Hydration motifs of ammonium bisulfate clusters show complex temperature dependence

Atmospheric new particle formation (NPF), the process by which trace atmospheric gases cluster and grow, is an important factor in understanding and modeling climate change.¹ Aerosol–radiation interactions are the largest source of error in climate change models,² and estimates suggest that up to half of all aerosols are formed by NPF.^3,4 While key contributors to NPF, primarily including sulfuric acid, nitrogen-containing bases, organic, and biogenic molecules,^4–21 have been identified, NPF remains among the most difficult aspects of climate change to model accurately.²² This difficulty is further compounded by the unknown role of water, which, while abundant in the atmosphere, is difficult to probe both experimentally and computationally.

While water is thought to participate in NPF, NPF rates have been shown to have a minimal dependence on relative humidity that is highly dependent upon cluster composition.^23,24 Structural information on NPF-relevant clusters is hard to obtain, as the small number density achievable precludes many traditional techniques.^25–27 Furthermore, experimental studies of hydrated NPF-relevant clusters have been hampered by the fact that ambient sampling techniques typically impart sufficient energy to evaporate water from the clusters.^28–30 

While sulfuric acid–water complexes have been well studied,^31–35 the explicit inclusion of water in quantum chemical methods probing NPF-relevant clusters with two or more compounds is not yet common, owing largely to the increased complexity water poses.^36–38 Atmospherically relevant clusters are increasingly being studied using algorithmic-based structure search techniques, such as the ABCluster program,^39,40 that use lower cost methods to more quickly explore the potential energy surface (PES) for the cluster configuration.^41,42 Structure searches of NPF-relevant clusters grow exponentially more difficult as the cluster size increases,⁴³ and water can greatly increase the complexity of the configurational space that needs to be sampled.^36–38 Additionally, the potential energy surface for water-including NPF-relevant clusters is quite shallow.³⁸ Currently, most sophisticated methods rely on sampling water at evenly distributed points across the entire cluster.^36–38 A more complete understanding of the structural role of water may allow for ways to tune structure search algorithms to scale more efficiently or at least limit the number of positions that need to be sampled.

Recent photoelectron and vibrational spectroscopic studies have begun to probe the binding motifs of water in small NPF-relevant clusters, showing that water acts as a hydrogen bond acceptor when interacting with cations and as a hydrogen bond donor when interacting with anions.^{13,14,44–47} Recent mass spectrometric studies of NPF-relevant clusters from our group show that the hydration of clusters correlates well with the number of surface hydrogen-bond donors,⁴⁸ and vibrational spectroscopic results on a prototypical ammonium bisulfate cluster show that the first several water adducts integrate into the cluster in a bridging motif, accepting hydrogen bonds from free NH sites and donating hydrogen bonds to nearby bisulfate oxygens.⁴⁹ Computational structural surveys of sulfuric acid–water clusters have found a similar motif, finding that the lowest energy structures typically contain water in a bridging configuration, where the water accepts a hydrogen bond from an SOH group and donates a hydrogen bond to an SO group.³⁸ 

In this work, we synthesize ammonium and alkylaminium bisulfate clusters, hydrate them using a multi-ion trap approach,⁵⁰ and analyze their structures using cryogenic ion vibrational predissociation (CIVP) spectroscopy and infrared multiphoton dissociation (IRMPD) spectroscopy. We then identify types of binding motifs that water participates in and how those binding motifs change as a function of cluster size and composition, allowing us to correlate several unique motifs with spectral markers of different water configurations.

All clusters studied herein were generated via an atmospherically isolated electrospray ionization (ESI) source described previously.²⁶ Ammonium bisulfate clusters were generated through ESI of ∼1 mM ammonium sulfate salt dissolved in 25/75 water/acetonitrile into a dry, ∼1 bar N₂ atmosphere. Clusters containing alkylamines were obtained by the addition of small amounts of alkylamine directly into the electrospray solution. These methods generate cluster distributions that are similar to those seen in chemical ionization mass spectrometry studies using the CLOUD chamber at CERN.¹⁰ 

All experiments were conducted on a home-built guided ion beam/tandem-ion trap/tandem time-of-flight (TOF) photofragmentation mass spectrometer described previously.²⁵ Briefly, ions generated from the ESI source are introduced into the vacuum system and are accumulated in an octopole ion trap before being guided through a quadrupole mass filter into a liquid nitrogen-cooled octopole ion trap similar to that described by Marsh et al.⁵⁰ To generate hydrated clusters, this trap is held between 150 K and 170 K and water vapor is introduced with helium buffer gas through a pulsed valve. A fraction of the hydrated clusters are then extracted at 10 Hz into a cryogenic trap held between 10 K and 15 K to facilitate the adsorption of D₂ molecules to the surface of the cluster. Ions are ejected into an orthogonal acceleration tandem time-of-flight mass spectrometer consisting of a linear stage and a reflectron stage. At the focus of the linear stage, the ions of interest are intersected by a pulse from an Nd:YAG pumped infrared optical parametric oscillator/optical parametric amplifier (OPO/OPA) laser system (LaserVision). Fragments are separated from undetached parents and guided into a microchannel plate detector that generates a signal, which is then amplified and digitized for analysis. The signal is integrated for both the “parent” ion of interest and the “fragment” ion, which is generated by interaction with the laser pulse. Laser power is monitored by directing a back reflection from the KBr window that couples the laser pulse in the vacuum system to a pyroelectric energy meter. The integrated peak intensities are recorded as a function of laser wavelength, and a linear action spectrum is created by dividing the integrated intensity of the fragment by the sum of the integrated intensities of the parent and the fragment, and a laser power correction is applied by dividing by the laser pulse energy.

Variable-temperature infrared multiphoton dissociation (VT-IRMPD) spectra were carried out in a similar fashion. In these spectra, the fragment is generated by ejection of a water molecule, using higher laser power. Hydrated clusters are generated in the liquid nitrogen-cooled octopole ion trap and extracted into the cryogenic ion trap as described above. Spectra are then recorded with the cryogenic ion trap held at various temperatures.

Hole burning spectra of these clusters were obtained by introducing a second laser, fixing it to the frequency of a vibrational signature that is likely unique to one or several isomers, and intersecting the ions in the cryogenic ion trap. This evaporates the tags off of only the isomers containing that spectral feature, and the experiment can proceed as originally explained, with the remaining tagged isomer being interrogated by the original laser after a time-of-flight mass separation. As it was not possible to entirely deplete either of the populations, a linear combination methodology that has been used previously was employed to obtain single-isomer spectra.⁵¹ 

All calculations were carried out at either the CAM-B3LYP/aug-cc-pVDZ (for clusters containing at least four bases and three acid moieties) or the CAM-B3LYP/aug-cc-pVTZ (for clusters containing less than four bases and three acid moieties) level of theory using the Gaussian 16 suite of programs.⁵² Both these levels of theory have been used previously and have been shown to produce both computed vibrational spectra with good agreement to experimental data and energetic orderings of isomers that match experimental results,^27,53 and several benchmark studies demonstrate the validity of these choices.^54,55 A scaling factor of 0.955 was applied to the frequency of vibrational modes above 3200 cm⁻¹ for all hydrated clusters to achieve a better match between computed results and experimental data solely for ease of viewing. Computed vibrational spectra are presented convolved with 5 cm⁻¹ Gaussian functions to reproduce experimental linewidths.

Cluster notation is given by the formula (m, n) = (NH₄⁺)_m(HSO₄⁻)_n for clusters containing ammonia and sulfuric acid and by (k, m, n)_l = $(alkylaminium+)k$(NH₄⁺)_m(HSO₄⁻)_n, where l = methylaminium (MA), dimethylaminium (DMA), or trimethylaminium (TMA) for clusters containing alkylamines.

The (3,2) cluster provides a useful initial framework for understanding hydration sites in ammonium bisulfate clusters, as it possesses most of the structural features of larger particles while dry but its smaller size allows it to be experimentally and computationally accessible. Its spectrum, which has been previously assigned, features bands that are useful for tracking structural changes as water uptake occurs. Previous mass spectrometric work probing the hydration of the (3,2) complex suggested that the key structural feature governing water uptake is free NH sites.⁴⁸ Additional spectroscopic work probing the structures of hydrated (3,2) complexes further reinforced the importance of free NH sites and explored which free NH sites tend to be populated.⁴⁹ While NH is an attractive hydrogen bond donor, the site must be able to accommodate water in a way that allows the water oxygen to act as a hydrogen bond acceptor from an NH and a water hydrogen to act as a hydrogen bond donor to a bisulfate oxygen. This binding motif is similar to a common binding motif found in several computational configurational sampling studies^38,56,57 and has been found experimentally in clusters of similar anions and monotonic cations.^58–60 Here, we present the spectrum of the (3,2)·(H₂O) cluster across a greater spectral range (600 cm⁻¹–4000 cm⁻¹ as compared to the original experiments 3300 cm⁻¹–3800 cm⁻¹), with a less perturbative tag (D₂ as compared to the original work’s N₂ tagged spectra), and compare it to the CIVP spectrum of the dry cluster, as shown in Fig. 1. We confirm the original structural assignment and note that the fingerprint region suffers only minor perturbations from the bound water. The fingerprint region (600 cm⁻¹–1800 cm⁻¹) is well-matched to the computed harmonic spectrum from the previously proposed structure, as shown in Fig. S1. The free and bound water OH stretches serve as useful spectral markers associated with the water that are sensitive to the exact binding environment of the water. Changes to these spectral markers as a function of cluster composition are likely to be indicative of different water binding motifs, providing a point of comparison to facilitate exploration of other hydrated ammonium sulfate clusters. With a structural assignment for the (3,2)·(H₂O) cluster in hand, we seek to see if this framework for hydration persists in clusters of different sizes. Of particular interest is whether this binding framework will be observed in the smallest cationic cluster, with its smaller size and limited flexibility.

Figure 2 presents the spectrum of (2,1)·(H₂O) compared to the dry spectrum, as well as the calculated spectrum of the lowest energy isomer found for (2,1)·(H₂O) and the calculated spectrum of the dry cluster for our previously reported structure.²⁷ We find that, such as the (3,2) cluster, the spectral features are largely preserved upon water binding, with the majority of the spectral features maintained or only lightly perturbed. However, as compared to the (3,2) cluster spectral changes, we find a clear change in the relative intensity pattern in the bisulfate region (between about 1000 cm⁻¹ and 1400 cm⁻¹) of the (2,1) hydrated spectrum compared to the unhydrated spectrum. This region, in particular, the SO₃ asymmetric stretches, has been shown previously to be sensitive to the exact hydrogen-bonding arrangement of the cluster,²⁷ suggesting that the bound water interferes with the pre-existing hydrogen bond network in the (2,1) cluster more directly than that in the (3,2) cluster. The lowest energy structure found, with a computed harmonic spectrum in relatively strong agreement with the experimental data, features a water binding motif in which the water does not bind to a free NH but rather inserts into a weak (∼2.2 Å in length) pre-existing hydrogen bond, as shown in Fig. 2. This suggests that while a strong hydrogen bond donor on the cluster is important for binding water, it is also important for the water to be able to act as a good hydrogen bond donor itself, donating back to an oxygen on a nearby bisulfate. For the (3,2) cluster, several free NH sites are oriented such that minimal structural deformation is needed for the water to both accept and donate hydrogen bonds; however, the smaller size of the (2,1) cluster results in the distance and angle between the free NH groups and bisulfate oxygens to be too large such that the site is not feasible for a water molecule to occupy. Instead, we see a weak hydrogen bond broken and replaced with two presumably stronger hydrogen bonds. Additionally, we note that the ammonium-related bands between about 1300 cm⁻¹ and 1600 cm⁻¹ become sharper upon water binding, suggesting that the water “locks” the cluster into a less-flexible framework.

Substitution of ammonia by alkylamines such as methylamine, dimethylamine, and trimethylamine, which are known to increase the growth rates of clusters,^9,10,61 has been shown to influence water uptake by atmospheric particles, particularly by modulating the number of free NH sites.^48,49 It was previously shown that these alkylamine-substituted clusters maintain the structure of the unsubstituted (2,1) cluster. As shown in Fig. S2, where the D₂-tagged CIVP spectra of the fully methylamine- and dimethylamine-substituted (2,0,1)·(H₂O) clusters are compared with the unsubstituted (2,1)·(H₂O), the spectral features and relative intensity patterns are maintained across all clusters, suggesting that the hydration motif is similar to the unsubstituted cluster for all fully substituted clusters. The lowest energy structures found, shown in Figs. S3 and S4, show the water to be in a similar weak NH insertion position for both these clusters, despite the varied amount of free NH sites between the clusters.

Evidence suggests that a new binding motif may emerge in the trimethylamine-substituted cluster. As shown in Fig. S5, we present tag-free IRMPD spectra of these alkylamine-substituted clusters. Of note is the absence of the bisulfate free OH stretching band in the trimethylamine-substituted spectrum, which is present in all the other tag-free spectra, suggesting that the water is located on the free OH of the bisulfate. The unhydrated trimethylamine-substituted cluster is proposed to largely maintain the structure of the (2,1) cluster, but it is only capable of making the two strong hydrogen bonds as each trimethylamine has only one hydrogen bond donor. When the weak NH bonds do not exist, we find that water binds to the free OH of a bisulfate, which is also consistent with previous work on (3,0,2)_DMA·(H₂O). It is of note that when all free NH sites are blocked in (3,0,2)_DMA, there is no evidence of an insertion into an already existing hydrogen bond, suggesting that while insertion into a weak hydrogen bond can be favorable, breaking a strong hydrogen bond seems to be quite unfavorable, and it may be that water only inserts into a structure when it can act as a form of strain-relief, taking a long, non-optimal hydrogen bond and replacing it with two shorter, stronger bonds.

Figure 3 shows the CIVP spectra of the dry (4,3) cluster compared to the (4,3)·(H₂O) cluster. There are two peaks present in the free OH water stretching region, suggesting that a mixture of at least two isomers is present. To establish this, hole burning spectra of these clusters were recorded, as shown in Fig. 4. The hole burning spectra show that these isomers appear to have unique SO₃ asymmetric stretches and free OH/NH stretches, suggesting that the isomer is not simply tag-related, but is in fact due to different positions of water. While the lack of a definitive structure for the (4,3) cluster makes elucidation of the exact structure of (4,3)·(H₂O) currently unfeasible, we can probe the structural motifs to which the water binds.

The spectrum of isomer 1, shown in green in Fig. 4, contains a single free OH water stretch at a lower wavenumber than any of the clusters studied above, suggesting that the position of the water in this cluster is unique. The spectrum also lacks a key stretch at around 1350 cm⁻¹, denoted with an asterisk, present in isomer 2 and shown previously to be indicative of a direct bisulfate–bisulfate hydrogen bond in the cluster.²⁷ This feature is a new structural motif that emerges at the (4,3) cluster size and persists through larger clusters, so it seems likely that if a new hydration motif is to emerge at this cluster size, it may be related to this feature.

The spectrum of isomer 2, shown in red in Fig. 4, contains a single free water OH stretch that is similar in wavenumber to the free water OH stretch of the hydrated (2,1) and (3,2) clusters, and the bisulfate–bisulfate stretch at ∼1350 cm⁻¹ is intact, suggesting that this isomer more closely resembles the NH-bound hydration motifs explored for the smaller clusters.

A computational structural survey finds that there are many structures within 0.5 kcal of being the lowest energy structure, as shown in Fig. 5. We find that these lowest energy structures all contain either NH hydration or bisulfate–bisulfate insertion motifs. We also find that the computed harmonic spectra, shown in Fig. 6, predict shifts in the free OH water stretching region that are diagnostic of the water binding arrangement. It is predicted that water bound to NH sites has a free OH water stretch at a slightly higher wavenumber compared to water inserted into bisulfate–bisulfate sites, as shown by the colored bands in Fig. 6. This shift is likely attributed to the fact that the water in the bisulfate–bisulfate insertion motif acts as a double H-bond acceptor (structures A and D in Fig. 5), whereas it is always a single acceptor in the surveyed NH sites. The bisulfate–bisulfate insertion hydration sites also feature a shift in the bisulfate–bisulfate bend, but the direction of that shift is sensitive to the exact binding orientation, and a general trend about the direction cannot be inferred from the current data. It is worth noting that the nearly degenerate NH hydration sites that show the same free OH_W peak position can fall into either of the motifs we explored in the (2,1)·(H₂O) and (3,2)·(H₂O) clusters, resembling either a bridge formed on a free NH or an insertion into a very weak pre-existing H-bond. While no absolute structures can be definitively assigned, we conclude that the (4,3) cluster forms two isomers upon hydration—one in which the water is bound to a surface NH group and one in which the water inserts into the bisulfate–bisulfate bond.

We have also probed how alkylamine substitution may influence water uptake, which is of particular interest in a cluster that has competing hydration sites of different natures. As shown in Fig. S6, for the (4,0,3)_MA cluster, surprisingly, we find that there is predominantly one free OH water stretching peak. This peak aligns with the free OH water stretching peak around 3650 cm⁻¹ of (4,3)·(H₂O) that we have assigned to a bisulfate–bisulfate insertion hydration motif. The bisulfate–bisulfate OH bend feature that is maintained from (4,3) to (4,0,3)_MA (depicted with asterisks in Fig. S6) is notably absent in the (4,0,3)_MA·(H₂O) spectrum, further suggesting that this hydration site interacts with the bisulfate–bisulfate hydrogen bond feature and that methylation has blocked the NH site.

It is currently unclear why the substitution of methlyamine seems to cause this bisulfate–bisulfate hydration motif to become much more favorable, as there is no evidence that the substitution induces any large structural changes, and it is expected that some free NH sites should still be maintained. In particular, it is unclear if the bisulfate–bisulfate site has become more energetically favorable compared to that same site in the (4,3)·(H₂O) cluster. It is possible that the methylamine substitution results in fewer weak hydrogen bonds than those present in the (4,3) structure, precluding an NH insertion mechanism similar to that of (2,1)·(H₂O). Other possibilities include the selective loss of free NH sites that are particularly amenable to the formation of the acceptor–donor hydration motif of (3,2), leaving only free NH sites in which the hydrogens of the water are unable to form stable bridging hydrogen bonds with bisulfate oxygens. It is also possible that the substitution introduces strain through shorter bond lengths or limited hydrogen bond donors that is alleviated by an insertion of water into the bisulfate–bisulfate bond. Regardless, the similarity between the experimental spectra of the ammonia-containing and methylamine-containing clusters suggests that methylamine substitution of the (4,3) cluster likely induces no large structural arrangements but either significantly changes the relative energetics of water binding sites or completely blocks it. The determination of exact structures for each isomer of these clusters will be necessary to confirm this hypothesis.

The presence of two peaks, each specific to presumably one isomer, can provide a useful window into the relative energetics of the binding sites of the (4,3) cluster. By recording spectra at varying temperatures, we would expect the relative intensities of the peaks to change in accordance with changes in the Boltzmann populations of the different isomers. As shown in Fig. 7, IRMPD spectra were taken at temperatures across the range of 4 K–150 K with the hope of using the relative intensities of the peaks to determine the relative energies of the isomers. Surprisingly, beginning at around 65 K, a much sharper peak emerges slightly to the red of the lower energy free OH_W peak. This new feature emerges and persists only through a narrow temperature range, predominantly between 65 K and 100 K. These findings were confirmed in two runs of increasing and one run of decreasing temperatures. Given that no second OH stretching peak appears in this range, we infer that this feature arises from the free OH stretch of a third isomer with a water molecule donating one hydrogen bond to a bisulfate oxygen. Returning to Figs. 5 and 6, we see that structure D does, indeed, feature an OH stretching peak in this region that is red shifted from that in structure A by 5 cm⁻¹. Notably, there is no difference in connectivity between isomers A and D, and thus, isomer D must be considered a metastable structure with an unknown but presumably low isomerization barrier to isomer A.

We track the temperature evolution of these peaks by fitting the spectra in this region at each temperature with three Gaussians, one centered at 3683 cm⁻¹ with a Gaussian root mean square width of 4.247, one centered at 3688 cm⁻¹ with a Gaussian root mean square width of 8.069, and one centered at 3723 cm⁻¹ with a Gaussian root mean square width of 12.019. These fittings are shown underlying the experimental spectra in Fig. 7. The fact that the constants chosen for each Gaussian peak fit well over the full temperature range suggests that there is minimal evolution of the band shapes as a function of temperature, and thus, the integrated areas of each peak are good estimates of the population of the isomer giving rise to it. In Fig. 8, we show the integrated area of the Gaussians as a function of temperature, demonstrating the abrupt emergence and disappearance of the third spectroscopic feature.

Assuming that each peak corresponds to one class of isomers, we conclude that the isomers giving rise to the peaks at 3688 cm⁻¹ and 3722 cm⁻¹ maintain a nearly constant ratio of integrated intensities, suggesting that they lie very close in energy. This finding is consistent with the observation of multiple structures with very similar energies in Figs. 5 and 6. Putting these observations together, a potential explanation emerges in which isomers such as A and B (and potentially C) in Fig. 5 coexist at low temperatures, while isomer D begins to populate at temperatures >50 K. At higher temperatures, the barrier to isomerization between A and D is surpassed, and the unique spectroscopic signature of D is lost by 125 K. Notably, this is similar to the temperature recently found for the onset of rapid mixing of water cluster isotopomers.⁶² Given that comparisons between the intensities of IR transitions are difficult to quantify, we are not able to directly extract these energies, but the temperatures of the onset and loss of the new isomer correspond to energies of 0.1 kcal/mol and 0.2 kcal/mol, respectively. While these energies are well within the expected uncertainties of the calculations presented here, they are consistent with this explanation. Clearly, however, if there are other hydration isomers lower in energy than those explored in our calculations, this analysis should be revisited. The complexity of these thermal populations calls for a much more sophisticated consideration using higher accuracy thermochemical calculations or ab initio molecular dynamics.

Looking at both the (2,1) and (4,3) dry and singly hydrated cluster spectra, it appears as though the addition of water to these clusters is rather complex. Hydration is controlled by a balance between a good donor and a good acceptor for the incoming water. We have identified spectral markers to help determine how both the (2,1) and (4,3) clusters accommodate water and have found there to be several possible motifs in which the incoming water could bind. The water can insert between a bisulfate OS and ammonium NH, it can hydrogen bond to a free NH or a free bisulfate OH, or it can insert into a bisulfate–bisulfate hydrogen bond. The (2,1) cluster appears to favor the bisulfate OS/ammonium NH insertion, whereas the (4,3) cluster hydration appears to show an isomeric mixture in which a roughly equal fraction of clusters contains a bisulfate–bisulfate insertion and some contain either the water hydrogen-bonding to a free NH or a bisulfate OS/ammonium NH insertion. When the (4,0,3)_MA cluster is hydrated, the water prefers the bisulfate–bisulfate insertion location, suggesting that the amine-substituted clusters dictate water uptake sites more specifically. The (4,3)·(H₂O) cluster is the smallest hydrated cluster where isomers are present at low temperatures, and as temperature is increased, new sites appear to become populated and complex dynamics are found even in relatively low temperature ranges. The next step now is to begin a more detailed probing of the energetics of these clusters over a larger temperature range.

See the supplementary material for comparison of computed and experimental spectra and coordinates for all computed structures.

The authors acknowledge support from the National Science Foundation of the USA under Grant Nos. CHE-1566019 and CHE-1905172.

The data that support the findings of this study are available from the corresponding author upon reasonable request.