Micellized Tris(bipyridine)ruthenium Catalysts Affording Preparative Amounts of Hydrated Electrons with a Green Light-Emitting Diode

Article abstract of DOI:10.1002/chem.201801955

We have explored alkyl substitution of the ligands as a means to improve the performance of the title complexes in photoredox catalytic systems that produce synthetically useable amounts of hydrated electrons through photon pooling. Despite generating a super-reductant, these electron sources only consume the bioavailable ascorbate and are driven by a green light-emitting diode (LED). The substitutions influence the catalyst activity through the interplay of the quenching parameters, the recombination rate of the reduced catalyst OER and the ascorbyl radical across the micelle-water interface, and the quantum yield of OER photoionization. Laser flash photolysis yields comprehensive information on all these processes and allows quantitative predictions of the activity observed in LED kinetics, but the latter method provides the only access to the catalyst stability under illumination on the timescale of the syntheses. The homoleptic complex with dimethylbipyridine ligands emerges as the optimum that combines an activity twice as high with an undiminished stability in relation to the parent compound. With this complex, we have effected dehalogenations of alkyl and aryl chlorides and fluorides, hydrogenations of carbon–carbon double bonds, and self- as well as cross-coupling reactions. All the substrates employed are impervious to ordinary photoredox catalysts but present no problems to the hydrated electron as a super-reductant. A particularly attractive application is selective deuteration with high isotopic purity, which is achieved simply by using heavy water as the solvent.

Full text of DOI:10.1002/chem.201801955

A Journal of
Accepted Article
Title: Micellized tris(bipyridine)ruthenium catalysts affording preparative
amounts of hydrated electrons with a green light-emitting diode
Authors: Robert Naumann, Florian Lehmann, and Martin Goez
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Chem. Eur. J. 10.1002/chem.201801955
Link to VoR: http://dx.doi.org/10.1002/chem.201801955
Supported by

10.1002/chem.201801955
Chemistry - A European Journal
Micellized tris(bipyridine)ruthenium catalysts affording preparative
amounts of hydrated electrons with a green light-emitting diode†
By Robert Naumann, Florian Lehmann and Martin Goez∗
Abstract: We have explored alkyl substitution of the ligands as a means to improve the
performance of the title complexes in photoredox catalytic systems that produce synthetically
useable amounts of hydrated electrons through photon pooling. Despite generating a super-
reductant, these electron sources only consume the bioavailable ascorbate and are driven
by a green light-emitting diode (LED). The substitutions influence the catalyst activity through
the interplay of the quenching parameters, the recombination rate of the reduced catalyst
OER and the ascorbyl radical across the micelle–water interface, and the quantum yield of
OER photoionization. Laser flash photolysis yields comprehensive information on all these
processes and allows quantitative predictions of the activity observed in LED kinetics, but
the latter method provides the only access to the catalyst stability under illumination on the
timescale of the syntheses. The homoleptic complex with dimethylbipyridine ligands emerges
as the optimum that combines an activity twice as high with an undiminished stability in
relation to the parent compound. With this complex, we have effected dehalogenations of
alkyl and aryl chlorides and fluorides, hydrogenations of carbon–carbon double bonds, and
self- as well as cross-couplings. All the substrates employed are impervious to ordinary
photoredox catalysts but present no problems to the hydrated electron as a super-reductant.
A particularly attractive application is selective deuteration with high isotopic purity, which is
achieved simply by using heavy water as the solvent.
1 Introduction
•−
Using the hydrated electron
e
as a•r−elay allows extending photoredox catalysis[1–6] to
aq
aqueous solutions with a vengeance: eaq provides the reducing power of metallic potassium
yet is persistent enough to be scavenged near-quantitatively by mM concentrations of
•−
substrates (standard electrode potential E◦ and unquenched lifetime of
e
,
−
2.9 V and
aq
1–2 µs);[7] and in doing its work on the substrate it self-destructs without leaving a by-product.
In a recent communication,[8] we have presented the first photoredox catalytic system
•−
that generates
e
on a laboratory scale while being both user-friendly and sustainable
aq
through operating with a green light-emitting diode (LED) instead of a high-power laser[9]
and consuming only a bioavailable sacrificial species. After an attempt at replacing the latter
∗Robert Naumann, Florian Lehmann, Prof. Dr. Martin Goez,
Martin-Luther-Universität Halle-Wittenberg, Institut für Chemie, Kurt-Mothes-Str. 2, D-06120 Halle (Saale),
Germany.
E-mail: martin.goez@chemie.uni-halle.de
†Supporting Information available: Comprehensive experimental details, absorption and emission spectra
in the visible range, additional laser flash photolysis results, LED photolysis background information, and
•−
NMR spectra for the LED-induced transformations via e . See DOI: ...
aq
1
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
to accommodate a specific class of synthetic applications,[10] here we focus on improving
the catalyst.
Our electron sources rely on the sequence absorption — reductive quenching — photo-
ionization, which cycles a tris(bipyridine)ruthenium catalyst through its ground state GS
,
metal-to-ligand charge-transfer excited state MLCT, and one-electron reduced complex
[11]
OER
.
Schematic structural formulas of the catalyst and ligand are displayed as the
insets of Figure 1; and the substitution pattern of the ligands has been compiled in Table 1,
below, together with the substance abbreviations used throughout this work. The ubiquitous
antioxidant and approved food additive ascorbate serves as the sacrificial donor. To increase
its reductive ability, we employ it in the form of its dianion Asc2−, which necessitates working
in strongly basic solution (pH 12.7). Its radical Asc•− is very unreactive; in particular, it
•− [12]
hardly intercepts e
.
aq
Lα
R
R
N
N
N
N
x
N—L —N
=
Ru2+
N
Lβ
GS:
N
N
Asc•−
OER
N
L
γ
η KSV
Asc2−
krec
εeff(OER)ϕion
(MLCT)
εeff(GS)
•−
e
GS
aq
•−
Figure 1: Introducing the green-light driven cyclic source of hydrated electrons e
.
The schematic representation of the reaction
aq
mechanism allows comparative kinetic analysis of laser flash photolysis and preparative LED illumination. Light-driven steps are shown
as green arrows. The micellized ruthenium catalyst (for its structural formula and that of the ligands, see the insets and Table 1) in its
ground state GS undergoes photoinduced electron-transfer quenching via its metal-to-ligand charge-transfer excited state MLCT by the
ascorbate dianion Asc2− through the micelle–•w−ater interface. The obtained one-electron reduced form OER is then ionized by a second
green photon, ejecting the hydrated electron eaq into the aqueous bulk and regenerating GS. All kinetic parameters needed for analysis
are given at the reaction arrows. The storage-limiting recombination of OER and the ascorbyl radical Asc•− is displayed in red.
The two-photon mechanism solves the problem that the energy of a single green photon
•−
(2.33 eV at 532 nm) does not suffice to eject
e
from any stable precursor molecule known
aq
to date.[13] Compared to its variant with post-ionization regeneration of the catalyst,[14] it
also optimizes the storage of the energy bestowed by the first photon: OER as a ”hidden”
radical anion[15] does not undergo photophysical deactivation and is expected to decay
predominantly by a second-order process, which is intrinsically more favourable than first-
order deactivation; and compartmentalization by an anionic micelle augments this benefit
considerably.[8]
When the second photon arrives after the ”expiry date” of the stored first photon, no
photoionization is possible. Hence, the storage losses by the bimolecular recombination
•−
of OER and Asc•− fundamentally limit the
e
production at the low photon densities an
aq
LED delivers. As a strategy to minimize this recombination, we have increased the micellar
shielding of the catalyst through substituting the ligands with alkyl groups in this work. We
investigate the consequences of these modifications on short and long timescales, by laser
flash photolysis and by the kinetics under LED illumination, relegating all experimental
2
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
details to Section 1 of the Supplementary Information (hereafter abbreviated to as SI–1,
etc.) for conciseness. This comparative study pays heed to recently voiced concerns that
mechanistic and kinetic studies on photoredox catalytic systems are much too scarce.[16–18]
As will emerge, no monocausal relationship between the structure of the catalyst and its
•−
performance in an
e
source exists, especially because the activity (i.e., the maximum rate
aq
•−
at which
e
can be produced) is not the only property that counts; equally important is the
aq
catalyst stability under long-time illumination. This duality is also the reason why laser flash
photolysis does not provide a complete picture but needs to be complemented by the LED
kinetics. Our approach has resulted in identifying a significantly better catalyst; and we will
demonstrate the successful application of this catalyst to a number of synthetic procedures
•−
on substrates that are extremely difficult to reduce but are readily attacked by e
.
aq
2 Results and discussion
2.1 Mechanism and kinetics
2.1.1 Laser flash photolysis
Our 532 nm laser excites the lowest-energy band of GS (SI–2.2) at its red edge. When Ru(II)
polypyridine complexes have absorbed a photon in the visible range, their long-lived MLCT
excited states are formed by intersystem crossing within sub-ps.[15] During ns laser pulses,
stimulated emission is thus noncompetitive such that a pulse of sufficient intensity can
achieve complete conversion of GS into the luminescent MLCT. The left inset of Figure 2a
demonstrates that we can drive our systems well into this limit with our setup.
(a)
(b)
1.0
0.8
0.6
0.4
0.2
0
1
0.8
0.4
0
0.1
0
10
5
1
0
0
0.5
t / µs
0
1
t / ms
0
I
800
532 / (mJ cm−2
)
0
0
0
1
2
0
4
8
0
250
I
500
)
750
t / µs
t / ms
532 / (mJ cm−2
Figure 2: Characterizing the system of Figure 1 by laser flash photolysis. Common experimental parameters, 50 µM catalyst in 50 mM
aqueous SDS at pH 12.7, 75 mM Asc2− except for the yellow curve and left inset in graph (a). Catalyst (see Table 1) colour codes
and symbols, where applicable: RuBpy, gray, circles; RuDmb, green, triangles; RuTbb, orange, squares; RuMdnb, cyan, pentagons;
RuBdnb, magenta, inverted triangles. Graph (a), single-flash experiments on OER formation and decay. The main plot displays, on
two timescales as indicated by the arrows at the curves, the unquenched (yellow) and quenched (red) luminescence and the OER con-
centration upon quantitative excitation (flash intensity, 728 mJ cm−2) of RuDmb; left inset, MLCT concentration, from the unquenched
luminescence intensity directly after the flash, as function of the excitation intensity; right inset, linearization demonstrating the bimolec-
ular recombination of OER and Asc•− for all catalysts. Graph (b), two-flash experiments (first flash, fixed intensity for quantitative
excitation of GS; 5 µs delay for reaching the quenching end point; second flash, variable intensity I532) for determining the e•aq− yield
•−
from OER; inset, e
trace obtained from RuDmb at the maximum I532, and superimposed best-fit first-order decay (rasterized curve).
aq
3
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
We found strictly monoexponential decays of MLCT in the micellar solutions (SI–3.1).
The unquenched lifetimes τ0 have been compiled in Table 1. In the microheterogeneous
medium, τ0 is noticeably longer than in water, indicating that in all our experiments the
complexes are completely micelle-bound. There is no discernible correlation between the
— with 720 ns ± 25 % generally similar — values of τ0 and the lipophilicity of the catalyst.
The expected quenching of MLCT by Asc2− manifests itself by a shortening of the MLCT
lifetime; and the MLCT decay is now accompanied by a rise of the OER absorption with
the same rate (see, left half of Figure 2a). The MLCT decay remains monoexponential
(SI–3.1) and exhibits the same initial amplitude as in the absence of Asc2−, which identifies
the quenching by the strongly hydrophilic dianion as a purely dynamic process across the
micelle–water interface. Table 1 lists the quenching rate constants kq; the dependence on
the alkyl substituents will be discussed below.
Table 1: Investigated complexes: abbreviations and substitution pattern; standard electrode potentials; and photokinetic parameters
obtained by laser flash photolysis.
c
catalyst
abbreviation
substituents R in ligand a
E◦ b
[V]
τ0
kq
−1s−1
η
krec
−1s−1
ϕion
[%]
Lα
Lβ
Lγ
[ns]
[107
M
]
[107
M
]
[19]
[19]
[20]
RuBpy
RuDmb
RuTbb
RuMdnb
RuBdnb
H
Me
tert-Bu
H
H
Me
tert-Bu
H
H
Me
tert-Bu
n-C9H19
n-C9H19
−1.07
−1.21
−1.20
830(620)
541(300)
880(380)
856(490)
725( n/a )f
5.6
1.8
0.63
2.9
1.4
0.64
0.55
0.49
0.51
0.52
11
12
4.6
7.0
3.6
1.1
2.3
0.90
1.1
0.69
−1.13 d
H
n-C9H19
−1.17 e
a See ligand structure in Figure 1. b Standard electrode potential of the couple GS/OER in acetonitrile. c Values in homogeneous
aqueous solution given in brackets. d Unavailable value approximated by the average of E◦ for (4,40-dimethyl-2,20-bipyridine)-bis-
(2,20-bi-pyridine)-ruthenium-(II)[19] and (4,40-di-tert-butyl-2,20-bipyridine)-bis-(2,20-bipyridine)-ruthenium-(II)[20] e Unavailable value ap-
.
proximated by the average of E◦ for bis-(4,40-dimethyl-2,20-bipyridine)-(2,20-bipyridine)-ruthenium-(II)[19] and bis-(4,40-di-tert-butyl-2,20-
bipyridine)-(2,20-bipyridine)-ruthenium-(II)[20] f Insufficiently soluble and self-associating in water.
.
Neither Asc2− nor its radical Asc•− absorb in the green; and all our catalysts possess an
isosbestic point between GS and MLCT around 510 nm, which almost coincides with the
maximum of the OER band (SI–2.4). Hence, that isosbestic wavelength permits not only
isolated but also very sensitive monitoring of OER. We calibrated the OER spectra by using
a quencher (4-methoxy phenolate) that affords a more favourably absorbing radical than
Asc•− and by exploiting an isosbestic point between GS and OER that occurs in the vicinity
of 410 nm for all our complexes. Details are again found in SI–2.4. In conjunction with the
quantitative conversion of GS into MLCT and the Stern–Volmer constant KSV calculated
from kq and τ0,
ꢀ
ꢁ−1
KSV
=
1 + 1/(kqτ0[Asc2−])
(1)
the absolute OER concentrations after the quenching directly yield the efficiency
η
of OER
formation. The resulting values have also been included in Table 1. They are seen to be
very similar with the exception of RuBpy, for which η is about 20 % higher.
The right half of Figure 2a focuses on the decay of OER, which takes place on a disparately
longer timescale. The clear second-order kinetics during the first ms — as established by the
linearizations in the inset, which directly give the values of krec in Table 1 — naturally suggest
interpreting this decay as the expected recombination of OER and the quenching by-product
4
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
Asc•−. This identification as a reaction between unlike species formed in equal amounts is
validated by different decay rate constants krec when OER of the same catalyst is accessed
with different quenchers, e.g., Asc2− vs. the urate dianion.[10] As a secondary effect that
only becomes visible in wider observation windows, we found an apparent decrease of
the recombination rate constant over time. We attribute this to the additional gradual
removal of the recombination partner Asc•− by its disproportionation.[21] Corroboration of
this explanation was obtained by halving the initial radical concentration through a lower laser
intensity: this led to a much later onset of the deceleration (SI–3.2). Even though a complete
kinetic description is hopeless on account of the highly complex secondary chemistry of
Asc•− [21]
the disproportionation is obviously beneficial to an LED driven electron generator
,
because it prolongs the availability of OER for absorbing the ionizing photon; in other words,
because it increases the ”shelf life” of the energy stored in the system by the first photon,
which becomes more and more crucial when the photon flux is lowered.
The emission spectra (SI–2.3) confine the MLCT energies to an interval only
±
0.03 eV
wide around the value for RuBpy (2.12 eV)[15]; the standard electrode potentials E◦ (Table 1)
are clustered in the narrow range
−
1.14
±
0.07 V; and E◦ of the couple Asc•− Asc2− is
/
+0.05 V.[21] Hence, both the quenching of MLCT by Asc2− and the recombination of OER
and Asc•− are strongly exergonic, with relatively small variations of the driving force by the
ligands. This suggests the accessibility of the micellized complex as the major influence on
the kinetics. Most exposed in the series RuBpy, RuMdnb, and RuBdnb is an unsubstituted
bipyridine ligand; rising lipophilicity of the other ligands will draw the complex more into the
interior of the micelle. Comparing quenching and recombination, the Coulombic effects
partly cancel: the dication MLCT sticks out more into the polar Stern layer than does the
monocation OER, but the quenching dianion Asc2− is repelled more strongly by the negative
surface charge of the micelle than is the recombining monoanion Asc•−. Consistent with this
reasoning, these three complexes exhibit a proportionality between krec and kq with the ratio
krec/kq only about half as large as the variation of each rate constant in that series (SI–3.3).
For the homoleptic complexes RuDmb and RuTbb, the exposed ligand is also different, and
shielding by the aliphatic substituents can additionally modulate the accessibility to Asc2−
and Asc•−. Hence, stronger effects are expected, and the constant of proportionality is
indeed found to be more than twice as large as the individual variability.
The pivotal step of the electron source of Figure 1 is the ionization of OER with a (second)
green photon, which excites OER practically at the maximum of its lowest-energy band
(SI–2.4). We studied that process by two-pulse experiments, in which we used a high-
intensity first pulse to achieve quantitative formation of MLCT, ensured completion of the
OER generation by an interpulse delay of 5
the intensity of the second pulse. Furthermore, we observed
µ
s (compare, left half of Figure 2a), and varied
•−
e
in isolation by difference
aq
experiments without and with N2O saturation (SI–1), a procedure that succeeds even in the
presence of strongly absorbing other transients and/or ”pathological” background artifacts in
micellar systems.[13,22]
•−
The
e
decay (for a representative trace, see the inset of Figure 2b) is monoexponen-
aq
tial, and a catalyst-independent lifetime of 194 ns rules out an influence of the ruthenium
•−
complexes on the further fate of
e
after its generation. This is consistent with our earlier
aq
5
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
•−
investigations,[8,9,14] which pinpointed the capture of
e
by residual ascorbate monoanion
aq
HAsc− as the lifetime-limiting factor in the absence of deliberately added scavengers. The
pKa of HAsc− is 11.74;[21] hence, the HAsc− concentration is about 7.5 mM in the experi-
ments of this work. Although the elusive[23] product of the reaction between•−
e
and HAsc−
•−
aq
interferes neither spectroscopically nor mechanistically with our desired eaq -induced syn-
•−
theses, it lowers the
e
utilization through (fortunately only inefficient) kinetic competition.
aq
The convolution of the fast decay with the 5 ns generating pulse noticeably rounds the cusp
•−
ideally expected for the
e
trace. To obtain the true initial heights, we therefore extrapolated
aq
the electron decays back to the center of the laser flash, as seen superimposed on the trace.
•−
The main plot of Figure 2b displays the intensity dependences of the
e
yield. Their
aq
initial linear rise confirms the expected monophotonic ionization of OER; but this is only
the first-order term of the dependence over the full range of intensities, a saturation curve
1
− exp
(
−κI) whose parameter
κ is proportional to ϕion and to the extinction coefficient
of OER as the only catalyst-specific quantities.[24] The values of ϕion extracted with the
calibrated OER spectra of SI–2.4 have been collected in Table 1. At first glance, their trend
seems to reflect a competition between alkyl substitution facilitating the ionization through
a higher energy of OER and impeding it through a ”deeper” localization of OER inside
the micelle. However, the much more electron-rich yet less lipophilic tris-(4,40-dimethoxy-
2,20-bipyridine)-ruthenium(II) disproves this expla•n−ation: although OER of this derivative
(E◦(GS/OER) =
−
1.25 V)[25] should be the best eaq emitter on both counts, we found ϕion
to be only 1.6 % (and the quenching parameters so unfavourable as to render that complex
useless for our purposes). This suggests that the properties and deactivation pathways of
the excited OER play the dominant role.
The laser experiments of this Section have provided an in-depth characterization of our
electron sources on short timescales, from µs to ms. In the next Section, we will contrast
them with LED photolyses, the durations of which are typically measured in hrs; nevertheless,
the information they yield is largely equivalent, as we will show.
2.1.2 LED photolysis
The spectra of GS (SI–2.2) and OER (SI–2.4) show that the LED excites the same transitions
as the laser did. The effective extinction coefficients with the polychromatic LED (for its
spectrum, see SI–2.1) differ noticeably (SI–4.1) from the ones at the laser wavelength, which
comes as a welcome side benefit in the case of GS.
The light intensity of our LED is 8 orders of magnitude lower than that of our laser during
•−
a pulse such that the tiny steady-state concentrations of
e
in preparative illumination
aq
require a quantification method other than direct observation. A viable alternative is provided
by adding a substrate that does not interfere with the electron generator but scavenges
•−
e
in a stoichiometrically defined reaction to give a stable product. The concentration of
aq
this accumu•la−ted product as function of the illumination duration is proportional to the total
•−
amount of eaq generated within that period; hence, its derivative is proportional to the
generation rate, that is, to the catalyst activity.
e
aq
Chloroacetate ClAc is suited extremely well for this purpose: neither it nor its products
6
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
attack the catalyst; it quenches neither MLCT nor excited OER; but it reac•ts− r[a8,p9]idly (rate
•−
constant, 1.9
×
109 M−1s−1 at the high ionic strength of our samples) with eaq
.
The
e
aq
capture is a dissociative electron transfer,[7]
•−
+ H•
+e
aq
ClCH2COO− −−−→ Cl− +• CH2COO− −−−−→ Cl− + CH3COO−
(2)
residual HAsc− can serve as the hydrogen donor for neutralizing the resulting carboxymethyl
radicals, and the liberated chloride ion is inert. As a further asset, this reaction also
•−
represents an environmentally relevant application of e because ClAc is a bulk chemical
aq
(annual production in excess of 106 tonnes) with strong acute toxicity yet high persistence in
the aquifer.[26]
In conjunction with the mechanism of Figure 1, Equation 2 thus suggests the formation of
one chloride ion to be accompanied by the consumption of two ascorbate molecules. An ion-
sensitive electrode provides a convenient means to measure the chloride concentration, and
the ascorbate concentration can be selectively determined by using a chemical assay (SI–1).
As the inset of Figure 3a demonstrates, there is indeed a fixed stoichiometric relationship
between the two species over the whole conversion range; but the stoichiometric factor
is somewhat smaller, namely, 1.6 in the case of RuBpy and 1.3 in that of RuDmb, which
will emerge to be the least active and the most active of our catalysts. The discrepancy
is caused by the partial involvement of the known[21] disproportionation of Asc•− as a
secondary reaction, which would recover one of the two ascorbate molecules if it occured
quantitatively, but obviously none if it did not take place at all. The more active the catalyst is,
the more Asc•− it produces per unit time, increasing the disproportionation probability and
thus lowering the ascorbate consumption. This behaviour on long timescales is consistent
with the above-mentioned deceleration over time of the recombination between OER and
Asc•−, which gains prominence at higher radical concentrations (SI–3.2).
(a)
(b)
0.8
0.4
0
1.0
0.5
0
40
0
0
25
−
[
]
Cl / mM
0
2
4
0
4
8
12
t / h
t / h
Figure 3: Full characterization of the electron sources of Figure 1 with the chloroacetate assay. Common experimental parameters,
50 µM catalyst, 75 mM Asc2−, and 25 mM ClAc in 50 mM aqueous SDS at pH 12.7; colour codes and symbol shapes for the catalysts
listed in Figure 2. Graph (a); time dependence of the chloride concentration liberated from ClAc, relative to the starting ClAc concentra-
tion (main plot), overlaid with best-fit functions a(1 − exp[−bt]); and ascorbate consumption per chloride produced (inset) for the least
and most active catalyst; slopes of the regression lines, 1.6 (RuBpy) and 1.3 (RuDmb). Graph (b), catalyst decomposition as function of
the illumination time; overlaid with best-fit functions exp[−t/τCat]. Further explanation, see text.
The main plot of Figure 3a displays the time dependence of ClAc decomposition according
7
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
to Equation 2 for our catalysts. Despite the deceptive simplicity of the curves, which can
bt]) over the limited
be approximated very well by monoexponential functions a(1 − exp −
[
temporal range of the Figure, modelling the kinetics over the whole extent of the reaction is
only possible numerically because the system composition changes in a nonlinear manner.[8]
Fortunately, such an elaborate procedure is not at all necessary for comparing catalysts:
instead, the desired information is already contained in the initial slopes of the curves, which
•−
correspond to the maximum generation rate of
e
for each system in its starting state,
aq
that is, with precisely known concentrations of all ingredients. These initial slopes directly
•−
yield (SI–4.2) the turnover frequencies for electron generation TOF(
e
), which describe
aq
the intrinsic activity of each catalytic system at the specified standard composition and
irradiation conditions of this work. Table 2 lists their values, which clearly mark RuDmb
as outstanding. Their connection with the photokinetic parameters obtained by laser flash
photolysis will be discussed below.
Table 2: Comparison of the intrinsic performance of the investigated catalysts.
RuBpy
RuDmb
RuTbb RuMdnb RuBdnb
•−
TOF(e
τCat
)
[h−1
[h]
]
78
9.70
1510
170
9.28
3160
113
3.10
700
106
4.20
890
100
2.28
460
aq
•−
TON(e
)
aq
However, catalytic activity is only one side of the coin; the other, equally important one is
catalyst stability. Figure 3b juxtaposes the decays of our catalysts, as measured through
their luminescence (SI–1), during preparative illumination. They are represented very well
by monoexponential functions exp
[ t/τCat] with the lifetimes τCat compiled in Table 2. The
−
catalyst loss is thought to occur through thermal population of a dissociative dd state slightly
[27]
higher in energy than MLCT
.
This was recently corroborated by our investigation on
which established that the decomposition rate is directly
[8]
the parent complex RuBpy
,
proportional to the LED power — and, by the same token, to the extinction coefficient
of GS — as well as to the quenched lifetime of MLCT. However, in comparing different
catalysts these trends can be totally obscured by the substance-specific efficiencies of ligand
loss from the dd state, as the example of RuBpy and RuDmb shows. The energy gaps
between their MLCT and dd states are 40 kJ/mol and 24 kJ/mol,[27] the effective extinction
coefficient of GS with our LED (SI–4.1) is 50 % higher for RuDmb, and the lifetime of MLCT
under our standard conditions (75 mM Asc2−) is 70 % longer in the case of RuDmb. All
these factors work in the same direction and should make RuDmb much less stable than
RuBpy; yet, these two complexes possess values of τCat that differ only by 5 % (Table 2).
We ascribe this deviation from the expected trend to a lower dissociation tendency of the
substituted ligand on account of its better σ-donating ability.
On the basis of the relationships that we have derived theoretically and demonstrated ex-
perimentally for the mechanism of Figure 1,[8,10] the influence of the chemical, spectroscopic,
8
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
•−
and photokinetic parameters on TOF(e ) is given by
aq
r
KSV × η
•−
TOF(e
)
∝
εeff(GS) × [Cat] ×
× εeff(OER) × ϕion
(3)
aq
krec
Equation 3 is derived by setting up a steady-state approximation for OER and eliminating
the intermediacy of MLCT through the quenching parameters.[8] At the low photon fluxes an
LED delivers, both these simplifications incur negligible errors. The square root arises from
the second-order recombination of OER and Asc•− and applies to all processes except the
ionization step.
•−
•−
The turnover number TON(
e
), which specifies the maximum amount of
e
a catalyst
aq
aq
can deliver, comprises both the activity and the stability. As our systems do not exhibit
•−
catalyst poisoning, TON(e ) can be estimated by[8]
aq
•−
TON(e
)
=
2 × τCat × TOF(e•a−q
)
(4)
aq
and the resulting values have been included in Table 2. Although they are only valid for the
system composition (through [Cat] and KSV) and irradiation conditions (through the εeff(X)
and τCat) of this work, they still allow a performance comparison of the catalysts. It is seen
that the highly lipophilic catalysts RuTbb, RuMdnb, and RuBdnb fall behind RuBpy by a
considerable amount, contrary to the hopes that their higher lipophilicity would minimize krec
,
but that RuDmb clearly outranks the field.
Equation 3 is the central link between the experiments on short and long timescales
provided that the catalyst concentration is kept constant and the ground-state spectra are
known, requirements that are trivial to meet. Laser flash photolysis yields the OER spectra
and the data of Table 1; and with these quantities, Equation 3 allows calculating relative
•−
values of TOF(
e
). Figure 4a demonstrates the success. Nevertheless, this approach
aq
has two inherent limitations. First, secondary reactions such as a disproportionation of the
quencher-derived radicals can cause deviations; and there is little chance of extrapolating
such effects from the timescale of laser flash photolysis to that of preparative photolysis.
Ascorbate is not very critical in this respect, although the inset of Figure 3a indicates
noticeable discrepancies, but with urate the disproportionation participates to such an
extent that the square-root dependence no longer holds.[10] Second, the stability of our
catalysts cannot be captured by laser flash photolysis because the decomposition quantum
yields are too small to be measurable in pulsed experiments but become important through
accumulating over the hours of the LED illumination.
When the reaction vessel has a larger diameter than the LED beam has, the question
arises whether the beam should be magnified to fill the entrance window more completely
but with lower intensity per area, or be collimated to illuminate only a small cylindrical
subvolume but with higher intensity per area. Two opposite influences of this size change
need to be considered. First, as long as the contents are continuously mixed by stirring,
the only effect of partial illumination is a reduction of the overall excitation rates by the
ratio of the illuminated volume to the total volume or, assuming a cylindrical vessel, by
the corresponding area ratio of the beam and the entrance window.[14] Second, with the
9
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
(a)
(b)
2
1.0
0.5
0
1.0
0.5
0
1
0
0
1
2
0
2
4
6
8
10
•−
A
LED / cm−2
TOF(e
)
pred, rel
aq
Figure 4: Predicting the activity and optimizing the performance of the catalytic electron sources for preparative LED photoredox catal-
ysis. Catalyst colour codes and symbol shapes, see Figure 2. Graph (a), experimental turnover frequency TOF of electron production
(Table 2) as function of the TOF calculated from the parameters obtained by laser flash photolysis (Table 1), both relative to the value for
RuBpy. Graph (b), influence of collimating the LED to different areas A at constant power and solution volume on the electron generation
•−
rate TOF(e ) and the catalyst lifetime τCat, both normalized to the values at an area of 1 cm2. The arrows at the curves denote the
aq
pertaining vertical scale. Experimental parameters, 18 ml of 50 µM RuDmb, 75 mM Asc2−, and 25 mM ClAc in 50 mM aqueous SDS
ꢀ
ꢁ
LED/(1 cm2) −1/2; τCat, right scale, open symbols, dotted
•−
at pH 12.7. TOF(e ), left scale, filled symbols, solid fit curve given by
A
aq
constant line with height 1.0. Further explanation, see text.
mechanism of Figure 1 the rate of catalyst decay is directly proportional to the LED radiative
•−
power P, but the rate of electron generation or the associated quantity TOF(
e
) exhibits the
aq
unusual dependence on P3/2, as we have already verified experimentally.[8] Taken together,
optically magnifying or reducing the beam diameter should not affect the catalyst stability in
•−
any way but should scale TOF(e ) with the inverse square root of the beam area.
aq
Figure 4b tests these predictions and proves them to hold true, in particular, that for the
electron generation rate, which might appear counterintuitive at first glance. In preparative
work with the mechanism of Figure 1, concentrating the LED beam to the smallest possible
diameter will thus give the best result.
2.2 Synthetic applications
RuDmb has the highest activity of our catalysts, yet — together with RuBpy — the longest
lifetime under LED illumination; therefore, we carried out all the following syntheses with this
•−
complex. Its improved TOF(
e
) can be exploited for shortening the irradiation duration or
aq
for increasing the amount of substrate. In this work we chose the second option, typically
[8]
by doubling the substrate concentration compared to our previous work on RuBpy
,
and
occasionally also by upscaling the sample volume.
2.2.1 Dehalogenations
In Section 2.1.2, we have extensively used ClAc as a test compound for investigating the
electron source of Figure 1 on the timescale of its applications. Here, ClAc additionally
serves to demonstrate the feasibility of upscaling the reaction. To that end, we increased
the solution volume by about an order of magnitude, from 3.8 ml to 36 ml, and partly
10
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
compensated the expected lengthening of the reaction time by doubling the optical path
length, which approximately doubles the absorbed light because the solution is optically thin.
As the NMR spectra in SI–5.1 bear out, near-quantitative turnover of the larger amount was
achieved within 48 h.
The dehalogenation can also be employed to effect deuteration, and two of its features are
very advantageous in that respect. First, the reaction is site-specific because the hydrogen
atom takes the place of the chlorine atom. Second, it is easy and inexpensive; as the only
requirement, one needs to work in D2O instead of H2O because the hydrogen (or deuterium)
donor is the ascorbate, which exchanges all its transferable protons with the solvent. The
upper traces of Figure 5a juxtapose the NMR spectra at the illumination end points for the
two reaction media. The substrate is quantitatively dechlorinated in both cases, but the 1:1:1
triplet in the deuterated solvent evidences the conversion into CDH2COO−. The minute but
clearly discernible singlet of CH3COO− at slightly lower field is larger than expected on the
basis of the ascorbate weight-in concentration and the isotopic purity of the solvent, which
suggests that the substrate and/or the SDS contribute as hydrogen donors to a small extent.
Another deuteration example will be given below. The simplicity, selectivity, near-quantitative
isotope substitution (as opposed to other methods in the literature)[28], and potentially large
•−
number of substrates that can be dechlorinated by
e
could make this an attractive method
aq
for deuteration and — with no changes in mechanism expected — tritiation, which is of
importance for investigations of pharmaceuticals.[28–31] Using
e
as mediator seems to
•−
aq
obviate the necessity of polarity matching by adding a thiol:[32] with the mechanism of this
work, the ascorbate species efficiently serve as both the electron and the hydrogen-isotope
donors in these photoredox-driven atom transfers.[31]
(a)
(b)
1
1
D2O
H2O
0
1
0
1
0
1
0
0
4.2
4.1
2.0
1.9
3.7
2.8
2.7
2.3
1.0
δ / ppm
δ / ppm
Figure 5: Dehalogenations of 10 mM aliphatic substrates with RuDmb and 75 mM Asc2− in 50 mM aqueous SDS at pH 12.7. Shown
are the 1H-NMR spectra before (bottom traces) and after 14 h of illumination with the green LED (all other traces), with the NMR
resonances of the respective substrate coloured cyan. Graph (a), chloroacetate ClAc in H2O and D2O, 50 µM RuDmb; products acetate
(green, 1.98 ppm, s) and monodeuteroacetate (orange, 1.97 ppm, t 1:1:1). Graph (b), 3-chloropropionate ClPr, 100 µM RuDmb; product
propionate (orange; 2.29 ppm, q; and 0.98 ppm, t). All spectra were recorded after acidification to pH 1; and each set of spectra has
been normalized to the global maximum signal. Further explanation, see text.
As an even more demanding aliphatic substrate than ClAc, we selected 3-chloropropionate
ClPr. The insertion of the sp3 centre between the carbon atoms bearing the functional
groups removes any through-bond activation of the C—Cl bond by the carboxylate sub-
stituent with respect to single-electron reduction. This is reflected by a more than threefold
11
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
decrease of the rate constant of electron capture (to 4.4
×
108 M−1s−1, even lower than for
•−
1-chloropropane)[7] compared to ClAc, which diminishes the fraction of
e
used produc-
aq
tively. By doubling the catalyst concentration to 100 µM (1 mole%), however, we were able to
convert ClPr to propionate with a yield of 90 % (Figure 5b), with the only cost incurred being
a lower TON. This is consistent with our recent experiments on syntheses with a laser-driven
•−
•−
catalytic
e
source, where we found
e
scavenging by HAsc− to be the underlying reason
aq
aq
for a trade-off between maximizing either TON or substrate conversion.[9]
In contrast to aliphatic chlorides, chloroarenes lose halogen in a consecutive mecha-
•−
nism.[33] Their radical anions resulting from
e
capture live long enough for character-
aq
ization before they cleave to give the chloride ion and a
σ radical. For the monochlori-
nated benzoates o-ClBz
,
m-ClBz, and p-ClBz, the cleavage rate constants differ widely
•−
(8
×
107 s−1, 1.2
×
106 s−1, and 4
×
107 s−1),[33,34] whereas the rate constants of
e
aq
capture lie within a much more narrow range (1•.4−
×
109 M−1s−1, 4.8
×
109 M−1s−1, and
6
×
109 M−1s−1)[33]. Progressive scavenging of eaq by the reaction product benzoate (rate
constant, 3.2
×
109 M−1s−1)[7] complicates the kinetics but does not render the reaction
self-limiting because the benzoate radical anion is thermodynamically capable of reducing
the chlorinated substrates, which are less electron-rich than is benzoate.
We determined the dehalogenation yields given in Scheme 1 by the NMR spectra (SI–5.1),
and in parallel also by measuring the chloride concentration, as in Section 2.1.2. Their trend
•−
reflects the interplay of
e
capture and radical-ion stability. Concentrating on the chlorine
aq
loss rather than on the stable end product benzoate is justified because it is equivalent to
quantifying the formation of the highly reactive carboxyphenyl radicals, which can be used
for cross coupling (see also below, Section 2.2.3).[10] The trapping of these radicals by a
coupling product or by a hydrogen donor is strongly dependent on the concentration of
the intercepting agent and on the structure of the substrate. Not unexpected for o-ClBz
the steric hindrance by the neighbouring carboxylate strongly suppresses additions of its
carboxyphenyl radical to surplus substrate such that the only product is benzoate (see the
spectrum in SI–5.1). With the other two isomers, the benzoate formation can be maximized
by raising the concentration of the hydrogen donor HAsc− through lowering the pH to 11.6,
concomitant with an increase of the ascorbate weigh•t-−in concentration such as to keep
,
constant the Asc2− concentration, and thus the TON(eaq ); we have already demonstrated
•− [9]
the feasibility of this strategy in our previous investigation with laser-generated e
.
aq
•−
As the other examples of Scheme 1 show, the impressive reducing power of
e
makes
aq
these arene dehalogenations equally feasible in the case of 4-fluorobenzoate and 4-chloro-
phenylacetate. The latter compound no longer experiences the slight activation with respect
•−
to
e
capture that is caused by the direct attachment of the carboxylate substituent to the
aq
aromatic ring; to compensate for this, we had to halve the substrate concentration.
2.2.2 Hydrogenations of carbon–carbon double bonds
•−
Olefinic double bonds are highly susceptible to attack by
fumarate Fu and cinnamate Ci are 7.5
e
(e.g., the rate constants for
×
aq
×
109 M−1s−1 and 1.4
1010 M−1s−1).[7] Hence,
the transformations of this Section are already complete within 6 h; and the yields of
12
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
X
H
RuDmb, SDS, Asc2−
520 nm, 14 h, H2O
R
R
H
H
H
O
O
O
H
O−
O−
O−
O−
O
(95 %, 86 %*)
Scheme 1: Dehalogenation of aryl halides by e
(58 %)
(76 %)
(51 %)
•−
produced with an LED. Experimental conditions, 100 µM RuDmb and 75 mM Asc2−
aq
in 50 mM aqueous SDS at pH 12.7; substrate concentrations, 5 mM (4-chlorophenylacetate) and 10 mM (all other compounds). The
dehalogenation yields specified in parentheses pertain to the substrate with X in the position indicated by the orange H in the substitution
product. Value with asterisk, X = F; all others, X = Cl. Further explanation, see text.
hydrogenated products are very good (Scheme 2 and SI–5.2). By control experiments with
laser flash photolysis as in the right half of Figure 2a, but in the presence of the respective
•−
substrate, we ascertained that these reactions rely on the reducing power of
be effected by OER itself.
e
and cannot
aq
R
R
RuDmb, SDS, Asc2−
520 nm, 6 h, H2O
O
O
O−
O−
Cl
−O
S
O
O
O
O
O
O−
O−
O−
O−
(100 %)
Experimental conditions, 100 µM RuDmb and 75 mM Asc2− in 50 mM
(95 %)
(85 %)
(80 %)
•−
Scheme 2: Olefin hydrogenations via LED-generated e
.
aq
aqueous SDS at pH 12.7; substrate concentrations, 10 mM; yields given in parentheses. Further explanation, see text.
An example for the simultaneous presence of two substructures potentially transformable
•−
by
e
is provided by p-chlorocin•n−amate ClCi. Because the carbon–carbon double bond
aq
reacts much more readily with eaq than does an arylic carbon–chlorine bond, we antic-
ipated that the double bond can be selectively hydrogenated while leaving unchanged
the chlorinated ring; and Figure 6 demonstrates that this is indeed the case. As the only
requirement, the illumination should not continue for too long after the olefinic substrate
has been consumed, but this is not a crucial point as the initial product with an aliphatic
•−
side chain has about the same low reactivity towards
(compare Figure 5).
e
as has 4-chlorophenylacetate
aq
In analogy to ClAc (above, Figure 5a), deuterations of the olefinic bonds in ClCi and in its
unsubstituted parent cinnamate are possible (top trace of Figure 6, and SI–5.2) simply by
carrying out the reactions in D2O. This is established by the aliphatic proton signals of the
products in this medium, which exhibit the required halving of the integrals, a minute upfield
13
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
1
D2O
H2O
0
1
0
1
0
7.6 7.4 7.2
6.4 2.8 2.6 2.4
δ / ppm
Figure 6: Addition of hydrogen or deuterium to the olefinic double bond of p-chlorocinnamate (10 mM) upon illumination with a green LED;
catalytic system, 100 µM RuDmb and 75 mM Asc2− in 50 mM aqueous SDS at pH 12.7; illumination duration, 6 h. Starting compound,
cyan (aromatic protons at 7.47 ppm, d, and 7.32 ppm, d; olefinic protons at 7.24 ppm, d, and 6.39 ppm, d); product 3-(4-chlorophenyl)-
propionate, orange (aromatic protons at 7.23 ppm, d, and 7.14 ppm, d; aliphatic protons in the protiated compound at 2.76 ppm, t, and
2.36 ppm, t; and in the deuterated compound at 2.73 ppm, d, and 2.33 ppm, d, both broadened by unresolved couplings to geminal and
vicinal D). Bottom trace, 1H-NMR spectrum before the illumination; centre and top traces, after the illumination in H2O and D2O. The
vertical scales are identical. Further explanation, see text.
shift, and doublet patterns with the same coupling constant as in the protiated product;
in contrast to ClAc, however, the couplings to deuterium cannot be resolved. We ascribe
the somewhat lower yields for the additions of deuterium (ClCi, 80 %; cinnamate, 73 %)
compared to those of hydrogen (Scheme 2) to the isotope effect for abstracting a deuterium
or a hydrogen atom from the ascorbate monoanion. The slower abstraction in the case
of deuterium increases the chance of deactivating the intermediate radical anion by back
electron to the catalyst, that is, nonproductively.
2.2.3 Carbon–carbon bond formations
In line with expectation, and also corroborated by several of the examples in the preceding
Sections, the final products and their yields reflect the competition between the different
trapping reagents in the sample for the intermediate radicals or radical anions. The out-
standing hydrogen-donating properties of HAsc− — whose presence in mM quantities is
•−
unavoidable when Asc2− is the sacrificial electron donor in the catalytic
e
sources, unless
aq
the already strongly basic pH is raised to impractical values — causes a natural bias towards
substitution by, or addition of, hydrogen. As we recently reported,[10] the replacement of
ascorbate by urate can help circumvent this problem for the catalyst RuBpy; but this solution
is not feasible with the other catalysts of this work because urate quenches their MLCT
states much too inefficiently.
Here, we explore two other approaches that exploit the high activity of the catalyst RuDmb
With the first, we generate relatively stable radical anions in amounts sufficient for their
dimerization to become competitive. With the second, we make use of the leeway this highly
efficient catalyst provides us with, namely, by decreasing the Asc2− concentration such that
.
•−
the
e
source still works but radical interception by HAsc− is minimized and the balance
aq
can be swung by a high concentration of another trapping agent.
•−
Acetophenone AcPh illustrates the first strategy. This substrate captures
e
in a diffusion
aq
14
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
controlled process (rate constant, 2.4
×
1010 M−1s−1)[7] but is inert towards OER of our
ruthenium complexes[35] for thermodynamical reasons (the standard redox potential of
AcPh is
−
1.8 V;[36] i.e., more negative than that of RuDmb by 0.73 V, compare Table 1).
Once formed, the radical anions quantitatively undergo self-coupling.[37] Figure 7 evidences
•−
that our catalytic
e
source affords 2,3-diphenyl-2,3-butanediol as the only product and
aq
in excellent yield (see Scheme 3; and SI–5.3 shows of the successful upscaling of this
experiment to ten times the volume). As the only modification of our standard procedure, we
had to lower the SDS concentration to 30 % of its usual value because AcPh associates with
•−
the micelles, which would shield it from attack by
e
. In accordance with the literature,[37] we
aq
found a dl:meso ratio of 54:46, confirming that the dimerization of the radical anions exhibits
no significant diastereoselectivity. The value of our method lies in obviating the necessity
of an expensive iridium catalyst for a pinacol coupling through photoredox catalysis;[35] our
two-photon approach yields the same result with a much more affordable ruthenium catalyst,
but can still be done with an LED.
1
0
1
x 0.2
0
7.8
7.4
7.0
δ / ppm
2.5
1.4
•−
•−
Figure 7: Homocoupling of acetophenone (10 mM), as effected by producing e
with an LED. Catalytic e
source, 100 µM RuDmb
aq
aq
and 75 mM Asc2− in 15 mM aqueous SDS at pH 12.7. Bottom trace, 1H-NMR spectrum before the illumination; top trace, after illumina-
tion for 14 h. The vertical scales are identical. Starting compound, cyan (aromatic protons at 7.89 ppm, d, 7.55 ppm, t, and 7.43 ppm, t;
aliphatic protons at 2.55 ppm, s, signal multiplied by 0.2 in the spectrum before irradiation); product 2,3-diphenyl-2,3-butanediol, orange
(superposition of dl and meso forms for the aromatic protons at 7.09 ppm, d, 7.03 ppm, t, and 6.93 ppm, d; aliphatic protons of the dl
form at 1.52 ppm, s; of the meso form, 1.38 ppm, s). The quintet of the SDS protons is shifted from 1.47 ppm in the starting spectrum to
1.47 ppm in the final spectrum because acetophenone and the diol associate differently with the micelles. Further explanation, see text.
RuDmb, SDS, Asc2−
520 nm, 14 h, H2O
Substrate + Trapping reagent
Product
O
HO
O
O
O−
N
1
2
3
(88 %)
(37 %)
(62 %)
O
−O
−O
OH
O−
•−
Scheme 3: Coupling reactions induced by e
that are accessed through LED irradiation. Experimental conditions: 100 µM RuDmb
aq
and 75 mM / 30 mM (1 / 2 and 3) Asc2− in 15 mM / 50 mM (1 / 2 and 3) aqueous SDS at pH 12.7; 10 mM / 5 mM (1 / 2 and 3) substrate;
200 mM trapping reagent (2 and 3 only). The yields shown in parentheses were determined from the 1H-NMR spectra (Figure 7 and
SI–5.3). Further explanation, see text.
The examples of our second s•tr−ategy employ the reactive carboxyphenyl radicals procured
in high yields by dissociative eaq attachment to p-ClBz (Scheme 1). To push back their
interception by hydrogen donation, we lowered the ascorbate weight-in concentration to
15
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
40 % of the standard concentration, which made it possible to add the desired trapping
reagent in seventy-fold excess over HAsc− at pH 12.7. As trapping reagents that avoid
the formation of several cross-coupling products, we selected mesitylacetate MesAc and
N-methylpyrrole-2-carboxylate NMPCA. With the former, all ring positions except meta are
occupied; with the latter, it is well known that the addition of radicals to pyrroles almost
exclusively occurs at the ring position adjacent to the nitrogen atom.[38–43] Proof of principle
for the viability of our approach is obtained from the NMR spectra in SI–5.3. In the case
of MesAc, the deceleration of the traget reaction by the overcrowded trapping reagent only
allows a moderate yield of the coupling product, with benzoate remaining the main product;
however, the sterically less demanding NMPCA turns the table and approximately reverses
the product distribution (Scheme 3).
•−
On the basis of these results it can be concluded that for cross couplings via
e
a catalyst
aq
should be optimized with respect to both intrinsic activity and stability. RuDmb represents a
step in that direction. The better the catalyst fares with respect to these properties, the more
it will permit reducing the amount of sacrificial donor until the unwanted hydrogen abstraction
by the substrate radicals becomes a minor side channel. This reduction decreases the
•−
e
yield, hence is tantamount to systematic catalyst poisoning; but as long as the catalyst
aq
survives the proportionally larger number of excitations that are needed for the same
turnover, the associated photophysical losses squander only photons but do not consume
any chemicals.
3 Conclusions
The kinetic investigations of this work have juxtaposed a microscopic and a macroscopic
•−
strategy for characterizing our LED-driven catalytic
e
sources. Laser flash photolysis
aq
monitors the individual steps on the timescales associated with them, which typically
means s to ms; and we have shown that in its more sophisticated variant of two-pulse
µ
experiments it is capable of observing each key process of our complex mechanisms in
isolation. Preparative LED kinetics with detection through a coupled transformation caused
•−
by e captures the time-integrated effect of all steps lumped together.
aq
Both approaches have their specific strengths and weaknesses. From a practical per-
spective, the laser method needs special equipment as well as expertise and is very
time-consuming, whereas the LED method uses the same setup as the actual syntheses,
and the experiments are easy and fast to perform. On a fundamental level, the laser method
provides direct and detailed information on practically all processes that determine the
activity of a catalyst, which the LED method can only supply indirectly (through Equation 3)
and with much more restricted possibilities of separating the influences.
This would make laser flash photolysis far superior were it not for two inherent limitations.
First, it cannot properly characterize slow processes that play only a marginal role on
fast timescales but may no longer be ignored during preparative illumination, such as the
secondary chemistry of the radicals derived from the sacrificial donor. Second, it is blind
to effects that are below the detection limit in pulsed experiments, where each absorbing
16
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
molecule is only excited once per pulse, but build up to large sum totals over the course of a
preparative illumination, where the same molecule experiences a vast number of excitations:
the catalyst decay falls into this category. In contrast, the LED kinetics is sensitive to both
issues, and directly characterizes the second.
In summary, both methods overlap only to some extent and neither provides a complete
picture; laser flash photolysis is indispensable when targeted improvement of the catalyst
activity by structural modification is an aim, as in this work, but must always be complemented
by the LED kinetics to determine the catalyst decay rate; when the focus is on optimizing
the overall performance for reactions performed with the output of the catalytic system,
investigating the LED kinetics as a stand-alone is both necessary and sufficient; and, in any
case, the LED kinetics supplies much more detailed information on a transformation than
does a single end-point measurement.
The remarkable fact that a low-power light source such as an LED can drive a two-photon
process on a benchtop scale hinges on the feature that the chemically cached energy of the
first photon ”leaks” only through a second-order process with respect to the intermediate.
Utilizing the energy added by the second photon is by no means restricted to electron
ejection: if the excited storage medium were instead quenched by an external species, the
kinetics of product formation would still obey Equation
3 provided that ϕion is replaced by the
Stern–Volmer term pertaining to this quenching. Hence, all the dependences on process
variables derived herein (e.g., on the light intensity; see Figure 4b) and previously[8,10] are of
complete generality for any such two-photon processes.
•−
The applications of
e
that we have presented illustrate that this ”super reductant” greatly
aq
extends th•e−range of substrates that can be transformed. Owing to the composition of our
catalytic eaq sources, the presence of mM concentrations of a good hydrogen donor is
unavoidable. Through competition, this decreases the likelihood of self- and cross-couplings,
although we have presented examples. However, substitutions of chlorine or fluorine by
hydrogen as well as hydrogenations of double bonds are within easy reach. As a particular
forte, the substrates can be deuterated selectively and with very high isotopic purity; and
this feat can be performed in an extremely simple way: all that it takes is the replacement of
the solvent H2O by D2O.
17
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
References
[1] J. M. R. Narayanam and C. R. J. Stephenson, Chem. Soc. Rev., 2011, 40, 102–113.
[2] C. K. Prier, D. A. Rankic and D. W. C. MacMillan, Chem. Rev., 2013, 113, 5322–5363.
[3] Chemical Photocatalysis, ed. B. König, DeGruyter, Berlin, 2013.
[4] D. Ravelli, M. Fagnoni and A. Albini, Chem. Soc. Rev., 2013, 42, 97–113.
[5] M. Reckenthäler and A. G. Griesbeck, Adv. Synth. Catal., 2013, 355, 2727–2744.
[6] N. A. Romero and D. A. Nicewicz, Chem. Rev., 2016, 116, 10075–10166.
[7] G. V. Buxton, C. L. Greenstock, W. P. Heiman and A. B. Ross, J. Phys. Chem. Ref.
Data, 1988, 17, 513–886.
[8] R. Naumann, F. Lehmann and M. Goez, Angew. Chem. Int. Ed., 2018, 57, 1078–1081.
[9] R. Naumann, C. Kerzig and M. Goez, Chem. Sci., 2017, 8, 7510–7520.
[10] R. Naumann and M. Goez, Chem. Eur. J., 2018, DOI: 10.1002/chem.201800626.
[11] M. Goez, C. Kerzig and R. Naumann, Angew. Chem. Int. Ed., 2014, 53, 9914–9916.
[12] M. Brautzsch, C. Kerzig and M. Goez, Green Chem., 2016, 18, 4761–4771.
[13] T. Kohlmann, R. Naumann, C. Kerzig and M. Goez, Photochem. Photobiol. Sci., 2017,
16, 185–192.
[14] T. Kohlmann, R. Naumann, C. Kerzig and M. Goez, Photochem. Photobiol. Sci., 2017,
16, 1613–1622.
[15] S. Campagna, F. Puntoriero, F. Nastasi, G. Bergamini and V. Balzani, Top. Curr. Chem.,
2007, 280, 117–214.
[16] S. P. Pitre, C. D. McTiernan and J. C. Scaiano, Acc. Chem. Res., 2016, 49, 1320–1330.
[17] R. Martinez-Haya, M. A. Miranda and M. L. Marin, Eur. J. Org. Chem., 2017, 2164–
2169.
[18] M. Marchini, G. Bergamini, P. G. Cozzi, P. Ceroni and V. Balzani, Angew. Chem. Int.
Ed., 2017, 56, 2–4.
[19] A. Juris, V. Balzani, P. Belser and A. von Zelewsky, Helv. Chim. Acta, 1981, 64, 2175–
2182.
[20] H. B. Ross, M. Boldaji, D. P. Rillema, C. B. Blanton and R. P. White, Inorg. Chem., 1989,
28, 1013–1021.
[21] M. B. Davies, J. Austin and D. A. Partridge, Vitamin
C: Its Chemistry and Biochemistry,
The Royal Society of Chemistry, Cambridge, 1st edn., 1991.
[22] M. Goez, D. von Ramin-Marro, M. Schiewek and M. H. O. Musa, Z. Phys. Chem.
(Muenchen, Ger.), 2014, 228, 193–207.
[23] R. McAlpine, M. Cocivera and H. Chen, Can. J. Chem., 1973, 51, 1682–1686.
[24] M. Goez and V. Zubarev, Chem. Phys., 2000, 256, 107–116.
[25] A. Juris, V. Balzani, F. Barigelletti, S. Campagna, P. Belser and A. von Zelewsky, Coord.
Chem. Rev., 1988, 84, 85–277.
18
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
[26] M. Rossberg, W. Lendle, G. Pfleiderer, A. Tögel, E.-L. Dreher, E. Langer, H. Rassaerts,
P. Kleinschmidt, H. Strack, R. Cook, U. Beck, K.-A. Lipper, T. R. Torkelson, E. Löser,
K. K. Beutel and T. Mann, in Ullmann’s Encyclopedia of Industrial Chemistry, Wiley-VCH
Verlag GmbH & Co. KGaA, 2000.
[27] W. J. Dressick, J. Cline, J. N. Demas and B. A. DeGraff, J. Am. Chem. Soc., 1986, 108
,
7567–7574.
[28] R. P. Yu, D. Hesk, N. Rivera, I. Pelczer and P. J. Chirik, Nature, 2016, 529, 195–199.
[29] P. H. Allen, M. J. Hickey, L. P. Kingston and D. J. Wilkinson, J. Label. Compd. Rad.,
2010, 53, 731–738.
[30] C. S. Elmore and R. A. Bragg, Bioorg. Med. Chem. Lett., 2015, 25, 167–171.
[31] Y. Y. Loh, K. Nagao, A. J. Hoover, D. Hesk, N. R. Rivera, S. L. Colletti, I. W. Davies and
D. W. C. MacMillan, Science, 2017, 358, 1182–1187.
[32] X. Guo and O. S. Wenger, Angew. Chem. Int. Ed., 2017, 57, 2469–2473.
[33] R. Zona, S. Solar, N. Getoff, K. Sehested and J. Holcman, Radiat. Phys. Chem., 2008,
77, 162–168.
[34] V. Konovalov, A. Raitsimring, Y. Tsvetkov and I. Bilkis, Chem. Phys. Lett., 1989, 157
,
257–264.
[35] M. Nakajima, E. Fava, S. Loescher, Z. Jiang and M. Rueping, Angew. Chem. Int. Ed.,
2015, 54, 8828–8832.
[36] G. J. Gores, C. E. Koeppe and D. E. Bartak, J. Org. Chem., 1979, 44, 380–385.
[37] S. Okamoto, K. Kojiyama, H. Tsujioka and A. Sudo, Chem. Commun., 2016, 52
,
11339–11342.
[38] Y. M. Osornio, R. Cruz-Almanza, V. Jiménez-Montaño and L. D. Miranda, Chem.
Commun., 2003, 2316–2317.
[39] O. Guadarrama-Morales, F. M
4515–4518.
éndez and L. D. Miranda, Tetrahedron Lett., 2007, 48,
[40] C. Iuga, S. O. Uribe, L. D. Miranda and A. Vivier-Bunge, Int. J. Quantum Chem., 2010,
110, 697–705.
[41] L. Marzo, I. Ghosh, F. Esteban and B. König, ACS Catal., 2016, 6, 6780–6784.
[42] J. I. Bardagi, I. Ghosh, M. Schmalzbauer, T. Ghosh and B. König, Eur. J. Org. Chem.,
2018, 34–40.
[43] M. Neumeier, D. Sampedro, M. Májek, V. A. de la Peña O’Shea, A. Jacobi von Wangelin
and R. Pérez-Ruiz, Chem. Eur. J., 2018, 24, 105–108.
19
This article is protected by copyright. All rights reserved.

10.1002/chem.201801955
Chemistry - A European Journal
Table of contents entry
Add more life to years ! A ligand substitution retains the light-stability of a micellized
metal-complex Cat but boosts the activity of its ionizable form CAT for producing hydrated
•−
electrons
e
with a green LED, which are used for reductive transformations of otherwise
aq
unreactive substrates on a laboratory scale.
CAT
stability
activity
50
20
0
0
0
12
0
4
t / h
t / h
•−
e
Cat
aq
20
This article is protected by copyright. All rights reserved.