Kangli
Li‡
a,
Gabin
Gbabode
a,
Marine
Vergé-Depré
b,
Benoit
Robert
b,
Maria
Barrio
c,
Jean-Paul
Itié
d,
Josep-Lluis
Tamarit
c and
Ivo B.
Rietveld
*ae
aLaboratoire SMS-EA3233, UFR des Sciences et Techniques, Universite de Rouen Normandie, Place Emile Blondel, 76821 Mont-Saint-Aignan, France. E-mail: ivo.rietveld@univ-rouen.fr
bSanofi R&D, Global Chemical Manufacturing and Controls/Synthetics Platform/Early Development Department, 13 quai Jules Guesde, F-94400 Vitry sur Seine, France
cGrup de Caracterització de Materials, Departament de Física and Barcelona Research Center in Multiscale Science and Engineering, Universitat Politècnica de Catalunya, EEBE, Campus Diagonal-Besòs, Av. Eduard Maristany 10-14, E-08019 Barcelona, Catalonia, Spain
dBeamline Psiché, Synchrotron SOLEIL – L'Orme des Merisiers Saint-Aubin – BP 48 91192 Gif-sur-Yvette, France
eFaculté de Pharmacie, Université Paris Cité, 4 Avenue de l'Observatoire, 75006 Paris, France
First published on 31st May 2022
The phase behaviour of drug molecules is important for the control over the desired polymorph in drug formulations, whether it is to ensure better stability or better solubility. In the case of pyrazinamide, a drug against tuberculosis, stability studies have been complicated due to the very slow transition kinetics observed in DSC measurements. Using vapour pressure measurements, in which the reluctance of phase transformation is in fact an advantage, all solid–solid phase transformation temperatures have been determined. This method has been key to map the phase behaviour of pyrazinamide. The use of high-pressure measurements with synchrotron X-ray diffraction has allowed the construction of the pressure–temperature phase diagram of the four solid phases of pyrazinamide and the liquid phase. The α form was found to be the stable form at room temperature. One striking feature of pyrazinamide is that one polymorph, the δ form, has a very large thermal expansion and extreme compressibility not found in the other three forms. This gives rise to curved solid–solid transition equilibria in the pressure–temperature phase diagram, which is not commonly observed in the pressure range of 0 to 1 GPa. Using the phase diagram, polymorph β could be obtained in its stable temperature domain.
The solid-state behaviour of pyrazinamide (PZA) has received a lot of attention, because of its four known solid phases, α, β, γ, and δ, and the fact that it has not been clear whether form α or form δ is the most stable form at room temperature.2–12 Moreover, the phase transitions between these phases exhibit large hysteresis, making it difficult to interpret the thermodynamics behind the transition behaviour.4,9,10,13 A consensus exists in the literature that form γ is the high-temperature form and that form α is stable at lower temperatures, probably down to room temperature, while form δ is most likely stable below room temperature.2,9,10 It is not clear whether the β form possesses a stable domain, but it has not been considered likely.9,10 Recently, the melting point of form α has been reported and the equilibrium temperature between α and γ has been established as low as 392 K (Table 1),2 however, except for the melting temperature of γ at 462 K, which has been known for some time,9,10 none of the other phase equilibrium temperatures between the four solid phases or between the solid phases and the liquid have been reported.
Property | Value | Unit |
---|---|---|
a The data from this table originates from ref. 2. b Estimated error, nonetheless dP/dTα→L must be larger than dP/dTγ→L (see ref. 2). | ||
T γ→L | 462.0 ± 0.5 | K |
ΔHγ→L | 240.6 ± 1.0 | J g−1 |
dP/dTγ→L | 3.5 ± 0.3 | MPa K−1 |
T α→L | 457 ± 1 | K |
ΔHα→L | 253 ± 4 | J g−1 |
dP/dTα→L | 4.0 ± 1b | MPa K−1 |
T α→γ | 392 ± 1 | K |
ΔHα→γ | 13.2 ± 0.6 | J g−1 |
dP/dTα→γ | −2.7 ± 0.9 | MPa K−1 |
Considering the large hysteresis in transition temperatures for pyrazinamide samples studied by DSC (differential scanning calorimetry),2,9,10 the high sublimation pressure of pyrazinamide14–17 can be used to obtain information about which phase is the most stable at a given temperature, provided that the solid phase does not transform into another phase. The latter condition appears to be met for pyrazinamide because of the large hysteresis in the system, so this disadvantage in the DSC becomes an advantage for the sublimation pressure measurements.
In the current paper, the phase behaviour of all four phases will be studied and a pressure–temperature phase diagram will be proposed involving all equilibria between all known phases in the system: α, β, γ, δ, the liquid, and the vapour phase. The analysis is based on vapour pressure measurements, but also uses studies of the thermal expansion of the different phases and of compression under pressure obtained from laboratory and synchrotron X-ray diffraction, respectively. In Table 1, previously published data on the phase behaviour of forms α and γ in relation to each other and the liquid can be found.2
The vapor (V) pressures of forms α and γ reported previously are mentioned here as they will be used in the phase diagram calculations:2
α → V![]() ![]() ![]() ![]() | (1) |
γ → V![]() ![]() ![]() ![]() | (2) |
Similarly, the specific volumes of forms α and γ have been published previously leading to the following equations:2
vα = 0.6624(7) + 6.8(6) × 10−5 T + 8.44(1.06) × 10−8 T2 | (3) |
vγ = 0.6520(7) + 4.8(5) × 10−5 T + 1.22(9) × 10−7 T2 | (4) |
β → V![]() ![]() ![]() ![]() | (5) |
δ → V![]() ![]() ![]() ![]() | (6) |
The δ to α transition is harder to interpret (Fig. S1c and d†), because several transitions may occur at the same time and therefore the identity of the final state of the sample cannot be established with certainty. In the DSC measurements, an enthalpy difference, ΔHδ→α, of 5.7 ± 0.4 J g−1 is found, but the low error is more a sign of the few data points than of a high precision. Taking the data from Table 1 on the enthalpy of the α to γ transition, ΔHα→γ, of 13.2 J g−1 and the difference between δ and γ discussed above, the enthalpy difference between α and δ should be between 3.3 and 4.8 J g−1. Thus, the value of 5.7 J g−1 obtained from integrating the observed DSC peak is most likely too high, probably due to partial transformation from δ into γ.
The DSC observations of form β (Fig. S1e†) are possibly even more complicated as demonstrated by the X-ray diffraction data of β as a function of the temperature in Fig. S1b in the ESI,† where both α and δ are observed on heating β. In the DSC curves, often a double peak is observed with an onset at around 370 K. The second peak cannot always be separated and has an onset at about 390 K, when separation is possible. Considering that melting of these β samples invariably occurs at the melting point of form γ (see Table S3†), it can again be surmised that, neglecting the heat capacity, all solid–solid transformations from the initial β form up to the melting point together must reflect the enthalpy difference between β and γ, ΔHβ→γ. Taking the average of the total enthalpy for the solid–solid transitions, 17.9 ± 2.3 J g−1 is found. Taking only the enthalpies obtained using single crystals of β, as the phase purity of the powder cannot be guaranteed, leads to 19.1 ± 2.6 J g−1. Because the double peak observed for the solid–solid transitions involving β cannot be properly separated, the enthalpy difference between α and γ of 13.2 J g−1 can be used leading to an enthalpy difference between α and β, ΔHβ→α, of 4.7–5.9 J g−1. This also suggests that the enthalpy difference between β and δ, ΔHβ→δ, is in the order of −0.1 to 2.6 J g−1 with β in all likelihood possessing a higher enthalpy content than form δ.
vβ = 0.6537(6) + 1.074(23) × 10−4 T, | (7) |
vδ = 0.6519(9) + 5.6(8) × 10−5 T + 1.44(16) × 10−7 T2. | (8) |
![]() | ||
Fig. 1 The specific volumes of forms α (red circles), β (green triangles), γ (black squares), and δ (blue diamonds) as a function of temperature under ordinary pressure. |
As can be seen in Fig. 1, where the specific volumes of all four solid phases are presented, form γ has the smallest specific volume and thus the highest density, form α has the largest specific volume and δ has the largest thermal expansion (see also Table S15†).
![]() | ||
Fig. 2 The specific volumes of forms α (red circles), β (green triangles), γ (black squares), and δ (blue diamonds) as a function of pressure at room temperature. |
Polymorph β possesses a P21/c space group with four molecules per unit cell (Fig. 4) and one molecule in the asymmetric unit. The four molecules form four “classic” dimers through two strong N–H⋯O hydrogen bonds with molecules outside the unit cell graph set R22(8). However, a C11(4) chain along the c-axis of slightly longer hydrogen bonds with the same groups links these dimers together. Along the c-axis a corrugated chain of dimers exists; however, due to the angles of the dimers these corrugated chains are interconnected through their C11(4) interactions in the direction of the b-axis forming a dimer-wide slab along the bc plane with the thickness of the a-axis. In the direction of the a-axis, these slabs are interconnected through weaker hydrogen bonds C–H⋯N in the aromatic ring. The maximum positive and negative expansion directions and compression direction are indicated in Fig. 4. The maximum thermal expansion and compression possess different directions, however, overall, the thermal expansion and the compression tensors remain very similar as can be seen in Fig. S10 and S15,† respectively. The positive thermal expansion is linked to the complex way that the corrugated bands in the direction of the c-axis are interconnected in the direction of the b-axis, apparently providing sufficient room for expansion along the b-axis. The compression (Fig. 4b) may be related to a mix between a displacement of the complex hydrogen network in the slabs and the weaker interactions between the slabs, while the negative thermal expansion is most likely a slight contraction due to the expansion along the b-axis and an expansion along the inverted compression direction −
.
![]() | ||
Fig. 4 Unit cell of form β with (a) the maximum thermal expansion (red arrow) and (b) the compression (blue arrow) and the negative thermal expansion (NTE, black arrow) directions. Many other hydrogen bonds are present that do not link the molecules within the unit cell; a figure with an overview can be found in the ESI† Fig. S17. |
Polymorph γ crystallizes in the non-centrosymmetric space group Pc with two molecules in the unit cell and one in the asymmetric unit. The molecules form linear chains through strong N–H⋯N hydrogen bonds with graph set C11(6). The two molecules in the unit cell are connected by a slightly longer hydrogen bond between N–H⋯O of their amide moieties. In addition, they possess a weak hydrogen bond C–H⋯N between their respective pyrazine rings. The planes through the rings of the two molecules in the unit cell form an angle between each other of 47.6°. Due to this angle the molecules form corrugated sheets parallel to the ac plane of the unit cell. These different sheets are interconnected through multiple weak hydrogen bonds involving the C–H and N from the pyrazine rings and the O of the amide moiety. The maximum expansion and compression directions are along the b axis as shown in Fig. 5, which is the overall direction of the complex network of multiple weaker hydrogen bond interactions involving the C–H in the rings.
![]() | ||
Fig. 5 Unit cell of form γ and the maximum thermal expansion and compression directions (purple arrow). |
Polymorph δ crystallizes in the triclinic system with the space group P with two molecules per unit cell and one in the asymmetric unit. A cyclic dimer, graph set R22(8), exists in form δ through the strong hydrogen bonds N–H⋯O of the amide moiety. Slightly longer hydrogen bonds exist between the oxygen and N–H of another pyrazinamide molecule, creating quite flat 2D bands of dimers as shown from the side in Fig. 6a. These 2D bands are reinforced by weaker hydrogen bonds between the C–H and the N in the respective pyrazine rings. The bands are parallel to the b-axis and are approximately found on the (101) plane. These slabs of dimers that are bound through strong hydrogen bonds form extended sheets through weaker hydrogen bonds, C–H⋯N, in the pyrazine rings of different slabs that lie in the same plane. These sheets are extremely flat. Short contacts involving mainly aromatic interactions (pi–pi interactions) keep the different sheets together in the
+
direction. The maximum expansion and compression directions are roughly perpendicular to the slabs/layers as indicated in Fig. 6b. As can be seen in Fig. S10d and S15d and also in Tables S15 and S19,† these layers are in fact extremely sensitive to compression and expansion.
![]() | ||
Fig. 6 (a) Unit cell of form δ and (b) the maximum thermal expansion and compression directions (purple arrow). |
LT↓\HT→ | δ | α | γ | L |
---|---|---|---|---|
a The phase on the left is the low temperature phase (LT), while the phase on top is the high temperature (HT) phase for each equilibrium. b Stable equilibria are indicated in italic bold. | ||||
β | 254; 2 × 10 −6 ; 1 | 290; 8 × 10−4; 5 | 314; 2 × 10−2; 18 | 414; 275; 260 |
δ | — | 299; 3 × 10 −3 ; 3 | 326; 8 × 10−2; 17 | 423; 403; 258 |
α | — | 399; 41; 13.2 | 456; 1459; 253 | |
γ | — | 462.0; 1816; 240.6 |
Because the sublimation curve for the γ form is known (eqn (2)) and the melting quantities of γ are known too (Table 1), an estimate for the vapour pressure of the liquid phase can be obtained:
L → V![]() ![]() ![]() ![]() | (9) |
The results obtained with the vapour pressure are not perfect, as the intersection temperatures suffer a high uncertainty due to the fact that the vapour pressure equations have similar slopes; a slight shift of those lines can have large effects on the calculated transition temperature. Therefore, temperatures listed in Table 2 may not fully coincide with the equilibrium temperatures, but they are certainly the best estimates, as long as better data are not available. Preference will be given to directly measured quantities if available.
With Table 2, the temperature domains for the stability of the different forms can be determined. It is known that the γ form has the highest melting temperature at 462 K and thus γ is the high temperature polymorph. The highest equilibrium temperature involving γ and another solid is that with α at 399 K (although direct experiments indicate that this temperature is 392 K) (see also Table 1).2 The subsequent highest equilibrium temperature involving α is that with δ at 299 K and thus below this temperature δ is the most stable form. Moreover, the β–δ equilibrium temperature can be observed at 254 K, indicating that below this temperature β is the most stable form.
In particular the β–δ and the δ–α equilibrium temperatures will need to be considered with care. Unfortunately, in the lowest temperature region where vapour pressure data of α and δ have been obtained (Table S1†), α is still the most stable form, but it is also clear that the vapour pressure values are very close around room temperature and that it is therefore very likely that the equilibrium temperature between α and δ is located between 0 and 30 °C; their vapor pressures and thus their Gibbs free energies are virtually the same in this temperature domain.
The equilibrium temperature between β and δ contains considerably more uncertainty due to the necessary extrapolations of the vapour pressure expressions to obtain this temperature, however, the fact that the vapour pressure of δ approaches that of β indicates that the β form appears to possess a stable domain at low temperature, something that had not been expected previously.9,10 Moreover, the proximity of the respective transitions of δ and β with the other phases, not more than 10 degrees apart, indicates that their Gibbs free energies are not far apart. To verify the stability of β at low temperatures, crystallisation experiments in acetonitrile have been carried out at 252 K, which is below the predicted phase equilibrium temperature between β and δ. The β form crystallised with possibly a small δ impurity (peak at 16.415° 2θ, see ESI† Fig. S16).
The fact that all polymorphs of pyrazinamide possess a stable temperature domain, as can be concluded from the vapor pressure data, is corroborated by the enthalpies that have been obtained by DSC (Table 2). While the enthalpy between β and δ is about +1 J g−1 and for δ to α about +3 J g−1, the enthalpy difference between α and γ was measured to be +13.2 J g−1 and the melting of γ involves +240.6 J g−1. This series clearly indicates that all transitions with increasing temperature are endothermic, which is necessary for fully enantiotropic systems. The enthalpy data in Table 2 are rounded values based on the averages discussed in section 3.2.
An energy plot as a function of the temperature based on the results discussed above can be found in Fig. 7.
![]() | (10) |
HT→\LT↓ | δ | α | γ | L |
---|---|---|---|---|
a The phase on the left is the low temperature phase (LT), while the phase on top is the high temperature (HT) phase for each equilibrium. b Stable equilibria are indicated in italic bold. | ||||
β | 0.0062; −0.0054; −1.15 | 0.0233; 0.0044; 5.26 | 0.0570; −0.0082; −6.94 | 0.578; 0.097; 5.94 |
δ | — | 0.0171; 0.0087; 1.97 | 0.0508; −0.0049; −10.3 | 0.572; 0.104; 5.50 |
α | — | 0.0337; −0.0124; −2.72 | 0.554; 0.133; 4.17 | |
γ | — | 0.521; 0.148; 3.53 |
The full consequences in relation to phase theory are not entirely clear. The change in the sign means that for a given pressure and temperature Δv equals zero, indicated by the orange dash-dot line in Fig. 8d. Below this line γ possesses the highest density, and above this line, δ possesses the highest density. It appears therefore that δ is the ultimate high-pressure form in the system. Nonetheless, the orange dash-dot line is not the phase equilibrium, which will be lying higher as next to the volume difference, also the enthalpy difference is of importance. At the point where the δ–γ phase equilibrium intersects the orange dash-dot line, the slope of the equilibrium is expected to be infinity (see eqn (10)) and then bend back to a positive finite slope.
In the case of gases and liquids, Δv = 0 implies that both phases are the same and therefore that also ΔS = 0. This in turn implies a critical point, as the two phases have become indistinguishable. In the present case, there is no trace of a phase transition, neither in the δ form nor in the γ form. Both phases have a clear own identity based on different space groups and unit-cell dimensions. It can therefore not be stated that the two phases are one and the same and due to their different molecular configurations, the energetic and entropic states must be different too. This would imply that for these two intersecting densities in the solid state, with two distinct crystallographic phases, no critical point exists.
As the situation is caused by the extreme compressibility of the δ form, the effect on its Gibbs energy surface may have the form of a ripple perpendicular to the pressure axis, which in the current situation with four phases, happens to intersect with the Gibbs energy surface of the γ form, causing the extremely curved phase equilibrium. At the same time, the compressibility of δ must also have an effect on the equilibrium curves of δ–α and β–δ. δ has already a smaller volume with respect to α and thus Δδ→αv > 0, which remains the case when δ is compressed. Due to the faster compression of δ over α however, Δδ→αv will increase and thus the slope of the equilibrium curve will be more curved and decrease, lowering the position of the triple point α–δ–L. A similar effect is expected for the β–δ equilibrium for which the inequality Δβ→δv < 0 is expected to become more negative and thus the equilibrium will be curved towards lower negative slopes. It should be clear however that in both cases the sign of the volume inequality remains the same, so the expected behaviour, obtained by extrapolation, remains the same, from a topological point of view.
In fact, it is likely that the α–δ phase equilibrium is found at about 0 °C. Considering Fig. 9 with the calculated α–δ phase equilibrium (Tables 2 and 3) leads to an intersection with the α–γ phase equilibrium (blue line) somewhat above 350 K. However, taking into account that this is a supposed α–δ–γ triple point, the δ–γ phase equilibrium must intersect here too, whereas the latter equilibrium (red line) actually has a negative slope at 0 MPa and does not approach the intersection between α–δ and α–γ. Nonetheless, the observation that the δ–γ phase equilibrium must be curved solves part of this problem, bringing its intersection with the α–γ phase equilibrium closer to the α–δ and α–γ intersection. Using the uncertainty in the α–δ equilibrium temperature and shifting this equilibrium down to 0 °C allows for a consistent phase diagram of the stable phases with a properly defined α–δ–γ triple point at around 340 K and 150 MPa (Fig. 9). In this adjustment only the α–δ transition temperature at 0 MPa was changed, but other adjustments to the slopes of the α–δ and δ–γ phase equilibria may also be necessary. The slope of the α–γ equilibrium was obtained by measurement and its uncertainty is considered smaller than that of the other two equilibria.
Another result that can be derived from the current phase diagram is that the observation by Tan et al.8 that form β is obtained by putting form γ under pressure is probably a misinterpretation of their Raman data, as the γ form gets stabilized with respect to β under pressure. If anything, transformation of γ into δ is much more likely under pressure, if γ changes at all.
Comparing the volumetric thermal expansion of the polymorphs, it can be observed that form δ possesses a larger thermal expansion than the other polymorphs. This result is most likely due to its layered structure in accordance with observations by Saraswatula et al.22 Form δ also exhibits the largest compression most likely due to the weakly interacting layers. Interestingly, forms α and β, which both exhibit negative thermal expansion to some level, are only stable at low pressures.
The most stable form at room temperature appears to be form α on a par with form δ. At room temperature both solids must have the same solubility, as their vapour pressure is virtually the same. Preference should probably be given to developing α as it will be the most stable form at temperatures above room temperature, where the risk of phase transformation is larger than at temperatures below room temperature, where δ will be more stable. Considering the studies carried out under pressure, tabletting will not affect a phase change of the formulated form, as no phase transformation has been observed under pressure.
The extreme curvature in the δ–γ equilibrium curve (see curved partially dashed red line in Fig. 9) leading to a minimum in the temperature for the curve represents a challenge for the topological approach to establish phase diagrams using only ordinary pressure data and the Clapeyron equation (eqn (10)). From a large number of experiments, it is clear that most solid–solid equilibria are straight lines,23,24 so the current finding appears to be a rare case. In addition, it is already clear from the ordinary pressure measurements that the thermal expansion of δ was much larger than that of the other forms, which may therefore function as an indication that also under the influence of pressure large changes can occur.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d2ce00484d |
‡ Current address: Zhejiang Shaoxing Institute of Tianjin University, 312![]() |
This journal is © The Royal Society of Chemistry 2022 |