You are on page 1of 11

RESEARCH ARTICLES

Milling-Induced Disorder of Pharmaceuticals:
One-Phase or Two-Phase System?
BRIAN S. LUISI,
1
ALES MEDEK,
1
ZHOUYUAN LIU,
2
PRAVEEN MUDUNURI,
1
BRIAN MOULTON
1
1
Vertex Pharmaceuticals, Cambridge, Massachusetts 02139
2
Department of Pharmaceutics, College of Pharmacy, The University of Arizona, Tucson, Arizona
Received 4 November 2011; revised 2 December 2011; accepted 9 December 2011
Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.23035
ABSTRACT: During milling, components are subjected to shear and tensile stresses, which
can result in physical phase transformations. The purpose of the work described in this report
is to understand the pathway by which two test compounds, D-salicin and (-indomethacin, un-
dergo a crystalline to amorphous transformation during cryomilling. The results show that the
transformation cannot be described by a standard one-phase or two-phase disordering mecha-
nism. In the one-phase model, a continuous set of states exist, linking perfect crystalline with
completely amorphous material, whereas the two-phase model of disorder depicts the mate-
rial as a binary mixture of crystalline and amorphous fractions. Instead, a model is proposed
where two one-phase regions, defected crystalline and amorphous regions, are separated by a
distinct transition. © 2011 Wiley Periodicals, Inc. and the American Pharmacists Association
J Pharm Sci
Keywords: crystal defects; amorphous; milling; physical characterization; processing
INTRODUCTION
Organic solids can exist in a variety of solid phys-
ical forms consisting of crystalline polymorphs, de-
fected structures, and amorphous states that have
distinguishing energies and densities.
1
This in turn
can have important implications for physical prop-
erties and subsequent performance.
1–4
Polymorphs
represent different solid-state packing arrangements
of the same chemical composition and are described
by unique crystal structures with characteristic free
energies.
4,5
Converting from one polymorph to an-
other requires passing through a phase transforma-
tion and involves a discontinuous change in the free
energy.
5
The introduction of lattice defects into a
given polymorph results in an increase in energy of
the material relative to the perfect, defect-free crystal
structure. Disruption of the crystal lattice by defect
formationand the creationof amorphous content have
been shown to occur for a variety of pharmaceuticals
and excipients through processing operations such as
milling.
6–10
Additional Supporting Information may be found in the online
version of this article. Supporting Information
Correspondence to: Brian S. Luisi (Telephone: +617-444-6354;
Fax: +617-444-6920; E-mail: Brian Luisi@vrtx.com)
Journal of Pharmaceutical Sciences
© 2011 Wiley Periodicals, Inc. and the American Pharmacists Association
Analogous to polymorphism, the presence of a
first-order phase transition that separates differ-
ent solid amorphous states indicates the presence
of polyamorphism.
11,12
Although the identification
of true polyamorphism for organic glasses in the
pharmaceutical sciences is still a matter of debate,
in a recent article published by Winkel et al.
13
on
solid carbonic acid, a phase transition separating two
glassy states was inferred from computational and
structural X-ray powder diffraction (XRPD) and in-
frared work.
12,13
Polyamorphous, high-density and
low-density glasses are known to exist within the
broader materials science community and have been
demonstrated for several classes of substances includ-
ing inorganic glasses, semiconductors, and ice.
14–17
Because glasses are heterogeneous at the molecu-
lar level, so it is possible to envision cases wherein
differences in local order gives rise to amorphous
structures that are not necessarily separated by
a phase transition.
11,12
Such amorphous structures
would have unique energy and density values but in-
terconversion between them would be a matter of lo-
cal molecular rearrangements without a first-order
phase transition.
Figure 1 shows a volume temperature diagram for
a glass-forming and crystal-forming species from the
liquid melt. Tf1 and Tf2 denote the respective fictive
JOURNAL OF PHARMACEUTICAL SCIENCES 1
2 LUISI ET AL.
Figure 1. Volume–temperature diagram of a glass-
forming and crystal-forming species from the liquid melt.
Tf1 and Tf2 denote the respective fictive temperature in the
T
g
region associated with forming a higher or lower density
glass based upon cooling rate. T
m
is the melting point asso-
ciated with a given crystalline polymorph. Overlaid on the
diagram are disordered crystal states that can arise as a
consequence of processing a particular polymorphic form.
temperatures in the glass transition temperature (T
g
)
region associated with forming a lower or higher den-
sity glass based upon cooling rate. In the idealized
case, the fictive temperature is defined as the tem-
perature at which the liquid structure is “frozen in”
to the solid state. It is also the equivalent tempera-
ture at which a glass would attain thermal equilib-
rium upon heating. A higher density, lower energy
glass is formed at slower cooling rates or following
relaxation of higher energy structures. If the cooling
rate is slow enough then crystallization to a particu-
lar polymorph is induced at the melting point (T
m
).
The perfect, defect-free, stable polymorph lies at the
energetically lowest state in the diagram. As defects
are progressively introduced into a crystal, the sys-
tem increases in energy. An unanswered question is
whether amorphous content arrived at through pro-
cessing crystalline forms equates to the same amor-
phous structure arrived at through melt quenching or,
indeed, any other processing technique such as spray
drying, lyophilization, solvent evaporation, or vapor
deposition. A second question involves the nature of
the transitions involved from increasing or decreas-
ing levels of crystallinity in a sample. Froma practical
perspective, the answers to both questions will dictate
the way disorder could be characterized qualitatively
or quantitatively by analytical techniques.
In a two-phase model of disorder, the material is de-
picted as a binary mixture of two unique states com-
prising amorphous and crystalline fractions.
9
This
model asserts that a decreasing percentage of crys-
tallinity corresponds to a decreasing crystalline frac-
tion in an increasingly large portion of amorphous
matrix. The one-phase model of disorder, in con-
trast, maintains a continuous set of states exists
between perfectly crystalline and completely amor-
phous material.
9
Decreasing crystallinity in the one-
phase model means a reduction of symmetry opera-
tors and an increase in defects, leading to disordered
states lying somewhere, energetically, between per-
fectly crystalline and totally random and amorphous.
A goal of the work described herein is to describe
the characteristics of milling-induced disorder for two
test compounds.
Crystallite size reduction, altering molecular mo-
bility, and an increase in defect density are all po-
tential changes brought about by milling that can re-
sult in modifying the Gibbs energy of a system.
10,18,19
An investigation of the possible pathways of defect
formation that can occur during milling or process-
ing of crystalline material can be aided by an anal-
ysis of their crystal structures. A previous publi-
cation by Wildfong et al.
10
attempted to develop a
thermodynamic-structure model for predicting com-
plete mechanically induced disorder in organic small
molecule crystals from a knowledge of the crystal
structure and physical properties.
10
Their attempt
adapted a dislocation defect mechanismthat has been
applied in mineralogy to organics.
10,20
Dislocations
are one-dimensional line defects that are assumed to
be induced in a crystal subjected to the shear stresses
involved during milling. Complete mechanical disor-
dering is defined as the point that the free energy
change associated with defect formation (G
d
) equals
the free energy change associated with transitioning
from the solid to the liquid state (G
am
) from Eqs. 1
and 2.
20
G
d
≈ ρ
d
M
v
b
2
µ
s

ln

2
b

ρ
d

(1)
G
am

H
f
T
m

T
m
−T
expt

(2)
From Eq. 1, the material properties affecting G
d
are the shear modulus (µ
s
); the molar volume (M
V
);
the Burgers vector (b); and ρ
d
, which is the disloca-
tion density.
20
The Burgers vector describes the di-
rection and unit displacement of a line defect in a
lattice. The shear modulus expresses a structure’s re-
sistance to strain from a perpendicular stress. In Eq.
2, H
f
is the heat of fusion, T
m
is the melting tempera-
ture, and T
exp
is the experimental temperature so that
the free energy change calculated for a crystalline to
amorphous transition relies on the assumption that
the amorphous solid is structurally and energetically
similar to the liquid phase.
10,20
In the model, the
JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps
TRANSFORMATIONS IN SOLID-STATE PROCESSING 3
entropy effects of line dislocations are assumed to be
negligibly small relative to the change in enthalpy
(H
d
) so G
d
≈ H
d
and the magnitude of H
d
is
a consequence of the magnitude of the strain energy
imposed by the defect.
20
The increase in energy that
arises with the introduction of a defect results from
the strain that is imposed on the crystal structure.
21
However, the strain that is imposed by one defect can-
not overlap with the strain of another defect. As a
consequence, dislocations can actually serve to block
the motion of further dislocation motion at a criti-
cal value, which is approximated as 1/(2b)
2
for small
molecule organics.
10
The question that then arises is
whether, as a consequence of defect pileup, the critical
dislocation density that results in a physical change
in the material that corresponds to G
d
= G
am
is
reached before the prohibitive dislocation density (ρ

).
At that point, the material is expected to produce
an amorphous XRPD pattern and a T
g
. Previously,
Wildfong et al.
10
showed that the model predicted
amorphization of sucrose and (-indomethacin and re-
tention of crystallinity for aspirin, acetaminophen,
and salicylamide. These predictions were born out
successfully with experimental results.
10
Although
amorphization was demonstrated experimentally, the
nature of the transitions in terms of a continuous or
discontinuous pathway was not investigated.
The purpose of the results described in this paper is
aimed toward developing an understanding of the dis-
ordering transitions and structural impact of milling
two test compounds, D-salicin and (-indomethacin.
The analysis is undertaken in the context of both clas-
sical one-state and two-state organic models of dis-
ordering along with models of inorganic amorphous
materials mentioned previously in the Introduction.
EXPERIMENTAL
All sample handling was completed in a nitrogen
glove bag to eliminate water sorption and plasticiza-
tion of the samples. This includes synthesis as well as
preparation of samples for analysis.
Materials and Sample Preparation
D-Salicin and (-indomethacin were obtained from
Sigma–Aldrich Inc. (St. Louis, Missouri) and used as
received. Samples were cryogenically ground at 77K
in a SPEX Sample Prep 6770 Freezer Mill (Metuchen,
New Jersy) following a method adapted from Crow-
ley and Zografi.
6
One gram of sample was placed in
a cylindrical polycarbonate holder with a steel im-
pact rod. The sample chamber was submerged in liq-
uid nitrogen and equilibrated for 2min. Milling was
performed at a frequency of 15impacts per second in
2-min intervals separated by 1-min cool down peri-
ods to allow the chamber to dissipate any heat gener-
ated during milling. Liquid nitrogen melt quenching
of each sample was performed in a nitrogen glove bag.
The samples were heated to melting in a 20-mL scin-
tillation vial and then submerged in liquid nitrogen.
In situ melt quenching of each sample was performed
by differential scanning calorimetry (DSC) by plac-
ing approximately 2mg of material in an aluminum
pinhole hermetic DSC pan with crimped lid. The pan
was heated to the T
m
of each sample and held there
isothermally for 3min. The sample was then cooled at
a rate of 30

C/min to −50

C for D-salicin and −40

C
for (-indomethacin and held there isothermally for
5min before ramping back up at a rate of 10

C/min.
DSC and XRPD measurements were performed im-
mediately after each sample was prepared. Sample
handling was completed in a nitrogen glove bag. If
necessary, the samples were then stored at −20

C
under desiccant before solid-state nuclear magnetic
resonance spectroscopy (SSNMR) was performed.
Methods
X-ray Powder Diffraction
X-ray powder diffraction patterns were recorded at
room temperature with Cu radiation of wavelength
1.540562 Å in reflection mode using a Bruker D8
Discover system equipped with a sealed tube source
and a Hi-Star area detector (Bruker AXS, Madison,
Wisconsin). The instrument was operated at 40kV
with a current of 35mA. The powder sample was
placed in an aluminum holder. Two frames were reg-
istered with an exposure time of 120 s each. The data
are subsequently integrated over the range of 4

–40

2θ with a step size of 0.02

and merged into one con-
tinuous pattern.
Differential Scanning Calorimetry
Differential scanning calorimetry was conducted on
a TA Instruments model Q2000 V24.3 calorimeter
(New Castle, Delaware). Approximately 2mg of solid
sample was placed in an aluminum pinhole hermetic
DSC pan with a crimped lid. The sample cell was
heated under nitrogen purge at a rate of 10

C/min.
Heat capacity (Cp) measurements were performed
on about 10–15 mg of samples hermetically sealed
in Tzero aluminum hermetic pans. Modulated DSC
was employed, with typical modulation parameters of
±1

C every 120 s with an underlying heating rate of
1

C/min. Crystalline and amorphous Cps were mea-
sured by subjecting the same sample to several heat-
ing and cooling cycles, and the data from the heat-
ing cycles were used for further analysis. Starting
materials were heated from subambient to approxi-
mately 20

C–30

C below their T
m
s, cooled at 10

C/
min, allowed to equilibrate, and heated again using
the modulated parameters described above. The crys-
talline samples were subjected to two such heating cy-
cles and heated through their melting temperatures
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES
4 LUISI ET AL.
on the third heating cycle. The molten samples were
then cooled at 10

C/min to subambient to prepare the
amorphous form in situ. The amorphous sample was
then subjected to similar heating and cooling cycles
three times to measure the Cp of the amorphous mate-
rial. Separate measurements on three different sam-
ples were performed to collect replicate data. Con-
figurational Cp (Cp conf), defined as the difference
between the amorphous and crystalline Cp, was de-
termined by subtracting the average Cp values de-
termined from the reversible heat flow signals of the
crystalline form from the amorphous form. Data col-
lected above the T
g
were fit to a linear equation and
extrapolated to the T
m
. Heat of crystallization (H
c
)
as a function of temperature was then determined by
subtracting the integral of the Cp conf from the T
m
to the temperature of interest from the heat of fusion
(H
f
).
Solid-State Nuclear Magnetic Resonance Spectroscopy
Solid-State nuclear magnetic resonance spectroscopy
characterization was conducted on Bruker-Biospin
9.4T (399.2 MHz
1
H Larmor frequency) wide-bore
spectrometer using Bruker-Biospin 4mm HFX probe
(Bruker AXS). Samples were packed into 4mm ZrO
2
rotors and spun under magic angle spinning (MAS)
condition with typical spinning speed of 12.5 kHz.
The sample temperature was regulated using Bruker
BCU-X cooling unit (Bruker AXS) and was set to
275 Kor lower depending on the T
g
of the material an-
alyzed. The proton longitudinal relaxation time (T
1H
)
was measured using saturation recovery relaxation
experiment. On the basis of the observed T
1H
, a proper
recycle delay for the
13
C cross-polarization (CP) MAS
experiment was chosen, typical about 1.3 × T
1H
. The
CPcontact time of carbon CPMAS experiment was set
to 2ms. ACPproton pulse with linear ramp (from50%
to 100%) was employed. The Hartmann–Hahn match
was optimized on external reference sample (glycine).
TPPM15 decoupling sequence was used with the field
strength of approximately 100 kHz.
Computational Determination of Material Properties
Accelrys Materials Studio version 5.5 (San Diego,
California) was used for the computational work and
contains all mentioned modules, forcefields and func-
tionals. The CASTEP module within Materials Stu-
dio was utilized to calculate the shear modulus of
D-salicin and (-indomethacin fromtheir crystal struc-
tures with DFT gradient corrected Perdew-Burke-
Ernzerhof (PBE) functional. (CSD RefCode CIZWOK
for D-salicin and INDMET for (-indomethacin).
22
Be-
cause the protons on the six-member aromatic ring
of the published CIZWOK structure of D-salicin were
not included in the structure, they were assigned geo-
metrically. The molecules in each structure were then
geometrically optimized with constant cell parame-
ters prior to shear modulus determination. The shear
modulus was determined by systematically straining
the crystal structures and calculating the Voigt av-
erages of the resulting stress. The Burgers vector
was determined by calculating the interplanar at-
tachment energies (E
A
) with the COMPASS27 molec-
ular mechanics force field as part of a growth mor-
phology calculation in the morphology module.
23,24
Forcefield assigned parameters were used for each
calculation. The molecules in each structure were ge-
ometrically optimized with constant cell parameters
with COMPASS27 in the Forcite module prior to the
growth morphology run.
24
The slip system (plane and
coplanar vector) of dislocation movement is assumed
to occur in those hkl planes with the lowest cohesive
E
A
and hence the lowest barrier to slip and flow, and
with a vector in a nonsterically hindered crystallo-
graphic direction. The Burgers vector was then calcu-
lated from the distance between molecular centers of
mass in the slip system.
RESULTS AND DISCUSSION
Defects that are described by dislocation motion oc-
cur within a specified slip system, which comprises
the plane and a coplanar slip vector, and is assumed to
occur in those hkl planes with the lowest cohesive E
A
s
and hence the lowest barrier to slip and flow. E
A
calcu-
lations on D-salicin identified the (020) as having the
lowest interplanar E
A
of −192.34 kJ/mol. Inspection
of the crystal structure shows that steric hindrance
within the (020) limits the coplanar slip vector to the
[001]. The Burgers vector was then calculated from
the distance between the molecular centers of mass
within the (020)[001] slip system as 0.765 nm. Fig-
ure 2 shows an image of the packing of D-salicin at
the (020) looking down the [001]. D-Salicin actually
contains several sterically unhindered slip systems
listed in Table S1 in the Supplementary Information,
the next lowest energy of which is the (110)[001] with
an E
A
of −351.52 kJ/mol. A plot of G
d
in kJ/mol as a
function of dislocation density for D-salicin is shown
in Figure 3. The individual model parameters used to
generate the plot are listed in the Supplementary In-
formation in Table S2. The calculated defect density
at which G
d
equals G
am
at 77 K(cryomilling condi-
tions) is 3.78 × 10
17
m
−2
compared with a prohibitive
dislocation density of 4.07 × 10
17
m
−2
. As a conse-
quence, the plot of free energy change with increasing
dislocation concentration intersects with G
am
before
ρ

and so D-salicin is predicted to become amorphous
when subjected to cryomilling for a sufficient time.
Provided that the energy imparted from cryomilling
is sufficient to overcome the energetic barrier to de-
fect formation in more than just the lowest E
A
slip
system, then multiple structurally distinct defected
states will arise.
JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps
TRANSFORMATIONS IN SOLID-STATE PROCESSING 5
Figure 2. View of the packing of D-salicin at the (020).
The coplanar [001] is tunneling into the page.
For comparison, the Burgers vector and shear mod-
ulus model parameters for (-indomethacin were also
calculated and compared with the published litera-
ture values of b = 1.097 nm in (10-1)[010] and µ
s
=
4.3GPa.
10
In this work, the lowest E
A
was calculated
at the (011) with E
A
= −133.99 kJ/mol, whereas the
published (10-1) was calculated to have the second
lowest E
A
of −157.80 kJ/mol. This yields a newly de-
termined slip system of (011)[100] with b = 0.919 nm.
The E
A
s calculated previously by using the general
Dreiding 2.21 forcefield with the charge equilibration
method used to assign charges and an older version
of Accelrys software Cerius
2
v.4.6.
10
E
A
s in this work
were calculated with the more recent Accelrys Ma-
terials Studio version 5.5 software and Compass27
forcefield with forcefield assigned charges resulting in
Figure 3. Plot of the change in free energy with increas-
ing dislocation density (G
d
) for D-Salicin. The plot crosses
G
amorphous
calculated at 77K (at critical ρ
crit
of 4.02× 10
17
m
−2
) before reaching the prohibitive dislocation density


= 4.31× 10
17
).
different results obtained from the calculations. Re-
gardless, both Burgers vector values yield the same
prediction of cryomilling, leading to the formation of
amorphous material. Indeed, (-indomethacin and D-
salicin both contain several sterically unhindered slip
systems listed in Table S1 in the Supplementary In-
formation. The shear modulus was calculated to be
4.7GPa, which is in good agreement with the pub-
lished value of 4.3GPa that was determined by qua-
sistatic indentation into powder compacts.
10
Samples of D-salicin and (-indomethacin were
analyzed by XRPD, DSC, and SSNMR following
cryomilling. Supplementary Figure S1 shows a com-
parison of the calculated XRPD patterns from the
single-crystal structures of each compound obtained
fromCSDentries CIZWOXfor D-salicin and INDMET
for (-indomethacin with the XRPD patterns of the
starting material. The overlays show that the two
compounds were received as single-phase systems,
with structures corresponding to the known poly-
morphs. Cryogrinding of each sample was performed
in order to decouple the effects of mechanical impact
without interference from thermal events. It is as-
sumed that at liquid nitrogen temperatures, the ther-
mally induced molecular rearrangements that could
lead to crystallization are absent. Additionally, any
process-induced disorder is assumed to be due to me-
chanical impact alone with negligible thermal im-
pact. Melt-quenched samples were also included in
the experiments to generate an amorphous reference
and to study the effects that the route taken to the
generation of amorphous material has on physical
properties.
Several polymorphs of (-indomethacin are known
to exist and have been crystallized in the solid state
from amorphous material.
25
Throughout the exper-
iments, XRPD and SSNMR data (discussed below)
on samples of (-indomethacin are consistent with
the presence of (-indomethacin as the only observed
crystalline form that is present following milling. Al-
though a variety of polymorphs have previously been
obtained from thermal treatment of amorphous (-
indomethacin, crystallization of milled material in
DSC scans in this work resulted in the formation of
crystalline material with a T
m
consistent with that of
(-indomethacin.
Figures 4 and 5 show patterns obtained at selected
time intervals following cryomilling as compared with
the starting materials. Following 1min of cryomilling
D-salicin, a reduction in crystallinity was observed
by XRPD through broadening of the diffracted peaks.
Broadening of diffraction lines relative to a reference
material (in this case, the as-received unprocessed
material) is indicative of the introduction of crys-
tal defects.
26
Nanocrystalline powders or nanosized
crystalline domains may also result in broadening
of Bragg peaks. However, SSNMR (discussed below),
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES
6 LUISI ET AL.
Figure 4. XRPD pattern comparison of starting mate-
rial with cryomilled D-salicin material at selected time
intervals.
which is sensitive to local order, does not show the
crystalline shifts that would be consistent with the
presence of nanoscale crystallinity and broad diffrac-
tion lines. As a consequence, we can conclude that
nanoscale crystallinity is not responsible for peak
broadening. After 60min of milling, evidence of crys-
talline diffraction is absent from the pattern, indicat-
ing destruction of long-range periodicity. The XRPD
patterns remain characteristic of material lacking
long-range order upon further milling of the samples
up to 180 min. All diffracted peaks in the XRPD pat-
terns can be analyzed in terms of the starting poly-
morph of D-salicin with no evidence of a polymorphic
change occurring with milling.
Milling of (-indomethacin results in the broadening
of some peaks in the diffraction pattern after 1min
of cryomilling, indicating a loss of crystallinity. The
majority of the reflections corresponding to the start-
ing polymorph are still present after 1min. However,
following 10 min of milling, more significant peak
broadening is observed along with the loss of several
diffraction lines. After 40min of milling, evidence of
crystalline diffraction is absent, indicating the loss of
Figure 5. XRPD pattern comparison of starting material
with cryomilled (-indomethacin material at selected time
intervals.
Figure 6. DSC scan comparisons of D-salicin starting ma-
terial with melt-quenched and cryomilled samples at se-
lected time intervals.
long-range order in the material. Further milling to
180min produces no discernable change in the XRPD
pattern. Both D-salicin and (-indomethacin showed
a reduction in crystallinity by XRPD with increasing
milling time until complete loss of crystallinity was
observed in both cases by XRPD.
Comparisons between DSC scans of cryomilled and
melt-quenched samples of D-salicin with the respec-
tive starting materials are shown in Figure 6. Un-
milled D-salicin and sample milled for 1min show a
single thermal feature in the DSC scan correspond-
ing to an endothermic melt at approximately 200

C.
Milling samples at intervals between 10 and 180 min
result samples that show an exothermic event in
the DSC, which occurs between 76.9

C and 92.4

C,
noted as the peak maxima in Figure 6, depending on
milling time. The exothermic event is attributed to
crystallization (annealing) of disorder, either through
amorphous content or defects that were introduced
into the sample by milling. The exothermic event
occurs at a temperature above the T
g
(∼59

C) of
amorphous material prepared by melt quenching in
liquid nitrogen for all samples. All milled samples
show the characteristic melting endotherm associ-
ated with the starting polymorph, indicating that the
disorder crystallized back to the original polymorph.
The temperature at which crystallization occurs (T
c
)
and the magnitude of the enthalpy of crystalliza-
tion (H
c
) increase with milling time, the trends of
which are shown in Figure 7. The enthalpy values
were corrected for the shift in T
c
as described in the
Experimental section. Glass transition events are ap-
parent at ∼59

C in both melt-quenched samples. The
liquid nitrogen melt-quenched sample showed the
highest T
c
and largest H
c
values of all the samples,
whereas the in situ melt-quenched sample showed
only a T
g
and did not undergo crystallization through
the final heating cycle. A glass transition event was
not observed for the milled samples, which can occur if
JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps
TRANSFORMATIONS IN SOLID-STATE PROCESSING 7
Figure 7. (a) Change in temperature of crystallization
(T
c
) and heat of crystallization (H
c
) with milling time
for D-salicin as compared with liquid nitrogen (LN2) melt-
quenched D-Salicin. (b) H
c
as a function of crystallization
temperature with the line representing the calculated max-
imum H
c
from the configurational heat capacity.
the change in Cp associated with the glass transition
event of the amorphous component is small and below
the detection limit. As a consequence, crystallization
and melting were the only detected events. Although
milling for 1min did produce an observable loss of
crystallinity in the XRPD pattern, crystallization was
not observed in the DSC, which can be attributed
to an exothermic magnitude below the limit of
detection.
Differential scanning calorimetry scans of start-
ing material, cryo-milled, and melt-quenched (-
indomethacin samples are shown in Figures 8 and
9. Starting (-indomethacin shows a single thermal
feature in the DSC corresponding to an endother-
mic melting event at approximately 162

C. Com-
parable to D-salicin, milling results in exothermic
events attributed to crystallization of milling-induced
disorder. The exothermic events occur at tempera-
tures above the T
g
(∼47

C) of amorphous material
prepared by melt quenching in liquid nitrogen. At
milling times between 60 and 180 min, however, a
second exothermic event is observed, indicating that
crystallization of disorder is occurring over two sep-
arate thermal events in those samples. All of the
milled samples show the characteristic melting en-
Figure 8. DSCscan comparisons of (-indomethacin start-
ing material with cryomilled samples at selected time
intervals.
dotherm associated with the starting material, indi-
cating that the induced disorder crystallized back to
the (-polymorph.
6
The T
c
and the magnitude of the
H
c
for (-indomethacin were also observed to be de-
pendent upon milling time, the trends of which are
shown in Figure 10. The H
c
is reported as the sum
of the two exothermic events for samples milled be-
tween 60 and 120 min. Glass transition events were
also observed at ∼47

C for melt-quenched samples.
The liquid nitrogen melt-quenched sample shows only
a single exothermic event with a peak at 134

C, which
closely corresponds to the second event observed in
samples with milling times between 120 and 180 min.
In contrast, the in situ melt-quenched sample shows
only a T
g
withno evidence of crystallizationor melting
as was the case with in situ melt-quenched D-salicin.
Although both samples were prepared from the melt,
the rates of cooling and thermal history are differ-
ent for each technique, which can be assigned to the
observed differences during DSC analysis. With the
Figure 9. DSC scan comparisons of (-indomethacin cry-
omilled samples at 60, 120, 150, and 180min compared
with liquid nitrogen (LN2) melt-quenched and in situ melt-
quenched material.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES
8 LUISI ET AL.
Figure 10. (a) Change in temperature of crystallization
(T
c
) and heat of crystallization (H
c
) with milling time for (-
indomethacin as compared with liquid nitrogen (LN2) melt-
quenched (-indomethacin. (b) H
c
as a function of T
c
with
the line representing the calculated maximum H
c
from
the configurational heat capacity.
exception of the sample milled for 1min, all milled
and melt-quenched (-indomethacin samples also
show a T
g
(magnification of T
g
region in Supple-
mentary Figure S2). (-Indomethacin milled for 1min
shows a crystallization event at 63

C and no T
g
. It
is possible that the Cp difference associated with the
T
g
in this case is too small to be observed or, alterna-
tively, that the exothermic event is the result of the
crystallization of defects that are unassociated with
glass transition.
In both test cases, there is a trend of increasing T
c
accompanied by an increase in the H
c
with milling
time. A plot comparing the measured and calculated
maximum H
c
values at each temperature is shown
in Figure 8 for (-indomethacin and Figure 10 for
D-salicin, with the line representing the calculated
values using DSC in situ prepared amorphous form
as the reference material. The maximumH
c
was de-
termined from the integration of the configurational
Cp as described in the Experimental section (Supple-
mentary Figure S3). The difference between the mea-
sured and calculated values represents the amount
of disorder that crystallized. The magnitude of the
H
c
is indicative of the amount of disordered mate-
rial in J/g that is undergoing crystallization. The T
c
for the same DSC scan rate, is an indication of the
resistance, or kinetic stability to crystallization with
temperature. As a consequence, because the extent of
disorder in each material is increasing (determined
by the magnitude of H
c
) concurrently with the T
c
, it
can be concluded that the stability to crystallization
is increasing as the extent of crystalline order in each
material is decreasing. Additionally, the liquid nitro-
gen melt-quenched samples show the largest values
for both the T
c
and the H
c
. This indicates that melt-
quenching produces both a greater extent of disorder
along with an increase in the stability of that disorder
as compared with milling.
A two-phase model analysis of our data states that
a decreasing percentage of crystallinity with milling
is the consequence of progressively smaller crystalline
domain sizes. A one-phase model analysis of our data
asserts that increasing disorder with milling is a con-
sequence of progressively introducing defects into the
crystal lattice, which results in a loss of the local
structure that is represented in the crystal. Topo-
logically disordered materials require bond breaking
and formation processes (in our case, hydrogen and
noncovalent bonds) in order to facilitate molecular
rearrangements that generate the long-range order
found in crystalline material.
27
Topological disorder-
ing in amorphous networks has been reported to re-
sult in an increased resistance to crystallization in
coamorphous naproxen–(-indomethacin materials as
well as inorganic glasses.
27,28
The number and type
of local rearrangements necessary for reproduction of
the crystal form are the result of the extent and type
of the defects that are introduced. Increasing topolog-
ical disorder will necessitate a larger total number
of molecular rearrangements, which potentially re-
sults in improved stability, relative to material with
local order resembling that of the crystalline phase, as
the extent of topological ordering that needs to occur
increases.
In order to further probe the state of the disorder
in milled samples, SSNMR analysis was conducted.
Figures 11 and 12 show the
13
C SSNMR spectra and
corresponding longitudinal proton T
1
relaxation data
for milled D-salicin and (-indomethacin. A spectral
line shape that can be well reproduced by linear com-
bination of the initial, fully crystalline, and the final,
fully amorphous spectra was observed with milling
for both samples. The intensities of crystalline and
amorphous peaks are decreasing and increasing, re-
spectively, with milling time. This is indicative of
a phase transition from a crystalline to an amor-
phous state.
29,30
After 60min, of milling, evidence of
crystallinity is absent from the
13
C SSNMR spectra
with no observable change upon further milling. How-
ever, the proton relaxation data for both compounds
shows, at short milling times, intermediary relaxation
rates, which is demonstrative of material structurally
JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps
TRANSFORMATIONS IN SOLID-STATE PROCESSING 9
Figure 11.
13
C SSNMR spectra and proton T
1
relaxation
data for cryomilled and starting D-salicin material.
distinct from, and in between, crystalline starting ma-
terial and the ones milled for longer time periods.
29
In terms of the one-phase or two-phase model, con-
tinuously changing relaxation time is clear evidence
of distinct intermediary disordered states that cannot
be described by a simple two-phase system. However,
the discrete phase transformation observed by chemi-
cal shift changes in the
13
C spectra contradicts a com-
plete one-phase model that links crystalline through
to amorphous material by a continuous set of disor-
dered and defected states.
If crystals of progressively smaller domain sizes,
eventually approaching the molecular level, were
formed by milling, we would expect the carbon peaks
to show progressive line broadening instead of the
observed physical mixture-like linear combination of
the crystalline and amorphous peaks. The progressive
line broadening would be caused by increasing frac-
tion of molecules located at the boundaries of the crys-
talline domains, which would be experiencing chem-
ical shift perturbations arising from intramolecular
conformations as well as intermolecular packing that
would be different from one boundary molecule to the
other as well as from the molecules inside the crys-
talline domains. Note that the strictly local effect of
Figure 12.
13
C SSNMR spectra and proton T
1
relaxation
data for cryomilled and starting (-indomethacin material.
chemical shifts (on the order of approximately 0.5nm)
also means that the SSNMR length scale would be
smaller than that of XRPDcrystalline size-dependent
line broadening.
26
As a consequence, the data support
the observation that structurally distinct amorphous
states are resulting in the shift to higher temper-
atures of crystallization, in DSC experiments, with
milling, even beyond the point of both D-salicin and (-
indomethacin becoming completely amorphous as ev-
idenced by both XRPD and SSNMR, rather than a ki-
netically diminished seeding effect as a consequence
of reducing crystalline domain size.
Hypotheses such as the presence of small seed crys-
tals in amorphous material that are present below
the detection limit of solid-state methods as well as
differences in specific surface area have been pro-
posed in the literature that are alternative to the
one put forward, here regarding different amorphous
structures.
6
In such a hypothesis, the small seed crys-
tals are intimately mixed with the amorphous frac-
tion and are responsible for seeding conversion to the
crystalline polymorph upon thermal treatment. An
attempt at modeling this effect for griseofulvin was
published by Feng et al.
7
by mixing crystalline with
amorphous material and performing a DSC analysis.
The data showed thermal properties consistent with
pure crystalline and amorphous fractions of grise-
ofulvin without evidence of a seeding effect. This
method was used to test a D-salicin sample that
had been milled for 60 min and was amorphous by
XRPD and SSNMR. Indeed, a 50:50 physical mixture
of melt-quenched D-salicin with material cryomilled
for 60min shows DSC thermal signatures consistent
with those of the pure components, as shown in Figure
13. However, it can be argued that a bulk mixture of
two phases as is the case with these physical mixtures
does not adequately represent the “small seeding”
phenomenon that may be present in the alternative
hypothesis, wherein crystalline domains may be dis-
persed along a smaller length scale, and so the reader
is referred to the references above. We believe that
differences in amorphous structure, as seen across
other disciplines, require consideration when analyz-
ing data from amorphous organic materials and are
responsible for the properties presented in this paper.
CONCLUSIONS
X-ray powder diffraction, DSC, and SSNMR charac-
terization of cryomilled D-salicin and (-indomethacin
indicate the existence of defected crystalline and
amorphous regions separated by a phase transi-
tion. The presence of a phase transition that oc-
curs upon converting crystalline material to amor-
phous was demonstrated by SSNMR. A complete
one-phase model of disorder that maintains that
a continuous set of states exists between perfectly
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES
10 LUISI ET AL.
Figure 13. DSC scan of a 50:50 binary mixture of crystalline D-salicin, 60min cryomilled, and
liquid nitrogen melt-quenched D-salicin. The mixture can be described from a combination of
the melt-quenched and cryomilled material.
crystalline and totally amorphous material does not
account for such a phase transition. However, a two-
phase model wherein the material is considered as a
binary mixture of two states comprising amorphous
and crystalline fractions is also incomplete. XRPD
shows peak broadening and SSNMR shows interme-
diate proton relaxation times, which are indicative
of crystal defects in both milled materials. DSC data
showing changes in T
c
for both melt-quenched and
amorphous cryomilled samples taken in conjuction
with the SSNMR data indicate differences in amor-
phous structures. Although a direct measurement of
the enthalpy or volume/density of the processed solid
amorphous material has not been made (Fig. 1), we
can infer the existence of different amorphous struc-
tures from the observed changes in the thermal be-
havior in DSCmeasurements coupled with XRPDand
SSNMR data. This observation parallels reported dif-
ferences in amorphous structures in other fields rang-
ing from silica glass and semiconductor materials to
polyamorphism in ice.
14
A model that depicts a phase
transition that separates two continuous one-phase
systems, defected crystalline from amorphous, is thus
proposed. Additionally, DSC experiments show that
as the extent of disorder is increasing the kinetic sta-
bility to crystallization is also increasing. The asser-
tion that distinct amorphous states exist for these ma-
terials means that there are different degrees of amor-
phous structural stability with respect to crystalliza-
tion. The work has important implications for analyti-
cal method development, understanding performance
through structure–property relationships and assess-
ing the potential impact that routes of processing can
have on stability. Further work in assessing the role
that kinetics plays in the rate of disordering versus
crystallization at a given temperature needs to be per-
formed in order to develop a comprehensive model of
the effects of processing a given polymorph.
ACKNOWLEDGMENTS
The authors thank Patrick Connelly for comments
and critical reading of the manuscript and Timothy
Barder for rendering equations and the defect density
graph for publication.
REFERENCES
1. Cui Y. 2007. A material science perspective of pharmaceutical
solids. Int J Pharm 339:3–18.
2. Hancock BC, Zografi G. 1997. Characteristics and significance
of the amorphous state in pharmaceutical systems. J Pharm
Sci 86:1–12.
3. Huttenrauch R, Fricke S, Zielke P. 1985. Mechanical activa-
tion of pharmaceutical systems. Pharm Res. 302–306.
4. Bernstein J. 2002. Polymorphism in molecular crystals.
Oxford: Clarendon.
5. Herbstein FH. 2004. Diversity amidst similarity: A multidis-
ciplinary approach to phase relationships, solvates and poly-
morphs. Cryst Grow Des 4:1419–1429.
6. Crowley KJ, Zografi G. 2002. Cryogenic grinding of in-
domethacin polymorphs and solvates: Assessment of amor-
phous phase formation and amorphous phase physical sta-
bility. J Pharm Sci 91:492–507.
7. Feng T, Pinal R, Carvajal MT. 2008. Process induced disor-
der in crystalline materials: Differentiating defective crys-
tals from the amorphous form of griseofulvin. J Pharm Sci
97:3207–3221.
8. Otte A, Carvajal MT. 2010. Assessment of milling-induced
disorder of two pharmaceutical compounds. J Pharm Sci
100:1793–1804.
9. Charthy SP, Pinal R. 2008. The nature of crystal disorder in
milled pharmaceutical materials. Colloids Surf A 331:68–75.
10. Wildfong PLD, Hancock BC, Moore MD, Morris KR. 2006. To-
wards an understanding of the structurally based potential for
mechanically activated disordering of small molecule organic
crystals. J Pharm Sci 95:2645–2656.
11. Shalaev E, Zografi G. 2002. The concept of “structure” in amor-
phous solids from the perspective of the pharmaceutical sci-
ences. In Amorphous food and pharmaceutical systems; Levine
H, Ed. Cambridge: RSC.
12. Hancock BC, Shalaev EY, ShamblinSL. 2002. Polyamorphism:
A pharmaceutical science perspective. J Pharm Pharmacol
54:1151–1152.
13. Winkel K, Hage W, Loerting T, Price SL, Mayer E. 2007. Car-
bonic acid: Frompolyamorphismto polymorphism. J AmChem
Soc 129:13863–13871.
14. McMillan PF. 2004. Polyamorphic transformations in liquids
and glasses. J Mater Chem 14:1506–1512.
15. Deb SK, Wilding MC, Somayazulu M, McMillan PF.
2001. Pressure-induced amorphisation and an amorphous–
amorphous transition in densified porous silicon. Nature
414:528–530.
JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps
TRANSFORMATIONS IN SOLID-STATE PROCESSING 11
16. Mishima O, Calver LD, Whalley E. 1985. An apparently first-
order transition between two amorphous phases of ice induced
by pressure. Nature 314:76–78.
17. Smith KH, Shero E, Chizmeshya A, Wolf GH. 1995. The
equation of state of polyamorphic germania glass: Two-
domain description of the viscoeleastic response. J Chem Phys
102:6851–6857.
18. Morup S, Jiang JZ, Bodker F, Horsewell A. 2001. Crystal
growth and the steady-state grain size during high-energy ball
milling. Europhys Lett 56:441–446.
19. Descamps M, Willart JF, Dudognon E, Caron V. 2007.
Transformation of pharmaceutical compounds upon milling
and comilling: The role of T
g
. J Pharm Sci 96:1398–
1407.
20. Tromans D, Meech JA. 2001. Enhanced dissolution of min-
erals: Stored energy, amorphism and mechanical activation.
Miner Eng 14:1359–1377.
21. Callister Jr. WD. Materials science and engineering. 3rd ed.
New York: John Wiley and Sons Inc.
22. Clark SJ, Segall MD, Pickard CJ, Hasnip PJ, Probert MJ,
Refson K, Payne MC. 2005. First principles methods using
CASTEP. Z Kristallogr 220:567–570.
23. Docherty R, Clydesdale G, Roberts KJ, Bennema P. 1991.
Application of Bravais–Friedel–Donnay–Harker attachment
energy and ising models to predicting and understanding
the morphology of molecular crystals. J Phys D: Appl Phys
24:89–99.
24. Sun H. 1998. COMPASS: An ab initio forcefield opti-
mized for condensed-phase applications—Overview with de-
tails on alkane and benzene compounds. J Phys Chem B
102:7338–7364.
25. Yoshioka M, Hancock BC, Zografi G. 1994. Crystallization of
indomethacin from the amorphous state below and above its
glass transition temperature. J Pharm Sci 83:1700–1705.
26. Jenkins R, Snyder RL. 1996. Introduction to X-ray powder
diffractometry. New York: John Wiley and Sons Inc.
27. Gupta PK. 1996. Non-crystalline solids: Glasses and amor-
phous solids. J Non-Crystalline Solids 195:158–164.
28. Lobmann K, Laitinen R, Grohganz H, Gordon KC, Strachan
C, Rades T. 2011. Co-amorphous drug systems: Enhanced
physical stability and dissolution rate of indomethacin and
naproxen. Mol Pharm 8:1919–1928.
29. Cadars S, Lesage A, Pickard CJ, Sautet P, Emsley L. 2009.
Characterizing slight structural disorder in solids by combined
solid-state NMRand first principles calculations. J Phys Chem
A 113:902–911.
30. Lubach JW, Xu D, Segmuller BE, Munson EJ. 2007. Investi-
gation of the effects of pharmaceutical processing upon solid-
state NMR relaxation times and implications to solid-state
formulation stability. J Pharm Sci 96:777–787.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES