Xenon is a hydrophobic, highly polarizable but rather inert gas. The 129Xe isotope (natural abundance of 26.4%) has a spin of I = 1/2 and its NMR parameters including chemical shift and exchange kinetics are very sensitive to surrounding environments. Meanwhile, application of hyperpolarization (hp) technique can increase the signal by more than 104-fold, thus allowing for very sensitive detection1,2. Consequently, xenon has long been used as an inert probe to determine cavity environments of porous materials and organisms such as proteins3,4,5,6. Promise has also been shown for xenon-based magnetic resonance imaging (MRI)7 and target-specific biosensing when a suitable xenon-trapping host is applied8,9,10,11,12,13.

Even with the important applications, especially recent exciting biosensing applications, our understanding of the host-guest chemistry of xenon is limited and development of efficient xenon receptors is highly desirable. As an inert spherical monoatomic gas, xenon lacks an apparent binding site; van der Waals interaction, specifically London dispersion force, is thought to be the main driving force. This makes xenon binding very challenging. Several macrocycle and cage compounds, including hemicarcerands14,15, cyclodextrins16,17, calixarenes18,19,20,21, cryptophanes22,23,24,25,26,27,28,29,30, cucurbiturils31,32,33,34,35,36,37, and recent porous imine cages38,39,40 have been explored for xenon binding, in solution or solid state. Cryptophanes are shown to be outstanding in terms of high affinity and competence for biosensing applications8,9,13,41,42, though, tedious synthesis and sometimes chiral resolution of the intrinsic enantiomers are required for chirality-sensitive detection43,44,45. On the other hand, assembled hosts have the advantages of reversible formation and easy modular accessibility, however, they are barely explored for xenon binding. In the only few reported examples, earlier Rebek’s rigid, hydrogen-bonded “tennis ball” was found to include xenon on 1H NMR observation46. Recently a cyclic peptide-based columnar assembly was shown able to confine xenon in crystalline form47. Meanwhile, metal-assembled Fe4L648 and Co4L649 cages were recently shown able to encapsulate xenon, with a binding constant of 16 M−1 determined for the former.

Herein we report a stable dimeric assembly of a chiral bisurea-bisthiourea macrocycle and its favorable, adaptive-binding ability toward xenon. The assembly represents a tight, completely sealed yet rather soft capsule and encapsulates xenon with a very high affinity for an assembled host. The adaptive-binding dynamics due to induced fit of the surrounding macrocyclic skeletons is revealed not only in crystal form, but also in solution through the unique spectral features of the dimeric assembly.


Design and synthesis

Recently we have developed a set of chiral tetraamino-bisthiourea macrocycles for substrate-induced assembly asymmetric catalysis50. We envisioned that introducing additional carbonyl groups to the diamine moieties on both sides will lead to a type of bisurea-bisthiourea chiral macrocycle M which incorporates multiple hydrogen bond donors (NH) and acceptors (C=S, C=O) simultaneously, and would favor intra- or intermolecular assembly (Fig. 1). The synthesis of the macrocycle is straightforward. After several simple transformations, the 1 + 1 condensation between bis-amine 1 and bis-isothiocyanate 2 furnished multi-grams of the chiral macrocycle in one batch with an excellent yield of 71% (Fig. 1 and Supplementary Methods).

Fig. 1: Synthesis of the chiral bisurea-bisthiourea macrocycle M.
figure 1

The structures of the macrocycle and its hydrogen-bonded dimeric form are drawn in the bottom.

Structures of the dimeric capsule

High-quality crystals of the dimeric capsule M2 were reproducibly obtained by diffusion of n-hexane to a macrocycle toluene solution. Two forms of crystallographically independent dimeric capsules were observed. As shown by the first form in Fig. 2, the two macrocycles adopt a bowl-shaped conformation and crossly overlay to each other through fourfold intermolecular C=O∙∙∙H–N hydrogen bonds (2.81–2.83 Å). The thiourea moieties lie in the bottom and adopt an unusual syn-anti configuration. The forming of two pairs of parallel intramolecular C=S∙∙∙H–N hydrogen bonds (3.39–3.40 Å) in the meantime helps lock the bowl-shaped macrocyclic conformation, leaving four exo NH sites for intermolecular hydrogen bonding. All the phenyl rings from different macrocycles stack onto each other, which may provide further stabilization.

Fig. 2: Crystal structure of the dimeric capsule M2.
figure 2

a Top view. b Side view. c Space filling view. Only the first form (Form 1) of the two crystallographically independent dimeric capsules is shown. H-bonding distances (given by the heavy atom distances): C=O∙∙∙H–N 2.808, 2.832, 2.808, and 2.832 Å; C=S∙∙∙H–N 3.393, 3.399, 3.393, and 3.399 Å.

The buckling of the two bowl-shaped macrocycles produces a completely sealed dimeric capsule (Fig. 2c). The cavity volume was detected to be 72 Å3 for the first form (Fig. 3a). Surprisingly, while the overall interacting motif holds, the second form shows a much smaller cavity (36 Å3; Fig. 3b and Supplementary Table 4). The cavity reduction is due to flattening of the bowl-shaped macrocyclic conformation as reflected by shortened (S=)C---C(=S) and lengthened (C=)O---O(=C) dimensions (Fig. 3b vs 3a), suggesting considerable flexibility of the macrocyclic skeleton. The inter- and intramolecular hydrogen bonding, however, is not significantly affected. The existence of the two forms of dimeric capsules is probably caused by crystal packing effect, as the macrocyclic component has a rather flexible skeleton. In both forms, the solvent molecules were found too large to occupy the cavity.

Fig. 3: Cavity volume of the dimeric capsule M2.
figure 3

a Form 1 (72 Å3). b Form 2 (36 Å3). The cavity volume is determined by a spherical probe of 1.4 Å using SwissPdbViewer54 and depicted in red.

Dimeric assembly in solution

The formation of dimeric assembly in solution was further studied. 1H NMR in (CDCl2)2/DMSO-d6 (5:1) shows a set of simple signals, consistent with an overall D2-symmetric free macrocycle structure in this competitive solvent system (Fig. 4b). In contrast, in pure (CDCl2)2, the spectrum totally changed to two sets of signals at equal intensity, in line with a reduced C2 macrocycle symmetry (Fig. 4c). The NH signals moved largely to downfield direction, suggesting favorable hydrogen bonding formation. By a combination of 1H, 13C and various 2D NMR (Supplementary Figs. 714), a dimeric capsular structure similar to that in crystal can be elucidated with all signals well assigned. The two sets of protons (in blue and purple) can exchange their positions through overturn of the syn-anti thiourea configuration as shown by 2D EXSY NMR (Supplementary Fig. 14; vide infra). As the solvent molecule is too large to fit, most capsules are likely to be empty except those occupied by N2 as indicated by the set of minor N2 inclusion peaks (Fig. 4c). High stability of the assembly was demonstrated by concentration-variable NMR as no essential spectral changes observed until dilution to 0.05 mM (Supplementary Fig. 23). A large dimerization constant of (1.9 ± 0.3) × 104 M−1 was further determined by isothermal titration calorimetry (ITC; Supplementary Fig. 26). The dimeric assembly was also confirmed with high-resolution CSI-MS where a peak at m/z = 2111.5522 corresponding to [M2−H] dominated (Supplementary Fig. 27).

Fig. 4: 1H NMR (298 K, 500 M) of the macrocycle M in different solvent systems with or without xenon.
figure 4

a The chemical structures of the macrocycle and dimeric capsule with related protons labeled. b 1H NMR in (CDCl2)2/DMSO-d6 (5:1). The macrocycle exists as monomeric form. c 1H NMR in (CDCl2)2. The macrocycle exists as dimeric capsule form. Asterisk denotes the minor N2 inclusion peaks (for details see Supplementary Fig. 7). d 1H NMR in (CDCl2)2 after bubbling of xenon. The major peaks correspond to Xe M2. In all cases, the initial concentration of the macrocycle ([M]initial) for preparing each solution is 10 mM.

Xenon binding

The macrocycle can congruously form a stable, sealed dimeric capsule in both solid state and solution with a cavity possibly suited for xenon (dw = 4.3 Å, V = 42 Å3). To our delight, through bubbling xenon gas, the original 1H NMR signals of the free dimeric capsule almost disappeared, instead a new group of sharp and well-resolved signals dominated, which can correspond to Xe M2 (Fig. 4d). All signals were clearly assigned by a combination of 1H, 13C, and various 2D NMR (Supplementary Figs. 1522). The upfield shift of NH signals and drastic changes of other proton signals may reflect a shielding effect of large electron cloud of xenon and an induced structural change due to tight inclusion (vide infra). The remarkable spectral difference is distinguished from the usually reported, relatively rigid hosts, suggesting the dimeric capsule is rather soft (as also shown in crystal) and can adaptively include the incoming xenon guest. The 129Xe NMR showed a similar slow exchange process and the signal of the bound xenon appeared at 169.6 ppm, largely upfield-shifted comparing to the free xenon in solution (Δδ = −53.4 ppm; Fig. 5). A binding constant of 99 ± 4 M−1 was determined for xenon by 1H NMR integration (see Supplementary Methods and Supplementary Fig. 15). To the best of our knowledge, this is the highest xenon affinity reported to date for assembled hosts. As compared, a previous only known binding constant of 16 M−1 was reported for a metal-assembled Fe4L6 host48.

Fig. 5: 129Xe NMR (298 K, 138 MHz) of xenon saturated in (CDCl2)2 without or with the dimeric capsule.
figure 5

a The blank xenon solution. b In the presence of the dimeric capsule M2 (20 mM). The chemical shift of the free xenon was referenced according to the literature22.

The existence of two sets of position-exchangeable protons in the macrocyclic skeleton provides a unique probe for exploring the adaptive xenon binding dynamics. As intracapsular, simultaneous overturn of the many locked syn-anti configurated thiourea motifs seems unlikely, the position exchange is probably accomplished through a disassembly-reassembly pathway, i.e., once disassembled the conformation of the free macrocycle can be readily converted (Fig. 6a). Accordingly, the position-exchange rate can reflect the assembly-disassembly dynamics. The influence of xenon binding can thus be probed by comparing the temperature-variable 1H NMR before and after xenon inclusion (Fig. 6b, c). Upon temperature increasing, the two sets of signals started to coalesce and Xe M2 showed an obviously higher coalescence temperature Tc and accordingly a higher activation energy ΔG than M2 (17.4 ± 0.1 vs 16.4 ± 0.2 kcal mol−1; Supplementary Figs. 24 and 25 and Supplementary Table 6). This suggests upon xenon inclusion, the dimeric assembly becomes more kinetically stable; in other words, xenon binding could have induced a tighter capsular structure by fitting the soft macrocyclic skeletons. It is worth noting that even at elevated temperatures (up to 373 K), the assembly was still kept intact, once again showing its high thermodynamic stability.

Fig. 6: Determination of the xenon binding dynamics.
figure 6

a Proposed exchange mechanism of the two sets of macrocycle protons (depicted in blue and purple) in the dimeric capsule. b, c Partial temperature-variable 1H NMR (500 MHz, (CDCl2)2) of M2 and Xe M2 respectively. Asterisk denotes the emerging peaks of free M2. In both cases, [M]initial = 10 mM. For a complete set of spectra, see Supplementary Figs. 24 and 25.

Structures of Xe  M2

Two different high-quality crystals of Xe M2 were independently obtained by diffusion of n-hexane to a macrocycle solution in chloroform or toluene saturated with xenon (Supplementary Tables 25). The first crystal crystallized in orthorhombic system as the free dimeric capsule with similar cell parameters (vide supra). As shown by the first form of the two crystallographically independent complexes (Fig. 7), the xenon was completely sealed in the cavity with a perfect packing coefficient of 52%51. In order to fit xenon, the cavity volume was expanded to 81 Å3. The xenon atom shows multiple contacts with the surrounding macrocyclic skeletons, including with the four endo nitrogen (3.83–3.87 Å; the sum of van der Waals radii of nitrogen and xenon is 3.71 Å52) and four carbon atoms (3.98–4.01 Å; the sum of van der Waals radii of carbon and xenon is 3.86 Å52) from the thiourea groups, and with the interior carbon (3.86–4.02 Å) and hydrogen atoms (3.23–3.42 Å; the sum of van der Waals radii of hydrogen and xenon is 3.36 Å52) from four diagonal phenyl rings (Fig. 7a). These contacts, especially that with polar nitrogen atoms, are different to those rigid aromatic faces or sole C–H sites in the usually reported xenon hosts. The second form showed an overall similar structure with a cavity of 69 Å3 and xenon packing coefficient of 61% (Supplementary Tables 4 and 5). The cavity size is largely expanded comparing to that free form of 36 Å3 (vide supra), reflecting the adaptive ability of this tight yet soft capsule. Again, the adaptation is accomplished through the adjust of the bowl-shaped macrocyclic conformation, rather than the change of hydrogen bonding distances.

Fig. 7: Crystal structure of Xe  M2.
figure 7

a Top view with the atoms contacting to xenon labeled (only the upside is labeled). b Side view with the front urea moiety cut off for clarity. c Side view with the macrocycle molecular surfaces shown. d Space filling view.

The second crystal crystallized in a different monoclinic system, but exhibited a similar capsular structure and xenon binding motif (Supplementary Tables 4 and 5). The contacts of xenon to the relevant thiourea nitrogen atoms are even shorter (3.69–3.82 Å). It is worth noting that, even both crystals were grown under 1 atm xenon pressure, the averaged xenon occupancy in the cavity reaches 0.85 and 0.94 respectively, suggesting favored binding53. The sealed but reversible assembly capsule system could thus have potentials for sequestration and separation of xenon (see Supplementary Note 1 for detailed discussion).

In conclusion, a chiral bisurea-bisthiourea macrocycle was efficiently synthesized. This macrocycle can form stable dimeric capsular assembly in both solid state and solution. The capsule possesses a tight, sealed cavity, and effectively binds xenon. Different from other rigid hosts, the capsule is rather soft and can adaptively include xenon through induced fit of the surrounding macrocyclic skeletons. Multiple contacts including several unusual contacts to polar nitrogen atoms are engaged. The adaptive binding dynamic was demonstrated not only in crystal, but also in solution by taking the unique spectral features of the dimeric assembly. These knowledges should have advanced our current understanding on the host–guest chemistry of xenon. The sealed but reversible assembly could also have potentials for developing host materials toward xenon sequestration and separation. With the easy synthesis, modification, robust yet adjustable assembly features, further optimization of the xenon affinity, and 129Xe NMR related parameters should be feasible toward biosensing applications. The enantiopure form of the chiral assembly could also have advantages on chirality-sensitive detection.