Noncovalent interaction guided selectivity of haloaromatic isomers in a flexible porous coordination polymer

Porous, supramolecular structures exhibit preferential encapsulation of guest molecules, primarily by means of differences in the order of (noncovalent) interactions. The encapsulation preferences can be for geometry (dimension and shape) and the chemical nature of the guest. While geometry-based sorting is relatively straightforward using advanced porous materials, designing a “chemical nature” specific host is not. To introduce “chemical specificity”, the host must retain an accessible and complementary recognition site. In the case of a supramolecular, porous coordination polymer (PCP) [Zn(o-phen)(ndc)] (o-phen: 1,10-phenanthroline, ndc: 2,6-naphthalenedicarboxylate) host, equipped with an adaptable recognition pocket, we have discovered that the preferential encapsulation of a haloaromatic isomer is not only for dimension and shape, but also for the “chemical nature” of the guest. This selectivity, i.e., preference for the dimension, shape and chemical nature, is not guided by any complementary recognition site, which is commonly required for “chemical specificity”. Insights from crystal structures and computational studies unveil that the differences in the different types of noncovalent host–guest interaction strengths, acting in a concerted fashion, yield the unique selectivity.


Introduction
6][7][8] These are made by connecting specic geometry nodes via dynamic crosslinking bonds. 9As a result, empty spaces or voids can be designed with desirable dimensions and chemical functionalities.By judiciously combining metal ions and complementary organic linkers, porous coordination polymers (PCPs) 10 and metal-organic cages 11 can be obtained; using organic linkers alone, covalent organic frameworks (COFs), 12 conjugated microporous polymers (CMPs), 13 and organic cages 14 can also be designed.With the advantages of easy structural tailorability, it is evident that pores of specic shapes and sizes can be made for desired chemical separations. 15,16This scheme works in a straightforward manner, but in the case of a mixture of molecules with very similar physical and chemical properties, selectivity decreases drastically.This is because of the fact that molecules are discriminated on the basis of dimension, shape, or chemical nature, not all three at once.Molecular recognition by a combination of all three is non-trivial to design and realize.
One way to introduce a more advanced recognition system in these porous materials is by allowing the pores to adjust to the guest molecules.The adjustment will depend on the dimension, shape as well as chemical nature of the guests.Among PCPs, a sub-category, i.e., exible or so PCPs exists. 17This class of PCPs is known to adapt its pore structure to the environment. 18,19These structures have exible metal-nodes or geometrically exible organic linkers.These features allow the structure to undergo contraction, expansion, or some specic distortion [20][21][22][23][24][25][26][27][28][29] in response to changes in temperature and pressure, or upon lling of guest molecules in the voids. 30,31hese structural responses are neither easy to predict nor to tune.3][34] In this work, we have realized selectivity by exploiting noncovalent interactions, acting in a concerted way, in a exible PCP (Scheme 1).Noncovalent interactions, such as hydrogen bonding, aromatic p-p, and C-H-p interactions are much weaker compared to coordination or covalent bonds, and hence allow geometric exibility, and can also be chemical-nature specic.We explore this possibility in a exible porous coordination polymer containing a supramolecular pocket with tunable void space. 35,36o test our hypothesis, we chose a library of positional haloaromatic isomers containing planar benzene rings.8][39][40][41][42][43] The recognition and separation of these are of high value and remain a grand challenge. 44,45The dihalobenzene isomers have (i) different geometries due to the difference in positions of halogen substitution and (ii) different halogens (uorine: F, chlorine: Cl, and bromine: Br) having different polarizabilities and sizes. 46,47onsidering these differences, we have tested the separation capabilities in a Zn-based PCP {[Zn(o-phen)(ndc)]$(DMF)}, having aromatic-philic supramolecular pockets (o-phen: 1,10phenanthroline, ndc: 2,6-naphthalenedicarboxylate, DMF: dimethylformamide). 48The preference for aromatic planar molecules is due to the unique pore-environment created by two sandwiching o-phen.Such an arrangement offers optimum space to host planar guest molecules like benzene, toluene, xylene, and substituted anilines. 48,49The combination of p/p, C-H/p, and hydrogen bonding interactions in this PCP yields unique molecular preferences.By performing in situ crystallization experiments on isomer-mixtures and single crystal structure determination, it was revealed that the host PCP prefers para-isomers (1,4-substituted) by geometric consideration and Br-substituted guests by chemical preference.As a result, ∼100% selective encapsulation was observed for the para-dibromo isomer from the mixture of para/ortho dibromo isomers, while for the para-diuoro-and dichloro-isomers, the host showed lower selectivities.The cooperativity between geometric and chemical preference improves the selectivity factor, and in the following section, we unveil the key parameters responsible, through structural insights of guestencapsulated crystal structures and rst-principles computational studies.

Results and discussion
A large number of exible PCPs are available; 50 we have selected one that can specically host substituted benzene molecules.The selected PCP is constructed by linking Zn(II)-o-phen node with an NDC linker.This 1D structure, as illustrated in Fig. S1, † self-assembles into a 3D porous structure held together by noncovalent interactions (reported elsewhere, 48 see Fig. S1 †).This creates void spaces between a pair of o-phen of neighboring chains, and hosts solvent molecules (DMF or aromatic molecules) as shown in Fig. 1a.In earlier work, it has been observed that the PCP is porous in nature with intrinsic guest induced exibility.The removal of the guest leads to the shrinkage of the pores, which reverts back to it's original porous structure on adsorption of the guest, conrmed by in situ adsorption-PXRD measurements. 48In the presence of DMF, the o-phen/o-phen distance is ∼8.5 Å, while for toluene, it reduces to about 7.5 Å. 48Hence the pore size is adjusted in accordance with the chemical nature of the guests.Similar changes were observed earlier, in the case of xylene isomers, i.e., structural adjustments induced by the geometry of the guest. 49The above features of the PCP prompted us to study its selectivities for haloaromatic isomers.
Two classes of halo-substituted benzene isomers are considered for separation using the selected PCP: ternary or binary mixtures of (i) monohalobenzenes, namely, uorobenzene (FB), chlorobenzene (CB), and bromobenzene (BB) and (ii) dihalobenzenes with halogen substitutions in ortho, meta and para congurations.For the latter class, only mixtures containing positional isomers of a given dihalobenzene, namely, diuorobenzene (DFB), dichlorobenzene (DCB), and dibromobenzene (DBB) are used.The PCP is assessed for a more challenging task of separating isomers of a given dihalobenzene having similar physicochemical properties rather than those with different halogen substitutions whose boiling points differ substantially.To probe the selectivities we have devised an in situ multicomponent crystallization-based experiment, as illustrated in Fig. 1b.In this method, the PCP precursors and the mono-/di-halobenzene isomer mixtures are mixed together to crystallize in the rst set of experiments, and for each of the three dihalobenzenes, we tested ortho/meta, ortho/para, and meta/para mixtures (see ESI section S3 and Table S1, S2 † with nomenclature guest@1).Aer the in situ crystallization experiments, the crystals were washed thoroughly and digested for characterization using 1 H-NMR experiments (see ESI section S4 and Fig. S3-S7 †).The relative amounts of the isomers estimated by the 1 H-NMR experiments are shown in Fig. 1b-d (see Fig. S3-S7 †).For all three dihalobenzenes, a common preference of ortho < meta < para is observed.However, there are two noticeable differences in the case of DBB compared to DFB or DCB: (i) the ortho/meta and ortho/para selectivity factors are nearly ∼99 (see ESI Table S3 †) for DBB and (ii) the meta/para selectivity factor is lower for DBB (Fig. 1b).Similar experiments were carried out using the ternary mixtures, i.e., with mixtures of ortho, meta, and para isomers of each dihalobenzene (See Fig. S6 †).The selectivity trend is ortho < meta < para, consistent with the binary mixture experiments (Fig. S3-S6 †).It is worth noting that ortho isomers of DCB and DBB are exclusively rejected in their respective series.These experiments demonstrate that the PCP shows (i) a geometric preference for the para isomers and (ii) relative concentrations as well as selectivity factors, in general, improve with larger halogen size.To conrm this chemical preference, we have carried out in situ multi-component crystallization experiments for the mixture of uorobenzene (FB), chlorobenzene (CB), and bromobenzene (BB) (see Fig. S7(a-c) †).The preference order is FB < CB < BB (Fig. 1d).This observation is in accordance with the results observed for binary and ternary dihalobenzene isomer mixtures.Such a concurrent geometry and chemical preference is unique.Additionally, the PCP is capable of selectively adsorbing halobenzene isomers, which has been demonstrated from the liquid-phase batch reactions using activated PCPs (see section S5, Fig. S14-S18 and Table S4 †).The relative concentrations and selectivity values precisely match with the binary/ternary in situ crystallization results.Furthermore, at a relatively lower concentration of p-DBB in the mixtures of p-DBB and o-DBB, p-DBB is preferentially adsorbed by the PCP, as can be seen from Fig. S18.† To understand the unique selectivity properties of the PCP, it is important to gain insights into the host-guest noncovalent interactions.To this end, we have carried out (i) crystal structure determination of the individual isomers encapsulated in the supramolecular pores using SCXRD experiments, (ii) periodicdensity functional theory (DFT) calculations of the determined crystal structures to quantify the binding energies and study the extent of noncovalent interactions, and (iii) estimation of the enthalpy of guest binding (DH) using differential scanning calorimetry (DSC) experiments.
We could synthesize and characterize the individual monohalobenzene and dihalobenzene isomer-encapsulated PCP single crystals (guest@1) (CCDC no.2266788-2266798).In all cases, the guest molecules are sandwiched between the o-phen rings, as illustrated in Fig. 2 (see ESI section S6 †).This spatial geometry allows for facile p-p stacking to stabilize the guest@host crystal structures.In Fig. 2a-c, the three isomers of DFB encapsulated in the supramolecular pocket are shown with details of the orientation and p-p stacking.For o-DFB@1, the C-F bond is oriented inwards with a twist angle of 40°(angle between the a-axis and the C-F bond) and the guest is not sandwiched symmetrically by the two o-phen rings with centroid-to-centroid distances of 3.986 and 4.397 Å.This is because the aromatic planes of the guest and o-phen are not parallel (Fig. 2a).However, for both m-DFB@1 and p-DFB@1, the two centroid-to-centroid distances are identical and are shorter than that for o-DFB@1, yielding distances of 3.921 and 3.970 Å, respectively.This clearly indicates that the p-p stacking interactions are weaker for the ortho isomer and hence it is  (d-f) dichlorobenzenes, i.e., o-DCB@1, m-DCB@1 and p-DCB@1, (g-i) dibromobenzenes, i.e., o-DBB@1, m-DBB@1 and p-DBB@1.Color schemeframework: hydrogen (white), carbon (silver), nitrogen (blue), oxygen (red), and zinc (cyan).Guests: carbon (gray), fluorine (green), chlorine (orange), and bromine (purple).Hydrogen atoms of the guest molecules are not shown for clarity.The guest molecules are represented by a thicker "ball-and-stick" representation.
least preferred (in accordance with the observed selectivity).The spatial geometries of the ortho isomers of DCB and DBB are slightly different compared to the case of DFB, as shown in Fig. 2d and g.The C-Cl/C-Br bonds are projected outward with twist angles of 20°(Cl) and 30°(Br) and the aromatic planes of the guest and o-phen are nearly parallel.However, for the meta and para isomers, spatial positions are very similar to those of any of the halogen substitutions, except for small differences in the centroid-to-centroid distance (see Fig. 2).For the monohalobenzenes, no distinct difference in the guest spatial geometry was observed (Fig. S19 †).These insights provide a qualitative explanation of the guest encapsulation and spatial geometry.The origin of preferential encapsulation of guests however, remains unclear.
To rationalize the observed order of guest preference, a quantication of the host-guest interaction strengths and specic host-guest interactions are required.To this end, we have employed periodic-density functional theory (DFT) calculations using the Gaussian plane wave 51 method implemented in Quickstep 52,53 of the CP2K-7.1 package 53,54 (see the ESI for details).First, periodic-DFT-based cell optimizations (CO) of supercells of all the experimentally determined crystal structures of guest@1 were performed.Later, the optimized structures were used to generate electron density difference maps from single point energy calculations (see section S8 †) as a visual aid to identify the regions of PCP which interact with the guest molecules (Fig. 3-5 and S26 †).Among DFB isomers, ortho-DFB exhibits four possible interactionsone hydrogen (HB) bonding and three C-H/p interactions (see Fig. 3a).For each of the meta and para isomers, eight interactions are seen, as illustrated in Fig. 3b and c.Although the types and number of interactions are the same for the meta and para isomers, p-DFB and m-DFB show prominent gain in electron density around hydrogen atoms in their respective crystal structures (Fig. 3e  and f).The calculated binding energies (ESI section S8, Table S17 †) are −124.90,−124.70 and −108.60 kJ mol −1 for para, meta, and ortho DFB, respectively, and this order is in accordance with the experimental observation (see 1 H NMR results and the selectivity order, Fig. 1c).
In the case of DCB, the ortho isomer exhibits two types of interactions: two of its hydrogen atoms have close contact with the framework, one forming a C-H/p and the other forming a HB with NDC and carboxylate oxygen of the framework, respectively.In the case of meta and para isomers, the interactions are more in number, as shown in Fig. 4.However, a noticeable difference in the case of o-DCB is the presence of a halogen bonding interaction with the PCP.When a halogen atom, 'X', is bonded to another group 'R', X suffers a depletion of electron density near its polar region with respect to the R-X bond, called a "s-hole".A halogen bond emerges when an electron-rich group 'A' (A = lone pair, p-electrons, etc.) interacts with the electron-depleted region of X.Such a setting demands :R−X/A = 180°. 55This geometry-based criterion for halogen bonding identication is a necessary condition but not a sufficient one, especially for the case of guests conned within MOF/ PCP pores. 56The electronic structure, however, denitively reveals the presence of such interactions.Hence, to elucidate this specic interaction (blue-dashed line in Fig. 4a), we have calculated maximally localized Wannier functions (MLWFs), and their corresponding centersmaximally localised Wannier function centers (MLWFCs), [56][57][58] MLWFs (violet surfaces) and Fig. 3 (a-c) The guest encapsulated cavities of (a) o-DFB@1, (b) m-DFB@1, and (c) p-DFB@1, respectively.o-phen rings sandwiching the guest molecules are omitted for clarity.The colored dashed-lines represent the type of supramolecular interactions (green and magenta lines represent C-H/p and hydrogen bonding interactions, respectively).Details of atom names, bond lengths, and angles associated with each interaction are shown in Table S18.† (d-f) Electron density differences in the vicinity of the guest in a cavity of (a) o-DFB@1, (b) m-DFB@1, and (c) p-DFB@1, respectively.Violet and magenta surfaces represent electron density gain and loss of magnitude 5.0 × 10 −4 a.u., respectively.The guest encapsulated cavities shown here were obtained from the guest@1 super cells optimized using periodic-DFT.The color scheme for the atoms is identical to that in Fig. 2.
MLWFCs (yellow spheres) of C]O1 (Fig. 4g-h) show that the pelectrons of the double bond face the C-Cl axis.
Such an orientation indicates the interactions between the C-Cl s-hole and the electron-rich p-clouds of the C]O1 bond.This observation explains the difference in the spatial orientation of o-DCB and o-DFB.However, this additional halogen bonding is not substantial enough to affect the selectivity trend noted earlier.Furthermore, among the stacked-benzene dimers, the "slip" geometry is more favored than the "faceon" geometry, and the energy difference is about 4 kJ mol −1 . 59,60ence, p-DCB, the only DCB isomer having a slip w. r. t. the ophen rings is more stabilized in the cavity than the other two isomers.Furthermore, binding energy calculations corroborate the selectivity preference of ortho < meta < para, with binding energies of −134.40,−142.70, and −144.00 kJ mol −1 , in the same order (see Table S17 and Fig. S27 †).For the DBB isomers, the nature of interactions is very similar to that of DCB.The o-DBB isomer, again, shows halogen bonding interactions; however the total possible interactions are still fewer compared to those of meta or para isomers.This is illustrated in Fig. 5.A noticeable difference is the change in the halogen bonding angle; for Br the halogen bond angle C-O-Br/Cl is larger, close to the ideal 180°(see Fig. S25 †).This indicates stronger halogen bonding for Br, and this is in accordance with the polarizability trend of the halogens. 55o-DBB also has a small region of electron density in the vicinity of the other halogen atom (Fig. 5d) alluding to another halogen bonding interaction not present for o-DCB in its crystal structure.Irrespective of this, the selectivity trend remains identical to that of DFB or DCB with binding energies of −138.50, −150.70, and −151.90 kJ mol −1 , for o-, m-, and p-DBB, respectively (ESI, Table S17, Fig. S27 †).We noted that due to the nonpolarizable nature of uorine, 55 o-DFB does not participate in halogen bonding interactions with the PCP.Consequently, the position of o-DFB within the cavity is distinctively different from those of either o-DCB or o-DBB.On the other hand, the corresponding orientations of meta and para isomers within the PCP cavity are similar for all three dihalobenzenes, and so are the guest-framework interactions.However, differences in the binding energies among meta (or para) isomers of dihalobenzenes are signicant for dihalobenzenes with different halogen substitutions (Table S17 †).Halogen atoms of meta or para dihalobenzenes interact via hydrogen bonding with the framework.Due to the increase in polarizability with halogen size, the strength of hydrogen bonding interactions is in the order F < Cl < Br; 55 thus the binding energies are in the order p-DFB < p-DCB < p-DBB (similarly for meta isomers) (ESI, Table S17 †).In addition to differences in the hydrogen bonding strengths, other factors can also bring about differences in the binding energies.Thus, the PCP not only discriminates positional isomers of dihalobenzenes, but also on the basis of the halogen or, simply, based on the chemical nature.
Additional conrmation for the selectivity was obtained through differential scanning calorimetry by experimental quantication of the guest binding enthalpy (DH).We observed that the order of binding enthalpy is identical to the selectivity trend obtained in the in situ crystallization experiments.For the dihalobenzenes, from ortho to para, the binding enthalpy increased linearly.However, the energy difference between the different halogen isomers is negligible, e.g., the enthalpy values for the ortho DFB, DCB, and DBB are 69.53 ± 4.39, 76.86 ± 2.28 and 71.92 ± 1.14 J g −1 (see Fig.

Conclusion
In conclusion, we have demonstrated selective encapsulation of halobenzene isomers using a exible porous coordination polymer.The current work showcases the prominence of concerted supramolecular interactions for the optimal and selective placement of guest molecules in a geometrically ordered position.This resulted from the directionally specic, short-range noncovalent interactions inside the nanospace.The adaptive nature of the PCP preferentially sorts para-> meta-> ortho dihalobenzene geometrical isomers and bromo-> chloro-> uoroin the case of monohalobenzenes.This preference is experimentally demonstrated by crystallization and separation experiments.All the experimental results were further corroborated by extensive computational studies such as ab initio cell optimizations, electron density difference mappings and periodic density functional theory (DFT) analyses.Additionally, theoretical binding energies and experimentally determined enthalpies unequivocally supported the order of selectivity obtained by the crystallization experiments.This unique preference of the supramolecular framework is attributed to the strength of the noncovalent interactions inside the pocket, but not to the type or number of interactions.The ndings presented here can be a tool guide to design more advanced supramolecular systems, which can perform easy recognition and separation of challenging isomers and isotopes.

Fig. 1
Fig. 1 (a) Illustration of a nanospace created between the o-phen of adjacent 1D chains of PCPs, (b) experimental procedure for in situ crystallization of binary aliquot mixtures of dibromobenzenes (DBB), and the right panel shows reaction vessel containing PCP precursors i.e., 2,6-NDC, o-phen, and z Zn(NO 3 ) 2 $6H 2 O in DMF,bar diagram showing relative amounts of isomers encapsulated after binary in situ crystallization of positional isomers of (c) dihalobenzene and (d) monohalobenzenes.[note: the components of each binary mixture for difluorobenzenes (DFB), dichlorobenzenes (DCB) and monohalobenzenes are mentioned at the top and bottom of the respective plots along with their relative concentrations].

Fig. 4
Fig.4The guest encapsulated cavities of (a) o-DCB@1, (b) m-DCB@1, and (c) p-DCB@1, respectively.o-phen rings sandwiching the guest molecules are omitted for clarity.Details of the atom names, bond lengths and angles associated with each interaction are shown in TableS19.† (d-f) Electron density differences in the vicinity of the guest in a cavity of (a) o-DCB@1, (b) m-DCB@1, and (c) p-DCB@1, respectively.The coloring scheme used in this figure is identical to that in Fig.2.(g and h) Reproduction of Fig.4(a) with maximally localised Wannier functions (MLWFs; violet and magenta surfaces) and maximally localised Wannier function centres (MLWFCs; yellow spheres) for the C-Cl s-bond (of o-DCB) and electron-pair in p-bonds (PB) of the framework C]O1 bonds.Violet and magenta surfaces represent MLWFs of opposite phases with an isovalue of 5.0 × 10 −2 a.u., with an exception of the s-bond of C-Cl which is shown for an isovalue of 1.5 × 10 −1 a.u.:C-Cl/PB are 125.04°and 146.10°, confirming that the p-bond electrons indeed interact with the s-hole of the chlorine atom.All the guest encapsulated cavities shown in this figure were extracted from the respective guest@1 supercells of periodic DFT calculations, post cell optimization.
S20 †).The difference due to the presence of different halogen atoms becomes clear by comparing the binding enthalpy values of the monohalobenzenes.These are 39.55 ± 1.81, 103.65 ± 2.35, and 116.90 ± 4.50 J g −1 for FB, CB, and BB, respectively (Table S16 †).The order of binding enthalpies measured by DSC experiments and the binding energies obtained by periodic-DFT calculations are identical for various guest molecule encapsulations in the PCP.