Communication: Vibrational spectroscopy of atmospherically relevant acid cluster anions: Bisulfate versus nitrate core structures

Negative ionic clusters containing sulfuric acid, nitric acid, and water were first measured in the atmosphere in 1978.^1–3 The smallest mixed composition ions are present in significant proportions in the stratosphere, with HSO₄⁻·HNO₃ and HSO₄⁻·(HNO₃)₂ comprising ∼16% and ∼5% of the total observed negative ions at 35 km.² Ions in the atmosphere are proposed precursors to binary and ternary aerosol formation.⁴ These ternary aerosols themselves play an important role, most notably in the formation of polar stratospheric clouds⁵ that catalyze ozone depletion during the polar spring. Despite the extensive mass characterization of the small precursor ions, relatively little is known experimentally about their structure. In particular, the question of charge localization has not been fully addressed. In previous experiments relying on the charge to mass ratio for ion detection,^1–3,6–8 the bisulfate ion (HSO₄⁻) was simply assumed to carry the negative charge due to the greater relative gas phase acidity of sulfuric acid versus nitric acid (Δ_acidG = −1265 kJ/mol vs −1329.7 kJ/mol).⁹ In this paper, we apply infrared multiple photon dissociation (IRMPD) spectroscopy to mass-selected, cryogenically cooled ions in the 550–1850 cm^‑1 spectral range in order to probe the structure of mixed sulfuric and nitric acid deprotonated cluster anions.¹⁰ 

Previous IRMPD spectra have been reported for hydrated clusters of nitrate¹¹ and bisulfate.¹² The bisulfate clusters showed evidence for a continuous elongation of the OH bond with increasing hydration, indicative of the beginnings of water-mediated dissociation, while the nitrate results revealed a high symmetry structure for the NO₃⁻·(H₂O)₃ cluster, similar to the D_3h structure of bare NO₃⁻.¹³ In both cases, rings of water molecules maximized hydrogen bonding, stabilizing the weakly acidic proton for HSO₄⁻ and binding to multiple oxygen sites with partial negative charge for both species. Kinetics experiments reacting H₂SO₄ with NO₃⁻·(HNO₃)_n showed rapid production of ions with the mass of HSO₄⁻·(HNO₃)_n for n = 1–3.^7,8

In this work, IRMPD experiments were carried out on a previously described ring electrode trap/time-of-flight (TOF) mass spectrometer^14,15 using the “Free Electron Laser for Infrared eXperiments” (FELIX).¹⁶ Ions were produced by electrospray of a solution of 25 mM H₂SO₄ and trace amounts of HNO₃ in a 1:4 water:acetonitrile mix. The negative ions were then transferred into a high vacuum system. Parent ions were mass-selected in a quadrupole mass filter and focused into a ring electrode ion trap. To allow for continuous ion loading and ion thermalization, the trap was continuously filled with He gas at an ion-trap temperature of 15 K. After filling the trap for 99 ms, all ions were extracted from the ion trap, focused both temporally and spatially into the center of the extraction region of an orthogonally mounted linear TOF mass spectrometer, and irradiated with a single FELIX macropulse (50 mJ/pulse and ∼0.25% rms bandwidth). Under these conditions, most of the ions are thermalized prior to IR irradiation.^17,18 The IRMPD spectra were recorded by monitoring all ion intensities simultaneously as the laser wavelength is scanned. The photodissociation cross section σ was determined from the relative abundances of the parent and photofragment ions, I_p(ν) and I_f(ν), and the frequency-dependent laser power P(ν) using σ = −ln[I_p(ν)/(I_p + I_f)]/P(ν).

The IRMPD spectra for three mixed negatively charged clusters and HSO₄⁻ are shown as black traces in Fig. 1. These spectra are constructed from multiple data sets over different regions. The strongest peaks appear in the 1100–1500 cm^‑1 region as in the HSO₄⁻·(H₂O)_n and NO₃⁻·(H₂O)_n spectra,^11–13 and correspond to stretching of the S=O and N=O bonds. Some lower intensity peaks appear at lower frequencies, and an isolated peak appears around 1650 cm^‑1. Bulk optical constants from the 2002 results of Norman et al.¹⁹ for laboratory-generated ternary aerosols (median aerosol size <1 μm) are shown in Fig. 1 for comparison. The bulk curves correspond to the imaginary refractive index, k. The aerosol data are characterized by their weight ratio of H₂SO₄:HNO₃:H₂O: the orange trace corresponds to a higher sulfuric acid concentration, 40:15:45 wt. %, while the blue trace corresponds to a higher nitric acid concentration, 10:45:45 wt. %. All features in the bulk spectra can be assigned to absorption bands of the individual ionic and molecular constituents (HSO₄⁻, HNO₃, and H₂O in this case).^5,19 These contributions are shown as dotted lines in Fig. 1: orange for HSO₄⁻ and blue for HNO₃.

All three clusters show some resemblance to the bulk aerosol results. Spectra 1 and 2 show a strong peak around 1170 cm^‑1, matching the prominent HSO₄⁻ peak at 1195 cm^‑1 in spectrum 4 and the bulk aerosol mode at 1150 cm^‑1. This feature is missing in spectrum 3. The most intense peak in spectrum 3 at 1297 cm⁻¹ lines up well with a bulk HNO₃ mode. Corresponding features are seen in spectra 1 and 2, as well. Another key spectral feature is the peak from the combined N=O stretch and NOH bending mode, occuring at 1672 cm^‑1 in the bulk. This peak is well separated from other transitions and is characteristic of intact HNO₃. It appears in bulk solution as well as in all three cluster spectra.

This initial comparison indicates that all three clusters contain HNO₃. In addition, HSO₄⁻ is present in spectra 1 and 2 but not in 3, implying that NO₃⁻ is the charge carrier for this cluster. Spectrum 1 shows a similar pattern of peaks as the bulk results, while spectrum 2 shows all of the aerosol peaks with the addition of two intense transitions at 1224 and 1338 cm^‑1, an indication that a sum of HSO₄⁻ and HNO₃ contributions is not sufficient to describe spectrum 2, and that multi-centered or H₂SO₄ vibrations are necessary.

In order to test these qualititative inferences of cluster structure, electronic structure calculations were performed using the GAUSSIAN 09 program.²⁰ Initial structures were constructed using a combination of chemical intuition and results from MD conformational searching using Macro-Model 9.9 (Ref. 21) and the OPLS_2005 (Ref. 22) force field. Optimized geometries and harmonic frequencies at the B3LYP/ 6-311+G(d,p) level of theory were used to determine preliminary energetic ordering and vibrational frequencies, and the lowest energy and best-fit structures were further optimized at the B3LYP/6-311++G(2df,2dp) level. Structures are labeled according to their relative energies, including zero-point energy corrections. Except for 1b, only the lowest energy structures are shown in Fig. 1 (many more structures are reported in Figs. S1–S4 of the supplementary material²³). The simulated IR spectra were derived from Gaussian convolutions (FWHM = 15 cm^‑1) of the stick spectra associated with the harmonic vibrational frequencies and linear absorption intensities. We compare these simulations to our multiple photon experimental results with the caveat that the linear absorption intensities can only approximate the true IRMPD intensities. No frequency scaling factor was used, since the NO and SO vibrations require different corrections to agree with experiment.^24,25

Several trends are observed in the simulated structures. First, increasing the number of hydrogen bonds to the negative charge center increases the stability of the cluster. Then, for the same number of hydrogen bonds, lower energy structures tend to have a more symmetric solvation of the negative charge with acid molecules on either sides of the charge carrier. This results in “centrally solvated” motifs for the larger clusters 2 and 3. Finally, families of similar isomers appear at similar energies, showing the same pattern of hydrogen bonds with small differences in geometry such as the orientation of a terminal HNO₃ molecule.

For cluster 1, the presence of an intact HNO₃ molecule is indicated by the experimental peak at 1621 cm^‑1. The presence of HSO₄⁻ is suggested by the peak at 1186 cm^‑1 matching spectrum 4. 1b has the appropriate form, HSO₄⁻·HNO₃. This structure has a single H-bond and a free bisulfate OH. The three dominant peaks in the 1100–1500 cm^‑1 region are simulated at 1164, 1316, and 1478 cm^‑1, and are assigned to the three dominant experimental peaks at 1186, 1310, and 1400 cm^‑1 (see Fig. 1). The peak at 1164 cm^‑1 corresponds to a bisulfate S=O stretching and SOH bending mode, analogous to the molecular HNO₃ mode simulated at 1652 cm^‑1, while the other two peaks are from N=O or combined N=O and S=O stretches. The simulated peaks at 990 cm^‑1 (combined N=O and S=O stretching mode) and at 1098 cm^‑1 (wag of the H-bonded HNO₃ proton) occur between two experimental peaks. These vibrations might correspond to the experimental peaks at 942 and 1031 cm^‑1, respectively, though this assignment does not conform to the bulk peak assignments and does not account for the experimental peak at 1483 cm^‑1. An alternate assignment would require the presence of another isomer to account for the low intensity experimental peaks at 890, 1031, and 1483 cm^‑1. The lowest energy structure for cluster 1 is 1a, which lies 24.3 kJ/mol below 1b, and can account for these missing peaks. 1a has the form NO₃⁻·H₂SO₄ involving two hydrogen bonds to the nitrate ion. We conclude that both 1a and 1b are present in the experimental results. The relative stability of these two structures is of interest, given that the reaction H₂SO₄ + NO₃⁻ → HSO₄⁻ + HNO₃ is exothermic by 63 kJ/mol.⁹ However, the addional hydrogen bond in 1a compared to 1b (see Fig. 1) compensates for the asymptotic energies, stabilizing 1a with respect to 1b.

For cluster 2, the best fit structure 2a features an overall H₂SO₄·HSO₄⁻·HNO₃ geometry with a centrally solvated charge. While a large number of isomers were found including structures with NO₃⁻ charge localization and structures with two hydrogen bonds between HSO₄⁻ and HNO₃, only those structures with the triply hydrogen bound HSO₄⁻·H₂SO₄ motif present in 2a had energies below 5 kJ/mol. These four structures, differing only in the placement of the HNO₃, showed reasonable agreement with experiment, though the assignment to 2a is preferred for energetic reasons.²³ The single HSO₄⁻·HNO₃ hydrogen bond in 2a is similar to that found in 1b, but the OH group on the bisulfate is hydogen-bonded to the adjacent H₂SO₄.

As mentioned above, the IR spectrum of cluster 3 is quite different from the other two. The best-fit and lowest energy structure 3a has a negatively charged NO₃⁻ moiety that is centrally solvated in the cluster. The geometry of 3a includes the NO₃⁻·H₂SO₄ dimer motif with two hydrogen bonds, similar to 1a. The intensities for the peaks at 945 and 1650 cm^‑1 are underestimated in the simulations, but these experimental peaks are also narrower than the others, making them appear stronger. The broadened peak at 1116 cm^‑1 in the experimental results is not well simulated in 3a, with the closest candidate being the H₂SO₄ combined SO stretch and SOH bending mode at 1185 cm^‑1. The presence of a small amount of one of the HSO₄⁻·(HNO₃)₂ isomers 3b or 3d (Fig. S3 of the supplementary material) could explain the 1116 cm^‑1 peak. However, these structures also show an intense HNO₃ peak around 1482 cm^‑1, characteristic of the donor-acceptor HNO₃, which is not apparent in the experiment. For 3, a large number of low-energy isomers were found, including two interesting families with the same number of hydrogen bonds: HNO₃·HSO₄⁻·HNO₃ with a central HSO₄⁻ charge carrier (>2.7 kJ/mol energy range, includes 3b and 3d) and NO₃⁻·H₂SO₄·HNO₃ with a terminal NO₃⁻ charge carrier (>3.4 kJ/mol). The terminal NO₃⁻ family does not agree with experiment as well as 3a, while the central HSO₄⁻ family and the 3a isomer both show some agreement and some discrepancy with experiment. Ultimately, we prefer to assign the major species in spectrum 3 to the centrally solvated nitrate structure 3a because of its low energy and complete ion solvation, whereby every NO₃⁻ oxygen atom receives a hydrogen bond. Cluster 3 highlights the ability of a particularly stable hydrogen-bonded structure to overcome the intrinsic gas phase acidities of the component moieties.

Both of the best fit structures 2a and 3a are also the lowest energy isomers. So why does 1b appear to dominate the spectrum when it is 24 kJ/mol above the global minimum? Previous gas phase IR experiments^26–28 have shown evidence for multiple isomers, and trapping in a local minimum is likely here, too. No other minima are present between 0 and +24 kJ/mol: 1a is the only stable structure with two hydrogen bonds (all of our calculations of the HSO₄⁻·HNO₃ structure with two hydrogen bonds and different starting geometries converged to 1a). While 1b has greater entropy than 1a, this cannot account for the full 24 kJ/mol energy difference, even at room temperature.²³ However, formation of HSO₄⁻·HNO₃ type structures in the precursor droplets is likely since (1) the bulk starting solution contains much HSO₄⁻ and little NO₃⁻ by acid-base equilibrium arguments and (2) statistically, many more distinct random starting geometries are possible with a single hydrogen bond than with two. Prior to trapping, we expect some relaxation of the ions but conclude that the barrier to interconversion of 1b to 1a is sufficient to trap a portion of the isomers in the 1b conformer (barrier ∼2.7 kJ/mol (Ref. 23)). In addition, it is possible that our experiment is less sensitive to stucture 1a than 1b, owing to the requirement that multiple photons be absorbed to observe signal. As this occurs, the cluster heats up. If a vibrational mode of 1a is being excited, one of the hydrogen bonds will eventually break, at which point isomerization to 1b will be very facile, since the proton affinity of NO₃⁻ is greater than that of HSO₄⁻. This isomerization can shut off additional photon absorption if the vibrational resonances for vibrationally hot 1b are different from those for hot 1a.

An underlying assumption in many published studies on atmospheric ions is that deprotonated clusters of sulfuric and nitric acid behave according to the measured thermodynamics of the individual components.^3,6,7,29 Our results challenge the assumption that the HSO₄⁻ anion is always the charge carrier in these small mixed clusters by showing compelling evidence that NO₃⁻ carries the charge in 3: H₂SO₄·NO₃⁻·HNO₃. In the kinetics results of Viggiano et al.,^7,8 the existence of clusters of the type NO₃⁻·H₂SO₄ was not considered based on the thermodynamic arguments above. Our overall findings show that the nitrate ion may indeed be formed within this cluster. These results may motivate further exploration of the structures of these small atmospheric clusters. Clusters 1, 2, and 3 are prime targets for OH stretching^30,31 and double-resonance measurements.³² Such experiments could further characterize the hydrogen bonding networks and separate isomers such as 1a and 1b.

In this paper, we have assigned the charge localization in the deprotonated cluster anions of sulfuric and nitric acid. We find that these small clusters behave quite differently than their individual components in the gas phase or in bulk water, where quantities like gas phase acidity and pH govern interactions. In clusters, these measures of acidity compete with the stabilization energy from multiple H-bonds and from the distribution of negative charge in “centrally solvated” geometries. This is clear for cluster 3 where calculated energetics and spectral assignments show that the negative charge is located on the nitrate moiety rather than the sulfuric acid moiety. Spectrum 2 is assigned to a structure with a solvated HSO₄⁻ ion, as expected from the bare ion thermodynamics and bulk acid-base considerations. For spectrum 1, the situation is more complex: the trapped local minimum structure HSO₄⁻·HNO₃ is observed predominatly together with small amounts of NO₃⁻·H₂SO₄. These results provide enhanced understanding of the ions present in our atmosphere, and call into question previous assumptions about their structures.

We gratefully acknowledge Jia Zhou, Gabriele Santambrogio, Mathias Brümmer, and Ludger Wöste for their aid in the preliminary measurements on HSO₄⁻. We thank the Stichting voor Fundamenteel Onderzoek der Materie (FOM) for granting the required beam time and highly appreciate the skill and assistance of the FELIX staff. This research is funded by the European Community's Seventh Framework Program (FP7/2007-2013, Grant No. 226716) and the (U.S.) Air Force Office of Scientific Research (USAFOSR) (FA9550-09-1-0343). T.I.Y. thanks the National Science and Engineering Research Council of Canada (NSERC) for a postgraduate scholarship. C.H. thanks the German Academic Exchange Service (DAAD) for a postdoctoral scholarship. Calculations were performed at the Molecular Dynamics and Computational Facility at the University of California, Berkeley (National Science Foundation (NSF) Grant No. CHE-0840505).