Introduction

The term “solubility” is a basic concept for substances to describe their compatibility with specific solvents and defined as the maximum amount of solute that can dissolve in a given amount of solvent at a certain temperature1. This property is significant in the field of pharmacy2,3, pigment4, environmental predictions5,6, agrochemical design7, protein folding8,9, and so on. In most cases, the simple and empirical rule “like dissolves like” can direct us to qualitatively estimate the compatibility between solutes and solvents10, which means substances with similar chemical characteristics will dissolve in each other.

Among these, aqueous solubility is of fundamental interest to both academia and industry11,12. This is mainly attributed to the following reasons: (1) water is the most conveniently available solvent; (2) green chemistry concern; (3) the characteristic hydrophobic effect; (4) water constitutes the basis of media for a biological system. However, most existing chemicals have undesirable compatibility with water and many approaches have been developed to enhance the hydrophilicity or aqueous solubility of specific substances2. For instance, the cosolvent method that adding an organic solvent to the aqueous solution is one of the most common and effective ways for solubility enhancement13; the formation of hydrochloride as salt for targeted substance is widely applicable in pharmaceuticals to fulfill biological compatibility14; the surfactant or host structures (e.g., cyclodextrin, capsule, coordination cage) as hydrotropic agent can transfer hydrophobic substances to water phase through forming inclusion complex12,15,16,17,18. In addition, chemists can use their synthetic toolbox to modify molecules with functional groups to achieve aqueous solubility enhancement purpose (e.g., decorating hydrophilic polyethylene glycol long-chain on molecular skeleton)19.

Polycyclic aromatic hydrocarbons (PAH) widely exist in chemical libraries and found applications in sensing, luminescence, electrochemistry, material science, etc. 20,21,22,23. Moreover, the B/N heteroatoms are often doped to PAH skeleton to tune their electronic structures and related physicochemical properties24,25,26, seldom focusing on the influence of aqueous solubility considering most PAH are superhydrophobic and insoluble in water due to their nonpolar and lipophilic nature27. Nevertheless, the subcomponents of PAH with N/O do** show entirely different hydrophilic performances. As shown in Fig. 1a, compared with benzene possessing poor aqueous solubility, the corresponding N-doped pyridine and pyrazine are freely soluble in water. N/O heteroatoms do** to the cyclopentadiene skeleton also greatly enhance the aqueous solubility. For medium-sized molecules (diphenyl, phenanthrene), we can still observe an extent of aqueous solubility improvement derived from the doped N atom on the panels. Along this line, we intend to explore this trend and propose a useful strategy that properly do** N/O heteroatoms on the polycyclic aromatic skeletons with relatively large sizes may greatly improve their hydrophilicity and aqueous solubility.

Fig. 1: The evolution of N/O do** from simplicity to complexity.
figure 1

a Aqueous solubility of simply common aromatic hydrocarbons and their corresponding N/O do** analogs at 25 °C (data in parentheses was obtained from hazardous substances data bank). b Molecular structure design via diversified aromatic polycycles fusion. c Combination of programmed N/O do** towards pristine C1 structure (highlighted molecules demonstrate excellent aqueous solubility).

Full size image

To verify our hypothesis, we chose a polycyclic aromatic skeleton (C1: 2-phenyl-1H-cyclophen[l]phenanthrene28) with more complexity as modeling compound (Fig. 1b). Two main reasons lead to this selection: (1) C1 fuses with 5-membered, 6-membered and condensed aromatic ring on its skeleton, which seems like a molecular splicing process using all subcomponents summarized in Fig. 1a; (2) ease to synthesize N/O doped analogs with only one or two-steps workup procedure. Eventually, 12 combinatorial molecules (C2-C12) with variable N/O do** numbers and positions were designed and successfully synthesized (presented in Fig. 1c). Compared with pristine C1, the hydrophilicity of these derived molecules is mostly improved. To our delight, three molecules with optimized N/O do** show excellent aqueous solubility (C10-C12). Furthermore, these water-soluble molecules underwent an aggregation process and form nanoparticles in aqueous media to facilitate the hydrogen bonding interactions, which results in a series of unexpected phenomena including NMR signals change, luminescence shift, and others. These results greatly support our anticipation at the beginning for hydrophilic properties improvement purposes.

Results and discussion

Synthesis and solubility parameters

We have employed various diones and aldehydes as starting materials to successfully obtain the desired N/O-doped structures as demonstrated in Fig. 1c. These molecules can be classified into two groups with solely N do** (group 1) and dual N/O do** (group 2). The one-step procedure of condensation reactions between diones and aldehydes in acetic acid directly gave rise to C2-C6. For C7-C12, an additional step of reducing one ketone to imine under ammonia gas atmosphere was required. Thus, the half-reduced intermediate can react with aldehydes to give the target molecules. These synthesized molecules have been fully characterized by 1H/13C NMR spectroscopy and HRESI-MS spectrometry (Supplementary Methods 1.1 and 1.2).

With the 12 combinatorial molecules in hand, we initially examined their hydrophilicity and aqueous solubility. The water-contact angle test gave a clear hydrophilic assessment for these molecules (Table 1 and Supplementary Table 1). The conventional shake-flask method was employed to obtain the thermodynamic aqueous solubility data29. Although we did not synthesize the model molecule C1, the nonpolar aromatic hydrocarbon backbone reflects its superhydrophobic nature and neglectable aqueous solubility. In group 1, C2 holding two N atoms on the imidazole subcomponent seems to remain hydrophobicity with the water-contact angle of 140.6°. When two more N atoms are introduced to the phenanthrene panel, C3 (0°) displays a jump from hydrophobicity to hydrophilicity compared with C1 and C2. It’s reasonable that the following C4-C6 with another N atom do** on different positions of terminal benzene ring share the same water contact angle of 0°. It is obvious that with the stepwise do** N atom to different parts of C1, these derived molecules demonstrate improved hydrophilicity. In contrast, C2-C6 all show relatively low aqueous solubility (<0.1 mg/mL), and little improvement is observed as we expected. In pursuit of both improving the hydrophilicity and aqueous solubility, we planned to introduce dual N/O heteroatoms to C1 skeleton as listed in group 2. Initially, when do** N/O atoms to two sides of cyclopentadiene subcomponent to form oxazole ring, C7 remains high hydrophobicity with a 135.2° water contact angle. Additional do** N atom on the terminal benzene ring for C8 slightly reduces the angle to 128.6°, yet is hydrophobic as well. The turning point comes to C9 that do** two N atoms on the phenanthrene subcomponent besides the oxazole part. A water contact angle of 0° indicates the high hydrophilicity of C9. Consequently, the following compounds C10-C12 with more N atom do** on the terminal benzene ring are also highly hydrophilic. The aqueous solubility for group 2 molecules showed distinct results that are different from group 1. C7 and C8 are barely soluble in water due to their high hydrophobicity. C9 is slightly soluble that implies an improving trend. Inspiringly, C10-C12 display good to excellent aqueous solubility with a maximum of 150 mg/mL (C12).

Table 1 Solubility-related parameters of C1–C12.
Full size table

To figure out why the doped N/O heteroatoms on the aromatic skeletons can both influence their hydrophilicity and aqueous solubility, we employed the Hansen solubility parameters (HSP) to interpret this phenomenon. This method developed by Charles Hansen et al.30 has found broad use both in academia and industry for predicting the compatibility or affinity between two substances31,32,33,34. The key elements of HSP approach are three partial solubility parameters consisting of δD, δP, and δH, in which δD represents the dispersion solubility parameter, δP represents the polar solubility parameter, and δH represents the hydrogen bonding solubility parameter, respectively.

The three partial parameters define a three-dimensional coordinate for substances in a virtual solubility space (Hanse space). In principle, the closer the two coordinates of HSP are, the more likely the substances will dissolve in each other, which is consistent with the notion of “like dissolves like” rule. To better describe the degree of closeness, another solubility parameter “distance” (Ra) is introduced as follows:

$${R}_{{\rm {a}}}^{2}=4{({\delta }_{{\rm {D}}2}-{\delta }_{{\rm {D}}1})}^{2}+{({\delta }_{{\rm {P}}2}-{\delta }_{{\rm {P}}1})}^{2}+{({\delta }_{{\rm {H}}2}-{\delta }_{{\rm {H}}1})}^{2}$$
(1)

Meanwhile, each substance has an intrinsic distance parameter R0, together with HSP as a center point that would define a solubility sphere in Hansen space. If the HSP coordinate of one substance is located inside the sphere of another substance, it indicates a high affinity between them. For solute and solvent cases, a dissolution process may happen. Otherwise, a low affinity is inferred if one substance is excluded from the sphere of another.

The total HSP data set of C1-C12 was experimentally tested and optimized from a combination of solvents library and listed in Table 1 (Supplementary Methods 1.3, Supplementary Fig. 1). As is seen, the calculated intrinsic distance parameter R0 is gradually increased along with the extent of N/O do** from C1 to C12, indicative of improved compatibility with more organic solvents. For the unique H2O, there are three sets of HSP for different conditions. The first is the HSP of single-molecule (pure) water (15.5, 16.0, 42.3), in which the δH is more salient than δD and δP, it well explains the strong hydrogen bonding interactions between water molecules. However, its use in the prediction of solubility in water is not deemed appropriate in most cases. The second and third sets of HSP for water are those derived from solubility data and should be chosen case by case. The second set of HSP (15.1, 20.4, 16.5) was derived from experimental data set exceeding 1% soluble compounds in water, which may be appropriate for the diluted solute case (denoted as Water1). The third set of HSP (18.1, 17.1, 16.9) was derived from experimental data set of complete miscible compounds with water, which may be appropriate for the dense solute case (denoted as Water2).

In practice, Water1 and Water2 were both used for C1–C12 to fulfill calculations for comparison. For the group of only N do** (C1–C6), although the distances between designed molecules and H2O (Ra) are considerably reduced, both HSP conditions give the criterion parameter RED >1.0 as the boundary for C1–C3, which is consistent with the experimental water solubility testing results; while Water1 and Water2 give relatively smaller RED below 1.0 for C4–C6, especially under Water2 condition which even generates a low value of 0.5 for C4–C5. Considering the bad water solubility of C4–C6 from experimental results, the derived Ra and RED parameters may be underestimated under Water2 condition. However, the Water1-based analysis still gives a paradoxical prediction of good water compatibility versus the bad experimental water solubility performance, which may be attributed to the complexity of solubility issues, especially in water conditions. For the group of dual N/O do** (C7–C12), Water1 and Water2-based analysis give distinct results in comparison. Under the Water1 condition, all the REDs exceed the boundary value of 1.0, indicative of bad compatibility with water. By contrast, under the Water2 condition, the REDs of C9–C12 are equal or below 1.0 while C7 and C8 still exceed the boundary value that is consistent with the Water1 result. As mentioned above, the two sets of water HSP should be chosen case by case for specific analysis. Considering C7–C9 with bad/poor aqueous solubility and C10–C12 with good to excellent aqueous solubility from the perspective of experimental results, the Water1 parameters shall be chosen for C7–C9 as a dilute solute case, while the Water2 parameters shall be chosen for C10–C12 for the dense solute case. Under this guiding principle, the calculation results of water solvent excluded from C7–C9 solubility sphere and included in C10–C12 solubility sphere using RED as an indicator match well with their experimental aqueous solubility performance. It is noteworthy that although C10–C12 could include H2O in their solubility sphere, the corresponding REDs (0.9–1.0) are still near the boundary condition and should be treated carefully. The following investigations of molecular behaviors in solution and intermolecular hydrogen bonding may provide certain mechanistic insight into the remarkable aqueous solubility performance.

Self-assembly behaviors in aqueous solution

Alternatively, we employed NMR technology to probe the existing state of soluble molecules C10–C12 in aqueous solution. Taking C11 for instance (Fig. 2a), with the increase of water content in MeOD-d4/D2O mixture, all the signals moved upfield and the overlapped peaks split into discrete ones. Finally, 10 distinguishable single peaks appeared corresponding to 10 inequivalent protons on C11 skeleton in pure D2O. The major upfield chemical shifts exceeded 1.00 ppm with a maximum of 1.68 ppm for peak d. These results suggested a strong chemical shielding effect between adjacent aromatic panels in close proximity, which may be caused by molecular aggregation in aqueous solution. Furthermore, the measured kinetic radii from DOSY testing showed a significant increase from 0.65 nm in MeOD-d4 to 3.98 nm in D2O, which again confirmed our anticipation that an aggregation process gradually occurs with the increase of water content (Supplementary Figs. 5, 8 and 11). Similar phenomena also happen for C10 and C12 (Supplementary Figs. 2–4, 6, 7, 9, 10 and 12). In comparison, we selected C9 with relatively poor aqueous solubility (0.9 mg/mL) to demonstrate the changing process under the same operation (Fig. 2b). When aliquots of D2O were introduced to C9 in MeOD-d4, all the signals uniformly moved to upfield until the 3:2 ratio (MeOD-d4/D2O) and a maximal shift of 0.84 ppm assigned to peak c’ were identified, indicating the aggregation process was occurring. However, with the continuous increase of D2O content, there were solids precipitated out of the solution. Up to 1:4 ratio, little C9 existed in the solution and the residual signals moved back to the low field being similar to the pattern of starting state in MeOD-d4. These suggested that C9 failed to form a stable aggregate in an aqueous solution due to its less sufficient hydrophilicity.

Fig. 2: Investigation of self-assembly behaviors in aqueous solution.
figure 2

a Stacked 1H NMR spectra of C11 in MeOD-d4/D2O mixture with variable volume ratios (each proton was assigned and labeled along with different solvent ratios, the kinetic radii of ensembles under each condition were measured by DOSY spectra and calculated by Stokes-Einstein equation). b Stacked 1H NMR spectra of analogous C9 in MeOD-d4/D2O mixture with variable volume ratios. c, AFM analysis of C11 in aqueous condition (a height profile of selected nanoparticle was inserted). d Statistical distribution of C11-based nanoparticle number versus size. e Left: NOE build-up curves for C11 in D2O, right: calculated intermolecular adjacent proton distances with peak d derived from the NOE growth rate analysis using peak g-d correlation as internal reference.

Full size image

To further confirm the aggregation-induced nanoparticles formed in water, we carried out atomic force microscope (AFM) measurement of C10-C12 in wet conditions35 (Fig. 2c and Supplementary Figs. 29–31). Typically, a drop of aqueous solution of C11 (1.5 mM) was cast on a freshly cleaved mica surface and subjected to test. The pristine AFM images showed uniform spherical nanoparticles dispersed in water. 2D and 3D height profiles gave a particle size around 8 nm. Statistical analysis provided sectional size distribution of C11-based nanoparticles with an average diameter of 8.32 nm (Fig. 2d), which is comparable to the DOSY radius (3.98 nm, Supplementary Fig. 5).

At this stage, we could confirm that rather than an expected molecular-level solvated state in solution, these optimized C10–C12 molecules undergo aggregation and self-assembly process to form nanoparticles in aqueous solution. Thus, the above obtained HSPs for C10–C12 could not represent the real conditions for these nanoparticles. Meanwhile, it is unworkable to directly obtain their HSPs neither from conventional experimental method because these nanoparticles only exist in aqueous media but disassemble once dissolve in organic solvents, nor from calculations because of the complex and unknown precise nanostructures. However, the compatibility between molecular C10–C12 and H2O disclosed by HSP analysis is a prerequisite to bringing these molecules into water phase as the first step, thus forming the thermodynamic-favored nanoparticles. This phenomenon also well explains why the C10–C12 holding not small enough but near to boundary condition REDs (0.9–1.0) could show good to excellent aqueous solubility performance.

To gain more insight into the structure and stacking configuration of these nanoparticles, two-dimensional NMR spectroscopy was adopted to in situ investigate this problem. Among versatile NMR technologies, linear nuclear overhauser effect (NOE) growth versus mixing time in a suitable region is used for quantitative distance determination in solution36 and expressed by the following equation:

$${r}_{{{\rm {AB}}}}={r}_{{{\rm {ref}}}}{({\sigma }_{{{\rm {ref}}}}/{\sigma }_{{{\rm {AB}}}})}^{1/6}$$
(2)

where rref is the known interproton distance as reference, σref and σAB are the NOE growth rates for reference and interprotons (A and B), rAB is the unknown distance to be determined.

A series of remote NOE correlations with proton d were specifically observed for C11 in D2O compared with MeOD-d4 condition, indicative of close packing induced spatial proximity in the aggregation state (Supplementary Fig. 11). The intramolecular protons g and d were selected as the reference and the distance was measured to be 4.08 Å from single crystal analysis. Quantification of the NOE growth rates and introducing to Eq. (2) yield a series of distances between proton d and adjacent phenanthroline protons (Fig. 2e). Together with all the calculated values, we can deduce that the distance between stacked molecules is <5 Å within the nanoparticle structure in water. Considering the height of nanoparticle is about 8.0 nm, there are at least 16 layers of molecular panel stacking among the nanostructure. Identically, C10 and C12 share a similar packing mode in an aqueous solution (Supplementary Figs. 15 and 16). Such a large ensemble with a considerable number of planar molecules as subcomponents in water is reminiscent of protein behaviors in solution, which hide their hydrophobic area in the core through winding and folding and expose the hydrophilic surface as exterior to interact with water molecule8. Under the same principle, these nanoparticles could reduce the area of hydrophobic parts exposed to water via close packing and forming efficient hydrogen bonding interactions in the interface as proved by the following crystal structure analysis.

Physicochemical properties of self-assembled nanoparticles

Attempts to obtain the single crystals from water that are suitable for X-ray diffraction analysis only succeeded for C11 through slow evaporation (Supplementary Data2). For C10 and C12, the excellent aqueous solubility may be detrimental to crystal growth. Alternatively, both crystals were obtained under evaporation conditions from methanol and chloroform, respectively (Supplementary Data 1, Data 3, Supplementary Table 2). Taking C11 for instance, displays ideal aromatic planarity from crystal analysis. Remarkably, the hydrogen bonds are ubiquitously existing across the structure. In detail, one C11 molecule closely contacts three H2O via hydrogen bonding interactions (Fig. 3a). Specifically, N2 and N4 atoms as acceptors form relatively strong hydrogen bonds with neighbored water molecules with distances of 3.00 and 2.93 Å and angles of 147.70° and 170.04°, respectively. Relatively weak hydrogen bonds are formed for N1 and O1 with distances of 3.51 and 3.74 Å and angles of 141.13° and 82.48°. In the crystal lattice, the molecular panels adopt a face-to-face packing mode via π-π stacking along the c-axis (Fig. 3a). The adjacent layer distance is measured to be 3.38 Å that is shorter than aqueous aggregation conditions, indicative of a more compact packing manner in the crystalline state. The water molecules show a zigzag distribution along the interlayers to act as hydrogen bonding donors. Similar π–π stacking phenomena are also found for C10 and C12 in their lattices (Supplementary Fig. 32). These widely existing hydrogen bonds in crystalline states provide direct evidence to visualize the intermolecular interactions and emphasize their significance toward hydrophilicity and aqueous solubility.

Fig. 3: Physiochemical properties of optimized N/O do** molecules with excellent aqueous solubility.
figure 3

a Crystal structure of C11 (left: hydrogen bonding connecting mode around one C11 molecule, the neighbored N-O distances forming hydrogen bonding were labeled, right: molecular packing along c-axis, the interlayers distances were labeled). b Apparent association constants for C10-C12 in CHCl3 and H2O were estimated via 1H NMR fitting, respectively. c Comparison of 13C NMR spectra for C10 in different existing states (bottom: MeOD-d4, middle: D2O, up: solid-state 13C MAS). d Left: luminescent emission spectra of C10 upon 365 nm excitation in solution with viable solvent ratios and solid states (inserted is the photograph of the solid sample when excitation), right: photograph of C10 emission upon 365 nm excitation in MeOH and H2O, respectively.

Full size image

As is known, high concentration as a driving force to promote the molecular aggregation in solution is ubiquitous and well recognized37,38. To compare the high concentration induced aggregation phenomenon with our case, we measured their apparent association constants (Ka) by 1H NMR shift analysis after simplifying the aggregates to a dimeric model39 (Supplementary Figs. 17–28). As listed in Fig. 3b, C10-C12 in chloroform display relatively weak association ability induced by high concentration factor. In contrast, their apparent association constants are increased by two orders of magnitude in hydrophilic media, indicative of a more stable and less dissociative ensemble. Meanwhile, the 13C NMR spectra gave more informative details about C10-C12 in different existing states (Fig. 3c, Supplementary Figs. 13 and 14). For instance, C10 displayed characteristic sharp 13C nuclear resonance signals in MeOD-d4. In comparison, these signals became broadened and moved to upfield with a maximal 3.19 ppm shift in D2O conditions, which was almost in consistent with the solid-state 13C magic-angle spinning (MAS) NMR spectrum despite its much broader signature. Moreover, the luminescent emissions of C10-C12 in various solvent ratios and solid-state were also screened (Fig. 3d, Supplementary Figs. 33 and 34). Along with the water content increasing in MeOH/H2O mixture, the emission peak showed a redshift from the ultraviolet region to the blue-green region (e.g., from 393 to 472 nm for C10), which may be caused by competitive nonradiative relaxation process between neighbored molecules in close packing fashion. Unexpectedly, the maximal emissions in water are close to the solid state of C10-C12 (e.g., 472 nm in water versus 486 nm in solid for C10), indicative of a homologous excited-state energy transfer pathway. We shall speculate that these self-assembled nanoparticles have quasi-solid behaviors in water and possess preliminary solid-like properties from NMR and luminescence perspectives.

Conclusions

Inspired by the simple aromatic heterocycles with desired hydrophilicity, we have expanded this trend and developed an efficient strategy of precise do** N/O heteroatoms on a predesigned polycyclic aromatic skeleton to greatly enhance its hydrophilicity and aqueous solubility. A series of analogous N/O do** molecules demonstrate that both properties are closely related to the do** species, numbers, and positions. The HSP calculations depending on two sets of water HSP under dilute and dense solutes conditions can give a considerable prediction of improved hydrophilicity and aqueous solubility along with the stepwise N/O do**, which is consistent with the experimental results on the whole. Specifically, the enhanced δH representing hydrogen bonding factor plays a significant role in improving hydrophilic performance. Unexpectedly, three molecules with optimal N/O do** achieve excellent aqueous solubility via a self-assembly process to form nanoparticles. The single crystal X-ray analysis proves the widely existing hydrogen bonding between N/O heteroatoms and water that contribute much to the aqueous solubility performance. Interestingly, these nanoparticles demonstrate quasi-solid properties in water from NMR and luminescence perspectives, which may derive from compact packing and large size of nanoparticles. In brief, we provide a useful mind to construct hydrophilic and even water-soluble polycyclic aromatics and give insight into the mechanism of their self-assembly behaviors in solution.

Methods

All details on syntheses, solubility-related parameters determination, NMR studies, AFM measurements, single crystal X-ray analysis, and luminescent emission are provided in the Supplementary Information.