Minimizing polymorphic risk through cooperative computational and experimental exploration

Article abstract of DOI:10.1021/jacs.0c06749

We combine state-of-the-art computational crystal structure prediction (CSP) techniques with a wide range of experimental crystallization methods to understand and explore crystal structure in pharmaceuticals and minimize the risk of unanticipated late-ap

Full text of DOI:10.1021/jacs.0c06749

Subscriber access provided by CORNELL UNIVERSITY LIBRARY
Article
Minimizing polymorphic risk through cooperative
computational and experimental exploration
Christopher R. Taylor, Matthew T. Mulvee, Domonkos S. Perenyi,
Michael R. Probert, Graeme M. Day, and Jonathan W. Steed
J. Am. Chem. Soc., Just Accepted Manuscript • Publication Date (Web): 08 Sep 2020
Downloaded from pubs.acs.org on September 8, 2020
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 15
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
Minimizing polymorphic risk through cooperative computational and
experimental exploration
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Christopher R. Taylor, ‡ a Matthew T. Mulvee,‡ b Domonkos S. Perenyi,b Michael R. Probert,*c Graeme
M. Day*a and Jonathan W. Steed*b
a Computational Systems Chemistry, School of Chemistry, University of Southampton, Southampton, SO17 1NX, UK. E-
mail: g.m.day@soton.ac.uk
b Department of Chemistry, Durham University, South Road, Durham, DH1 3LE, UK. E-mail: jon.steed@durham.ac.uk
c Chemistry, School of Natural and Environmental Sciences, Newcastle University, Newcastle Upon Tyne, NE1 7RU, UK E-
mail: michael.probert@newcastle.ac.uk
KEYWORDS: Polymorph screening, molecular crystals, crystal structure prediction, high-pressure crystallography, free
energy calculations, gel-assisted crystallography
ABSTRACT: We combine state-of-the-art computational crystal structure prediction (CSP) techniques with a wide range of exper-
imental crystallization methods to understand and explore crystal structure in pharmaceuticals and minimize the risk of unanticipated
late-appearing polymorphs. Initially, we demonstrate the power of CSP to rationalize the difficulty in obtaining polymorphs of the
well-known pharmaceutical isoniazid and show that CSP provides the structure of the recently obtained, but unsolved, Form III of
this drug despite there being only a single resolved form for almost 70 years. More dramatically, our blind CSP study predicts a
significant risk of polymorphism for the related iproniazid. Employing a wide variety of experimental techniques, including high-
pressure experiments, we experimentally obtained the first three known non-solvated crystal forms of iproniazid, all of which were
successfully predicted in the CSP procedure. We demonstrate the power of CSP methods and free energy calculations to rationalize
the observed elusiveness of the third form of iproniazid, the success of high-pressure experiments in obtaining it, and the ability of
our synergistic computational-experimental approach to “de-risk” solid form landscapes.
polymorph of ROY9,10 have all been reported. Such a high de-
gree of polymorphism is uncommon, however, and some com-
Introduction
The significance of late-appearing polymorphism
pounds are monomorphic, i.e. they have only one known pure
crystal structure e.g. Pigment Yellow 74,11 fenamic acid12 and
Polymorphism is the existence of multiple crystal structures
the bromo derivative of ROY.13 In the ideal case, this is
with identical chemical compositions for a particular chemical
achieved through close packing to give a dense crystal with all
compound.1–3 Many pharmaceutical drugs display polymor-
potential directional intermolecular interactions being optimal,
phism and the different structures have different physicochem-
with no alternatives of comparable stability.14
ical properties such as solubility, hydration stability, etc., which
However, it can be premature to assume that a compound is
can markedly impact the overall drug efficacy.4 In addition, fac-
monomorphic based on empirical evidence alone, as there exist
tors such as crystal morphology and particle size can influence
a number of examples of late-appearing polymorphs of com-
the drug’s formulation and processing properties.5 Thus, a thor-
ough understanding of the solid form landscape of a particular
pounds previously considered to have only a single form, some-
times with significant consequences. Until very recently, isoni-
compound provides the opportunity to fine-tune material prop-
azid (ISN) had only one resolved solid form, and the difficulty
erties.
in obtaining any other crystal structures, despite the molecule
Polymorph screening, using a wide range of experimental pa-
being known for almost 70 years, led to it being characterised
rameters and procedures, has become routine across the solid-
as monomorphic.15 Recently, two new forms were discovered
state sciences. Famously, McCrone stated that ‘‘the number of
via crystallization from the melt,16 finally revealing its polymor-
forms known for a given compound is proportional to the time
phism and confirming hints of additional forms suggested by
and energy spent in research on that compound.’’6 Indeed, there
are many compounds for which an extraordinary number of
thermomicroscopy experiments decades prior.17
The most famous case of late-appearing polymorphism is that
forms have been discovered. For instance, the tenth form of
of ritonavir, an antiretroviral drug synthesized by Abbott
galunisertib7, the twelfth form of aripiprazole8 and fourteenth
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 2 of 15
Laboratories with a late-appearing polymorph that resulted in a It is well-known that the “energy landscapes” provided by
two-year halt in production and $250 million in lost sales.1,18,19
Bučar et al. have discussed 10 other cases of elusive or disap-
pearing polymorphs.1 For example, the dopamine agonist rotig-
otine, first produced in 1985, was only known to exist in one
crystal structure.20 However, almost 30 years after its discovery,
the appearance of a new crystalline form in the Neupro®
patches stopped its clinical use. The patches were eventually re-
formulated as a stable amorphous matrix, but during this pro-
cess, patches were unavailable for four years. Clearly, it is ex-
tremely desirable to avoid these late-appearing forms and to be
confident that all likely polymorphs of a compound have been
discovered and, ideally, fully characterized.
CSP – the set of predicted stable crystal structures, ranked by
their static lattice energy – feature many more unique structures
than have been observed even for highly-polymorphic mole-
cules like ROY.36 Clearly, not every structure predicted by a
CSP procedure is a feasible polymorph, so to be useful, CSP
must suggest which structures are in the “danger zone” on the
landscape, i.e. those that are likely to crystallize alongside or
instead of the known or desired form(s).
1
2
3
4
5
6
7
8
9
Rationalizing polymorph risk through computed energet-
ics
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The late appearance of thermodynamic forms may be due to
unfavourable crystallization kinetics, but once an energetically
preferred form crystallizes it can inhibit the formation of previ-
ously known metastable forms.21 Interestingly, both ritonavir
and rotigotine are examples of conformational polymorphism,
in which different molecular conformations are adopted in the
polymorphs and the need for conformational change may con-
tribute to the nucleation barrier for a particular form.
To define this “danger zone”, previous work has demon-
strated that computed lattice energy differences between exper-
imentally-observed polymorphs for a wide range of organic
molecules are usually smaller than 2 kJ/mol, less than 7.2
kJ/mol in 95% of cases and extending beyond 10 kJ/mol in only
rare cases.32 These statistics can be applied to assess the likeli-
hood of observing predicted polymorphs of a target molecule
based on their lattice energy difference relative to the lowest-
energy structure. From an experimental perspective, this energy
window could be measured relative to the lowest energy ob-
served structure. However, this makes the interpretation of CSP
results dependent on the extent of experimental screening, and
thus system-dependent and subject to change if a lower energy
structure is observed. With the results of CSP in hand, assuming
that the computational exploration for structures is complete
and that the energetic ranking is accurate, we know the structure
and the energy of the lowest energy possible crystal structure.
This global energy minimum should always be assumed to be
an observable crystal structure, even if the kinetics of crystalli-
zation to this structure are unfavourable. Thus, where CSP has
been performed, the most consistent choice of reference for the
energetic window of polymorphism is the predicted global min-
imum, ensuring that the analysis is applied consistently where
CSP has been performed “blind”, in advance of any known
structures, and to systems with crystal structures already
known.
The discovery of novel forms is aided by increasingly sophis-
ticated crystallization techniques e.g., the templating of the
catemeric Form V of carbamazepine via sublimation onto the
isostructural dihydrocarbamazepine,22 the crystallization of
novel β-coronene under an external magnetic field23 and the
crystallization of two novel forms of adefovir dipivoxil in ionic
liquids.24 These kinds of experiments are unlikely to feature in
a traditional polymorph screen and hence combined experi-
mental and computational modelling approaches using a broad
range of techniques are needed to minimize the risk of a late-
occurring, stable solid form.
It is in this context that computational methods for crystal
structure prediction (CSP) can be employed to understand these
risks.25–28 At a basic level, predicting, enumerating, and ranking
the stable crystalline forms of a given molecule can provide a
picture of what crystal forms are possible and whether the most
highly ranked structures (by a chosen scoring function, typi-
cally the lattice energy) have all already been experimentally
characterized. The results of CSP can, therefore, motivate addi-
tional effort in exploring crystallization conditions.29 At a more
sophisticated level, there is the possibility of guiding the exper-
imental polymorph search, both by predicting new forms’ ex-
istence and by suggesting conditions under which they might be
produced – either to streamline efforts to obtain them or to high-
light conditions that should be avoided to minimize the risk of
their formation.
While the main focus of polymorph risk analysis is normally
the identification of polymorphs that are thermodynamically
more stable than structures that are already known,7,37 higher
energy polymorphs are of interest for a complete understanding
of a molecule’s solid form diversity. Desolvation of solvates has
been demonstrated as a route to high energy polymorphs, but
their instability and difficulties in obtaining high quality crys-
tals of these forms may make them underrepresented in studies
of the energy range of polymorphism, which rely on fully de-
termined crystal structures. The relative energies of such
desolvated structures can vary widely. Forms II and III of
galunisertib, for example, are 8-10 kJ/mol higher in energy than
the (unobserved) CSP global minimum and only accessible via
desolvation.7 In other cases, desolvated structures’ relative en-
ergies can range from +15 kJ/mol38 to +25 kJ/mol39, and even
up to +50 kJ/mol in the case of porous organic frameworks.40
Clearly, if desolvation or other high energy processes for ob-
taining alternative forms are under consideration, then the en-
ergy window considered relevant in CSP must widen signifi-
cantly.
Computational approaches can also offer insight into the like-
lihood of the formation of novel polymorphs under non-ambient
conditions, particularly high pressure, as was performed post
hoc for 2-fluorophenol and 4-chlorophenol,30 and more recently
used a priori to predict high-pressure polymorphs of the phar-
maceutical dalcetrapib.21 However, experimental crystalliza-
tion of high-pressure polymorphs is often not thermodynami-
cally-controlled,31 frustrating direct comparison between com-
puted high-pressure thermodynamics and high-pressure crystal-
lization experiments. Calculation of crystal structure free en-
ergy rather than static lattice energy,32–35 can close the gap be-
tween the simulation environment and experimental conditions,
but these require expensive and sensitive dynamical calcula-
tions.
Regardless, to make any confident prediction of monomor-
phism (or completely characterized polymorphism), the sam-
pling of possible structures must be sufficiently extensive to
ACS Paragon Plus Environment

Page 3 of 15
Journal of the American Chemical Society
have found all the relevant (low-energy) candidates and the predicting “blindly” (i.e. with no prior crystal structure infor-
1
2
3
4
5
6
7
8
method for obtaining their relative energies must be sufficiently
accurate.25 The sampling problem is one of the biggest chal-
lenges of CSP due to the high dimensionality of the search space
and is made even more difficult when molecular flexibility adds
to the number of degrees of freedom. A routine, rudimentary
approach for treating flexibility is to sample the crystal packing
possibilities of multiple molecular conformers generated from
isolated-molecule optimizations. This approach assumes that
the in-crystal conformation is nearly equivalent to a stable gas-
phase conformer, but allows for less stable conformers that
might lead to more favourable packing interactions to be con-
sidered in the CSP procedure. The advantage of this CSP ap-
proach is that the cost of the procedure increases only linearly
with the number of conformers considered; the disadvantage is
that it does not allow for significant conformational distortions
of a gas-phase minimum that might be stabilized (or kinetically-
trapped) by subsequently favourable packing arrangements.
mation) the CSP landscape of ISN and IPN. Armed with this
knowledge, we attempt to crystallize as many as possible of the
forms that appear from the CSP landscapes to be experimentally
“at risk” of formation. We determine this risk based in part on
the 7.2 kJ/mol energy window for likely polymorphism, but
with particular emphasis on locating the global energy mini-
mum predicted structures. Our experimental screen incorpo-
rates a wide range of techniques – typical solvent screening
methods, templated sublimations, gel-phase crystallizations and
high-pressure experiments – to maximise our ability to locate
any elusive, metastable, or previously unobserved polymorphs.
Finally, we combine the information from both approaches to
characterize the observed forms and assess the risk of late-ap-
pearing polymorphism for both compounds.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The severity of this rigid-molecule approximation can be re-
lieved by optimizing the final crystal structures using a method
that allows intramolecular degrees of freedom to relax. The
most commonly-employed such method is periodic density
functional theory (DFT).25 Refinement of predicted structures
with DFT also often provides improved energetic rankings and
computed properties of the structures compared to the force
fields used in the initial structure generation and minimization,
albeit at a considerably increased computational cost.
Figure 1: Molecular structures of isoniazid (ISN) and iproniazid
(IPN).
Computational Methods and CSP Procedure
Generation of hypothetical structures via CSP
In this work, we combine CSP with both traditional and non-
traditional experimental crystallization approaches to explore
the polymorphism of two related molecules, ISN and iproniazid
(IPN), Figure 1. As described, until very recently,16 only one
non-solvated crystal structure of ISN was reported despite
screening; our previous, more targeted attempts at gel-assisted
crystallizations (with two mimetic gelators) as well as micro-
emulsion crystallization experiments did not obtain any new
forms.15 Even so, there were hints of other possible forms (from
the aforementioned thermal microscopy experiments) that had
not been further described in almost fifty years.17 The recent
work of Zhang et al.16 has located two metastable forms (Forms
II and III) of ISN via crystallization from the melt. Of these,
only Form II’s structure could be fully solved, and a unit cell
was proposed for Form III from powder X-ray diffraction data.
The elusiveness of these forms of ISN in previous experiments,
combined with the tractability of ISN from a CSP perspective
as a small, conformationally-rigid molecule, make it an ideal
candidate for further experimental screening combined with
computational study.
Initial molecular conformers for both molecules were gener-
ated via a combined molecular mechanics sampling and DFT
optimization procedure (described more fully in the ESI). This
procedure yielded a single conformer for ISN, and for IPN a set
of five conformers labelled A through E in order of decreasing
relative stability (see ESI for diagrams). All five conformers lay
within 5 kJ/mol of the global gas-phase minimum – i.e. with
conformational energies well within the expected energy
bounds for observed crystal structures.42
For each conformer of each molecule, hypothetical crystal
packing arrangements were generated with rigid molecular ge-
ometries, in our global lattice energy explorer method,43 details
of which are described fully in previous work. The search was
restricted to the 25 most common space groups (see ESI) ob-
served in the Cambridge Structural Database44 for one molecule
in the asymmetric unit (Zʹ=1) and, for isoniazid, also Zʹ=2. Such
restrictions on the complexity of the asymmetric unit are a cru-
cial assumption in CSP for maintaining tractable computational
cost – however, it precludes the identification of higher Zʹ struc-
tures and those in unusual space groups. The recently discov-
ered ISN Form II has unusually low symmetry, with four inde-
pendent molecules in the asymmetric unit (Zʹ=4) and represents
a considerable challenge for any CSP effort, due to the number
of independent degrees of freedom. Given that there was lim-
ited evidence of any polymorphs of ISN at the outset of our
work, a search up to Zʹ=2 represents a reasonable compromise
between exploration and affordability.
The structural analogue IPN was the first monoamine oxidase
inhibitor and the isopropyl group adds to its conformational
flexibility. This may give rise to conformational polymorphism
in the solid state in a manner that is unavailable to ISN, as well
as possible differences in observed hydrogen bonding motifs
due to having one fewer terminal hydrazine hydrogen compared
to ISN. Thus, we were interested in how this chemical change
impacts the crystal structure landscape. Furthermore, IPN has
no reported crystal structures beyond a phosphate salt,41 and is
therefore an excellent candidate with which to undertake a fresh
cooperative experimental and computational polymorph screen
more akin to the screening process in the pharmaceutical indus-
try.
The hypothetical packing arrangements were optimized in a
multi-stage process of successively higher-accuracy energy
minimization methods, progressing from rigid-molecule pair-
wise force field models using the DMACRYS software45 (see
ESI for a complete description) to a periodic DFT optimization
and ranking of the most stable structures. Duplicate structures
were removed by automated comparison of computed PXRD
patterns obtained via the PLATON46 program. For the final
We aim to predict and characterize the polymorph landscapes
for these compounds as completely as possible. We begin by
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 4 of 15
optimization in periodic DFT, we used the PBE functional47 and
solution was used to dissolve the gelator (1 w/v%). Then the
solutions were also left to cool slowly in the heating blocks.
Grimme’s D3 dispersion correction48 with Becke-Johnson
damping (GD3BJ)49 and a plane-wave basis set, using the
VASP50–53 software package. All atomic degrees of freedom
and unit cell parameters were relaxed in the final stage of the
procedure, which introduces a description of the molecular flex-
ibility in response to the crystal packing arrangement.
1
2
3
4
5
6
7
8
Templated sublimations
The powder of the sample being sublimed was placed on a
glass slide and then on a Linkam LTS420 heating stage. The
templating crystal was affixed to a borosilicate glass coverslip
with a small amount of Vaseline and then separated from the
glass slide with a small rubber o-ring. The powders were then
heated to a temperature to achieve sublimation, 131 °C for IPN
and 141 °C for ISN at either 5 or 10°C/min for 6 hours for to
ensure that all the sample sublimes and then the sample was left
at the set temperature overnight to allow for crystal growth.
Predicting free energies
9
The energy landscapes obtained via the above methods are
computed under the assumption of a static crystal structure – no
thermal effects or zero-point energy are included. This is a sig-
nificant approximation, but necessary to manage computational
cost. Once a sufficiently small number of plausible structures
have been identified, it becomes tractable to predict free energy
differences. For ordered crystal structures, the most important
contribution to free energies beyond the static energy arises
from the dynamics of the crystal structure: the zero-point vibra-
tional energy and the phonon modes populated at finite temper-
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
High-pressure experiments
High-pressure experiments were conducted by compressing
crystals that were grown at ambient pressure in a Merrill–Bas-
sett diamond anvil cell (DAC)56 using Fluorinert™ FC-70 as an
inert pressure transmitting fluid. A 250 μm thick stainless steel
gasket was pre-indented to ca. 150 μm and drilled with a 300
μm precision hole to create the sample chamber between the
two diamond anvils, culet size of 800 μm. The pressure inside
the cell was measured after equilibration using a ruby sphere
included in the sample chamber by the R1 ruby fluorescence
method.57 The diamond anvil cell was directly attached to a go-
niometer head and mounted on the diffractometer. Data were
collected using the XIPHOS II diffractometer at Newcastle Uni-
versity, using a four-circle Huber Eulerian goniometer with off-
set chi cradle fitted with a Bruker APEXII CCD area detector
and an Ag− Kα IμS generator. Data collections of crystals in
DACs are poor at locating H-atom positions due to shading by
the gasket and DAC, which reduces the completeness of the da-
taset.58–60
( )
푇 to the
ature, which contribute the vibrational term 퐹
vib
( )
Helmholtz free energy 퐴 푇 :
( )
퐴 푇 = 퐸latt + 퐹
( )
푇
vib
( )
푇 is calculated from the phonon frequencies derived
퐹
vib
from DFT (PBE-GD3BJ). We employed the Phonopy54 pack-
age to obtain the phonon frequencies via finite displacements,
calculating energies in VASP for a supercell of each structure;
the supercell dimensions are chosen to correspond to sampling
reciprocal space q-points of at most 0.12 Å-1 spacing, which has
been demonstrated32,33 to be sufficiently converged for poly-
morph vibrational energy differences. Phonopy employs a var-
iant of the Parlinski-Li-Kawazoe method55 for interpolating be-
tween explicitly sampled q-points. It is critical to ensure that
structures are at true minima to obtain reliable frequencies of
vibration about the atomic equilibrium positions; hence struc-
tures were reoptimized with significantly more stringent con-
vergence criteria (1000 eV basis set cut-off, 0.005 eV/Å in
forces) before the dynamical matrix was calculated.
Novel crystal structures of iproniazid are reported, Forms I
and II were produced by slow cooling and Form III was pro-
duced by high-pressure experiments. A more detailed analysis
of these novel forms, as well details on gelator synthesis, char-
acterization methods and computational processes (molecular
conformer generation, Space group selection for structure gen-
eration and structure minimization) and further results of the
free energy calculations are all available in the ESI.
Free energy calculations were improved by employing the
Debye model to describe the acoustic mode contribution to the
phonon density of states from the Brillouin zone centre to the
nearest sampled q-point.33 We obtain the elastic tensor from the
DFT calculations and calculate an orientationally-averaged ve-
locity of sound in the crystal, from which a Debye frequency
was determined and used to calculate a correction term for long-
Results and Discussion
Isoniazid: crystal structure prediction
( )
wavelength acoustic contributions to 퐹 푇 . Details are pro-
vib
vided in our previous work.33
Experimental Methods
Experimental methods were undertaken continuously in
unison with the computational studies.
Solution crystallizations and gel phase crystallizations
Solution crystallizations were performed by the heating of a
saturated solution of either ISN or IPN until completely dis-
solved. The solutions were left to cool slowly in a heating block.
These were carried out in parallel with gel-phase crystalliza-
tions under the same conditions, but in which the heated
ACS Paragon Plus Environment

Page 5 of 15
Journal of the American Chemical Society
Such a clear thermodynamic preference for an experimen-
tally observed structure in a CSP landscape is unusual61,62
1
–
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
molecules typically exhibit multiple low-lying structures on the
CSP landscape well inside the energy window of risk (even if
these structures have not been observed), and in cases where a
crystal structure is known a priori, it is not necessarily the
global lattice energy minimum (which may indicate a risk of
spontaneous formation of this more stable polymorph). The ex-
istence of Form I as the global minimum in lattice energy and
the size of the energy gap between it and other structures (at this
level of theory and within the space groups and Zʹ values con-
sidered) helps to rationalize the previously long-held empirical
conclusion that this molecule had only one accessible crystal
structure. If ISN were a novel compound undergoing screening,
we would suggest on this basis that the risk of late-appearing,
stable polymorphs would be low, particularly if high-energy
processes such as desolvation of solvatomorphs are not being
considered.
Form III
Form II
Form I
Figure 2: The CSP landscape for isoniazid (ISN), in which the low-
est-energy predicted structure (circled in solid black) matches the
longest-known experimentally-observed form. Orange crosses in-
dicate structures generated assuming only one isoniazid molecule
in the asymmetric unit (Zʹ=1), while blue triangles indicate Zʹ=2.
The global energy minimum structure was located in searches with
Z`=1 and Z`=2 (the two points at –134 kJ/mol, 1.48 g/cm3). The
pink circle shows where the recently discovered Zʹ=4 Form II lies
when optimized using the same methods, while the broken circle
indicates the CSP structure whose lattice parameters and computed
PXRD pattern closely match those of the unsolved Form III. Ener-
gies and densities are obtained from PBE+GD3BJ periodic DFT.
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3: Comparison of experimental PXRD pattern (black,
originally presented by Zheng et al.) for Form III of ISN and the
computed PXRD pattern (blue) for the second-lowest energy CSP
structure. Intensities of each pattern have been scaled to the value
of the largest peak in each case. The red arrows highlight peaks in
the experimental pattern that we do not predict but which
correspond to reflections in the PXRD pattern of Form I of ISN.
The CSP results for isoniazid are shown in Figure 2: The CSP
landscape for isoniazid (ISN), in which the lowest-energy pre-
dicted structure (circled in solid black) matches the longest-
known experimentally-observed form. Orange crosses indicate
structures generated assuming only one isoniazid molecule in
the asymmetric unit (Zʹ=1), while blue triangles indicate Zʹ=2.
The global energy minimum structure was located in searches
with Z`=1 and Z`=2 (the two points at –134 kJ/mol, 1.48
g/cm3). The pink circle shows where the recently discovered
Zʹ=4 Form II lies when optimized using the same methods,
while the broken circle indicates the CSP structure whose lattice
parameters and computed PXRD pattern closely match those of
the unsolved Form III. Energies and densities are obtained from
PBE+GD3BJ periodic DFT.. The pronounced global minimum
energy structure accurately reproduces the historically ob-
served, thermodynamically most stable structure Form I, known
since 1954. All of the other structures generated by CSP lie at
least 6.5 kJ/mol higher, at the upper reaches of the expected en-
ergy window for polymorphism. The global minimum of the
Zʹ=2 search is also isostructural to Form I, but in a lower-sym-
metry space group – relaxing the symmetry constraints has not
yielded any lower-energy structures other than Form I. Dupli-
cate structures occurring within the Zʹ sets have been removed,
but duplicates between sets (e.g. Form I, the global minima in
each) have been retained to emphasise that removing the sym-
metry constraints still locates many of the same low-energy
structures as in the Zʹ=1 search, providing reassurance of suffi-
ciently thorough sampling.
The recent work of Zhang et al16 in obtaining two new isoni-
azid polymorphs (Forms II and III) occurred contemporane-
ously with the present study, and refutes these previous conclu-
sions of monomorphism. Form III is a highly metastable poly-
morph that converts rapidly to Form I and proved too short-
lived for full characterization. The unit cell proposed for Form
III (3.931(2) Å, 9.754(5) Å, 8.568(4) Å) from synchrotron pow-
der diffraction is in excellent agreement with those of the sec-
ond-lowest energy predicted structure (a = 3.751 Å, b = 17.164
Å, c = 9.692 Å, β= 95.67, P21/c, Zʹ=1), i.e. the most plausible
polymorphic candidate from CSP (Figure 2, broken circle, 6.5
kJ/mol above the global minimum), apart from cell doubling in
one direction (b from our CSP structure is double c from PXRD
indexing).16 The computed PXRD pattern for our CSP structure
is also in good agreement with the experimental pattern (Figure
3); reflections present in the experimental pattern but absent
from our prediction (noted by red arrows in Figure 3) corre-
spond to Form I reflections, due to conversion during sample
preparation (see supplementary information of ref. 16).
Hence, we propose that our predicted structure corresponds to
the experimental Form III.
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 6 of 15
Form II, with Zʹ=4, would be unlikely to be found in a typical energy landscape as an indication of multiple likely, stable pol-
1
2
3
4
5
6
7
8
blind CSP attempt. Lattice energy minimization of the structure
reported by Zhang et al. places Form II 5.5 kJ/mol above the
global minimum (Figure 2, pink circle), within the expected en-
ergy range for polymorphism. The high calculated energies of
Forms II and III, relative to Form I, agree with the experimen-
tally observed metastability of these polymorphs, which trans-
form rapidly to Form I, and the energetic ordering agrees with
the room temperature stability ranking (I > II > III) from exper-
iment.16
ymorphs of IPN.
Comparison of predicted hydrogen bond motifs
Isoniazid: The global minimum structure of ISN from the
CSP landscape, which matches the experimental structure Form
I, features hydrogen bonding between the pyridyl group of one
ISN molecule and the terminal amine group of another, forming
a 퐶11(8) chain (Figure 5a). These chains are in turn linked to
each other with N-H∙∙∙N bonds between the terminal amine and
the hydrazide nitrogen atom in an adjacent molecule. However,
the carbonyl group does not participate in any H-bonding in
Form I.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Iproniazid: crystal structure prediction
The CSP structure matching Form III, in contrast, features the
same terminal amine N-H∙∙∙N(pyridyl) chains instead linked by
(hydrazide)N-H∙∙∙O(carbonyl) bonds (Figure 5c). While this
and several of the predicted higher-energy structures feature H-
bonding motifs that involve the carbonyl group as an acceptor
(always in a carbonyl-hydrazide H-bond, sometimes sharing the
oxygen atom with a carbonyl-amine H-bond), it appears that in-
volving the carbonyl is unnecessary to form a particularly stable
crystal structure of ISN, in opposition to what might be ex-
pected based on Etter’s rules.63
Form II
Form I
Form III
Figure 4: The CSP landscape for iproniazid (IPN). Data point color
denotes space group number, while the shapes of points indicate
the gas-phase conformer used to generate the initial crystal struc-
ture, with A being the gas-phase global minimum conformation. In
subsequent crystallization experiments, Forms I and II of IPN
matched the structures in the solid and dashed black circles, respec-
tively. The global static lattice energy minimum, in the dotted cir-
cle, would eventually be located via high-pressure experiments as
Form III. Energies and densities are obtained from PBE+GD3BJ
periodic DFT.
The CSP results for IPN, with Z' = 1 structures generated 5
distinct molecular conformers, are shown in Figure 4. As is
common in blind CSP, we posit that the global minimum energy
structure (in space group P21/c), which is also the global density
maximum, is the most likely to be observed experimentally.26
A close-lying predicted structure (space group P21) is approxi-
mately 1 kJ/mol higher in energy, and we propose as a likely
polymorph. These are followed by a gap, then a more varied,
nearly continuous distribution of possible structures starting ap-
proximately 5 kJ/mol above the global minimum.
Figure 5: Hydrogen bonding for (a) ISN-I, (b) ISN-II, (c) ISN-III
and (d) the global minimum energy structure of IPN. Blue lines in-
dicate intermolecular hydrogen bond interactions.
Form II of ISN displays noticeably different hydrogen bond-
ing to Form I or Form III in a more complex arrangement, ow-
ing to its low symmetry (Zʹ=4). Only 2 of the 4 symmetry-in-
equivalent molecules exhibit pyridyl-acceptor H-bonding, both
via a neighbouring hydrazide N-H donor rather than pyridyl-
amine chains. However, in all 4 molecules the H-bonding in-
volves all the available protons (terminal amine and hydrazide),
and the carbonyl oxygen (Figure 5b).
All but two of the 13 structures within the characteristic pol-
ymorphic energy window of 7.2 kJ/mol feature either the gas-
phase minimum conformer A or the 4th-lowest energy con-
former D, indicating that the latter can compensate for its higher
conformational energy with improved intermolecular interac-
tions or denser packing. Indeed, the global minimum energy
(and maximum density) crystal structure features conformer D.
Unlike in the case of ISN, there is an alternative structure (fea-
turing conformer A) within a very small energy gap of the
global minimum. While not every static lattice energy mini-
mum on a CSP landscape is guaranteed to be an observable pol-
ymorph,36 we consider the presence of two predicted structures
based on different molecular conformers at the bottom of the
Iproniazid: In contrast with ISN, many of the low-energy
predicted structures (e.g. 12 of 13 structures within the 7.2
kJ/mol window) display hydrogen bonding involving the car-
bonyl group (most commonly to the hydrazide N-H), suggest-
ing that the carbonyl acting as an acceptor is energetically opti-
mal. The variety of H-bonding patterns available in predicted
structures of IPN is reduced compared to ISN by the lack of the
ACS Paragon Plus Environment

Page 7 of 15
Journal of the American Chemical Society
accessible terminal amine group, e.g. the pyridyl group appears
less likely to participate in H-bonding due to the isopropyl
1
2
3
4
group precluding the formation of end-to-end molecular chains.
Instead the second N-H typically H-bonds to other infinite H-
bonding chains of IPN (Figure 5d).
5
6
7
Experimental crystallization of ISN and IPN
The crystallization of ISN was carried out in 26 solvents via
slow cooling of saturated solutions and through the slow evap-
oration of ISN solutions. In all cases, the known form of ISN
was produced, as first reported in 1954.64 As a result, sublima-
tion experiments were also undertaken as they have previously
been used to crystallize metastable forms.65–67 ISN was sub-
limed by heating on a microscope stage below its melting point
(171.4°C) at relatively high heating rates (5 and 10 °C/min). A
borosilicate glass coverslip was placed above the sample and
the ISN vapour crystallized on the colder coverslip. However,
no new forms were obtained.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 6: Photos of IPN crystals (a) Form I grown from slow cool-
ing a saturated nitrobenzene solution and (b) Form II produced via
the sublimation of IPN powder on a borosilicate coverslip. Unit cell
(c) Form I and (d) Form II.
For iproniazid, slow cooling and slow evaporation in 22 sol-
vents resulted in 14 samples exhibiting diffraction quality crys-
tals. The crystals are all colourless and plate-like (Figure 6).
Single crystal X-ray diffraction shows that almost all of the
crystals are of the same, new form, designated Form I (IPN-I),
in the monoclinic space group P21.
Gel-phase crystallizations
Gel phase crystallization is an emerging technique for ex-
panding the polymorphism search space70,71 and was undertaken
for ISN and IPN with a series of gelators (Figure 7). Gelator 1
mimics the structure of ISN and IPN and parallels the use of
other drug-mimetic gelators that have previously been shown to
stabilize metastable forms over the thermodynamically fa-
voured polymorph,70 as well as resulting in the discovery of new
solvates.72 Gelators 2 and 3 were chosen to mimic co-formers
known to form co-crystals with ISN. The co-crystals of ISN are
polymorphic and exhibit different H-bonding motifs with the
carbonyl and pyridyl groups both being H-bond acceptors.73–75
Thus, these gelators may be able to template forms with differ-
ent H-bonding motifs. Gelator 4 forms gels in aromatic solvents
and therefore was chosen in an attempt to recover Form II in
toluene.
In one instance, crystallization from slow cooling of a satu-
rated toluene solution of IPN produced a second form of IPN,
designated Form II (IPN-II), in space group Pbca; as with IPN-
I, the structure has Zʹ=1. Despite multiple attempts, it was not
possible to reproduce this form through solvent crystallization
methods, suggesting that IPN-II is likely to have a high barrier
to nucleation in solution, is metastable with respect to or is out-
grown by IPN-I.68 Indeed, slurry experiments, in which both
forms were added to a saturated solution, resulted in only IPN-
I, as confirmed by PXRD (Fig. S1), demonstrating that IPN-I is
the more stable form under ambient conditions. IPN-II was,
however, reproducibly obtained by the sublimation of IPN pow-
der onto borosilicate coverslips.
Comparison of the structural information and geometric
overlays using Mercury69 (Fig. S2) of both IPN Forms I and II
with the CSP structures reveals excellent matches with two pre-
dicted structures, located 5.8 (Form II) and 1.1 (Form I) kJ/mol
above the global minimum. Therefore, both these forms of IPN
were correctly predicted by CSP, with energies consistent with
the statistics for observed polymorphs. The metastable IPN-II
is located higher in the energy landscape, while IPN-I matches
the second-lowest energy predicted structure; thus, the pre-
dicted energetic ordering of these two forms is consistent with
the experimental observation that IPN-I is more readily ob-
tained and more stable than IPN-II.
However, these solution and sublimation crystallization ex-
periments did not yield crystals corresponding to the global en-
ergy minimum on the CSP landscape. Furthermore, the land-
scape (Figure 4) contains multiple structures that are of very
similar energy to IPN-II. It is not apparent why IPN-II is exper-
imentally observed while similarly metastable structures and
the global minimum are not obtained. Both observations indi-
cate that the risk of late-appearing polymorphism of IPN re-
mains after conventional solvent screening and sublimation.
Figure 7: Chemical structures of gelators used in this study.
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 8 of 15
These experiments were carried out at the same solute con- under higher pressure, suggesting high pressure crystallization
1
2
3
4
5
6
7
8
centration as the slow cooling crystallizations. For each experi-
ment, either ISN or IPN and then the gelator were dissolved
with heating and sonication, and left to cool under ambient con-
ditions. In all cases, Form I of both ISN and IPN were obtained,
demonstrating the strong preference for these forms to crystal-
lize over any other predicted structures.
as a means to obtain this form experimentally.
Geometry optimizations under pressure show that the energy
difference between IPN-I and the global minimum CSP struc-
ture widens from -0.7 at zero applied pressure to -7.6 kJ/mol at
P = 2.4 GPa (Figure 8). This is expected, as the global minimum
is already the densest predicted structure and therefore no
“cross-over” in ranking is expected with increased pressure.
While this trend agrees with physical intuition, it provides no
clarity as to why this global minimum structure is not found in
conventional crystallization experiments. Evidently, static lat-
tice energy calculations alone are not enough to rationalize the
elusiveness of the high-density form.
Templated sublimations
9
Previous work has demonstrated that it is possible to template
the growth of a particular form and discover new polymorphs
by subliming onto different surfaces, e.g. polycrystalline pow-
ders,76 siloxane-coated glass65 and crystals with related struc-
ture.32,33. Both Forms I and II of IPN display different hydrogen
bonding patterns to the known form of ISN and hence crystals
of one compound could be used to template the growth of new
forms of the other analogue.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Computed free energies of iproniazid
Given the increased possibility of obtaining the elusive high-
density predicted structure of IPN under elevated pressures, a
more useful quantity to compute is the free energy change as a
function of both temperature and pressure. However, the latter
is a computationally expensive proposition, as each step in pres-
sure requires new phonons to be calculated as the crystal struc-
ture is compressed. To provide some estimate of the free energy
difference between the forms under higher pressure, we make
the approximation that the pressure and vibrational effects are
independent and purely additive, i.e. we compute vibrational
contributions to the free energy only for the ambient pressure
structures, and add these to the PV contribution. These approx-
imate free energies are shown in Figure 8.
In the CSP landscapes, the lowest energy IPN structure with
ISN-like H-bonding is 7.0 kJ/mol above the global minimum.
Conversely, on the ISN landscape, the lowest energy structure
displaying IPN-like H-bonding (i.e. involving the carbonyl
group) is the CSP structure that we propose matches Form III
of ISN, 6.5 kJ/mol higher in energy than Form I. While these
energy differences are towards the higher end of the energy
range of expected polymorphism based on experimental statis-
tics, they remain plausible risks. This assessment is borne out
in the case of ISN with the recent experimental production of
Form III, but is also significant for IPN, as the observable Form
II is only 1.2 kJ/mol more stable than the aforementioned low-
est-energy ISN-like structure.
Sublimation of each compound was attempted using crystals
of either ISN or IPN as a template for the other by crystalliza-
tion directly from the vapour phase. Crystals of both com-
pounds did grow on top of the surface of the parent template
(Fig. S3) but in both cases the same polymorph was produced
(Form I ISN or Form II IPN) as sublimation crystallization in
the absence of template. These experiments further qualify the
risk of further metastable forms of ISN/IPN with unusual H-
bonding pattern; these calculated forms -- along with the global
minimum form of IPN -- appear to be inaccessible through these
methods.
We also attempted to seed melt crystallisations with using ei-
ther ISN or IPN as a template for the other. However, we did
not observe changes to crystal morphology (using polarised op-
tical microscopy) or changes of melting temperatures during the
reheating measurements and concluded that seeding was unsuc-
cessful. Hence, we did not pursue such experiments further.
Figure 8: The computed energy difference, ΔE, between IPN-I and
the global minimum in lattice energy (reference value) as a function
of pressure. Grey squares are static lattice energies including the
PV contribution; the remaining data are Helmholtz free energies
presented at three different temperatures: 0 K (black circles) i.e.
only ZPE contributions, 100 K (blue crosses), and 300 K (red diag-
onal crosses). Values were calculated via periodic PBE+GD3BJ;
vibrational contributions were corrected with a Debye model for
low-frequency modes.
An elusive high-density form of iproniazid
The calculated global energy minimum form of IPN is nota-
bly denser than either of the forms experimentally obtained.
While it is not unprecedented that the first-discovered or most
experimentally-accessible structure is not the global energy
minimum on a CSP landscape, the lower energy and signifi-
cantly greater density together single this prediction out for fur-
ther efforts. From the perspective of confidence in solid form
screening, an unobserved CSP global minimum represents a
major risk and a priority for experimental work. A higher-den-
sity polymorph should become comparatively even more stable
With dynamical effects incorporated into the calculations, the
static CSP global minimum becomes metastable with respect to
IPN-I at ambient pressure, being 2.8 kJ/mol higher in free en-
ergy at 300 K. In fact, the missing high-density form of IPN is
computed to be higher in free energy at ambient pressure than
IPN-I across all temperatures; the vibrational zero-point energy
(ZPE) difference of 1.8 kJ/mol between the two forms re-ranks
ACS Paragon Plus Environment

Page 9 of 15
Journal of the American Chemical Society
them even before any thermal contributions. It is known that
High-pressure experiments
1
2
3
4
5
6
7
8
vibrational ZPE can be important in relative polymorph rank-
ing, particularly for some hydrogen-bonded species.77,78 In these
cases, an accurate treatment of lattice dynamics is necessary to
obtain an estimate of the ZPE and hence polymorph rankings
that are consistent with experimental observations, as appears
to be the case for IPN.
The global lattice energy minimum on the IPN landscape
contains the molecule in a less-stable conformation than that of
IPN-I or IPN-II. This conformation may have a sufficiently high
nucleation barrier that it cannot be crystallized by solution
phase, gel phase or sublimation screening as its crystallization
is kinetically hindered. Similar observations were made for ri-
tonavir and rotigotine whereby the thermodynamically fa-
voured forms were conformationally different from the meta-
stable forms and were not discovered for many years.
Having considered both temperature and pressure, the elu-
siveness of the high-density form is more easily explained – this
structure is metastable at standard temperature and pressure
with respect to IPN-I, with a considerably larger energy differ-
ence between structures than the static calculations alone indi-
cated. As expected, increased pressure stabilizes this high-den-
sity form, with higher temperatures requiring higher applied
pressures to make it more favourable than IPN-I. While our
approximation of additivity of the vibrational and PV terms
compounds the uncertainty in the free energy, the results indi-
cate that at 300 K an applied pressure in excess of 1.0 GPa
would be required to make the high-density structure most fa-
vourable.
9
As the CSP static global minimum structure is significantly
denser than IPN-I and IPN-II, it is stabilized by higher pres-
sures, due to the smaller pressure-volume contribution to the
free energy; this is shown by the free energy calculations
(Figure 8). High-pressure experiments are known to be capable
of effecting conformational change in crystal structures.79,80
Thus, both high-pressure recrystallization and compression of
crystals grown at ambient pressures were undertaken in Mer-
rill–Bassett diamond anvil cells (DACs).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
All attempts to recrystallize IPN from solution at various
pressures produced polycrystalline samples. This may be due to
the many surfaces on which nucleation can occur inside the
DAC, e.g. the ruby spheres, the edge of the tungsten gasket, or
the faces of the diamond.81 As a result we turned to compression
of crystals grown at ambient pressure.
Given this re-ordering, we should consider whether other pre-
dicted structures could similarly become competitive with IPN-
I in free energy terms with increasing temperature or pressure.
The expense and sensitivity of calculating the periodic DFT
phonon frequencies makes it desirable to estimate in which
cases re-ordering is likely without carrying out the full calcula-
tion on many structures. Our previous work has demonstrated
that the vibrational contribution to polymorphic free energy dif-
ferences rarely exceeds 2 kJ/mol at room temperature.32 Given
that the third lowest energy predicted structure of IPN is 3.8
kJ/mol higher in static lattice energy than IPN-I, any of the
higher energy structures are considered unlikely to become
more stable in free energy at room temperature, and so we re-
strict our full free energy treatment to IPN-I and the high-den-
sity global minimum structure. (The expected range in magni-
tude of this entropic contribution also justifies our decision to
consider only static lattice energies for ISN, as no predicted
structures are likely to become competitive with Form I at room
temperature.)
At lower pressures (≤0.3 GPa), only slight reductions in the
unit cell dimensions were observed for IPN-II (Table S1). How-
ever, the compression is anisotropic, affecting mostly the b axis.
At higher pressures (ca. 0.5-0.8 GPa), the crystal breaks per-
pendicular to the longest axis, indicating that this form is unable
to compress further or transform to relieve the stress caused by
the elevated pressure.
When IPN-I was compressed up to ca. 2.1 GPa, no changes
to the crystal habit could be observed visually. However, single
crystal X-ray diffraction confirmed that under hydrostatic pres-
sure IPN-I (P21) undergoes a single crystal-to-single crystal
transformation to produce a new structure, IPN-III (P21/c).
Pressure-mediated transformations without the destruction or
dissolution of the crystal are rare but can be achieved for mole-
cules with conformational flexibility.81 Single crystal-to-single
crystal transformations have been achieved for other organic
molecules through conformational changes, e.g. β-glycine to δ-
glycine,82 glutathione-I to glutathione-II,58 and di-p-tolyl disul-
phide α form to β form.83 In the present case this transformation
results in a conformational change such that the structure
closely matches the predicted conformer D. Structural overlay
(Fig. S4) confirms that this new form corresponds to the pre-
dicted global lattice energy minimum structure of IPN.
Assessing the polymorphic risk of the IPN landscape based
on calculated free energies, the global lattice energy minimum
is now less of a risk than indicated by the static calculations
alone. This demonstrates the power of computational work to
identify risks in crystallization processes, but also highlights the
importance of thorough, rigorous application of advanced meth-
ods to provide accurate insight. If assessments of risk are made
on the basis of “black-box” CSP approaches alone, using static
lattice energies (as in Figure 4), then this maximal density struc-
ture would be perceived as a serious risk as a late appearing,
stable polymorph because it is the global minimum on such a
landscape. This risk is diminished somewhat when free energy
(including quantum vibrational effects) is considered, as dy-
namical contributions re-rank the two structures and suggest
IPN-I is in fact most stable. However, the free energy differ-
ence between the two remains well within the “danger zone” of
polymorphism at ambient pressure, and hence it is prudent to
explore whether high-pressure experiments can indeed obtain
this form as the free energy trends in Figure 8 indicate.
Upon decompression to ambient pressure, the crystallinity of
the sample was significantly reduced because of damage to the
crystal from compression, decompression and X-ray irradiation.
However, it proved possible to obtain an ambient pressure
structure determination, which while of low precision and data
completeness unambiguously identifies that Form III is retained
after decompression and removal from the DAC. Comparison
of the crystallographic data for the Form III high-pressure data
and the data for the crystal recovered from the DAC are availa-
ble in the ESI (Table S2). This key result indicates that Form III
represents a considerable risk as a late-appearing polymorph, if
created in an industrial setting under milling conditions, for ex-
ample.
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 10 of 15
1
2
Crystal packing analysis
3
4
5
6
7
8
9
Differences in the packing of the polymorphs of IPN explain why
Form I rather than Form II transforms to Form III and why Form
III is the densest polymorph. The three IPN polymorphs exhibit a
similar H-bonding motif - a 퐶11(4) chain with the carbonyl group as
an acceptor and the donor being the N-H adjacent to the carbonyl
of the next molecule (Figure 9 and Figure 10). These are the short-
est contacts for all forms, and they appear as two spikes on the 2D
fingerprint plots of the IPN Hirshfeld surfaces (Figs. S5 and S6),
which were used to visualize the differences in intermolecular in-
teractions. For IPN-I the H-bonding chains run parallel to the b-
axis and only a small compression of the b-axis is observed upon
transition to IPN-III. This is commensurate with only a small re-
duction in the H-bond length in IPN-III (Table S2). Similarly, these
chains run parallel to the a-axis for IPN-II and only a modest re-
duction of this axis is recorded after compression in the DAC (Ta-
ble S1). This is in line with previous studies that the shortest con-
tacts remain unchanged and transformations instead rely on a rear-
rangement of longer contacts.84,85
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 9: Hydrogen bonding for (a) IPN-I, (b) IPN-II and (c) IPN-
III. Blue lines indicate interactions between molecules.
Accordingly, there are some subtle differences between
forms in the interactions of the pyridyl groups in adjacent H-
bonding chains. All IPN forms exhibit a short contact between
the pyridyl groups via a (pyridyl)C-H···N(pyridyl) interaction.
For IPN-I and IPN-II these are 퐶11(3) chains, which stack in a
herringbone-like manner. These contacts in IPN-II are longer
than in IPN-I (Table S3), indicating poorer packing. The com-
pression of IPN-I significantly reduces the a-axis (the c-axis in
IPN-III), such that the pyridyl groups are less offset from one
another. The conformational change also results in the two
pyridyl rings becoming nearly coplanar. These changes trans-
form these contacts to 푅22(6) rings in IPN-III which are notice-
ably shorter than in the other forms (Fig. S8g, Table S3); fur-
ther, the H-bonding chains now run anti-parallel to one another
(Figure 9).
Table 1: Crystallographic data for the three polymorphs of IPN
Crystal Form
Form I
Form II
C9H13N3O
179.219
Form III
C9H13N3O
179.219
Formula
C9H13N3O
179.219
Molecular
weight /g mol–1
Crystal System
Monoclinic
P21
Orthorhombic
Pbca
Monoclinic
P21/c
Space Group
T/K
120
120
291
Pressure/ GPa
ambient
8.1440(8)
5.0966(5)
11.7207(13)
90
ambient
4.9971(7)
16.831(3)
22.850(3)
90
2.21
a/Å
11.240(5)
5.043(3)
15.171(14)
90
b/Å
c/Å
α/°
β/°
107.122(4)
90
90
109.14(7)
90
γ/°
90
V/Å3
464.93(8)
2
1921.7(5)
8
812.4(10)
4
Z
Zʹ
1
1
1
ρcalc /g cm–3
1.280
1.239
1.465
independent re-
flections
3067
[Rint =0.0731]
1577
[Rint =0.0918]
287
[Rint =0.0851]
goodness-of-fit
1.024
1.060
1.082
final R indexes R1 = 0.0501
R1 = 0.0398
R1 = 0.0987
[I ≥ 2σ(I)]
wR2 = 0.1033
wR2 = 0.0857
R1 = 0.0612
wR2 = 0.2524
R1 = 0.1434
final R indexes R1 = 0.0719
[all data]
Figure 10: H-bonding sheets of IPN visualized perpendicular to the
H-bonding chains with alternative sheets colored red and blue. Blue
lines represent interactions between IPN molecules. (a) IPN-I, (b)
IPN-III and (c) IPN-II.
wR2 = 0.1111
wR2 = 0.0945
2011026
wR2 = 0.2863
2011027
CCDC code
2011025
Together, the H-bonding chains and the pyridyl-pyridyl contacts
produce sheets that connect the IPN molecules (Figure 10) in all
three forms. These sheets are additionally held together by (hydra-
zine)N-H···N(hydrazine) H-bond interactions (Fig. S9a and S10).
ACS Paragon Plus Environment

Page 11 of 15
Journal of the American Chemical Society
The shorter pyridyl-pyridyl contacts in IPN-III allow the sheets to
experimental screening process, including solvent-based, gel-
1
2
3
4
5
6
7
8
stack closer together, while the anti-parallel chains and the confor-
mational change allows rotation of the isopropyl group for denser
packing within these sheets (Fig S9d); both effects contribute to
IPN-III being the densest. Further analysis of the packing and
presentation of the Hirshfeld fingerprint plots can be found in the
ESI.
phase and sublimation crystallization, successfully obtained
two of the predicted structures: the stable Form I and metastable
Form II. However, the global minimum on the static energy
landscape, the densest predicted structure, remained elusive de-
spite its predicted thermodynamic stability. It was only through
diamond anvil compression experiments that this structure,
Form III, was obtained experimentally. Detailed free-energy
calculations representing the state-of-the-art in CSP techniques
rationalize the stability relationship between Forms I and III.
Form III is not the global energetic minimum when thermal ef-
fects are considered but is obtainable when experimental
searching is guided by CSP. Once formed by compression,
Form III persists at ambient pressure and hence CSP has re-
vealed a high-risk late appearing polymorph and indicated the
conditions under which it is formed.
In IPN-II, the b-axis is normal to the plane of these sheets
(Fig. S10), and it is this axis along which IPN-II compresses
before breaking under pressure. It is speculated that these
sheets compress together rather than the H-bonding chains to
accommodate the pressure increases.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Though all three forms contain these H-bonded sheets, the shape of
the sheets in IPN-I and IPN-III is notably different from those in
IPN-II. The sheets of both Forms I and III are relatively planar,
with gentle undulations. In contrast, those of Form II are more cor-
rugated, having an increased roughness (rugosity) with neighbour-
ing sheets interpenetrating more (Figure 10). As these sheets in
IPN-II are very different to those in IPN-III, there is no obvious
route for transformation to IPN-III, in contrast to IPN-I which only
needs to undergo a comparatively small change in the sheet struc-
ture. Topological rugosity has been linked to slip planes for crys-
tals which affect the mechanical properties and thus the tabletabil-
ity of different forms.86,87 Therefore, it is expected that the different
forms of IPN may behave differently during formulation, particu-
larly if tabletting pressures cause the transformation from IPN-I to
IPN-III that can persist at ambient pressure. Given the structural
differences, however, we do not expect IPN-II to transform due to
pressure during the tableting process.
This work demonstrates the power of combining exhaustive
experimental screening with modern CSP methods to elucidate
the risk of late-appearing polymorphism. We emphasize the
synergy between the two fields; CSP of iproniazid suggested a
significant risk of polymorphism, which justified a thorough ex-
perimental screening that obtained Forms I and II. When the
most likely structure according to CSP eluded this screening,
further computational analysis via free energy calculations ra-
tionalized its elusiveness, while high-pressure experiments mo-
tivated by the predicted maximal density of this structure suc-
cessfully yielded Form III. The crystallization and characteri-
zation of three polymorphs of iproniazid, for which no pure
crystal structure was previously available, is tangible evidence
of the value of this combined approach for exploring and “de-
risking” solid form landscapes.
Conclusions
We have demonstrated the power of computational CSP
methods to rationalize the difficulty in obtaining polymorphs of
isoniazid, due to the absence of any competitive predicted ther-
modynamic minima (assuming Zʹ≤2). Solvent and sublimation
based crystallization methods did not reveal any further forms
beyond the long-known Form I, consistent with previous
screening.15
ASSOCIATED CONTENT
Extended description of experimental procedures (sublimation,
gelator synthesis, and high-pressure experiments); extended de-
scription of computational CSP procedure (conformer generation
and space group sampling); crystallographic data and analysis
(PXRD patterns, Hirshfeld and packing analysis); crystal structure
files (CIF format). This material is available free of charge via the
Internet at http://pubs.acs.org.
The contemporaneous discovery by Zhang et al. of two new
forms from melt crystallization16 is remarkable, but consistent
with our computational findings in both cases. Form II would
not be predicted in a typical CSP due to its high Z`, making a
search prohibitively expensive in the general case. However,
the structure of Form II is ranked favourably in energy when
treated with our computational methods, and within the range
of likely energy differences between observed polymorphs.
Similarly, the second-lowest energy CSP structure of ISN lies
somewhat higher in energy but still within the usual range for
polymorphism, and matches the PXRD pattern of Form III,
which could not be fully characterized experimentally. There-
fore, although our initial assumptions precluded our CSP pro-
cedure from predicting Form II, Form III was predicted as the
most likely polymorphic structure, albeit with an energy that
suggests it would be challenging to isolate (and arguably agree-
ing with its experimental instability). Thus, our approach suc-
cessfully predicted the only other Zʹ=1 structure of ISN yet dis-
covered and has revealed its crystal structure.
AUTHOR INFORMATION
Corresponding Authors
Graeme M. Day -- Computational Systems Chemistry, School of
Chemistry, University of Southampton, Southampton, SO17 1NX,
UK. E-mail: g.m.day@soton.ac.uk
Jonathan W. Steed -- Department of Chemistry, Durham Univer-
sity, South Road, Durham, DH1 3LE, UK. E-mail:
jon.steed@durham.ac.uk
Michael R. Probert -- Chemistry, School of Natural and Environ-
mental Sciences, Newcastle University, Newcastle Upon Tyne,
NE1 7RU, UK E-mail: michael.probert@newcastle.ac.uk
Author Contributions
‡These authors contributed equally to the completion of this work
and preparation of this manuscript.
More significantly, a blind CSP of the analogue iproniazid
predicted a polymorphic system with at least two notably low-
energy structures and several metastable ones. An exhaustive
Notes
The authors declare no competing financial interest.
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 12 of 15
(13) Lutker, K. M.; Tolstyka, Z. P.; Matzger, A. J. Investigation
Acknowledgements
of a Privileged Polymorphic Motif: A Dimeric ROY Derivative. Cryst.
Growth Des. 2008, 8, 136–139. https://doi.org/10.1021/cg700921w.
(14) Price, S. L. From Crystal Structure Prediction to Polymorph
Prediction: Interpreting the Crystal Energy Landscape. Phys. Chem.
Chem. Phys. 2008, 1996–2009. https://doi.org/10.1039/b719351c.
(15) Kennedy, S. R.; Jones, C. D.; Yufit, D. S.; Nicholson, C. E.;
Cooper, S. J.; Steed, J. W. Tailored Supramolecular Gel and Micro-
emulsion Crystallization Strategies – Is Isoniazid Really Monomor-
1
2
3
4
5
6
7
8
We thank the Engineering and Physical Sciences Research
Council for a project grant (EP/R013373/1). We acknowledge
the use of the IRIDIS High Performance Computing Facility,
and associated support services at the University of Southamp-
ton, in the completion of this work. Via CRT and GMD’s mem-
bership of the UK's HEC Materials Chemistry Consortium,
which is funded by EPSRC (EP/L000202, EP/R029431), this
work used the ARCHER UK National Supercomputing Service
(http://www.archer.ac.uk).
phic?
CrystEngComm
2018,
20,
1390–1398.
https://doi.org/10.1039/C8CE00066B.
(16) Zhang, K.; Fellah, N.; Shtukenberg, A. G.; Fu, X.; Hu, C.;
Ward, M. D. Discovery of New Polymorphs of the Tuberculosis Drug
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
We thank Dr Dmitry S. Yufit, the Department of Chemistry,
Durham University, UK for assistance with diffraction meas-
urements, and Dr Charlie McMonagle, Chemistry, Newcastle
University, UK for assistance with the high-pressure experi-
ments. We thank Prof. Mike Ward, New York University, USA
for provision of PXRD data of Form III of isoniazid for com-
parison.
Isoniazid.
CrystEngComm
2020,
22,
2705–2708.
https://doi.org/10.1039/D0CE00440E.
(17) Kuhnert‐Brandstätter, M. Thermomicroscopy in the Analysis
of Pharmaceuticals; International Series of Monographs in Analytical
Chemistry; Pergamon Press: Oxford, 1971; Vol. 45.
(18) Cruz-Cabeza, A. J.; Reutzel-Edens, S. M.; Bernstein, J.
Facts and Fictions about Polymorphism. Chem. Soc. Rev. 2015, 44,
8619–8635. https://doi.org/10.1039/c5cs00227c.
(19) Bauer, J.; Spanton, S.; Henry, R.; Quick, J.; Dziki, W.; Por-
ter, W.; Morris, J. Ritonavir: An Extraordinary Example of Conforma-
tional Polymorphism. Pharm. Res. 2001, 18, 859–866.
https://doi.org/10.1023/A:1011052932607.
(20) Mortazavi, M.; Hoja, J.; Aerts, L.; Quéré, L.; van de Streek,
J.; Neumann, M. A.; Tkatchenko, A. Computational Polymorph
Screening Reveals Late-Appearing and Poorly-Soluble Form of Rotig-
otine. Commun. Chem. 2019, 2. https://doi.org/10.1038/s42004-019-
0181-9.
(21) Neumann, M. A.; Van De Streek, J.; Fabbiani, F. P. A. A.;
Hidber, P.; Grassmann, & O.; Grassmann, O. Combined Crystal Struc-
ture Prediction and High-Pressure Crystallization in Rational Pharma-
ceutical Polymorph Screening. Nat. Comm. 2015, 6, 7793.
https://doi.org/10.1038/ncomms8793.
References
(1)
Bučar, D. K.; Lancaster, R. W.; Bernstein, J. Disappearing
Polymorphs Revisited. Angew. Chem. Int. Ed. 2015, 54, 6972–6993.
https://doi.org/10.1002/anie.201410356.
(2)
Polymorphism, What It Is and How to Identify It: A Systematic Re-
view. RSC Adv. 2013, 3, 16905–16931.
https://doi.org/10.1039/c3ra41559g.
(3) Bernstein, J. Polymorphism in Molecular Crystals; Oxford
University Press, 2002.
Brog, J.-P. P.; Chanez, C.-L. L.; Crochet, A.; Fromm, K. M.
(4)
Hilfiker, R. Polymorphism: In the Pharmaceutical Industry;
Hilfiker, R., Ed.; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim,
FRG, 2006.
(22) Arlin, J.-B. B.; Price, L. S.; Price, S. L.; Florence, A. J. A
Strategy for Producing Predicted Polymorphs: Catemeric Carbamaze-
(5)
Su, C. S.; Liao, C. Y.; Jheng, W. De. Particle Size Control
pine
Form
V.
Chem.
Comm.
2011,
47,
7074.
and Crystal Habit Modification of Phenacetin Using Ultrasonic Crys-
tallization. Chem. Eng. Technol. 2015, 38, 181–186.
https://doi.org/10.1002/ceat.201300573.
https://doi.org/10.1039/c1cc11634g.
(23) Potticary, J.; Terry, L. R.; Bell, C.; Papanikolopoulos, A. N.;
Christianen, P. C. M.; Engelkamp, H.; Collins, A. M.; Fontanesi, C.;
Kociok-Köhn, G.; Crampin, S.; Da Como, E.; Hall, S. R. An Unfore-
seen Polymorph of Coronene by the Application of Magnetic Fields
during Crystal Growth. Nat. Comm. 2016, 7, 11555.
https://doi.org/10.1038/ncomms11555.
(24) An, J.-H.; Kim, W.-S. Antisolvent Crystallization Using
Ionic Liquids As Solvent and Antisolvent for Polymorphic Design of
Active Pharmaceutical Ingredient. Cryst. Growth Des. 2013, 13, 31–
39. https://doi.org/10.1021/cg300730w.
(6)
McCrone, W. C. Physics and Chemistry of the Organic Solid
State. In Physics and Chemistry of the Organic Solid State; Fox, D.,
Labes, M. M., Weissberger, A., Eds.; Wiley-Interscience: New York,
1965; Vol. 2, 725–767.
(7)
Bhardwaj, R. M.; McMahon, J. A.; Nyman, J.; Price, L. S.;
Konar, S.; Oswald, I. D. H.; Pulham, C. R.; Price, S. L.; Reutzel-Edens,
S. M. A Prolific Solvate Former, Galunisertib, under the Pressure of
Crystal Structure Prediction, Produces Ten Diverse Polymorphs. J. Am.
Chem.
https://doi.org/10.1021/jacs.9b06634.
(8) Zeidan, T. A.; Trotta, J. T.; Tilak, P. A.; Oliveira, M. A.;
Soc.
2019,
141,
13887–13897.
(25) Day, G. M. Current Approaches to Predicting Molecular Or-
ganic Crystal Structures. Crystallography Reviews 2011, 17, 3–52.
https://doi.org/10.1080/0889311X.2010.517526.
Chiarella, R. A.; Foxman, B. M.; Almarsson; Hickey, M. B. An Un-
precedented Case of Dodecamorphism: The Twelfth Polymorph of Ar-
ipiprazole Formed by Seeding with Its Active Metabolite.
CrystEngComm
https://doi.org/10.1039/c5ce02467f.
(9) Levesque, A.; Maris, T.; Wuest, J. D. ROY Reclaims Its
Crown: New Ways to Increase Polymorphic Diversity. J. Am. Chem.
Soc. 2020, 142, 11873–11883. https://doi.org/10.1021/jacs.0c04434.
(10) Tyler, A. R.; Ragbirsingh, R.; McMonagle, C. J.; Waddell,
P. G.; Heaps, S. E.; Steed, J. W.; Thaw, P.; Hall, M. J.; Probert, M. R.
Encapsulated Nanodroplet Crystallization of Organic-Soluble Small
(26) Reilly, A. M.; Cooper, R. I.; Adjiman, C. S.; Bhattacharya,
S.; Boese, A. D.; Brandenburg, J. G.; Bygrave, P. J.; Bylsma, R.; Camp-
bell, J. E.; Car, R.; Case, D. H.; Chadha, R.; Cole, J. C.; Cosburn, K.;
Cuppen, H. M.; Curtis, F.; Day, G. M.; DiStasio Jr, R. A.; Dzyab-
chenko, A.; van Eijck, B. P.; Elking, D. M.; van den Ende, J. A.; Facelli,
J. C.; Ferraro, M. B.; Fusti-Molnar, L.; Gatsiou, C.-A.; Gee, T. S.; de
Gelder, R.; Ghiringhelli, L. M.; Goto, H.; Grimme, S.; Guo, R.; Hof-
mann, D. W. M.; Hoja, J.; Hylton, R. K.; Iuzzolino, L.; Jankiewicz, W.;
de Jong, D. T.; Kendrick, J.; de Klerk, N. J. J.; Ko, H.-Y.; Kuleshova,
L. N.; Li, X.; Lohani, S.; Leusen, F. J. J.; Lund, A. M.; Lv, J.; Ma, Y.;
Marom, N.; Masunov, A. E.; McCabe, P.; McMahon, D. P.; Meekes,
H.; Metz, M. P.; Misquitta, A. J.; Mohamed, S.; Monserrat, B.; Needs,
R. J.; Neumann, M. A.; Nyman, J.; Obata, S.; Oberhofer, H.; Oganov,
A. R.; Orendt, A. M.; Pagola, G. I.; Pantelides, C. C.; Pickard, C. J.;
Podeszwa, R.; Price, L. S.; Price, S. L.; Pulido, A.; Read, M. G.; Reuter,
K.; Schneider, E.; Schober, C.; Shields, G. P.; Singh, P.; Sugden, I. J.;
Szalewicz, K.; Taylor, C. R.; Tkatchenko, A.; Tuckerman, M. E.; Va-
carro, F.; Vasileiadis, M.; Vazquez-Mayagoitia, A.; Vogt, L.; Wang,
Y.; Watson, R. E.; de Wijs, G. A.; Yang, J.; Zhu, Q.; Groom, C. R.
2016,
18,
1486–1488.
Molecules.
Chem
2020,
6,
1755-1765.
https://doi.org/10.1016/j.chempr.2020.04.009.
(11) Braga, D.; Grepioni, F. Crystal Engineering: From Mole-
cules and Crystals to Materials; Springer Netherlands, 1999; 421–441.
(12) Uzoh, O. G.; Cruz-Cabeza, A. J.; Price, S. L. Is the Fenamate
Group a Polymorphophore? Contrasting the Crystal Energy Land-
scapes of Fenamic and Tolfenamic Acids. Cryst. Growth Des. 2012,
12, 4230–4239. https://doi.org/10.1021/cg3007348.
ACS Paragon Plus Environment

Page 13 of 15
Journal of the American Chemical Society
Report on the Sixth Blind Test of Organic Crystal Structure Prediction
Random Sampling. J. Chem. Theory Comput. 2016, 12, 910–924.
https://doi.org/10.1021/acs.jctc.5b01112.
(44) Allen, F. H. The Cambridge Structural Database: A Quarter
of a Million Crystal Structures and Rising. Acta Crystallogr., Sect. B:
Struct. Sci., Cryst. Eng. Mater. 2002, 58, 380–388.
https://doi.org/10.1107/S0108768102003890.
(45) Price, S. L.; Leslie, M.; Welch, G. W. A.; Habgood, M.;
Price, L. S.; Karamertzanis, P. G.; Day, G. M. Modelling Organic Crys-
tal Structures Using Distributed Multipole and Polarizability-Based
Model Intermolecular Potentials. Phys. Chem. Chem. Phys. 2010, 12,
8478–8490. https://doi.org/10.1039/C004164E.
(46) Spek, A. L. Structure Validation in Chemical Crystallog-
raphy. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2009, 65, 148–155.
https://doi.org/10.1107/S090744490804362X.
Methods. Acta Crystallogr., Sect. B: Struct. Sci., Cryst. Eng. Mater.
2016, 72, 439–459. https://doi.org/10.1107/S2052520616007447.
(27) Price, S. L. Is Zeroth Order Crystal Structure Prediction
(CSP_0) Coming to Maturity? What Should We Aim for in an Ideal
Crystal Structure Prediction Code? Faraday Discuss. 2018, 211, 9–30.
https://doi.org/10.1039/C8FD00121A.
(28) Oganov, A. R. Crystal Structure Prediction: Reflections on
Present Status and Challenges. Faraday Discuss. 2018, 211, 643–660.
https://doi.org/10.1039/C8FD90033G.
1
2
3
4
5
6
7
8
(29) Cui, P.; McMahon, D. P.; Spackman, P. R.; Alston, B. M.;
Little, M. A.; Day, G. M.; Cooper, A. I. Mining Predicted Crystal Struc-
ture Landscapes with High Throughput Crystallisation: Old Molecules,
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
New
Insights.
Chem.
Sci.
2019,
10,
9988–9997.
https://doi.org/10.1039/C9SC02832C.
(47) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradi-
ent Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865–
3868. https://doi.org/10.1103/PhysRevLett.77.3865.
(48) Grimme, S.; Antony, J.; Ehrlich, S.; Krieg, H. A Consistent
and Accurate Ab Initio Parametrization of Density Functional Disper-
sion Correction (DFT-D) for the 94 Elements H-Pu. J. Chem. Phys.
2010, 132, 154104. https://doi.org/10.1063/1.3382344.
(49) Grimme, S.; Ehrlich, S.; Goerigk, L. Effect of the Damping
Function in Dispersion Corrected Density Functional Theory. J. Comp.
Chem. 2011, 32, 1456–1465. https://doi.org/10.1002/jcc.21759.
(50) Kresse, G.; Hafner, J. Ab Initio Dynamics for Liquid Metals.
Phys. Rev. B: Condens. Matter Mater. Phys. 1993, 47, 558–561.
https://doi.org/10.1103/PhysRevB.47.558.
(51) Kresse, G.; Hafner, J. Ab Initio Molecular-Dynamics Simu-
lation of the Liquid-Metal–Amorphous-Semiconductor Transition in
Germanium. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 49,
14251–14269. https://doi.org/10.1103/PhysRevB.49.14251.
(52) Kresse, G.; Furthmüller, J. Efficiency of Ab-Initio Total En-
ergy Calculations for Metals and Semiconductors Using a Plane-Wave
(30) Oswald, I. D. H.; Allan, D. R.; Day, G. M.; Motherwell, W.
D. S.; Parsons, S. Realizing Predicted Crystal Structures at Extreme
Conditions:ꢀ The Low-Temperature and High-Pressure Crystal Struc-
tures of 2-Chlorophenol and 4-Fluorophenol. Cryst. Growth Des. 2005,
5, 1055–1071. https://doi.org/10.1021/cg049647b.
(31) Boldyreva, E. High-Pressure Polymorphs of Molecular Sol-
ids: When Are They Formed, and When Are They Not? Some Exam-
ples of the Role of Kinetic Control. Cryst. Growth Des. 2007, 7, 1662–
1668. https://doi.org/10.1021/cg070098u.
(32) Nyman, J.; Day, G. M. Static and Lattice Vibrational Energy
Differences between Polymorphs. CrystEngComm 2015, 17, 5154–
5165. https://doi.org/10.1039/C5CE00045A.
(33) Nyman, J.; Pundyke, O. S.; Day, G. M. Accurate Force
Fields and Methods for Modelling Organic Molecular Crystals at Finite
Temperatures. Phys. Chem. Chem. Phys. 2016, 18, 15828–15837.
https://doi.org/10.1039/C6CP02261H.
(34) Hoja, J.; Ko, H.-Y.; Neumann, M. A.; Car, R.; DiStasio, R.
A.; Tkatchenko, A. Reliable and Practical Computational Description
of Molecular Crystal Polymorphs. Sci. Adv. 2019, 5, eaau3338.
https://doi.org/10.1126/sciadv.aau3338.
(35) Reilly, A. M.; Tkatchenko, A. Understanding the Role of Vi-
brations, Exact Exchange, and Many-Body van Der Waals Interactions
in the Cohesive Properties of Molecular Crystals. J. Chem. Phys. 2013,
139, 024705. https://doi.org/10.1063/1.4812819.
(36) Price, S. L. Why Don’t We Find More Polymorphs? Acta
Crystallogr., Sect. B: Struct. Sci., Cryst. Eng. Mater. 2013, 69, 313–
328. https://doi.org/10.1107/S2052519213018861.
(37) Nyman, J.; Yu, L.; Reutzel-Edens, S. M. Accuracy and Re-
producibility in Crystal Structure Prediction: The Curious Case of
Basis
Set.
Comput.
Mater.
Sci.
1996,
6,
15–50.
https://doi.org/10.1016/0927-0256(96)00008-0.
(53) Kresse, G.; Furthmüller, J. Efficient Iterative Schemes for
Ab Initio Total-Energy Calculations Using a Plane-Wave Basis Set.
Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54, 11169–11186.
https://doi.org/10.1103/PhysRevB.54.11169.
(54) Togo, A.; Tanaka, I. First Principles Phonon Calculations in
Materials
Science.
Scr.
Mater.
2015,
108,
1–5.
https://doi.org/10.1016/j.scriptamat.2015.07.021.
(55) Parlinski, K.; Li, Z. Q.; Kawazoe, Y. First-Principles Deter-
mination of the Soft Mode in Cubic ZrO2. Phys. Rev. Lett. 1997, 78,
4063–4066. https://doi.org/10.1103/PhysRevLett.78.4063.
(56) Merrill, L.; Bassett, W. A. Miniature Diamond Anvil Pres-
sure Cell for Single Crystal X‐ray Diffraction Studies. Rev. Sci. In-
strum. 1974, 45, 290–294. https://doi.org/10.1063/1.1686607.
(57) Piermarini, G. J.; Block, S.; Barnett, J. D.; Forman, R. A.
Calibration of the Pressure Dependence of the R1 Ruby Fluorescence
Line to 195 Kbar. J. Appl. Phys. 1975, 46, 2774–2780.
https://doi.org/10.1063/1.321957.
(58) Moggach, S. A.; Lennie, A. R.; Morrison, C. A.; Richardson,
P.; Stefanowicz, F. A.; Warren, J. E. Pressure Induced Phase Transi-
tions in the Tripeptide Glutathione to 5.24 GPa: The Crystal Structure
of Glutathione-II at 2.94 GPa and Glutathione-III at 3.70 GPa.
CrystEngComm 2010, 12, 2587. https://doi.org/10.1039/c001254h.
(59) Moggach, S. A.; Allan, D. R.; Parsons, S.; Warren, J. E. In-
corporation of a New Design of Backing Seat and Anvil in a Merrill–
Bassett Diamond Anvil Cell. J. Appl. Crystallogr. 2008, 41, 249–251.
https://doi.org/10.1107/S0021889808000514.
ROY.
CrystEngComm
2019,
21,
2080–2088.
https://doi.org/10.1039/C8CE01902A.
(38) Braun, D. E.; McMahon, J. A.; Koztecki, L. H.; Price, S. L.;
Reutzel-Edens, S. M. Contrasting Polymorphism of Related Small
Molecule Drugs Correlated and Guided by the Computed Crystal En-
ergy Landscape. Cryst. Growth Des. 2014, 14, 2056–2072.
https://doi.org/10.1021/cg500185h.
(39) Cruz-Cabeza, A. J.; Day, G. M.; Jones, W. Predicting Inclu-
sion Behaviour and Framework Structures in Organic Crystals. Chem.
-
Eur.
J.
2009,
15,
13033–13040.
https://doi.org/10.1002/chem.200901703.
(40) Pulido, A.; Chen, L.; Kaczorowski, T.; Holden, D.; Little,
M. A.; Chong, S. Y.; Slater, B. J.; McMahon, D. P.; Bonillo, B.; Stack-
house, C. J.; Stephenson, A.; Kane, C. M.; Clowes, R.; Hasell, T.;
Cooper, A. I.; Day, G. M. Functional Materials Discovery Using En-
ergy–Structure–Function Maps. Nature 2017, 543, 657–664.
https://doi.org/10.1038/nature21419.
(41) Chieh, P. C.; Palenik, G. J. Crystal Structure of Iproniazid
(60) Katrusiak, A. High-Pressure Crystallography. Acta Crystal-
Phosphate.
J.
Chem.
Soc.
A
1971,
576–579.
logr.,
Sect.
A:
Found.
Adv.
2008,
64,
135–148.
https://doi.org/10.1039/J19710000576.
https://doi.org/10.1107/S0108767307061181.
(42) Thompson, H. P. G.; Day, G. M. Which Conformations
Make Stable Crystal Structures? Mapping Crystalline Molecular Ge-
ometries to the Conformational Energy Landscape. Chem. Sci. 2014, 5,
3173–3182. https://doi.org/10.1039/C4SC01132E.
(43) Case, D. H.; Campbell, J. E.; Bygrave, P. J.; Day, G. M.
Convergence Properties of Crystal Structure Prediction by Quasi-
(61) Day, G. M.; Chisholm, J.; Shan, N.; Motherwell, W. D. S.;
Jones, W. An Assessment of Lattice Energy Minimization for the Pre-
diction of Molecular Organic Crystal Structures. Cryst. Growth Des.
2004, 4, 1327–1340. https://doi.org/10.1021/cg0498148.
(62) Day, G. M.; Motherwell, W. D. S.; Jones, W. Beyond the
Isotropic Atom Model in Crystal Structure Prediction of Rigid
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 14 of 15
Molecules:ꢀ Atomic Multipoles versus Point Charges. Cryst. Growth
(75) Grobelny, P.; Mukherjee, A.; Desiraju, G. R. Drug-Drug Co-
Des. 2005, 5, 1023–1033. https://doi.org/10.1021/cg049651n.
(63) Etter, M. C. Encoding and Decoding Hydrogen-Bond Pat-
terns of Organic Compounds. Acc. Chem. Res. 1990, 23, 120–126.
https://doi.org/10.1021/ar00172a005.
(64) Jensen, L. H. The Crystal Structure of Isonicotinic Acid Hy-
drazide1. J. Am. Chem. Soc. 1954, 76, 4663–4667.
https://doi.org/10.1021/ja01647a053.
Crystals: Temperature-Dependent Proton Mobility in the Molecular
Complex of Isoniazid with 4-Aminosalicylic Acid. CrystEngComm
2011, 13, 4358–4364. https://doi.org/10.1039/c0ce00842g.
(76) Kamali, N.; O’Malley, C.; Mahon, M. F.; Erxleben, A.;
McArdle, P. Use of Sublimation Catalysis and Polycrystalline Powder
Templates for Polymorph Control of Gas Phase Crystallization. Cryst.
1
2
3
4
5
6
7
8
Growth
Des.
2018,
18,
3510–3516.
(65) Solomos, M. A.; Capacci-Daniel, C.; Rubinson, J. F.; Swift,
J. A. Polymorph Selection via Sublimation onto Siloxane Templates.
https://doi.org/10.1021/acs.cgd.8b00268.
(77) Rivera, S. A.; Allis, D. G.; Hudson, B. S. Importance of Vi-
brational Zero-Point Energy Contribution to the Relative Polymorph
Energies of Hydrogen-Bonded Species. Cryst. Growth Des. 2008, 8,
3905–3907. https://doi.org/10.1021/cg800524d.
(78) Heit, Y. N.; Beran, G. J. O. How Important Is Thermal Ex-
pansion for Predicting Molecular Crystal Structures and Thermochem-
istry at Finite Temperatures? Acta Crystallogr., Sect. B: Struct. Sci.,
Cryst.
Growth
Des.
2018,
18,
6965–6972.
9
https://doi.org/10.1021/acs.cgd.8b01151.
(66) Liu, Z.; Li, C. Solvent-Free Crystallizations of Amino Ac-
ids: The Effects of the Hydrophilicity/Hydrophobicity of Side-Chains.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Biophys.
Chem.
2008,
138,
115–119.
https://doi.org/10.1016/j.bpc.2008.09.011.
(67) Liu, Z.; Zhong, L.; Ying, P.; Feng, Z.; Li, C. Crystallization
of Metastable β Glycine from Gas Phase via the Sublimation of α or γ
Form in Vacuum. Biophys. Chem. 2008, 132, 18–22.
https://doi.org/10.1016/j.bpc.2007.10.003.
(68) Yu, L. Survival of the Fittest Polymorph: How Fast Nu-
cleater Can Lose to Fast Grower. CrystEngComm 2007, 9, 847–851.
https://doi.org/10.1039/b709260c.
(69) Macrae, C. F.; Sovago, I.; Cottrell, S. J.; Galek, P. T. A.;
McCabe, P.; Pidcock, E.; Platings, M.; Shields, G. P.; Stevens, J. S.;
Towler, M.; Wood, P. A. Mercury 4.0: From Visualization to Analysis,
Design and Prediction. J. Appl. Crystallogr. 2020, 53, 226–235.
https://doi.org/10.1107/S1600576719014092.
(70) Foster, J. A.; Damodaran, K. K.; Maurin, A.; Day, G. M.;
Thompson, H. P. G.; Cameron, G. J.; Bernal, J. C.; Steed, J. W. Phar-
maceutical Polymorph Control in a Drug-Mimetic Supramolecular Gel.
Chem. Sci. 2016, 8, 78–84. https://doi.org/10.1039/C6SC04126D.
(71) Andrews, J. L.; Pearson, E.; Yufit, D. S.; Steed, J. W.; Ed-
kins, K. Supramolecular Gelation as the First Stage in Ostwald’s Rule.
Cryst.
Eng.
Mater.
2016,
72,
514–529.
https://doi.org/10.1107/S2052520616005382.
(79) Johnstone, R. D. L.; Ieva, M.; Lennie, A. R.; McNab, H.;
Pidcock, E.; Warren, J. E.; Parsons, S. Pressure as a Tool in Crystal
Engineering: Inducing a Phase Transition in a High-Z’ Structure.
CrystEngComm
https://doi.org/10.1039/b917290d.
(80) Moggach, S.; Parsons, S. Molecular Solids at Extreme Pres-
sure. CrystEngComm 2010, 12, 2515.
https://doi.org/10.1039/c0ce90003f.
2010,
12,
2520–2523.
(81) Fabbiani, F. P. A.; Allan, D. R.; Parsons, S.; Pulham, C. R.
An Exploration of the Polymorphism of Piracetam Using High Pres-
sure.
CrystEngComm
2005,
7,
179–186.
https://doi.org/10.1039/b418976k.
(82) Dawson, A.; Allan, D. R.; Belmonte, S. A.; Clark, S. J.; Da-
vid, W. I. F.; McGregor, P. A.; Parsons, S.; Pulham, C. R.; Sawyer, L.
Effect of High Pressure on the Crystal Structures of Polymorphs of
Glycine.
Cryst.
Growth
Des.
2005,
5,
1415–1427.
Cryst.
Growth
Des.
2018,
18,
7690–7700.
https://doi.org/10.1021/cg049716m.
(83) Sobczak, S.; Katrusiak, A. Colossal Strain Release by Con-
formational Energy Up-Conversion in a Compressed Molecular Crys-
https://doi.org/10.1021/acs.cgd.8b01539.
(72) Dawn, A.; Andrew, K. S.; Yufit, D. S.; Hong, Y.; Reddy, J.
P.; Jones, C. D.; Aguilar, J. A.; Steed, J. W. Supramolecular Gel Con-
trol of Cisplatin Crystallization: Identification of a New Solvate Form
Using a Cisplatin-Mimetic Gelator. Cryst. Growth Des. 2015, 15,
4591–4599. https://doi.org/10.1021/acs.cgd.5b00840.
tal.
J.
Phys.
Chem.
C
2017,
121,
2539–2545.
https://doi.org/10.1021/acs.jpcc.6b11030.
(84) Patyk-Kaźmierczak, E.; Kaźmierczak, M. A New High-
Pressure Benzocaine Polymorph — towards Understanding the Molec-
ular Aggregation in Crystals of an Important Active Pharmaceutical In-
gredient (API). Acta Crystallogr., Sect. B: Struct. Sci., Cryst. Eng. Ma-
ter. 2020, 76, 56–64. https://doi.org/10.1107/S2052520619016548.
(85) Kaźmierczak, M.; Katrusiak, A. Quantitative Estimate of
Cohesion Forces. CrystEngComm 2015, 17, 9423–9430.
https://doi.org/10.1039/C5CE01942G.
(73) Aitipamula, S.; Banerjee, R.; Bansal, A. K.; Biradha, K.;
Cheney, M. L.; Choudhury, A. R.; Desiraju, G. R.; Dikundwar, A. G.;
Dubey, R.; Duggirala, N.; Ghogale, P. P.; Ghosh, S.; Goswami, P. K.;
Goud, N. R.; Jetti, R. R. K. R.; Karpinski, P.; Kaushik, P.; Kumar, D.;
Kumar, V.; Moulton, B.; Mukherjee, A.; Mukherjee, G.; Myerson, A.
S.; Puri, V.; Ramanan, A.; Rajamannar, T.; Reddy, C. M.; Rodriguez-
Hornedo, N.; Rogers, R. D.; Row, T. N. G.; Sanphui, P.; Shan, N.;
Shete, G.; Singh, A.; Sun, C. C.; Swift, J. A.; Thaimattam, R.; Thakur,
T. S.; Kumar Thaper, R.; Thomas, S. P.; Tothadi, S.; Vangala, V. R.;
Variankaval, N.; Vishweshwar, P.; Weyna, D. R.; Zaworotko, M. J.
Polymorphs, Salts, and Cocrystals: What’s in a Name? Cryst. Growth
Des. 2012, 12, 2147–2152. https://doi.org/10.1021/cg3002948.
(74) Swapna, B.; Maddileti, D.; Nangia, A. Cocrystals of the Tu-
berculosis Drug Isoniazid: Polymorphism, Isostructurality, and Stabil-
(86) Bryant, M. J.; Maloney, A. G. P.; Sykes, R. A. Predicting
Mechanical Properties of Crystalline Materials through Topological
Analysis.
CrystEngComm
2018,
20,
2698–2704.
https://doi.org/10.1039/C8CE00454D.
(87) Sun, C.; Grant, D. J. W. Influence of Crystal Structure on the
Tableting Properties of Sulfamerazine Polymorphs. Pharm. Res. 2001,
18, 274–280. https://doi.org/10.1023/A:1011038526805.
ity.
Cryst.
Growth
Des.
2014,
14,
5991–6005.
https://doi.org/10.1021/cg501182t.
ACS Paragon Plus Environment

Page 15 of 15
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
15
ACS Paragon Plus Environment