Dipali
Ahuja
a,
Michael
Svärd
*ab and
Åke C.
Rasmuson
*ab
aSynthesis and Solid State Pharmaceutical Centre, Bernal Institute, Department of Chemical Sciences, University of Limerick, Co. Limerick, Ireland
bDepartment of Chemical Engineering, KTH Royal Institute of Technology, Teknikringen 42, SE-10044 Stockholm, Sweden. E-mail: micsva@kth.se; akera@kth.se
First published on 2nd April 2019
The influence of temperature and solvent on the solid–liquid phase diagram of the 1:1 sulfamethazine–salicylic acid co-crystal has been investigated. Ternary phase diagrams of this co-crystal system have been constructed in three solvents: methanol, acetonitrile and a 7:3 (v/v) dimethylsulfoxide–methanol mixture, at three temperatures. The system exhibits congruent dissolution in acetonitrile and the co-crystal solubility has been determined by a gravimetric technique. The Gibbs energy of co-crystal formation from the respective solid components has been estimated from solubility data, together with the corresponding enthalpic and entropic component terms. The Gibbs energy of formation ranges from −5.7 to −7.7 kJ mol−1, with the stability increasing with temperature. In methanol and the DMSO–methanol mixture, the co-crystal dissolves incongruently. It is shown that the solubility ratio of the pure components cannot be used to predict with confidence whether the co-crystal will dissolve congruently or incongruently. The size of the region where the co-crystal is the only stable solid phase is inversely related to the pure component solubility ratio of salicylic acid and sulfamethazine.
For design and operation of a crystallization process for the manufacture of co-crystals, a complete and detailed phase diagram is crucial as it reveals the stability regions for the different crystalline phases. Based on a proper phase diagram identifying the region where the co-crystal is the stable solid phase, the conditions for manufacturing can be determined. Parameters like solvent and temperature can significantly affect the solubility of the co-crystal and alter the shape of the phase diagram.28,42 It has been suggested that a large solubility difference between the two components most likely leads to an incongruent system.42 Of general importance is whether the co-crystal dissolves congruently or not, the width of the region where the co-crystal is stable, and the co-crystal solubility. The wider the co-crystal region, the more robust the process of manufacturing becomes. If the co-crystal dissolves congruently a simple cooling crystallization can be performed, and the design of the process is even more facilitated. However, if the starting solution composition is adjusted appropriately,19,20 a non-congruent system is no major obstacle as long as the co-crystal region is not too narrow.
It is important to understand the co-crystal stability and formation in terms of key thermodynamic parameters like Gibbs energy. In the literature, only a few studies specifically treat thermodynamics of co-crystals. Nehm et al. defined the solubility product of a co-crystal as the product of the component concentrations and demonstrated it for the carbamazepine–nicotinamide co-crystal system.43 In 2007, Chiarella et al. showed that crystallization from a solution containing stoichiometric amounts of pure components might or might not form pure co-crystal based on solvent choice, and explained this on the basis of phase diagrams for the 1:1 trans-cinnamic acid–nicotinamide co-crystal.28 The factors responsible for the formation and stability of co-crystals with different stoichiometry using carbamazepine–4-aminobenzoic acid as a model system have been identified by Rodríguez-Hornedo et al.44 ter Horst et al. utilized thermodynamic principles to develop a method for co-crystal screening.45 In 2012, Leyssens et al. showed the importance of the solvent for synthesis and stability of diverse stoichiometric caffeine–maleic acid co-crystals.46 In acetone where the relative solubility between the pure components is high, the 2:1 co-crystal is inaccessible, whereas in ethyl acetate with a reduced relative solubility, this zone becomes accessible. Croker et al. reported the formation of a new co-crystal with p-toluenesulfonamide and triphenylphosphine and studied the effect of the solvent by constructing phase diagrams in two solvents: acetonitrile and dichloromethane.19 Zhang and Rasmuson studied the thermodynamics and crystallization of theophylline–oxalic acid and theophylline–glutaric acid 1:1 co-crystals.47,48 They estimated Gibbs energies of co-crystal formation from solubility data and investigated the effect of polymorphism on the phase diagrams. In 2017, Bacchi et al. constructed ternary phase diagrams for a liquid API, propofol, with solid coformers: bipyridine and phenazine.49 They employed co-crystallization to stabilize the liquid drug in a crystalline form.
Sulfonamides are considered able to form co-crystals, as they possess both hydrogen bond donor and acceptor groups. Sulfamethazine (SMT), a sulfonamide drug, is an antimicrobial and an anti-infective agent. It is used as a veterinary medicine to treat a variety of infections. In humans, it is used for the treatment of urinary tract infection, chlamydia, malaria, rheumatoid fever and toxoplasmosis.50 It belongs to the BCS class II, i.e. it is known to have a high permeability but a low solubility, and consequently a low bioavailability. For pure solid SMT, only one pure component crystal structure has been reported, belonging to the monoclinic crystal system.51,52 SMT can form co-crystals with several carboxylic acids. A 1:1 co-crystal between SMT and salicylic acid (SA) has been reported.53 The chemical structures of SMT and SA are depicted in Fig. 1.
In this work, the SMT–SA co-crystal system is used as a model for construction and analysis of phase diagrams. We have investigated the thermodynamics of the SMT–SA co-crystal in three different solvent systems. Ternary phase diagrams have been constructed in methanol and acetonitrile at 10, 20, 30 °C and in a 7:3 (v/v) dimethyl sulfoxide–methanol mixture at 20, 30 and 40 °C. The objectives of this work include identifying the stability regions of the co-crystal in the three solvent systems, and studying the effect of coformer solubility and temperature on the appearance, shape, and symmetry of the phase diagrams. We have also estimated the Gibbs energy of co-crystal formation from the solubility data. The volumetric productivity and co-crystal yields in the three systems are discussed.
Solvents were selected with the ambition to cover congruent as well as incongruent conditions, with the starting point in the expectation that a symmetric and congruent system would be obtained when the pure components had similar solubility, and asymmetric and incongruent systems obtained for larger solubility differences. Preliminary solubility experiments were carried out, based on which the three solvent systems were selected. The solvents evaluated were methanol, acetonitrile, water, chloroform, acetone, ethyl acetate, dimethyl sulfoxide, N,N-dimethyl acetamide, N,N-dimethyl formamide, and dimethyl sulfoxide–methanol mixtures of different ratios. The preliminary solubility experiments revealed a high SA to SMT solubility ratio in ethyl acetate and in methanol. Methanol was chosen as an example of a likely incongruent dissolution over ethyl acetate on the basis of a higher SMT solubility. In acetonitrile the solubility ratio was lower than in acetone, and was selected as a likely congruent dissolution system. In chloroform, the settling velocity of solid was low because of similar densities of SMT, SA and chloroform, and in water the problem was the same. Hence, for practical reasons, water and chloroform were not selected. SMT exhibits about the same low solubility in all solvents tested except for DMSO, DMA and DMF. However, SMT forms solvates in pure DMSO, DMF and DMA.54 A few DMSO–methanol mixtures were evaluated with the intent to keep the DMSO content high. A 7:3 (v/v) DMSO–methanol mixture, prepared by mixing 7 volume parts of DMSO and 3 volume parts of methanol, was chosen as it fulfils the three targets, i.e. high SMT solubility, no solvate formation and a low solubility ratio.
Because of the high boiling point (189 °C) and accordingly low volatility of DMSO, the gravimetric method was not feasible for DMSO–methanol solutions. Solution concentrations were determined by HPLC for the determination of the solubility of SMT and SA in this solvent system. This entailed the construction of calibration curves i.e. peak area vs. concentration using stock solutions of known concentrations of SMT and SA in DMSO–methanol. The calibration lines showed good linearity (R2 = 0.99). The saturated solution samples were filtered into clean glass vials. Peak areas for the saturated solutions of SMT and SA at different temperatures were obtained and the corresponding concentrations in turn obtained using the calibration curves.
Fig. 2 Experimental and calculated PXRD patterns for SMT–SA co-crystal (a), main hydrogen-bond motif in the crystal structure of the 1:1 SMT–SA co-crystal (b).51 |
Solvent | T (°C) | Solubility (g solute/g solvent) | Solubility (mol L−1) | Standard deviation (mol L−1, n = 4) | Solubility ratio (SA/SMT) (M/M) | |||
---|---|---|---|---|---|---|---|---|
SMT | SA | SMT | SA | SMT | SA | |||
Methanol | 10 | 0.0092 | 0.4788 | 0.0262 | 2.7456 | 0.0010 | 0.0014 | 104.7 |
20 | 0.0140 | 0.5688 | 0.0399 | 3.2615 | 0.0011 | 0.0015 | 81.7 | |
30 | 0.0228 | 0.7007 | 0.0650 | 4.0178 | 0.0013 | 0.0018 | 61.8 | |
Acetonitrile | 10 | 0.0102 | 0.0634 | 0.0291 | 0.3742 | 0.0014 | 0.0019 | 12.8 |
20 | 0.0151 | 0.0881 | 0.0431 | 0.5051 | 0.0018 | 0.0016 | 11.7 | |
30 | 0.0209 | 0.1265 | 0.0592 | 0.6827 | 0.0015 | 0.0017 | 11.5 | |
7:3 (v/v) DMSO–methanol | 20 | 0.4762 | 0.7011 | 1.7237 | 5.1152 | 0.0084 | 0.0079 | 2.96 |
30 | 0.5029 | 0.7345 | 1.8202 | 5.3727 | 0.0074 | 0.0081 | 2.95 | |
40 | 0.5232 | 0.7620 | 1.8940 | 5.5602 | 0.0078 | 0.0071 | 2.93 |
Fig. 3 Van't Hoff plots for SMT and SA in 7:3 (v/v) DMSO–methanol mixture, where xSMT and xSA correspond to mole fraction solubility of SMT and SA, respectively. |
In methanol, the phase diagram (Fig. 4a) is rather asymmetric as would be expected given the high solubility ratio of SA to SMT (Table 1), and the co-crystal region is significantly skewed towards the more soluble component, SA. In addition, the region where the co-crystal is the stable solid phase is very narrow. A very narrow co-crystal region makes the manufacturing process more difficult to design and operate. The dissolution of the co-crystal is incongruent, i.e. it is not possible to establish a solid–liquid equilibrium between the co-crystal solid phase and a stoichiometric solution. For this reason, the solubility of the co-crystal cannot be determined by traditional methods. Continued dissolution of this co-crystal would tend to move the solution composition to the point where the stable solid phase is a mixture of SMT and co-crystal, i.e. the invariant point. The phase diagram at three temperatures as a function of SMT and SA concentrations has been depicted in Fig. 4b.
In acetonitrile, the solubility of SMT is slightly higher while that of SA is lower, leading to a reduced solubility ratio between the two components (Table 1). The phase diagram is quite symmetric and the co-crystal dissolves congruently. Since the solubility of SA is still approximately 12 times higher than that of SMT, the co-crystal region is slightly skewed towards the SA side of the diagram, i.e. towards the more soluble component (Fig. 5a). The co-crystal region is clearly broader than in methanol. Since the co-crystal dissolves congruently the solubility of the co-crystal can be gravimetrically determined (Table 2). The SMT concentration at equilibrium with the co-crystal is slightly higher than the SMT concentration at equilibrium with pure SMT (Fig. 6a), whereas the SA concentration at equilibrium with the co-crystal is lower than the SA concentration at equilibrium with pure SA. The van't Hoff plot of the solubility data for the co-crystal is shown in Fig. 6b, from the slope (−ΔH/R) of which the van't Hoff enthalpy of solution is determined to be +22.2 kJ mol−1 (1:1 complex). The corresponding values for the pure components SMT and SA in acetonitrile are +25.4 and +20.8 kJ mol−1, respectively. The co-crystal value is not too far from the average of the values of the pure components.
Fig. 5 Zoom-in view of the ternary phase diagram of SMT–SA co-crystal system in acetonitrile at 30 °C. Values are in mass fractions. The blue dotted line is the 1:1 stoichiometric line. Regions and various points in the diagram are as same as marked in Fig. 4. The points A1 and A2 (black filled diamonds) represent starting compositions for co-crystal yield and volumetric productivity determination (a), phase diagrams at 10 (red), 20 (light blue) and 30 °C (green) in acetonitrile. The horizontal lines are the solubilities of SMT at 10, 20, 30 °C in acetonitrile. The points (filled circles) depict the invariant points at the three temperatures (b). |
Solvent | T (°C) | Solid phases at equilibrium | Invariant point (mole fraction) | K sp (M2) | S AB (M) | ΔGa (kJ mol−1) | |||
---|---|---|---|---|---|---|---|---|---|
x SMT | x SA | x solvent | HPLC | Gravimetry | |||||
a Gibbs energy of co-crystal formation. | |||||||||
Methanol | 10 | SMT + co-crystal | 0.0015 | 0.0056 | 0.9929 | 5.8 × 10−3 | −6.0 | ||
SA + co-crystal | 0.0014 | 0.0066 | 0.9920 | 5.4 × 10−3 | |||||
20 | SMT + co-crystal | 0.0020 | 0.0072 | 0.9908 | 8.9 × 10−3 | −6.5 | |||
SA + co-crystal | 0.0018 | 0.0080 | 0.9902 | 8.2 × 10−3 | |||||
30 | SMT + co-crystal | 0.0030 | 0.0092 | 0.9878 | 1.7 × 10−2 | −6.8 | |||
SA + co-crystal | 0.0027 | 0.0103 | 0.9870 | 1.6 × 10−2 | |||||
Acetonitrile | 10 | SMT + co-crystal | 0.0022 | 0.0010 | 0.9968 | 7.9 × 10−4 | −5.9 | ||
SA + co-crystal | 0.0004 | 0.0059 | 0.9936 | 1.0 × 10−3 | |||||
Co-crystal | 0.0015 | 0.0015 | 0.9969 | 8.6 × 10−4 | 0.0293 | 0.0304 | |||
20 | SMT + co-crystal | 0.0025 | 0.0012 | 0.9963 | 1.1 × 10−3 | −6.9 | |||
SA + co-crystal | 0.0006 | 0.0064 | 0.9929 | 1.7 × 10−3 | |||||
Co-crystal | 0.0017 | 0.0017 | 0.9966 | 1.5 × 10−3 | 0.0388 | 0.0404 | |||
30 | SMT + co-crystal | 0.0033 | 0.0015 | 0.9952 | 1.8 × 10−3 | −7.2 | |||
SA + co-crystal | 0.0010 | 0.0075 | 0.9914 | 3.0 × 10−3 | |||||
Co-crystal | 0.0026 | 0.0026 | 0.9948 | 2.7 × 10−3 | 0.0524 | 0.0534 | |||
DMSO–methanol (7:3, v/v) | 20 | SMT + co-crystal | 0.0678 | 0.0708 | 0.8613 | 1.59 | −6.8 | ||
SA + co-crystal | 0.0031 | 0.1949 | 0.8019 | 0.23 | |||||
30 | SMT + co-crystal | 0.0696 | 0.0738 | 0.8566 | 1.71 | −6.3 | |||
SA + co-crystal | 0.0042 | 0.2035 | 0.7923 | 0.33 | |||||
40 | SMT + co-crystal | 0.0721 | 0.0758 | 0.8521 | 1.84 | −6.2 | |||
SA + co-crystal | 0.0051 | 0.2143 | 0.7806 | 0.44 |
Fig. 6 The experimental solubility of pure SMT and the co-crystal in acetonitrile (a), van't Hoff plot of the co-crystal in acetonitrile (b). |
A 1:1 co-crystal ‘AB’ equilibrates with the API ‘A’ and the coformer ‘B’ in the saturated solution as per eqn (1). The corresponding equilibrium constant can be expressed in terms of thermodynamic activities. In eqn (2), Ksp refers to the solubility product of the co-crystal, when the activity of the solid co-crystal is taken as unity. Assuming that the contributions from activity coefficients (γ) can be neglected, Ksp can be approximated by the product of concentrations of its co-crystal components, with concentrations in mol L−1. This assumption is valid for ideal solutions, and approximately so for dilute solutions where γA and γB are independent of concentration. The constant Ksp reflects the strength of interactions between the API and coformer in the co-crystal relative to interactions with the solvent in solution.43 The co-crystal intrinsic solubility (SAB) can be estimated from eqn (3) at 10, 20 and 30 °C in acetonitrile (congruent case) using concentrations obtained from HPLC well matching those determined by the gravimetric method (Table 2).
(1) |
Ksp = aA,liqaB,liq = γA[A]γB[B] ≈ [A][B] | (2) |
(3) |
A(s) + B(s) → AB(s) | (4) |
(5) |
Based on the solubility data, the Gibbs energy of formation of the co-crystal from its pure solid components (eqn (4)) can be determined by eqn (5), where and denote the activities of the solute in a solution in equilibrium with the pure co-crystal components respectively. aAliq and aBliq are the activities of the co-crystal components in a solution in equilibrium with pure co-crystal.56 By approximating the activities with the concentrations in mol L−1, the free energy change can be estimated. Using the co-crystal solubility data in acetonitrile (a congruent system), the Gibbs energy of co-crystal formation at 10, 20 and 30 °C is estimated to be −5.7, −7.1 and −7.7 kJ mol−1, respectively. The Gibbs energy of formation has also been estimated using the average Ksp in the three solvent systems; the data is reported in Table 2. The values are all quite close to the values obtained from the co-crystal solubility data. Altogether, the negative value of the Gibbs energy change reveals that the formation of the 1:1 co-crystal from pure solid SMT and SA is a spontaneous process, and that the co-crystal is thermodynamically stable compared to a physical mixture of pure SMT and SA solid phases. With increasing temperature, the free energy change becomes more negative, signifying an increased stability of the co-crystal.
The entropic (eqn (7)) and enthalpic (eqn (8)) components of the Gibbs energy of formation can be determined:
ΔG = ΔH − TΔS | (6) |
(7) |
(8) |
The calculated Gibbs energies are plotted in the appropriate coordinates in Fig. 7, from which estimates of the entropy and the enthalpy of formation are determined from the slopes. The entropy of co-crystal formation is found to be 0.1015 kJ K−1 mol−1, i.e. the co-crystal formation is associated with a positive entropy change. The estimated co-crystal enthalpy of formation is +23.1 kJ mol−1, which agrees closely with the average value of +22.8 kJ mol−1 obtained using eqn (6). Hence, the SMT–SA co-crystal formation from its solid components is shown to be an endothermic processes, i.e. energy needs to be provided to synthesize the co-crystal. Obviously, the conclusion is that the formation of the co-crystal is entirely driven by a favorable entropy increase.
Fig. 7 A plot of ΔG vs. T to find the entropy of SMT–SA co-crystal formation (a), Gibbs–Helmholtz plot to determine the enthalpy of SMT–SA co-crystal formation (b). |
The 7:3 (v/v) mixture of DMSO and methanol was chosen to reach a higher solubility of SMT. The solubility ratio between the two co-crystal components in this solvent is very low (∼2.9), and hence this system is expected to be congruent. However, as shown in Fig. 8a, the system is in fact shown to be incongruent, even though the co-crystal region is very broad and only slightly skewed away from the 1:1 stoichiometric line (Fig. 8a). The effect of temperature on the phase diagram is shown in Fig. 8b.
Fig. 8 Full-scale ternary phase diagram of the SMT–SA co-crystal system in 7:3 (v/v) DMSO–methanol mixture at 30 °C. Values are in mass fractions. The blue dotted line is the 1:1 stoichiometric line. Regions and various points in the diagram areas are same as marked in Fig. 4. The points D1 and D2 (black filled diamonds) represent starting compositions for co-crystal yield and volumetric productivity determination (a), phase diagrams at 20 (red), 30 (light blue) and 40 °C (green) in 7:3 (v/v) DMSO–methanol. The horizontal lines are the solubilities of SMT at 20, 30, 40 °C in 7:3 DMSO–methanol. The points (filled circles) depict the invariant points at the three temperatures (b). |
In accordance with the SA to SMT solubility ratio, the co-crystal shows incongruent dissolution in methanol where the ratio is high and congruent dissolution in acetonitrile where the ratio is low. However, for an even lower solubility ratio in 7:3 (v/v) DMSO–methanol the co-crystal unexpectedly shows incongruent behaviour. The nature of co-crystal dissolution was confirmed by separate co-crystal dissolution experiments. In methanol and DMSO–methanol, the originally pure cocrystal solid phase transformed into a mixture of solid SMT and co-crystal, whereas in acetonitrile, a pure co-crystal solid was maintained. So even if the co-crystal dissolves nearly congruent in 7:3 (v/v) DMSO–methanol it is perfectly clear that the system is incongruent. In addition, the co-crystal region for this system does not change systematically towards the SMT axis compared to the acetonitrile system as would have been expected from the difference in the solubility ratio. Accordingly, it can be concluded that the coformer to API solubility ratio for the SA–SMT system cannot be safely used as a guide to the nature of the co-crystal dissolution behaviour. However, quite clearly the solvent has a major influence on the nature of co-crystal dissolution and the overall appearance of the phase diagram.
With increasing temperature the solubility of all the solid phases increase, which leads to a shift of the various solid state regions down towards the solid SMT–SA axis in the ternary phase diagram. This leads to a larger region for the solution phase region (see Fig. S9–S11 in the ESI†). Temperature changes did not bring about any remarkable changes on the overall appearance of the phase diagram.
The experimentally determined invariant points for the three solvents in terms of mole fractions are given in Table 2, together with the corresponding Ksp values obtained from eqn (3). Obviously, the value depends on the solvent. The order in which Ksp varies is DMSO–methanol > methanol > acetonitrile. In each solvent, the Ksp value and the corresponding co-crystal solubility increase with temperature. In methanol and acetonitrile, there is just a small difference in the Ksp values obtained from the two invariant points. However, in the DMSO–methanol mixture, the Ksp difference is much higher, most likely because at higher concentrations the error associated with neglecting the activity coefficients becomes larger. For the congruent acetonitrile case, the Ksp for the co-crystal is between the Ksp values obtained from the two invariant points.
The width of the region where the co-crystal is the most stable phase can be measured as the linear distance between the two invariant points as per eqn (9).
(9) |
Based on this, the width of the co-crystal region decreases in the order DMSO–methanol > acetonitrile > methanol, and is inversely proportional to the solubility ratio between the two co-crystal components (SA/SMT, Table 1) i.e. the smaller the solubility ratio, the wider the co-crystal region (Fig. 9).
The co-crystal yield and productivity results are given in Table 3. The starting overall compositions of pure solid SMT, SA and solvent in the experiments are marked in Fig. 4, 5 and 8 by (M1, M2), (A1, A2) and (D1, D2), respectively. The farther away from the solid–liquid equilibrium line this initial overall composition point is placed, the greater is the surplus of material that can transform into solid co-crystal, and hence the higher the potential co-crystal yield and productivity. The difference in yield for the different solvents primarily depends on the co-crystal solubility. A high solubility leaves a greater amount dissolved in the solution at the end of the process. The yield can be improved by adding more pure solid components at a stoichiometric ratio. The very narrow co-crystal region in methanol requires a high precision in dosing the components. In spite of a large co-crystal region in DMSO–methanol, a limiting factor for DMSO–methanol is the high boiling point, which makes it difficult to completely remove the toxic solvent. Irrespective of the system being congruent or incongruent, the pure co-crystal can be synthesized by slurry co-crystallization as long as the liquid composition starting point is along the curve where the co-crystal is in equilibrium with the solution. It may be noted that it is favourable from a yield point of view if the phase diagram is skewed towards the coformer axis, since this corresponds to a lower concentration of SMT in the solution.
Pointa | Input reagents (mass fraction) | Co-crystal mass (g) | Co-crystal yield (g g−1) | Volumetric productivity (g mL−1) | ||
---|---|---|---|---|---|---|
SMT | SA | Solvent | ||||
a M (methanol), A (acetonitrile), D (DMSO–methanol). | ||||||
M1 | 0.0700 | 0.0600 | 0.8700 | 0.1423 | 0.55 | 0.06 |
M2 | 0.1530 | 0.0980 | 0.7490 | 0.4574 | 0.91 | 0.21 |
A1 | 0.0620 | 0.0380 | 0.9000 | 0.1460 | 0.73 | 0.06 |
A2 | 0.1968 | 0.1032 | 0.7000 | 0.5819 | 0.96 | 0.26 |
D1 | 0.2100 | 0.2430 | 0.5470 | 0.0910 | 0.10 | 0.05 |
D2 | 0.3540 | 0.2780 | 0.3680 | 0.4857 | 0.38 | 0.29 |
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c9ce00124g |
This journal is © The Royal Society of Chemistry 2019 |