Understanding the Effects of Ionicity in Salts, Solvates, Co-Crystals ...

Perspective pubs.acs.org/crystal

Understanding the Effects of Ionicity in Salts, Solvates, Co-Crystals, Ionic Co-Crystals, and Ionic Liquids, Rather than Nomenclature, Is Critical to Understanding Their Behavior Published as part of the Crystal Growth & Design virtual special issue In Honor of Prof. G. R. Desiraju Steven P. Kelley, Asako Narita, John D. Holbrey,‡ Keith D. Green, W. Matthew Reichert,§ and Robin D. Rogers* Center for Green Manufacturing and Department of Chemistry, The University of Alabama, Tuscaloosa, Alabama 35487, United States S Supporting Information *

ABSTRACT: The incorporation of active pharmaceutical ingredients (APIs) into multicomponent solid forms (such as salts and co-crystals) or liquid forms (such as ionic liquids (ILs) or deep eutectic mixtures) is important in optimizing the efficacy and delivery of APIs. However, there is a current debate regarding the classification of these multicomponent systems based on their ionicity which could interfere with their consideration in important applications. Multicomponent systems of intermediate ionicity can show a combination of properties, leading to behavior that is neither strictly typical of either purely ionic or purely neutral compounds, nor easily described as intermediate between the two. In this perspective, we attempt to illustrate the problems in classifying multicomponent APIs based on one of two categories by discussing selected literature regarding solid and liquid multicomponent APIs and presenting the crystal structures of some relevant systems as case studies. It is clear that a focus on restrictive nomenclature carries with it the risk that a thorough examination of the physicochemical properties of the compounds will be overlooked.

I

point trend vs molar composition similar to that of deep eutectic solvents,6 where across a certain molar composition the sample no longer freezes but remains liquid until its glass transition, far below the freezing point of the pure components. FT-IR and NMR spectroscopy showed that the acid and base both shared the acidic proton rather than becoming fully ionized. This issue with determining the ionicity of low-melting multicomponent APIs is mirrored in the solid state. For years, controlling the physical properties of solid APIs through modification of the solid phase without changing the molecular structure of the API itself has remained a prime research interest of academic and industrial research groups.7 One of the most important and most used methods of noncovalent modification of APIs is salt formation, in which the API is neutralized by an acid or a base to make a salt.8 More recently, co-crystallization (broadly defined as the crystallization of two noncovalently interacting neutral compounds in the same crystal lattice, although some prefer to restrict this term to cases where both molecules are solids at ambient temperatures) has

onic liquids (ILs) are functional materials which have been applied in a diverse array of fields, especially in the past 10 years.1 In 2007 our group proposed using ILs composed of biologically active ions as active pharmaceutical ingredients (APIs) and prepared several examples of low melting salts of acidic or basic APIs with various counterions.2 By converting APIs into pure salts that are liquid at or below body temperature, some problems typically associated with solid APIs, such as low solubility and polymorphism, might be overcome. Additionally, a second functionality such as enhanced transport or additional biologically activity could be added by selecting the appropriate counterion. ILs have been loosely defined as salts that melt below 100 °C; however, such a classification has resulted in a host of nomenclature issues and, as a result, misconceptions about the physical, chemical, and biological properties of the compounds themselves. Initially, it was conceived that in the molten state ILs were composed only of dissociated ions. However, further understanding of older and newer data suggests that many protic ILs are in equilibrium with their neutral conjugate acids and bases to a much greater extent than predicted by aqueous pKa values.3,4 For example, we observed that the melting point of APIs could be dramatically lowered through hydrogen bonding between only partially ionized mixtures of acid and bases.5 Mixtures of lidocaine and oleic acid show a melting © 2013 American Chemical Society

Received: January 8, 2013 Revised: February 10, 2013 Published: February 13, 2013 965

dx.doi.org/10.1021/cg4000439 | Cryst. Growth Des. 2013, 13, 965−975

Crystal Growth & Design

Perspective

also been used for this same purpose.9,10 Co-crystallization has even more scope than salt formation because there appears to be no theoretical limit on the types of APIs that can be incorporated into co-crystals. Researchers have been able to make new co-crystals year after year11 by combining design strategies such as the concepts of supramolecular synthons12 with high throughput screening methods such as solventassisted grinding.13 The application of these relatively new methods of noncovalent modification of APIs has sparked interest in their regulatory and intellectual property ramifications, as well as the technological aspects. New crystalline salts and solvates of a known API are considered patentable new forms. Cocrystals are also considered patentable, novel forms if they improve upon drug properties.14 Co-crystals are not yet heavily marketed as drugs, but there is ample reason for optimism. The high number of citations for one of the first studies comparing the pharmacokinetic profile of a novel co-crystal to a marketed drug,15 published in 2007, captures the overall interest in economic aspects of the field. Indeed, some APIs currently marketed as salts, such as caffeine citrate,16 may in fact be cocrystals. As a result of the very high recent interest and growing numbers of patent applications in the field of API co-crystals, the FDA recently released draft guidance outlining their proposed regulatory classification of co-crystals.17 Under this guidance, pharmaceutical solids would be grouped in three categories: salts, co-crystals, and polymorphs. Co-crystals are considered “crystals with two or more neutral molecules in the lattice,” salts are crystalline solids which contain ions, and polymorphs are different solid forms of the same API including amorphous forms and solvates. Under the guidance, co-crystals would be considered analogous to API-excipient systems such as inclusion compounds of drugs and so would not be considered new drugs. Rather, they would be eligible for approval under an Accelerated New Drug Application (ANDA), which only requires proving bioequivalence to a market drug. By contrast, the FDA regulates novel salts of an API as entirely new drugs which must undergo full testing. Since its release, large numbers of experts in solid state chemistry have voiced concerns regarding the draft guidance,18 based largely on two observations. First, many multicomponent APIs could not be unambiguously classified by the three mutually exclusive categories. For instance, it is not clear how to treat a co-crystal of a salt, such as the well-known co-crystals of fluoxetine HCl with carboxylic acids,19 given the different regulatory treatment of co-crystals vs. salts. Second, the starkly different treatment of salts and co-crystals does not reflect the equilibrium between the two forms that actually exists in many cases. The FDA-proposed classifications have similar ramifications for the development and use of multicomponent liquid APIs. The design of ILs has a lot in common with the design of cocrystals (even the principles of rational design of ILs are sometimes called anti-crystal engineering20), and systems such as the low melting eutectics of lidocaine and oleic acid (quite similar to a co-crystal but in liquid form) cannot be neatly described by any of the proposed FDA classifications. The attempts to classify (or even regulate) these compounds with nomenclature is presumably based on physical and chemical similarities being present among compounds in the same category due to some common chemical feature. The regulations proposed by the FDA suggest that ionicity is

responsible for the different behavior of one class of compounds over another. ILs are also defined at least in part by their ionicity,21 and quantifying the effects of ionicity has been a major issue in this field.3,22 Purely ionic and purely neutral multicomponent compounds have been viewed as two extremes of a spectrum on the basis of charge separation, with ions being fully charge separated and neutral species being fully charge delocalized. However, in terms of intermolecular interactions and packing, ions and neutral molecules are different to the point where there is no spectrum. Mixtures of ions and neutrals have combined properties, and the ionic and co-crystalline traits are not easily inferred from the stoichiometry and empirical formula. While ionicity clearly affects properties, it is too early to begin grouping all possible mixtures of ions and neutral species into one of two groups. A focus on restrictive nomenclature carries with it the risk that a thorough examination of the physicochemical properties of the compounds will be overlooked. One might ask if there is in fact a set of properties that vary systematically with ionicity and if so, how? In order to help us answer this question in our field of interest, we have decided to take advantage of the research in this area by the solid state community. By understanding how the structures of multicomponent solids vary with ionicity, we can develop better models for the multicomponent liquids such as ILs where the system is more dynamic and harder to characterize. In this Perspective, we compare the literature precedent for classifying multicomponent solids vs. liquids based on ionicity from selected references and describe the intermediate cases for both. We also use the structures of several crystals isolated in our laboratories as case studies to illustrate the complex relationship between structure, ionicity, and nomenclature. Mixtures of Neutral Molecules and Ions. Salts and cocrystals are often viewed as extremes of a continuous spectrum of multicomponent solids ranging from fully ionized to fully nonionized. ILs and molecular liquids can be considered the extremes for the liquid analogue of this spectrum, with solutions of salts in molecular liquids (which would include partially ionized acid/base mixtures in the absence of “solvent”) occupying the middle. However, the intermediate ionized states of both solids and liquids can be tricky to classify because they contain both ions and neutral species. Things can become even more complicated when it is not possible to determine which molecules are ionized due to tautomerization or if two formally noncovalently bonded species are so strongly associated (for instance, through a low-barrier hydrogen bond) that they behave as a single ion or single neutral moiety. One group of these intermediate compounds, sometimes called ionic co-crystals, consists of multicomponent solids with both neutral molecules and an electrically neutral ion combination. The co-crystals of fluoxetine·HCl (Prozac) and various organic acids reported by Childs et al. are important examples.19 However, purely inorganic salts such as alkali and alkaline earth metal halides have been co-crystallized with a wide range of weakly acidic amides such as barbituric acid.23 These compounds are also related to salt solvates, an extremely well-known class of crystalline solids which also contain ions and neutral molecules (usually water or the solvent from which the crystal was grown), and some researchers now define cocrystals as being composed of two or more normally solid compounds in order to distinguish co-crystals from solvates.9,18 When inorganic ions are present in the co-crystal, the 966



Perspective

Scheme 1. Schematic Representation of Formation of Oligomeric Ions Formed by Adding Excess of Lidocaine (Left) or Salicylic Acid (Right) to Lidocaine Salicylate (Center)

was linearly correlated with ΔpKa across the range of −1 to 4.28 However, the crystal structures were only sorted into one of two categories: those that showed no proton transfer at all versus those where at least one proton was fully transferred. Crystals containing excess acid or base (ionic co-crystals) are classified the same way as fully ionized salts. Also, disordered crystals (such as those in which the proton may be disordered between the acid and the base) were excluded completely. For ILs, more techniques are available to determine the extent of proton transfer, and properties of ILs correlated with proton transfer (e.g., excess boiling point, ionic conductivity, etc.) increase continuously with ΔpKa for values far greater than 3.3 The ΔpKa values only serve to rank acid−base pairs in terms of their likelihood to form salts in both the cases of ILs and crystalline solids. It is not possible to reliably predict the internal structure or stoichiometry of these systems based solely on ΔpKa. Low, positive ΔpKa values (between 0 and 3, for instance) show an especially pronounced amount of structural variability. A combination review/study by Childs et al. published in 2007 summarized numerous examples of how the correlation of properties with ΔpKa breaks down in this range.29 For instance, crystal structures of multicomponent crystals containing phenols show a wide range of C−O bond distances in this ΔpKa window.30 A similar trend is observed for the degree of asymmetry in carboxylate C−O bonds, which are totally symmetric at the limit of ionicity and most dissymmetric when the molecule is protonated. However, between the ΔpKa range of 0−3 the differences vary widely and are not correlated with ΔpK a . A neutron diffraction study of a co-crystal of pentachlorophenol and 4-methylpyridine found that the average hydrogen position shifted as a function of temperature, being more ionic in character at low temperatures and more cocrystalline at higher temperatures.31 ILs made by reacting Brønsted acids and bases are typically referred to as “protic ILs”. The properties of protic ILs are also affected by their degree of ionicity.3 Partially ionized protic ILs have higher vapor pressures and lower conductivities than fully ionized, aprotic ILs. While ΔpKa rules are used to guide selection of acids and bases when designing a protic IL, the ΔpKa rule for protic ILs is quite different than that noted above. Miran et al. found that protic ILs of the superbase 1,8diazabicycloundec-7-ene (DBU) with various strong acids only show identical behavior to aprotic, fully ionized ILs when ΔpKa was greater than 15.4

compound can be clearly described as having both ions and neutral species, and Braga et al. have advocated reserving the term ionic co-crystal for these structures. Defining the ions in an ionic co-crystal becomes more complicated if there are multiple possible ionization states for the components, and the stoichiometry implies that some must be neutral and others must be ionized. Mixtures of ILs and formally neutral molecules are even harder to describe this way because ILs are liquids and thus dissolve or mix with other substances across a range of concentrations. For instance, deep eutectic solvents are typically formed by mixing an ionic salt, such as choline chloride or zinc chloride, with a neutral molecule such as urea.24 Deep eutectic solvents show IL-like behavior, but they cannot be purely ionic because the eutectics form across a range of stoichiometries. We have also prepared low-melting multicomponent APIs salts by mixing an API salt with an excess of either the acid or base to form oligomeric ions, such as lidocaine salicylate·salicylic acid (Scheme 1).25 Mixtures of Brønsted Acids and Bases. Weak Brønsted acid and base pairs are commonly used to make solid salts, cocrystals, and ILs. However, even in the crystalline state the ionized and nonionized forms of the components can be in equilibrium. Additionally, the oligomerization of neutral acid or base molecules with their conjugates can also occur, and products are not necessarily what would be predicted based on the stoichiometry of the reactants. The ΔpKa values between acid and base (ΔpKa = pKa[HB]+ − pK a [HA]) are commonly used to predict whether neutralization can occur. Crystalline solids are usually described as ionic if ΔpKa > 3 (in aqueous solutions, this corresponds to a 1000 to 1 ratio of products to reactants, or a 99.9% complete proton transfer). A study on co-crystals of 4,4′-bipyridine and various carboxylic acids found that if ΔpKa < 0, the crystal is likely to be totally nonionized.26 Gilli et al. have used a combination of theoretical and empirical evidence to create a ΔpKa “slide rule” which predicts both the degree of charge separation and the strength of a hydrogen bond interaction between a wide range of acid−base pairs.27 However, as ΔpKa values are based on dilute aqueous solutions at equilibrium, ΔpKa rules are only qualitative predictors for solids, let alone the more dynamic liquids. This is not to say that no correlation exists between ionization in the solid state and ΔpKa. A recent survey of over 6000 structures from the CSD showed that the number of crystal structures reported to contain ionized versus nonionized acid base pairs 967



Perspective

Figure 1. Packing of 1,2-dimethylimidazolium picrate down the b axis (left) vs packing of the 1:1.5 co-crystal of 1-methyl-4,5-dicyanoimidazole and picric acid down the c axis (right). Figure prepared using coordinates from ref 33.

equivalents of acid or base than needed to form the neutral salt). Packing and interactions can also be used to justify ionicity, as seen in the crystal structures of the 1:1 co-crystal of pyridine and formic acid vs the 1:1:3 salt solvate of pyridinium formate and formic acid.35 In the 1:1:3 salt solvate, neutral formic acid molecules act as both hydrogen bond donors and acceptors in a manner which is only consistent with the pyridinium cation being protonated and the central formate ion being deprotonated (Figure 2, right).

The lack of stoichiometric restrictions on the composition of the liquid leads to other interesting phenomena as well. For instance, mixtures of N-methylpyrrolidine and acetic acid actually become more ionic when the acid is present in excess.32 This is not simply due to the excess of a reactant driving the reaction farther to completion, but also due to strong hydrogen bonding between acetic acid molecules and acetate anions which forms more noncoordinating, oligomeric anions. Effects of Ionicity on Properties. It is reasonable to assume that multicomponent substances composed solely of ions will be different from those composed solely of neutral species. Ions exert long-range Coulombic forces in every direction, so an ion often shows different interactions from its neutral counterpart. Also, ionic compounds must be composed of the appropriate stoichiometry of ions for electroneutrality and must have some sort of charge ordering. Crystalline salts and ILs both show structural features associated with charge ordering. By contrast, co-crystals are not required to contain the components at any particular stoichiometry. The components usually have some complementary interaction which allows them to crystallize together, but they are not required to alternate with each other. This is illustrated in the comparison of the crystal packing between a salt, 1,2dimethylimidazolium picrate, and a 1:1.5 co-crystal of 1methyl-4,5-dicyanoimidazole and picric acid.33 The salt packs in alternating sheets of cations and anions (Figure 1, left), while the co-crystal packs in zones of imidazole and picric acid molecules interacting with each other separated by zones of self-associating picric acid molecules (Figure 1, right). Investigation into the differences between salts and cocrystals has been primarily based on molecular structure. (The review by Childs et al. and references therein are good examples.29) This is because of an emphasis on crystallographic data in which the position of a hydrogen atom along a single hydrogen bond determines the ionicity of the structure, or cases in which spectroscopy and molecular structure are the most reliable indicators of ionization state. If the only differences between salts and co-crystals were the structures of the molecules vs ions, then a compound could be assigned its position on the continuum based solely on the experimental determination of the hydrogen atom position or ratio of ions to neutral molecules. However, evidence exists for systematic packing differences between salts and co-crystals. A CSD survey by Aakeröy et al. of around 230 salts and co-crystals found that co-crystals were much more likely to crystallize with the expected stoichiometry (based on the complementary interaction used to make the cocrystal) than salts.34 Salts were much more likely to form solvates and crystals with unexpected stoichiometries (i.e., more

Figure 2. Hydrogen bonding in the asymmetric units of the 1:1 cocrystal of pyridine and formic acid (left) vs. the 1:1:3 salt solvate of pyridinium formate and formic acid (right). Figure produced using coordinates from ref 35.

Ionicity in ILs manifests itself through its effects on different physical properties. Protic ILs often have higher vapor pressures and lower conductivities than aprotic ILs due to the presence of nonionized species.3 Charge ordering is expected to lead to higher structural order in ILs than molecular solvents.36 However, it would seem that the ion conductivity of the molten salt would be the physical property with the strictest quantitative relationship to ionicity. The relationship between ionicity of ILs and their conductivity can be inferred by the Walden plot method.37 In this method, the log of molar conductivity of the liquid is plotted as a function of the log of viscosity. The method is based on the Walden rule, which predicts that this plot should have a slope of 1 and pass through the origin. By using a reference which obeys the Walden rule, usually aqueous KCl, the y-intercept of the plot indicates how freely the ions can migrate. Those with y-intercepts lower than the KCl reference line are effectively less ionic. However, this method is affected by both the concentration of nonionized species and ion association (after all, crystalline salts are not poor conductors because they are not ionic) and so is a qualitative, rather than quantitative, indicator of ionicity. It is worth noting that very small, nonremovable amounts of certain species may be present in a “pure” liquid due to selfreactions and can have a major influence on bulk physical or even chemical properties. The hydronium and hydroxide ions in liquid water are a prime example. This phenomenon is 968



Perspective

[Mor][HMal] are salts, [Mor][Succ]·H2Succ contains both ions and a neutral molecule and can be considered a type of ionic co-crystal. The crystal structure of [Mor][OAc] is exemplary of full ionicity. The ΔpKa42 value for morpholine (pKa = 8.36)43 and acetic acid (pKa = 4.76) is 3.60, suggesting this compound should be fully ionized at 1:1 stoichiometry. The hydrogen atoms were both located within bonding distance to the nitrogen atom on morpholine and at tetrahedral angles to each other (Figure 4, top). The morpholinium and acetate ions form alternating hydrogen bonded chains which interact with each other to make the 3-D, charge ordered lattice. While maleic acid is diprotic, pKa2 is much greater than pKa1 (6.58 vs 1.9344), so it would be expected that all maleic acid molecules should be singly deprotonated before double deprotonation can occur. Because ΔpKa of morpholinium and maleic acid is 6.43, complete formation of the 1:1 salt is expected. Although the compound crystallizes with two symmetry-unique formula units (Figure 5, left), both are singly ionized as indicated by the position of the hydrogen atoms located from the difference map, as well as the packing. The packing is somewhat more complicated because the [HMal]− ions engage in dimeric stacking. Incidentally, because all molecules in the system are charged, this requires two morpholinium cations to pair with each other. They do so by packing with the uncharged parts of the molecule facing each other so the positively charged ammonium head groups face anions (Figure 5, right). The double cations and double anions pack with each other in a salt-like fashion with all ammonium and carboxylate groups involved in charge-ordered hydrogen bonding (Figure 6). For succinic acid, pKa1 and pKa2 are closer to each other (4.19 and 5.48, respectively45). The actual pKa values in a crystal will be different due to packing forces, and succinic acid molecules can exist in multiple ionization states within the same lattice in a manner that does not correlate with the base used.46 The 1:1 morpholinium/succinic acid compound we isolated has four protonated morpholinium cations, two doubly deprotonated succinate dianions, and two neutral succinic acid molecules per asymmetric unit, giving it the overall formula [Mor]2[Succ]·H2Succ with Z′ = 2. While [Mor][OAc] and [Mor][HMal] formed large single crystals, [Mor]2[Succ]·H2Succ consistently formed intergrown masses of tiny crystals or syrups when recrystallized from a number of different solvents. This is likely due to the larger number of symmetry unique molecules inhibiting crystallization. As seen in the ORTEP of the asymmetric unit (Figure 7), the succinate anions accept strong hydrogen bonds from both the cations and the succinic acid molecules. The carboxylate bond lengths of the succinic acid molecules are highly asymmetric and match those found in the crystal structure of pure succinic acid well,47 indicating that they do not carry any of the negative charge. Hence, this crystal structure can be described as a mixture of discrete ions and neutral molecules. The cations and anions form charge-ordered hydrogen bonded chains as expected, but the succinic acid molecules have no charge ordering requirement. The succinic acid molecules always donate hydrogen bonds to succinate anions, and as a result the hydrogen bonding is markedly different than in the other two salts. The bridging succinic acid molecules link the cation−anion chains into a 3-D open network of strong hydrogen bonding (Figure 8, left). The

responsible for the unusual properties of 1,3-dialkylimidazolium acetate ILs. Although dialkylimidazolium cations are normally considered aprotic, it is well-known that they can be deprotonated at the C2 position by superbases to form Nheterocyclic carbenes.38 The possibility of deprotonation of the imidazolium cation by acetate has been predicted to influence properties such as the vapor pressure of 1,3-dialkylimidazolium acetate ionic liquids.39 We have shown the influence of N-heterocyclic carbenes on the chemical properties of 1,3-dialkylimidazolium acetate salts through their reactions with chalcogens40 and carbon dioxide.41 Although the carbene is not present at spectroscopically detectable levels, these reactions proceed to yields of 50− 90%. The fact that reactions with acids such as CO2 proceed without addition of base indicate that deprotonation of the imidazolium cation by acetate is the source of the steady-state concentration of carbene in the IL. The deprotonation of imidazolium by a comparatively weak base such as acetate (compared, for example, to hydroxide) can be rationalized by the ability of two acetate anions to stabilize the acidic proton through a low-barrier hydrogen bond, as seen in the crystal structure of the 1-ethyl-3-methylimidazolium acetate-CO2 adduct (Figure 3). While this striking example is not due to

Figure 3. Asymmetric unit of the co-crystal of neutral 1-ethyl-3methylimidazolium-2-carboxylate (the −CO2 adduct of [C2mim][CH3COO]) and [C2mim][H(CH3COO)2]. Figure produced using coordinates from ref 41.

intermediate ionicity per se, it does clearly illustrate how the dynamic nature of the IL allows physically and chemically important equilibria to proceed which could otherwise be negligible in solids. It is clear from the literature that the effects of ionicity have been of concern to researchers, even if a systematic relationship between ionicity and properties remains elusive. Let us turn to a few examples of crystal structures of mixed ionicity from our laboratories. Structural Effects of Mixed Ionicity: Salts of Morpholine with Carboxylic Acids. As mentioned earlier, if salts and co-crystals always packed similarly then the effects of ionicity on the crystal structure would be straightforward. To demonstrate how intermediate ionicity affects the packing, we present the crystal structures of morpholinium acetate ([Mor][OAc]), morpholinium hydrogen maleate ([Mor][HMal]), and morpholinium succinate succinic acid ([Mor]2[Succ]·H2Succ). All three were prepared by evaporating 1:1 mol:mol aqueous solutions of morpholine and the corresponding acid, and all three crystallize with morpholine and the acid at a 1:1 ratio. However, while [Mor][OAc] and 969



Perspective

Figure 4. 50% Probability ellipsoid plot of the asymmetric unit of [Mor][OAc] (top), hydrogen bonded chains (bottom left), and packing down the ac diagonal (bottom right, green = cations, blue = anions).

Figure 5. 50% probability ellipsoid plot of the asymmetric unit of [Mor][HMal] (left) and ball-and-stick plot showing cation−cation and anion− anion aggregations (right).

molecular recognition that results in co-crystallization. Rather than being intermediate, or an average, between a salt and cocrystal, [Mor]2[Succ]·H2Succ is better described as being a combination, or a sum of a salt and a co-crystal. The crystal structures of these salts help explain some of the phenomena observed in ionic liquids. For instance, a study by Fukaya et al. reported the synthesis of choline salts of acetate, hydrogen maleate, and hydrogen succinate.48 [Choline][HMal] had a lower freezing point than [Choline][OAc], which was explained as being due to [HMal]− being a weaker base than [OAc]−. However, this logic implies that the hydrogen succinate salt should have a freezing point in between the other two, as its pKa is closer to that of acetic acid. Instead, the hydrogen succinate salt was reported not to crystallize at all.

packing can be described as the interpenetration of two of these cages (Figure 8, right). These three cases show an interesting relationship between composition, ionicity, and packing. First, the presence of mixed ionization states in [Mor]2[Succ]·H2Succ shows that neither the ΔpKa rule nor attempting to control the reaction through stoichiometry can guarantee control over the system. Second, while the ionicity of each of the molecules and ions in [Mor]2[Succ]·H2Succ is fairly unambiguous, the packing shows a mixture of ionic and co-crystalline effects. A certain supramolecular moiety, 1-D hydrogen bonded cation−anion chains, is governed by charge ordering. The succinic acid bridges, however, are governed by the ability of acids to form strong bonds to their conjugate bases, which is the type of 970



Perspective

Compared to the lattices of [Mor][OAc] and [Mor][HMal], the complex lattice shown by [Mor]2[Succ]·H2Succ is likely more kinetically and entropically disfavored relative to the liquid state. The presence of [Succ]2− and H2Succ, possibly in equilibrium with [HSucc]−, are likely responsible for the low melting point and supercooling of choline hydrogen succinate. Imidazolium Ionic Co-Crystals/Salt Solvates with Oligomeric Ions. Multicomponent crystals of the formula [HB][A]·HA or [HB][A]·B are well-known and usually include strong, charge assisted hydrogen bonds between conjugate acid−base pairs. This hydrogen bond can be strong and partially covalent in nature, so it is possible for hydrogenbonded oligomers to behave as large, single ions. We will illustrate this here with the co-crystal 1,2-dimethylimidazolium hydrochloride·1,2-dimethylimidazole ([HDmim][Cl]·Dmim) and the salt 1,2,3-trimethylimidazolium salicylate·salicylic acid ([Tmim][H(Sal)2]). [HDmim][Cl]·Dmim was made by reacting an ethanol solution of 1,2-dimethylimidazole with hydrochloric acid and allowing the product to crystallize on standing at RT. The compound crystallizes with one neutral Dmim molecule, one [HDmim]+ cation, and one chloride anion. Because Cl− is weakly basic it can be assumed that one of the Dmim molecules must be protonated. The hydrogen atom was located from the difference map and shown to participate in a [Dmim-H]+− [Dmim] hydrogen bond, but it is much closer to one molecule than the other. The location of the hydrogen and classification of this multicomponent system is supported by the C−N bonds within each ring. The protonated imidazolium cation has two symmetric C−N bonds, which is typical for the electronically delocalized cation. The nonprotonated imidazole has asymmetric C−N bonds, which is typical of neutral imidazole molecules where one nitrogen atom is sp2-hybridized and the other is sp3-hybridized. The asymmetric unit is shown in Figure 9. On the basis of molecular structure, this compound could be considered a co-crystal; however, both molecules play very similar roles in crystal packing. Both make hydrogen bonds to chloride ions through the acidic ring hydrogen atoms and the methyl group α- to the nitrogen atom. Each ring π-stacks with two other rings in infinite chains of alternating cations and neutral molecules. The hydrogen-bound cation-neutral dimer essentially behaves as a single cation in a salt-like packing (Figure 10).

Figure 6. Hydrogen bonding (top) and packing down the b axis (bottom, green = cations, blue = anions) in [Mor][HMal].

Figure 7. 50% probability ellipsoid plot of the asymmetric unit of [Mor]2[Succ]·H2Succ.

Figure 8. Hydrogen bonding network (left) and packing down the c axis (right, green = cations, blue = anions, black = neutral) of [Mor]2[Succ]·H2Succ. 971



Perspective

[HDmim][Cl]·Dmim and [Tmim][H(Sal)2] illustrate more of the complex ways in which crystals can exhibit both salt and co-crystalline traits. [HDmim][Cl]·Dmim can be described as an ionic co-crystal, while [Tmim][H(Sal)2] might better be described as a salt of an oligomeric anion. However, [HDmim][Cl]·Dmim packs with a salt-like charge-ordered lattice, while [Tmim][H(Sal)2] has co-crystalline character where one of the salicylate molecules does not appear to be restricted by charge ordering. The differences above can be rationalized on the basis of the different shapes of the ions involved. The localization of the positive charge to one of the imidazolium cations in [HDmim][Cl]·Dmim allows π-stacking to occur with less cation−cation repulsion. The large [H(Sal)2]− ion appears to twist to allow it to pack more efficiently and engage in more short contacts. Because of this twisting, a lattice of stacked cations and anions similar to the lattice of [HDmim][Cl]·[Dmim] does not form. We have previously reported the mixing of solid salts with excess free acid or base as a liquefaction strategy for solid salts of Brønsted acids and bases, including mixing salicylate salts with salicylic acid (see, for example, Scheme 1).25 The reaction had notable effects on the physical properties of the compounds, particularly melting points, and 1H NMR evidence indicated that the acidic proton was bound in strong hydrogen bonds to both the acid and its conjugate base. We call this strategy the “confused proton” approach, as it lowers the melting point by allowing ions to oligomerize through hydrogen bonding in which the acidic hydrogen atoms are free to move between molecules, increasing charge delocalization and disorder, and it was found to work with the addition of excess base as well as acid (Scheme 1). The ionicities of mixtures of salicylate salts and salicylic acid were lower than those of the pure molten salts, which is consistent with the formation of oligomers and charge delocalization. However, a report by Stoimenovski et al. on mixtures of a base with excess acetic acid showed that the formation of oligomeric acetic acid−acetate ions was associated with greater ionicity than the one-to-one mixture of acid and base.32 Interestingly, the packing of [Tmim][H(Sal)2] appears to show less ionicity compared to the packing of [HDmim][Cl]·Dmim. In the case of the liquid mixtures of acetic acid− acetate which formed oligomeric ions, ionicity increased because the primary cation−anion interaction, hydrogen bonding, was weakened by decreased basicity of the oxygen atoms. However, while hydrogen bond basicity of the

Figure 9. 50% probability ellipsoid plot of the asymmetric unit of [Dmim][Cl]·Dmim.

The salicylate salt, [Tmim][H(Sal)2], was made by reacting 1,2,3-trimethylimidazolium carbonate with salicylic acid and allowing the product to crystallize on cooling. [Tmim][H(Sal)2] crystallizes with two salicylate molecules sharing a single proton and one [Tmim]+ cation in the asymmetric unit (Figure 11, left), with the cation disordered over two conformations. Again, the cation has no basic sites, so the stoichiometry suggests one of the salicylate molecules must be ionized while the other is neutral. On the basis of the location of the hydrogen atom and the asymmetry of the carboxylate C−O bonds in both salicylate molecules, they both appear to be carrying some of the negative charge. This compound could then be described as the oligomeric (in this case dimeric) salt of [Tmim]+ and [H(Sal)2]−, where each salicylate is neither entirely charged nor entirely neutral. They do not share the charge equally though; one of the salicylate molecules is significantly closer to the hydrogen atom and has more asymmetric C−O bonds than the other (and is thus less ionized). The packing, however, is quite unlike that of a typical salt. The phenyl groups of the two salicylate molecules behave differently. The more ionic salicylate molecule participates in charge-ordered stacking with the imidazolium cations. The phenyl group of the other salicylate molecule accepts C−H···π interactions from the imidazolium cation and the more ionized salicylate. The effect of these different packing roles results in a more complicated lattice composed of charge-ordered columns bridged by salicylate molecules (Figure 11, right).

Figure 10. Short contact environment around the cation-neutral dimer (left) and packing down the c axis (right) of [HDmim][Cl]·Dmim. 972



Perspective

Figure 11. 50% probability ellipsoid plot of asymmetric unit (left) and packing down the c axis (right) of [Tmim][H(Sal)2]. The disorder in the cation is excluded.

Figure 12. Equilibrium between fully ionized, fully nonionized, and oligomerized species which can be present in solid or liquid mixtures of Brønsted acids and bases.

Scheme 2. Equilibrium of States between Lidocaine (pKa = 7.9) and Oleic Acid (pKa = 4.2)

state by inference. Furthermore, it may not be possible to trap many of the IL systems in a crystalline state, especially those such as liquid co-crystals in which the melting point suppression is related to a dynamic equilibrium. In these systems, especially protic systems, it is possible to find ionized and nonionized species, oligomeric ions, and indeed a continuum of partially charged states (Figure 12). This equilibrium has different implications for liquids than it does for solids. In a crystalline solid, acidic protons are mobile while the molecules themselves are more firmly locked in place. Changes in ionicity can occur without changing the packing, and if the solid is dissolved the strong hydrogen bonds will be broken. Hence, from a regulatory standpoint the salt co-crystal continuum may be about predicting whether or not the API will behave as a neutral molecule or an ion when in solution. An IL, on the other hand, does not have to dissolve to cross membranes. In the case of a mixture such as lidocaine and oleic acid (Scheme 2), the equilibrium is a continuum of states, not just a continuum of ionicity. While fully ionized and fully neutral molecules can dissociate in solution, the hydrogen bond holding the two molecules together at intermediate ionicity may be strong enough that they remain associated in the liquid

carboxylate groups is also going to be decreased by oligomerization of the salicylate ions, the interactions involving the phenyl group will not be as strongly affected. The increasing importance of C−H···π interactions in light of reduced hydrogen bonding may explain why this crystal deviates from salt-like packing. Hence, similar interactions may be occurring in the molten salicylic acid−salicylate mixtures. These crystal structures imply that even though oligomerization appears to generally result in melting point reduction, the restructuring mechanism associated with oligomerization is dependent on the functional groups on the ion. It is particularly noteworthy that the structures of the molecules themselves imply that [Tmim][H(Sal)2] is more ionic as it consists only of ions (if the oligomeric salicylate moiety is considered a single ion and less than fully ionized molecules are considered ions), while [HDmim][Cl]·Dmim clearly has neutral molecules and ions. These cases point out an interesting correlation between ionicity in the solid state and observed ionicity in the liquid state. However, it is naı̈ve to assume that crystal structures can always be used to determine detailed properties of the molten 973



Perspective §

state. The movement of the two molecules through solution or across a membrane can be correlated, meaning that the API is not described adequately by an ionic or a neutral model. Instead, the API is now part of a larger complex whose transport properties and charge are more significantly affected by the counterion than is assumed to occur in a salt or a cocrystal. Each of these species will affect physical properties and electronics in a different way, and it may be possible to determine which species are present by combining a wide range of measurements. This is certainly necessary to further understand what these compounds are and how they result in melting point suppression; however, in the context of preparing more effective APIs, the only thing that may matter is whether or not the compound or mixture works as intended. Again, the original goal of preparing API ILs was simply to alleviate problems associated with solid APIs by lowering their melting points. While a variety of speciation changes may result in this melting point reduction, the phenomenon itself is easy to measure. Attempting to regulate them by ionicity may be fruitless in and of itself, since it seems unlikely that a simple relationship between ionicity and properties such as bioequivalence and toxicity exists.

(W.M.R.) Department of Chemistry, University of South Alabama, Mobile, AL 36688, USA. Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. Funding

AFOSR Award No. F49550-10-1-0521. NSF-SURP Grant No. CHE 0243883. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This material is based upon work supported by the Air Force Office of Scientific Research under AFOSR Award No. F4955010-1-0521. K.G. was supported by The University of Alabama’s 2003 NSF Summer Undergraduate Research Program grant (NSF-SURP Grant No. CHE 0243883).

■ ■

DEDICATION This paper commemorates Professor Gautam Desiraju’s 60th birthday.

■

CONCLUSIONS Purely ionic and purely neutral multicomponent compounds have been viewed as two extremes of a spectrum, but reality is comprised of a continuum of everything in between. While ions and neutral molecules are indeed two ends of this continuum based on charge separation, with ions being fully charge separated and neutral species being fully charge delocalized, attempts to extend this to bulk physical properties of solids and liquids are unlikely to capture the full scope of possible useful combinations. Mixtures of ions and neutrals have combined and sometimes unpredictable properties, whether solid or liquid. While ionicity clearly affects properties, the outcome of its effect on structural or even physical properties (much less biological properties) of possible mixtures of ions and neutral species cannot always be predicted and may be quite unexpected. A focus on restrictive nomenclature at this point will lead to misassumptions about physicochemical properties at a time in which careful study is required to master control over these promising but very subtle materials.

■

ASSOCIATED CONTENT

S Supporting Information *

Crystallographic information files and experimental details on the synthesis and characterization of [Mor][OAc], [Mor][HMal], [Mor]2[Succ]·H2Succ, [HDmim][Cl]·Dmim, and [Tmim][H(Sal)2]. This material is available free of charge via the Internet at http://pubs.acs.org. Crystallographic information files are also available from the Cambridge Crystallographic Data Center (CCDC) upon request (http://www.ccdc.cam.ac. uk, CCDC deposition numbers 918420−918424).

■

REFERENCES

(1) (a) Angell, C. A.; Ansari, Y.; Zhao, Z. Faraday Discuss. 2012, 154, 9−27. (b) Forsyth, S. A.; Pringle, J. M.; MacFarlane, D. R. Aust. J. Chem. 2004, 57, 113−119. (c) Welton, T. Chem. Rev. 1999, 99, 2071− 2078. (2) Hough, W. L.; Smiglak, M.; Rodríguez, H.; Swatloski, R. P.; Spear, S. K.; Daly, D. T.; Pernak, J.; Grisel, J. E.; Carliss, R. D.; Soutullo, M. D.; Davis, J. H., Jr. New. J. Chem. 2007, 31, 1429−1436. (3) Yoshizawa, M.; Xu, W.; Angell, C. A. J. Am. Chem. Soc. 2003, 125, 15411−15419. (4) Miran, M. S.; Kinoshita, H.; Yasuda, T.; Susan, M. A. B. H.; Watanabe, M. Phys. Chem. Chem. Phys. 2012, 14, 5178−5186. (5) Bica, K.; Shamshina, J. L.; Hough, W. L.; MacFarlane, D. R.; Rogers, R. D. Chem. Commun. 2011, 47, 2267−2269. (6) Abbott, A. P.; Boothby, D.; Capper, G.; Davies, D. L.; Rasheed, R. K. J. Am. Chem. Soc. 2004, 126, 9142−9147. (7) (a) Berge, S. M.; Bighley, L. D.; Monkhouse, D. C. J. Pharm. Sci. 1977, 66, 1−19. (b) Wei-Qin, T.; Whitesell, G. Pharm. Dev. Technol. 1998, 3, 215−223. (c) Datta, S.; Grant, D. J. W. Nat. Rev. Drug Discovery 2004, 3, 42−57. (d) Morisette, S. L.; Almarsson, Ö .; Peterson, M. L.; Remenar, J. F.; Read, M. J.; Lemmo, A. V.; Ellis, S.; Cima, M. J.; Gardner, C. R. Adv. Drug Delivery Rev. 2004, 56, 275− 300. (8) Stahl, P. H.; Wermuth, C. G. In Handbook of Pharmaceutical Salts; Stahl, P. H., Wermuth, C. G., Eds.; Verlay Helvetica Chimica Acta & Wiley-VCH: Zurich, 2008. (9) Shan, N.; Zaworotko, M. J. Drug Discovery Today. 2008, 13, 440− 446. (10) Frišcǐ ć, T.; Jones, W. J. Pharm. Pharmacol. 2010, 62, 1547− 1559. (11) (a) Brittain, H. G. Profiles Drug Subst., Excipients, Relat. Methodol. 2010, 35, 373−390. (b) Brittain, H. G. Profiles Drug Subst., Excipients, Relat. Methodol. 2011, 36, 361−381. (c) Brittain, H. G. Cryst. Growth. Des. 2012, 12, 1046−1054. (12) Desiraju, G. Angew. Chem., Int. Ed. 1995, 34, 2311−2327. (13) Shan, N.; Toda, F.; Jones, W. Chem. Commun. 2002, 2372−3. (14) Trask, A. V. Mol. Pharmaceutics 2007, 4, 301−309. (15) Hickey, M. B.; Peterson, M. L.; Coppettuolo, L. A.; Morrisette, S. L.; Vetter, A.; Guzman, H.; Remenar, J. F.; Zhang, Z.; Tawa, M. D.; Haleay, S.; Zaworotko, M. J.; Almarsson, Ö . Eur. J. Pharm. Biopharm. 2007, 67, 112−119. (16) Karki, S.; Frišcĭć, T.; Jones, W.; Motherwell, W. D. S. Mol. Pharmaceutics 2007, 4, 347−354.

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Present Addresses ‡

(J.D.H.) School of Chemistry and Chemical Engineering, Queens University Belfast, Belfast BT9 5AG, Northern Ireland, United Kingdom. 974



Perspective

(17) Guidance for Industry: Regulatory Classification of Pharmaceutical Co-crystals; Food and Drug Administration: Silver Spring, MD, December 2011. (18) Atipamula, S.; Banerjee, R.; Bansal, A. K.; Biradha, K.; Cheney, M. L.; Choudhury, A. R.; Desiraju, G. R.; Dikundwar, A. G.; Dubey, R.; Duggirala, N.; Ghogale, P. P.; Ghosh, S.; Goswami, P. K.; Goud, N. R.; Jetti, R. R. K. R.; Karpinski, P.; Kaushik, P.; Kumar, D.; Kumar, V.; Moulton, B.; Mukherjee, A.; Mukherjee, G.; Myerson, A. S.; Puri, V.; Ramanan, A.; Rajamannar, T.; Reddy, C. M.; Rodriguez-Hornedo, N.; Rogers, R. D.; Row, T. N. G.; Sanphui, P.; Shan, N.; Shete, G.; Singh, A.; Sun, C. C.; Swift, J. A.; Thaimattam, R.; Thakur, T. S.; Thaper, R. K.; Thomas, S. P.; Tothadi, S.; Vangala, V. R.; Variankaval, N.; Vishweshwar, P.; Weyna, D. R.; Zaworotko, M. J.. Crys. Growth. Des. 2012, 12, 2147−2152. (19) Childs, S. L.; Chyall, L. J.; Dunlap, J. T.; Smolenskaya, V. N.; Stahly, B. C.; Stahly, G. P. J. Am. Chem. Soc. 2004, 126, 13335−13342. (20) Dean, P. M.; Turanjanin, J.; Yoshizawa-Fujita, M.; MacFarlane, D. R.; Scott, J. L. Cryst. Growth. Des. 2009, 9, 1137−1145. (21) MacFarlane, D. R.; Seddon, K. R. Aust. J. Chem. 2007, 60, 3−5. (22) Tokuda, H.; Tsuzuki, S.; Susan, M. A. B. H.; Hayamizu, K.; Watanabe, M. J. Phys. Chem. B. 2006, 110, 19593−19600. (23) (a) Braga, D.; Grepioni, F.; Maini, L.; Prosperi, S.; Gobetto, R.; Chierotti, M. R. Chem. Commun. 2010, 46, 7715−7717. (b) Braga, D.; Grepioni, F.; Lampronti, G. I.; Maini, L.; Turrina, A. Cryst. Growth. Des. 2011, 11, 5621−5627. (c) Braga, D.; Grepioni, F.; Maini, L.; Capucci, D.; Nanna, S.; Wouters, J.; Aerts, L.; Quéré, L. Chem. Commun. 2012, 48, 8219−8221. (24) (a) Abbott, A. P.; Barron, J. C.; Ryder, K. S.; Wilson, D. Chem.Eur. J. 2007, 13, 6495−6501. (b) Abbott, A. P.; Capper, G.; Davies, D. L.; Rasheed, R. K.; Tamyrajah, V. Chem. Commun. 2003, 70−71. (25) Bica, K.; Rogers, R. D. Chem. Commun. 2010, 46, 1215−1217. (26) Bhogala, B. R.; Basavoju, S.; Nangia, A. CrystEngComm. 2005, 7, 551−562. (27) Gilli, P.; Pretto, L.; Bertolasi, V.; Gilli, G. Acc. Chem. Res. 2009, 42, 33−44. (28) Cruz-Cabeza, A. CrystEngComm 2012, 14, 6362−6365. (29) Childs, S. L.; Stahly, P.; Park, A. Mol. Pharma. 2007, 4, 323− 328. (30) Majerz, I.; Malarski, Z.; Sobczyk, L. Chem. Phys. Lett. 1997, 274, 361−364. (31) Steiner, T.; Majerz, I.; Wilson, C. C. Angew. Chem., Int. Ed. 2001, 40, 2651−2654. (32) Stoimenovski, J.; Izgorodina, E. I.; MacFarlane, D. R. Phys. Chem. Chem. Phys. 2010, 12, 10341−10347. (33) Smiglak, M.; Hines, C. C.; Reichert, W. M.; Vincek, A. S.; Katritzky, A. R.; Thrasher, J. S.; Sun, L.; McCrary, P. D.; Beasley, P. A.; Kelley, S. P.; Rogers, R. D. New. J. Chem. 2012, 36, 702−722. (34) Aakeröy, C. B.; Fasulo, M. E.; Desper, J. Mol. Pharmaceutics 2007, 4, 317−322. (35) Wiechart, D.; Mootz, D. Angew. Chem., Int. Ed. 1999, 38, 1974− 1976. (36) Hardacre, C.; Holbrey, J. D.; Mullan, C. L.; Nieuwenhuyzen, M.; Youngs, T. G. A.; Bowron, D. T. Acc. Chem. Res. 2008, 112, 8049− 8056. (37) (a) MacFarlane, D. R.; Forsyth, M.; Izgorodina, E. I.; Abbott, A. P; Annat, G.; Fraser, K. Phys. Chem. Chem. Phys. 2009, 11, 4962−4967. (b) Angell, C. A.; Byrne, N.; Belieres, J. −P. Acc. Chem. Res. 2007, 40, 1228. (38) Nair, V.; Bindu, S.; Sreekumar, V. Angew Chem. Int. Ed. 2004, 43, 5130−5135. (39) Hollóczki, O.; Gerhard, D.; Massone, K.; Szarvas, L.; Németh, B.; Veszprémi, T.; Nyulászi, L. New J. Chem. 2010, 34, 3004−3009. (40) Rodríguez, H.; Gurau, G.; Holbrey, J. D.; Rogers, R. D. Chem. Commun. 2011, 47, 3222−3224. (41) Gurau, G.; Rodríguez, H.; Kelley, S. P.; Janiczek, P.; Kalb, R. S.; Rogers, R. D. Angew. Chem. Int. Ed. 2011, 50, 12024−12026.

(42) pKa values and references taken from the table of pKa data compiled by Williams, R., accessed online at http://research.chem. psu.edu/brpgroup/pKa_compilation.pdf; last accessed on 10/25/12. (43) Hall, H. K., Jr. J. Am. Chem. Soc. 1957, 79, 5441. (44) Dawson, R. M. C. et al. Data for Biochemical Research; Clarendon Press: Oxford, 1959. (45) Brown, H. C. et al. In Determination of Organic Structures by Physical Methods; Braude, E. A., Nachod, F.C., Eds.; Academic Press: New York, 1955. (46) Sridhar, G. P.; Vijayan, M. Acta Crystallogr. 1991, B47, 927−935. (47) Thalladi, V. R.; Nüsse, M.; Boese, R. J. Am. Chem. Soc. 2000, 122, 9227−9236. (48) Fukaya, Y.; Iizuka, Y.; Sekikawa, K.; Ohno, H. Green Chem. 2007, 9, 1155−1157.

975
