On the Effect of Secondary Nucleation on Deracemization through Temperature Cycles

Article abstract of DOI:10.1002/chem.201904239

Herein, the pivotal role of secondary nucleation in a crystallization-enhanced deracemization process is reported. During this process, complete and rapid deracemization of chiral conglomerate crystals of an isoindolinone is attained through fast microwave-assisted temperature cycling. A parametric study of the main factors that affect the occurrence of secondary nucleation in this process, namely agitation rate, suspension density, and solute supersaturation, confirms that an enhanced stereoselective secondary nucleation rate maximizes the deracemization rate. Analysis of the system during a single temperature cycle showed that, although stereoselective particle production during the crystallization stage leads to enantiomeric enrichment, undesired kinetic dissolution of smaller particles of the preferred enantiomer occurs during the dissolution step. Therefore, secondary nucleation is crucial for the enhancement of deracemization through temperature cycles and as such should be considered in further design and optimization of this process, as well as in other temperature cycling processes commonly applied in particle engineering.

Full text of DOI:10.1002/chem.201904239

A Journal of
Accepted Article
Title: On the Effect of Secondary Nucleation on Deracemization via
Temperature cycles
Authors: Fabio Cameli, Joop H. ter Horst, René R. E. Steendam,
Christos Xiouras, and Georgios D. Stefanidis
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.201904239
Link to VoR: http://dx.doi.org/10.1002/chem.201904239
Supported by

10.1002/chem.201904239
Chemistry - A European Journal
On the Effect of Secondary Nucleation on Deracemization via
Temperature cycles
Fabio Cameli1, Joop H. ter Horst2, René R. E. Steendam3, Christos Xiouras4, *, and Georgios D.
Stefanidis1,*
1Process Engineering for Sustainable Systems (ProcESS), Department of Chemical Engineering KU
Leuven, Celestijnenlaan 200F, 3001 Leuven, Belgium.
2University of Strathclyde, EPSRC Centre for Innovative Manufacturing in Continuous Manufacturing
and Crystallization (CMAC), Strathclyde Institute of Pharmacy and Biomedical Sciences, Technology
and Innovation Centre, 99 George Street, Glasgow G1 1RD, United Kingdom.
3Ardena, Solid State Research, Meibergdreef 31, 1105 AZ, Amsterdam, Netherlands.
4Crystallization Technology Unit (CTU), Janssen Pharmaceutical Companies of Johnson & Johnson,
Janssen Research & Development, Tournhoutseweg 30, 2340, Beerse, Belgium.
*Corresponding authors: cxiouras@its.jnj.com, georgios.stefanidis@kuleuven.be
Keywords: Chiral resolution, crystal engineering, secondary nucleation, temperature cycling,
microwave heating
Abstract
Herein, the pivotal role of secondary nucleation in a crystallization-enhanced deracemization process is
reported. During this process, complete and rapid deracemization of chiral conglomerate crystals of an
isoindolinone is attained through fast microwave-assisted temperature cycling. A parametric study of
the main factors that affect the occurrence of secondary nucleation in this process, namely agitation
rate, suspension density and solute supersaturation, confirms that an enhanced stereoselective secondary
nucleation rate maximizes the deracemization rate. Analysis of the system during a single temperature
cycle shows that while stereoselective particle production during the crystallization stage leads to
enantiomeric enrichment, undesired kinetic dissolution of smaller particles of the preferred enantiomer
occurs during the dissolution step. Therefore, secondary nucleation is crucial for the enhancement of
deracemization through temperature cycles and as such should be considered in further design and
optimization of this process, as well as in other temperature cycling processes commonly applied in
particle engineering.
Introduction
The need for obtaining enantiopure compounds is widely increasing in fine chemicals manufacturing
as well as in pharmaceutical industry.[1] While asymmetric synthesis of a target enantiomer may imply
1
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
lengthy protocols, costly catalysts and low product recovery, downstream separations of racemic
mixtures offer a valuable and easily applicable alternative for the manufacture of single enantiomers.
Among the various purification techniques established so far, chiral resolution via crystallization
appears to be the most favorable choice in terms of solid yield (therefore productivity) and operational
costs.[1c] A recent striking development in crystallization processes for separating enantiomers are the
solid-state deracemization methods[2-3], which feature complete conversion of the unwanted enantiomer
to the preferred enantiomer with an unprecedented maximum 100% yield. The solid state
deracemization processes are most often realized by applying bead[2b-e] and/or ultrasound grinding[2g-h]
or temperature cycling[3] to a racemic suspension of conglomerate crystals in the presence of a liquid-
phase racemization reaction. Temperature cycling-induced deracemization (TCID), in which repeated
partial dissolution and recrystallization cycles of the chiral species lead to enantiopurity, is perhaps a
more promising deracemization approach compared to the bead grinding technique as it circumvents
the need for further separation of grinding media from the solid product.[2b]
In TCID, temperature sweeps are applied to a suspension containing racemic conglomerate crystals and
a solution in which solute racemization is catalyzed. Although TCID features simple steps in its
realization, the key elements of the controlling mechanism of this relatively recent deracemization
technique have not been completely established yet. Operating parameters such as the impact of solid
mass involved in the thermal cycles,[3b-d] the amount of catalyst,[3c] the cooling rate,[3b,c,e] the suspension
volume,[3d-e] the initial enantiomeric excess[3c,e,f], the amplitude of the sweeps,[3c,e,g] the racemization
rate,[3h] and the use of additives as growth inhibitors[3i] have been studied previously. Moreover, several
computational studies[4] have attributed the observed chiral enrichment to the crystallization step
occurring during the cooling stage. The observed increase in chiral enrichment in the solid phase has
been ascribed to a difference in the growth rate dispersion between the two enantiomeric crystal size
distributions[4a-b] or to a crystal growth mechanism which involves (partial) incorporation of small chiral
clusters to crystals of the same chirality.[4c] Another recent publication, which simulates TCID using a
population balance model, claims that growth, dissolution and racemization processes alone can explain
deracemization without the need for breakage or agglomeration, as long as both temperature and size
dependence of solubility (through the Gibbs-Thomson equation) are taken into account.[4d] Indeed,
alternative mechanisms to conventional crystal growth following McCabe’s ΔL law have been sought
to explain deracemization because the latter, featuring equal growth/dissolution kinetics for the different
chiral populations, would lead to no changes in enantiomeric enrichment during temperature cycling
even in the presence of fast racemization[4a]. Alternatively, the reason for the observed amplification of
the enantiomeric excess in TCID might be found in a mechanism related to the bead grinding process
in which the presence of small solid enantiopure clusters is deemed responsible for the chiral
enrichment. A possible mechanism for the formation of small enantiopure clusters in TCID could be
secondary nucleation, which is considered to be the main nucleation mechanism in suspensions.
2
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
Secondary nucleation involves the generation of new nuclei facilitated through the presence of particles
of the same substance in the suspension.[5] The rate of production of secondary nuclei is mainly
dependent on three process parameters, namely agitation intensity, suspension density and solute
supersaturation.[5c] Numerous pathways for the generation of secondary nuclei have been discovered[6]
and extensively investigated[7] to date. While several studies have highlighted the importance of
secondary nucleation in chiral crystallization[8] as early as in the 1950s with the work of Havinga[8c] and
later in the work of Kondepudi et al.[8d], there is still no clear evidence reported about the interplay
between the generation of secondary nuclei and the enantiomeric enrichment in deracemization
processes involving thermal cycles. Experimental evidence of secondary nucleation taking place in the
recrystallization step during TCID have been reported as secondary nucleation was observed when
extremely fast cooling was attained within a microwave apparatus[9] and during stirrer-induced
secondary nucleation featuring intense stirring through a homogenizer.[3d] However, previous
mechanistic studies on TCID have not considered the effects of secondary nucleation during the
cooling/low temperature stage of the process. As such, the exact interplay between the generation of
secondary nuclei and the enantiomeric enrichment in deracemization processes involving thermal
cycles remains elusive.
Herein, the effect of secondary nucleation on a TCID process is studied experimentally in a microwave
apparatus that allows for unprecedented heating and cooling rates. This unique approach enables the
dissolution/re-crystallization stages to proceed isothermally at high driving forces. In order to identify
potential links between the magnitude of secondary nucleation and the deracemization process, we
investigated the effect of three operating parameters that are considered to influence secondary
nucleation in stirred crystallization processes, namely agitation rate, suspension density and solute
supersaturation. Finally, by comprehensively looking into the evolution of the particle size, liquid phase
concentration and solid phase enantiomeric excess of the system during both dissolution and
recrystallization stages of the process in just two thermal cycles, we elucidate potential mechanisms
which drive TCID. The results of this work advance the understanding into the phenomena involved in
deracemization and pave the way towards future industrial applications.
Results
In order to investigate the potential of deracemization for 2-isopropyl-3-hydroxy-3-phenylisoindolin-1-
one, 1, in toluene with catalytic DBU (Scheme 1), a series of batch experiments were carried out at
various conditions, reported in Table 1. Prior to the application of temperature cycles, an isothermal
experiment (run #1, Table 1) was performed using intense stirring to test whether deracemization is
possible even in the absence of temperature cycles. The results of this experiment, reported in ESI,
showed no substantial increase in the enantiomeric excess. As expected, in the absence of grinding or
temperature cycling, enantiomeric excess does not evolve.
3
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
In order to design appropriate temperature profiles for deracemization, the solubility of racemic 1, in
toluene with catalytic DBU (Scheme 1) was measured at various temperatures and reported in Figure 1
(left). The solubility data as a function of temperature can be adequately described by the empirical
Equation 1 (C(T); mg gsolvent-1, T; K):
Scheme 1. Racemization reaction of 1 in the presence of catalytic DBU in toluene.
ꢇꢈꢉꢈꢋ
ꢁ ꢃ
ꢀ ꢂ ꢄ exp 12.3 ꢆ
(1)
ꢅ
ꢊ
Based on this information, temperature cycling experiments can be designed featuring different cycle
amplitudes (∆ꢂ ꢄ ꢂꢌ ꢆ ꢂ ) between the high and low temperatures, Th and Tl, respectively. In this study,
ꢍ
different cycle amplitudes were investigated by keeping Tl the same in all experiments at 35 °C, but
varying Th between 50-80 °C (Figure 1, right). In turn, the amplitude of the thermal cycles dictates the
solid mass which dissolves during each cycle, and with that the mass of crystals remaining upon partial
dissolution. The theoretical fraction of solid mass after heating at the respective Th can be quantified
according to[3d]
:
∗
ꢁ ꢃ
ꢊ
ꢒ
ꢏ ꢑꢏ
ꢐ
ꢎ ꢄ
(2)
∗
ꢁ
ꢃ
ꢊ
ꢓ
ꢏ ꢑꢏ
ꢐ
∗
∗ꢁ
ꢃ
where ꢀ ꢂꢌ and ꢀ ꢁꢂ ꢃ are respectively the equilibrium concentrations of the solution during the
ꢍ
isothermal dissolution and crystallization periods and where ꢀꢔ is the initial total dry mass of crystals
used.
Since in this work we were mostly interested in the effect of secondary nucleation on temperature
cycling induced deracemization, we focused on studying individually the factors that influence the rate
of secondary nucleation during the cooling step, namely agitation rate (runs #2 - #4 in Table 1),
supersaturation level (runs #4 - #7 in Table 1) and suspension density (runs #4 and #8 - #11 in Table
1). While the effect of the cooling rate was not investigated in this study, the fastest cooling rate that
was experimentally possible was selected in all experiments. Hence, the thermal cycles can be
considered to evolve via isothermal dissolution and crystallization stages.
4
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
Figure 1. Left: Solubility of rac-1 in toluene with DBU (33 mM). Solubility limits of the applied cycles
with different top temperatures indicated by the blue (Th = 50 °C), red (Th = 70 °C) and orange (Th = 80
°C) lines. The grey line represents the solubility at the low temperature (Tl = 35 °C). Right: Thermal
cycles applied with different Th values and same low temperature limits.
Table 1. Overview of the process conditions during the experiments.
Experimental
run #
Supersaturation Suspension
Agitation
rate, N
(rpm)
Fraction of
mass surviving
dissolution, w
(%)
∆ꢂ (°C)
at ꢂ = 35 °C
density at ꢂꢌ
(mg gsolvent )
ꢍ
-1
Si (-)
1
2
3
4
5
6
7
8
9
35
1.0
2.6
2.6
2.6
1.5
3.3
4.1
2.6
2.6
2.6
2.6
56.8
25.7
25.7
25.7
25.7
25.7
0.0
1000
550
700
100.0
45.2
45.2
45.2
70.6
36.4
0.0
17.4
35-70
35-70
35-70
35-50
35-80
35-90
35-70
35-70
35-70
35-70
1000
1000
1000
1000
1000
1000
1000
1000
6.5
16.1
63.9
73.7
34.1
67.2
70.3
10
11
Effect of agitation rate
A key parameter in establishing the effect of secondary nucleation on a deracemization process
employing temperature cycles is the agitation rate through its influence on contact nucleation (collisions
between crystal-impeller or crystal-crystal) and fluid shear. In order to investigate qualitatively the
extent of secondary nucleation in our system and the influence of the hydrodynamic conditions, we
designed three temperature cycling experiments (runs #2 - #4, Table 1), among which the agitation rate
is varied between 550 and 1000 rpm. The evolution of the volume-based particle size distributions
during these experiments (measured at the end of the temperature cycles) is presented in Figure 2
together with an estimated evolution of the specific surface area per cycle. Note that the specific surface
area values were calculated based on the PSDs assuming constant density, spherical non-porous
5
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
particles with smooth surfaces and may not quantitatively correspond to the actual specific surface area
of the particles at the respective conditions. While these values can only be viewed as indicative, they
still provide a helpful means to compare the evolution of the surface area between the different
experiments.
The results in Figure 2 reveal that in those experiments, temperature cycling leads to a progressive
decrease in crystal sizes and a concomitant increase in the specific surface area. The observed trend in
crystal size and specific surface area is ascribed to the formation of secondary nuclei during the
crystallization stage. If new particles did not form, the observed trend in size and specific surface area
could only be explained by much faster (partial) dissolution compared to crystal growth, which in turn
would also lead to a lower suspension density than the theoretical one (based on the solubility at Tl) at
the end of each cycle. However, in all experiments the yield at the end of the process was between 91-
98% of the theoretical one, with losses pertaining mostly due to filtration. Additionally, the decrease in
mean size and increase in specific crystal surface area are more pronounced as the agitation rate (and
with that, the secondary nucleation rate) increases form 550 rpm to 1000 rpm. Moreover, while all PSDs
exhibit a bimodal trend with several particles < 10 μm, the trend is more pronounced in the case of
intense agitation (1000 rpm) with a clear progression of the mean population size towards the small
sizes classes, indicating that, as expected, the generation of secondary nuclei is enhanced at higher
agitation rates.
6
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
Figure 2. Particle size distributions (PSDs) of temperature cycling experiments with different agitation
rates: run #2 (top left), run #3 (top right) and run #4 (bottom left). Bottom right: Estimated evolution of
the specific surface area for the same experiments, calculated by the PSDs assuming constant density,
spherical non-porous particles with smooth surfaces. Blue triangles, green circles and red squares
correspond to runs #2, #3 and #4, respectively.
Furthermore, SEM images (Figure 3) taken after 10 cycles for the three experiments indicate the
presence of small particles next to larger ones, which is a typical result of the occurrence of secondary
nucleation. Run #4 at 1000 rpm results in the formation of very small particles coexisting with larger
ones, in which no substantial signs of mechanical breakage can be detected (Figure 3c). A similar
qualitative trend is visible in the pictures featuring the crystals obtained from run #3 at 700 rpm (Figure
3b) and run #2 at 550 rpm (Figure 3a). However, in the latter images it is noticeable that these
experimental conditions resulted in larger mean sizes of particles compared to those of run #4, which
reflects the data of the PSD shown in Figure 2. The differences in sizes of the small crystals reveal how
some of the newly formed nuclei grow larger and may survive the dissolution step.
Figure 3. SEM images of crystals at the end of the 10th cycle in runs #2, #3 and #4, at 550 rpm (a), 700
rpm (b) and 1000 rpm (c), respectively.
Figure 4 shows the evolution of the solid phase enantiomeric excess for runs #2 - #4. While all three
experiments feature a considerable deracemization rate for TCID processes, resulting in an enantiopure
outcome within 16 cycles, a clear trend in the rate of the deracemization can be observed, with an
increased agitation rate leading to faster deracemization. It is possible that the faster deracemization
rate observed at a higher degree of agitation is linked to the faster production of secondary nuclei in a
stereoselective manner. An increase in the agitation rate will introduce secondary nucleation of both
enantiomer crystal populations. However the autocatalytic nature of the secondary nucleation process,
that sustains the imbalance in enantiomer crystal population (i.e. formation of a very large number of
secondary nuclei of the same handedness from the surface of a crystal in a stirred solution)[6a,7d-e,8a]
,
ensures that the enantiomeric crystal population in excess will be amplified faster than the minority
population. On the other hand, only isothermal stirring cannot deliver enrichment of the chiral
suspension via attrition-enhanced deracemization for this system as shown by the isothermal run #1
reported in ESI. In fact, in the absence of any supersaturation, formation of smaller particles must rely
7
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
on mechanical breakage, which in our experiments (absence of grinding) is limited. Therefore, the
observed trends in Figure 4 cannot be attributed to Viedma ripening[2b]
.
A deeper study on the evolution of the newly formed secondary nuclei with regards to birth and
dissolution along the duration of two thermal cycles is presented in a following section.
Figure 4. Evolution of the enantiomeric excess E for experiments at different agitation rates. Blue
triangles, green circles and red squares correspond to runs #2, #3 and #4, respectively.
Effect of supersaturation
In order to study the effect of supersaturation on secondary nucleation during temperature cycling-
enhanced deracemization, runs #4 - #6 were performed using different thermal profiles (Figure 1b and
Table 1), which in turn induce different supersaturation levels during the crystallization stage at 35 °C.
Table 1 reports the supersaturation ratios (Si) attained at the beginning of the isothermal step at the low
temperature, assuming negligible desupersaturation due to nucleation and growth during the fast
cooling stage (~2 min for all experiments). As secondary nucleation during crystallization is influenced
by both supersaturation and the amount of seed crystals (suspension density), the initial mass of the
enantiomers used for runs #4 - #6 was varied to maintain the same suspension density at the end of the
partial dissolution phase in each cycle, in order to isolate and study the effect of supersaturation alone.
Nevertheless, the undersaturation conditions for dissolution would inevitably change by varying the
thermal sweeps, hence affecting possible preferential dissolution of small particles.
The evolution of the enantiomeric excess for runs #4 - #6, is reported in Figure 5 and shows that higher
initial supersaturation ratio Si during the crystallization phase enhances the deracemization rate in the
case of sweeps between 35 and 70 °C (red curve) compared to the lower supersaturation ratio attained
by running the dissolution stage at 50 °C (blue curve). Indeed, the experiment with sweeps to 70 °C is
practically enantiopure after 7 cycles, while the lower supersaturation conditions lead to only 42%
enantiomeric excess at the same number of cycles. However, a further increase of the initial
supersaturation ratio up to 3.3 (dissolution at 80 °C, orange curve), results in a substantially decreased
deracemization rate as compared to the other two experiments. It is also noted that in the extreme case
8
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
of Si = 4.1, enabled by complete dissolution of the crystals at 90 °C (run #7) before cooling, stochastic
primary nucleation occurred and no deracemization was observed (ESI).
The observed non-monotonic dependence of the deracemization rate on the initial supersaturation may
be a result of the competition between stereoselective and non-stereoselective nucleation occurring via
different mechanisms. Although the precise mechanism of secondary nucleation in this system is
currently unknown, we can speculate that at lower supersaturation ratios, secondary nucleation
originates directly from the seeds through attrition and/or detachment of surface irregularities grown on
the seed crystals, thereby leading to the stereoselective production of secondary nuclei and the observed
strong enantiomeric enrichment. Obviously, such a mechanism is enhanced by increasing
supersaturation, as more attrition fragments form and surface irregularities would grow faster and in a
more irregular manner. However, as supersaturation increases further, the concentration and size of
subcritical clusters in the solution increases substantially, which after a certain point may lead to non-
stereoselective enhanced secondary nucleation through for instance the embryo coagulation
mechanism[6b], or even to non-stereoselective primary nucleation, slowing down the deracemization
rate. Although this is mainly a hypothesis, evidence of extensive nucleation during the crystallization
phase in the high supersaturation experiment (run #6) was indeed observed by the analysis of the PSD
evolution reported in ESI together with snapshots of the suspension. The sluggish, slightly irregular
evolution of the enantiomeric excess during that experiment, seen in Figure 5, may indicate that multiple
nucleation mechanisms that differ in their stereoselective outcomes are simultaneously competing for
the consumption of the supersaturation. A similar dependence of the mechanism and stereoselective
outcome of secondary nucleation on supersaturation was also reported in the work of Denk and Botsaris
for sodium chlorate[5a]
.
Figure 5. Enantiomeric excess (E) trend for experiments run under different initial supersaturations
(Si), but with the same suspension density at the beginning of the cooling stage. Red squares, blue circles
and orange triangles correspond to runs #4, #5 and #6, respectively.
9
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
Effect of suspension density
The third key parameter for secondary nucleation in TCID is the suspension density at the initial stages
of the crystallization phase, as contact nucleation through the collision among particles or with the
stirrer is deemed as one of the most effective routes for the generation of new crystals. Additionally,
shear-induced secondary nucleation is also assumed to benefit from the higher crystal surface area
provided by increasing suspension density. In fact, contact among particles is inevitable in agitated
suspensions and can induce detachment of surface irregularities growing on top of larger crystals as
well as the displacement of the coagulated solute layer at the solution side of the surface of the parent
crystal. Figure 6 presents the deracemization trend of run #4 and runs #8 - #11 performed under the
same temperature profile and thus the same initial supersaturation during the crystallization phase, but
with different values for the initial suspension density at the beginning of the crystallization step. As
compared to run #4 at 25.7 mg gsolvent-1, run #8 at 6.5 mg gsolvent-1 showed a slower deracemization rate
from the early cycles and a subsequent decline in the solid-state enantiomeric excess, which is likely
due to the formation of crystals of both chiralities. This behavior might be a consequence of a low
suspension density during the start of the crystallization stage as the fraction of mass dissolved per cycle
is higher than the other cases. The relatively small amount of crystals left after the dissolution stage
(mass fraction surviving dissolution, w, is only 17.4% in run #8, as compared to 45% in run #4) cannot
consume the applied supersaturation through stereoselective secondary nucleation alone. Instead, non-
stereoselective nucleation and/or primary nucleation also occur, leading to the formation of both chiral
forms and a decrease in enantiomeric excess. The observed increase in number of particles during
application of thermal cycles for run #8 is reported in ESI. Hence, a threshold suspension density exists
below which non-stereoselective secondary nucleation takes place in addition to stereoselective
secondary nucleation. Such a threshold suspension density was also found for sodium bromate[3d] and
is expected to be system dependent.
On the other hand, increasing the suspension density up to 63.9 mg gsolvent-1 (runs #4, #9 and #10) does
not significantly influence the observed behavior of the enantiomeric excess: the deracemization rate as
a function of suspension density reaches a plateau value. Taking into account that at a higher suspension
density more solid mass is required to be deracemized whilst the thermal sweeps involve the same
quantity of crystal mass, it can be concluded that a higher suspension density results in enhanced
specific deracemization rate [% chiral enrichment / (number of cycles × suspension density)].
Moreover, given that the same amount of solvent is employed in these experiments, it can be concluded
that higher suspension density results in enhanced process productivity in terms of mass of target
enantiomer delivered per unit of time. Under these conditions of increasing suspension density, the
deracemization rate must be enhanced by the larger magnitude of secondary nucleation, which balances
the drawback of involving the same absolute amount of mass in the thermal sweeps. Besides, the
relatively low percentage of mass dissolved in each cycle at high suspension densities could favor the
10
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
survival of the freshly formed nuclei of the preferred enantiomer, which can further boost
stereoselective secondary nucleation in the subsequent cycle.
Figure 6. Enantiomeric excess E as a function of the number of temperature cycles for different
suspension densities. Blue triangles, black squares, red circles, orange rhombi and cyan inverted
triangles, correspond to runs #4 and #8 - #11, respectively.
-1
However, by further increasing the suspension density up to 73.7 mg gsolvent in run #11, the
deracemization rate does not increase further, but instead it is found to decrease. An increase in
suspension density also leads to a larger amount of solids to be deracemized. This effect leads to a
decrease in deracemization rate and counterbalances the effect of secondary nucleation, thus making
suspension density a valuable process parameter required for future process modelling of TCID
systems.
Thus, the suspension density strongly dictates the deracemization rate as it controls the balance between
the different mechanisms taking place. At too low suspension densities, non-stereoselective processes
may occur which lead to inhibition of deracemization. At too high suspension densities, the amount of
solid material to be deracemized is large and requires a long deracemization time.
Investigation into a single temperature cycle
All measurements reported so far referred to the low temperature isothermal stage of the thermal cycles.
Monitoring the behavior of the system during the heating, cooling and high-temperature phases can
prove helpful in understanding the phenomena leading to enantiomeric enrichment. Usually this is not
possible, as for most compounds and temperature profiles, the changes in the enantiomeric excess
obtained in a single cycle are very small and thus difficult to detect. However, this is not a limitation in
the present work as appreciable deracemization occurs during a single temperature cycle. Thus, the
evolution of the populations of the two enantiomeric crystals over just two temperature cycles is an
efficient means to understand how the enantiomeric enrichment unfolds and which physical phenomena
are controlling it.
11
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
In order to investigate the behavior of the system during the various stages of the temperature cycle,
two-cycle experiments were run between 35 and 70 ºC and an agitation rate of 1000 rpm with additional
intermediate sampling during the isothermal holding steps at Th and Tl. Figure 7 shows the evolution of
the solid phase enantiomeric excess for different experiments featuring different values of the initial
enantiomeric excess (E0=0.5, 21 and 41%) and equal value of suspension density at the end of the
dissolution stage (25.7 mg gsolvent-1). As expected, for the experiments featuring a higher E0, E exhibits
a pronounced non-monotonic behavior during a single cycle, with a steep increase during the dissolution
step and a more constant period during the crystallization step. Notably, no decrease in the enantiomeric
excess is observed during the crystallization step, even though the total mass is increasing, which
illustrates that the mass of the excess enantiomer increases faster than that of the counter enantiomer.
Figure 7. Evolution of the enantiomeric excess (E) in the various stages of the temperature cycling
process for three experiments featuring different initial enantiomeric excesses (orange curves with
triangles correspond to an initial excess of 0.5%, blue curves with circles feature an initial excess of
21% while the initial enantiomeric excess of the red curves with squares was 41%. In all cases w is
equal to 45.2%. Black line refers to the applied temperature profile. All the experiments were run at
1000 rpm.
In order to obtain additional insight into the behavior of the system during the two cycles, the
enantiomeric excess data of Figure 7 are combined with independent estimates of the total mass of the
crystals (estimated by a mass balance using liquid phase concentration measurements at the same times)
to obtain the total crystal mass concentration of R and S species during the isothermal dissolution and
crystallization steps for the different experiments (Figure 8, left). The results indicate that when the two
enantiomeric crystal populations are very similar in mass (i.e. E0=0.5%) at the beginning of the
experiment, they evolve almost symmetrically over the dissolution and crystallization phases leading
to only small enantiomeric enrichment. This implies that in the dissolution stage the two enantiomers
dissolve equimolarly as expected for a conglomerate-forming compound. Similarly, the crystallization
12
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
phase is not significantly biased towards one chiral species yielding the close trends of R and S crystal
concentrations represented by the orange curves in Figure 8 (left).
Conversely, when a significant mass imbalance is present in the initial chiral mixture of the
enantiomeric crystal populations, crystal dissolution no longer proceeds in an equimolar fashion, which
becomes apparent in the two stages of dissolution at high temperature reported in Figure 8 left (blue
and red curves). Relatively more mass of the excess enantiomer (R species in this case) dissolves
compared to the S enantiomer. In the experiment with E0=21%, the crystal mass of R which dissolves
at a high temperature is 1.7 times higher than the mass of S while in the case of E0=41%, 3.2 times more
of the R enantiomer crystals dissolve. These amounts very roughly correspond to the ratio of the
enantiomer masses at the start of the two processes (1.53 and 2.4, respectively). Thus, neglecting the
possible differences in the PSDs of the two enantiomers, we assume that the non-equimolar dissolution
is attributed to kinetic dissolution of the preferred enantiomer that is present in a larger amount and thus
possesses a larger available surface area for dissolution. Once the preferred enantiomer dissolves first,
the fast racemization reaction converts it to the counter enantiomer maintaining the driving force for
dissolution equal for each enantiomer. However, it is worth mentioning that even though undesired
conversion of the preferred enantiomer to the counter enantiomer may occur during the dissolution
phase, the overall enantiomeric excess can still increase (evidently from 39% to 55% and from 58% to
69%, for the two experiments, as visible in the second cycle), since the total solid mass (mR + mS)
decreases more than the difference between the enantiomer solid masses (mR - mS). Hence, the
dissolution process contributes substantially in the increase of the enantiomeric excess during
deracemization by temperature cycles even when it is biased toward the preferred enantiomer.
During the cooling and isothermal holding steps at the lower temperature, for the same experiments at
higher initial enantiomeric excess (blue and red curves in Figure 8 left), both enantiomers crystallize
leading to an increase in the suspension density of both chiral crystalline phases, with the average rate
of mass increase of the enantiomer in excess being higher during that period. In fact, when only 0.5%
initial enantiomeric excess is employed, the solid mass increase of the preferred enantiomer is only 4%
higher than the counter enantiomer, while in the experiment with E0=41%, 77% of the crystallized mass
is of the target enantiomer. Thus, the autocatalytic nature of the crystallization process becomes
apparent when the mass imbalance between the two enantiomers is higher at the start of the isothermal
step at low temperature. It is worth pointing out that this value refers only to the isothermal
crystallization step, although part of the crystallization must take place also during the fast cooling stage
(~2 min), in which it is assumed that a similar trend in enantioselective crystallization is present.
Furthermore, it is apparent from the different trends in Figure 8a that the main portion of the
crystallizing mass appears at the beginning of the crystallization period, when the supersaturation is
higher and then follows an asymptotic behavior until the end of the isothermal period.
13
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
This large increase in the difference between the mass of the enantiomers in the crystallization stage of
the temperature cycle may be explained by the occurrence of stereoselective secondary nucleation.
Figure 8b shows the evolution of the specific surface area of both enantiomers for the same experiments,
measured by laser diffraction. For every experiment, the specific surface area constantly increases
during the crystallization stage, which confirms the occurrence of secondary nucleation, since a purely
growth-dominated crystallization would lead to the opposite behavior, i.e. a decrease in the specific
surface area. Furthermore, the values of specific surface area calculated from laser diffraction greatly
underestimate the presence of fines and surface defects that could play a central role during the
crystallization process. Conversely, during the dissolution stage the specific surface area decreases,
which indicates the disappearance of some of the smaller particles formed during the crystallization
stage. While the increase in the specific surface area (and thus the generation of secondary nuclei)
appears similar during the crystallization stage for the three experiments, the handedness of the
secondary nuclei, as shown in Figure 8a, which in turn dictates the enrichment achieved in a single
cycle, heavily depends on the amounts of the two enantiomers that seed the crystallization process in
each step. Hence, higher enantiomeric imbalance at the beginning of the crystallization stage results in
more secondary nuclei of the excess chirality produced.
In Figure 9, it is shown that in the extreme case where all the particles undergoing temperature cycling
are of one handedness, the formation of new particles (as indicated by the slightly increasing specific
surface area) unfolds only through stereoselective secondary nucleation, as the chirality of the particles
remained the same throughout the process. This conclusion must hold for supersaturation conditions
that do not prompt non-stereoselective nucleation as observed in Figure 5. Thus, the trend displayed in
Figure 9 highlights the importance of the chiral bias represented by the enantiomeric seeds that
seemingly steer the stereoselective formation of new particles. Therefore, the same mechanism must
govern the chiral enrichment in Figure 8a, in which the experiments featuring a higher spread between
the two chiral populations deliver higher deracemization rate through autocatalytic formation of
secondary nuclei.
Besides, as the specific surface area is inversely proportional to the mean particle diameter, partial
dissolution of the crystals would lead to an increase of the specific surface area as in the case shown in
Figure 9. Therefore, the dissolution phase for the enantiopure particles, that featured similar sizes
among each other, mainly resulted in crystal shrinkage rather than kinetic dissolution of small particles
as in the experiments displayed in Figure 8b. A more detailed representation of the specific surface area
distribution for these experiments, reported in ESI, shows that the dissolution step yields significant
dissolution of small particles generated in the crystallization phase for the experiments starting with
scalemic conditions.
14
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
Figure 8. Left: Evolution of the crystal mass concentration of the R enantiomer (solid lines, full
symbols) and S enantiomer (dashed curves, empty symbols) during two thermal sweeps. Different initial
enantiomeric excesses of the R form were employed: orange curves with triangles correspond to an
initial excess of 0.5%, blue curves with circles feature an initial excess of 21% while the initial
enantiomeric excess of the red curves with squares was 41%. Right: Trend of the total (R + S) specific
surface area during two thermal sweeps for the same experiments. Lines are drawn as a guide to the
eye. Black lines represent the applied temperature profile.
Figure 9. Enantiomeric excess (red squares) and specific surface area values (blue circles) of an initially
enantiopure suspension during the application of two temperature cycles. Lines are drawn as a guide to
the eye.
Discussion
The results presented in this work indicate that stereoselective secondary nucleation occurs during the
cooling/isothermal stage of a temperature cycling-enhanced deracemization process. Furthermore, the
magnitude of secondary nucleation seems to directly influence the deracemization rate, with higher
secondary nucleation rates leading to faster deracemization. Although the operating conditions of our
experiments can be viewed as quite extreme (thermal sweeps of ± 35 °C at ~ 18 °C min-1), thereby
favoring secondary nucleation, we believe that the occurrence of the latter process cannot be excluded
in other systems reported in the literature employing temperature cycling for deracemization at milder
15
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
conditions. For example, in the study of Suwannasang et al.[3g], wherein deracemization of a chiral
compound was performed employing temperature cycles at a maximum of ± 10 °C and very slow
cooling rates in the order of ~0.2 °C min-1, it was shown by SEM microscopy that in many experiments
the particle size progressively reduced after application of temperature cycling, similar to our results.
A possible explanation to this observation could be the formation of small particles due to secondary
nucleation. It is noted that in the absence of new particle formation, temperature cycling should in
principle affect a population of crystals similar to Ostwald ripening, i.e. increase in the mean population
size and decrease in the number of particles.
Indeed, the secondary nucleation threshold (SNT), i.e. threshold supersaturation below which secondary
nucleation is negligible, can in certain systems/operating conditions be extremely low[5b], while contact
secondary nucleation often requires a surprisingly small impact energy to be effective.[7d,10] Coupled
with the fact that most temperature cycling deracemization experiments were performed in small scale
equipment using magnetic stirring bars, the probability for secondary nucleation may be considered
high even at seemingly mild conditions (small amplitude of temperature cycles and/or cooling rates).
Indeed, in larger crystallizers, contact nucleation is less efficient and it was observed that under those
conditions, temperature cycling alone was not sufficient to deracemize sodium bromate[3d]. For sodium
bromate it was required to have a sufficiently high suspension density which would avoid non-
stereoselective nucleation. A high suspension density was previously realized through crystal breakage
with the use of a homogenizer. In the present work, a high suspension density was maintained through
secondary nucleation which was induced through rapid cooling. However, in the present study it was
also shown that some of the secondary nuclei of the preferred enantiomer formed in excess during the
crystallization step, being significantly smaller than the crystals of the counter enantiomer, may dissolve
preferentially in the dissolution step. To alleviate this undesired effect, one could homogenize (e.g. by
wet-milling) the crystals prior to the dissolution phase in order to bring the size of the larger counter
enantiomer crystals closer to the size of the newly-formed crystals of the preferred enantiomer.
Unfortunately, such apparatus is unsuitable for the small microwave cavity employed in this study.
Among the operating parameters affecting the rate of secondary nucleation, solute supersaturation and
suspension density have proven to affect the deracemization rate in a non-monotonic fashion, whereas
increasing agitation intensity proportionally enhances deracemization rate. An overview of the effect of
solute supersaturation and suspension density on the deracemization rate per cycle is shown in Figure
10, where one could ideally find optimal values of the process parameters to achieve the highest
deracemization rate. The values for the deracemization rate were obtained through linear fits of the
enantiomeric excess vs. number of cycles data. It is seen from Figure 10 that deracemization rates
between 10.7 and 11.6 % per cycle are obtained when supersaturation ratio is equal to 2.6 and
suspension density varies between 25.7 and 63.9 mg gsolvent-1. Beyond these operating limits, decreased
deracemization rates per cycle are observed with the lowest value of 2.5% per cycle detected for the
16
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
case of supersaturation ratio of 3.3 and suspension density of 25.7 mg gsolvent-1. Therefore, for this
specific system, the set of operational parameters that delivered the highest deracemization rate (11.6
% per cycle) comprises of suspension density at high temperature of 25.7 mg/gsolvent, solute
supersaturation ratio of 2.6 and agitation rate equal to 1000 rpm. It is worth pointing out that in the
present study the influence of the racemization kinetics on the deracemization process was not
investigated. However, as shown previously[3h], the latter can also be an important factor for optimizing
deracemization by temperature cycles.
Figure 10. Effect of increasing suspension density and solute supersaturation ratio on the cycle-based
deracemization rate.
Currently, secondary nucleation has not been incorporated in the proposed process models that attempt
to capture the mechanisms behind deracemization via temperature cycles[4]. The present work
demonstrates that stereoselective secondary nucleation is a key feature of temperature cycling-enhanced
deracemization and the deracemization pathway is therefore similar to Viedma ripening[2b] since both
processes rely on continuous surface area formation through the generation of new particles in a
stereoselective manner (through particle breakage in Viedma ripening and stereoselective secondary
nucleation in temperature cycling). However, since preferential dissolution of the small secondary
nuclei of the preferred enantiomer can occur, an additional nuclei survival mechanism may be required
for complete deracemization. Although not yet experimentally verified, such a mechanism could be an
enantioselective agglomeration phenomenon, as proposed for Viedma ripening, in which secondary
nuclei are incorporated in the larger crystals during the growth stage of the cycle in parallel to secondary
nucleation.[11]
Conclusions
In conclusion, we achieved complete and rapid solid-state deracemization of a chiral isoindolinone by
microwave-assisted temperature cycling under conditions which strongly favor the occurrence of
17
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
secondary nucleation. The effect of secondary nucleation on deracemization by temperature cycles was
elucidated by studying the effects of operating parameters that are linked with the generation of
secondary nuclei in seeded crystallization processes, namely agitation rate, suspension density and
solute supersaturation. Careful measurements of the evolution of the particle size and enantiomeric
excess during these experiments reveals that an increase in the agitation rate results in the enhancement
of stereoselective secondary nuclei formation, which in turn leads to an enhanced deracemization rate.
An increase in either supersaturation at a constant suspension density or an increase in suspension
density at constant supersaturation during the isothermal stages at low temperatures also leads to an
increase in deracemization rate initially. Interestingly, this effect reverses upon exceeding a threshold
value of the supersaturation or suspension density. In the first case, non-stereoselective nucleation
lowers the enantiomeric enrichment at high solute supersaturation ratios, whilst high suspension density
levels imply lower percentage of mass involved in the dissolution/crystallization cycles thus lower
deracemization capacity. Finally, a close inspection of the behavior of the system over the duration of
just two temperature cycles for various experiments reveals that both dissolution and re-crystallization
are heavily biased towards the enantiomer in excess, with re-crystallization being the most important
factor for chiral enrichment achieved during a single cycle. Since the presence of secondary nucleation
is seemingly beneficial for this process, it would be relevant to investigate the emergence of secondary
nucleation in other more common systems (at milder conditions) in order to assess the system specificity
of the findings of this work and subsequently focus on the optimization of such phenomenon, if it is
proven to be indeed a general leading mechanism in temperature cycling-induced deracemization.
Experimental section
Materials
2-isopropyl-3-hydroxy-3-phenylisoindolin-1-one, 1, was synthesized following the procedure reported
in ESI. The experimental conditions for deracemization were chosen on the basis of previous successful
attempts of solid state deracemization of 1 via attrition-enhanced deracemization[12,13]. These process
conditions included 3 mL of toluene (Alfa Aesar), 15 µL of catalytic racemization agent 1,8-
diazabicyclo[5.4.0]undec-7-ene (DBU, Sigma Aldrich) and 200 mg (unless otherwise stated) of
crystalline 1.
Solubility measurements
The temperature-dependent solubility of 1 in toluene with catalytic amounts of DBU (same
concentration as in the deracemization experiments) was obtained by clear point measurements at
various concentrations by adding a known amount of racemic 1 in 0.872 g of toluene and DBU in a 1.5
mL vial. The solubility in the absence of the catalyst was also measured via gravimetric analysis
(reported in ESI) showing an overall lower solubility of 1 in pure toluene over temperature as compared
18
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
to the case including DBU. The vials were placed in a Crystal16 apparatus (Technobis, Crystallization
Systems) and three consecutive temperature cycles were applied to all vials. Each cycle consisted of
heating the suspension to 90 °C at 0.2 °C min-1, holding at 90 °C for 10 min, cooling to 0 °C at 5 °C
min-1 and holding at 0 °C for 2 h. For each cycle, the saturation temperature was recorded for each vial
as the temperature at which 100% transmissivity of the detecting laser was achieved.
Temperature cycling experiments
All deracemization experiments were performed in a microwave reactor (Anton Paar Monowave 300),
which allows for both rapid heating via microwave irradiation and fast cooling through a jet of
compressed air (5 bar) into the microwave cavity. During the experiments, the internal suspension
temperature was measured and controlled by an optical fiber probe, while the reactor wall temperature
was measured by an IR sensor. In all experiments, the low temperature of the cycle (Tl) was kept the
same (35 °C), while the high temperature (Th) was varied between 50-80 °C. The suspension was
agitated throughout the experiments with an oval-shaped magnetic stirring bar (ø=0.5 cm, L=1 cm),
stirring at constant values between 550 and 1000 rpm.
Sampling
In most experiments, samples of the suspension were taken at the end of the cycles (i.e. at the end of
the isothermal crystallization stage). For some experiments, samples were also taken at intermediate
stages of the isothermal dissolution and crystallization stages of the thermal cycles. In every sampling
event, 100 µL of the suspension was withdrawn from the reactor and filtered on a glass filter (Porosity
5, ø=1-1.6 µm) to separate the solid from the liquid phase and both phases were subjected to analysis.
After the end of the experiments, the overall solid mass retrieved by filtering the samples was used to
calculate the total solid yield (Y) defined as the total mass collected from the process over the theoretical
mass available at the cold stage of the temperature cycles:
ꢖꢗ
ꢕ ꢄ
ꢁ3ꢃ
∗ꢁ ꢃ
ꢖꢘ ꢆ ꢀ ꢂ ∗ ꢙꢗꢚꢍꢛꢜꢝꢔ
ꢍ
Where Ms is the total mass collected by filtration, M0 is the initial dry mass of crystals, C*(Tl) is the
-1
solubility at the low temperature step expressed in mg gsolvent and msolvent is the total mass of solvent
employed.
Solid and liquid phase characterization
The solid phase enantiomeric excess (E) of the dried crystals was determined by chiral HPLC analysis
after dissolution in 2-propanol. The enantiomers of 1 were separated by a Lux Amylose-1
(Phenomenex) column running at 0.5 mL min-1 of a mixture of n-hexane/ethanol 95/5 (v/v%). After
separation, the enantiomers were subsequently detected by a Diode Array Detector (DAD) at a
wavelength of 254 nm (retention times: 16 min for R and 20 min for S). The enantiomeric excess was
19
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
quantified as the relative difference between the areas of the integrated peaks displayed in the
|ꢟ
chromatograms: ꢞ ꢄ ꢑꢟ | where AR and AS represent the areas of the integrated curves respectively
ꢠ
ꢡ
ꢟ
ꢢꢟ
ꢡ
ꢠ
for the R and S enantiomers. Therefore, from the enantiomeric excess trend an estimation of the
deracemization rate per unit of cycle [% chiral enrichment / number of cycles] can be calculated to
compare the different experiments. This definition equals a time-based deracemization rate as the cycle
duration remains constant in all the experiments while a specific deracemization rate [% chiral
enrichment/ (number of cycles * suspension density)] can be introduced when the suspension density
varies in the experiments. Additionally, part of the dried crystals withdrawn from the suspension was
subjected to particle size analysis by laser diffraction (Malvern 3000, Mastersizer), in order to determine
the crystal size distribution (CSD) of the sample, which was assumed to be representative of the whole
crystal population in the reactor. n-hexane was used as a dispersant in a Hydro SV cell of a Malvern
Mastersizer 3000 laser diffractometer. A magnetic stirring bar (1800 rpm) was used to ensure dispersion
of the crystals and to avoid agglomeration without inducing crystal breakage. The absence of crystal
breakage during the measurements was ascertained by monitoring the obscuration (percentage of laser
intensity decrease after passing through the particle field), which remained essentially constant during
the duration of the measurements. In some experiments, the liquid phase concentration of 1 was also
measured via UV-Vis analysis (Lambda 365, Perkin Elmer) by firstly producing a calibration curve that
linked values of known solute concentrations with absorbance levels recorded by the instrument at the
same conditions as the experiments.
Acknowledgement
This project has received funding from the European Union’s Horizon 2020 research and innovation
programme under the Marie Skłodowska-Curie grant agreement No 721290. This publication reflects
only the author’s view, exempting the Community from any liability. Project website: http://cosmic-
etn.eu/. Fabio Cameli thanks Prabhat Ranjan for synthesizing the compound used in the experiments.
J.H.t.H. thanks the EPSRC Centre for Innovative Manufacturing in Continuous Manufacturing and
Crystallization (http://www.cmac.ac.uk) for support (EPSRC funding under Grant Reference:
EP/I033459/1).
References
[1]
a) E. Francotte, W. Lindner, Chirality in Drug Research, VCH, Weinheim, 2006. b) J.-F. C.
Guo-Qiang Lin, Qi-Dong You, Chiral Drugs. Chemistry and Biological Action, John Wiley &
Sons, Inc., Hoboken, New Jersey, 2011. c) H. Lorenz, A. Seidel-Morgenstern, Angew. Chem.
Int. Ed. 2014, 53, 1218–1250.
[2]
a) C. Rougeot, J. E. Hein, Org. Process Res. Dev. 2015, 19, 1809–1819. b) C. Viedma, Phys.
Rev. Lett. 2005, 94, 3–6. c) C. Viedma, B. J. V Verkuijl, J. E. Ortiz, T. De Torres, R. M. Kellogg,
D. G. Blackmond, Chem. Eur. J. 2010, 16, 4932–4937. d) W. L. Noorduin, W. J. P. Van
20
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
Enckevort, H. Meekes, B. Kaptein, R. M. Kellogg, J. C. Tully, J. M. Mcbride, E. Vlieg, Angew.
Chem. Int. Ed. 2010, 49, 8435–8438. e) W. L. Noorduin, H. Meekes, A. A. C. Bode, W. J. P.
Van Enckevort, B. Kaptein, R. M. Kellogg, E. Vlieg, Cryst. Growth Des. 2008, 8, 1675–1681.f)
C. Xiouras, E. Van Cleemput, A. Kumpen, J. H. Ter Horst, T. Van Gerven, G. D. Stefanidis,
Cryst. Growth Des. 2017, 17, 882–890. g) C. Xiouras, A. Fytopoulos, J. Jordens, A. G.
Boudouvis, T. Van Gerven, G. D. Stefanidis, Ultrason. Sonochem. 2018, 43, 184-192. h) C.
Xiouras, A. Fytopoulos, J. H. Ter Horst, A. G. Boudouvis, T. Van Gerven, G. D. Stefanidis,
Cryst. Growth Des. 2018, 18, 3051-3061. i) C. Xiouras, J. H. Ter Horst, T. Van Gerven, G. D.
Stefanidis, Cryst. Growth Des. 2017, 17, 4965-4976.
[3]
a) C. Viedma, P. Cintas, Chem. Commun. 2011, 47, 12786. b) K. Suwannasang, A. E. Flood, C.
Rougeot, G. Coquerel, Cryst. Growth Des. 2013, 13, 3498–3504. c) W. W. Li, L. Spix, S. C. A.
De Reus, H. Meekes, H. J. M. Kramer, E. Vlieg, J. H. Ter Horst, Cryst. Growth Des. 2016, 16,
5563–5570. d) R. R. E. Steendam, J. H. ter Horst, Cryst. Growth Des. 2018, 18, 3008–3015. e)
F. Breveglieri, G. M. Maggioni, M. Mazzotti, Cryst. Growth Des. 2018, 18, 1873–1881. f) G.
M. Maggioni, M. P. Fernández-Ronco, M. van de Meijden, R. M. Kellogg, M. Mazzotti,
CrystEngComm. 2018, 20, 3828–3838. g) K. Suwannasang, A. E. Flood, C. Rougeot, G.
Coquerel, Org. Process Res. Dev. 2017, 21, 623–630. h) R. Oketani, M. Hoquante, P. Cardinael,
G. Coquerel, Cryst. Growth Des. 2018, 18, 6417–6420. i) G. Belletti, H. Meekes, F. P. J. T.
Rutjes, E. Vlieg, Cryst. Growth Des. 2018, 18, 6617–6620.
[4]
a) K. Suwannasang, G. Coquerel, C. Rougeot, A. E. Flood, Chem. Eng. Technol. 2014, 37, 1329–
1339. b) R. Uchin, K. Suwannasang, A. E. Flood, Chem. Eng. Technol. 2017, 40, 1252–1260.
c) H. Katsuno, M. Uwaha, Phys. Rev. E 2016, 93, 1–10. d) B. Bodák, G. M. Maggioni, M.
Mazzotti, Cryst. Growth Des. 2018, 18, 7122−7131.
[5]
[6]
[7]
a) E. G. Denk, G. D. Botsaris, J. Cryst. Growth. 1972, 14, 493–499. b) T. L. Threlfall, S. J.
Coles, CrystEngComm. 2016, 18, 369-378. c) S. G. Agrawal, A. H. J. Paterson, Chem. Eng.
Commun. 2015, 202, 698–706.
a) R. Y. Qian, G. D. Botsaris, Chem. Eng. Sci. 1998, 53, 1745–1756. b) R. Qian, G. D. Botsaris,
Chem. Eng. Eng. Sci. 1997, 52, 3429–3440. c) E. G. Denk, G. D. Botsaris, J. Cryst. Growth.
1972, 15, 57–60.
a) Y. Cui, A. S. Myerson, Cryst. Growth Des. 2014, 14, 5152–5157. b) M. L. Briuglia, J. Sefcik,
J. H. Horst, Cryst. Growth Des. 2019, 19, 421−429 c) R. R. E. Steendam, P. J. Frawley, Cryst.
Growth Des. 2019, 19, 3453−3460. d) J. Anwar, S. Khan, L. Lindfors, Angew. Chem. Int. Ed.
2015, 54, 14681–14684. e) A. Cacciuto, S. Auer, D. Frenkel, Nature. 2004, 428, 404–406.
a) M. McBride, R. L. Carter, Angew. Chem. Int. Ed. Engl. 1991, 30, 293–295. b) J. Cartwright,
J. García-Ruiz, O. Piro, C. Ignacio Sainz-Díaz, I. Tuval Phys. Rev. Lett. 93, 035502. c) E.
Havinga, Biochim. Biophys. Acta. 1954, 13, 171–174. d) D. K. Kondepudi, R. J. Kaufman, N.
Singh, Science. 1990, 250, 975–976.
[8]
21
This article is protected by copyright. All rights reserved.

10.1002/chem.201904239
Chemistry - A European Journal
[9]
F. Cameli, C. Xiouras, G. D. Stefanidis, CrystEngComm. 2018, 20, 2897.
[10] R. W. Rousseau, F. P. O’Dell, Ind. Eng. Chem. Process Des. Dev. 1980, 19, 603–608.
[11] M. Mcbride, J. C. Tully, Nature, 2008, 452, 161-162..
[12] F. Yagishita, H. Ishikawa, T. Onuki, S. Hachiya, T. Mino, M. Sakamoto, Angew. Chem. Int. Ed.
2012, 51, 13023–13025.
[13] R. R. E. Steendam, M. C. T. Brouwer, E. M. E. Huijs, M. W. Kulka, H. Meekes, W. J. P. Van
Enckevort, J. Raap, F. P. J. T. Rutjes, E. Vlieg, Chem. Eur. J. 2014, 20, 13527–13530.
22
This article is protected by copyright. All rights reserved.