The anomeric effect in 1,3-benzodioxole: additional evidence from the rotational, vibration–rotation and rovibronic spectra

The millimetre wave pure rotation and vibration–rotation spectrum of 1,3-benzodioxole has been studied and the vibrational spacing between the two lowest vibrational states in the five membered ring puckering potential has been determined with considerable accuracy, at ΔE01 = 259 726.035(10) MHz or 8.663 5280(4) cm−1.

In addition, rotationally resolved analysis of two hot bands near the reported origin band in the laser-induced fluorescence spectrum has been carried out and was found to rule out the current literature assignment of this spectrum.

The new information has been used to reassign the lowest vibrational transitions in the far-infrared and the Raman spectrum of 1,3-benzodioxole in terms of a one dimensional potential with a central barrier of 108 cm−1, which successfully accounts for all states up to υ = 5, and for the observed variation in rotational constants.

Considerable reassignment of the published spectra, as well as new experimental work is still necessary for a confident determination of the low-energy region of the molecular potential of 1,3-benzodioxole.


Determination of the effects which are responsible for the preferred structural conformations of molecules is often less than straightforward.

It is useful to keep in mind that the understanding of such a basic feature, as the staggered geometry of ethane, has come only recently,1 from a careful computational study showing that this geometry results from hyperconjugation and not from steric repulsion.

In cases of molecules which have conformers of comparable energy separated by small energy barriers the situation is often much more complex.

Even the geometry at the global minimum, and not only its origin may be uncertain.

Several spectroscopic studies of 1,3-benzodioxole (BDO) have been carried out, and were devoted to elucidation of details of its large amplitude motion, which is a convolution of ring puckering, flapping and ring twisting.

The molecule is shown in Fig. 1, where the parameters used to describe the ring puckering and flapping motions are also indicated.

Earlier investigations led to conflicting conclusions.

The results of a far infrared (FIR) investigation2 were interpreted in terms of planarity of the molecule, whereas a single vibronic level fluorescence (SVLF) study3 suggested that the molecule is non-planar, with the barrier to planarity attributed to a twisting motion at the oxygen atoms in positions 1 and 3.

The pure rotational spectra of the ground state and of several, mainly ring puckering, vibrational satellites, also indicated unambiguously that the molecule is non-planar, but that the barrier to planarity is due to the puckering motion of the five membered ring.4

The puckered conformation of BDO has been interpreted5–7 to originate from the anomeric effect, which is a stereoelectronic effect that involves the donation of electron density from a lone pair on one oxygen atom to the adjacent carbon-oxygen bond.

Such interaction, believed to be the result of n–s* overlap, reaches its maximum value when the –C–O–C–O torsional angle is at 90° and is excluded in case of a planar ring.

Since the puckering of the ring increases the magnitude of the anomeric effect, there can be a stabilization of a nonplanar structure.

The studies reported in refs. 5–7 combined the analysis of far-infrared and Raman spectra with that of laser-induced fluorescence (LIF) from electronic transitions between the ground (S0) and the first excited electronic (S1) states.

In ref. 6 a partial reassignment was made of the fundamental transition of the butterfly motion, with respect to refs. 2 and 4 while in ref. 7 the potential energy surface of the ring puckering and butterfly motions was determined for the first time in the S1 state.

Following the experimental results, two theoretical studies confirmed the anomeric effect to be the main effect in determining the bent conformation of BDO in both S0 and S1 states.8,9

We will use throughout the text the notation υp, υf and υt to label the vibrational states of ring-puckering, flapping (or butterfly) and twisting motions, respectively.

In normal mode numbering the pertinent quantum numbers are υ39, υ38 and υ20, respectively.

The electronic transitions will be labelled using the notation Nn1n0, where N labels the normal mode, n0 is the vibrational quantum number for this mode in S0, and n1 is the corresponding value in S1.

Analysis using a 2D model comprising of ring-puckering and butterfly motions, was performed in refs. 6 and 7 although some unusually large discrepancies between the observed and the calculated transition frequencies were obtained.

Since rotationally resolved techniques allow for unambiguous identification of vibrational states, some of us measured the rotationally resolved spectrum of several strong, mainly cold bands of the LIF spectrum of the S1 ← S0 (π* ← π) electronic transition of BDO.10

It was found that, for S1, the rotational constants of state υp = 2 obtained from the 3920 and 3921 transitions as assigned in ref. 7, were incompatible with each other.

A new assignment of the ring-puckering states in S1, needed to understand the molecular geometry and the anomeric effect, is still missing.

In order to succeed in the reassignment, we need to be sure of the assignment in S0.

Since the υp = 1–0 (ΔE01) splitting has not been measured experimentally, but deduced from the centrifugal distortion effects,4 this was the primary target of additional measurements.

At the same time we searched for the rotationally resolved spectrum of the 3911 transition, measuring some weak hot bands close to the band at 34 783.53 cm−1, so far attributed to the origin.

Experimental details

BDO (99% purity) was purchased from Aldrich, and was used without any further purification.

The millimetre wave spectra were recorded in Warsaw, and the rovibronic spectra in Florence, as described below.

Rotational and vibration–rotation spectra

We used the broadband millimetre wave spectrometer based on backward wave oscillator (BWO) sources, which has been described in ref. 11 and references cited therein.

The spectrometer uses source modulation, and the frequency of the source is locked to the controlling 3 GHz synthesizer by means of two phase lock loops.

The millimetre wave spectrum was recorded at frequencies from 170 to 279 GHz, by means of two different mm-wave sources.

Measurements were made on a static sample kept at room-temperature and pressure of 20–50 mTorr, which was held in a 3 m long absorption cell of 10 cm diameter.

Rovibronic spectra

The fluorescence excitation spectra were recorded using the experimental apparatus in Florence, described in detail elsewhere.12

The sample was introduced directly into the molecular beam source and was heated to 50 °C, to reach a suitable concentration of BDO in the molecular beam, while the 50 μm-diameter nozzle was maintained at a slightly higher temperature (55 °C) to avoid condensation and clogging of the nozzle.

The carrier gas was pure helium at a pressure of 6 × 105 Pa for the cold bands, and 8 × 104 Pa for the hot bands.

The molecular beam was obtained by selecting the central part of the expansion with a 400 μm-diameter skimmer, placed at 10 mm from the nozzle.

On passing through the skimmer the molecular beam entered a second, differentially pumped, chamber in which it could interact with the UV radiation and where the LIF signal was collected.

The excitation source was a ring dye laser (Coherent 699-21) with Rhodamine 6G as lasing medium pumped by an argon ion laser.

The ring laser was operating in single mode, and was actively frequency stabilized.

To obtain the UV radiation the frequency of the ring laser was doubled using an external cavity (Wavetrain, LAS), and we obtained up to 20 mW of UV power, at the crossing point with the molecular beam.

The UV linewidth was 3 MHz.

The fluorescence radiation was collected by a system of two spherical mirrors and two lenses, and it was focused on the photocathode of a photomultiplier tube (EMI 9893Q/350).

The photomultiplier tube signal was then sent to a photon counter (SRS 400).

The frequency scan of the laser was monitored by a 150 MHz free spectral range étalon (to check the regularity of the scan and to control that the laser does not jump from the selected cavity mode), and by a wavelength-meter (Burleigh Wavemeter) to check the absolute calibration.

The accuracy of the laser wavelength measurement is of 0.04 cm−1 in the UV.

The data acquisition and the laser scan were controlled by a personal computer.


Rovibronic spectra

Upon realising from measurement of bands displaced from 94 to 383 cm−1 from the origin band10 that the assignment of ref. 7 was in question, we considered it crucial to assign the 3911 transition, expected to lie close to the origin band.

This would have supplied the ΔE01 splitting in S1, to compare with the value of ΔE01 in S0.

We found, in the proximity of the origin at 34 785.53 cm−1, two candidate hot bands, one displaced by −14.5 cm−1, and a second one by +1.5 cm−1 (Fig. 2).

The results of fitting the measured frequencies for these two bands, using a rigid rotor Hamiltonian and the program Jb95,13 are given in Table 1.

The measured and the obs. − calc. frequencies are listed in supplementary Tables S1 and S2. The rotational constants of the lower states for the two new bands are compared in Table 2 with those for the lowest excited vibrational states of S0 known from the microwave investigation.4

It can be seen that the B and C constants differ by close to ten times the experimental errors from the values in the S0υp = 1 state.

Neither of the two new transitions appears to originate from that state, so none of these bands can be satisfactorily assigned to the 3911 transition on the basis of rotational constants.

In addition assignment of the lower frequency band to 3911 would require a considerable difference between the ΔE01 values in S1 and S0, and a much larger ΔE01 spacing in the S0 state (by at least 15 cm−1) than the reported value of 8.1–8.3 cm−1.4

On the other hand assignment of the +1.5 cm−1 band would require the 3921 band to be within about 8.2 + 1.5 = 9.7 cm−1 of the 3920 band, whereas candidate bands according to the assignment of ref. 7 and the measurements in ref. 10 are 7.53(11) cm−1 apart.

If we compare the S0 rotational constants obtained from the two new bands with the values for the lowest lying vibrational satellites in S0 obtained from the pure rotational spectrum, we can see that they most closely match those of υp = 2 and of the combination state (υp = 2, υt = 1).

A correspondence with υt = 1 also cannot be excluded.

For any of these to be the case the 000 band, the origin, would actually have to be at a considerably lower value, by ca. 300 cm−1, than the reported frequency.

We have indeed seen some indications of the presence of very weak bands in that region.

The inconsistencies discussed above provide further evidence, in addition to that already reported in ref. 10 that the assignment of the ring puckering progression in S1, as given in ref. 7 is most probably incorrect.

It appears that the key 3911 band, and even the actual band origin are still to be identified.

Since the value ΔE01 = 8.4 cm−1 in the S0 state was not an experimental one, but was derived from experimental centrifugal distortion constants we made an attempt to measure it directly, with millimetre/submillimetre wave spectroscopy.

The results are given in the following section.

Millimetre wave rotational and vibration–rotation spectra

The measurements of the pure rotational spectrum in the ground, υp = 0, and in the first excited ring puckering state, υp = 1, were first extended to the millimetre wave region by using predictions based on the rotational constants of .ref. 4

The spectrum is relatively weak, since BDO has a small dipole moment equal to 0.237(5) D and 0.222(4) D in the ground, and in the first excited state, respectively.4

The dipole moment is in the form of the μa dipole moment component, so that only a-type rotational transitions would be expected.

Since the asymmetry of the molecule is appreciable, with the asymmetry parameter κ = −0.643 for the ground state, the spectrum was expected to be devoid of many of the characteristic features that normally serve to aid the assignment.

Nevertheless, as the molecule is close to planarity it gives rise to the type-II bandheads,14,15 which are characteristic of high-J transitions in such molecules.

The bandheads are produced by specific overlaps of aR-type transitions with differing values of J and K−1, such that J decreases and K−1 increases by 1 away from the leading line in the band.

Such bands were indeed observed for both states, as visible in the specimen spectrum reproduced in Fig. 3, and their measurement allowed for rapid expansion of the pure rotational data set.

An attempt to fit such data demonstrated that it was no longer viable to carry out single state fits as was possible with the less extensive data sets of .ref. 4

In the present case it was necessary to carry out simultaneous fits of data for υp = 0 and υp = 1, by taking account of the high order, Coriolis-type interaction connecting such states.

We used the block diagonal Hamiltonian of the formH = ΔE01 + Hr(0) + Hr(1) + Hc(01),where Hr(0) and Hr(1) are single state Watson’s asymmetric rotor terms in representation Ir and reduction A.16

The interstate coupling term Hc(01) has been expressed in the reduced axis system Hamiltonian suggested by Pickett17Hc(01) = (Fac + FacJP2 + FacKPa2)(PaPc + PcPa),which is equivalent to a second order Coriolis coupling term.

The empirical centrifugal distortion expansion of the Fac coupling constant proved to be necessary for fitting vibration–rotation transitions.

The measured frequencies were fitted using Pickett’s SPFIT program.18,19

Pure rotational aR0,1-type transitions were measured for υp = 0 and υp = 1 states at frequencies up to 210 GHz, which are near the maximum of the rotational absorption profile, see Fig. 4.

The values of rotational quantum numbers ranged up to J = 90 and K−1 = 40.

The pure rotational data set by itself already allowed the ΔE01 vibrational energy level spacing to be determined to within several tens of MHz, at which point location of vibration–rotation transitions became feasible.

Since the dipole moment change induced by the puckering vibration will be perpendicular to the ring plane and along the c-inertial axis, the vibration–rotation transitions are expected to follow c-type selection rules.

Their initial prediction was made by assuming that the associated vibrational dipole moment was at least comparable in magnitude to the permanent dipole moment.

The calculated intensity profiles are shown in Fig. 4 and take the form of a considerably distorted P, Q, R structure.

The compact Q branch at 260–270 GHz was thought to offer the best chance for observation, while the comparably intense R branch was outside the frequency range of the available spectrometer.

The prediction proved to be sufficiently accurate to readily locate the Q-branch interstate transitions, which were indeed found to be considerably more intense than the pure rotational transitions, see Fig. 5.

The vibrational transition moment proved to be sufficiently high to observe also the P-type interstate transitions, some of which are marked in Fig. 3.

We assigned and measured a total of 181 υp = 1 ← 0 transitions and fitted them together with pure rotational transitions in υp = 0 and υp = 1 states, also using the SPFIT program.

The resulting spectroscopic constants are reported in Table 3, and the measured transition frequencies and obs. − calc. differences are collected in Table S3 of the ESI. The vibrational spacing has been determined with a very high accuracy, ΔE01 = 259 726.035(10) MHz, that is 8.663 5280(4) cm−1.

Part of the reason for this accuracy is the low level of intercorrelations affecting ΔE01.

It is interesting to compare relatively small values of the ΔJK constants in Table 3 with 0.644(10) and −0.575(11) kHz determined, respectively for υp = 0 and υp = 1 states in single state fits.4

This provides direct evidence for the level of effective centrifugal distortion contributions introduced by the inversion potential, which was the basis for the derivation of additional information on the potential carried out in ref. 4.

Potential function

For the first time, we have a precise experimental value for the ground state ring puckering splitting in S0, whereas in previous analyses4,6 this could only be determined indirectly.

The present value of 8.663 5280(4) cm−1 is found to be intermediate between previous estimates.

Two values, 8.1 and 8.31 cm−1 were determined from potentials fitted to observables from pure rotation spectra4 and far-infrared data of .ref. 2

A value of 9.6 cm−1 was also inferred upon reinvestigation of the far-infrared spectrum,6 and was found to be consistent with the potential fitted therein to far-infrared, Raman, and fluorescence spectra.

Until now assignment of transitions involving higher quanta was considered reliable, such as that of υp = 2 ← 0, for which values of 99.1 and 100 cm−1 were derived from far-infrared and dispersed fluorescence spectra,6 respectively.

In view of the uncertainty concerning the reliability of the assignment we decided to re-examine the available data by going back to the simple one-dimensional analysis.

An isolated ring-puckering mode would be expected to be well described by a double minimum potential, as has been the case for 1,3-dioxole.20

An economic form of such a potential, such as the reduced quartic–quadratic21V(z) = A(z4 + Bz2),is defined by only two parameters, and central barrier to planarity, V0, is simply given by AB2/4.

Even in the presence also of other low-frequency normal modes this type of 1D-potential section should provide a sufficiently accurate description, at least in the energy region below excitation of other modes.

Since there is a consensus between theory and experiment that both the twisting and the flapping modes are well above 200 cm−1 in frequency4,6,10 there should be several puckering energy levels in the intermolecular potential below the first excited states of the twisting or flapping motions.

In addition the several operable (and for practical purposes mutually exclusive) selection rules, Δυp = 1,3 in the far-infrared, and Δυp = 2,4 in Raman20,22 are expected to generate quite a sizable number of observable transitions between only the several lowest vibrational states.

We used the program ANHARM from the PROSPE web-site23 for these calculations, together with the new ΔE01 and vibrational frequencies from .ref. 6

We decided to retrace the assignment from scratch and the results are collected in Table 4.

For a double minimum potential the compression of the first two vibrational states by the presence of the central barrier at the planar configuration is accompanied by a somewhat smaller compression of the next higher pair of states, which leads to the second lowest vibrational frequency.

It was immediately found that for the ab initio estimate V0 = 150(50) cm−1 for BDO9 the use of the current value of ΔE01 leaves only the 58.2 cm−1 line in the infrared spectrum as the only realistic candidate for υp = 3 ← 2.

This is in fact the assignment of .ref. 6

The use of these two lines to predict the υp = 2 ← 1 transition leads to a frequency of 76 cm−1.

The most probable assignment is therefore the nearest line to high frequency, at 79.5 cm−1, since there is no candidate to low frequency.

In fact the 58.2 cm−1 line is rather broad and structured, so it is better to use 8.66 and 79.5 cm−1 transitions to define the potential, whereupon we immediately find that the 148.2 cm−1 line can be assigned to υp = 3 ← 0.

The only other transition involving quanta up to υp = 3, namely υp = 3 ← 1 is indeed also found in a consistent position in the Raman spectrum, at 138.5 cm−1.

We have thus satisfactorily assigned all transitions up to υp = 3, and we note that the agreement for υp= 3 ← 2 would be improved if the mean frequency of its broad profile and not the low frequency peak were used.

Nevertheless assignment of transitions involving the υp = 4 level meets difficulties, while there are plausible candidates for transitions to the υp = 5 level, but not involving υp = 4.

We note that the calculated energy of 231 cm−1 for υp = 4 is close to the estimated positions υt = 1 and of υf =.14,9

There could therefore be several mechanisms by which the position of this level could be perturbed.

If we assume a small 6 cm−1 upward shift in the position of υp = 4 then it becomes possible to assign five more vibrational transitions, as seen in Table 4.

It is noted that all of the assigned transitions are consistent with those with the selection rule Δυp = 1,3 being observable in the infrared, and those for Δυp = 2,4 being observable in Raman, and that relative intensities are in fair agreement with the calculated ones.

A further test of the derived 1D potential was carried out by checking its ability to reproduce a second reliable piece of experimental data namely changes of rotational constants and inertial planar moments with υp measured in .ref. 4

For this purpose we used a different form of the double minimum potentialV(τ) = V0[1 − (τ/τe)2]2,where V0 is the barrier to inversion and τe is the equilibrium value of the inversion angle (see Fig. 1).

This potential was augmented with a flexible model, available for one and two-dimensional motions,24,25 and the molecular framework was constructed by using the parameters listed in .ref. 4

Vibrational energies in Table 4 were reproduced with V0 = 108.5(11) cm−1, τe = 22.50(6)° and the changes in rotational constants and second moments of inertia required two additional adjustable parameters describing the variation in the CCO valence angle α with puckering excitationα(τ) = α0 + Δα(τ/τe)2.The results of this calculation are compared with experimental values in Table 5.

It should be noted that observation of the zigzag behaviour in rotational constants in the range υp = 1–3 is indicative straightaway that these levels are in the vicinity of the central barrier.

In the derived 1D potential υp = 0 and 1 are 50 and 41 cm−1 below the barrier, respectively, while υp = 2 is 38 cm−1 above the barrier.

Reproduction of the planar moments and rotational constants as documented in Table 5 is seen to be excellent.

The A rotational constant is a possible exception, but its behaviour is known to be the most difficult to reproduce, since it is associated with the smallest moment of inertia.

The barrier of 108 cm−1 corresponding to the present 1D assignment is somewhat lower than analogous 1D values of 124 cm−1 (ref. 4) and 128 cm−1.6

It is also lower than barrier heights of 126 cm−1,4 and 164 cm−1,5 derived on application of a more sophisticated two-dimensional model.

We have, in fact, also sought a 2D model solution that would encompass the 1D solution described above.

We have found that the density of lines in the spectra is such that if the criterion for reproducing frequencies is reduced to, say, 3 cm−1, and the relative intensity criteria are also relaxed, then it is not possible to achieve a unique fit.

The number of parameters in the 2D-fit and their flexibility is such that they are able to equally successfully encompass many radically different assignments of transitions.

A confident assignment of the υf = 1 ← 0 transition would be the minimum requirement for an attempt at a more reliable analysis with that model.

On the other hand observation of the υp = 2 ← 0 transition in the Raman spectrum near 88 cm−1 would be enough to confirm the present 1D analysis.


The present work has delivered several new, precise pieces of information pertaining to the molecular dependence of the chemically relevant anomeric effect.

Augmentation of the available lower resolution data by information from rotationally resolved spectroscopy revealed considerable deficiencies in previously published spectral analysis, suggesting that a considerable re-assignment of the ring puckering transitions in 1,3-benzodioxole is required for both the S0 and the S1 states.

The origin of the S1 ← S0 electronic transition almost certainly has to be re-assigned, it is expected to lie at a lower frequency than that reported in refs. 3 and 5–7 and further work on this problem is in progress in the Florence laboratory.

It is now clear that, while there is little doubt that the puckering barrier in 1,3-benzodioxole is considerably lower than the value of 275 cm−1 in 1,3-dioxole,20 its precise elucidation is proving rather difficult, as evidenced by the number of previous reports devoted to this problem.

While a full 2D analysis of the dynamics of the five-membered ring would be more appropriate, it is mandatory to first reach a confident assignment of several more states, in particular those with υf > 0.

On the other hand the success of the 1D analysis suggests that near the minimum of the potential the dynamics is dominated by the puckering.

It is also worth noting how close the actual value of ΔE01 = 8.663 5280(4) cm−1 turned out to be to the value of 8.31 cm−1 calculated on previous application of the 1D model to all data from pure rotational spectroscopy4.