Computational NMR Spectra of o-Benzyne and Stable Guests and their Hemicarceplexes

The incarceration of o-benzyne and 27 other guest molecules within hemicarcerand 1, as reported experimentally by Warmuth, and Cram and co-workers, respectively, has been studied by density functional theory (DFT). H-NMR chemical shifts, rotational mobility and conformational preference of the guests within the supramolecular cage were determined, which showed intriguing correlations of the chemical shifts with structural parameters of the host-guest system. Furthermore, based on the computed chemical shifts reassignments of some NMR signals are proposed. This affects in particular the putative characterization of the volatile benzyne molecule inside a hemicarcerand, for which our CCSD(T) and KT2 results indicate that the experimentally observed signals are most likely not resulting from an isolated o-benzyne within the supramolecular host. Instead, we show that the guest reacted with an aromatic ring of the host, and this adduct is responsible for the experimentally observed signals.


Introduction
One of the most exciting and challenging research fields in chemistry emerged in 1985 with Cram's synthesis of a molecule capable of trapping other molecules in its interior. [1] Since then, the chemistry of molecular container compounds has become a challenging and rewarding field of organic chemistry. [2][3][4][5][6] Early container molecules were shown to be able to encapsulate in their cavity almost any component present in the reaction mixture [1] and were therefore called carcerand (from the Latin word carcer, i.e. prison). In these supramolecular hosts the encapsulated guest cannot leave the molecular prison, not even at high temperatures. In contrast, so-called hemicarcerands trap guests that can be liberated at elevated temperatures, with the combination of the host and guest called a hemicarceplex. The process of switching from encapsulation to liberation of the guest in these hemicarcerands was defined by Houk and co-workers as gating, [5,[7][8][9] which involves a change in conformation of the supramolecular container molecule. Moreover, the (de)complexation processes are controlled by a process known as constrictive binding, which is related to the activation barrier required to trap the guest molecule inside the host cavity through a size-restricting portal. [10] Several hemicarcerands have been synthesized by joining two cavitands with several linkers [3,6] and a large variety of compounds have been incarcerated inside hemicarcerands cavities ranging from Xe [10] to large molecules such as ferrocene, adamantine, camphor [11] or C60. [12] Nowadays, hemicarcerands and other classes of molecular containers can be used for many different applications in molecular recognition, catalysis, drug delivery, and storage. [13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29] For instance, Cram and co-workers synthesized the highly reactive cyclobutadiene inside a hemicarcerand, [30] and investigated the binding properties of hemicarcerands that can undergo chemical reactions without guest-release. [31] In the literature, other examples have been given of unstable compounds that exhibit a high stability when encapsulated inside host molecules. [3,[32][33] The finding that two benzene molecules perfectly fit into the cavity of some hemicarcerands, raised the idea of using the latter host molecules to perform bimolecular reactions. [34] Kang and Rebek successfully performed a Diels-Alder reaction between p-benzoquinone and cyclohexadiene in a self-assembling molecular capsule, [4] demonstrating that significant acceleration in the rates of chemical reactions may occur inside container molecules, just as was shown before inside cyclodextrins. [35][36] In addition, a study by Piatnitski and Deshayes demonstrated that photochemical radiation is not only able to initiate reaction inside a hemicarceplex but it is also able to release guests from the host in a controlled manner when the host is designed to be susceptible to photolysis. [37] Hence, the interior of a hemicarcerand has therefore been shown to be a suitable environment in which to synthesize and stabilize highly reactive compounds from thermal and photochemical reactions. The o-benzyne molecule (see Figure 1), another vulnerable species that does not survive in solution or the gas phase, posed one of the most intriguing systems. Its existence was shown [38] by NMR experiments on the photochemically generated o-benzyne molecule inside a molecular container. Soon after, the existence of o-benzyne was substantiated by low-level quantum chemistry calculations of the NMR spectra. [39] More accurate calculations at coupled-cluster level [40] and DFT level [41] however showed significant deviations (ca. 1.0-1.5 ppm) from the experimental data, even though these methods usually give a much smaller deviation (ca. 0.3-0.5 ppm). The most likely origin for this difference is probably the fact that these calculations were done on the isolated molecule, and not as a guest in the molecular container. However, given the high sensitivity of NMR, it cannot be completely discarded that the experimentally observed spectra do not belong to the o-benzyne molecule. Furthermore, benzyne was found to react with the hemicarcerand, so that NMR spectra attributed to benzyne are suspect. 43 It was also shown that the assignment of a molecular structure by experimental data alone can be tricky and may lead to wrong assumptions. [42]  So far, the theoretical investigations of structures and dynamics of hemicarcerands have mainly involved force field (MM2, MM3, AMBER) and semiempirical calculations. [9,12,[43][44][45][46][47][48] While these methods calculate geometries and energies with sufficient accuracy to predict complexation behavior, it has been shown to have problems describing the unusual environment inside a hemicarcerand or hemicarceplex cavity accurately. This immediately suggests that more sophisticated methods are required on the full host-guest system to accurately analyze its electronic structure. [49] Moreover, reports on the calculations of NMR parameters for hemicarceplexes are sparse, even though the 1 H-NMR spectroscopy has been proven to be a very valuable tool for determining the presence of guest inside a host and the stoichiometry of complexation. With this in mind, the overall aim of this work is to study computationally the structure and 1 H-NMR chemical shifts of hemicarcerand 1 (see Figure 2) and its corresponding hemicarceplexes with o-benzyne and a variety of 27 acyclic, cyclic or aromatic guests for which experimental data are available for comparison. [38,50] In particular, we have predicted the 1 H and 13 C-NMR chemical shift constants of the isolated and incarcerated guests. Additionally, the rotational mobility and the conformational preference are described for the guest molecules.

Results and Discussion
The well-known hemicarcerand 1 is globular-shaped and is composed by attaching two tetraaryl bowls to one another and their rims through four -O(CH2)4O-hemispheric bridges (see Figure 2). Likewise, four R groups (R= C6H5CH2CH2, CH3(CH2)4) are attached to each bowl at their bases in 1. [50] X-ray structural determination has not been reported for empty 1, and therefore our initial model system for the host is based on all the key features of the reported X-ray structures for the host-guest systems 1@G (where G is any of 6H2O, 1,4-I2C6H4, p-xylene, C6H5NO2, (CH3)2NCO(CH3) or 2-BrC6H4OH). [50] To reduce the computational effort, the R groups on the outside of 1 were replaced with methyl groups since they are expected to have a minor influence on the binding properties once the guest is inside. Indeed, the optimized structures at the PBE-D/TZ2P level of hemicarcerand 1 (R= C6H5CH2CH2, CH3) showed that the effect of these groups on the core structure of the molecular container is negligible. Likewise, our model structure of host 1 (R= CH3) presents a conformation where the methylene bridges -(CH2)4-are all distributed on the outside of the container, and the upper hemisphere is twisted by 17º with respect to the lower hemisphere, which agree well with the twist angle of 15º reported for the X-ray structure of 1@6H2O. [50] In contrast, a difference of 6º was found if compared with the optimized structure reported by Liddell and co-workers (twist of 23º, semiempirical AM1 method). [49] The distance between the two oxygen atoms of each -O(CH2)4O-bridge varies between 2.80 and 2.82 Å and the separation between the two parallel square planes defined by the aryl carbons (bonded to H), used to define the length of the polar axis, is around 9.85 Å. For the geometry optimizations of hemicarceplexes 1@G, the guest molecules were introduced into the host starting from several initial host-guest geometrical configurations, placing the guests along the long polar and shorter equatorial axis of the host.

Guest molecules inclusion within host
First, we focus on the 28 different guest molecules that were incarcerated in the host 1. Besides o-benzyne, the guest molecules can be divided into classes A-F regarding their shapes. Class A contains acyclic aliphatic compounds of three to six non-hydrogen atoms containing zero to two branches, class B includes five-membered ring compounds and classes C-F contain aromatic guests in which all the structures tend to be planar except for the methyl fragments of 1,4-(CH3O)2C6H4 (16d) (see Figure S1). The particular case of o-benzyne will be discussed separately. Hence, in the following we present the results for the other 27 guest molecules: Table 1 shows the calculated 1 H-NMR chemical shift values (δ) of the classes A-F guests, both on their own as well as when encapsulated inside host 1; in it also are indicated the changes (Δδ) that occur upon incarceration of the guests, and for comparison also the experimental values are included. [50] In general, we found an excellent agreement between the calculated and experimental shift values, especially for the isolated molecules where 21 of the 27 guests show differences in chemical shifts of the order of ±0.3 ppm or less in all their shifts (see Δδ of G, Table 1). On the other hand, only 7 of the 27 incarcerated guests (1@G) were found with similarly low Δδ values (see Δδ of 1@G, Table 1). Accordingly, it appears that the host-guest interactions can alter the 1 H-NMR shift values dramatically and 10.1002/chem.201904756

Chemistry -A European Journal
This article is protected by copyright. All rights reserved. Table 1. Calculated and experimental 1 H-NMR chemical shifts (δ) of free and incarcerated guests and their spectra changes in the chemical shifts of guest protons caused by incarceration (Δδ).

Calculated
Experimental Accepted Manuscript

Chemistry -A European Journal
This article is protected by copyright. All rights reserved.

FULL PAPER
4 the orientation of the guest within host 1 might play a crucial role in some cases. Here, we report the results for the most favorable orientation of the guests inside the molecular container among the different possible orientations that we explored (see Supporting Information). Typically, the chemical shifts of the encapsulated guests are 1-4 ppm shifted up-field, which likely depends on guest orientation and perhaps dynamics. [51] The origin for this up-field shift is the magnetic shielding resulting from the benzene rings in the polar caps, and -OCH2O-and -OCH2Ar-groups, that form the framework of the host. As a result, the 1 H-NMR shift values of class A (acyclic molecules) guests decrease upon complexation, ranging from Δδ= 1.35 (for the Hb proton of (CH3)2NCOCH3,) to Δδ = 4.18 ppm (for the Ha proton of CH3COCH2CH3) (see Figure  S1 for atom labeling

Chemistry -A European Journal
This article is protected by copyright. All rights reserved.  Table 1). Analysis of the 1 H-NMR signals of class B guests (five-membered ring structures) showed a different behavior. First, the large changes for the isolated 4-butyrolactone (6b) in experiment vs. theory, suggests strongly that in the experiment the isolated compound has either another conformation or another electronic structure. We ruled out an orientational effect, thus we varied the model geometry to another realistic structure, which is its protonated form (protonation of the carbonyl oxygen). Our results for the protonated guest show again the anticipated small differences between theory and experiment (Δδ values of ±0.19 or less; see Table S1, SI), which suggests that in this case the protonated form is the one that was detected in the experiment. In the case of cyclopentanone (7b), our calculations clearly indicate an inversion of the assignment of the Ha and Hb spectral shifts. Thus, the experimentally observed peaks correspond to Ha= 2.15 and Hb= 1.94 ppm (isolated guest). A similar case was found for 2-cyclopenten-1-one (9b), where the Hc and Hd signals were also assigned inversely in the experimental study. Applying these changes to the experimental results, we obtained again small differences in the chemical shifts of the order of ±0.42 ppm or less in all their shifts (see Δδ of G, Table 1). The signal assignments of the class B incarcerated guests (1@G) turned out to be more complicated. The large Δδ values may indicate that in the experiment the guests have different conformation or electronic structure when encapsulated (see Δδ of 1@G, Table 1). Similar to empty host 1, these hemicarceplexes possess twisted conformations with angles that vary from 9º (8b and 10b guests) to 16º (9b guest). Moreover, our molecular models, in conjunction with the large decrease of the chemical shifts caused by incarceration (see Δδ, Table 1), suggest that these small guests may have strong interactions with the -O(CH2)4O-bridges of the host. Furthermore, there exists the possibility that larger compounds, including disubstituted (classes C, D and E) and trisubstituted phenyl derivatives (class F), would remain in an extended form along the long polar axis of the host. In this study, the aromatic guests are aligned correctly inside the host (see Figure S2, SI), as reported in Reference 50. For example, a good agreement between the calculated and experimental twist angle was obtained for the C6H5NO2 (12c) guest (0.8º difference between the optimized PBE-D/TZ2P structure and the reported X-ray structure). The class D compounds are the longest and most tightly held rigid guests. In particular, the cavity of the host 1 is spacious enough and complementary for the inclusion of 1,4-(CH3O)2C6H4 (15d). The two -CH3O-groups of the guest nicely fit into the two hemispheres of the host, achieving stabilizing van der Waals interactions with the aromatic rings of the host. This is also consistent with the 1 H-NMR observations that the methyl protons certainly occupy the polar caps (Δδ= 3.80) and whose Aryl-H atoms occupy the equatorial zone (Δδ= 1.30) of the host. It is well-known that phenol derivatives exhibit rotational isomerism. [52] Hence, particular attention has been focused on the aromatic compounds with hydroxyl substituents: 17d, 21e, 23e, 25f and 27f guests. No deviations from experiment were observed for the chemical shifts of the isolated 21e, 25f, and 27f guests suggesting a strong preference to only one isomer. However, the 17d and 23e guests showed large discrepancies according to the experimental values (see Δδ of G, Table 1).
In 2-bromophenol (23e), there exist two isomers (cis and trans) originated from the orientation of the OH group with respect to the Br substituent (see Figure 3). Our results at the PBE-D/TZ2P corroborate the greater stability of the cis over the trans form, in agreement with the results reported in the literature. [53] According to the calculations, these two forms are close in energy in gas phase (ΔE= 3.7 kcal•mol -1 ) and solvent, e.g. in chloroform solution using the COSMO model, has only a small effect on the energy gap (ΔE= 2.0 kcal•mol -1 ). Therefore, the populations of these two forms should be close or comparable and the consideration of only one form to calculate the chemical shifts may not be enough. The analysis of the 1 H-NMR shifts for the cis isomer of the 2-bromophenol (23e) shows that the calculated Ha proton value overestimates the experimental signal by 0.7 ppm. Further on, if to take into account that the calculation for the trans isomer predicts the Ha proton shift at higher field (underestimate the signal by -0.4 ppm), then the experimental value lies somewhere in between these two isomers which might be in fast exchange in the NMR time scale (see Table S2, SI). In the rest of the proton signals that form the aromatic ring, the calculations indicate that the two different pairs of signals, (Hb and He) and (Hc and Hd), were assigned inversely in the experimental study. Thus, applying these changes to the experimental results we obtained again small differences of the order of ±0.18 ppm or less in all the chemical shifts (see Δδ of G, Table 1). Similar considerations can be applied for the incarcerated 2-bromophenol (1@23e), however, in this case the assignment of the signals becomes more difficult because the Ha peak is hidden in the experimental study. For benzene-1,4-diol (17d), the consideration of both cis and trans isomers does not help to reproduce the observed Ha shifts (see Table S2, SI). Therefore, we also explored solvent effects by adding two explicit water molecules that interact with the hydroxyl groups (see Table S1, SI). Interestingly, inclusion of the solvent molecules leads to a large deshielding of the Ha proton NMR signals; the obtained results are now in excellent agreement with experiment, with only 0.1 (Ha) and 0.2 (Hb) ppm difference. This finding indicates for 17d the solvent effects play a larger role than the rotational isomerism.

Chemical shifts of incarcerated hosts 1@G
Even though the guest signals are much more sensitive than the host signals to incarceration, the 1 H-NMR spectra of the hemicarcerands themselves also might provide conclusive evidence for the incarceration of a guest within the container. The

Chemistry -A European Journal
This article is protected by copyright. All rights reserved.
host signals show sets of eight chemically different protons for the inward-(Hi) and outward-pointing acetal protons (Ho), the aryl protons (Ha), and the methine protons (Hm), of which we analyzed the Hi and Ho acetal protons that are the most sensitive to the presence of the guest (See Figure 2 for proton labels). A comparison of the calculated and experimental Hi and Ho chemical shifts of free and incarcerated hosts 1@G (δ), their chemical shift changes caused by incarceration (Δδ), and the corresponding differences between these values are summarized in Table S3 of the Supporting Information. According to the experimental spectra, [50] the Hi and Ho signals are equivalent when a symmetrical guest is encapsulated or when an unsymmetrical guest that is able to freely tumble inside the host in the 1 H-NMR time scale and makes the two halves of the host symmetrical. The important point to note is that for C6H5I (13c), 4-CH3C6H4OCH3 (18d), 2,4-Cl2C6H3CH3 (24f), and 3,4-Cl2C6H3CH3 (26f), the host signals (Hi and/or Ho) are split into two. This suggests that in the presence of the above guests the two halves of the host are not identical on the 1 H-NMR time scale. Thus, in these cases the inability of the guests to rotate around the short axis of the host causes the northern and southern hemispheres to be slightly different. This behavior was also identified in our calculations and, similar to the experimental signals, the deviations between the upper and lower hemispheres are of 0.3-0.5 ppm for the Hi shifts and 0.1-0.2 ppm for the Ho shifts (see δ, Table S3, SI). If we compare our results and those reported by Cram and coworkers, [50] we note a remarkable agreement between the calculated and experimental Hi and Ho shifts for the isolated host 1 and the incarcerated hosts 1@G. The maximum deviations from the experimental values are of ±0.6 or ±0.5 ppm in the Hi and Ho shifts, respectively (see Δδ, Table S3, SI). This reinforces the excellent agreement between experimental and theoretical determination of the 1 H-NMR chemical shifts of these hemicarcerands.

Encarceration of o-benzyne
In 1997, Warmuth made use of guest incarceration inside hemicarcerand 1 to stabilize the o-benzyne in solution. [38] The 1 H-NMR spectra signals for incarcerated o-benzyne appear at 4.99 and 4.31 ppm. The first signal was assigned to Ha, and the latter to Hb. Stronger evidence for incarcerated o-benzyne was obtained from its 13 C-NMR spectrum. Three carbon signals for o-benzyne were found, as shown in Table 2. Additionally, an estimate for the chemical shifts of isolated o-benzyne in solution were obtained by taking the chemical shift differences Δδ (H) and Δδ (C) of incarcerated benzene as a measure for the host shielding and adding them to the observed chemical shifts of incarcerated o-benzyne (see experimental δ of G, Table 2).
In the case of the isolated o-benzyne, the calculated 1 H and 13 C-NMR shifts reported by Jiao and co-workers [39] at the SOS-DFPT-PW91/III level are in reasonable agreement if compared with the data reported by Warmuth. However, discrepancies were found with the NMR calculations of Helgaker and co-workers, where the 1 H chemical shifts differ from the experimental values by more than is usual for such constants and the ordering of the observed protons is in inverse mode between theory (Ha<Hb) and experiment (Ha>Hb). [54] It is important to note that all calculations reported until now were done on the isolated molecule. Computational studies for the o-benzyne as a guest inside a molecular container have not been reported yet. With this in mind, we explored the incarceration of o-benzyne within hemicarcerand 1 (see Figures 1-2), starting from several initial host-guest geometrical configurations and placing the molecule along the long polar and shorter equatorial axes of the hosts. A stable conformation was achieved with the guest aligned along the long polar axis of the host (see Figure S3), in agreement with the minimum energy conformer reported by Cram and coworkers using molecular dynamics. [45] At the KT2/ET pVQZ level, the 1 H-NMR shift constants obtained for the incarcerated obenzyne are of 5.06 (Ha) and 5.33 ppm (Hb). If compared with the spectra signals reported by Warmuth, [38] the calculated 1 H-NMR shift values are of 0.07 (Ha) and 1.02 ppm (Hb) larger than the experimental values (see Δδ of 1@G, Table 2). Unlike experimental data, the host-guest (1@G) 1 H-NMR shifts indicate that Ha is more shielded than Hb. The isolated o-benzyne has a "triple" bond length of 1.25 (at PBE-D/TZ2P) and 1.26 Å (at CCSD(T)/aug-cc-pVTZ), in good agreement with the bond length of 1.24 ± 0.02 Å obtained by Orendt and co-workers by a simulation of the 13 C dipolar NMR spectrum, [55] and the bond length of 1.27 Å reported at the CCSD(T)/6-31G** level. [56] A comparison of the NMR chemical shift results for the isolated o-benzyne molecule are shown in Table 2. Our 1 H-NMR chemical shift calculations obtained with both KT2/ET-pVQZ and CCSD(T)/pcS-3 methods, like those of Helgaker and co-workers, [54] differ from the experimental values by more than is usual for such constants. At the KT2/ET-pVQZ level, the 1 H-NMR shift constants are of 6.77 (Ha) and 7.60 ppm (Hb). If compared with the experimental data reported by Warmuth, [38] we found discrepancies of -0.92 and 0.59 ppm for the Ha and Hb protons, respectively (see Δδ of G, Table 2). Interestingly, it is not only for the encapsulated o-benzyne that we find discrepancies, but also for the isolated form do we observe substantial differences. One should bear in mind that the experimental values of isolated o-benzyne were obtained by assuming the same large incarceration shift of 2.7 ppm for the two 7 protons in o-benzyne, equal to the incarceration shift in benzene. [38] Moreover, the conformational and dynamic effects may play an important role in this case. According to the dimensions of the guest and cavity of the host, the o-benzyne has space to move inside the host molecule and a static optimized geometry may not be enough for obtain accurate chemical shift values. Note also that the results can be complicated by the fact that o-benzyne can react with an aromatic ring of its hemicarcerand host. [45,[57][58][59][60][61] For this reason, we examined the addition product of a Diels-Alder (DA) reaction between the o-benzyne and hemicarcerand 1. o-Benzyne adds to one of the aryl ether units of 1 (diene component) to give the germinal para adduct (see Figure S4, SI). In the 1 H-NMR spectrum of the 1@o-benzyne DA adduct, the proton signals originating from o-benzyne were identified at 6.14 (Ha), 5.21 (Hb), 4.58 (Hc), and 3.12 (Hd). [61] We optimized the 1@o-benzyne DA adduct and calculated the 1 H-NMR chemical shifts. Interestingly, our results indicate an inversion of the assignment of the Hb and Hc spectral shifts. Thus, the experimentally observed peaks correspond to Hb= 4.58 and Hc= 5.21 ppm. Applying these changes, we found an excellent agreement between the calculated and experimental shift values, with small differences of the order of ±0.53 ppm or less in all their shifts (see Δδ of G, Table S4, SI).

Conclusion
In the present investigation, we have analyzed in detail the 1 H-NMR chemical shifts of the hemicarceplexes formed by obenzyne and a variety of 27 guests within hemicarcerand 1 (1@G). Our study, via density functional theory (DFT) at the PBE-D/TZ2P level for geometries and KT2/ET-pVQZ level for the NMR shift constants, provides a new strategy to characterize these challenging host-guest complexes. In the particular case of the o-benzyne guest, we obtained in both isolated and host-guest (1@o-benzyne) cases significant deviations from the 1 H-NMR experimental data and we cannot draw any definite conclusions regarding the assignment of the NMR chemical shifts. Surprisingly, we have shown that the discrepancies between theory and experiment are not due to the incarceration as asserted in earlier studies. The results shown here indicate that it cannot be conclusively determined whether the experimentally observed spectra belong to the o-benzyne molecule or to an adduct with an aromatic ring of the hemicarcerand. An investigation of the conformational and solvent effects combining molecular dynamic simulations and average NMR chemical shift calculations (similar to a recent study on a platinum complex [62] ) is currently underway for the incarcerated o-benzyne.

Computational Methods
All electronic-structure calculations of the isolated guests and the host-guest hemicarceplexes were performed using DFT. Equilibrium geometries were computed in the gas-phase with the Amsterdam Density Functional (ADF) program, [63][64] using the QUILD program [65] with the dispersion-corrected PBE-D [66][67] functional in conjunction with uncontracted Slater-type orbitals (STOs) of triple-ζ quality plus one set of polarization functions (TZ2P). [68] In particular for the guest structures that contain iodide atoms, scalar (SR) and spin-orbit (SO) relativistic effects were included using the zeroth-order regular approximation (ZORA). [69][70][71][72][73] Moreover, and additional optimization of the o-benzyne guest was computed at the CCSD(T)/aug-cc-pVTZ [74][75] method with the CFOUR program. [76][77] All the 1 H and 13 C-NMR chemical shift constants were calculated using the KT2 functional and the conductor-like screening model (COSMO) [78][79][80] for simulating bulk solvation in chloroform. The ET-pVQZ all electron basis set was used for all atoms except iodine, for which we used the TZ2P basis set and the SR and SO relativistic effects were included using ZORA. [69][70][71][72][73] All chemical shift values are reported with respect to tetramethylsilane (TMS) and were obtained with the GIAO method. [81]