¹⁷O NMR AND DFT STUDY OF HYDROGEN BONDING: PROTON SHARING AND INCIPIENT TRANSFER

1. Introduction

Hydrogen bonding (H-bonding) is an important structural motif in a wide variety of molecular systems. It is still not completely understood and is constantly producing new surprises [1]. This particularly holds for the cases of very short H-bonds [2, 3]. The well-known effects are expressed there in a radically different manner relative to their analogs with moderate or long H-bond lengths.

The sensitivity of ¹⁷O chemical shifts and nuclear quadrupole coupling constants to H-bonding has attracted attention for several decades [4–8]. Due to difficulties arising from the  low natural abundance of ¹⁷O the investigations based on the ¹⁷O properties have been gaining a wider publicity only recently [9–13]. However, investigations of examples with very short H-bonds are still rather scant [3, 4, 9, 14, 15]. On the  other hand, such systems are of great interest to biochemical processes and material properties. Recent advances have shown that, with the  availability of ultrahigh magnetic fields (e.g. 21 T), ¹⁷O NMR becomes applicable in studies of biological macromolecules in both solution and solid states, see [16, 17] and references therein.

However, a systematic ¹⁷O NMR investigation of the effects of a progressive proton displacement from the  proton donor to the  acceptor, including the  proton transfer influence on increasing H-bond strength, does not seem to be carried out yet. Such a gradual increase of bond strength resulting from changes in either the proton donating or accepting moieties can be achieved by appropriate substituting in the H-bonding partners whereby the groups forming the H-bond remain unchanged.

An appropriate series of gradually changing H-bonds can be formed by azine N-oxide complexes with acids. Moreover, compounds with corresponding intramolecular H-bonds lend themselves to comparative investigations of this type of bonding with the intermolecular ones that are in the same range of O…O distances. Azine N-oxide complexes with carboxylic acids have been fairly extensively investigated in the past [18–20]. These works were aimed mostly at demonstrating the relationships between the  differences of pK’s of the  components and the  spectroscopic evidence of the strength of the resulting H-bonds, such as the centre of gravity of the broad OH absorption in the FTIR spectra as well as ¹H and ¹³C chemical shifts. The results were reviewed in Ref. [20].

Meanwhile, precise X-ray and neutron diffraction studies of representative N-oxide complexes with carboxylic acids and molecules with corresponding intramolecular H-bonds [21–24] have shown that the  H-bonds are indeed extremely short, they are in the  range between 2.41 and 2.44 Å, and that the protons may assume positions to either nearer acid oxygen or to N-oxygen, but in any case close to the middle of the O…O vector.

¹⁷O liquid-state NMR is a useful high-resolution technique to solve structural problems for small organic molecules. The  fast reorientation of the  molecules in solutions results in narrow spectral lines due to the averaging of quadrupolar interaction. Therefore, it appeared attractive to investigate H-bonding in a  solution as reflected by ¹⁷O chemical shifts. For this work we chose the  complexes of pyridine N-oxide (PyO) with the  following acids  –  acetic (AA), cyanoacetic (CyA), propiolic (PA), trichloroacetic (TCA), trifluoroacetic (TFA), hydrochloric (HCl) and methanesulfonic (MSA) – as well as some related molecules with intramolecular H-bonds (derivatives of picolinic acid N-oxide, PANO) in acetonitrile solutions.

The acidity of the donor increases in the above order and thus the  series of complexes covers the extreme cases without and with proton transfer. The predominant position of the proton was determined from the FTIR frequencies of protonation sensitive pyridine ring modes [25]. The ¹³C chemical shifts of the ring carbon turned out to also be reliable indicators of the state of protonation [24, 26].

In addition to the  experimental NMR work, DFT calculations have been done for the magnetic shielding tensor at a fairly high level of theory [26]. The  calculations were carried out on complexes taking into account the  solvent effect by a  polarized continuum model using the  integral equation formalism (IEFPCM).

2. Experiment

Acetonitrile (ACN), anhydrous, 99.8% from Sigma-Aldrich, was used as a solvent. In some cases, ACN was kept over the molecular sieves (3 Å) for several days. Commercial acids  –  acetic (AA), cyanoacetic (CyA), propiolic (PA), trichloroacetic (TCA), trifluoroacetic (TFA), hydrochloric (HCl), methanesulfonic (MSA) and picolinic acid N-oxide (PANO) – were preliminary purified by standard methods. Water from TFA was removed by adding trifluoroacetic acid anhydride and subsequent distillation. The solutions of 1.0 M concentration in ACN were prepared by weighing (±0.1 mg) the components in a nitrogen filled dry box. Samples with PyO-HCl were prepared by direct dissolving of PyO-HCl crystals. The limited solubility in ACN allowed us to prepare the samples of 0.1  M concentration of PyO-HCl, 0.5  M for the PyO complex with MSA and ca. 0.3 M for PANO.

The 4-nitro derivative of PANO (nitroPANO) was synthesized in the  Department of Organic Chemistry of Vilnius University by nitration of PANO using a  slightly modified procedure than that reported in Ref. [27]. A  mixture of 2-picolinic acid N-oxide (1 g, 6.36 mmol), conc. H₂SO₄ (3 mL), and conc. HNO₃ (2.2 mL, d = 1.5 g/cm³) was heated at 408 K under stirring for 3 h. After cooling to room temperature, the reaction mixture was poured to a crushed ice. The formed precipitate was filtered off, washed with water to pH = 7 and recrystallized from ethanol to give 0.4 g (31%) of 4-nitropicolinic acid N-oxide, m. p. 422–423 K, Ref. [27]: m. p. 421 K. The 4-methoxy derivative of PANO (methoxyPANO) was synthesized in the  National Institute of Chemistry (Ljubljana, Slovenia) by known procedures [27].

¹⁷O NMR experiments were carried out at room temperature (T = 298 K) on a Bruker Avance III 400 WB NMR spectrometer operating at 400 and 54 MHz for ¹H and ¹⁷O, respectively, using a BBO probe-head. D₂O in the capillary insert was used for reference in the ¹⁷O NMR spectra. In order to get a better signal-to-noise ratio the natural abundance ¹⁷O spectra were recorded using 10  mm tubes. The π/2 pulses of 22.5 µs at the transmitter power of 70 W were applied using the acquisition time of 0.02 s and the recycle delay of 0.4 s. The number of scans was 400 000. The NMR spectra were processed using the Topsin 3.2 software. Additionally, the signal shapes and graphics were processed using the Microcal Origin package.

3. DFT calculations

Quantum chemical density functional theory calculations (DFT) were performed to gain the ¹⁷O NMR chemical shift information of the  investigated molecular systems. After some testing, the  hybrid gradient corrected functional of Perdew, Burke and Ernzerhof PBE1PBE [28], also known as PBE0, and the  6-311++G** basis set [29] were adopted for full geometry optimizations and magnetic shielding tensor (σ) calculations. Magnetic shielding tensor was calculated by using the  gauge independent atomic orbital (GIAO) approach [30]. In the case of quantum chemical calculations for systems involving PyO, solvent effects were taken into account by a  polarized continuum model using the integral equation formalism (IEFPCM) [31] along with the universal force field (UFF) atomic radii for the construction of the cavity as implemented in the Gaussian 03, Revision B.01 [32] suite of programs. In the case of quantum chemical calculations for systems involving PANO, solvent effects were modelled by the PCM [33] with the default settings as implemented in Gaussian 16, Revision A.03 [34].

The isotropic ¹⁷O NMR chemical shifts of the  investigated molecules (H-complexes) were transformed to the  δ-scale using the  Wasylishen proposed formula δ  =  307.9  –  σ_iso, where σ_iso is the  isotropic part of the  calculated magnetic shielding tensor [35].

This approach was proven to be adequate in various cases earlier. For example, a  fairly good agreement between the calculated and experimentally measured ¹H and ¹³C NMR chemical shifts was obtained for the molecular systems involved in a strong H-bonding, but also for rather ‘inert’ species, e.g. –CH₃ protons [24–26, 36, 37].

4. Results and discussion

Besides the  H-bond systems studied experimentally, additional DFT calculations for the  PyO complexes with water, tribromoacetic (TBA) and sulphuric (H₂SO₄) acids were carried out for completeness. The complete series is presented in Fig. 1. The sequence of subjects was approximately ordered increasing the available values of acid dissociation constants (pKa) as well as the FTIR frequencies of PyO ring modes and the ¹³C chemical shifts of ring carbons [24–26]. It is clearly seen that the series splits into two parts – the neutral B…H–A bonds with r(O–H)/R(O…O) < 0.5 and the ionic pairs BH⁺…A^– with r(O–H)/R(O…O) > 0.5 that result from the proton transfer (PT). The PyO complexes with TCA and TFA occupy a special intriguing place in the series. Two stable structures were deduced for them by DFT calculation that correspond to the H-bond complexes with and without PT (Table 1). The PyO-TCA complex was noted by Hadži et al. as a peculiar system modelling the ‘proton kicking’ along the hydrogen bond [38].

The ¹⁷O chemical shifts were calculated after the  full geometry optimization in ACN media. The  calculated and experimental values are presented in Table  1. Some experimental ¹⁷O NMR spectra of the PyO and its complexes with acids are shown in Fig. 2.

Two signals appear in all ¹⁷O NMR spectra – one fairly sharp and the other rather broad signal (except for PyO-PA, in which only one signal appears). It is necessary to find out which of the  observed signals belongs to N-oxide oxygen. This is easily accomplished by considering the ¹⁷O spectrum of a pyridine–carboxylic acid complex. Pyridine (Py) and its N-oxide have comparable proton accepting propensities in the solvents of a low polarity [39]. Therefore, the  ¹⁷O chemical shifts of a  carboxylic acid group should be similar with both acceptors, while the N-oxide signal should not appear with the pyridine complex. Indeed, only one fairly narrow signal (∆ν_1/2  ≈  190±5  Hz) was observed in the  ¹⁷O NMR spectrum of Py-TFA at 269  ppm. The  chemical shift is quite similar to one of the narrower signal in the spectrum of PyO-TFA at 262 ppm (Fig. 2(e), Table 1). Then the wider signal at ~240 ppm belongs to N-oxide.

Next to be discussed is the reason for the appearance of a single ¹⁷O NMR signal of the car-boxylic group instead of two. Two carboxylic sig nals would correspond to a unidentate bonding in the absence of averaging by a process that is fast on the  NMR time scale. Static bidentate or dynamic averaging with a  unidentate bonding will result in a single signal. Hence, one can conclude that in the studied cases the averaging process is fast on the NMR time scale.

The measured ¹⁷O chemical shifts of the  car-boxylic group cover the  range between 262 and 278  ppm (Table  1). The  experimental shift for nitroPANO drops to 324  ppm. This exceptional case can be attributed to the NO₂ substitution effect. The  measured shits are, roughly, between the shift of the hydroxylic oxygen (180 ppm) and one of the  carbonyl oxygen (312  ppm) observed in the crystalline phthalic acid [9]. However, note that in benzoic acid only one ¹⁷O signal appears (270 ppm, [40]) both in the crystal and the solution.

From the limited experimental data concerning carboxylic acids and carboxylates in various environments and hydrogen bonding strengths it was concluded that the bond strength expressed in bond shortening increases magnetic shielding [41]. It is then somewhat surprising that the ¹⁷O magnetic shielding is the same in dimeric benzoic acid and potassium hydrogen dibenzoate despite the large difference in H-bond lengths (2.65 and 2.52 Å, respectively) [42]. However, these two examples differ in charge (one is neutral and the other anionic) and it is quite possible that competing influences are operating.

For COOH signals in the studied series, there is no obvious correlation between the ¹⁷O chemical shifts and progressive proton displacement from the acid to the accepting base oxygen as it is estimated from the calculated O–H and O...O distances (Fig. 3). This means that the ¹⁷O NMR signal of COOH is a bad spectral marker determining proton location.

Considering the  N–O signal, there is a  clear trend towards increasing the  17O magnetic shielding and thus decreasing ¹⁷O chemical shift with the  increase of H-bond strength (Table  1). The observed increasing shielding corresponds to the previously noted effects of protic solvents on pyridine N-oxides [43, 44]. Notable is the parallel trend of the H-bonding and the substituent effects observed in the  series of 4-substituted pyridine N-oxides [43, 45]. The result that δ (N–O) vs r(O–H)/R(O…O) was found being a  continuous and almost linear function (Fig. 4) makes the N–O signal a proper spectral marker for the determination of proton location and thus for proton transfer.

There are several reasons that it can be even a  better marker than the  ¹H  NMR signal of the  proton directly involved in the  H-bonding. It is logical to suppose that the most sensitive to the geometry of H-bond should be the ¹H chemical shift of the bridge proton, δ (OH). However, its precise measurements often fail because of drying difficulties. In some systems, water molecules are very difficult to remove even applying molecular sieves, working in dry boxes or using other tricks (degassing, argon bubbling, etc.) [26, 46]. It is well-known that δ (OH) in strong H-bonds is high (∼14–20 ppm) [20, 24, 26]. If the moisture is present, due to the dynamic averaging with the signal of water molecules (∼5 ppm) the experimentally measured δ (OH) values can be shifted far up-field from the ‘true’ values. The lack of precision can influence the experimental data so seriously that the fine comparative evaluation of the H-bond state in the series is unsafe or even contradictory.

The next source of trouble is the  shape of the  function δ  (OH)  =  f(r(O–H)/R(O…O)). It contains the up- and down-branches that correspond to the neutral A–H…B bonds and the ionic pairs A^–…H⁺B, respectively, with the maximum at the middle point A…H…B (Fig. 4, also see [20, 26]). Therefore, sometimes it is puzzling to decide which of the two branches the  experimentally measured δ  (OH) should be attributed to. Such situation is shown in Fig. 4 for the PyO-TCA and PyO-TFA complexes, i.e. the  most challenging subjects in the studied series.

The monotonous, i.e. with no extremum, dependence of the  ¹⁷O chemical shift on r(O–H)/R(O…O) (Fig. 4) allows one to solve this problem unambiguously. It is clearly seen that the experimental point δ (N–O) = 276 ppm for PyO-TCA fits to the group of the points of neutral H-bonds. Among the PyO complexes with obviously weaker acids (AA, PA, CyA), it is the closest point to r(O–H)/R(O…O)  =  0.5, however, less than 0.5. Hence, the PyO-TCA complex in the ACN solution can be considered as H-bonding with incipient PT, whereas it is known from neutron diffraction data that in its crystalline state PT occurs [21].

In the case of PyO-TFA, the best fit of the experimental point δ  (N–O)  =  240  ppm brings it to r(O–H)/R(O…O) = 0.5, i.e. to the location of a proton at the middle point. According to literature, this is a rare case and has to be discussed in more detail. A  gape ∆r(O–H)/R(O…O) ≈  0.1 is present around the centred position, i.e. the symmetric O…H…O bonding (Fig. 4). This indicates that two minima on the potential surface that correspond to the  complexes with and without PT are separated by ca 0.2 Å. The thorough reviews of crystallographic data [47, 48] also point out that a certain gape should exist in the O–H vs O…O distances correlation for symmetric bonds. This is related with the two-site proton separation that was estimated as ~0.2 Å [48]. It was discussed that the change from the two-site proton to the centred proton occurs when the  amplitude of thermal vibration along the H-bond reaches the top of the  potential barrier. The  direct experimental evidence of thermally induced proton migration was obtained for the H-bond complex of pentachlorophenol with 4-methylpyridine by neutron diffraction experiments carried out in the range of 20–200  K [49]. The proton gradually migrates with temperature, from being closer to the O atom (at high temperature) to being closer to the N atom (at low temperature). The middle point is passed through at ~90 K. In the present work, the ¹⁷O NMR experiments were carried out at room temperature. This is sufficient to intensify the proton sharing between the donor and the acceptor in the PyO-TFA complex bringing the proton to the middle point.

The group of the acids with intramolecular H-bond (nitroPANO, PANO and methoxy-PANO) deviates from the general trend (Fig. 5). First of all, note a much steeper slope of δ (N–O) vs r(O–H)/R(O…O). It is nice to see that this result of DFT calculations is confirmed by the experiment. The steeper slope can be explained to be due to the strong substituent effect that acts beside the H-bonding and PT. The overlap of both effects was deduced for 4-substituded PyOs and differentiated in the  ¹³C NMR spectra [19]. However, the  physical origin of significant discrepancies between the  calculated and measured δ  (N–O) values for the  PANO group is not clear. There are several important factors that could cause it. The discrepancy could be related with a bad solubility of these acids in ACN (~0.3 M or even less). The outcome is that a part of material can persist as small (nano-sized?) crystallites that are mobile enough to contribute to the high resolution NMR signal. The solvent field effect modelled in the DFT calculations by PCM works improperly for such structures. As shown in [24] analyzing the solid-state ¹³C NMR spectra, the crystal force field (CFF) in PANO forces the proton from donor to acceptor atoms much more strongly than the solvent reaction field (SRF).

The next factor can be related with the referencing of ¹⁷O chemical shifts. In the present work, the δ-scale proposed by Wasylishen [35] was used. It can appear that this referencing is not the optimal one for the derivatives of PANO, and then it should be refined. The above questions were beyond the scope of the present paper and can be considered as a goal for future works.

5. Concluding remarks

The liquid-state ¹⁷O NMR spectra of pyridine N-oxide (PyO) complexes with the  series of acids as well as some related molecules with intramolecular H-bonds (derivatives of picolinic acid N-oxide) were studied in the acetonitrile (ACN) solution. The N–O signal of proton acceptor was found to be a relevant spectral marker for the determination of proton location in the H-bond and proton transfer.

The full geometry optimization carried out using DFT has revealed that the studied series can be divided into two parts – the neutral B…H–A bonds for weaker acids and the ionic pairs BH⁺… A^– for stronger acids. The  PyO complexes with TCA and TFA occupy a  challenging place. Two stable structures were deduced for them that correspond to the H-bond complexes with and without PT. A gape ∆r(O–H)/R(O…O) ≈ 0.1 present around the symmetric O…H…O bonding means that two minima on the potential surface that correspond to the complexes with and without PT are separated by ca 0.2 Å.

The ¹⁷O NMR results have revealed that PyO in a solution forms with TCA the H-bond complex with incipient PT only, whereas it is known from neutron diffraction that in the crystalline state PT occurs. This can be considered as one more confirmation of the crucial influence of environment on the  preferred proton location facilitating the  PT process and that the crystal field forces the proton to move from the donor to acceptor atoms much stronger than the solvent reaction field.

The proton location in the PyO-TFA complex was found at the middle point. As the ¹⁷O NMR experiments were carried out at room temperature, this can be due to the thermally induced proton sharing. To our knowledge, this would be the first observation of the centred O...H...O bonding in liquid solutions.

It is likely that the discrepancy between the calculation and experimental ¹⁷O NMR data for PANO-series can be due to the persistence of nano-crystallites in the solution because of low solubility. The calculations of magnetic shielding using the cluster and periodic density functional theory with plane-wave basis sets (DFT/PW) as well as Car–Parrinello molecular dynamics (CPMD) [22] would provide more information to elucidate this problem.

Acknowledgements

In large part, the present work has been inspired by the  ideas and suggestions of the  pioneer of H-bonding research, Professor Dušan Hadži (1921–2019). We thank Professor Liudvikas Kimtys for his criticism and comments. The authors acknowledge the Center for Spectroscopic Characterization of Materials and Electronic/Molecular Processes (‘Spectroversum’, http://www.spectroversum.ff.vu.lt) for the  use of spectroscopic equipment. Computations were performed on the  resources at the  high performance computing infrastructure of Vilnius University (‘HPC Saulėtekis’, http://supercomputing.ff.vu.lt).

References

VANDENILINIO RYŠIO TYRIMAS TAIKANT ¹⁷O BMR IR TANKIO FUNKCIONALO TEORIJĄ: PROTONO DALYBOS IR PRADINĖ PERNAŠA

Santrauka

Šiame darbe buvo eksperimentiškai ištirti piridino N-oksido vandenilinio ryšio (H-ryšio) kompleksų su rūgštimis (acto (AA), cianacto (CyA), propiolo (PA), vandenilio chloridu (HCl), metansulfonrūgštimi (MSA)) bei kai kurių panašių molekulių su vidiniais H-ryšiais (4-pakeisti pikolino rūgšties N-oksidai) ¹⁷O BMR spektrai acetonitrilo (ACN) tirpaluose. Siekiant nustatyti protono lokalizaciją ir H-ryšių geometriją bei terpės reakcijos lauko įtaką išmatuotiesiems ¹⁷O cheminiams poslinkiams, pasitelkus kvantinės mechanikos tankio funkcionalo teoriją (DFT), buvo visiškai optimizuota H-ryšio kompleksų geometrija ir apskaičiuoti deguonies magnetinio ekranavimo tenzoriai. Skaičiavimai atlikti taikant modifikuotąjį hibridinį Perdew, Burke ir Ernzerhof funkcionalą (PBE1PBE) kartu su 6-311++G** bazinių funkcijų rinkiniu bei GIAO atominių orbitalių artinį. Dielektrinės terpės (ACN) reakcijos laukas buvo įskaitytas pritaikant poliarizuojamojo kontinuumo modelį ir naudojant integralinių lygčių formalizmą (IEFPCM). PyO kompleksams su TCA ir TFA buvo nustatyta po dvi energetiškai stabilias struktūras, kurios priklauso H-ryšiams su protono pernaša (PP) ir be jos. Atstumas, skiriantis šių struktūrų potencinės energijos duobes, yra apie 0,2 Å. Eksperimentiškai išmatuoti ¹⁷O BMR spektrai leidžia teigti, kad ACN tirpale PyO-TCA kompleksas gali būti apibūdintas kaip H-ryšys su pradine PP. Tai esmingai skiriasi nuo jų H-ryšio kristale, kur iš neutronų difrakcijos duomenų galima matyti, kad PP vyksta. PyO-TFA komplekse dėl termiškai sukeltų protono dalybų protonas yra lokalizuotas ties H-ryšio centru (O...H...O). ¹⁷O BMR duomenys, gautieji molekulėms su vidiniais H-ryšiais (4-pakeisti pikolino rūgšties N-oksidai), skiriasi nuo bendrosios schemos, nustatytos PyO-rūgšties kompleksams. Darbe aptartos galimos to priežastys, tarp kurių – cheminių pakaitų ir H-ryšio efektų persiklojimas, tirpinamųjų medžiagų nanokristalitai, kurie išlieka tirpale dėl blogo tirpumo ir kt.