MICROPOROUS CRYSTALS

Self-assembly of lattices with highstructural complexity from ageometrically simple moleculeHiroshi Yamagishi1*†, Hiroshi Sato1, Akihiro Hori2, Yohei Sato3, Ryotaro Matsuda2,Kenichi Kato4, Takuzo Aida1,5†

Here we report an anomalous porous molecular crystal built of C–H···N-bonded double-layered roof-floor components and wall components of a segregatively interdigitatedarchitecture. This complicated porous structure consists of only one type of fully aromaticmultijoint molecule carrying three identical dipyridylphenyl wedges. Despite its highsymmetry, this molecule accomplishes difficult tasks by using two of its three wedges forroof-floor formation and using its other wedge for wall formation. Although a C–H···N bond isextremely labile, the porous crystal maintains its porosity until thermal breakdown of theC–H···N bonds at 202°C occurs, affording a nonporous polymorph. Though this nonporouscrystal survives even at 325°C, it can retrieve the parent porosity under acetonitrile vapor.These findings show how one can translate simplicity into ultrahigh complexity.

The research field of crystal engineering wasinitiated by Desiraju and co-workers, whoestablished its basic concept in the late1980s by using small organic molecules ofgeometrical simplicity (1). Since then, organ-

ic molecules of further structural complexity havebegun to be used for crystal engineering in con-junction with coordinating metal ions to obtainmore-complex crystals that have tailored func-tions (2–4). During our curiosity-driven study onthe solution behaviors of hyperbranched, multi-

joint, fully aromatic molecules, we made a seren-dipitous finding that may be antithetic to thecurrent trend. Like dendrimers (5–7), such aro-matic molecules are characterized by their non-planarmorphology and conformational flexibility,so that they are noncrystalline in the solid state,except for a few examples (8, 9). Py6Mes is a newlydesignedD3h-symmetric molecule that consistsexclusively of aromatic rings that are connectedvia single bonds (Fig. 1A). The nonplanarity ofthis multijoint aromatic molecule stems from a

large steric repulsion between themethyl groupsof themesitylene (Mes) core and connected phenyl-ene rings (Fig. 1B). Py6Mes has a propeller shapewith a symmetry group of D3h, which carries apolar outer shell composed of six pyridyl (Py)units for sterically protecting the nonpolar aro-matic core (Fig. 1, A to C). We were curious aboutwhether Py6Mes, which has a nonpolar solvo-phobic core, is soluble in highly dielectric mediaas a discrete spherical amphiphile because of itssolvophilic shell. However, despite this character-istic, Py6Mes readily precipitated in acetonitrile(MeCN) and 2-propanol (isoPrOH), affording theporous crystal Pyopen, which included crystallo-graphically disordered solvent molecules in itsnanopores (Fig. 1D). Although Pyopen comprisesonly one type of geometrically simple molecule,it accomplishes very complicated tasks (Fig. 2B).As shown in Figs. 1D and 2D, Py6Mes uses two ofits three dipyridylphenyl wedges for the construc-tion of C–H···N-bonded double-layered roof-floorcomponents, adopting a cofacial orientation. In

RESEARCH

Yamagishi et al., Science 361, 1242–1246 (2018) 21 September 2018 1 of 4

1Department of Chemistry and Biotechnology, School ofEngineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku,Tokyo 113-8656, Japan. 2Department of Applied Chemistry,Graduate School of Engineering, Nagoya University, Furo-cho,Chikusa-ku, Nagoya 464-8603, Japan. 3Institute for IntegratedCell-Material Sciences (WPI-iCeMS), Kyoto University,Yoshida, Sakyo-ku, Kyoto 606-8501, Japan. 4RIKEN SPring-8Center, 1-1-1 Kouto, Sayo-cho, Sayo-gun, Hyogo 679-5148,Japan. 5RIKEN Center for Emergent Matter Science,2-1 Hirosawa, Wako, Saitama 351-0198, Japan.*Present address: Division of Materials Science, Faculty of Pureand Applied Sciences, University of Tsukuba, 1-1-1 Tennodai,Tsukuba, Ibaraki 305-8571, Japan.†Corresponding author. Email: aida@macro.t.u-tokyo.ac.jp (T.A.);yamagishi@macro.t.u-tokyo.ac.jp (H.Y.)

Fig. 1. Molecularstructures ofPy6Mes and crystalpacking diagramsof Pyopen⊃MeCN,Pyopen⊃isoPrOH,PyVDW⊃CHCl3, andPyVDW⊃THF. (A toC) Molecular struc-ture (A), wireframe(B), and CPK (C)representations ofPy6Mes. The pyridylrings are colored inblue and green,whereas the mesity-lenyl and phenylrings are colored inblack and white.(D to G) Wireframerepresentations ofthe crystal-packingdiagrams ofPyopen⊃MeCN (D),Pyopen⊃isoPrOH (E),PyVDW⊃CHCl3 (F),and PyVDW⊃THF (G).

on August 9, 2020

http://science.sciencem

ag.org/D

ownloaded from

an orthogonal direction to the roof-floor compo-nents, Py6Mes uses its residual one dipyridyl-phenyl wedge together with the mesitylenyl corefor the construction of wall components formedby a solvophobically interdigitated architecture(Fig. 2E). Such difficult tasks, as accomplished byPy6Mes, illustrate how ultrahigh structural com-plexity can be achieved through self-assembly of asingle molecule of ultrahigh symmetry. Further-more, even though Pyopen has C–H···N-bondedparts, it has an exceptionally high thermal sta-bility (Fig. 3B) and, at the same time, has an ex-cellent self-healing ability (Fig. 3C). These findingsalso allow us to tackle a current issue inmaterialsscience: how thermally robustmaterials aremadeto self-heal (10–15).C–H···N bonds are far more labile than con-

ventional hydrogen bonds (16–20). However, incontrast to hydrogen bonds, C–H···N bonds cansurvive even in highly dielectricmedia because theirattraction force is given essentially by a disper-sion force (21–23). To the best of our knowledge,Pyopen reported herein is the most thermostableporous crystal among those comprising non-classical C–H···X bonds (24–26) and amongall healable porous organic and metal-organiccrystals reported thus far (10–12). Py6Mes, thecomponent of Pyopen, carries three dipyridylphenylwedges at its mesitylene core (Fig. 1A; details forthe synthesis of Py6Mes are available in the sup-

plementary materials). In MeCN that has anexceptionally high relative permittivity (er = 37.5),Py6Mes self-assembles into single crystals that aresuitable for x-ray crystallography. The as-receivedcrystal adopts a space group of P21/c and includescrystallographically disordered MeCN moleculesin its zigzag-shaped one-dimensional (1D) nano-pores, which have diameters of 6 Å (Pyopen⊃MeCN;Fig. 1D and fig. S13). As expected, each dipyridyl-phenyl wedge tilts almost perpendicularly to theconnected mesitylenyl core, adopting a dihedralangle in the range of 70 to 90°, but none of thethree dipyridylphenyl wedges are crystallograph-ically identical. Despite a large number of aro-matic rings, only one pair of pyridyl rings thatcontain N3 atoms likely forms a cofacial p-stackwith an interplanar distance of 3.36 Å (fig. S14B).Among the five pairs of geometrically close Hand C atoms, only two pairs (H8A···C18 andH18···C13; fig. S14A and table S1) form a C–H···p bond. Five out of the six pyridyl rings inPy6Mes are involved in the formation of theC–H···N bonds (Fig. 2, A to D, and table S1).Overall, the back-to-back connected double-layered roof-floor components, supported byC–H···N bonds, are formed along the crystal-lographic bc plane (Fig. 2, B to D), whereas thewall components that are formed by the inter-digitation of the dipyridylphenylmesitylenyl partsare constructed along the crystallographic b axis

(Fig. 2E). Thermogravimetric analysis (TGA) ofPyopen⊃MeCN showed that the included MeCNmolecules began tobe released fromthenanoporeseven at room temperature, affording guest-freePyopen at 50°C, which is a much lower temper-ature than the boiling point (82°C) of MeCN(black curve in fig. S7). These observations indicatethat the guest MeCN molecules are only weaklytrapped in the nanopores of Pyopen. As expected,the powdery sample of guest-free Pyopen displayeda typical type I sorption isotherm for N2, with asteep slope in the low relative pressure (P/P0) re-gion and aBrunauer-Emmett-Teller (BET) surfacearea of 219m2 g–1 (red circles in Fig. 3D). The poresize distribution, as estimated bymicropore analysis(27),was unimodalwith amaximumpeakat 0.6nm(fig. S8), which is in excellent agreement with thediameter of the 1D nanopore observed in Fig. 1D.As has been discussed for porous crystals in

the literature (28), the relative permittivity of thecrystallization solvent largely affects the mode ofsolute-solvent interactions and therefore affectsthe crystal structures.When Py6Mes was allowedto assemble in highly dielectric isoPrOH (er = 18.3),the formation of solvent-including porous crys-tal Pyopen⊃isoPrOH (Fig. 1E) occurred, which isisomorphic to Pyopen⊃MeCN (Fig. 1D), and itsnanopores, again, included crystallographicallydisordered isoPrOH molecules. Considering alsothat theTGAprofile of Pyopen⊃isoPrOHwith respect

Yamagishi et al., Science 361, 1242–1246 (2018) 21 September 2018 2 of 4

Fig. 2. Modes of C–H···N bondsin Pyopen⊃MeCN and crystal-packing diagrams of Pyopen⊃MeCN.(A) Drawings representing themodes of C–H···N bonds inPyopen⊃MeCN. (B to E) CPKrepresentations of the crystal-packing diagrams of Pyopen⊃MeCN.Orange outlines indicate theC–H···N-bonded roofs andfloors in Pyopen⊃MeCN [(B) and(C)], and dashed purple outlinesindicate the wall connected tothe roof and floor [(B), (C),and (E)].

RESEARCH | REPORTon A

ugust 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

to the release of isoPrOH (red curve in fig. S7)is analogous to that observed for Pyopen⊃MeCN,we conclude that Pyopen weakly traps isoPrOH,similarly to whenMeCN is used, in its nanopores.In sharp contrast to the above cases in highlydielectric media, when Py6Mes was allowed toassemble in moderately dielectric media suchas chloroform (CHCl3, er = 4.8) and tetrahydro-furan (THF, er = 7.6), classical van der Waalsinclusion crystals denoted as PyVDW⊃CHCl3 andPyVDW⊃THF formed, respectively (Fig. 1, F andG;for their crystal structures, see the supplementarymaterials). As summarized in table S2, we ob-served a clear correlation between the relative per-mittivities of the crystallization solvents and spacegroups of the resulting polymorphs (29). Inclusioncrystals PyVDW⊃CHCl3 (Fig. 1F) and PyVDW⊃THF(Fig. 1G) are isomorphic to each other and havea space group of P21/n, but they are geometri-cally different from Pyopen⊃MeCN (Fig. 1D) andPyopen⊃isoPrOH (Fig. 1E), both of which adopt aspace group of P21/c. It is likely that Py6Mes is nolonger amphiphilic inmoderately dielectric CHCl3and THF; that is, these solvents are affinitivetoward both the shell and core parts of Py6Mes.Consequently, Py6Mes does not form a solvo-phobically interdigitated structure, as observedfor the wall components in Pyopen⊃MeCN andPyopen⊃isoPrOH (Fig. 1, D and E). In PyVDW⊃CHCl3(Fig. 1F) and PyVDW⊃THF (Fig. 1G), the includedsolvent molecules are essential constituents forsupporting the crystal structures. Hence, as con-firmed by TGA (fig. S7) and x-ray diffraction

(XRD) (fig. S10), the crystals were easily demol-ished after heating at 100°C for 3 hours to re-move the included solvent molecules.Porous crystals, particularly when guest-free,

are intrinsically unstable unless they use metalcoordination bonds or dynamic covalent bonds.In the literature (24–26), none of the reportedporous crystals that are composed of nonclassicalC–H···X bonds can survive above 130°C. As de-scribed earlier, Pyopen turned out to be far morethermally robust than the reported examples.In a differential scanning calorimetry (DSC) pro-file over a wide temperature range from 40° to300°C, guest-free Pyopen in the first heating pro-cess displayed a single exothermic peak at 202°Cdue to a crystalline phase transition (red curve infig. S5). Upon subsequent cooling, neither an exo-thermic peak nor an endothermic peak appeared(blue curve in fig. S5), indicating that the phasetransition at 202°C is irreversible. Accordingly, itspowder XRD (PXRD) profiles did not change upto 202°C (red, yellow, and green curves in Fig. 3B)and remainedessentially identical to that simulatedfrom the single-crystal structure of Pyopen⊃MeCN(fig. S12). Upon further heating to induce thephase transition, the PXRD profile changedabruptly (blue and purple curves in Fig. 3B) andirreversibly (fig. S11), affording a new crystallinephase. Although the crystals thus formed wereheavily cracked and no longer eligible for single-crystal x-ray structural analysis, we successfullyidentified the crystalline structure by means ofRietveld analysis for the PXRD profile measured

at 50°C (see supplementarymaterials). In contrastto Pyopen, the product, denoted nonporous Pyclose

adopting a space group of P�1 , has no solvent-accessible 1D channels (Fig. 3A and fig. S15, A toD). Consistently, Pyclose showed a higher density(1.192 g cm–3) than porous Pyopen (1.022 g cm–3)and barely adsorbed N2 (blue circles in Fig. 3D).The Rietveld analysis indicated that none of theC–H···N bonding pairs that were originally pres-ent in Pyopen were shorter than sum of the vander Waals radii of the H and N atoms (table S3),suggesting that the crystalline phase transitionat 202°C was triggered by thermal breakdownof the C–H···N bonds in the double-layered roof-floor components (fig. S15, A to D).Although self-healing is one of the attractive

features in materials science, an essential chal-lenge, in the case of solid materials, is to addressthe general issue that high thermal robustnessand excellent healing ability are mutually exclu-sive (13–15). Pyclose is thermally more robust thanPyopen and can survive even at 325°C (Fig. 3A).Nevertheless, under MeCN vapor, Pyclose self-healed to retrieve its parent porosity at ambienttemperatures. Typically, crystalline Pyclose was putin a glass vial that was filled with MeCN vaporand incubated at 20°C. As shown in Fig. 3C, a setof new PXRD peaks assignable to Pyopen⊃MeCNgradually emerged at the expense of the originaldiffractions due to Pyclose, where the completerecovery of the original diffractions required7 hours (Fig. 3C). The total volume ofMeCN vaporeventually adsorbed by Pyclose on its recovery was

Yamagishi et al., Science 361, 1242–1246 (2018) 21 September 2018 3 of 4

Fig. 3. Transfor-mations amongPyopen⊃MeCN, Pyopen,and Pyclose, and theirN2 adsorption iso-therms. (A) Wire-frame representationsof the crystal-packingdiagrams ofPyopen⊃MeCN andPyclose. (B) PXRDprofiles of Pyopen

measured at differenttemperatures uponheating [wavelength(l) = 1.07965 Å].q, angle of scattering.(C) Time-dependentPXRD profiles ofPyclose incubated in acuvette filled with MeCNvapor (l = 1.54 Å).The black line repre-sents a PXRD profilesimulated from thesingle-crystal struc-ture of Pyopen⊃MeCN.(D) N2 adsorptionisotherms of Pyopen

(red circles) andPyclose (blue circles)at 77 K.

RESEARCH | REPORTon A

ugust 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

nearly equal to the original pore volume ofPyopen (fig. S17). Guest-free Pyopen obtained fromPyopen⊃MeCN thus recovered again showed aN2 adsorption behavior characteristic of micro-porous materials (fig. S16). The observed recov-ery is not the consequence of a trivial processinvolving partial dissolution of Pyclose inMeCN followed by recrystallization of Py6Mes,because a finely ground amorphous powder sam-ple of Pyopen, when likewise treated underMeCNvapor at 20°C, did not transform back into Pyopen

but gradually exerted a totally different XRDpattern (fig. S9) andMeCN adsorption profile(fig. S18). Namely, the reversible transformationoccurs only between Pyopen⊃MeCN and Pyclose.Here we demonstrate that Pyopen⊃MeCN con-

secutively transforms into guest-free Pyopen andthen nonporous Pyclose, which transforms backinto Pyopen⊃MeCN under MeCN vapor. Figure 4shows a possible energy diagram for the overallcrystalline transformation based on the DSC andTGA profiles. As described above, the release ofguest MeCNmolecules from Pyopen⊃MeCN to af-ford Pyopen likely requires only a negligibly smallactivation energy, as evidenced by the TGAprofileof Pyopen⊃MeCN (fig. S7). By contrast, the trans-

formation of Pyopen into Pyclose requires consider-able heating and features a single exothermicpeak in DSC at 202°C (fig. S5), for which thechange in enthalpy (DH) value was evaluated tobe 15.3 kJmol–1. Bymeans of the Kissinger meth-od (30), the activation energy for this processwasevaluated to be as high as 320 kJmol–1 (fig. S6). Onthe other hand, the regeneration of Pyopen⊃MeCNfrom Pyclose under MeCN vapor occurs autono-mously (Fig. 3C), suggesting that this transforma-tion is energetically downhill and has a smallactivation energy to overcome, even at ambienttemperatures. As is apparent in Fig. 3A, thisprocess requires only a moderate reorientationof the interdigitating dipyridylphenylmesitylenylunits to effectively cancel their dipoles. We sup-pose that MeCN vapor would play an interme-diating role in this reorientation event.As exemplified by the formation of PyVDW in

moderately dielectric media (Fig. 1, F and G), as-sembled structures usually reflect the symmetryof constituent molecules. In this sense, it may bedifficult to imagine that high-symmetry Py6Mesis the sole constituent for porous Pyopen, whoseroof-floor components are structurally and ori-entationally different from its wall component(Fig. 1, D and E). Apart from the exciting physicalproperties of Pyopen that may change the pre-conception of healable solid materials (Fig. 4),the most intriguing result presented here is thatPy6Mes, in highly dielectric media for construct-ing Pyopen, uses its three equivalent wedgesfor totally different tasks (Fig. 2B). This exampledemonstrates that even a single molecule of geo-metrical simplicity could assemble into a highlycomplicated exotic material of low symmetry.

REFERENCES AND NOTES

1. G. R. Desiraju, Angew. Chem. Int. Ed. Engl. 34, 2311–2327(1995).

2. M. Irie, T. Fukaminato, K. Matsuda, S. Kobatake, Chem. Rev.114, 12174–12277 (2014).

3. H. Furukawa, K. E. Cordova, M. O’Keeffe, O. M. Yaghi, Science341, 1230444 (2013).

4. S. Horike, S. Shimomura, S. Kitagawa, Nat. Chem. 1, 695–704(2009).

5. D. A. Tomalia et al., Polym. J. 17, 117–132 (1985).6. S. M. Grayson, J. M. J. Fréchet, Chem. Rev. 101, 3819–3868 (2001).7. C. J. Hawker, J. M. J. Fréchet, J. Am. Chem. Soc. 112,

7638–7647 (1990).8. R. E. Bauer, V. Enkelmann, U. M. Wiesler, A. J. Berresheim,

K. Müllen, Chem. Eur. J. 8, 3858–3864 (2002).9. S. Nakajima et al., Chem. Commun. (Camb.) 54, 2534–2537

(2018).10. Y. Sakata et al., Science 339, 193–196 (2013).11. Y.-G. Huang et al., Nat. Commun. 7, 11564 (2016).12. J. Tian, P. Thallapally, J. Liu, G. J. Exarhos, J. L. Atwood, Chem.

Commun. (Camb.) 47, 701–703 (2011).13. P. Cordier, F. Tournilhac, C. Soulié-Ziakovic, L. Leibler, Nature

451, 977–980 (2008).

14. Y. Chen, A. M. Kushner, G. A. Williams, Z. Guan, Nat. Chem. 4,467–472 (2012).

15. Y. Yanagisawa, Y. Nan, K. Okuro, T. Aida, Science 359, 72–76(2018).

16. U. Samanta, P. Chakrabarti, J. Chandrasekhar, J. Phys. Chem.A 102, 8964–8969 (1998).

17. F. A. Cotton, L. M. Daniels, G. T. Jordan IV, C. A. Murillo, Chem.Commun. (Camb.) 0, 1673–1674 (1997).

18. M. Mascal, Chem. Commun. (Camb.) 0, 303–304 (1998).19. V. R. Thalladi, A. Gehrke, R. Boese, New J. Chem. 24, 463–470

(2000).20. E. Arunan et al., Pure Appl. Chem. 83, 1619–1636 (2011).21. G. R. Desiraju, T. Steiner, The Weak Hydrogen Bond in

Structural Chemistry and Biology (Oxford Univ. Press, 1999).22. M. Nishio, Y. Umezawa, K. Honda, S. Tsuboyama, H. Suezawa,

CrystEngComm 11, 1757–1788 (2009).23. G. R. Desiraju, Acc. Chem. Res. 35, 565–573 (2002).24. P. Sozzani, S. Bracco, A. Comotti, L. Ferretti, R. Simonutti,

Angew. Chem. Int. Ed. 44, 1816–1820 (2005).25. K. J. Msayib et al., Angew. Chem. Int. Ed. 48, 3273–3277

(2009).26. C. G. Bezzu, M. Helliwell, J. E. Warren, D. R. Allan,

N. B. McKeown, Science 327, 1627–1630 (2010).27. R. S. Mikhail, S. Brunauer, E. E. Bodor, J. Colloid Interface Sci.

26, 45–53 (1968).28. N. B. McKeown, J. Mater. Chem. 20, 10588–10597 (2010).29. I. M. Smallwood, Handbook of Organic Solvent Properties

(Butterworth-Heinemann, Oxford, 2012).30. H. E. Kissinger, Anal. Chem. 29, 1702–1706 (1957).

ACKNOWLEDGMENTS

H.Y. thanks A. I. Cooper and M. A. Little in the Department ofChemistry and Materials Innovation Factory, University ofLiverpool, for x-ray diffraction measurements. Synchrotronradiation experiments were performed at BL44B2 in SPring-8 withthe approval of the RIKEN SPring-8 Center (proposal 20160024).Funding: This work was supported by the Japan Society for thePromotion of Science (JSPS) through its Grant-in-Aid for ScientificResearch (S) (18H05260) on “Innovative functional materialsbased on multi-scale interfacial molecular science.” H.Y. thanks theJSPS for a Young Scientist Fellowship and Leading GraduateSchools (MERIT) and the Leverhulme Trust (Leverhulme ResearchCentre for Functional Materials Design) for providing fundingduring a research placement at the University of Liverpool. Authorcontributions: H.Y. performed and interpreted all the experimentsassociated with molecular synthesis, crystal growth, and structuralcharacterization. H.S. performed the sorption experiments. K.K.conducted synchrotron x-ray studies. A.H., Y.S., and R.M. carriedout crystallographic studies. All authors contributed to the writingand editing of the manuscript. H.Y. and T.A. conceived the project,designed experiments, and directed the research. Competinginterests: None declared. Data and materials availability:Crystallographic data reported in this paper are listed in thesupplementary materials and archived at the CambridgeCrystallographic Data Centre under reference numbers CCDC1857527 to 1857531. All other data needed to evaluate theconclusions in the paper are present in the paper or thesupplementary materials.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/361/6408/1242/suppl/DC1Materials and MethodsFigs. S1 to S18Tables S1 to S3References (31–38)

20 March 2018; accepted 23 July 201810.1126/science.aat6394

Yamagishi et al., Science 361, 1242–1246 (2018) 21 September 2018 4 of 4

Fig. 4. An energy diagram for the consecu-tive transformation cycle involvingPyopen⊃MeCN, Pyopen, and Pyclose. Energylevels of Pyopen⊃MeCN, Pyopen, and Pyclose wereestimated on the basis of the DSC and TGAprofiles. The arrows indicate the direction of thecrystal transformations. Ea, activation energy.

RESEARCH | REPORTon A

ugust 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

Self-assembly of lattices with high structural complexity from a geometrically simple moleculeHiroshi Yamagishi, Hiroshi Sato, Akihiro Hori, Yohei Sato, Ryotaro Matsuda, Kenichi Kato and Takuzo Aida

DOI: 10.1126/science.aat6394 (6408), 1242-1246.361Science 

, this issue p. 1242Sciencewas stable up to 202°C and could be recovered after collapse by exposure to solvent vapor.

H?N bonds and van der Waals forces. Despite the weakness of these interactions, the porous structure−from labile C and crystallized it from highly dielectric solvents. Porous crystals formed with complex pore-wall structures that resulted

synthesized an aromatic molecule that bears a symmetrical outer shell of three dipyridylphenyl wedges et al.Yamagishi Organic materials can exhibit high porosity, but the structures often collapse or decompose at high temperatures.

Robust assembly of aromatic molecules

ARTICLE TOOLS http://science.sciencemag.org/content/361/6408/1242

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2018/09/19/361.6408.1242.DC1

REFERENCES

http://science.sciencemag.org/content/361/6408/1242#BIBLThis article cites 35 articles, 4 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Science. No claim to original U.S. Government WorksCopyright © 2018 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of

on August 9, 2020

http://science.sciencem

ag.org/D

ownloaded from