pubs.acs.

org/OPRD Article

Practical Ferrioxalate Actinometry for the Determination of Photon

Fluxes in Production-Oriented Photoflow Reactors
Bavo Vandekerckhove, Nicola Piens, Bert Metten, Christian V. Stevens, and Thomas S. A. Heugebaert*
Cite This: Org. Process Res. Dev. 2022, 26, 2392−2402 Read Online

ACCESS Metrics & More Article Recommendations *

sı Supporting Information
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ABSTRACT: Accurate determination of the photon flux is of major importance to evaluate and characterize photochemical reactor
setups. Knowing the photon flux ensures reproducible reactor operation and facilitates predictable scale-up. Over the past years, flow
reactors have proven to be the key enabling technology for photochemistry to become relevant on production scales. This is mainly
Downloaded via UNIV OF SANTO TOMAS on August 24, 2023 at 01:01:22 (UTC).

due to the mitigation of the limited penetration depth of photons in typical batch reactors. However, due to the practical drawbacks
of the widely accepted standard for photon flux determination (ferrioxalate actinometry) concerning precipitation and gas formation
at higher conversion, reliable actinometry in flow reactors is still challenging. In this paper, three practical approaches for the
ferrioxalate-based determination of the photon flux are presented, which address these problems. These “dimmed emitter,”
“segment-based,” and “time-resolved” methods thus allow photon flux determination in flow reactors with higher irradiated volumes
and more powerful light sources, which is of utmost importance in the context of future scale-up.
KEYWORDS: actinometry, ferrioxalate, photochemistry, flow chemistry, photon flux, HANU 2X 15 reactor, Vapourtec UV-150 reactor

■ INTRODUCTION
The use of light (UV−vis) in chemical reactions was already
put forward in the early 1900s as one of the seeds of the green
chemistry principles. Unlike chemical reagents or catalysts,
photons leave no waste or excess of atoms behind, and
reactions can often proceed rapidly at ambient temperature
and pressure.1 After remaining remarkably underdeveloped for
many decades, the interest in photochemical synthesis has
increased exponentially in the last few years.2−5 According to
the Beer−Lambert law, the intensity of an electromagnetic
wave inside a reaction mixture decreases exponentially, causing
only the first layers to be irradiated efficiently.6 Therefore,
continuous-flow setups using small pathlengths have been the
key enabling technology for photochemistry to become
industrially relevant.7 Scaling-up photochemical flow processes
can now be performed by flow systems that retain the short
pathlengths but have higher irradiated volumes and more
powerful lamps to increase material throughput and reaction
conversions.
In most cases, light is a reagent that needs to be added in
stoichiometric amounts to the reaction mixture. However,
when evaluating literature reports on photochemical reactions,
light sources are rarely described in a way that provides
sufficient information on the number of useful out-the-front
photons they produce, and/or their spectral distribution.8
Therefore, to further advance the field of applied photo-
chemistry, the determination of the incident photon flux
(moles of photons/s or Einstein/s) in the used system is of
major importance. Quantification of the photon flux is typically
accomplished with chemical actinometry, i.e., the measurement
of the radiant flux using the yield of a photochemical reaction.
Actinometry has been investigated for decades in batch, but
accurate and reliable measurement of the incident photon flux
in flow reactors is still far from trivial.9
Considering the importance of the photon flux to evaluate
and compare the performance of reactors for photon-limited
chemical processes, the practical description of actinometry in
photoflow reactors is strongly desired. For each reactor system,
the incident photon flux reaching the reaction mixture depends
on many different factors, e.g., reactor type (batch vs flow),
reactor design (tubular vs plate reactors), reactor materials and
thickness, pathlength, reaction setup, wavelength, lamp type,
lamp power, etc. Therefore, it is of high importance to quantify
the number of photons provided to the reaction mixture in the
reactor. Only then, the efficiency of the performed process in a
particular reactor setup can be evaluated and reactor setups can
be compared so that the process becomes reproducible for
other labs and scaling-up can be performed.10
Some research has been performed before on the
determination of the photon flux using chemical actinometry
in photoflow reactors. Aillet et al. performed reliable
actinometry in microreactors with an irradiated volume
below 1 mL.11 The widely accepted standard ferrioxalate
actinometer was used to determine the photon flux. However,
the authors reported difficulties as the large concentrations
required to ensure full absorption led to the formation of solids
at higher conversions. Therefore, some studies were focused
on the design of new, soluble actinometers for the
Received: March 15, 2022
Published: June 10, 2022

© 2022 American Chemical Society https://doi.org/10.1021/acs.oprd.2c00079

2392 Org. Process Res. Dev. 2022, 26, 2392−2402
Organic Process Research & Development
determination of the photon flux in continuous-flow systems to
overcome this problem.12 However, these new actinometers
are often quite expensive, not broadly characterized, and are
not as easy to use and analyze as the aforementioned
ferrioxalate.13,14 As such, reproducible methods that allow
the use of the widely accepted ferrioxalate standard for
actinometry in larger and more productive photoflow reactors
are still lacking.
■ FERRIOXALATE ACTINOMETRY AND ITS
LIMITATIONS
Several different actinometers have been characterized and
used depending on their applicability range: Reinecke salt
(316−750 nm), Aberchrome 540 (436−546 nm), DPA (400−
550 nm), o-nitrobenzaldehyde (300−410 nm), 2,4-dinitro-
benzaldehyde (365 and 440 nm), etc. However, because of its
wide applicability (250−500 nm), the ease of analysis
(spectrophotometric), and the broad availability of quantum
yields from multiple, independent sources, ferrioxalate is the
most frequently used actinometer for the determination of
photon fluxes.12 The principle of this actinometer is based on
the irreversible light-induced redox reaction of ferrioxalate.
Under the influence of light (250−500 nm), FeIII+ is converted
into FeII+ according to the equations below (Scheme 1).15 The
Scheme 1. Reaction Mechanism of the Light-Induced Redox
Reaction of Ferrioxalate
determination of the conversion of the ferrioxalate reaction is
performed spectrophotometrically. The FeII+ formed during
the photoredox reaction is added to a 1,10-phenanthroline
solution where a quantitative complexation process results in
the formation of the deeply red ferroin. Subsequently, the
conversion can be determined by absorbance measurement at
510 nm.
Despite the widespread use of this ferrioxalate actinometer,
this system displays some physical and chemical application
limits that are underexposed in the literature, especially when
considering flow systems. Wriedt and Ziegenbalg mention a
limit for the ferrioxalate actinometer under intensive irradiation
(high-power light) where a drop of quantum yield leads to
systematically wrong measurements.15 The formed iron(II
+)oxalate has a very low solubility in water (0.097 g/100 mL),
causing the formation of solids. These precipitates are highly
undesired, particularly in flow reactors, where this creates
clogging of the tubing or reactor channels. Precipitation not
only causes shadowing of the reactants but also affects the
chemical equilibria of the actinometer system. In addition,
CO2 gas is released during the reaction causing gas bubbles,
resulting in unreliable residence times, which impedes proper
2393
pubs.acs.org/OPRD Article
calculation of the reaction kinetics. This gas and precipitate
formation at higher conversions thus leads to systemically
wrong measurements.16 As a consequence, characterization of
photoflow reactors using ferrioxalate is only possible at low
conversion grades. To obtain sufficient kinetic data for a
reliable photon flux calculation, multiple residence times are
needed below the one where the product starts to precipitate.
However, when implementing these restrictions, the improved
high-intensity lamps and higher reactor volumes (particularly
important for scale-up) in larger flow systems lead to
unrealistically short residence times and unrealistically high
flow rates. In this paper, we present and compare three
practical approaches to address these issues.
■ DETERMINATION OF THE PHOTON FLUX
The determination of the conversion of the ferrioxalate
reaction X (%) is performed spectrophotometrically. In our
experiments, 2 mL (V0) of a collected sample was first diluted
with 50 mL of deionized water (V1 = 52 mL). Next, 2 mL (V2)
of this diluted sample was mixed with 5 mL of o-
phenanthroline (Co‑phenanthroline = 1 g/L), 5 mL of buffer
solution (36 mL of sulfuric acid 1 mol/L and 60 mL of sodium
acetate 1 mol/L, diluted to 100 mL with water), and finally 5
mL of water (V3 = 17 mL). Samples were left to equilibrate in
the dark for 1 h before absorbance measurement. The
conversion was calculated according to the formula below,
where A510 represents the absorbance of the mixture (at 510
nm), CA0 is the actinometer concentration, and k is the slope of
the calibration curve A510 = f(CFe2+) obtained using standard
solutions of FeII+ and o-phenanthroline. In our case, the slope
of the calibration curve k was identified to be 10 792 L/mol
(see the Supporting Information).11
C FeII + 1 A510 V3 V1
X (%) = =
CA0 CA0 k V2 V0 (1)
Using this conversion and the corresponding residence time,
the experimental rate of the reaction can be determined. This
reaction rate is also described in the mathematical model for
photochemical conversion rates (eq 2). It consists in
discretising the range of wavelengths covered by the lamp
emission into several elementary intervals Δλi in which the
wavelength-dependent parameters (quantum yield (φ), ma-
terial transmittance Tλ, and Napierian molar absorption
coefficient κλ) can be averaged and considered constant.11
The theoretical rate is then calculated using the following
formula
dX qp,0
= [T i i g i f i ]
dt CA0 Vr i
f=1 e( A e)
=1 e( k iCAl)
(2)
The model contains several parameters, where qp0 is the
incident photon flux, CA0 corresponds to the actinometer
concentration, Vr is the reactor volume, and gλi represents the
wavelength-dependent density function of the light source
emission. The fλi factor is defined as the fraction of incident
light absorbed at wavelength λ, calculated using the above
equation, where κλ is the Napierian molar absorption
coefficient (see the Supporting Information), CA is the residual
actinometer concentration, and l is the optical pathlength.
Complete absorbance is approximated, whereby the calculation
and the variation of actinometer concentration with residence
https://doi.org/10.1021/acs.oprd.2c00079
Org. Process Res. Dev. 2022, 26, 2392−2402
Organic Process Research & Development
time become independent of the actual optical pathlength.
Therefore, the average inner diameter can be used for further
calculations (see the Supporting Information).12 This model
can be used for polychromatic and monochromatic lamps, as
long as the density function of the light source is known.
Preferably, five data points are needed to perform a reliable
and accurate calculation of the photon flux, by minimizing the
summed quadratic error between the experimental and the
modeled conversion for every data point as shown in the
following equation
R= [(X )imodel (X )iexp ]2
i (3)
Through minimizing R, the photon flux below 500 nm (qp0
parameter, eq 2) is obtained, which creates the best fit between
the experimental and modeled reaction rates. Subsequently,
the total photon flux is calculated. If the light source emits
wavelengths above 500 nm, the photon flux is corrected for the
complete spectrum as a percentage of light below 500 nm
(obtained via the light source density function gλi).11
■ BACKGROUND INFORMATION ON DEVICE AND
DESIGN EFFICIENCIES
When purchasing a light source, different device characteristics
should be mentioned in the manual. The power input (W) of a
light source is defined as the amount of electrical power that
enters the device from the electricity grid. Usually, a radiant
power (W) is mentioned as well, which is defined as the
amount of light emitted by the light source. The ratio of the
radiant power and the power input gives information on how
efficiently the light source device transforms the incoming
electricity into light energy, which is called the device
efficiency.
The photon flux, expressed in mol of photons/s or Einstein/
s, can be transformed to the Watt unit using Planck’s law (E =
hc/λ). This way, the photon flux now presents the amount of
light energy that enters the reaction mixture, expressed in
power. The ratio of this photon flux (W) and the radiant input
(W) refers to the fraction of light that actually reaches the
reaction mixture compared to the amount of light emitted by
the light source. This is called the reactor system efficiency,
and it is a measure of how well the lamp design (orientation
and shape of light beams) matches the reactor design
(orientation and shape of reactor channels). These concepts
are schematically shown below (Figure 1).
Figure 1. Schematic representation of the device efficiency and the
reactor system efficiency related with photoreactors.
■ RESULTS AND DISCUSSION
Since the influence of several process parameters on the
ferrioxalate reaction is not unambiguously reported, fine-tuning
of the actinometry conditions was first performed using a small
flow reactor equipped with a low-power light-emitting diode
(LED) lamp. Second, in the context of scale-up and
intensification of photochemical processes, experiments have
been performed to identify the practical implications of high-
2394
pubs.acs.org/OPRD Article
intensity lamps in a small reactor (Vapourtec UV-150 reactor)
on the actinometry experiments for the determination of the
photon flux (q0). Afterward, the influence of reactor volume
increase on the actinometry was studied using an oscillatory
plug flow reactor (HANU 2X 15).
Control Experiments: Tuning the Actinometry Re-
action Conditions (Vapourtec UV-150 Reactor). Given
the variability in reported actinometry conditions, some time
was first spent on the optimization of the actinometry reaction
itself with a low-power LED, before heading to more powerful
lamps and larger reactor volumes. It became clear from the
literature that although the ferrioxalate actinometry reaction
was already extensively investigated, some contradicting
conclusions were made. Several parameters were reported to
influence the reliability of the actinometry performance.
Depending on the cited source, a 0.05 or 0.2 M concentration
of H2SO4 is reported to be optimal in the reagent mixture.
Apart from the reaction pH, Wriedt et al. also mentioned the
degradation of the ferroin complex in the final analytical buffer
at a very low pH (<1.3) and recommends an optimal analytical
pH range of 3.7−5.7.17 Additionally, some uncertainty exists
on the influence of the timing of the absorbance measurement
on the data, leading to some contradicting conclusions in the
literature.11,18 In combination with the mention of the need for
mixing right before analysis, a thorough evaluation seems to be
of major importance to obtain reliable data.10 In addition, since
the scope of this research is the use of powerful lamps, also the
phenanthroline concentration becomes important to ensure
complete complexation of FeII+ when higher conversions are
obtained. As such, a screening was performed to determine the
influence of the aforementioned parameters using the
commercial Vapourtec UV-150 flow reactor (irradiated volume
= 2 mL), combined with a low-power polychromatic cold
white LED lamp (power input = 60 W, radiant power = 10.6
W, further details in the Supporting Information).
Previous literature reports consistently indicate that the
analytical mixture should be left to equilibrate for at least 1 h.
To gain insight into the further influence of the timing of
absorbance measurement, three longer equilibration times
were added. One closer to the 1 h measurement (5 h) to
visualize fast changes in absorbance and two to account for
longer equilibration times: 20 and 30 h. For both the 0.05 M
and 0.2 M H2SO4 concentrations, this influence of the timing
of absorbance measurement was evaluated at four different
phenanthroline concentrations (0.1, 0.5, 1, and 2 g/L). The
calculated photon fluxes are shown in Tables 1 and 2,
respectively. The flow rates needed in this 2 mL coil were
between 12 and 60 mL/min. These settings are within the
range of the system’s pumps, and all experiments could be
performed without the formation of solids or CO2 bubbles. For
every residence time experiment, two samples were collected
and the average absorbance was used for further calculations.
Immediately obvious is the fact that poor data is obtained
for the experiments using a 0.05 M H2SO4 concentration
(Figure 2A). Even upon removal of the obvious outliers, all
0.05 M H2SO4 experiments resulted in inconsistent conversion
plots with R2 below 0.93, and calculated photon fluxes far
below those of the more consistent 0.2 M experiments (see the
Supporting Information). It can thus be concluded that a
sulfuric acid concentration of 0.05 M delivers unreliable data.
For the 0.2 M H2SO4 experiments, good conversion plots and
consistent calculated photon fluxes were achieved for the 0.5,
1, and 2 g/L phenanthroline concentrations. As expected,
https://doi.org/10.1021/acs.oprd.2c00079
Org. Process Res. Dev. 2022, 26, 2392−2402
Organic Process Research & Development pubs.acs.org/OPRD Article

Table 1. Photon Flux (μEinstein/s) Determined Using Four Different o-Phenanthroline Concentrations with a Sulfuric Acid
Concentration of 0.05 M in the Start Mixturea
0.05 M H2SO4 concentration 0.1 g/L 0.5 g/L 1 g/L 2 g/L
1h 2.87 (3: 3.50) 6.73 (4: 7.31) 4.40 (3: 4.92) 6.69 (4: 5.96)
5h 3.57 (3: 4.18) 8.45 (4: 9.43) 4.27 (3: 5.07) 5.61
20 h 3.88 10.77 (4: 12.23) 3.43 (3: 4.03) 5.78
30 h 5.81 10.46 (4: 11.76) 3.93 (3: 4.94) 4.54 (2: 7.33)
a
The absorbance is measured at four different times after o-phenanthroline addition. Five equally spaced residence times were used to determine
the conversion rate and to calculate each photon flux. Values obtained after manual removal of apparent outliers are shown in parentheses,
alongside the remaining number of data points.

Table 2. Photon Flux (μEinstein/s) Determined Using Four
Different o-Phenanthroline Concentrations with a Sulfuric
Acid Concentration of 0.2 M in the Start Mixturea
0.2 M H2SO4 concentration 0.1 g/L 0.5 g/L 1 g/L
1h 4.25 13.14 13.72
5h 5.64 13.34 13.67
20 h 4.35 13.30 14.59
30 h 3.88 13.98 14.64
a 6
The absorbance is measured at four different times after o-
phenanthroline addition. Five equally spaced residence times were
used to determine the conversion rate and to calculate each photon
flux.
using a 0.1 g/L phenanthroline concentration results in an
underestimation of the received photon flux, due to incomplete
complexation of FeII+. The timing of the analysis seems to be
of very limited effect. For the experiments using 0.5 and 1 g/L
phenanthroline concentrations, a maximum increase of 6.5%
over a 30 h storage is measured, while a similar decrease is
observed for the 2 g/L experiments. Note that surrounding
light can still convert a significant amount of ferrioxalate, even
in the analytical solution, and as such, interfere with the
absorbance measurements (see the Supporting Information).
This could possibly explain the small increase in absorbance
measured over time, as the samples need to be opened for each
consecutive analysis to fill the cuvettes.
As we conclude that higher acidity is needed during the
photoredox reaction, but Wriedt reported degradation of the
ferroin complexes under too acidic conditions, the pH of the
final analytical samples was studied as well (Table 3). For the
starting mixtures containing 0.05 M and 0.2 M H2SO4, pHs of
Table 3. pH of Analytical Samples during an Actinometry
Experiment Using 0.05 and 0.2 M H2SO4 Concentrations,
Respectively
2 g/L residence time (s) 0.05 M H2SO4 0.2 M H2SO4
13.15 0 1.97 2.10
12.72 2 1.99 2.11
13.17 3 1.99 2.12
11.14 4 1.97 2.10
1.98 2.07
8 1.97 2.12
1.3 and 0.81, respectively, were measured. The undiluted buffer
used during the analysis was identified to have a pH of 1.55.
The pH of an analytical sample (5 mL buffer, 5 mL water, 5
mL o-phenanthroline solution, 2 mL sample) for every
residence time was measured after 1 h (see Table 3). From
this data, it is clear that the acid concentration in the reaction
mixture does not significantly influence the final pH in the
analytical samples, and this pH is well above the mentioned
limit of 1.3, even for the higher sulfuric acid concentration.
Additionally, the pH remained constant over time, and since
the same value is found for each residence time, the amount of
ferrous(II+)oxalate does not influence the final pH either.
Quite remarkably, a slightly higher final pH was observed for
the higher initial acid concentration. The reasons for this effect
are still unclear, but we can conclude that the final acidity of
the analytical samples is not the main cause of the inconsistent
data of the 0.05 M H2SO4 experiments.
From the above experiments, a sulfuric acid concentration of
0.2 M in combination with measuring the absorbance 1 h after
the addition of a 1 g/L phenanthroline solution, was defined to

Figure 2. Conversion plots of the 0.05 M (A) and 0.2 M (B) H2SO4 concentration experiments using a phenanthroline concentration of 1 g/L,
where the ferroin absorbance is measured after 1 h. For every residence time, two samples are taken and the average value is used for calculations.

2395 https://doi.org/10.1021/acs.oprd.2c00079
Org. Process Res. Dev. 2022, 26, 2392−2402
Organic Process Research & Development
be the most optimal set of parameters for future actinometry
experiments. Using these settings, the final photon flux of the
polychromatic cold white LED lamp was determined to be
13.7 μEinstein/s.
Small Reactor with a High-Intensity Lamp (Vapourtec
UV-150 Reactor). With these optimized parameters in hand,
further screening of high-intensity lamps was performed. To
investigate the practical implications, a high-power 450 nm
LED (power input = 150 W, radiant power = 108 W) was used
in the Vapourtec UV-150 reactor with a 2 mL reactor coil.
Initially, residence times of 20−60 s were used to perform the
actinometry experiments. However, the reactor coil was filled
with solids and clogged after only 1.5 min. The yellowish solid
was identified as the ferrous(+II)oxalate product (Figure 3).
Figure 3. Photograph of the formation of the yellow solids in the 2
mL reactor coil (Vapourtec UV-150).
The presence of solids should be avoided at all times
because of their physical and chemical drawbacks, as
mentioned before. Shorter residence times were thus chosen
to perform the reaction: 2, 4, 6, 8, and 10 s. Since these are
extremely short and the shortest achievable residence time in
this 2 mL coil using the native Vapourtec V-3 peristaltic pumps
is 12 s, the pump was replaced with a Knauer K1800 plunger
pump to access higher flow rates (1−250 mL/min). Never-
theless, performing the reaction with a 10 s residence time
again resulted in the formation of gas pockets and solids and
Figure 4. Conversion plot of ferrioxalate as a function of residence time.
2396
pubs.acs.org/OPRD Article
subsequently clogging. As five data points are desired for
accurate calculations, new residence times were chosen: 2, 3, 4,
6, and 8 s (Figure 4). At the highest residence time, some very
small gas pockets were still present, causing a somewhat
shorter residence time and a small shift of the data point to the
right. Since this could have a small influence on the reliability,
this data point was not taken into account for the final photon
flux calculation.
Despite the fact that the practical issues could be largely
avoided, the collected samples of the three highest residence
time experiments still contained very small solid particles.
Since the used residence times were already extremely short,
further shortening these was not possible. If solids are absent in
the collected sample, 2 mL can be used for further dilution and
the standard analytical protocol can be performed as described
in the theoretical background. Since small particles are now
present, it is important to take these into account during
further analysis. Consequently, to ensure representative
sampling, the complete sample should be used for further
analysis to ensure a reliable absorbance measurement.
Therefore, the volume of the collected sample is calculated,
using the net mass of the sample and the density of the
mixture, to determine the appropriate amount of water needed
for the 1:25 dilution. Subsequently, before the addition of 2
mL of the diluted sample to the analytical vial with 5 mL of
water, 5 mL of buffer, and 5 mL of phenanthroline, it was
sonicated for 20−30 min to bring all solids into solution. Of
the final 17 mL of analytical solution, 1 mL was transferred to a
cuvette for absorbance measurement. The inconvenience of
this sonication procedure exemplifies the problems associated
with the determination of high photon fluxes. Using the
calculation model as described above, the total photon flux
received in the Vapourtec UV-150 reactor for a 2 mL coil and a
high-power 450 nm LED was found to be equal to
q0 = 31.3 Einstein/s
Comparison of the Reactor System Efficiency within
the Vapourtec UV-150. To date, the photon flux of two
different light sources was determined in the same small-
volume Vapourtec UV-150 reactor. Since the reactor coil
remained constant for both cases, the reactor system efficiency
can be used to evaluate changes in the light source design and
https://doi.org/10.1021/acs.oprd.2c00079
Org. Process Res. Dev. 2022, 26, 2392−2402
Organic Process Research & Development
its match to the reactor channel geometry. The reactor system
efficiency was found to be 28.5% for the cold white LED lamp,
but only 7.5% for the high-power 450 nm lamp (see the
Supporting Information). The difference in lamp design�
indicating a more inhomogeneous irradiation for the 450 nm
lamp (Figure 5)�could explain the difference in efficiency of
21%. It is clear that photoreactor design improvements should
go beyond a simple increase of radiant power.
Figure 5. Comparison of lamp design: cold white LED (left) and 450
nm LED lamp (right).
High-Volume Reactor with High-Intensity Lamps
(HANU 2X 15−15 mL). To study practical actinometry in
higher-volume reactors with powerful lamps, we next moved to
the HANU 2X 15. This patented photoreactor is a continuous-
flow reactor based on the COSTA technology (Figure 6), with
Figure 6. Visualization of the HANU 2X 15 photoflow reactor.
an internal volume of 15 mL. The process space consists of a
split-and-recombine flow pattern (using static mixers) operated
in conjunction with a pulsator, a device that superimposes a
periodic oscillation on the net flow generated by a metering
pump. This oscillatory flow regime and repeated split-and-
recombine flow path ensure adequate mixing regardless of the
net flow rate. Consequently, pulsator amplitude and frequency
are introduced as new process parameters. Light sources for
2397
pubs.acs.org/OPRD Article
the HANU 2X 15 reactor are customized and purchased from
Peschl Photoreactors.19,20
Despite the higher irradiated volume of this HANU 2X 15
reactor, the standard procedure for actinometry as mentioned
above is still achievable, though challenging in the context of
the required flow rates. However, due to the availability of this
photon flux as a reference value, the reactor is a perfect model
to study simplified actinometry methods, which can then be
used on even larger reactors. In addition, the HANU 2X 15
reactor has control over the residence time distribution (RTD)
by adapting the pulsator characteristics. This allows an accurate
study of the influence of different RTDs on the actinometry
results.
One of the determining parameters needed during the
calculation of the photon flux is the reactor volume. Therefore,
the irradiated volume of the HANU 2X 15 reactor was
calculated to ensure that dead volumes are not included in
further calculations. When taking the volume of the static
mixers into account, the irradiated volume was determined to
be 12.03 mL.
As described above, short residence times are a requirement
for good and reliable actinometry in photoflow reactors with
high-power light sources. Since the irradiated volume is now 6
times larger, very high flow rates are needed to achieve such
residence times. Limited by the achievable pumping rate, 6, 8,
10, 12, and 14 s were chosen. Although the lowest residence
time was tripled with regard to the Vapourtec experiments, one
should note that this still corresponds with a maximum flow
rate of 120 mL/min, which is near the maximal achievable rate
for the HANU 2X 15 with the available pumps. For the three
highest residence times, very small solids were visible in the
collected sample, and the complete sample was developed into
the analytical mixture to avoid unrepresentative sampling due
to sedimentation. Although excellent conversion plots were
obtained, it is preferred to avoid these additional handling
steps and the associated excessive use of reagents.
Since the pulsator amplitude and frequency are now
introduced as new process parameters, optimal conditions
should be picked. Using tracer experiments (Thymol blue), the
residence time distribution (RTD) in the HANU 2X 15
reactor could be determined (see the Supporting Information).
The effect of different pulsator amplitudes on multiple flow
rates is demonstrated in Figure 7. A value of 15% (which
corresponds to a stroke volume of 0.062 mL) and a maximum
frequency (3 Hz) were found to give the most desirable narrow
RTD. An amplitude of 70% (which corresponds to a stroke
volume of 0.29 mL) induces a shift of the RTD curve to the
right, with this effect being most abundant at lower flow rates.
It should be noted that the measured RTDs cannot be used to
calculate the actual residence time in the irradiated volume,
due to physical constraints (nonilluminated volumes of
connectors and tubing, pressure equilibration upon tracer
injection, etc.). As such, the theoretical residence time
(irradiated volume/flow rate) must be used in the photon
flux calculations. Figure 7 thus only provides an indication of
the width and location of the RTDs for different pulsator
amplitudes at different flow rates.
Using the optimal pulsator amplitude (15%), the photon flux
was calculated for the monochromatic 365 nm LED lamp
(power input = 100 W, radiant power = 43.2 W)
q0 = 31.3 Einstein/s
https://doi.org/10.1021/acs.oprd.2c00079
Org. Process Res. Dev. 2022, 26, 2392−2402
Organic Process Research & Development
Figure 7. Visualization of the residence time distribution in the HANU 2X 15 reactor. The effect of different pulsator amplitude values at multiple
flow rates is demonstrated. Equal tracer concentrations were used for all experiments. The distribution curves were normalized based on their
respective maxima to improve readability.
The usage of a nonoptimal pulsator amplitude of 70%, and
consequently a wide residence time distribution, on the
actinometry experiment was studied as well. To avoid the
very small particles, the residence times were slightly changed
to 4, 6, 8, 10, and 12 s (180 mL/min up to 60.5 mL/min) for
these experiments. Practically, 4 s is the shortest residence time
possible in the HANU 2X 15 reactor with the available pumps.
Now, only two experiments (10 and 12 s) still contained some
very small particles in the collected sample. Hereby, the
photon flux was determined to be
q0 = 33.0 Einstein/s
Despite the use of pulsator parameters to obtain a wide RTD,
the calculated photon flux changed only slightly. In Figure 7, it
is visible that using an amplitude of 70% causes a shift of the
RTD curve to the right, explaining the small overestimation.
However, the effect of different pulsator amplitudes is found to
be very small at these high flow rates (Figure 7). It is important
to mention that although it does not cause any analytical
issues, it is not possible to obtain five homogeneous samples
during the actinometry experiments in the HANU 2X 15
reactor. This once again emphasized the need for an easier and
more practical method to perform actinometry in higher-
volume flow reactors.
Design and Study of Different Concepts to Facilitate
Actinometry in Larger Systems. Dimmed Emitter Actino-
metry in Larger Systems (DETAILS). As mentioned before,
within ferrioxalate actinometry experiments, it is needed to
keep the conversion below 40% to avoid the formation of gas
pockets and solids. This is inevitably accompanied with
unachievably high flow rates, as production-oriented systems
are not designed to evaluate the experimental space leading to
low conversions. Certainly, when the volumes of photoflow
reactors continue to increase in the future to facilitate and
realize upscaling, there is an urgent need for a simpler way to
perform reliable actinometry. One very simple approach is to
make use of the dimmability of LED lamps. As the reactor
2398
pubs.acs.org/OPRD Article
design efficiency is independent of the radiant power, it is
possible to decrease the photon flux to 50% (or lower) to
reduce the conversion grade and allow full actinometry
experiments with (relatively) low flow rates. For LED modules
within their recommended operational range and at constant
temperatures, this correlation is conveniently linear.21 The
obtained photon flux for this lower light intensity, can then be
extrapolated to the 100% light intensity, provided that a known
correlation is present between the radiant power emitted and
the electrical power consumed by the light source. Upon
performing actinometry for a light intensity at 50% (5
residence times, 2 samples, 2 technical repeats), the calculated
photon flux was indeed reduced by a factor of 2 and as
opposed to the actinometry at 100% intensity, solid formation
was completely avoided. The photon flux was now determined
to be 14.7 μEinstein/s, which extrapolates to 29.4 μEinstein/s
at full intensity, a close approximation of the previously
determined total photon flux (31.3 μEinstein/s). It should be
noted however that the inaccuracy of the dimmer’s setpoint (1
kΩ potentiometer: relative intensity 0.5 ± 0.1) was reflected in
the technical repeats. As shown in Figure 8, the associated
reaction rates diverged quite heavily, and based on either one
of these repeats, the estimated photon flux could have been
over- (36.8 μEinstein/s) or underestimated (22.2 μEinstein/s).
Once linearity between the photon flux and the light
intensity is confirmed, this concept can be used to facilitate the
determination of the photon flux in larger reactors, while
avoiding the unrealistic high flow rates and/or solid formation.
If the setup allows it, the DETAILS concept is operationally
very simple. It is applicable to any pump rate and reactor
volume combination. The dimming percentage can be adjusted
as required to keep the conversion below 40% to avoid solid
and gas formation. Nevertheless, still high volumes of reagents
are needed to perform all five experiments in the complete
reactor volume. In addition, the concept depends on a reliably
dimmable light source, and this may need to be confirmed
https://doi.org/10.1021/acs.oprd.2c00079
Org. Process Res. Dev. 2022, 26, 2392−2402
Organic Process Research & Development pubs.acs.org/OPRD Article

below the actual photon flux of 31.3 μEinstein/s. When

inspecting the LED strip at the selected region, the presence of
a screw was observed, which may have caused a lower local
light intensity and consequently an underestimation of the
total photon flux. Additionally, selecting the very end of the
reactor may be suboptimal due to larger shadowing effects of
the stainless-steel reactor ridges (see Figure 9).
As such, the irradiated segment was moved closer to the
center of the reactor, ensuring the absence of LED screws in
that region (Figure 10). Because this area does not contain the

Figure 8. Conversion of ferrioxalate using the DETAILS concept at

50% light intensity. Two technical repeats, which consist of manually
resetting the dimming percentage, are shown as separate conversion
plots.

through experiments at different light intensities or through a Figure 10. Shielded HANU 2X 15 reactor with the irradiated volume
number of technical repeats. in the middle of the reactor.
Segment Narrowed Actinometry in Larger Systems
(SNAILS). When accurate reduction of radiant power (dimming output of the reactor, the irradiated volume was now 1.54 mL.
of the light source) is not available, a second option is to Consequently, extrapolation should now be performed by
maintain full light intensity but reduce the effectively irradiated multiplying the measured photon flux by 7.8 (12.03 mL/1.54
volume. This can easily be achieved by shielding part of the mL). After actinometry, a photon flux of 3.09 μEinstein/s was
reactor from the lamp. Knowing the reduced volume factor found for the irradiated volume, which resulted in a flux of
(original volume/new volume), and assuming a homogeneous 24.12 μEinstein/s for the entire HANU 2X 15 reactor. Despite
local volumetric rate of photon absorption (LVRPA) across the improvement after relocating the irradiated volume, the
the entire reactor, the total photon flux can be estimated by extrapolated value is still a significant underestimate. As the
extrapolation of the segment’s calculated photon flux to the previously used cardboard is quite thick to ensure full blockage
original reactor volume. In addition, if the smaller volume of the photons, this does induce some undesired shadowing of
segment can be chosen near the end of the reactor (avoiding the mixture at the start and end of the selected segment.
large dead volumes), the amount of wasted reagents to achieve To counter this effect, black electrical tape (<0.1 mm) was
steady-state conditions for each residence time and to next used for shielding instead of cardboard (Figure 11). Using
subsequently collect representative samples is greatly reduced.
To verify the possibilities of this extrapolation route, the
length of the HANU illumination window was reduced to give
an irradiated volume of 1.42 mL. This volume was located at
the end of the reactor, and cardboard with a thickness of
approximately 0.5 cm was used for the shielding (Figure 9).
After performing the actinometry experiment (5 residence
times, 2 samples each), a photon flux of 2.44 μEinstein/s was
obtained. This value was multiplied by 8.5 (12.03 mL/1.42
mL) to extrapolate the photon flux to the complete reactor
volume. As such, an estimated photon flux of 20.7 μEinstein/s
is obtained for the entire reactor. This value is significantly Figure 11. Shielded HANU 2X 15 reactor with black electrical tape.

the same irradiated volume, a photon flux of 3.65 μEinstein/s

was obtained. After extrapolation, this corresponds to a flux of
28.5 μEinstein/s for the complete reactor. Comparing this to
the true photon flux of 31.3 μEinstein/s (standard method),
this is still an underestimation of 8.9%. Practically however,
this is the best approximation we could achieve for this system
using this methodology. Ideally, the shielding material should
be as close as possible to the reaction mixture. A better
approximation could possibly be obtained by applying the tape
under the glass, right above the reaction mixture. However, this
is not practically possible in this system.
Figure 9. Shielded HANU 2X 15 reactor with the irradiated volume Extrapolation from smaller reactor segments is clearly not
at the end of the reactor. straightforward. Due to differences in the LVRPA across the
2399 https://doi.org/10.1021/acs.oprd.2c00079
Org. Process Res. Dev. 2022, 26, 2392−2402
Organic Process Research & Development
reactor, several factors such as the location of the irradiated
segment, small dimensional inconsistencies, shielding material
and its thickness, reactor wall thickness, etc. all have a non-
negligible role. When applying this concept for photon flux
determination, a good estimate can only be obtained if a
representative reactor area is used. Another potential strategy
to further increase the reliability of this method is the
calculation of an average photon flux from multiple experi-
ments where the irradiated area is shifted to different locations
on the reactor (with the disadvantage of extra work). However,
the SNAILS principle is applicable to any system because the
segment can be made as small or as large as needed to fit the
pump capabilities and lower volumes of reagents are needed
before representative samples can be collected for each data
point.
Time-Resolved Actinometry in Larger Systems (TRAILS).
Since the SNAILS concept inevitably involves a small error in
the final estimate of the photon flux, a third concept was
studied as well. This approach is based on the concepts
presented by Mozharov et al., where kinetic studies were
performed by flow rate manipulation to achieve multiple
residence time data points in a single experiment.22 In their
study, the flow rate is programmed to increase from a low to a
high flow rate. Following this increase, several samples are
collected correlating to decreasing residence times. When
applying this approach to photochemical reactors, it is further
simplified, as the reaction can be stopped instantly by
switching off the lamp. Achieving data for lower residence
times can then be obtained by maintaining the flow rate in the
dark and collecting the corresponding fractions. It should be
noted that this is in fact a further refinement of the previously
described SNAILS principle, as each collected sample
represents the conversion in a certain segment of the reactor,
with a different irradiation time. However, because these
volumetric segments now span the entirety of the reactor,
differences in LVRPA will be averaged out in the final
calculation.
Practically, the longest residence time without gas formation
or precipitation is determined. This can be performed by
starting at the system’s highest pumping rate and gradually
decreasing it. This flow rate is then used to perform a single
actinometry experiment. After achieving steady state, the lamp
is shut off. The reactor is now completely filled with the
reaction mixture and contains molecules that all have different
residence times, depending on their location in the reactor
(Figure 12). Molecules located toward the end of the reactor
have the highest residence time, while molecules near the inlet
have only been irradiated very shortly. Next, pumping is
continued at a low flow rate which allows collecting the reactor
volume in different fractions (see the Supporting Information).
Every fraction now corresponds to a data point of which the
conversion can be calculated.
Because the TRAILS methodology relies on collecting data
concerning the reaction conversion near the inlet but does so
minutes later at the outlet of the reactor, the RTD becomes an
important influencing parameter. As such, the influence of a
narrow and wide RTD, both at the stage of steady-state
irradiation and the stage of sample collection, was included in
the evaluation.
First, the pulsator was switched off during the collection of
the different fractions (green curve in Figure 7). During the
steady-state ferrioxalate reaction, the pulsator parameters were
set to pulsator amplitudes of 15 and 70%, respectively, to
2400
pubs.acs.org/OPRD Article
Figure 12. Visualization of the TRAILS concept: the HANU 2X 15
reactor is split up into six parts, and the residence time boundaries of
every fraction are shown. The average residence times of these
boundaries are used for further calculations.
obtain a narrow and wide RTD. After performing the reaction
with both a narrow and wide RTD, the photon fluxes were
calculated
q0, narrow RTD = 24.4 Einstein/s
q0, wide RTD = 49.3 Einstein/s
While the obtained results were sub-par, the use of a narrow
RTD results in a better approximation of the expected 31.3
μEinstein/s. Despite the high flow rate during the reaction, a
narrow RTD during the illuminated stage is of importance for
a more accurate approximation.
Next, the effect of the RTD during sample collection was
studied. In this case, the pulsator was switched on during the
collection of the different fractions, once with 15% pulsator
amplitude for a narrow RTD and once with 70% pulsator
amplitude for a wide RTD. During the photoreaction, the
pulsator parameters were set to achieve a narrow RTD in both
cases. The following photon fluxes were obtained
q0, narrow RTD = 31.8 Einstein/s
q0, wide RTD = 5.5 Einstein/s
The effect of a poor residence time distribution during sample
collection is quite clearly detrimental to the accuracy of the
final results. However, the obtained photon flux when
imposing a narrow RTD during both the photoreaction stage
and the in-the-dark collection stage results in a very accurate
approximation of the photon flux (31.8 vs 31.3 μEinstein/s,
error of 1.6%, conversion plot shown in Figure 13). Also the
small amounts of precipitate observed in the samples collected
during the standard procedure are now completely avoided.
In conclusion, the time-resolved methodology provides
accurate values for photon flux determination. In addition, it
is very fast and easy to perform as only one experiment is
needed for the determination of the reaction kinetics.
Nevertheless, this method is only applicable if conversions
below 40% are obtained at the system’s maximum flow rate.
■ CONCLUSIONS
In conclusion, three new concepts were successfully designed
and developed to simplify actinometry experiments in newly
developed photoflow reactors. Based on an initial optimization
of the ferrioxalate actinometry analytical procedure (evaluating
a number of contradicting literature statements), the dimmed
https://doi.org/10.1021/acs.oprd.2c00079
Org. Process Res. Dev. 2022, 26, 2392−2402
Organic Process Research & Development
Figure 13. Conversion of ferrioxalate using the TRAILS concept with a narrow RTD during reaction and collection of the different fractions.
emitter (DETAILS), segment-narrowed (SNAILS), and time-
resolved (TRAILS) actinometry concepts have shown their
advantages and disadvantages. Each has its own applicability
range for photon flux determination in larger systems. The
latter, however, was found to give an accurate approximation of
the photon flux, by determining the kinetics of a photo-
chemical reaction in a single experiment.
■
*
ASSOCIATED CONTENT
sı Supporting Information
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.oprd.2c00079.
Description of the experimental work and a calculation
template; additional technical information on the
different light sources and experimental details of the
different procedures; and charts representing all data
(PDF)
Calculation example: HANU 2X 15−365 nm−15 mL
(XLSX)
■ AUTHOR INFORMATION
Corresponding Author
Thomas S. A. Heugebaert − SynBioC Research Group,
Department of Green Chemistry and Technology, Faculty of
Bioscience Engineering, Ghent University, B-9000 Ghent,
Belgium; orcid.org/0000-0003-3297-2798;
Email: Thomas.Heugebaert@UGent.be
Authors
Bavo Vandekerckhove − SynBioC Research Group,
Department of Green Chemistry and Technology, Faculty of
Bioscience Engineering, Ghent University, B-9000 Ghent,
Belgium
Nicola Piens − Ajinomoto Bio-Pharma Services, 9230
Wetteren, Belgium
Bert Metten − Ajinomoto Bio-Pharma Services, 9230
Wetteren, Belgium
Christian V. Stevens − SynBioC Research Group, Department
of Green Chemistry and Technology, Faculty of Bioscience
Engineering, Ghent University, B-9000 Ghent, Belgium;
orcid.org/0000-0003-4393-5327
Complete contact information is available at:
2401
pubs.acs.org/OPRD Article
https://pubs.acs.org/10.1021/acs.oprd.2c00079
Author Contributions
B.V. conducted all wet lab actinometry experiments, analysis,
and calculations N.P. supported building of the lab actino-
metry setup and supervised experiments. B.M. supervised and
set up wet lab actinometry experiments. C.V.S. designed and
supervised the project and edited the manuscript. T.S.A.H.
designed and supervised the project and contributed to the
implementation of the research, analysis of the results, and
writing of the manuscript.
Funding
The authors are indebted to VLAIO�Belgium (Baekeland
HBC.2021.0198) for financial support.
Notes
The authors declare no competing financial interest.
■ REFERENCES
(1) Ballini, R.Eco-Friendly Synthesis of Fine Chemicals || Chapter 2.
The Contribution of Photochemistry to Green Chemistry, 2009 .
(2) Li, P.; Terrett, J. A.; Zbieg, J. R. Visible-Light Photocatalysis as
an Enabling Technology for Drug Discovery: A Paradigm Shift for
Chemical Reactivity. ACS Med. Chem. Lett. 2020, 11, 2120−2130.
(3) Lipp, A.; Badir, S. O.; Molander, G. A. Stereoinduction in
Metallaphotoredox Catalysis, Angewandte Chemie International
Edition. Angew. Chem., Int. Ed. 2021, 60, 1714−1726.
(4) Liu, J.; Lu, L.; Wood, D.; Lin, S. New Redox Strategies in
Organic Synthesis by Means of Electrochemistry and Photochemistry.
ACS Cent. Sci. 2020, 6, 1317−1340.
(5) Williams, J. D.; Kappe, O. C. Recent advances toward sustainable
flow photochemistry. Curr. Opin. Green Sustainable Chem. 2020, 25,
No. 100351.
(6) Wypych, G. Handbook of UV Degradation and Stabilisation, 2nd
ed.; ChemTec Publishing, 2015.
(7) Cambié, D.; Bottecchia, C.; Straathof, N. J. W.; Hessel, V.; Noël,
T. Applications of Continuous-Flow Photochemistry in Organic
Synthesis, Material Science and Water Treatment. Chem. Rev. 2016,
116, 10276−10341.
(8) Bonfield, H. E.; Knauber, T.; Lévesque, F.; Moschetta, E. G.;
Susanne, F.; Edwards, L. J. Photons as a 21st century reagent. Nat.
Commun. 2020, 11, No. 804.
(9) Kuhn, H. J.; Braslavsky, S. E.; Schmidt, R. Chemical actinometry
(IUPAC Technical Report) Pure and Applied Chemistry. Pure Appl.
Chem. 2004, 76, 2105.
https://doi.org/10.1021/acs.oprd.2c00079
Org. Process Res. Dev. 2022, 26, 2392−2402
Organic Process Research & Development pubs.acs.org/OPRD Article

(10) Kayahan, E.; Jacobs, M.; Braeken, L.; Thomassen, L.; Kuhn, S.;
Van Gerven, T.; Leblebici, M. Dawn of a new era in industrial
photochemistry: the scale-up of micro- and mesostructured photo-
reactors. Beilstein J. Org. Chem. 2020, 16, 2484−2504.
(11) Aillet, T.; Loubiere, K.; Dechy-Cabaret, O.; Prat, L. Accurate
Measurement of the Photon Flux Received Inside Two Continuous
Flow Microphotoreactors by Actinometry. Int. J. Chem. React. Eng.
2014, 12, 257−269.
(12) Roibu, A.; Fransen, S.; Leblebici, M. E.; Meir, G.; Van Gerven,
T.; Kuhn, S. An accessible visible-light actinometer for the
determination of photon flux and optical pathlength in flow photo
microreactors. Sci. Rep. 2018, 8, No. 5421.
(13) Pitre, S.; McTiernan, C.; Vine, W.; DiPucchio, R.; Grenier, M.;
Scaiano, J. C. Visible-Light Actinometry and Intermittent Illumination
as Convenient Tools to Study Ru(bpy)3Cl2 Mediated Photoredox
Transformations. Sci. Rep. 2015, 5, No. 16397.
(14) Ji, Y.; DiRocco, D. A.; Hong, C. M.; Wismer, M. K.; Reibarkh,
M. Facile Quantum Yield Determination via NMR Actinometry. Org.
Lett. 2018, 20, 2156−2159.
(15) Wriedt, B.; Ziegenbalg, D. Application Limits of the
Ferrioxalate Actinometer. ChemPhotoChem 2021, 5, 947−956.
(16) Wriedt, B.; Ziegenbalg, D. Common pitfalls in chemical
actinometry. J. Flow Chem. 2020, 10, 295−306.
(17) Wriedt, B.; Kowalczyk, D.; Ziegenbalg, D. Experimental
Determination of Photon Fluxes in Multilayer Capillary Photo-
reactors. ChemPhotoChem 2018, 2, 913−921.
(18) Hamai, S.; Hirayama, F. Actinometric determination of
absolute fluorescence quantum yields. J. Phys. Chem. A 1983, 87,
83−89.
(19) Debrouwer, W.; Kimpe, W.; Dangreau, R.; Huvaere, K.;
Gemoets, H. P. L.; Mottaghi, M.; Kuhn, S.; Van Aken, K. Ir/Ni
Photoredox Dual Catalysis with Heterogeneous Base Enabled by an
Oscillatory Plug Flow Photoreactor. Org. Process Res. Dev. 2020, 24,
2319−2325.
(20) Creaflow. https://www.creaflow.be/hanutm-flow-reactors (ac-
cessed Feb 2).
(21) Ozuturk, E. Formulation of the LED’s radiant power-current Recommended by ACS
characteristic at pulsed high currents. Optik 2013, 124, 3549−3553.
(22) Mozharov, S.; Nordon, A.; Littlejohn, D.; Wiles, C.; Watts, P.;
Dallin, P.; Girkin, J. M. Improved Method for Kinetic Studies in Effects of Cascading Optical Processes: Part III. Impacts on
Microreactors Using Flow Manipulation and Noninvasive Raman Spectroscopic Measurements of Fluorescent Samples
Spectrometry. J. Am. Chem. Soc. 2011, 133, 3601−3608. Max Wamsley, Dongmao Zhang, et al.
JUNE 29, 2023
ANALYTICAL CHEMISTRY READ

Extending Photocatalyst Activity through Choice of Electron

Donor
Felicity Draper, Timothy U. Connell, et al.
JANUARY 11, 2023
THE JOURNAL OF ORGANIC CHEMISTRY READ

High-Throughput Determination of Stern–Volmer

Quenching Constants for Common Photocatalysts and
Quenchers
Rachel N. Motz, Serge Ruccolo, et al.
JUNE 29, 2023
ACS ORGANIC & INORGANIC AU READ

Overcoming Back Electron Transfer in the Electron

Donor–Acceptor Complex-Mediated Visible Light-Driven
Generation of α-Aminoalkyl Radicals from Secondary Ani...
August Runemark and Henrik Sundén
DECEMBER 08, 2022
THE JOURNAL OF ORGANIC CHEMISTRY READ

Get More Suggestions >

2402 https://doi.org/10.1021/acs.oprd.2c00079
Org. Process Res. Dev. 2022, 26, 2392−2402