Introduction

Porous materials comprise a diverse subset of molecular ensembles, such as metal-organic frameworks (MOFs)1,2,3,4, covalent organic frameworks (COFs)5,6,7, hydrogen-bonded organic frameworks (HOFs)8,9,10, and zeolites11,12,13. The selectivities of MOFs, COFs, HOFs, and zeolites in molecular adsorption are determined by the sizes, shapes, and dimensions of the pores that are tightly retained in the crystalline structures. Therefore, porous materials have often been used for nonheat-driven separations in industry14.

Macrocyclic molecules often form porous crystals in which various guest molecules are adsorbed within molecular cavities15,16,17,18,19,20,21. In macrocyclic molecular crystals, porosity-without-pores22 behavior has often been reported. The macrocyclic molecular cavities in the solvent-free crystalline state are occupied by interpenetrating structures or collapse to reduce the surface area; therefore, the internally quenched cavities result in latent pores with the loss of porosity. The latent pores remain dormant and are restored by encapsulating complementary guest molecules. Thus, guest encapsulation is primarily restricted by the sizes, shapes, and dimensions of the latent pores. For example, calix[n]arene crystals prepared under vacuum at high temperature contain no mappable pores; however, when they are exposed to liquids and/or vapors, a single-crystal to single-crystal (SCSC) transition23,24 is observed to form crystals containing guest molecules within newly emerged pores, for which the sizes, shapes, and dimensions of the macrocyclic pores produce selective encapsulation25,26,27,28,29,30,31,32,33.

In contrast, extrinsic intermolecular pores are frequently found in nonmacrocyclic molecular crystals. Close-packed molecular compartments preferentially reduce the surface area and maximize favorable intermolecular interactions in the crystalline state34; thus, sterically bulky three-dimensional structures, which encapsulate small guest molecules, are required to maintain residual voids as extrinsic intermolecular pores35,36,37,38. However, porosity-without-pore behavior with selective guest encapsulation is uncommon in nonmacrocyclic molecular crystals due to a lack of structurally determined latent pores. Thus, it remains challenging to establish porosity-without-pore behavior in selective guest encapsulation using nonmacrocyclic molecular crystals.

Tris(phenylisoxazolyl)benzene derivatives are planar molecules that form stacked supramolecular polymers through a combination of dipole–dipole interactions and π–π stacking interactions39. We envisioned that the two intermolecular interactions would determine the molecular arrangement of flat molecule 1 (Supplementary Figs. 1 and 2) in the crystalline state to generate well-defined latent pores (Fig. 1). Herein, we report stereoselections in the encapsulation of cis-decalin from a cis-/trans-decalin mixture based on planar tris(phenylisoxazole)benzene derivative 1 via porosity-without-pore behavior.

Fig. 1: Graphical summary of the decalin encapsulation.
figure 1

Schematic representation of the sequential phase transitions of 1 in the solid state.

Full size image

Results

Characterization of CHCl3@1, 1α, 1β, and mixed-phase state

Tris(phenylisoxazolyl)benzene 1 showed good crystallinity with CHCl3. X-ray diffraction-grade single crystals of CHCl3@1 with the C2/c space group were grown by slow evaporation of chloroform at room temperature (Fig. 2a and Supplementary Fig. 3). Two molecules of 1 formed a pair (colored in red in Fig. 2a), in which alternating deposition of the isoxazole rings caused the local dipole moments of isoxazoles to cancel each other out (Fig. 2b)40. The pairs were slip-stacked to form columnar assemblies with an interplanar distance of 3.4 Å (Fig. 2c)41. In the stacked structures, the remaining two isoxazole rings were slightly tilted to facilitate head-to-tail dipole‒dipole interactions along the stacking direction. Figure 2a and d illustrate another pair, comprising two molecules of 1 colored in purple, which adopted a perpendicular arrangement with a center-to-center aromatic ring distance of 14.6 Å and an intermolecular dihedral angle of 78.6°; a pore enclosed a pair of chloroform molecules. The size, shape, and dimension of the pores were fitted to the exterior of the paired chloroform molecules (Fig. 2d). The close contacts between 1 and the chloroform molecules determined by Hirshfeld surface analysis and the fingerprint plots are summarized in Fig. 2e and Supplementary Figs. 4 and 5. One of the two chloroform molecules in the porous cavity exhibited intermolecular Cl/H (55%), Cl/Cl (23%), H/C (6%), Cl/O (4%), and H/N (4%) contacts (Fig. 2f and Supplementary Fig. 5). The close contacts H(CHCl3)–C(1) and Cl(CHCl3)–H(1) suggested the presence of intermolecular CH/π and CH/Cl interactions in the crystal. Accordingly, the two chloroform molecules formed a molecular exterior, complementary to the interiors of the porous cavities in the crystal, where CH/π and CH/Cl interactions participated in the stabilization of the crystal structure of CHCl3@1.

Fig. 2: Studies of CHCl3@1.
figure 2

a Single-crystal structure of CHCl3@1 along the crystallographic b-axis generated from a chloroform solution of 1. b The structure of the dimeric pair of 1, highlighted in red as depicted in a. The dipole moment induced by an isoxazole ring is shown as a blue arrow. c The single-crystal structure of CHCl3@1 along the crystallographic c-axis. The stacked structure comprises a pair of 1, which is highlighted in red, as depicted in a. d Top, side, and front views of the single-crystal structure of CHCl3@1. The two perpendicularly arranged molecules of 1, highlighted in purple, are stacked in piles. The pair of chloroform molecules is encapsulated within the tubular pore. e Hirshfeld surface of CHCl3@1 mapped with dnorm. f Relative contributions to the Hirshfeld surface area for intermolecular contacts calculated from the structure are shown in e. The labels of the contact atoms denote Cl(CHCl3)–Cl(CHCl3), Cl(CHCl3)–O(1), Cl(CHCl3)–H(1), H(CHCl3)–C(1), and H(CHCl3)–N(1), from top to bottom. The labels indicate close contacts between other atoms; the details are shown in the Supplementary Information (Supplementary Fig. 5). Minimally disordered CHCl3 molecules were omitted for clarity.

Full size image

Heating a single crystal of CHCl3@1 at 60 °C under vacuum for 3 h induced an SCSC transition, which resulted in solvent-free single crystals of 1α and 1β (Fig. 3a, b and Supplementary Figs. 6–8) (see the Methods section for the detection of the two phases). The space groups of the single crystals 1α and 1β were Cc and Pca21, respectively, and no apparent pores were found, which was confirmed by N2 sorption experiments in the powder state (the estimated BET surface area was 2.52 m2 g–1, Supplementary Fig. 9 and Supplementary Note 1). The methoxyphenylisoxazole moiety in the single-crystal structure of 1α was disordered and showed two orientations with occupancies of approximately 50%, and the center-to-center aromatic ring distance of the perpendicularly arranged pair of 1 was determined to be 10.0 Å with an intermolecular dihedral angle of 77.3° (Fig. 3a, c and Supplementary Figs. 8 and 10). Thus, pair 1 became closer and more squeezed to fill the residual pores. In addition, the rotational disorder of the isoxazole ring also reduced the size of the residual pores (Supplementary Fig. 8). In the single-crystal structure of 1β, the paired molecules of 1 were arrayed along the crystallographic a-axis to form a densely packed structure without disorder (Fig. 3b, d, Supplementary Figs. 8 and 11). The center-to-center distance between the two nearest molecules of 1 was measured to be 12.7 Å. The two perpendicularly arranged molecules of 1 generated a dihedral angle of 76.2°. The pairs of 1 were squeezed and slightly rotated to quench the residual pore between the two molecules of 1 (Supplementary Fig. 8).

Fig. 3: Studies of 1α and 1β.
figure 3

Single-crystal structures of (a) 1α and (b) 1β along the crystallographic b-axis. The disordered parts are highlighted in blue with an equal probability for the methoxyphenylisoxazole moiety. c Top, side, and front views of the single-crystal structure of 1α. The two perpendicularly arranged molecules of 1, highlighted in purple, are stacked in piles. No tubular pores were found. The disordered units were omitted for clarity. d Top, side, and front views of the single-crystal structure of 1β. The two perpendicularly arranged molecules of 1, highlighted in purple, are stacked in piles. No tubular pores were found. e PXRD patterns of powder 1 (A) as-prepared, (B) after heating at 60 °C for 3 h in vacuo, and (C) after heating at 60 °C for 6 h in vacuo. (D) PXRD pattern of a 53:47 mixture of the 1α and 1β phases estimated from Rietveld refinement. Simulated PXRD patterns shown in (E) and (F) were generated from the single crystals 1α and 1β, respectively.

Full size image

To study the phase behaviors in the powder state of 1, an as-prepared 1 was heated at 60 °C under vacuum for 3 h. The resulting solvent-free powder was subjected to powder X-ray diffraction (PXRD) measurements, which illustrated a PXRD pattern that differed from that of the as-prepared powder 1. The newly observed PXRD pattern was not consistent with the PXRD pattern for either the 1α or 1β phases simulated from their single-crystal structures (Supplementary Fig. 12). The PXRD pattern was reproduced with fair precision from a Rietveld refinement of the mixture of 1α and 1β based on the single-crystal structures of 1α and 1β and varying their ratio within the angular range of 2–50°. The Rietveld refinement allowed us to determine the unit cell parameters and hkl indices with a reasonable Rwp value of 7.27% (Fig. 3e, Supplementary Figs. 13). The unit cell parameters were consistent with those obtained from single crystals of 1α and 1β. The hkl indices were assigned to a mixture of space groups Cc and Pca21, which corresponded to those of the single crystals of 1α and 1β, respectively (Supplementary Tables 2, 3, and 7). Assuming that the solvent-free powder contained only the 1α and 1β phases, a 53:47 ratio between the 1α and 1β phases in the mixture was estimated with a two-phase model42 involving a Cc phase and a Pca21 phase with Rigaku SmartLab Studio II software.

The thermodynamic stability of the solvent-free powder with the 1α and 1β phases was studied. When the as-prepared 1 was heated at 60 °C in vacuo for 6 h, the 1α and 1β phases were completely converted to the 1β phase (Fig. 3e and Supplementary Fig. 14), suggesting that the 1β phase is thermodynamically more stable than the 1α phase, which is most likely a metastable phase.

Characterization of decalin@1

As-prepared 1 was dissolved in cis-decalin and in trans-decalin at 140 °C in glass vials. The resulting solutions were slowly cooled under ambient conditions, which generated X-ray diffraction-grade single crystals of cis-decalin@1 and trans-decalin@1. The molecular arrangements of 1 within the crystal structures of both cis-decalin@1 (Fig. 4a, c, Supplementary Figs. 15 and 17) and trans-decalin@1 (Fig. 4b, d, Supplementary Figs. 16 and 17) were reasonably consistent with those in the single-crystal structures of CHCl3@1 (Fig. 2a). The two molecules of 1 adopted a perpendicular arrangement to form a pair, which was stacked in a columnar fashion. The cis- and trans-decalin molecules were located within pores formed by the columnar arrangement in the crystalline state.

Fig. 4: Studies of cis-decalin@1 and trans-decalin@1.
figure 4

Images of single-crystal structures of (a) cis-decalin@1 and (b) trans-decalin@1 along the crystallographic b-axis generated from the cis-decalin and trans-decalin solutions of 1, respectively. c Cis-decalin and (d) trans-decalin encapsulation within the pores between two molecules of 1. Hirshfeld surfaces of (e) cis-decalin@1 and (g) trans-decalin@1 mapped with dnorm. Relative contributions to the Hirshfeld surface area for intermolecular contacts calculated from the structures shown in (f) Panel e and (h) Panel g. The labels on the contact atoms denote H(decalin)–H(1 or decalin), H(decalin)–C(1), H(decalin)–N(1), and H(decalin)–O(1), from top to bottom. (i) PXRD patterns for solvent-free powder with the 1α and 1β phases when exposed to (A) cis-decalin vapor and (B) trans-decalin vapor. PXRD patterns of (C) cis-decalin@1 and (D) trans-decalin@1 estimated with Rietveld refinement based on simulated PXRD patterns of (E) cis-decalin@1 and (F) trans-decalin@1 generated from their single-crystal structures.

Full size image

Hirshfeld surface analysis and fingerprint plots of the single-crystal structures of cis-decalin@1 revealed that approximately 80% of the hydrogen atoms in cis-decalin were located near the hydrogen atoms of surrounding molecule 1 and neighboring cis-decalin molecules, whereas 13%, 2%, and 5% of the hydrogen atoms in the cis-decalin molecule contacted the C, N, and O atoms of molecule 1, respectively (Fig. 4e, f, Supplementary Figs. 18 and 19). The close contacts of H(cis-decalin)–C(1), H(cis-decalin)–N(1), and H(cis-decalin)–O(1) atoms contributed to the formation of intermolecular CH/π or hydrogen bonds in the molecular structure. Concordant results were obtained with Hirshfeld surface analyses and fingerprint plots for the single-crystal structure of trans-decalin@1, as summarized in Fig. 4g, h and Supplementary Figs. 18 and 20. More than 80% of the hydrogen atoms in trans-decalin were in close contact with the hydrogen atoms of the surrounding 1 molecule and neighboring trans-decalin molecules, and the remaining 20% of the hydrogen atoms of the trans-decalin molecule were located near the C, N, and O atoms of molecule 1. Accordingly, the formation of cis-decalin@1 and trans-decalin@1 was directed by the shape complementarity between the guests and the porous cavity formed in crystalline 1 during cocrystallization.

Common porous crystals often encapsulate and release their guests at the solid/liquid interface in thermodynamic equilibrium; thus, guest exchange is observed when porous crystals are exposed to guest molecules that are more complementary than those encapsulated in the pores43,44. To discuss the stabilities of the cocrystals, single crystals of cis-decalin@1 and trans-decalin@1 were soaked in trans-decalin and cis-decalin, respectively. Both specimens were left standing for two days under ambient conditions. No guest exchange was observed in the cocrystals (Supplementary Fig. 21), implying that large energy barriers prohibited in-and-out guest exchange at the solid/liquid interface. Accordingly, the guest encapsulation of 1 is most likely a kinetic process without any guest exchange45.

Exposing a solvent-free powder with the 1α and 1β phases to cis- and trans-decalin vapors for 6 days in glass vessels resulted in new PXRD patterns (Fig. 4iA, B and Supplementary Fig. 22), consistent with those of cis-decalin@1 and trans-decalin@1 (Fig. 4iC, D, Supplementary Figs. 23 and 24), which were obtained by Rietveld refinements based on simulated PXRD patterns of single crystals of cis-decalin@1 and trans-decalin@1 (Fig. 4iE and F). Thus, the solvent-free powder with the 1α and 1β phases encapsulated decalin molecules to form crystals of cis-decalin@1 and trans-decalin@1 by opening the collapsed porous cavity between the two molecules of 1. The cis-decalin@1 and trans-decalin@1 powders were exposed to trans- and cis-decalin vapors, respectively, for 6 days, but no discernible changes were observed in the PXRD patterns. The guest molecules in the cis-decalin@1 and trans-decalin@1 powders did not undergo exchange at the solid/vapor interface for at least 6 days, which was consistent with the results observed for the single crystals (Supplementary Fig. 25).

Selectivity in guest encapsulation

Cis- and trans-decalin molecules have the same elemental compositions; thus, stereoselective encapsulation of either cis- or trans-decalin requires shape complementarity between the guest molecules and the porous cavity. The stereoselective crystallization of 1 with the guest was studied. As-prepared 1 was dissolved in a 1:1 mixture of cis- and trans-decalin at 140 °C, which enabled cooling at room temperature under ambient conditions. X-ray diffraction-grade single crystals were obtained and subjected to single-crystal X-ray diffraction (SXRD) analysis. The crystal structure exhibited cocrystallization of 1 with cis-decalin (cis-decalin@1) (Supplementary Fig. 26), which indicated that the pores between two molecules of 1 were well defined and encapsulated by cis-decalin rather than trans-decalin. To determine the limit of selection (LOS), single crystals of 1 were grown from a 3:7 mixture, a 2:8 mixture, a 1:9 mixture, and a 0.1:9.9 mixture of cis- and trans-decalin. X-ray diffraction analyses of the single crystals based on the What is this? program46 revealed that cis-decalin@1 was selectively formed from the 3:7 mixture and the 2:8 mixture of cis- and trans-decalin (Supplementary Fig. 27). In contrast, trans-decalin@1 formed single crystals when grown in a 1:9 mixture of cis- and trans-decalin. Accordingly, the formation of single-crystal cis-decalin@1 was preferred even in the presence of excess trans-decalin.

The vapor adsorbed by 1 resulted in stereoselective encapsulation. The solvent-free powder with the 1α and 1β phases was prepared by heating the as-prepared 1, which was exposed to a vaporous cis-/trans-decalin mixture (1:1, w/w). A change in the PXRD pattern for solvent-free powder with the 1α and 1β phases was clearly observed when exposed to the vaporous cis-/trans-decalin mixture for 6 days. The PXRD pattern obtained from the cis-/trans-decalin mixture was fully consistent with that of 1 exposed to cis-decalin vapor (Fig. 5 and Supplementary Fig. 28). Although the vapor pressure of trans-decalin was reported to be slightly greater than that of cis-decalin47, 1 selectively encapsulated cis-decalin from a vaporous cis-/trans-decalin mixture. To determine the LOS, a solvent-free powder with the 1α and 1β phases was exposed to vaporous (3:7, 1:9, 0.5:9.5, 0.4:9.6, 0.3:9.7, 0.2:9.8, and 0.1:9.9) cis-/trans-decalin mixtures (w/w) for 6 days (Supplementary Fig. 29). The formation of cis-decalin@1 was indicated by the PXRD pattern after exposure to the vaporous 3:7, 1:9, and 0.5:9.5 cis-/trans-decalin mixtures, whereas the PXRD pattern for 1 exposed to vaporous 0.1:9.9 cis-/trans-decalin mixtures was not consistent with that of cis-decalin@1. Judging from a plot of the residual sum of squares (RSS) between the PXRD patterns of the mixtures and cis-decalin against the proportion of cis-decalin in the cis-/trans-decalin mixture, the selective formation of cis-decalin@1 required at least 4 w/w% cis-decalin in the trans-decalin solution (Supplementary Fig. 30 and Table 1). It is striking that the encapsulation of a cis-decalin molecule into 1 was kinetically more favorable than that of a trans-decalin molecule, with an activation energy difference of approximately 1.6 kcal mol–1. Time-dependent PXRD monitoring further suggested the encapsulation mechanism of 1. The solvent-free powder with the 1α and 1β phases was exposed to vaporous cis-decalin, resulting in changes in the PXRD patterns (Fig. 6a and Supplementary Fig. 31). As the diffraction peaks attributed to the 1α and 1β phases gradually decreased, those of cis-decalin@1 increased. The changes in intensity of PXRD peaks were plotted against time, which showed that the intensity of PXRD peaks of 1α and 1β decreased at approximately the same ratio, while the intensity of those of cis-decalin@1 increased with time (Fig. 6b and Supplementary Fig. 31). The sum of the normalized intensity of 1α, 1β, and cis-decalin@1 was observed to be more than 0.9 at each time point, ruling out the presence of an intermediate phase. Thus, the solvent-free powder with the 1α and 1β phases was instantaneously converted to cis-decalin@1 by the direct crystal-to-crystal phase transition.

Fig. 5: Selective cis-decalin encapsulation of 1.
figure 5

PXRD patterns for the solvent-free powder with the 1α and 1β phases after exposure to a vaporous 1:1 cis-/trans-decalin mixture. The bottom line represents the difference between the PXRD patterns for 1 after exposure to a vaporous 1:1 cis-/trans-decalin mixture and vaporous cis-decalin (Fig. 4iA).

Full size image
Table 1 Summary of the cis-/trans-selectivity of 1 for decalin encapsulation
Full size table
Fig. 6: Time-dependent decalin encapsulation of 1.
figure 6

a Time-dependent changes in the PXRD pattern for the solvent-free powder with the 1α and 1β phases when exposed to a 1:1 mixture of vaporous cis- and trans-decalin for (A) 0 days, (B) 1 day, (C) 2 days, (D) 5 days, (E) 7 days, and (F) 9 days. The red and blue arrows in Panel b indicate increasing and decreasing trends, respectively. b Plot of normalized PXRD peak intensity of (square) 1α (6.24°), (triangle) 1β (7.62°), and (circle) cis-decalin@1 (7.14°) against time. The bar graph indicates the sum of the normalized intensity, where green, magenta, and blue correspond to 1α, 1β, and cis-decalin@1, respectively.

Full size image

To restore the guest-free powders for reuse, the powder cis-decalin@1 was placed in a glass vial (capacity of 5 mL) and heated at 60 °C in vacuo. The complete removal of the cis-decalin molecule required heating for 6 h, resulting in a solvent-free powder with the 1β phase (Fig. 7a and Supplementary Fig. 31). The removal of encapsulated solvent molecules was also observed in CHCl3@1 and trans-decalin@1 by heating (Supplementary Fig. 32). When the solvent-free powder with the 1β phase was exposed to a vaporous cis-/trans-decalin mixture (1:1, w/w) for 6 days, no discernible change in the PXRD pattern was observed, which suggested that powder 1β did not encapsulate a decalin molecule (Fig. 7a and Supplementary Fig. 31). Thus, the solvent-free powder with the 1β phase is in a solvent-free inert state (Fig. 1). Given that solvent-free powder with the 1α and 1β phases encapsulate decalin molecules, the presence of the 1α phase most likely promotes the inert 1β phase to facilitate the encapsulation of decalin molecules. The 1β-rich solvent-free powder with a 1α:1β ratio of 2:98 (estimated by Rietveld refinement with an Rwp value of 9.07%) was subjected to a guest encapsulation experiment by exposure to a vaporous cis-/trans-decalin mixture (1:1, w/w) for 6 days (Fig. 7b and Supplementary Fig. 33). The resulting powders contained cis-decalin@1 and 1β at a ratio of 29:71, respectively, as estimated by Rietveld refinement, with an Rwp value of 8.45% (Fig. 7b and Supplementary Fig. 34). The amount of cis-decalin@1 formed is significantly greater than that of the 1α phase. Thus, all of the 1α phase was consumed in the encapsulation of cis-decalin, while the 1α phase promoted a portion of the 1β phase to encapsulate decalin molecules yielding cis-decalin@1. To detail the transition from the solvent-free powder with the 1α and 1β phases to cis-decalin@1, the solvent-free powder with the 1α and 1β phases (1α:1β = 53:47) was mixed with 1β in a ratio of 8:2, which is herein defined as post-mixed powder whereas the solvent-free powder with 1α and 1β obtained by heating the as-prepared 1 is defined as the pre-mixed powder, in the following section, where above-mentioned all the solvent-free powder with the 1α and 1β phases denote pre-mixed powder. The resulting post-mixed powder (1α:1β = 42:58) was ground by a mortar and pestle. The post-mixed solvent-free powder with the 1α and 1β phases was exposed to a vaporous 1:1 cis-decalin mixture for 6 days, which resulted in the mixture of cis-decalin@1 and 1β left (Fig. 7c and Supplementary Fig. 31). This observation suggests that the pre-mixed state is crucial for the decalin encapsulation. It is most likely that the 1α phase is in adequate interfacial contact with the 1β phase in the pre-mixed powder, which promotes the 1β phase to encapsulate decalin molecules. These results indicate that the pre-mixed state with the 1α and 1β phases is crucial for the encapsulation of decalin molecules in the latent pore, where the 1α phase most likely promotes a fair amount of the 1β phase to encapsulate decalin molecules.

Fig. 7: Decalin encapsulation behaviors of 1.
figure 7

a (A) PXRD pattern for cis-decalin@1. (B) PXRD pattern for cis-decalin@1 shown in Panel A after heating at 60 °C for 6 h in vacuo. (C) Simulated PXRD pattern generated from single-crystal structure of 1β. (D) PXRD pattern for the resulting powder shown in Panel B after exposure to a vaporous cis-/trans-decalin (1:1) mixture for 6 days. b (A) PXRD pattern for as-prepared 1 (80 mg) after heating at 60 °C for 3 h. (B) PXRD pattern for the resulting powder shown in Panel A after being exposed to a vaporous cis-/trans-decalin mixture (1:1, w/w) for 6 days (for detailed procedures, see the Methods section). (C) PXRD pattern of a mixture of 1α and 1β and (D) PXRD pattern of a mixture of cis-decalin@1 and 1β estimated by Rietveld refinement. Simulated PXRD patterns of (E) 1α, (F) 1β, and (G) cis-decalin@1 generated from their single-crystal structures. c (A) PXRD pattern for the post-mixed powder with the 1α and 1β phases (1α:1β = 42:58). (B) PXRD pattern for the post-mixed powder shown in Panel A after exposure to vaporous cis -decalin for 6 days. The green and black arrows indicate peaks corresponding to 1β and cis -decalin@1, respectively. Simulated PXRD patterns generated from single-crystal structure of (C) 1α, (D) 1β, (E) cis -decalin@1. d (A) PXRD pattern of the solvent-free powder with the 1β phase after it was dissolved in CH2Cl2, concentrated in vacuo, washed with EtOH, and then dried in vacuo at room temperature. (B) PXRD pattern for the resulting powder shown in Panel A after heating at 60 °C for 3 h. Simulated PXRD patterns of (C) 1α and (D) 1β generated from their single-crystal structures.

Full size image

To reuse the inert powder with the 1β phase, this powder was dissolved in CH2Cl2 with gentle heating in a water bath and then concentrated in vacuo. The resulting powder was washed with EtOH and then heated at 60 °C in vacuo for 3 h to yield the active solvent-free powder with the 1α and 1β phases, demonstrating the repeatability of the decalin encapsulation and release process (Fig. 7d and Supplementary Fig. 31).

Discussion

In summary, we report that planar tris(phenylisoxazolyl)benzene derivative 1 formed cocrystalline CHCl3@1, which produced latent porous crystals of 1α and 1β. Although there were no viable pores in either the 1α or 1β phase, they favorably encapsulated cis-decalin over trans-decalin from a cis-/trans-mixture of decalin within the emerging pores, which was established through a combination of single-crystal X-ray diffraction and PXRD analyses. Hirshfeld surface analysis and fingerprint plots revealed that the selectivity of solvent encapsulation was driven by the sizes, shapes, and dimensions of the well-defined pores. The time-dependent PXRD monitoring of the solvent-free powder with the 1α and 1β phases upon exposure to vaporous cis-decalin revealed that the 1α and 1β phases were converted to cis-decalin@1 by the direct crystal-to-crystal phase transition. Finally, the post-mix experiment revealed that the pre-mixed state of the 1α and 1β phases are crucial for the decalin encapsulation, leading to the conclusion that the 1α phase is in adequate interfacial contact with the 1β phase in the pre-mixed powder, which promotes the 1β phase to encapsulate of the decalin molecules. Such a consideration of the guest encapsulation behavior should be useful to deepen the understanding of small molecules exhibiting the porosity-without-pore behavior.

In general, macrocyclic assemblies exhibit selectivity for guest binding within their cavities. The sizes, shapes, and dimensions of guest molecules must be complementary to the cavity interiors to achieve high selectivity in guest binding. On this basis, flat molecules are hard to show selectivity in guest binding. Although flat molecule 1 resulted in solvent-filled pores, the perpendicular arrangement of two molecules of 1 rigidified the porous cavities via intermolecular interactions, which was confirmed by the fingerprint plot of CHCl3@1, cis-decalin@1, and trans-decalin@1 (Supplementary Figs. 5, 19, and 20). In fact, the size, shape, and dimension of the pores occupied by CHCl3 in CHCl3@1 are consistent with those observed in cis-decalin@1 and trans-decalin@1, which is characterized by the observation of the concordant void volumes within the unit cell of the pores of CHCl3@1, cis-decalin@1, and trans-decalin@1 (Supplementary Fig. 35), resulting in the excellent selectivity coming from the rigid feature of the porous cavity even though they are latent without guest molecules.

After heating of the crystals of 1, the latent pores were retained and remained dormant. Upon exposing the crystals to complementary guest molecules, the latent pores encapsulated the guest molecules and maintain the structural rigidity responsible for the high selectivity. Accordingly, the active crystals of 1α and 1β possessed latent intermolecular pores, exhibiting porosity-without-pore behaviors. 1α (Fig. 3a) ideally possesses porous cavity derived from the disordered moiety. However, N2 adsorption experiments confirmed that solvent-free powder with the 1α and 1β phases is nonporous (Supplementary Fig. 9 and Supplementary Note 1). We thus conclude that porosity-without-pore is the most suitable description for the observed phenomenon. Although porosity-without-pore behaviors commonly require macrocyclic molecular structures with large molecular cavities, the simple, nonmacrocyclic, planar molecule 1 clearly exhibited porosity-without-pore behavior and remarkable stereoselectivity between cis-/trans-decalin. This work represents a potential advantage of using flat molecules as porous materials. The pores provided by the flat molecule do not collapse in response to environmental stimuli because intermolecular pores emerge only at the moment of guest molecule encapsulation. Thus, the present work is expected to set the stage for the creation of porous materials that show separation capability with improved environmental stress tolerance.

Methods

All reagents and solvents were commercial reagent grade and were used without further purification except where noted. Triethylamine (98.0%) and CH2Cl2 (98.0%) were purchased from Nacalai Tesque, Inc., JAPAN. Dry triethylamine and dry CH2Cl2 were obtained by distillation over KOH and CaH2, respectively. 1,3,5-triethynylbenzene48 and benzohydroximoyl chrolide49 were synthesized according to reported procedures. 1H and 13C NMR spectra were recorded on a JEOL ECA-500 spectrometer. Chemical shifts were reported in the delta scale in ppm relative to residual chloroform (δ = 7.26 and 77.2 for 1H and 13C, respectively). High Resolution ESI mass spectrometry was performed using a Thermo Fisher Scientific LTQ Orbitrap XL hybrid FTMS. Melting points (M.p.) were measured with a Yanagimoto micro melting point apparatus. N2 gas sorption experiment was performed on a BELSORP-maxII034VP-MZ (MicrotracBEL). X-ray crystallographic data were obtained with a Rigaku XtaLAB Synergy R, DW system, HyPix diffractometer. The crystals were kept at 100 K during data collection. Using Olex250, the structure was solved with the SHELXD51 structure solution program using Dual Space or SHELXT52 structure solution program using Intrinsic Phasing and refined with the SHELXL53 refinement package using Least Squares minimization. The crystals were irradiated with CuKα radiation (λ = 1.54184 Å).

Characterization

The synthesis and characterization of 1 are described in full detail in the Supplementary Information, which contains the 1H NMR and 13C NMR spectra of the compound.

Preparation of single crystals of CHCl3@1

First, 1 mg of as-prepared 1 and 1 mL of commercial-grade CHCl3 were mixed in a glass vial (capacity of 5 mL) at room temperature. The specimen was heated at 50 °C for 2 min by oil-bath heating without stirring under an open-air atmosphere. Then, the heater was turned off to reach room temperature. The diffraction grade of the single crystals increased as the temperature gradually decreased.

Preparation of single crystals of 1α and 1β

Approximately 50 single crystals of CHCl3@1 were randomly selected and placed on a glass plate by a metallic spatula. Then, the glass plate with the crystals was heated by a hot plate at 60 °C for 3 h in vacuo, resulting in diffraction-grade plates and block crystals corresponding to 1α and 1β, respectively. The crystals were separated by using a needle and subjected to SXRD analysis.

Preparation of single crystals of cis-decalin@1 and trans-decalin@1

As-prepared 1 (3 mg) and 2 mL of commercial grade cis- or trans-decalin were mixed in a glass vial (capacity of 5 mL) at room temperature. The specimens were heated at 140 °C for 5 min by heating in an oil bath with stirring under an open-air atmosphere. Then, the heater was turned off to reach room temperature. The diffraction grade of the single crystals increased as the temperature gradually decreased.

Hirshfeld surface analysis

Hirshfeld surface analysis was carried out using CrystalExplorer software (version 21.3)54.

Rietveld refinement

Rietveld refinement was carried out using Rigaku SmartLab Studio II software (version 4.5.286.0). The initial crystallographic parameters were acquired from the single-crystal structures. The Rietveld refinement in this work was performed to assign hkl indices; thus, atomic coordinate refinements were excluded from the Rietveld refinement, resulting in relatively large S values that did not impact the authors’ conclusions.

Preparation of solvent-free powders

First, 1.0 g of as-prepared 1 was placed in a glass vial (capacity of 50 mL), which was heated at 60 °C for 3 h in vacuo, resulting in a pre-mixed solvent-free powder with the 1α and 1β phases. The resulting mixture (80 mg) was placed in a glass vial (capacity of 5 mL), which was heated at 60 °C for 3 h in vacuo, producing a solvent-free powder with the 1β phase.

Preparation of a 1β-rich mixture of 1α and 1β

As-prepared 1 (80 mg) was placed in a glass vial (capacity of 5 mL), which was heated at 60 °C for 3 h in vacuo, resulting in a pre-mixed 1β-rich solvent-free powder with the 1α and 1β phases.

Experimental procedure for selective encapsulation in the single-crystalline phase

As-prepared 1 (3 mg) and 2 mL of commercial grade cis- and trans-decalin mixture was mixed in a glass vial (capacity of 5 mL) at room temperature. The specimens were heated at 140 °C for 5 min by heating in an oil bath with stirring under an open-air atmosphere. Then, the heater was turned off to reach room temperature. The diffraction grade of the single crystals increased as the temperature gradually decreased.

Experimental procedure for selective encapsulation in the powder phase

As-prepared 1 (80 mg) was placed in a glass vial (capacity of 5 mL), which was placed in a larger glass vial (capacity of 50 mL) with 8 mL of a cis- and trans-decalin mixture, cis-decalin, or trans-decalin. The larger glass vial was sealed tightly, and the specimens were left standing for 6 days. The resulting powders were subjected to PXRD analysis.

Experimental procedure for time-dependent PXRD measurements

The pre-mixed solvent-free powder with the 1α and 1β phases was placed on a glass plate. The specimen was placed on a glass stand in a desiccator (capacity of 500 mL). Cis-decalin (300 mL) was poured into the desiccator. The desiccator containing the specimen was covered with a lid. The specimen was left standing at ambient temperature, and the aged specimen was subjected to PXRD measurements.

Experimental procedure for the gas sorption experiment

The N2 gas sorption experiment was performed on a BELSORP-maxII034VP-MZ (MicrotracBEL). As-prepared 1 (30 mg) was heated at 60 °C for 3 h in vacuo to form the solvent-free powder with the 1α and 1β phases prior to the measurement. The N2 isotherms were measured using a liquid nitrogen bath (77 K).