Roesky’s ketone: a spectroscopic study

We present a joint experimental-theoretical spectroscopic study of 5-oxo-1,3,2,4-dithiadiazole, also known as Roesky’s ketone.

The theoretical results of a vibrational analysis, calculated at the DFT/B3LYP/6-311+G* level of theory, of the title compound have been compared with experimental data, consisting of Raman and IR frequencies in different phases, and the bands have been assigned to the normal vibrations of the molecule.

Additionally, an analysis of the origin of the high intensity of the band assigned to the CO stretching mode was performed based on calculated stockholder charges and atomic dipoles.

The results of theoretical calculations of the 13C and 14N NMR chemical shifts are compared to experimentally obtained shifts.


Compounds containing an –NSN–S– fragment are relatively new and their molecular properties, most importantly reactivity and opto-electronic features, differ substantially from those of the corresponding hydrocarbons.

This makes these compounds very interesting for further experimental and theoretical studies.

During the past few years a considerable body of work has been performed on these kinds of compounds, in the context of a multidisciplinary study in the framework of the search for new materials and their possible applications (see ref. 1–6).

Within this collection of new compounds, ring structures in which the –NSN–S– fragment is closed into a cyclic system by one or more atoms have assumed a prominent role since issues like delocalisation of π-electrons, ring currents and aromaticity become highly interesting subjects of study, both from a theoretical and an experimental point of view.

In recent years we have been focussing on five- and six-membered rings such as 1,3λ4δ2,2,4-benzodithiadiazine and its fluorinated derivates7,8 and the title compound, Roesky’s ketone or 5-oxo-1,3,2,4-dithiadiazole.9

We are currently performing a similar study on an organometallic (SN)2 compound containing a cobalt atom.

We recently subjected Roesky’s ketone to an intensive computational study to gain a deeper insight in the appropriate methods and basis sets for the most reliable description of the compound’s properties, in particular its geometry.9,10

Additionally, a detailed analysis of the bonding and structure of the ketone was performed with the emphasis on orbital topologies, atomic charges, atomic and molecular dipoles, aromaticity parameters and Fukui functions.

In the current paper we present a joint theoretical and experimental study on the spectroscopic properties of Roesky’s ketone: the infrared (IR), Raman and nuclear magnetic resonance (NMR) spectra are discussed and a detailed analysis of the origin of the high intensity of the band assigned to the CO stretching mode was performed based on calculated stockholder charges and atomic dipoles.


Roesky’s ketone was synthesised as previously reported.9

Vibrational spectroscopy

The IR spectra of the compound, dispersed in KBr pellets, were recorded on a Bruker IFS 66v Fourier transform spectrometer, using a Globar source in combination with a Ge/KBr beam splitter and a broadband mercury cadmium telluride (MCT) detector.

The interferograms were averaged over 200 scans, Happ-Genzel apodized and Fourier transformed using a zero filling factor of 4 to yield spectra at a resolution of 2 cm−1.

FT-Raman spectra of the compound were recorded on a Bruker IFS 66v interferometer equipped with a FT-Raman FRA 106 module.

The sample was excited by the 1064 nm line of a Nd/YAG laser operating at 150 mW.

For each spectrum, 2000 scans, using a resolution of 4 cm−1, were averaged and Fourier-transformed with a zero filling factor of 4.

NMR spectroscopy

13C NMR and 14N NMR spectra were recorded on a Jeol DELTA GSX270 spectrometer in CDCl3 and C6D6 with the chemical shift δ of the carbon atom referenced to external tetramethylsilane (TMS) and that of the nitrogen atoms to nitromethane.

All measurements were performed at 25 °C.

Computational methods

All calculations were performed on isolated molecules in Cs symmetry (see ref. 9 for further details) using the Gaussian 9811 and BRABO12 packages, at the DFT/B3LYP13 and the Hartree–Fock (HF)14 levels of theory with the 6-311+G* 15,16 and Dunning’s cc-pVTZ17,18 basis sets.

Geometry optimisations were performed as previously reported.9

Spectroscopic properties were calculated starting from optimised geometries with the same method/basis set combination as the one used for the corresponding geometry optimisation.

Harmonic force field calculations were performed at the DFT/B3LYP/6-311+G* level of theory, yielding the frequencies used in the vibrational analysis.

No attempts were made to scale the force field.

The Cartesian force field was transformed into one described in a set of pseudosymmetry coordinates19,20 to construct the potential energy distribution (PED), which was subsequently used to derive approximate descriptions of the normal modes.

Stockholder charges qA were calculated as previously reported,21 based on the Hirshfeld partitioning of space.22

Components of the atomic dipole moments μA were calculated according to μA,n = −∫rnΔρA(r⃑)dr⃑, with n = x,y,z, r⃑ the position vector of the atom and Δρ the deformation density as usually defined.23

Chemical shielding factors were calculated at all atomic positions at the HF/6-311+G*, the DFT/B3LYP/6-311+G* and the DFT/B3LYP/cc-pVTZ levels of theory, at the corresponding geometries, using the GIAO method24–28 implemented in Gaussian 98.

The chemical shift for the carbon atom was obtained by subtracting the chemical shielding value of this atom from the one calculated for tetramethylsilane (TMS) which is 195.8637 ppm at the HF/6-311+G* level, 184.0193 ppm at the B3LYP/6-311+G* level and 184.4807 ppm at the B3LYP/cc-pVTZ level, based on the corresponding geometries.

Chemical shifts for the nitrogen atoms were obtained by subtracting the chemical shielding values of the two atoms from that calculated for nitromethane (CH3NO2) which is −184.7452 ppm at the HF/6-311+G* level, −152.4366 ppm at the B3LYP/6-311+G* level and −147.6395 ppm at the B3LYP/cc-pVTZ level, based on the corresponding geometries.

The molecular framework and atomic numbering are shown in Fig. 1.

Results and discussion

Vibrational spectroscopy

Table 1 compares the theoretical vibrational frequencies, calculated at the B3LYP/6-311+G* level, with the experimental ones found in the gas-phase IR spectrum,29 the Nujol IR spectrum,30 the Raman spectrum in dichloromethane30 and with new solid-state IR and Raman data.

They have been ordered according to the Herzberg numbering scheme for the normal modes.

The experimental values can be easily assigned based on the accordance between the frequencies and, in the case of the Nujol and the KBr spectra, also the intensities.

The assignments have been derived from the PEDs of the normal modes.

The molecule has twelve normal coordinates, which can be subdivided into six stretching modes ν, along each of the six bonds, an in-plane CO bending mode β, an out-of-plane CO bending mode γ, two in-plane ring deformation modes Δ and two out-of-plane ring torsion modes Γ.

It is interesting to note that the seven bands at the low-wavenumber end of the spectrum (ν6ν12) and the CO stretching mode (ν1) correspond to very pure normal modes according to the calculations: contributions from other symmetry coordinates than the principal one are limited to 18%.

For the four stretching modes at higher wavenumbers (ν2ν5) considerable mixing with other symmetry coordinates such as the ring bending modes Δ and other stretching modes, is observed, as is to be expected in ring systems.

Nevertheless, there is always one major contribution and this has been listed in Table 1.

The only bands that can be clearly distinguished in the spectra are the ones with a considerable calculated intensity, and the most intense band in the IR spectra is νCO (ν1), as expected.

Some of the frequencies appear in all of the spectra and correlate very well with the calculated spectrum, whereas some of the calculated frequencies are obviously lacking in the experimental spectra.

As can be seen, the bands at 874 and 515 cm−1 can not be found in any of the experimental spectra.

These bands may be too weak to be observed or, in the case of the band at 515 cm−1, accidentally degenate with the one predicted at 522 cm−1.

As expected, the calculations overestimate the higher wavenumber for the νCO band considerably, but under about 1000 cm−1, the correlation is quite acceptable and the differences remain small.

The poor correlation between the experimental and theoretical values of the calculated band at 307 cm−1 needs to be commented on, since the correlation between experiment and theory is generally better for lower wavenumbers.

Previously, we discussed the poor correlation between the CS bond length calculated at the B3LYP/6-311+G* level, and the corresponding interatomic distance found in the crystal by low-temperature XRD:9 the calculations [1.9410 Å] grossly overestimate the bond length found in the solid state [1.8305(17) Å].

The expression of this inconsistency in the vibrational spectrum is then straightforward: an overestimated bond length means an underestimated force constant for the corresponding stretching mode, which means that the frequency will be likewise underestimated, as can be seen in Table 1.

Nevertheless, the Raman frequencies observed at 377 cm−1 in the liquid and at 390 cm−1 in the solid can only be assigned to the calculated value of 307 cm−1.

The intensity of the CO stretching mode

The intensity of an IR band is proportional to the square of the change of the dipole moment μ with the normal coordinate Q describing the vibrational mode,The stockholder partitioning scheme21,22 allows the interpretation of the dipole moment in terms of atomic charges and atomic dipole moments.23,31

The nth component of the molecular dipole, μn with n = x, y, z, can be written aswhich is the sum of the contributions of charge transfer (first term, with qA the stockholder charge and RA,n the coordinate of atom A) and of the intra-atomic charge polarization, calculated by the atomic dipole moment components μA,n (second term).

By doing this for different values of Q an interpretation of the change of the dipole moment with the normal coordinate in terms of changes of atomic dipoles and charges becomes possible, and the origin of dμ/dQ can be studied.

The high purity of the CO stretching mode (ν1), henceforth abbreviated QCO, and the high intensity of the corresponding band in the experimental IR spectrum, makes this the ideal candidate for such an analysis.

The relevant data has been graphically presented in Figs. 2 and 3.

The CO stretching vibration was simulated in seven discrete steps (ΔrCO = 0.1 Å) and these are represented in the horizontal axes in Figs. 2 and 3, going from the shortest (point 1) CO distance to the longest (point 7) via the equilibrium position in point 4.

Fig. 2a displays the variation of the total molecular dipole moment and of its components, which were calculated using eqn. (1); the values are given in Debye (D) and were recalculated from the corresponding values in atomic units (au) using 1 D = 0.393 426 58 au.

The separate atomic contributions to the total molecular dipole, as defined in eqn. (1), can now be discussed.

Fig. 2b shows the changes in the atomic positions with the CO stretching coordinate: the two atoms comprising the CO bond show the largest displacements while the other remain virtually stationary – this confirms the purity of the normal mode.

A closer look at the changes of the atomic charges for each of the positions of the vibration, which are given in Fig. 2c, reveals that there are large variations in the charges on all six atoms, but that the largest changes in stockholder charge can be found on the oxygen atom, O(5), followed by C(5) and S(1); the charges on the two nitrogen atoms and S(3) undergo the smallest variation.

Upon stretching the CO bond the oxygen and carbon atoms both gain negative charge (0.322 and 0.156 |e|, respectively) while the four other atoms gain positive charge: the largest increase is found for S(1) (0.232 |e|).

Thus, the negative charge which is added to O(5) during stretching seems to come primarily from S(1).

Fig. 2d shows that there are large fluctuations in the atomic dipole moments, |A|, for all the atoms during the stretching of the CO bond.

The variations in the two components of this property in the molecular plane, μA,x and μA,y, have been presented in Figs. 3a and 3b, respectively, and the figures show some interesting trends.

The value of the total atomic dipole on N(2) remains virtually constant during the stretching, as do its components, at least around the equilibrium point.

In contrast, the value of the atomic dipole on N(4) decreases substantially, as do its components (especially μA,x) indicating that the atomic dipole on this atom rotates in the plane of the molecule.

In comparison, the atomic dipoles on S(1), S(3) and C(5) are small and again their fluctuations are related to the changes in the two components.

The atomic dipole of O(5) becomes very small in point 2, i.e. for a situation where the CO bond is shortened and, as can been seen from Fig. 2a, at the exact point where dμ/dQ changes its sign (vide infra).

At the equilibrium position (point 4) dμ/dQ has a positive sign.

When the CO bond is shortened, the molecular dipole becomes very small and for further reduction of the bond length, dμ/dQ has a negative sign.

The most important contribution to the total dipole in this region comes from μy because μx is very small.

The fact that the contribution of the y-component is the most important can easily be interpreted because the CO bond lies virtually parallel to the y axis (see Fig. 1); consequently, when the bond is stretched, μy becomes larger.

To explain what the specific contribution of μy is caused by, the graphs in Figs. 3c and 3d need to be analysed.

From equation 1 it is known that μn is the sum over all the atoms of the terms μA,n and qRA,n.

The combination of Figs. 3a and 3c on the one hand and Figs. 3b and 3d on the other indicates that the contributions to qxA + μx and qyA + μy due to the components of the atomic dipole moments, μx and μy, are several times smaller than contributions due to the components of the charge transfers, qxA and qyA.

For the latter two the most important contribution to the variation of the components of the molecular dipole with the normal coordinate QCO comes from the oxygen atom, O(5).

The major contribution to qxA + μA,x and qyA + μA,y for O(5) must therefore come from the qxA and qyA terms of O(5), or indirectly the variation of the stockholder charge on the oxygen atom.

The latter is then the cause of the high intensity of the band of the CO stretching mode (ν1) in the vibrational spectrum.

NMR spectroscopy

The experimental NMR data of Roesky’s ketone, in particular 13C and 14N chemical shifts, have been compiled in Table 2.

Regarding the 13C NMR data, only one singlet can be found in the spectrum at 189.91 ppm in CDCl3 and at 188.90 ppm in C6D6.

The calculated NMR shift values correlate well with the experimental values.

Regarding the 14N NMR data, two singlets are found in the spectrum which means that the spins do not couple with any other nuclei, and this makes the experimental assignment difficult, even though it is known that N(2) is generally deshielded with respect to N(4) for these types of compounds.8

The theoretical calculations confirm this general trend in the case of Roesky’s ketone.

As can be seen, the correlation of the absolute values between theory and experiment is far less satisfactory than for the carbon spectrum.

The difference between the calculated and experimental values for N(2) amounts to more than 300 ppm.

For N(4) the calculations predict a strongly shielded nitrogen atom with respect to the one in the reference compound, in contrast to the experimental value.

When these data are compared with previous studies on substituted 1,3λ4δ2,2,4-benzodithiazines,8 it is clear that the unsatisfactory numerical agreement obtained here is not unusual and this confirms the complicated nature of calculated nitrogen chemical shifts and their comparison with experimental data.

Furthermore, our findings for Roesky’s ketone indicate that the values are quite dependent on the basis set used.


In order to perform an in-depth spectroscopic study of Roesky’s ketone, new 13C and 14N NMR spectra were recorded, as well as solid-state Raman and infrared spectra.

The newly recorded vibrational spectra together with the data already present in the literature, comprising IR and Raman spectra in different phases, have been assigned based on the results of force field calculations at the DFT/B3LYP/6-311+G* level.

The analysis of the vibrationally induced changes of the molecular dipole moment due to the CO stretching mode reveals that the charge transfer of the oxygen atom contributes most to the change of the dipole moment during the vibration, causing the high intensity of the IR band.

Observed trends in NMR parameters of the title compound can be theoretically predicted based on calculated NMR shielding constants.