A Validated "pool and Split" Approach to Screening and Optimization of Copper-Catalyzed C-N Cross-Coupling Reactions

Article abstract of DOI:10.1021/acs.joc.0c02392

A general method for the quick identification of effective catalytic systems for copper-catalyzed C-N cross-couplings is described. This is based on evaluating mixtures of copper sources, ancillary ligands, and bases in different solvents followed by two

Full text of DOI:10.1021/acs.joc.0c02392

pubs.acs.org/joc
Article
A Validated “Pool and Split” Approach to Screening and
Optimization of Copper-Catalyzed C−N Cross-Coupling Reactions
Raphael R. Steimbach, Philipp Kollmus, and Marco Santagostino*
Cite This: J. Org. Chem. 2021, 86, 1528−1539
Read Online
sı
*
Metrics & More
Article Recommendations
Supporting Information
ACCESS
ABSTRACT: A general method for the quick identification of effective
catalytic systems for copper-catalyzed C−N cross-couplings is described.
This is based on evaluating mixtures of copper sources, ancillary ligands,
and bases in different solvents followed by two deconvolution procedures,
which aim at identifying the most proper reagent combination in only three
distinct steps. Despite being a high-throughput approach in nature, the
proposed method utilizes only frugal technological platforms such as 24-
well microplates while offering a screening efficiencythe number of
executed experiments vs the total number of possible experimentshigher
than 95%. To facilitate visualization and mining of the high-throughput
experimentation (HTE) data, Visual Basic scripts have been developed,
which allow streamlining the extraction of raw HPLC data into TIBCO
Spotfire for the graphical display in the form of pie charts. The unique capabilities of this “pool and split” approach have been
demonstrated by applying it to literature known cross-coupling reactions. In every case, the described experimental setup was
validated by retrieving the original literature conditions in addition to exposing several additional solutions with a minimum number
of parallel experiments. Finally, examples are provided for the successful application of this HTE screening workflow to internal
projects.
formation.6 Such an approach would be of special benefit for
INTRODUCTION
■
copper-catalyzed cross-coupling reactions, which are proven to
be quite recalcitrant toward optimization as it often appears
that the success of a given protocol depends more on the
careful optimization of the catalytic system as a whole. Thus, in
addition to the choice of the ancillary ligand, the proper
selection of copper source, base, and solvent plays a central
role. Contributing to this is the fact that copper-catalyzed
reactions are recognized to involve a pool of dynamically
interconvertible complex species of a poorly predictable nature,
implying that the general design principles based on the steric
and electronic characteristics of the ligand appear not to hold
true as in the case of palladium catalysis.7 While a possible
solution to the problem of optimizing a transformation in such
a complex reaction space would be simply to screen as many
reagent combinations as possible via HTE, the adoption of
simpler combinatorial approaches of lower technological
profile could retain the appeal of a large multidimensional
The flourishing plethora of new synthetic methods coupled to
a constantly increasing substrate complexity poses significant
challenges to the skillful navigation in a multidimensional
reaction space. High-throughput experimentation (HTE)
techniques maximize experimental efficiency by increasing
the number of experiments per time unit while minimizing
material consumption.1 Toward a more effective exploration,
most recent efforts have focused on extreme reduction of the
reaction scale while increasing the density of experiments2 as
well as on microfluidics-based strategies.3 While these
approaches are undoubtedly of great value, they often require
nontrivial engineering solutions to cope with the wide diversity
of chemistry to be executed, including the compelling necessity
of homogeneity for the flow chemistry format, and intrinsically
generate a huge number of data points calling for the adoption
of advanced high-throughput analysis methods such as MISER
or MALDI4 and for more efficient protocols for data extraction
and mining. Simplified approaches that can substantially
reduce the number of reactions required to obtain a hit result
and that do not need specialized high-throughput instrumen-
tation are therefore appealing. Screening of mixtures of
reagents or substrates followed by deconvolution steps to
isolate the reactivity mode of interest can be particularly
effective to this end. However, such strategies have been
seldom used for discovering new reactivity patterns5 or for
identifying effective catalyst systems for a given trans-
Received: October 9, 2020
Published: December 31, 2020
https://dx.doi.org/10.1021/acs.joc.0c02392
© 2020 American Chemical Society
J. Org. Chem. 2021, 86, 1528−1539
1528

The Journal of Organic Chemistry
pubs.acs.org/joc
Article
a
Chart 1. Four Ligand Sets That Were Used in the Present “Pool and Split” Approach
a
The ligand sets were prepared by mixing equimolar amounts of the indicated ligands. Some amount of solvent was added as necessary. Ligand Set
CuLG1 is a liquid. Ligand Set CuLG2 was prepared as a DMSO 45% w/w solution. Ligand Set CuLG3 is a mixture of solids. Ligand Set CuLG4
was prepared as a methanol 50% w/w solution.
screen while minimizing the investments associated with
complex engineering and high-throughput analysis.
ligands8 L19−L24, respectively. This first screening step (step
A) aimed at selecting the best combination of ligand set/
solvent and was normally performed in a 24-well microplate on
a 20−30 mg scale. The subsequent screening step (step B) is
also a 24-reaction run and implies concomitant deconvolution
of the most active ligand set and of the copper source mixtures
in the solvent selected in the first discovery screening run. As
both steps use the three-base mixture, the last step (step C)
concludes the deconvolution protocol focusing on one ligand/
copper source/solvent combination and explores in three
additional experiments the effect of the base variation normally
on a larger scale (50−100 mg) in a 10-well parallel reaction
block. For this last step, it turned out to be often more practical
and advantageous to exploit the free positions in the reaction
block to include, in addition to the three initially selected
bases, a larger variety of both inorganic and organic bases.
During the deconvolution process, the copper catalyst loading
was reduced from 30 mol % in the discovery screening to 10
mol % in the final deconvolution step. In more challenging
cases, a careful optimization of the continuous experimental
variables via a statistical design of experiment (DoE) approach
was carried out.
RESULTS AND DISCUSSION
■
Taking inspiration from Wieland and Breit’s6a and Moran et
al.’s,6b6c combinatorial approaches, we devised a simplified
experimental setup that would allow an easy identification of
optimal conditions for copper-catalyzed C−N cross-coupling
reactions that is based on screening and deconvoluting
mixtures of catalytic components. Arrays of catalyst mixtures
were tested in standard parallel equipment, such as 24-well
aluminum reactor blocks equipped with 1 mL glass microvials.
Thus, 24 copper ligands, pooled in four sets of six, were
combined with mixtures of four copper sources (CuI, CuCl,
Cu2O, and CuO) and three bases (K2CO3, K3PO4, and
CsOAc) and tested in six mostly polar, high-boiling solvents.
This allowed 1728 combinations to be evaluated in only 24
parallel experiments in the first discovery screening run. The
pooling strategy for the 24 copper ligands takes into
consideration structural and reactivity similarities and it was
meant to limit excessive dispersion of positive hits across the
plate (Chart 1). The physical state of the ligands was also
considered in order to facilitate dosing of the ligand mixtures,
which was carried out using commercially available solid and
liquid handling technologies. Ligand Set 1 contained standard
aliphatic diamines L1−L6 (CuLG1). Ligand Set 2 was mainly
composed of O-ligands such as polyols and dicarbonyls L7−
L12 (CuLG2). Ligand Sets 3 (CuLG3) and 4 (CuLG4) were
composed of various common copper O, N ligands (L13−
L18) and of some of the most prominent Ma et al.’s oxalamide
In situ assay yields were determined by HPLC using external
standards of key reaction components, specifically the desired
reaction product and starting material, but quantification of
known side products can be also be included in the workflow,
if desired. An internal standard was added during HPLC
sample preparation to normalize for dilution errors. In order to
streamline the comparative analysis of the reaction mixtures
while minimizing manual data manipulation, we developed an
1529
https://dx.doi.org/10.1021/acs.joc.0c02392
J. Org. Chem. 2021, 86, 1528−1539

The Journal of Organic Chemistry
pubs.acs.org/joc
Article
Figure 1. Streamlined data processing and data mining workflow for HTE.
automated data processing workflow using Visual Basic scripts
that extract the peak areas of preselected peaks from all the
HPLC chromatograms of the screening runs, saved as comma
separated value (.csv) files, into standardized Excel spread-
sheets. TIBCO Spotfire software eventually extracts the
numerical data and text strings from the Excel files and in
addition associates the primary HPLC chromatogramsin
form of Enhanced Metafiles (EMF files)to the correct
position in the plate for manual inspection, if needed. Thus,
overviews of the 24-well microplates are dynamically
generated, most commonly in the form of pie charts, which
display graphically the content of each well, providing a
convenient and flexible visualization for data analysis and
mining (Figure 1). Each pie sector indicates the amount, in
terms of assay yield, of key reaction components. The
occurrence of decomposition pathways or side reactions is
revealed by a total pie area below 100%.
We decided to validate the present “pool and split” approach
on four C−N cross-coupling reactions reported in the
literature. The selected transformations cover cross-couplings
of both NH acidic and NH basic components with aryl iodides
and bromides. Before embarking on the “pool and split”
optimization approach, the literature reactions were repro-
duced to examine the impurity profile and to access the
reaction products as well as the most relevant process
impurities.
The first reaction that was evaluated was the N-arylation of
3-methyl-pyridin-2(1H)-one 2 with iodoaniline 1, which is
reported to occur in 78% yield in the presence of 15 mol %
CuI/15 mol % oxine L16/K2CO3 in DMSO at 130 °C (Figure
2).9 The described conditions were repeated twice on a 0.8
and 4.0 mmol scale, and the N-arylated product 3 was isolated
in 60−65%. The main reason for the moderate yield was the
incomplete consumption of iodoaniline 1 and the formation of
a number of unidentified impurities in small quantities (see the
Supporting Information).
This first “pool and split” discovery screening was run in
duplicate in order to verify the plate-to-plate reproducibility
using a total of 30 mol % copper salt loading (7.5 mol % of
each copper source) and 30 mol % ligands (5% of each ligand)
at 100 °C for 18 h. An additional set of 24 parallel experiments
was also run at the same copper loading using this time a 1:2
copper/ligand ratio (i.e., 60 mol % ligands). In all three cases,
quite similar results were observed, with some more
Figure 2. Literature conditions9 for the N-arylation of 2 with 2-fluoro-
4-iodoaniline 1 and yield after “pool and split” optimization.
pronounced scattering only in the case of the relatively low-
boiling acetonitrile and no consistent effect of the copper/
ligand ratio (Supporting Information). Thus, all the subse-
quent screening activities were run using a more practical
copper/ligand ratio of 1:1. The results of the initial 24
experiments indicated complete or almost complete con-
sumption of 1 and good assay yield of 3 using a number of
different catalyst mixtures. Appealingly, with the Ligand Set
CuLG3, which contains the reported ligand oxine L16, an
assay yield of 85% was measured when working in DMSO.
This initial hit was deconvoluted in two runs, first by splitting
the ligand set (20 mol % per ligand) and the copper mixture
(20 mol % per copper source) and finally by examining the
effect of the single bases at 10 mol % catalyst loading. At the
end of the deconvolution process, the mixture 10 mol % CuI/
10 mol % oxine/K2CO3 in DMSO at 100 °C was found to
deliver about 75% assay yield together with some 10% starting
material 2, which is in line with what was measured under the
literature conditions (Scheme 1, blue pathway). Perhaps more
intriguing, this series of experiments revealed a number of far
more efficient and practical catalytic systems. Considering
scalability10 and cost11 aspects, we chose to follow up a result
from the aliphatic amines Ligand Set CuLG1 in the process
friendly solvent t-AmOH. Complete deconvolution of this hit
and the inclusion of a wider variety of bases in the last
deconvolution step led to the discovery of three catalytic
systems delivering around 90−95% assay yield, namely,
combinations of 10% CuI and 10% trans-N,N′-dimethylcyclo-
hexane-1,2-diamine L2 in t-AmOH in the presence of CsF,
K2CO3, or K3PO4 (Scheme 1, red pathway). The screening
results were confirmed on a 4.2 mmol scale using the low-
1530
https://dx.doi.org/10.1021/acs.joc.0c02392
J. Org. Chem. 2021, 86, 1528−1539

The Journal of Organic Chemistry
pubs.acs.org/joc
Article
Scheme 1. Screening and Deconvolution Strategy for the N-Arylation of 2 with 1a
a
Spotfire pie charts for the HTE results of the screening and deconvolution steps. Step A (gray box): selection of the solvent and the ligand set.
Step B (yellow box): selection of the ligand and the copper source. Step C (green box): selection of the base. The size of each pie sector indicates
the content of key reaction components: limiting starting material 1 (yellow sectors) and desired product 3 (green sectors). The numbers reported
in the colored boxes specify the assay yield of the product. Blue arrows and boxes in the left part of the scheme indicate the deconvolution workflow
leading to the literature reported conditions. The red arrows and boxes lead to the determination of improved conditions.
molecular-weight and nonhygroscopic K2CO3. With this base
(2 equiv), the N-arylation reaction was complete within 5 h in
refluxing t-AmOH and 3 was isolated in 92% yield (98% assay
yield).
The second case study was the more challenging N-arylation
of 3-methyl-pyridin-2(1H)-one 2 with the 4-iodobenzonitrile
4. This is reported to occur with the same catalytic system as
above (30 mol % CuI/30 mol % oxine/K2CO3 in DMSO) in
58% isolated yield (Figure 3).
We were able to reproduce this result and to isolate the
product 5 in a comparable yield (48%) on a 0.8 and 4.4 mmol
scale. From the reaction mixture, we could also isolate and
characterize the three major impurities, namely, the O-arylated
product 6 (15 area%) and the two side compounds 7 (19 area
%) and 8 (3 area%), coming from partial hydrolytic pathways
of the nitrile moiety under basic conditions. For this cross-
coupling again, the screening and deconvolution approach
exposed the originally reported combination (Scheme 2, blue
pathway). Interestingly enough, since the deconvolution was as
a rule performed at a lower temperature than the literature
conditions (100 °C instead of 130 °C), HTE assay yields of 5
were better and around 80−85% as a much reduced formation
of the hydrolyzed side products 7 and 8 (<5%) was generally
observed. However, the N/O arylation ratio (i.e., 5/6), as
determined by HPLC, could not be much improved and stayed
stable at around 4:1 with both K2CO3 (the base utilized in the
literature procedure) and K3PO4 using oxine L16 in DMSO.
As in the previous case, the amines Ligand Set CuLG1 in t-
AmOH offered a practicable alternative to the literature
conditions. Starting with an assay yield of around 70% in the
first discovery screening step, essentially quantitative assay
yields were registered after two deconvolution steps with the
catalytic systems 10% CuI and 10% trans-N,N′-dimethylcyclo-
hexane-1,2-diamine L2 in the presence of 2 equiv of either CsF
or K2CO3 in t-AmOH after stirring overnight at 100 °C
(Scheme 2, red pathway). Under these conditions, complete
regioselectivity and only trace amounts of the previously noted
side products, including the O-arylated 6, were observed. The
reaction scaled up well and, on a 4.4 mmol scale, was complete
Figure 3. Literature conditions9 for the N-arylation of 2 with 4-
iodobenzonitrile 4 and the major side products isolated from the
reaction mixture and yield after the “pool and split” optimization.
1531
https://dx.doi.org/10.1021/acs.joc.0c02392
J. Org. Chem. 2021, 86, 1528−1539

The Journal of Organic Chemistry
pubs.acs.org/joc
Article
Scheme 2. Deconvolution Strategy for the N-Arylation of 2 with 4a
a
Spotfire pie charts for the HTE results for the screening and deconvolution steps leading to the literature reported conditions (blue pathway) and
to the improved conditions (red pathway).
already after 2 h, returning a 92% assay yield (74% isolated
yield).
and the reaction product 10 could be isolated in 49% yield
after chromatography (vs 86% lit).
With 4-bromoanisole 4, the initial discovery screening step
confirmed the peculiarity of the oxalamide ligandsLigand Set
CuLG4which were active in three out of the six tested
solvents. Interestingly, the subsequent deconvolution step in
DMSO pointed to DMPAO L19 as the most active ligand in
the CuLG4 group, either in the presence of CuI (43% assay
yield) or Cu2O (52% assay yield). Focusing on the literature
copper source CuI, a final base refinement in a 10-well format
revealed a positive effect of CsOAc (27%) and CsF (43%) in
addition to the literature base (K3PO4, 18% assay yield)
(Scheme 3). The generally lower assay yields observed in this
second deconvolution step performed in a 10-well SK233
reactor block were attributed to the partial separation of
important amounts of bromoanisole from the reaction mixture
in the form of a dense liquid, which condensed in the
headspace of the 10 mL reaction vials used at this stage. To
overcome this phase separation issue and considering the good
reaction performances also observed with the lower boiling
solvent t-AmOH (Scheme 3), a DoE-driven optimization step
was undertaken using the more active base CsF and utilizing
the solvent mixture DMSO/t-AmOH (1:1) at 120 °C.12 With
this solvent combination and working above the refluxing
temperature of t-AmOH, the phase separation issue was
completely avoided. Complete consumption of bromoanisole
was indeed observed and an assay yield of 86% (64% isolated
yield on a 5.0 mmol scale) could be obtained after 20 h under
the following optimized conditions: 10% CuI, 10% DMPAO, 3
equiv of CsF in 10 vol DMSO/t-AmOH (1:1) at 120 °C.
The coupling of N-benzylmethylamine with the N-
unprotected 5-bromoindole 11, which is reported to occur
with a moderate yield of 60% using combinations of CuI,
The two subsequent cases that we set out to examine
regarded the far more difficult copper-catalyzed cross-coupling
of aryl bromides. Specifically, the reactions of N-benzylmethyl-
amine with both 4-bromoanisole 9 and with the N-unprotected
bromoindole 11 were evaluated. The reported conditions for
these cross-couplings utilize combinations of 10% CuI and
20% of the oxalamide DMPAO (L19) in DMSO at 90 °C and
in n-BuOH at 110 °C,8a both in the presence of K3PO4. The
N-arylation of N-benzylmethylamine with 4-bromoanisole
proceeded uneventfully on a 7.0 mmol scale under the
literature conditions without formation of notable side
products, besides traces (<5%) of benzaldehyde (Figure 4).
However, after the reported reaction time and despite excess
N-benzylmethylamine, conversion of 4-bromoanisole into
product 10 was unsatisfactory and about 45% at 230 nm
(notably however, a 98% conversion was measured at 254 nm)
Figure 4. Literature conditions8a for the N-arylation of N-
benzylmethylamine with 4-bromoanisole and yield after the “pool
and split” optimization.
1532
https://dx.doi.org/10.1021/acs.joc.0c02392
J. Org. Chem. 2021, 86, 1528−1539

The Journal of Organic Chemistry
pubs.acs.org/joc
Article
Scheme 3. Deconvolution Strategy for the N-Arylation of N-Benzylmethylamine with 4-Bromoanisolea
a
Spotfire pie charts for the HTE results for the screening and deconvolution steps leading to the literature reported conditions (blue pathway). A
DoE-based optimization step was added in this case to reach optimal conversions using CsF as a base and DMSO/t-AmOH (1:1) as a solvent
mixture. A partial deconvolution in t-AmOH was also carried out, showing the effectiveness of this solvent using Cu(I) sources (red pathway).
oxalamide DMPAO (L19), and K3PO4 in boiling n-BuOH,8a
returned in our hands a comparable 44% yield on a 10 mmol
scale using the literature conditions. Despite the excess of N-
benzylmethylamine, this amination was characterized by a
significant cross reactivity of the reaction product 12 with the
bromide starting material 11 leading to important amounts
(HPLC ratio 12/13 was 2:1) of the N-arylated indole 13
(Figure 5).
initial ligands/solvents screen returned in fact very few positive
hits. Interestingly enough, best results were with the
oxalamides Ligand Set CuLG4, however with negligible assay
yields: 4% in t-AmOH and 3% in DMSO. In general, the
reactions were largely incomplete after 18 h at 100 °C and
important formation of several side products was observed.
Besides this, the alcohol solvent selected in the literaturen-
butanolwas not part of our prototypical “pool and split”
experimental setting, making the discovery screening more
challenging. Thus, while with the first deconvolution step in t-
AmOH, we could identify conditions similar to those reported
(CuI/DMPAO L19 / K3PO4), the yield remained unsat-
isfactorily low around 10% (Scheme 4). Interestingly however,
focusing on the more active pair Cu2O/DMPAO and
completing the deconvolution steps by including a wider
variety of bases, CsF was identified as a pertinent additive,
returning a 43% assay yield at 10% copper loading. As before, a
more careful optimization of the quantitative parameters was
necessary to improve conversion and selectivity to an
acceptable level with a 64% assay yield was achieved by
applying a response surface-based optimization method13 so
that the product 12 was isolated after 45 h at 110 °C in 60%
yield. The new conditions compared thus very well with the
literature results and are more selective as formation of 13 is
much reduced (ratio 12/13 is 4:1). However, formation of
some amounts of arylated DMPAO ligand 14 was observed
together with some unreacted starting material 5-bromo-1H-
indole 11.
This cross-coupling turned out to be utmost challenging for
the application of the present “pool and split” approach. The
Having demonstrated the general validity of the present
“pool and split” approach for literature known examples of
copper-catalyzed C−N cross-couplings by rapidly exposing
effective reagent combinations, we decided to apply the same
Figure 5. Literature conditions8a and side products for the N-
arylation of N-benzylmethylamine with 5-bromo-1H-indole 11 and
yield after the “pool and split” optimization.
1533
https://dx.doi.org/10.1021/acs.joc.0c02392
J. Org. Chem. 2021, 86, 1528−1539

The Journal of Organic Chemistry
pubs.acs.org/joc
Article
Scheme 4. Deconvolution Strategy for the N-Arylation of N-Benzylmethylamine with 5-Bromo-1H-indolea
a
Spotfire pie charts for the HTE results for the screening and deconvolution steps. Improved conditions based on combinations of Cu2O and CsF
in t-AmOH (red pathway) were found after complete deconvolution and DoE-based optimization.
strategy to some internal projects. As prototypical examples,
we report below “pool and split” screening results for the N-
arylation of the N-alkyl-substituted piperazine 16 with the
commercially available electron-rich aryl bromide 15 (Figure
6) and of tert-butyl carbazate with the substituted iodopyridine
18 (Figure 7).
discovery screening step, combinations of the polar aprotic
solvents DMSO or NMP and O-ligands L7−L12 (CuLG2)
offered acceptable solutions with assay yields of about 15%. A
further combinationt-AmOH/oxalamide ligands of
CuLG4was also exposed but was not further pursued at
this stage. The last deconvolution steps centered around the
DMSO/CuLG2 system and allowed the rapid identification, in
only two separate steps, of conditions based on the TMHD
ligand L7 and CsOAc as the optimum base. This combination
was found to deliver on a screening scale (0.4 mmol) a quite
clean coupling reaction with 42% assay yield and 45%
conversion after 18 h at 100 °C using only a limited excess
of the N-alkyl-substituted piperazine 16 (Scheme 5).
Using standard literature conditions 50 mol % CuI and
Cs2CO3 in DMF, the Ullmann cross-coupling of the
iodopyridine 18 with tert-butyl carbazate (Figure 7) proceeded
sluggishly and the desired product 19 was obtained in only
moderate yield (50%) while significant amounts of des-
iodopyridine were formed.
Figure 6. Best screening results (42% assay yield) for the N-arylation
of the N-alkyl-substituted piperazine 16 with the aryl bromide 15.
Again, application of the standard “pool and split” protocol
identified quickly several active catalytic systems, delivering
complete conversions and improved selectivity toward
deiodination. In particular, these are combinations of 1,2-bis-
amines (Ligand Set CuLG1) in DMSO or, to a lesser extent,
1,3-diketones or polyalcohols (Ligand Set CuLG2) in NMP
(Scheme 6). The former combination was completely
deconvoluted, leading to final screening conditions delivering
a 93% assay yield and only 5% deiodination: 10% CuO, 10%
TMEDA, and K2CO3 (2 equiv) in DMSO at 100 °C.
Figure 7. Best screening results (93% assay yield) for the N-arylation
of tert-butyl carbazate with the iodopyridine 18.
According to the workflow previously defined, the first 24-
well screening run was performed at 30 mol % copper loading
for 18 h at 100 °C followed by two deconvolution steps in a
24-well plate and 10-well reaction block at reduced copper/
ligand loadings of 20 mol % and 10 mol %, respectively. In the
These conditions were finally refined by applying a “micro-
DoE” in 24-well microplates on a 10 mg/well scale with 5 mol
% CuO loading. To assure precise and reproducible copper
catalyst dosing at these very low amounts, CuO was added
1534
https://dx.doi.org/10.1021/acs.joc.0c02392
J. Org. Chem. 2021, 86, 1528−1539

The Journal of Organic Chemistry
pubs.acs.org/joc
Article
16 runs) was utilized to evaluate the effect of changing the
reaction temperature (80−100 °C) as well as the stoichiometry
of K2CO3 (1.0−3.0 equiv), tert-butyl carbazate (1.1−2.0
equiv), and TMEDA (5−15 mol %). Gratifyingly, an excellent
reproducibility of the results was observed on this very low
scale. The best balance between conversion and selectivity was
obtained at low base levels, high tert-butyl carbazate levels, and
high temperatures. The effect of TMEDA (i.e., the L/Cu ratio)
was not significant instead, showing that this ligand can be
varied between 5 and 15% without significant impact. The
“sweet spot” indicating the best compromise for essentially
complete conversion (>97%) and reduced deiodination (<3%)
is pictured in the plot below (Figure 8), and these results were
confirmed on a larger scale with a significant scaling up factor
(1000×).
Scheme 5. Deconvolution Strategy for the N-Arylation of N-
Alkyl-substituted Piperazine 16 with the Aryl Bromide 15a
a
Spotfire pie charts for the HTE results for the screening and
deconvolution steps. The deconvolution workflow leading to proper
initial coupling conditions is highlighted.
Scheme 6. Deconvolution Strategy for the N-Arylation of
tert-Butyl Carbazate with the Iodopyridine 18a
Figure 8. Design Expert overlay plot for the N-arylation of the tert-
butyl carbazate (Boc-hydrazine) with the iodopyridine 18. Design
Expert overlay plot showing the “sweet spot” i.e., the operating
window framed by an acceptable parameter rangefor achieving less
than 3% deiodination and more than 97% conversion: 5% CuO, 5%
TMEDA, and K2CO3 (1−1.5 equiv) in DMSO at 95−100 °C for 20
h. The overlay plot is a graphical optimization tool that allows to
identify an area of the design space where multiple response criteria
are met. The yellow area defines the acceptable factor settings, while
the gray area the unacceptable ones. The red points indicate the
position of experimental runs within the reaction space under
evaluation. Additional details are given in the Supporting Information.
In summary, we have developed an operationally simple yet
effective screening and optimization approach for copper-
catalyzed C−N cross-coupling reactions, which is based on
screening and deconvoluting mixtures of catalytic components
in three distinct steps using standard, inexpensive, and
commercially available equipment for parallel chemistry. In
the first discovery screening step, 24 copper ligands
appropriately pooled in four structurally homogeneous sets
of six ligands are combined in six solvents with mixtures of
copper sources and of bases in order to be able to evaluate
1728 combinations in one 24-well aluminum microplate. Two
subsequent deconvolution steps in the solvent selected in the
first step allow the identification of the best combinations of
a
Spotfire pie charts for the HTE results for the screening and
deconvolution steps. The deconvolution workflow leading to proper
initial coupling conditions is highlighted in blue. Assay yields for
relevant side compounds are also graphically depicted: des-
iodopyridine (red sectors) and hydrolyzed iodopyridine (violet
sectors).
absorbed on glass beads, a technique recently disclosed by
scientists at Abbvie.14 A replicated full Factorial Design (2 ×
1535
https://dx.doi.org/10.1021/acs.joc.0c02392
J. Org. Chem. 2021, 86, 1528−1539

The Journal of Organic Chemistry
pubs.acs.org/joc
Article
aluminum top and the reactor base with nine evenly placed screws
and closed within the glovebox before transferring to the heating
block. The reactions were heated and stirred on a heating block with a
tumble stirrer (V&P Scientific) using 5.08 mm-diameter uncoated stir
disks (V&P Scientific VP 721F) located outside the glovebox.
Reaction screenings in 10-well format were carried out under argon in
a 10-well STEM reaction block with aluminum inserts in low-volume
custom-made 10 mL vials (11 × 100 mm) equipped with a glass-inert
condenser system. Before heating was applied, the reaction mixtures
were degassed with vacuum/argon cycles (five times). Assay yields
were either measured using an internal standard (biphenyl) in the 24-
well format experiments or using external references in the 10-well
format screening.
Preparation of the Reagent Mixture. The base mixture was
prepared by grinding equimolar amounts of K2CO3, K3PO4, and
CsOAc into a fine powder under a nitrogen atmosphere. The copper
source mixture was prepared by grinding equimolar amounts of CuI,
CuCl, CuO, and Cu2O (half quantity) into a fine powder under a
nitrogen atmosphere. The ligand mixtures CuLG1−CuLG4 were
prepared by combining equimolar amounts of the ligands in Chart 1
into a homogeneous powder or by dissolving equimolar amounts of
the ligands in an appropriate degassed solvent: Ligand Set CuLG2 was
prepared as a DMSO 45% w/w solution and Ligand Set CuLG4 was
prepared as a methanol 50% w/w solution.
General Screening and Deconvolution Workflow. In the first
screening step (step A), a 24-well aluminum reaction plate with 1.0
mL vials and stirrers was charged in a glovebox with the aryl halide
(20−30 mg, 1.0 equiv), the amine/amide nucleophile (1.2−1.5
equiv), the base mixture (2.0 equiv in total, 0.5 equiv of each base),
the copper salt mixture (30 mol % of four copper species, 7.5 mol %
each), and the ligand mixture (30 mol % of six ligands, 5 mol % each).
The six different solvents (300 μL dry, degassed) were added as
indicated and the reaction plate was closed under a nitrogen
atmosphere before removing it from the glovebox and stirring at
100 °C for 18 h in a tumble stirrer equipped with an anodized
aluminum heating block. After cooling to room temperature, assay
yield was determined by diluting with 350 μL methanol containing 10
mg/mL biphenyl as an internal standard. After thorough mixing, 25
μL of aliquots was diluted in 1250 μL methanol, filtered if necessary,
and analyzed by HPLC. Product yield and the remaining starting
material were determined by signal integration relative to the internal
standard signal.
In the first deconvolution step (step B), the four copper salts are
separately screened in the solvent identified in step A against the six
components of the ligand group previously selected. This step uses
the same mixture of bases as in the step A. The experimental setup
and procedure are identical to those in Step A, but catalyst/ligand
loading is reduced to 20 mol % each.
The last deconvolution step (step C) uses 10% of the catalyst/
ligand combination selected in step B and may include a wider base
selection (up to 10−24 bases, 2 equiv each). This step is normally
performed on a higher scale (100 mg aryl halide) in a parallel
synthesis block with reflux condensers under an argon atmosphere or
again in a 24-well aluminum reaction plate on a lower scale (20−30
mg). After charging all the solid components, air was excluded using
vacuum/argon cycles followed by addition of the liquid reagents and
1.0 mL (300 μL for the 24-well plate) of solvent. Reactions were
stirred at 100 °C for 18 h, and assay yield was determined by diluting
to 100 mL with aqueous methanol and analyzing on an HPLC
instrument calibrated to external standards or as described above
(internal standards) for the 24-well format.
ligand/copper source/base with a reduced number of parallel
experiments (24 + 3−10 runs) and are open to further
customization, if required. Actually, the present “pool and
split” screening workflow is very flexible in nature as different
combinations of catalytic componentsbesides those used in
the described setupas well as different screening formats
(e.g., 96-well screening plates) may also be used. Despite using
frugal technological platforms, the current screening efficiency,
calculated as the number of actually executed experiments
(<60) vs the total number of possible experiments (1728 in the
present experimental set), is very high and around 95%. The
described experimental setup was validated on a number of
literature known cross-coupling reactions and was able to
retrieve in every case the original literature conditions. More
interestingly, the high information density of this approach was
usually able to expose several additional solutions, which could
be further followed up as desired. Finally, we demonstrated the
efficacy of the present “pool and split” approach in the case of
internal projects as the N-arylation of an N-alkyl-substituted
piperazine with an electron-rich aryl bromide and for the N-
arylation of the tert-butyl carbazate with an iodopyridine.
Currently, this screening approach represents our standard
procedure when approaching new copper-catalyzed cross-
coupling reactions and appears to deliver consistently high-
quality screening results, even in the presence of various
functional groups or on more complex substrates (such as
amides, esters, arylchlorides, indoles, or indazoles), which are
used as a robust basis for further implementation on a larger
scale. We anticipate that this strategy will enable the screening
and optimization of a broader range of chemical trans-
formations.
EXPERIMENTAL SECTION
■
General. All reactions mixtures were prepared under a purified
nitrogen atmosphere in a glovebox. All reagents and starting materials
were used as purchased from commercial suppliers (Sigma Aldrich
and ABCR) and used without further purification. Solvents were
anhydrous, sure-seal quality, and used with no further purification.
The ligands and additives were purchased from commercial sources
(Sigma Aldrich and TCI) or prepared as described in the literature
and stored in the glovebox. Nuclear magnetic resonance (NMR)
spectra were recorded on a Bruker Avance 400 or Bruker Avance 600
spectrometer with tetramethylsilane as an internal reference.
Chemical shifts are reported as parts per million (ppm), and coupling
constants (J) are given in Hertz (Hz). HPLC assays were carried out
using a C-18 reversed-phase column (Waters, XBridge BEH C18, 50
× 3.0 mm; 3.5 μm) eluting with 0.2% NH3(aq) and acetonitrile
(MeCN). Compound elution was monitored between 210 and 320
nm (DAD). Flash chromatography was performed using a silica gel 60
(0.040−0.063 mm, 230−400 mesh). HRMS experiments were
performed on a Thermo Scientific LTQ Orbitrap XL or Waters
Synapt G2-Si spectrometer using electrospray ionization in positive
ion mode (ESI+).
High- and Medium-Throughput Experimentation Equip-
ment and Materials. Reaction screenings in 24-well format were
carried out in 1.0 mL vials (8 × 30 mm) in a 24-well plate aluminum
reactor block (from Analytical Sales and Services). Liquid chemicals
or solutions were dosed using electronic multi- or single-channel
pipettes inside the glovebox. Undesired addition solvent was removed
using a Christ vacuum centrifuge. Solid chemicals were dosed in the
Microsoft Excel and TIBCO Spotfire were used for data analysis
and visualization.
Further reaction optimization (DoE, kinetic experiments) was done
in a parallel synthesis block and the results quantified against external
standards. DoE runs were planned and analyzed using Design Expert
10.
Preparation of 1-(4-Amino-3-fluorophenyl)-3-methylpyridin-
2(1H)-one 3.9 In a 20 mL reactor, 2-fluoro-4-iodoaniline 1 (1.00 g,
4.21 mmol, 1.00 equiv), 3-methyl-pyridin-2(1H)-one 2 (0.553 g, 5.07
́
́
1.0 mL vials using a Protege dosing unit (Unchained Labs) located
inside the glovebox. Below each reactor vial in the aluminum 24-well
plate was a 0.062 mm-thick silicon-rubber gasket. Directly above the
glass vial reactor tops was a Teflon perfluoroalkoxy copolymer resin
sealing gasket and, above that, two more 0.062 mm-thick silicon-
rubber gaskets. The entire assembly was compressed between an
1536
https://dx.doi.org/10.1021/acs.joc.0c02392
J. Org. Chem. 2021, 86, 1528−1539

The Journal of Organic Chemistry
pubs.acs.org/joc
Article
mmol, 1.20 equiv), K2CO3 (1.17 g, 8.43 mmol, 2.00 equiv), and CuI
(80 mg, 0.42 mmol, 0.10 equiv) were combined with t-AmOH (10
mL) and the mixture was degassed with vacuum/argon cycles.
(1R,2R)-(−)-N,N′-dimethylcyclohexane-1,2-diamine L2 (60 mg, 0.42
mmol, 0.10 equiv) was added, and the reaction mixture was heated to
100 °C and left stirring under argon for 5 h. The reaction mixture was
diluted with H2O (30 mL) and extracted with EtOAc (4 × 20 mL).
The combined extracts were dried over Na2SO4, and the solvent was
removed in vacuo. The dark residue was purified by column
chromatography (cyclohexane/EtOAc, 20:80) to give the product 3
as a white solid (0.86 g, 92%). (DMSO-d6, 600 MHz): δ 7.41 (dd,
1H, J = 6.84, 1.90 Hz), 7.30−7.37 (m, 1H), 7.08 (dd, 1H, J = 12.17,
2.28 Hz), 6.84−6.90 (m, 1H), 6.76−6.84 (m, 1H), 6.17 (t, 1H, J =
6.78 Hz), 5.35 (s, 2H), 2.01 (s, 3H). 13C{1H} NMR (DMSO-d6, 150
MHz): δ 161.9, 149.3, 137.0, 136.7, 136.3, 129.2, 128.7, 122.7, 115.2,
113.8, 104.8, 17.0; HRMS (TOF MS ES+) m/z: [M + H]+ Calculated
for C12H12N2OF 219.0934; Found 219.0936.
to 120 °C and left stirring under argon overnight (an 88% conversion
was measured already after 4 h reaction time). The reaction mixture
was diluted with a diluted solution of NaHCO3 (30 mL) and
extracted with EtOAc (4 × 20 mL). The combined extracts were
dried over Na2SO4, and the solvent was removed in vacuo. The dark
residue was purified by column chromatography (cyclohexane/
EtOAc, 95:5 to 90:10) to give the product 10 as a yellow oil (0.78
1
g, 64%). H NMR (DMSO-d6, 400 MHz): δ 7.26−7.33 (m, 2H),
7.16−7.25 (m, 3H), 6.75−6.82 (m, 2H), 6.66−6.74 (m, 2H), 3.64 (s,
3 H), 2.85 (s, 3H). 13C{1H} NMR (DMSO-d6, 100 MHz): δ 151.0,
144.0, 139.2, 128.2, 127.0, 126.6, 114.5, 114.0, 56.5, 55.2, 38.9; MS
(MS ES+) m/z: [M + H]+ 228.
Preparation of N-Benzyl-N-methyl-1H-indol-5-amine 12.8a In a
20 mL reactor, 5-bromo-1H-indole 9 (1.00 g, 5.10 mmol, 1.00 equiv),
N-methybenzylamine (0.989 g, 8.16 mmol, 1.60 equiv), CsF (1.94 g,
12.74 mmol, 2.50 equiv), Cu2O (36 mg, 0.255 mmol, 0.05 equiv), and
DMPAO L19 (98 mg, 0.51 mmol, 0.10 equiv) were combined with t-
AmOH (10 mL) and the mixture was degassed with vacuum/argon
cycles. The reaction mixture was heated to 110 °C and left stirring
under argon 45 h (a 64% assay yield and 87% conversion of the
starting material was measured after this time). The reaction mixture
was diluted with a diluted solution of NaHCO3 (30 mL) and
extracted with EtOAc (4 × 20 mL). The combined extracts were
dried over Na2SO4, and the solvent was removed in vacuo. The dark
residue was purified by column chromatography (cyclohexane/
EtOAc, 100:0 to 90:10) to give the product 12 as a yellowish oil
Preparation of 4-(3-Methyl-2-oxopyridin-1(2H)-yl)benzonitrile
5.9 In a 20 mL reactor, 4-iodobenzonitrile 4 (1.00 g, 4.37 mmol,
1.00 equiv), 3-methyl-pyridin-2(1H)-one 2 (0.571 g, 5.23 mmol, 1.20
equiv), K2CO3 (1.21 g, 8.73 mmol, 2.00 equiv), and CuI (83 mg, 0.44
mmol, 0.10 equiv) were combined with t-AmOH (10 mL) and the
mixture was degassed with vacuum/argon cycles. (1R,2R)-(−)-N,N′-
dimethylcyclohexane-1,2-diamine L2 (62 mg, 0.44 mmol, 0.10 equiv)
was added, and the reaction mixture was heated to 100 °C and left
stirring under argon for 2 h. The reaction mixture was diluted with
H2O (30 mL) and extracted with EtOAc (4 × 20 mL). The combined
extracts were dried over Na2SO4, and the solvent was removed in
vacuo. The yellow residue was purified by column chromatography
(cyclohexane/EtOAc, 80:20 to 20:80) to give the product 5 as a beige
solid (0.68 g, 74%).1H NMR (DMSO-d6, 400 MHz): δ 7.95−8.03
(m, 2H), 7.61−7.70 (m, 2H), 7.47−7.56 (m, 1H), 7.36−7.44 (m,
1H), 6.28 (br t, 1H, J = 6.81 Hz), 2.05 (br s, 3H). 13C{1H} NMR
(DMSO-d6, 100 MHz): δ 161.3, 144.8, 137.6, 135.5, 133.1, 129.2,
128.0, 118.2, 110.7, 105.6, 16.8; IR (ATR) 2231, 1656 cm−1, HRMS
(TOF MS ES+) m/z: [M + H]+ Calculated for C13H11N2O 211.0871;
Found 211.0876.
1
(0.73 g, 60%). H NMR (DMSO-d6, 400 MHz): δ 10.68 (br s, 1H),
7.24−7.31 (m, 2H), 7.21−7.28 (m, 2H), 7.20−7.24 (m, 1H), 7.18−
7.22 (m, 1H), 7.18 (t, 1H, J = 2.7 Hz), 6.85 (d, 1H, J = 2.4 Hz), 6.79
(dd, 1H, J = 8.8, 2.4 Hz), 6.22 (ddd, 1H, J = 3.0, 2.0, 0.9 Hz), 2.86 (s,
3H), 4.44 (s, 2H). 13C{1H} NMR (DMSO-d6, 100 MHz): δ 143.8,
139.5, 130.0, 128.4, 128.1, 127.3, 126.5, 125.0, 111.6, 111.5, 103.6,
100.3, 57.8. MS (MS ES+) m/z: [M]+ 236.
The following two side compounds could also be isolated from the
same reaction mixture upon chromatography (cyclohexane/EtOAc,
100:0 to 90:10): N-benzyl-N-methyl-1′H-[1,5′-biindol]-5-amine 13,
1
beige gum (22 mg, 17% ca. 90% pure). H NMR (DMSO-d6, 400
Applying the literature methodology described in ref9 (CuI/oxine
L16/K2CO3 in DMSO at 130 °C, 2 g scale), the following side
compounds could also be isolated from the reaction mixture after
purification by column chromatography (cyclohexane/EtOAc, 80:20
to 20:80): 4-((3-methylpyridin-2-yl)oxy)benzonitrile 6 white gum
MHz): δ 11.25 (br s, 1H), 7.63 (d, 1H, J = 2.1 Hz), 7.54 (dt, 1H, J =
8.5, 0.7 Hz), 7.44 (d, 1H, J = 3.1 Hz), 7.44−7.46 (m, 1H), 7.33 (d,
1H, J = 9.0 Hz), 7.26−7.31 (m, 2H), 7.23−7.29 (m, 2H), 7.20−7.23
(m, 1H), 7.18−7.24 (m, 1H), 6.93 (d, 1H, J = 2.3 Hz), 6.85 (dd, 1H,
J = 9.0, 2.5 Hz), 6.51 (ddd, 1H, J = 2.9, 2.0, 0.9 Hz.), 6.45 (dd, 1H, J
= 3.1, 0.7 Hz), 4.50 (s, 2H), 2.93 (s, 3H). 13C{1H} NMR (DMSO-d6,
100 MHz): δ 144.3, 139.4, 134.2, 131.5, 129.9, 129.5, 128.9, 128.2,
128.0, 127.2, 126.8, 126.5, 117.8, 115.0, 112.1, 111.6, 110.6, 103.6,
101.6, 101.4, 57.3. HRMS (TOF MS ES+) m/z: [M + H]+ Calculated
for C24H22N3 352.1814; Found 352.1815. N1-(2,6-dimethylphenyl)-
N2-(1H-indol-5-yl)oxalamide 14, reddish gum (12 mg, 1%) at 303 K
1
(320 mg, 17%). H NMR (DMSO-d6, 400 MHz): δ 7.99−8.03 (m,
1H), 7.84−7.90 (m, 2H), 7.76−7.80 (m, 1H), 7.25−7.31 (m, 2H),
7.15 (br dd, 1H, J = 7.28, 4.87 Hz), 2.29 (br s, 3 H). 13C{1H} NMR
(DMSO-d6, 100 MHz): 160.1, 158.1, 144.7, 140.7, 134.0, 121.9,
121.4, 120.2, 118.6, 106.4, 15.2; IR (ATR) 2227 cm−1, HRMS (TOF
MS ES+) m/z: [M + H]+ Calculated for C13H11N2O 211.0871; Found
211.0877. 4-(3-methyl-2-oxopyridin-1(2H)-yl)benzamide 7 white
1
as a 6:4 mixture of rotamers. H NMR (400 MHz, DMSO-d6): δ
1
10.99−11.16 (br s + br s, 1H), 7.25−7.51 (m, 4.9H), 7.18−7.25 (m,
1.1H), 6.99−7.15 (m, 3.6H), 6.95 (dd, 0.6H, J = 8.7, 2.0 Hz), 6.63
(br s, 1H), 2.26 (br s, 2.4H), 2.13 (br s, 3.6H). 13C{1H} NMR
(DMSO-d6, 100 MHz): δ 169.7/169.1*, 167.3/166.9*, 140.7/140.4*,
137.9*/136.2, 133.5/132.4*, 133.5/133.0*, 128.1/128.1*, 127.2/
127.1*, 127.5*/126.8, 126.0/125.8*, 118.2/118.2*, 115.2/114.5*,
110.9/110.6*, 101.1/101.1*, 18.4*/17.9 (* minor rotamer). HRMS
(TOF MS ES+) m/z: [M + H]+ Calculated for C18H17N2O3
309.1239; Found 309.1240.
gum (253 mg, 12%). H NMR (DMSO-d6, 400 MHz): δ 8.05 (br
s, 1H), 7.93−8.02 (m, 2H), 7.35−7.55 (m, 5H), 6.26 (br t, 1H, J =
6.78 Hz), 2.04 (br s, 3H). 13C{1H} NMR (DMSO-d6, 100 MHz): δ
167.1, 161.5, 143.3, 137.4, 136.0, 133.7, 129.1, 128.1, 126.6, 105.3,
16.9; IR (ATR) 2227 cm−1, HRMS (TOF MS ES+) m/z: [M + H]+
Calculated for C13H11N2O 211.0871; Found 211.0877. 4-((3-
methylpyridin-2-yl)oxy)benzamide 8, beige gum (18 mg, 1%). 1H
NMR (DMSO-d6, 400 MHz): δ 7.97 (dd, 1H, J = 4.8, 1.5 Hz), 7.93
(br s, 1H), 7.90 (m, 2H), 7.74 (ddq, 1H, J = 7.3, 1.5, 0.7, 0.7 Hz),
7.29 (br s, 1H), 7.13 (m, 2H), 7.09 (dd, 1H, J = 7.3, 4.8 Hz), 2.30 (s,
3H). 13C{1H} NMR (DMSO-d6, 100 MHz): δ 167.2, 160.7, 156.7,
144.5, 140.3, 129.9, 121.5, 120.2, 119.6, 15.3; HRMS (TOF MS ES+)
m/z: [M + H]+ Calculated for C13H12N2O2 229.0977; Found
229.0972.
ASSOCIATED CONTENT
* Supporting Information
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.joc.0c02392.
■
sı
Preparation of N-Benzyl-4-methoxy-N-methylaniline 10.8a In a
20 mL reactor, 4-bromoanisole 9 (1.00 g, 5.34 mmol, 1.00 equiv), N-
methybenzylamine (0.907 g, 7.48 mmol, 1.40 equiv), CsF (2.44 g,
16.04 mmol, 3.00 equiv), CuI (102 mg, 0.53 mmol, 0.10 equiv), and
DMPAO L19 (103 mg, 0.53 mmol, 0.10 equiv) were combined with a
1:1 mixture of DMSO/t-AmOH (10 mL) and the mixture was
degassed with vacuum/argon cycles. The reaction mixture was heated
Plate-to-plate reproducibility results, DoE optimization
details, copies of NMR spectra for all isolated
compounds (PDF)
Visual Basic script used for data extraction from HPLC
chromatograms (TXT)
1537
https://dx.doi.org/10.1021/acs.joc.0c02392
J. Org. Chem. 2021, 86, 1528−1539

The Journal of Organic Chemistry
pubs.acs.org/joc
Article
(5) (a) Robbins, D. W.; Hartwig, J. F. A simple, multidimensional
approach to high-throughput discovery of catalytic reactions. Science
2011, 333, 1423−1427. (b) Troshin, K.; Hartwig, J. F. Snap
deconvolution: An informatics approach to high-throughput discovery
of catalytic reactions. Science 2017, 357, 175−181.
AUTHOR INFORMATION
Corresponding Author
Marco Santagostino − Chemical Process Development,
Boehringer Ingelheim Pharma GmbH & Co. KG, Biberach
an der Riß 88397, Germany; orcid.org/0000-0002-7724-
1641; Phone: +49 (7351) 54-7746;
■
(6) (a) Wieland, J.; Breit, B. A combinatorial approach to the
identification of self-assembled ligands for rhodium-catalysed
asymmetric hydrogenation. Nat. Chem. 2010, 2, 832−837.
(b) Wolf, E.; Richmond, E.; Moran, J. Identifying lead hits in catalyst
discovery by screening and deconvoluting complex mixtures of
catalyst components. Chem. Sci. 2015, 6, 2501−2505. (c) Richmond,
E.; Moran, J. Harnessing Complex Mixtures for Catalyst Discovery.
Synlett 2016, 27, 2637−2643.
Email: marco.santagostino@boehringer-ingelheim.com
Authors
Raphael R. Steimbach − Chemical Process Development,
Boehringer Ingelheim Pharma GmbH & Co. KG, Biberach
an der Riß 88397, Germany
Philipp Kollmus − Chemical Process Development, Boehringer
Ingelheim Pharma GmbH & Co. KG, Biberach an der Riß
88397, Germany
(7) Beletskaya, I. P.; Cheprakov, A. V. The Complementary
Competitors: Palladium and Copper in C−N Cross-Coupling
Reactions. Organometallics 2012, 31, 7753−7808.
(8) (a) Zhang, Y.; Yang, X.; Yao, Q.; Ma, D. CuI/DMPAO-catalyzed
N-arylation of acyclic secondary amines. Org. Lett. 2012, 14, 3056−
3059. (b) Zhou, W.; Fan, M.; Yin, J.; Jiang, Y.; Ma, D. CuI/Oxalic
Diamide Catalyzed Coupling Reaction of (Hetero)Aryl Chlorides and
Amines. J. Am. Chem. Soc. 2015, 137, 11942−11945. (c) Fan, M.;
Zhou, W.; Jiang, Y.; Ma, D. CuI/Oxalamide Catalyzed Couplings of
(Hetero)aryl Chlorides and Phenols for Diaryl Ether Formation.
Angew. Chem., Int. Ed. Engl. 2016, 55, 6211−6215. (d) Gao, J.;
Bhunia, S.; Wang, K.; Gan, L.; Xia, S.; Ma, D. Discovery of N-
(Naphthalen-1-yl)-N’-alkyl Oxalamide Ligands Enables Cu-Catalyzed
Aryl Amination with High Turnovers. Org. Lett. 2017, 19, 2809−
2812.
Complete contact information is available at:
https://pubs.acs.org/10.1021/acs.joc.0c02392
Author Contributions
The manuscript was written through contributions of all
authors. All authors have given approval to the final version of
the manuscript.
Notes
The authors declare no competing financial interest.
(9) Filipski, K. J.; Kohrt, J. T.; Casimiro-Garcia, A.; van Huis, C. A.;
Dudley, D. A.; Cody, W. L.; Bigge, C. F.; Desiraju, S.; Sun, S.; Maiti,
S. N.; Jaber, M. R.; Edmunds, J. J. A versatile copper-catalyzed
coupling reaction of pyridin-2(1H)-ones with aryl halides. Tetrahe-
dron Lett. 2006, 47, 7677−7680.
(10) Despite DMSO offering often a boost in reactivity in copper-
catalyzed cross-couplings, possibly due to its optimal solvating
properties and its high Brønsted basicity, this solvent should generally
to be regarded with caution and thorough safety investigations are
necessary before any significant upscaling using this solvent. DMSO is
in fact known to decompose highly exothermically near its boiling
point and the onset temperature of decomposition can be reduced
significantly in reaction mixtures in the presence of contaminants such
as strong acids, or bases, or halogen anions. See Yang, Q.; Sheng, M.;
ACKNOWLEDGMENTS
■
We acknowledge Dr. J. Fordham for reading and formally
improving the content of this manuscript. We thank
Boehringer Ingelheim Pharma GmbH & Co. KG for funding
a 6 month internship of R.S.
REFERENCES
■
(1) (a) Shevlin, M. Practical High-Throughput Experimentation for
Chemists. ACS Med. Chem. Lett. 2017, 8, 601−607. (b) Schmink, J.
R.; Bellomo, A.; Berrit, S. Scientist-Led High-Throughput Exper-
imentation (HTE) and Its Utility in Academia and Industry.
Aldrichimica Acta 2013, 46, 71−80. (c) Mennen, S. M.; Alhambra,
C.; Allen, C. L.; Barberis, M.; Berritt, S.; Brandt, T. A.; Campbell, A.
́
́
Li, X.; Tucker, C.; Vasquez Cespedes, S.; Webb, N. J.; Whiteker, G.
T.; Yu, J. Potential Explosion Hazards Associated with the
Autocatalytic Thermal Decomposition of Dimethyl Sulfoxide and Its
Mixtures. Org. Process Res. Dev. 2020, 24, 916−939. and references
therein
́
D.; Castanon, J.; Cherney, A. H.; Christensen, M.; Damon, D. B.;
̃
Eugenio de Diego, J.; García-Cerrada, S.; García-Losada, P.; Haro, R.;
Janey, J.; Leitch, D. C.; Li, L.; Liu, F.; Lobben, P. C.; MacMillan, D.
W. C.; Magano, J.; McInturff, E.; Monfette, S.; Post, R. J.; Schultz, D.;
Sitter, B. J.; Stevens, J. M.; Strambeanu, I. I.; Twilton, J.; Wang, K.;
Zajac, M. A. The Evolution of High-Throughput Experimentation in
Pharmaceutical Development and Perspectives on the Future. Org.
Process Res. Dev. 2019, 23, 1213−1242.
(11) The mixture of Cu2O and 3,4,7,8-tetramethyl-1,10-phenanthro-
line in DMSO delivered in the first deconvolution run an almost
quantitative cross-coupling assay yield. The price of the ligand is
however considerably high at ca. 75€/g.
(2) Buitrago Santanilla, A.; Regalado, E. L.; Pereira, T.; Shevlin, M.;
Bateman, K.; Campeau, L.-C.; Schneeweis, J.; Berritt, S.; Shi, Z.-C.;
Nantermet, P.; Liu, Y.; Helmy, R.; Welch, C. J.; Vachal, P.; Davies, I.
W.; Cernak, T.; Dreher, S. D. Nanomole-scale high-throughput
chemistry for the synthesis of complex molecules. Science 2015, 347,
49−53.
(3) (a) Perera, D.; Tucker, J. W.; Brahmbhatt, S.; Helal, C. J.;
Chong, A.; Farrell, W.; Richardson, P.; Sach, N. W. A platform for
automated nanomole-scale reaction screening and micromole-scale
synthesis in flow. Science 2018, 359, 429−434. (b) Schneider, G.
Automating drug discovery. Nat. Rev. Drug Discovery 2018, 17, 97−
113.
(4) For a critical review of limits and opportunities of uHT reaction
screening approaches, see: Lin, S.; Dikler, S.; Blincoe, W. D.;
Ferguson, R. D.; Sheridan, R. P.; Peng, Z.; Conway, D. V.; Zawatzky,
K.; Wang, H.; Cernak, T.; Davies, I. W.; DiRocco, D. A.; Sheng, H.;
Welch, C. J.; Dreher, S. D. Mapping the dark space of chemical
reactions with extended nanomole synthesis and MALDI-TOF MS.
Science 2018, 361, eaar6236.
(12) A fractional factorial design in four factors (24‑1) was applied at
this stage. The continuous factors taken into consideration were the
amount of ligand DMPAO [−1 = 5%/+1 = 15%], amount of N-
benzylmethylamine [−1 = 1.2 equiv/+1 = 1.6 equiv], amount of CsF
[−1 = 1.0 equiv/+1 = 3.0 equiv], and amount of added water [−1 = 0
mol %/+1 = 2 mol %]. The following factors were not varied: CuI
[10%]; DMSO/t-AmOH, 1:1 [10 vol]; and reaction temperature
[120°C] (Supporting Information).
(13) A rotatable (α = 1.414) Central Composite design in three
factors was applied. The continuous factors taken into consideration
were the amount of ligand DMPAO [−1 = 5%/+1 = 15%], amount of
N-benzylmethylamine [−1 = 1.2 equiv/+1 = 1.6 equiv], and amount
of CsF [−1 = 1.0 equiv/+1 = 3.0 equiv]. The following factors were
not varied: Cu2O [5%]; t-AmOH, 1:1 [10 vol]; reaction temperature
[110°C]; and reaction time [24 h] (Supporting Information).
(14) Martin, M. C.; Goshu, G. M.; Hartnell, J. R.; Morris, C. D.;
Wang, Y.; Tu, N. P. Versatile Methods to Dispense Submilligram
Quantities of Solids Using Chemical-Coated Beads for High-
1538
https://dx.doi.org/10.1021/acs.joc.0c02392
J. Org. Chem. 2021, 86, 1528−1539

The Journal of Organic Chemistry
pubs.acs.org/joc
Article
Throughput Experimentation. Org. Process Res. Dev. 2019, 23, 1900−
1907.
1539
https://dx.doi.org/10.1021/acs.joc.0c02392
J. Org. Chem. 2021, 86, 1528−1539