Dispersion-controlled docking preference: multi-spectroscopic study on complexes of dibenzofuran with alcohols and water

The structural preferences within a series of dibenzofuran—solvent complexes have been investigated by electronic, vibrational, and rotational spectroscopic methods probing supersonic jet expansions. The experimental study is accompanied with a detailed theoretical analysis including dispersion-corrected density functional theory, symmetry adapted perturbation theory, as well as coupled cluster approaches. The complementary, multi-spectroscopic results reveal a preferred OH∙∙∙O structure for the smallest complex of dibenzofuran–water, whereas for the methanol complex an OH∙∙∙π isomer is simultaneously observed. For the largest complex, dibenzofuran– tert -butyl alcohol, only a π-bound structure is found. These comprehensive investigations show that a completely inverse trend regarding the docking preference is observed by comparing the present results with the ones for analogous diphenylether complexes. This can be rationalized on the basis of the planarity/non-planarity and rigidity/flexibility of the different systems, providing valuable insight into the interplay between different non-covalent interactions. This analysis is a further step towards a quantitative description of very delicate energetic balances with the overall goal of yielding reliable structural predictions for non-covalently bound systems.


Introduction
Non-covalent attraction plays a key role in molecular recognition and aggregation as fundamental processes governing (bio)chemical processes [1][2][3][4] . Therefore, already small changes within the interplay of intermolecular forces can affect these processes significantly. London dispersion forces are one of the major contributors, along with Keesom and Debye forces, together which are known as van der Waals forces. These forces contribute with different proportions to various types of hydrogen bond interactions, and they often compete with each other due to their respective directionality. However, they might also reinforce each other instead of rivaling. Analyzing this interplay between non-covalent interactions on a molecular level provides detailed insight into these fundamental driving forces for molecular recognition and aggregation. London dispersion forces are ideally suited to control the formation of a certain molecular structure, as their magnitude sums with increasing interaction surface. This enables the competition with stronger interactions of electrostatic nature and might even lead to the predominance of dispersive attrac-tion over Pauli repulsion 5,6 . The structures of neutral, non-covalently bound molecular complexes involving aromatic moieties and water or alcohol molecules have been extensively studied in the gas phase (cf. e.g. Refs. [7][8][9] and references therein). Among these studies, several complexes involving heteroaromatic moieties are found including works on indole-water 10 , 7-azaindole-water 11 , pyrrole- 12 and carbazole-solvent complexes 13,14 . Moreover, several studies on furan derivatives were carried out, including a comparative FTIR jet and theoretical study on 2,5dimethylfuran-as well as the 2,3-benzofuran-methanol complexes 15 . The latter have additionally been studied by laser induced fluorescence and IR fluorescence dip spectroscopy, including the respective water complexes 16 . For the 2,5dimethylfuran-methanol complex, the OH•••O binding motif was identified as the preferred structure, which exhibits additional CH•••π stabilization by the interaction of the methyl group with the π-cloud. The OH•••π-bound isomer was also observed as a slightly less stable structure, although it turned out to cause a stronger OH stretching red-shift than the OH•••O-bound structure, i.e. the π-cloud causes a stronger distortion of the OH bond than the lone pairs of the ether oxygen. In the case of 2,3-benzofuran-water 16 , a balanced situation was found with coexisting, nearly isoenergetic OH•••O and OH•••π isomers. Within that study, no preferred site could be identified, which agrees with the theoretical predictions of less than 0.5 kJ/mol energy difference for the applied methods , second "analysis" part in preparation). Several spectroscopic studies on dibenzofuran (DBF) have been performed in the condensed phase [18][19][20][21][22][23] and in the gas phase [23][24][25][26][27][28][29][30] , including works on the DBF dimer 27,28,31 and mixed dimers of DBF with fluorene and benzofuran 28,32,33 . Auty et al. 24 recorded fluorescence excitation spectra of DBF and the DBF-water complex. They assumed that the complex is hydrogen-bonded to the oxygen atom of DBF based on the spectral shift of the fluorescence excitation bands. However, there is a lack of ab initio calculations supporting this assumption as well as further, structurally more sensitive spectroscopic experiments.
In previous studies, we established a multi-spectroscopic approach in order to elucidate the preferred binding sites in different aromatic ether-alcohol and -water complexes. [34][35][36][37][38] Within the series of diphenyl ether (DPE) complexes, we have shown that water and small alcohol molecules prefer the πdocking site, whereas larger alcohols preferably bind to the ether oxygen atom. This observation contradicts the intuitive expectation of a preferred π docking, when the size of the alcohol increases. In that study, the respective contribution of London dispersion to the interaction energy for the different complexes was analyzed in order to explain the observed trend. In addition to that, the distortion of the DPE structure, caused by a twist of the phenyl rings upon aggregation of an alcohol or water molecule, was identified as another major aspect influencing the trend.
In order to gain a deeper understanding on the influence of structural deformation upon complex formation, a systematic change in the structure of the ether is valuable. One possible change is the direct connection of the two phenyl rings of DPE, which leads to dibenzofuran (DBF). By doing so, the initial flexibility of DPE is entirely disabled since DBF is planar and rigid. Considering the fact that the π system is delocalized over both phenyl rings via the furan ring, the molecule is expected to remain planar upon solvent aggregation in order to maximize aromaticity. Therefore, there is no deformation of the ether geometry within the solvent complexes that might influence the binding preference, contrary to the case of DPE complexes in which deformation plays a substantial role 37 . Moreover, the twisted structure of DPE was shown to enable CH•••O contacts between ortho CH moieties and the oxygen atom of the alcohol or water molecule. This had a significant influence on the structural preference as well. Since for DBF only inplane CH groups are available for CH•••O contacts, structures mainly interacting via the π-cloud should not be affected by CH•••O contacts. Whatever the relative importance of such qualitative concepts may be, they add up to a computable and experimentally verifiable energy difference between competing solvent docking sites.
Experimental verification of predicted structural preferences of such molecular complexes requires studies on a mo-lecular level, where the isolated molecular aggregates can be investigated without the influence of any environment. These conditions can be fulfilled by molecular beam investigations, allowing the formation of molecular complexes and clusters in a supersonic expansion. A variety of spectroscopic methods can be combined with molecular beam experiments, including FTIR spectroscopy 15,34,35,38 , mass-and isomer-selective IR/UV laser spectroscopy (IR/R2PI 8,[39][40][41][42] ) and chirped-pulse Fourier transform microwave (CP-FTMW [43][44][45] ) spectroscopy. The combination of these different spectroscopic techniques yields complementary results, providing valuable experimental data ideally suited for benchmarking theoretical approaches.
In the present paper, we investigate a series of DBF complexes with water, methanol and tert-butyl alcohol (ROH with R=H, Me, t-Bu) by a multi-spectroscopic strategy, including FTIR, IR/UV and CP-FTMW spectroscopy. The experimental study is accompanied with a detailed theoretical analysis including dispersion-corrected density functional theory as well as wave function-based methods.

FTIR spectroscopy
Linear FTIR spectra were recorded using the 'popcorn' jet setup. DBF (alfa aesar, ≥99%) was deposited on molecular sieve and exposed to carrier gas pulses in a heatable sample compartment enclosed by two poppet valves (opening at 70 mbar differential pressure upstream and either 690 or 350 mbar downstream). Helium was used as the carrier gas at 1.5 bar. A gas pulse from a 0.069 m 3 reservoir picked up the sample and was supersonically expanded into a 3.6 m³ buffer volume. A sufficiently low background pressure is ensured by a pumping system operating at 500 m³/h. Two nozzle variants were applied: a 2×10×0.5 mm double-slit nozzle and a newly designed 60×0.2 mm heatable 'V-nozzle', which is angled (162°) to approximately fit the focused IR beam shape. The alcohols (MeOH (Sigma Aldrich, ≥99.8%), MeOD (eurisotop, 99% D), t-BuOH (Roth, ≥99%)) were introduced upstream of the gas reservoir by a coolable saturator or by using premixed gas bottles. Each gas pulse was probed by a single synchronized scan of a Bruker IFS 66v/S FTIR spectrometer. 100-400 scans were averaged to obtain the final spectrum. More details can be found in Refs. 34,46.

IR/UV spectroscopy
The experimental set-up for the combined IR/UV experiments is described in detail elsewhere 47,42 , thus only a brief description is given here. The experiments were carried out in a molecular beam apparatus consisting of a differentially pumped linear time-of-flight (TOF) mass spectrometer with a pulsed valve (Series 9 with pulse driver Iota One, General Valve, 500 µm orifice) for skimmed jet expansion. DBF was purchased from Merck (≥97.0%). MeOH (Sigma-Aldrich, ≥99.7%) and t-BuOH (Sigma-Aldrich, ≥99.7%) were each supplied via cooled reservoirs and co-expanded with DBF (held at room temperature) using the carrier gas neon (2.5-3.0 bar).
For the R2PI and IR/R2PI experiments, two tunable nanosecond laser systems were necessary, including one independent UV laser system and one IR laser system. The UV laser radiation is obtained via second harmonic generation in a BBO crystal using the output of a dye laser (Cobra-Stretch, Sirah). The latter is pumped by the second harmonic (532 nm) of a Nd:YAG laser (SpitLight 600, Innolas). The IR laser radiation in the range of 3520-3750 cm −1 is generated by difference frequency mixing (DFM) in a LiNbO 3 crystal using the fundamental (1064 nm) of a Nd:YAG laser (Quanta-Ray Pro-230, Spectra-Physics) and the output of a second dye laser (PrecisionScan, Sirah), which is pumped by the second harmonic (532 nm) of the same Nd:YAG laser. The resulting IR radiation is amplified in an optical parametric amplification (OPA) process in a further LiNbO 3 crystal using the output of the DFM process and the fundamental (1064 nm) of the Nd:YAG laser. For the IR/R2PI spectra, the IR laser was irradiated 50 ns prior to the UV laser.

CP-FTMW spectroscopy
The rotational spectra of the DBF-ROH complexes were recorded with the Hamburg CP-FTMW spectrometer COMPACT, which is operated between 2−8 GHz. Experimental details are given elsewhere 48,49 . DBF (stated purity ≥ 98%) was purchased from Sigma-Aldrich and used without further purification. The molecules were seeded into a supersonic expansion using a modified pulse nozzle (Parker General Valve, Series 9, 1.1 mm orifice diameter) equipped with a heatable reservoir. DBF was placed into the reservoir close to the valve orifice and heated to 100°C. The solvent (ROH) was placed in an external reservoir upstream of the valve at a second set of tubing to regulate the amount of carrier gas that was flown over it and thus to regulate the amount of solvent. For all of the experiments, neon (3 bar backing pressure) was used as a carrier gas to form a supersonic expansion into the vacuum chamber. Additional experiments with helium as a carrier gas (3 bar backing pressure) were performed for DBF-MeOH. For each gas pulse, the ensemble of molecules was polarized with a series of eight microwave chirps of 4 µs duration spanning 2-8 GHz, following the fast-frame approach 50 . The chirps were generated with an arbitrary waveform generator, amplified by a 300 W travelling wave tube amplifier, and transmitted into the vacuum chamber with a horn antenna. Following each excitation chirp, 40 µs of the free induction decay (FID) of the macroscopic ensemble of polarized molecules was recorded, yielding a frequency resolution of 25 kHz. For the experiments, a total of 5 million averages (for DBF-H 2 O) and 2 million averages (for DBF-MeOH and DBF-t-BuOH, respectively) were co-added and Fourier transformed with a Kaiser-Bessel window function to give the broadband rotational spectrum in the frequency domain.
All spectra were first fit to an asymmetric rotor Hamiltonian using the JB95 program 51 . The transition frequencies were then refined using the AABS program suite, and the final asymmetric rotor Hamiltonian fits were completed with SPFIT 52 . Line lists for all three dimers are provided in the ESI. An analysis of the observed tunneling splitting arising from internal rotation of the methanol methyl group in the dibenzofuran-methanol complex was performed using the XIAM program 53 . XIAM is a least squares fitting program specifically designed for analyzing spectra of molecules exhibiting internal rotors by employing the combined axis method of Woods to account for internal rotation through a potential barrier.

Computational Methods
Input structures were manually constructed with Avogadro 54 using the MMFF94s force field 55 66,67 . Similarly, calculations were performed with the M06-2X functional 68 including the D3 correction and the def2-TZVP basis set. The SCS-CC2 calculations were carried out with the aug-cc-pVDZ 64 and def2-TZVP basis sets using Turbomole 7.3, while correspondingly aug-cc-pVDZ-cbas 69 and def2-TZVP-cbas 69 were chosen as the auxiliary Coulomb fitting basis sets (cbas) required by the ricc2 module for the RI approximation. Harmonic vibrational frequencies at the SCS-CC2 level were calculated with the NumForce script of Turbomole 7.3. All geometries were confirmed to be minima with only real harmonic vibrational frequencies. All DFT and SCS-CC2 energies were corrected for the basis set superposition error (BSSE) by applying the counterpoise correction method 70 . DLPNO-CCSD(T) single-point calculations for the B3LYP-D3(BJ)/aug-cc-pVTZ geometries were carried out with ORCA 4.0.1 71 using the cc-pVTZ and cc-pVQZ basis sets 64 with corresponding cc-pVTZ/C and cc-pVQZ auxiliary basis sets 69 for the RI approximation. Additionally, the "TightPNO" 72 and "TightSCF" options were applied. For comparison of zeropoint-vibrational-energy-(ZPE-)corrected energies, harmonic ZPE corrections obtained at the B3LYP-D3(BJ)/aug-cc-pVTZ level were added to the DLPNO-CCSD(T)/cc-pVTZ and DLPNO-CCSD(T)/cc-pVQZ energies. Furthermore, a local energy decomposition (LED) scheme 73,74 was applied within the DLPNO-CCSD(T)/cc-pVQZ calculations. This was mainly used for extracting physically meaningful dispersion contributions to the total interaction energies. For comparison, second order SAPT(0) calculations 75 were carried out with the truncated juncc-pVDZ basis set 64,76,77 , using the PSI4 program 78 . In order to analyze the respective contributions to the interaction energy of the investigated complexes, SAPT(0)/juncc-pVDZ calculations were performed (cf. Table S1) as well as more sophisticated DLPNO-CCSD(T)/cc-pVQZ calculations for which a local energy decomposition (LED) scheme was applied (cf. Table S2). As expected, both approaches yield a larger dispersion contribution in the OH•••π motifs and significantly more electrostatic contribution for the OH•••Op structure. This supports the finding that the OH•••O hydrogen bond, dominated by electrostatics (cf. Table S1), combined with the CH•••O contact leads to a stronger stabilization than the two OH•••π contacts within the other isomers.
For the DBF-MeOH complex, the calculated minimum structures at the B3LYP-D3(BJ)/def2-TZVP level are depicted in the second row of Fig. 1. In this case, two oxygen-bound isomers are found: within the C s -symmetric OH•••Ot isomer (on top), the methyl group of the methanol molecule is positioned above the furan ring, enabling CH•••π interactions. The second oxygen-bound structure is denoted as OH•••Op and exhibits an OH•••O hydrogen bond in the DBF plane with the methyl group pointing away from the DBF plane. Therefore, the OH•••Op isomer is lacking CH•••π interactions in contrast to the OH•••Ot arrangement. However, the in-plane hydrogen bond allows for a stabilizing CH•••O contact between the MeOH oxygen atom and a CH group of DBF (cf. Fig. 1), similar to the OH•••Op isomer of DBF-H 2 O. The OH•••π6 isomer is bound via an OH•••π contact, and it is stabilized by CH•••π interactions of the methyl group with the π-cloud. In contrast to the related systems 2,5dimethylfuran-MeOH and 2,3-benzofuran-MeOH 16,15 , no minimum structure is found with an OH•••π interaction involving the five-membered furan ring. This might allow for the conclusion that within the extended π system of DBF, the sixmembered benzene rings are better hydrogen bond acceptors than the furan moiety. As discussed in previous works on 2methylfuran, 2,5-dimethylfuran, and 2,3-benzofuran 16,15,17 , the furan oxygen acceptor site loses attractiveness upon the introduction of methyl groups or one phenyl moiety.
Analyzing the different energy contributions shows that the largest dispersion contribution is in the OH•••π6 isomer, followed by the symmetric OH•••Ot structure (cf. Tables S1 and S2). Both arrangements contain CH•••π stabilization. Accordingly, the dispersion contribution is significantly lower in OH•••Op, and the structure is clearly dominated by the electrostatic   Table S3, ESI), indicating a clearly weakened hydrogen bond as it largely deviates from an ideal linear hydrogen bond. This significantly affects the OH stretching frequencies, which will be discussed in the Experimental Results section.
The optimized minimum structures for the t-BuOH complex are shown in the last row of Fig. 1. Similar to the methanol complex, two OH•••O structures and one OH•••π arrangement are found as minimum geometries. The OH•••Ot isomer is C ssymmetric, identical to the OH•••Ot isomer of the corresponding MeOH complex. The t-Bu moiety is positioned on top of the furan ring leading to CH•••π interactions with the π-cloud. In contrast to the analogous MeOH complex, the hydrogen bond is less bent (177° for t-BuOH vs. 155° for MeOH, B3LYP-D3(BJ)/def2-TZVP level, cf. Table S3, ESI).
In the OH•••Op isomer, the t-BuOH moiety is tilted to one side, which indicates a slight CH•••O interaction between the alcohol oxygen atom and a neighboring CH group, resembling a somehow distorted version of the OH•••Op isomer of DBF-MeOH with the alcohol being located rather above the π plane due to stronger CH•••π interactions. Comparing all non-C s symmetric OH•••Op structures, the solvent molecule increasingly approaches the π-cloud above the DBF plane going from water to t-BuOH. In the respective OH•••π6-bound isomer, the t-Bu moiety is in closer proximity to the DBF π-cloud, resulting in a larger interaction surface for CH•••π interactions compared to the OH•••Op isomer. However, CH•••π interactions should be of similar magnitude in the OH•••π and the C s symmetric OH•••Ot isomer. This is in line with dispersion contributions obtained at the SAPT(0)/jun-cc-pVDZ and the DLPNO-CCSD(T)/cc-pVQZ levels, which are similar for the two isomers, but clearly smaller for the OH•••Op isomer. As hydrogen bond interaction is indicated by the OH stretching red-shift, calculated OH stretching frequencies can be compared for the competing structures. Calculations at the B3LYP-D3 level suggest a stronger OH•••O hydrogen bond compared to OH•••π. However, the contrary is predicted at the M06-2X/def2TZVP and SCS-CC2 levels, indicating a stronger OH•••π acceptor compared to OH•••Ot. This aspect will be discussed later in the Experimental Results section.
The calculated relative energies for all DBF-ROH complexes at different levels of theory are found in Table 1. The values for the B3LYP-D3(BJ) and SCS-CC2 levels result from geometry optimizations and harmonic frequency calculations, whereas single point calculations were performed at the DLPNO-CCSD(T)/cc-pVQZ level using the geometries obtained at the B3LYP-D3(BJ)/aug-cc-pVTZ level. No anharmonic treatments of the ZPE were used, as anharmonic corrections are expected to be small (assumed to be in the order of <0.5 kJ/mol), and they have furthermore proven to perform non-systematically in relative energy predictions for similar systems 17 . As shown in previous studies 15 , the structures with a rather localized OH•••O hydrogen bond contain more ZPE than OH•••π-bound structures. This is reflected in a consistent OH•••Op destabilization on the order of 0.7-2.3 kJ/mol with respect to OH•••π isomers when electronic (∆E) and ZPE-corrected energies (∆E 0 ) are compared.
For the water complex, the prediction of the energetic order is almost uniform: the OH•••Op structure is preferred by 0.6 up to 3.9 kJ/mol, depending on the theoretical level and basis set. The highest applied level suggests the oxygen site to be preferred by about 2 kJ/mol, which might raise questions about the population of a π-bound structure in molecular beam experiments unless major isomerization barriers prevent relaxation. The SCS-CC2 and M06-2X/def2-TZVP calculations (cf. Table S4, ESI) prefer the OH•••π5 structure. Regarding the DBF-MeOH complex, a rather undecided situation is found with an oscillation of the energetic order between an OH•••π6, OH•••Op and even OH•••Ot preference. The ZPE destabilization of the oxygen-bound structures compared to the OH•••π6 Table 1 Relative energies for DBF-ROH complexes with (∆E0) and without ZPE correction (∆E) obtained at different levels of theory. All values are given in kJ/mol and include BSSE correction.  Table 1). Considering the energetic range of 0.5 kJ/mol for all three binding motifs obtained at the DLPNO-CCSD(T)/cc-pVQZ level -being certainly within the error bar of the method -would not exclude the presence of more than one isomer in molecular beam experiments. In the case of DBF-t-BuOH, the predicted binding preference is undecided as well among the different applied computational approaches: the symmetric OH•••Ot isomer is preferred at the B3LYP-D3(BJ)/def2-TZVP level, whereas the larger aug-cc-pVTZ basis leads to an OH•••π6 preference, together with the SCS-CC2 approach. Finally, the DLPNO-CCSD(T) approach favors the OH•••Op structure by 0.5 kJ/mol over OH•••π6. Overall, the relative ZPE-corrected energies of all three binding motifs are predicted to be within a range of 1.6 kJ/mol. Thus, the simultaneous presence of more than one isomer cannot be excluded within supersonic jet experiments. In order to elucidate the aspect of possibly co-existing isomers, being relevant for all investigated DBF-ROH complexes, the analysis of interconversion barriers can be helpful, aside from considering only the relative energies of the isomers. Therefore, transition state calculations were performed with the QST3 method as well as the woelfling module based on transition state guesses from the GFN-xTB method. The obtained interconversion barriers and transition state structures are shown in Fig. S1 (ESI). In the case of the water complex, the calculated barrier of less than 1 kJ/mol between the two πbound isomers OH•••π5 and OH•••π6 suggests that interconversion occurs under the supersonic expansion conditions. However, barriers of about 5 kJ/mol between the π-bound structures and the OH•••Op isomer might allow a kinetic trapping of oxygen-and π-bound isomers, respectively, in the case that they are both initially populated.
For the MeOH complex, a low barrier of about 1 kJ/mol is predicted between the two oxygen-bound isomers OH•••Ot and OH•••Op, suggesting that interconversion occurs under the experimental conditions. Similar to the water complex, the isomerization barriers between OH•••O and OH•••π binding motifs are larger than the ones between the same binding motifs, yet they are slightly lower than for DBF-H 2 O at about 3 kJ/mol. Nevertheless, kinetic trapping of the respective lower energy isomer can be expected, in case that more than one isomer is initially populated. Furthermore, the TS calculations suggest that the interconversion of the OH•••Op structure into the OH•••π6 isomer involves the C s -symmetric OH•••Ot structure as an intermediate state.
Regarding the tert-butyl alcohol complex, the barrier between the oxygen-bound isomers OH•••Ot and OH•••Op is calcu-lated to be <1 kJ/mol, suggesting interconversion. Similar to the case of DBF-MeOH, the TS calculations suggest that conversion of the OH•••Op structure into the OH•••π6 isomer occurs via the intermediate OH•••Ot arrangement. The predicted isomerization barrier from the OH•••Ot to the OH•••π6 isomer is approximately 2 kJ/mol. Hence, interconversion between the binding motifs should not be excluded as well. A discussion of these aspects with respect to the experimental findings will be continued in the Experimental Results section.
For a comparison of theory and experiment, the structurally sensitive OH stretching vibration can serve as a spectroscopic probe to be compared to calculated harmonic OH stretching wavenumbers. In some cases, particularly if two competing structures with the same binding motif are present, the OH stretching vibrations might be indistinguishable. Therefore, a comparison of the experimental rotational constants obtained from rotational spectroscopy combined with calculated dipole moment components can lead to an unambiguous structural assignment. All calculated values relevant for comparison to the experiments are found in Tables S5, S6 and S9-11 (ESI) and are discussed in the Experimental Results section. In the end, comparison to the experiments will reveal the individual performance of each theoretical approach.

DBF-H 2 O IR/UV Results
For all investigated systems, R2PI spectra were recorded, revealing isomer-specific electronic excitation energies of the respective complexes (cf. Fig. S2, ESI). Based on these findings, IR/R2PI spectra were measured in the OH-stretching region (3520-3750 cm −1 ) for different excitation energies of the respective complexes.
The R2PI spectrum of the DBF-H 2 O complex reveals a S 1 ←S 0 transition that is blue-shifted by +171 cm −1 compared to the DBF monomer transition (cf. Fig. S2, ESI). This is in agreement with the findings of fluorescence excitation spectra 24 . No additional, red-shifted transitions with respect to the monomer were detected. The experimentally observed shift of +171 cm −1 is in qualitative agreement with the predicted blue-   24 is confirmed by our mass-selective R2PI experiments combined with predicted S 1 ←S 0 excitation energies at the SCS-CC2/def2-TZVP level. These calculations have proven to yield robust predictions 57 .
In order to obtain additional structural information, the OH stretching vibration is used as spectroscopic probe for identifying the docking motif of the complex. Therefore, an IR/R2PI spectrum was recorded via the electronic resonance at +171 cm −1 , which is shown in Fig. 2  for OH•••π structures. Moreover, the changes of the splitting with respect to the splitting of free water clearly suggest the  OH•••Op isomer to be the observed structure. Note that the splittings obtained from the SCS-CC2 calculations are not considered here, as they are found to be unable to reproduce the frequency splitting of the free water molecule correctly.

CP-FTMW results
Rotationally resolved spectroscopy can provide unambiguous proof of the observed clusters a) via comparison of the experimental rotational constants with the results of quantum chemical calculations and b) via structure determination using isotopic substitution, either in natural abundance or using enriched samples. The experimental rotational constants for DBF-H 2 O obtained from broadband CP-FTMW spectroscopy are summarized in Table 2 together with the results from quantum-chemical calculations. The comparison clearly identifies the observed complex as the OH•••Op structure, in agreement with the (IR/)R2PI studies, with the rotational constants of the OH•••π5 isomer being clearly different. Other complexes were not observed under the experimental conditions using neon as a carrier gas. Note that we report an average fit, i.e., fitting the center frequencies of a doublet splitting arising from the internal motion of the water molecule with respect to the DBF moiety. A more detailed analysis of this internal motion is beyond the scope of the present study and will be reported elsewhere. The spectrum is dominated by a-and b-type transitions, while no c-type transitions were observed, which points to averaging due to the internal motion.

DBF-MeOH
In contrast to the clear-cut case of DBF-H 2 O, where the different experimental and most theoretical methods match nicely in finding a single dominant isomer, a more difficult case is found for DBF-MeOH, where the theoretical methods are rather undecided between up to three different structures.

FTIR results
FTIR spectra of the co-expanded DBF-MeOH mixture using the double-slit nozzle are shown in Fig. 3 (b). The methanol concentration of 0.15% is chosen such that almost no monomer (3686 cm −1 ) or homodimer (3575 cm −1 ) are visible. A distinct band at 3594 cm −1 is observed, but the red-shift upon complexation seems too large to be associated with a heterodimer. Indeed, when comparing to the previously measured spectrum of 2,3-benzofuran-MeOH 15 ( Fig. 3(a)) this band lies within the region of larger clusters. Taking the strong cohesion and excess of DBF into account, a trimer including one methanol and two DBF molecules is the most probable assignment. Further discussion on this trimer can be found in the ESI. Searching for spectral features in proximity to the dimer bands of 2,3benzofuran-MeOH, two peaks might be located at 3639 and 3646 cm −1 , hardly distinguishable from noise level. In an attempt to alter the expansion conditions to form more mixed dimers, spectrum (c) was recorded. The methanol concentration was increased about two-fold, while the DBF concentration was slightly decreased. However, the major change was the use of the newly developed V-nozzle, which nominally enhances the absorption path by a factor of about 6, and a . A further increase of the methanol concentration (spectrum (d) in Fig. 3) did not seem to enhance the dimer abundance any further. Given the weakness and broadness of these bands, only a vague assignment to a specific isomer could be made, but in comparing the dimer band positions of 2,3-benzofuran-MeOH, it is pausible that two dimers are observed due to an OH•••π isomer further red-shifted than an OH•••O isomer. The peak intensity of the further red-shifted isomer is at best two-fold higher, but given the lower predicted IR intensity of OH•••π isomers, the actual excess in abundance may be larger, even in the weakly relaxing helium expansion employed. This tentative assignment called for confirmation by complementary spectroscopic methods.  IR/R2PI spectra, thus no third isomer is found). The electronic excitation spectrum itself contains valuable information: a comparison to calculated S 1 ←S 0 excitation energies at the SCS-CC2/def2-TZVP level suggests that the blue-shifted transition arises from the OH•••Ot isomer with a qualitatively matching predicted shift of +126 cm The two additional features marked with an asterisk (*) originate from ionization-induced fragmentation of a mixed DBF-MeOH-H 2 O cluster (cf. Fig. S3, ESI). Given the overlap situation in the FTIR experiment and its relatively high nozzle temperature, as well as the different carrier gas, the wavenumber agreement between the two IR experiments is satisfactory. The observed isomer splitting of 5 cm −1 is probably more reliable than the 7 cm −1 deduced from the FTIR spectrum.

IR
The relative order of calculated OH stretching wavenumbers for the different OH•••O and OH•••π structures turns out to be ambiguous: DFT calculations using the B3LYP-D3(BJ) functional predict a red-shifted OH stretch for both OH•••Obound structures compared to the OH•••π6 isomer. Interestingly, calculations at the SCS-CC2 level suggest a switch of the order: the OH•••π6 structure is predicted to have a lower OH stretching frequency than the OH•••O equivalent, which suggests the π-cloud to be the stronger acceptor site. The same is observed for calculations with the functional M06-2X (cf. Table  S6, ESI). Note, however, that these two methods failed in predicting the correct complex with water.
Given the very small OH stretching frequency differences between the observed species, the prediction of the frequency order for a distinct theoretical method is ambiguous. Finally, based on the clear isomer assignment via the electronic resonances, the OH•••Ot isomer is found to exhibit the less redshifted OH stretching vibration compared to the OH•••π6 isomer. This has been observed for similar systems as well (cf. discussion in the ESI and Refs. 15,16).
Drawing conclusions on relative populations from the electronic resonances in the R2PI spectrum is difficult in this specific case, as the R2PI spectrum of the DBF-MeOH mass trace is influenced by very strong resonances of the DBF monomer (for further explanation cf. Fig. S2, ESI). Additional structural and abundance insight will be gained by rotational spectroscopy.

CP-FTMW results
In the rotational spectrum of the DBF-MeOH mixture, we observed two DBF-MeOH complexes (cf. Fig. 5). The experimentally obtained rotational parameters are summarized in Table 3. The rotational constants for the two complexes are clearly different, and the spectra also differ in the type of rotational transitions observed (i.e., only a-and b-type transitions but no c-type transitions for one complex and only b-and ctype transitions but no a-type transitions for the other complex). Such observations provide additional input for assigning the structures. Based on a comparison of the rotational constants and the observed type of transitions vs. calculated dipole-moment components (cf.   Both complexes show internal rotation splitting due to the internal rotation of the methyl group of the methanol moiety, which results in characteristic doublets for each rotational line. The fact that this internal rotation leads to sizeable splittings and is not locked points to only a loose interaction of the methyl group with DBF. For OH•••π6, two sets of rotational constants are presented. Fit 1 corresponds to a fit to an asymmetric rotor Hamiltonian including only the A states due to methyl group internal rotation, thus presenting effective rotational constants. Using the program XIAM, these line splittings can be analyzed, resulting in Fit 2. It includes the analysis of the methyl group internal rotation and thus also provides information about the torsional barrier V 3 . The V 3 barrier determined from the experimental line splitting into A and E states is V 3 (exp)=4.055(11) kJ/mol (Table 3), which is in decent agreement with the calculated barrier of about 5 kJ/mol (level of theory). The wealth of experimental information is thus well suited to identify and further characterize the respective molecules under study and can also be used to benchmark the theoretical models employed. For OH•••Ot, only an A-states fit is presented. A global fit including internal rotation resulted in large standard deviations, potentially pointing to a second internal motion. The obtained rotational parameters, however, allow a clear identification of the respective isomers. Based on our experimental line intensities, the OH•••π6 complex is found to be about 10 times stronger than the one for the OH•••Ot complex. Considering the stronger dipole moment for the OH•••Ot complex, this points to a clear energetic preference for the OH•••π6 complex, which is also the global minimum at the B3LYP-D3(BJ)/def2-TZVP level and the B3LYP-D3(BJ)/aug-cc-pVTZ level, in the latter case it is isoenergetic with the OH•••Ot isomer, and the SCS-CC2/def2-TZVP level. This finding agrees qualitatively with the FTIR evidence of a higher abundance of the more red-shifted species and the corresponding results for 2,3-benzofuran 15 . It also agrees with the IR/UV experiment, considering the mentioned intensity uncertainty within the R2PI spectra. Interestingly, the third complex, OH•••Op, was not observed in our microwave study despite intense analysis, even though it is predicted to be the global minimum by the DLPNO-CCSD(T) approach (cf. Table 1). Thus, a low interconversion barrier from the OH•••Op complex to one or both of the other complexes, OH•••Ot and OH•••π6, can be assumed.

FTIR results
The FTIR spectrum of DBF-t-BuOH (cf. Fig. 6), measured in helium carrier gas with the V-nozzle, shows similar features as DBF-MeOH. The monomer and homodimer bands of tert-butyl alcohol are observed at 3643 cm −1 and 3499 cm −1 , respectively. These values are slightly blue-shifted to those reported previously 82 , which hints at warmer expansion conditions, probably due to the heated nozzle. Fortunately, the proposedly mixed dimer signals are more pronounced than for methanol, revealing one band at 3607 cm −1 with a weak shoulder at 3613 cm −1 . Given the similarity of the experimental data, an analogous assignment to the DBF-MeOH clusters is suggestive. Therefore, the band at 3607 cm −

IR/UV results
The IR/UV analysis of the DBF-t-BuOH complex yielded the presence of one single isomer in a conformationally colder neon expansion. Regarding the recorded R2PI spectrum (cf. Fig. S2 . Table S5, ESI). Therefore, an assignment of the OH•••π6 isomer can already be made based on the shift of the electronic origin. Fig. 7 shows the IR/R2PI spectrum obtained via the electronic transition at −39 cm −1 with respect to the DBF monomer transition. It exhibits a single OH stretching vibration at 3605 cm −1 . The spectra obtained via all further transitions observed in the R2PI spectrum (cf. Fig. S2, ESI) exhibit the same vibrational transition. Therefore, the presence of a second isomer is unlikely (cf. also IR fixed /R2PI spectrum in Fig. S6   oxygen-bound structures compared to the OH•••π isomer (cf. Fig. 7b and Additional insight regarding the structural assignment is provided by rotational spectroscopy. As pointed out in the theoretical results section, the isomerization barrier between the OH•••Op and the OH•••π structure of the DBF-t-BuOH complex is calculated to be about 2 kJ/mol, whereas the barrier between OH•••Ot and OH•••Op is smaller than 1 kJ/mol. Since only one isomer is found in the experiment with neon as carrier gas, it might be concluded that the isomerization barriers are too low for both binding motifs to be stabilized during the supersonic expansion. This would lead to the exclusive population of the global minimum structure in the molecular beam. Comparing the IR/UV and FTIR investigations (cf. Fig. 6 and 7), the shoulder at 3613 cm −1 exclusively observed in the FTIR spectrum seems to originate either from a less stable DBF-t-BuOH isomer, populated due to different expansion conditions, or from a larger cluster.

CP-FTMW results
The analysis of the broadband rotational spectra for the DBFt-BuOH mixture also reveals the presence of only one strong spectrum, for which 162 rotational lines, distributed over a-, b-, and c-type transitions, could be identified and fitted to an asymmetric rotor Hamiltonian, with a-type transitions dominating the spectrum. The resulting molecular parameters are summarized in Table 5, together with the results from quantum chemical calculations. The widely different rotational constants for the three DBF-t-BuOH complexes allow their identification as the OH•••π6 isomer, which is stabilized by secondary CH•••π interactions, in agreement with the IR/UV spectroscopic results. As in the t-BuOH monomer and in other complexes involving t-BuOH, no internal rotation splitting due to internal rotation of the three methyl groups is observed because of the high barrier hindering this motion. The OH•••π6 structure is predicted to be the global minimum by several quantum chemical methods (including ZPE correction, Table  1). Note the interesting basis set dependence for the dispersion corrected B3LYP-D3(BJ) approach: using the aug-cc-pVTZ basis set, the correct global minimum (after ZPE correction) is predicted, while usage of the def2-TZVP basis set leads to the OH•••Ot isomer as the energetic minimum structure. The fact that only one species is observed with CP-FTMW and IR/UV spectroscopy employing neon as a carrier gas, while FTIR spectroscopy using helium observes weak evidence for a second isomer gives an indication that the barrier between the OH•••Ot and OH•••π6 structures, calculated to be 1.8 kJ/mol (Fig. S1, ESI), is indeed sufficiently low to be overcome in a neon expansion.

Conclusions
A detailed multi-spectroscopic and theoretical analysis on a series of isolated dibenzofuran-alcohol and -water complexes is presented. By combining FTIR, IR/UV and CP-FTMW spectroscopy, the unambiguous assignment of the preferred structures for the respective complexes could be achieved. The most valuable contribution of the FTIR approach, for which DBF complexes are currently at the technological limit, is a survey over the minimum number of relevant complexes under warmer expansion conditions. The IR/UV approach is less limited in molecular size. It provides conformationally resolved IR spectra, and the UV shift from the monomer gives valuable information on the docking position, O vs. π, of the OH group. This is crucial because the spectral shifts between these two docking positions are extremely subtle such that theoretical harmonic predictions remain ambiguous. The CP-FTMW approach provides a firm structural assignment of dominant and also secondary complexes, which goes beyond the O vs. π contact information. It discriminates between O docking geometries, which exploit secondary interactions with either peripheric C-H groups (p) or aromatic π clouds (t) in the planar acceptor molecule. The comparison to theory revealed deficiencies and strengths of different theoretical approaches. For the DBF-H 2 O complex an oxygen bound structure was identified by electronic, vibrational and rotational spectrosco-py, building on the early work of Auty et al. 24 Despite the prediction of nearly isoenergetic π-bound structures, no second isomer is found. The DLPNO-CCSD(T)/cc-pVQZ method as well as B3LYP-D3(BJ)/aug-cc-pVTZ calculations yield reasonable relative energies that explain the experimental observations. Regarding the methanol complex, two isomers were identified in the molecular beam experiments. The species were identified as the OH•••π6 isomer and the OH•••Ot isomer. For the oxygen-bound structure, an interconversion of OH•••Op to OH•••Ot is expected due to a low isomerization barrier. Based on the broadband rotational spectroscopic results, the OH•••π6 isomer is found to be more strongly populated than the OH•••Ot structure, which is confirmed by the FTIR results and is also reasonable within the uncertainty of the R2PI signal intensities. Nevertheless, within the error of the methods, the predicted relative energies at the DLPNO-CCSD(T)/cc-pVQZ and B3LYP-D3(BJ)/aug-cc-pVTZ level are in agreement with the experimental findings since both theoretical methods indicate two nearly isoenergetic structures. Furthermore, the chosen theoretical approaches largely deviate in predicting OH stretching wavenumbers, which even leads to a switch in the order between the OH stretches of the two docking motifs. This has also been observed for the related 2,5-dimethylfuranmethanol 15 and 2,3-benzofuran-methanol complexes 15,16 . The only approaches, which correctly predict a red-shifted OH stretching vibration for the OH•••π6 isomer compared to OH•••Ot are the M06-2X functional as well as calculations at the SCS-CC2 level. This probably indicates a deficiency of established theoretical approaches including the harmonic approximation, which should be considered in future developments.
The tert-butyl alcohol complex, representing the largest ether-solvent complex in this study, was shown to form only one stable isomer in the molecular beam experiments. Based on a red-shifted S 1 ←S 0 transition, a π-bound structure was identified. Rotational spectroscopy clearly confirmed the observed structure to be the OH•••π6 isomer. By comparing all investigated DBF-solvent complexes, we observe a binding preference that switches from oxygen via a balanced situation to the π site when going from small solvent molecules to larger ones. This is inverse to the trend that has been observed for the previously studied series of diphenyl ether-solvent complexes. The stabilization due to London dispersion is found to be more pronounced in π-bound structures than in oxygen motifs, indicated by dispersion contributions extracted from both SAPT(0) and DLPNO-CCSD(T) calculations. An exception is found for the t-BuOH complex, where the C s symmetric OH•••Ot and the OH•••π6 isomers are found to have similar dispersion contributions. We moreover found the influence of CH•••O contacts on OH•••Op structures to decrease from the small solvent molecules to the larger ones, while simultaneously enabling stronger CH•••π interactions in this series. However, their magnitude in the OH•••O arrangements is always outweighed by the one in the respective OH•••π structures. Thus, the additional CH•••O contact leads to preferred oxygen-binding for the smaller solvent molecules, whereas for the larger t-BuOH, London dispersion finally outbalances the CH•••O-stabilization, leading to preferred π-binding. Regarding relative energies, the overall performance of the B3LYP-D3(BJ)/aug-cc-pVTZ approach is satisfactory and it is, at a first glance surprisingly, even superior to the DLPNO-CCSD(T) approach that seems to slightly underestimate the stability of OH•••π6 complexes compared to OH•••Op. Given the fact that the relative energies are mostly below 1 kJ/mol and therefore in the order of ZPE corrections, this slight inconsistency might be attributed to the usually less relevant fact that the geometries are not optimized at the DLPNO-CCSD(T) level. Furthermore, neglected anharmonic contributions to the ZPE do not allow for a safe relative ranking of the two electronic structure methods at this subtle level.
Upon comparison to the series of diphenyl ether-solvent complexes 37 and also the phenyl vinyl ether-methanol complex 38 , the absence of backbone deformation in the DBF complexes proves to be true, as no structures were found involving a non-planar DBF structure, which is not surprising since aromaticity is preserved. The influence of CH•••O contacts -playing a decisive role in DPE and PVE complexes -partly remains in the DBF-solvent structures as well. However, it is constrained to the oxygen-bound motifs, as the CH moieties of DBF available for CH•••O contacts are within the DBF plane and therefore not in vicinity to the alcohol oxygen atoms within OH•••π arrangements. This finally leads to a trend regarding the binding preference that is contrary to the series of DPEsolvent complexes. A further interesting aspect is that the central, five-membered furan moiety in DBF is only a competitive π docking site for the water complex. In the MeOH and t-BuOH complexes, only structures involving the six-membered benzene moieties as π acceptors were observed.
One parameter governing the rigidity of the aromatic chromophore (DBF vs. DPE) completely changes the preferred binding site within a series of solvent complexes. The big challenge is to quantify the corresponding small effect induced by small energy differences between different structures. We succeeded by a comprehensive combination of different experimental and theoretical methods, which finally offers a clear structural assignment. Within the investigated series of DBF-solvent complexes, we found a variable interplay between non-covalent interactions among which London dispersion forces make the difference in determining the final docking preference.          ) that it reduces its actual intensity within a range of about ±30 cm −1 around the monomer transition (grey dotted line in Fig. S2). On the other hand, the transition of the OH•••Ot structure at +135 cm −1 is almost unaffected by the monomer as there is no close monomer transition. Moreover, the OH•••π6 transition is "contaminated" by fragmentation of larger clusters (cf. Fig. S3, ESI), which is reflected in the corresponding IR/R2PI spectrum (cf. Fig. 4b). In contrast, the OH•••Ot transition is not influenced by fragmentation indicated by its clean IR/R2PI spectrum (cf. Fig. 4a)

DBF-MeOH
It is known e.g. from investigations on 1-indanol 2 or benzyl alcohol and other systems 3 Fig. 3 and 4). Calculations at the B3LYP/6-311++G(d,p) and B3LYP-D3(BJ)/aug-cc-pVTZ level suggested the OH•••O structure to show the larger OH stretching red-shift, whereas the functional M05-2X as well as MP2 calculations, both using the 6-311++G(d,p) basis set, predicted a reverse order. By analyzing the intermolecular modes 1 and by performing experiments with MeOD instead of MeOH 4 , the less shifted band could be clearly assigned to the weakly populated OH•••O isomer, confirming the π cloud to be the stronger hydrogen bond acceptor site. Consequently, the frequency shifts were predicted wrongly using the B3LYP functional (independent on the use of dispersion corrections), whereas the M05-2X functional as well as MP2 predicted the right order of shifts. Building on these findings for the very related system of 2,3-benzofuran, the similarity of a different OH stretching frequency order predicted by B3LYP-D3(BJ) calculations than for other approaches is obvious. The M06-2X functional is known to perform worse than B3LYP-D3(BJ) for predicting the energetic order of such intermolecular energy balances as it exhibits a systematic bias towards π-bound structures. 5 Nevertheless, this does not necessarily affect the performance of vibrational frequency predictions. Therefore, based on the intriguing analogy to the 2,3-benzofuran-methanol complex, the frequency order obtained by M06-2X/def2-TZVP and the SCS-CC2 calculations is regarded as the most probable one. Hence, the transition at 3642 cm −1 is assigned to the OH•••Ot structure, whereas the transition at 3637 cm −1 is assigned to the OH•••π6 isomer.     Figure 3 (main article), the DBF-MeOD spectrum has been measured with identical parameters and aligned in wavenumber to the non-deuterated spectrum by shifting to the MeOH monomer band (+968 cm −1 ) and scaling by a factor of √2. The predicted frequencies are scaled (×0.9675) to the methanol monomer OH stretching frequency.
Discussion on the trimer band: The excess of DBF in the expansion hints at a trimer origin of the band at 3594 cm −1 composed of one methanol and two DBF molecules. This is further supported by the fact that the binding energy for a DBF homodimer is predicted to be larger (38.9 kJ/mol) than for the heterodimer (20.1 kJ/mol) (computed at B3LYP-D3(BJ, abc)/def2-TZVP level using ORCA 4.0 including zero point vibrational energy). Furthermore, the binding for methanol to a DBF dimer (28.4 kJ/mol) is also predicted to be stronger than to a single DBF molecule. Since the red-shift seems large for a cluster containing only one hydrogen bond, some trial structures have been computed at B3LYP-D3(BJ, abc)/def2-TZVP level using ORCA 4.0 (cf . Table S7). Interestingly, an oxygen-bound isomer is favored by 2.0 kJ/mol over the combination of the preferred DBF dimer 6 and DBF-MeOH (π-bound) geometries. The predicted spectral shift (see Fig. S7) is in good agreement, thus the band is tentatively assigned to this oxygenbound structure, although the predicted shift for the π-bound isomer would also be in agreement.
The assignment to an oxygen-bound cluster is supported by the slightly larger red-shift in the spectrum of DBF-MeOD (see upper panel of Figure S7), when arranging the spectra such that the monomer bands are aligned after stretching the OD region by √2. It is a signature of larger anharmonicity than in free methanol, which has been observed before for methylated furans. 7 The oxygen-bound dimers similarly showed this larger red-shift, while π-bound clusters were slightly less red-shifted.