An integrated console for capsule-based, automated organic synthesis

Article abstract of DOI:10.1039/d1sc01048d

The current laboratory practices of organic synthesis are labor intensive, impose safety and environmental hazards, and hamper the implementation of artificial intelligence guided drug discovery. Using a combination of reagent design, hardware engineering, and a simple operating system we provide an instrument capable of executing complex organic reactions with prepacked capsules. The machine conducts coupling reactions and delivers the purified products with minimal user involvement. Two desirable reaction classes-the synthesis of saturated N-heterocycles and reductive amination-were implemented, along with multi-step sequences that provide drug-like organic molecules in a fully automated manner. We envision that this system will serve as a console for developers to provide synthetic methods as integrated, user-friendly packages for conducting organic synthesis in a safe and convenient fashion. This journal is

Full text of DOI:10.1039/d1sc01048d

Chemical
Science
View Article Online
View Journal
EDGE ARTICLE
An integrated console for capsule-based,
automated organic synthesis†
Cite this: DOI: 10.1039/d1sc01048d
a
Tuo Jiang,ab Samuele Bordi,a Angus E. McMillan,
Kuang-Yen Chen,a
*
All publication charges for this article
have been paid for by the Royal Society
of Chemistry
Fumito Saito, a Paula L. Nichols,ab Benedikt M. Wanner ab and Jeffrey W. Bode
a
*
The current laboratory practices of organic synthesis are labor intensive, impose safety and environmental
hazards, and hamper the implementation of artificial intelligence guided drug discovery. Using a combination
of reagent design, hardware engineering, and a simple operating system we provide an instrument capable of
executing complex organic reactions with prepacked capsules. The machine conducts coupling reactions
and delivers the purified products with minimal user involvement. Two desirable reaction classes – the
synthesis of saturated N-heterocycles and reductive amination – were implemented, along with multi-step
sequences that provide drug-like organic molecules in a fully automated manner. We envision that this
system will serve as a console for developers to provide synthetic methods as integrated, user-friendly
packages for conducting organic synthesis in a safe and convenient fashion.
Received 22nd February 2021
Accepted 9th April 2021
DOI: 10.1039/d1sc01048d
rsc.li/chemical-science
Introduction
Synthetic organic chemistry has long been practiced with an
assortment of standard – but not standardized – equipment
including stir plates, heating devices, cooling baths, glass asks,
and a menagerie of funnels, lters, extractors, and traps. While
this exibility has enabled synthetic chemistry to be executed with
minimal equipment at relatively low cost, the lack of standards for
reagents and apparatus contributes to poor reproducibility and
a reputation of capriciousness. More signicantly, traditional
practices impose inherent safety hazards and stymie efforts at
automation of basic techniques. As noted by Cronin,1 IBM,2,3 Jen-
sen, Jamison,4 and others the deciency of standard equipment
precludes the “digitalization” of organic synthesis, which is
necessary for mechanizing not only the production of valuable
organic molecules but also collecting and annotating data that will
make complex molecule synthesis more predictable and reliable.
The widely recognized desire to automate synthesis is man-
ifested in a delightful array of devices and contraptions for
organic synthesis.5,6 Innovative and robust chemical operating
systems for the iterative synthesis of oligomeric molecules, such
as peptides7,8 and DNA,9 have led to widespread deployment of
automated synthesizers for access to these molecular classes.
Researchers have also made outstanding progress on extending
aLaboratory for Organic Chemistry, Department of Chemistry and Applied Biosciences,
¨
¨
ETH Zurich, 8093 Zurich, Switzerland. E-mail: bode@org.chem.ethz.ch
b
¨
Synple Chem AG, 8093 Zurich, Switzerland. E-mail: wanner@synplechem.com
† Electronic supplementary information (ESI) available: Crystal structures,
experimental procedures, analytical data for all new compounds, spectra for all
new compounds and NMR spectra of compounds obtained directly from
machine. For ESI and crystallographic data (CCDC 2064034 and 2064033) in CIF
or other electronic format see DOI: 10.1039/d1sc01048d
Fig. 1 Approaches towards automatic organic synthesis with specially
developed instruments.
© 2021 The Author(s). Published by the Royal Society of Chemistry
Chem. Sci.

View Article Online
Chemical Science
Edge Article
these concepts to more demanding oligomers, including
carbohydrates10 and polyolens11 (Fig. 1).
More specialized targets that require diverse – and oen
operationally demanding
– experimental conditions have
proven more challenging to access in an automated fashion,
especially for de novo synthesis without an established route.
Current approaches include inspiring endeavours to add
elements of automation to traditional organic synthesis
equipment, as exemplied by the ChemKonzert12 and the
Chemputer,13 as well as grander implementations such as Eli
Lilly's Life Science Studio.14 At the current stage of development,
however, these approaches require considerable manual inter-
vention for setup, cleaning, and programming, precluding their
operation by non-specialists.
As an alternative to automating classical techniques for
organic synthesis, numerous practitioners have advocated new
modalities. For example, ow chemistry takes a fundamentally
different approach to organic reactions, which is particularly
well-suited for reliable scale up and production of optimized
routes.15,16 Highly specialised, self-contained systems are used
to prepare radioactive tracers in tiny quantities.17 Others have
promoted alternatives to traditional techniques for workup,
separation, and purication. This is perhaps best exemplied
by the use solid supported reagents and scavengers for complex
molecule synthesis, as champanioned by Ley and co-
workers.18,19 These solid-supported reagents and scavengers
have evolved into disposable kits for reaction screening/workup
that are widely used in the discovery phase of the pharmaceu-
tical industry. Despite the welcome step away from some of the
unsavory and time-consuming rituals of organic chemistry, the
lack of a standard platform and equipment for implementing
these concepts has hampered their widespread adoption.
In this report, we document a compact, capsule-based, fully
automated console for executing organic synthesis, including
the formation of saturated N-heterocycles, reductive amination,
and multi-step combinations. By recognizing that preparative
organic synthesis can be reduced to a few basic operations such
as mixing, heating, and transfers coupled with pre-packaged
reagents and resins, reaction protocols can be scripted and
executed with high reproducibility. The resulting union of
hardware engineering, soware development, reaction innova-
tion, and purication technologies provides a hands-free
approach to preparing organic compounds prized in drug
discovery. This ‘operating system’ for automating organic
synthesis obviates the need to directly handle most organic
reagents and waste, resulting in a safe, user friendly platform
suitable for unsupervised operation. We envision that future
organic methodologies will be implemented using the combi-
nation of optimized reagent capsules and short soware scripts,
thereby both improving safety and enabling their direct adop-
tion by end users, regardless of their technical knowledge of
preparative organic chemistry.
Fig.
2
Fluidics setup of the automated console. The setup is
composed of two valves, two syringe pumps and the heatable capsule
holder. Through different valve settings solvents can be pumped from
the solvent reservoirs or sample vial into the individual compartments
of the capsule or into the waste. The connected nitrogen supply
facilitates preserving an inert environment.
to construct it from off-the-shelf components. Accordingly, our
design began by combining commercial rotary valves and
syringe pumps with Teon or glass components (Fig. 2) that
were compatible with organic solvents. With this straightfor-
ward setup we can use up to ve solvents or mixtures, while
maintaining a container for waste collection. Although this
modest and somewhat inexible conguration will preclude
some reaction types, the majority of useful methodologies are
operationally compatible with this design. These choices also
facilitate construction within dened specications and reduce
set-up parameters that might limit operation by non-specialist
users. The key liquid handling elements were supplemented
by an electromagnetic stirrer, heating elements in the vial and
capsule holders, a touch screen interface, and an RFID scanner
(Fig. 3).
An automated reaction sequence is comprised of a series of
sequentially executed fundamental instructions (Table 1). Only
these seven instructions were sufficient to operate all the
components of the device – pumps, valves, heaters, stirrers and
capsule holder – to constitute a synthetic sequence. Additional
parameters are used to dene a specic operation. The
Results and discussion
Fig. 3 The user adds the starting material to the reaction vial, which is
loaded together with the cartridge into the console. The automated
reaction is started on the touchscreen. After completion the user can
In order to facilitate consistent assembly and operation of
a modular console for automated organic synthesis, we sought pick up the product in the reaction vial.
Chem. Sci.
© 2021 The Author(s). Published by the Royal Society of Chemistry

View Article Online
Edge Article
Chemical Science
Table 1 Set of instructions for the script that comprise a reaction sequence
Instruction
Parameters
Description
cmd_start
cmd_pump
cmd_valve
cmd_stir
cmd_heat
cmd_delay
cmd_end
—
Run start sequence (e.g. close capsule holder and prime pump)
Run the pump for a certain time with the dened pump speed and syringe volume
Turn valve 1 or 2 to the desired port number
Turn stirrer off or on with dened speed
Set one of the heaters to dened temperature
(Pump-speed, pump-time, pump-volume)
(Valve-number, port-number)
(Stir-on/off, stir-speed)
(Heater-number, temperature)
(Time)
Wait for dened period of time
Run end sequence (e.g. open capsule holder and reset instrument for next reaction)
—
microprocessor is capable of storing a multitude of reaction separate manipulations, each of which was assigned to
sequences including a self-cleaning program, which enables a compartment. Different compartment sizes were selected
rapid consecutive runs of the console, much like modern, depending on the quantity of reagent required. A case was
capsule-based automated chromatography systems or coffee designed to combine the four compartments into a single
makers. With an eye towards preventing operational mistakes, capsule. On top of the capsule assembly a Viton seal ensures the
the capsules include an RFID chip read by the console, which tightness of the connections, aer the insertion and automatic
ensures the correct program is loaded and that the console is connection of the capsule inside the console.
properly congured before the automated sequence begins
(Fig. 4).
As a candidate reaction for complete automation, we
selected the synthesis of substituted, saturated N-heterocycles
Aer several prototypes, we developed a custom engineered from aldehydes using stannyl amine protocol (SnAP) chem-
capsule-holder assembly and a housing unit that allows the istry,20–22 which is recognized as a versatile approach to
capsule to be easily exchanged without manual attachment of substituted morpholines, piperazines, oxazepanes, and other
uidic lines (Fig. 4). This key element ensures that reactions can attractive scaffolds. The chemistry proceeds with a broad
be setup in seconds from prepackaged capsules, guaranteeing substrate scope under consistent conditions (Cu(OTf)2, 2,6-
safe operation and cleaning of the console.
lutidine, CH2Cl2/HFIP (1,1,1,3,3,3-hexauoro-2-propanol)), but
For designing the capsule, we rst addressed the selection of utilizes potentially toxic tin reagents and requires labor inten-
materials considering existing supply chains. Polyethylene sive steps, including weighing of the reagents, complexation of
reservoirs were attractive as capsule compartments due to their the copper and ligand, preparation of an imine intermediate, an
low cost, commercial availability and resistance against intricate aqueous workup, and column chromatography. In
commonly used solvents and reagents. Packing solid reagents providing a fully automated implementation of SnAP chemistry,
in these reservoirs between two porous polyethylene frits facil- we considered all the steps and dedicated one compartment of
itates storage and allows for the immediate deployment on the the capsule to a specic task: (1) imine formation of the SnAP
console. The intended chemistry applications required four reagent with the aldehyde, (2) copper-mediated cyclization
using Cu(OTf)2/2,6-lutidine complex in HFIP, (3) workup to
remove copper salts, and (4) purication using silica-based
chromatography (Scheme 1). We planned
a system for
research scale (up to 0.5 mmol) operation, in which users need
only to add the aldehyde starting material; all other reagents
would be encapsulated into a sealed capsule; no reagents would
need to be weighed, and the console would deliver the puried
N-heterocyclic product with no further user involvement.
For imine formation, we generated solid-supported imino-
phosphoranes from polymeric triphenylphosphine and the
corresponding SnAP azides23 and packed this material into the
rst compartment. Extensive testing identied the optimal
conditions to be continuous circulation of a solution of the
aldehyde in CH2Cl2 over 2–3 equivalents of the iminophos-
phorane while heating at 50 ꢀC. The standard conditions for
cyclization to the saturated N-heterocycles requires the addition
of the imine to preformed Cu(OTf)2/2,6-lutidine complex in
HFIP. This was best achieved by loading solid 2,6-lutidinium
triate and Cu(OTf)2 into the second compartment, separated
by a polypropylene frit to ensure stability upon storage. Given
Fig. 4 The capsule consists of four compartments that are prefilled
with the necessary reagents. A chemically resistant seal on top and luer
connections at the bottom interface with the console. After the the low solubility of Cu(OTf)2 in the reaction media, the imine
capsule is inserted into the capsule holder, the fluidic lines connected
automatically via motorized sledge.
solution was recirculated through the Cu(OTf)2/2,6-lutidinium
© 2021 The Author(s). Published by the Royal Society of Chemistry
Chem. Sci.

View Article Online
Chemical Science
Edge Article
triate compartment to effect the cyclization and generate the
N-heterocycles. The cyclization was complete within 3 hours
when the compartment and the reaction vial were heated to
ꢀ
40 C.
For the workup and purication of the N-heterocycles, we
implemented a “catch & release” strategy. Aer cyclization, the
console loaded the reaction mixture onto silica-gel to remove
most of the copper salts and the ltrate was transferred onto
silica-supported propylsulfonic acid (SCX-2, 4 equivalents,
compartments 3 and 4) which protonated the basic N-
heterocycle and retained it on the support. This material also
acted as a metal scavenger and successfully removed the
remaining copper salts. Following washing of the resin with
MeOH to expel the non-basic byproducts, including any
unreacted aldehyde and most of the tin species, the desired
product was released in high purity (>90%) using a 2.5 M N,N-
diisopropylamine solution in THF in the majority of examples.
Typically, only a minimal amount of tin residue (<5%) was
observed (1H NMR analysis).
Using this instrument and the appropriate capsules, we
performed the automated synthesis of morpholines, oxaze-
panes, piperazines and diazepanes using aryl, heteroaryl and
alkyl aldehydes (Scheme 1). The corresponding substituted N-
heterocycles were obtained as THF solutions, which were
evaporated to afford the products in good yield and high purity
($90%). In general, aryl aldehydes bearing electron-
withdrawing groups, such as halides, uorinated alkyls, esters
Scheme
1 Automated formation of saturated N-heterocycles. A
broad scope of aldehydes are transformed to various saturated N-
heterocycle using SnAP reagent capsules. In most cases a purity of
>90% was obtained. See ESI† for detailed evaluation of the reaction
product after automated sequence (a) quantity of the desired product
directly after the automated sequence, calculated from1 H-NMR using
0.25 mmol (0.5 equiv.) of mesitylene as internal standard is given (b)
quantity of isolated product after column chromatography is given.
Scheme 2 Reductive amination process and products. Aldehydes and
ketones are coupled with primary or secondary amines with a capsule
for reductive amination. Yields of the reaction product (purity >95%)
directly after the automated synthesis are given.
Chem. Sci.
© 2021 The Author(s). Published by the Royal Society of Chemistry

View Article Online
Edge Article
Chemical Science
Scheme 3 Automated, multistep, capsule-based organic synthesis. The SnAP capsule and reductive amination capsule can be combined in two
consecutive machine runs to generate more complex structures. Yields were calculated from isolated material after purification by chroma-
tography; diastereomers -where present-were resolved in a later step by preparative HPLC for characterization; see ESI† for details.
and nitriles, gave good yields. Electron rich aldehydes – which the vial placed on the same console, and the script and capsule
are slower to form the imine intermediate – or alkyl aldehydes – for the reductive amination were loaded. The fully automated
which are prone to enamine formation – gave the cyclized sequence gave 6a in 30% overall yield as a mixture of two dia-
products with lower yield, but still in high purity.
stereomers. Several other attractive products were prepared (6b
By changing only the capsule contents and soware and 6c) by this automated sequence (Scheme 3). The protected
instructions, the same console can be used for other common carbonyl can be included in either the SnAP reagent (6a, 6b, 6c)
organic reactions. For example, we implemented a capsule- or the aldehyde substrate (6d, 6e and 6f).
based approach to reductive amination by loading one
Capsule designs and scripts for a number of other organic
compartment with solid supported cyanoborohydride, and transformations have been developed25 and console upgrades,
a second with SCX-2, to effect the purication using the same including the ability to employ low temperature and photo-
solid-phase extraction method described for the SnAP chem- chemical conditions, will make this approach suitable for many
istry. In this case, the user selects both coupling partners (up to other reaction classes. Ongoing innovations in facilitating the
a maximum of 0.5 mmol) and places them in the reaction vial execution of organic synthesis, such as reagents coated onto
together with the reaction solvent. Optimization of the reaction glass beads26 and immobilized catalysts and reagents,27 will
conditions identied CH2Cl2/HFIP (4 : 1) as the preferred facilitate the implementation of other important reactions onto
solvent system for the broadest range of test substrates. For this automated, consoled-based format.
reductive aminations between secondary amines and ketones,
We are aware that not every type or reaction encountered in
toluene at 40 ꢀC proved to be the preferred solvent and phe- organic synthesis can be implemented on such an instrument.
nylacetic acid – loaded into a third compartment – was used as The aim of this tool is to increase efficiency in the laboratory by
an activator. A fourth compartment was charged with solid automating routine steps, allowing more time to spend on
supported benzaldehyde, used as a scavenger for any unreacted challenging reactions that are harder or impossible to auto-
primary amine. Depending on the choice of carbonyl (aldehyde mate. The console also greatly facilitates standardization of
vs. ketone) and amine (1ꢀ or 2ꢀ) compounds different pre- reaction conditions and reagents, enhancing reproducibility
programmed reaction sequences where implemented; the user and reliability. With some innovation, we believe that capsules
selects the appropriate program on the touch display. A variety for reaction classes can be developed and innovations in
of compounds were synthesized with the optimized conditions, organic methods will obviate some of the more dangerous and
leading in most of the cases to high yields and purity within 3 to unpleasant practices that cannot be easily implemented.
5 hours (Scheme 2).
To execute fully automated, multistep organic synthesis, we
employed reagents containing a protected carbonyl and loaded
them into capsules.24 By simply adapting the script and SnAP
capsules to include a deprotection step on a polystyrene sup-
is easily observed with the widespread adoption of automated
ported acid we could telescope the N-heterocycle formation and
carbonyl deprotection. Aer the cyclization and deprotection
were complete, the solvent was removed, an amine was added,
require little user involvement. Our console/capsule/script
Conclusions
The desire for automation in medicinal chemistry laboratories
purication instruments, which use pre-packed silica-gel
cartridges for the purication of organic compounds and
© 2021 The Author(s). Published by the Royal Society of Chemistry
Chem. Sci.

View Article Online
Chemical Science
Edge Article
approach brings the same level of convenience, reliability, and
automation to the setup, execution, workup, and purication of
organic reactions, using a compact, user-friendly setup. We
envision instruments such as the ones described here will serve
as standardized platforms upon which automated chemistry
will be developed. To make an analogy, the console serves as the
computer and operating system while the capsules are the
applications. Organic chemists will become the developers who
package their methods into integrated, user-friendly packages
that can be used immediately in a safe and convenient fashion.
H. Gao, R. W. Hicklin, P. P. Plehiers, J. Byington,
J. S. Piotti, W. H. Green, A. J. Hart, T. F. Jamison and
K. F. Jensen, Science, 2019, 365, eaax1566.
5 M. Trobe and M. D. Burke, Angew. Chem., Int. Ed., 2018, 57,
4192.
6 S. V. Ley, D. E. Fitzpatrick, R. M. Myers, C. Battilocchio,
R. J. Ingham, et al., Angew. Chem., Int. Ed., 2015, 54, 10122.
7 R. B. Merrield, J. Am. Chem. Soc., 1963, 85, 2149.
8 M. Amblard, J.-A. Fehrentz, J. Martinez and G. Subra, Mol.
Biotechnol., 2006, 3, 239.
9 B. E. Kaplan, Trends Biotechnol., 1985, 3, 253.
10 A. Abragam Joseph, A. Pardo-Vargas and P. H. Seeberger, J.
Am. Chem. Soc., 2020, 142, 8561.
Author contributions
J. W. B., B. M. W., and P. L. N. conceived of the idea. B. M. W.
built the console and designed the soware. T. J., K.-Y. C.
implemented and developed the SnAP capsules. S. B. developed
the reductive amination. A. E. M. optimized and performed
multi-step reactions. F. S. designed and synthesized the iSnAP
reagent. All authors contributed to data acquisition and anal-
ysis. J. W. B, B. M. W., A. E. M and P. L. N. wrote the manuscript.
11 J. Li, S. G. Ballmer, E. P. Gillis, S. Fujii, M. J. Schmidt,
A. M. E. Palazzolo, J. W. Lehmann, G. F. Morehouse and
M. D. Burke, Science, 2015, 347, 1221.
12 K. Machida, Y. Hirose, S. Fuse, T. Sugawara and
T. Takahashi, Chem. Pharm. Bull., 2010, 58, 87.
13 S. Steiner, J. Wolf, S. Glatzel, A. Andreou, J. M. Granda,
G. Keenan, T. Hinkley, G. Aragon-Camarasa, P. J. Kitson,
D. Angelone and L. Cronin, Science, 2019, 363, eaav2211.
14 A. G. Godfrey, T. Masquelin and H. Hemmerle, Drug
Discovery Today, 2013, 18, 795.
Conflicts of interest
This technology was developed with the support of funding
from the Swiss Commission for Technology Innovation, the
ETH Pioneer Fellowships, and the European Research Council
and licensed to the ETH spinoff company Synple Chem AG.
B. M. W., P. L. N., K.-Y. C. and J. W. B. are listed as inventors of
a patent (WO Pat., WO2017121724A1, 2016) related to this
technology. B. M. W., P. L. N., T. J. and J. W. B. are co-founders
of Synple Chem AG.
´
15 A.-C. Bedard, A. Adamo, K. C. Aroh, M. G. Russell,
A. A. Bedermann, J. Torosian, B. Yue, K. F. Jensen and
T. F. Jamison, Science, 2018, 361, 1220.
16 A. Adamo, R. L. Beingessner, M. Behnam, J. Chen,
T. F. Jamison, K. F. Jensen, J.-C. M. Monbaliu,
A. S. Myerson, E. M. Revalor, D. R. Snead, T. Stelzer,
N. Weeranoppanant, S. Y. Wong and P. Zhang, Science,
2016, 352, 61.
17 S. J. Oh, D. Y. Chi, C. Mosdzianowski, J. Y. Kim, H. S. Gil,
S. H. Kang, J. S. Ryu and D. H. Moon, Nucl. Med. Biol.,
2005, 32, 900.
18 S. V. Ley and I. R. Baxendale, Nat. Rev. Drug Discovery, 2002,
1, 573.
19 J. Tulla-Puche and F. Albericio, The Power of Functional Resins
in Organic Synthesis, WILEY-VCH, 2008.
20 C.-V. T. Vo, G. Mikutis and J. W. Bode, Angew. Chem., Int. Ed.,
2013, 52, 1705.
21 C.-V. T. Vo, M. U. Luescher and J. W. Bode, Nat. Chem., 2014,
6, 310.
22 M. U. Luescher, K. Geoghegan, P. L. Nichols and J. W. Bode,
Aldrichimica Acta, 2015, 48, 43.
Acknowledgements
Financial support was provided by the European Research
Council (ERC-2012-StG_20111012 – Starting Grant; ERC-2015-
¨
PoC – Hetspresso), ETH Zurich Pioneer Fellowship and Inno-
suisse (27406.1 PFLS-LS). We thank the LOC MS Service for
analysis and the LOC machine shop (Christoph Bartschi) for
¨
engineering support, Vijay Pattabiraman for resin evaluation
and project students who contributed to this work (Oliver Gro-
ninger, Leran Zhang, Konrad Leopold and Ralph Werner). X-ray
services were provided by SMoCC – The Small Molecule Crys-
tallography Center of ETH Zurich (www.smocc.ethz.ch).
23 W.-Y. Siau and J. W. Bode, J. Am. Chem. Soc., 2014, 4, 17726.
24 F. Saito, N. Trapp and J. W. Bode, J. Am. Chem. Soc., 2019,
141, 5544.
25 See https://www.synplechem.com/solutions/cartridges for
additional reaction classes and capsules.
26 N. P. Tu, A. W. Dombrowski, G. M. Goshu, A. Vasudevan,
S. W. Djuric and Y. Wang, Angew. Chem., Int. Ed., 2019, 58,
7987.
27 A. Kirschning, H. Monenschein and R. Wittenberg, Angew.
Chem., Int. Ed., 2001, 40, 650.
References
1 S. H. M. Mehr, M. Craven, A. I. Leonov, G. Keenan and
L. Cronin, Science, 2020, 370, 101.
2 A. C. Vaucher, F. Zipoli, J. Geluykens, V. H. Nair, P. Schwaller
and T. Laino, Nat. Commun., 2020, 11, 3601.
3 IBM RXN for Chemistry, https://rxn.res.ibm.com, accessed Jan
05, 2021.
4 C. W. Coley, D. A. Thomas, J. A. M. Lummiss, J. N. Jaworski,
C. P. Breen, V. Schultz, T. Hart, J. S. Fishman, L. Rogers,
Chem. Sci.
© 2021 The Author(s). Published by the Royal Society of Chemistry