azepine 21 with a pro-
ductivity of 5.82 g/100 min. We found that the batch reaction
could be irradiated for a total of 210 min allowing the isolation
of product in a record yield of 69%, albeit at a lower produc-
tivity. This product was also less pure; a tell-tale red coloura-
tion indicating photodegradation from over-irradiation.
The Paternꢂ–Bꢃchi reaction of benzaldehyde and dihydrofur-
an 16 (entry 8) had previously been optimized in a detailed
study by Griesbeck et al. by using a falling film reactor with
a high power 3 kW XeCl excimer lamp (75%, 130 g, 14 h).[41] In
our hands, scale-up under batch conditions at 0.3m concentra-
tion gave a 67% isolated yield (14.05 g, 3 h) of oxetane 17 by
using a 400 W lamp. Under flow conditions, an optimized 72%
yield was obtained with a productivity of 20.52 g at the batch
end point of 3 h. This represented an almost 50% increase in
productivity for the flow reactor in this particular case.
Two factors are likely at play here: firstly, at this higher con-
centration, the sensitive product is more effectively “screened”
by the starting maleimide. Secondly, because the irradiation
time is longer at higher concentration, the reaction was much
easier to monitor by NMR spectroscopy, and therefore, the
point between maximum conversion and over-irradiation
Moving away from four-membered ring formation, we next
studied the performance of the maleimide chromophore in the
previously developed intramolecular [5+2] cycloaddition to
&
&
Chem. Eur. J. 2014, 20, 1 – 8
www.chemeurj.org
4
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Paper
could be more accurately determined compared to our earlier
studies. This example clearly illustrates the importance of
matching substrate concentration with photon flux (lamp
power) in batch reactions, that is, too much power risks prod-
uct degradation and too little leads to slow and incomplete
conversion. Performing the reaction in flow using a three-layer
system (0.1m, 4.2 mLminÀ1) resulted in essentially identical
yield and productivity to batch (5.90 g/100 min, 60% yield).
Entry 11 describes an acetophenone-sensitized di-p-methane
rearrangement first reported by Edman.[44,45] We found this re-
action to be exceptionally efficient both in flow and in batch
and actually struggled to identify its limits. For example, this
was run at a very high concentration of 0.6m to give almost
quantitative yields in both batch and flow, with the latter prov-
ing marginally more productive at the batch end time.
Table 2. Comparison of residence time versus productivity in three differ-
ent reactors (n/a=not applicable).
Reactor Type Volume Flow rate
Residence
time [min]
Productivity Yield
[mL]
[mLminÀ1
]
[ghÀ1
]
[%]
1
2
batch 418
n/a
3
180
48
2.20
2.85
68
68
3-
145
layers
1-
3
45
2
23
1.90
68
layer
Entry 12 describes a reaction previously developed by us in-
volving the irradiation of pyrrole 24 to tricyclic aziridine 25
through a complex photochemical two-step process.[17] Consis-
tent with the rest of the 1-layer/36 W reactors used in this
study, the yields between batch and flow were essentially iden-
tical, with a minor increase in batch productivity when the
power correction was factored in.
this approach to be highly misleading, because it is immedi-
ately apparent that the three-layer reactor is superior for prep-
arative chemistry (Table 2). For example, running the reaction
in Reactor 2 for 22 h gave 63 g of isolated product 3. In the
same time period, Reactor 3 would give approximately 40 g of
product.
Finally, we studied the scale-up of the photochemical ring
opening of propellane 26 with butane-2,3-dione to diketone
28, which was previously described in batch (58%, 26.5 g,
8 h).[46,47] After optimization in flow, this reaction proceeded in
an identical 58% yield, but with greatly increased productivity
to deliver 51.8 g of pure 28 in a single 70 min run (projected
1.07 kg/24 h). Due to the potentially explosive nature (Fig-
ure S19 in the Supporting Information) of 26, we declined to
carry out a comparative batch reaction on the same scale. The
ability to safely irradiate large quantities (1400 mL) of the po-
tentially hazardous solution of 26 through a relatively small re-
actor (53 mL volume) represents a significant safety advance
for this useful reaction (Scheme 1).
This highly rewarding study has uncovered a number of fea-
tures of both batch and flow photochemistry for a wide variety
of reactions. When comparing both modes at the same photo-
lysate concentrations, UV lamp powers and wavelengths, three
important and general trends are apparent:
1) Yields for batch and flow reactors in synthetic photochem-
istry are essentially the same at full conversion.
2) Three-layer FEP reactors have on average 20% higher pro-
ductivity compared to the same batch end point.
3) One-layer FEP reactors have on average 20% lower produc-
tivity compared to the same batch end point.
It is straightforward to understand the lower productivity of
a one-layer FEP reactor compared to batch due to the shorter
path lengths, gaps between channels and the absorption of
light by FEP. For example, in scenario A (Figure 4), light enter-
ing a tube would encounter some absorption by FEP polymer,
before either exciting a substrate molecule or leaving the FEP
tube. In scenario B, light could simply pass through the gaps
between the FEP tubing. Overall, this would result in less effi-
cient capture of photons in the one-layer reactor compared to
batch. Moving to the three-layer system, this reactor consis-
tently exceeded the productivity of the batch reactor by an
average of 20%. This enhancement over the one-layer system
reflects increased residence time due to the larger reactor
volume and the resultant improvement in the capture of pho-
tons previously lost through gaps in the channels (scenario B).
This leads to an increase in the effective path length, because
any light leaving the inner layer would be able to penetrate
the second and third layers. It is also interesting to compare
the path length of the batch reactor (8.7 mm) with the three-
layer flow reactor. Although it is tempting to suggest that the
flow reactor path length is similar (3ꢁ2.7=8.1 mm) to batch,
Scheme 1. Photochemical ring opening and reaction of propellane 26 with
butane-2,3-dione in flow.
As was mentioned previously, residence times are commonly
used in flow chemistry to assess the efficiency of a reactor. In
our own experience with flow photochemistry, we have found
residence time to be a less relevant parameter than productivi-
ty, especially when comparing batch and flow. For example,
the data collected for the cycloaddition of 1 to 3, can be used
to compare residence time and productivity as metrics of the
efficiency of a reactor. On residence time alone, the Reactor 3
would appear to be superior, that is, two times faster than Re-
actor 2, and eight faster than Reactor 1 in achieving the same
maximum 68% yield. However, analysis of productivities shows
Chem. Eur. J. 2014, 20, 1 – 8
www.chemeurj.org
5
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
&
These are not the final page numbers! ÞÞ

Full Paper
tions possible, then it is difficult to make meaningful conclu-
sions between the two. This in turn risks leading to dogmatic
assertions over the superiority of one mode over another. This
study has been highly revealing and has helped overturn as-
sumptions we have held over our own previous work in the
area. Whilst these conclusions hold for the range of photoreac-
tions described in this study, it is unclear whether similar out-
comes would be observed for other types of reaction involving
light, for example, photocatalysis, photooxygenation and
Single Electron Transfer. We hope that similar studies in these
areas and other non-photochemical areas of flow chemistry
are carried out so that the field can blossom into a technique
that is used in conjunction with established batch modes
rather than attempt to compete against it.
Figure 4. Light-transmission scenarios through one-layer and three-layer FEP
reactors.
Figure 4 shows that due to the offset layering of the FEP
tubing, the effective path length is likely to be significantly
less. Combining this with the expected absorption of FEP
tubing, one might expect batch to be superior to even the
three-layer FEP reactor. Clearly, this is not the case. At present,
the superior performance of the FEP reactor is not fully under-
stood and may reflect more complicated factors, such as differ-
ent concentration gradients between tubes and possible light
scattering effects affecting photon path lengths (Figure 4, C).
Experimental Section
General details
1
Batch reactions were analysed over time by H NMR using an inter-
nal standard (see the Supporting Information for detail). The pro-
ductivity of a batch reactor was then determined by the amount
of product isolated over the reaction time. For short-flow photo-
chemistry runs, at least one reactor volume was allowed to pass
through the irradiated reactor (to achieve “steady state”) before
photolysate was collected for a determined run time. For longer
runs of several hours and beyond, the total photolysate was col-
lected. In each case, the product was isolated and purified, and the
amount obtained was then extrapolated up or down to match the
batch end point.
Conclusion
In light of this study, the answer to the question—which is
better, batch or flow?—is a complex one, as we have shown
that both modes can do the same chemistry in essentially
identical yields and similar productivities for a range of reac-
tions. Ultimately, scale will likely dictate which reactor the user
chooses. For example, batch photochemical reactors are ideal
for first-time reactions (and users) due to lower cost, ease of
operation, reaction monitoring and ease of optimization. To
avoid over-irradiation in batch, it is essential to match the
power of the lamp with the concentration of substrate (e.g.,
entry 10, Table 1). It is likely that in the past, many poorly per-
forming small-scale batch reactions have been unnecessarily
discounted due to over-irradiation with an unsuitably matched
UV source. Batch reactors using 400 W lamps and 400 mL
volume immersion wells are well suited for reactions up to 10–
15 g scales, and in some cases, even larger (e.g., entry 11,
Table 1). FEP flow reactors are ideal for producing products on
scales of greater than 10 g in a single run, and for processing
larger volumes safely and efficiently (e.g., 26 to 28, Scheme 1).
FEP reactors are also ideal for irradiating large volumes of
dilute solutions, which would be impractical in batch mode,
especially when using large low-pressure lamps (e.g., en-
tries 5,7 and 12, Table 1).
Batch reactors (Figure 1a) were standard immersion well quartz re-
actors fitted with a Pyrex sleeve (36 mm o.d., 33 mm i.d.) that ac-
commodate a 400 W medium pressure Hg lamp and have a typical
filled volume of 400 mL, that is, the volume of photolysate that
covers the whole length of the Hg lamp. The reactor’s effective
path length was calculated to be approximately 8.7 mm. Cooling
was provided by either mains water or by a glycol/water-chilled
circulator. When monochromatic radiation was used, 36 W (40 cm)
single ended PL-L lamps (254 or 365 nm) were inserted into the
unfiltered quartz immersion well (i.e., no Pyrex sleeve) resulting in
approximately 50% immersion (see the Supporting Information for
exact power conversions in each case).
400 W flow reactors (Figure 1a) have been described previously by
us and involve 1–3 layers of FEP tubing (2.7 mm i.d., 3.1 mm o.d.)
wrapped around a customized water- or glycol-cooled Pyrex im-
mersion well (with sufficient tubing to cover the whole length of
the Hg lamp, ca. 15 cm) containing a 400 W medium pressure Hg
lamp.[11] Reactor volumes: 45 mL (one layer); 145 mL (three layer).
36 W flow reactors (Figure 1b) have been described previously by
us and involve one layer of FEP tubing (2.7 mm i.d., 3.1 mm o.d.)
wrapped around a custom quartz tube containing a low pressure
36 W single ended PL-L lamp (254 or 365 nm) cooled by air.[16] Re-
actor volume: 90 mL (one layer).
This study concludes that where productivity is concerned
both batch and flow modes of irradiation are equally impor-
tant in synthetic photochemistry. The choice of one mode over Acknowledgements
the other depends on what is required from a particular ex-
periment. We hope that these observations will enable first-
time users of photochemistry to choose appropriately. Finally,
this study highlights clearly that unless careful steps are taken
to compare batch versus flow under the most equitable condi-
We thank Novartis (L.D.E.), EPSRC EP/K006053/1 (P.J.K.), EPSRC
EP/L003325/1 (L.D.E.), University of Bristol (J.P.K.) for Postdoc-
toral funding, and the EPSRC Bristol Chemical Synthesis Doc-
toral Training Centre (EP/G036764/1), AstraZeneca, Syngenta
&
&
Chem. Eur. J. 2014, 20, 1 – 8
www.chemeurj.org
6
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Paper
[23] S. Cludius-Brandt, L. Kupracz, A. Kirschning, Beilstein J. Org. Chem. 2013,
9, 1745–1750.
[24] D. T. McQuade, A. G. O’Brien, M. Dorr, R. Rajaratnam, U. Eisold, B. Mon-
nand, T. Nobuta, H. D. Lohmannsroben, E. Meggersc, P. H. Seeberger,
Chem. Sci. 2013, 4, 4067–4070.
for PhD studentship funding (K.G.M., M.J.R.). We also thank P.
Dinham, D. Mitchell and D. Tarling (University of Bristol) for re-
actor construction, and C. Heuberger (Novartis, Basel) for DSC
analysis of 26.
[25] B. G. Anderson, W. E. Bauta, R. C. Cantrell Jr., Org. Process Res. Dev. 2012,
16, 967–975.
[26] R. S. Andrews, J. J. Becker, M. R. Gagnꢄ, Angew. Chem. 2012, 124, 4216–
4219; Angew. Chem. Int. Ed. 2012, 51, 4140–4143.
Keywords: batch
photochemistry · rearrangement · scale-up
reactors
·
cycloaddition
·
flow
[27] Y. S. M. Vaske, M. E. Mahoney, J. P. Konopelski, D. L. Rogow, W. J. McDo-
nald, J. Am. Chem. Soc. 2010, 132, 11379–11385.
[28] T. Horie, M. Sumino, T. Tanaka, Y. Matsushita, T. Ichimura, J. Yoshida, Org.
Process Res. Dev. 2010, 14, 405–410.
[29] F. Lꢄvesque, P. H. Seeberger, Angew. Chem. 2012, 124, 1738–1741;
Angew. Chem. Int. Ed. 2012, 51, 1706–1709.
[30] D. Kopetzki, F. Lꢄvesque, P. H. Seeberger, Chem. Eur. J. 2013, 19, 5450–
5456.
[31] M. Fischer, Angew. Chem. 1978, 90, 17–27; Angew. Chem. Int. Ed. Engl.
1978, 17, 16–26.
[32] For original batch reactions for Table 1 entries 1 and 3, see R. F. Childs,
A. W. Johnson, J. Chem. Soc. C 1967, 874–876.
[33] K. I. Booker-Milburn, J. K. Cowell, F. D. Jimꢄnez, A. Sharpe, A. J. White,
Tetrahedron 1999, 55, 5875–5888.
[34] K. I. Booker-Milburn, J. R. Baker, I. Bruce, Org. Lett. 2004, 6, 1481–1484.
[35] P. J. Hickford, J. R. Baker, I. Bruce, K. I. Booker-Milburn, Org. Lett. 2007, 9,
4681–4684.
[36] R. C. Cookson, E. Crundwell, R. R. Hill, J. Hudec, J. Chem. Soc. 1964,
3062–3075.
[37] S. D. Banister, M. Manoli, M. L. Barron, E. L. Werry, M. Kassiou, Bioorg.
Med. Chem. 2013, 21, 6038–6052.
[38] A. M. Kenwright, J. D. Sellars, Mag. Res. Chem. 2012, 50, 803–808.
[39] T. Aillet, K. Loubiere, O. Dechy-Cabaret, L. Prat, Chem. Eng. Prog. 2013,
64, 38–47.
[1] J. C. Pastre, D. L. Browne, S. V. Ley, Chem. Soc. Rev. 2013, 42, 8849–8869.
[2] I. R. Baxendale, J. Chem. Technol. Biotechnol. 2013, 88, 519–552.
[3] J. Wegner, S. Ceylan, A. Kirschning, Adv. Synth. Catal. 2012, 354, 17–57.
[4] J. Yoshida, H. Kim, A. Nagaki, ChemSusChem 2011, 4, 331–340.
[5] J. Wegner, S. Ceylan, A. Kirschning, Chem. Commun. 2011, 47, 4583–
4592.
[6] R. L. Hartman, J. P. McMullen, K. F. Jensen, Angew. Chem. 2011, 123,
7642–7661; Angew. Chem. Int. Ed. 2011, 50, 7502–7519.
[7] D. Webb, T. F. Jamison, Chem. Sci. 2010, 1, 675–680.
[8] X. Y. Mak, P. Laurino, P. H. Seeberger, Beilstein J. Org. Chem. 2009, 5, 19.
[9] J. P. Knowles, L. D. Elliott, K. I. Booker-Milburn, Beilstein J. Org. Chem.
2012, 8, 2025–2052.
[10] M. Oelgemoeller, Chem. Eng. Technol. 2012, 35, 1144–1152.
[11] B. D. A. Hook, W. Dohle, P. R. Hirst, M. Pickworth, M. B. Berry, K. I. Booker-
Milburn, J. Org. Chem. 2005, 70, 7558–7564.
[12] C. Birr, W. Lochinger, G. Stahnke, P. Lang, Justus Liebigs Ann. Chem.
1972, 763, 162–172.
[13] A. M. Braun, M-T. Maurette, E. Oliveros, Photochemical Technology. Chap-
ter 4, Wiley, Chichester, 1991.
[14] M. D. Lainchbury, M. I. Medley, P. M. Taylor, P. Hirst, W. Dohle, K. I.
Booker-Milburn, J. Org. Chem. 2008, 73, 6497–6505.
[15] F. R. Bou-Hamdan, F. Lꢄvesque, A. G. O’Brien, P. H. Seeberger, Beilstein J.
Org. Chem. 2011, 7, 1124–1129.
[40] L. D. Elliott, M. Berry, A. J. Orr-Ewing, K. I. Booker-Milburn, J. Am. Chem.
Soc. 2007, 129, 3078–3079.
[41] A. G. Griesbeck, N. Maptue, S. Bondock, M. Oelgemçller, Photochem.
Photobiol. Sci. 2003, 2, 450–451.
[42] K. I. Booker-Milburn, N. J. Costin, R. F. Dainty, D. Patel, A. Sharpe, Tetrahe-
dron Lett. 1998, 39, 7423–7426.
[16] K. G. Maskill, J. P. Knowles, L. D. Elliott, R. W. Alder, K. I. Booker-Milburn,
Angew. Chem. 2013, 125, 1539–1542; Angew. Chem. Int. Ed. 2013, 52,
1499–1502.
[17] D. C. Harrowven, M. Mohamed, T. P. GonÅalves, R. J. Whitby, D. Bolien,
H. F. Snedon, Angew. Chem. 2012, 124, 4481–4484; Angew. Chem. Int.
Ed. 2012, 51, 4405–4408.
[43] K. I. Booker-Milburn, C. E. Anson, C. Clissold, N. J. Costin, R. F. Dainty, M.
Murray, D. Patel, A. Sharpe, Eur. J. Org. Chem. 2001, 1473–1482.
[44] J. R. Edman, J. Am. Chem. Soc. 1966, 88, 3454–3455.
[45] J. R. Edman, J. Am. Chem. Soc. 1969, 91, 7103–7107.
[46] P. Kaszynski, J. Michl, J. Org. Chem. 1988, 53, 4593–4594.
[47] I. J. Lochert, Ph.D Thesis, Flinders University of South Australia, 1996.
[18] A. Yavorskyy, O. Shvydkiv, N. Hoffmann, K. Nolan, M. Oelgemçller, Org.
Lett. 2012, 14, 4342–4345.
[19] M. Hurevich, J. Kandasamy, B. M. Ponnappa, M. Collot, D. Kopetzki, D. T.
McQuade, P. H. Seeberger, Org. Lett. 2014, 16, 1794–1797; M. Collot, D.
Kopetzki, D. T. McQuade, P. H. Seeberger, Org. Lett. 2014, 16, 1794–
1797.
[20] H.-Q. Do, S. Bachman, A. C. Bissember, J. C. Peters, G. C. Fu, J. Am. Chem.
Soc. 2014, 136, 2162–2167.
[21] D. Cantillo, O. de Frutos, J. A. Rincon, C. Mateos, C. O. Kappe, J. Org.
Chem. 2014, 79, 223–229.
Received: July 10, 2014
[22] T. P. Willumstad, O. Haze, X. Y. Mak, T. Y. Lam, Y. P. Wang, R. L. Danheiser,
J. Org. Chem. 2013, 78, 11450–11469.
Published online on && &&, 0000
Chem. Eur. J. 2014, 20, 1 – 8
www.chemeurj.org
7
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
&
These are not the final page numbers! ÞÞ

Full Paper
FULL PAPER
&
Flow Photochemistry
L. D. Elliott, J. P. Knowles, P. J. Koovits,
K. G. Maskill, M. J. Ralph, G. Lejeune,
L. J. Edwards, R. I. Robinson, I. R. Clemens,
B. Cox, D. D. Pascoe, G. Koch, M. Eberle,
M. B. Berry, K. I. Booker-Milburn*
&& – &&
Spot the difference! By careful match-
ing of reaction parameters, the per-
formance of 13 different photochemical
reactions were compared in both batch
and flow reactors. Surprisingly, the
yields obtained in the different reactor
modes were essentially identical. Simi-
larly, the productivity differences be-
tween the two reactor modes, under
the same time scales, were relatively
small (see figure).
Batch versus Flow Photochemistry: A
Revealing Comparison of Yield and
Productivity
&
&
Chem. Eur. J. 2014, 20, 1 – 8
www.chemeurj.org
8
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!