Photocatalytic Oxidative [2+2] Cycloelimination Reactions with Flavinium Salts: Mechanistic Study and Influence of the Catalyst Structure

Article abstract of DOI:10.1002/cplu.202000767

Flavinium salts are frequently used in organocatalysis but their application in photoredox catalysis has not been systematically investigated to date. We synthesized a series of 5-ethyl-1,3-dimethylalloxazinium salts with different substituents in the positions 7 and 8 and investigated their application in light-dependent oxidative cycloelimination of cyclobutanes. Detailed mechanistic investigations with a coumarin dimer as a model substrate reveal that the reaction preferentially occurs via the triplet-born radical pair after electron transfer from the substrate to the triplet state of an alloxazinium salt. The very photostable 7,8-dimethoxy derivative is a superior catalyst with a sufficiently high oxidation power (E=2.26 V) allowing the conversion of various cyclobutanes (with E_ox up to 2.05 V) in high yields. Even compounds such as all-trans dimethyl 3,4-bis(4-methoxyphenyl)cyclobutane-1,2-dicarboxylate can be converted, whose opening requires a high activation energy due to a missing pre-activation caused by bulky adjacent substituents in cis-position.

Full text of DOI:10.1002/cplu.202000767

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
1
2
3
4
5
6
7
8
9
Photocatalytic Oxidative [2+2] Cycloelimination Reactions
with Flavinium Salts: Mechanistic Study and Influence of
the Catalyst Structure
Tomáš Hartman,[a] Martina Reisnerová,[a] Josef Chudoba,[b] Eva Svobodová,[a]
Nataliya Archipowa,[c] Roger Jan Kutta,*[d] and Radek Cibulka*[a]
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Dedicated to Professor František Liška on the occasion of his 80th birthday.
Flavinium salts are frequently used in organocatalysis but their
application in photoredox catalysis has not been systematically
investigated to date. We synthesized a series of 5-ethyl-1,3-
dimethylalloxazinium salts with different substituents in the
positions 7 and 8 and investigated their application in light-
dependent oxidative cycloelimination of cyclobutanes. Detailed
mechanistic investigations with a coumarin dimer as a model
substrate reveal that the reaction preferentially occurs via the
triplet-born radical pair after electron transfer from the
substrate to the triplet state of an alloxazinium salt. The very
photostable 7,8-dimethoxy derivative is a superior catalyst with
a sufficiently high oxidation power (E*=2.26 V) allowing the
conversion of various cyclobutanes (with Eox up to 2.05 V) in
high yields. Even compounds such as all-trans dimethyl 3,4-bis
(4-methoxyphenyl)cyclobutane-1,2-dicarboxylate can be con-
verted, whose opening requires a high activation energy due to
a missing pre-activation caused by bulky adjacent substituents
in cis-position.
Introduction
subject of several theoretical investigations.[2,3] In principle,
[2+2] cycloeliminations are forbidden in the ground state.[4]
However, a few thermally activated processes have been
described for specifically substituted cyclobutanes having strain
energy originating from their four ring structure (ca.
109 kJmolÀ1; ref.[5]) increased by steric effects of substituents.[6]
Recently, the use of cyclobutane ring opening has been
intensively studied in mechanochemistry.[7] Examples demon-
strating the usefulness of [2+2] cycloeliminations in organic
synthesis have also been reported.[8]
Photolyases are photoenzymes that are common in all king-
doms of life repairing photo-damaged DNA.[1] Photo-damaged
DNA typically consists of cyclobutanes arising from [2+2]
cycloadditions after UV light absorption. This damage is
reversed by light-induced [2+2] cycloeliminations via excited
fully reduced and deprotonated flavins. Cycloelimination reac-
tion of cyclobutanes resulting in their corresponding alkenes
([2+2] cycloelimination, [2+2] cycloreversion) has been a
In the excited state, cyclobutanes open without a significant
activation barrier. However, excitation typically requires UV light
even when it occurs via sensitization.[2] The splitting of a
thymine dimer or 1,2-diphenylcyclobutane utilizing cyanoar-
enes and UV light (313 or 366 nm) is proposed to occur via
electron-transfer in an exciplex formed between an excited
photocatalyst and a substrate.[9] Some [2+2] cycloeliminations
based on photoredox catalysis employing a dye and light in the
visible spectral range have also been reported: splitting of
quinolone and coumarin dimers mediated by pyrylium and
trityl salts,[10] cycloreversion of tetraphenylcyclobutanes or
triphenyloxetane with (thia)pyrylium salts,[11] and thymine dimer
opening by riboflavin tetraacetate (1) occurring in the presence
of perchloric acid.[12] These systems work via oxidative cleavage
starting with an electron transfer from a dimer to an excited
dye thus differing from the process in flavin-dependent photo-
lyases which is initiated by the cyclobutane reduction.[1a,13]
Recently, we have shown that flavin-based photooxidative
splitting is more robust when using the flavinium salt 2b
(Figure 1b) instead of 1 with perchloric or trifluoromethane
sulfonic acid (Figure 1a).[14] Because of the positive charge,
flavinium salts are strong enough to initiate the oxidative
cycloelimination of substrates with high oxidation potentials
[a] Dr. T. Hartman, M. Reisnerová, Dr. E. Svobodová, Prof. Dr. R. Cibulka
Department of Organic Chemistry
University of Chemistry and Technology, Prague
Technická 5, 166 28, Prague 6 (Czech Republic)
E-mail: cibulkar@vscht.cz
[b] Dr. J. Chudoba
Central Laboratories
University of Chemistry and Technology, Prague
Technická 5, 166 28, Prague 6 (Czech Republic)
[c] Dr. N. Archipowa
Manchester Institute of Biotechnology and School of Chemistry
The University of Manchester
Manchester M1 7DN (United Kingdom)
[d] Dr. R. J. Kutta
Institute of Physical and Theoretical Chemistry
University of Regensburg
93040 Regensburg (Germany)
E-mail: roger-jan.kutta@chemie.uni-regensburg.de
Supporting information for this article is available on the WWW under
https://doi.org/10.1002/cplu.202000767
This article is part of a Special Collection on “Chemistry in the Czech Re-
public”.
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH. This is
an open access article under the terms of the Creative Commons Attribution
Non-Commercial NoDerivs License, which permits use and distribution in
any medium, provided the original work is properly cited, the use is non-
commercial and no modifications or adaptations are made.
ChemPlusChem 2021, 86, 373–386
373
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
1
2
3
4
5
6
7
Scheme 1. [2+2] Cycloelimination with a model substrate 3a.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Photo-physics and -chemistry of 2b
First, the photo-physics of 2b in acetonitrile was investigated
by time-resolved absorption spectroscopy (Figure 2a). Impor-
tantly, 2b exists in an equilibrium to its pseudo-base 2bÀOH
with a pKR+ value of 4.8 in water (see Figure 3c).[20a] The
formation of 2bÀOH depends on the moisture inside the
system and it is manifested by an absorption spectrum with a
first absorption maximum at 350 nm (see ESI). The formation of
2bÀOH is suppressed in the presence of an acid, e.g. TfOH or
AcOH. Nevertheless, the pseudo-base does not absorb at
450 nm allowing for a selective excitation of only 2b.
2b shows comparable photo-physics as observed for
isoalloxazines. The excitation of 2b initially leads to the excited
singlet state of 2b, which converts with a lifetime of 4.8 ns
partially back to the ground state and partially into the triplet
state. The spectra of the S1 and the T1 states are very similar to
those known for isoalloxazines.[21] The excited singlet state
fluorescence quantum yield is determined to Φfl =11% and the
triplet yield is determined to ca. 89% (this value was obtained
from spectra modelling, see ESI). With a lifetime of 425 ns in
non-degassed acetonitrile, the triplet state returns back to the
ground state via back inter system crossing and energy transfer
quenching with molecular oxygen (Figure 2a,c).
Figure 1. Systems for oxidative [2+2] cycloelimination reactions mediated
by flavin photocatalysts that are excited by light.
even under neutral conditions. This allowed the extension of
the substrate scope demonstrating that the [2+2] cycloelimina-
tion is a valuable tool in organic synthesis.
In contrast to neutral flavins,[15,16] flavinium salts are novel in
photo-redox catalysis.[17] However, application of flavinium salts
in “dark” organo-catalysis has been known for several
decades.[18] Flavinium salts are mainly used in monooxygenation
reactions where they act via flavin hydroperoxides analogously
to flavin-dependent monooxygenases.[19] In this area, the
structure of the flavinium catalyst is crucial for its activity.[20]
However, a deep knowledge of the structure to activity relation-
ship is completely missing among flavinium photocatalysts.
Herein, following our preliminary results, we report a mecha-
nistic investigation on the [2+2] cycloelimination reaction
mediated by the excited flavinium salt 2b. Further, we have
prepared a series of flavinium salts 2 in order to study the
influence of substitution on their photocatalytic efficiency. We
figured out some disadvantages of catalyst 2b, e.g. low activity
for less strained substrates. This allowed us the development of
a robust and general flavinium-based light-driven procedure for
[2+2] cycloeliminations.
Mechanism of the cycloelimination of 3a mediated by 2b
Transient absorption (TA) spectra of 2b in the presence of 3a
show that the excited states react with the substrate seen in
reduced lifetimes and formation of new absorption features
(Figure 2b–d). At a concentration of 10 mM of 3a the excited
singlet state lifetime of 2b is quenched from 4.8 to 3.4 ns giving
a bi-molecular quenching rate of 1keT =8.6·109 MÀ1sÀ1. However,
corresponding spectral features of the potential radical pair are
not observed, indicating that the singlet-born radical pair
Results and Discussion
Efficient cycloelimination mediated by 2b
*
*
1[2b (+) :3a +] recombines faster than it is formed ending in a
non-productive reaction. The T1 state lifetime of 2b is also
quenched in the presence of 10 mM 3a from 425 ns to 170 ns
The photocatalytic cycloelimination reaction of 3a with the
flavinium salt 2b is quantitative within 10 minutes of irradiation
at 450 nm light.[14] It is proposed to occur by photoinduced
electron transfer (PET) from a substrate to 2b in an excited
state. This proposal was based on i) quenching experiments of
the 2b fluorescence with the coumarin dimer 3a and ii) the
free Gibbs energy for PET, which is À0.51 eV for 3a at maximum
(assuming the S1 state of 2b is involved; EredS1(2b)=2.56 V).
Herein, we studied the photocatalytic cleavage of the coumarin
dimer 3a to 4a by 2b (Scheme 1) in more detail.
giving a bi-molecular quenching rate of keT =3.5·108 MÀ1sÀ1. In
3
*
*
this case, a triplet-born radical pair 3[2b (+) :3a +] is formed
which is spin-forbidden for recombination, thus, allowing its
detection. The triplet-born radical spectrum consists mainly of
*
2
+
contributions of the protonated 2b species (see absorption
band at ca. 400 nm in Figure 2 and the theoretically calculated
absorption spectrum in Figure S18 in the Supporting Informa-
*
2
+
tion) since the decay of the 3a is in the order of magnitude
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
374
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Figure 2. Transient absorption data of 2b (303 μM) in acetonitrile in the absence (a) and presence (b) of 10 mM 3a. a,b: False colour representation of the
transient absorption data. c,d: Decay associated difference spectra (DADS) resulting from global fits on the data in a and b with corresponding lifetimes as
indicated by the tables in the inset. e: Species associated spectra and corresponding concentration time profiles in the inset: grey=2b ground state (S0),
*
black=12b* excited singlet state (S1), blue=32b triplet state (T1), red=22b , orange=4a product, cyan=global fit. The mole fraction for 4a is scaled by 0.5
since one 3a decomposes into two 4a molecules.
*
+
of its formation so that 23a
does not accumulate to a
detectable level. The protonation arises potentially from the
moisture in the solvent (to note, the residual water content is
expected to be lower in the preparative cycloeliminations. Thus,
Scheme 2. Light induced dealkylation of 2b.
*
2
participation of the neutral radical 2b under these conditions
cannot be excluded). The yield of the triplet-born radical pair
can be determined to 60% via ΦeT =1Àk0/kq, where k0 is the
total decay rate constant of the triplet in the absence and kq in
the presence of 3a. Simultaneously with the formation of the
triplet-born radical pair another species is formed. Its spectrum
has a single absorption band in the detection window peaking
at 310 nm and can be assigned to the final product of the
cycloelimination 4a as confirmed by UV/Vis and 1H NMR
absorption spectra of pure 4a (orange spectrum in Figure 2e).
On time scales longer than 400 μs the ground state bleach is
still observed. Ca. 10% of the initially excited molecules 2b
photo-degrade by dealkylation to form 5b (Scheme 2), which is
not resolved in these data. This will be discussed in more detail
(O2) recovering 2b and not forming any further 4a represents a
pure loss channel in terms of the total conversion. This is further
confirmed by a significantly enhanced quantum yield for this
*
+
system in the absence of O2 (see Table S2) so that 22b is
*
+
oxidized back to 2b by 24a forming more 4a. All known
spectra identified and determined in other experiments, i.e. the
1
3
2b ground state, the 2b* excited singlet state, the 2b triplet
*
2
+
state, the 2b radical cation, and the 4a photo-cleaved final
product, allow perfect decomposition of the transient
a
absorption data into physically meaningful concentration-time
profiles as shown in the inset of Figure 2e. As evident the build-
up of the product 4a shows a mole fraction of ca. 1.4 indicating
a partial radical chain reaction after initial eT under the
*
+
below. The formed 22b that reacts with molecular oxygen
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
375
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Figure 3. Photocatalytic cleavage of the coumarin dimer 3a by the excited flavinium salt 2b to the corresponding coumarin monomer 4a. General scheme
(a), reaction mechanism based on time-resolved spectroscopic studies (b) and alloxazinium salt 2b/pseudobase 2bÀOH equilibrium with a pKR+ of 4.8 in
water (c).
concentration conditions used in the TA experiment (In
comparison to conditions used for the data presented in
Table 1, the photocatalyst concentration was 4.6 times higher in
the TA experiment). At this point it should be pointed out that
the quantum yield of the system is significantly dependent on
the concentrations of the participating components (see
Table 1). Figure 3 summarizes the proposed reaction mecha-
nism.
Photostability of 2b
The photostability of 2b was investigated by recording
sequences of absorption spectra after a single blue light pulse,
or with continuous illumination. When 2b is illuminated by a
Table 1. Comparison of cycloelimination reaction of 3a to 4a with 2b and
5b.[a]
*
+
single blue light pulse, 2b is formed. 2b partially recovers on
S1[b]
Catalyst
Ered
[V]
λexc
[nm]
Time
[min]
Yield
[%][c]
Quantum yield Φ[d]
a minute time scale, but, interestingly, the recovery is not
*
+
complete. On prolonged illumination, 2b decomposes further
into another species with an absorption spectrum peaking at
ca. 310 and 380 nm, while the latter absorption band shows
vibrational fine structure (Figure 4). This photo conversion is
5 mM 3a
10 mM 3a
2b
5b
2.56
2.37
450
400
10
60
quant.
18
0.31 (0.75)
0.003
0.47
0.006
1
entirely irreversible. The H NMR of this photo-product reveals
[a] Reaction conditions: 3a (0.02 mmol), 2b or 5b (2.5 mol-%), acetonitrile
(V=1 mL), room temperature, under Ar atmosphere. [b] See next section
for the determination. [c] Yields determined by H NMR. [d] Determined
with 66 μM photocatalyst and 5 mM or 10 mM 3a in non-degassed
acetonitrile with expected c(O2)=2.4 mM[22] or in degassed acetonitrile by
cycles of freeze, pump thaw at a vacuum of 10À6 mbar (value in brackets).
that 2b decomposes upon illumination mainly via de-alkylation
to alloxazine 5b and acetaldehyde (Scheme 2; see ESI). The
absorption spectrum of pure 5b is in good agreement with the
obtained spectrum of the photo-product of 2b (Figure 4e,f).
1
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
376
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Figure 4. Sequence of stationary absorption spectra and the corresponding concentration-time profiles of formed species, when 2b dissolved in acetonitrile is
either incubated in the dark (a–b), illuminated by one blue light pulse (Δt=1 s; λmax =450 nm; c–d), or stepwise illuminated with blue light (λmax =450 nm; e–
f).
Contribution of 5b in cycloeliminations mediated by 2b
concentration of 5 mM. Thus, under the preparative condition
with reaction times on the minutes scale using 2b, the
contribution of 5b can be neglected. However, for cyclobutane
substrates with less steric hindrance, the conversion efficiency
might be significantly lower so that longer irradiation times
may be required. Therefore, in such cases the efficiency of the
cycloelimination will be mainly controlled by 5b.
Under the conditions used in the preparative cycloelimination
reaction with 2b, the neutral alloxazine 5b (formed by deal-
kylation), does not decompose significantly, which raised the
question, whether 5b could also contribute to the cyclo-
elimination. Considering the redox potentials of excited 5b
(2.37 V in maximum) and ground state 3a, the electron transfer
should be energetically allowed. And indeed, using 5b under
otherwise identical reaction conditions resulted in 18% of
cycloelimination of 3a after 60 minutes of irradiation at 400 nm.
However, using 5b as photocatalyst is significantly less efficient
than using 2b which gives quantitative conversion within 10
minutes of irradiation at 450 nm.[14] Further, the determination
of the quantum yields for the corresponding cycloelimination
using either 2b or 5b as photocatalyst in non-degassed
acetonitrile (Table 1) reveal that the conversion is by 2 orders of
magnitude smaller for 5b compared to 2b at a substrate
Synthesis
The instability observed for 2b started our search for more
suitable alloxazinium catalysts. In order to study a relationship
between the structure and the corresponding activity in the
series of alloxazinium salts 2, we prepared a 7-trifluoromethyl
derivative 2a
– regioisomer of the originally tested 8-
trifluoromethyl salt 2b. With the aim to introduce a heavy atom
into the alloxazine core, we prepared a pair of SF5-substituted
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
377
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
derivatives 2c and 2d. Then, we synthesized an unsubstituted
derivative 2e and a series of methyl- and methoxy- substituted
alloxazinium salts 2f–k as representatives of alloxazines with
electron-donating group(s) (see Scheme 3 for the structures).
Alloxazinium salts 2 were prepared from the corresponding
alloxazines 5 via reductive alkylation into position 5 using
acetaldehyde followed by oxidation of the flavin skeleton with
sodium nitrite (Scheme 3a). In contrast to literature
reports,[18a,20a, 23] the reductive alkylation step was performed in
dichloromethane providing good yields of the salts 2 (38–74%).
Two approaches were applied to synthesize alloxazines 5. The
first approach (Scheme 3b) utilized the condensation of alloxan
with an aromatic diamine 6, readily available from the
corresponding nitroanilines 7 via reduction. The condensation
reaction provided a mixture of 7- and 8- substituted alloxazines
8, which were alkylated and then separated.
Since the mixture of isomers 5f and 5g obtained via the
first approach could not be separated, an alternative method
using anilino uracils 9 prepared from 6-chlorouracil and 3- or 4-
methylaniline was used (Scheme 3c). Anilino uracil 9a was
cyclized to alloxazines using oxidative nitrosation with sodium
nitrite in acetic acid[24] yielding a mixture of 7-methylalloxazine
(8f) and 7-methylalloxazine-N-oxide (10a) which was immedi-
ately alkylated using methyl iodide to get a mixture of 1,3,7-
trimethylalloxazine (5f) and 1,3,7-trimethylalloxazine-N-oxide
(11a) in a moderate yield. The mixture was reduced using H2/
Pd to 5f. Cyclization of 9b gave a mixture of 6- and 8-
methylalloxazine 8g and 8l (1:9) with traces of the correspond-
ing N-oxides. After methylation and reduction, the mixture of
regioisomers 5l and 5g was separated by column chromatog-
raphy.
1
2
3
4
5
6
7
8
9
A series of truxinates 12 possessing various substituents
was prepared as model substrates, in order to investigate the
ability of various flavin photocatalysts to drive the correspond-
ing oxidative [2+2] cycloeliminations. Synthesis started from
cinnamic acids 13 which were transformed to anhydrides 14
with triphosgene (Scheme 4). Anhydrides were transformed
into cyclobutane 15 via photocatalytic cycloaddition by using
7,8-dimethoxy-1,3-dimethylalloxazine (5k) and illumination at
400 nm.[25] The sensitized cycloaddition reaction led to the
mixtures of products containing small amounts of anhydrides
(E,Z)-14 and (Z,Z)-14, which were formed by isomerization of 14,
isomeric cyclobutane (anti-trans product), and the major
product 15. Cyclobutane anhydrides 15 were typically obtained
as oils and used without purification, due to their instability on
silica. Thus, 15 were hydrolyzed to acids and transformed via
dichloride into esters 12. The final ester products 12 were
purified by column chromatography, which allowed its separa-
tion from the monomer esters and the minor cyclobutane
esters 16, as side products. The synthesis of the other cyclo-
butane substrates 17–24 is described in the ESI.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Spectral and electrochemical properties
All salts 2 absorb blue light with absorption maxima for the S1
S0 transition ranging from 424 to 479 nm (see Figure S19).
While the monomethoxy derivatives 2h and 2i absorb more in
Scheme 3. Synthesis of flavinium salts 2 (a) and synthetic approaches towards flavins 5 (b, c).
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
378
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Scheme 4. Synthesis of truxinates 12.
the blue part of this range, the dimethoxy derivative 2k absorbs
in the reddest part of this range (see Table 2 and Figure S19).
The emission maxima of the salts 2 for the S1!S0 transition lie
in the range from 525 to 540 nm with the exception of 2f, 2h,
and 2i with maxima at 550, 586, and 589 nm, respectively. In
contrast to the alloxazinium salts the alloxazines have blue
shifted absorption maxima for the S1 S0 transition in the range
of 372 to 424 nm. The corresponding emission maxima range
from 426 to 475 nm (see Table 3).
The quaternization of the alloxazines 5 resulting in the salts
2 strongly influenced their redox properties as observed by
significant shifts of the first reduction potential by ca. 0.7 V. The
substitution in positions 7 and 8 for both series of salts and
neutral alloxazines changed the reduction potential by ca. 0.3 V.
The redox properties of the excited singlet state E*(S1) can be
determined from the Rehm-Weller equation,[26] which are
summarized in Tables 2 and 3. However, the oxidation power
should be smaller, since the cycloelimination is initiated via the
triplet state of the flavinium catalyst. For instance, E*(T1) for 2b
was calculated quantum chemically to 2.20 V.
Table 2. Spectroscopic and electrochemical data for flavinium salts 2.
E0À0
E*
°
Flavinium salt
λabs
[nm][a]
λF
[nm][b]
E (1)
[V][c]
[eV][d]
[V][e]
2a
2b
2c
2d
2e
2f
2g
2h
2i
442
450
442
450
451
463
447
424
425
479
525
528
528
526
535
550
533
586
589
540
0.01
0.02
0.02
0.04
À0.10
À0.09
À0.09
À0.14
À0.07
À0.28
2.56
2.54
2.56
2.54
2.52
2.45
2.53
2.46
2.45
2.54
2.57
2.56
2.58
2.58
2.42
2.36
2.44
2.32
2.38
2.26
Cycloeliminations
The cycloelimination mediated by various alloxazinium salts 2
was investigated for a series of truxinates 12 with the oxidation
potential tuned by the substitution in position 4- of its phenyl
groups (Table 4). For the methyl derivative 12b next to the
cycloreversion also significant side reactions were observed that
most probably arose from benzyl radical formation leading to a
methyl group transformation. Very electron deficient substrates
12g and 12h (not shown in Table 4) with a CF3 or NO2 group,
2k
[a] Lowest energy band in the absorption spectra. [b] The maximum of the
fluorescence emission spectrum when excited at λabs. [c] First reduction
wave vs. SCE (reversible). [d] Approximated from absorption and emission
maxima; [e] Excited state reduction potential E*=E +E (ref. [26]).
0–0
°
Table 3. Spectroscopic and electrochemical data for alloxazines 5.
Table 4. Cycloreversion by salts 2.
E0–0
E*
°
Alloxazine
λabs
[nm][a]
λF
[nm][b]
E (1),
[V][c]
[eV][d]
[V][e]
Catalyst
Substrate[a] (Eox[V])
12a[b]
12c[b]
12d[c]
12f[c]
12g[c]
3a[b]
5a
5b
5c
5d
5e
5f
5g
5h
5i
380
373
374
382
378
376
388
424
372
391
426
432
428
440
435
429
447
475
475
442
À0.72
À0.71
À0.33
À0.43
À0.96
À0.88
À0.84
À0.91
À0.9
3.08
3.08
3.10
3.02
3.05
3.08
2.97
2.76
2.92
2.98
2.36
2.37
2.77
2.59
2,09
2.20
2.13
1.85
2.02
1.94
(1.5)
(1.8)
(2.1)
(>2.2)
(>2.2)
(2.05)
2a
2b
2c
2d
2e
2f
2g
2h
2i
72
67
22
23
100
59
22
80
83
100
72
86
78
93
100
52
58
76
68
100
8
11
12
5
0
13
0
0
0
0
10
0
0
0
0
0
0
0
0
0
0
quant.
92
70
55
trace
0
5
0
22
13
24
23
9
9
5k
À1.05
10
13
5
0
[a] Lowest energy band in the absorption spectra. [b] The maximum of the
fluorescence emission spectrum when excited at λabs. [c] First reduction
wave vs. SCE (reversible). [d] Approximated from absorption and emission
maxima. [e] Excited state reduction potential E*=E +E (ref. [26]).
2k
quant.
[a] Substrate (0.02 mmol), 2 (2.5 mol%), acetonitrile (1 mL), 450 nm, room
temperature, under Ar atmosphere. [b] 10 min. [c] 60 min.
0–0
°
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
379
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
respectively, were not opened with salts 2. In contrast, electron-
enriched truxinates 12a and 12c with a methoxy or tert-butyl
group were opened by all salts 2 with high conversion within
10 minutes. The substrates 12d and 12f with oxidation
potentials between 2.0 V and 2.2 V were opened with electron-
deficient alloxazinium salts 2a–d, however, a prolonged reac-
tion time (60 min) was necessary to achieve significant con-
versions. Interestingly, cycloelimination of 12d and 12f
occurred also with the unsubstituted salt 2e and with the
electron-enriched dimethoxy derivative 2k. Analogous results
were also obtained for the coumarin dimer 3a which was
effectively opened with 2a–d and 2k after 10 min.
1
2
3
4
5
6
7
8
9
Further, the performance of the dimethoxy derivative 2k in
comparison to 2b for a series of substrates was investigated
both on an analytical and preparative scale (Table 5). Although,
2k compared to 2b has lower oxidation properties in terms of
the excited state reduction potential (see Table 2), comparable
efficiencies on cycloeliminations were obtained for both photo-
catalysts. Both, 2k and 2b, split the coumarin dimer 3a
(Entry 1), the cyclic coumarin aza-analogue 3b (Entry 2), the
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Table 5. Comparison of cycloreversion mediated by either 2b[14] or 2k using selected cyclobutane derivatives.
Entry
Cyclobutane
Eox
Product
λexc
[nm]
Catalyst
Analytical scale[a]
Preparative scale[b]
[V][c]
Time
[min]
conv.[d]
[%]
Time
[min]
Yield[e]
[%]
450
450
530
530
2b
2k
2b
2k
10
10
60
480
quant.
quant.
85
30
30
–
92
93
–
1
2
3
2.05
1.34
2.14
quant.
–
–
450
450
2b
2k
10
10
quant.
quant.
30
30
88
91
450
450
2b
2k
60
60
39[f]
6
240
–
52
–
450
450
2b
2k
60
60
0
0
–
–
–
–
4
>2.2
450
450
2b
2k
10
10
quant.
quant.
10
10
86
83
5
6
7
1.71
1.77
1.55
450
450
2b
2k
10
10
quant.
80
10
10
91
86
450
450
2b
2k
60
60
quant[f]
71
120
60
75
87
450
450
2b
2k
60
60
47
45
60
60
50
57
8
9
2.02
1.57
450
450
2b
2k
10
10
quant.
quant.
10
10
94
92
450
450
2b
2k
60
60
7[f]
quant.
–
60
–
78
10
11
12
1.73
1.69
1.35
450
450
2b
2k
10
10
quant.
quant.
10
10
86
91
450
450
2b
2k
10
10
78
quant.
10
10
88[g]
94
[a] Substrate (0.02 mmol), 2b or 2k (2.5 mol%), acetonitrile (1 mL), room temperature, under Ar atmosphere. [b] Substrate (0.4 mmol), 2b or 2k (5 mol%),
acetonitrile (20 mL), room temperature, under Ar atmosphere. [c] Values vs. SCE. [d] Conversions from 1H NMR. [e] Yield of isolated product. [f] 10 mol% of 2.
[g] Contained 10% of Z-isomer.
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
380
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
cyclic anhydride 15a (Entry 5), the model of the thymine dimer
17 (Entry 6), the cycloadducts with tetramethylethylene 18 and
19 (Entries 7 and 8), the truxilate 12a (Entry 9), its sensitive
furan-containing analogue 21 (Entry 11), and the cis, trans-
isomer of tetrakis(4-methoxyphenyl)cyclobutane 22 (Entry 12).
The dimer 3c is only opened with 2b, however, a prolonged
reaction time is needed. The cyclic anhydride 15d with non-
substituted phenyl rings is beyond the limit of both flavinium
photocatalysts, which is in line with its high oxidation potential
(Entry 4).
Substrate 20 (all-trans isomer of 12a; Entry 10) needed a
prolonged irradiation time, although it has a relatively low
oxidation potential. This can be explained by a missing cis-
effect[27] (see below for further details). Since salt 2b photo-
decomposes (see first section) only a low cycloelimination yield
of 20 could be achieved even with a high catalyst loading up to
10%. In contrast, the photostable 2k splits 20 effectively.
The different photostability of 2b and 2k is also reflected in
the course of the cyclobutane 18 photocycloelimination, which
has no high oxidation potential, but is still relatively difficult to
split for both photocatalysts (Figure 5). The conversion rate for
the splitting of 18 into 25 in case of 2b is initially fast but stops
after several minutes at a level of ca. 70% indicating a
significant decomposition of 2b (black line in Figure 5). In
contrast, the cycloelimination of 18 using 2k is throughout
significantly slower, but is quantitative within 60 minutes (red
line in Figure 5) indicating a negligible photodecomposition of
2k on this time scale. Further, the reusability of 2b and 2k was
investigated for the model reaction with 3a, which is split
quantitatively by both photocatalysts. While 2k could be used
several times, 2b could only be used once due to photo-
decomposition (Table 6).
1
2
3
4
5
6
7
8
9
In the absence of any substrate only 2k showed significant
photostability. 2b was decomposed completely already after
2 min of irradiation, while only ca. 10% of 2k decomposed after
60 min of irradiation under identical conditions (see ESI for
more details). Therefore, the dimethoxyalloxazinium salt 2k
with its sufficiently high redox potential for oxidative splitting
of most substrates in combination with its high photostability
represents a valuable photocatalyst for cycloeliminations.
Finally, we noticed a significant influence of the substrate
configuration on the cycloelimination rate, known as the cis-
effect,[27] i.e. an acceleration of the cyclobutane splitting due to
steric repulsion between neighboring substituents in cis-
position, e.g. mainly phenyl rings. This was already shown on
the compounds 12a and 20. Interestingly, using 2k even the
cycloelimination of the all-trans derivative 20 is observed.
Furthermore, cycloelimination of the other trans-diarylcyclobu-
tanes without aryl group repulsion could be split by 2k
(Table 7), although less efficient compared with their cis-
derivatives.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Cycloaddition vs. cycloelimination
The neutral dimethoxyalloxazine 5k was described as a useful
mediator of intramolecular [2+2] photocycloadditions via
energy transfer.[25] Thus, we speculated, whether the corre-
sponding alloxazinium perchlorate 2k could also trigger this
reaction. Indeed, 2k catalyzes the [2+2] cycloaddition of
selected anhydrides 14 similarly to the neutral alloxazine 5k
(Table 8). Almost all tested cinnamic anhydrides 14 were
transformed to cyclobutanes 15 with high stereoselectivity in
moderate to quantitative yields in acetonitrile with 2k (it should
be noted that only the cyclobutane products with anti/cis
configuration were observed). Only for anhydride 14a (Table 8,
Entry 1) no formation of its cycloadduct was observed, which
may be explained by a 2k-initiated oxidative cycloelimination
of 15a back to 14a which is faster compared to the formation
of 15a (Scheme 5a). In contrast, oxidative cycloelimination of
15c, 15d, 15f and 15g with 2k was not observed in agreement
with their too high oxidation potentials as shown above (cf
Table 5, Entry 4).
Figure 5. Course of photocycloelimination of 18 mediated by 2b and 2k.
Table 6. Consecutive photocycloelimination experiments with 2b and 2k
on the model substrate 3a forming 4a.[a]
Entry
Cat.
Conv. [%]
2nd round
1st round
3rd round
1
2
2b
2k
100
100
0
100
–
75
[a] Substrate 3a (0.02 mmol), 2b or 2k (2.5 mol%), acetonitrile (1 mL),
exc =450 nm, room temperature, under Ar atmosphere; 10 minutes each
λ
Scheme 5. Cycloaddition/cycloelimination in the 14/15 system mediated by
2k and irradiation at 450 nm. For conversions and conditions, see Table 8.
round.
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
381
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
Table 7. Cycloreversion of cis- and trans-diarylcyclobutanes mediated by 2k.[a]
1
2
3
Entry
Time
[min]
cis-isomer
Conv.
[%][b]
trans-isomer
Conv.
[%][b]
4
5
6
7
8
1
2
10
1
quant.
37
5
6
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
3
4
10
10
quant.
quant.
40
32
5
10
quant.
15
1
[a] Substrate (0.02 mmol), 2k (2.5 mol%), acetonitrile (1 mL), λexc =450 nm, room temperature, under Ar atmosphere. [b] Conversions from H NMR.
Table 8. [2+2] Cycloadditions of anhydrides 14 sensitized by 2k.[a]
addition of 14a with 2.5% of 2k. Indeed, under these
conditions 89% of cyclobutane 15a were formed after 10
minutes of irradiation (Scheme 5b).
Entry
1
Anhydride
Product
Time
[min]
Conv.[b]
[%]
10
0
Conclusion
We described the detailed mechanism of the [2+2] photo-
cycloelimination of the coumarin dimer 3a as a model substrate
mediated by the salt 2b. After photoexcitation, 2b forms
efficiently its triplet state, i.e. ca. 89%, in which it can abstract
an electron from a substrate (dimer) initiating the cyclobutane
ring opening to form the final product coumarin (4a) partly via
its radical cation 4a and oxidizing 2b back to 2b. Depend-
ing on the concentration conditions the conversion is also
accelerated by an additional radical chain mechanism.
Simultaneously, it turned out that 2b undergoes fast
dealkylation forming the relatively photostable neutral allox-
azine 5b which also drives the cycloelimination, however, with
a two orders of magnitude lower quantum yield. Despite its
strong oxidizing properties, the lower stability of the electron
deficient salt 2b as well as its tendency to form the pseudobase
2bÀOH by reaction with water, e.g. due to residual moisture in
the solvent, can limit its practical applications in photocatalysis.
The photocycloelimination was tested for various differently
substituted alloxazinium salts 2 on the conversion of the
coumarin dimer 3a and on a series of truxinates 12. The 7,8-
dimethoxyalloxazinium derivative 2k showed a similar effi-
ciency as the 8-trifluoromethyl analogue 2b. Importantly, the
flavinium salt 2k is very photostable allowing the cycloelimina-
tion to proceed to high conversions even for substrates whose
opening requires to overcome a high activation barrier due to
steric hindrance as in cyclobutane 20. Additionally, 2k also
2
3
4
6
10/60
10/60
60
70/46
100/50
76
*
*
+
+
60
42
[a] Substrate (0.02 mmol), 2k (2.5 mol%), acetonitrile (1 mL), λexc =450 nm,
room temperature, under Ar atmosphere. [b] Conversions from H NMR.
1
The cycloelimination is only efficient in acetonitrile or
nitromethane.[14] In contrast, the cycloaddition does not depend
on the solvent significantly (cycloadditions with alloxazines
work well in dimethylsulfoxide, dimethylformamide, alcohols,
acetone, toluene, chlorinated solvents, and best in
acetonitrile).[25a] In order to avoid the cycloelimination we
selected dimethylsulfoxide as a suitable solvent for the cyclo-
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
382
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
initiates the intramolecular [2+2] cycloaddition of electron
deficient and electron neutral cinnamic anhydrides 14 which do
not undergo back oxidative cycloelimination because of their
high oxidation potential. With 4-methoxycinnamic anhydride
14a, both cycloaddition to 15a and its back cycloelimination to
14a occur in the presence of 2k. Interestingly, the preference
of either the cycloelimination or cycloaddition product of this
reversible process mediated by a single catalyst can be tuned
by the solvent environment.
To our knowledge this is the first example of a detailed
investigation of positively charged flavin derivatives as photo-
catalysts. The most promising flavinium catalyst for photo-
oxidative processes turned out to be 2k, and our findings
demonstrate also its potential for broader applications in
various other photocatalytic transformations.
Electrochemical measurements
1
2
3
4
5
6
7
8
9
Electrochemical measurements were performed using a standard
three-electrode system in a methrom type electrochemical cell with
a glassy carbon working electrode, silver wire pseudo-reference
electrode, and a platinum wire auxiliary electrode.[16d,17a] Cyclic
voltammograms were collected by PGSTAT128 N. The cyclic
voltammetry measurements were carried out in acetonitrile con-
taining a cyclobutane or alloxazinium salt 65 (c=1×10À3 molLÀ1
)
and
tetrabutylammonium
hexafluorophosphate
(c=1×
10À1 molLÀ1) as supporting electrolyte under argon atmosphere.
The scan rate was 100 mVsÀ1. The measured redox potentials were
converted into values relative to the standard calomel electrode
(SCE) using the standard redox couple Fc+/Fc as suggested by
Addison and others.[29] Conversion of the measured values into
those vs. SCE involved subtraction of the difference between
experimental E1/2 values for the standard redox couple Fc+/Fc
measured after each experiment (relative to the Ag wire) and E1/2
value for the standard redox couple Fc+/Fc measured against SCE
at the same apparatus (+0.381 V, the value obtained as an average
of 5 measurements).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Experimental Section
General comments to the starting material and synthesis
Stationary UV/Vis absorption and emission spectroscopy
Starting materials and reagents were obtained from commercial
suppliers and used without further purification. The solvents were
purified and dried using standard procedures.[28] Flavinium salts 2,
alloxazines 5, and substrates 12–22 were prepared and character-
ized as described in the ESI. NMR spectra were recorded on a
All UV/Vis absorption (Agilent Cary 60 or Shimadzu UV-1800
spectrophotometer) and UV/Vis emission spectra (Varian Cary
Eclipse or Jobin Yvon Fluorolog-3 fluorescence spectrofluorometer)
were recorded at room temperature. Fluorescence quantum yields
were determined with a Hamamatsu C9920-02 system equipped
with a Spectralon integrating sphere. The quantum yield accuracy
is <10% according to the manufacturer.
1
Varian Mercury Plus 300 (299.97 MHz for H, 75.44 MHz for 13C, and
282.23 MHz for 19F) or Agilent 400-MR DDR2 (399.94 MHz for 1H and
100.58 MHz for 13C) at 298 K unless otherwise stated. Chemical
shifts δ are given in ppm, using residual solvent or tetrameth-
ylsilane as an internal standard. Coupling constants J are reported
in Hz. High-resolution mass spectra were obtained on Q-Tof Micro
(Waters), equipped with a quadrupole and time-of-flight (TOF)
analyser and subsequent a multichannel plate (MCP) detector. Thin
layer chromatography (TLC) analyses were carried out on a DC
Alufolien Kieselgel 60 F254 (Merck). Preparative column chromatog-
raphy separations were performed on a silica gel Kieselgel 60
0.040–0.063 mm (Merck). Melting points were measured on a
Boetius melting point apparatus and are uncorrected.
Stepwise illumination and recording of UV/Vis absorption
spectra
The relative product quantum yield (PQY) in the presence of
molecular oxygen was determined by recording absorption spectra
of a particular system after step-wise temporally and geometrically
defined illumination (pulse width Δt=1 s, λexc =390/460 nm) in
10×10 mm quartz cell filled with 3 mL sample that was continu-
ously stirred when the sample was illuminated. These sequences of
spectra showing the product build-up upon illumination were
decomposed into the contributing species spectra and correspond-
ing concentration-time profiles (see ESI for all data). The initial slope
of the product build-up over illumination time is proportional to
the product quantum yield. For comparison within all measure-
ments the illumination times where corrected by the corresponding
overlap integrals between the normalized spectrum of the
excitation source and the absorption spectrum of the correspond-
ing chromophore, that was excited. The absolute quantum yield
was calculated from recording also the product formation of the
known photo-oxidation of 4-methoxybenzylic alcohol (MBA) to the
corresponding aldehyde by tetra-acetyl riboflavin (TARF) as a
reference system with a known quantum yield.[21a]
General procedures for cycloelimination reaction
Experiments on analytical scale: The cyclobutane derivative
(0.02 mmol) and the alloxazinium salt 2 or alloxazine 5 (0.5 μmol)
were placed in a Schlenk tube and anhydrous acetonitrile (1 mL)
was added. The homogenous solution was deoxygenated three
times via the freeze-pump-thaw technique. The solution in the
Schlenk
tube
was
irradiated
(LED,
Luxeon
STAR/0,
1030 mW@700 mA, 450 nm for 2 or LED, LED Engin LZC-70UA00
400 nm for 5) for a certain time, usually for 10 minutes. The reaction
mixture was analyzed by 1H NMR.
Experiments on preparative scale: The cyclobutane derivative
(0.4 mmol) and the alloxazinium salt 2 (2b or 2k; 0.02 mmol) were
placed in a Schlenk tube and anhydrous acetonitrile (20 mL) was
added. The homogenous solution was deoxygenated three times
via the freeze-pump-thaw technique. The solution in the Schlenk
tube was irradiated (6×LED, Luxeon STAR/0, 1030 mW@700 mA,
450 nm for 2 or 6×LED, LED Engin LZC-70UA00 400 nm for 5) for a
certain time depending on the cyclobutane substrate. After
irradiation, the solvent was removed, and the crude product was
purified by column chromatography (see ESI for details and
products characterization).
Time-resolved UV/Vis emission spectroscopy
A self-constructed Time Correlated Single Photon Counting (TCSPC)
setup[16h,30] was used to record emission decay data at single
detection wavelength. A quartz cuvette with four optical windows
of the dimension 2 mm×10 mm was used. The sample was excited
along the 2 mm pathlength and the emission was recorded
orthogonally to this. The optical density of the sample was set to
ca. 0.1 at the excitation wavelength over 2 mm pathlength.
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
383
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
the sub-ps transient absorption setup 10 mL of the sample were
cycled through an in-house build cell with a pathlength of ca.
100 μm for pump and probe beams. In the sub-ns transient
absorption setup 10 mL of the sample were cycled through a flow
cell (Starna) with a pathlength of 2 mm for pump and probe beams.
In all cases all scans resulted in reproducible data sets. Additionally,
the integrity of the sample was checked by recording stationary
absorption spectra before and after each measurement. No photo-
catalyst degradation was observed under the used conditions. The
shown data correspond to one representative measurement. No
smoothing or filtering procedures were applied to the data.
Time-resolved UV/Vis absorption spectroscopy
1
2
3
4
5
6
7
8
9
The ns to ms transient absorption spectroscopy was recorded by
a streak camera setup as described previously.[16h,21a,30–31] In brief, the
third harmonic of a Nd:YAG laser (10 Hz, Surelite II, Continuum)
pumping an Optical Parametric Oscillator (OPO, Continuum) tuned
to 355/460 nm (10 mJ, ca. 10 ns) was used for sample excitation. As
a probe light a pulsed 150 W Xe-flash lamp (Applied Photophysics)
was used which was focused three times via toric mirror optics: i)
before probe shutter, ii) into sample, iii) into spectrograph. The
entire white light probe pulse was analyzed by a combination of a
spectrograph (200is, Bruker) and a streak camera (C7700, Hamamat-
su Photonics). The use of mechanical shutters enabled the record-
ing of a sequence of three individual data sets: i) an image (DFL)
with both flash lamp and laser, ii) an image (D0) without any
incoming light, and iii) an image (DF) only with the flash lamp. 100
of such sequences were recorded and corresponding data sets
were averaged. Then, the TA was calculated as [Eq. (1)]:
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Transient absorption data analysis and modelling
SVD-based rank analysis and global fitting were performed using an
in-house written program described previously.[21b,30–31] In brief, the
linear least squares problem in Equation (2)
�
�
2
c2 ¼kDAÀFBk¼Min
DF À D0
DFL À D0
(2)
DOD ¼log
(1)
~
is solved, where A is the time-resolved absorption data matrix, F
is the matrix containing the analytical functions accounting for the
temporal changes in the data, i.e. exponential decays (convoluted
with the instrument response, typically a Gaussian function), and B
is the matrix with the to be determined spectra. Further
optimization of c2 is achieved by optimizing the rate constants in F
by a nonlinear least squares algorithm. As a result of such fits so-
called decay associated difference spectra (DADS in matrix B) and
their associated optimized rate constants are obtained. These are
the unique result of the global fit and this treatment does not
require any model for the kinetics involved in the transient
processes. The number of exponentials in the global fit is
determined by the SVD-based rank analysis, which is described
elsewhere.[33] The model that relates the actual species kinetics to
the elementary function is applied afterwards resulting in species
associated spectra (SAS). The shape of the SAS in terms of identity
with well-known spectra or following physical laws decides about
the appropriateness of the model. This step does not change the c2
value found in the global fit and, therefore, this procedure has the
advantage that all interpretation is performed with the same quality
of fit.
A 10 mL sample was stepwise cycled by a peristaltic pump (ecoline,
ISMATEC) through a flow cell with a pathlength of 2 mm for pump
and 10 mm for probe beams (dimensions: 2 mm×10 mm×30 mm,
Starna) ensuring a total replacement of the sample prior to each
individual measurement. No photocatalyst degradation was ob-
served under the used conditions.
The Sub-ps Pump/Supercontinuum-Probe Spectroscopy were
carried out using UV/Vis pump-supercontinuum probe spectrom-
eter at 1 kHz repetition frequency as described in ref.[16h] In brief, a
Ti-sapphire amplifier system (Coherent Libra) was used to generate
800 nm with 1.2 mJ pulses at 1 kHz. The output was split into three
parts of which only two were used: 1) Ca. 50% of the 800 nm
pulses were used to pump a colinear Optical Parametric Amplifier
(OPA, TOPAS-800-fs, Light Conversion) tuned to pump pulses
centered at ca. 450 nm (ca. 100 fs, ca. 400 nJ at the sample position)
for sample excitation. 2) Ca. 10% were used to pump a non-
colinear Optical Parametric Amplifier (NOPA, In-house build) tuned
to pulses centered at ca. 530 nm (ca. 100 fs, ca. 5 μJ at the CaF2
position) for generation of supercontinuum white light probe
pulses by focusing into a moving CaF2 disc of 1 mm thickness
giving a probe spectrum ranging from 310 to 700 nm. Pump pulses
were delayed via a motorized delay line equipped with an open
corner cube mirror up to 2 ns. Two complementary high-speed
spectrographs (Entwicklungsbüro EB Stresing) for signal and
reference recording were used. The pump and probe pulses were
focused colinearly into the sample to spot sizes of ca. 80 μm and
60 μm full width at half maximum (FWHM), respectively. For longer
As an alternative analysis, known species spectra, taken either from
literature or recorded in this work, were taken in order to
decompose the recorded time-resolved data matrix using the
transpose of the data matrix in [Eq. (2)] and using the basis spectra
instead of analytical functions. The resulting concentration-time-
profiles inform about the appropriateness of the basis spectra and
the physical reasonability, i.e. total sum of species being constant
to 1.
delays reaching out from ns to μs time ranges
a similar
spectrometer was used in which the pump laser was electronically
delayed relative to the probe laser. A detailed description can be
found in ref.[32] The relative polarizations between the pump and
probe were set by a half-wave plate in the pump-beam path to
Quantum chemical calculations
Quantum-chemical calculations on the excited singlet and triplet
°
magic angle (54.71 ) for observations of pure population changes.
states of 2b as well as on the radical species of 2b that potentially
*
contribute to the time-resolved absorption signals, i.e. 2b and
*
The averaged pre-t0 laser scatter signal was subtracted from the
data and the ca. 1 ps chirp of the white light was corrected for prior
to data analysis using the coherent artefact as an indicator for time
zero at each wavelength. Throughout the probe range, the spectral
resolution was better than 4 nm and the temporal resolution was
better than 150 fs. 10 individual scans with averaging 100 spectra
per time point were typically recorded. The time axis – within total
500 points – was linear between À1 and 2 ps and logarithmic from
2 ps to the maximum time delay ensuring that the dynamics on
every timescale will have equal weighting in the fitting analysis. In
2b +, were performed using the Firefly QC package,[34] which is
partially based on the GAMESS (US)[35] source code. All ground state
structures were optimized on the level of restricted open shell
density functional theory (ROHF-DFT) using the B3LYP functional
and the split-valence 6–31G(p,d)+ + basis set. In terms of the radical
spectrum calculation, complete active space self-consistency field
(CASSCF) theory was used in order to calculate the static correlation
energy. The most intense electronic transitions of 2b radicals, i.e.
DX D0 are of π–π* and n–π* type. Thus, 13 contributing π
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
384
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
1256; c) T. Majima, C. Pac, H. Sakurai, J. Am. Chem. Soc. 1980, 102, 5265–
5273; d) T. Majima, C. Pac, J. Kubo, H. Sakurai, Tetrahedron Lett. 1980, 21,
377–380.
electrons were included into the CAS, distributed over 12 molecular
orbitals (MO), and energy averaging over 10 states with equal
weights, i.e. CASSCF(13,12)10, was performed. In order to calculate
the dynamic correlation energy extended multi-configuration
quasi-degenerate perturbation theory (XMCQDPT) was used on top
of the CASSCF optimized MOs.[36] In all cases an intruder state
avoidance (ISA) denominator shift of 0.02 was used. Solvent effects
were taken into account with the polarized continuum model
(PCM).
1
2
3
4
5
6
7
8
9
[10] K. Okada, K. Hisamitsu, T. Mukai, Tetrahedron Lett. 1981, 22, 1251–1254.
[11] M. A. Miranda, M. A. Izquierdo, F. Galindo, J. Org. Chem. 2002, 67, 4138–
4142.
[12] a) K. Miyake, Y. Masaki, I. Miyamoto, S. Yanagida, T. Ohno, A. Yoshimura,
C. Pac, Photochem. Photobiol. 1993, 58, 631–636; b) C. Pac, K. Miyake, Y.
Masaki, S. Yanagida, T. Ohno, A. Yoshimura, J. Am. Chem. Soc. 1992, 114,
10756–10762.
[13] R. Epple, T. Carell, Angew. Chem. Int. Ed. 1998, 37, 938–941; Angew.
Chem. 1998, 110, 986–989.
[14] T. Hartman, R. Cibulka, Org. Lett. 2016, 18, 3710–3713.
[15] a) J. B. Metternich, R. J. Mudd, R. Gilmour, in Photocatalysis in Organic
Synthesis, 2019 ed. (Ed.: B. König), Georg Thieme Verlag, Stuttgart, 2019,
pp. 391–404; b) B. König, S. Kümmel, E. Svobodová, R. Cibulka, in
Physical Sciences Reviews, Vol. 3, 2018; c) I. K. Sideri, E. Voutyritsa, C. G.
Kokotos, Org. Biomol. Chem. 2018, 16, 4596–4614.
[16] a) N. P. Ramirez, B. König, J. C. Gonzalez-Gomez, Org. Lett. 2019, 21,
1368–1373; b) T. Hering, B. Mühldorf, R. Wolf, B. König, Angew. Chem.
Int. Ed. 2016, 55, 5342–5345; Angew. Chem. 2016, 128, 5428–5431;
c) L. M. Bouchet, A. A. Heredia, J. E. Argüello, L. C. Schmidt, Org. Lett.
2020, 22, 610–614; d) M. März, M. Kohout, T. Neveselý, J. Chudoba, D.
Prukała, S. Niziński, M. Sikorski, G. Burdziński, R. Cibulka, Org. Biomol.
Chem. 2018, 16, 6809–6817; e) T. Neveselý, E. Svobodová, J. Chudoba,
M. Sikorski, R. Cibulka, Adv. Synth. Catal. 2016, 358, 1654–1663; f) B.
Mühldorf, R. Wolf, Angew. Chem. Int. Ed. 2016, 55, 427–430; Angew.
Chem. 2016, 128, 437–441; g) J. B. Metternich, R. Gilmour, J. Am. Chem.
Soc. 2016, 138, 1040–1045; h) A. Graml, T. Neveselý, R.-J. Kutta, R.
Cibulka, B. König, Nat. Commun. 2020, 11, 3174.
[17] It was shown that the usage of ethylene-bridged flavinium salts allow
aerobic photooxidations of benzylic substrates, see: a) J. Zelenka, E.
Svobodová, J. Tarábek, I. Hoskovcová, V. Boguschová, S. Bailly, M.
Sikorski, J. Roithová, R. Cibulka, Org. Lett. 2019, 21, 114–119; b) J.
Zelenka, R. Cibulka, J. Roithová, Angew. Chem. Int. Ed. 2019, 58, 15412–
15420; c) A. H. Tolba, F. Vávra, J. Chudoba, R. Cibulka, Eur. J. Org. Chem.
2020, 2020, 1579–1585.
[18] a) T. Sakai, T. Kumoi, T. Ishikawa, T. Nitta, H. Iida, Org. Biomol. Chem.
2018, 16, 3999–4007; b) R. Cibulka, Eur. J. Org. Chem. 2015, 2015, 915–
932; c) H. Iida, Y. Imada, S. I. Murahashi, Org. Biomol. Chem. 2015, 13,
7599–7613.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Acknowledgements
This project was supported by the Czech Science Foundation
(Grant No. 18-15175S). R.J.K. thanks the Deutsche Forschungsge-
meinschaft for a Research Fellow stipend (KU 3190/1-1, KU
3190/2-1) supporting this project. The authors are grateful to Prof.
Eberhard Riedle (BioMolekulare Optik, Ludwig-Maximillians Uni-
versität München) for allowance to use of the ns to μs transient
absorption apparatus. T.H. and R.C. thank Dr. Petr Beier for
providing pentafluorosulfanyl precursors.
Conflict of Interest
The authors declare no conflict of interest.
Keywords: cyclobutanes · cycloelimination reactions · electron
transfer · photocatalysis · time-resolved spectroscopy
[1] a) A. Sancar, Angew. Chem. Int. Ed. 2016, 55, 8502–8527; Angew. Chem.
2016, 128, 8643–8670; b) A. Sancar, Chem. Rev. 2003, 103, 2203–2238;
c) K. Brettel, M. Byrdin, Curr. Opin. Struct. Biol. 2010, 20, 693–701.
[2] a) E. Schaumann, R. Ketcham, Angew. Chem. Int. Ed. 1982, 21, 225–247;
Angew. Chem. 1982, 94, 225–229; b) K. Mizuno, C. Pac, CRC Press, 1995,
pp. 358–374.
[3] a) D. Davis, V. P. Vysotskiy, Y. Sajeev, L. S. Cederbaum, Angew. Chem. Int.
Ed. 2011, 50, 4119–4122; Angew. Chem. 2011, 123, 4205–4208; b) N.
Paul, M. Jiang, N. Bieniek, J. L. P. Lustres, Y. Li, N. Wollscheid, T. Buckup,
A. Dreuw, N. Hampp, M. Motzkus, J. Phys. Chem. A 2018, 122, 7587–
7597; c) K. Shima, J. Kimura, K. Yoshida, M. Yasuda, K. Imada, C. Pac, Bull.
Chem. Soc. Jpn. 1989, 62, 1934–1942.
[4] a) R. B. Woodward, R. Hoffmann, Angew. Chem. Int. Ed. 1969, 8, 781–853;
Angew. Chem. 1969, 81, 797–869; b) R. Hoffmann, R. B. Woodward, J.
Am. Chem. Soc. 1965, 87, 2046–2048.
[5] P. v. R. Schleyer, J. E. Williams, K. R. Blanchard, J. Am. Chem. Soc. 1970,
92, 2377–2386.
[19] a) W. J. H. Van Berkel, N. M. Kamerbeek, M. W. Fraaije, J. Biotechnol.
2006, 124, 670–689; b) F. G. Gelalcha, Chem. Rev. 2007, 107, 3338–3361.
[20] a) P. Ménová, H. Dvořáková, V. Eigner, J. Ludvík, R. Cibulka, Adv. Synth.
Catal. 2013, 355, 3451–3462; b) Y. Imada, H. Iida, S. Ono, Y. Masui, S.-I.
Murahashi Chem. Asian J. 2006, 1, 136–147; c) A. A. Lindén, N.
Hermanns, S. Ott, L. Krüger, J.-E. Bäckvall, Chem. Eur. J. 2005, 11, 112–
119; d) P. Ménová, R. Cibulka, J. Mol. Catal. A 2012, 363–364, 362–370.
[21] a) U. Megerle, M. Wenninger, R.-J. Kutta, R. Lechner, B. Konig, B. Dick, E.
Riedle, Phys. Chem. Chem. Phys. 2011, 13, 8869–8880; b) R. J. Kutta, N.
Archipowa, N. S. Scrutton, Phys. Chem. Chem. Phys. 2018, 20, 28767–
28776.
[22] C. Franco, J. Olmsted, Talanta 1990, 37, 905–909.
[23] Y. Arakawa, T. Oonishi, T. Kohda, K. Minagawa, Y. Imada, ChemSusChem
2016, 9, 2769–2773.
[24] Y. Sakuma, S. Matsumoto, T. Nagamatsu, F. Yoneda, Chem. Pharm. Bull.
[6] a) J. Liese, N. Hampp, J. Phys. Chem. A 2011, 115, 2927–2932; b) C. Pac,
Trends Phys. Chem. 1990, 1, 15–31; c) M. Yasuda, K. Yoshida, K. Shima, C.
Pac, S. Yanagida, Bull. Chem. Soc. Jpn. 1989, 62, 1943–1950.
1976, 24, 338–341.
[25] a) V. Mojr, E. Svobodová, K. Straková, T. Neveselý, J. Chudoba, H.
Dvořáková, R. Cibulka, Chem. Commun. 2015, 51, 12036–12039; b) V.
Mojr, G. Pitrová, K. Straková, D. Prukała, S. Brazevic, E. Svobodová, I.
Hoskovcová, G. Burdziński, T. Slanina, M. Sikorski, R. Cibulka, Chem-
CatChem 2018, 10, 849–858; c) M. Jirásek, K. Straková, T. Neveselý, E.
Svobodová, Z. Rottnerová, R. Cibulka, Eur. J. Org. Chem. 2017, 2017,
2139–2146.
[26] D. Rehm, A. Weller, Ber. Bunseng. Phys. Chem. 1969, 73, 834–839.
[27] a) N. Yonezawa, T. Yoshida, M. Hasegawa, J. Chem. Soc. Perkin Trans. 1
1983, 1083–1086; b) C. Pac, T. Ohtsuki, Y. Shiota, S. Yanagida, H. Sakurai,
Bull. Chem. Soc. Jpn. 1986, 59, 1133.
[7] a) Z. S. Kean, Z. Niu, G. B. Hewage, A. L. Rheingold, S. L. Craig, J. Am.
Chem. Soc. 2013, 135, 13598–13604; b) Z. Chen, X. Zhu, J. Yang, J. A. M.
Mercer, N. Z. Burns, T. J. Martinez, Y. Xia, Nat. Chem. 2020, Ahead of
Print; c) M. F. Pill, K. Holz, N. Preusske, F. Berger, H. Clausen-Schaumann,
U. Luening, M. K. Beyer, Chem. Eur. J. 2016, 22, 12034–12039; d) M. J.
Robb, J. S. Moore, J. Am. Chem. Soc. 2015, 137, 10946–10949.
[8] a) J. D. White, Y. Li, J. Kim, M. Terinek, Org. Lett. 2013, 15, 882–885; b) W.
Oppolzer, Acc. Chem. Res. 1982, 15, 135–141; c) J. D. White, J. Kim, N. E.
Drapela, J. Am. Chem. Soc. 2000, 122, 8665–8671; d) M. T. Crimmins,
J. M. Pace, P. G. Nantermet, A. S. Kim-Meade, J. B. Thomas, S. H.
Watterson, A. S. Wagman, J. Am. Chem. Soc. 2000, 122, 8453–8463;
e) B. A. Boon, A. G. Green, P. Liu, K. N. Houk, C. A. Merlic, J. Org. Chem.
2017, 82, 4613–4624.
[28] D. D. Perrin, Purification of Laboratory Chemicals, 4th Ed., Elsevier
Science Ltd., Oxford, 1996.
[29] H. G. Roth, N. A. Romero, D. A. Nicewicz, Synlett 2016, 27, 714–723.
[30] R. J. Kutta, Blitzlichtphotolyse – Untersuchung zu LOV-Domänen und
photochromen Systemen, Universität Regensburg 2012.
[9] a) C. Pac, T. Ohtsuki, Y. Shiota, S. Yanagida, H. Sakurai, Bull. Chem. Soc.
Jpn. 1986, 59, 1133–1139; b) C. Pac, Pure Appl. Chem. 1986, 58, 1249–
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
385
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH

Full Papers
doi.org/10.1002/cplu.202000767
ChemPlusChem
[31] a) R.-J. Kutta, T. Langenbacher, U. Kensy, B. Dick, Appl. Phys. B 2013, 111,
203–216; b) B. Dick, U. Kensy, R.-J. Kutta, in 15 Transient Absorption, De
Gruyter, 2013, p. 295.
[35] M. W. Schmidt, K. K. Baldridge, J. A. Boatz, S. T. Elbert, M. S. Gordon, J. H.
Jensen, S. Koseki, N. Matsunaga, K. A. Nguyen, S. Su, T. L. Windus, M.
Dupuis, J. A. Montgomery, Jr., J. Comput. Chem. 1993, 14, 1347–1363.
[36] A. A. Granovsky, J. Chem. Phys. 2011, 134, 214113.
1
2
3
4
5
[32] B. Baudisch, Time resolved broadband spectroscopy from UV to NIR,
Ludwig-Maximilians-Universität München 2018.
[33] K. Lanzl, M. v. Sanden-Flohe, R.-J. Kutta, B. Dick, Phys. Chem. Chem. Phys.
2010, 12, 6594–6604.
[34] A. A. Granovsky, 1997, p. http://classic.chem.msu.su/gran/firefly/in-
dex.html.
Manuscript received: December 1, 2020
Revised manuscript received: December 22, 2020
Accepted manuscript online: December 23, 2020
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
ChemPlusChem 2021, 86, 373–386
www.chempluschem.org
386
© 2020 The Authors. ChemPlusChem published by Wiley-VCH GmbH