Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

MINIREVIEW

Crystallization from the Amorphous State: Nucleation–Growth


Decoupling, Polymorphism Interplay, and the Role of Interfaces
MARC DESCAMPS, EMELINE DUDOGNON

University Lille 1, UMET (Unité Matériaux et Transformations) UMR CNRS 8207, Bat P5 F-59655, Villeneuve d’Ascq, France

Received 26 February 2014; revised 21 April 2014; accepted 23 April 2014


Published online 5 June 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.24016

ABSTRACT: The physical stability of the amorphous state is governed by crystallization, which results from the complex interplay of
nucleation and growth processes. These processes can be further complicated by the preferred initial nucleation of less-stable phases, and
interpretation requires the evaluation of the relative roles of structure, dynamics, and thermodynamics on the kinetics of the recrystallization.
As a contribution to this issue, we reanalyze data sets concerning recrystallization of two pharmaceutical compounds: L-arabitol and RS
ibuprofen. These compounds share the property of being good glass formers and present monotropic polymorphism. In the present analysis,
we are mainly focusing on the localization of nucleation and growth zones and the role of a transient crystallization of the metastable
phase. On the basis of the elementary theories, the results offer the opportunity to discuss the impact of interfacial energies, molecular
mobility, crystal disorder, liquid short-range order, and crack formation in the glass.  C 2014 Wiley Periodicals, Inc. and the American

Pharmacists Association J Pharm Sci 103:2615–2628, 2014


Keywords: crystallization; amorphous; nucleation; kinetics; polymorphism; crystal growth

INTRODUCTION stability is that these rates both have a maximum at some de-
gree of undercooling. The existence of such maxima is essential
Preparing poorly soluble drugs in the amorphous state is con-
for the capability of vitrifying a melt by cooling it at sufficiently
sidered to be one of the more attractive methods of enhanc-
high rate. The two maxima do not necessarily occur at the same
ing dissolution and dilution properties.1,2 The most important
temperature (see Fig. 1b). The relative temperature positioning
drawback of using pharmaceuticals in the amorphous state is
of the nucleation and growth zones is the basic feature that de-
their inherent instability to crystallization upon storage, which
termines whether a compound is or is not a good glass former.
eventually leads to the loss of the beneficial dissolution proper-
Simple evidence of this is that often undercooled pharmaceu-
ties. Moreover, avoiding crystallization is of course the decisive
ticals are more prone to crystallization upon reheating than
stage in the amorphization process itself.
upon cooling because of the maximum rate of nucleation being
Amorphization can be achieved via several routes in addition
at a significantly lower temperature than that of growth. At
to the conventional quench cooling of the melt (milling, fast con-
any temperature, the global kinetics of recrystallization will be
densation, spray drying, and so on).3 Whatever the amorphiza-
dictated by the rates of both nucleation and growth processes.
tion route, recrystallization of an amorphous compound can
Controlling the stability of an amorphous pharmaceutical, and
occur throughout the temperature domain of the undercooled
optimizing the amorphization process, thus requires the iden-
melt (T < melting temperature, Tm ), which is a metastable
tification of the temperature domain of preferred occurrence of
state of matter (this state can even be unstable below the glass
each process and the temperatures corresponding to maximum
transition temperature, Tg ). An amorphous undercooled com-
rates. Such a determination can provide a better insight into
pound can be maintained in a metastable state for an amount
the rate of global crystallization and kinetic laws themselves,
of time that depends strongly on temperature. An understand-
but is difficult in practice because as the critical size of nuclei
ing of the lifetime of metastability requires the investigation of
is of the order of a few nm, the initial stage of nucleation is not
the kinetics of recrystallization at each temperature. Crystal-
observable in situ.
lization of an undercooled amorphous melt in fact results from
The crystallization process is complicated by the fact that
the complex interplay of nucleation and growth processes (see
molecular compounds are often polymorphic. Nucleation does
Fig. 1a). Nucleation is the process by which crystalline clus-
not necessarily correspond initially to that of the more stable
ters having a minimum critical size appear randomly, but at
phase. This has been popularized by the empirical Ostwald’s
some given rate N [with dimension t−1 L−3 in the usual three-
rule of stages,4,5 which states-without theoretical foundation–
dimensional (3D) space], within the amorphous matrix. Once
that the first phase to appear is the one which is closest in free
formed, the crystalline grains grow with a domain wall velocity
energy to the parent phase.
V (with dimension t−1 L). The rates of these two processes have
The investigation of the crystallization of indomethacin poly-
their proper temperature variations. The most important fea-
morphs from the amorphous state by Zografi and coworkers6,7
ture with regard to amorphization capability and amorphous
is particularly illustrative of the main factors that influence
recrystallization. By a direct microscopic observation of crys-
tal particle number density (see Figs. 1a and 1b), it was
Correspondence to. Marc Descamps (Telephone: +33-3-20-43-49-79;
E-mail: marc.descamps@univ-lille1.fr) demonstrated that the nucleation rate of phase " has a max-
Journal of Pharmaceutical Sciences, Vol. 103, 2615–2628 (2014) imum above Tg , whereas the growth rate maximum is lo-

C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association cated at higher temperature somewhere between Tg and Tm .

Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014 2615


2616 MINIREVIEW

Figure 1. Photomicrograph of the indomethacin "-crystal form grown from the amorphous phase. Reproduced from Andronis and Zografi7 with
permission from Elsevier. (b) Nucleation and growth rates for the indomethacin "-crystal as a function of temperature. (Adapted from Figs. 7
and 8 of Andronis and Zografi7 with permission from Elsevier).

Furthermore, it was shown that these positions are different MATERIALS AND METHODS
for the two known polymorphs of indomethacin, which is the
Compounds were purchased from Sigma–Aldrich and used
essential feature of their selection during crystallization. Fi-
without further purification.
nally, the potential important role of the crystal—amorphous
Differential scanning calorimetry experiments were carried
interface energy in this polymorph selection was stressed.
out in DSC Q10 and DSC Q1000 calorimeters from TA Instru-
In line with these previous works, the purpose of the present
ments with a heating rate of 10 K/min. A small amount of sam-
paper is to exemplify the recrystallization process of two molec-
ple (less than 5 mg) was enclosed in a hermetic aluminum pan.
ular compounds studied in our group: L-arabitol8,9 and RS
Measurements were performed under dry helium (at flow rate
ibuprofen.10,11 These compounds share the property of being
of 25 mL/min) to improve the thermal conductivity. A liquid
good glass formers and present monotropic polymorphism.
nitrogen cooling system was used in order to reach temper-
More specifically, we will pay attention to a possible separation
atures as low as −130◦ C. Temperatures and enthalpies were
between zones of preferred nucleation and preferred growth
calibrated using indium at the same heating rate and the same
and to the interplay of crystallization of the metastable poly-
environmental conditions as the experiments.
morphs.
Powder X-ray diffraction (PXRD) experiments were per-
This new analysis offers an opportunity:
formed with an INEL CPS 120 diffractometer (8Cu K"1 = 1.54056
Å) equipped with a 120◦ curved position sensitive detector cou-
pled to a 4096 channel analyzer. A Cryostream Plus controller
r to draw a TTT (temperature, time, transformation rate) from Oxford Cryosystems was used to regulate the tempera-
diagram, which is a familiar concept in the domain of met- ture. Samples were placed into Lindemann glass capillaries
allurgy but less so in pharmaceutical science; (Ø = 0.7 mm).
r to show the interest of a scaling analysis of the transfor-
mation curves for identifying changes of kinetic regimes;
r to define a simple differential scanning calorimetry (DSC) RECRYSTALLIZATION OF L-ARABITOL
protocol which allows the preferred zones of nucleation L-Arabitol is liquid above Tm I = 101◦ C8,9 and can be deeply
and growth to be located; undercooled even at rates as low as 2◦ C/min to give rise to a
r to highlight the role of a transient polymorphic form as a calorimetric glass transition at Tg ≈ −14◦ C (scan rate 5◦ C/min).
facilitator of nucleation and (or) growth of the stable form;
r to point out the role of cracks as sources of heterogeneous
Recrystallization can only be observed upon reheating after su-
percooling to a sufficiently low temperature and using a suffi-
nucleation. ciently slow heating rate (see, for example, Fig. 3).

Overall Crystallization Kinetics of the Undercooled Melt


RS ibuprofen is particularly illustrative of the resistance of To determine the overall crystallization rates, amorphous L-
an ordered racemic compound against nucleation in contrast arabitol was prepared by quenching the melt at different tem-
with the relative ease of nucleating a more disordered form at peratures ranging from Tg to Tm . The recrystallization was fol-
low temperature. Results obtained from the two compounds al- lowed isothermally using DSC. In this condition, the exothermic
low the important role played by the interfacial energy in the heat flow is directly proportional to the enthalpy release rate
separation of nucleation and growth, and on phase selection, and thus to the crystallization kinetic rate. At each anneal-
to be stressed. The order/disorder contrast between the crys- ing temperature, the samples were followed for periods of up to
tallized phase and the parent undercooled melt is certainly of 20 h. For T > 50◦ C, no recrystallization could be detected during
great importance in modifying the entropic contribution of the a reasonable annealing time. At lower temperatures, however,
free energy surface. A possible link with modification of molec- the annealing duration was enough to recrystallize totally the
ular mobility at the interface will be briefly addressed. initial amorphous phase. Figure 2 shows a typical isothermal

Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014 DOI 10.1002/jps.24016
MINIREVIEW 2617

Figure 2. Typical DSC isothermal thermogram (T = 25◦ C) showing recrystallization of L-arabitol and corresponding time dependence of the
fractional volume X(t) recrystallized.

T (°C)
Maximum rate
60
of growth

50 ● ●●
5% ● ● ● 90%
40 ● ● ●

● ● ●

30 ● ●●

●● ●

20 ● ● ●

● ● ●
Figure 3. Temperature dependency of the time evolution of the frac-
tional volume X(t) of recrystallized L-arabitol in phase I. The different 10
temperatures are that reached after a quench from the melt (obtained Maximum rate
from DSC isothermal measurements). Nmax
of nucleation

DSC signal (at T = 25◦ C) with a peak in the enthalpy release at 200 400 600 800
time tpeak ≈ 115 min, which corresponds to the maximum rate
of crystallization. Time integration and normalization of this Time (min)
signal gives the time (t) evolution of the fraction of material
Figure 4. (L-Arabitol) TTT (Temperature, Time, Transformation
transformed X(T,t), which has a typical sigmoidal shape. X(t) rate) diagram derived from Figure 3. Each line corresponds to a given
varies from 0 to 1 during the crystallization. The observed time value (5%, 50%, 90%) of the fraction of material transformed.
evolutions of X(t) for various temperatures between 15◦ C and
50◦ C are summarized in Figure 3. The progress of the isother-
mal crystallization as a function of time and temperature is con- tion time then decreases when temperature decreases because
veniently represented by the TTT diagram shown in Figure 4. the thermodynamic driving force increases. At temperatures
The TTT diagram is constructed from the X(T,t) curves and lower than that of the nose, the crystallization time increases
shows a nose-shaped feature of most rapid isothermal crystal- as molecular mobility decreases when approaching Tg .
lization at about 30◦ C, where 1 h is enough to fully crystallize
the DSC sample. When cooling the melt, the temperature range Localization of Preferred Temperatures of Nucleation and Growth
of this nose must be passed quickly in order to successfully vit-
Scaling Analysis of the Growth Curves
rify the sample. The overall behavior can be understood by the
fact that the crystallization rate is obviously zero at the equilib- There are several factors that combine to determine X(T,t) and
rium melting temperature Tm , where an infinite time would be the overall crystallization kinetics. Among these factors are the
needed to crystallize the sample because the thermodynamic nucleation rate (N) and the growth rate (V = linear velocity).
driving force disappears at this temperature. The crystalliza- When both events occur together during the recrystallization,

DOI 10.1002/jps.24016 Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014
2618 MINIREVIEW

1.2

X (t / t0.5)
1.0

0.8

0.6
0.5
0.4
15°C
20°C
0.2 25°C
30°C
15°C T 40°C 35°C
0.0 40°C
45°C 50°C 45°C
50°C
–0.2
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2

t / t0.5
t/t(1/2)

Figure 5. (L-Arabitol) The scaled curves of X(t) (corresponding to those shown in Fig. 3) plotted in terms of the scaled time t/t0.5 .

the most simple model for X(t), which takes into account at the from which crystallization could propagate by growth (in that
same time the effective decrease of the nucleation efficiency as case an Avrami exponent d + 1 < 4 would be expected). This
the untransformed fraction 1 − X(t) decreases and the impinge- simple scaling analysis provides insight into the mechanism of
ment of neighboring grains at the end of crystallization, is that crystallization at the highest temperature: the nucleation rate
of JMAK (Johnson, Mehl, Avrami, and Kolmogorov).12 is very small: about 7 h are needed for one nucleus to appear
For homogeneous nucleation and isotropic 3D growth: in the DSC sample. Growth from this nucleus is comparatively
fast and the sample is fully crystallized in less than 1 h. From
X(t) = 1 − exp[−6(t/t0 )4 ](6 = 4B/3) this analysis, we can conclude that N and V have their maxima
in distinct temperatures ranges.
This is a universal sigmoidal function of the time scale:
Maxima of Nucleation and Growth Rates of the Stable Phase I
3 −1/4
t0 = (NV ) We have used a DSC instrument working in temperature scan-
ning mode to locate the approximate temperature ranges of the
where 4 (= d + 1 where d is the dimensionality of the space in respective maxima of N(T) and V(T). In the following experi-
which the nucleation and growth transformation is proceeding) ments, crystalline L-arabitol samples were heated and held at
is the Avrami exponent. X(t) is an almost symmetric sigmoid 117◦ C (≈Tm + 15◦ C) for 15 min to ensure complete melting.
on both sides of the half transformation point (occurring for Melted samples were then cooled to different temperatures Ta
t0.5 ≈ t0 ). As a result, we expect that X(t) considered at different < Tm at the highest achievable cooling rate allowed by the DSC
temperatures—with very different values for N and V giving (>20◦ C/min). The samples were then kept for 60 min at Ta and
different values of t0 —can be scaled onto a universal curve afterward reheated at 2◦ C/min by DSC.
once plotted versus t /t0.5 . Figure 6 shows the DSC curves recorded upon heating sam-
Figure 5 shows the X(T,t) curves corresponding to those of ples cooled to temperatures Ta ranging from −83◦ C to 82◦ C.
Figure 3 plotted against a scaled time parameter (t/t0.5 ). At the No calorimetric events can be observed on the DSC heat-
lowest temperatures (15◦ C ≤ T ≤ 40◦ C), scaled curves super- ing curves for samples annealed at Ta > 35◦ C. The absence of
impose rather well showing that scaling is well satisfied. This a melting endotherm in particular shows that no crystalliza-
suggests that in this temperature domain, the recrystallization tion process occurred during the cooling/heating cycle. For Ta
kinetics obey the same law and correspond to a nucleation and < −10◦ C, we observe a very small endothermic manifestation
growth process. This plot reveals that kinetics recorded at the of melting, which shows that only an extremely limited recrys-
highest temperatures (45◦ C and 50◦ C) deviate strongly from the tallization (less than 3%) occurs during the cooling/heating cy-
universal sigmoidal symmetric shape. The 50◦ C growth curve cle. Throughout this Ta range, the enthalpy of melting is very
has a pronounced step like behavior associated to a long time small and independent of Ta preventing any variability from
lag followed by a very fast crystallization. This could only be de- being detected. Furthermore, crystallization exotherms on the
scribed by an unphysically high value of the Avrami exponent DSC curves are not detectable. On these curves, we can see
d + 1 > 10. Such behavior indicates that on approaching Tm nu- the localization of the glass transition temperature at Tg ≈
cleation and growth do not occur simultaneously in a continu- −12◦ C. These results show that the sample was totally amor-
ous process, and also shows that there are no preexisting nuclei phized during cooling, which is consistent with what can be

Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014 DOI 10.1002/jps.24016
MINIREVIEW 2619

Figure 7. (L-Arabitol) DSC heating scans after applying isothermal


annealing of different short durations ta to samples quenched from the
melt to Ta = 20◦ C.

mations that have occurred at Ta , the annealed samples were


heated to 70◦ C and held at this temperature for 30 min where
Figure 6. (L-Arabitol) DSC curves recorded upon heating (heating
growth only is expected. In the hope of learning something
rate of 2 K/min) for samples undercooled at temperatures Ta ranging
about the nucleation mechanism itself, ta durations smaller
from −83◦ C to 82◦ C and annealed for 60 min at this temperature.
than that which was used in the production of Figure 6 were
employed. After a short annealing time, a small melting peak
is observed at Tm II = 80◦ C (onset), which is significantly lower
predicted from the global TTT representation and the position
than the value observed initially. Increasing ta gives rise to
of the nose of most rapid crystallization. In particular, no signif-
the progressive appearance of a melting peak at Tm I and a de-
icant nucleation occurred in this low temperature domain. On
crease in the amplitude of that occurring at Tm II . These results
the contrary, for annealing temperatures Ta ranging between
show that annealing initially induces nucleation of metastable
−5◦ C and 30◦ C, pronounced crystallization exotherms, followed
form II of L-arabitol, which melts at lower temperature. When
by melting endotherms, are observed. The exotherms have their
ta is increased, the nuclei of phase II progressively disappear,
maxima located not far below the melting point, in the range
whereas the nuclei of the stable phase I form and cause melting
60◦ C–95◦ C, that is, at temperatures clearly separated from the
at the higher temperature. The existence of a metastable phase
annealing Ta domain. The preceding analysis of isothermal ki-
for L-arabitol has been postulated previously.13
netics - particularly that recorded at 50◦ C demonstrates that
A sample fully converted to form II can be obtained by
above 50◦ C, we enter a temperature range where growth ve-
quenching L-arabitol from the melt to a low temperature (20◦ C)
locity is increasing strongly, and nucleation in a volume as
and holding it there for 2 min followed by a rapid increase to
small as that of the DSC sample becomes negligible during
70◦ C (Tm II – 10◦ C) where growth of phase II is maximum. Full
the time taken to cross the recrystallization exotherm. The in-
crystallization at this temperature occurs in less than 2 h. No
terpretation is that the exotherms correspond to the growth of
contamination by phase I is observed during this process, which
crystals nucleated at significantly lower temperature. If so, as
once again confirms that nucleation of phase I only operates at
exotherms occur in the same temperature range, V(T) values
low temperature and after a long enough time. Phase II has
are expected to be similar. The measured enthalpy of recrystal-
lower melting temperature (Tm II = 80◦ C), lower melting en-
lization is roughly proportional to the number of nuclei formed
thalpy (Hm ≈ 28.8 kJ/mol) and melting entropy (Sm ≈ 81
at lower temperature. The maximum enthalpy of recrystalliza-
J/mol K) than phase I (Tm I ≈ 101◦ C, Hm ≈ 36.5 kJ/mol and
tion is obtained for an annealing performed at Ta = 5◦ C, and so
Sm ≈ 98 J/mol K). This is sufficient information to conclude
this is the temperature where the maximum of N is expected.
that phases I and II form a monotropic set.
To get an insight into the features of this transient
Transient Metastable Phase
metastable phase, long isothermal (at Ta = 0◦ C) time-resolved
Careful investigations of the crystallization conditions allow X-ray measurements were performed for durations of up to 14 h.
the detection of the transient appearance of a new metastable In order to increase the intensities of Bragg peaks and to help
form of L-arabitol (which will be referred to below as phase II). in resolving the kinetics, a procedure was employed to enhance
We report two experiments (PXRD and DSC) that display this nucleation. The melted sample was quenched in liquid nitrogen
additional complexity and the influence of this less stable form (to well below Tg ) before investigation. The associated thermal
on the crystallization of form I. shock produces cracks in the glassy sample, which then sup-
Figure 7 shows DSC heating scans obtained after applying port heterogeneous nucleation upon reheating. The occurrence
isothermal annealing of different durations ta to samples of L- and effect of cracks as nucleation facilitators will be discussed
arabitol quenched from the melt to Ta = 20◦ C (a temperature and visualized in more detail below for the case of Ibupro-
situated in the domain of favored nucleation of phase I). Be- fen. X-ray results are shown in Figure 8a. During the isother-
fore recording the DSC scans, in order to amplify the transfor- mal annealing, peaks characteristic of the metastable phase II

DOI 10.1002/jps.24016 Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014
2620 MINIREVIEW

Figure 9. (RS ibuprofen) DSC curves recorded upon heating (heating


rate 10 K/min) for samples undercooled at different temperatures and
annealed for 360 min at this temperature.

to the maximum value observed for phase II during anneal-


ing). Clearly, phase II develops first and its rate of evolution
decreases when phase I starts to grow. Phase I then develops
to the detriment of phase II. Finally, phase I only remains and
continues to crystallize even after phase II has disappeared.

RECRYSTALLIZATION OF RS IBUPROFEN
The investigation of the recrystallization and polymorphism of
racemic ibuprofen (ra-Ibu)10,11 provides additional understand-
ing and insight into factors that have an impact on nucleation
and growth. The typical feature distinguishing the crystal state
of the initial compound (hereafter designated by phase I) is
that the unit cell contains an equal number of molecules of
each enantiomer S(+) and R(−), which are positioned in an or-
dered manner. ra-Ibu is representative of a large class of active
ingredients. The racemic compound I has significantly higher
melting temperature (Tm I ≈ 71◦ C), greater melting enthalpy
Figure 8. (a) (L-Arabitol) Time sequence of the X-ray diffraction pat-
(Hm ≈ 26.9 kJ/mol) and melting entropy (Sm ≈ 78 J/mol K)
terns during the recrystallization at Ta = 0◦ C. To speed up the evo- than each S(+) or R(−) crystalline isomer (Tm ≈ 46◦ C–52◦ C,
lution, the melted sample was previously quenched in liquid nitrogen Hm ≈ 18 kJ/mol and Sm ≈ 55 J/mol K).14 In addition to the
to induce cracks and heterogeneous nucleation before recording. Peaks stable ordered racemic compound I, the existence of another
arrowed in blue and red are for phase II and phase I, respectively. metastable crystalline form II that melts at Tm II ≈ 17◦ C has
(b) (L-Arabitol) Time evolutions of the sum of the integrated intensities been shown.10,11,15 We present a brief account of the crystalliza-
of the representative Bragg peaks shown in Figure 8a (normalization tion conditions of these two phases and their relationship10,11
to the max observed for phase II). with the aim of deriving trends linked to the bicomponent na-
ture of the compound.
appear initially and then disappear when Bragg peaks of phase
Crystallization of Stable Phase I
I progressively increase. To investigate quantitatively the time
evolutions of the amounts of phases II and I, we have selected In order to locate the preferred zones of nucleation and growth
three specific Bragg peaks for each phase, which are sufficiently of phase I, we have applied the protocol described above for L-
spatially resolved and intense. Peaks arrowed in blue and red in arabitol (see Fig. 6) consisting of a cycle of melting—quenching
the figure were used for phase II and phase I, respectively. The to various annealing temperatures Ta and holding at that tem-
time evolution of the amounts of these two phases is thus repre- perature for a given duration ta and reheating. Figure 9 shows
sented by the evolutions of the sum of the integrated intensities that an exotherm of recrystallization is detectable at high tem-
of the three representative Bragg peaks. Figure 8b shows the perature (between 35◦ C and 65◦ C) upon heating, only if the
parallel evolutions of these quantities (normalized for clarity sample has been previously annealed for a sufficient period

Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014 DOI 10.1002/jps.24016
MINIREVIEW 2621

effective only between −130◦ C and approximately −40◦ C. As


may be deduced from Ref.16, nucleation of phase II well above
Tg (at 0◦ C) is much more homogeneous and slow. The long,
irreproducible time lag preceding crystallization after a direct
quench to 0◦ C, and the following 8–10 h needed to grow nucleus
of phase II, are in fact consistent with a rare, homogeneous nu-
cleation event. The rather short recrystallization exotherm at
0◦ C obtained after cooling a sample to low temperature (see
Fig. 11a) is thus mainly linked to the growth process of a rela-
tively highly nucleated amorphous sample. Interestingly, such
a growth process occurs not far below the melting temperature
of phase II (Tm II ≈ 17◦ C). The present experiment further sug-
gests that even if the rate of nucleation of phase II is increased
by cracks, it may occur below Tg .

Relationship Between Crystallizations of Stable (I) and Metastable


(II) Phases
It should be noted that even if heterogeneities promote nu-
cleation of phase II, they do not invert the order of appear-
ance of the phases. The initial transformation to phase II, how-
Figure 10. (a) (RS ibuprofen) DSC heating curve obtained after un- ever, accelerates the subsequent transformation toward phase
dercooling of the melt well below Tg (typically to −130◦ C) then reheat-
I. Isothermal DSC (Fig. 11a) shows the chain of the different
ing to 0◦ C for a 30 min. annealing. It reveals successive melting of
phase II and I. (b) (RS ibuprofen) Gibbs diagram.
steps when annealing is continued. The exotherm takes a bi-
modal aspect whose second bump corresponds to the II to I
conversion. The latter develops on a longer timescale than the
between −40◦ C and −10◦ C. This recrystallization is that of initial crystallization of phase II. Hot-stage microscopy (Fig.
the racemic crystal (phase I) as melting occurs at 71◦ C. These 11b) allows the visualization of the crystallizations and their
results show that nucleation and growth zones are very well intertwining at 0◦ C after a similar melt-cool-heat treatment.
separated, which is consistent with the very good glass forming Part of the sample is sandwiched between two cover slips (be-
nature of ra-Ibu (Tg ≈ −45◦ C). Nmax is located at about −30◦ C, tween the dashed lines in the Fig. 11b) and was free of cracks
not far above Tg . Vmax is obviously impossible to locate precisely that instead appear in the nonconfined area. This setup allows
in a DSC heating scan, but the temperature range where rapid visualization of homogeneity or heterogeneity of the nucleation
growth is expected is situated not far below melting. It should and growth process. The succession of slides shows that a ma-
be noted that an annealing duration as long as ta = 360 min was jority of crystals of phase II develop in the outer part and pro-
necessary to trigger a noticeable nucleation. As will be shown gressively grow into the confined amorphous sample (in black).
below, the initial crystallization of phase II can significantly After 2 h, phase I starts developing at the expense of phase
increase further formation of phase I. II (interface delimited by dotted line). Heating the partially
converted sample allows the identification of the two phases
Crystallization of Metastable Phase II by their melting temperatures. After a direct quench to 0◦ C,
the same succession of transformations is observed, but over a
A new crystalline polymorph (II) was found for which nucle-
much longer time scale.16
ation and growth from the melt were favored at temperatures
significantly lower than that which control crystallization of
Specificities of Phase II
phase I. Figure 10a shows the most efficient operational method
of rapidly and completely recrystallizing the quenched liquid Phase II has a significantly lower melting temperature (Tm II
into form II.10 This method consists of the following steps: un- = 17◦ C), smaller melting enthalpy (Hm ≈ 8.8 kJ/mol) and
dercooling of the melt well below Tg (typically to −130◦ C) then melting entropy (Sm ≈ 30 J/mol K) than racemic phase I.
heating to 0◦ C for a 30 min. period of annealing. It is this proto- Figure 10b shows a schematic Gibbs free energy (also known
col that was initially used to prepare powder samples suitable as free enthalpy on account of its functional definition: G =
for structural analysis of phase II.15 It should be noted that H – T·S) diagram, which situates phase I and II as well as
phase II can also be crystallized after directly cooling to 0◦ C, the liquid phase. Slopes were calculated from the evaluation
but this process takes about a day and involves a long time lag of the melting entropies of the crystalline phases. Schematic
of unpredictable duration (typically 14 h).16 The difference in curves are plotted assuming that the entropies are tempera-
crystallization behavior with these two protocols is associated ture independent. The picture is that of a monotropic situation
with the fact that deep undercooling below Tg induces cracks where the metastable phase appears to be unusually close in
in the glass, which favor heterogeneous nucleation during the entropy to the liquid. There are a number of features that show
reheating step. The presence of cracks in a glassy sample has that the crystalline structure of the metastable phase II of RS
been recognized as being a factor which significantly affects Ibuprofen is highly disordered. Refinement of the crystalline
nucleation.17 The DSC sample is thus more highly nucleated structure from X-ray data using the hypothesis of an ordered
and less than 1 h is enough to crystallize phase II at 0◦ C. As arrangement was only possible by introducing an anomalously
cracks are cured not far from Tg = −45◦ C, the efficient - crack high effective Debye Waller factor.15 Disorder is indicated in X-
assisted - heterogeneous nucleation process is expected to be ray diffractograms, which show a very rapid decrease of peak

DOI 10.1002/jps.24016 Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014
2622 MINIREVIEW

Figure 11. (a) (RS ibuprofen) DSC thermogram for isothermal recrystallization recorded at 0◦ C after deep undercooling to −130◦ C (induction of
cracks). Successive crystallizations of phase II and I are appearing. (b) (RS ibuprofen) Corresponding time sequence of micrographies. Dash lines
delimit part of the sample between two microscope slides. Dotted lines delimit the growing phase I. The figure shows the progressive development
first of phase II then of phase I at the expense of phase II.

intensities with increasing Bragg angle. Low-frequency Raman neous time-scale evolutions where it is the physical law itself
data of the metastable phase also show the features of a disor- that is changing.
dered crystal.18 Analysis of the first stage of crystallization in the tem-
perature domain where the maximum of nucleation Nmax of
phase I of both compounds (typically not far above Tg ) is ob-
served gives insights into the nucleation pathway of the stable
DISCUSSION form. In both cases, it generally involves a first step of forma-
Remarks about Behaviors Identified for L-Arabitol and RS tion of a more disordered metastable form (phase II), which
Ibuprofen only emerges transiently. This is another example of cross-
nucleation between polymorphs.19 We may however suspect a
The two compounds exhibit monotropic polymorphism. The double effect: that of a secondary nucleation on the less-stable
metastable phase II of ra-Ibu, however, has a significantly lower phase, which occurs concomitantly with a II to I transforma-
Tm II with regard to Tm I and a much smaller melting enthalpy tion. Phase II thus facilitates nucleation of form I. However, the
(and entropy). This is an indication of the significant degree temperature where the maximum rate of the latter is found to
of disorder characteristic for phase II of ibuprofen. Concern- occur is largely decoupled from temperatures of fast growth
ing the recrystallization from the melt of the stable forms, it velocity.
was concluded in both cases that the maxima of N and V (Nmax In both cases, we have seen that initial formation of cracks
and Vmax ) are set well apart from each other. Nmax is situated in the deeply quenched glass is able to promote substantial
slightly above Tg , whereas Vmax is located slightly below the heterogeneous nucleation even below Tg , particularly with ra-
melting temperature. This separation is clearly the origin of Ibu. This is not incompatible with more homogeneous, less fre-
the high propensity of both compounds to form glasses. The quent, nucleation above Tg . Heterogeneities simply amplify the
kinetics of recrystallization of the stable phase is, however, sig- nucleation rates in the temperature domain where homoge-
nificantly faster with L-arabitol. As a result, a TTT diagram neous nucleation slows down significantly because of the lack
for crystallization of phase I could be plotted, and its nose lo- of molecular mobility. The presence of cracks should certainly
calized, which would be practically impossible with ra-Ibu. It is be taken into consideration when investigating the stability of
important to note that Nmax and Vmax are situated on either side solid amorphous compounds.
of the TTT nose. This is a consequence of the fact that the char-
acteristic times for nucleation and growth combine in an actual
effective characteristic time t0 for global recrystallization. This
Remarks about Mechanisms and Data Analysis
shows clearly that it will not be possible to extract activation
energy for the global phenomena as is sometimes attempted. The kinetics of first order transformations such as recrystal-
We have shown the interest of using a scaling analysis de- lizations is a very complex and challenging question both in
rived from a simple model of nucleation and growth to local- physics and in material science. It is out of the question to at-
ize temperature ranges where kinetics obey the same phys- tempt an in-depth analysis of our observations. Nevertheless,
ical law. It is very useful to identify scaling violations, that they can be used to gain insights using the most elementary
reveal changes of regimes, to prevent the attribution of erro- models.

Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014 DOI 10.1002/jps.24016
MINIREVIEW 2623

Classical Nucleation and Growth Picture in Brief spond different modes of crystal growth: normal (or continuous)
growth for diffuse interfaces, lateral growth for sharp inter-
Homogeneous nucleation is a thermally activated process
faces. The latter can be itself subdivided in two main different
by which some crystalline clusters have to be formed in order
types: spiral (or dislocation) and 2D nucleation.24,25 These dif-
to enable a phase transformation to occur.20–22 In the conven-
ferent mode of growth give rise to different expression of the
tional Becker-Döring capillarity theory (see for example 23 ), the
temperature-dependent linear growth rate V(T). V(T) can be ex-
free enthalpy of formation of the cluster is assumed to result
pressed in a compact form, which summarizes the results given
from the competition between an unfavorable positive interfa-
in the literature7,24 for the different modes of crystal growth. It
cial free enthalpy (() and the negative bulk driving force (G).
is derived from the initial semi-phenomenological expression of
G is the free enthalpy (Gibbs free energy) difference between
Turnbull20 and can be written in the form:
the parent melt and product crystalline phase. Interfacial en-
ergy prevails for small crystalline clusters (nuclei). The latter
V(T) ∝ V0 (T) ·  · [1 − exp(−G/RT)]
have a good chance of growing if they exceed a certain criti-

cal size r3D (T) and overcome the nucleation barrier g∗3D (T).
Nucleation events thus occur with a steady rate: To cross the interface, a molecule must overcome an acti-
vation free energy Ga , which is roughly that of the diffu-
sion coefficient of a molecule in the liquid, so that: V0 (T) ∝
N(T) = f 0 (T)exp(−g∗ /RT)
exp(−Ga /RT)].  is a function of G, which depends of the
specific mechanism of growth. The term inside the brackets ex-
where f0 is an “attempt frequency” for the addition of molecules presses the influence of the thermodynamic driving force G
from the metastable melt to the stable crystal across the inter- on the net flux from liquid to crystal.
face. For a spherical 3D nucleus:
r For normal growth, molecules are directly incorporated

r3D = 2(/G and g∗3D = 16B(3 /3G2 at the growth sites of the crystal surface. In that case: 

= const. ( proportional to the relative number of growth
If we assume that the driving force G is proportional to the sites). This specific form of the equation predicts a quasi
undercooling T = Tm – T: linear increase in growth velocity with decreasing tem-
perature (increasing T) at small undercoolings and a
G ≈ Sm · T = Hm · T/Tm decrease at much lower temperatures as the viscosity in-
creases. As for N(T), a maximum Vmax is thus expected
where Hm and Sm are respectively the latent heat and en- at some undercooling. It should be noted that, contrary to
tropy of melting, then: the case of nucleation, the interface free energy ( has no
explicit effect on the normal growth velocity, and thus on
the position of Vmax .

r3D (T) ∼ 1/T and g∗3D (T) ∼ 1/T2 . r For lateral growth of a dislocation clean surface, nucle-
ation of a 2D disc-shaped layer is a prerequisite for the
In a first approximation, and on a short temperature inter- continuation of the process. Critical radius and nucle-
val, we can consider that f0 (T) decreases with temperature in an ation barriers associated to this 2D process are now given
Arrhenius way and scales roughly with the diffusion constant by:
of the bulk liquid. The effect of f0 (T) becomes significant near
the glass transition. The two terms of N(T) vary with tempera- ∗ (
r2D = and g∗2D = B(2 h/G
ture in opposite directions and a maximum Nmax is expected at G
some temperature above Tg .
If the nucleus forms in contact with other phases, impurity where h (of the order of nm) is the height of the 2D nuclei.
sites, or containers walls, the interfacial free enthalpy term is The limitation associated to the 2D nucleation is taken into
reduced, and the heterogeneous nucleation barrier is smaller account in the expression of  ∼ = const. × exp(−g∗2D /RT).
than g∗3D by a catalytic potency factor f < 1. As will be discussed The interface energy ( may that time have an influence on
below, cracks, which may form in the molecular vitreous solid, the position of Vmax . Vmax is expected at a temperature some-
can strongly enhance crystal nucleation in a temperature do- what lower than for normal growth. To get an idea of the pos-
main (T < Tg ) where slow molecular mobility is expected to sible influence of the interface energy on the proximity of Nmax
freeze the nucleation process. and Vmax , we calculate the ratio of the nucleation barriers:
Growth Process. Once a viable supercritical nucleus is
formed, it will continue to grow during isothermal annealing. In g∗2D /g∗3D = (3/16)(G × h)/(
the following, we consider only the case where density change
and component redistribution are negligible (a process con- Taking the numerical values available for the " phase of
trolled by the interface). There are two types of crystal/liquid indomethacin (Fig. 9 of Ref. 7): 0 (at Tm ) < G (kJ/m3 ) < 2.5 ×
interfaces: an interface which is rough (or diffuse) at the molec- 104 (at Tg ), ( ∼
= 0.02 J/m2 , h ∼
= 1 nm, we find:
ular level or a flat (or smooth) sharply defined interface24,25 .
Taking into account the estimations made by Jackson,26,27 flat 0(at Tm ) < g∗2D /g∗3D < 1/4(at Tg )
interface is expected for compounds having high entropy of
melting (Sm > 4R ∼ = 33 J/mol K), whereas rough interfaces This simple evaluation shows that the 2D nucleation barrier
are expected for substances having low Sm values (Sm < is significantly lower than the 3D one on all the temperature
2R ∼= 16.6 J/mol K). To these different types of interfaces corre- range situated between melting and glass transition. The 2D

DOI 10.1002/jps.24016 Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014
2624 MINIREVIEW

nucleation which is associated to a lateral growth should thus


have a limited influence on the position of Vmax with regard to
the position of Nmax .

Impact of the Interface Free Enthalpy


Nmax and Vmax are both the result of the competition between
the increasing thermodynamic driving force and the decreas-
ing molecular mobility. Their temperature positions, however,
differ mainly because of the effect of the surface tension (. As
we just saw, this should be valid whatever the growth mech-
anism. Concerning recrystallization, ( has a double influence:
first on the separation between Nmax and Vmax then on the phase
selection.
( has a strong impact on the nucleation rate via g∗3D (T), but
has no (or limited for lateral growth) explicit influence on the Crystal Liquid
growth rate (except in the very early stage of growth). Because
of (, the decrease in growth velocity due to the slowing down of Hl
molecular mobility occurs at a smaller undercooling than the TMSl
corresponding decrease in nucleation rate. Nmax is expected at a TM S
T S (interf)
temperature lower than Vmax . A high value of ( and thus of g∗3D
(T) strongly reduces the influence of the driving force G(T) TM S (bulk)
on nucleation and contributes to an increase in the distance = HM
between the two maxima. As a result, undercooling and glass
formation become easier. Hc
The phase selection during recrystallization of an under-
TMSc
cooled melt is strongly dependent on the structure dependence
of (. For a stable phase, the driving force G(T) is higher than Figure 12. Schematic representation of the crystal/liquid interface
for a metastable phase. At a given T < Tm , the nucleation bar- showing short-range molecular ordering. Corresponding evolution of
rier may, however, be smaller for the metastable phase if ((T) is entropy that illustrates the negentropic approach of Spaepen (see
sufficiently small, resulting in the preferred nucleation of the Ref. 28)
metastable phase.
According to this picture, the stability of the amorphous state
Figure 12 shows a schematic representation of the evolution
(via nucleation/growth decoupling), phase selection, and ulti-
of molecular ordering in the liquid, at the interface, and of the
mately the Ostwald rule of stages have mainly a kinetic origin
corresponding physical quantities: enthalpy H and entropy S.
modulated by the interface.
For simplicity, the temperature is taken to be Tm so that there
At a given T < Tm , the mobility of molecules within the bulk
is a balance between bulk high enthalpy and high entropy:
amorphous state is the same irrespective of the crystallized
Hm = Tm Sm . At the interface, the balance is no longer sat-
phase, and has thus a priori little influence on this modulation.
isfied, which gives rise to the excess free energy (. Obviously,
We will mention below that molecular mobility in the liquid
the maximum value of ( is Hm = Tm Sm (bulk). Molecular or-
near the interface can, however, be modified as a result of the
ganization in the interface locally reduces the configurational
structural short-range organization of the molecules needed to
entropy jump, which becomes:
adjust to the configurational constraints of the crystal plane.

Influence of Disorder and Correlations on the Interface Free S(interface) < S(bulk).
Enthalpy
At Tm , the difference in free enthalpy (per molecule at the
It must be remembered that the surface tension ( is a free interface) of the liquid, which develops an interface, and that
enthalpy (Gibbs free energy) and as such has an energetic and of the reference bulk is: Tm [S(bulk) − S(interface)].
entropic contribution. Turnbull,20 then Spaepen28 and Oxtoby29 Spaepen28 proposed the following formal expression for (:
pointed out that atoms or molecules of the liquid have to in-
crease ordering near the crystal boundary in order to optimize ( ∝ " T Sm with "
fitting with the translational order of the crystal and even-
tually to allow an ordered embryo to grow. The entropy loss = [S(bulk) − S(interface)]/Sm (< 1)
corresponding to this interface ordering is mainly at the origin
of the crystal melt interfacial free enthalpy. Spaepen justified This model has been successfully used mainly for atomic
the fact (at least for atomic compounds) that entropy rises more systems and to our knowledge only once for molecular sys-
slowly than enthalpy when going from crystal to bulk liquid. tems: n-alkane liquids. Turnbull remarked that the relative
The only available model for ( which takes into account struc- undercoolings for several n-alkane liquids are almost an order
tural aspects is the negentropic model,28,30 which provides an of magnitude smaller than those for other simple molecular
interesting guide to understanding the important potential role melts or for Hg.31 This is related to a very high crystal nucle-
of relative disordering in the crystal embryo and relative order- ation frequency and interfacial energies an order of magnitude
ing in the surrounding liquid. smaller than for Hg or H2 O.21,32,33 An immediate consequence

Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014 DOI 10.1002/jps.24016
MINIREVIEW 2625

of the small undercooling is the difficulty of forming glasses the phenomenon. It has been shown recently36,37 that at the sur-
of the n-alkanes. Turnbull and Spaepen attribute this effect to face of a glassy molecular compound, diffusion of molecules can
the high probability for occurrence of linear configurations in be higher than that in the bulk by six orders of magnitude. This
the melt. Such correlations existing in the bulk mean that the surface mobility is ascribed to the fact that surface molecules
entropy loss in adjusting to the crystal plane is minimized, re- obviously experience interactions that are different from those
sulting in rather small values of ". A negentropic model28,30 in the bulk because of a smaller number of neighbors.
based on this model34 can give a satisfying estimate of the
crystal-melt tension.
If ( is small, crystallization is easier. From the negentropic Growth Velocity and a Possible Change of Mobility Near the
model this can be understood for two main reasons: (1) if a liq- Interface
uid is locally organized, (in a way which mimicks the crystalline Both systems studied show that metastable (more disordered)
order), it does not need much further organization at the inter- crystals have not only a tendency to be more rapidly nucleated,
face, leading to a weak value for ", which for n-alkanes was but also to grow more rapidly than those of the stable phase.
found to decrease with increasing n. (2) The other possibility is Several reasons can be invoked. To simplify, we consider the
that of a disordered crystal giving a relatively low value of Hm crystal growth of the metastable form at a point away from the
= TSm , meaning that the first nuclei to form are those of the very early stage where curvature and capillary effect could slow
metastable phase, as is found for phase II of our systems. The down the growth.
latter have the lowest melting enthalpies and thus the highest
crystalline entropies (as visualized by slopes in the Gibbs dia- r If the Sm of the metastable phase is much lower than that
gram). This applies in particular to the highly disordered phase
II of ra-Ibu, which appears prior to phase I. In contrast, the cor- of the stable phase, a change in the mode of growth mech-
responding liquid phase is a multicomponent one (mixture of R anism can be expected, to become (more) continuous and
and S species), which increases the melting entropy. Ordering faster. It could be the case for phase II of ra-Ibu. For phase
at the interface to adjust to the ordered racemic structure I is II, Sm ∼ = 30 J/mol K. It is a value intermediate between
all the more costly in entropy, which hinders nucleation of this 4R and 2R. For such substances, a transition temperature
from lateral to normal growth is to be expected.24
stable phase. r In the case of L-arabitol, the Sm of phase II is lower than
that of phase I. However, both are much greater than 4R.
Nucleation from the Melt and from the Glass: Influences of Lateral growth is to be expected for both phases. For phase
Cracks and Mobility I, the driving force G(T) is higher than for phase II, which
would contribute to a decrease of the growth velocity of the
A metastable amorphous melt (Tg < T < Tm ) and an unsta-
metastable phase. The 2D nucleation barrier for surface
ble amorphous glass (T < Tg ) obviously differ by their molec-
nucleation may, however, be smaller for the metastable
ular mobilities. Molecular compounds are often fragile glass
phase if ((T) is sufficiently small, resulting in a faster
formers35 and exhibit a very fast slowing down of the main
lateral growth of the metastable phase.
" relaxation at the approach of Tg . Below Tg , the amorphous
system becomes solid with ultraslow bulk " relaxations and re-
maining secondary $ relaxations. Molecular mobility can never From the Turnbull equation for interface-controlled growth,
be a driver of nucleation, which is thermodynamic in nature. It the remaining factor that influences growth velocity is that of
only plays the role of a facilitator. Obviously, some mobility is the “jump” frequency of liquid molecules across the liquid–solid
necessary for crystallization to occur, but simply because mobil- interface. Molecules must overcome an activation energy Ga
ity slows down with temperature, this cannot explain by itself by thermal activation, which, in a first approximation is that
that some nucleation rates seem to rise somewhere in the glassy which opposes the diffusion of molecules in the liquid, that is,
state. When attempting to find or to predict the origin of the in- that involved in viscosity.
stability of an amorphous glass with regard to crystallization, it The jump frequency is proportional to exp(−Ga /RT). How-
is important to pay attention to the homogeneous or heteroge- ever, such a term has to be multiplied by the probability (p)
neous nature of the nucleation. The examples presented above for a molecule to be accommodated on the crystal surface. It is
help to show that, because of its brittle nature, the amorphous expected that because of the crystallographic structure, not all
glass is prone to crack some tens of ◦ C below Tg . This is a fre- these molecules will find a suitable site to stick to the surface.
quently observed phenomenon with molecular compounds that The effective flux of molecules from the liquid to the crystal
can give spikes in the DSC thermogram. Crack generation sub- will be ∝ p exp(−Ga /RT). This induces an apparent modifi-
stantially increases the specific surface area of the compound cation of the activation energy which depends on the degree
which in turn enhances the possibilities of heterogeneous nu- of molecular organization on both sides of the interface. There
cleation. Heterogeneous nucleation rates can be considerably will be a similar flux in the reverse direction, but modified by
accelerated by the effective reduction of interfacial energy.21,22 the thermodynamic driving force G. The situation for racemic
As a result, even if a homogeneous bulk nucleation rate is in- ibuprofen is easy to imagine. The liquid is a disordered mixture
trinsically slow in the glassy state, the presence of cracks can of the two enantiomer molecules (R) and (S). At a given T < Tm ,
multiply this rate substantially. This is to be expected when the activation jump barrier may be effectively smaller for the
reheating the cracked glass and approaching Tg from below.17 metastable phase II if p is large enough. This would result in
Such an effect will no longer occur if cracks are cured near Tg a faster growth velocity for the metastable phase. Indeed, we
when fluidity starts appearing. Nucleation assisted by cracks can imagine that the probability of molecular accommodation
can only occur at temperature lower that the curing tempera- is larger if there is a substitutional disorder between R and S
ture. It must be mentioned that molecular mobility may amplify molecules on the crystal side. The positional molecular order

DOI 10.1002/jps.24016 Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014
2626 MINIREVIEW

which is proper to phase I is much less easy to accommodate. r Because Sm ≤ 2R (16.6 ∼ J/mol K), normal (or continu-
These different considerations show that several factors can ous) growth of rough interfaces is the only possible mech-
combine to make growth of metastable phases faster. anism. Plastic crystals grow rapidly from the melt.
There is another factor that may slow down mobility near the
interface to a greater or lesser extent. As some molecular or-
Such a behavior is expected for recrystallization in a phase
dering of the liquid is expected near the crystal boundary, the
which is highly disordered. Recrystallization of phase II of RS
free energy of activation for boundary migration is certainly
ibuprofen is close to this situation.
higher than the free energy Ga for self-diffusion within the
liquid. The mobility involved locally in the growth mechanism
is thus decreased depending both of the degree of accommo- 2. Other type of bad glass formers are systems that behave
dation needed to achieve a jump and of the degree of short like the n-alkanes with n ranging from 16 to 32. As men-
range ordering in the liquid side of the interface. Mobility at tioned above, the resistance to crystal nucleation of these
the free surface of an amorphous system is the object of intense compounds is very weak. They behave in a way which is
activity.36,37 It would be highly useful to improve understand- closely similar to that of plastic crystals and show very
ing of the mobility at the crystal/liquid interface. In the case reduced undercooling. However, their melting entropy is
of confinement of molecular liquids in nanoporous materials, it high: for Eicosane (n = 20) Sm ∼ = 226 J/mol K.40 In that
is known that some slowing down of the molecular motions at case, it is the entropy loss incurred by the melt when
the interface with the confining material can be observed. In a it adjusts to the constraints of the crystal plane, which
fully similar way, we may expect a modification of the molec- determines the surface tension. If similar intermolecular
ular mobility at the interface with the growing crystal, which correlations exist in the bulk and in the interface, they
depends on the nature of the crystalline polymorph. will not reduce the interface entropy relative to its bulk
value. At Tm , for example, the interface entropy term (Tm
Sm ) will quite fully compensate enthalpy (Hm ) as it
CONCLUDING REMARKS does in the bulk. It results in a low surface tension and a
low nucleation barrier. The result is a weak value of the
Many factors can influence the recrystallization from the melt
scaled interfacial tension. It is the situation considered
and thus the stability of the amorphous systems. If we try to
in the negentropic model of Spaepen30,34 (see Fig. 12).
categorize these factors, we can say broadly that:
3. The third type is that we have found for phases I of L-
arabitol and RS ibuprofen. It is also the situation found
1. Thermodynamic is the driver. for salol and other good molecular glass formers. The sit-
2. Molecular mobility is the facilitator. uation of salol is well documented because it is a model
3. Interface energy is the modulator. of fragile glass former. Crystallization from the melt is
4. Heterogeneities and cracks are amplificators. slow. For this compound, Sm ∼ = 60 J/mol K, Tm ∼ = 41◦ C,

Tg = −58 C◦ 38,41
. Substantial nucleation rate was detected
at temperatures close to Tg (∼ = −53◦ C).42 At this tem-
From the discussion above, it becomes clear that the molec-
perature, undercooled salol was found to crystallize by
ular entropy of melting Sm , which measures the degree of
coalescence of small crystalline embryos rather than by
similarity between crystal and the melt, plays a pivotal role in
usual growth of one initial super critical cluster. Normal
controlling several above quoted factors.
crystal growth was found to occur at higher tempera-
Three types of systems provide useful benchmarks for esti-
tures. This behavior shows that maxima of nucleation
mating the tendency of a molecular compound to recrystallize
and growth rate are largely decoupled with nucleation
from the melt more or less easily. Unlike the last type, the first
rate having a maximum near Tg . As for our systems, eas-
two are that of bad glass formers:
ier recrystallization can be observed on heating. Large
interface energy (connected to large Sm ) leads to the
1. The first type corresponds to molecular crystals, which necessity to undercool the compound deeply in order to
have a low entropy of melting such as: attain a sufficiently small value of the nucleation critical
size and nucleation barrier. As is expected for such high
value of Sm , the growth rate of salol at small under-
Cyclohexanol (Sm ∼ = 5.72 J/mol K, Tm ∼ = 25◦ C38 ); CBr4
cooling exhibits a lateral growth and surface nucleation-
(Sm ∼ = 10.88 J/mol K, Tm ∼ = –10◦ C38 ); succinonitrile (Sm ∼
=
limited behavior. As a consequence, the growth rate
11.1 J/mol K, Tm ∼ = 61◦ C39 ).These compounds have all a plas-
increases more slowly than linear with decreasing tem-
tic crystal phase at high temperature (rotationally disordered
perature (increasing T) at small undercoolings. Neu-
crystalline phase). It is this phase which melts. Melting entropy
mann and Micus27,43 have however shown that a maxi-
is low because the crystalline state is already partly disordered.
mum rate of growth rate (Vmax ) due to viscosity increase
All these systems show very poor undercooling ability for two
is reached at about 24◦ C (T ∼ = 18◦ C). The maximum
reasons:
of growth rate thus occurs at a rather low value of un-
dercooling. It shows that even if growth is limited by
r The nucleation barrier is small, as Sm , which fixes the 2D nucleation, the corresponding barrier g∗2D is not high
maximum value of the surface tension, is small. Molecules enough to push the position of Vmax to temperatures as
of the melt situated at the crystal/melt interface adapt low as that of Nmax . Salol is another example of separa-
with relative ease to the disordered molecules of the tion of the two maxima, which is at the origin of the good
crystal. glass forming ability of the compound.

Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014 DOI 10.1002/jps.24016
MINIREVIEW 2627

To this last type of good glass formers can be added a less 9. Carpentier L, Filali K, Derollez P, Guinet Y. 2013. Crystallization
frequently found situation, which gives rise to a very stable un- and polymorphism of l-arabitol. Thermochim Acta 556:63–67.
dercooled liquid or glass. It is typically that of meta-toluidine.17 10. Dudognon E, Danède F, Descamps M, Correia NT. 2008. Evi-
It was found that the undercooled liquid phase of this com- dence for a new crystalline phase of racemic Ibuprofen. Pharm Res
25(12):2853.
pound persists during extremely long times without showing
11. Dudognon E, Correia NT, Danède F, Descamps M. 2013. Solid-solid
any sign of recrystallization. No crystallization is observed
transformation in racemic Ibuprofen. Pharm Res 30:81–89.
even at extremely low cooling rate of 10◦ C/day. Even cycling 12. Christian JW. 1975. The theory of transformation in metals and
the system between Tg and Tm does not reveal recrystalliza- alloys. Oxford, UK: Pergamon Press.
tion. This allowed to investigate the temperature evolution of 13. Diogo HP, Pinto SS, Moura Ramos JJ. 2007. Slow molecular mobil-
the amorphous structure using very long X-ray experiments.44 ity in the crystalline and amorphous solid states of pentitols: a study
Crystallization requires prior deep temperature excursion well by thermally stimulated depolarisation currents and by differential
below Tg where cracks are formed and trigger the formation scanning calorimetry. Carbohydr Res 342:961–069.
of a very small number of nuclei. Crystallization is then de- 14. Romero AJ, Rhodes CT. 1993. Stereochemical aspects of the molec-
tectable upon reheating in the temperature domain where ular pharmaceutics of ibuprofen. J Pharm Pharmacol 45:258; Xu F,
Sun L-X, Tan Z-C, Li R-L, Tian Q-F, Zhang T. 2005. Low temperature
growth is fast enough (i.e., slightly below Tm ). X-ray44 and
heat capacity of (S)-ibuprofen. Acta Phys Chim Sin 21(1):1–5.
neutron scattering45 investigations of undercooled m-toluidine
15. Derollez P, Dudognon E, Affouard F, Danède F, Correia NT,
show the existence of a pre-peak in the structure factor, which Descamps M. 2010. Ab initio structure determination of phase II of
reveals an important short-range clustering of the molecules. racemic ibuprofen by X-ray powder diffraction. Acta Cryst B66:76.
The origin of the difficulty involved in homogeneous nucleation 16. Williams PA, Hughes CE, Harris DM. 2012. New insights into the
is most probably linked to the need to destroy this local orga- preparation of the low-melting polymorph of racemic ibuprofen. Cryst
nization in order to allow adaptation to the crystal in the crys- Growth Des 12(12):5839.
tal/liquid interface. A significant local increase of entropy could 17. Legrand V, Descamps M, Alba-Simionesco C. 1997. Glass-forming
help to explain the apparent effective high value of the interface meta-toluidine: A thermal and structural analysis of its crystalline
energy. polymorphism and devitrification. Thermochim Acta 307:77.
18. Hédoux A, Guinet Y, Derollez P, Dudognon E, Correia NT. 2011.
Raman spectroscopy of racemic ibuprofen: Evidence of molecular dis-
order in phase II. Int J Pharm 421(1):45–52.
ACKNOWLEDGMENTS 19. Yu TJ. 2006. Kinetics of Cross-Nucleation between Polymorphs.
J Phys Chem 13,110(14):7098.
The authors thank the EU INTERREG IV A 2 Mers-Seas- 20. Turnbull D. 1956. Phase Changes. Solid State Phys 3:225.
Zeeën Crossborder Cooperation Programme for funding. They 21. Kelton K. 1991. Crystal Nucleation in Liquids and Glasses. Solid
are very grateful to Professor G. Zografi for giving permission to State Phys 45:75–177.
use a nice photomicrograph of amorphous indomethacin in the 22. Christian JW. 1975. The theory of transformation in metals and
course of crystallization (from Ref. 5). They thank colleagues of alloys. Oxford, UK: Pergamon.
the group (F. Affouard, N. Correia, L. Carpentier, F. Danède, P. 23. Debenedetti PG. 1996. Metastable liquids concepts and principles.
Derollez, A. Hedoux, Y. Guinet, L. Paccou, and J.F. Willart) for New Jersey: Princeton University Press..
their cooperation and interaction. They also thank the master’s 24. Gutzow IJ. 1977. The mechanism of crystal growth in glass forming
systems. Crystal Growth 42:15; Gutzow IJ. 1980. Kinetics of crystal-
students C. Avellaneda and K. Filali for their participations
lization processes in glass forming melts. Cryst Growth 48:589.
in the measurements. The authors are deeply indebted to Dr. 25. Porter DA, Easterling KE. 1988. Phase transformations in metals
Mark Eddleston for his kind attentive improvement of the En- and alloys. Berkshire, England: VNR International.
glish version. 26. Jackson KA, Uhlmann DR, Hunt JDJ. 1967. On the nature of crys-
tal growth from the melt. Cryst Growth 1:1.
27. Jackson KA. 2004. Kinetic processes. Weinheim: Wiley-VCH.
28. Spaepen F. 1975. A structural model for the solid-liquid interface
REFERENCES
in monatomic systems. Acta Met 23:729.
1. Hancock BC, Zografi G. 1997. Characteristics and Significance of the 29. Oxtoby DW, Haymet ADJ. 1982. A molecular theory of the solid-
Amorphous State in Pharmaceutical Systems. J Pharm Sci 86:1. liquid interface. II. Study of bcc crystal-melt interfaces. J Chem Phys
2. Baird JA, Taylor LS. 2012. Evaluation of amorphous solid disper- 76:6262.
sion properties using thermal analysis techniques. Adv Drug Deliv Rev 30. Spaepen F, Meyer RB. 1976. The surface tension in a structural
64:396–421. model for the solid-liquid interface. Scripta Met 10:257.
3. Willart JF, Descamps M. 2008. Solid state amorphization of phar- 31. Turnbull D, Cormia RL. 1961. Kinetics of crystal nucleation in some
maceuticals. Mol Pharm 5(6):905–920. normal alkane liquids. J Chem Phys 34:820.
4. Ostwald W. 1897. Studien uber die Bildung und Umwandlung fester 32. Turnbull D. 1942. Kinetics of Solidification of Supercooled Liquid
Korper. Z Phys Chem 22:289. Mercury Droplets. J Chem Phys 20:411.
5. Threlfall T. 2003. Structural and Thermodynamic Explanations of 33. Wood GR, Walton AG. 1970. Homogeneous nucleation kinetics of
Ostwald’s Rule. Org Process Res Dev 7:1017. ice from water. J Appl Phys 41:3027.
6. Yoshioka M, Hancock BC, Zografi G. 1997. Crystallization of in- 34. Turnbull D, Spaepen F. 1978. Crystal nucleation and the crys-
domethacin from the amorphous state below and above its glass tran- talmelt interfacial tension in linear hydrocarbons. J Polym Sci
sition temperature. J Phar Sci 83(12):1700–1705. 63:237.
7. Andronis V, Zografi G. 2000. Crystal nucleation and growth of in- 35. Böhmer R, Ngai KL, Angell CA, Plazek DJ. 1993. Nonexponential
domethacin polymorphs from the amorphous state. J Non-Cryst Solids relaxations in strong and fragile glass formers. J Chem Phys 99:4201.
271:236–248. 36. Zhu L, Brian CW, Swallen SF, Straus PT, Ediger MD, Yu L. 2011.
8. Carpentier L, Desprez S, Descamps M. 2003. Crystallization and Surface self-diffusion of an organic glass. Phys Rev Let 106:256103.
glass properties of pentitols: Xylitol, adonitol, arabitols. J Therm Anal 37. Capaccioli S, Ngai KL, Paluch M, Prevosto D. 2012. Mechanism of
Calorim 73:577–586. fast surface self-diffusion of an organic glass. Phys Rev E 86:051503.

DOI 10.1002/jps.24016 Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014
2628 MINIREVIEW

38. Domalski ES, Hearing ED. 1996. Heat Capacities and Entropies of 42. Hikima T, Hanaya M, Oguni M. 1995. Discovery of a po-
Organic Compounds in the Condensed Phase. Volume III. J Phys Chem tentially homogeneous-nucleation-based crystallization around the
Ref Data 25:1. glass transition temperature in salol. Solid State Commun
39. Rai US, Singh OP, Singh NB. 1987. Some thermodynamic as- 93(8):713.
pects of organic eutectics, succinonitrile-phenanthrene system. Indian 43. Neumann K, Micus G. 1954. Die Lineare Kristallisations-
J Chem 26A:947. geschwindigkeit des Salols in dünnen Schichten Z Phys Chem 2:25;
40. Chickos JS, Nichols G. 2001. Simple relationships for the estima- see also [27] pp 274–275.
tion of melting temperatures of homologous series. J Chem Eng Data 44. Descamps M, Legrand V, Guinet Y, Amazzal A, Alba C, Dore J.
46:562. 1997. “Pre-peak” in the structure factor of simple molecular glass for-
41. Moura Ramos JJ, Correia NT, Diogo HP. 2004. Vitrification, nucle- mers. Prog Theor Phys Suppl 126:207.
ation and crystallization in phenyl-2-hydroxybenzoate (salol) studied 45. Morineau D, Alba-Simionesco C. 1998. Hydrogen-bond-induced
by Differential Scanning Calorimetry (DSC) and Thermally Stimulated clustering in the fragile glass-forming liquid m-toluidine: Experiments
Depolarisation Currents (TSDC). Phys Chem Chem Phys 6:793. and simulations. J Chem Phys 109:8494.

Descamps and Dudognon, JOURNAL OF PHARMACEUTICAL SCIENCES 103:2615–2628, 2014 DOI 10.1002/jps.24016

You might also like