The reductive conversion of CO2 into industrial products (e.g., oxalic acid, formic acid, methanol) can occur via aqueous CO2 as a transient intermediate. While the formation, structure, and reaction pathways of this radical anion have been modelled for decades using various spectroscopic and theoretical approaches, we present here, for the first time, a vibrational spectroscopic investigation in liquid water, using pulse radiolysis time-resolved resonance Raman spectroscopy for its preparation and observation. Excitation of the radical in resonance with its 235 nm absorption displays a transient Raman band at 1298 cm−1, attributed to the symmetric CO stretch, which is at ∼45 cm−1 higher frequency than in inert matrices. Isotopic substitution at C (13CO2) shifts the frequency downwards by 22 cm−1, which confirms its origin and the assignment. A Raman band of moderate intensity compared to the stronger 1298 cm−1 band also appears at 742 cm−1 and is assignable to the OCO bending mode. A reasonable resonance enhancement of this mode is possible only in a bent CO2(C2v/Cs) geometry. These resonance Raman features suggest a strong solute-solvent interaction, the water molecules acting as constituents of the radical structure, rather than exerting a minor solvent perturbation. However, there is no evidence of the non-equivalence (Cs) of the two CO bonds. A surprising resonance Raman feature is the lack of overtones of the symmetric CO stretch, which we interpret due to the detachment of the electron from the CO2 moiety towards the solvation shell. Electron detachment occurs at the energies of 0.28 ± 0.03 eV or higher with respect to the zero point energy of the ground electronic state. The issue of acid-base equilibrium of the radical, which has been in contention for decades, as reflected in a wide variation in the reported pKa (−0.2 to 3.9), has been resolved. A value of 3.4 ± 0.2 measured in this work is consistent with the vibrational properties, bond structure, and charge distribution in aqueous CO2.

An in-depth understanding of the chemical steps in the reduction of carbon dioxide (CO2) is of immense importance in biological, environmental, and industrial processes.1 It is a linear molecule with negative adiabatic electron affinity (AEA).2 Therefore, its reduction involving one electron addition requires a change in molecular geometry, which can be induced in an appropriate chemical environment.3 

The molecular mechanism of the CO2 reduction in water has been widely investigated in recent years using gas phase clusters as models.4–7 A variety of experimental data and theoretical calculations have become available on electron attachment to CO2–(H2O)n clusters. However, a recent molecular dynamics simulation has emphasized that the reaction of electron with CO2 in an aqueous environment can display a dynamic behavior quite different from that in clusters in the femtosecond (∼10−15 s) time domain.8 Thus, the importance of spectroscopic information on the radical anion of CO2 in liquid water can hardly be overemphasized. Infrared spectroscopy has been applied to explore the structure of CO2 in low temperature inert matrices.9 However, there has been no vibrational spectroscopic investigation of this important reaction intermediate in liquid water. Unlike gas-phase or inert matrices, the water interaction with a solute anion can be strong and can profoundly affect its bond properties and chemistry. The radical is fairly reactive and cannot be prepared in sufficient concentrations for its steady state observation in room temperature water. We present, here, a time-resolved resonance Raman study of the carbon dioxide radical anion in water (hereafter aqueous CO2) for the first time. The hydration effect on the radical anion properties and a rare experimental signature of short-time dynamics of electron detachment are discussed.

The vibrational spectroscopy of molecular ions embedded in gas-phase water clusters often probes the water molecules only.5 The information thus obtained is used in modelling the nature and interactions of the ion with the bulk water. It has been shown that the CO2 ⋅ (H2O)6 cluster excited by a photon energy of >1500 cm−1 can lead to electron transfer and formation of CO2 ⋅ (H2O)n, the value of n being variable.7 The intermediate steps involve reorganization of water molecules in a few picoseconds to impart a small amount of excess electron charge to linear CO2. The subsequent molecular bending and stabilization of charge on CO2 occurs in less than a picosecond.7 In contrast, a recent ab initio molecular dynamics study suggests that the excitation of bending vibration of CO2 in itself is sufficient to induce electron addition to CO2–H2O clusters, which happens in a few tens of femtoseconds, due to the enhanced attraction of electron to a bent CO2 molecule.8 While the time interval from electron addition to its stabilization on CO2 in clusters varies between 40 and 80 fs, it is predicted to be faster in aqueous solution and to occur in about 20 fs.8 

It is well known that aqueous CO2 can be conveniently prepared by radiation-chemical methods.10 Hydrated electrons produced upon radiolysis of oxygen-free water can readily add to dissolved CO2 to yield a transient absorption at 235 nm.10 An identical absorption is also seen on reaction of the OH radical, formed upon water radiolysis, with carbon monoxide (CO + OH → HOCO → CO2 + H+). The same absorbing species appears upon H atom abstraction from formic acid/formate (HCO2/HCO2H → CO2) by the OH radical and H atom (a minor product of water radiolysis). Thus, the identification of the 235 nm absorption with the aqueous CO2 is well established.10–13 Therefore, Raman scattering studies were performed in resonance with this electronic absorption. Oxidation of formic acid was used for the preparation of CO2 in water. The oxidative preparation of aqueous CO2 for resonance Raman investigation allows us to look for a signature of femtosecond molecular dynamics of electron detachment, thus complementing the dynamics of electron attachment to CO2 in liquid water on a comparable time scale.

The time-resolved resonance Raman observation of aqueous CO2 has allowed us to follow the pH dependence of the radical population in acidic solutions without interference from other species affecting the measured value. The protonation behavior (CO2 + H+ ↔ HOCO), monitored by the well-resolved structure-specific Raman bands, has settled the controversy persisting for almost four decades about the value of the pKa, an important thermo-chemical parameter.

Sodium formate (HCOONa; ACS reagent, ≥99.0%) as well as sodium formate-13C (99 at. % 13C) was purchased from Sigma Aldrich and used without further purification. Aqueous solutions of formate salts (5-10 mM) were made with water purified in a Millipore Milli Q system. The pH of solutions was adjusted either by addition of sodium hydroxide to formic acid or sodium formate. pH measurements were made with an Orion 811 pH meter calibrated with Fisher reference buffers. All solutions were saturated with nitrous oxide (N2O) in order to convert hydrated electrons into OH radicals in less than 50 ns. The concentration of N2O at room temperature was assumed to be about 25 mM.

Pulse and probe transient resonance Raman experiments were performed using a 2 MeV Van de Graaff electron accelerator providing 300 ns electron pulses with a 7.5 Hz repetition rate as an ionization source (pulse source). The existing experimental setup14 was modified for the Raman excitation in the ultraviolet region. In particular, a dye laser (Scanmate, Lambda Physik) tuned to 470 nm, pumped with a 308 nm excimer laser (Compex 102, Lambda Physik), was frequency doubled and used as a probe source. A pair of dielectric mirrors and a focusing lens were used in the laser delivery line. A typical laser energy used during the experiment was ∼0.4 mJ/pulse. The front portion of a submillimeter liquid jet of aqueous solutions was used as a sampling volume. To sustain a stable jet, two ISCO 500D syringe pumps in a continuous flow mode were employed. The position of the jet was adjusted to achieve the best overlap of the focused electron beam with the focused laser ray crossing the jet. Backscattered light was collimated by a lens adjusted to collect the Raman signal from the volume of the jet in which the laser and electron beam were crossing. In order to avoid X-rays (bremsstrahlung) registering on the intensified charge-couple device (ICCD) camera, the detection system was placed outside of the accelerator vault. A collimated beam of scattered light was directed by three flat Al/MgF2 mirrors outside of this vault through a hole in the wall and focused onto a 850 mm spectrograph (SPEX 1402) equipped with a 2400 lines/mm grating blazed at 240 nm (Richardson Gratings). A gated GEN2 ICCD camera (Unigen, PI-MAX, Princeton Instruments) was used as the detector. The camera gate (80 ns) and laser pulse were set to coincide and occur following the electron pulse in order to observe the generated transient species at their highest concentration accessible but avoid unwanted Cherenkov radiation, which would saturate the detector. Raman frequencies were determined from the dispersion of our instrument by reference to measurements of ethanol and acetonitrile Raman shifts. Instrumental dispersion was better than 2.5 cm−1 per optical channel and the centers of the bands were measurable with a probable error of ±2.5 cm−l.

The radiation-chemical methods of CO2 preparation in aqueous solution are well documented in the literature. In brief, the electron pulse irradiation of oxygen-free water produces transient radicals comprising of hydrated electron (eaq ∼ 45%), hydrogen atom (H ∼ 10%), and hydroxyl radical (OH ∼ 45%) on the 100 ns time scale.15 In N2O saturated (∼25 mM) solutions, eaq converts into OH radicals in <50 ns (eaq + N2O → OH + OH + N2), so that the reaction is complete on termination of the 300 ns electron pulse.15 In our experimental setup, the dose of absorbed radiation is enough to produce an OH radical concentration of ∼10−4 M. In moderately basic solutions (pH <11), the OH radical reacts with formate anion by H atom abstraction from HCOO to produce CO2 (OH + HCO2→ CO2 + H2O⋅k ∼ 4.3 × 109 M−1 s−1).16 The H atom in water reacts by a similar mechanism, but at a slower rate (k ∼ 2.1 × 108 M−1 s−1).17 The aqueous CO2 prepared by these radical chemical methods exhibits an absorption peak (λmax) at 235 nm. The extinction coefficient (ε235nm) has been estimated as ∼2200 M−1 cm−1.12,18 The radical reacts by a second order process, at a rate constant of ∼7 × 108 M−1 s−1.12 Therefore, the decay is negligible in measurements made within 1 μs after the electron pulse.

The raw Raman spectra (700-2600 cm−1) obtained under different chemical conditions, probed at 235 nm, and collected before (1A, 1B, 1C) and 100 ns after (1D) the electron pulse are depicted in Figure 1. Each spectrum is an average of 72 000 pulse sequences acquired at 7.5 Hz. 1A is a spectrum recorded for Ar-saturated water. The broad spectral features seen in the figure are due to liquid water. The spectrum in 1B was obtained on replacing Ar by N2O to convert hydrated electrons into OH in the solution. A band at 1280 cm−1 appears that can be readily attributed to aqueous N2O.19 The spectrum in 1C was obtained on dissolving 10 mM sodium formate in the solution of 1B at pH 10. An additional band beyond that in 1B, due to formate anion, appears at 1353 cm−1. The Raman spectrum obtained on irradiation of the 1C solution by ∼300 ns electron pulses and spectral data collected 100 ns after each pulse is furnished in Figure 1(D). The new Raman features present in Figure 1(D) are attributable to aqueous CO2.

FIG. 1.

Raman spectra of aqueous solutions probed at 235 nm: A—Ar saturated water, B—N2O saturated water, C—10 mM solution of sodium formate in N2O saturated water, D—like C but recorded 100 ns after 300 ns long electron pulse. Spectra A, B, C where vertically shifted for clarity.

FIG. 1.

Raman spectra of aqueous solutions probed at 235 nm: A—Ar saturated water, B—N2O saturated water, C—10 mM solution of sodium formate in N2O saturated water, D—like C but recorded 100 ns after 300 ns long electron pulse. Spectra A, B, C where vertically shifted for clarity.

Close modal

The resonance Raman spectrum of CO2 resulting from subtraction of the identical number of averaged Raman scattering signals collected before and 100 ns after the 300 ns electron pulse is presented in Figure 2. The spectrum consists of two principal vibrations: a weak vibration at 742 cm−1 and a relatively stronger vibration at 1298 cm−1. Additionally, one can notice a negative peak at 1353 cm−1, resulting from over subtraction of the 1353 cm−1 formate band, done in order to eliminate broad and intense water Raman signals and very strong stray background from the processed spectrum in Figure 2. The 1298 cm−1 and 742 cm−1 vibrational frequencies combine to produce a band at 2029 cm−1. This combination band confirms that both vibrations belong to the same chemical species. Additional evidence for the origin of the 1298 cm−1 band as the transient oxidation product of formate anion comes from the OH oxidation of isotopic H13COO. The resonance Raman spectrum of 13CO2 is presented in Figure 3. The Raman bands of the parent H13COO and its transient oxidation product 13CO2 both shift downwards in frequency. The shift measured for the 1298 cm−1 band in 13CO2 was 22 cm−1.9 

FIG. 2.

Resonance Raman spectrum of CO2 radical anion probed at 235 nm in water.

FIG. 2.

Resonance Raman spectrum of CO2 radical anion probed at 235 nm in water.

Close modal
FIG. 3.

Resonance Raman spectrum of 13CO2 radical anion probed at 235 nm obtained after 300 ns electron pulse in N2O saturated 5 mM H13COO solution at pH 10, with small amount of oxygen impurity present (the bands at 1147 and 2266 cm−1 are due to O2).26 

FIG. 3.

Resonance Raman spectrum of 13CO2 radical anion probed at 235 nm obtained after 300 ns electron pulse in N2O saturated 5 mM H13COO solution at pH 10, with small amount of oxygen impurity present (the bands at 1147 and 2266 cm−1 are due to O2).26 

Close modal

Several calculations on CO2 ⋅ (H2O)n species are available in the literature that can be used to derive useful information on the bond lengths and bond angles based on the observed Raman frequencies. Here, we consider the density functional theory (DFT) calculations by Beyer, Bondybey, and co-workers at the B3LYP/6-311++G(3df,3pd) level of theory that was applied by them for interpreting experimental data on gas-phase reactions of hydrated electron with CO2.6 It should be mentioned, however, that the calculated vibrational frequencies furnished by them are harmonic frequencies which we compare with experimental frequencies that are anharmonic. When that difference is taken into account, the calculated symmetric CO stretching mode in CO2 ⋅ (H2O)5, 1298 cm−1, and 1294 cm−1 in CO2 ⋅ (H2O)4 both provide an excellent agreement with the observed Raman frequency of 1298 cm−1 in water. Since the free energy difference (ΔG) predicted for the reaction CO2 ⋅ (H2O)5 → CO2 ⋅ (H2O)4 + H2O at room temperature is negative, the hydration shell having four water molecules would be more stable in the liquid state.6 The OCO bending mode in both clusters is predicted at 744 cm−1, which compares well with our experimental Raman frequency of 742 cm−1 in liquid water. The calculation predicts these frequencies for bare CO2 as 1218 cm−1 and 679 cm−1.6 Therefore, an upward shift of about 76-80 cm−1 in symmetric CO stretch and that of ∼65 cm−1 in OCO bend is expected on radical hydration. The low temperature inert matrices are assumed to be weakly interacting with solutes. However, a moderately strong interaction of CO2 with its environment is indicated by the infrared frequencies (1254 and 714 cm−1) measured in Ne matrix (shifts of 36 and 30 cm−1) where other ions and a trace of water were also present.9 The calculated frequencies of CO2 ⋅ H2O (1257 and 708 cm−1) suggest that the solvent interaction in matrices is comparable to that of a single water molecule.6,9

The observation of the OCO bending mode in Raman is a firm spectroscopic evidence of the loss of center of symmetry of CO2 on electron attachment in water. Its proximity with that in the formate anion suggests a bent molecular geometry with the OCO angle in the vicinity of ∼130°. The OCO bending vibration is symmetry forbidden in Raman in a linear molecular geometry (D∞h) and the symmetric CO stretch is forbidden in infrared. The aqueous CO2 molecular geometry is, therefore, qualitatively similar to that in matrices and also bare CO2. The very strong water interaction, evidenced by large frequency shifts, should manifest in shorter CO bonds and greater OCO angle, i.e., a structure tending towards that of CO2. In the B3LYP/6-311++G(3df,3pd) molecular geometry of CO2 ⋅ (H2O)4, these parameters are ∼1.24 Å and ∼138°, respectively.6 Because of large frequency shifts from vacuum or matrices, aqueous CO2 is better represented as [CO2 ⋅ (H2O)4], a small fraction of charge being shared by the water molecules.

The solute-solvent interactions can be visualized in terms of static and dynamic solvent effects. The static solvent effect may involve the solvent reaction field on the solute embedded in a dielectric medium and the interactions of solvent atoms in close proximity with the solute atoms. The static solvent effect may alter the molecular geometry and bond properties of the solute, and also the bond properties of the solvent molecules in contact, as reflected in their vibrational spectra.7 In vibrational spectroscopy of clusters, the focus is on the water molecules in the hydration shell. Here, we are concerned with the solute radical anion. The water molecules in the hydration shell are too few in number, compared to the bulk water (∼1:105), and therefore, indistinguishable. The frequency shifts predicted by the B3LYP/6-311++G(3df,3pd) theory may comprise of static as well as dynamic solvent effects. Hydrogen bonding largely contributes towards the static component.

A bent molecular geometry of CO2 would normally belong to C2v point group, with equivalent CO bonds. However, a random distribution of solvent molecules in its proximity would render the sites available for its occupation to have a lower symmetry (e.g., Cs). For an observable distortion of the solute molecular geometry due to site symmetry, the solute-solvent interactions have to be strong, which is the case with the aqueous environment. Irrespective of C2v or Cs molecular geometry, all three normal vibrations of the radical, i.e., symmetric and asymmetric CO stretch and OCO bend would have infrared as well as normal Raman activity, and be, in principle, observable by both spectroscopic methods. Therefore, a discrimination based on solely symmetric argument and its likely spectroscopic repercussion cannot be made. However, Raman scattering in resonance with an allowed electronic transitions displays only totally symmetric modes, which have favorable Raman Franck-Condon factors. Non-totally symmetric vibration can be seen only when resonance is with a weak electronic transition that borrows absorption intensity from a nearby strongly allowed transition by vibronic mixing via excitation of that vibration. The 235 nm electronic transition of aqueous CO2 is moderately strong and there is no close by strong electronic transition from which it can borrow intensity. The asymmetric CO stretching mode is a non-totally symmetric vibration, Franck-Condon forbidden, and cannot be observed in 235 nm resonance Raman, unless the molecular geometry is distorted from C2v. A Cs symmetry implies non-equivalence of the two CO bonds. Therefore, we looked carefully for the presence of any Raman signal in the 1500-1800 cm−1 region (asymmetric CO stretch in inert matrices at ∼1658 cm−1)9 attributable to the CO2 radical, but no such signal could be located. Therefore, we conclude that the radical retains a C2v symmetry even in water, and departure if any is unobservable. The lack of any apparent effect of solvent on the molecular symmetry, but otherwise a large effect on bond properties, are valuable considerations in testing the validity of theoretical models.

Dynamic solvent effects are indicated by interactions that involve motion of the solvent and solute molecules. In the case of a strong solute-solvent bonding, the solvent internal modes may couple with the solute modes. The solvent vibration ∼1208 cm−1 of heavy water (bending mode) may couple with the 1298 cm−1 symmetric vibration of CO2 because of proximity in frequency if the displacement vectors have favorable orientation. Since no frequency shift is observed on isotopic substitution in the solvent (H2O vs D2O), such an interaction must be negligible (Fig. 4). In the case of CO2 ⋅ D2O and CO2 ⋅ (D2O)2, the CO stretching and DOD bending oscillations have nearly perpendicular displacements in the theoretical description, but the solvent motions become more complex with additional water molecules.6 The absence of solvent isotope shifts puts an additional restriction on the hydrogen bonded water network models that can realistically solvate CO2. Symmetrical hydrogen bonding by a single water molecule,6 enclosed in the hydration cage, i.e., [(CO2 ⋅ H2O) ⋅ (H2O)n], is a configuration in which the C2v symmetry distortion is unlikely to be observed.

FIG. 4.

Comparison of the resonance Raman spectra of CO2 in light and heavy water.

FIG. 4.

Comparison of the resonance Raman spectra of CO2 in light and heavy water.

Close modal

The possible electronic transitions in CO2 are discussed in several publications.1 Here, we limit ourselves to the bond properties of the radical in the excited state, as evidenced by its resonance Raman. In C2v CO2, molecular geometry of bare and bent CO2, the ground electronic state acquires 2A1 symmetry and the excited electronic states can be of 2B1 or 2A1 symmetry. Based on earlier works, the 235 nm transition can be ascribed to 2A12A1, the transition moment in the molecular plane.20 The prominence of the 1298 cm−1 symmetric CO stretching mode in the resonance Raman spectrum is a clear indication that the CO bond undergoes a significant change in the resonant excited state. Since the absorbed electronic energy can be accommodated by the molecule only by weakening the CO bonds, we conclude that CO bonds get considerably elongated in the 2A1 state. On the other hand, the bending mode is also fairly strong, suggesting that the bending coordinate also changes. Thus, there is complete change of the molecular geometry in the excited state. If the excited electronic state involved a drastic change in the equilibrium geometry of solvent molecules, the low frequency solute-solvent modes that contribute towards the resonance Raman bandwidths would be H2O/D2O sensitive. Since no such effect was seen, we conclude that there is no evidence of a significant charge-transfer component in the excited electronic state like has been invoked earlier.18 

Resonance Raman scattering being an extremely fast physical process (≪10−16 s), it can provide a snapshot of molecular dynamics at very early times that is not accessible by currently available shortest pulse lasers. We believe, a signature of such a dynamics is present in the resonance Raman spectrum of aqueous CO2 obtained by us.

As discussed earlier, the CO symmetric stretching mode has the most favorable Raman Franck Condon factor, and therefore, is expected to exhibit overtone and combination modes. While we observe a combination frequency of 1298 and 742 cm−1 vibrations at 2029 cm−1, no overtone of the highly enhanced 1298 cm−1 vibration or any other combination band is seen at higher frequencies. The combination band frequency is ∼11 cm−1 lower than the frequency if no significant anharmonicity was involved and is an indication of a shallow potential well or significant distortion close to its bottom and a likely shift in the equilibrium molecular geometry. The absence of overtones of the 1298 cm−1 mode is surprising. A logical explanation would be that the scattered photons access a continuum of states with onset at about 2250 (±150) cm−1 (0.28 ± 0.2 eV). The fundamental vibrations 742 and 1298 and their combination 2029 cm−1 are characteristics of a structurally stable species, even if chemically reactive, while the continuum arises due to fast instability. The continuum of states cannot be ascribed to CO bond dissociation in CO2. A probable mechanism of dissociation that would involve the straightening of the molecular geometry and pulling CO and O apart would require several eV, in view of the strong bonds21 (bond length ∼1.24 Å; bond order ∼1.5).10 Therefore, CO2 bond dissociation as an explanation can be readily ruled out. The dissociation of successive water molecules from the hydration shell is predicted to occur with an average energy of ∼0.5 eV, (∼4000 cm−1). However, no such loss is anticipated in the liquid state. Instead, the exchange of water molecules in the hydration shell with that in the bulk can occur, causing a small fluctuation in the CO stretching frequency. Since the corresponding Raman band is not excessively broad, the exchange of water molecules between the hydration shell and the bulk water cannot account for the absence of overtone. An interesting possibility, consistent with the molecular dynamics simulations, is electron displacement from the CO2 moiety of [CO2 ⋅ (H2O)n], on the time scale of the scattering process (≪10−16 s), when ∼0.28 eV or higher energy is absorbed. The CO2 and the hydration shell motions that follow correspond to a changed electronic and subsequent nuclear configurations, and not to those of structurally stable [CO2 ⋅ (H2O)n].22 If the absorbed energy is totally converted into the kinetic energy of the nuclear motions, the chemical/electronic nature of the species would remain intact, but it would be in a vibrationally excited state, and the overtone of the CO symmetric stretching vibration (1298 cm−1) vibration should be observed. The recent ab initio molecular dynamics simulation predicts a similar snapshot of chemical events on the femtosecond time scale, when the excess electron is in close proximity of bent CO2 before complete attachment. In view of the simulation, the aqueous CO2 as a stable chemical species does not exist at times <20 fs from the moment an electron is added to CO2 ⋅ (H2O)n. Such an unstable state can be accessed by the Raman scattering process, when an appropriate amount of energy is imparted by the scattered photons to the aqueous CO2 and discrete vibrations above that energy become undetectable. Therefore, we consider our observation an experimental confirmation of the predicted dynamics. Implicit in this assertion is the assumption that the intermediate states leading to complete electron attachment to CO2 are identical to the states that would be encountered when electron detachment from CO2 occurs. In other words, the dynamics associated with reverse electron transfer is traced when the radical is instantaneously excited by an energy of 0.28 eV or higher. Its identity as a distinct chemical species is lost in less than 13 fs after excitation to an energy state that is ∼2250 cm−1 higher than the zero point energy ground state.

In order to further clarify the nature of the continuum and the sequence of chemical events leading to the electron ejection from the aqueous CO2 to bulk water, we represent the [CO2(H2O)n] structure as

a[CO2(H2O)n]+b[CO2(H2O)n](ab),

where CO2(H2O)n and CO2(H2O)n are resonance structures for the ground electronic state configuration of the molecular nuclei. The onset of continuum, as described above, corresponds to a b/a ratio different from the ground electronic state and the corresponding nuclear configurations. Such a state is clearly a solute to solvent, non-stationary charge-transfer electronic state that the molecular dynamics simulations model, albeit in a reverse order. In fact, a continuum of states for different values of b/a, terminating into the highest energy state corresponding to b = 1, a = 0, i.e., [CO2(H2O)n], are intermediary steps in forward as well as reverse electron transfer dynamics. The [CO2(H2O)n] configuration represents a complete solute to solvent electron transfer, and its fragmentation into CO2 and the aqueous electron should lead to a linear CO2.7 

A qualitative model of the continuum of states intermediate between the aqueous CO2 ground electronic state and CO2 + eaq is depicted in Figure 5. In this model, the aqueous CO2 ceases to exist as a distinct chemical species when excited at energies >2250 cm−1 (time period ∼13 fs) by the process of Stokes Raman scattering.22 As discussed earlier, [CO2(H2O)4] provides a good description of the vibrational properties and thermodynamical stability of the aqueous CO2. It has been claimed recently that the aqueous electron can be similarly well described by (H2O)4.23 Even the preferential binding of an excess electron with the two H atoms of a single water molecule can produce most of the vibrational frequency shifts observed for the (H2O)n clusters.7 The [CO2(H2O)4] and (H2O)4 species will obviously differ in the organization of the hydrating molecules. The vertical detachment energy (VDE) of an (H2O)6 isomer is reported as 0.48 eV and AEA as 0.12 eV.7 A somewhat similar or slightly lower values can be assumed for the (H2O)4 cluster. The calculated VDE of CO2 ⋅ (H2O)4 is ∼1.93 eV and that of CO2 as ∼0.43 eV.6 Thus, the hydration energy of CO2 in CO2 ⋅ (H2O)4 is about ∼1.5 eV. Since the aqueous CO2 exists as a distinct chemical species up to ∼0.28 eV, roughly 1.65 eV is used in the reorganization of CO2 and water molecules to stabilize the excess electron of which about ∼1.22 eV could be ascribed to water molecules. While the structural and dynamic parameters for clusters used here need not accurately describe the aqueous state, they conform to the qualitative model we have presented based on experimental observations.

FIG. 5.

Visualization of potential energy surfaces in function of OCO bending angle for the systems (CO2)aq/(CO2 + e)aq based on information available in literature.2,8 The continuum of states corresponds to numerous overlapping potential energy surfaces (not shown), with changes in CO bond lengths and OCO bond angle.

FIG. 5.

Visualization of potential energy surfaces in function of OCO bending angle for the systems (CO2)aq/(CO2 + e)aq based on information available in literature.2,8 The continuum of states corresponds to numerous overlapping potential energy surfaces (not shown), with changes in CO bond lengths and OCO bond angle.

Close modal

The acid base equilibrium of aqueous CO2 and its conjugate acid has been studied by several investigators using electronic absorption,10–12 conductivity,11,13 reaction products,13 and electron paramagnetic resonance (EPR)24 and widely different results have been obtained. It has been discussed earlier that the formal electronic charge on the O atom in a molecule determines its bond properties and protonation behavior.25 If a unit negative electronic charge is localized on a single O, the protonation would occur at a pKa of ∼15. On the other hand, if charge is equally distributed at two O, the pKa should be about ∼5. This correlation is based on the assumption that the electronic charge delocalization in an anion contributes towards structural delocalization far more than in its neutral counterpart, which is often negligible in comparison. The correlation works fairy well in oxygen containing molecules, whether closed shell or radical.25 It has been used to infer the correct structure of a radical anion in water when the spectral and structural assignments based on purely spectroscopic arguments greatly differed from the calculation based assignments. The pKa of the formic acid is 3.7, as the charge on the formate anion is almost equally distributed between the two O atoms. One would expect a similar protonation behavior for CO2, if a small charge polarization due to the presence of H atom bonded to C in formate anion is ignored. The pKa values of alpha carbon centered radicals of monocarboxylic acids are very similar to pKa values of their parent monocarboxylic acids.12 Consistent with these arguments, an early measurement on HOCO reported a pKa of 3.9.11 However, the later determinations based on optical absorption, EPR, and product formation reached to drastically different values.12,13,24 To resolve this discrepancy persisting for several decades, we measured the pH dependence of the Raman band intensities of aqueous CO2 as shown in Figure 6. The spectral intensity does not change between pH 5-10, but it rapidly decreases below pH 4. In view of the observed pH dependence and the experimental uncertainty in determining the Raman band intensities, we estimate the pKa of the acid form of CO2 as 3.4 ± 0.2. We assume protonation on the O site, and not C. The radical-radical reaction of CO2 produces oxalic acid (HO2C–CO2H), which implies that the unpaired electron spin is largely localized on C and charge shared by two oxygens, consistent with our assumption. Thus, we believe, a long standing issue has been settled. At this point, it is not clear why drastically different values of pKa were obtained depending on the monitoring methods. One possibility is the degree of interference from the chemical isomer HCO2 in some experiments, which will have a much lower pKa than HOCO. Bimolecular reaction of HCO2 will prohibit oxalic acid formation and favor disproportionation, leading to CO2 and formic acid as products.13 Since we monitored only the spectral features associated with CO2 for pKa determination, our determination should not suffer from the possible contamination of HCO2 at lower pHs.

FIG. 6.

Dependence of relative C–O stretching Raman band (1298 cm−1) peak intensity on pH.

FIG. 6.

Dependence of relative C–O stretching Raman band (1298 cm−1) peak intensity on pH.

Close modal

Complementing the reduction pathway for preparing the radical anion of carbon dioxide (CO2) by electron addition that has been applied for most spectroscopic studies in inert matrices or water clusters, we preferred the oxidation method to obtain the resonance Raman spectra of this transient species in liquid water. In addition to experimental convenience, this approach provides a snapshot of early molecular dynamics in the electron detachment from aqueous CO2 as is manifested in the resonance Raman spectra. Thus, a qualitative comparison is possible with the electron attachment dynamics predicted by the recent ab initio molecular dynamics simulations.

This study shows that water interacts very strongly with the solute CO2, as evidenced by a large frequency increase of ∼80 cm−1 of the CO symmetric stretching mode with reference to the calculated frequency in bare CO2 by the B3LYP/6-311++G(3df,3pd) procedure. This DFT calculation, reported in the literature on [CO2 ⋅ (H2O)4,5], provides an excellent agreement with the aqueous CO2 frequencies observed in this work for the CO stretching and OCO bending modes. The effect of inert matrices (Ar, Ne) on these frequencies is equivalent to that of a single H2O on CO2. In spite of strong solvent interaction, there is no evidence of symmetry distortion beyond C2v, or coupling of the solvent motion with the internal modes of the radical. The excited state equilibrium geometry in 235 nm electronic transition of the radical is considerably displaced along the CO stretching coordinate. However, there is little change in the solute-solvent coordinate, and therefore, the role of solvent in the electronic excitation seems negligible. A surprising feature of the resonance Raman spectra of aqueous CO2 is the absence of overtone and combination bands above 2200 cm−1. At this energy, the photon energy lost to the radical by the scattered photon, at the time scale of the scattering process, is spent in the charge delocalization from the bent CO2 towards the hydration shell, followed by reorganization of the CO2 bond lengths and bond angle, along with that of the solvent in about 13 fs. This scenario resembles, and complements, the electron stabilization on the bent CO2 in water in ∼20 fs predicted by a recent ab initio MD simulation.

Several pKa values were reported in literature, measured by transient conductivity, optical absorption, EPR, and product formation. The value determined by monitoring the Raman spectra of aqueous CO2 in acidic solutions is ∼3.4, closer to the value first measured by the transient conductivity method.

We would like to express our thanks to the supporting staff in NDRL for their help in implementing improvements to the time-resolved Raman setup over the last year. In particular help from our glassblower Kiva Ford and our machinist Joseph Admave is greatly appreciated. Thanks are also due for clarifying discussions with Dr. Ian Carmichael, Dr. David M. Bartels, and Dr. Gordon Hug.

The research described herein was supported by the Division of Chemical Sciences, Geosciences, and Biosciences, Basic Energy Sciences, Office of Science, United States Department of Energy (Grant No. DE-FC02-04ER15533). NDRL-5102 is the contribution number from the Notre Dame Radiation Laboratory.

1.
J. M.
Weber
,
Int. Rev. Phys. Chem.
33
(
4
),
489
(
2014
).
2.
R. N.
Compton
,
P. W.
Reinhardt
, and
C. D.
Cooper
,
J. Chem. Phys.
63
(
9
),
3821
(
1975
);
G. L.
Gutsev
,
R. J.
Bartlett
, and
R. N.
Compton
,
J. Chem. Phys.
108
(
16
),
6756
(
1998
).
3.
C. E.
Klots
and
R. N.
Compton
,
J. Chem. Phys.
67
(
4
),
1779
(
1977
);
C. E.
Klots
and
R. N.
Compton
,
J. Chem. Phys.
69
(
4
),
1636
(
1978
);
A. R.
Rossi
and
K. D.
Jordan
,
J. Chem. Phys.
70
(
9
),
4422
(
1979
).
4.
C. E.
Klots
,
J. Chem. Phys.
71
(
10
),
4172
(
1979
);
S. T.
Arnold
,
R. A.
Morris
,
A. A.
Viggiano
, and
M. A.
Johnson
,
J. Phys. Chem.
100
(
8
),
2900
(
1996
).
5.
L. A.
Posey
,
M. J.
Deluca
,
P. J.
Campagnola
, and
M. A.
Johnson
,
J. Phys. Chem.
93
(
4
),
1178
(
1989
).
6.
O. P.
Balaj
,
C. K.
Siu
,
I.
Balteanu
,
M. K.
Beyer
, and
V. E.
Bondybey
,
Chem. - Eur. J.
10
(
19
),
4822
(
2004
).
7.
K. J.
Breen
,
A. F.
DeBlase
,
T. L.
Guasco
,
V. K.
Voora
,
K. D.
Jordan
,
T.
Nagata
, and
M. A.
Johnson
,
J. Phys. Chem. A
116
(
3
),
903
(
2012
).
8.
P.
Liu
,
J.
Zhao
,
J. X.
Liu
,
M.
Zhang
, and
Y. X.
Bu
,
J. Chem. Phys.
140
(
4
),
044318
(
2014
).
9.
W. E.
Thompson
and
M. E.
Jacox
,
J. Chem. Phys.
111
(
10
),
4487
(
1999
).
10.
P.
Neta
,
M.
Simic
, and
E.
Hayon
,
J. Phys. Chem.
73
(
12
),
4207
(
1969
).
11.
A.
Fojtik
,
G.
Czapski
, and
A.
Henglein
,
J. Phys. Chem.
74
(
17
),
3204
(
1970
).
12.
G. V.
Buxton
and
R. M.
Sellers
,
J. Chem. Soc., Faraday Trans. 1
69
(
3
),
555
(
1973
).
13.
R.
Flyunt
,
M. N.
Schuchmann
, and
C.
von Sonntag
,
Chem. - Eur. J.
7
(
4
),
796
(
2001
).
14.
G. N. R.
Tripathi
,
ACS Symp. Ser.
236
,
171
(
1983
);
I.
Janik
and
G. N. R.
Tripathi
,
J. Chem. Phys.
138
(
4
),
044506
(
2013
).
[PubMed]
15.
G. V.
Buxton
, in
Radiation Chemistry. From Basics to Applications in Material and Life Sciences
, edited by
M.
Mostafavi
,
M.
Spotheim-Maurizot
,
T.
Douki
, and
J.
Belloni
(
EDP Sciences
,
Les Ulis, France
,
2008
).
16.
A. J.
Elliot
,
D. R.
McCracken
,
G. V.
Buxton
, and
N. D.
Wood
,
J. Chem. Soc., Faraday Trans.
86
(
9
),
1539
(
1990
).
17.
G. V.
Buxton
,
C. L.
Greenstock
,
W. P.
Helman
, and
A. B.
Ross
,
J. Phys. Chem. Ref. Data
17
(
2
),
513
(
1988
).
18.
M.
Lin
,
Y.
Katsumura
,
Y.
Muroya
,
H.
He
,
T.
Miyazaki
, and
D.
Hiroishi
,
Radiat. Phys. Chem.
77
(
10-12
),
1208
(
2008
).
19.
R. G.
Dickinson
,
R. T.
Dillon
, and
F.
Rasetti
,
Phys. Rev.
34
(
4
),
0582
(
1929
).
20.
G. W.
Chantry
and
D. H.
Whiffen
,
Mol. Phys.
5
(
2
),
189
(
1962
);
T.
Habteyes
,
L.
Velarde
, and
A.
Sanov
,
J. Chem. Phys.
126
(
15
),
154301
(
2007
).
[PubMed]
21.
L.
Pauling
,
The Nature of the Chemical Bond and the Structure of Molecules and Crystals: An Introduction to Modern Structural Chemistry
, 3rd ed. (
Cornell University Press
,
1960
).
22.

The situation is analogous to the resonance Raman scattering of a vibrationally excited molecule.

23.
A.
Kumar
,
J. A.
Walker
,
D. M.
Bartels
, and
M. D.
Sevilla
,
J. Phys. Chem. A
119
(
34
),
9148
(
2015
).
24.
A. S.
Jeevarajan
,
I.
Carmichael
, and
R. W.
Fessenden
,
J. Phys. Chem.
94
(
4
),
1372
(
1990
).
25.
G. N. R.
Tripathi
,
J. Phys. Chem. A
102
(
13
),
2388
(
1998
).
26.
I.
Janik
and
G. N. R.
Tripathi
,
J. Chem. Phys.
139
(
1
),
014302
(
2013
).