Switching Polymorph Stabilities with Impurities: A Thermodynamic Route to Benzamide Form III

We investigate the polymorphic behavior of benzamide, the first compound known to exhibit polymorphism, in the presence of small amounts of nicotinamide in the crystallization environment. A previous study by Emmerling et al. 1 showed that the presence of nicotinamide promotes the transformation of the thermodynamic polymorph I of benzamide into its metastable polymorph III via mechanochemistry. We show that this transformation is the result of a thermodynamic switch between these two polymorphic forms driven by the formation of solid solutions with a small amount of nicotinamide. The presence of nicotinamide in the crystallization environment promotes the robust and exclusive crystallization of the elusive form III. These results represent a promising route to the synthesis and utilization of elusive polymorphs of pharmaceutical interest. ABSTRACT: We investigate the polymorphic behavior of benzamide, the first compound known to exhibit polymorphism, in the presence of small amounts of nicotinamide in the crystallization environment. A previous study by Emmerling et al. 1 showed that the presence of nicotinamide promotes the transformation of the thermodynamic polymorph I of benzamide into its metastable polymorph III via mechanochemistry. We show that this transformation is the result of a thermodynamic switch between these two polymorphic forms driven by the formation of solid solutions with a small amount of nicotinamide. The presence of nicotinamide in the crystallization environment promotes the robust and exclusive crystallization of the elusive form III. These results represent a promising route to the synthesis and utilization of elusive polymorphs of pharmaceutical interest. Abstract: We investigate the polymorphic behaviour of benzamide, the first compound known to exhibit polymorphism, in the presence of small amounts of nicotinamide in the crystallization environment. A previous study by Emmerling et al. 1 showed that the presence of nicotinamide promotes the transformation of the thermodynamic polymorph I of benzamide into its metastable polymorph III via mechanochemistry. We show that this transformation is the result of a thermodynamic switch between these two polymorphic forms driven by the formation of solid solutions with a small amount of nicotinamide. The presence of nicotinamide in the crystallization environment promotes the robust and exclusive crystallization of the elusive form III. These results represent a promising route to the synthesis and utilization of elusive polymorphs of pharmaceutical interest.

Molecular crystals are essential constituents of everyday consumer products such as paints, foods or medicines. The quality, safety and efficacy of such products depend greatly on the properties and structure of the crystals exhibited by their active ingredients. Since most compounds can and will crystallize in various forms (known as polymorphs), 2 discovering, understanding and controlling polymorphism is a necessity in product development. In the world of pharmaceuticals, over 50% of compounds exhibit polymorphism. 3 Whilst some polymorphs are easy to discover and consistently crystallized, others remain elusive and only appear (or disappear) by chance after decades of research on them. 4,5 The quest for new drugs requires, amongst other things, the synthesis, crystallization and characterization of new chemical entities in drug discovery and of the scale up and crystallization process design in drug development. Along the journey towards a final drug product, it is common to observe the appearance of metastable polymorphs first which eventually -transform to other more stable polymorphs. This solid form evolution has often been linked to an increase in chemical purity. When chemical purity is low, compounds can be hard to crystallize or do so in unwanted forms. As the synthetic routes are improved and the compounds are subjected to crystallization development, their purity increases which may lead to new crystal forms. 6 Whilst crystallization is a very selective process (thus widely used as a purification technique), sometimes the tiniest amount of impurities can have a significant impact on the outcome. This is especially true when the impurity has a close molecular similarity with the compound under development. Impurities are well-known to modify crystal habit 7,8 impact the kinetics of crystallization 9 and/or inhibit nucleation, crystal growth, and conversion kinetics between polymorphs 10 thus promoting or preventing the observation of a particular crystal form. The incorporation of impurities in crystal forms is also known to impact crystal properties such as elasticity or hardness. 11 Two notorious examples of drugs changing polymorphic forms in the presence of impurities are aspirin and paracetamol. Despite the fact that aspirin has been crystallized in thousands of tones since the 19th century, its second form of was only discovered in 2005 12 when it was crystallized in the presence of impurities (form II can now be consistently produced in the presence of aspirin anhydride). [13][14][15] Paracetamol, similarly, can only be crystallized as its form II in the presence of various impurities one of which being metacetamol. 16 Despite the impact that impurities have on polymorphism and crystallization, there are various molecular mechanisms by which they can act 6 some of which are not well understood. Pioneering early work by the Weizmann group showed how through preferential adsorption on specific crystal faces, impurities are able to modify crystal habits but also inhibit nucleation and allow for polymorph control. 17,18 More recent work by Liu et al. 19 has shown how adsorption of metacetamol on form I surfaces in the wrong orientation inhibits its nucleation and growth thus allowing form II to crystallize. In that regard, previous work has mostly focused on using impurities to block the nucleation and growth of stable polymorphs in order to allow for the metastable polymorph to appear. Here we explore a different molecular route of action: rather than blocking the nucleation and growth of the stable form, we explore how other impurities may be able stabilize the metastable polymorph through the formation of solid solutions. Understanding whether or not such routes are possible, and the molecular modes of action would allow us to achieve greater control of polymorphism during crystallization. The polymorphism of benzamide (BZM, Figure 1), the first molecular compound ever reported to be polymorphic, was noted by Wöhler and Liebig in 1832 20 only to be forgotten for over 170 years. Studies on form I (the stable polymorph) abounded and its structure was first reported in 1959. 21 Because of the elusive nature and metastability of the other polymorphs (II and III) and the impossibility to grow large single crystals of good quality suitable for single X-ray diffraction, the solution of the structure of form II and form III had to wait until the advancement of synchrotron radiation 22,23 and crystal structure solution from powder 24 methods. More recently, the structure of a metastable and disordered form IV has been proposed. 25 The targeted, robust and exclusive growth of crystals of form III (the form of interest in this contribution) still remains a challenge since this form grows poorly and always concomitantly with form I. 26 In a recent study by Emmerling and co-workers 1 , a failed attempt to produce 1:1 cocrystals of BZM with its close relative nicotinamide (NCM, Figure 1) by ball-mill grinding led to the observation of form III BZM. The solid state BZM form I to form III conversion was consistently observed in a range of BZM:NCM concentrations. The authors mentioned in their conclusions that "it was assumed the nicotinamide molecules were included in the crystal system of benzamide, triggering formation of benzamide III". If, rather than staying at the crystal surfaces blocking growth, impurities are incorporated in lattice positions within crystal structures, they form what it is known as solid solutions. 27 Inspired by these results, we ventured to investigate this historically important compound further in order to shed light on the driving force of the BZM form I to form III conversion.
First, we performed ball-mill liquid assisted grinding (LAG) with ethanol of pure BZM and BZM in the presence of NCM in a range of compositions. LAG of pure BZM affords form I whilst LAG of BZM in the presence of NCM (5-30 mol%) results in BZM form III (ESI S2.3.1), consistent with Emmerling's work. At concentrations of NCM above 30 mol%, diffraction peaks corresponding to solid NCM start appearing suggesting that the solid solubility of NCM in BZM form III lies around that value. When ethanol is replaced with isopropanol, the LAG results do not change. Interestingly, ball mill neat grinding (NG) of pure BZM resulted in pure BZM form III without the need of any NCM at all (ESI S2.3.2). The observation of form III upon NG of pure BZM may be explained as a crystal size effect. 28 Why small amounts of NCM also induce a BZM form III conversion is less clear since it could potentially be due to a combined size reduction and impurity effect.
In order to explore the possibility of solid-solution formation and its impact on the stability of forms I and III of BZM, we performed lattice energy calculations (PBE-d) for both crystal polymorphs as a function of NCM concentration ( Figure 2). For pure BZM, form I is computed to be more stable than form III by only 0.2 kJ/mol. The small energy difference between the forms is consistent with the fact that forms I and III often crystallize concomitantly, with form I known to be the most stable. As the concentration of NCM in the lattice increases, the relative stability of the two polymorphs changes and above 10 mol%, BZM form III becomes more stable than BZM form I. The changes in the lattice energies are, however, very subtle. We note that, since the lattice energies of forms I and III are so close, no computational method available would be able to provide energy differences of such accuracies (within 0.2 kJ/mol). To confirm the revelations of the above simulations, the predicted forms stability changes were assessed experimentally by means of solvent-mediated phase transformation experiments (slurries). 29 If slurry experiments are given enough time, the lower solubility of the more stable form at each given composition would eventually lead the system to thermodynamic equilibrium and thus the more stable form at such composition. Excess solid of BZM form I was stirred in saturated isopropanol (IPA) solutions at 25°C in the presence of various amounts of NCM (ESI S2.4.2). 2.5 g of mixed BZM form I with NCM form I solids (at 2, 4, 6, 10 and 20mol% content of NCM) were added to 5 g of IPA and stirred for one week at which point the solids were characterized. Figure 3 shows the PXRD results for the excess solids of such slurries. Remarkably, the slurry experiments lead to the switch in forms from form I to form III in the presence of NCM. Under these experimental conditions, a total of 4mol% NCM was required to observe a form I to form III switch in the slurries. When the slurry experiments were repeated in ethanol, the form III conversion was also observed but at slightly higher total concentrations of NCM (7 mol%). These experiments were also performed at higher temperatures, 45 ˚C, with similar results obtained (ESI S2.4.1). The amount of impurity present in the slurry had an important effect on the kinetics of the conversion with kinetics accelerating at higher total concentrations of NCM. For example, slurries from ethanol show that when the concentration of NCM is approximately 10 mol%, the conversion to BZM form III starts after 30 min and completes within 4 h whilst when the NCM concentration is approximately 20 mol% the conversion to form III completes in just 15 minutes (ESI S2.4.1). In all cases, the resulting BZM form III was confirmed to be a solid solution with NCM -as suggested by the absence of diffraction peaks of pure NCM forms, shifts in PXRD peaks and changes in the DSC thermographs. Increasing the concentration of NCM resulted in a linear depression of the melting point of BZM form III from 125.8 °C (10 mol % NCM) to about 121.2 °C (30 mol% NCM) (ESI S2.4.1). Crash cooling experiments (to 25°C, 1.2 supersaturations) of BZM in the presence of NCM (approximately 10, 20 and 30 mol%) were also performed in isopropanol. Consistently with the LAG and the slurry experiments, BZM form III was also obtained via crash cooling in the presence of NCM at all concentrations studied (ESI S2.5).
In the presence of NCM, thus, we are able to consistently and exclusively obtain well-formed form III BZM crystals without the presence of form I crystals, something never achieved before from solution crystallization. This can be achieved both via solvent-mediated phase conversion of form I into form III (slurries) or via crash cooling from solution in the presence of NCM ( Figure 4). We notice that the morphologies of our form III are more equant and similar to those of pure form I and very different to the needles of pure form III BZM. 24  Until now, we have referred to the amount of NCM content (mol%) with respect to the total amount of solids (BZM+NCM) used in the experiments. In order to estimate the NCM threshold which results in the thermodynamic lattice energy switch (Figure 2), we require to characterize the amount of NCM incorporated in forms I and III BZM crystals. This was quantified by retrieving the solids from the slurry experiments in IPA and characterising them with 1H-NMR. The fraction of the NCM incorporated in the crystals over the total NCM fraction in the experiment (in mol%) was calculated to be 0.7 and 0.8 for BZM forms I and III respectively (see ESI). Thus, NCM incorporates with very equal efficiency in both lattices. We observe that at around 3mol% of NCM incorporation in the BZM crystals, form I starts converting to form III, thus the concentration threshold for the stability switch lies around that value. The discrepancy observed between the calculated 10 mol% and the experimental 3 mol% limits can understandably be attributed to the difficulty of simulating energy changes that are so subtle.
The comparison of the crystal packing of BZM forms I and III reveals a high similarity, with both structures featuring a common 2-D arrangement of centrosymmetric hydrogen-bonded dimers propagating along the a and b crystallographic directions and connected via a second set of N-H…O hydrogen bonds and other non-directional contacts (ESI S2.2). The only difference between structures occurs along the c direction, where the common 2-D arrangements are packed differently and connected via different sets of weaker contacts. The stability switch of the forms in the presence of NCM is driven by two concomitant but opposite effects: i) small destabilization experienced by BZM form I and ii) the increase of stability of BZM form III upon incorporation of small amounts of NCM. Given the high structural similarity between the two polymorphs, the incorporation of NCM in the crystal lattice is only subtly impacting weak and non-directional contacts. In this case, the "switch" in stability takes at very small impurity incorporations (3 mol%) because of the close relative lattice energy of the two pure polymorphs.
In conclusion, we have demonstrated that the relative stability of polymorphs can be inverted by using selective impurities able to form solid solutions. This has been shown for benzamide forms I and III in the presence of nicotinamide. Lattice energy calculations performed on both crystal polymorphs as a function of NCM content suggested a change in the relative stability of the two polymorphs as the concentration increases. This observation was confirmed experimentally from results of solvent mediated phase transformations that show full conversion from BNZ form I to BNZ form III starting from concentrations of NCM in the solid lattices of just 3 mol%. The concentration of NCM in the solution was shown to impact the kinetics of the form I to form III conversion, increasing the transformation rates at increasing concentrations of NCM. PXRD, NMR and thermal analysis confirmed the formation of solid solutions. Although impurities are well-known to impact crystal habit, slow nucleation and crystallization kinetics and impact polymorphism, 7-10 their role in forming solid solutions and the resulting impact on polymorphism has been explored only rarely. 30-36 To the best of our knowledge, this study shows for the first time experimentally and computationally that impurities can invert the thermodynamic stability of polymorphs through the formation of solid solutions. This mechanism may be able to explain the appearance and disappearance of some polymophs. 37 Through its understanding, we are now seeking to rationally design such impurities in order to achieve reliable access of elusive or computationally predicted polymorphs though this type of "solid-solution thermodynamic switch". This novel concept opens a new route to new polymorphs and their control, which will be of major interest to the pharmaceutical industry.

S1.2.1 Liquid Assisted Grinding (LAG)
LAG experiments in isopropanol were carried out using the  Table S1 details the compositions used in both LAG experiments.

Geometry optimisations
All crystal lattices were subjected to the same optimisation procedure using the software VASP (version 5.3.3 [2][3][4][5]. For this, the PBE functional 6 with PAW pseudopotentials 7,8 was used together with the Grimme's van der Waals corrections 9 . For the planewaves, a kinetic energy cut-off of 520 eV was used. The Brillouin zone was sampled using the Monkhorst-Pack approximation 10 using k-points separated by approximately 0.04 Å (see Table S2). Similar methodologies were used in the literature for the computational study of BZM polymorphs 11,12 .
The optimisation cycle performed consisted of two steps: i) a full geometry optimisation allowing for the unit-cell parameters to change followed by, ii) a geometry optimisation with the unit cell fixed.
Structural relaxations were halted when the calculated force on every atom was less than 0.003 eV/Å.
The energy obtained from this process is the electronic energy of the supercell being simulated (E e supercell).

Crystal structure models
The crystal structures of benzamide forms I and III were retrieved from the Cambridge Structural Database (CSD refcodes BZAMID07 and BZAMID08 respectively). A number of supercells were then generated for both forms I and III structures (Table S2). A single molecule of BZM in such supercells was then replaced for a molecule of NCM. By generating different supercells with different number of molecules, a single BZM to NCM substitution allows for the simulation of the various solid-solution stoichiometries (Table S2). Since NCM has a nitrogen atom in meta position from the amide group, two configurations are possible depending on which of the meta position is occupied by the nitrogen (configuration A and B, Figure S1 left). NCM substitutions were performed in both such configurations A and B.

Calculation of gas-phase energy of BZM and NCM
A single molecule of BZM was placed in a 20 Å x 20 Å x 20 Å supercell. The molecule was then allowed to geometry optimise freely (the cell parameters being fixed). The VASP energy model described above was used and the electronic energy of BZM was calculated (E e BZM).
For NCM (configuration A), the same process was repeated and thus the electronic energy of a single NCM molecule was calculated (E e NCM).

Calculation of lattice energies
The lattice energies of forms I and III with its various levels of NCM incorporated were calculated from the electronic energies of the supercell and the molecules in the gas-phase using the equation below: where Nsupercell, NBZM and NNCM are the total number of molecules, number of BZM molecules and number of NCM molecules in the simulation cell.
This allowed for the calculation of the lattice energy for forms I and III BZM as a function of NCM incorporation in both the configurations A and B. The results are plotted in Figure S1 right. In both forms I and III, configurations B always resulted in lower calculated lattice energies and thus results on configurations B are given in the main manuscript.

S1.5.1 Isopropanol based slurries
All slurry experiments were conducted using a total amount of solid mixture of 2. were stirred continuously for 1 week by magnetic stirrers to ensure that solid-liquid equilibrium is achieved. After a week the slurries were filtered under vacuum and the resulting powder was immediately characterised via PXRD.

S1.5.2 Ethanol based slurries
Slurries were performed in a jacketed vessel at 25 °C and at 45°C in 10 g of ethanol using a total load of 5 g of solid.
The different amount of NCM tested were of 2-5 %, 7-10 %, 20 %, 30 %, 40 % and 50 % (expressed as wt% with respect to total solid load of 5 g). Table S3 contains all compositions of the slurries in isopropanol and ethanol.

S1.7.4 Optical Microscopy
Samples from slurries and crash cooling were analysed using Zeiss Axioplan 2 microscope and images were obtained using the INFINITY Analyse and Capture Software version 6.5.6.

S2.1 Identification of polymorphs BZM
The crystalline structure of BZM powders/crystals were identified by comparing the experimental PXRD patterns with the calculated patterns from the single crystal structures obtained from CSD: BZAMID01 (form I) 14 and BZAMID08 (from III) 15 .

NCM
The crystalline structure of commercial NCM powder was confirmed by comparing the experimental PXRD pattern with the calculated patterns from the single crystal structures obtained from CSD: NICOAM05 (form I) 16 .

S2.2 Crystal packing comparison of Form I and Form III
A crystal structure comparison of BZM I and III has been carried out using the XPac procedure 17 using the whole molecule of BZM I as the common ordered set of points (COSP). Results of the comparison, including the dissimilarity index (χ), are reported in Table S5. Figure S2 shows the structural similarity and a description of the crystal packing of BZM I and III.  14 and BZAMID08 (from III) 15 . respectively. In all the cases, powder patterns show a good fit with that of BZM form III, calculated from the single-crystal structure (BZAMID08). At NCM concentration of 40%, extra peaks corresponding to NCM form I appear (blue arrows) in the powder pattern, suggesting that 30-40% represents the solubility limit of NCM into BZM.
Figures S3 and S4 also show linear shifts of the PXRD peaks to higher values of 2θ as the concentration of NCM increases (see also Figure S5). This suggests a decrease of the planar spacing as BZM is partially replaced by NCM, resulting in a solid solution.  Figure S6 shows a comparison of the PXRD patterns of BZM obtained from neat grinding with BZM form I and BZM form III patterns calculated from the single crystal-structures (BZAMID01 and BZAMID08, respectively).

S2.3.2 Neat grinding
The results clearly show that neat grinding of pure BZM promotes the conversion of BZM form I to BZM form III. Figure S6. The PXRD pattern of BZM exposed to neat grinding, and its comparison to BZM form I and BZM form III patterns obtained from CSD.   Figure S11. Single endothermic event can be observed, which is a typical behaviour for solid solution formation. Furthermore, as the doping level of NCM increases, melting point depression occurs, also characteristic of solid solutions. The latter along with the enthalpy of fusion changes are shown in Figure   S12. For comparison, Figure S13 shows the DSC thermographs of manually prepared physical mixtures of BZM and NCM. Two peaks are clearly observed, further confirming a solid solution formation when BZM and NCM system undergoes slurrying.

S2.4.2 Slurries in isopropanol
In order to ascertain whether a different solvent also promotes the conversion of BZM form I to BZM form III in the presence of NCM, slurry experiments in isopropanol were carried out. Figure S14 shows the PXRD patterns of samples obtained by slurrying BZM form I in the presence of NCM concentrations of 2 -30% (wt.%). The results show that the conversion to BZM form III starts at NCM doping level of around 4% (wt.%). Figure S14. PXRD patterns of slurry products obtained from mixtures of BZM and NCM in isopropanol at different wt.% doping levels.   Figure S16 shows the differences between the incorporation of NCM in the crystal lattice BZM during slurry and crash cooling experiments in isopropanol. Higher incorporations of NCM occur during slurry experiments, corresponding to a higher segregation coefficient, which is indicated by the slope of the trendline. Figure S17 demonstrates the differences between the incorporation of NCM in BZM form III and BZM form I at slurring conditions (isopropanol slurries). It was observed that more NCM incorporates in BZM form III lattice than in form I lattice at equilibrium conditions. This is also demonstrated by the estimated segregation coefficients, being 0.8 for NCM in BZM form III and 0.7 for NCM in BZM form I, which are however very close.

S2.7 Optical Microscopy images
Figures S18 and S19 show microscopy images of crystallites obtained from slurries in isopropanol and crash cooling crystallisation, respectively. The initial solution concentration for both samples had 20 wt.% NCM (with respect to total solid added).