Water network-mediated, electron-induced proton transfer in [C₅H₅N ⋅ (H₂O)_n]⁻ clusters

As an isolated species, the radical anion of pyridine, Py⁻, exists as an unstable transient negative ion,^1–9 while in aqueous environment it is known¹⁰ to undergo rapid protonation to form the neutral pyridinium radical PyH⁽⁰⁾ along with hydroxide. Consequently, the presence of water plays two essential roles underlying the chemical behavior of Py⁻: one to provide a dielectric medium that can electrostatically prevent the valence anion from undergoing electron autodetachment, and the other to facilitate proton transfer by stabilizing the nascent hydroxide ion. Such a scenario provides an interesting opportunity to explore the detailed manner by which these two processes unfold as water molecules are sequentially added to the ionic ensemble [Py ⋅ (H₂O)_n]⁻. This study is motivated by the earlier cluster work of Desfrancois and coworkers,¹ who made the important observation that the [Py ⋅ (H₂O)_n]⁻ distribution begins promptly at n = 3 when formed using Rydberg electron transfer. Estimates^1,2,4–9,11 of the Py electron affinity range from −0.48 to −0.79 eV, which means that the isolated anion is unstable with respect to electron detachment. Desfrancois and coworkers¹ attributed the onset of the distribution at n = 3 in the hydrated clusters to the greater stabilization of the Py⁻ anion (relative to the neutral) with increasing hydration, as indicated schematically in Fig. 1. This raises the question, however, of whether such a radical ion can be prepared and isolated as a locally stable arrangement in the early hydration regime prior to the proton transfer event. Indeed, the values for the vertical detachment energies of the untagged [Py ⋅ (H₂O)_n≥3]⁻ clusters measured in recent negative ion photoelectron spectra¹² have been interpreted to indicate that both PyH⁽⁰⁾ created by electron-induced proton transfer and reactant hydrated Py⁻ anions were present at n = 3, while the spectra of the larger clusters were most consistent with the PyH ⋅ (OH)⁻ ⋅ (H₂O)_n−1 species.¹² This leaves open the questions of how these conclusions depend on the internal energies of the ions, as well as what internal rearrangements drive formation of the proton-transferred products. In particular, these results point to an interesting scenario where the proton is effectively titrated from water to Py⁻ according to the number and arrangement of water molecules in the spirit of our recent study of network-mediated, intracluster proton transfer in the conversion of nitrosonium NO⁺ to nitrous acid HONO.¹³ Our approach involves an integrated analysis of the photoelectron angular distribution, vibrational predissociation, and photo-induced vibrational autodetachment spectroscopies of the Ar-tagged products from the reaction between neutral Py and the negatively charged water clusters,

and thus extends the recent photoelectron work of Wang et al.¹² We then explore the mechanism by which an excess electron, initially delocalized in a diffuse orbital of the system,^14,15 becomes localized into a valence molecular orbital through the cooperative interplay between the water network and the key proton transfer reaction coordinate: N⋯H⋯OH.

Anionic clusters of H₂O with Py were synthesized in an electron-impact ionized supersonic jet expansion of Ar carrier gas (60 psi) seeded with trace water and Py vapors, which were co-expanded through a single pulsed valve (Parker Hannifin General Valve Series 9, 0.030 in. aperture). A 1 keV electron beam counterpropagating along the axis of the supersonic expansion generated negatively charged cluster ions in the resulting neutral plasma. Conditions were optimized for attachment of weakly bound Ar atom “tags” to the cluster anions, thus enabling vibrational predissociation spectroscopy by photoevaporation of the Ar atoms,

using the Yale double focusing, tandem time-of-flight photofragmentation mass spectrometer described previously.¹⁶ 

The parent [Py ⋅ (H₂O)_n]⁻ ⋅ Ar_p ion and the [Py ⋅ (H₂O)_n]⁻ ⋅ Ar_p−q photofragment were separated by a secondary (reflectron) mass spectrometer. The [Py ⋅ (H₂O)₃]⁻ ⋅ Ar_p cluster also exhibited vibrational autodetachment above 3000 cm⁻¹,

The action spectrum for electron ejection through Eq. (3) was recorded as a function of photon energy by monitoring the electron signal on microchannel plates positioned below the laser interaction region.

All vibrational predissociation and photo-induced vibrational autodetachment spectra between 2250 and 3800 cm⁻¹ were obtained using a LaserVision optical parametric oscillator/amplifier mid-infrared system pumped by a Nd:YAG laser (10 Hz, 8 ns pulse width). Radiation in the 600-2500 cm⁻¹ region was produced by generating the difference frequency between the ∼1.5 and ∼3 μm beams from the OPO/OPA in a AgGaSe₂ crystal. The spectra of the low and high frequency regions (600-2500 cm⁻¹ vs. 2250-3800 cm⁻¹) were combined by normalizing the intensities of transitions that occur in the overlap region from 2250 to 2500 cm⁻¹ that can be reached with both configurations. In addition, the photofragment signal was divided by the laser pulse energy to account for changes over the course of a scan. Experimental spectra represent the summation of 15-20 individual scans.

Photoelectron spectra were acquired by velocity map imaging at an excitation energy of 2.33 eV using an instrument described in detail elsewhere.^17–21 Images were converted to photoelectron spectra using the BASEX transformation.¹⁹ The vibrational structure of the NO⁻ spectrum was used to calibrate the spectrometer.

To aid in assigning the vibrational spectra, the geometries of various possible structures of the [Py ⋅ (H₂O)_n]⁻, n = 2-5 clusters were optimized using M06-2X²² density functional method with the 6-31++G(d,p)^23–25 basis set. This approach was chosen because it has recently been shown to accurately reproduce experimental photodetachment energies of hydrated azabenzene anions.¹² The X-H (X = N,O,C) stretching fundamentals were scaled by 0.980 for this level of theory, while the lower energy, non-stretching modes were not scaled. These calculations were performed using the Gaussian09 program.²⁶ 

In order to gain insight into the rearrangement dynamics, we also carried out BOMD simulations of [Py ⋅ (H₂O)₃]⁻ starting from the structure corresponding to the dipole-bound anion. The simulations were carried out in the NVE ensemble with an excess energy corresponding to T = 270 K, randomized over all atoms. This temperature corresponds to ∼180 cm⁻¹ per degree of freedom, whereas excitation of an OH stretch would deposit ∼120 cm⁻¹ into each intermolecular degree of freedom (assuming validity of the equipartition theorem and that no energy is deposited into the high-frequency intramolecular modes after randomization). Thus, the simulations were carried out at an energy content in excess of that in play under the experimental conditions, which leads to accelerated dynamics. The BOMD simulations were carried out using the CP2K²⁷ code and were conducted using the BLYP density functional method,^28,29 including the D2 long-range interaction correction of Grimme³⁰ and employing the Goedecker-Teter-Hutter pseudopotentials,³¹ aug-DZVP-GTH basis sets,^31,32 and a plane wave cutoff of 280 Ry. The Martyna Tuckerman Poisson solver³³ was employed in the calculation of the electrostatic interactions. A cubic box of 18 Å sides was employed, and a time step of 1.0 fs was used.

The mass spectrum obtained from the ionized free jet, as described above, is displayed in Fig. 2, which shows the distribution of [Py ⋅ (H₂O)_n]⁻ clusters (red) and their complexes with an argon atom (blue). Bare hydrated electron clusters with n = 6 and 7 (colored gray) and their Ar adducts (bronze) are also generated in abundance. The abrupt onset of [Py ⋅ (H₂O)_n]⁻ anions at n = 3 is in accordance with the earlier observations by Desfrancois and co-workers,¹ as discussed in the Introduction. While not our focus in this work, we note in passing that we were able to generate the smaller n = 2 species using Ar-mediated condensation,

in a different source scheme where the Py vapor is introduced with a second pulsed valve and entrained into the electron-beam ionized expansion. That approach was used earlier to form very low adiabatic electron affinity (AEA) anions [e.g., (H₂O)₄⁻] as well as metastable anions with diffuse excess electrons.^34–37 

It is useful to consider the origin of the onset of the cluster distribution at n = 3 in the context of the behavior of related systems. For example, the (H₂O)_n⁻ distribution, at high water partial pressures,^38,39 displays an abrupt onset at n = 15, which has been traced to a condition where the AEA becomes roughly equal to the water binding energy at this cluster size. In that case, the onset occurs because the smaller cluster anions are neutralized due to electron ejection when the condensation energy released upon attachment of a water monomer is larger than the AEA (i.e., a type of associative detachment reaction).⁴⁰ The onset of the [Py ⋅ (H₂O)_n]⁻ cluster distribution at n = 3 therefore indicates that the AEA increases much more rapidly with increasing cluster size than for neat water clusters.

To assess the nature of the excess electron orbital in the [Py ⋅ (H₂O)_n]⁻ clusters, the angular distributions of photoelectrons were obtained by velocity map imaging using an excitation wavelength of 532 nm (2.33 eV). These results for n = 3 are given in Fig. 3, where the raw image and photoelectron energy distributions (after BASEX transformation)¹⁹ are given in traces (a) and (b), respectively. The spectrum is very broad, consistent with that reported by Wang et al.,¹² displaying a full width at half maximum (FWHM) of 0.85 eV, indicating that the vertical electron detachment energy (VDE) is much greater than the AEA. The peak photoelectron yield of the distribution occurs at 1.53 eV, corresponding to the VDE, which is in close agreement with the VDE of 1.56 eV reported by Wang et al.¹² Most importantly, the image in Fig. 3(a) exhibits a significantly negative anisotropy parameter⁴¹ of about −0.5, indicating that photoelectrons are preferentially ejected perpendicular to the direction of the electric field vector (double-headed white arrow in Fig. 3(a)). This behavior is opposite to that displayed by the hydrated electron clusters, (H₂O)_n⁻, for which β values are observed close to +2,^42,43 the limiting value expected for a s → p electron ejection scenario. Furthermore, anionic water clusters in the size range around n = 5 have much lower VDEs (<0.5 eV)^34,44–46 with narrow photoelectron distributions closer resembling that displayed by diffuse, dipole-bound electrons.⁴⁷ These results indicate that the excess electron in the [Py ⋅ (H₂O)₃]⁻ cluster occupies a valence molecular orbital. Photoelectron spectra of the n = 4 and n = 5 clusters are also given in the supplementary material (Fig. S1).⁴⁸ As reported earlier,¹² the binding energy increases with hydration such that the band becomes distorted due to threshold effects^49–51 when the VDE approaches the photon energy.

The calculated [M06-2X/6-31++(d,p)] minimum energy structures for the n = 3-5 clusters identified earlier by Wang et al.¹² are reproduced in Fig. 4. Interestingly, the n = 3 cluster has two nearly isoenergetic isomers: one that exhibits proton-transfer from a water molecule to the pyridine leaving a solvated hydroxide ion (Fig. 4(a)) and another that contains a hydrated Py⁻ ion (Fig. 4(b)). Thus, the n = 3 cluster appears to represent a tipping point in the electron-driven intracluster proton transfer reaction and indeed introduces the interesting possibility that vibrational excitation of the cold cluster can promote large amplitude rearrangements associated with this process.

The broad photoelectron spectra do not allow structural determination of the molecular nature of the species that bear the negative charge. To help clarify this situation, we obtained the vibrational predissociation spectra of the [Py ⋅ (H₂O)_n]⁻ clusters tagged with Ar, with the results presented in Fig. 5. The average number of Ar atoms lost is found to be linearly dependent on the photon energy (see the supplementary material, Fig. S3),⁴⁸ establishing that each tag has a similar binding energy on the order of 750 cm⁻¹/Ar. In the case of the smallest persistent cluster at n = 3, it was only possible to observe fragment anions corresponding to Ar loss below 3300 cm⁻¹, while strong features were observed in the higher energy range for the n = 4 and 5 species. This type of behavior has been observed in (H₂O)_n⁻ ⋅ Ar_m clusters,⁵² for which the nascent anionic fragment, (H₂O)_n^−*, is unstable with respect to electron autodetachment,²⁰ 

unless it is sufficiently cooled by Ar evaporation. We, therefore, also monitored the electron photodetachment channel

with the results displayed in the inset in Fig. 5(a) for the Ar-tagged n = 3 cluster. The “missing” bands are indeed recovered in the electron loss channel, and these transitions can also be observed in the photofragmentation spectrum of [Py ⋅ (H₂O)₃]⁻ tagged with 4 Ar atoms (see the supplementary material, Fig. S2).⁴⁸ This observation of vibrational autodetachment is significant in that it emphasizes the close proximity of the adiabatic electron continuum and the significant rearrangement that is required to access the intersection between the anionic and neutral potential surfaces.

A striking qualitative aspect of the vibrational spectra is that they are extremely broad. For n = 3 and 4, for example, diffuse absorptions span most of the 900-3300 cm⁻¹ range in contrast to the sharp features displayed by the water cluster anions in this size regime.^53,54 This behavior is also unusual for the microhydration of small molecular anions (e.g., CO₂⁻ ⋅ (H₂O)_n)³⁵ of similar size. Even the intramolecular HOH bending mode is obscured in the n = 3 spectrum, and the only features that can be assigned by inspection are the relatively sharp bands near 3700 cm⁻¹ due to the free OH stretching modes and those near 3000 cm⁻¹ arising from the ring CH stretching fundamentals. Interestingly, the transitions in the lower energy range are much better defined in the larger (n = 5) cluster spectrum, with the onset of the diffuse band shifted toward higher energy, beginning at 2400 cm⁻¹.

The calculated harmonic spectra for the minimum energy structures displayed in Fig. 4 are included in Fig. 5 as inverted traces. Perhaps the most important conclusion from this comparison is that the spectra are so dominated by strong anharmonicities that even the most cursory assignments are not obvious. Rather than engage a very difficult exercise in simulating the anharmonic spectra of floppy systems clearly at play here, we find it more useful to empirically evaluate the chemical compositions of the [Py ⋅ (H₂O)_n]⁻ clusters by comparing their vibrational spectra to that of the OH⁻ ⋅ (H₂O)₄ cluster reported earlier,⁵⁵ and included at the top of Fig. 6. Interestingly, a similarly diffuse band is also found in this species, where three water molecules are directly attached to the hydroxide ion and a fourth lies in the second solvation shell as illustrated in the inset structure. This diffuse band has been attributed to excitation of the OH groups engaged in H-bonds to the ion, while the sharp pattern highest in energy arises mostly from non-bonded OH groups or OH groups of the water molecule in the second solvation shell in a double H-bond donor (DD) configuration (ν_α). Note that the intensity of the hydroxide OH fundamental has been found to be dramatically suppressed upon completion of its hydration shell.⁵⁶ The fact that the spectral signatures of all three classes of OH environments (primary shell, secondary shell, and free) in hydrated hydroxide clusters are present in [Py ⋅ (H₂O)_4,5]⁻ strongly suggests that these clusters adopt PyH ⋅ (OH⁻) ⋅ (H₂O)_n−1 arrangements that would be consistent with an intracluster variation of the proton transfer process observed in solution (Eq. (1)).

The reduction in complexity of the n = 5 spectrum in the fingerprint region warrants an effort to identify key well-defined peaks in the context of the hydroxide product, specifically the sharp features labeled ν_β and ν_γ in Fig. 5. This region of the spectrum is expanded in Fig. 7. The bands in question are quite close in frequency to those predicted for the ring CH bending modes of PyH⁽⁰⁾, denoted $ν23Py$ and $ν21Py$ and colored green and orange, respectively, in Fig. 7. Here, the mode numbering is that for isolated PyH⁽⁰⁾ with C_2v symmetry (numbered in order of highest to lowest symmetry and highest to lowest energy).

In light of the difficulties with the harmonic predictions discussed above, it is again useful to compare the observed spectra with empirical data to further assess the viability of the PyH ⋅ (OH⁻) ⋅ (H₂O)_n−1 motif. Figure 7 compares the vibrational predissociation spectrum of [Py ⋅ (H₂O)₅]⁻ to the IR absorption spectrum of the PyH⁽⁰⁾ radical isolated in a matrix of frozen para-hydrogen⁵⁷ and to the gas phase vibrational spectrum of neutral Py.⁵⁸ PyH⁽⁰⁾ in the para-hydrogen matrix was synthesized by Lee et al. by electron bombardment and photolysis of Cl₂/Py/p-H₂. For a simplified comparison to our data, we have simulated their spectrum using the tabulated wavenumbers and intensities that were assigned to the PyH⁽⁰⁾ species (convoluted using a Lorentzian with a FWHM of 4 cm⁻¹). Both ν₂₃ and ν₂₁ of the PyH⁽⁰⁾ species are very close to the observed sharp bands in the [Py ⋅ (H₂O)₅]⁻ vibrational predissociation spectrum (Fig. 7(a)) with similar relative intensities, albeit with a small red shift (14 and 30 cm⁻¹ for ν₂₃ and ν₂₁, respectively) in the solid p-hydrogen matrix. Most importantly, these bands are quite different from the pattern displayed by the neutral Py molecule, which has been obtained from vapor-phase measurements and is included in Fig. 7(c).

The empirical evidence points to a situation where the system explores both Py⁻ ⋅ (H₂O)_n and PyH ⋅ (OH⁻) ⋅ (H₂O)_n−1 components with the latter dominating but with the relative weights of the two terms depending on the degree of solvation of the OH⁻. Qualitatively, as the hydroxide-based clusters accommodate additional water molecules on the emergent network, the configuration with the excess charge on the OH is stabilized, while that with the charge on the Py becomes less important. Thus, in the n = 3 case, the proton is more strongly shared in the N⋯H⋯O motif, giving rise to very broad spectral features typical for shared protons.⁵⁹ At smaller n, excitation of the nominal NH stretch samples the Py⁻ ⋅ (H₂O)_n configuration, in essence driving the proton transfer reaction with concomitant excess charge displacement. Such a scenario has been demonstrated to be at play in the binary complex F⁻ ⋅ HOH,⁶⁰ where excitation of the bridging proton stretch samples a region in the potential surface corresponding to HF⋯OH⁻. When this occurs in a system where the associated water network responds to the change in the charge distribution, it is expected that vibrational excitation will induce substantial heavy particle rearrangements, thus rationalizing the nearly continuous nature of the spectrum for n = 3.

To better quantify the nature of the potential surfaces governing the heavy particles and excess electron, we calculated one-dimensional potential energy curves for proton transfer in [Py ⋅ (H₂O)₃]⁻ and Py ⋅ (H₂O)₃. The potentials were calculated starting from the M06-2X optimized structure shown in Fig. 4(a) with a PyH ⋅ (OH⁻) ⋅ (H₂O)₂ configuration, and then carrying out a series of single-point calculations for the anion and neutral for different values of the NH distance, assuming linear motion between the N atom and the O of the hydroxyl, and keeping all other degrees of freedom frozen. The resulting anion potential energy curve, shown in Fig. 8, is exceedingly anharmonic, such that even the zero-point level samples the Py⁻ ⋅ (H₂O)₃ structure. The vibrational zero-point level of the neutral potential along this slice lies above the v = 1 level of the anion; however, relaxation of the geometry of the neutral causes its zero-point level to fall below the v = 1 level of the anion (as evidenced by the experimentally observed opening of the electron autodetachment channel). We note that the actual energy gap between the proton-transfer and non-proton transfer structures may be somewhat larger than predicted by the M06-2X method. In particular, preliminary results using the restricted open-shell MP2 method^61,62 predict the non-proton transfer structure to be about 477 meV less stable than the proton-transfer structure, rather than being nearly isoenergetic as predicted by the M06-2X method. In either case, however, it is still anticipated that excitation of the v = 1 OH manifold will induce significant large amplitude motions that mediate interconversion between these two structures.

As the charge becomes more stabilized on the OH⁻, the spectra become simpler as excitation of one quanta of OH stretch no longer leads to sampling of the Py⁻ ⋅ (H₂O)_n class of structures. A primitive manifestation of this type of blue-shifting by differential solvation of charge-localized, asymptotic partners (in this case Py⁻ ⋅ H₂O vs. PyH ⋅ OH⁻) has, in fact, been routinely invoked to understand the rare gas tag effects in binary complexes.^63–65 The case of solvation by water can thus be viewed as an extreme example of a generic solvent response by an anionic, proton bound complex.

The preceding discussion emphasized the close energetic proximity of the PyH ⋅ (OH⁻) ⋅ (H₂O)_n−1 and Py⁻ ⋅ (H₂O)_n valence-bond arrangements. There is actually a third low-lying valence-bond structure that corresponds to a dipole-bound Py ⋅ (H₂O)_n⁻ anion with a diffuse excess electron, which is expected to be involved in early stages of (H₂O)_n⁻ + Py condensation. Our calculations indicate that excitation of one quanta into the NH or OH stretch vibrations of the most stable PyH ⋅ (OH⁻) ⋅ (H₂O)₂ species would deposit sufficient energy for the system to sample such dipole-bound structures as exit channel configurations leading to electron autodetachment. To further evaluate the dynamics of the interconversion between the Py ⋅ (H₂O)₃⁻ and PyH ⋅ (OH⁻) ⋅ (H₂O)₂ structures, we performed BOMD simulations on the [Py ⋅ (H₂O)₃]⁻ cluster (see Section II C for details). Ten trajectories, four between 1 and 5.5 ps and six for between 14.7 and 19.3 ps, were run. Of these only two gave rise to the solvated OH⁻ ion, one near 16 ps and the other near 1.6 ps. In these simulations, the Py ⋅ (H₂O)₃⁻ isomer was heated to 270 K and changes to the geometry and electronic structure were recorded in 1 fs increments. An example video of one of these trajectories is given in the supplementary material.⁴⁸ Figure 9 plots the electron binding energy along one of the trajectories together with the differences in the charge densities of the anion and neutral, calculated at the MP2/aug-cc-pVDZ^66,67 level, augmented with a 7s7p set of diffuse functions centered at the middle of the ring for the final, key intermediate, and product structures (MP2-level charge densities are employed rather than BLYP^28,29 charge densities as the MP2 method is expected to give more reliable charge distributions). For the indicated trajectory, the electron remains dipole bound to the (H₂O)₃ water network for about 1.65 ps, before evolving into an intermediate structure of the form Py⁻⋯H⁺⋯OH⁻(H₂O)₂. This is followed by rapid formation of the pyridinium radical PyH ⋅ (OH⁻) ⋅ (H₂O)₂. In the product structure, both an electron and a proton have transferred to the Py, with the former occupying the π^∗ LUMO of the neutral Py molecule. Evolution to the product PyH ⋅ (OH⁻) ⋅ (H₂O)₂ structure requires the two H₂O monomers to adopt an arrangement that effectively solvates the OH⁻. For the indicated trajectory, the system rapidly evolves from the intermediate to the product, although for other trajectories (not shown) there are fluctuations between the intermediate and reactant-like Py ⋅ (H₂O)₃⁻ structures before the product is accessed. As seen in Fig. 9, the electron binding energy increases as the PyH ⋅ (OH⁻) ⋅ (H₂O)₂ product is formed, approaching a VDE similar to that observed from the photoelectron spectrum (1.53 eV).

Gas phase cluster chemistry, together with analyses of the predissociation and photoelectron spectroscopies with electronic structure calculations, has been used to clarify the chemical rearrangements that occur when Py combines with hydrated electron clusters (H₂O)_n⁻ for n = 3-5. The photoelectron spectrum of the resulting [Py ⋅ (H₂O)_n]⁻ species supports the presence of a valence anion, in which the VDE is much greater than its AEA. Analysis of the vibrational patterns then confirms the formation of neutral pyridinium radical PyH⁽⁰⁾ and hydrated hydroxide, where the neutral radical occupies one of the sites in the primary solvation shell around the hydroxide anion: PyH ⋅ (OH⁻) ⋅ (H₂O)_n−1. Synthetic access to this arrangement now presents an attractive opportunity to follow the reaction chemistry of the PyH⁽⁰⁾ radical in the microhydration regime. A particularly timely choice in this regard is the activation of CO₂ by proton coupled electron transfer, which has been proposed for the catalytic activity of Py by Musgrave⁶⁸ and Bocarsly.⁶⁹ 

This work was supported by the Mississippi Center of Supercomputing Research and the National Science Foundation under Grant Nos. CHE-1338056 (G.S.T.) and CHE-0955550 (N.I.H.). Both G.S.T. and N.I.H. acknowledge NSF EPSCoR support under Grant No. EPS-0903787. M.A.J. and K.D.J. want to thank the U.S. Department of Energy for support under Grant Nos. DE-FG02-06ER15800 and DE-FG02-06ER15066. M.A.J. also expresses gratitude to the staff and facilities of the Yale University Faculty of Arts and Sciences High Performance Computing Center, and NSF Grant No. CNS 08-21132, which partially funded acquisition of the facilities. K.D.J. also acknowledges use of computational facilities in the University of Pittsburgh’s Center for Simulation and Modeling.