High-throughput Synthesis and Screening of Iridium(III) Photocatalysts for the Fast and Chemoselective Dehalogenation of Aryl Bromides

Article abstract of DOI:10.1021/acscatal.0c02247

A high-throughput optical screening method for the photocatalytic activity of a structurally diverse library of 1152 cationic iridium(III) complexes ([Ir(C^N)2(N^N)]+), corresponding to all combinations of 48 cyclometalating (C^N) and 24 ancillary (N^N) ligands, was developed. This rapid assay utilizes the colorimetric changes of a high contrast indicator dye, coumarin 6, to monitor the photo-induced electron transfer from a sacrificial amine donor to the metal complex excited state. The resulting [Ir(C^N)2(N^N)]0 can then reduce an aryl bromide to form the highly reactive aryl radical intermediate. The rate of this reaction is dictated by the molecular structure of both coordinating ligands. Relative reaction rate constants determined via this method correlated closely with 19F NMR measurements obtained using a fluorinated substrate. A simple model that expresses the rate constant as a product of a single ″strength″ parameter assigned to each of the 72 ligands can well account for the 1152 measured rate constants. The best performing complexes exhibit much higher reactivity than the benchmark photocatalysts commonly used in photoredox transformations. The catalysts were also successfully tested for their chemoselectivity. The developed screening methodology can enable generation of the large data sets needed to use modern data science to extract structure-activity relationships.

Full text of DOI:10.1021/acscatal.0c02247

Subscriber access provided by Uppsala universitetsbibliotek
Article
High-Throughput Synthesis and Screening of Iridium(III) Photocatalysts
for the Fast and Chemoselective Dehalogenation of Aryl Bromides
Velabo Mdluli, Stephen DiLuzio, Jacqueline Lewis, Jakub Kowalewski,
Timothy U. Connell, David Yaron, Tomasz Kowalewski, and Stefan Bernhard
ACS Catal., Just Accepted Manuscript • Publication Date (Web): 28 May 2020
Downloaded from pubs.acs.org on May 28, 2020
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
High-Throughput Synthesis and Screening of Iridium(III) Photocatalysts for the Fast and
Chemoselective Dehalogenation of Aryl Bromides
Velabo Mdluli, Stephen Diluzio, Jacqueline Lewis, Jakub F. Kowalewski, Timothy U. Connell,
David Yaron, Tomasz Kowalewski, Stefan Bernhard*
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Department of Chemistry, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, USA
ABSTRACT
A high-throughput optical screening method for the photocatalytic activity of a structurally
+
diverse library of 1152 cationic iridium(III) complexes ([Ir(C^N) (N^N)] ), corresponding to all
2
combinations of 48 cyclometalating (C^N) and 24 ancillary (N^N) ligands, was developed. This
rapid assay utilizes the colorimetric changes of a high contrast indicator dye, coumarin 6, to
monitor the photo-induced electron transfer from a sacrificial amine donor to the metal complex
0
excited state. The resulting [Ir(C^N) (N^N)] can then reduce an aryl bromide to form the highly
2
reactive aryl radical intermediate. The rate of this reaction is dictated by the molecular structure of
both coordinating ligands. Relative reaction rate constants determined via this method correlated
1
9
closely with F NMR measurements obtained using a fluorinated substrate. A simple model that
expresses the rate constant as a product of a single "strength" parameter assigned to each of the 72
ligands can well account for the 1152 measured rate constants. The best performing complexes
exhibit much higher reactivity than the benchmark photocatalysts commonly used in
photoredox transformations. The catalysts were also successfully tested for their
chemoselectivity. The developed screening methodology can enable generation of the large data
sets needed to use modern data science to extract structure-activity relationships.
1
ACS Paragon Plus Environment

ACS Catalysis
Page 2 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
INTRODUCTION
The absorption and conversion of visible light into chemical energy is an enviromentally
sustainable and powerful tool in organic synthesis that provides access to reactive, open-shell
intermediates under mild conditions. Commonly employed pathways to such reactive
intermediates include the use of harsh conditions, such as either strong acids or bases, toxic
initiators, or high-energy ultraviolet illumination of precursors, that may lead to unwanted side
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
reactions. In contrast, photoredox catalysis exploits the increased oxidative and reductive power
of an excited, molecular photocatalyst to promote directed, light-driven electron transfers as a
pathway to generate such high-energy, highly reactive species. This attractive strategy, rendered
even more powerful when operating in concert with two-electron transition metal catalysts, has
driven the rapid expansion of this field over the last decade.2-4 Photoredox methodology has also
been embraced by the chemical manufacturing industry, particularly in the preparation of high-
value chemicals including pharmaceuticals and agrochemicals, due to its unparalleled capacity for
chemoselective late-stage functionalization.5-7
Despite increasing reports that utilize organic dyes and first row transition metal complexes,8-9
6
the most widely used catalysts are still coordination complexes of d transition metals, specifically
ruthenium(II) and iridium(III). The strong ligating power of cyclometalating ligands and the
observed photo and redox stability of the resulting iridium(III) complexes is a highly valuable asset
in organic photochemical transformations. It is noteworthy that the unique photophysical
+
properties of heteroleptic iridium(III) complexes of the type [Ir(C^N) (N^N)] , where C^N is a
2
cyclometalating and N^N is an ancillary 1,2-diimine ligand, are attractive in a wide range of
applications in addition to photoredox catalysis, including: water reduction,10-12 biological
imaging13-15and optoelectronic devices.16-18 Irradiation of these complexes generates a long-lived
2
ACS Paragon Plus Environment

Page 3 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
triplet excited state, facilitated by the large spin-orbit coupling constant of the heavy metal atom.
Spatially partitioned frontier orbitals also allow for judicious tuning of redox and optical
properties; the highest occupied molecular orbital (HOMO) centers on the C^N ligand while the
lowest unoccupied molecular orbital (LUMO) is located on the N^N ligand. Independent control
of these orbitals (Figure 1) should allow a 'push-pull' design approach to tune the HOMO-LUMO
gap and redox potentials, enabling careful selection of a photocatalyst that matches the energy
demands of a specific chemical substrate or process. It is therefore surprising that the vast scope
of synthetic photoredox methodology is derived from a limited number of complexes, centered
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
+
+
around the commercially available, cationic [Ir(ppy) (dtbbpy)] and [Ir(dF(CF )ppy) (dtbbpy)]
2
3
2
complexes.
Optimal catalyst selection can be accelerated by high-throughput synthesis and screening
HTSS) techniques. Cost-efficient screening of multiple reactions using conventional analytical
(
methods is both difficult and time consuming, but colorimetric reagents provide a simple, yet
robust optical response that has been successfully applied in electrocatalysis,19-21 organic
synthesis,22-26 olefin polymerization27 and the preparation of solid state materials.28 Online
monitoring of parallelized reactions provides even greater value by revealing chemical reaction
kinetics under standardized conditions.29-32 Finally, HTSS generates large amounts of data that
enables the growing field of data-driven science, described as the fourth approach to chemical
3
3
discovery, in addition to experimentation, theory and computation. Such methodology requires
the pioneering development of more revealing data visualization techniques and powerful
statistical approaches in the chemical sciences; the use of machine learning algorithms is one
example of such a thrust that was employed to better understand structure-reactivity relationships
in catalytic systems.34-37
3
ACS Paragon Plus Environment

ACS Catalysis
Page 4 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
In this work, we report an innovative methodology that allowed for HTSS of a structurally
+
diverse library of 1152 distinct [Ir(C^N) (N^N)] photocatalysts (the explored C^N and N^N
2
structures are shown in the Supporting Information). This was achieved by monitoring the drastic
quenching of an indicator dye’s luminescence by protons generated during a model photochemical
reaction. Several of the complexes were subsequently prepared in batch via conventional
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
methodologies and compared to two iridium(III) photocatalysts, [Ir(ppy) (dtbbpy)]PF and
2
6
[
Ir(dF(CF )ppy) (dtbbpy)]PF , commonly used in photoredox synthesis. Not only did some of the
3 2 6
newly discovered Ir(III) chromophores exhibit faster reaction kinetics, but also presented choices
for pursuing chemoselective photochemical reactions.
Figure. 1. The frontier orbitals (highest and lowest occupied molecular orbitals) are well-
+
partitioned in [Ir(C^N) (N^N)] photocatalysts, enabling the judicious tuning of photophysical and
2
redox properties via structural modification of the cyclometalating (C^N) and ancillary (N^N)
ligands.
4
ACS Paragon Plus Environment

Page 5 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
EXPERIMENTAL METHODS
General Considerations
All reagents and solvents were commercially sourced and used without further purification,
unless otherwise stated. Nuclear magnetic resonance (NMR) spectra were obtained using Bruker
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
13
Avance and NOE 500 MHz spectrometers; H and C spectra were referenced to residual solvent
19
peaks, F NMR used the PF anion signal (δ -71.11 ppm) as an internal standard. Electrospray
6
ionization mass spectrometry (ESI-MS) was performed on a Thermo-Fisher LCQ instrument.
Synthesis of cyclometalating ligands (C^N) and iridium(III) dimers followed previously published
procedures.38-42
Photoreactor Design
SOLIDWORKS design files and Python control software for custom-built photoreactors
are supplied in the Supporting Information. High intensity illumination (Chanzon 100 W Royal
Blue LEDs) of a 96-well plate equipped with 1 mL shell vials containing iridium(III) luminophores
was monitored with a AuviPal 5 Megapixel Raspberry Pi Camera (Module 1080p OV5647 Sensor)
mounted between the light sources (Figure 2). Reactions were run under an inert argon atmosphere,
maintained at positive pressure and monitored using a Winsen ME2-O -Ф20 oxygen sensor. Top-
2
down illumination for photography utilized six pairs of colored Chanzon 10 W LEDs: warm white
(
3000K-3500K), ultraviolet (375nm), blue (440-450nm), green (520-525 nm), yellow (590-592
nm) and red (650-660nm). The lighting was controlled with a Raspberry Pi computer through an
eight-channel relay board that also controlled the USB camera.
5
ACS Paragon Plus Environment

ACS Catalysis
Page 6 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 2: Rendering of a SOLIDWORKS design plan for the custom-built 96-well photoreactor
that was used for monitoring the reaction progress of reductive debromination reactions. Two
100W blue LED light sources illuminate shell vials through the bottom. Reaction progress is
monitored by camera using six 10W LEDs of various colors to illuminate samples from above.
Colorimetric Screening
Solutions of [Ir(C^N) Cl] dimers (100 µL, 0.3 mM) and N^N ancillary ligand (100 µL, 0.6
2
2
mM) in dimethylsulfoxide were pippetted into 1 mL 9800-840 glass shell vials across twelve
columns and eight rows of a 96-well plate, respectively. The sealed solutions were heated at 150
o
+
C overnight to produce heteroleptic iridium(III) complexes ([Ir(C^N) (N^N)] ) as previously
2
3
9
reported. The cleavage of [Ir(C^N) Cl] dimers by diimine ligands to form catinic irdium (III)
2
2
complexes proceeds with ease and high efficiency. Once cooled, 400 µL of a dimethylformamide
solution containing varied amounts of aryl bromide, sacrificial electron donor and coumarin 6 was
added to each vial. The resulting solution was mixed throughly by resealing and shaking the plate
6
ACS Paragon Plus Environment

Page 7 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
for several minutes. The vials were transferred to the reactor and degassed under argon for two
hours. When the oxygen sensor readings reached zero, vials were irradiated with high intensity
blue LEDs for two minutes, followed by image capture under the 10 W LED lights. This
irradiation/imaging cycle repeated continuously for five hours. Representative images of the
synthesis, reactor set-up and raw experimental pictures are included in the Supporting
Information. Compiled photos were analyzed using a custom-written Wolfram Mathematica
algorithm that quantifies the changes of the individual reaction well in a designated color space.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Computational Methods
The electronic structures of iridium(III) complexes were calculated via density functional
theory (DFT), through geometry optimizations with B3LYP functional and LANL2DZ basis set
on Gaussian 09. Analysis of photocatalyst reaction rate constants was done with Python SciKit-
learn package.43
Phosphorescence Lifetime and Emission Screening
Luminophores were irradiated with a pulsed 365 nm LED, powered by a 40 ns square pulse
from a Siglent SDG1052 function generator. Emission was detected with a Hamamatsu H7732-11
photomultiplier tube connected to a Tektronix TDS3032B digital oscilloscope interfaced to a
Raspberry Pi 3 Model B+ computer. Scattered light from the excitation source was removed from
acquired spectra using a plexiglass filter (365 nm long pass). Emission spectra were measured
concurrently using a StellarNet BLACK-Comet concave grating spectrometer .
Electrochemical Screening
Cyclic voltammetry was performed with a CH-Instrument Electrochemical Analyzer 600C
potentiostat. Measurements were conducted in a three-electrode system consisting of silver wire
7
ACS Paragon Plus Environment

ACS Catalysis
Page 8 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
pseudo-reference electrode, a platinum coil counter electrode, and a glassy carbon working
electrode. Solutions (1 mM analyte and 0.1 M tetra-n-butylammonium hexafluorophosphate
supporting electrolyte) in acetonitrile were degassed and measured under argon. Voltammograms
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
-1
were collected under positive scan polarity with a scan rate of 0.1 V s and redox potentials
0
+
44
reported relative to a ferrocene internal standard (E (Fc /Fc) = 0.40 V vs SCE).
8
ACS Paragon Plus Environment

Page 9 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
RESULTS AND DISCUSSION
Method Development
We selected the reductive dehalogenation of aryl halide45-47 as a robust model for screening
the photocatalytic reactivity of iridium(III) photocatalysts. Following irradiation, the excited state
photocatalyst is quenched by a sacrificial electron donor (commonly a tertiary amine), liberating
a proton from the oxidized amine. An electron is subsequently transferred from the reduced
photocatalyst to an electron acceptor (i.e. aryl halide), affording catalyst turnover. The reactive
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
8
aryl radical generated can participate in a range of C–C and C–X coupling pathways, however
without a suitable nucleophile or electrophile, the radical abstracts a hydrogen atom from the
solvent system affording the dehalogenated product (Scheme 1).
Scheme 1. The photoredox reaction being monitored via colorimetric changes of coumarin 6
indicator dye.
1
9
Using a fluorinated aryl bromide reagent allows the reaction to be monitored via F NMR,
showing an exponential decay in the amount of starting material and a corresponding increase in
9
ACS Paragon Plus Environment

ACS Catalysis
Page 10 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
the amount of product (see Supporting Information). Because the studied reactions are complete
1
9
within a couple of hours, monitoring the reaction progress on a 96-well plate with F NMR is
unattainable. To compare the rates of different photocatalysts, a faster, more efficient and low-cost
online screening method that uses an appropriate indicator dye was sought. Bromocresol green is
an acid sensitive, colorimetric reagent previously used in post-reaction titration to determine the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
9
effectiveness of aryl halide reduction. Unfortunately, it is also susceptible to debromination
under photoredox conditions and therefore unsuitable for a time-based assay. Instead we selected
coumarin 6 (Scheme 1), which exhibits an intense absorption band (λabs = 444 nm, ε = 54 000 M-
1
-1
cm ) and high fluorescent quantum yield (λ = 510–560 nm, Φ = 0.78) that is drastically
em
quenched in the presence of protons.50-52 Such a definitive record of the sharp colorimetric change
provides a diagnostic handle that allows accurate comparison of photoinduced electron transfer
rates among the hundreds of photocatalysts being screened.
The suitability of coumarin 6 as an indicator for comparing photoredox reaction rates was
evaluated using twelve heteroleptic iridium(III) complexes as photocatalysts for the
dehalogenation of 2’-bromo-4’-flouroacetophenone, with N,N-diisopropylethylamine (DIPEA)
serving as a sacrificial reductant (Figure 3). The in-situ prepared photocatalysts all contained the
ancillary ligand 4,4′-di-tert-butyl-2,2′-bipyridine (NN6) with representative cyclometalating
ligands selected from a larger combinatorial library employed in our laboratory. The concentration
of coumarin 6 varied down each photocatalyst column (0–1.4 mM in 0.2 mM increments).
10
ACS Paragon Plus Environment

Page 11 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 3. Images of a 96-well plate showing the colorimetric changes of the indicator dye in
parallelized photoredox reactions of 12 distinct iridium(III) photocatalysts. Each vial in a column
contained 4mM 2’-bromo-4’-flouroacetophenone, 40 mM DIPEA, 0.1 mM photocatalyst and
coumarin 6 (C6) indicator dye in increasing concentration down the plate. A video of the reaction
progress can be found in the Supporting Information.
Under blue illumination, the neutral dye produces bright green luminescence that is easily
observed during imaging. Reaction progress is monitored when this luminescence is quenched by
the produced protons, with a drastic colorimetric change observed when no neutral coumarin 6
remains; instead only the blue illumination light is transmitted through a reaction well and detected
by the camera. The time required for the sufficient release of protons to reach the equivalence point
increases along with increasing amounts of dye through the column. Since the eight reaction wells
of each column contain the same photocatalyst, the elementary rate law describing proton
11
ACS Paragon Plus Environment

ACS Catalysis
Page 12 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
production can be deduced; an exponential decay function fitted for each column (see Supporting
Information) revealed that the formation of protons, caused by catalyst-mediated photoinduced
electron transfer, followed first order kinetics.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Rate constants determined via HTSS were corroborated by integration of the reactant and
19
product signals obtained through F NMR spectroscopy (Figure 4). The strong linear correlation
2
(
R = 0.98) of these two techniques confirms the validity of the developed system for comparing
the relative photocatalytic activity of iridium(III) complexes.
Figure 4. a. Stacked 19F NMR indicating the terminal amounts of reagent (2’-bromo-4’-
19
flouroacetophenone) and product (4’-fluoroacetophenone) measured via F NMR spectroscopy
from the wells of a row on the reaction plate catalyzed by different photocatalysts (Figure 3). b.
The ratio ln([A] /[A] ), where [A] /[A] is the fraction reactant remaining, measured by NMR, is
t
0
t
0
linearly correlated to the rate constant obtained via image-based analysis of the colorimetric
technique.
Using the developed colorimetric technique, we set out to determine the relative catalytic
+
activity of 1152 distinct [Ir(C^N) (N^N)] photocatalysts, comprising 48 C^N and 24 N^N ligands
2
(structures shown in
Supporting Information). For this set, reduction of
12
ACS Paragon Plus Environment

Page 13 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
bromopentafluorobenzene, with its accessible reduction potential,53 ensured that the results
maximized information about relative activity of photocatalysts, rather than being limited by
insufficient potential to transfer electrons to a challenging substrate. The reaction conditions (4
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
+
mM aryl halide, 40 mM DIPEA, 0.4 mM coumarin 6 and 0.1 mM [Ir(C^N) (N^N)] ) allowed the
2
calculation of rate constants from the time required to reach an equivalence point representative of
3
0% substrate reduction.
Prior to the screening of all the hundreds of photocatalysts, a second type of
calibration was used to confirm that colorimetric response of the indicator occurs uniformly across
the 96-well plate (see video in Supporting Information). This was evaluated by reacting uniform
plates with the wells consisting of 96-identical reactions across the entire plate (4mM
+
bromopentafluorobenzene, 40 mM DIPEA, 0.4 mM coumarin 6 and 0.1 mM [Ir(C^N) (N^N)] ).
2
The PF salts of five traditionally synthesized and isolated Ir (III) complexes were selected for five
6
uniform plates that were expected to exhibit a range of rates constants. Analysis of the plate
revealed well-to-well variation of 6 – 14%, in the red-green-blue (RGB) color space, for the all
plates. This relative standard deviation within the wells of the five plates was reduced to 4 –
1
1% when targeted analysis was done along the blue channel of RGB, since the indicator dye
detection point is a visible appearance of a blue color. Ultimately, using the Internationale de
l’eclairage La*b* (CIELab) which decouples the chromatic part of color from its brightness,
resulted in the error range being minimized further still to 2 – 8%. The rate constants for the five
uniform plates was calculated to be 0.0 ± 0.0, 0.3 ± 0.02 , 0.8 ± 0.06, 1.5 ± 0.12 and 2.2 ± 0.16 h-
1
+
+
+
+
for [Ir(CN46) (NN6)] , [Ir(CN11) (NN6)] , [Ir(CN1) (NN6)] , [Ir(CN70) (NN6)] and
2
2
2
2
+
[
Ir(CN34) (NN6)] , respectively. With uniformity across the 96-well plate established, we
2
+
screened the relative catalytic activity of the 1152 [Ir(C^N) (N^N)] photocatalysts, in the
2
13
ACS Paragon Plus Environment

ACS Catalysis
Page 14 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
reduction of bromopentafluorobenzene; all image analysis for this work was done within the
CIELab color space. Figure 5 confirms that the rate constants of the five traditionally synthesized
[Ir(C^N) (N^N)]PF complexes meausred in these uniform plates and the results obtained with in-
2
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
situ synthesized photocatalysts used for the HTSS (see Supporting Information, attached csv file,
for all the rate constants) are within the margins of error.
+
Figure 5. Comparison of the catalytic activity of [Ir(C^N) (N^N)] complexes utilizing
2
traditionally and in-situ synthesized photocatalysts under identical dehalogenation conditions.
14
ACS Paragon Plus Environment

Page 15 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
Measuring Structure-Activity Relationships
+
The HTSS rate constants of the 1152 [Ir(C^N) N^N] complexes used in the photo-
2
driven dehalogenationation of bromopentafluorobenzene are attached in Supporting Information.
Unlike most studies that synthesize and evaluate a small number of chemically related metal
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
complexes, this large, standardized data set provides a unique opportunity to generalize structure-
+
activity relationships in [Ir(C^N) N^N] photocatalysts. Figure 6 was constructed using a select,
2
but representative, group of C^N and N^N ligands to show some of the major trends.
Figure 6. A 3D bar chart, sorted by increasing average rate constants, for a representative set
cyclometalating (C^N) and ancillary ligands (N^N) ligands, highlighting some of the major trends
+
of the [Ir(C^N) N^N] photocatalysts. See Supporting Information for array plot.
2
Alkylation, especially on the pyridyl moiety of the C^N ligand, significantly increased the
rate constant of the resulting photocatalyst by donating electron density, as reported in literature.54
Consider starting with CN1, adding a methyl group at the 4- position to the pyridyl ring improves
15
ACS Paragon Plus Environment

ACS Catalysis
Page 16 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
the resulting photocatalyst (CN2), by an average 15% increase in the calculated rate constants. It
was quite interesting to discover that using an ethyl substituent (CN70) resulted in noticeably
greater (35%) enhancement compared to the methyl substitution. Since the electronic structure of
the metal complex will not be significantly altered by a change from a methyl to an ethyl
substituent, it is likely that the increased steric bulk prevents a N^N ligand dissociation, a known
degradation pathway in these photocatalysts.55 Similarly, the large conjugated systems in 2-
phenylbenzothiazole ligands (CN30 and CN34) result in photocatalysts with relatively higher
reactivity compared to their 2-phenylpyridine counterparts (CN30 is nearly two-fold faster than
CN1). The extent of π conjugation in the C^N ligand is known to drastically impact the nature of
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
56-58
the excited state of iridium(III) complex, transforming it to a predominately LC excited state.
Certain substitutions on the C^N ligand seem to be detrimental to resulting photocatalyst
activity, such as electron withdrawing groups (CN11), and methoxy groups at the 5-position of the
phenyl ring (CN46 and CN89). The diminished ability observed for highly fluorinated
photocatalysts can be partly rationalized by considering photocatalyst redox properties:
+
0
0
[
Ir(CN1) NN6] /[Ir(CN1) NN6] (E = -1.51 V) is a stronger reductant than its fluorinated
2 2
+
0
0
cousin, [Ir(CN11) NN6] /[Ir(CN11) NN6] (E = -1.44 V). This implies that electron transfer
2
2
from the latter to the aryl halide acceptor is slower, assuming it does not fall into the Marcus-
5
9
inverted region. With methoxy substituents, para- substituted cyclometalating ligands (CN34)
exhibit the well understood activation of the resulting photocatalyst, while meta-substitution
(CN46 and CN89) produced strongly deactivated photocatalysts; this trend with m-methoxy
substituents was also observed when related complexes were employed in light emitting
electrochemical cells.60
16
ACS Paragon Plus Environment

Page 17 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
61
For the ancillary ligands (N^N), the solvated complexes of dimethyl sulfoxide provided
a negative control, and exhibited rate constants close to zero for all tested C^N derivatives. With
the diimines, the data reveals that most bipyridine derivatives outperform those of phenanthroline.
Alkylation of the N^N ligands typically enhanced photocatalyst activity, but for methyl groups at
α-position to the nitrogen donor atoms of an N^N ligand, alkylation is beneficial to the
phenanthroline-based (NN34) while being detrimental to the bipyridine ones (NN4). Another
characteristic effect observed with N^N was that the presence of ketone functional groups, e.g.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
9
,10-phenanthrenedione (NN14) significantly reduces photocatalyst activity, likely due to a
dramatic decrease in the HOMO-LUMO gap and the absence of a long-lived excited state.
Data Modeling
The HTSS measured the rate constants for 1152 catalysts consisting of all pairings of 48
cyclometalating (C^N) and 24 ancillary (N^N) ligands. We will denote these experimental rates
as Rate(cn,nn), where cn and nn span the ranges 1 to 48 and 1 to 24, respectively. A striking aspect
of the observed reaction rates is that they appear to be decomposable into contributions from the
individual C^N and N^N ligands. We will refer to the ability of an individual ligand to promote
catalysis as the “strength” of that individual ligand and explore a number of ways to extract these
strengths from the measured rates. A particularly simple way to estimate the strength of a ligand
is to sum the observed rate over all complexes that include that ligand (Eqn. 1),
Smargin (cn) =∑ Rate(cn,nn) ;
Smargin (nn) = ∑ Rate(cn,nn)
(1)
푛푛
푐푛
We refer to these as Smargin because these would be the margins of a two-dimensional data table
formed from Rate(cn,nn). A heat map of the reaction rates, with the C^N and N^N ligands sorted
17
ACS Paragon Plus Environment

ACS Catalysis
Page 18 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
by Smargin (Figure 7a), shows the tendency of Rate(cn,nn) to increase with Smargin of the individual
ligands.
To help quantify this tendency, we considered two models that are based on the general
idea of ligand strength but use alternative measures of this strength. The first is a sum model, in
which the rate is written as a sum of the ligand strengths (Eqn. 2),
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Rate(cn,nn) = Ssum (cn) + Ssum (nn)
The second model assumes the rate is a product of ligand strengths (Eqn. 3),
Rate(cn,nn) = Sproduct (cn) * Sproduct (nn)
(2)
(3)
For both models, a least squares procedure is used to fit the model parameters, Ssum or Sproduct, to
the observed rates. Each model has 72 parameters, one each for the 48 C^N and 24 N^N ligands,
which can be fit to the rates measured for the 1152 catalysts. Separating the data into train and test
sets has little impact on the quality of the model predictions (see Supporting
Information). Although both the sum (Figure 7b) and product (Figure 7c) model can account for
2
a large proportion of the variance in the data, the product model performs better (r of 0.87 for
product model versus 0.74 for sum model).
Figure 7. a. Heatmap of rate constants as a function of C^N and N^N ligands, with ligands sorted
from poor to good performers, as measured by Smargin. The scatterplots show the predicted rate
versus the experimental rate from the sum model (b) the product model (c).
18
ACS Paragon Plus Environment

Page 19 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
The product model also gives more reasonable results for catalysts with low performance,
with predicted rates for these complexes having a smaller variance than for the sum model and
with all predicted rates remaining positive. This reflects the tendency, apparent in the heat map,
for poorly-performing ligands to lead to poorly-performing catalysts, even when paired with a
strong partner ligand. The variables, Smargin, S , and Sproduct, provide quantitative measures of the
strength of each ligand. The rankings of ligands, from poor to strong performers, obtained for Ssum
are identical to those with Smargin, while slight reordering occurs with Sproduct (see Supporting
Information). Our use of the sum and product models is primarily intended to quantify the
tendency of the rate to decompose into contributions from the individual ligands. However, it is
interesting to note that the product model is what would be expected if the relevant chemical
properties of the ligands were associated with energies or other properties that, in the relation for
the rate constant, appear in an exponential.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
sum
The ability of these models to explain a large proportion of the variance in the experimental
data validates that our measured rates are not simply results of random changes in indicator dye
but are instead meaningful trends in catalysts’ properties. It is also interesting to observe that
properties of the ligands alone are able to establish meaningful correlations, as seen in other
chemical fragment-based approaches employed in data-driven analysis of catalytic systems.62-65
The above models are only able to make predictions for complexes containing already studied
ligands. In ongoing work, we are using DFT-derived features to enable the prediction of
photocatalyst activity exclusively from chemical structure.
19
ACS Paragon Plus Environment

ACS Catalysis
Page 20 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Conventional Tests with Selected Photocatalysts
The HTSS identified several iridium(III) complexes with impressive reactivities compared
to the most commonly used commercial photoredox catalysts. To further explore the results we
+
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
selected four complexes to traditionally synthesize and fully characterize: [Ir(CN40) (NN6)] ,
2
+
+
+
[
Ir(CN35) (NN6)] , [Ir(CN35) (NN16)] and the highly active photocatalyst [Ir(CN63) (NN6)]
2
2
2
(see Figure 8 for structures). The optical, electrochemical and DFT calculated properties of the
+
photocatalysts were compared to the most commonly used photoredox catalysts [Ir(CN1) (NN6)]
2
+
and [Ir(CN14) (NN6)] ([Ir(ppy) (dtbbpy)]PF and [Ir(dF(CF )ppy) (dtbbpy)]PF ).
2
2
6
3
2
6
+
Figure 8. Normalized room temperature emission spectra of the six [Ir(C^N) (N^N)]
2
photocatalysts. The argon degassed samples (20 µM acetonitrile solutions) were excited at λ = 365
nm, and their luminescence spectrum was measured at room temperature.
Absorbance spectra were measured in acetonitrile and are analogous to other reported
heteroleptic cationic iridium(III) complexes.66 Below 300 nm all compounds exhibit intense
ligand-centered (LC) transitions, while absorptions at longer wavelengths are assigned to metal-
ligand-to-ligand charge transfer (MLLCT) transitions. Emission spectra revealed that
+
+
[
Ir(CN35) (NN6)] and [Ir(CN35) (NN16)] exhibited a broad structureless emission indicative
2 2
3
67
of a predominately MLLCT excited state. The vibronic substructure observed in the other
20
ACS Paragon Plus Environment

Page 21 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
3
synthesized complexes indicates the often observed mixing with an LC excited state. This is also
reflected in the longer lifetimes of these complexes, compared to the relatively short measured
3
+
lifetime for the mostly MLLCT excited state complex [Ir(CN1) (NN6)] (Table 1).
2
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Table 1. Photophysical, electrochemical and density functional theory (DFT) calculated data of
the selected iridium(III) complexes. Also, the HTSS rate constant values from the dehalogenation
the bromopentafluorobenzene are included.
+
Electrochemical analysis showed varied behavior, from [Ir(CN40) (NN6)] with a
2
+
strongly positive reversible oxidation potential, similar to the state-of-art [Ir(CN14) (NN6)] used
2
+
in systems that proceed via an oxidative quenching mechanism to [Ir(CN63) (NN6)] with
2
reversible and quasi-reversible reductions at -1.46, -1.90 and -2.07 V vs SCE (Figure 9). With
only the six photocatalysts, no direct correlation between the luminescence maximum
energy, redox potentials or calculated HOMO-LUMO gaps can be found, indicating that
extensive photophysical and electrochemical data and DFT calculations are necessary to establish
such correlations.
21
ACS Paragon Plus Environment

ACS Catalysis
Page 22 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 9. Comparison of the electrochemistry and DFT of two iridium(III) complexes;
+
[
Ir(CN40) (NN6)] shows a strongly positive oxidation potential, while the second compound
2
+
([Ir(CN63) (NN6)] ) exhibit multiple reduction peaks. The DFT calculations document the
2
widely different energy of the involved frontier orbitals which highlights that structure-activity are
much more complex and, consequently, a HTSS approach is necessary until better correlations
with calculated features allow the accurate prediction of reactivity from quantum mechanics.
Insight obtained from the developed rapid screening approach may assist the identification
of photocatalysts that demonstrate thermodynamic and kinetic reaction control. Chemoselective
68
dehalogenation of trisubstituted aryl halides is reported with organic dye photocatalysts but the
modular synthesis of metal complexes allows comparatively facile tuning of redox and
photophysical properties. The variation in activity of the six synthesized complexes in the
19
debromination of 1,2-dibromo-4,5-difluorobenzene was monitored by F NMR (Figure 10).
22
ACS Paragon Plus Environment

Page 23 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 10. 19F NMR analysis of the dehalogenation of 1,2-dibromo-4,5-difluorobenzene (A,
singlet peak, δ -137.3 ppm), catalyzed by six different photocatalysts. The first dehalogenation
product, 4-bromo-1,2-difluorobenzene (B, doublet, δ -135.8 and -140.4 ppm) and the fully
debrominated product 1,2-difluorobenzene (C, singlet, δ -139.7 ppm) can be obtained exclusively
and in almost quantitative yields by two different photocatalysts (species B by
+
+
[
Ir(CN35) (NN16)] and C by [Ir(CN63) (NN6)] ) .
2
2
The three complexes containing either 2-phenylbenzothiazole or 2-phenylbenzoxazole
cyclometalating ligands all produced the singly debrominated product in high yield after 16 hours
+
irradiation, while [Ir(CN63) (NN6)] quantitatively produced 1,2-difluorobenzene. In
2
+
+
comparison, both commercial photoredox catalysts ([Ir(CN1) (NN6)] and [Ir(CN14) (NN6)] )
2
2
yielded a mixture of products. This promising result suggests similar chemoselectivity in C–C
bond forming reactions, comparable to that observed with some organic dyes can be achieved.69-
7
0
Zeitler’s work with the organic dyes highlights this importance of tuning the photocatalysts from
highly oxidizing to highly reducing potentials to achieve targeted catalysis. We believe the
heteroleptic Ir(III) photocatalysts offer a much more accessible platform for synthetically tuning
their activity and selectivity compared to organic dye molecules. Also, the very efficient and fast
23
ACS Paragon Plus Environment

ACS Catalysis
Page 24 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
+
chromophore [Ir(CN63) (NN6)] is comparatively easy to synthesize and does not contain any
2
expensive and environmentally problematic fluorine substitutions.
CONCLUSION
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
The presented work allowed the development of an indicator dye-based methodology for the online
monitoring of the catalytic performance of 1152 in-situ synthesized cationic iridium(III)
complexes in photoredox processes. The utility of this technique was demonstrated through rapid
screening of the C^N and N^N ligands parameter space to pinpoint champion catalysts for the
dehalogenation of aryl bromide. In addition to accelerating chemical discovery, this colorimetric
approach is a significant step towards providing large sets of standardized experimental data that,
in future work, can be combined with computationally derived features to build towards highly
predictive machine learning models. Results obtained from this data-driven inquiry were
confirmed via traditional synthesis and characterization techniques. The reactivity of some newly
synthesized photocataysts exceeded those of the most well-studied and commercially-used
cationic Ir(III) complexes by a factor of 3-4. It was also possible to tailor the discovered catalysts
for chemoselectivity.
24
ACS Paragon Plus Environment

Page 25 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
ASSOCIATED CONTENT
Supporting Information
All supporting figures and tables, Wolfram Mathematica code for image analysis and data
processing, Python code for photoreactor control and data modeling, as well as a description of
data modeling with train-test splits, are presented in a PDF document. Also attached is the csv file
of the 1152 rate constants and the two videos of real-time reaction progress mentioned in the main
text. This material is all available free of charge at http://pubs.acs.org.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
AUTHOR INFORMATION
Corresponding Author
*
bern@cmu.edu
Notes
The authors declare no competing financial interest.
ACKNOWLEDGEMENTS
We are grateful for the financial support of the US NSF (CHE-1764353) and Carnegie
Mellon's Manufacturing Futures Initiative that is funded by the Richard King Mellon Foundation.
NMR instrumentation at Carnegie Mellon University was partially supported by the NSF (CHE-
1039870 and CHE-1726525).
25
ACS Paragon Plus Environment

ACS Catalysis
Page 26 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
REFERENCES
1
.
Rossi, R. A.; Pierini, A. B.; Peñéñory, A. B. Nucleophilic Substitution Reactions by Electron
Transfer. Chem. Rev. 2003, 103, 71-167.
2
2
3
.
Reckenthäler, M.; Griesbeck, A. G. Photoredox Catalysis for Organic Syntheses. Adv. Synth. Catal.
013, 355, 2727-2744.
Wang, C.-S.; Dixneuf, P. H.; Soulé, J.-F. Photoredox Catalysis for Building C–C Bonds from
.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
C(sp )–H Bonds. Chem. Rev. 2018, 118, 7532-7585.
Prier, C. K.; Rankic, D. A.; MacMillan, D. W. C. Visible Light Photoredox Catalysis with
Transition Metal Complexes: Applications in Organic Synthesis. Chem. Rev. 2013, 113, 5322-5363.
Nicholls, T. P.; Leonori, D.; Bissember, A. C. Applications of Visible Light Photoredox Catalysis
to the Synthesis of Natural Products and Related Compounds. Nat. Prod. Rep. 2016, 33, 1248-1254.
Douglas, J. J.; Sevrin, M. J.; Stephenson, C. R. J. Visible Light Photocatalysis: Applications and
New Disconnections in the Synthesis of Pharmaceutical Agents. Org. Process Res. Dev. 2016, 20, 1134-
147.
7.
4
.
5
.
6
.
1
DiRocco, D. A.; Dykstra, K.; Krska, S.; Vachal, P.; Conway, D. V.; Tudge, M. Late-Stage
Functionalization of Biologically Active Heterocycles Through Photoredox Catalysis. Angew. Chem. Int.
Ed. 2014, 53, 4802-4806.
8.
10166.
9
Romero, N. A.; Nicewicz, D. A. Organic Photoredox Catalysis. Chem. Rev. 2016, 116, 10075-
.
Larsen, C. B.; Wenger, O. S. Photoredox Catalysis with Metal Complexes Made from Earth-
Abundant Elements. Chem. Eur. J. 2018, 24, 2039-2058.
0. Goldsmith, J. I.; Hudson, W. R.; Lowry, M. S.; Anderson, T. H.; Bernhard, S. Discovery and High-
Throughput Screening of Heteroleptic Iridium Complexes for Photoinduced Hydrogen Production. J. Am.
Chem. Soc. 2005, 127, 7502-7510.
1
1
1.
Generation. Chem. Soc. Rev. 2017, 46, 603-631.
2. Mills, I. N.; Porras, J. A.; Bernhard, S. Judicious Design of Cationic, Cyclometalated Ir(III)
Complexes for Photochemical Energy Conversion and Optoelectronics. Acc. Chem. Res. 2018, 51, 352-
Yuan, Y.-J.; Yu, Z.-T.; Chen, D.-Q.; Zou, Z.-G. Metal-Complex Chromophores for Solar Hydrogen
1
3
1
64.
3.
Lo, K. K.-W.; Louie, M.-W.; Zhang, K. Y. Design of Luminescent Iridium(III) and Rhenium(I)
Polypyridine Complexes as in Vitro and in Vivo Ion, Molecular and Biological Probes. Coord. Chem. Rev.
010, 254, 2603-2622.
14. Caporale, C.; Massi, M. Cyclometalated Iridium(III) Complexes for Life Science. Coord. Chem.
Rev. 2018, 363, 71-91.
15. Connell, T. U.; Donnelly, P. S. Labelling Proteins and Peptides with Phosphorescent d Transition
Metal Complexes. Coord. Chem. Rev. 2018, 375, 267-284.
16. Flamigni, L.; Barbieri, A.; Sabatini, C.; Ventura, B.; Barigelletti, F. Photochemistry and
Photophysics of Coordination Compounds: Iridium. In Photochemistry and Photophysics of Coordination
Compounds II, Balzani, V.; Campagna, S. Eds. Springer Berlin Heidelberg: Berlin, Heidelberg, 2007; pp
43-203.
7. Choy, W. C. H.; Chan, W. K.; Yuan, Y. Recent Advances in Transition Metal Complexes and
Light-Management Engineering in Organic Optoelectronic Devices. Adv. Mater. 2014, 26, 5368-5399.
8. Kagalwala, H. N.; Chirdon, D. N.; Bernhard, S. Solar Fuel Generation. In Iridium(III) in
Optoelectronic and Photonics Applications, 2017; pp 583-615.
9. Reddington, E.; Sapienza, A.; Gurau, B.; Viswanathan, R.; Sarangapani, S.; Smotkin, E. S.;
Mallouk, T. E. Combinatorial Electrochemistry: A Highly Parallel, Optical Screening Method for
Discovery of Better Electrocatalysts. Science 1998, 280, 1735-1737.
2
6
1
1
1
1
2
0.
Jun, Y. J.; Park, S. H.; Woo, S. I. Combinatorial High-Throughput Optical Screening of High
Performance Pd Alloy Cathode for Hybrid Li − Air Battery. ACS Comb. Sci. 2014, 16, 670-677.
26
ACS Paragon Plus Environment

Page 27 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
2
1.
Park, S. H.; Choi, C. H.; Koh, J. K.; Pak, C.; Jin, S.-a.; Woo, S. I. Combinatorial High-Throughput
Screening for Highly Active Pd − Ir − Ce Based Ternary Catalysts in Electrochemical Oxygen Reduction
Reaction. ACS Comb. Sci. 2013, 15, 572-579.
2
Assay for High-Throughput Screening of Transaminases in Liquid or Solid-Phase. Chem. Commun. 2015,
5
2.
Baud, D.; Ladkau, N.; Moody, T. S.; Ward, J. M.; Hailes, H. C. A Rapid, Sensistive Colorimetric
1, 17225-17228.
3. Shaughnessy, K. H.; Kim, P.; Hartwig, J. F. A Fluorescence-Based Assay for High-Throughput
2
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Screening of Coupling Reactions. Application to Heck Chemistry. J. Am. Chem. Soc. 1999, 121, 2123-
2
2
132.
4.
Crabtree, R. H. Combinatorial and Rapid Screening Approaches to Homogeneous Catalyst
Discovery and Optimization. Chem. Commun. 1999, 1611-1616.
25. Copeland, G. T.; Miller, S. J. A Chemosensor-Based Approach to Catalyst Discovery in Solution
and on Solid Support. J. Am. Chem. Soc. 1999, 121, 4306-4307.
26. Shah, R.; Pratt, D. A. Determination of Key Hydrocarbon Autoxidation Products by Fluorescence.
J. Org. Chem. 2016, 81, 6649-6656.
27. Boussie, T. R.; Diamond, G. M.; Goh, C.; Hall, K. A.; Lapointe, A. M.; Leclerc, M.; Lund, C.;
Murphy, V.; Shoemaker, J. A. W.; Tracht, U.; Turner, H.; Zhang, J.; Uno, T.; Rosen, R. K.; Stevens, J. C.
A Fully Integrated High-Throughput Screening Methodology for the Discovery of New Polyolefin
Catalysts: Discovery of a New Class of High Temperature Single-Site Group (IV) Copolymerization
Catalysts. J. Am. Chem. Soc. 2003, 125, 4306-4317.
2
8.
Mitrovic, S.; Soedarmadji, E.; Newhouse, P. F.; Suram, S. K.; Haber, J. A.; Jin, J.; Gregoire, J. M.
Colorimetric Screening for High-Throughput Discovery of Light Absorbers. ACS Comb. Sci. 2015, 17, 176-
1
2
81.
9.
Curtin, P. N.; Tinker, L. L.; Burgess, C. M.; Cline, E. D.; Bernhard, S. Structure−Activity
Correlations Among Iridium(III) Photosensitizers in a Robust Water-Reducing System. Inorg. Chem. 2009,
4
3
8, 10498-10506.
0. Lee, J.-H.; Cheglakov, Z.; Yi, J.; Cronin, T. M.; Gibson, K. J.; Tian, B.; Weizmann, Y. Plasmonic
Photothermal Gold Bipyramid Nanoreactors for Ultrafast Real-Time Bioassays. J. Am. Chem. Soc. 2017,
1
3
Wagner, B. K.; Clemons, P. A.; Booker-Milburn, K. I.; Boskovic, Z. V.; Schreiber, S. L. Real-Time
Biological Annotation of Synthetic Compounds. J. Am. Chem. Soc. 2016, 138, 8920-8927.
32.
S.; Kitchin, J. R.; Ulissi, Z. W.; Millstone, J. E.; Bernhard, S. Parallelized Screening of Characterized and
DFT-Modelled Bimetallic Colloidal Co-Catalysts for Photocatalytic Hydrogen Evolution. ACS Catal.
2
3
39, 8054-8057.
1. Gerry, C. J.; Hua, B. K.; Wawer, M. J.; Knowles, J. P.; Nelson Jr, S. D.; Verho, O.; Dandapani, S.;
Lopato, E. M.; Eikey, E. A.; Simon, Z. C.; Back, S.; Tran, K.; Lewis, J.; Kowalewski, J. F.; Yazdi,
020, 10, 4244-4252.
3. Hey, T.; Tansley, S.; Tolle, K. The Fourth Paradigm Data-Intensive Scientific Discovery. Microsoft
Research: Redmond, WA, 2009.
4. Bonk, B. M.; Weis, J. W.; Tidor, B. Machine Learning Identifies Chemical Characteristics That
Promote Enzyme Catalysis. J. Am. Chem. Soc. 2019, 141, 4108-4118.
5. Bai, Y.; Wilbraham, L.; Slater, B. J.; Zwijnenburg, M. A.; Sprick, R. S.; Cooper, A. I. Accelerated
3
3
Discovery of Organic Polymer Photocatalysts for Hydrogen Evolution from Water through the Integration
of Experiment and Theory. J. Am. Chem. Soc. 2019, 141, 9063-9071.
3
6.
ACS Catal. 2019, 9, 8383-8387.
7. Masood, H.; Ying Toe, C.; Yang Teoh, W.; Sethu, V.; Amal, R. Machine Learning for Accelerated
Discovery of Solar Photocatalysts. ACS Catal. 2019, 9, 11774−11787.
8. Mills, I. N.; Kagalwala, H. N.; Bernhard, S. Cyano-Decorated Ligands: A Powerful Alternative to
Palkovits, R.; Palkovits, S. Using Artificial Intelligence to Forecast Water Oxidation Catalysts.
3
3
Fluorination for Tuning the Photochemical Properties of Cyclometalated Ir(III) Complexes. Dalton Trans.
2016, 45, 10411-10419.
27
ACS Paragon Plus Environment

ACS Catalysis
Page 28 of 30
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
3
9.
Lowry, M. S.; Hudson, W. R.; Pascal, R. A.; Bernhard, S. Accelerated Luminophore Discovery
through Combinatorial Synthesis. J. Am. Chem. Soc. 2004, 126, 14129-14135.
0. Ghodbane, A.; D'Altério, S.; Saffon, N.; McClenaghan, N. D.; Scarpantonio, L.; Jolinat, P.; Fery-
4
Forgues, S. Facile Access to Highly Fluorescent Nanofibers and Microcrystals via Reprecipitation of 2-
Phenyl-benzoxazole Derivatives. Langmuir 2012, 28, 855-863.
4
A Simple and Efficient Mechanochemical Route for the Synthesis of 2-Aryl Benzothiazoles and Substituted
Benzimidazoles. RSC Adv. 2014, 4, 39606-39611.
1.
Banerjee, M.; Chatterjee, A.; Kumar, V.; Bhutia, Z. T.; Khandare, D. G.; Majik, M. S.; Roy, B. G.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
2.
Relevant 2-Phenylbenzothiazoles Substituted on the Benzothiazole Ring. Tetrahedron 2011, 67, 7743-
747.
43.
Weekes, A. A.; Bagley, M. C.; Westwell, A. D. An Efficient Synthetic Route to Biologically
7
Pedregosa, F.; Varoquaux, G.; Gramfort, A.; Michel, V.; Thirion, B.; Grisel, O.; Blondel, M.;
Prettenhofer, P.; Weiss, R.; Dubourg, V.; Vanderplas, J.; Passo, A.; Cournapeau, D.; Brucher, M.; Perrot,
M.; Duchesnay, É. Scikit-Learn: Machine Learning in Python. J. Mach. Learn. Res. 2011, 12,
2825−2830.
44.
1
Connelly, N. G.; Geiger, W. E. Chemical Redox Agents for Organometallic Chemistry. Chem. Rev.
996, 96, 877-910.
5.
4
Devery, J. J.; Nguyen, J. D.; Dai, C.; Stephenson, C. R. J. Light-Mediated Reductive
Debromination of Unactivated Alkyl and Aryl Bromides. ACS Catal. 2016, 6, 5962-5967.
6. Pal, A.; Ghosh, I.; Sapra, S.; König, B. Quantum Dots in Visible-Light Photoredox Catalysis:
Reductive Dehalogenations and C-H Arylation Reactions Using Aryl Bromides. Chem. Mater. 2017, 29,
4
5
4
225-5231.
7. Shon, J.-H.; Sittel, S.; Teets, T. S. Synthesis and Characterization of Strong Cyclometalated Iridium
Photoreductants for Application in Photocatalytic Aryl Bromide Hydrodebromination. ACS Catal. 2019, 9,
8
4
646-8658.
8. Ghosh, I.; Ghosh, T.; Bardagi, J. I.; König, B. Reduction of Aryl Halides by Consecutive Visible
Light-Induced Electron Transfer Processes. Science 2014, 346, 725-728.
9. Poznik, M.; König, B. Fast Colorimetric Screening for Visible Light Photocatalytic Oxidation and
Reduction Reactions. React. Chem. Eng. 2016, 1, 494-500.
0. Mina, M. V.; Puzyk, I. P.; Puzyk, M. V. The Effect of Acids on Fluorescence of Coumarin-6 in
Organic Solvents. Opt. Spectrosc. 2013, 114, 244-246.
51. Vasylevska, A. S.; Karasyov, A. A.; Borisov, S. M.; Krause, C. Novel Coumarin-Based Fluorescent
4
5
pH Indicators, Probes and Membranes Covering a Broad pH Range. Anal. and Bioanal. Chem. 2007, 387,
2131-2141.
5
2.
Mason, M. D.; Ray, K.; Pohlers, G.; Cameron, J. F.; Grober, R. D. Probing the Local pH of Polymer
Photoresist Films Using a Two-Color Single Molecule Nanoprobe. J. Phys. Chem. B 2003, 107, 14219-
1
5
4224.
3.
Meyer, A. U.; Slanina, T.; Yao, C.-J.; König, B. Metal-Free Perfluoroarylation by Visible Light
Photoredox Catalysis. ACS Catal. 2016, 6, 369-375.
4. Garces, F. O.; King, K. A.; Watts, R. J. Synthesis, Structure, Electrochemistry, and Photophysics
of Methyl-Substituted Phenylpyridine Ortho-Metalated Iridium(III) Complexes. Inorg. Chem. 1988, 27,
5
3
5
464-3471.
5. Tinker, L. L.; McDaniel, N. D.; Curtin, P. N.; Smith, C. K.; Ireland, M. J.; Bernhard, S. Visible
Light Induced Catalytic Water Reduction without an Electron Relay. Chem. Eur. J. 2007, 13, 8726-8732.
6. Zhao, Q.; Liu, S.; Shi, M.; Wang, C.; Yu, M.; Li, L.; Li, F.; Yi, T.; Huang, C. Series of New Cationic
5
Iridium(III) Complexes with Tunable Emission Wavelength and Excited State Properties: Structures,
Theoretical Calculations, and Photophysical and Electrochemical Properties. Inorg. Chem. 2006, 45, 6152-
6
160.
57.
Zeng, X.; Tavasli, M.; Perepichka, I. F.; Batsanov, A. S.; Bryce, M. R.; Chiang, C. J.; Rothe, C.;
Monkman, A. P. Cationic Bis-Cyclometallated Iridium(III) Phenanthroline Complexes with Pendant
28
ACS Paragon Plus Environment

Page 29 of 30
ACS Catalysis
1
2
3
4
5
6
7
8
9
Fluorenyl Substituents: Synthesis, Redox, Photophysical Properties and Light-Emitting Cells. Chem. Eur.
J. 2008, 14, 933-943.
5
8.
Lamansky, S.; Djurovich, P.; Murphy, D.; Abdel-Razzaq, F.; Lee, H. E.; Adachi, C.; Burrows, P.
E.; Forrest, S. R.; Thompson, M. E. Highly Phosphorescent Bis-Cyclometalated Iridium Complexes:
Synthesis, Photophysical Characterization, and Use in Organic Light Emitting Diodes. J. Am. Chem. Soc.
2
5
001, 123, 4304-4312.
9. Marcus, R. A. Electron Transfer Reactions in Chemistry: Theory and Experiment (Nobel Lecture).
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Angew. Chem. Int. Ed. 1993, 32, 1111-1121.
0. Hasan, K.; Bansal, A. K.; Samuel, I. D. W.; Roldán-Carmona, C.; Bolink, H. J.; Zysman-Colman,
6
E. Tuning the Emission of Cationic Iridium(III) Complexes Towards the Red Through Methoxy
Substitution of the Cyclometalating Ligand. Sci. Rep. 2015, 5, 12325.
61.
Peloquin, A. J.; Smith, M. B.; Balaich, G. J.; Iacono, S. T. Crystal Structure of Chlorido(dimethyl
3
2
sulfoxide-ĸS)bis[4-(pyridin-2-yl)benzaldehyde-ĸ C ,N]iridium(III) Acetonitrile Monosolvate. Acta
Crystallogr., Sect. E, Cryst. Commun. 2017, 73, 1279-1281.
62.
Janet, J. P.; Liu, F.; Nandy, A.; Duan, C.; Yang, T.; Lin, S.; Kulik, H. J. Designing in the Face of
Uncertainty: Exploiting Electronic Structure and Machine Learning Models for Discovery in Inorganic
Chemistry. Inorg. Chem. 2019, 58, 10592-10606.
6
3.
and Improving Single-Site Polymerization Catalysts. A Review. Organometallics 2014, 33, 2944-2959.
4. Tortorella, S.; Marotta, G.; Cruciani, G.; De Angelis, F. Quantitative Structure–Property
Cruz, V. L.; Martinez, S.; Ramos, J.; Martinez-Salazar, J. 3D-QSAR as a Tool for Understanding
6
Relationship Modeling of Ruthenium Sensitizers for Solar Cells Applications: Novel Tools for Designing
Promising Candidates. RSC Adv. 2015, 5, 23865-23873.
6
5.
Fey, N.; Orpen, A. G.; Harvey, J. N. Building Ligand Knowledge Bases for Organometallic
Chemistry: Computational Description of Phosphorus(III)-Donor Ligands and the Metal–Phosphorus
Bond. Coord. Chem. Rev. 2009, 253, 704-722.
6
6.
Photonic Applications. In Iridium(III) in Optoelectronic and Photonics Applications, 2017; pp 1-69.
7. Li, Y.; Dandu, N.; Liu, R.; Hu, L.; Kilina, S.; Sun, W. Nonlinear Absorbing Cationic Iridium(III)
Deaton, J. C.; Castellano, F. N. Archetypal Iridium(III) Compounds for Optoelectronic and
6
Complexes Bearing Benzothiazolylfluorene Motif on the Bipyridine (N^N) Ligand: Synthesis,
Photophysics and Reverse Saturable Absorption. ACS Appl. Mater. Interfaces 2013, 5, 6556-6570.
68.
Poelma, S. O.; Burnett, G. L.; Discekici, E. H.; Mattson, K. M.; Treat, N. J.; Luo, Y.; Hudson, Z.
M.; Shankel, S. L.; Clark, P. G.; Kramer, J. W.; Hawker, C. J.; Read de Alaniz, J. Chemoselective Radical
Dehalogenation and C–C Bond Formation on Aryl Halide Substrates Using Organic Photoredox Catalysts.
J. Org. Chem. 2016, 81, 7155-7160.
6
9.
Color Regulation of Redox Potentials. Angew. Chem. Int. Ed. 2016, 55, 7676-7679.
0. Speckmeier, E.; Fischer, T. G.; Zeitler, K. A Toolbox Approach To Construct Broadly Applicable
Ghosh, I.; König, B. Chromoselective Photocatalysis: Controlled Bond Activation through Light-
7
Metal-Free Catalysts for Photoredox Chemistry: Deliberate Tuning of Redox Potentials and Importance of
Halogens in Donor–Acceptor Cyanoarenes. J. Am. Chem. Soc. 2018, 140, 15353-15365.
29
ACS Paragon Plus Environment