Automated and Continuous-Flow Platform to Analyze Semiconductor-Metal Complex Hybrid Systems for Photocatalytic CO2Reduction

Article abstract of DOI:10.1021/acscatal.1c02921

Sunlight-driven CO₂reduction is increasingly considered as a promising approach to contribute toward a carbon-neutral fuel cycle, but most photocatalyst systems are currently studied individually under batch conditions with manual, labor-intensive analytical procedures. Here, we present the advantages of a continuous-flow setup to study photocatalytic CO₂to CO reduction systems, which also benefits from aspects of automation (using programmed in-line gas quantification of multiple samples in parallel). The capabilities of the methodology are demonstrated using a state-of-the-art light absorber platform based on ZnSe quantum dots (QDs) in combination with a series of molecular co-catalysts based on Ni and Co for visible-light-driven CO₂reduction in aqueous ascorbate solution. A newly synthesized Co-tetraphenylporphyrin featuring three sulfonate groups and one amine group (Co(tppS3N1)) is identified to exhibit a benchmark photocatalytic activity (18.6 μmol of CO, 79.7 mmol of CO g_ZnSe^-1, TON_Co(CO) of 619, external quantum efficiency (EQE) >5%). The utility of our methodology is further shown by applying the setup to study the photocatalyst systems under lower light intensities, low CO₂concentration, and aerobic conditions, which impact the photocatalytic activity and selectivity. Overall, this work reports an improved methodology for studying photocatalytic CO₂reduction alongside advancing the understanding of QD molecular co-catalyst hybrids using ZnSe QDs as a versatile light absorber based on earth-abundant components that operate under fully aqueous conditions.

Full text of DOI:10.1021/acscatal.1c02921

pubs.acs.org/acscatalysis
Research Article
Automated and Continuous-Flow Platform to Analyze
Semiconductor−Metal Complex Hybrid Systems for Photocatalytic
CO2 Reduction
Constantin D. Sahm, Geani M. Ucoski, Souvik Roy, and Erwin Reisner*
Cite This: ACS Catal. 2021, 11, 11266−11277
Read Online
sı
*
Metrics & More
Article Recommendations
Supporting Information
ACCESS
ABSTRACT: Sunlight-driven CO2 reduction is increasingly
considered as a promising approach to contribute toward a
carbon-neutral fuel cycle, but most photocatalyst systems are
currently studied individually under batch conditions with manual,
labor-intensive analytical procedures. Here, we present the
advantages of a continuous-flow setup to study photocatalytic
CO2 to CO reduction systems, which also benefits from aspects of
automation (using programmed in-line gas quantification of
multiple samples in parallel). The capabilities of the methodology
are demonstrated using a state-of-the-art light absorber platform
based on ZnSe quantum dots (QDs) in combination with a series
of molecular co-catalysts based on Ni and Co for visible-light-
driven CO2 reduction in aqueous ascorbate solution. A newly
synthesized Co-tetraphenylporphyrin featuring three sulfonate groups and one amine group (Co(tppS3N1)) is identified to exhibit a
benchmark photocatalytic activity (18.6 μmol of CO, 79.7 mmol of CO gZnSe−1, TONCo (CO) of 619, external quantum efficiency
(EQE) >5%). The utility of our methodology is further shown by applying the setup to study the photocatalyst systems under lower
light intensities, low CO2 concentration, and aerobic conditions, which impact the photocatalytic activity and selectivity. Overall, this
work reports an improved methodology for studying photocatalytic CO2 reduction alongside advancing the understanding of QD
molecular co-catalyst hybrids using ZnSe QDs as a versatile light absorber based on earth-abundant components that operate under
fully aqueous conditions.
KEYWORDS: artificial photosynthesis, CO2 reduction, photocatalysis, quantum dots, automation
experimental conditions, which results in high numbers of
permutations and therefore individual experiments.
INTRODUCTION
■
Photocatalytic solar fuel systems are most commonly studied in
batch reactors with gaseous reaction products (H2, CO, CH4)
accumulating in the reactor headspace.1−5 While such batch
systems are easy to set up and offer a high product sensitivity,
they possess a number of drawbacks such as buildup of gaseous
reaction products, which can lead to overpressure and
subsequently aid product leakage, accumulating products
reducing photocatalytic efficiency by undergoing back-reactions
or catalyst poisoning as well as the requirement of labor-
intensive manual sampling and gas analysis. Some of the most
successful photocatalytic systems for CO2 reduction comprise
two distinct functional materials, a solid-state light absorber
combined with an additional catalyst based on transition-metal
complexes.4,6,7 Such hybrid photocatalysts are promising
because they combine the favorable photophysics and stability
of solid semiconductors with the selectivity of (molecular)
catalysts, herein referred to as co-catalyst. Within the search for
new hybrid photocatalysts, it is necessary to study large libraries
of light absorbers and co-catalysts as well as optimize
Hence, high-throughput techniques utilizing continuous-flow
methodologies and an increasing degree of automation with
reliable product analysis are desirable for the development of
next-generation photocatalysts and optimization of conditions.
While continuous-flow photocatalysis is used in the synthesis of
organic pharmaceuticals,8 photodegradation of organic com-
pounds,9 and photoredox catalysis,10 there has been relatively
little activity in the production of solar fuels.11 Some continuous-
flow photocatalytic CO2 reduction systems have been reported
previously, but they were generally limited to gas-phase
reactions employing low-yield titania photocatalysts without
Received: June 29, 2021
Revised: August 4, 2021
https://doi.org/10.1021/acscatal.1c02921
© XXXX American Chemical Society
ACS Catal. 2021, 11, 11266−11277
11266

ACS Catalysis
pubs.acs.org/acscatalysis
Research Article
Figure 1. Continuous-flow photocatalytic CO2 reduction with automated in-line gas chromatography. (A) Schematic diagram of the automated
continuous-flow setup. (B) Schematic representation of the photocatalyst system consisting of ligand-free ZnSe-BF4 QDs (yellow sphere; BF4− anions
are omitted for clarity) in combination with molecular co-catalysts for visible-light-driven CO2 to CO reduction in water using ascorbic acid (AA) as
the sacrificial electron donor. (C) Overview of the molecular co-catalysts studied in depth during this work.
consideration to match light absorbers and co-catalysts in the
assembly of photocatalytic hybrid architectures.12−14 Progress
has also been made in high-throughput screening of molecular
dyes with H2 evolution co-catalysts, but the detection method-
ology was limited to colorimetric methods,15 whereas a different
study utilized an automatic H2 monitoring system consisting of
mass flow controller (MFC) and photoreactor directly coupled
to a gas chromatograph (GC).16 An automated robotic station
for the screening of organic polymer photocatalysts for
photocatalytic H2 evolution was presented recently, but this
system was also based on a batch operation with the products
accumulating in the headspace.17
Here, we present a continuous-flow methodology to study
photocatalytic CO2 to CO reduction alongside H2 evolution in
aqueous solution. The developed setup uses automated in-line
gas chromatography of multiple samples in parallel (Figure 1),
and features a high measurement frequency (up to 0.25 min−1)
with a remarkable sensitivity (10−12 mol of CO s−1). The
capabilities of the continuous-flow system are investigated for
semiconductor-metal complex hybrid photocatalysts, using
ZnSe quantum dots (QDs) as a versatile light-absorbing
platform (Figure 1B). We show that ZnSe QDs can drive a
range of molecular CO2 reduction co-catalysts based on metal
porphyrins, cyclams, and quaterpyridines (Figure 1C) in
aqueous ascorbate solution. Finally, we illustrate the versatility
of the system by studying the influence of the light intensity, the
effect of exposure to O2 on the hybrid photocatalysts, and the
consequence of varying the composition of feed gas from pure
CO2 to dilute CO2, which paves the way to study simulated real-
life conditions.
RESULTS AND DISCUSSION
■
Method Development. The continuous-flow setup was
assembled as depicted in Figures 1 and 2, and the technical
details can be found in the Experimental Section. Gas cylinders
containing balance gas (e.g., CO2, N2, air) were connected to a
set of three-way valves, which were attached to a set of MFCs
that provide a defined gas stream to each photoreactor. The
MFCs allow for precise control of the flow rates, whereas the
valve enables convenient switching between the desired balance
gas. The outlets of the MFCs were connected to gas saturators
filled with water to humidify the gas stream and then to the
photoreactor with Teflon tubing. The photoreactors (volume
7.9 mL) consist of glass vials containing stirrer bars, sealed with
septa, and filled with photocatalyst solution (3 mL). The gas
11267
https://doi.org/10.1021/acscatal.1c02921
ACS Catal. 2021, 11, 11266−11277

ACS Catalysis
pubs.acs.org/acscatalysis
Research Article
Figure 2. (A) Technical diagram of the continuous-flow setup. Photographs of (B) the photoreactors illuminated by the solar light simulator (three out
of six photoreactors in operation), (C) the mass flow controller (MFC) setup, and (D) the gas chromatograph (GC) equipped with a flow-selection
valve.
stream is routed through the entire vial by bubbling through the
photocatalyst solution. Each sample outlet was connected with
Teflon tubing through a particulate filter and shutoff valve to a
flow-selection valve controlled by the GC. The flow-selection
valve can be programmed to select one sample stream at a time,
purge it through the sample loop, and inject it into the GC, after
which the next sample stream can be selected.
deviating flow rate. The GC consists of two separate columns, a
precolumn and a main column, connected to a high-sensitivity
barrier discharge ionization detector. The precolumn made of a
polymeric material (HayeSep D) separates the (target) gases
(H2, CO, CH4, N2, O2) from CO2 and moisture and blocks the
latter two components from accessing the main column, which is
made from Molsieve. The raw data acquired by the GC provide
This setup enables the analysis of multiple samples in parallel
with a single experimental setup by sampling through every
photoreactor one after the other. The measurement frequency is
thereby determined by the number of samples and the run time
of a GC analysis, which is approximately 4.25 min. In a typical
experiment, six samples (each sample injected every 25.5 min)
were measured in parallel, which was found to be a good balance
between sample throughput and time resolution from GC
analysis. The outlet of the sample loop was connected to a mass
flow meter to verify the set flow rate prior to an experiment and
detect potential leakage. The system is very sensitive to leakage,
and verifying the flow rate prior to starting an experiment is
required to ensure that the setup was correctly assembled. The
automated nature of the system eliminates the human error of
manual syringe injection and leaves a deviation from the set flow
rate due to leaks in the setup as the key origin of error. An
internal standard would not provide an added benefit because its
concentration in the sample loop is unchanged, independent of a
gas (in mol s−1) for each
̇
the momentary product evolution rate n
injection, which is calculated using the formula given in the
Experimental Section.
The flow rate is a key variable that requires careful
consideration because of its interplay with the sample loop
size and the GC’s sensitivity. A high flow rate results in increased
dilution of product and a low flow rate is therefore favorable for
higher product accumulation within the sample loop to ensure
sufficient sensitivity for the GC analysis. On the other hand, very
low flow rates result in unstable gas flows, a delay in product
detection, and difficulty to avoid leakage (which can be observed
by residual air peaks in the GC analysis). To increase the
sensitivity, the sample loop size may be increased from 25 μL to
2 mL, but care must be taken to avoid overloading of the
precolumn with CO2. A flow rate between 3.5 and 4.0 standard
cubic centimeters per minute (sccm) at a sample loop size of 1
mL was found to be the optimum between high sensitivity and a
stable, reproducible gas stream without significant leakage in our
11268
https://doi.org/10.1021/acscatal.1c02921
ACS Catal. 2021, 11, 11266−11277

ACS Catalysis
pubs.acs.org/acscatalysis
Research Article
Figure 3. Survey of continuous-flow photocatalytic CO2 reduction in the presence of ZnSe-BF4 and various co-catalysts Co(qpy), Ni(cycP), and
Co(tppS3N1). (A) H2 evolution rate, (B) CO evolution rate, (C) total amount of evolved H2, and (D) total amount of evolved CO. Conditions: 0.5
μM ZnSe-BF4, 10 μM co-catalyst, 0.1 M AA, pH 4.5 (Co(tppS3N1)), pH 5.5 (Ni(cycP) and Co(qpy)), AM 1.5G, 100 mW cm−2, λ > 400 nm, CO2
(4.0 sccm), 25 °C.
setup with an excellent limit of quantification of 1.26 × 10−12
mol s−1.
fied with either four sulfonate groups (Co(tppS4)) or three
sulfonate groups and one amine (Co(tppS3N1)) (Figure 1C
and Table S1). The latter catalyst is a novel CO2 reduction co-
catalyst and an analogue of the recently reported Co(tppS4),
which was shown to be a benchmark molecular co-catalyst for
photocatalytic CO2 to CO reduction in aqueous solution when
sensitized with a Ru(2,2′-bipyridine)3 photosensitizer.20 We
introduced an amine moiety into Co(tppS3N1) as this strategy
was previously reported to aid coulombic stabilization of CO2
reduction intermediates24 while preserving the general structure
of the molecule.
A standard gas mixture for calibration was used containing
each 2000 ppm H2, CO and CH4 in CO2 as balance gas. The gas
mixture was further diluted with CO2 (by mixing two gas
streams with different flow rates) to reach concentration levels of
2000, 1600, 1200, 800, 400, and 200 ppm, which were used to
generate a calibration curve, as depicted in Figure S1. The slope
of the linear fit of the curve gives the response factor, for a given
sample loop size. The delay of a given gas in the photoreactor to
reach the GC was quantified by manually injecting a defined
amount of gas into the reactor headspace followed by GC
analysis and was found to be <4 min, which is below the time
resolution of the GC analysis (Figure S2).
Survey of ZnSe QDs with Molecular Co-catalysts. We
studied the capabilities of the continuous-flow system for
photocatalysis using state-of-the-art ZnSe QDs as a light
absorber in combination with molecular CO2 reduction co-
catalysts based on transition-metal complexes. ZnSe allows for
utilization of parts of the visible spectrum of sunlight due to its
bulk band gap of 2.7 eV and provides an ample reductive driving
force for common molecular catalysts due to its conduction
band position at approximately −1.4 V vs normal hydrogen
electrode (NHE) (at pH 5.5).18 Previously, ZnSe QDs have
been shown to drive co-catalysts based on the Ni(cyclam)
family,19 and here, we screened a series of molecular co-catalysts
based on earth-abundant metals including porphyrins,20
quaterpyridines,21 terpyridines,3 and phthalocyanines.22,23 See
the Experimental Section for the synthesis and characterization
of ZnSe QDs and the molecular co-catalysts (Figures S3 and
S4). An initial prescreening was conducted to reduce the
number of possible combinations and revealed promising
activity for Co(quaterpyridine) (Co(qpy)), phosphonic acid-
modified Ni(cyclam) (Ni(cycP)), and Co(porphyrins) modi-
The photocatalytic activity of the selected co-catalysts
Co(qpy), Ni(cycP), and Co(tppS3N1) was subsequently
studied with ZnSe QDs under continuous-flow of CO2 (4
sccm) with our system described above in aqueous ascorbic acid
(AA) solution (0.1 M, 3 mL) by irradiating the samples with
UV-filtered simulated solar light (λ > 400 nm, AM 1.5G, 100
mW cm−2; see the Experimental Section for details). The data
processing during continuous flow is described in the following
and exemplarily shown for photocatalytic CO formation with
ZnSe|Ni(cycP) (Figure S5). The raw data acquired from the GC
̇
analysis, ngas, is depicted in Figure S5A and gives the product
evolution rate. Integration of the formation rate over irradiation
time provides the total amount of evolved product. This was
achieved by numerical integration of the formation rate using
the trapezoidal method for each sample individually (Figure
S5B). The three independent replicates of identical conditions
were averaged by calculating the mean and standard deviation
over irradiation time (see the Experimental Section for details).
For visual display, the values of each individual sample are
plotted as transparent scatter (50% transparency), whereas the
mean is represented as a continuous line (compare also to Figure
3). To display the uncertainty of each experimental condition,
the standard deviation is visualized as shaded area surrounding
11269
https://doi.org/10.1021/acscatal.1c02921
ACS Catal. 2021, 11, 11266−11277

ACS Catalysis
pubs.acs.org/acscatalysis
Research Article
the mean (mean standard deviation) where the transparency
is proportional to the standard deviation. The result appears as a
vertical “cloud” of uncertainty and is performed for both product
formation rate and total amount of product (Figure S5C,D).
This example underscores the excellent data quality and
resolution of the data acquired using the continuous-flow
methodology, which significantly exceeds many previously
reported studies on hybrid photocatalysts.3,19 Additionally, the
automated sampling technique enables significantly higher
throughput of light absorber−catalyst combinations compared
to manual sampling because human input is only required for
setting up the photoreactors and the tubing, and hereafter, the
system operates independently.
ZnSe-BF4 QDs in the absence of a co-catalyst are highly active
toward the H2 evolution reaction by reducing aqueous protons
(Figure S6) and only form traces of CO from CO2 as reported
previously.25,26 Significant amounts of CO are only produced in
the presence of co-catalysts, and the best-performing co-catalyst
(Co(qpy), Ni(cycP), and Co(tppS3N1)) of the three catalyst
families (quaterpyridine, cyclam, porphyrin) were analyzed in
detail (see Figure 3 and Table S2). All QD-co-catalyst hybrids
were optimized with respect to co-catalyst loading (optimum 20
molco‑catalyst molQD−1) and pH (4.5 for Co(tppS3N1) and 5.5 for
Ni(cycP) and Co(qpy), Figure S7). ZnSe|Co(qpy) reaches the
maximum CO production rate after approximately 100 min,
after which the CO evolution activity declines rapidly and
mostly H2 is generated (Figure 3). In contrast, the
ZnSe|Ni(cycP) system featuring a molecular co-catalyst with
phosphonate-anchoring group displays much better longevity,
which may be attributed to catalyst immobilization. There is a
small induction period for both H2 and CO evolution, but the
generation of CO remains relatively steady over the course of
hours and only declines slowly, presumably due to a high
quantity of attached/undamaged Ni(cycP) on the surface.
Additionally, the initial high rate of Co(qpy) suggests it being
the more efficient co-catalyst but less stable under turnover
conditions. The time for both catalyst systems needed to reach
peak turnover is likely caused by the photophysics of the QDs
because it is also visible when the unfunctionalized QDs are used
for proton reduction in the absence of a molecular co-catalyst
(Figure S6).
CO originates from CO2 and no other carbon sources contribute
to product generation (Figure S8). No other reaction products
(e.g., methane, methanol, formate) were detected via GC and
1H NMR spectroscopy, respectively. Only traces of CO and
mostly H2 were produced in the absence of a co-catalyst, and no
gaseous products were detected in a series of deletional control
experiments in the dark or absence of QDs or AA, demonstrating
that all components of the photocatalytic system are required
(Table S3).
The attachment of the co-catalysts on the QD surface was
quantified via UV−vis spectroscopy by comparing the
absorption profile of the co-catalyst in solution with the profile
of the supernatant after exposure to QDs and subsequent
removal of the QDs via centrifugation (Figure S9 and Table S4).
The absorption spectra of Co(qpy) are nearly unchanged,
confirming its diffusional nature and lack of immobilization. The
attachment of Ni(cycP) to ZnSe QDs was previously quantified
using ion-coupled plasma optical emission spectroscopy as
Ni(cycP) does not feature a distinct UV−vis absorption profile
and resulted in immobilization of 7.8% (1.6 molco‑cat
molQD−1).19 Co(tppS3N1) exhibits a drastically lowered
absorption intensity (based on the Soret band at 420 nm),
which was translated to an attachment of 84% of the amount of
employed co-catalyst (16.8 molco‑cat molQD−1). Electrokinetic ζ
potential measurements confirm that ZnSe-BF4 exhibits a
positive potential (+20 mV), which rationalizes the attachment
of negatively charged Co(tppS3N1). This electrostatic assembly
also shows a higher attachment than Ni(cycP) with its dedicated
phosphonate surface-bound anchoring group. As a result of the
different anchoring strategies, the photocatalytic activity of the
different catalysts is not strictly comparable, and the selection of
catalysts aims to demonstrate the versatility of the continuous-
flow methodology as well as ZnSe QDs as the light-absorbing
platform.
Transmission electron microscopy images of the particles
after photocatalysis show the formation of some larger
aggregates but retain a nanocrystalline morphology (Figure
S10A). The stability of the particles was further confirmed by
collecting UV−vis spectra after irradiation of the particles.
Increased scattering implies particle agglomeration, but the
absorption onset of the ZnSe-BF4 QDs retains unchanged
compared to the stock solution, which indicates an unchanged
size (Figure S10B). While those analyses do not contain
information about the molecular co-catalyst, they demonstrate
that the ZnSe particles are stable (within the time frame of the
experiments conducted) and retain their key morphological and
photophysical properties. Hence, we conclude that the declining
photocatalytic activity observed for all QD-co-catalyst hybrids is
mainly determined by accumulation of noninnocent oxidation
products (DHA),27,28 a reduction in overall surface area as a
result of agglomeration processes and/or loss and decom-
position of the molecular co-catalyst.
In contrast, ZnSe|Co(tppS3N1) exhibits a significantly longer
induction period for both proton and CO2 reduction. This is a
surprising finding and suggests a strong anchoring of the co-
catalyst to the QD surface as otherwise more H2 evolution would
be expected in the initial irradiation time of the experiment due
to the high activity of the bare QDs toward proton reduction
(Figure S6; see below for a more detailed analysis of the
induction period). After approximately 120 min, the hybrid
photocatalyst starts to evolve both H2 and CO with a local
maximum rate at approximately 260 min and an absolute
maximum after 500 min. It is likely that the evolved H2
originates not from the ZnSe surface but from the porphyrin
co-catalyst itself as H2 evolution activity is known for
Co(tppS4).20 The CO evolution rate peaks eventually and
In-Depth Study of ZnSe|Co(tppS3N1). We subsequently
explored the origin of the distinct induction period of
Co(tppS3N1) during photocatalysis by comparing it with
Co(tppS4). While Co(tppS4) does not exhibit such an
induction period, Co(tppS3N1) eventually evolves CO at
higher rates and outperforms Co(tppS4) (Figure S11). After
1000 min of continuous irradiation, ZnSe|Co(tppS3N1)
produces 18.6 3.3 μmol of CO (79.7 14.0 mmol of CO
gZnSe−1, TONCo = 619), whereas ZnSe|Co(tppS4) generates 7.0
surpasses the two other catalyst assemblies, producing 18.6
−1
3.3 μmol of CO (79.7
14.0 mmol of CO gZnSe
,
TONCo(tppS3N1) (CO) of 619) after 1000 min of irradiation. In
comparison, ZnSe|Co(qpy) and ZnSe|Ni(cycP) reach a
TONco‑cat (CO) of 71 and 116, respectively, after the same
irradiation time.
All QD-co-catalyst assemblies evolve solely 13CO under an
atmosphere of isotopically labeled 13CO2, supporting that the
0.7 μmol of CO (30.1 3.2 mmol of CO gZnSe−1, TONCo
=
233). Both molecular co-catalysts attach equally well to the QDs
11270
https://doi.org/10.1021/acscatal.1c02921
ACS Catal. 2021, 11, 11266−11277

ACS Catalysis
pubs.acs.org/acscatalysis
Research Article
Figure 4. Investigation of the induction period of Co(tppS3N1). (A) Reduction of Co(tppS4) in the presence of ascorbate as previously reported.20
The reduction of CoIII to CoII is accompanied by the removal of an axial H2O ligand (d6 octahedral to d7 square-pyramidal). (B) UV−vis absorption
spectra of Co(tppS4) and Co(tppS3N1) (1.7 μM in water) in the presence of AA (0.1 M, pH 4.5) recorded every 1 and 5 min, respectively. “Start”
refers to the spectrum in purely aqueous solution in the absence of AA. “Priming” refers to stirring the components in solution prior to irradiation. (C)
Priming Co(tppS3N1) in the presence of AA and its influence on the photocatalytic activity using ZnSe-BF4. The dashed vertical line indicates the
catalytic onset of the green and blue traces. Conditions: 0.5 μM ZnSe-BF4, 10 μM co-catalyst, 0.1 M AA, pH 4.5 (Co(tppS3N1)), AM 1.5G, 100 mW
cm−2, λ > 400 nm, CO2 (4.0 sccm), 25 °C.
with approximately 86% (Co(tppS4)) and 84% (Co-
(tppS3N1)), respectively, of the amount of employed co-
catalyst (Table S4). This finding indicates that the affinity of the
co-catalyst to the QD surface is not the cause of the different
performance. Electrochemical characterization of the two co-
catalysts via cyclic voltammetry (CV) did not yield unambig-
uous results to explain the different activities and induction
periods (for details, see Figure S12).
ligand charge transfer bands in CoIII vs CoII complexes. The shift
is followed by a reduction in the intensity of the 412 nm signal
accompanied by a broad feature growing at 585 nm, which is
presumably caused by a reduced Co species.29,30 In comparison,
Co(tppS3N1) exhibits only a marginal shift of the main
absorption feature at 427 nm upon addition of AA, which is
gradually lowered in intensity in subsequent time intervals. A
broad growing feature at 580 nm is also visible, likewise to
Co(tppS4). Notable is the timescale of both reactions: While
Co(tppS4) requires approximately 25 min to reach a steady-
state signal, Co(tppS3N1) takes significantly longer (120 min)
to reach the steady absorption profile. This period (120 min)
matches with the induction period of Co(tppS3N1) under
photocatalytic conditions in which no activity is observed
(Figure S11).
Co(tppS4) was previously shown to occur initially as CoIII in
aqueous solution and required reduction to CoII, facilitated by
AA (independent of light), to enter the catalytic cycle20 (Figure
4A), accompanied by a distinct blueshift of the absorption
spectrum. This initial CoIII to CoII was monitored via UV−vis
spectroscopy by recording a spectrum of the co-catalyst in AA
solution at regular time intervals (Figure 4B). The spectrum of
Co(tppS4) indeed shows the blueshift of the main absorption
feature from 425 to 412 nm (Soret band) within 1 min of
addition of AA, presumably due to changes in the metal-to-
It was then investigated if the induction period observed in
photocatalysis can be accelerated/eliminated by priming the co-
catalyst in the presence of AA. Therefore, the photocatalyst
11271
https://doi.org/10.1021/acscatal.1c02921
ACS Catal. 2021, 11, 11266−11277

ACS Catalysis
pubs.acs.org/acscatalysis
Research Article
Figure 5. Influence of (A) low light intensity, (B) low CO2 concentration, and (C) O2 on the photocatalytic CO2 reduction activity in the presence of
ZnSe-BF4 QDs and various co-catalysts (600 min irradiation). CO selectivity is defined as n (CO)/(n (CO) + n (H2)). 100% light intensity refers to
100 mW cm−2. Conditions (if not otherwise stated): 0.5 μM ZnSe-BF4, 10 μM co-catalyst, 0.1 M AA, pH 4.5 (Co(tppS3N1)), pH 5.5 (Ni(cycP) and
Co(qpy)), AM 1.5G, 100 mW cm−2, λ > 400 nm, CO2 (4.0 sccm), 25 °C. Jitter was added for each data point manually for plots (B) and (C) for clarity
and to avoid excessive overlap. Note: the sample ZnSe|Co(tppS3N1) at 20% light intensity showed only marginal activity for both H2 and CO (see the
Supporting Information); hence, the CO selectivity constitutes a large standard error.
system was assembled and stirred in the dark in the presence of
all components (Prime ZnSe|Co(tppS3N1) with AA, Figure 4C,
green trace) for 2 h before irradiation. A different sample
consisted of only co-catalyst dissolved in AA solution (Prime
only Co(tppS3N1) with AA, red trace, Figure 4C), which was
left for 2 h prior to addition of ZnSe and irradiation. In
comparison to the regular experiment (blue trace, Figure 4C),
stirring the fully assembled catalyst system did not affect the
induction period. However, when Co(tppS3N1) was pre-
reduced in the presence of AA (and absence of ZnSe), the
induction period vanished, and the hybrid catalyst showed an
almost instant onset after addition of ZnSe followed by
irradiation. This observation supports that the induction period
originates from the slow initial reduction of CoIII to CoII. It is
also observed that there is no local maximum followed by a
decrease/increase of the product evolution curves, compared to
the samples that were not pre-reduced, which could be related to
different types of Co species that initiate at various times
possibly related to a different coordination geometry of
Co(tppS3N1) on ZnSe. It can therefore be concluded that
Co(tppS3N1) exhibits a different redox behavior when it is in a
unique chemical environment (immobilized on the QD surface)
compared to when it freely diffuses in the bulk solution.
Additionally, the in-depth detection and analysis of the
induction period highlight the capabilities of the continuous-
flow system with frequent, automated GC sampling.
Performance Evaluation. The external quantum efficiency
(EQE) was exemplarily determined for the best-performing
photocatalytic system, ZnSe|Co(tppS3N1). At 1.5 mW cm−2
and 400 nm monochromatic illumination, ZnSe|Co(tppS3N1)
achieved an EQECO of 5.1 0.5% (see the Experimental Section
for details and Table S5, average over 4 h), which significantly
exceeds the previous benchmark on ZnSe QDs (ZnSe|Ni(cycP),
EQECO = 3.4
0.3%).19 The performance reported herein
(TONco‑cat > 600, EQECO > 5%) obtained by hybrid
ZnSe|Co(tppS3N1) is among the highest within reported
11272
https://doi.org/10.1021/acscatal.1c02921
ACS Catal. 2021, 11, 11266−11277

ACS Catalysis
pubs.acs.org/acscatalysis
Research Article
colloidal photocatalyst systems that operate free of precious
metals in aqueous solution.7,19,31 Additionally, the activity
translates to a TON (CO) per QD (i.e., sensitizer,
approximately 1100 Zn ions per QD) above 104, which
underscores the QDs as platform for CO2 reduction and is in
contrast to many homogeneous systems, which often employ
much higher concentration of photosensitizer than co-catalyst.20
A higher activity was reported for a mercaptopropionic acid-
modified CdS nanocrystal in combination with a dinuclear
cobalt catalyst, which reached a TONco-cat of 1380 (22 h) with an
impressive CO selectivity of 95% in fully aqueous solution.7
However, the study did not report the EQE and the light
intensity used during photocatalysis, which does not allow for a
direct comparison. The Co(tppS4) analogue reached a
TONco-cat of 926 (EQE = 0.81%) in homogeneous aqueous
solution although only in the presence of large quantities of a
precious-metal [Ru(2,2′-bipyridine)3]2+ photosensitizer
(TONRu ∼ 19).20 Photosystems consisting of earth-abundant
materials have reached high activities (TONco‑cat = 155, EQE >
4%) in organic solution, in particular the Co(quarterpyridine)
(and its Fe analogue) when linked to mesoporous carbon nitride
(TONco-cat = 500, EQE = 4%).6,32 Higher performances have
been achieved with precious-metal-based catalysts for CO2
photoreduction to formate. Anchoring a dinuclear Ru complex
on Ag-loaded graphitic carbon nitride showed enhanced
performance (TON > 2000, up to 98% selectivity, EQE =
0.2%) in water and a TON > 33 000 when organic solvents were
used instead.4,33
Influence of Light Intensity, CO2 Concentration, and
Oxygen. The continuous-flow methodology for the photo-
catalytic experiments allows for the rapid study of different
experimental conditions. Lowering the light intensity (I), i.e.,
changing the flux of available photons and, consequently,
electrons generated, can impact the ratio of evolved H2 vs CO,
due to a delicate balance of charge separation/recombination
kinetics and turnover rate of the co-catalyst. Lowering the light
intensity for the diffusional assembly ZnSe|Co(qpy) affects the
activity of both proton and CO2 reduction, but CO is less
affected than H2 (Figure 5A, full data set in Figure S13). At a
given time point (600 min), H2 evolution is marginally below a
near-linear trend (reduction) in activity, whereas CO deviates
from this relation significantly, thus enhancing the overall CO
selectivity from 7.9% (I = 100%) close to 5-fold to 39% (I =
25%). Nevertheless, the TONCo(qpy) is not improved and
saturates between TON < 70 (I = 100%) and TON < 40 (I =
20%) (Figure S13). The anchored co-catalyst hybrid
ZnSe|Ni(cycP) also exhibits enhanced CO selectivity, but less
pronounced than Co(qpy). The activity at lower light intensities
for both H2 and CO is greatly diminished in the case of the
electrostatic assembly ZnSe|Co(tppS3N1), (non-primed). This
is presumably related to its long induction period. The number
of available electrons may not be sufficient to generate the active
species within the time frame of this experiment. This finding
indicates that the induction period (CoIII to CoII) is facilitated
by light and proceeds from photogenerated QDs and not from
AA, when the co-catalyst is immobilized on the QD surface. This
is in line with the experiments from priming the photocatalyst in
the dark or in the presence of AA (and absence of QDs). The
experiments above demonstrate that excess photogenerated
electrons on ZnSe QDs are utilized toward the H2 evolution
reaction. Lowering the light intensity in the case of the weakly
anchoring Ni(cycP) and diffusional Co(qpy) allows for higher
proportion of electrons to be transferred to the co-catalyst for
CO2 reduction rather than for competing H2 evolution.
We then studied the ability of the QD-co-catalyst hybrids to
operate under a stream of low concentration of CO2 (20% in
balance gas N2, similar to flue gas; pH was kept constant in all
experiments in regard to the optimized pH determined for each
catalyst above) because the supply of concentrated CO2 for the
generation of solar fuels is an obstacle due to the energy costs
associated with concentrating atmospheric levels of CO2 (ca.
415 ppm).34 H2 evolution is effectively unaffected at 20% CO2
(within experimental error), whereas the relative CO activity is
drastically lowered to 26−38% (Figure 5B, full data set in Figure
S14). The loss in CO activity (at 20% CO2) suggests that the
availability of CO2 is limiting at this concentration. Although
photocatalysts were reported to reduce CO2 with only minimal
loss in activity down to 10%,35 those photocatalyst systems
achieved this activity via insertion of CO2 (“capture”) into the
metal−ligand bond of triethanolamine coordinated to a CO2-
reducing Re metal center.
The effect of O2 on performance is another interesting aspect
in colloidal CO2 photoreduction due to the long-term goal to
couple CO2 reduction with water oxidation (instead of using a
sacrificial electron donor), which would produce O2 as a side
product or issues from real-world applications from leakage of
atmospheric O2 into the system.36 This tolerance toward O2 is
particularly challenging because O2 can quench the photo-
excited light absorber and photogenerated electrons and is
therefore considered parasitic in most cases as it competes with
CO2 reduction. In particular, the formation of singlet O2 is a
feasible pathway and observed for most photosensitizers in the
presence of molecular O2.37 Other reactive oxygen species
•−
include O2 and H2O2, resulting from O2 reduction and may
decompose components of the photocatalytic systems.38 For all
studied hybrid photocatalysts, the presence of O2 led to a
diminished photocatalytic activity (Figure 5C, the full data set
can be found in Figure S15). While ZnSe|Co(qpy) shows 30%
relative CO activity in the presence of O2, ZnSe|Ni(cycP) shows
a reduction to approximately 11%. ZnSe|Co(tppS3N1) does not
exhibit any activity in the presence of O2. The instability of the
porphyrin is presumably governed by the formation of singlet O2
species, which can gradually decompose the molecular unit via a
ring-opening pathway, as reported for Co-porphyrin ana-
logues,39,40 although the latter catalyst family has also been
shown to display some tolerance toward O2 in a different
study.20 For Co(qpy), H2 evolution is suppressed more than
CO2 reduction, thereby enhancing the CO selectivity from 8%
(anaerobic) to 16% in the presence of O2. For Ni(cycP), the CO
selectivity remains unaffected by the presence of O2. This
observation may be explained by the diffusional nature of
Co(qpy), which leaves the QD surface accessible for O2
reduction from the ZnSe conduction band. Nevertheless, O2 is
detrimental to all photocatalyst systems studied here and is
presumably related to the formation of reactive oxygen species.
CONCLUSIONS
■
A continuous-flow setup for CO2 reduction photocatalysis was
developed to study semiconductor−metal complex systems,
which enables automated, high-frequency measurements of
multiple samples in parallel while retaining a high sensitivity.
The capabilities of the continuous-flow methodology were
exemplified using a state-of-the-art light-absorbing platform
based on ZnSe-BF4 QDs, which was shown to drive a range of
molecular co-catalysts based on Ni or Co for visible-light-driven
11273
https://doi.org/10.1021/acscatal.1c02921
ACS Catal. 2021, 11, 11266−11277

ACS Catalysis
pubs.acs.org/acscatalysis
Research Article
CO2 reduction. The QDs were thereby able to supply
photogenerated electrons to molecular co-catalysts that are
either of diffusional nature, possess a phosphonate-anchoring
group, or assemble on the surface through electrostatic
interactions. The different anchoring strategies were compared
and quantified with the electrostatic assembly leading to near-
quantitative immobilization. A novel CO2 reduction catalyst, a
Co-tetraphenylporphyrin featuring three sulfonate groups and
one amine group (Co(tppS3N1)), thereby exhibited the
benchmark photocatalytic activity in combination with ZnSe-
BF4, evolving 18.6 μmol of CO (79.7 mmol of CO gZnSe−1) and
reaching a TONCo (CO) of 619 after 1000 min of irradiation
with a CO selectivity of >40%. This photocatalytic activity is the
highest obtained using ZnSe QDs and among the highest in
colloidal photocatalytic CO2 reduction using earth-abundant
materials in aqueous solution. The distinct induction period of
this benchmark photocatalyst was assigned to slow CoIII to CoII
reduction as a prerequisite to enter the catalytic cycle and could
be accelerated by priming (pre-reducing the co-catalyst in AA
solution in the dark). All photocatalyst systems were studied
under low light intensities, low CO2 concentration, and aerobic
conditions, which was shown to affect the ratio of H2 evolution
vs CO formation. The screening method presented here allows
for faster, more reliable, and accurate analysis of photocatalytic
systems for aqueous CO2 reduction and further demonstrates
the underexplored potential of rationally assembled semi-
conductor−metal complex hybrids constructed from abundant
elements for demanding photocatalytic transformations at high
performance.
samples were digested in HNO3 and diluted with ultrapure
water to 1−10 ppm analyte. Blank samples of diluted HNO3
were recorded as background. 1H NMR spectra (to investigate
the potential production of formate) were recorded on a Bruker
400 MHz Avance III HD Smart Probe Spectrometer with water
pre-saturation (64 scans).
Synthesis and Characterization of ZnSe QDs. Ligand-
free ZnSe-BF4 QDs were prepared as reported previously.19 The
mean particle size was determined from transmission electron
microscopy images (d = 4.65 nm) and the particles feature a
good visible-light absorption onset (λmax = 412 nm). To
calculate the QD concentration in the stock solution, the Zn2+
and Se2− concentration determined by ion-coupled plasma
optical emission spectroscopy was divided by the number of Zn
atoms per QD based on the mean particle diameter and the bulk
density of ZnSe (5.262 g cm−3). The full characterization of the
QDs can be found in Figure S3.
Synthesis and Characterization of Molecular Co-
catalysts. Ni(cycH) and Ni(cycP),41,42 Co(pcS4)43 and
Co(pcTMA4),23 Ni(terpyS) and Ni(terpyP),3 and Co-
(tppS4)20 and Co(qpy)21 were prepared and characterized
according to literature procedures.
Co(tppS3N1). The free base ligand, tppS3N1 (41.7 mg), was
prepared as described in the literature44,45 and added to a 125
mL, three-neck flask containing 10 mL of methanol, and the
resulting solution was stirred at 60 °C under nitrogen for 10 min.
A solution of cobalt(II) acetate tetrahydrate (57.1 mg) in 5 mL
of methanol and 15 mL of chloroform was then added, and the
progress of the stirred reaction was monitored by thin-layer
chromatography and UV−vis spectroscopy. After 8 h of
reaction, the flask was opened to air and the solvent removed
under reduced pressure with a rotatory evaporator, and the red
solid residue was purified by column chromatography on silica
gel (gradient of methanol and methanol/acetic acid (9:1) as
eluent) to remove the excess Co salt. The crude product was
then passed through an ion-exchange resin (Dowex 50W-X8, H+
form, 50−100 mesh), followed by filtration over Celite using
MeOH (3×). The product was precipitated with an excess of
acetonitrile and collected as a dark purple solid and dried in
vacuo (yield 84%). Mass spectrometry (MS) (electrospray
ionization (ESI), positive mode) (m/z) calcd for
[C44H27N5O9S3Co]2+: 462.0157; found 462.0174. Elemental
analysis for C44H26CoN5Na3O9S3·9H2O: C, 45.76; H, 3.84; N,
6.06; found: C, 45.67; H, 3.51; N, 5.81.
Sample Preparation for Photocatalysis. A ZnSe-BF4
stock solution (64.1 μM in dimethylformamide (DMF), 23.4
μL) and a co-catalyst solution (5.0 mM in H2O, typically 6 μL)
were added to a Pyrex glass photoreactor (Chromacol 10-SV,
Fisher Scientific) containing a magnetic stirrer bar. The mixture
was diluted with AA (0.1 M in water, pH adjusted to 4.5−6.5
with NaOH and NaHCO3) to a total solution volume of 3 mL.
Specifically, pH 4.5 (after saturating with CO2) was achieved by
adjusting the pH of the 0.1 M AA solution with NaOH to 4.5. A
final pH of 5.5 (after saturating with CO2) was achieved by
adjusting the pH of the 0.1 M AA solution with NaOH to 6.5,
which decreases to 5.5 upon purging with CO2. A final pH of 6.5
was achieved by neutralizing the pH of the 0.1 M AA solution
and adding 0.1 M NaHCO3 buffer, which gives a pH of 6.5 upon
saturation with CO2. The photoreactor was then sealed with a
rubber septum and pierced with two needles (inlet and outlet).
Constant-Flow Setup with Automated Product Quan-
tification. The inlet of the photoreactor was connected to a
MFC (Brooks GF040) supplying a stream of CO2 (CP grade,
EXPERIMENTAL SECTION
■
Materials. Zinc stearate (purum, Sigma-Aldrich), octade-
cene (90% techn., Sigma-Aldrich), and selenium powder (99%,
Sigma-Aldrich) for the ZnSe QD synthesis; L-ascorbic acid
(99%, Sigma-Aldrich), trimethyloxonium tetrafluoroborate
(≥97.0%, Sigma-Aldrich), and organic solvents were used as
received. Anhydrous solvents were supplied from Acros
Organics. All aqueous experimental solutions were prepared
with ultrapure water (DI water; Milli-Q, 18.2 MΩ cm). 13CO2
(>99 atom % 13C) was purchased from Sigma-Aldrich.
Physical Characterization. Powder X-ray diffraction was
conducted using an X’Pert PRO by PANalytical BV instrument
using Cu Kα irradiation. Transmission electron microscopy
images were collected using a Thermo Scientific (FEI) Talos
F200X G2 instrument, operating at an accelerating voltage of
200 kV. Samples were prepared by drop-casting a dilute QD-
solution on holey-carbon-coated Cu grids followed by
evaporation of the solvent. Gas-phase infrared spectra of the
photoreactor headspace were recorded on a Thermo Scientific
Nicolet iS50 FT-IR spectrometer in transmission mode. UV−vis
spectra were recorded on an Agilent Cary 60 UV−vis
spectrophotometer using quartz glass cuvettes (1 cm path
length). ζ-Potential measurements of ZnSe-BF4 (0.5 μM, in
water) were conducted using a Malvern Zetasizer Nano ZS90
instrument at 25°C. High-resolution mass spectra were recorded
using a Thermo Scientific Orbitrap Classic mass spectrometer.
Elemental analysis was carried out by the Microanalysis Service
of the Yusuf Hamid Department of Chemistry, University of
Cambridge, using an Exeter Analytical CE-440 Elemental
Analyzer. Inductively coupled plasma optical emission spectros-
copy was carried out by the Microanalysis Services, Yusuf
Hamied Department of Chemistry, University of Cambridge,
using a Thermo Scientific iCAP 7400 spectrometer. The
11274
https://doi.org/10.1021/acscatal.1c02921
ACS Catal. 2021, 11, 11266−11277

ACS Catalysis
pubs.acs.org/acscatalysis
Research Article
BOC, humidified with a water saturator) with a flow rate of 4.0
sccm. The flow rate at the GC outlet was verified prior to the
experiment with an Alicat gas flow meter to avoid gas leakage.
The outlet of the photoreactor was connected to a flow-selection
valve controlled by a Shimadzu Tracera GC-2010 Plus GC for
product quantification of the gaseous reaction products (see
below). Six samples (two triplicates of identical conditions)
were typically analyzed in parallel. Upon purging with a constant
stream of CO2, the solution pH decreased to 5.5 due to
saturation with CO2 (pH 6.5 prior to purging) or remains at 4.5
(pH 4.5 prior to purging). The photoreactor was purged for a
further 45 min in the dark and sampled via online GC
quantification. The first two injections of each sample were used
to determine a “background” peak, which was subtracted from
further injections. The photoreactor was then placed in a water
bath maintained at 25 °C, stirred, and irradiated by a solar light
simulator (Newport Oriel, AM 1.5G, 100 mW cm−2). The six
samples were evenly distributed within the light simulator to
account for possible variations of the light intensity depending
on the position in the simulator. UV irradiation was filtered with
a 400 nm cutoff filter (UQG).
GC Setup. The Shimadzu Tracera GC-2010 Plus GC used a
barrier discharge ionization detector, kept at 300 °C, and was
equipped with a HayeSep D (2 m × 1/8 in. OD × 2 mm ID, 80/
100 mesh, Analytical Columns) precolumn and an RT-Molsieve
5A (30 m × 0.53 mm ID, Restek) main column to separate H2,
O2, N2, CH4, and CO while preventing CO2 and H2O to reach
the Molsieve column. He carrier gas (grade 5.0, BOC) was
purified (HP2−220, VICI) prior to entering the GC. The
column temperature was kept at 85 °C. The gaseous flow from
the flow-selection valve was passed through a loop (volume 1.0
mL) and injected approximately every 4.25 min into the GC. Six
samples were analyzed in parallel, with each individual sample
being injected every 25.5 min. The GC calibration was
performed with a known standard for H2, CO, and CH4 (2040
ppm H2/2050 ppm CO/2050 ppm CH4 in balance gas CO2,
BOC) by diluting the mixture with pure CO2.
deviation. Specifically, the calculated standard deviation is
used to compute a Gaussian density for that standard deviation,
plotting a cloud with the opacity proportional to the density.
This appears as a vertical “cloud” of uncertainty.48 The
maximum of the uncertainty cloud is set to 1 standard deviation.
Isotopic Labeling Studies. For 13C isotopic labeling,
photocatalysis experiments were performed as described above,
but with accumulating products in the headspace under steady-
state conditions and using 13CO2 as the headspace gas. After
1000 min (16.7 h), the photoreactor headspace was transferred
to an evacuated gas infrared cell (SpecAc, 10 cm path length,
equipped with KBr windows) and a high-resolution gas-phase
transmission spectrum was collected.
EQE Determination. Photocatalysis samples were prepared
with a modified procedure using a flat-sided quartz cuvette (1
cm path length, airtight) as the photoreactor in batch mode.
Co(tppS3N1) was first dissolved in AA solution (1.2 mL, 0.1 M,
pH adjusted to 4.5). After 2 h, the ZnSe-BF4 QD stock solution
was added to give a 0.5 μM concentration, and the sample was
purged with CO2 (containing 2% CH4 as internal standard). The
sample was preirradiated for 200 min with a solar light simulator
to approach the highest activity. The cuvette was then purged
again with CO2/CH4 (2%) and irradiated with monochromatic
light (λ = 400 5 nm, A = 0.80 cm2) using an LOT Quantum
Design MSH 300 monochromator. Aliquots of headspace gas
were taken periodically and analyzed by gas chromatography.
The EQE was calculated according to the formula
2n × N × h × c
tirr × λ × I × A
A
EQE (%) =
× 100
where n is the amount of produced CO, NA is Avogadro’s
constant, h is Planck’s constant, c is the speed of light, tirr is the
irradiation time, λ is the irradiation wavelength, I is the
irradiation intensity, and A is the irradiated area. The factor 2 in
the numerator is used because two electrons (and consequently
two photons) are required for the conversion of CO2 to CO.
Electrochemical Characterization. CV was performed in
a one-chamber electrochemical cell using a glassy carbon
working electrode. The cell was assembled by placing the
working electrode together with a Ag/AgCl reference and a Pt
mesh counter electrode into a solution of the co-catalyst (1.0
mM) dissolved in 0.1 M aqueous NaClO4 (1 mL total volume).
(Note, NaClO4 can be explosive, and care needs to be taken
when handling it. A stock solution should never be left to dry.)
The cell was sealed and purged with CO2 for 15 min. CV scans
were recorded at room temperature using a BioLogic VSP
potentiostat with a scan rate of 100 mV s−1. The potentials were
converted from Ag/AgCl/KCl(satʼd) to normal hydrogen
electrode (NHE) by adding +0.2 V.
Data Analytics. The data were processed and visualized
using the statistical programming language R with the tidyverse
library.46,47 First, the flow rates were corrected by subtracting a
“background” peak obtained in the dark prior to irradiation (we
noticed a marginal CO background peak depending on the
residual amount of oxygen present in the sample streama
feature of the gas chromatograph and not the sample). Second,
̇
the momentary product evolution rate (ngas) corresponding to
each injection was calculated using the following formula.
area GC
̇
p × V ×
fi
n
=
̇
gas
R T
where p is the pressure in the photoreactor (ambient pressure,
101 325 Pa), V is the flow rate (4.0 sccm), R is the universal gas
ASSOCIATED CONTENT
* Supporting Information
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acscatal.1c02921.
■
̇
sı
constant, T is the temperature (298 K), and fi is the response
factor for each gas determined by the calibration procedure.
Third, the total amount of evolved product was calculated using
trapezoidal integration of the product evolution rates. The three
independent replicates of identical conditions were averaged by
calculating the mean and standard deviation over irradiation
time and sample. For visual display, the actual values for each
sample are plotted as transparent scatter, whereas the mean is
represented as a continuous line. In addition, the standard
deviation is visualized by the shaded area surrounding the mean
where the transparency is proportional to the standard
Results from prescreening of various molecular CO2
reduction co-catalysts (Table S1); results of the photo-
catalytic CO2 reduction experiments and photocatalytic
control experiments (Tables S2 and S3); quantification of
co-catalyst attachment (Table S4); external quantum
efficiency determination (Table S5); calibration curve of
the GC and raw gas chromatograms (Figure S1);
continuous-flow photocatalysis method development
11275
https://doi.org/10.1021/acscatal.1c02921
ACS Catal. 2021, 11, 11266−11277

ACS Catalysis
pubs.acs.org/acscatalysis
Research Article
(2) Maeda, K.; Kuriki, R.; Sekizawa, K.; Ishitani, O. Metal-Complex/
Semiconductor Hybrids for Carbon Dioxide Fixation. Sol. Hydrogen
Nanotechnol. X 2015, 9560, No. 956006.
(3) Kuehnel, M. F.; Orchard, K. L.; Dalle, K. E.; Reisner, E. Selective
Photocatalytic CO2 Reduction in Water through Anchoring of a
Molecular Ni Catalyst on CdS Nanocrystals. J. Am. Chem. Soc. 2017,
139, 7217−7223.
(4) Kuriki, R.; Matsunaga, H.; Nakashima, T.; Wada, K.; Yamakata, A.;
Ishitani, O.; Maeda, K. Nature-Inspired, Highly Durable CO2
Reduction System Consisting of a Binuclear Ruthenium(II) Complex
and an Organic Semiconductor Using Visible Light. J. Am. Chem. Soc.
2016, 138, 5159−5170.
(5) Rao, H.; Schmidt, L. C.; Bonin, J.; Robert, M. Visible-Light-Driven
Methane Formation from CO2 with a Molecular Iron Catalyst. Nature
2017, 548, 74−77.
(Figure S2); characterization of ZnSe-BF4 QDs (Figure
S3); overview of screened molecular co-catalysts (Figure
S4); data processing and visualization for evolved CO
(Figure S5); photocatalytic H2 evolution (Figure S6); co-
catalyst optimization (Figure S7); 13C isotopic labeling
(Figure S8); co-catalyst attachment quantification
(Figure S9); postcatalysis characterization (Figure S10);
benchmarking the photocatalytic CO2 reduction activity
(Figure S11); CV scans (Figure S12); influence of light
intensity on photocatalytic CO2 reduction (Figure S13);
influence of low CO2 concentration on photocatalytic
CO2 reduction (Figure S14); and influence of the
presence of O2 on photocatalytic CO2 reduction (Figure
S15) (PDF)
(6) Ma, B.; Chen, G.; Fave, C.; Chen, L.; Kuriki, R.; Maeda, K.;
Ishitani, O.; Lau, T.-C.; Bonin, J.; Robert, M. Efficient Visible-Light-
Driven CO2 Reduction by a Cobalt Molecular Catalyst Covalently
Linked to Mesoporous Carbon Nitride. J. Am. Chem. Soc. 2020, 142,
6188−6195.
AUTHOR INFORMATION
■
Corresponding Author
(7) Bi, Q.-Q.; Wang, J.-W.; Lv, J.-X.; Wang, J.; Zhang, W.; Lu, T.-B.
Selective Photocatalytic CO2 Reduction in Water by Electrostatic
Assembly of CdS Nanocrystals with a Dinuclear Cobalt Catalyst. ACS
Catal. 2018, 8, 11815−11821.
(8) Lima, F.; Grunenberg, L.; Rahman, H. B. A.; Labes, R.; Sedelmeier,
J.; Ley, S. V. Organic Photocatalysis for the Radical Couplings of
Boronic Acid Derivatives in Batch and Flow. Chem. Commun. 2018, 54,
5606−5609.
(9) Ritter, A.; Reifler, F. A.; Saini, S. Quick Screening Method for the
Photocatalytic Activity of Nanoparticles and Powdery Materials. Appl.
Catal., A 2009, 352, 271−276.
(10) Seo, H.; Liu, A.; Jamison, T. F. Direct β-Selective Hydro-
carboxylation of Styrenes with CO2 Enabled by Continuous Flow
Photoredox Catalysis. J. Am. Chem. Soc. 2017, 139, 13969−13972.
(11) Hurtado, L.; Natividad, R.; García, H. Photocatalytic Activity of
Cu2O Supported on Multi Layers Graphene for CO2 Reduction by
Water under Batch and Continuous Flow. Catal. Commun. 2016, 84,
30−35.
(12) Dilla, M.; Schlögl, R.; Strunk, J. Photocatalytic CO2 Reduction
Under Continuous Flow High-Purity Conditions: Quantitative
Evaluation of CH4 Formation in the Steady-State. ChemCatChem
2017, 9, 696−704.
Erwin Reisner − Yusuf Hamied Department of Chemistry,
University of Cambridge, Cambridge CB2 1EW, United
Kingdom; orcid.org/0000-0002-7781-1616;
Email: reisner@ch.cam.ac.uk; http://www-
reisner.ch.cam.ac.uk/.
Authors
Constantin D. Sahm − Yusuf Hamied Department of
Chemistry, University of Cambridge, Cambridge CB2 1EW,
United Kingdom
Geani M. Ucoski − Yusuf Hamied Department of Chemistry,
University of Cambridge, Cambridge CB2 1EW, United
Kingdom
Souvik Roy − Yusuf Hamied Department of Chemistry,
University of Cambridge, Cambridge CB2 1EW, United
Kingdom; orcid.org/0000-0003-0146-5283
Complete contact information is available at:
https://pubs.acs.org/10.1021/acscatal.1c02921
Notes
(13) Dilla, M.; Becerikli, A. E.; Jakubowski, A.; Schlögl, R.; Ristig, S.
Development of a Tubular Continuous Flow Reactor for the
Investigation of Improved Gas−Solid Interaction in Photocatalytic
CO2 Reduction on TiO2. Photochem. Photobiol. Sci. 2019, 18, 314−318.
(14) Ali, S.; Lee, J.; Kim, H.; Hwang, Y.; Razzaq, A.; Jung, J.-W.; Cho,
C.-H.; In, S.-I. Sustained, Photocatalytic CO2 Reduction to CH4 in a
Continuous Flow Reactor by Earth-Abundant Materials: Reduced
Titania-Cu2O Z-Scheme Heterostructures. Appl. Catal., B 2020, 279,
No. 119344.
(15) Motz, R. N.; Lopato, E. M.; Connell, T. U.; Bernhard, S. High-
Throughput Screening of Earth-Abundant Water Reduction Catalysts
toward Photocatalytic Hydrogen Evolution. Inorg. Chem. 2021, 60,
774−781.
(16) Ozawa, H.; Haga, M.; Sakai, K. A Photo-Hydrogen-Evolving
Molecular Device Driving Visible-Light-Induced EDTA-Reduction of
Water into Molecular Hydrogen. J. Am. Chem. Soc. 2006, 128, 4926−
4927.
(17) Burger, B.; Maffettone, P. M.; Gusev, V. V.; Aitchison, C. M.; Bai,
Y.; Wang, X.; Li, X.; Alston, B. M.; Li, B.; Clowes, R.; Rankin, N.; Harris,
B.; Sprick, R. S.; Cooper, A. I. A Mobile Robotic Chemist. Nature 2020,
583, 237−241.
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
The authors gratefully acknowledge financial support by the
Christian Doppler Research Association, the Austrian Federal
Ministry for Digital and Economic Affairs, the National
Foundation for Research, Technology and Development and
the OMV Group (C.D.S, E.R.), the European Commission for a
Horizon 2020 Marie Skłodowska-Curie Individual Fellowship
(S.R., GAN745604, CO2RED), and the Conselho Nacional de
Desenvolvimento Científico e Tecnológico (G.M.U., CNPq
Grant 204123/2017-8). Dr. Gaia Neri and Dr. Alexander
Cowan are gratefully acknowledged for providing Ni(cycP). The
authors thank Dr. Heather Greer for analyzing transmission
electron microscopy samples and thankfully acknowledge the
EPSRC Multi-User Equipment Call (EP/P030467/1). They
also thank Dr. Andreas Wagner for help during the method
development and Dr. Erwin Lam and Dr. Shannon A. Bonke for
helpful discussions and proofreading of the manuscript.
(18) Kaneko, H.; Minegishi, T.; Nakabayashi, M.; Shibata, N.; Kuang,
Y.; Yamada, T.; Domen, K. A Novel Photocathode Material for
Sunlight-Driven Overall Water Splitting: Solid Solution of ZnSe and
Cu(In,Ga)Se2. Adv. Funct. Mater. 2016, 26, 4570−4577.
(19) Kuehnel, M. F.; Sahm, C. D.; Neri, G.; Lee, J. R.; Orchard, K. L.;
Cowan, A. J.; Reisner, E. ZnSe Quantum Dots Modified with a
REFERENCES
■
(1) Dalle, K. E.; Warnan, J.; Leung, J. J.; Reuillard, B.; Karmel, I. S.;
Reisner, E. Electro- and Solar-Driven Fuel Synthesis with First Row
Transition Metal Complexes. Chem. Rev. 2019, 119, 2752−2875.
11276
https://doi.org/10.1021/acscatal.1c02921
ACS Catal. 2021, 11, 11266−11277

ACS Catalysis
pubs.acs.org/acscatalysis
Research Article
Ni(Cyclam) Catalyst for Efficient Visible-Light Driven CO2 Reduction
in Water. Chem. Sci. 2018, 9, 2501−2509.
(36) Wakerley, D. W.; Reisner, E. Oxygen-Tolerant Proton Reduction
Catalysis: Much O2 about Nothing? Energy Environ. Sci. 2015, 8, 2283−
2295.
(37) Schweitzer, C.; Schmidt, R. Physical Mechanisms of Generation
and Deactivation of Singlet Oxygen. Chem. Rev. 2003, 103, 1685−1758.
(38) Mondal, B.; Sen, P.; Rana, A.; Saha, D.; Das, P.; Dey, A.
Reduction of CO2 to CO by an Iron Porphyrin Catalyst in the Presence
of Oxygen. ACS Catal. 2019, 9, 3895−3899.
(39) Nakazono, T.; Parent, A. R.; Sakai, K. Improving Singlet Oxygen
Resistance during Photochemical Water Oxidation by Cobalt
Porphyrin Catalysts. Chem. − Eur. J. 2015, 21, 6723−6726.
(40) Nakazono, T.; Parent, A. R.; Sakai, K. Cobalt Porphyrins as
Homogeneous Catalysts for Water Oxidation. Chem. Commun. 2013,
49, 6325−6327.
(41) Beley, M.; Collin, J.-P.; Ruppert, R.; Sauvage, J.-P. Nickel(II)-
Cyclam: An Extremely Selective Electrocatalyst for Reduction of CO2
in Water. J. Chem. Soc., Chem. Commun. 1984, 2, 1315−1316.
(42) Neri, G.; Forster, M.; Walsh, J. J.; Robertson, C. M.; Whittles, T.
(20) Call, A.; Cibian, M.; Yamamoto, K.; Nakazono, T.; Yamauchi, K.;
Sakai, K. Highly Efficient and Selective Photocatalytic CO2 Reduction
to CO in Water by a Cobalt Porphyrin Molecular Catalyst. ACS Catal.
2019, 9, 4867−4874.
̀
(21) Guo, Z.; Cheng, S.; Cometto, C.; Anxolabéhere-Mallart, E.; Ng,
S.-M.; Ko, C.-C.; Liu, G.; Chen, L.; Robert, M.; Lau, T.-C. Highly
Efficient and Selective Photocatalytic CO2 Reduction by Iron and
Cobalt Quaterpyridine Complexes. J. Am. Chem. Soc. 2016, 138, 9413−
9416.
(22) Weber, J. H.; Busch, D. H. Complexes Derived from Strong Field
Ligands. XIX. Magnetic Properties of Transition Metal Derivatives of
4,4′,4″,4’’’-Tetrasulfophthalocyanine. Inorg. Chem. 1965, 4, 469−471.
(23) Schipper, E. T. W. M.; Heuts, J. P. A.; Piet, P.; Beelen, T. P. M.;
German, A. L. Effects of Complexation of Oppositely Charged Water-
Soluble Cobaltphthalocyanines on the Catalytic Mercaptoethanol
Autoxidation. J. Mol. Catal. 1994, 87, 161−176.
(24) Azcarate, I.; Costentin, C.; Robert, M.; Savéant, J. M. Through-
Space Charge Interaction Substituent Effects in Molecular Catalysis
Leading to the Design of the Most Efficient Catalyst of CO2-to-CO
Electrochemical Conversion. J. Am. Chem. Soc. 2016, 138, 16639−
16644.
(25) Kuehnel, M. F.; Creissen, C. E.; Sahm, C. D.; Wielend, D.;
Schlosser, A.; Orchard, K. L.; Reisner, E. ZnSe Nanorods as Visible-
Light Absorbers for Photocatalytic and Photoelectrochemical H2
Evolution in Water. Angew. Chem., Int. Ed. 2019, 58, 5059−5063.
(26) Sahm, C. D.; Mates-Torres, E.; Eliasson, N.; Sokołowski, K.;
Wagner, A.; Dalle, K. E.; Huang, Z.; Scherman, O. A.; Hammarström,
L.; García-Melchor, M.; Reisner, E. Imidazolium-Modification
Enhances Photocatalytic CO2 Reduction on ZnSe Quantum Dots.
Chem. Sci. 2021, 12, 9078−9087.
(27) Martindale, B. C. M.; Joliat, E.; Bachmann, C.; Alberto, R.;
Reisner, E. Clean Donor Oxidation Enhances the H2 Evolution Activity
of a Carbon Quantum Dot−Molecular Catalyst Photosystem. Angew.
Chem., Int. Ed. 2016, 55, 9402−9406.
̀
J.; Farras, P.; Cowan, A. J. Photochemical CO2 Reduction in Water
Using a Co-Immobilised Nickel Catalyst and a Visible Light Sensitiser.
Chem. Commun. 2016, 52, 14200−14203.
(43) Yang, Y.-C.; Ward, R. J.; Seiders, R. P. Dimerization of Cobalt(II)
Tetrasulfophthalocyanine in Water and Aqueous Alcoholic Solutions.
Inorg. Chem. 1985, 24, 1765−1769.
(44) Arlegui, A.; El-Hachemi, Z.; Crusats, J.; Moyano, A. 5-Phenyl-
10,15,20-Tris(4-Sulfonatophenyl)Porphyrin: Synthesis, Catalysis, and
Structural Studies. Molecules 2018, 23, No. 3363.
(45) Sansiaume, E.; Ricoux, R.; Gori, D.; Mahy, J.-P. Oxidation of
Organic Molecules in Homogeneous Aqueous Solution Catalyzed by
Hybrid Biocatalysts (Based on the Trojan Horse Strategy).
Tetrahedron: Asymmetry 2010, 21, 1593−1600.
(46) R Core Team: RA Language and Environment for Statistical
Computing; R Foundation for Statistical Computing: Vienna, Austria,
2021. Retrieved from www.R-Project.Org/.
(47) Wickham, H.; Averick, M.; Bryan, J.; Chang, W.; McGowan, L.;
Franco
̧
is, R.; Grolemund, G.; Hayes, A.; Henry, L.; Hester, J.; Kuhn, M.;
Pedersen, T.; Miller, E.; Bache, S.; Mu
̈
ller, K.; Ooms, J.; Robinson, D.;
(28) Bachmann, C.; Probst, B.; Guttentag, M.; Alberto, R. Ascorbate
as an Electron Relay between an Irreversible Electron Donor and
Ru(II) or Re(I) Photosensitizers. Chem. Commun. 2014, 50, 6737−
6739.
(29) Araullo-McAdams, C.; Kadish, K. M. Electrochemistry, Spec-
troscopy, and Reactivity of (Meso-tetrakis(1-methylpyridinium-4-
yl)porphinato)Cobalt(III,II,I) in Nonaqueous Media. Inorg. Chem.
1990, 29, 2749−2757.
Seidel, D.; Spinu, V.; Takahashi, K.; Vaughan, D.; Wilke, C.; Woo, K.;
Yutani, H. Welcome to the Tidyverse. J. Open Source Software 2019, 4,
No. 1686.
(48) Pav, S. E. Documentation for Geom_cloud Function, 2020.
www.Rdocumentation.org.
(30) Behar, D.; Dhanasekaran, T.; Neta, P.; Hosten, C. M.; Ejeh, D.;
Hambright, P.; Fujita, E. Cobalt Porphyrin Catalyzed Reduction of
CO2. Radiation Chemical, Photochemical, and Electrochemical
Studies. J. Phys. Chem. A 1998, 102, 2870−2877.
(31) Lian, S.; Kodaimati, M. S.; Weiss, E. A. Photocatalytically Active
Superstructures of Quantum Dots and Iron Porphyrins for Reduction
of CO2 to CO in Water. ACS Nano 2018, 12, 568−575.
(32) Cometto, C.; Kuriki, R.; Chen, L.; Maeda, K.; Lau, T.-C.; Ishitani,
O.; Robert, M. A Carbon Nitride/Fe Quaterpyridine Catalytic System
for Photostimulated CO2-to-CO Conversion with Visible Light. J. Am.
Chem. Soc. 2018, 140, 7437−7440.
(33) Kuriki, R.; Yamamoto, M.; Higuchi, K.; Yamamoto, Y.; Akatsuka,
M.; Lu, D.; Yagi, S.; Yoshida, T.; Ishitani, O.; Maeda, K. Robust Binding
between Carbon Nitride Nanosheets and a Binuclear Ruthenium(II)
Complex Enabling Durable, Selective CO2 Reduction under Visible
Light in Aqueous Solution. Angew. Chem., Int. Ed. 2017, 56, 4867−
4871.
(34) Osterloh, F. E. The Low Concentration of CO2 in the
Atmosphere Is an Obstacle to a Sustainable Artificial Photosynthesis
Fuel Cycle Based on Carbon. ACS Energy Lett. 2016, 1, 1060−1061.
(35) Nakajima, T.; Tamaki, Y.; Ueno, K.; Kato, E.; Nishikawa, T.;
Ohkubo, K.; Yamazaki, Y.; Morimoto, T.; Ishitani, O. Photocatalytic
Reduction of Low Concentration of CO2. J. Am. Chem. Soc. 2016, 138,
13818−13821.
11277
https://doi.org/10.1021/acscatal.1c02921
ACS Catal. 2021, 11, 11266−11277