Efficient Screening for Ternary Molecular Ionic Cocrystals Using a Complementary Mechanosynthesis and Computational Structure Prediction Approach

Article abstract of DOI:10.1002/chem.201904672

The discovery of molecular ionic cocrystals (ICCs) of active pharmaceutical ingredients (APIs) widens the opportunities for optimizing the physicochemical properties of APIs whilst facilitating the delivery of multiple therapeutic agents. However, ICCs are often observed serendipitously in crystallization screens and the factors dictating their crystallization are poorly understood. We demonstrate here that mechanochemical ball milling is a versatile technique for the reproducible synthesis of ternary molecular ICCs in less than 30 min of grinding with or without solvent. Computational crystal structure prediction (CSP) calculations have been performed on ternary molecular ICCs for the first time and the observed crystal structures of all the ICCs were correctly predicted. Periodic dispersion-corrected DFT calculations revealed that all the ICCs are thermodynamically stable (mean stabilization energy=?2 kJ mol^?1) relative to the crystallization of a physical mixture of the binary salt and acid. The results suggest that a combined mechanosynthesis and CSP approach could be used to target the synthesis of higher-order molecular ICCs with functional properties.

Full text of DOI:10.1002/chem.201904672

A Journal of
Accepted Article
Title: Efficient screening for ternary molecular ionic cocrystals using a
complementary mechanosynthesis and computational structure
prediction approach
Authors: Abeer F. Shunnar, Bhausaheb Dhokale, Durga Prasad
Karothu, David H. Bowskill, Isaac J. Sugden, Hector H.
Hernandez, Panče Naumov, and Sharmarke Mohamed
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Chem. Eur. J. 10.1002/chem.201904672
Link to VoR: http://dx.doi.org/10.1002/chem.201904672
Supported by

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
Efficient screening for ternary molecular ionic cocrystals using a
complementary mechanosynthesis and computational structure
prediction approach
[a]
[a]
[b]
[c]
Abeer F. Shunnar, Bhausaheb Dhokale, Durga Prasad Karothu, David H. Bowskill, Isaac J.
[
c]
[d]
[b]
[a]
Sugden, Hector H. Hernandez, Panče Naumov and Sharmarke Mohamed*
solid forms[2] using the principles of crystal engineering, often
aided by computational insight.[ Despite this fact, it remains
impossible to be certain that all crystal forms have been
discovered in a solid form screen since there are always more
variables to investigate than the time and resources available.[
Nevertheless, the solid form screen is considered a success if it
culminates in the discovery of one or more crystal forms with
desirable properties (e.g. solubility, stability or bioavailability).
Within the class of multicomponent crystal forms, binary
cocrystals comprising two molecular components have received
significant attention[6] over the past two decades. However, recent
reports of the advantages[7] offered by ionic cocrystals (ICCs)
derived from the cocrystallization of APIs with atomic inorganic
[3]
Abstract: The discovery of molecular ionic cocrystals (ICCs) of active
pharmaceutical ingredients (APIs) widens the opportunities for
optimizing the physicochemical properties of the API whilst facilitating
the delivery of multiple therapeutic agents. However, ICCs are often
observed serendipitously in crystallization screens and the factors
dictating their crystallization are poorly understood. We demonstrate
that mechanochemical ball-milling is a versatile technique for the
reproducible synthesis of ternary molecular ICCs in less than 30
minutes of grinding with or without solvent. Computational crystal
structure prediction (CSP) calculations were performed on ternary
molecular ICCs for the first time and the observed crystal structures
of all ICCs were correctly predicted. Periodic DFT-D calculations
reveal that all ICCs are thermodynamically stable (mean stabilization
4]
5]
-
1
energy: -2 kJ mol ) relative to the crystallization of a physical mixture
of the binary salt and acid. The results suggest that a combined
mechanosynthesis and CSP approach could be used to target the
synthesis of higher-order molecular ICCs with functional properties.
2
salts (e.g. NaCl and CaCl ), suggest that the range of practically
accessible solid forms for optimizing the properties of APIs,
continues to widen.[8] This is in the context of recent reports of the
synthesis of a quaternary cocrystal[9] and the first reported
synthesis of a six-component molecular solid.[
10]
Here, we focus on the systematic crystallization of ternary
−
+
molecular ICCs that conform to the empirical formula A BH CH,
Introduction
−
+
where A and BH are molecular ions derived from acid (AH)-
base (B) proton transfer and CH is an acidic molecular coformer.
We define molecular ICCs as solid forms that: 1) are derived from
the cocrystallization of organic molecular species (e.g. AH, B and
CH) that are all solids under room temperature and pressure
conditions; 2) are sustained by charge-assisted hydrogen
bonding interactions that arise as a consequence of one or more
proton transfer events. These requirements provide a contrast
with ICCs derived from organic molecules and atomic inorganic
salts (i.e. the so called “organic-inorganic ICCs”)[ since such
systems do not arise from proton transfer events and display
structures that are largely distinguished on the basis of the
coordination patterns observed around a metal ion. The above
framework also allows us to draw a sharp distinction between
molecular ICCs and solvates[12], which unlike ICCs are derived
following inclusion of one or more volatile liquids. The solvent
molecules can often be removed in non-stoichiometric solvates
without affecting the structural integrity of the resulting desolvate.
By contrast, molecular ICCs are defined by structures featuring
three or more molecular species that are all integral to the crystal
packing and display a single sharp melting point. Since it is
The ability to select the optimal crystal form of an active
pharmaceutical ingredient (API) has important economic,
regulatory and clinical consequences. The past three decades
has seen significant progress in our ability to synthesize complex
[
1]
[
a]
A. F. Shunnar, Dr. B. Dhokale, Dr. S. Mohamed*
Department of Chemistry
11]
Khalifa University of Science and Technology
PO Box 127788, Abu Dhabi (UAE)
E-mail: sharmarke.mohamed@ku.ac.ae
Dr. D. P. Karothu, Prof. P. Naumov
New York University Abu Dhabi
[
b]
c]
PO Box 129188, Abu Dhabi (UAE)
D. H. Bowskill, Dr. I. J. Sugden
[
Molecular Systems Engineering Group
Center for Process Systems Engineering
Department of Chemical Engineering
Imperial College London,
London, SW7 2AZ (UK)
[d]
Dr. H. H. Hernandez
Department of Biomedical Engineering
Center for Membrane and Advanced Water Technology
Khalifa University of Science and Technology
Masdar Campus
−
+
possible to cocrystallize a molecular salt (A ∙ BH ) with a variety
of acidic coformers (CH), we classify ternary molecular ICCs into
two categories based on the chemical structures of AH and CH:
conjugate acid-base ICCs (denoted “CAB-ICCs”)[13] and non-
conjugate acid-base ICCs (denoted “NCAB-ICCs”). CAB-ICCs
PO Box 127788, Abu Dhabi (UAE)
Supporting information for this article is given via a link at the end of
the document.
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
assisted grinding (OP-LAG) synthetic protocol (Scheme 1). The
S-LAG approach proceeds via the mechanosynthesis of the 1:1
binary salt complex followed by a subsequent step involving
stoichiometric amounts of this salt and an acid coformer. By
contrast the OP-LAG approach proceeds via the direct
cocrystallization of the ICC using a 2:1 (AH:B; CAB-ICC) or 1:1:1
(
AH:CH:B; NCAB-ICC) stoichiometric ratio of the reactants.
Initially, the mechanosynthesis of CAB-ICCs and NCAB-ICCs
reported exclusively from solution crystallization screens[14, 18] are
targeted as a validation of our proposed mechanochemical rapid
screening protocol. This is followed by mechanosynthesis
experiments targeting a novel CAB-ICC of 2-chlorobenzoic acid
and 4-dimethylaminopyridine using a coformer replacement
strategy based on molecular size-matching. The thermal
stabilities of the ICCs are compared as we screen various
combinations of AH/CH pairs. We also assess the role that
solvent plays in driving the self-assembly of molecular ICCs
during the mechanosynthesis experiment. CAB-ICCs and NCAB-
ICCs are synthesized (Scheme 1) via the cocrystallization of 4-
dimethylaminopyridine (4-DMAP) with 2-chlorobenzoic acid (2-
CLBZAH), 4-chlorobenzoic acid (4-CLBZAH), 2-hydroxybenzoic
acid (2-HBZAH) or 4-hydroxybenzoic acid (4-HBZAH).
Despite the proven value of computational crystal structure
prediction (CSP) methods in facilitating the discovery of
previously unknown solid forms,[4c, 19] molecular solids comprising
three or more chemical entities are not routinely studied in such
work. This reflects the demands on computing resources and the
complexity of the search space, which scales exponentially[ with
the degrees of freedom. Indeed, the chloride salt hydrate (target
XXIV) proposed in the sixth blind test[21] of CSP, had the lowest
number of attempted predictions by participants of the blind test
and only one group was able to generate the experimental
structure in their list of predicted structures. Here, we report on
the CSP of ternary molecular ICCs for the first time. Initially, we
test the success of the computational model in reproducing the
crystal structures of known CAB-ICCs and NCAB-ICCs displaying
rigid molecular conformations. This is followed by a more
extensive blind[21] CSP study on two conformationally flexible
systems: a binary salt and a ternary CAB-ICC derived from 2-
CLBZAH and 4-DMAP. We use the predicted crystal form
landscapes (CFLs) to assess the preferred packing modes of
ternary molecular ICCs and the degree to which CSP methods
can guide the selection of ICCs in crystallization screens. We also
test the hypothesis that the crystallization of ternary molecular
ICCs is under thermodynamic control[22] and that the use of
dispersion-corrected density functional theory (DFT-D) energies
is diagnostic enough to guide the experimental discovery of
ternary molecular ICCs.
20]
Scheme 1: (a) Molecular structures of the base (B=4-dimethylaminopyridine; 4-DMAP)
and acid (AH) coformers used to target ternary molecular ionic cocrystals (ICCs) with
-
+
the empirical formula, A BH CH. AH is 2-chorobenzoic acid (2-CLBZAH), 4-
chlorobenzoic acid (4-CLBZAH), 2-hydroxybenzoic acid (2-HBZAH) or 4-hydroxybenzoic
acid (4-HBZAH). The species CH is one of the four acid coformers shown. (b) Reaction
schemes for the one-pot (OP-LAG) and serial (S-LAG) liquid-assisted grinding routes for
2
2
the synthesis of ternary molecular ICCs. (c) Illustration of the common 퐷 (5) hydrogen
bond heterosynthon observed in most of the ICCs studied in this work.
are derived from experiments where AH = CH such that A− and
CH are conjugate acid-base pairs. By contrast, NCAB-ICCs are
−
derived from experiments where AH ≠ CH, such that A and CH
are not conjugate acid-base pairs. NCAB-ICCs potentially offer
more scope[ for tuning the solid-state properties of ionizable
APIs due to the unrestricted number of AH/CH pairs that can be
used to target such solid forms.
14]
Mechanosynthesis is increasingly being recognized as an
efficient and environmentally friendly methodology for a range of
chemical transformations.[15] Recent efforts that aim to synthesize
ICCs[7a, 7c, 7e, 7f] have focused on the cocrystallization of APIs with
metal halides and there have been some documented successes
in the mechanosynthesis of such organic-inorganic ICCs.[16]
However, the mechanosynthesis of ternary molecular ICCs Results and Discussion
remains an underexplored area of research. Here, we
demonstrate the potential for using mechanochemical ball milling
Mechanosynthesis and solid-state properties of ICCs
(
under neat grinding (NG) or liquid-assisted grinding (LAG)
Solution crystallization of ternary molecular ICCs is complicated
by the differences in the solubilities of the reactants, which if
significant can lead to undesired products or a physical mixture of
the starting reactants. This is reflected in the serendipitous
conditions)[ for the facile screening of CAB-ICCs and NCAB-
ICCs. LAG mechanosynthesis of ICCs is explored using a serial
two-step liquid-assisted grinding (S-LAG) or a one-pot liquid-
17]
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
Figure 1: Chemical composition, numbering system and significant hydrogen bonding interactions (colored cyan) for all solid forms synthesized in this work. The atoms are colored
by element for all solid forms expect for 1a, where the species in the asymmetric unit is colored by symmetry equivalence. Images of the bulk powder sample obtained following
mechanosynthesis of each target solid form is shown where applicable. Cambridge Structural Database (CSD)[23] reference codes for ICCs (3‒6) previously reported[14, 18] following
solution crystallization screens are indicated in capital letters. * The solvent is either MeOH or i-PrOH for this step.
solution crystallization of ternary molecular ICCs in experiments
targeting the binary salt using a 1:1 stoichiometric ratio of the acid
and base.[24] By contrast, mechanosynthesis (Scheme 1) of ICCs
is attractive in minimizing (OP-LAG) or completely removing (NG)
the role of the solvent during the synthesis of the ICC. We set out
to test the potential for rapid mechanochemical screening of ICCs
using 4-DMAP and the set of acid coformers highlighted in
Scheme 1. The diversity of solid forms obtained from our
mechanosynthesis screens are illustrated in Figure 1. A salt
anhydrate (1), two salt hydrates (1a, 1b) and a CAB-ICC (2) were
discovered in mechanosynthesis experiments involving 2-
CLBZAH and 4-DMAP. The ICCs 3‒6 (Figure 1) previously
reported[14, 18] from solution crystallization screens were all shown
to be amenable to mechanosynthesis (see SI, Figures S1-S4).
OP-LAG experiments targeting the CAB-ICC with the composition
KUJDIE),[18] generating the same target solid form under neat
grinding (NG), S-LAG or OP-LAG conditions. In most cases, the
resulting products following NG or OP-LAG experiments were the
same. However, the effect of solvent was found to be significant
for experiments involving 4-HBZAH. For OP-LAG experiments
involving a 2:1 stoichiometric ratio of 4-HBZAH and 4-DMAP, the
NG product (7-II) was characterized to be Form II of the reported
CUKNUT structure whilst the OP-LAG (MeOH solvent) product
did not match CUKNUT and presented challenges in cell indexing
due to the poor crystallinity of the sample. This apparent failure in
the mechanosynthesis of 7-I using MeOH is in the context of this
being the reported[ solvent of choice for growing single crystals
of 7-I. In most cases, OP-LAG experiments using a protic solvent
led to detectable improvements in the crystallinity and phase
purity of the ICC as monitored via PXRD measurements. For all
target ICCs (2‒7), the solid form could be synthesized in as little
as 15 minutes of grinding but favorable conversion from reactants
to the product was detected at 30 minutes of grinding using the
OP-LAG approach. All ICCs appeared to be stable to
amorphization even after grinding for 60 minutes.
14]
-
+
4
-HBZA ∙4-DMAPH ∙4-HBZAH (CSD Refcode: CUKNUT; Form I,
hereafter 7-I) did not yield a solid form matching the reported[
CUKNUT structure. Instead a polymorph of this CAB-ICC (Form
II; hereafter 7-II) was isolated under NG conditions.
Mechanosynthesis of ICCs was shown to be feasible with or
14]
without addition of a polar solvent (MeOH, i-PrOH or H
2
−ꢀ
O; using
Molecular ICCs display supramolecular features typically
a solvent/solute ratio (휂)[25] of approximately 0.1 μL mg for OP-
LAG experiments). ICCs were also shown to be amenable to
mechanosynthesis via the serial (S-LAG) or one-pot (OP-LAG)
liquid-assisted grinding routes (Scheme 1). Figure 2 illustrates the
success in the mechanosynthesis of CAB-ICC 4 (CSD Refcode:
associated with salts as they comprise formally charged ions (A
−
+
and BH ) that engage in charge-assisted hydrogen bonding
interactions. However, ICCs are also a type of cocrystal since they
−
+
are derived from the cocrystallization of a binary salt (A BH ) with
another species (e.g. CH). These observations suggest the need
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
Table 1: Properties of the ICCs obtained following mechanosynthesis
[
a]
∆푽퐯퐝퐖[a]/Åퟑ
[b]/%
ICC
ICC Composition
ICC
Type
ICC H-bond
Graph Set
∆퐩푲
퐚
∆푪
퐤
Melting Point
Onset/°C
Melting
Point
Peak/°C
Predicted Energetic
ICC Stabilization
Rank on CFL[
c]
Energy (∆푬퐈퐂퐂
/
Number
풌푱 풎풐풍− )[d]
ퟏ
-
+
ꢀ(ꢁ)
2(5)
2(5)
2(5)
2(5)
ꢀ(ꢁ)
2
3
4
5
6
2-CLBZA ∙4-DMAPH ∙2-CLBZAH
CAB
CAB
퐷
ꢀ
5.71
4.71
5.99
5.99
5.99
4.40
5.14 (2.41)
5.33 (2.46)
10.68 (2.41)
10.68 (7.81)
10.68 (2.33)
10.76 (2.41)
0
91.85
152.09
108.00
148.10
147.78
162.49
94.96
156.43
110.47
150.72
151.21
164.77
5
1
-0.11
-1.45
-0.69
-0.63
-4.86
-4.42
-
+
4-CLBZA ∙4-DMAPH ∙4-CLBZAH
퐷
2
-0.2
0.6
1.2
1.1
2.8
-
+
2-HBZA ∙4-DMAPH ∙2-HBZAH
CAB
퐷
2
6
-
+
2-HBZA ∙4-DMAPH ∙4-CLBZAH
NCAB
NCAB
CAB
퐷
2
2
-
+
2-HBZA ∙4-DMAPH ∙4-HBZAH
퐷
2
21
-
-
+
7-II
4-HBZA ∙4-DMAPH ∙4-HBZAH
퐷
ꢀ
[
a]
These properties were calculated using the calculator plug-ins in MarvinSketch.[32] The ∆푉vdW column lists the absolute difference in the calculated[32] van der
Waals volumes for the cation and anion (and in brackets for the acid and anion) in each ICC. [b] The Kitaigorodsky packing coefficients (퐶k) were calculated using
PLATON[33]. ∆퐶k = 퐶k(ICC) ꢂ 퐶k(sꢃlt). For each ICC, the salt crystal structures were retrieved from the CSD by searching for structures comprising the molecular
ions found in the ICC (see SI Table S3 for CSD refcodes). [c] The energetic rank indicates the relative stability of the experimental ICC on the computed crystal form
landscape (CFL). The lowest possible rank is 1 which corresponds to the global minimum in lattice energy. The energetic ranking for 7-II could not be computed
since the newly determined Form II structure (7-II) was outside the scope of the rigid-body searches for hypothetical crystal structures. [d] The ICC stabilization
energy (∆퐸퐼ꢄꢄ ) relates to the DFT-D estimate of the lattice energy of each ICC relative to the energies of the binary salt and acid (see Equation 2 and the
accompanying discussion in the main text).
for a nuanced perspective when attempting to classify ICCs as
their properties do not conform to the seemingly binary choice
between a salt or a cocrystal.[6c, 26] Some attempts have already
been made to classify[27] multicomponent solid forms in the
context of the wider set of possible crystal forms. However,
variations in classification continue to exist with some reports
[28]
[27b]
describing ICCs as “salt cocrystals” , “cocrystal salts”
or
even “acid solvates”[29]. Despite the lack of a widely accepted
classification scheme for such solid forms, what is clear is that the
successful synthesis of ternary or higher-order ICCs offers
opportunities[
11]
for optimizing the solid-state properties of
ionizable or non-ionizable molecules. All the ICCs shown in Figure
1
1
(2‒7) are derived from acid-base pairs with ∆p퐾 values (Table
) in the range 4.40‒5.99. For NCAB-ICCs 5 and 6, the more
a
acidic component (2-HBZAH) ionizes in the process of
assembling the ternary ICC, in agreement with chemical intuition.
However, in the context of the “salt-cocrystal continuum”[30], ICCs
defy the predictive power of existing empirical rules (i.e. “∆p퐾a
rule”) for targeting neutral or charged complexes of molecules
since they comprise neutral and ionized forms of the same
molecule(s).[31]
Whilst supramolecular synthons[3] are valuable tools for
understanding the most significant intermolecular interactions
driving the assembly of binary cocrystals,[34] the complicated
interplay of competing weak intermolecular interactions in higher-
order crystals, makes it difficult to predict the dominant
intermolecular interactions in such systems. For cocrystals
displaying competing hydrogen bond donor and acceptor sites on
the constituent molecules, Hunter has proposed[35] a method for
estimating the likelihood of cocrystallization on the basis of the
differences in the calculated interaction site pairing energies for
the cocrystal and the constituent molecules. Such interaction site
pairing energies can be derived from the molecular electrostatic
potentials (MEPs) of the molecules that make up the cocrystal.
This approach has been used to successfully predict not only
whether a cocrystal could form[35b] but also the dominant
[
36]
supramolecular synthons in cocrystals.
For all ICCs in our
Figure 2: (a) Overlay of the experimental room-temperature PXRD patterns for
the neat grinding (NG), one-pot (OP-LAG, MeOH) and serial (S-LAG, MeOH)
liquid-assisted grinding products in experiments targeting CAB-ICC 4 (Refcode:
KUJDIE). PXRD patterns for 4-DMAP (Refcode: BUKJOG11), 2-HBZAH
-
+
series (2‒7), the R COO ∙∙∙H- NR hydrogen bond is observed
1
2
between the molecular ions and this is expected given the strong
coulombic forces that stabilize this donor-acceptor pairing. The
MEPs for the acid coformers (Figure 3) used in our
mechanosynthesis screens were calculated (Figure 3) in an
attempt to quantify the relative strengths of the remaining set of
possible hydrogen bond donor-acceptor pairings. For all ICCs in
(Refcode: SALIAC16) and KUJDIE are simulated from the reported single
crystal structures. Diagnostic Bragg reflections for the targeted CAB-ICC are
2
2
indicated in asterix (*). (b) Crystal packing for 4, indicating the 퐷 (5) synthon
between the anion (blue), cation (red) and acid (green) in the ICC structure.
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
-
1
Figure 3: (a) Molecular electrostatic potential (MEP) surfaces for 2-CLBZAH, 4-CLBZAH, 2-HBZAH and 4-HBZAH (potential values in kJ mol ). Only the potentials
for hydrogen bond donor (blue) and acceptor (red) groups capable of engaging in intermolecular hydrogen bonding are indicated. (b) Comparison of the hydrogen
bonding interactions (colored cyan) in CAB-ICC 3 (4-CLBZA ∙ 4-DMAPH+ ∙ 4-CLBZAH) with those observed in NCAB-ICC 6 (2-HBZA ∙ 4-DMAPH+ ∙ 4-HBZAH). In
-
-
the packing diagrams for both ICCs the anion (blue), cation (red) and acid (green) are colored by symmetry equivalence.
our series, the carboxylic acid OH donor of the coformer engages
in discrete hydrogen bonding interactions with one of the
carboxylate oxygens of the anion. This is expected for ICCs such
as 3 (Figure 3) where the acid coformer only bears one hydrogen
bond donor (Figure 3). However for ICCs such as 6 that comprise
the 4-HBZAH coformer, the acid now possesses two donor sites
capable of engaging in intermolecular hydrogen bonding with the
molecular ions. The calculated MEP surface for 4-HBZAH shows
significance of this observation. However, the consistency in the
differences in the intrinsic molecular volumes across the series
(Table 1) is worthy of further investigation as it may point towards
the necessary conditions for favorable coformer exchange
leading to NCAB-ICCs such as 5 and 6.
Comparison of the Kitaigorodsky packing coefficients (퐶 )[
37]
k
of the molecular ICCs in our series with the packing coefficients
of their respective molecular salts as reported in the CSD (see SI
Table S3 for refcodes), show that only 5, 6 and 7-II display > 1%
(
Figure 3) that the carboxylic acid OH donor has a potential
energy that is approximately 45 kJ mol-1 less than that of the para-
substituted OH donor, implying the latter is a better hydrogen
bond donor. However, the observed hydrogen bonding
interactions in 6 (Figure 3) show that the carboxylic acid OH donor
interacts with the carboxylate oxygen acceptor. This suggests that
whilst the principle of the strongest hydrogen bond donor
interacting with the strongest hydrogen bond acceptor may be
sufficient in predicting binary cocrystal formation,[35b] the dominant
role of electrostatics in stabilizing ICCs means that MEP surfaces
derived in vacuo are not adequate for predicting all the observed
supramolecular synthons in ICCs.
improvement (Table 1) in 퐶 upon switching from the salt to the
k
ICC. The finding that 2‒4 show little or no change (Table 1) in the
packing efficiency of the species in the crystal as we switch from
the binary salt to the ternary ICC, is consistent with previous
observations[22] that poor crystal packing of the salt ions is not a
satisfactory explanation for rationalizing ICC formation. CSP
studies (rigid molecular conformations) on the set of ICCs with
previously reported crystal structures (3‒6 and 7-I) has shown
(see SI, Tables S6-S10) that all experimentally determined solid
forms could be predicted within an energy range of 20 kJ mol-1
with respect to the global minimum (GM) structure. The newly
characterized 7-II polymorph was outside the scope of the rigid-
body CSP search for 7 due to the molecular flexibility exhibited by
7-II. However, the Form I CUKNUT polymorph (7-I) was predicted
at rank 28 during the search for hypothetical rigid-body ICC
structures. The relatively high energy for 7-I is consistent with
DFT-D estimates of the relative energies for the two polymorphs
(see below) which has shown that 7-II is approximately 2.56 kJ
The physical appearance (Figure 1) and melting points
(
Table 1) of the ICCs were found to be sensitive to the chemical
identities of the AH/CH pairs used to crystallize the ICC. All ICCs
displayed sharp melting endotherms (see SI, Figure S6) with
onsets (Table 1) ranging from 91.85 °C (2) to 162.49 °C (7-II).
Comparison of the van der Waals volumes ( ∆푉
) of the
vdW
+
−
−
cation/anion (BH /A ) and coformer/anion (CH/A ) pairs in ICCs
3
3
-1
(
Table 1) shows negligible differences (ꢁ Å < ∆푉vdW < 11 Å ) in
mol more stable than 7-I. Overall, the finding that most of the
the intrinsic molecular volumes of all three hydrogen-bonded
species in the series 2‒7. The small sample size considered here
prevents us from drawing general conclusions about the
ICCs (2-5) display CFLs where the experimental crystal structure
is amongst the ten most stable structures and the fact that all
previously characterized ICC structures (3‒6 and 7-I) could be
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
Figure 4: (a) Overlay of the experimental PXRD patterns following NG and LAG (MeOH, i-PrOH or H
2
O) mechanosynthesis experiments involving a 1:1
stoichiometric ratio of 2-CLBZAH and 4-DMAP. PXRD patterns for 1, 1a, 1b, CLBZAC03 (2-CLBZAH) and BUKJOG11 (4-DMAP) are simulated from the single
crystal structures. Diagnostic Bragg reflections for facilitating comparison of the PXRD patterns obtained following NG or LAG (MeOH or i-PrOH) experiments with
the PXRD pattern simulated from the single crystal structure of 1 are indicated in diamond () whilst reflections for comparing the PXRD pattern of the LAG (H
product with the pattern simulated from the single crystal structure of 1a are indicated in asterix (*). (b) Final Rietveld refinement fit for the binary salt 2-CLBZA ∙4-
2
O)
-
+
DMAPH (1). (c) Crystal packings for 1, 1a and 1b illustrating the role that water plays in satisfying the unused hydrogen bond acceptors in 1 (indicated in black
circles).
predicted, suggest that computational CSP methods can play an
important role in the discovery of ternary molecular ICCs.
crystallization conditions meant that diffraction quality single
crystals of the salt anhydrate could not be grown. Instead, the
crystal structure of 1 (see SI, Table S2) was solved from the
PXRD data using the Monte Carlo simulated annealing method[
with the final Rietveld[39] fit (Figure 4b) converging at an 푅푤푝 of
5.72%. It is clear from the experimental crystal structure of 1
(Figure 4c) that each symmetry related anion displays a single
38]
Experimental crystal structures
Prior to our work, there were no reported binary or ternary
multicomponent solid forms derived from the 2-CLBZAH/4-DMAP
acid-base pair. The 2-CLBZA ∙ 4-DMAPH+ (1) binary salt was
successfully synthesized in quantitative yield under NG and LAG
-
unused hydrogen bond acceptor. Comparing
1 with the
(
MeOH or i-PrOH) conditions (Figure 4a) using
a
1:1
corresponding hydrate crystal structures shows that this
“frustration” in hydrogen bonding is resolved (Figure 4c) by the
inclusion of water, leading to solvent molecules bridging isolated
units of hydrogen-bonded molecular ions in 1a and 1b.
stoichiometric ratio of the acid:base pair. Attempts to grow
diffraction-quality single crystals of the mechanosynthesis product
for 1 using a range of solvents failed, with the exception of
experiments involving i-PrOH, which led to hydrated crystals
containing 1.33 (1a) or one (1b) water molecule(s) per mole of the
ion pair (see SI, Table S1). LAG mechanosynthesis of 1 using
water led to a sample whose PXRD pattern matched (Figure 4a)
that simulated from the single crystal structure of 1a. However
LAG using the same i-PrOH solvent used to grow single crystals
The comparable van der Waals volumes of 2-HBZAH and
[
18]
2-CLBZAH (Table 1) coupled with the existence of a reported
-
+
CAB-ICC (4, KUJDIE) with composition 2-HBZA ∙4-DMAPH ∙2-
HBZAH allowed us to target the synthesis of 2-CLBZA ∙4-
DMAPH ∙2- CLBZAH CAB-ICC (2) using a coformer replacement
-
+
strategy based on molecular size-matching. Suitable colorless
crystals of 2 with block morphology were grown using i-PrOH
solvent and the crystal structure of the ICC (see SI, Table S1) was
determined using single crystal X-ray diffraction (SXRD) methods.
Comparison of the supramolecular synthons of 2 and 4 (Figure 5)
1
a and 1b led instead to the salt anhydrate 1, suggesting that the
milling product is sensitive to the water activity in the
crystallization medium. Heating a powder sample of 1a to a
temperature of 110 °C led to facile dehydration (see SI, Figure
S5), furnishing 1. The facile hydration of 1 under solution
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
arrangements to CUKNON, implies minimal risk of packing
polymorphism in this ICC system.
NG experiments targeting the CAB-ICC 7-I (CUKNUT)[14]
led instead to the characterization of a new polymorph (7-II) of this
ICC (Figure 7a). The crystal structure (see SI, Table S2) for 7-II
was solved from the PXRD data of the NG product using the
[
38]
Monte Carlo simulated annealing
method (Figure 7b: final
Rietveld[ fit had 푅푤푝 = 3.60%). The 7-II polymorph crystallizes
39]
in the same 푃ꢁ space group observed for 7-I. Variable-cell DFT-
ꢀ
D geometry optimizations confirmed 7-II to be a true lattice energy
minimum with sensible hydrogen bond geometries. The DFT-D
optimizations were used to support the identification of the acidic
proton positions in 7-II given the low scattering factor of the
hydrogen atom. In both polymorphs, all hydrogen bond donors
and acceptors are satisfied (Figure 7c) and the structures display
Figure 5: Observed hydrogen bonding (colored cyan) motifs in CAB-ICCs 2 and
4
. Despite the comparable molecular volumes of 2-HBZAH and 2-CLBZAH,
significant differences in the crystal packings and solid-state properties (Table
) of 2 and 4 are observed. The centroid-centroid distance (4.32 Å) in the 휋— 휋
ꢀ
1
discrete 퐷 (ꢁ) synthons between the acid-anion and cation-anion
ꢀ
stacking interactions between the acid and cation of 4 are indicated. Anions are
pairs. In addition to these discrete interactions, 7-II displays a
colored blue, cations in red and acid coformers in green.
2
2
퐶 (16) chain motif between the acid and anion (Figure 7c), with
ꢀ
ꢀ
show however that 2 adopts a 퐷 (ꢁ) synthon whilst 4 adopts the
2
2
common 퐷 (5) synthon that is observed for most of the ICCs (3‒
6
) in our series. Inspection of the molecular conformations
−
2
adopted by the A /CH pairs in ICCs that display the 퐷 (5)
2
synthon (3‒6), show that they are all approximately planar. The
planar molecular conformations in these ICCs, facilitates close
molecular association marked by 휋— 휋 interactions between the
acid and cation with centroid-centroid distances (Figure 5)
ranging from 3.60—4.32 Å. By contrast, the torsions defining
rotation of the carboxylic acid and carboxylate groups in 2 are -
71° and -48° respectively, leading to significant deviations from
planarity (Figure 5) and by extension the absence of the close
molecular association that appears to be the hallmark of those
2
2
ICCs displaying the 퐷 (5) synthon. Thus, whilst a molecular-size
matching approach may be successful in targeting CAB-ICCs that
−
differ in the identities of A /CH pairs, these observations illustrate
the challenges in attempting to extrapolate the dominant
supramolecular synthons of ICCs on the basis of the synthons
observed in structurally related systems.
-
+
The CFL of 4-CLBZA ∙ 4-DMAPH ∙ 4-CLBZAH (3, Refcode:
CUKNON)[14] was analyzed (Figure 6) to gain
deeper
a
understanding of the alternative packing preferences of the
2
2
molecular components in ICCs displaying the 퐷 (5) synthon.
The experimental CUKNON structure of 3 is the GM in lattice
energy and is approximately 3.4 kJ mol-1 more stable than the
rank 2 structure on the CFL. The top 3 ranked ICC structures all
2
2
display the same 퐷 (5) synthon. Comparing the hydrogen bond
geometry in each predicted polymorph of 3 with the geometry in
the reference GM structure has revealed that 31% of all structures
within 10 kJ mol-1 of the GM, adopt the 퐷 (5) synthon with a high
2
2
Figure 6: (a) Crystal form landscape (CFL) for CAB-ICC 3 (CUKNON) illustrating
the diversity of possible packing polymorphs classified in terms of the root-
mean-squared deviation (RMSD / Å) of the predicted hydrogen bond motif for
each polymorph relative to that of the experimental global minimum (GM)
structure. ∆퐸ꢅaꢆꢆ is the energy of each polymorph relative to the GM structure.
The hydrogen bond similarities were judged according to the following cut-offs:
High: 0 ≤ 푅푀푆퐷 ≤ 0.ꢇ ; Medium: 0.ꢇ < 푅푀푆퐷 ≤ 0.8 ; Low: 푅푀푆퐷 > 0.8 . (b)
Overlay of the crystal packings in the predicted GM structure (colored by
element) and the experimental CUKNON structure (colored green) after
matching 15 molecules in the coordination spheres of the two structures.
degree of similarity (root-mean-squared deviation: 0‒0.3 Å ).
Within the usual caveats[40] of static lattice energy calculations,
the computational model therefore suggests that the reported[14]
experimental structure of 3 is energetically favored and there
appears to be an energetic driving force for nucleation of crystals
2
2
that display the 퐷 (5) synthon. Since most of the predicted low
energy polymorphs belong to the same cluster of structures
2
2
displaying the 퐷 (5) synthon, it follows that the absence of
energetically competitive structures with alternative packing
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
Figure 7: (a) Comparison of the simulated PXRD patterns from the single crystal structures of 7-I (CUKNUT), 7-II, 4-HBZAH (JOZZIH01) and 4-DMAP (BUKJOG11)
with the experimental PXRD pattern obtained following neat grinding (NG) of a 2:1 molar ratio of 4-HBZAH and 4-DMAP. * Indicates diagnostic Bragg reflections for
comparing the experimental PXRD pattern of the NG product with the simulated PXRD pattern of 7-II. (b) Final Rietveld refinement fit for Form II of 4-
dimethylaminopyridinium 4-hydroxybenzoate 4-hydroxybenzoic acid (7-II). (c) Comparison of the hydrogen bond motifs in 7-I (CUKNUT) and 7-II. Anions are colored
blue, cations in green and acid coformers in red.
one cation trapped inside each chain. The packing coefficient (퐶k)
of 7-II is approximately 2.8% higher (Table 1) than that observed
for the 4-HBZA ∙4-DMAPH+ binary salt (CSD refcode:
the crystal structure for 1 is in support of the experimentally
determined structure and highlights the value of CSP methods for
aiding structure solution from PXRD data. Notably, the CFL for 1
contains many low-energy/high-density polymorphs within 20 kJ
mol-1 that are more stable and efficiently packed than the
experimental structure of 1 (Figure 8a). The facile hydration of 1
to form a non-stoichiometric (1a) hydrate and the observation of
a dense CFL with many more stable polymorphs of the anhydrate,
therefore suggest that there is scope for the discovery of more
stable anhydrous polymorphs of 1. By contrast, the crystal
-
[
41]
SOLGUX ), whilst the 퐶 of 7-I is identical to that of the salt. The
k
7-II polymorph displays the highest onset of melting (162.49 °C)
of all the ICCs in the series 2‒7 (Table 1). Challenges in the
mechanosynthesis of pure 7-I prevented us from characterizing
its physical properties relative to 7-II.
Blind structure prediction and periodic DFT-D estimates of
the energetic driving force for ICC formation
th
structure of CAB-ICC 2 was successfully predicted as the 5 most
-
+
stable polymorph, with a relative energy of 6.5 kJ mol-1 with
respect to the GM structure. The CFL (Figure 8b) for this system
is by comparison much less dense in the low-energy region with
fewer stable lattice energy minima. Notably, CAB-ICC 2 is
predicted to have a CFL with a range of polymorphs (including the
experimental structure) that are more stable than the sum of the
lattice energies for the most stable salt and acid polymorphs (i.e.
below the green line in Figure 8b), indicating a thermodynamic
preference for ICC formation. The successful prediction of the
The CFLs for the 2-CLBZA ∙4-DMAPH binary salt (1) and 2-
-
+
CLBZA ∙4-DMAPH ∙2-CLBZAH ternary CAB-ICC (2) were
calculated (Figure 8) from first principles under blind test[21]
conditions. Both systems display molecular flexibility and reflect
challenging targets for current methods of CSP. The blind nature
of the CSP work allowed us to test the predictive value of
computed CFLs for guiding the selection of ternary molecular
th
ICCs. The binary salt structure of 1 was predicted as the 19 most
stable polymorph in the CFL (Figure 8a) and was approximately
5
.4 kJ mol-1 above the GM structure. The successful prediction of
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
Figure 8: Predicted crystal form landscapes (CFLs) for the (a) binary salt 1 and (b) CAB-ICC 2. Each black circle represents a hypothetical polymorph on the lattice
-
1
energy landscape. The experimental structures for 1 and 2 are indicated by the red triangles. The blue line in each CFL indicates an energy range of 20 kJ mol
relative to the predicted global minimum (GM) structure. The green line in (b) indicates the sum of the lattice energies for the GM structures in the CFLs of the acid
and salt. (c) Molecular degrees of freedom for each component of 1 and 2 that were optimized during the final ranking of the crystal lattice energies using
CrystalOptimizer.70 Curly arrows indicate torsions and square arrows indicate angles. Structural overlays between the experimental and matching predicted crystal
structures are shown for the (d) binary salt 1 and (e) CAB-ICC 2.
experimental structure of this flexible ternary ICC system under
ꢊꢋꢌꢌ
2 ꢊSꢎꢏT
ꢀ ꢊꢎꢌꢋD
blind test conditions, suggests that CSP methods are now at a
stage of maturity where they can be used to quantify the risks of
polymorphism and aid in the discovery of higher-order
multicomponent crystal forms.
∆
퐸ꢈꢉꢉ = 푍ꢋꢌꢌ ꢂ ꢍ3 푍SꢎꢏT ꢐ 3 푍ꢎꢌꢋD
ꢑ
(1)
where 퐸 are the electronic structure energies of the unit cells and
푖
In an attempt to obtain accurate lattice energies for probing
the thermodynamic driving force for ICC formation, the
experimental crystal structures of all ICCs in our series (2‒7) were
subjected to variable-cell DFT-D geometry optimizations using
the energies are normalized by the number of molecules in the
unit cell (ꢒ) with the comparison made between a mole of ICC and
2/3 and 1/3 of a mole of salt and acid respectively. For all ICCs in
our series (2‒7), we find that ∆퐸ꢈꢉꢉ is negative suggesting that
the TPSS meta-GGA functional[42] with D3 dispersion
crystallization of the ICC is energetically favored (Figure 9). The
correction.[
43]
Similar optimizations were performed for the
-1
mean value of ∆퐸ꢈꢉꢉ is -2 kJ mol across the series. However, the
corresponding salt and acid components used to construct the
ICC (see SI, Table S3). The energetic driving force for crystallizing
an ICC (∆퐸ꢈꢉꢉ ) when compared to the competing process of
crystallizing a physical mixture of the binary salt and acid was
estimated (Figure 9) according to the following equation
-1
ICC is energetically favored by less than 1 kJ mol for 2, 4 and 5
suggesting that the use of DFT-D relative energies alone, would
not have been sufficient to rule out formation of a physical mixture
of the binary salt and acid. For 3, 6 and 7, ∆퐸 ranges from -1.45
ꢈꢉꢉ
kJ mol-1 to -4.86 kJ mol-1. Although the thermodynamic driving
force for the crystallization of ICCs has yet to be studied in a
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
sufficient to target the ternary ICC, although LAG experiments did
lead to improved phase purity and crystallinity for most systems.
ICCs display thermal stabilities that can be tuned depending on
the chemical identities of AH and CH. The binary salt complex of
4-dimethylaminopyridinium
2-chlorobenzoate
(1)
proved
challenging to crystallize under solution crystallization conditions
given the relatively high affinity of this salt for water, which led to
hydrated crystals containing 1.33 (1a) and one (1b) water
molecule(s) per ion pair. The salt complex of 1 was synthesized
in quantitative yield under NG or LAG conditions using the ball-
milling technique. A novel CAB-ICC of 4-dimethylaminopyridinium
2-chlorobenzoate 2-chlorobenzoic acid (2) was synthesized using
a
coformer replacement strategy based on molecular-size
matching. All ICCs with rigid molecular conformations (3‒6 and 7-
I) were shown to have crystal packings that could be predicted
from first principles using computational methods of CSP. The
crystal structures for the conformationally flexible binary salt
complex (1) and ternary CAB-ICC (2) determined in this work
were also successfully predicted under blind test conditions using
no prior information other than the input molecular structures. A
Figure 9: Estimate of the energetic driving force (∆퐸퐼ꢄꢄ) for crystallizing each of
the ternary molecular ICCs (2‒7) in our study relative to the competing process
of crystallizing a physical mixture of the binary salt and acid. We assume
crystallization is under thermodynanic control and that negative values for ∆퐸퐼ꢄꢄ
imply favorable ICC formation. 7-I refers to the Form I CUKNUT structure whilst
Form
II
structure
for
4-dimethylaminopyridinium
4-
hydroxybenzoate 4-hydroxybenzoic acid (7-II) was synthesized
from NG experiments using a 2:1 molar ratio of the acid and base.
This polymorph was shown to have the highest thermal stability
of all ICCs surveyed and its discovery illustrates the importance
of mechanochemical screening for higher-order cocrystals using
NG conditions as the absence of solvent can provide access to
solid forms not seen in solution. DFT-D calculations on all ICCs
surveyed has shown good agreement with experimental
observation in predicting that all ICC systems are
thermodynamically stable relative to the crystallization of a
physical mixture of the salt and acid. However, the mean
stabilization energy of ICCs in our limited series was found to be
-2 kJ mol , which is lower than the -8 kJ mol previously reported
for binary cocrystals using a comparable DFT-D model. Overall,
the successful computational prediction and experimental
realization of previously unknown ternary molecular ICCs suggest
that a complementary mechanosynthesis and crystal structure
prediction approach could aid in the rapid screening and selection
of functional molecular ICCs with improved physicochemical
properties.
7
-II is the Form II structure for this CAB-ICC determined in this work.
rigorous manner, several previous reports have attempted to
quantify the thermodynamic driving force for the crystallization of
binary cocrystals. Issa et al[44] studied 26 cocrystals of 4-
aminobenzoic acid, succinic acid and caffeine using a multipole
electrostatic model combined with an empirical potential for
describing the dispersion-repulsion interactions. In general, they
found that the experimentally observed cocrystals tended to be
more stable than the pure components. However, several
exceptions to this rule were observed, such as the finding that an
experimentally observed urea-succinic acid cocrystal was
-1
-1
-1
calculated to be 12.09 kJ mol higher in energy relative to its pure
components. A more recent study by Taylor and Day[45] employed
DFT-D calculations using the PBE functional and D3 dispersion
correction to evaluate the relative stabilities of 350 cocrystals.
This larger study concluded that 95% of experimentally observed
binary cocrystals are thermodynamically stable relative to their
pure component energies. The average stabilizing energy of
these systems was found to be -8 kJ mol . This is in contrast to
our findings for the thermodynamic stabilization of ternary
molecular ICCs in our limited set, which appears to be in better
agreement with the energy differences of pure component
polymorphs.[40b, 46]
-1
Conclusions
A rational approach for the discovery of multicomponent crystal
forms is critical to the formulation of active ingredients with optimal
solid-state properties. For ternary molecular ICCs, the differences
in the solubilities of the reactants can be significant and this can
limit the success of solution crystallization screens. Our work has
−
+
shown that ternary molecular ICCs of the type A ∙ BH ∙ CH can
be synthesized using mechanochemical ball-milling in as little as
30 minutes. For most of the systems surveyed, NG conditions are
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
Experimental Section
targeting NCAB-ICCs employed 1:1:1 molar ratios of
1 2
acid :acid :base. OP-LAG experiments were performed using
methanol (MeOH), isopropanol (i-PrOH) or distilled water using a
List of abbreviations and symbols
solvent/solute ratio (휂)[ of approximately 0.1 μL mg . For S-
LAG experiments, 30 μL of solvent was added to a stoichiometric
ratio of the reactants in each step. Unless otherwise stated,
assume that the solvent of choice in S-LAG or OP-LAG
experiments is MeOH. For all acid-base pairs, the calculated[
aqueous ∆p퐾a was in the range 4.40‒5.99, indicating high
probability[26, 30, 47] of ionization to form charged molecular ions
during the mechanosynthesis experiment. Mechanosynthesis
experiments were performed using a Retsch MM200 mixer mill
equipped with a 10 mL capacity stainless steel grinding jars and
two 7 mm diameter stainless steel grinding balls per jar. All milling
experiments were performed at a frequency of 25 Hz. To test the
reproducibility of the mechanosynthesis protocol as well as the
stability of the ICC towards amorphization, the grinding
experiments were repeated for each reaction over successively
longer time periods spanning 15, 30 or 60 minutes. PXRD data
were collected on the milling products obtained after 60 minutes
of grinding.
25]
−ꢀ
Abbreviation/symbol
Definition
2
2
4
4
4
-CLBZAH
-HBZAH
-CLBZAH
-DMAP
2-chlorobenzoic acid
2-hydroxybenzoic acid
4-chlorobenzoic acid
4-dimethylaminopyridine
4-hydroxybenzoic acid
32]
-HBZAH
AH/CH
Acid coformer (2-CLBZAH, 2-
HBZAH, 4-CLBZAH or 4-HBZAH)
Base (4-dimethylaminopyridine)
B
CAB-ICC
Conjugate acid-base ionic
cocrystal
Crystal form landscape
CFL
CSD
퐶푘
Cambridge Structural Database
Kitaigorodsky packing coefficient
Crystal structure prediction
CSP
DFT-D
Dispersion-corrected density
functional theory
Global minimum
Single crystal X-ray diffraction
The OP-LAG (i-PrOH solvent) products obtained after milling 1:1
or 1:2 molar ratios of 4-DMAP:2-CLBZAH were used to grow
single crystals of the salt hydrates (1a, 1b) and CAB-ICC (2)
respectively. Diffraction quality single crystals of 1a/1b and 2 were
grown via the slow solvent evaporation of i-PrOH using 0.718
mmol or 0.689 mmol of the ground reactants dissolved in 5 mL of
i-PrOH. The i-PrOH solvent was allowed to evaporate slowly in a
ventilated fume hood over a period of 7 days. The experiment
targeting the binary salt complex led to a mixture containing single
crystals of 1a and 1b. Single crystal X-ray diffraction data for 1a,
GM
ICC
Ionic cocrystal
i-PrOH
LAG
Isopropanol
Liquid-assisted grinding
Methanol
MeOH
MEP
Molecular electrostatic potential
NCAB-ICC
Non-conjugate acid-base ionic
cocrystal
Neat grinding
1b and 2 were collected using a Bruker Duo three-circle
diffractometer equipped with a Cobra cooling device (Oxford
Cryosystems) with graphite-monochromated 푀표 퐾훼 ( 휆 =
NG
OP-LAG
PXRD
One-pot liquid-assisted grinding
Powder X-ray diffraction
0.7107ꢇ Å) radiation and Photon 100 CMOS area detector. The
optimum strategy for data collection involved different sets of 휙
and 휔 scans with 0.5 ° steps in 휙/휔. Data collection, integration,
scaling and absorption corrections were performed using Bruker
Apex 3 software.[48] Data reduction was performed using SAINT[49]
and XPREP.[50] All the data were corrected for Lorentzian,
RMSD
n
Root mean-squared deviation (Å)
for matching at least n molecules
in the coordination spheres of
two crystal structures
[
51]
polarization and absorption effects using the SADABS program.
The structures were solved using SHELXT-2014/5[ and refined
S-LAG
SXRD
푉vdW
휂
Serial liquid-assisted grinding
52]
Single-crystal X-ray diffraction
van der Waals volume (Å3
Solvent/solute ratio (μL mg−ꢀ
used in mechanosynthesis
experiments
2
by full-matrix least-squares methods based on 퐹 against all
reflections using SHELXL-2014/7[53]. The Olex2 graphical user
interface[54] was used to visualize and manipulate the structure
solution and refinement process. Two water molecules in 1a
exhibited disorder, which was modelled and treated by using a set
of constraints, namely EADP and EXYZ. All non-hydrogen atom
positions were located using difference Fourier methods and
refined anisotropically. The positions of hydrogens bonded to
carbon were set using the HFIX command, whilst all other
hydrogen atom positions were located from the difference Fourier
map and freely refined. Publication quality images were
)
)
Mechanosynthesis of ICCs
The chemicals 4-DMAP, 4-CLBZAH, 2-HBZAH and 4-HBZAH
were purchased from Sigma-Aldrich ( ≥ 98% GC), whilst 2-
CLBZAH was purchased from Acros Organics (≥ 98% GC). All
chemicals were used in experiments as supplied. CAB-ICCs were
synthesized using 2:1 acid:base molar ratios whilst experiments
[55]
X-Seed[56] and POV-Ray.
[57]
generated using Mercury 4.0.0,
A
summary of the crystallographic information for 1a, 1b and 2 can
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
be found in SI Table S1. CCDC 1956859-1956863 contain all the
supplementary crystallographic data for this paper. These data
can be obtained free of charge from the CCDC
Differential scanning calorimetry and thermogravimetric
analysis
Differential scanning calorimetry and thermogravimetric analysis
were performed using a NETZSCH STA 449F3 Jupiter and
NETZSCH STA 401 F1 Pegasus instruments respectively. A
ceramic crucible was used in both experiments. Samples were
purged by a stream of dry nitrogen gas and heated at a rate
5°C/min over the temperature range 30‒300 °C.
(
www.ccdc.cam.ac.uk/conts/retrieving.html).
Powder X-ray diffraction
Powder X-ray Diffraction (PXRD) data were collected on a
PANalytical Empyrean diffractometer equipped with an
X’Celerator RTMS detector. Diffraction experiments were
performed in Bragg-Brentano reflection geometry using Ni-filtered
Computational crystal structure prediction
퐶푢 퐾 radiation. PXRD data were collected in the angular range
Two sets of crystal structure prediction (CSP) data were
generated; 1) Set 1 consisted of the predicted crystal form
landscapes (CFLs) for known rigid CAB-ICCs and NCAB-ICCs 3‒
7; 2) Set 2 consisted of generating the CFLs for the flexible binary
salt 1 and CAB-ICC 2. For the set 1 data, the assumed molecular
conformations were calculated at the M06/6-31G(d,p) level of
theory whilst for set 2, the calculations were performed at the
PBEPBE/6-311+G(d,p) level of theory. In all cases, molecular
conformations were calculated in the gas phase using
GAUSSIAN09. Searches for hypothetical crystal structures for 3‒
7 assumed the calculated gas phase conformational energy
minima as input and these conformations were constrained
throughout the simulation. For systems 1 and 2, flexible torsions
were determined through second derivatives and finite difference
perturbations. Local Approximate Models (LAMs) were
constructed using a uniform grid along the one-dimensional DOFs,
훼
4‒50° in ꢁ휃 but due to the lack of any significant diffraction peaks
at the extremes, the range 5‒45° ꢁ휃 was used for comparison
with the simulated PXRD patterns as well as for structure solution
of
1
and 7-II. Cell indexing and systematic absence
determinations for 1 (MeOH LAG product) and 7-II (NG product)
were performed using the TREOR90[58] and X-Cell[59] programs
respectively as implemented in the Reflex module of BIOVIA
Materials Studio 8.0.[60] The background was fitted using a six-
order polynomial and the peak profiles were modelled using a
Pseudo-Voigt function. Pawley refinement was performed on the
indexed unit cell using the Reflex module. The starting molecular
geometries in 1 and 7-II were calculated in the gas phase at the
M06/6-31G(d,p) level of theory using GAUSSIAN09[61]. Forcite[60]
was used to optimize the relative positions of the species in the
indexed unit cell using the Fine (1) or Ultra-fine (7-II) quality
setting for the convergence criteria of the energy, forces and
displacement. During this initial Forcite geometry optimization,
rigid-body constraints were imposed on the molecular
conformations and the unit cell parameters were not allowed to
vary. The Dreiding force field was used during the geometry
optimization and atomic charges were derived using the
Gasteiger method. The crystal structures for 1 and 7-II were
solved using the Reflex PowderSolve[38] module of Materials
Studio 8.0, which uses a Monte Carlo simulated annealing
technique. In addition to the degrees of freedom (DOFs) defining
the positions and relative orientations of the molecular units in the
asymmetric unit, additional DOFs for the torsions defining rotation
at 30° increments. The global search was performed using
CrystalPredictor II,[
62]
using the smoothed intramolecular
potential[63], with 500 × 10 minimizations for 1 and 1 million for 2.
During the initial structure generation stage, the electrostatic
contributions towards the intermolecular forces were estimated
using atomic ESP charges derived from the ab initio
wavefunctions. Dispersion–repulsion contributions towards the
lattice energy were estimated using a Buckingham exp-6 function
3
using the FIT potential parameters for C, H
C
(hydrogen attached
to carbon), N, O and Cl from the work of Williams[ as well as the
64]
parameters for H
N O
(hydrogen attached to nitrogen)[65] and H
(hydrogen attached to oxygen)[66] that were subsequently
determined by fitting to crystal structures containing the N–
H···O=C interactions and carboxylic acid structures respectively.
For each search, a final clustering step was performed to remove
2
of the NMe , carboxylate or carboxylic acid groups were also
defined prior to initiating the Monte Carlo search. Hydrogen atoms
were explicitly modelled during the Monte Carlo search for the
global minimum structure in the 푅푤푝 surface. The number of
Monte Carlo steps required to converge on the correct solution
was set automatically by Reflex based on the specified DOFs for
each system, leading to a total of 24 million and 417 million steps
for 1 and 7-II respectively. Rietveld[39] refinement was performed
within Reflex and all relevant flexible torsions, bond lengths and
angles were refined. The March-Dollase preferred orientation
correction was applied. All non-hydrogen atom temperature
factors were refined anisotropically whilst hydrogen atoms were
treated as isotropic. The final Rietveld cycle yielded an 푅푤푝 of
-1
all duplicate structures with lattice energies within 0.2 kJ mol , cell
3
[55]
volumes within 1.0 Å and a powder pattern similarity index of
at least 0.97. The Crystal Packing Similarity module of Mercury
4.0.0[55] was used to match the predicted and experimental crystal
structures. This was done by estimating the root-mean-squared
deviation for matching at least 15 molecules (RMSD15) in the
coordination spheres of the experimental and predicted structures
using a tolerance of 20% for the distances and 20° for the angles.
For set 1, the 1000 most stable structures produced following
clustering were passed to DMACRYS[
67]
for lattice energy
minimization using a distributed multipole model[
68]
for the
5.72 % for 1 and 3.60 % for 7-II. A summary of the crystallographic
information for 1 and 7-II can be found in SI Table S2.
electrostatic contribution towards the lattice energy. Multipoles
were calculated up to rank 4 (hexadecapole) for all atoms by
performing a distributed multipole analysis of the ab initio charge
density using GDMA2.2.[ For set 2, CrystalOptimizer
69]
[70]
was
used to refine the lowest energy 1000 structures at the
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
PBEPBE/6-311+G(d,p) level of theory, with additional flexibility
introduced as indicated in Figure 8c. All lattice energies in the
reported CFLs are given per formula unit.
Keywords: crystal engineering • mechanosynthesis • molecular
ionic cocrystals • green chemistry • crystal structure prediction
Molecular electrostatic potentials
[
[
1]
2]
J. Bernstein, Polymorphism in Molecular Crystals, Vol. 30, 2nd Edition
ed., Oxford University Press, 2019.
The assumed molecular conformations for computing the
molecular electrostatic potentials (MEPs) were those calculated
following gas phase geometry optimization at the M06/6-31G(d,p)
level of theory using GAUSSIAN09. MEPs for all the acid
coformers in Scheme 1 were calculated in vacuo using the TPSS-
D3 density functional with the 6-311+G** basis set. Local maxima
and minima on the MEP surface (0.002 e/au isosurface) were
determined using a positive point charge in vacuum as a probe.
The calculations lead to the interaction energy (in kJ mol )
between the positive point probe and the surface of the molecule
at the point of contact. The MEP surfaces were computed using
Spartan’18.[71]
a) M. K. Corpinot, D.-K. Bučar, Cryst. Growth Des. 2019, 19, 1426-
1453; b) S. Kandambeth, K. Dey, R. Banerjee, J. Am. Chem. Soc.
2019, 141, 1807-1822; c) T. Hasell, A. I. Cooper, Nat. Rev. Mater.
2016, 1, 16053.
[
[
3]
4]
G. R. Desiraju, Angew. Chem. Int. Ed. Engl. 1995, 34, 2311-2327.
a) M.-A. Perrin, M. A. Neumann, H. Elmaleh, L. Zaske, Chem.
-1
Commun. 2009, 3181-3183; b) R. M. Bhardwaj, J. A. McMahon, J.
Nyman, L. S. Price, S. Konar, I. D. H. Oswald, C. R. Pulham, S. L.
Price, S. M. Reutzel-Edens, J. Am. Chem. Soc. 2019, 141, 13887-
13897; c) A. Pulido, L. Chen, T. Kaczorowski, D. Holden, M. A. Little, S.
Y. Chong, B. J. Slater, D. P. McMahon, B. Bonillo, C. J. Stackhouse, A.
Stephenson, C. M. Kane, R. Clowes, T. Hasell, A. I. Cooper, G. M. Day,
Nature 2017, 543, 657; d) M. Mortazavi, J. Hoja, L. Aerts, L. Quéré, J.
van de Streek, M. A. Neumann, A. Tkatchenko, Commun. Chem. 2019,
Solid-state periodic DFT-D geometry optimizations
All periodic DFT-D calculations were performed using VASP[72]
with planewave basis sets and PAW pseudopotentials. The
combination of the TPSS meta-GGA functional[42] and D3
dispersion correction[43] was chosen for optimization of the crystal
structures due to this combination’s good performance in previous
benchmark studies of the X23 test set[73] as well as its
2, 70; e) P. Cui, D. P. McMahon, P. R. Spackman, B. M. Alston, M. A.
Little, G. M. Day, A. I. Cooper, Chem. Sci. 2019.
[
[
5]
6]
J. Haleblian, W. McCrone, J. Pharm. Sci. 1969, 58, 911-929.
a) N. Shan, M. J. Zaworotko, Drug Discovery Today 2008, 13, 440-446;
b) N. K. Duggirala, M. L. Perry, O. Almarsson, M. J. Zaworotko, Chem.
Commun. 2016, 52, 640-655; c) C. B. Aakeröy, M. E. Fasulo, J.
Desper, Mol. Pharmaceutics 2007, 4, 317-322.
demonstrated ability to correctly determine relative polymorph
stabilities.[74] We make use of a large basis set with an energy cut-
off of 1000 푒푉 and a tight 퐾-point mesh with the maximum 퐾-point
distance set to ꢁ휋 × 0.0ꢁ5 Å−ꢀ using a Γ-centred Monkhorst-Pack
scheme. Convergence of each SCF cycle was set to 10 푒푉 and
the geometry was considered converged when all forces were
below 0.01 푒푉/Å. Starting from the experimentally determined
crystal structures, lattice parameters and the fractional
coordinates of the atom positions were optimized simultaneously.
Careful attention was paid to check that lattice parameters did not
change significantly during the optimization to mitigate the extent
of Pulay stresses.
−ꢓ
[7]
a) D. Braga, F. Grepioni, L. Maini, S. Prosperi, R. Gobetto, M. R.
Chierotti, Chem. Commun. 2010, 46, 7715-7717; b) D. Braga, F.
Grepioni, G. I. Lampronti, L. Maini, A. Turrina, Cryst. Growth Des. 2011,
11, 5621-5627; c) A. J. Smith, S.-H. Kim, N. K. Duggirala, J. Jin, L.
Wojtas, J. Ehrhart, B. Giunta, J. Tan, M. J. Zaworotko, R. D. Shytle,
Mol. Pharmaceutics 2013, 10, 4728-4738; d) N. K. Duggirala, A. J.
Smith, Ł. Wojtas, R. D. Shytle, M. J. Zaworotko, Cryst. Growth Des.
2
014, 14, 6135-6142; e) A. R. Buist, A. R. Kennedy, Cryst. Growth Des.
014, 14, 6508-6513; f) L. Casali, L. Mazzei, O. Shemchuk, K. Honer,
2
F. Grepioni, S. Ciurli, D. Braga, J. Baltrusaitis, Chem. Commun. 2018,
4, 7637-7640.
N. A. Mir, R. Dubey, G. R. Desiraju, Acc. Chem. Res. 2019, 52, 2210-
220.
M. Paul, G. R. Desiraju, Angew. Chem. Int. Ed. Engl. 2019, 58, 12027-
2031.
10] M. Paul, S. Chakraborty, G. R. Desiraju, J. Am. Chem. Soc. 2018, 140,
309-2315.
Acknowledgements
5
[
[
[
8]
9]
Financial support for this work was made possible by Khalifa
University of Science and Technology under the CIRA program
2
(
Project Code: CIRA-2018-068). DHB would like to acknowledge
1
financial support from Syngenta as well as the EPSRC grant
number EP/N509486/1. IJS would like to acknowledge financial
support from Eli Lilly and Company. The authors would like to
acknowledge the support of the research laboratories and
research computing departments of Khalifa University. SM would
like to thank Profs. C.C Pantelides and C.S. Adjiman (Imperial
College London) for providing access to the CrystalPredictor and
CrystalOptimizer codes. A subset of the theoretical calculations
reported in this work were performed using the high-performance
computing clusters at the Masdar and SAN campuses of Khalifa
University. The remaining theoretical calculations were performed
using the Imperial College Research Computing Service, DOI:
2
[11] D. Braga, F. Grepioni, O. Shemchuk, CrystEngComm 2018, 20, 2212-
2220.
[12] a) S. Mohamed, S. A. Barnett, D. A. Tocher, Acta Crystallogr., Sect. E:
Crystallogr. Commun. 2007, 63, o3575; b) S. Mohamed, Acta
Crystallogr., Sect. E: Crystallogr. Commun. 2016, 72, 1348-1352.
[
[
[
13] S. R. Perumalla, C. C. Sun, CrystEngComm 2013, 15, 5756-5759.
14] B. Lou, S. R. Perumalla, C. C. Sun, J. Mol. Struct. 2015, 1099, 516-522.
15] a) D. Cinčić, I. Brekalo, B. Kaitner, Cryst. Growth Des. 2012, 12, 44-48;
b) S. Karki, T. Friščić, W. Jones, W. D. S. Motherwell, Mol.
10.14469/hpc/2232.
Pharmaceutics 2007, 4, 347-354; c) J.-L. Do, T. Friščić, ACS Cent. Sci.
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
2
017, 3, 13-19; d) K. Preet, G. Gupta, M. Kotal, S. K. Kansal, D. B.
Swift, R. Thaimattam, T. S. Thakur, R. Kumar Thaper, S. P. Thomas, S.
Tothadi, V. R. Vangala, N. Variankaval, P. Vishweshwar, D. R. Weyna,
M. J. Zaworotko, Cryst. Growth Des. 2012, 12, 2147-2152; b) E.
Grothe, H. Meekes, E. Vlieg, J. H. ter Horst, R. de Gelder, Cryst.
Growth Des. 2016, 16, 3237-3243; c) M. Gryl, M. Koziel, K. M.
Stadnicka, Acta Crystallogr., Sect. B: Struct. Sci., Cryst. Eng. Mater.
2019, 75, 53-58.
Salunke, H. K. Sharma, S. Chandra Sahoo, P. Van Der Voort, S. Roy,
Cryst. Growth Des. 2019, 19, 2525-2530.
[
16] a) K. Linberg, N. Z. Ali, M. Etter, A. A. L. Michalchuk, K. Rademann, F.
Emmerling, Cryst. Growth Des. 2019, 19, 4293-4299; b) O. Shemchuk,
V. André, M. T. Duarte, D. Braga, F. Grepioni, CrystEngComm 2019,
21, 2949-2954.
[
[
17] T. Friščić, W. Jones, Cryst. Growth Des. 2009, 9, 1621-1637.
[28] W. Guo, S. Du, Y. Lin, B. Lu, C. Yang, J. Wang, Y. Zeng, New J. Chem.
2018, 42, 15068-15078.
18] B. Lou, S. R. Perumalla, C. C. Sun, Cryst. Growth Des. 2015, 15, 24-
2
8.
19] a) A. T. Hulme, S. L. Price, D. A. Tocher, J. Am. Chem. Soc. 2005, 127,
116-1117; b) M. A. Neumann, J. van de Streek, F. P. A. Fabbiani, P.
[29] M. Hemamalini, H.-K. Fun, Acta Crystallogr., Sect. E: Crystallogr.
Commun. 2010, 66, o479-o480.
[
1
[30] S. L. Childs, G. P. Stahly, A. Park, Mol. Pharmaceutics 2007, 4, 323-
338.
Hidber, O. Grassmann, Nat Commun 2015, 6; c) S. Mohamed, L. Li,
CrystEngComm 2018, 20, 6026-6039; d) S. Mohamed, D. P. Karothu,
P. Naumov, Acta Crystallogr., Sect. B: Struct. Sci., Cryst. Eng. Mater.
[31] O. N. Kavanagh, G. Walker, M. Lusi, Cryst. Growth Des. 2019, 19,
5308-5313.
2
016, 72, 551-561.
[32] Calculator Plugins were used for structure property prediction and
calculation, MarvinSketch v16.3.7.0, ChemAxon
[
[
20] A. R. Oganov, Faraday Discuss. 2018, 211, 643-660.
21] A. M. Reilly, R. I. Cooper, C. S. Adjiman, S. Bhattacharya, A. D. Boese,
J. G. Brandenburg, P. J. Bygrave, R. Bylsma, J. E. Campbell, R. Car,
D. H. Case, R. Chadha, J. C. Cole, K. Cosburn, H. M. Cuppen, F.
Curtis, G. M. Day, R. A. DiStasio Jr, A. Dzyabchenko, B. P. van Eijck,
D. M. Elking, J. A. van den Ende, J. C. Facelli, M. B. Ferraro, L. Fusti-
Molnar, C.-A. Gatsiou, T. S. Gee, R. de Gelder, L. M. Ghiringhelli, H.
Goto, S. Grimme, R. Guo, D. W. M. Hofmann, J. Hoja, R. K. Hylton, L.
Iuzzolino, W. Jankiewicz, D. T. de Jong, J. Kendrick, N. J. J. de Klerk,
H.-Y. Ko, L. N. Kuleshova, X. Li, S. Lohani, F. J. J. Leusen, A. M. Lund,
J. Lv, Y. Ma, N. Marom, A. E. Masunov, P. McCabe, D. P. McMahon, H.
Meekes, M. P. Metz, A. J. Misquitta, S. Mohamed, B. Monserrat, R. J.
Needs, M. A. Neumann, J. Nyman, S. Obata, H. Oberhofer, A. R.
Oganov, A. M. Orendt, G. I. Pagola, C. C. Pantelides, C. J. Pickard, R.
Podeszwa, L. S. Price, S. L. Price, A. Pulido, M. G. Read, K. Reuter, E.
Schneider, C. Schober, G. P. Shields, P. Singh, I. J. Sugden, K.
Szalewicz, C. R. Taylor, A. Tkatchenko, M. E. Tuckerman, F. Vacarro,
M. Vasileiadis, A. Vazquez-Mayagoitia, L. Vogt, Y. Wang, R. E.
Watson, G. A. de Wijs, J. Yang, Q. Zhu, C. R. Groom, Acta Crystallogr.,
Sect. B: Struct. Sci., Cryst. Eng. Mater. 2016, 72, 439-459.
(http://www.chemaxon.com), 2016.
[33] A. Spek, Acta Crystallogr. Sect. D. Biol. Crystallogr. 2009, 65, 148-155.
[34] T. R. Shattock, K. K. Arora, P. Vishweshwar, M. J. Zaworotko, Cryst.
Growth Des. 2008, 8, 4533-4545.
[35] a) C. A. Hunter, Angew. Chem. Int. Ed. 2004, 43, 5310-5324; b) D.
Musumeci, C. A. Hunter, R. Prohens, S. Scuderi, J. F. McCabe, Chem.
Sci. 2011, 2, 883-890.
[36] C. B. Aakeröy, J. Desper, M. M. Smith, Chem. Commun. 2007, 3936-
3938.
[37] A. Kitaigorodsky, Molecular crystals and molecules, Vol. 29, Elsevier,
2012.
[38] G. E. Engel, S. Wilke, O. Konig, K. D. M. Harris, F. J. J. Leusen, J.
Appl. Crystallogr. 1999, 32, 1169-1179.
[39] H. Rietveld, J. Appl. Crystallogr. 1969, 2, 65-71.
[40] a) S. Mohamed, Integrating Computed Crystal Energy Landscapes in
Crystal Form Discovery and Characterisation, in Understanding
Intermolecular Interactions in the Solid State: Approaches and
Techniques, D. Chopra, Monographs in Supramolecular Chemistry, The
Royal Society of Chemistry, 2018, pp. 1-31, 978-1-78801-079-5,
10.1039/9781788013086-00001; b) J. Nyman, G. M. Day,
CrystEngComm 2015, 17, 5154-5165.
[
[
22] S. Mohamed, A. A. Alwan, T. Friščić, A. J. Morris, M. Arhangelskis,
Faraday Discuss. 2018, 211, 401-424.
23] a) R. Taylor, P. A. Wood, Chem Rev 2019, 119, 9427-9477; b) C. R.
Groom, I. J. Bruno, M. P. Lightfoot, S. C. Ward, Acta Crystallogr., Sect.
B: Struct. Sci., Cryst. Eng. Mater. 2016, 72, 171-179.
[41] A. Mukherjee, G. R. Desiraju, Cryst. Growth Des. 2014, 14, 1375-1385.
[42] J. Tao, J. P. Perdew, V. N. Staroverov, G. E. Scuseria, Phys. Rev. Lett.
2003, 91, 146401.
[
[
[
[
24] B. Sarma, N. K. Nath, B. R. Bhogala, A. Nangia, Cryst. Growth Des.
[43] S. Grimme, J. Antony, S. Ehrlich, H. Krieg, J. Chem. Phys. 2010, 132,
154104.
2
009, 9, 1546-1557.
25] T. Friščić, S. L. Childs, S. A. A. Rizvi, W. Jones, CrystEngComm 2009,
1, 418-426.
[44] N. Issa, P. G. Karamertzanis, G. W. A. Welch, S. L. Price, Cryst.
Growth Des. 2009, 9, 442-453.
1
26] S. Mohamed, D. A. Tocher, M. Vickers, P. G. Karamertzanis, S. L.
Price, Cryst. Growth Des. 2009, 9, 2881-2889.
[45] C. R. Taylor, G. M. Day, Cryst. Growth Des. 2018, 18, 892-904.
[46] A. J. Cruz-Cabeza, S. M. Reutzel-Edens, J. Bernstein, Chem. Soc.
Rev. 2015, 44, 8619-8635.
27] a) S. Aitipamula, R. Banerjee, A. K. Bansal, K. Biradha, M. L. Cheney,
A. R. Choudhury, G. R. Desiraju, A. G. Dikundwar, R. Dubey, N.
Duggirala, P. P. Ghogale, S. Ghosh, P. K. Goswami, N. R. Goud, R. R.
K. R. Jetti, P. Karpinski, P. Kaushik, D. Kumar, V. Kumar, B. Moulton,
A. Mukherjee, G. Mukherjee, A. S. Myerson, V. Puri, A. Ramanan, T.
Rajamannar, C. M. Reddy, N. Rodriguez-Hornedo, R. D. Rogers, T. N.
G. Row, P. Sanphui, N. Shan, G. Shete, A. Singh, C. C. Sun, J. A.
[47] A. J. Cruz-Cabeza, CrystEngComm 2012, 14, 6362-6365.
[48] Bruker APEX3 v2019.5-2, Bruker AXS Inc., Madison, Wisconsin, USA,
2015.
[49] Bruker SAINT v8.34A, Bruker AXS Inc., Madison, Wisconsin, USA,
2013.
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
[
[
[
[
[
[
50] Bruker XPREP v2014/2, Bruker AXS Inc., Madison, Wisconsin, USA,
014.
51] Bruker SADABS v2014/5, Bruker AXS Inc., Madison, Wisconsin, USA,
014.
52] G. Sheldrick, Acta Crystallogr. Sect. A: Found. Crystallogr. 2015, 71, 3-
G. A. Voth, P. Salvador, J. J. Dannenberg, S. Dapprich, A. D. Daniels,
Ö. Farkas, J. B. Foresman, J. V. Ortiz, J. Cioslowski, D. J.
Fox,Gaussian, Inc., Wallingford, CT, USA, 2009.
2
2
[62] M. Habgood, I. J. Sugden, A. V. Kazantsev, C. S. Adjiman, C. C.
Pantelides, J. Chem. Theory Comput. 2015, 11, 1957-1969.
[63] I. J. Sugden, C. S. Adjiman, C. C. Pantelides, Acta Crystallogr., Sect. B:
Struct. Sci., Cryst. Eng. Mater. 2019, 75, 423-433.
8.
53] G. Sheldrick, Acta Crystallogr. Sect. C: Cryst. Struct. Commun. 2015,
1, 3-8.
7
[64] a) L.-Y. Hsu, D. E. Williams, Acta Crystallogr. Sect. A: Found.
Crystallogr. 1980, 36, 277-281; b) S. R. Cox, L.-Y. Hsu, D. E. Williams,
Acta Crystallogr. Sect. A: Found. Crystallogr. 1981, 37, 293-301; c) D.
E. Williams, S. R. Cox, Acta Crystallogr., Sect. B: Struct. Sci., Cryst.
Eng. Mater. 1984, 40, 404-417.
54] O. V. Dolomanov, L. J. Bourhis, R. J. Gildea, J. A. K. Howard, H.
Puschmann, J. Appl. Crystallogr. 2009, 42, 339-341.
55] C. F. Macrae, I. J. Bruno, J. A. Chisholm, P. R. Edgington, P. McCabe,
E. Pidcock, L. Rodriguez-Monge, R. Taylor, J. v. Streek, P. A. Wood, J.
Appl. Crystallogr. 2008, 41, 466-470.
[65] D. S. Coombes, S. L. Price, D. J. Willock, M. Leslie, J. Phys. Chem.
1996, 100, 7352-7360.
[
[
56] L. J. Barbour, J. Supramol. Chem. 2001, 1, 189-191.
57] Persistence of Vision Raytracer v3.6, Persistence of Vision Pty. Ltd.
[66] T. Beyer, S. L. Price, The Journal of Physical Chemistry B 2000, 104,
2647-2655.
(
http://www.povray.org/), Williamstown, Victoria, Australia., 2004.
58] P.-E. Werner, L. Eriksson, M. Westdahl, J. Appl. Crystallogr. 1985, 18,
67-370.
[
[67] S. L. Price, M. Leslie, G. W. A. Welch, M. Habgood, L. S. Price, P. G.
Karamertzanis, G. M. Day, PCCP 2010, 12, 8478-8490.
[68] a) A. J. Stone, Chem. Phys. Lett. 1981, 83, 233-239; b) A. J. Stone, M.
Alderton, Mol. Phys. 1985, 56, 1047-1064.
3
[
[
[
59] M. Neumann, J. Appl. Crystallogr. 2003, 36, 356-365.
60] Dassault Systèmes BIOVIA, San Diego: Dassault Systèmes, 2014.
61] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb,
J. R. Cheeseman, G. Scalmani, V. Barone, B. Mennucci, G. A.
[69] A. J. Stone, J. Chem. Theory Comput. 2005, 1, 1128-1132.
[70] A. V. Kazantsev, P. G. Karamertzanis, C. S. Adjiman, C. C. Pantelides,
J. Chem. Theory Comput. 2011, 7, 1998-2016.
Petersson, H. Nakatsuji, M. Caricato, X. Li, H. P. Hratchian, A. F.
Izmaylov, J. Bloino, G. Zheng, J. L. Sonnenberg, M. Hada, M. Ehara, K.
Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O.
Kitao, H. Nakai, T. Vreven, J. A. Montgomery Jr., J. E. Peralta, F.
Ogliaro, M. J. Bearpark, J. Heyd, E. N. Brothers, K. N. Kudin, V. N.
Staroverov, R. Kobayashi, J. Normand, K. Raghavachari, A. P. Rendell,
J. C. Burant, S. S. Iyengar, J. Tomasi, M. Cossi, N. Rega, N. J. Millam,
M. Klene, J. E. Knox, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R.
Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C.
Pomelli, J. W. Ochterski, R. L. Martin, K. Morokuma, V. G. Zakrzewski,
[71] Wavefunction Inc. , 18401 Von Karman Ave., Suite 370, Irvine CA
92612, 2019.
[72] J. Hafner, J. Comput. Chem. 2008, 29, 2044-2078.
[73] a) A. M. Reilly, A. Tkatchenko, J. Chem. Phys. 2013, 139, 024705; b) J.
Moellmann, S. Grimme, J. Phys. Chem. C 2014, 118, 7615-7621.
[74] J. G. Brandenburg, S. Grimme, Acta Crystallogr., Sect. B: Struct. Sci.,
Cryst. Eng. Mater. 2016, 72, 502-513.
This article is protected by copyright. All rights reserved.

Chemistry - A European Journal
10.1002/chem.201904672
FULL PAPER
FULL PAPER
Abeer F. Shunnar, Bhausaheb Dhokale,
Durga Prasad Karothu, David H.
Bowskill, Isaac J. Sugden, Hector H.
Hernandez, Panče Naumov, Sharmarke
Mohamed*
Page No. – Page No.
Mechanosynthesis of ternary molecular ionic cocrystals (ICCs) with significantly
different physicochemical properties was achieved in less than 30 minutes of
grinding. The crystal structures of the ICCs were successfully predicted using
computational methods. The results pave the way for the efficient screening of
higher-order multicomponent crystal forms with functional properties.
Efficient screening for ternary
molecular ionic cocrystals using a
complementary mechanosynthesis
and computational structure
prediction approach
This article is protected by copyright. All rights reserved.