Introduction

Separations of petrochemicals are exceedingly important1. Processes of utmost importance include separations of benzene, toluene, ethylbenzene, and xylenes, or the BTEX hydrocarbons1. BTEX compounds are used extensively in manufacturing and experience tremendous global consumption (i.e., >50 M tons/year). The separation of ethylbenzene from xylenes is troublesome owing to similar reactivities and physical properties (bps: o-xylene, 144 °C; m-xylene, 139 °C; p-xylene, 138 °C, ethylbenzene 136 °C). Rectification processes, for example, are energy intensive, using highly controlled and refined distillations. The removal of ethylbenzene in styrene (bp: 145 °C) production, as well as thiophene (bp: 84 °C) from benzene (bp: 80 °C), also suffer from similar separation issues2. Problems of separations are important from economic and sustainability perspectives, representing 10–15% of the world’s energy consumption1. Indeed, chemical manufacturing greatly benefits from separations of petrochemicals owing to widespread applications in materials and health sciences.

An approach to confront separations of petrochemicals is the crystallization of organic molecules3,4,5. Through molecular recognition, a crystallization can achieve a separation with up to a perfect degree of selectivity (e.g., perfect size exclusion). A molecular host can ‘shrink wrap’ around a guest to accommodate specific size, shape, and functionality demands of a targeted guest, with the crystallization preferably occurring at ambient temperatures. Given the various challenges that come with developing hosts to recognize, entrap, and sequester molecules, the identification of methods based on molecular and supramolecular designs to facilitate molecular separations represents an important ongoing problem.

Our interest in the molecule DEPN lies in its ability to force organic functional groups into close proximity to achieve targeted functions (e.g., catalysis, sensing) (Fig. 1)6,7. DEPN originates from a family of U-shaped molecules used in molecular recognition, catalysis, and chemical sensing. We expected the closely separated pyridyl groups of DEPN to coordinate with up to two B-atoms to afford the diboron N → B assembly DBP-DEPN. Steric considerations suggested that the closely spaced and coordinated catecholate groups would be rotated away from the plane of the naphthyl backbone. Inspired by work of Höpfl, the rigid and conjugated structure of DBP-DEPN would exhibit a capacity to allow for inclusion and capture of aromatic-rich hydrocarbons upon crystallization8,9.

Fig. 1: Schemes of chemical structures.
figure 1

DEPN (left) and DBP-DEPN (right).

Results and discussion

Molecular T-pentomino structure and packing

DBP-DEPN was generated through condensation of DEPN, phenylboronic acid, and catechol in dichloromethane. DBP-DEPN crystallizes from CH2Cl2 as thin yellow blades in the monoclinic space P21/c (Fig. 2). DEPN coordinates with two B-atoms of two catecholates (B → N 1.660(4), 1.664(4)) each of which is canted (117.2(1), 156.7(1)o) and disordered over two positions (occupancies: 88:12 and 85:15). The coordinated pyridyl groups are twisted from co-planarity (117.6(1), 135.7(1)o), with the alkene bonds pointing in opposite directions (Fig. 2a). The separation distance of the pyridyls (NN 5.998(4)Å) likely reflects steric strain imparted by the closely spaced catecholates. A result of the assembly process is that DBP-DEPN adopts an overall T-shaped geometry of edge lengths approximately 10.2 Å × and 12.0 Å and thickness of 8.1 Å (Fig. 2b). C-H∙∙∙O hydrogen bonds (CO: 3.448(4), 3.47(1) Å) reinforce the T-structure (Fig. 2c).

Fig. 2: X-ray structure of DBP-DEPN.
figure 2

a Side view of coordinated assembly, b T-shaped geometry, c intramolecular C-H∙∙∙O hydrogen bonds, d 2D packing, e tongue-and-groove stacking with intermolecular C-H∙∙∙O hydrogen bonds, and f voids situated between hosts (orange). Atom colors: O = red, N = blue, B = salmon.

DBP-DEPN self-organizes in the solid state within the crystallographic bc-plane into face-to-face π-stacks (plane-to-plane: 3.699(6)Å) (Fig. 2d). The pyridyls form head-to-tail dimers with the naphthyls, with the dimers being related by 2-fold rotation within the plane. The layers stack along the a-axis in a tongue-in-groove fit, with the layers connected by C-HO hydrogen bonds (3.579(5), 3.74(2) Å) (Fig. 1e). The overall arrangement gives rise to voids filled with CH2Cl2 solvent molecules (224 Å3) that are situated adjacent to the pyridyl and catecholate groups (Fig. 2f). The void volume corresponds to approximately 5% of the total crystal unit cell volume.

Consultation of plane figure and tiling models reveals the T-shaped molecular structure of DBP-DEPN to approximate the shape of T-pentomino (Fig. 3)10. A T-pentomino is a plane figure formed by joining five equal squares edge-to-edge in the shape of a T (Fig. 3a). An n-omino (n = 2, 3, 4, 5…) is a plane figure formed by n squares joined at the edges. The popular video game Tetris® is inspired for n = 4 and involves attempts to achieve intentional close packing of various tetrominoes. There are 12 possible ways in which the edge-sharing of squares can generate a pentomino (Fig. 3b). The naming convention of pentominoes corresponds to omino shape in relation to an alphanumeric letter (e.g., L, T, W, X). A feature common of n-ominoes is a limited ability of a given n-omino to tile a plane by isohedral packing. The resulting tiled structures are packed perfectly, being completely devoid of cavities and holes. Tiling models show the 2D packing of DBP-DEPN to conform to one of the six packings of a T-pentomino in a plane (see Supplementary Fig. 18). Four of the six packings, which are polymorphs11, involve the head-to-tail dimers present in crystalline DBP-DEPN (Fig. 3c). The structure reported here conforms to the denoted ‘2-fold rotated’ polymorph. The remaining three polymorphs involve dimers of T-pentominoes related by either translation (i.e., eclipsed translation, offset translation) in 1-isohedral packing or isolated in 2-isohedral packing12.

Fig. 3: T-pentomino structure and packing.
figure 3

a T-shape of DBP-DEPN and a T-pentomino, b the 12 pentominoes, and c relation of packing of DBP-DEPN to the 2-fold rotated tiling (polymorph). There are four T-pentomino polymorphs that consist of face-to-face stacked dimers. Naming of polymorphs is introduced here.

The 2D packing of T-shaped DBP-DEPN can be further understood from electronic considerations. Using DFT calculations, DBP-DEPN exhibits three electronically differentiated regions: (i) electron-poor pyridyls, (ii) electron-rich catecholates, and (iii) neutral to slightly electron-rich naphthyls (Fig. 4). Stacking of the electronic-rich catecholates, as encountered in the eclipsed translated and offset translated polymorphs, is generally avoided in the 2-fold rotated tiled structure, whilst head-to-tail pyridyl and naphthyl stacking is achieved. The presence of the voids despite the T-shape of DBP-DEPN can be attributed to deviations13 of the overall shape of DBP-DEPN from an ideal T-pentomino. For DBP-DEPN, the length of the vertical line (12.0 Å) that intersects with the T’s horizonal line (10.4 Å) differs by 1.5 Å (i.e., nearly a carbon-carbon bond length). Nevertheless, the ability of DBP-DEPN to tile the plane as a polymorph of a pentomino attests to robustness of the self-assembly process to achieve the prescribed pattern of close packing.

Fig. 4: DFT calculations.
figure 4

Electronically differentiated regions of T-shaped DBP-DEPN.

Inclusion of BTEX hydrocarbons

The tiling and crystal packing of T-shaped DBP-DEPN hosts the BTEX hydrocarbons benzene (B1), toluene (B2), and ethylbenzene (B3) (Fig. 5)1. Specifically, crystallization of DBP-DEPN from solutions of each of B1-B3 afforded long yellow needles. SCXRD determinations demonstrated the formation of BTEX-included solids that are isostructural with solvated DBP-DEPN (Table 1, Fig. 5a). There are significant changes to the shapes of the cavities as realized by contortions of T-shaped DBP-DEPN as reflected by twisting of the pyridyls and catecholates toward 180o versus the solvated structure. The volumes of the cavities shrink up to 17% for B1 versus the larger guests (Table 2). The guest-induced contortions and shrinkage likely reflect domination of attractive interactions between DBP-DEPN and the BTEX guests. Overall, the structural changes are consistent with a packed host with cavities that shrink wrap around the hydrocarbons5.

Fig. 5: DBP-DEPN solid-state behavior.
figure 5

a PXRD data of guest lattices and b 1H NMR data demonstrating selective guest uptake of B3 (asterisks) from xylene mixture.

Table 1 Crystallographic data for DBP-DEPN∙guest
Table 2 Geometric parameters describing of T-shape of DBP-DEPN∙guest

Highly selective inclusion

We have discovered DBP-DEPN to selectively include B3. Crystallization of DBP-DEPN from commercial xylenes (Oakwood Scientific) rapidly afforded yellow blade-like crystals. A 1H NMR spectrum revealed the solid to exclude each of the o-, m-, and p-xylene isomers. Instead, the spectrum showed the presence of only ethylbenzene in a solid of composition DBP-DEPNB3 (Fig. 5b). Commercial xylenes consists of approximately 20% ethylbenzene (B3), along with 40% m-xylene (B4) and up to 20% each of o- and p-xylene (B5-B6). B3 is commonly present in xylenes as a product of catalytic reforming of petroleum or by way of isolation of xylenes from coal tar14. A 1H NMR spectrum of our commercial sample revealed the xylenes to contain 20% of B3. The inclusion of B3 is consistent of DBP-DEPN being selective for B3 uptake from the mixture of xylenes. Crystallizations were also performed in simulated xylene mixtures with decreasing amounts of B3 (i.e., 15%, 10%, 5%). In each case, crystalline DBP-DEPN ∙ B3 exclusively formed when the host was crystallized from the preformed xylene mixtures (m-, o-, p-isomers 2:1:1). Up to 80% of the original host material was recovered in the xylene crystallizations with the recoveries occurring in practical time periods on the order of 10 min and at ambient conditions.

SCXRD and PXRD data revealed the isolated solid DBP-DEPNB3 to be isostructural with the T-pentomino packing arrangement, with B3 being disordered within the cavities. While the lowest R-refinement value was achieved using a solvent mask, the structures of the aromatic ring and ethyl group of B3 were clearly apparent. Selective BTEX guest uptake was also realized when DBP-DEPN was introduced to either thiophene or styrene15. Thiophene is a common impurity in benzene production and styrene is generated from dehydrogenation of B316. Crystallizations of DBP-DEPN from preformed solutions in competition experiments (1:1 molar ratios) containing either B1 and thiophene or B3 and styrene afforded exclusively crystalline DBP-DEPNB1 and DBP-DEPNB3. Each solid, thus, excluded thiophene and styrene, respectively. Up to 84% and 89% of the original host was recovered in the B1-thiophene and B3-styrene crystallizations, respectively. The packing arrangements were isostructural with the T-pentomino lattice. We also note that when competition experiments were performed involving solutions containing two and up to three components of B1-B3, a hierarchy of B3 > B2 > B1 was established that is consistent with the high selectively realized involving the xylenes. On the order of 80% of the host was recovered with the solids exhibiting preferential 6:3:1 uptake of B3:B2:B1. Variabilities of all cell parameters for the solids characterized by SCXRD were on the order 0.30%, which is consistent with the overall 2D packing not being appreciably pliable.

Design and polyomino framework theory

Within the field of host-guest chemistry, DBP-DEPN acts akin to an extrinsically porous organic molecule3. Such molecules are designed to feature unusual molecular and supramolecular structures and conformations that are expected to pack inefficiently17 so as to de facto generate cavities to accommodate guests. The structure of DBP-DEPN differs from those hosts in that the T-shape is a priori biased to exhibit packings that lack pores and cavities. In fact, the 2D packing of DBP-DEPN agrees with the theory of polyomino tiling that states that n-ominoes will exhibit limited packing arrangements that enclose space in two dimensions. Given that the solid DBP-DEPN also entraps molecules, it should now, in principle, be possible to synthesize molecules with shapes that conform to the various n-ominoes and, in line with the properties of DBP-DEPN, develop solids that - counterintuitively - exhibit host-guest behavior. Considerations of perfect 2D packing of organic molecules on surfaces in relation to the presence or absence of voids has been discussed18. For the current case, such materials can be expected to create small pores given the biased nature of an omino to enclose space. The field of n-ominoes is also replete with examples wherein combinations19 of different n-ominoes pack to fill space (vis-a-vis, molecular cocrystals) (Fig. 6). The structures of n-ominoes, effectively, can provide a hitherto undocumented roadmap for the design and discovery of functional organic molecules and host-guest materials.

Fig. 6: Example of tiling of multiple ominoes.
figure 6

A rectangle involving a combination of five tetrominoes (not including mirror images for ‘L’ (blue) and ‘Z’ (green) combinations).

To conclude, we have reported on the diboron assembly DBP-DEPN to exhibit a structure that approximates a T-pentomino. The assembly packs to assume a structure that corresponds to a known polymorph devoid of cavities and holes. The resulting solid selectively entraps BTEX hydrocarbons. We are now exploring the synthesis of molecules with structures that conform to polyominoes, where there will be a focus on how similar a molecule must be to a polyomino for the crystal packing to be realized. It is also intriguing how the single host DBP-DEPN separates different guests selectively from different mixtures. The behavior generally contrasts those porous solids that exhibit fixed channels and pores (e.g., zeolites, metal-organic frameworks)20 and are usually designed for a particular separation. In that way, DBP-DEPN is attractive from sustainability and economic perspectives since needs to develop compositionally new host materials for different guests and separations can be obviated.

Methods

Chemicals

All materials were obtained from commercial sources and used as received unless indicated. Dibromonaphthalene and palladium bis-triphenylphosphine dichloride were purchased from A2B Chem. Catechol, 4-vinyl pyridine, dichloromethane (99%) toluene (99%), and triethylamine were all purchased from Sigma-Aldrich. Xylenes (95%) and ethylbenzene (99%) were purchased from Acros. Phenylboronic acid was purchased from Oakwood chemicals.

DEPN

To a 150 ml round bottom flask containing a degassed (argon, 10 min) solution of 50 ml acetonitrile and 75 ml of triethylamine, 1,8-dibromonaphthalene (20 mmol) was added and stirred until dissolved. Palladium bis-triphenylphosphine dichloride (10 mol%, 0.2 mmol) was added while under argon. After filtering through silica gel to remove polymerized material, 4-vinyl pyridine was added to the reaction mixture and stirred at reflux for 72 h. After TLC (ethyl acetate) showed no bromoarene remaining, the reaction mixture was cooled and rotary evaporated until solid remained. This material was eluted through a flash column using ethyl acetate until the solution was clear instead of yellow. The resulting liquid was again rotary evaporated to yield a solid. Recrystallization from ethyl acetate granted shiny, crystalline material after washing with acetonitrile. Yield: 60.3%, White-yellow crystals.

Assembly formation

DBP-DEPN

DEPN (0.9 mmol), phenyboronic acid (1.82 mmol), and catechol (1.82 mmol) are first ground with drops of dichloromethane/toluene (1:1 solution). The resulting powder was scraped from the grinding apparatus and placed into a 30 ml scintillation vial. The mortar-and-pestle were rinsed with another 4.0 ml (2 × 2.0 ml) of solution dropwise and ground to agitate any attached material, which was then added to the vial. The resulting suspension was heated until the powder had dissolved. Upon cooling, translucent bright yellow crystals appeared within 20 min in near quantitative yield. The resulting powder was oven-dried at 80 °C for a minimum of 4 h. Use of dichloromethane is not necessary for crystallization to proceed; however, it allows for ease of solids separation.

Crystallization experiments

Aromatic guest-free solid was grown in open air from dichloromethane, resulting in a single molecule inclusion species. For each crystallization with a BTEX hydrocarbon, a scintillation vial containing ground DBP-DEPN (0.06–0.08 mmol) was filled with 2.0 ml of BTEX hydrocarbon, gently heated, and allow to cool to room temperature.

Competitive crystallization experiments were performed in the same manner as above but with increased DBP-DEPN (0.13–0.16 mmol). Crystallization of DBP-DEPN with a 1:1 w/w mixture of thiophene and benzene (2.0 ml) following heating resulted in crystals containing only benzene, as determined by 1H NMR spectroscopy. PXRD data matched the host material. The procedure was repeated with both a 1:1 w/w mixture of ethylbenzene and styrene (2.0 ml) as well as a standard mixture (3.0 ml) of commercial xylenes (~ 2(m):1(p):1(o):1(EB)). In addition, competition between guests was undertaken with benzene, toluene, and ethylbenzene being mixed equal parts by mass (0.1 mmol of each, equimolar ratio) with 1:1 ratios for competition between benzene and toluene, benzene and ethylbenzene, toluene and ethylbenzene, and a 1:1:1 ratio of the three together. Crystallizations occurred within 5 min after a 1:20 mass ratio of host to guest was added to the respective solution in a 30 ml scintillation vial and gently heated to ensure dissolution. The crystalline product was removed, and excess solution was allowed to evaporate for 20 min.

Aromatic guest-free (CH2Cl2 solvate)

1H NMR (400 MHz, CDCl3) δ 8.56 (d, J = 5.5 Hz, 4H, Hf), 8.15 (d, J = 4, 16 Hz, 2H, Hh), 7.99 (d, J = 7.4 Hz, 4H, Hj), 7.75–7.73 (m, 4H, Hc), 7.72 (d, J = 3.3 Hz, 2H, Hk),7.49 (d, J = 5.5 Hz, 4H, Hg), 7.37 (m, J = 3.6 Hz, 4H, Hd), 6.98 (d, J = 4, 16 Hz, 2H, Hh), 6.88–6.87 (m, 2H, He), 6.86 (d, J = 2.0 Hz, 4H, Hb), 6.75 (d, J = 2.0 Hz, 4H, Ha), 5.36 (s, 1H).

Solvate characterization performed using a Exeter Analytical CE-440 addressing carbon, hydrogen, and nitrogen. Analysis calculated for C48H36N2O4, C: 79.36 H:4.99 N:3.86. Found: C: 73.44 H: 4.77 N:2.83.

Aromatic guest-free (dried in oven)

1H NMR (400 MHz, CDCl3) δ 8.49 (d, J = 5.5 Hz, 4H, Hf), 8.16 (d, J = 4, 16 Hz, 2H, Hh), 7.89 (dd, 4H,Hc), 7.65 (d, 2H, HJ), 7.52 (t, 2H, Hk), 7.45 (m-overlapped, 4H, Hd), 7.43 (m-overlapped, 2H, HL), 7.41 (m-overlapped, 2H, He), 7.31 (d, J = 5.5 Hz, 4H, Hg), 7.17 (m, J = 1.9 Hz, 4H, Hb), 6.99 (m, J = 1.9 Hz, 4H, Ha), 6.92 (d, J = 4, 16 Hz, 2H, Hi).

Molecular modeling

Electrostatic potential maps were generated for each DBP-DEPN complex by calculations at ground state in gas phase using Spartan ’18 V1.2.0 software. Calculations were performed with ωB97X-D/6-31 + G* using the density functional model in a vacuum. Charge value ranges were set to a standard range of +/− 200 KJ/mole with isovalues of 0.002 for each. Single crystal lattice data was used to produce atom coordinates with hydrogens relaxed to neutron distances. Solvent inclusions removed to limit complexity.

Thermal measurements

Thermogravimetric analysis data were recorded on a TA Instruments TAQ-500.