Hydration control in molecular salts through counterion selection Authors: Henry A. Holleb, Natalie E. Pridmore, Amy V. Hall, Pablo Martinez-Bulit, Aurora J. Cruz-Cabeza Published: 2026 ABSTRACT Predicting when and why molecular salts form hydrates has remained a significant problem in solid-state chemistry, despite its impact on the stability and performance of industrially relevant materials. Here, we identify counterion charge localization as a key determinant of hydrate formation in molecular salts, providing a framework for predicting and controlling solid-state hydration. By analyzing over 31,000 crystal structures and performing electron density calculations, we show that delocalized counterion charge suppresses hydrate formation, while localized charge promotes it. This principle makes hydration propensity a more tuneable property (from ~80% to <10%) through considered counterion selection. Guided by these insights and the ΔpKa rule, we experimentally demonstrate rational control of hydration in salts of 4-aminoacetanilide. Our results establish a direct link between molecular electrostatics and crystallization outcomes, affording predictive design of molecular salts with tailored hydration behavior. Introduction Crystallization is a fundamental process across chemical industries, enabling the purification and isolation of compounds1. Despite its central role in chemical research, development and manufacturing, crystallization remains poorly understood, with its outcomes notoriously difficult to anticipate2–5. Among the most unpredictable phenomena is the formation of solvates, crystals that incorporate solvent molecules into their crystal structure, with hydrates being particularly common6–14. There are several compelling reasons to investigate hydrate formation. First, water is a prolific solvate former due to its small size, simple geometry, and strong dipole moment7,10–12. Second, water is ubiquitous and can be absorbed from the atmosphere by many solvents, leading to unexpected hydrate formation even when crystallization occurs from nominally non-aqueous media15,16. Once incorporated into crystals, water molecules frequently become mobile in response to changes in temperature, pressure, or humidity, often resulting in solid-form changes involving full or partial dehydration7,10,11. Importantly, hydrates of various stoichiometries, their polymorphs, non-stoichiometric hydrates, anhydrates, and dehydrated forms all display distinct physicochemical properties, complicating product formulation7,12,17–20. Many molecular-based consumer products are formulated as crystalline salts, including food additives (e.g., sodium glutamate), agrochemicals (e.g., glyphosate isopropylamine), pharmaceuticals (e.g., fluoxetine hydrochloride), and organic conductors and semiconductors (e.g., the charge-transfer salt tetrathiafulvalene–tetracyanoquinodimethane). Notably, around 50% of drugs are salts, as this often improves solubility and related properties21–27; and charge-transfer salts constitute a central class of materials in organic conductors and semiconductors. Salt design and formulation are therefore critical across multiple industries. Given an active molecular ingredient (AMI) of interest, salt design involves two fundamental steps: (i) selection of a counterion (chemical design) and (ii) crystallization of the salt (materials design). Typically, the AMI is first produced as a neutral compound that is subsequently ionized to form the salt. For charge-transfer salts, ionization is achieved by pairing the AMI with a counter-component capable of donating or accepting an electron to/from the AMI, so that mixing results in partial charge transfer and salt formation. Here, the relative ionization potential of one component and the electron affinity of the other drive ionization. For salts involving acidic or basic AMIs (as is the case for most active pharmaceutical ingredients, APIs), salt formation is achieved by pairing the AMI with a counter-compound capable of proton transfer. Figure 1a illustrates salt design for basic and acidic APIs paired with suitable acid and base counter-components22. The likelihood of salt formation between an acid and a base pair can be estimated using the empirical ΔpKa rule, which compares the pKa of the protonated base with that of the acid (Fig. 1a, right). When ΔpKa > 4, salt formation is favored; when −1 <ΔpKa < 4, the outcome depends on the crystal environment; and when ΔpKa <−1, salts are unlikely to form28.Fig. 1Salt formation and hydration trends.a Examples of salt formation for basic and acidic APIs (exemplified with Piriton and Advil salts) and ΔpKa rule (right). b Hydration probability (Pw in %) of molecular crystals by type. c Hydration probability (Pw in %) of crystalline salts classified by counterion type (TFA: trifluoroacetate). Salt formation and hydration trends. a Examples of salt formation for basic and acidic APIs (exemplified with Piriton and Advil salts) and ΔpKa rule (right). b Hydration probability (Pw in %) of molecular crystals by type. c Hydration probability (Pw in %) of crystalline salts classified by counterion type (TFA: trifluoroacetate). Designing salts involves more than simply selecting a counterion based on ionization/electron accepting capabilities or applying the ΔpKa rule. The choice of counter-compounds and the resulting ionized counterions profoundly influences the salt’s crystallization behavior, crystal structure, physicochemical properties (e.g., melting point, hygroscopicity, mechanical strength), and manufacturability29–32. Factors, such as counterion size, shape, hydrogen-bonding ability, and charge delocalization, affect a salt’s properties in ways that are often unpredictable29,30,33,34. In practice, beyond ionization and pKa considerations, salt ‘design’ is largely driven by brute-force screening, testing numerous pairings and crystallization conditions to generate as many salts as possible, followed by characterization and property evaluation29–32. Thus, salt design remains a multidimensional optimization problem, still dominated by experimental trial-and-error approaches. The Cambridge Structural Database (CSD), containing over 1.3 million small-molecule crystal structures, consolidates global crystallographic knowledge in one resource35. Numerous studies have used the CSD to investigate hydrate formation in molecular crystals. Infantes et al. investigated water–water interactions in crystals describing commonly observed assemblies (rings, chains, and sheet motifs), while Gillon et al. quantified the most frequent water environments in crystals -with three hydrogen bonds being the most common36,37. Later, Infantes et al. reported that polarity and hydrogen-bond donors and acceptors increase a compound’s potential to form hydrates38. Recent studies making use of the CSD Python Application Programming Interface (CSD Python API) have enabled more complex analyses of hydrates in the CSD, with Werner and Swift reporting structural relationships between hydrate–anhydrate pairs39,40. Regarding salts, studies by Aakeroy et al. and Cruz-Cabeza et al. have shown that crystals containing ionic species hydrate significantly more frequently than those composed of uncharged molecules41,42. Indeed, around 21% of salts form hydrates compared to just 6% of cocrystals and 4% of neutral single-component crystals (Fig. 1b). The mechanistic reasons behind the heightened hydration propensity of crystals containing ionic species remain unclear, revealing a crucial gap in our understanding. To address this gap, we performed a systematic analysis of salts in the CSD, classifying them by common counterions. While salts overall are significantly more prone to hydration than neutral crystals (Fig. 1b), we find that hydration propensity varies markedly depending on the nature of the counterion (Fig. 1c). In particular, we observe that charge localization on the counterion plays a critical role: counterions capable of delocalizing charge form significantly fewer hydrates than those with localized charge (Fig. 1c). Motivated by this observation, we investigated the structural features governing charge localization in counterions and their relationship to hydrate formation. Using an extensive structural dataset and electron density calculations, we uncover a principle that enables the prediction of hydration probabilities for molecular salts. We demonstrate this principle experimentally by designing salts of a compound with previously unknown salt formation and hydration behavior. This work establishes design rules linking counterion selection to hydration behavior in molecular salts. Results and discussion Salts in the CSD were surveyed, and common counterions were identified and classified into families. All counterions appearing in at least 100 crystal structures were extracted by clustering based on SMILES representations, resulting in 71 common counterions split into four types (Fig. S1). Of these, 22 were excluded from further analysis because they were either (i) too bulky to allow direct interaction between water and the charged site (6 ions, e.g., tetrabutylammonium), or (ii) contained additional (non-intramolecular) hydrogen-bond donors beyond the charged XH+ group (16 ions, e.g., citrate), leading to more complex and less predictable hydration behavior. The remaining 49 counterions were considered for analysis, of which 46 were classified into five distinct families, with three left unclassified (pyridinium, picrate, and DMAP). This yielded a dataset of 46 common counterions across more than 31,000 molecular salts, grouped as inorganic anions (13 counterions, with thiocyanate considered as an inorganic due to its similarity to iodide), sulfonyl anions (5), carboxylate anions (10), inorganic cations (8), and ammonium cations (10), with all counterions analyzed shown in Fig. 2.Fig. 2Families of common counterion types in the CSD.31,401 salts were classified into five main families based on the structure of their most common counterion. The number of crystal structures in each family is given in parentheses. a The anionic families include sulfonyl anions, carboxylate anions, and inorganic anions. b The cationic families contain inorganic cations, and ammonium cations (1-PEA: 1-phenylethylammonium). Families of common counterion types in the CSD. 31,401 salts were classified into five main families based on the structure of their most common counterion. The number of crystal structures in each family is given in parentheses. a The anionic families include sulfonyl anions, carboxylate anions, and inorganic anions. b The cationic families contain inorganic cations, and ammonium cations (1-PEA: 1-phenylethylammonium). The most prevalent salt families in the CSD (classified by counterion type; Fig. 2) are those containing inorganic anions (21.1k structures) or inorganic cations (4.0k), followed by sulfonyl anions (2.8k), carboxylate anions (1.8k), and ammonium cations (1.7k). The most common anionic counterions are chloride (7707 structures), bromide (3791), iodide (2872), and triflate (1692), while the most common cationic counterions are potassium (1160), sodium (887), ammonium (799), and magnesium (310). While many of these counterions are widely used in pharmaceutical salt formulation, others serve various roles in different fields. For example, 1-phenylethylammonium (1-PEA+) is commonly used in studies of chiral systems, while succinate salts are widely used as food additives43–46. Next, we calculated the experimental propensity of hydration (Pw, %) for salts by counterion type using CSD data. This probability was determined by classifying each salt structure as anhydrous or hydrated (if it contained water) and calculating the fraction of hydrated structures (NHydrates) over the total number of salts (N) per ion type, as shown in Eq. (1):1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}(\%)=\frac{{N}_{{{\rm{Hydrates}}}}}{{N}_{{total}}}\times 100$$\end{document}Pw(%)=NHydratesNtotal×100 Hydration propensity varies dramatically when calculated per counterion type (Fig. 1c). Data for selected common anions are shown in Fig. 1c, with all data provided in Table S3. Mesylate salts are hydrates in ~20% of cases, whereas triflate salts are only hydrates in <5%, and salts containing tosylate ions show intermediate hydration propensities (Fig. 1c). Ionic charge is a dominant factor in hydration, with salts hydrating more frequently than cocrystals (21% vs. 6%). Yet this binary distinction oversimplifies the trend, with triflate salts forming fewer hydrates than typical uncharged cocrystals. Clearly, the role of charge in hydration is more nuanced than simply charged versus uncharged or even divalent versus monovalent. Salts in the CSD and crystalline material formulations frequently contain inorganic counterions, particularly the halides (29% of all CSD salts), which biases trends (Fig. 2). This bias means that any global average for salt hydration will be skewed by these dominant few counterions; thus, analyzing hydration behavior by counterion type is essential to mitigate data bias. Analysis of the propensities by counterion type revealed that charge localization within the counterion structure strongly influences the hydration propensity of salts (Fig. 1c). For example, among halides, hydration propensity decreases with increasing ionic radius: iodide salts form roughly half as many hydrates as chloride salts. Similarly, inductive effects reduce hydration propensity: with alkyl halide groups alpha to the oxyanion, withdrawing electron density and lowering hydration propensity. Resonance and intramolecular hydrogen bonding were also observed to play a role, with salicylate salts exhibiting lower hydration propensity than acetate salts. To quantify the impact of charge localization in counterions on salt hydration, we calculated molecular electrostatic potentials (ESP) of common counterions using electron densities obtained via density functional theory (DFT). The molecular ESP is a scalar field representing the electrostatic potential generated by a molecule in three-dimensional space, determined by its geometry and charge distribution (nuclear charges and electron density). Computationally, ESP is evaluated by first calculating the electron density and then determining the potential experienced by a positive test charge on a chosen molecular surface. This surface typically corresponds to the van der Waals envelope, defined by an isosurface of low electron density (0.002 e au⁻³ in our study). The maximum and minimum ESP values on this surface indicate the points of strongest interaction with negative and positive charges, respectively. In other words, these values quantify the strongest intermolecular interactions a counterion can engage in with a charge or dipole. Our goal was to correlate these interaction strengths with hydration propensity and explore how counterion structure influences ESP and, consequently, hydration behavior. Figure 3 shows ESP maps for representative ions, along with the minimum (for anions) or maximum (for cations) ESP values in kJ mol⁻¹, typically located on the atom bearing the formal charge. We examined how structural factors affect charge localization and ESP values across five ion types. For monoatomic ions, two variables were considered: charge and size. For molecular ions, four variables were explored: charge, electron-withdrawing and electron-donating groups (EWG and EDG), resonance, and intramolecular hydrogen bonding (intra-HB). The progression from highly localized to increasingly delocalized charges is illustrated in Fig. 3.Fig. 3Effects of molecular substituents on charge localization.Influence of counterion structural variations on charge localization on a anions and b cations. Ions at the top of each column have the highest charge localization, while those below show progressively lower localization. Electrostatic potential (ESP) values are reported in kJ mol⁻¹. Abbreviations: EDG electron-donating group, EWG electron-withdrawing group, intra-HB intramolecular hydrogen bond. Effects of molecular substituents on charge localization. Influence of counterion structural variations on charge localization on a anions and b cations. Ions at the top of each column have the highest charge localization, while those below show progressively lower localization. Electrostatic potential (ESP) values are reported in kJ mol⁻¹. Abbreviations: EDG electron-donating group, EWG electron-withdrawing group, intra-HB intramolecular hydrogen bond. For monoatomic ions (halides and alkali metals), reducing the formal charge from +2 to +1 (Ca²⁺ → Na⁺) decreases ESPmax by 725 kJ mol⁻¹ (1885 → 1126 kJ mol⁻¹). Increasing ionic radius also reduces the absolute value of ESP minimum or maximum, as seen in Na⁺ → K⁺ (decrease of 270 kJ mol⁻¹) and Cl⁻ → I⁻ (decrease of 96 kJ mol⁻¹). Thus, charge magnitude has the strongest effect on charge localization in monoatomic ions, followed by ion size. For polyatomic ions (sulfonates, carboxylates, ammoniums), removing charge also reduces ESP but less dramatically than in monoatomic ions. For example, SO₄²⁻ → MeSO₃⁻ lowers |ESPmin| by 461 kJ mol⁻¹, while piperazinium²⁺ → ammonium⁺ lowers |ESPmax| by 265 kJ mol⁻¹. Additional delocalization mechanisms further reduce the absolute value of ESP min or max: resonance with an aromatic ring decreases |ESPmin| by 45 kJ mol⁻¹ (acetate → benzoate), and intramolecular hydrogen bonding reduces |ESPmin| by another 51 kJ mol⁻¹ (benzoate → salicylate). Inductive effects, though smaller, accumulate. Adding fluorine substituents to mesylate reduces |ESPmin| by ~22 kJ mol⁻¹ per substituent, totaling ~67 kJ mol⁻¹ for three of them (mesylate → triflate). For ammoniums, electron-donating groups have a similar effect: replacing a proton with a methyl group reduces |ESPmax| by ~50 kJ mol⁻¹ per substitution (ammonium → dimethylammonium). The length of the alkyl substituent also affects the |ESPmax| with longer alkyl chains stabilizing the charge more effectively. Overall, counterions with divalent charge (Ca²⁺, piperazinium²⁺, sulfate²⁻) have the highest absolute ESP values (1000–2000 kJ mol⁻¹), followed by monovalent cations (~700–1200 kJ mol⁻¹) and monovalent anions (ESP ~ 500–700 kJ mol⁻¹). Because |ESPmin| and |ESPmax| reflect the strength of potential intermolecular interactions between counterions and their environment, we investigated whether these descriptors can predict the propensity of salts to hydrate47–49. Each salt in our dataset consists of a common counterion paired with an ionized counterpart, meaning each crystal contains both an |ESPmin | (from the anion) and |ESPmax | (from the cation). Smaller ions (typically corresponding to the common counterion types identified in Fig. 2) tend to bear the most extreme |ESPmin| and |ESPmax| values. Although our approach considers only one counterion when assessing hydration propensity, this simplification is justified. The common counterion is generally smaller, carries more localized charge, and has more extreme electrostatic potential, making it the primary driver of hydration behavior in molecular salts. Furthermore, when designing salts, the selection is often confined to the choice of counterion. Therefore, a focus on counterion behavior allows for proactive design through a considered screening process. Analysis of hydration statistics for molecular salts in the CSD revealed strong linear correlations between the probability of hydration (Pw) and the logarithm of the absolute ESP extrema, log|\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{ESP}}_{\min }$$\end{document}ESPmin| or log|\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{ESP}}_{\max }$$\end{document}ESPmax | , for common counterions (S3). The analysis was further refined by grouping counterions into families that share common geometries and charge distributions (e.g., sulfonates possess three partially charged oxygen atoms in a tetrahedral geometry, whereas carboxylates have two in a planar geometry). Inorganic anions and cations were grouped together, as they show similar behavior. This classification sharpens the overall trends, yielding strong linear correlations (R2 > 0.9) between Pw and log|ESPmin or max| across all families (Fig. 4).Fig. 4Correlation between electrostatic potential (ESP) extrema and hydration propensity.Relationship between ESP extrema and Pw for salts containing common counterions grouped by family: a carboxylate anions, b sulfonyl anions, c inorganic ions, and d ammonium cations. Selected counterions are labeled for illustration; full datasets are provided in Table S3. Correlation between electrostatic potential (ESP) extrema and hydration propensity. Relationship between ESP extrema and Pw for salts containing common counterions grouped by family: a carboxylate anions, b sulfonyl anions, c inorganic ions, and d ammonium cations. Selected counterions are labeled for illustration; full datasets are provided in Table S3. Comparison of the slopes reveals differences between families: sulfonyl anions (240) > ammonium cations (134) > inorganic ions (113) > carboxylates (101). The steeper slope observed for sulfonyl anions might result from the way the charge is distributed. In sulfonates, the ESP minima are present at three sites between the oxygens clustered around the sulfur. These multiple minima are all localized on one side of the ion, with the organic moiety on the other. Like sulfonyl anions, carboxylates also have their charge distributed on one side with an organic moiety on the other. However, because there is only a single ESP minimum located between the two oxygens, fewer interactions are required and result in a lower slope. In contrast, inorganic ions tend to have more uniform charge distributions across their surface, resulting in a greater area that can be interacted with and stabilized since they have no organic moiety. Within each counterion family, Pw decreases as the magnitude of the ESP extrema decreases, regardless of the mechanism of charge delocalization (e.g., ionic charge, ionic radius, electron-withdrawing or donating groups, intramolecular hydrogen bonding, or resonance effects). The linear relationships identified in Fig. 4a–d allow the hydration propensities for counterions with limited representation in the CSD to be predicted. We compiled a variety of acids and bases together with their acid strengths and predicted salt hydration propensities (Tables 1 and 2). These counterions were then ranked by predicted Pw (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred). For anions (Table 1), edisylic acid, sulfuric acid, and carbonic acid exhibit the highest predicted hydration propensities (~ 50%), whereas PF6-, BF4-, bistriflimide-, tribromide-, and triiodide- all show much lower hydration propensities (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred < 12%). For cations (Table 2), Mg²⁺ and Ca²⁺ are predicted to have the highest hydration propensity (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred > 70%), while diisopropylammonium, triethylammonium, and dicylohexylmmonium are among the lowest (Pw < 10%). A more complete table with additional acids and bases with their predicted hydration propensities can be found in tables S1 and S2.Table 1Collection of acids and their acid strengths (pKa), together with the corresponding anions, their family classification (fam), number of CSD observations (NCSD), computed ESPmin values, and predicted salt hydration propensities (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred)Acid\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p{K}_{a}$$\end{document}pKa [HA]AnionFamNCSDESPMin (kJ mol−1)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred(%)Edisylic acid−2.2, −1.5*Edisylate2-S22−82152Sulfuric acid−3.0, 1.9*Sulfate2-I434−106546Carbonic acid3.8, 10.3*Carbonate2-C57−116146H2SiF60.2SiF62-I158−102544Armstrong acid<1Armstrong2-S165−74242Oxalic acid1.2, 3.5Oxalate2-C205−101340Succinic acid4.2, 5.6Succinate2-C76−92736Fumaric acid3.0, 4.5Fumarate2-C95−92135Terephthalic acid3.5, 4.3Terephthalate2-C133−86132HF3.2Fluoride-I189−78431Acetic acid4.8Acetate-C223−69823Nitric acid−1.4Nitrate-I1099−66923Mesylic acid−1.9Mesylate-S207−60120Benzoic acid4.2Benzoate-C143−65320HCl−5.954Chloride-I7707−63220Nicotinic acid2.0Nicotinate-C37−63819HBr−8.854Bromide-I3791−60318Salicylic acid3.0Salicylate-C148−60217Perchloric acid−15.2Perchlorate-I1722−572163,5-Dinitrobenzoic acid2.73,5-Dinitrobenzoate-C162−57715Malonic acid2.8, 5.7Malonate-C103−57515HI−9.554Iodide-I2872−56715Tosylic acid−1.3Tosylate-S511−56914Trifluoroacetic acid−0.25TFA-C369−56614γ-Resorcylic acid1.1γ-Resorcylate-C56−56414Maleic acid1.9, 6.2Maleate-C215−56014Picric acid0.42Picrate-CSD595−56612Triflic acid−14.7Triflate-S1692−53512HPF6−2055PF6-I1415−51811HBr + Br2−8.854Tribromide-I130−4504Bistriflimide−10.2Bistriflimide-S207−5032HI + I2−9.554Triiodide-I107−409~0The equation used to calculate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred depends on the anion family classification, where C = carboxylates, S = sulfonyl anions, I = inorganic anions, and CSD means the Pw was obtained directly from the CSD. The table is ranked by anions from highest to lowest predicted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred. Where multiple acidic groups are present, multiple pKa values are reported, with the corresponding ESP and hydration propensity calculated for the specific anion shown.Note: pKa values were taken from various sources. Unless otherwise stated, pKa values were obtained from the National Institute of Health (NIH). (*) pKa was calculated using ChemAxon’s chemicalize, (‡) pka was taken from DrugBank.Table 2Collection of bases and the acid strengths of their conjugated protonated base (pKa), together with the corresponding cations, their family classification (fam), number of CSD observations (NCSD), computed ESPmax values, and predicted salt hydration propensities (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred)Base\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p{K}_{a}$$\end{document}pKa [BH + ]CationFamNCSDESPMax (kJ mol−1)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred(%)Mg2+(OH-)214.8Magnesium2+I266209279Ca2+(OH-)214.8Calcium2+I215190175Na+OH-14.8Sodium+I804114149Ethylenediamine7.5, 10.7Ethylenediaminium2+A160112938Piperazine5.6, 9.8Piperazinium2+A196111037K+OH-14.8Potassium+I101786736Ammonia9.3Ammonium+I79985431Rb+OH-14.8Rb+I14378431Cs+OH-14.8Cs+I22170626DMAP9.6DMAP+CSD16563125DABCO3.0, 8.8DABCO2+A7681119Dimethylamine10.6Dimethylammonium+A16974815Imidazole7.0Imidazolium+P14272215Benzylamine9.8Benzylammonium+A12773814Morpholine8.5Morpholinium+A10773413T-butylamine10.7T-butylammonium+A10273213Cyclohexamine10.6Cyclohexammonium+A92723131-PEA9.591-PEA+A34672212Trimethylammonia9.7Trimethylammonium+A7171412Pyridine5.2Pyridinium+CSD18271012Piperidine11.1Piperidinium+A9471011Diethylammonia11.0Diethylammonium+A10770711Diisopropylamine11.1Diisopropylammonium+A736759Triethylamine10.8Triethylammonium+A2636648Dicyclohexylamine10.4Dicylohexylammonium+A1496496The equation used to calculate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred depends on the cation family classification, where I = inorganic anions, A = ammonium cations, and CSD means the Pw was obtained directly from the CSD. The table is ranked by anions from highest to lowest predicted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred. Where multiple basic groups are present, multiple pKa values are reported for the conjugated bases, with the corresponding ESP and hydration propensity calculated for the specific cation shown (1-PEA: 1-phenylethylamine).Note: pKa values were taken from various sources. Unless otherwise stated, pKa values were obtained from the National Institute of Health (NIH). In hydroxides, water is the protonated base (with pKa = 14.8) and the metal acts as an observer in the deprotonation and salt formation. Table 1Collection of acids and their acid strengths (pKa), together with the corresponding anions, their family classification (fam), number of CSD observations (NCSD), computed ESPmin values, and predicted salt hydration propensities (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred)Acid\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p{K}_{a}$$\end{document}pKa [HA]AnionFamNCSDESPMin (kJ mol−1)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred(%)Edisylic acid−2.2, −1.5*Edisylate2-S22−82152Sulfuric acid−3.0, 1.9*Sulfate2-I434−106546Carbonic acid3.8, 10.3*Carbonate2-C57−116146H2SiF60.2SiF62-I158−102544Armstrong acid<1Armstrong2-S165−74242Oxalic acid1.2, 3.5Oxalate2-C205−101340Succinic acid4.2, 5.6Succinate2-C76−92736Fumaric acid3.0, 4.5Fumarate2-C95−92135Terephthalic acid3.5, 4.3Terephthalate2-C133−86132HF3.2Fluoride-I189−78431Acetic acid4.8Acetate-C223−69823Nitric acid−1.4Nitrate-I1099−66923Mesylic acid−1.9Mesylate-S207−60120Benzoic acid4.2Benzoate-C143−65320HCl−5.954Chloride-I7707−63220Nicotinic acid2.0Nicotinate-C37−63819HBr−8.854Bromide-I3791−60318Salicylic acid3.0Salicylate-C148−60217Perchloric acid−15.2Perchlorate-I1722−572163,5-Dinitrobenzoic acid2.73,5-Dinitrobenzoate-C162−57715Malonic acid2.8, 5.7Malonate-C103−57515HI−9.554Iodide-I2872−56715Tosylic acid−1.3Tosylate-S511−56914Trifluoroacetic acid−0.25TFA-C369−56614γ-Resorcylic acid1.1γ-Resorcylate-C56−56414Maleic acid1.9, 6.2Maleate-C215−56014Picric acid0.42Picrate-CSD595−56612Triflic acid−14.7Triflate-S1692−53512HPF6−2055PF6-I1415−51811HBr + Br2−8.854Tribromide-I130−4504Bistriflimide−10.2Bistriflimide-S207−5032HI + I2−9.554Triiodide-I107−409~0The equation used to calculate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred depends on the anion family classification, where C = carboxylates, S = sulfonyl anions, I = inorganic anions, and CSD means the Pw was obtained directly from the CSD. The table is ranked by anions from highest to lowest predicted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred. Where multiple acidic groups are present, multiple pKa values are reported, with the corresponding ESP and hydration propensity calculated for the specific anion shown.Note: pKa values were taken from various sources. Unless otherwise stated, pKa values were obtained from the National Institute of Health (NIH). (*) pKa was calculated using ChemAxon’s chemicalize, (‡) pka was taken from DrugBank. Collection of acids and their acid strengths (pKa), together with the corresponding anions, their family classification (fam), number of CSD observations (NCSD), computed ESPmin values, and predicted salt hydration propensities (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred) The equation used to calculate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred depends on the anion family classification, where C = carboxylates, S = sulfonyl anions, I = inorganic anions, and CSD means the Pw was obtained directly from the CSD. The table is ranked by anions from highest to lowest predicted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred. Where multiple acidic groups are present, multiple pKa values are reported, with the corresponding ESP and hydration propensity calculated for the specific anion shown. Note: pKa values were taken from various sources. Unless otherwise stated, pKa values were obtained from the National Institute of Health (NIH). (*) pKa was calculated using ChemAxon’s chemicalize, (‡) pka was taken from DrugBank. Table 2Collection of bases and the acid strengths of their conjugated protonated base (pKa), together with the corresponding cations, their family classification (fam), number of CSD observations (NCSD), computed ESPmax values, and predicted salt hydration propensities (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred)Base\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p{K}_{a}$$\end{document}pKa [BH + ]CationFamNCSDESPMax (kJ mol−1)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred(%)Mg2+(OH-)214.8Magnesium2+I266209279Ca2+(OH-)214.8Calcium2+I215190175Na+OH-14.8Sodium+I804114149Ethylenediamine7.5, 10.7Ethylenediaminium2+A160112938Piperazine5.6, 9.8Piperazinium2+A196111037K+OH-14.8Potassium+I101786736Ammonia9.3Ammonium+I79985431Rb+OH-14.8Rb+I14378431Cs+OH-14.8Cs+I22170626DMAP9.6DMAP+CSD16563125DABCO3.0, 8.8DABCO2+A7681119Dimethylamine10.6Dimethylammonium+A16974815Imidazole7.0Imidazolium+P14272215Benzylamine9.8Benzylammonium+A12773814Morpholine8.5Morpholinium+A10773413T-butylamine10.7T-butylammonium+A10273213Cyclohexamine10.6Cyclohexammonium+A92723131-PEA9.591-PEA+A34672212Trimethylammonia9.7Trimethylammonium+A7171412Pyridine5.2Pyridinium+CSD18271012Piperidine11.1Piperidinium+A9471011Diethylammonia11.0Diethylammonium+A10770711Diisopropylamine11.1Diisopropylammonium+A736759Triethylamine10.8Triethylammonium+A2636648Dicyclohexylamine10.4Dicylohexylammonium+A1496496The equation used to calculate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred depends on the cation family classification, where I = inorganic anions, A = ammonium cations, and CSD means the Pw was obtained directly from the CSD. The table is ranked by anions from highest to lowest predicted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred. Where multiple basic groups are present, multiple pKa values are reported for the conjugated bases, with the corresponding ESP and hydration propensity calculated for the specific cation shown (1-PEA: 1-phenylethylamine).Note: pKa values were taken from various sources. Unless otherwise stated, pKa values were obtained from the National Institute of Health (NIH). In hydroxides, water is the protonated base (with pKa = 14.8) and the metal acts as an observer in the deprotonation and salt formation. Collection of bases and the acid strengths of their conjugated protonated base (pKa), together with the corresponding cations, their family classification (fam), number of CSD observations (NCSD), computed ESPmax values, and predicted salt hydration propensities (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred) The equation used to calculate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred depends on the cation family classification, where I = inorganic anions, A = ammonium cations, and CSD means the Pw was obtained directly from the CSD. The table is ranked by anions from highest to lowest predicted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred. Where multiple basic groups are present, multiple pKa values are reported for the conjugated bases, with the corresponding ESP and hydration propensity calculated for the specific cation shown (1-PEA: 1-phenylethylamine). Note: pKa values were taken from various sources. Unless otherwise stated, pKa values were obtained from the National Institute of Health (NIH). In hydroxides, water is the protonated base (with pKa = 14.8) and the metal acts as an observer in the deprotonation and salt formation. Tables 1 and 2 are a resource to guide salt design. Table 1 compiles acidic compounds together with their corresponding anions, to design salts of basic AMIs (the most common case for pharmaceutical APIs), while Table 2 compiles basic compounds with their corresponding cations, for the design of salts for acidic AMIs. An initial screening for suitable acids or bases can be performed based on pKa (ensuring salt formation). Once a list of prospective acids or bases has been identified, considering \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred can help to screen a variety of hydration propensities rationally. To illustrate the practical application of our predictive framework, we performed a salt screening experiment on a basic molecule with unknown salt formation behavior: 4-aminoacetanilide (4-AM, Fig. 5). The pKa of the protonated compound was estimated to be 5. Four acids were selected, HCl, HBr, HI, and γ-resorcylic acid, based on their availability and required acid strength (pKa ≤ 1). 4-AM’s weak basicity limited the number of viable acids, as many carboxylic acids are not strong enough to protonate the primary amine. All choices ensured ΔpKa > 4, satisfying the proton transfer criterion for salt formation. Predicted hydration propensities for the corresponding anions were 20% (Cl⁻), 18% (Br⁻), 15% (I⁻), and 14% (γ-resorcylate), enabling rational counterion selection. Crystallization was carried out from isopropanol: water mixtures (3:2 v/v) under controlled cooling conditions, yielding good-quality single crystals for all four salts. Single-crystal X-ray diffraction confirmed salt formation in all cases (Figs. S4–S7) as anticipated by the ΔpKa rule. Excitingly, the two salts predicted to have the highest hydration propensities (chloride and bromide) crystallized as hydrates, whereas the two salts predicted to have the lowest propensity (iodide and γ-resorcylate) crystallized as anhydrous. While these four salts represent only a portion of every possible salt (and conceivably could have been all anhydrous), it demonstrates how the framework can be applied. If the hydrated bromide salt was obtained first, it would suggest going lower in propensity (as done with iodide and γ-resorcylate) to identify a potential anhydrous salt form.Fig. 5Experimental demonstration of ESP-based hydration probability predictions and their use for salt design.The compound 4-aminoacetanilide (4-AM) is weakly basic with pKa of 5, and unknown salt formation behavior. We proceeded to design our salt screening by: a using the ΔpKa rule to select acids of the correct strength, and b by exploring salts leading to different hydration propensities. These predictions informed the choice of acids and the experiments. Crystallization from isopropanol: water (3:2 v/v) yielded single crystals whose structures were confirmed by X-ray diffraction. As anticipated, all pairings led to salt formation. Excitingly, the two salts with the two highest hydration probabilities led to hydrated salts, whilst the two salts with the lowest hydration probabilities led to anhydrous salt. Experimental demonstration of ESP-based hydration probability predictions and their use for salt design. The compound 4-aminoacetanilide (4-AM) is weakly basic with pKa of 5, and unknown salt formation behavior. We proceeded to design our salt screening by: a using the ΔpKa rule to select acids of the correct strength, and b by exploring salts leading to different hydration propensities. These predictions informed the choice of acids and the experiments. Crystallization from isopropanol: water (3:2 v/v) yielded single crystals whose structures were confirmed by X-ray diffraction. As anticipated, all pairings led to salt formation. Excitingly, the two salts with the two highest hydration probabilities led to hydrated salts, whilst the two salts with the lowest hydration probabilities led to anhydrous salt. To build on our example of 4-AM and further understand the impact of the AMI structure on salt hydration (beyond that of the smaller common counterion from which \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred is calculated), we examined hydration trends in salts of eight AMIs (Fig. 6). Alongside 4-AM, we selected three basic and four acidic AMIs that form salts with a range of common counterions. We then evaluated whether the predicted salt hydration propensities \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred (calculated from the structure of the common counterion alone), translate into consistent behavior within individual AMI systems.Fig. 6Occurrence of hydrous and anhydrous salts for four basic (left) and four acidic (right) AMIs with multiple reported salts in the CSD.Each bar represents a crystal structure of an AMI paired with a common counterion. For example, quinine salts span from maleate⁻ (lowest yellow bar) to sulfate²⁻ (highest blue bar). Hydrated salts are shown in blue, anhydrous salts in yellow, and counterions exhibiting both hydrated and anhydrous forms in green. 4-AM⁺ refers to salts of 4-aminoacetanilide prepared in the screening experiments. The red line indicates the average \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred of the AMIs based on their ESP-derived values and corresponding hydration equations (ammoniums and carboxylates). Occurrence of hydrous and anhydrous salts for four basic (left) and four acidic (right) AMIs with multiple reported salts in the CSD. Each bar represents a crystal structure of an AMI paired with a common counterion. For example, quinine salts span from maleate⁻ (lowest yellow bar) to sulfate²⁻ (highest blue bar). Hydrated salts are shown in blue, anhydrous salts in yellow, and counterions exhibiting both hydrated and anhydrous forms in green. 4-AM⁺ refers to salts of 4-aminoacetanilide prepared in the screening experiments. The red line indicates the average \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred of the AMIs based on their ESP-derived values and corresponding hydration equations (ammoniums and carboxylates). Across the analyzed examples (Fig. 6), there seems to be a system-dependent hydration threshold in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred, denoted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred},\,*}[{AMI}]$$\end{document}Pwpred,*[AMI], above which hydration of salts generally occurs. Below this limit, predominantly anhydrous salts appear, while those above it tend to be hydrous. While the outcome will also depend on crystallization conditions, this observation is useful, as the formation of an anhydrous salt suggests that salts of counterions with lower values of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred are also likely to be anhydrous. Although exceptions exist and are expected (e.g., the cinnamate salt of metformin), the identified salts follow the ranking of common counterions. While specific interactions between the AMI and the counterion undoubtedly influence hydration outcomes, the relative ordering of salts based on counterion-derived \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred remains congruent with the broader CSD trends. These rankings are effective, as rationalized earlier, because common counterions are typically small and possess localized charges, resulting in more extreme ESP values than those of the larger AMIs typically involved in salt formation. For example, the average \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred}}$$\end{document}Pwpred computed using the ESP extrema of the eight AMIs studied in Fig. 6 is approximately 15% (shown as a red band), which is lower than that calculated from the structure of the most common counterions reported in Tables 1 and 2. It’s also worth noting that many of the AMIs in Fig. 6, are relatively simple and small compared to many AMIs. This means that 15% may be an overestimate of the average AMI hydration propensity. Ions with highly localized charge may be more prone to hydrate formation because they rely on external interactions to stabilize their charge. Strong electrostatic repulsions associated with ionization can be mitigated either internally, through inductive or resonance effects within the counterion, or externally, through solvation. Solvation is a well-established mechanism for stabilizing ionic species, with water being particularly effective due to its high dielectric constant, bidirectional hydrogen bonding ability, and small size41. Consequently, when intrinsic stabilization within the counterion and interionic stabilization from the oppositely charged species are insufficient, hydrate formation becomes favorable, as water effectively screens electrostatic repulsion. While ionization is known to promote hydrate formation, considering intramolecular charge stabilization provides a deeper understanding of how charge distribution governs hydration in molecular salts. This insight enables rational counterion selection. Counterions with greater intrinsic charge stabilization can suppress hydration, whereas those with less stabilized charge favor hydrous forms, as demonstrated for 4-AM. Although counterion selection offers significant control over hydration propensity, the threshold \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{w}^{{pred},\,*}[{AMI}]$$\end{document}Pwpred,*[AMI] remains system-dependent and relies on the interaction between the AMI and the counterion. Exceptions are therefore expected. While complete control over crystallization remains an aspiration, modeling charge density in counterions provides a framework for predicting hydration behavior in molecular salts. Future work exploring the interplay between AMI structure and counterion properties may reveal further insights into the hydration of molecular salts. Molecular salts hydrate more readily than non-ionized crystals, yet the precise role of electrostatics in driving this behavior has remained unclear. By analyzing over 31,000 crystal structures and modeling electron density, we reveal that charge localization governs hydrate formation in molecular salts. This outcome reflects not only the contribution of valence but also inductive effects that shape electron density. We further demonstrate that an averaged view of salt hydration conceals the substantial variability introduced by counterion choice, even within the same family of ions. Predicted hydration propensities span wide ranges (54–1% for anions and 79–6% for cations), providing considerable flexibility to design salts that are either resistant or deliberately prone to hydration. Combined with the ΔpKa rule, these insights offer guidance to advance the development of crystalline salts with tailored properties. Future studies will explore how specific interactions between paired ions influence hydration behavior, refining this framework and enabling greater control over salt design. Our findings provide a predictive basis for counterion selection and have the potential to transform salt design across multiple industries, including pharmaceuticals, where hydration propensity can be anticipated and directed. Methods CSD searches All searches of the CSD were carried out using version 5.45 (June 2024) of the CSD through ConQuest, applying the ‘best R-factor’ subset with the following filters enabled: 3D coordinates determined, no errors, single-crystal only, not polymeric, and organics only50. By searching within the best R-factor subset, we remove redeterminations while retaining polymorphs of salts and salt hydrates. For all the dataset searches, the allowed elements were restricted to C, N, O, S, B, H/D, P, Si, the alkali and alkaline earth metals, and halides. Within this general search, all suitable counterions (those with no intermolecular hydrogen bond donors and >100 unique entries) were identified using the CSD Python API (Active Programming Interface) using Simplified Molecular Input Line Entry System (SMILES) strings. SMILES representations for all components were extracted for each entry in the subset, and the occurrence of each unique structure was tallied across refcodes. Multiple instances of the same structure within a single refcode (e.g., two acetate anions within the same structure) were treated as a single occurrence. Overall, 71 counterions were identified, 49 of which met the inclusion criteria. Once a counterion was identified by SMILES string, it was searched for in ConQuest within the subset mentioned above. All counterions were searched for by molecular structure alone, except maleate and fumarate, which were also distinguished by text search. Overall, this represents a comprehensive screening of the most common counterions in the organic CSD. Once the 49 counterions were retrieved, they were classified into families. The anions consisted of carboxylates and sulfonyl-containing ions; the cations consisted of 1˚, 2˚, and 3˚ ammoniums, while the inorganics are unique in that they span both charges. Overall, this left 3ions uncategorized (picrate-, DMAP+, and pyridine+) while the remaining 46 belong to larger families of ions. The plot of ESP versus Pw showing all counterions together (including picrate-, DMAP+, and pyridine+) can be found in Figure S3. Not every refcode has retrievable SMILES strings. This means that the number of structures identified by ConQuest and SMILES strings is expected to differ slightly. All refcodes are retrievable directly from the CSD, and the refcodes used in our datasets are available in the repository under the Data Availability section. ESP calculations All ESP charges were calculated in Spartan20 software51. The ωB97X-D functional and 6-311 G + G(2df,2p) basis set were used to optimize the ion’s geometry at the ground state first, and then to compute the ESP plot52. All ESP plots were calculated on a standard 0.002 e au−3 molecular surface, with the resolution set to high. The molecules were initially generated through the input of SMILES strings. Riley et al. found the quality of ESP plots to be dependent on functional selection, but somewhat insensitive to the basis set, provided the basis set had polarization functions on atoms heavier than hydrogen47. Mehta and Martin identified the ωB97X-D functional to be one of the best-performing hybrid functionals for calculating electrostatic potentials, and the 6-311 G + G(2df,2p) basis set offers a good balance of computational cost and accuracy53. Further information on the calculations can be found in the ESI. Just as Riley et al. had concluded, changing the basis set made little difference to the computed ESP minima and maxima. For example, the ESP minimum for maleate was −559.96 kJ/mol with the 6-311 G + G(2df,2p) basis set and −558.51 kJ/mol using the Def2-TZVPPD basis set, whereas for chloride the values were −632.1 kJ/mol and −633.0 kJ/mol. Further comparisons of the methods are presented in Table S4. Crystallization experiments All reagents were purchased from Tokyo Chemical Industry (TCI) and used as received without further purification. Crystallizations were carried out in a mixed solvent of isopropanol and water (3:2 v/v). For each experiment, the selected acid was combined in a 1:1 molar ratio with 0.1 g (0.67 mmol) of 4-aminoacetanilide. The solvent mixture (isopropanol/water, 3:2 v/v) was then added to each vial, 3 mL for the chloride salt and 4 mL for the bromide, iodide, and γ-resorcylate salts. The vials were capped and placed on a hotplate and heated to 45 °C with 800 RPM stirring for approximately 1 h, until the solutions became clear with a faint pink hue. The stir bars were removed, and the vials were recapped. The hotplate was then set to 22 °C with the vials left undisturbed overnight, during which crystals formed. Single crystal X-ray diffraction (SCXRD) SCXRD was conducted for the structural elucidation of the chloride, bromide, iodide, and γ-resorcylate salt of 4-aminoacetanilinide. All salts of 4-aminoacetanilinide were collected at a temperature of 120.0(2) K using MoKα radiation (λ = 0.71073 Å) on a Bruker D8 Venture with a Photon III MM C7 or C14 CPAD detector, IμS-III-microsource, focusing mirrors diffractometer equipped with a Cryostream (Oxford Cryosystems 700) open-flow nitrogen cryostat. All hydrogens on heteroatoms were placed by difference map. Supplementary information Supplementary Information Transparent Peer Review file Supplementary Information Transparent Peer Review file