Chirality-dependent balance between hydrogen bonding and London dispersion in isolated ( )-1-indanol clusters †

The aggregation behavior of racemic and enantiopure 1-indanol has been studied by FTIR spectroscopy, resonant ion dip IR spectroscopy, and spontaneous Raman scattering in supersonic jets. This triple experimental approach, augmented by homology to related molecular fragments and dispersioncorrected DFT predictions, allows disentangling the complex spectroscopic signature in the OH stretch range. Evidence for chirality-sensitive aggregation via isolated OH p bonds in competition with cooperative OH OH p patterns is collected. An accurate description of London dispersion forces provides the key to its explanation.


Introduction
Hydrogen bonds and London dispersion interactions are key driving forces for supramolecular assemblies.For amphiphilic molecules, they may add to or compete with each other.3][4] In dimers of such aromatic alcohols, it is generally found that the OH groups are linked together by a strong hydrogen bond, while the aromatic rings simultaneously try to optimize their dispersion interaction.The acceptor OH may or may not engage in a residual weak hydrogen bond with one of the aromatic rings, depending on the flexibility of the linker. 2This will usually not dominate the interaction.However, there are cases where even an OHÁ Á ÁO hydrogen bond is sacrificed for better dispersion interaction. 5In complexes between an aromatic alcohol and protic solvents, it has been shown that the aromatic ring can efficiently compete with the OH group as a proton acceptor. 6,7Here, we explore 1-indanol, an aromatic alcohol which shows a delicate balance between three kinds of interaction, namely OHÁ Á ÁOH, pÁ Á Áp and OHÁ Á Áp contacts.We will see that the latter can potentially compensate the former two and thus dominate the interaction topology in the dimer.The balance is so delicate that it depends on the relative chirality of the two monomer units, a phenomenon called chiral recognition or chirality recognition. 8here are several examples in the literature where the subtle interplay of interactions in dimers gives way to a more robust hydrogen bond pattern in larger oligomers, due to the cooperative nature of the hydrogen bond. 5,8Chirality recognition effects may be enhanced, because the cyclic hydrogen bond arrangement in alcohol trimers, tetramers and pentamers 9 invites a dispersiondriven reinforcement by non-polar side groups.This is true, if all monomers have the same handedness and thus are able to be on the same face of the hydrogen bond plane. 10Such calix-like aggregation patterns react sensitively to enantiomeric mutation of single building blocks.However, it is conceivable that with growing importance and bulkiness of the non-polar residue, dispersion forces may partially win over hydrogen bonding and prevent ring closure of the hydrogen-bonded chain in these oligomers.The balance between such supramolecular design elements can be addressed experimentally and theoretically.
If one wants to have an intimate link between quantum chemical modeling and experiment, it is imperative to start the investigation of supramolecular assemblies in the gas phase.This typically rules out structure determination by diffraction methods and calls for spectroscopic techniques.Microwave (MW) spectroscopy 11 is powerful in deciding between different theoretical structure proposals, but only in the presence of a sizeable overall dipole moment.This may not be the case in symmetric dimers and oligomers.However, aromatic alcohol clusters are open to a powerful triple combination of conformationally selective IR/UV double resonance spectroscopy, 12 linear IR absorption spectroscopy 13 and spontaneous Raman scattering 14 in supersonic jets.This is what the present work achieves for the case of 1-indanol self-aggregation.To the best of our knowledge, such a triple combination of these complementary techniques has never before been applied to molecular recognition phenomena in the gas phase.
1-Indanol is an attractive test system for several reasons.It is comparatively rigid, except for the OH torsion and the ring puckering and flapping 15 motion (Fig. 1).Intramolecular interactions between the OH group and the aromatic ring are subtle.Chirality recognition is expected to be pronounced, judging from the melting point of 52-54 1C for the racemic compound, 73 1C for the enantiopure compound, and 37-38 1C for the metastable conglomerate. 16,17The IR spectra of the solids also differ significantly, which has been speculatively related to differences in the dimers. 16In particular, the racemate shows a more red-shifted OH-stretching vibration.Differences are also found for the dielectric constant of dilute CCl 4 solutions, 16 which is consistently lower in the racemic case.The effect disappears in benzene, indicating the importance of OHÁ Á Áp or pÁ Á Áp interactions in indanol self-aggregation.Besides OHÁ Á Áp interactions, weaker CHÁ Á Áp interactions 18 could be relevant and even influence the chirality recognition. 19,20The presence of both sand p-bonds in the carbon fragment allows testing of subtle dispersion energy differences. 21he monomer of 1-indanol has been studied several times using spectroscopic techniques.First indications for one dominant conformation in a cold R2PI experiment 22 have been extended to 2-3 conformations in conformationally warmer expansions 12 and completed in terms of the conformational interpretation 12,23 by LIF, IR/UV and MW spectroscopy.IR spectra in CCl 4 solution had indicated early on that the (pseudo-)axial 24 OH group vibration (ax) is shifted some 15 cm À1 to the blue of the (pseudo-)equatorial group (eq), which is energetically favored. 24Further isomers were identified some 25 cm À1 to the blue of the most stable conformation.Several mixed clusters of 1-indanol with solvent molecules have been studied, 12,22 but this is the first experimental investigation on self-aggregation.
The structure of this study is as follows.FTIR spectra are used to classify the signals according to monomers, dimers and larger clusters, as well as homoconfigurational and heteroconfigurational assemblies.IR/UV double resonance confirms the assignments and links different peaks to common conformations.Raman spectroscopy further confirms the assignments and provides symmetry information for the dimers and oligomers.
Based on this, a rather complete picture of the preferred hydrogen bond topologies is already obtained.This is compared to the results of quantum-chemical DFT calculations including semiempirical dispersion corrections, and finally to known jet IR data on fragments of 1-indanol, namely benzyl alcohol, cyclopentanol and propanol.More remotely related compounds have also been studied, 15,[25][26][27] but are somewhat less useful for an explicit comparison to the 1-indanol case.

Methods
Details of the heated nozzle FTIR setup (popcorn-jet) can be found elsewhere. 5,28,29In short, a helium gas pulse of E1.5 bar flows through the heated nozzle in which the substance is deposited on molecular sieve enclosed between two poppet valves.The gas mixture is probed perpendicularly to two parallel slits (10 Â 0.5 mm 2 each) at a distance of 4 AE 3 mm by synchronized FTIR scans of a Bruker IFS 66v/S spectrometer.After optical filtering an InSb detector collects the signal.Typical spectra contain 600-1000 coadded scans (300-500 gas pulses).The background pressure is kept low by a buffer volume (3.6 m 3 ) continuously evacuated using two Roots and rotary piston pumps.For some spectra, a high temperature pre-expansion attachment is used to suppress the formation of larger clusters in a controlled way for a given compound concentration in the carrier gas.It consists of two 200 mm long channels in which the gas mixture can be heated for about one millisecond up to a maximum of 500 1C.The racemate of 1-indanol (FLUKA, Z98%) appears to be slightly less volatile than the enantiopure compound ((S)-1-indanol, FLUKA, Z99%, S : R Z 98 : 2) -therefore its FTIR spectra are typically recorded with a 10 K higher sample temperature.This is reminiscent of the methyl mandelate case, 5 but in view of the mostly liquefied indanol samples not easy to rationalize.
Mass-resolved S 0 -S 1 spectra are obtained by means of onecolor Resonance-Enhanced Multiphoton Ionization (1c-REMPI).The skimmed supersonic beam is intersected by the laser in the ion-source region of a home-built linear one meter time-of-flight mass spectrometer.The ions are detected by a micro-channel plate detector (RM Jordan, 2.5 cm).The ion signal is averaged by an oscilloscope (Tektronix) and processed through a PC.The heated sample ((S)-1-indanol, 99%, Aldrich Chemicals; rac 1-indanol, 98%, Aldrich Chemicals) is seeded in neon at a pressure of E0.8 bar and expanded into a vacuum through a 0.30 mm pulsed nozzle (General Valve).Neon is selected as a carrier gas because the two main monomer REMPI transitions 12 are observed and it ensures good cooling conditions without formation of large clusters leading to background which could be the case in argon.The temperature of the sample is kept at 70 1C for optimizing the dimer formation.The delay between the valve opening and the laser is crucial for observing the dimer and has been carefully adjusted for maximizing the signal.[36] Fig. 1 Visualization of the important conformational degrees of freedom in the most stable (pseudo-)equatorial conformer (eq) of (S)-1-indanol.
The focal conditions are adjusted to minimize fragmentation induced by the UV laser (see above) and to obtain a smaller spot for the UV laser than for the IR.This is achieved by mildly focusing the UV laser with a 4000 mm focal length lens and adding an iris 1 m before it reaches the cold region of the jet to decrease the size of the spot.The IR laser is focused by combining CaF 2 lenses of 700 mm and 500 mm focal length.The UV radiation is obtained by mixing the frequency-doubled output of a dye laser (Quantel TDL90 equipped with LDS 722 dye) with the fundamental radiation of the same YAG laser (Quantel).The tunable IR source is a Euroscan OPO based on a KTP crystal.The IR pulse reaches the supersonic beam 50 ns before the UV pulse, ensuring that depletion of only the neutral S 0 state is obtained.The resolution of the UV laser is 0.02 cm À1 while that of the IR OPO is 3 cm À1 .All the delays are computer-controlled by a home-made digital pulse generator (G.I.N.I).
In the spontaneous Raman scattering measurements 29,37 helium at E0.7 bar is guided through a heatable saturator containing racemic (see FTIR section) or enantiopure 1-indanol.In the latter case, both enantiomers yield equivalent spectra, but most of the reported data are based on (R)-1-indanol (Aldrich, 99%).Controlled amounts of water can be added to the helium by flowing it through a coolable saturator filled with water and by mixing with pure helium prior to guiding it through the indanol saturator.Concentration control is possible by changing the water temperature and by changing the mixing ratio through pulsed magnetic valves.The mixture is expanded continuously through a slit nozzle (0.15 Â 4.0 mm 2 ).At a distance of 1.5 mm from the nozzle a cw Coherent Verdi V18 frequency-doubled Nd:YVO 4 laser beam (18 W, l = 532 nm) is focused onto the expansion.Scattered light is collected and collimated by a camera lens (Nikon, + = 50 mm, f/1.2) at 901 angle to jet and laser beam.It is focused onto the entrance slit of a monochromator (McPherson Model 2051 f/8.6, f = 1000 mm) and Rayleigh scattered photons are eliminated using a Raman edge filter.A CCD camera (PI Acton, Spec-10: 400 B/LN, 1340 Â 400 pixels) cooled with liquid nitrogen detects the dispersed photons.Cosmic ray signals are removed by comparing block-averaged spectra and a Ne I emission light source serves for wavelength calibration.Comparison to other systems studied 38,39 in this setup suggests rotational temperatures of 10-30 K, vibrational temperatures of 50-150 K and barrier-dependent conformational temperatures of 100-200 K or more. 38,39xploratory quantum chemical studies are performed at the B97D/TZVP/TZVPFit (in brief B97D) level with the Gaussian 09 program package. 40Local minima are reoptimized at the B3LYP/def2-TZVP level including empirical dispersion correction D3 (ref.41) (B3LYP-D3) with Turbomole V6.3.1. 42For wavenumber calculations a SMP parallelized version of aoforce 6.4 is used.It includes improved gradients for the dispersion correction D3, which reduces the occurrence of small imaginary vibrational wavenumbers.BSSE corrections are not applied.Differential Raman scattering cross sections are calculated at a 901 angle for a wavelength of 532 nm assuming a vibrational temperature of 100 K.The latter is not a critical assumption for Raman intensities of high frequency fundamentals, though. 43

Results and discussion
Experimental band positions and their assignments deduced in the following are summarized in Table 1.Our nomenclature classifies spectral bands (small first letters) and aggregates (capital first letters) to be either homochiral (hom/Hom) or heterochiral (het/Het) with respect to the configuration at Table 1 Band positions (ñ), aggregate sizes (M = monomer, D = dimer, T = trimer or tetramer), and relative chirality information (hom/het) deduced from the various experimental approaches; experimental assignments (the same letter (a-d) links bands of a given species, O/P denotes the binding partner of an OH hydrogen atom and is underscored for the dominant vibrational amplitude or subscripted with (a)s in the case of (anti-)symmetric coupling) are also given.The weak band at 3614 cm À1 could be due to noise or an impurity or a larger cluster carbon C 1 (Fig. 1).This is followed by information about the surrounding of the hydroxylic hydrogen atoms: ''O'' indicates hydrogen-bonding to an oxygen atom, ''P''describes interaction with an aromatic p-system, and ''F'' denotes hardly any interaction (''free'').Underscored letters indicate dominant normal mode amplitude of the coordinated hydrogen atom and subscripts a/s indicate anti-/symmetric coupling of the two OH stretches if present.

FTIR
We start with a discussion of FTIR jet spectra of racemic 1-indanol (Fig. 2a-c).In trace (a) one can see the dominant monomer peak eq at 3627 cm À1 and a second monomer band ax at 3650 cm À1 , both well known from the literature. 12A weak, rather structureless band in the range of E3450-3600 cm À1 may be present, but the high temperature of the preexpansion attachment efficiently prevents most of the aggregates.Reducing the temperature of the preexpansion attachment leads to the appearance of additional peaks (trace b) due to clusters.When no preexpansion attachment is used (c) these gain in intensity and even more peaks can be identified.Considering this behavior, peaks in the range of E3450-3600 cm À1 mostly stem from hydrogen bonded dimers (D) whereas bands below 3450 cm À1 are more probably due to trimers (T) or larger clusters, also matching typical red shifts in cooperatively hydrogen bonded aggregates. 9The large number of peaks especially in the dimer range indicates a high conformational and topological diversity.
To further analyze this diversity, enantiopure (S)-( 1)-indanol has also been studied (trace d).A smaller sample temperature was necessary to preserve a similar dimer concentration.The largest difference to trace c is probably that the bands at 3573 cm À1 and 3501 cm À1 do not appear which means that the species responsible for these bands are formed by two monomers of different chirality (heterochiral dimers).Accordingly, the peaks marked D and T in trace d of Fig. 2 all belong to homochiral species.Furthermore, the bands marked D have roughly doubled in intensity compared to trace c (less in the case of the band at 3584 cm À1 ), whereas those marked T have more than doubled in height.This meets the simplified statistical expectation that homochiral dimers compete with a similar number of heterochiral dimers, whereas homochiral trimers compete with a larger number of heterochiral trimers in a racemic mixture. 44Assuming largely irreversible dimerization in the jet, i.e. equal probabilities for a given monomer to form a heterochiral and a homochiral dimer, 44 bands belonging to homochiral dimers should double their intensity in the enantiopure spectrum.The larger enantiopure/racemic factor as a function of concentration for T bands may hint at homochiral trimers (or even tetramers). 44However, this classification is not rigorous, given the signal-to-noise ratio and the occasional observation of molecule-switching collisions, 45 i.e. sticking probabilities less than one.Note also the broad band in trace c underlying the dimer range that may affect relative intensities.Note finally the negative baseline in traces a and b where hot monomers would absorb.This is due to such absorptions in the reference spectra recorded prior to the pulse, caused by slow evaporation of indanol deposits near the heated channel.
Irrespective of these shortcomings, the direct absorption spectra provide an overview of the aggregation shifts without any risk of overlooking an important hydrogen-bonded cluster size or isomer, because alcohol clusters are regularly-behaved in terms of IR intensity and spectral width. 9While most of the entries in the column FTIR in Table 1 are thus fairly robust, it remains completely unclear which of the bands belong to the same cluster conformation or to different ones.This is where an input from a conformer-specific spectroscopy such as a doubleresonance technique is needed.

REMPI and IR/UV
3.2.1 Mass spectra.The indanol cluster ions formed by two-colour photoionization undergo extensive fragmentation because vertical ionization results in vibrationally-excited species.Fragmentation of the dimer in the ionic states results in the loss of water (I 2 -W fragment) or evaporation of a monomer with concomitant loss of hydrogen (I-1 fragment).The formation of I-1 is energetically favored and is predominant at low laser fluence, while the loss of water is observed when large laser fluence is used.Fragmentation efficiency amounts to almost 100% so that the dimer is never seen at its own mass but only at that of the I 2 -W or I-1 fragment.The electronic and vibrational spectra of the dimer have therefore been recorded monitoring the signal at either of the two fragment masses.They are identical whatever the fragment monitored.
Due to monomer evaporation in the ionic state, the trimer is seen not only at its own mass but also at that of the dimer.It is to be noted that none of the fragmentation processes observed for the dimer appears in the trimer.The electronic or vibrational spectra of the trimer have been recorded by monitoring the signal at either the mass of the trimer or that of the dimer.The spectra obtained in both ways are identical.
3.2.2Electronic spectroscopy.Fig. 3 presents the S 0 -S 1 spectrum of the homochiral (a) and heterochiral (b) dimers, and the homochiral trimer (c) of 1-indanol.Dimer.The S 0 -S 1 spectra of the dimers are very dense, which shows modification of the structure upon electronic excitation with an extended Franck-Condon factor distribution.They are red-shifted, as expected for an aromatic dimer in which dispersion is an important component of the binding energy.The homochiral dimer (trace a) shows intense peaks located at À249, À238, À231, À211, À181, À157, À148, À140, À136, À119, and À87 cm À1 .Smaller peaks are also seen in the spectrum; they will not be discussed in what follows.The spectrum of the racemic mixture (trace b) shows additional intense bands at À250, À225, À188, À172, À160, À144, À132, À97 and À90 cm À1 which can be safely assigned to the heterochiral dimer.In addition, a broad absorption background is observed.
Trimer.In contrast to what is observed for the dimer, the electronic transition is hardly shifted from that of the monomer.The transition origin is located at À24 cm À1 and the whole spectrum appears in between the electronic transitions of the two main monomer conformers.Moreover, a clear pattern is observed, with three equally-spaced bands separated by 9-10 cm À1 , in sharp contrast to the complicated pattern observed for the dimers.
3.2.3Vibrational spectroscopy.Vibrational spectra obtained by setting the probe on the most intense transitions observed in the S 0 -S 1 spectrum of the enantiopure and racemic samples have been recorded (Fig. 4).
Homochiral dimers.The spectrum obtained with the probe set at À181 cm À1 (c) shows two intense peaks in the region of the OH stretching vibration, located at 3466 and 3578 cm À1 .These bands match the bands observed at 3466 and 3584 cm À1 in the FTIR spectra (Fig. 4).Probing the bands at À140, À136 or À87 cm À1 leads to the observation of the same bands.A different vibrational spectrum is obtained when setting the probe laser on the transition located at À211 cm À1 (b).The band observed at 3545 cm À1 is also found in the FTIR spectrum (3549 cm À1 ), while the second band at 3529 cm À1 is very weak in the FTIR spectra.The same vibrational frequencies are observed when setting the probe at À238 and À249 cm À1 .The close spacing of the two bands indicates that the associated OH oscillators have a similar environment and are weakly coupled to each other.It seems therefore that two different conformers are abundant in the expansion, with different structures and different red shifts of their electronic transition.
Heterochiral dimer.The vibrational spectrum obtained by setting the probe on the intense band of the heterochiral dimer observed at À225 cm À1 (a) in the S 0 -S 1 spectrum is also shown in Fig. 4. It displays two intense bands located at 3573 and 3503 cm À1 .Both double resonance bands have counterparts in the racemic FTIR spectrum.The same spectrum is obtained when setting the probe on the bands located at À250, À188, À132, À97 and À90 cm À1 of the S 0 -S 1 spectrum.This indicates that all these features are due to the same ground-state conformer of the dimer.Because of their very low intensity and superposition with absorption of the homochiral dimer, the other bands could not be probed conclusively.However, no additional absorption band is observed when setting the probe on either of these bands or on the absorption background.
Homochiral trimer.The vibrational spectrum of the trimer (d) has been obtained by setting the probe on its most intense  band at À14 cm À1 , and monitoring the signal at the mass of the dimer.The spectrum obtained by setting the probe on the other members of the vibrational progression is identical.It displays two broad bands located at 3400 and 3526 cm À1 matching more narrow bands observed in the FTIR spectrum.In addition, a red-shifted absorption appears at 3303 cm À1 with a similar width to that in the FTIR spectra.The apparent weakness and possibly also reduced width of this band are due to the very strong decrease in intensity of the laser in this frequency range.
IR/UV summary.The most valuable new information from the IR/UV double resonance experiment is a rigorous connection of different bands to a given cluster conformation.The corresponding linking letters from Fig. 4 are given in Table 1 together with plausible OP labels sorted by the expected hydrogen bond strength in a coupled (O)OP chain.The more red-shifted bands are typical for OHÁ Á ÁO hydrogen bonds, whereas the less redshifted peaks indicate OHÁ Á Áp interactions.The explanation for the PP labels given to the more degenerate oscillators with intermediate shifts observed in isomer (c) profits from Raman spectra.

Raman
Fig. 5 adds such Raman spectra to the FTIR spectra of racemic and enantiopure 1-indanol on a vertical scale which makes the cluster signals comparable in size.Because the hydrogen bond-induced intensity enhancement is much less pronounced for Raman scattering than for IR, the two monomer bands eq and ax are now very prominent.Their intensity ratio (Fig. 6) is consistent with a free energy difference of about 2 kJ mol À1 , if the interconversion is frozen at an early stage of the expansion and the scattering strengths are similar.The invariance of the ax/eq intensity ratio to the change in concentration supports the frozen character of the ax/eq interconversion in the jet expansion, although this may change with other carrier gases than helium (cf.ref. 12 and 22).At 3638 and 3665 cm À1 there are weaker bands formally denoted eq 0 and ax 0 which could have a monomer origin.A small water impurity gives rise to a band marked ''*'' (Q-branch of the monomer symmetric stretch).To ensure that none of the bands is due to indanolwater aggregates we also recorded spectra with E15Â higher water concentration which did not change the observed bands except for * and did not lead to additional (hydrate) bands at all (see ESI †).
With two exceptions, all sizeable, moderately red-shifted cluster bands have similar relative intensities in the IR and Raman spectra.This suggests a low symmetry of the underlying species.The exceptions are hom PP a and hom PP s , the first one being strongly IR active and the second one strongly Raman active.Both bands belong to the same species based on the IR/UV results which indicates exciton coupling between locally degenerate or nearly degenerate OH stretching modes.For an antiparallel arrangement, in-phase motion generates Raman activity, whereas a 1801 phase shift induces a strong modulation of the dipole moment.
Some very weak bands also differ between the two spectroscopic techniques and impurity interpretations cannot be ruled out.The band at 3614 cm À1 is only IR-active, the bands at 3560 cm À1 and perhaps 3370 cm À1 are more Raman-active.The integrated scattering strength ratio between the axial (ax) and equatorial (eq) bands is given and seen to be stable with respect to expansion conditions.* marks a water band; see ESI. † Their hom/het character is somewhat undecided, whereas the band at 3340 cm À1 is clearly homochiral.
To further confirm the size of the homochiral species responsible for the bands discussed, we studied their concentration dependencies (Fig. 6).Only the two strong monomer bands eq and ax are visible at lowest concentration (trace a).At larger concentration (b) the main dimer bands homO À P, homPP s , and hom À OP can be recognized and in the monomer range blue shifted satellites eq 0 and ax 0 appear.homPP a and homOO À P appear at even larger concentration (c) while the other dimer bands grow more and start to exceed eq 0 and ax 0 .This continues at the highest concentration (d) indicating that eq 0 and ax 0 are dominantly monomer species.An overproportional increase of homOO À P similar to hom À OOP and homO À OP supports their assignment to a trimer or larger species.The weak bands marked D/T or T probably contain trimer or higher cluster contributions and will not be discussed further.

Spectroscopic summary
At this point, a summary of the insights into 1-indanol aggregation from the triple spectroscopic strategy and previous experimental evidence is appropriate.Two dominant monomer conformations are found with an energy difference of the order of 2 kJ mol À1 , which differ in their alicyclic ring conformation.Further conformations or hot bands are an order of magnitude less prominent.There is one non-symmetric heterochiral dimer, which features a classical OHÁ Á ÁOH hydrogen bond together with a fairly strong OHÁ Á Áp bond.An analogous structure is also found for a homochiral pair of 1-indanol molecules, together with a distinctly different dimer with two closely spaced OH stretching fundamentals, as shown by conformation selective double resonance.Their similar intensity in the IR/UV hole burning study is probably a consequence of saturation.The complementary intensities of the linear IR and Raman spectra reveal an exciton coupling between two symmetryequivalent or nearly symmetry-equivalent OH oscillators and thus remarkably large OHÁ Á Áp wavenumber shifts.Finally, there is a dominant non-symmetric homochiral trimer structure with OHÁ Á ÁOHÁ Á ÁOHÁ Á Áp topology along with a variety of heterochiral trimer combinations.An alternative tetramer assignment cannot be ruled out completely, but the usual completeness of linear FTIR and Raman spectra of alcohol clusters in terms of cluster sizes suggests a trimer origin.
Beyond these experimental insights, one can hope to obtain more detailed structural information from quantum-chemical calculations.The spectroscopic data promise a fairly rigorous judgement of the power and limitations of state-of-the-art theoretical approaches.

Quantum-chemical calculations
The system size and aromatic nature of 1-indanol clusters leave dispersion-corrected density functionals as the most powerful tools for a combined prediction of reliable structures and spectra.For vibrational spectroscopy, the hybrid B3LYP functional offers a good compromise between computational effort and reliable harmonic shifts due to hydrogen bonding.Combined with the most recent dispersion correction (D3 41 ), it is our favourite approach.For comparison, we also present results for the faster non-hybrid B97D method, which overestimates OH hydrogen bond shifts 5,29,38 and fails in some subtle conformational issues of OH groups. 46ig. 7 shows structures of particularly stable homo-and heterochiral dimers ordered by increasing electronic energy (decreasing dissociation energy D e ) at the B3LYP-D3 level (Table 2).Most listed dimer species are energetically close.Considering the subtle monomer conformational degrees of freedom (OH torsion, pseudo-axial and -equatorial orientation) this was to be expected.Comparing to the B97D results listed in Table 2, one is tempted to conclude that this is all one can say from an energetical viewpoint.This is even more true if one considers that supersonic jet expansions are not guaranteed to prefer the global cluster minimum structures for such ''sticky'' and structurally diverse monomers.We shall nevertheless briefly discuss some structural aspects of the located dimers.
Irrespective of their relative chirality, two to three dimer topologies are found.The most favorable among the low energy structures is a cooperative OHÁ Á ÁOHÁ Á Áp arrangement, where one monomer inserts its OH group into the intramolecular OHÁ Á Áp contact of the other.This is the most stable structure, although it forces one of the monomers into a less favorable ax conformation.HetOP is more stable than HomOP.With two axial conformations, the energy difference is smaller.All this happens in an energetic window of 4 kJ mol À1 (B3LYP-D3) or even 2.5 kJ mol À1 (B97D).The penalty of not forming an OHÁ Á Áp hydrogen bond (HomOF) is noticeable, despite the realization of stable eq conformations, an unconstrained OHÁ Á ÁOH bond, and CHÁ Á Áp contacts (Table 2).
An entirely different hydrogen bond topology emerges if the OHÁ Á ÁOH hydrogen bond is sacrificed for two equivalent OHÁ Á Áp interactions (PP).This allows for C i symmetry in heterochiral pairs and C 2 symmetry in the homodimers, if the monomer conformation is the same.Here, the situation depends significantly on the employed density functional and on the relative chirality of the monomers.At the B97D level, this topology is among the most stable dimer conformations, well within 1 kJ mol À1 of the best structures for the heterochiral pairing.At the B3LYP-D3 level, there is an energy gap of E2 kJ mol À1 for HomPP and E4 kJ mol À1 for HetPP.This is in qualitative agreement with the experimental observation of homPP bands, but no hetPP bands.One of the reasons for the somewhat unexpected competitiveness of two OHÁ Á Áp bonds with a cooperative OHÁ Á ÁOHÁ Á Áp arrangement may be sought in the conservation of stable eq monomer conformations in all PP structures shown.Besides the kinetic advantage which this offers in an eq : ax 5 : 2 jet abundance ratio (6 : 1 advantage for eq-eq encounters over ax-ax encounters which lead to stable HomOP dimers), it also allows for a closer approach of the dispersion-controlling heavy atoms.There are some subtle differences in the OHÁ Á Áp bonds.In the most stable HomPP structure the OH group points more or less at the ring center, whereas in HomPP 0 and HetPP, it points more to carbon atom no.7 or 4 (Fig. 1 and 7), respectively.However, the energetical bottom line is the competitiveness of HomPP structures, but not so much HetPP structures, at the B3LYP-D3 level, in full agreement with experimental evidence.
The inability of B97D to reproduce discrimination in the PP structures can be discussed at the level of dispersion energy gains DE D2 and DE D3 upon dimer formation, which are listed in Table 2. B97D predicts these stabilizing dispersion energies to be significantly larger for the PP structures, by about 1-5 kJ mol À1 , than for the OP structures.This advantage is completely lost or even inverted at the B3LYP-D3 level.The D2 parametrization tends to overestimate pÁ Á Ás dispersion, as it treats all carbon dispersion coefficients equally, independent of their coordination number.This is remedied in the D3 parametrization. 41Therefore, and because of a zero point energy advantage of 1-2 kJ mol À1 , the PP motif can compete with HomOP, but not with the more stable HetOP structure.
For all dimers and trimers, harmonic frequency calculations have also been performed (Table 3).The results for homochiral clusters are visualized in Fig. 8 as shifts relative to the most stable eq monomer wavenumber.It is worth mentioning that for B3LYP-D3, this harmonic monomer wavenumber is off by 143 cm À1 , close to twice the expected anharmonicity of the OH oscillator. 47B97D predicts a significantly softer OH oscillator and therefore only requires a shift by 80 cm À1 , far from twice the anharmonicity constants.Therefore, Fig. 8 concentrates on the B3LYP-D3 results.To encode both intensity and shift information, we have chosen to schematize the experimental Raman (bottom) and IR (top, inverted) spectra in the form of stick spectra, where the stick length is proportional to the integrated band intensity and the stick position is shifted relative to the experimental peak to match the proposed Table 2 Electronic (D e ) and vibrational zero point corrected (D 0 ) dissociation energies of the most stable dimers, trimers and tetramers in kJ mol À1 into eq monomers at various DFT levels, sorted according to decreasing D e (B3LYP-D3) energy.The largest values in their class are written in bold face.The monomer ring puckering conformation is given in the donor-acceptor sequence.Also shown are included dispersion energy contributions in version D3 (D E D3 ) and D2 (DE D2 ) in kJ mol À1 upon aggregation relative to non-interacting eq monomers.The dispersion energy is seen to be comparable in size to the total binding energy, in particular for the older D2 correction.Only a small fraction of the E D3 /E D2 difference is due to the more compact B97D structures theoretical assignment.The latter is assumed to be the most stable structure (D 0 at the B3LYP-D3 level) for each topology.The required shifts are visualized as diagonal connecting lines, clearly emphasizing the deficiencies of the harmonic predictions.

Conformations
Relative Raman and IR intensities for individual dimers and trimers are reproduced reasonably well, if we assume roughly half as many homochiral trimers as individual dimers in the expansions.The heterochiral dimers with OHÁ Á ÁOHÁ Á Áp-topology (not shown in Fig. 8 but listed in Table 3) have similar shifts from the monomer OH stretch.As in HomOP, the less shifted band always belongs to the vibration localized on the OH-bond interacting with the p-electron density, while the other OH-stretching vibration is more shifted, indicating a stronger hydrogen bond interaction as expected in OHÁ Á ÁO.Because there is a unique HetOP dimer in the experimental spectrum (Fig. 4), it serves as a calibration for the homochiral case.The average harmonic shift of the two modes from the eq monomer mode amounts to À(100 AE 5) cm À1 , which compares well to the experimental average shift of À90 cm À1 (Table 1).In general, each of the three predicted heterochiral dimers could thus be the species underlying the heterochiral bands het O À P and het À OP in Fig. 4 but the most stable one (HetOP) is more probable.Anyway, one can be sure that the ''true'' structure has the same hydrogen bonding topology.While the experimental splitting between the two bands amounts to 73 cm À1 (Table 1), the predictions range from 89 to 121 cm À1 .
The experimentally observed splitting between the wavenumbers of the p-bonded hydroxyl hydrogen atom and that involved in a classical OHÁ Á ÁO hydrogen bond is larger in the homochiral OHÁ Á ÁOHÁ Á Áp dimer (118 cm À1 ) than in its heterochiral counterpart (73 cm À1 ).This indicates that the OHÁ Á ÁO interaction is more favored in the homochiral species at the expense of the OHÁ Á Áp interaction.The calculated splittings do not reproduce perfectly the observed splittings.HomOP underestimates it (53 cm À1 ), HomOP 0 overestimates it (163 cm À1 ).Clearly, harmonic B3LYP-D3 predictions are too uncertain for an unambiguous assignment.Moreover, the energy difference between HomOP and HomOP 0 is not significant.From a spectroscopic point of view, it is tempting to assign the observed OHÁ Á ÁOHÁ Á Áp homochiral dimer to HomO À P 0 as the ordering of the experimentally observed bands, namely homO À P > hetO À P > het À OP > hom À OP, then matches the theoretical prediction for HomOP 0 and HetOP.
HomPP has its OH stretching wavenumbers in between the bands of HomOP 0 and closely spaced.Though the predicted Davydov splitting is smaller (Table 3), this is a reasonable candidate for the homochiral bands homPP a and homPP s based on its energy.HomPP 0 provides an even better spectral fit, but is likely to be metastable with respect to isomerization to HomPP.
The complementarity of the IR and Raman spectra of HomPP deserves further consideration.Although there are homochiral objects which can combine into inversion symmetric dimers (la coupe du roi 48 ), this is not possible for 1-indanol.Nevertheless, the pseudo-inversion symmetry of the two vibrational chromophores in this C 2 -symmetric homochiral dimer leads to a near-complementarity of the IR and Raman OH spectra.In line with the underestimated splitting, the intensity complementarity is more pronounced in theory.While we searched for energetically competitive ax-eq variations of HomPP, which might have a larger splitting and less intensity complementarity, we were unable to find them.
In summary, the dimer potential energy hypersurface is too corrugated to allow for an unambiguous structural assignment of the observed spectral peaks.While the hydrogen bond topology is rather clear (one OHÁ Á ÁOHÁ Á Áp dimer for both diastereomers and one sandwich-like double OHÁ Á Áp structure for the homochiral pairing), there are typically two or more slightly distorted minima which are consistent in terms of energy and the infrared spectrum, on the accuracy scale of the calculation.
Based on these assignments the elusiveness of the HetPP complex is still somewhat surprising.Its energy is very similar to that of HomPP, both are probably kinetically favored compared to other dimers, but only the homochiral species is formed, and not only as a ''byproduct'' but in similar amounts as the most stable homochiral dimer.This could be blamed on residual deficiencies of the harmonic B3LYP-D3 approach or an incomplete search for dimer structures.However, it may as well be the consequence of a lower conversion barrier to the OP structure, which is further downhill in the heterochiral pairing.
It should be noted that the overall distribution of calculated wavenumbers for dimers differs from the experimental dimer band distribution (Fig. 8).Theory predicts a ''gap'' without any band red-shifted by 40Á Á Á150 cm À1 , but that is not seen experimentally.Shifts of stretching wavenumbers in p-bonded hydroxy groups are systematically underestimated (bands homO À P, hetO À P, homPP a , homPP s ), while those of hydroxy groups involved in an OHÁ Á ÁO-interaction (bands het À OP and hom À OP) seem overestimated.Part of this may be due to different anharmonicity effects. 29However, because of the size of the systems, it was not possible for us to perform anharmonic frequency calculations at a reasonable quantum-chemical level.
With the limitations of harmonic wavenumber predictions at the dispersion-corrected density functional level for 1-indanol dimers in mind, one can briefly address the homochiral trimer bands observed in the IR and Raman spectra.The more or less equal spacing of three bands with similar IR and Raman relative intensities rules out C 3 -symmetry such as in HomOOO (or a C 2 /C 4 tetramer assignment).Two structurally different but energetically close chain structures with an OHÁ Á ÁOHÁ Á ÁOHÁ Á Áp motif (HomOOP and HomOOP 0 ) have similar predicted spectra.HomOOP is predicted to have a low Raman intensity for the central band, in qualitative agreement with experiment.The errors in the red-shift predictions are well within those observed for dimers.
In summary, the performance of the quantum-chemical spectral predictions is of reasonable quality, leaving room for improvements in the description of oscillator coupling.The relative energy sequences, on the other hand, seem to be consistent with experiment within an error margin of 2Á Á Á3 kJ mol À1 , in particular for the D3 dispersion correction applied to the B3LYP functional.

Homology assignment
As an alternative to theoretical calculations and for additional insight, it is rewarding to compare the spectra of 1-indanol and its dimer to those of fragment molecules and their dimers.
An obvious fragment is 1-propanol, 49 which is obtained when removing the benzene ring.Except for ring strain effects, its global minimum (Gt) corresponds to the axial conformation of 1-indanol and its OH group absorbs at 3682 cm À1 .Some of the red-shift found in 1-indanol (3650 cm À1 , Table 1) might be attributed to a weak influence of the aromatic ring, but much of it is expected to be due to the secondary carbon instead of primary carbon, to which the OH group is attached in 1-indanol.The aromatic influence would be expected to be larger in the equatorial conformation, because the 1-indanol OH group now points towards the ring.Notably, the observed OH stretching band (3627 cm À1 ) is only 23 cm À1 lower in wavenumber, similar to the lowering of 22 cm À1 in the underlying Tg 1-propanol conformation (3660 cm À1 ).This indicates that the OH red-shift from axial to equatorial 1-indanol should not be blamed predominantly to through-space aromatic influence, but rather to stereoelectronic effects within the 5-ring unit.This is confirmed by comparison to 2-propanol, 50 which describes the local environment of the OH group even better.The gauche conformation (3658 cm À1 ) corresponding to axial 1-indanol is again more blue-shifted (by 22 cm À1 ).In this case, the absolute wavenumber discrepancy to 1-indanol is reduced to 8-9 cm À1 , reflecting the influence of the secondary carbon that can be quantified to À24 cm À1 for both ax-like and eq-like conformations of 1-propanol (3682 and 3660 cm À1 ) and 2-propanol (3658 and 3636 cm À1 ).An even closer agreement can be expected for cyclopentanol. 51,52Here, the gauche conformation corresponding to axial 1-indanol is the only one surviving supersonic jet expansion and it absorbs at 3656 cm À1 , very close to axial 1-indanol (3650 cm À1 ).In weakly polar solution phase IR spectroscopy, 53 there is an exact wavenumber match between the gauche forms of cyclopentanol and 2-cyclopentenol, the latter being even closer to 1-indanol.Therefore, the spectral effect of the aromatic ring is seen to be very small.We note that the dominant benzyl alcohol conformation 2 absorbs at 3649 cm À1 and corresponds somewhat more to eq 1-indanol. 54Including the substitution effect for secondary alcohols (À24 cm À1 ), this is consistent with the band position assigned to eq 1-indanol (3627 cm À1 ).1-Phenylethanol also adopts an eq-like conformation, but its terminal methyl group points outside the ring plane, making a direct comparison to 1-indanol less straightforward. 4,55hile the OH stretching wavenumber is thus only weakly influenced by the aromatic ring, this seems to be different for the ax/eq energy sequence, which favors eq for 1-indanol by about 2 kJ mol À1 .This is qualitatively reversed for the corresponding 1-propanol conformations (Gt corresponding to axial is preferred 49 ), for 2-propanol (g corresponding to axial is preferred 50 ), and for cyclopentanol (axial is preferred, with the same OH group orientation as in one of the stable axial 1-indanol conformations 52 ).Thus, the equatorial preference in 1-indanol could either be caused by through-space aromatic interaction or by the sp 2 nature of two of the ring carbons.However, subtle ring strain effects can also not be ruled out.A jet investigation of 2-cyclopentenol would be helpful to clarify this.Low polarity solution spectroscopy 53 indicates that there is no significant preference for either puckering state, so indeed the aromatic ring could be responsible for the equatorial dominance in 1-indanol at low temperature.
For the dimers, we discuss red-shifts relative to the most stable monomer.In 1-propanol, the most stable dimer shows a red shift of 156 cm À1 , but shifts as small as 123 cm À1 are also observed. 49In 2-propanol, a red shift of about 141 cm À1 is found. 50Cyclopentanol 51,52 matches this with a shift of 140 cm À1 .Benzyl alcohol dimers show a shift of only 135 cm À1 , 2 which may possibly indicate some OHÁ Á ÁO hydrogen bond strain induced by competing aromatic interactions.The associated OHÁ Á Áp contact gives rise to a red shift of only 66 cm À1 . 2 Similar, but slightly larger values are found for the less strained naphthylethanol : methanol 6,7 and benzene : methanol 2 (ref.56) cases.In this framework, one would expect OHÁ Á ÁOH hydrogen bonds in 1-indanol dimers in the region of 3470Á Á Á3520 cm À1 (3627-156Á Á Á3650-135) and OHÁ Á Áp hydrogen bonds in a 3560Á Á Á3590 cm À1 (3627-66Á Á Á3650-66) range.These match the ranges of the observed OP 1-indanol dimer transitions (3466 and 3501 cm À1 for À OP, 3573 and 3584 cm À1 for O À P bands) quite well, whereas the new type of PP dimer transitions (3531 and 3549 cm À1 ) and the higher frequency trimer (3527 cm À1 ) transition fall in the gap.They seem to be lower in frequency than typical isolated OHÁ Á Áp vibrations.We can only speculate whether this is due to a more favorable geometry or due to dispersion-induced compression.
In summary, the complex OH stretching spectrum of 1-indanol monomers and dimers can be assigned quite well without resorting to quantum chemical calculations using a quadruple approach of FTIR, UV/IR, Raman and homology spectroscopy.Of course, the availability of quantum chemical calculations adds extra confidence to the assignments, in particular by providing relative energies.

Conclusions
We have presented a combined experimental FTIR, IR/UV, and Raman study of supersonic jet expansions of 1-indanol.Its dimers have a wide variety of conformational, configurational and hydrogen bond topological degrees of freedom available.Only a small selection is realized despite similar predicted binding energies and presumably wide barriers for interconversion.Remarkably, the combined experimental evidence is able to narrow down the spectral assignment to one heteroconfigurational OHÁ Á ÁOHÁ Á Áp structure and two homoconfigurational dimers.One of them resembles its heterochiral counterpart in optimizing pÁ Á Áp interactions, whereas the other forms a unique double OHÁ Á Áp structure without any hydrogen bond cooperativity or competition.There seems to be little structural diversity in homoconfigurational trimers, whereas heteroconfigurational mixtures give rise to broader spectral features indicative of structural heterogeneity.
Although the C 2 -symmetric double OHÁ Á Áp structure of the homoconfigurational indanol dimer is probably not the global minimum structure, its postulated presence in the expansion is still remarkable.Like the dimers of glycolaldehyde, 29 it represents a case where two isolated hydrogen bonds compete against one cooperative pair.The situation is actually quite analogous, although isolated OHÁ Á ÁOQC hydrogen bonds are much stronger than isolated OHÁ Á Áp hydrogen bonds.The reason is strong competition from intramolecular hydrogen bonds in glycolaldehyde, reducing the net strength of the intermolecular ones.In both systems, competition from non-hydrogen bond interactions (CQOÁ Á ÁCQO electrostatics in glycolaldehyde and pÁ Á Áp interactions in 1-indanol) appears to play a decisive role, together with possible kinetic arguments in terms of reaction barriers.More experimental examples will be required before systematic explanations can emerge.
The chirality of 1-indanol leads to an experimentally robust absence of heterochiral double OHÁ Á Áp structures, despite very different expansion conditions in the FTIR, IR/UV and Raman experiments and the clear presence of a homochiral counterpart.This chirality discrimination is not predicted at the B97D level, but consistent with the B3LYP-D3 predictions.It may serve as a sensitive test of different electronic structure and zero point energy treatments.

Fig. 4
Fig. 4 IR depletion spectra of indanol dimers and trimers in supersonic jets; numbers indicate the REMPI transition probed (cf.Fig. 3).FTIR spectra (traces c and d from Fig. 2) are shown for comparison in the lowest and second-highest trace (in blue).

Fig. 7
Fig. 7 Most stable homo-(a-f) and heterochiral (g-j) dimer structures and some homochiral trimers (k-m) and tetramers (n-p) optimized at the B3LYP-D3 level.Metastable structures with a given hydrogen bond topology are denoted with primes.

Table 3
Harmonic OH stretching wavenumbers ñ in cm À1 , IR intensities I in km mol À1 , and differential Raman scattering cross sections s 0 in 10 À36 m 2 sr À1 for the most stable monomers, dimers, trimers, and one tetramer calculated at B3LYP-D3 and B97D levels