CRYSTAL

GROWTH
Phase Solubility Diagrams of Cocrystals Are Explained by Solubility & DESIGN
Product and Solution Complexation
2006
Sarah J. Nehm, Barbara Rodrı́guez-Spong, and Naı́r Rodrı́guez-Hornedo* VOL. 6, NO. 2
Department of Pharmaceutical Sciences, UniVersity of Michigan, Ann Arbor, Michigan 48109-1065 592-600

ReceiVed July 13, 2005; ReVised Manuscript ReceiVed October 24, 2005

ABSTRACT: The purpose of this work was to develop a mathematical model that describes the solubility of cocrystals by taking
into consideration the equilibria between cocrystal, cocrystal components, and solution complexes. These models are applied to the
phase diagrams of carbamazepine/nicotinamide (CBZ/NCT) cocrystal in organic solvents. The CBZ/NCT (1:1) cocrystal solubility
was measured by suspending cocrystal in solutions of varying nicotinamide concentrations in ethanol, 2-propanol, or ethyl acetate.
Results show that the solubility of the cocrystal decreases with increasing nicotinamide concentration. Mathematical models demonstrate
that (1) the solubility of a cocrystal AB is described by the solubility product of cocrystal components and by solution complexation
constants, (2) these equilibrium constants can be determined from solubility methods, and (3) graphical representation of the cocrystal
solubility dependence on ligand concentration will serve as a diagnostic tool for the stoichiometry of solution complexes. CBZ/NCT
cocrystal Ksp values increase as the cocrystal solubility increases, while K11 values decrease. The dependence of cocrystal solubility
on solubility product and complexation constants provides a powerful approach to design cocrystal screening methods and to formulate
solutions with cocrystal components where crystallization does not occur.

Introduction In this report, we develop models that describe the solubility

of binary cocrystals with nonionizable components, by taking
Molecular interactions responsible for the formation of
into consideration the equilibria between cocrystal and cocrystal
molecular complexes of active pharmaceutical ingredients
components in solution. We demonstrate that (1) the solubility
(APIs) with other compounds are important not only because
of a 1:1 cocrystal AB is described by a solubility product, (2)
of the ability to control pharmaceutical properties without
the cocrystal solubility is increased by a constant value when
changing covalent bonds,1-4 but also because they can be used
there is 1:1 solution complexation, and (3) the cocrystal
in the design of new materials.5-8 In recent years, advances in
solubility goes through a minimum value when there are 1:1
crystal engineering and supramolecular chemisty have motivated
and 1:2 solution complexes. Thus, a graphical representation
research on the design of pharmaceutical materials by directing
of the cocrystal solubility dependence on ligand concentration
molecular assembly of different components in the crystalline
will serve as a diagnostic tool for the stoichiometry of solution
state to form cocrystals.9-16 Pharmaceutical properties of some
complexes. Experimental results for the solubility of carbam-
cocrystals have also been reported such as solubility, dissolution
azepine/nicotinamide 1:1 cocrystal in organic solvents are
rate, hygroscopicity, and chemical stability.9,17-19 The purpose
analyzed in terms of the theoretical arguments developed.
of the studies presented in this manuscript is to explain the
solubility behavior of cocrystals in solutions of cocrystal Solubility products and complexation constants are also deter-
components. mined.
Molecular recognition phenomena that give rise to complex- These findings have practical applications in the development
ation in solution was the focus of research by Higuchi and co- of crystallization methods for cocrystal screening and in the
workers in the 1950s2,3,20,21 and is the subject of various formulation of solutions with ingredients that can form co-
reviews.1,22 Much of this fundamental work was geared toward crystals.
understanding the mechanisms of complex formation and Fundamentals. Solubility Product of Cocrystals. The
determining the relative influence of intermolecular interactions solubility of a binary cocrystal of API (A) and ligand or cocrystal
in solutions (hydrogen bonds, charge-transfer forces, and other component (B), of composition AaBb where the cocrystal
electrostatic and induction forces) that favored complex forma- components do not ionize or form complexes in solution, is
tion. This early work was concerned with improving the aqueous given by the equilibrium reaction
solubility of poorly water soluble compounds via solution
complexation,2,3 and mathematical models were developed to AaBb solid h aAsolution + bBsolution (1)
evaluate complexation behavior from solubility studies of API
in solutions of ligand. As a result, fewer publications address
Subscripts refer to the stoichiometric number of molecules
insoluble molecular complexes that formed during phase
of A or B in the complex. The equilibrium constant for this
solubility studies of API crystals.3,4,21 It is noted that the terms
reaction is given by
molecular complex, solid-state complex, and molecular com-
pound have been used in the pharmaceutical literature since the
1950s, while the term cocrystal is found in the more recent aAaaBb
literature. Keq ) (2)
aAB
* Corresponding author. College of Pharmacy, 428 Church Street, Ann
Arbor, MI 48109-1065. Phone: 734-763-0101. Fax: 734-615-6162. and is proportional to the thermodynamic activity product of
E-mail: nrh@umich.edu. the cocrystal components. If the activity of the solid is equal to
10.1021/cg0503346 CCC: $33.50 © 2006 American Chemical Society
Published on Web 12/06/2005
Phase Solubility Diagrams of Cocrystals Crystal Growth & Design, Vol. 6, No. 2, 2006 593

1 or is constant, the cocrystal solubility can be described by a

solubility product

Ksp ) aAaaBb ≈ [A]a[B]b (3)

where [A] and [B] are the molar concentrations of each cocrystal
component at equilibrium, as long as the activity coefficients
are unity. This approximation applies to dilute solutions and
for practical purposes will be used in this paper to calculate
material balances and solution compositions.
If a binary cocrystal of 1:1 stoichiometry dissolves in pure
solvent into its individual components without further com-
plexation or ionization to form a saturated solution, the mass
balance for each component in solution can be expressed in
terms of the molar solubility of the cocrystal, S

[A] ) S and [B] ) S (4)

and substituting these in the solubility product eq 3 gives Figure 1. Effect of ligand concentration on the solubility of cocrystal
AB (solid line) and the solubility of single component crystal A (dashed
line) calculated from eq 9 with Ksp ) 0.0129 M2, and SA ) 0.09 M.
Ksp ) S2 and S ) (Ksp)1/2 (5) The transition ligand concentration [B]tr, occurs when the solubility of
A equals the solubility of AB.
Equations 4 and 5 apply only to solutions of stoichiometric
composition when the solution molar ratio is the same as that considerations, the solubility curves of cocrystal and single
of the cocrystal. component crystal intersect. Therefore, there is a cocrystal
For nonstoichiometric solution compositions, let C be the component concentration in solution, [B]tr, at which the solubil-
excess concentration of ligand so the mass balances when excess ity of cocrystal AB is equal to the solubility of crystal A and
B is added become above which the solubility of cocrystal AB is less than that of
crystal A.
[A] ) S and [B] ) S + C (6) The transition concentration of cocrystal component can be
and therefore predicted by substituting the single component crystal solubility,
SA, for the cocrystal solubility, S, in eq 7 and rearranging to
Ksp ) S(S + C) (7) give

In the case in which a large excess of ligand is present, such Ksp - SA2
that C . S, then Ctr ) (10)
SA
S ≈ Ksp/C (8)
where Ctr is the excess concentration of cocrystal component
A quadratic equation must otherwise be solved and gives at the transition concentration, and the total concentration of
cocrystal component at the transition, [B]tr, is given by
-C + xC2 + 4Ksp
S) (9) Ksp - SA2
2 [B]tr ) S + Ctr ) S + (11)
SA
This equation predicts that addition of either cocrystal compo-
nent to a solution in excess of S decreases the cocrystal solubility where S is the solubility of the cocrystal under stoichiometric
when the preceding conditions apply. conditions. This equation predicts that [B]tr increases as the
A plot of the solubility of cocrystal AB as a function of total solubility of A decreases or as Ksp increases, as shown in Figure
ligand in solution according to eq 9 is shown in Figure 1. 2.
Here, [A]T ) S and [B]T ) S + C as shown in eq 6, and the The phase diagram which includes the solubilities of the
subscript “T” stands for total concentration. This solubility cocrystal and single component crystal, Figure 1, also defines
behavior resembles that of the common ion effect in the case four domains representing regions of kinetic and thermodynamic
of sparingly soluble salts. But in contrast with the case of salts control for the dissolution or crystallization of the single and
and that of solvates where analogous equilibria have been multicomponent phases. Domain I is supersaturated with respect
considered,23-27 cocrystals dissociate into primary components to A but undersaturated with respect to cocrystal AB. Both A
(at least two different molecules) that can crystallize as single and AB are supersaturated in domain II, but undersaturated in
component phases. domain III. Domain IV is supersaturated with respect to AB
Also shown in Figure 1 is the solubility of single component but is undersaturated with respect to A. A theoretical plot such
crystal of A, as a function of the cocrystal component or ligand as this indicates regions of thermodynamic stability and which
(B) concentration in solution. This phase diagram is based on form(s) will dissolve or have the potential to crystallize. This
the following assumptions: (1) A is less soluble than B, (2) A information is important for the development of screening
is less soluble than AB in stoichiometric solutions (with respect methods for the crystallization of cocrystals, identifying condi-
to AB), (3) there is no complexation or ionization of cocrystal tions where phase transformations between crystal and cocrystal
components in solution, and (4) the solubility of A is indepen- occur, and controlling or preventing the crystallization of the
dent of the concentration of B in solution. Under these cocrystal in solutions of cocrystal components.
594 Crystal Growth & Design, Vol. 6, No. 2, 2006 Nehm et al.

[AB] is the solution concentration of complex and according

to eq 15 is

[AB] ) K11Ksp (18)

Thus, the solution concentration of complex is fixed by the

coupled equilibria. This presents an unusual and interesting
condition.
Substituting eqs 14 and 18 into eqs 16 and 17 one obtains

Ksp
[A]T ) + K11Ksp (19)
[B]
[B]T ) [B] + K11Ksp (20)

[A]T is the solubility of cocrystal AB, when measuring total A

in solutions under the equilibrium conditions described in eq
Figure 2. The transition ligand concentration as a function of the 1. By combining the above equations, the cocrystal solubility
solubility of the single component crystal, SA, calculated from eq 11 can be expressed in terms of the total ligand concentration, [B]T
with values of Ksp ) 0.0129 M2 (solid line) and Ksp ) 0.0032 M2
(dashed line).
according to

Figure 1 also illustrates the path of a solution that is initially Ksp

[A]T ) + K11Ksp (21)
undersaturated with respect to A and AB at point x and goes [B]T - K11Ksp
through saturated ([A]T ) SAB) and supersaturated ([A]T > SAB)
states with respect to AB as the ligand concentration in solution If K11Ksp , [B]T, then
increases to point y. The driving force for crystallization is the
supersaturation, or difference in chemical potential between y Ksp
and z. Since crystallization of AB will reduce [B], the reaction [A]T ) + K11Ksp (22)
will proceed until a saturated state at z′ is reached. [B]T
While the decrease in cocrystal solubility with increasing
ligand concentration provides a means for identifying conditions Therefore, both Ksp and K11 can be evaluated from a plot of
for preparing cocrystals, the path shown in Figure 1, x to y to [A]T versus 1/[B]T. If there are no higher order complexes in
z′, predicts the unexpected crystallization of AB from a solution, this plot is linear with slope ) Ksp and intercept )
formulation of compound A in solutions of B if only the K11Ksp under the conditions K11Ksp , [B]T.
solubility of A is considered during development. For instance, Equation 22 predicts that the cocrystal solubility decreases
in the case in which an API (A) is formulated in an undersatu- with increasing ligand concentrations and that the cocrystal
rated solution of composition x, addition of excipient B to the solubility is higher by a constant value, the product of K11 and
solution at a composition y may result in crystallization of Ksp, compared to the case in which there is no solution
cocrystal AB, since the concentration y in domain IV is complexation. This is shown in the lower two curves of Figure
supersaturated with respect to the cocrystal. 3a. The top two curves represent conditions with higher order
Effect of Complex Formation in Solution on the Solubility complexes, 1:2, and are discussed in the following section.
of Cocrystals AB (1:1). 1:1 Solution Complexation. When The K11 determined from the binary cocrystal solubility as a
dissolution of a 1:1 cocrystal of an API (A) and ligand (B) leads function of ligand concentration can also be used to predict the
to 1:1 complex formation in solution, the equilibrium reactions solubility increase of single component crystal (A) in solutions
are of ligand, according to

Ksp KS
ABsolid y\z Asoln + Bsoln (12) Asolid y\z Asoln (23)
K11
Asoln + Bsoln y\z ABsoln (13) K11
Asoln + Bsoln y\z ABsoln (24)
Equilibrium constants for these reactions are the solubility
product where Ks is the equilibrium constant for the solubility of crystal
A, and K11 is the binding constant for the formation of the
Ksp ) [A][B] (14) complex AB in solution. The equilibrium reaction for complex
formation in solution, eq 24, is the same as that in the case of
and the binding constant for the 1:1 complex formed in solution binary cocrystals presented above, eq 13. Using the mass balance
for total A and total B, the solubility of the single component
[AB] [AB] A is given by
K11 ) ) (15)
[A][B] Ksp
K11[A]o[B]T
From the mass balances for A and B in solution [A]T ) [A]o + (25)
1 + K11[A]o
[A]T ) [A] + [AB] (16)
where [A]o is the intrinsic solubility of crystal A in the absence
[B]T ) [B] + [AB] (17) of ligand. When the solubility of the cocrystal in solutions of
Phase Solubility Diagrams of Cocrystals Crystal Growth & Design, Vol. 6, No. 2, 2006 595

account the equilibrium constants for these reactions, K12 can

be expressed in terms of K11 and Ksp by

[AB2] [AB2]
K12 ) ) (27)
[AB][B] K11Ksp[B]

The mass balance of A now becomes

[A]T ) [A] + [AB] + [AB2] (28)

Substituting expressions developed for Ksp, K11, and K12 into

this equation leads to an expression for [A]T in terms of the
free ligand concentrations and equilibrium constants by

Ksp
[A]T ) + K11Ksp + K11K12Ksp[B] (29)
[B]
If [B] is not known, the expression for total ligand concentration
in solution

[B]T ) [B] + K11Ksp + 2K11K12Ksp[B] (30)

can be rearranged and substituted into the mass balance equation

for [A]T, eq 29 obtaining

Ksp(1 + 2K11K12Ksp)
[A]T ) + K11Ksp +
[B]T - K11Ksp
K11K12Ksp([B]T - K11Ksp)
(31)
1 + 2K11K12Ksp

This equation can be simplified for the case in which K11Ksp ,

[B]T, and 2K12K11Ksp , 1 to give

Ksp
[A]T ) + K11Ksp + K11K12Ksp[B]T (32)
[B]T

Figure 3. Effect of solution complexation on the solubility of (a) To gain information about the shape of this curve, the first
cocrystal AB and (b) single component crystal A as a function of total derivative d[A]T/d[B]T is examined and reveals that a plot of
ligand concentration. The cocrystal solubility was calculated from eq [A]T vs [B]T would be concave upward according to the change
22 or 32 depending on the solution complex stoichioimetries as in slope from negative to positive, corresponding to low values
discussed in the text: no complex, 1:1 complex, and 1:1 + 1:2 of [B]T and high values of [B]T, respectively.
complexes. Cocrystal Ksp ) 4.5 × 10-4 M2 and the complexation The total ligand concentration at which the cocrystal has the
constant values used are indicated in the graph. A minimum cocrystal
solubility is predicted when a 1:2 complex is formed, and this minimum solubility ([A]T has the minimum value) is calculated
concentration can be calculated from eq 35. Single component crystal from
solubility was calculated from eq 25 or 36 considering the same solution
complex stoichiometries. d[A]T Ksp
) 0 when K11K12Ksp ) (33)
ligand is lower than that of the single component crystal, direct
d[B]T [B]T2
measurement of the solubility of A may not be possible due to
the crystallization of the cocrystal. Therefore, this method and solving for [B]T gives
provides a means of predicting the solubility of A in the presence

x
1
of ligand. [B]T ) (34)
1:1 and 1:2 Solution Complexation. The dependence of K11K12
cocrystal AB solubility on ligand concentration will reveal the
presence of 1:1 and 1:2 solution complexes as described below. The minimum solubility of cocrystal is then obtained by
Consider a 1:2 complex of A and B formed by bimolecular substituting eq 34 in eq 32 and solving for [B]T to give
collisions where B binds to AB to form the soluble complex
AB2. The equilibrium expression for the formation of the AB2 [A]T,min ) Ksp(K11 + 2xK11K12) (35)
complex is
Thus, the concentration of ligand at which the minimum
K12 cocrystal solubility occurs is inversely proportional to the
ABsoln + Bsoln y\z AB2,soln (26) binding constants and independent of the solubility product,
whereas the minimum cocrystal solubility is directly proportional
The equilibria described by eqs 12 and 13 lead to the formation to the solubility product and the binding constants. This behavior
of the 1:2 complex in a stepwise fashion and by taking into is shown in the plot of [A]T vs [B]T for a hypothetical system
596 Crystal Growth & Design, Vol. 6, No. 2, 2006 Nehm et al.

using eqs 30 and 31 and assuming that Ksp ) 4.5 × 10-4 M2, Table 1. Solubilitiesa of Cocrystal CBZ/NCT and Single
K11 ) 12.7 M-1, and K12 ) 0.5 K11 or K12 ) 2 K11, Figure 3a. Component Crystals, CBZ(III) and NCT(I), in Organic Solvents at
T ) 25 °C
The solubility of the single component crystal of A is also
dependent on the complexation behavior in solution. For 1:1 compound ethanol (M) 2-propanol (M) ethyl acetate (M)
and 1:2 complex formation in solution and based on the CBZ/NCT 0.116b ( 0.003 0.044c ( 0.003 0.024d ( 0.001
equilibrium reactions (eqs 24 and 26), the solubility of A is (1:1)
given by CBZ(III) 0.1080b ( 0.0001 0.039c,e ( 0.003 0.0440d,e ( 0.0001
NCT(I) 0.841 ( 0.008 0.496 ( 0.004 0.098 ( 0.002

( K11[A]o + K12K11[A]o[B]
)
a Solubility values are the mean ( standard deviation of n ) 3.

[A]T ) [A]o + [B]T (36) b Statistically significant difference in solubilties, P < 0.05. c Statistically
1 + K11[A]o + 2K12K11[A]o[B] insignificant difference in solubilities, P > 0.10. d Statistically significant
difference in solubilities, P < 0.001. e Statistically insignificant difference
This equation predicts that the solubility of A increases in a in solubilities, P > 0.05.
nonlinear fashion as [B]T increases, since the slope is a function
of [B]. A plot of [A]T vs [B]T according to this equation is shown
filter (Osmonics, Minnetonka, MN), allowing at least 2 mL to saturate
in Figure 3b by calculating [B] from the mass balance equation the filter prior to sample collection. Samples were diluted with the same
and binding constants according to solvent in which the solubility analysis was performed. The solutions
equilibrated within 48 h and average sample concentration differed by
[B]T ) [B] + K11[A]o[B] + 2K11K12[A]o[B]2 (37) <2% at 24 and 48 h. CBZ concentrations were calculated by measuring
the absorbance of CBZ (λmax ) 284 nm) by UV/Vis spectroscopy
and solving for [B] from the quadratic equation. Values for the (Beckman DU-650, Fullerton, CA). The CBZ concentration corres-
ponds to the CBZ/NCT solubility based on the 1:1 molar ratio co-
binding constants are the same as those used for the prediction crystal.
of cocrystal solubilities and are K11) 12.7 M-1, K12 ) 0.5K11,
Because of the cocrystal to CBZ dihydrate transformation in water,17
or K12 ) 2K11. Theoretical models for evaluating binding the solubility in aqueous systems is not reported here. The solubility
constants from the solubility analysis of single component of CBZ/NCT in NCT solutions of ethanol, 2-propanol, or ethyl acetate
crystals were developed by Higuchi and Zuck3 for 1:1 complexes was determined by the same methods as described above. A large
and Higuchi and Bolton20 for higher order complexes and dilution factor (∼1000×) was necessary to maintain linearity according
reviewed by Connors and Higuchi.1,28 to Beer’s Law. Under these conditions, the complexation reported in
the results section is not detectable. NCT solutions were prepared by
diluting stock solutions of NCT with the respective solvents. The NCT
Materials and Methods concentrations studied varied from 30 to 50% of the NCT(I) solubility
in the solvents (Table 1).
Materials. Anhydrous monoclinic carbamazepine (CBZ(III); lot no. Solubility of Cocrystal Components in Organic Solvents. Equi-
093K1544 USP grade and 99.8% purity) was purchased from Sigma librium solubility of CBZ and NCT in ethanol, 2-propanol, or ethyl
Chemical Company (St. Louis, MO), stored at 5 °C over anhydrous acetate was determined from undersaturation by adding excess solid
calcium sulfate and used as received. Nicotinamide (NCT(I); lot no. CBZ(III) or NCT(I) to each solvent, respectively. Experiments were
122K0077) was purchased from Sigma Chemical Company (St. Louis, carried out in a similar manner as described for the cocrystal, and
MO) and used as received. Solid state forms were identified by X-ray samples were analyzed by UV/Vis spectroscopy for CBZ and HPLC
powder diffraction or differential scanning calorimetry. for NCT. The HPLC instrument (Agilent 1100 series, Palo Alto, CA)
Ethyl acetate and 2-propanol were of HPLC grade and were was equipped with a UV diode array detector (Agilent Technologies
purchased from Fisher Scientific (Fair Lawn, NJ). Anhydrous ethanol G1315B, Palo Alto, CA) and contained a Zorbax SB-Phenyl 2.1 ×
(200 proof) was USP grade and was purchased from Pharmco 150 mm column packed with 5 µm media (Agilent Technologies, Palo
(Brookfield, CT). Alto, CA). The maximum wavelength for absorbance for NCT was set
Preparation of Cocrystals. Carbamazepine/Nicotinamide (CBZ/ at 260 nm. NCT concentrations were analyzed following a 15 min
NCT). CBZ/NCT cocrystals (1:1 molar ratio) were prepared by either isocratic method using a 60/40 water/methanol mobile phase containing
a solvothermal method or by methods based on the phase diagrams 0.1% trifluoroacetic acid.
presented here, where supersaturation with respect to the cocrystal is X-ray Powder Diffraction (XRPD). Powder diffraction patterns
created by the effect of cocrystal components on reducing the solubility of solid phases were recorded with a Scintag X-ray diffractometer
of the molecular complex to be crystallized. In the latter method, the (Franklin, MA) using CuKR radiation (λ ) 1.54 Å), tube voltage of
cocrystal was prepared from solutions of a nonstoichiometric composi- 40 kV, and tube current of 20 mA. The intensities were measured at
tion of CBZ and NCT or by suspending solid CBZ(III) in NCT aqueous 2-theta values from 2° to 50° at a continuous scan rate of 5 °/min.
or organic solutions or by suspending CBZ(III) and NCT(I) in pure When steady-state concentrations were obtained during the solubility
solvents. No cooling or evaporation was necessary for cocrystallization. experiments, the solid phase was analyzed by X-ray powder diffraction,
In the solvothermal method, however, supersaturation was created by and results were compared to the diffraction patterns of each pure phase.
cooling a solution of NCT (0.35 g; 2.83 mmol) and CBZ (0.67 g; 2.83
mmol) in ethyl acetate (50 g) from 40 to 25 °C. The solid phases by Infrared (IR) Spectroscopy. FTIR spectra of solid phases were
any of the above methods were harvested by vacuum filtration and collected on a Bruker Vertex 70 FT-IR (Billerica, MA) unit equipped
dried at room temperature (22-23 °C) under reduced pressure (25 with a DTGS detector. Samples were placed on a zinc selenide (ZnSe)
mmHg) on Whatman #50 filter paper (Maidstone, England) for 30 min Attenuated Total Reflectance (ATR) crystal accessory and 64 scans
to remove loosely bound solvent. The solid phases were confirmed to were collected for each sample at a resolution of 4 cm-1 over a
be CBZ/NCT cocrystal by X-ray powder diffraction, FT-IR spectros- wavenumber region of 4000-600 cm-1.
copy, and differential scanning calorimetry. Cocrystals were stored at Differential Scanning Calorimetry (DSC). The thermal behavior
5 °C over anhydrous calcium sulfate. of solid phases was studied using a TA Instruments 2920 modulated
Solubility of Cocrystals in Organic Solvents. The equilibrium DSC (TA Instruments, New Castle, DE) with refrigerated cooling
solubility of CBZ/NCT cocrystal in pure organic solvents (ethanol, system (RCS) in standard mode. Approximately 5-10 mg samples were
2-propanol, or ethyl acetate) was determined from undersaturation by weighed into aluminum DSC pans, crimped, and heated at 10 °C/min.
adding excess cocrystal solid phase in each solvent. The suspensions A dry nitrogen purge was used at a flow rate of 50 cm3/min through
were stirred with magnetic stirrers in 20 mL glass vials at constant the DSC cell and 110 cm3/min through the RCS. The DSC was
temperature (25 ( 0.5 °C) maintained with a circulating temperature calibrated for temperature using n-dodecane (Tm ) -9.65 °C, Aldrich,
bath (Neslab RTE-110, Portsmouth, NH). Samples were drawn at Milwaukee, WI) and indium (Tm ) 156.60 °C, TA Instruments, New
various time intervals over 72 h and filtered using a 0.22 µm nylon Castle, DE) at 10 °C/min. Indium was used to calibrate the cell constant.
Phase Solubility Diagrams of Cocrystals Crystal Growth & Design, Vol. 6, No. 2, 2006 597

Figure 4. Solubility of CBZ/NCT cocrystal (1:1) and single component Figure 5. Total CBZ concentration in equilibrium with cocrystal, CBZ/
crystal of CBZ(III) at 25 °C as a function of total NCT concentration NCT, at 25 °C as a function of the inverse total NCT concentration in
in ethanol, 2-propanol, and ethyl acetate. The solid lines represent the (9) ethanol, (2) 2-propanol, and (b) ethyl acetate, showing the linear
predicted solubility, according to eq 43 with values for Ksp and K11 in dependence predicted by eq 43.
Table 3. Filled symbols are experimental cocrystal solubility values in
Table 2. Linear Regression Analysis According to Eq 43
(9) ethanol, (2) 2-propanol, and (b) ethyl acetate. The dashed lines
represent the predicted solubility of CBZ(III) according to eq 25 and equation of line (( standard error R2
the K11 values calculated from the cocrystal solubility analysis, Table solvent of slope and y-intercept) value
3. Open symbols are experimental CBZ(III) polymorph solubility values ethanol y ) 0.0129 (( 0.0006)x + 0.0042 (( 0.0037) 0.98
in pure solvent. 2-propanol y ) 0.0016 (( 0.0001)x + 0.010 (( 0.001) 0.96
ethyl acetate y ) 0.00045 (( 0.00003)x + 0.0057 (( 0.0008) 0.97

Results Table 3. CBZ/NCT Cocrystal Solubility Product, Ksp, and Solution

Complexation Constant, K11, in Organic Solvents
The solubilities of cocrystal CBZ/NCT (1:1) and the single solvent Ksp (M2) K11 (M-1)
component crystals of CBZ and NCT in three organic solvents
ethanol 0.0129 ( 0.0006a 0.3 ( 0.3a
are presented in Table 1. 2-propanol 0.0016 ( 0.0001 6.3 ( 0.7
The cocrystal solubility depends on the solvent as follows: ethyl acetate 0.00045 ( 0.00003 12.7 ( 1.8
ethanol > 2-propanol > ethyl acetate. CBZ is the least soluble a Standard error in each solvent system.
of the two cocrystal components in these solvents, and the
solubility of the most thermodynamically stable polymorph, the total CBZ concentration in solution versus the inverse of
monoclinic form (III) ranks as follows: ethanol > 2-propanol the total NCT concentration according to
) ethyl acetate. NCT solubility follows the same order as the
cocrystal solubility. Ksp
The solubility ratio of cocrystal to single component crystal [CBZ] ) (40)
[NCT]
is important when considering solvents and conditions for the
crystallization and isolation of cocrystals by solvent evaporation The results follow this linear dependence as shown in Figure
or cooling methods. Results show that the largest difference in 5.
the solubility of cocrystal and CBZ(III) is observed in ethyl Examination of the linear regression analysis in Table 2
acetate, with a solubility ratio (cocrystal/CBZ(III)) of 0.55. The reveals that in 2-propanol and ethyl acetate the y-intercepts are
solubilities of cocrystal and CBZ(III) are very similar in ethanol statistically different from zero, suggesting that a 1:1 solution
(ratio ) 1.07 and significant at P < 0.05), while in 2-propanol complex is in equilibrium with the dissolved single components
the solubilities are not significantly different (P > 0.10). and the cocrystal according to the equilibrium reactions given
The cocrystal solubility decreases nonlinearly as the total by eqs 12 and 13, that when applied to this system become
nicotinamide concentration increases in the solvents studied, as
shown in Figure 4. Ksp
CBZ/NCTsolid y\z CBZsoln + NCTsoln (41)
This behavior is anticipated from the solubility product
according to the equilibrium reaction, eq 1, in the absence of K11
solution complexation and the solubility of cocrystal given by CBZsoln + NCTsoln y\z CBZ/NCTsoln (42)
eq 9. When applied to this cocrystal the equilibrium reaction is
The solubility products and complexation constants were
Ksp calculated from eq 22 when K11Ksp , [B]T in the form of
CBZ/NCTsolid y\z CBZsoln + NCTsoln (38)
Ksp
and the solubility product becomes [CBZ]T ) + K11Ksp (43)
[NCT]T
Ksp ) [CBZ][NCT] (39) Table 3 presents the Ksp values calculated from the slopes and
the K11 values calculated from the y-intercepts according to eq
Solubility products were calculated from the slopes of plots of 43.
598 Crystal Growth & Design, Vol. 6, No. 2, 2006 Nehm et al.

component phases. Thus, in addition to explaining the solubility

behavior of binary cocrystals, the analysis presented here can
reveal complexation in solution and identify well-defined regions
in the parameter space from which to crystallize the desired
components and where transformations may occur.
Our results show that the CBZ/NCT cocrystal solubility
decreases as the ligand concentration, NCT, increases in organic
solutions. This behavior is expected based on the cocrystal Ksp
alone when solution complexation is negligible, as well as by
Ksp and K11 when there are complexes in solution with the same
stoichiometry as the cocrystal.
In contrast to the phase solubility analysis that studies the
solubility of substrate or API starting with the single component
crystal, here we consider the solubility of cocrystals when the
cocrystal under study is the only solid phase at equilibrium.
The theoretical analysis presented in this report shows that (1)
the solubility of a 1:1 cocrystal AB is described by a solubility
Figure 6. Comparison of experimental and calculated cocrystal product, (2) the cocrystal solubility is increased by a constant
solubilities as a function of ligand concentration, when solution value when there is 1:1 solution complexation, because the
complexation is neglected, according to eq 40, and using the Ksp values concentration of this complex is a constant (the product of
calculated from the slopes of the lines in Figure 5, Table 3. Symbols K11Ksp), and (3) the cocrystal solubility goes through a minimum
represent experimental solubility values in (9) ethanol, (2) 2-propanol,
value when there are 1:1 and 1:2 solution complexes. Thus, a
and (b) ethyl acetate.
graphical representation of the cocrystal solubility dependence
on ligand concentration will serve as a diagnostic tool for the
The cocrystal solubility curves predicted from this equation
stoichiometry of solution complexes.
and shown in Figure 4 are in very good agreement with the
experimental solubility values. Ksp values follow the same This analysis was applied to the solubility of the CBZ/NCT
relative order as the cocrystal solubilities: ethanol > 2-propanol (1:1) cocrystal and allowed for measurement of the solubility
> ethyl acetate, whereas K11 values follow a trend inverse to product and the 1:1 equilibrium constant for solution complex
the solubility of cocrystal. Although in ethanol the y-intercept formation. The Ksp values were dependent on solvent and were
was not statistically different from zero, the K11 was still proportional to the solubility of cocrystal in pure solvent. The
calculated; however, the small K11 value obtained and the large K11 values were inversely related to cocrystal solubility and to
standard error associated with it suggests that 1:1 complexation cocrystal component solubilities. This indicates that complex
is negligible in this solvent. formation is favored in solvents where these components have
Figure 6 compares the experimental and predicted cocrystal lower solubilities. Similar trends have been reported for solution
solubilities had solution complexation been neglected, according complexation of various drugs with ligands as reviewed by
to eq 40. This shows that 1:1 solution complexation of cocrystal Higuchi and Connors1 and have been explained by considering
components increases the solubility of a 1:1 cocrystal by a that solubility of a solute reflects the affinity of the solute with
constant, which is the product of Ksp and K11. Thus, if only the the solvent; thus, higher solubilities favor solute-solvent
Ksp calculated from the solubility of cocrystal in pure solvents interactions, whereas lower solubilities favor solute-solute
was considered to predict the cocrystal solubility dependence interactions.
on nicotinamide concentration, the cocrystal solubility would Given these observations, it is of interest to consider what
have been underestimated. drives the molecular complexes of CBZ and NCT in solution
Preliminary experiments to measure the solubility of the single and how these might relate to the supramolecular complex in
component crystal, CBZ(III), in nicotinamide solutions at the cocrystal. The structure of the cocrystal11 shows that
concentrations above the solubility of the cocrystal, S, showed nicotinamide links with the carbamazepine homodimer (car-
transformation of CBZ(III) to cocrystal. Figure 4 shows the boxamide dimer) by forming a hydrogen-bonded tape where
measured solubility of CBZ(III) in pure solvent29 and the donors and acceptors of both amide groups are fulfilled. The
predicted dependence on NCT solution concentration as a result nitrogen atom within the ring structure of nicotinamide does
of solution complexation. K11 values in Table 3 were used in not form strong hydrogen bonds. There are also π-π interac-
eq 25 to predict the solubility of CBZ(III) in nicotinamide tions between the carbamazepine and the nicotinamide rings.
solutions. The solubility of CBZ(III) and transformation to Studies of solution complexation of nicotinamide and other
cocrystal in solutions of cocrystal components are currently aliphatic amides with various APIs (diazepam, steroids, griseof-
under investigation in our laboratory. ulvin) concluded that the aromaticity of the pyridine ring
promotes plane-to-plane stacking of molecules in solution and
favors complex formation.32 Amides have also been shown to
Discussion
participate in hydrogen bonding with carboxylic acids in
Theoretical analysis and experimental results presented here solution.21 While we only studied three solvents, ethanol,
show that the solubility of cocrystal depends on the concentra- 2-propanol, and ethyl acetate, the hydrogen-bonding properties
tion of cocrystal components in solution and that this dependence differ and the ability of the solvent to interact as hydrogen bond
can be explained by a cocrystal solubility product and by the donor and acceptor appears to weaken the complex formation.
binding constants of complexes formed in solution. The solubil- K11 values calculated for CBZ and NCT in this study are
ity product behavior is analogous to that described for sparingly similar to those reported for drug nicotinamide complexes in
soluble salts;23,25,30,31 however, with the main difference that water at 25 °C in the range of 4 to 20 M-1. The strongest binding
cocrystals dissociate into molecules that can crystallize as single is exhibited by moricizine (17 M-1 in pH 6 and 8 M-1 in pH
Phase Solubility Diagrams of Cocrystals Crystal Growth & Design, Vol. 6, No. 2, 2006 599

7, both parabolic fits),33 halofantrine (5 to 9 and 18 to 20 M-1, complexation. We acknowledge Phil Zocharski for help with
depending on linear or parabolic fits),34 and nifedipine (9 M-1)35 the nicotinamide solubility experiments. We gratefully acknowl-
followed by diazepam, griseofulvin, progesterone, 17-β estradiol, edge a research gift from Boehringer Ingelheim and partial
testosterone,32 and oxytetracycline,20 all between 4 and 5 M-1. support from GM07767 NIGMS (S.J.N.) and the Fred W. Lyons
Although CBZ/NCT did not form a 1:2 complex in the Jr. Fellowship (B.R.S.) from the College of Pharmacy, Univer-
solvents studied, NCT has been reported to form a 1:2 complex sity of Michigan. The contents of this publication are solely
with other compounds, and the K12 values span a much larger the responsibility of the authors and do not necessarily represent
range than the K11 values. For example, the highest K12 value the official views of NIGMS and Boehringer Ingelheim.
observed is of the halofantrine and NCT complex (82 to 150
and 30 to 38 M-1, depending on linear or parabolic fits),34 References
followed by the steroids (progesterone, 17-β estradiol, and (1) Higuchi, T.; Connors, K. A. Phase-Solubility Techniques. In AdVances
testosterone) between 25 and 45 M-2,32 and moricizine exhibits in Analytical Chemistry and Instrumentation; Nurnberg, H. W., Ed.;
Wiley-Interscience: New York, 1965; pp 117-212.
the smallest K12 value of 1 M-1 in pH 6 and 5 M-1 in pH 7 (2) Higuchi, T.; Zuck, D. A. Investigation of Some Complexes Formed
(both parabolic fits).33 in Solution by Caffeine. II. Benzoic Acid and Benzoate Ion. J. Am.
Phase solubility methods have been applied to pharmaceutical Pharm. Assoc. 1953, 42, 132-138.
compounds since the 1950s to identify the solution complexation (3) Higuchi, T.; Zuck, D. A. Investigation of Some Complexes Formed
in Solution. III. Interactions Between Caffeine and Aspirin, p-
behavior of poorly water soluble drugs or substrates with a Hydroxybenzoic Acid, m-Hydroxybenzoic Acid, Salicylic Acid,
soluble ligand.2,3,20,21 In these reports, the solubility of the single Salicylate Ion, and Butyl Paraben. J. Am. Pharm. Assoc. 1953, 42,
component crystal of a drug or substrate is studied and the solid 138-145.
phase is examined for the formation of solid-state complexes (4) Higuchi, T.; Pitman, I. H. Caffeine Complexes with Low Water
that are less soluble than the single component crystal of Solubility: Synthesis and Dissolution Rates of 1:1 and 1:2 Caffeine-
Gentisic Acid Complexes. J. Pharm. Sci. 1973, 62, 55-58.
substrate. (5) Etter, M. C. Encoding and Decoding Hydrogen-Bond Patterns of
The focus of these earlier reports was to enhance the solubility Organic Compounds. Acc. Chem. Res. 1990, 23, 120-126.
of drugs in water and address the behavior of soluble solution (6) Leiserowitz, L.; Schmidt, G. M. J. Molecular Packing Modes. Part
complexes, or type A phase solubility diagrams. Formation of III. Primary Amides. J. Chem. Soc. A 1969, 2372-2382.
(7) Desiraju, G. R. Supramolecular Synthons in Crystal Engi-
solid-state complexes received less attention, although work was neering- -A New Organic Synthesis. Angew. Chem., Int. Ed. Engl.
done to identify the stoichiometry of complexes from the 1995, 34, 2311-2327.
solubility behavior. Models were developed that describe the (8) Aakeroy, C. B.; Beatty, A. M.; Helfrich, B. A.; Nieuwenhuyzen, M.,
solubility behavior in terms of solution complexation in the Do polymorphic compounds make good cocrystallizing agents? A
presence of insoluble complexes that resulted in type B solubility structural case study that demonstrates the importance of synthon
flexibility. Cryst. Growth Des. 2003, 3, 159-165.
diagrams.3,4,21 (9) Trask, A. V.; Motherwell, W. D. S.; Jones, W. Pharmaceutical
While cocrystal solubility is a useful operational variable its Cocrystallization: Engineering a Remedy for Caffeine Hydration.
value depends on the solution composition of cocrystal con- Cryst. Growth Des. 2005, 5, 1013-1021.
stituents, whereas under these conditions the solubility product (10) Childs, S. L.; Chyall, L. J.; Dunlap, J. T.; Smolenskaya, V. N.; Stahly,
B. C.; Stahly, G. P. Crystal Engineering Approach to Forming
is a constant. Solubility products are dependent on temperature, Cocrystals of Amine Hydrochlorides with Organic Acids. Molecular
solvent, and conditions that affect the activity coefficients of Complexes of Fluoxetine Hydrochloride with Benzoic, Succinic, and
cocrystal components in solution.23,27 In view of the fact that Fumaric Acids. J. Am. Chem. Soc. 2004, 126, 13335-13342.
the concentration of cocrystal components in solution can be (11) Fleischman, S. G.; Kuduva, S. S.; McMahon, J. A.; Moulton, B.;
varied independently, the solubility product is a meaningful Walsh, R. D. B.; Rodrı́guez-Hornedo, N.; Zaworotko, M. J. Crystal
Engineering of the Composition of Pharmaceutical Phases: Multiple-
parameter to consider when developing methods for cocrystal Component Crystalline Solids Involving Carbamazepine. Cryst.
screening or designing liquid solution formulations where Growth Des. 2003, 3, 909-919.
crystallization of cocrystal is to be avoided. Thus, an evaluation (12) Walsh, R. D. B.; Bradner, M. W.; Fleischman, S.; Morales, L. A.;
of Ksp from only the cocrystal solubility in pure solvents can Moulton, B.; Rodrı́guez-Hornedo, N.; Zaworotko, M. J., Crystal
Engineering of the Composition of Pharmaceutical Phases. Chem.
be used to estimate the cocrystal solubility dependence on ligand Commun. 2003, 186-187.
concentration. Although this may underestimate the cocrystal (13) Remenar, J. F.; Morissette, S. L.; Peterson, M. L.; Moulton, B.;
solubility (due to neglecting solution complexation), it is a useful MacPhee, J. M.; Guzman, H. R.; Almarsson, O. Crystal Engineering
and conservative prediction when developing solution formula- of Novel Cocrystals of a Triazole Drug with 1,4-Dicarboxylic Acids.
tions with components that form cocrystals. J. Am. Chem. Soc. 2003, 125, 8456-8457.
(14) Brader, M. L.; Sukumar, M.; Pekar, A. H.; McClellan, D. S.; Chance,
R. E.; Flora, D. B.; Cox, A. L.; Irwin, L.; Myers, S. R. Hybrid insulin
Conclusions cocyrstals for controlled release delivery. Nat. Biotechnol. 2002, 20,
In this paper we present mathematical models that explain 800-804.
(15) Bettinetti, G.; Caira, M. R.; Callegari, A.; Merli, M.; Sorrenti, M.;
the solubility behavior of cocrystals as a function of cocrystal Tadini, C. Structure and Solid-State Chemistry of Anhydrous and
component concentration in solution by considering the solubil- Hydrated Crystal Forms of the Trimethoprim-Sulfamethoxypyridazine
ity product and solution complexation of cocrystal components. 1:1 Molecular Complex. J. Pharm. Sci. 2000, 89, 478-489.
The immediate implication of the results is that cocrystal (16) Amai, M.; Endo, T.; Nagase, H.; Ueda, H.; Nakagaki, M. 1:1
solubility products and solution complexation constants can be Complex of Octadecanoic Acid and 3-Pyridinecarboxamide. Acta
Crystallogr., Sect. C 1998, 54, 1367-1369.
evaluated from solubility studies. This is shown for the (17) Rodrı́guez-Spong, B. Enhancing the Pharmaceutical Behavior of
carbamazepine/nicotinamide cocrystal in organic solvents. The Poorly Soluble Drugs Through the Formation of Cocrystals and
broader implications are that these concepts can be applied to Mesophases, Ph.D. Thesis., University of Michigan, 2005.
calculate crystallization diagrams for cocrystal preparation, to (18) Rodrı́guez-Spong, B.; Zocharski, P.; Billups, J.; McMahon, J.;
Zaworotko, M. J.; Rodrı́guez-Hornedo, N. Enhancing the Pharma-
develop cocrystal screening methods, and to formulate solutions ceutical Behavior of Carbamazepine Through the Formation of
with cocrystal components where crystallization is to be avoided. Cocrystals. AAPS J. 2003, 5, Abstract M1298.
(19) Zocharski, P.; Nehm, S.; Rodrı́guez-Spong, B.; Rodrı́guez-Hornedo,
Acknowledgment. We are grateful to Prof. Kenneth A. N., Can Solubility Products Explain Cocrystal Solubility and Predict
Connors for very helpful discussions and comments on solution Crystallization Conditions? AAPS J. 2004, 6, Abstract R6192.
600 Crystal Growth & Design, Vol. 6, No. 2, 2006 Nehm et al.

(20) Higuchi, T.; Bolton, S., The Solubility and Complexing Properties (29) Kelly, R. C. A Molecular Approach to Understanding the Directed
of Oxytetracycline and Tetracycline III. J. Am. Pharm. Assoc. 1959, Nucleation and Phase Transformation of Carbamazepine and Nitro-
48, 557-564. furantoin in Aqueous and Organic Solutions, Ph.D. Thesis, University
(21) Kostenbauder, H. B.; Higuchi, T. Formation of Molecular Complexes of Michigan, 2003.
by Some Water-Soluble Amides I. J. Am. Pharm. Assoc. 1956, 45, (30) Butler, J. N., Ionic Equilibrium: A Mathematical Approach; Addison-
518-522. Wesley: Massachusetts, 1964; pp 174-205.
(22) Grant, D. J. W.; Higuchi, T. Solubility BehaVior of Organic (31) Sohnel, O.; Garside, J. The Thermodynamic Driving Force for
Compounds; Wiley: New York, 1990; Vol. 21. Crystallization from Solution. J. Cryst. Growth 1979, 46, 238-240.
(23) Rodrı́guez-Clemente, R. Complexing and Growth Units in Crystal (32) Rasool, A. A.; Hussain, A. A.; Ditter, L. W. Solubility Enhancement
Growth from Solutions of Electrolytes. J. Cryst. Growth 1989, 98, of Some Water-Insoluble Drugs in the Presence of Nicotinamide and
617-629. Related-Compounds. J. Pharm. Sci. 1991, 80, 387-393.
(24) Shefter, E.; Higuchi, T., Dissolution Behavior of Crystalline Solvated (33) Hussain, M. A.; DiLuccio, R. C.; Maurin, M. B. Complexation of
and Nonsolvated Forms of Some Pharmaceuticals. J. Pharm. Sci. Moricizine with Nicotinamide and Evaluation of the Complexation
1963, 52, 781-791. Constants by Various Methods. J. Pharm. Sci. 1993, 82, 77-79.
(25) Nielsen, A. E.; Toft, J. M. Electrolyte Crystal Growth Kinetics. J. (34) Lim, L.-Y.; Go, M.-L. Caffeine and nicotinamide enhances the
Cryst. Growth 1984, 67, 278-288. aqueous solubility of the antimalarial agent halofantrine. Eur. J.
(26) Khankari, R. K.; Grant, D. J. W. Pharmaceutical Hydrates. Ther- Pharm. Sci. 2000, 10, 17-28.
mochem. Acta 1995, 248, 61-79. (35) Suzuki, H.; Sunada, H. Mechanistic Studies on Hydrotropic Solu-
(27) Tomazic, B.; Nancollas, G. H. The Kinetics of Dissolution of Calcium bilization of Nifedipine in Nicotinamide Solution. Chem. Pharm. Bull.
Oxalate Hydrates. J. Cryst. Growth 1979, 46, 355-361. 1998, 46, 125-130.
(28) Connors, K. A. Binding Constants: The Measurement of Molecular
Complex Stability; John Wiley & Sons: New York, 1987. CG0503346