Molecular Dynamics Simulations of the ... - ACS Publications

Sep 22, 2014 - Despite the rich literature on these systems, it is still unclear if these salt clusters form via the CRM or the IEM. The current study...
0 downloads 0 Views 3MB Size
Article pubs.acs.org/JPCB

Molecular Dynamics Simulations of the Electrospray Process: Formation of NaCl Clusters via the Charged Residue Mechanism Lars Konermann,* Robert G. McAllister, and Haidy Metwally Department of Chemistry, The University of Western Ontario, London, Ontario N6A 5B7, Canada S Supporting Information *

ABSTRACT: Electrospray ionization (ESI) produces desolvated ions from solution phase analytes for mass spectrometric detection. The final steps of gas phase ion formation from nanometer-sized solvent droplets remain a matter of debate. According to the ion evaporation model (IEM), analytes are ejected from the droplet surface via field emission, whereas the charged residue model (CRM) envisions that ions are released upon droplet evaporation to dryness. Exposure of salt solutions to ESI conditions produces a range of cluster ions. Despite the rich literature on these systems, it is still unclear if these salt clusters form via the CRM or the IEM. The current study explores the formation of NanClm(n−m)+ clusters from aqueous sodium chloride solution under positive and negative polarity conditions. Molecular dynamics (MD) methods are used for simulating the temporal evolution of charged NaCl-containing water droplets. A trajectory stitching approach is developed for continuously removing evaporated moieties from the simulation, thereby dramatically reducing computational cost. In addition, this procedure ensures adequate temperature control and eliminates evaporative cooling that would otherwise slow down the process. Continuous water evaporation leads to progressive droplet shrinkage, while the emission of solvated single ions ensures that the system remains at ca. 90% of the Rayleigh limit. Early during the process all ions in the droplet behave as freely dissolved species, but after a few nanoseconds at 370 K the systems gradually morph into amorphous wet salt aggregates. Ultimately, free NanClm(n−m)+ clusters form as the last solvent molecules evaporate. Our data therefore provide direct evidence that sodium chloride cluster formation during ESI proceeds via the CRM. The IEM nonetheless plays an ancillary role, as it allows the system to shed charge (mostly in the form of hydrated Na+ or Cl−) during droplet shrinkage. It appears that this study marks the first successful MD simulation of complete CRM processes.



INTRODUCTION Electrospray ionization (ESI) mass spectrometry (MS) is a widely used analytical technique that covers a diverse range of applications.1−3 The ESI process starts with analyte solution that is passed through a conductive capillary to which high voltage has been applied. Redox processes lead to the buildup of positive or negative charge (depending on the polarity used) in the solution as it passes through the capillary.4 This charge accumulation induces the formation of a Taylor cone at the capillary outlet from which a plume of highly charged droplets is emitted.5 These ESI droplets are exposed to a heated gas environment where they undergo rapid evaporation and Coulombically driven jet fission.6 Nanometer-sized progeny droplets generated by these events release analyte ions into the gas phase.5,7 The ions are then sampled by an atmospheric pressure-to-vacuum interface where collisional activation promotes the final desolvation steps.6,8,9 Ultimately, ion detection by a suitable analyzer produces electrical signals that are converted into a mass spectrum.10 The mechanism of gas phase ion formation from charged nanodroplets continues to be a controversial topic. Many researchers believe that small analyte ions undergo field © 2014 American Chemical Society

emission from the droplet surface, as envisioned by the ion evaporation model (IEM).5,11−14 Large globular analytes such as natively folded proteins likely follow the charged residue model (CRM) where free ions are formed upon droplet evaporation to dryness.5,13,15−19 However, the distinction of IEM vs CEM on the basis of analyte size is not universally accepted,20,21 and it has been proposed that the IEM can also apply to large analytes.22−25 Hybrid models involving elements of both the CRM and the IEM have been put forward as well.26,27 In addition, recent data support a chain ejection model (CEM) for disordered polymers such as denatured proteins.28−30 During the CEM macromolecular chains get expelled from the droplet by electrostatic and solvophobic effects.13 Infusion of salt solutions into an ESI source generates charged clusters such as NanClm(n−m)+.31−35 It seems possible that these species are IEM products, i.e., that cations and anions associate within the shrinking droplet prior to cluster ejection Received: July 29, 2014 Revised: September 15, 2014 Published: September 22, 2014 12025

dx.doi.org/10.1021/jp507635y | J. Phys. Chem. B 2014, 118, 12025−12033

The Journal of Physical Chemistry B

Article

from the surface.36,37 Alternatively, salt clusters might form via the CRM, following a scenario where ESI droplets evaporate to dryness.38−40 An unambiguous differentiation between the two mechanisms on the basis of existing measurements is difficult, considering that the process may depend on the type of salt and the cluster stoichiometry.41−46 Hence, the mechanism of salt cluster formation during ESI remains an open question. Molecular dynamics (MD) simulations have become an important tool for probing mechanistic aspects of ESI13,28,47−54 and related phenomena.55−60 An obvious approach for examining the salt cluster formation process, therefore, would be to model the temporal evolution of salt-containing ESI droplets. Simulations of this type could help settle the longstanding dispute of IEM vs CRM. Surprisingly, attempts to pursue such a strategy have been relatively scarce. MD simulations of the ESI process have demonstrated that the production of small solvated ions such as Na+ occurs in accordance with the IEM.13,47,49,51,52 Simulation studies dealing with CRM events have been much more difficult.13 One challenge is related to the time scale of these processes. Whereas IEM events can readily be observed in simulation windows of less than 1 ns,13,47,49,51,52 droplet evaporation to dryness (the hallmark of the CRM) takes orders of magnitude longer.13 Also, MD simulations on droplet systems are often conducted at constant energy where evaporative cooling continuously reduces the temperature, thereby bringing solvent evaporation to a halt.13,50,51 In ESI mass spectrometers this cooling is countered by heated background gas.6,61 A recent study simulated such a situation directly, by placing hydrated metal ions in a bath of Ar atoms. Complete loss of the ∼20 bound waters took place within ca. 2 ns.54 That study54 marked the first time that simulations produced a solvent-free analyte ion via solvent evaporation. Most ESI mechanistic studies, however, require investigations on droplets comprising thousands of solvent molecules. The use of an explicit gas environment for systems of this size severely increases computational cost.62 A more effective approach involves the use of a thermostat.52,62,63 Nonetheless, to our knowledge there are no reports in the literature that describe simulations of ESI nanodroplet evaporation to dryness. In other words, MD strategies that are capable of dealing with CRM scenarios have yet to be developed. The current work devises an approach to overcome the difficulties outlined above. We conduct temperature-stabilized MD simulations that culminate in the formation of dry NanClm(n−m)+ clusters from salt-containing droplets containing ∼2420 water molecules. Rapid solvent evaporation from these droplets is accompanied by the shedding of charge carriers, mostly in the form of single solvated ions. These charge emission events are consistent with the IEM. The majority of the ions, however, remain contained within the shrinking droplet to the very end. Desolvated NanClm(n−m)+ aggregates form as the last water molecules evaporate, implying that the CRM represents the dominant mechanism of salt cluster formation.

extraction cones were set to 45 and 4 V, respectively, and the ESI capillary was held at 3 kV. Measurements were conducted in positive and in negative polarity. Isotope models were generated using the MassLynx software package provided by the instrument manufacturer. MD Simulations: General Aspects. Simulations were conducted on desktop Linux computers using Gromacs 4.6.5 in double precision for leapfrog integration of Newton’s equations.64,65 Unless noted otherwise, the Amber99sb-ILDN force field66 was used. A number of MD runs were also conducted using the OPLS all-atom (OPLS/AA) force field.67 All simulations employed the widely used three-site TIP3P water model.68 Earlier work has demonstrated that the choice of water model is not critical for droplet simulations, as various frameworks ranging from simple SPC to five-site TIP5P models all exhibit very similar evaporation behavior.51 The use of polarizable models in droplet simulations increases computing times by a factor of ∼30, and problems with energy conservation have been noted.69 Similar to recent studies from other laboratories,28,47,48,51−57 we therefore restricted the simulations of the current work to nonpolarizable models. H2O bond distances and angles were constrained using the SETTLE algorithm.70 Simulations were performed in a vacuum environment without cutoffs for electrostatic or Lennard-Jones interactions.51 ESI droplets were generated by carving spheres of the desired radius from a pre-equilibrated bulk water coordinate file. Ions were incorporated into these systems by replacing random water molecules with Na+ or Cl−. The droplets were initially subjected to steepest descent energy minimization using an integration step size of Δt = 0.5 fs.51 Equilibration was conducted with coupling to a velocity rescaling thermostat which employs a modified Berendsen scheme.71 Initial equilibration was performed at 1 K for 10 ps with Δt = 0.5 fs. For all the following steps Δt was increased to 1 fs. Equilibration was continued at 150 K for 20 ps and subsequently at 250 K for 70 ps.52 Production runs were then started with velocity rescaling71 at 370 K, unless noted otherwise. Additional details regarding temperature coupling are provided below. All simulations were conducted without barostat. Both center-of-mass translation and rotation of the system were eliminated throughout the simulation. The recoil associated with particle ejection can nonetheless cause translation and rotation of the residual droplet.69 This effect is particularly pronounced for small droplets, where ejection events carry away a relatively high percentage of the system mass. Temperature Stabilization by Trajectory Stitching. As noted above, a common issue with ESI simulations is the occurrence of droplet freezing due to the loss of kinetic energy brought about by solvent evaporation.13,50,51 In 2 ns test runs this phenomenon was explored using droplets with a 2 nm radius consisting of 1081 H2O, 17 Na+, and 8 Cl− at an initial temperature of 370 K (Figure 1). Simulations conducted at constant energy (i.e., without thermostat) resulted in a droplet temperature decrease from 370 to 293 K. The droplets lost ∼240 water molecules during the 2 ns window. Most of these evaporation events occurred early during the simulations, when the temperature was still high. This behavior is consistent with earlier observations.51 Next, it was attempted to stabilize the droplet temperature using a velocity rescaling thermostat.71 The coupling strength between system and heat bath is governed by a user-defined



EXPERIMENTAL AND SIMULATION METHODS Mass Spectrometry. ESI mass spectra were acquired on a Synapt G2 time-of-flight instrument (Waters, Milford, MA). 10 mM NaCl in water was infused into a standard Z-spray interface at a flow rate of 5 μL min−1 using a syringe pump. ESI was conducted using a desolvation gas (N2) temperature of 473 K and a source block temperature of 353 K. The sample and 12026

dx.doi.org/10.1021/jp507635y | J. Phys. Chem. B 2014, 118, 12025−12033

The Journal of Physical Chemistry B

Article

this strategy is comparable to a crude version of the Andersen thermostat.74 This stitching approach provided temperature stabilization of the droplet over extended (150 ns) MD trajectories. Supporting Information Figure S1 demonstrates that temperature deviations of no more than ±5% were observed for the first 4 ns. For later time points the temperature standard deviation increases, reflecting the instantaneous kinetic energy fluctuations of the rapidly shrinking system. Nonetheless, Supporting Information Figure S1 clearly illustrates that trajectory stitching virtually eliminates problems associated with evaporative cooling. A second key advantage associated with trajectory stitching is that evaporated particles can be removed from the system after completion of each MD segment. The continuously decreasing system size for consecutive segments dramatically speeds up the simulations. Complete 150 nm trajectories for 2.6 nm droplets could be run overnight on a regular quad-core CPU desktop machine without GPU acceleration. Particles were considered to be evaporated when they were more than 7 nm away from the main droplet center of mass. For trajectory segments with droplet fission the algorithm was designed to ensure transfer of the larger fragment into the subsequent segment. Trajectory stitching was implemented by using a script that makes alternating calls to Gromacs and to a custom-designed droplet cleanup program. Temperature Profile. The heated gas environment in the ion source of typical ESI mass spectrometers promotes solvent evaporation and analyte desolvation.6 For mimicking these conditions the simulations discussed below were run at 370 K for the first 50 ns. The tendency of the system to suffer from evaporative cooling is most pronounced early during the process when most of the solvent is lost (see below for details). The need for adequate temperature stabilization thus necessitated the use of very short (100 ps) MD segments for the first 4 ns. This was followed by 500 ps segments up to t = 10 ns and then 5 ns segments up to t = 50 ns. Electrosprayed analytes experience collisional activation as they traverse the sampling region and ion guides of the mass spectrometer. Experiments have shown that this process can raise the analyte temperature up to a range of 450−800 K.9,75 We therefore followed the initial 50 ns/370 K regime of the simulations by 50 ns at 450 K and then another 50 ns at 700 K (both in 5 ns segments), for an overall simulation window of 150 ns (Supporting Information Figure S1). Droplet Composition. ESI simulations were conducted on NaCl-containing water droplets with an initial radius of 2.6 nm which comprised ∼2420 water molecules. Na+ and Cl− ions were added to these systems at two different concentrations, 0.29 M (“low salt”) and 0.38 M (“high salt”), corresponding to 13 and 17 Na+/Cl− pairs, respectively. The bulk analyte solutions employed in typical ESI-MS experiments are less concentrated. However, the systems simulated in this work are meant to represent late progeny droplets.5 Solvent evaporation taking place during the production of these droplets can cause a 102−103-fold increase in solute concentration.76 The droplet composition used in our simulations, therefore, is in line with experimental conditions. The upper limit of net droplet charge is given by the Rayleigh equation5,77

Figure 1. Evaporation behavior of positively charged droplets with an initial radius of 2 nm (1081 H2O, 17 Na+, and 8 Cl−) at an initial temperature of 370 K. MD simulations were run under four conditions: without thermostat and with velocity rescaling thermostat using τT = 1, 0.1, and 0.01 ps, as noted along the x-axis. (a) Temperature of the droplet (black) and temperature of the entire system (including evaporated ions and waters, gray) after 2 ns. (b) Cumulative number of evaporated water molecules after 2 ns.

time constant τT. Shorter time constants correspond to tighter coupling. τT = 0.1 ps is a common choice,51,52 but we also tested values of 1 and 0.01 ps. Velocity rescaling stabilized the temperature around 370 K, when averaged over the entire system (i.e., the droplet with all of the evaporated molecules, Figure 1a). However, when considering only the droplet, a considerable temperature decrease persisted under thermostated conditions. The extent of droplet cooling decreased with decreasing τT (Figure 1a). This trend was accompanied by an increased number of evaporated water molecules (Figure 1b). Nonetheless, it is apparent from these data that even with very tight coupling the thermostat was not capable of eliminating evaporative droplet cooling. Similar observations were made when using Berendsen72 and Nosé−Hoover73 coupling (data not shown). The Andersen thermostat74 is not implemented in Gromacs 4.6.5 and was therefore not tested. Evidently, the persistence of evaporative droplet cooling renders simulations of CRM processes difficult under any of the conditions in Figure 1. We therefore settled on a scheme whereby long MD trajectories were “stitched together” from many shorter run segments. The idea behind this strategy is that evaporative cooling remains almost negligible for short simulation windows (e.g., 100 ps). After each of these short runs all velocities were reassigned at random from a Maxwell distribution, and the system was recoupled to the velocity rescaling thermostat71 at the desired temperature. In essence,

zR = 12027

8π ε0γR3 e

(1)

dx.doi.org/10.1021/jp507635y | J. Phys. Chem. B 2014, 118, 12025−12033

The Journal of Physical Chemistry B

Article

× 1) as well as Na23Cl22+ (cuboid 3 × 3 × 5, Figure 2a).31,35 The main peaks observed in negative polarity ESI-MS correspond to singly charged NanCl(n+1)− clusters (Figure 2b). Interestingly, the dominance of magic number species is less pronounced than for positive polarity; e.g., the 3 × 3 × 3 Na13Cl14− cluster41 does not stand out very strongly in Figure 2b. The insets in Figure 2 reveal the fine structure of the MS signals, reflecting the 35Cl/37Cl isotope heterogeneity. Close inspection of the data also reveals the presence of doubly charged clusters with lower abundance (red distributions in Figure 2).33 Positive Droplet Simulations. We conducted a first set of ESI simulations on NaCl-containing water droplets with excess positive charge under low salt conditions (26 Na+ and 13 Cl−). Snapshots taken from a typical 150 ns trajectory are shown in Figure 3. The droplet initially maintains an approximately spherical shape, but with undulations and short-lived protrusions (Figure 3, 100 ps). Solvent loss predominantly occurs by evaporation of single H2O molecules. During the first few nanoseconds this is accompanied by the ejection of solvated Na+ ions, with transition states that tend to exhibit connecting water filaments (Figure 3, 116 ps; Supporting Information Movie S1). Ions within the droplet initially have no tendency to associate with one another; instead they act as freely dissolved species that are surrounded by their individual hydration shells. With ongoing water evaporation, however, Na+ and Cl− begin to experience encounters. This leads to disordered salt aggregates that undergo rapid dissociation/ association within the droplet (Figure 3, 1.4 ns and 3.3 ns). Over time, these ion−ion contacts become more permanent (Figure 3, 4 ns and 4.5 ns). The partially hydrated t = 4.5 ns structure of Figure 3 has the ion composition Na15Cl132+. Emission of one final Na+ produces a cubic Na14Cl13+ cluster which retains bound waters at its sodium corner points (Figure 3, 7.5 ns; Supporting Information Movie S2). These remaining waters are lost within tens of nanoseconds, ultimately producing the desolvated 3 × 3 × 3 species (Figure 3, 150 ns) that dominates the mass spectrum of Figure 2a. Although this salt cube undergoes vibrational motions, it preserves its overall shape over extended time periods (Supporting Information Movie S3). Five independent low salt MD simulations all generated the same Na14Cl13+ product. Summary statistics of these runs are provided in Figure 4a,b. Results very similar to these Amber99sb-ILDN data were obtained when the simulations were repeated using the OPLS/AA force field (Supporting Information Figure S2). As a next step we examined the behavior of high salt droplets, with 30 Na+ and 17 Cl−. The temporal evolution of these systems (Figure 4c,d) was similar to that described above, with the exception that droplet evaporation produced slightly larger Na19Cl172+ clusters, comprising a 3 × 3 × 3 core that is decorated with a 3 × 3 × 1 layer (Figure 5). In contrast to the magic number product generated in low salt simulations (Figure 3, 150 ns), desolvated Na19Cl172+ exhibits frequent structural transitions where two [Na+ Cl− Na+] rows compete for inclusion at one of the cluster core edges (Supporting Information Movie S4). This behavior is consistent with reports of facile rearrangements in NaCl nanoclusters.41 None of the positive droplet simulations in Figure 4 showed any loss of Cl−, as seen from the horizontal green profiles in panels b and d. Thus, the number of Cl− in the positive cluster simulation products was determined by the chloride content of the initial nanodroplet.

where zR is the number of elementary charges e, R is the droplet radius, and ε0 is the vacuum permittivity. The surface tension γ is 0.058 91 N m−1 for water at its boiling point.78 For R = 2.6 nm eq 1 predicts a net charge of 15. However, the droplet evolution during ESI typically takes place slightly below the Rayleigh limit.5,79−81 For our simulations we therefore chose an initial charge of 13+ (13− in negative polarity), which corresponds to 87% of zR. This excess charge was achieved by incorporation of additional Na+ or Cl−. In ESI experiments the excess droplet charge may comprise other species such as H+, OH−, or NH4+,4,5 but there are also conditions that result in the ternary H2O/Na+/Cl− droplet composition used in our simulations.82 Each of the MD runs (positive and negative, with high and low salt) was repeated five times with different initial droplet structures and different random seeds for velocity assignments and temperature rescaling. Images were rendered in Pymol (Schrödinger), and movies were generated using VMD.83



RESULTS AND DISCUSSION ESI-MS of NaCl Clusters. Consistent with previous reports,31−35 infusion of 10 mM aqueous NaCl solution into an ESI source produced a wide range of salt clusters. When using positive polarity, the mass spectrum is dominated by singly charged Na(n+1)Cln+ species (Figure 2a). The most intense signal corresponds to the magic number cluster Na14Cl13+, which has a cubic 3 × 3 × 3 structure.31,35 Other prominent magic number peaks include Na5Cl4+ (planar 3 × 3

Figure 2. Mass spectra obtained by electrospraying aqueous NaCl solution using (a) positive polarity and (b) negative polarity ESI. Individual peaks are annotated according to the dominant ion signals. The insets show close-ups of the Na14Cl13+ and Na13Cl14− magic number signals along with the corresponding theoretical isotope models (blue dots). Red dots refer to the isotope models of doubly charged clusters. 12028

dx.doi.org/10.1021/jp507635y | J. Phys. Chem. B 2014, 118, 12025−12033

The Journal of Physical Chemistry B

Article

Figure 3. MD simulation snapshots for the evaporation of a positively charged water droplet with an initial radius of 2.6 nm, containing 2424 H2O (oxygen: red; hydrogen: white), 26 Na+ (blue), and 13 Cl− (green). The time points corresponding to individual frames are indicated. Note that the zoom level increases as time proceeds.

density and thus a reduced desolvation enthalpy compared to Na+ (rion = 0.99 Å).78 IEM ejection of hydrated single ions is the dominant mechanism for shedding charge during droplet shrinkage in both positive and negative polarity (Na+ in Figure 4b,d; Cl− in Figure 4f,h). For negatively charged systems the ejection of NaCl2− represents a second pathway for losing charge, although this process is not very common. NaCl2− can be released as hydrated species from shrinking droplets (Figure 6a, Supporting Information Movie S5). This process bears strong similarities with the ejection of single ions and is therefore classified as an IEM event (compare t = 2.174 ns of Figure 6 and t = 116 ps of Figure 3).5,11−14 Another mechanism for generating NaCl2− is the fragmentation of desolvated salt clusters. Such fragmentation was only observed for highly charged (3−) precursors, which undergo extensive structural distortions prior to dissociation (Figure 6b, Supporting Information Movie S6). Both types of NaCl2− production events occurred in 3 out of 10 trajectories. Figure 2b confirms that NaCl2− is produced under experimental conditions. The possible occurrence of NaCl2− loss in negative polarity simulations translates into a range of product clusters after droplet evaporation to dryness. All of these products were doubly charged. Na11Cl132−, Na12Cl142−, and Na13Cl152− were

Returning to the key question of this work, we can now examine what these MD simulations reveal about the mechanism of salt cluster formation during ESI. The ejection of single solvated Na+ during droplet shrinkage is consistent with the IEM.5,11−14 However, the positive droplet simulations did not show a single instance of cluster desorption from the droplet surface. Instead, NanClm(n−m)+ species were always released upon droplet evaporation to dryness; i.e., these clusters are CRM products.5,13,38−40 Negative Droplet Simulations. We complemented the simulations of the preceding section with studies on negatively charged droplets, both under low salt (13 Na+, 26 Cl−) and high salt conditions (17 Na+, 30 Cl−). The summary statistics of Figure 4e−h reveal that the formation of negative salt clusters generally proceeds along the same lines as discussed above for positive polarity. In other words, large negative clusters are CRM products as well. Despite the overall similarities in their behavior, there are some subtle differences in the temporal evolution of positive and negative droplets. No solvent persists under negative conditions for t > 10 ns (Figure 4d,g), whereas a few residual waters remain attached to positively charged clusters for up to ∼60 ns (Figure 4a,d). We attribute this difference to the larger ionic radius of Cl− (1.81 Å), which implies a lower charge 12029

dx.doi.org/10.1021/jp507635y | J. Phys. Chem. B 2014, 118, 12025−12033

The Journal of Physical Chemistry B

Article

Figure 5. Na19Cl172+ cluster generated in MD simulations after evaporation of a high salt droplet with an initial composition of 2416 H2O, 30 Na+, and 17 Cl−. The dotted square indicates the cubic 3 × 3 × 3 cluster core.

events (Figure 4). It is of interest to explore how this ongoing charge reduction compares to the Rayleigh limit. We will focus on the initial 4 ns window during which the droplets evaporate down to ∼100 water molecules. For smaller systems it becomes unclear if using a bulk water surface tension in eq 1 remains adequate.69,84 Rayleigh’s theory applies to spherical systems,5,77 whereas ESI droplets undergo frequent distortions into nonspherical shapes. We therefore determined an “effective” droplet radius for each time point, corresponding to that of a sphere with the equivalent number of H2O, Na+, and Cl−. When conducting this analysis, it is seen that the relative droplet charge remains close to its initial value of 0.87zR throughout the shrinkage process for all conditions studied (Figure 7). These data compare favorably to experiments, where excursions to z ≫ zR or z ≪ zR are not observed.5,79−81



Figure 4. Summary of evaporation kinetics for nanodroplets with an initial radius of 2.6 nm (∼2420 H2O molecules). The individual panels display the number of water molecules and ions contained within the droplet as a function of time for four different initial ion compositions: (a, b) positive droplet, “low salt” with 26 Na+ and 13 Cl−; (c, d) positive droplet, “high salt” with 30 Na+ and 17 Cl−; (e, f) negative droplet, “low salt” with 13 Na+ and 26 Cl−; (g, h) negative droplet, “high salt” with 17 Na+ and 30 Cl−. Data for each condition were averaged over five MD runs using the Amber99sb-ILDN force field. Error bars represent standard deviations.

CONCLUSIONS For many years it has been unclear whether the ESI-mediated production of salt clusters proceeds according to the CRM or the IEM.36−46 The current work explored this questions via MD simulations, using a trajectory stitching approach that provides adequate temperature control while at the same time dramatically reducing computational cost. Simulations on the temporal evolution of salt-containing ESI droplets provide direct evidence that NanClm(n−m)+ clusters are CRM products. Water evaporation causes gradual droplet shrinkage without ejection of large salt clusters from the droplet surface. Instead, the final NanClm(n−m)+ products represent charged residues that are left behind as droplets evaporate to dryness. Despite the unambiguous conclusion that NanClm(n−m)+ clusters are CRM products, the IEM still plays a role during the formation of these species. Shrinkage of ESI nanodroplets is accompanied by the ejection of hydrated charge carriers such as Na+ or Cl− (also NaCl2− on rare occasions). This type of charge emission represents an IEM process.5,11−14 Overall, it is therefore concluded that NanClm(n−m)+ clusters are produced via the CRM but that the IEM plays an ancillary role during droplet shrinkage. This finding is consistent with earlier proposals that were developed on the basis of experimental observations.26,27,85 It is gratifying that our simulations readily generate the 3 × 3 × 3 magic number cluster that represents the dominant species in experimental positive polarity spectra (Figures 2a and 3).

formed under low salt conditions, whereas Na15Cl172−, Na16Cl182−, and Na17Cl192− were seen in high salt simulations. Examples of these product species are depicted in Figure 6 for t = 150 ns. Na12Cl142− (Figure 6a) resembles a 3 × 3 × 3 magic number cube, but it lacks a central Na+ and therefore represents a hollow structure. Na17Cl192− (Figure 6d) is analogous to the Na19Cl172+ cluster of Figure 5. In summary, all of the large t = 150 ns salt clusters generated from negative droplets represent CRM products. Only the “minicluster” NaCl2− can be released from shrinking droplets via the IEM, or it can be formed by fragmentation of larger precursor clusters. Temporal Evolution of Droplet Charge. Phase Doppler interferometry and other experimental techniques revealed that ESI droplets remain slightly below the Rayleigh limit throughout their life cycle.5,79−81 Hence, the initial droplet charge used for our 2.6 nm droplets was chosen to be 0.87zR (eq 1). During the simulations droplets shrink because of water evaporation, while at the same time charge is lost via IEM 12030

dx.doi.org/10.1021/jp507635y | J. Phys. Chem. B 2014, 118, 12025−12033

The Journal of Physical Chemistry B

Article

Figure 6. MD snapshots taken at different times during the evaporation of negative droplets. (a) Low salt droplet with an initial composition of 2424 H2O, 13 Na+, and 26 Cl− during ejection of a solvated NaCl2− cluster. Also shown is the final Na12Cl142− product obtained in this run at t = 150 ns. (b) Data obtained during another low salt simulation, involving desolvated Na12Cl153− which ejects NaCl2− at t = 106 ns. The resulting Na11Cl132− product is shown for t = 150 ns. The two bottom panels display additional examples of t = 150 clusters generated via evaporation of negative high salt droplets. (c) Na16Cl182− and (d) Na17Cl192−. Coloring of elements is as in Figure 3.

notwithstanding, it would be unrealistic to presume that simple MD simulations of the type conducted here can exactly reproduce experimental mass spectra. Any attempts in this direction would require additional insights into the droplet size distribution within the ESI plume and the corresponding actual salt concentrations. Also, some of the experimentally observed clusters may represent gas phase fragments of the initial CRM products.31,34 Doubly charged clusters are particularly prone to fragmentation which may help explain the prevalence of singly charged species in Figure 2.33 In the future it will be interesting to extend the approach developed here to other types of ESI-MS analytes. While the current study provides strong evidence for the formation of NanClm(n−m)+ clusters via the CRM, our results do not necessarily apply to other solutes that undergo ESI-mediated clustering.41−46 It will also be exciting to test if the CRM holds for compact macromolecular analytes as often predicted5,13,15−19 or if the ESI process for these species proceeds via different pathways.20−25 Work in this direction is currently ongoing in our laboratory.

Figure 7. Droplet charge z relative to the Rayleigh charge zR during ESI simulations under the four conditions of Figure 4. The time window refers to evaporation down to a droplet size of ∼100 water molecules. Each data set represents an average of five MD runs.

Our simulations did not generate the corresponding 3 × 3 × 3 species in negative ion mode, despite inoculating the initial droplets with the proper number of Na+. It is intriguing to speculate that this effect seen in our negative droplet simulations is related to the low abundance of the 3 × 3 × 3 product in the experimental spectrum of Figure 2b. This idea



ASSOCIATED CONTENT

S Supporting Information *

Figures S1, S2 and Movies S1−S6. This material is available free of charge via the Internet at http://pubs.acs.org. 12031

dx.doi.org/10.1021/jp507635y | J. Phys. Chem. B 2014, 118, 12025−12033

The Journal of Physical Chemistry B



Article

(20) Testa, L.; Brocca, S.; Grandori, R. Charge-surface correlation in electrospray ionization of folded and unfolded proteins. Anal. Chem. 2011, 83, 6459−6463. (21) Marchese, R.; Grandori, R.; Carloni, R.; Raugei, S. A computational model for protein ionization by electrospray based on gas-phase basicity. J. Am. Soc. Mass Spectrom. 2012, 23, 1903−1910. (22) Ogorzalek Loo, R. R.; Lakshmanan, R.; Loo, J. A. What protein charging (and supercharging) reveal about the mechanism of electrospray ionization. J. Am. Soc. Mass Spectrom. 2014, 25, 1673− 1693. (23) Siu, K. W. M.; Guevremont, R.; Le Blanc, J. C. Y.; O’Brien, R. T.; Berman, S. S. Is droplet evaporation crucial in the mechanism of electrospray mass spectrometry? Org. Mass Spectrom. 1993, 28, 579− 584. (24) Nguyen, S.; Fenn, J. B. Gas-phase ions of solute species from charged droplets of solutions. Proc. Natl. Acad. Sci. U. S. A. 2007, 104, 1111−1117. (25) Fenn, J. B.; Rosell, J.; Meng, C. K. In electrospray ionization, how much pull does an ion need to escape its droplet prison? J. Am. Soc. Mass Spectrom. 1997, 8, 1147−1157. (26) Hogan, C. J.; Carroll, J. A.; Rohrs, H. W.; Biswas, P.; Gross, M. L. Combined charged residue-field emission model of macromolecular electrospray ionization. Anal. Chem. 2009, 81, 369−377. (27) Allen, S. J.; Schwartz, A. M.; Bush, M. F. Effects of polarity on the structures and charge states of native-like proteins and protein complexes in the gas phase. Anal. Chem. 2013, 85, 12055−12061. (28) Chung, J. K.; Consta, S. Release mechanisms of poly(ethylene glycol) macroions from aqueous charged nanodroplets. J. Phys. Chem. B 2012, 116, 5777−5785. (29) Yue, X.; Vahidi, S.; Konermann, L. Insights into the mechanism of protein electrospray ionization from salt adduction measurements. J. Am. Soc. Mass Spectrom. 2014, 25, 1322−1331. (30) Larriba, C.; de la Mora, F.; Clemmer, D. E. Electrospray ionization mechanisms for large polyethylene glycol chains studied through tandem ion mobility spectrometry. J. Am. Soc. Mass Spectrom. 2014, 25, 1332−1345. (31) Hao, C. Y.; March, R. E.; Croley, T. R.; Smith, J. C.; Rafferty, S. P. Electrospray ionization tandem mass spectrometric study of salt cluster ions. Part 1 - investigations of alkali metal chloride and sodium salt cluster ions. J. Mass Spectrom. 2001, 36, 79−96. (32) Hop, C. E. C. A. Generation of high molecular weight cluster ions by electrospray ionization; implications for mass calibration. J. Mass Spectrom. 1996, 31, 1314−1316. (33) Zhang, D. X.; Cooks, R. G. Doubly charged cluster ions (nacl)(m)(na)(2) (2+): Magic numbers, dissociation, and structure. Int. J. Mass Spectrom. 2000, 195, 667−684. (34) Feketeova, L.; O’Hair, R. A. J. Comparison of collision- versus electron-induced dissociation of sodium chloride cluster cations. Rapid Commun. Mass Spectrom. 2009, 23, 60−64. (35) Wakisaka, A. Nucleation in alkali metal chloride solution observed at the cluster level. Faraday Discuss. 2007, 136, 299−308. (36) Meng, C. K.; Fenn, J. B. Formation of charged clusters during electrospray ionization of organic solute species. Org. Mass Spectrom. 1991, 26, 542−549. (37) Kebarle, P.; Peschke, M. On the mechanisms by which the charged droplets produced by electrospray lead to gas phase ions. Anal. Chim. Acta 2000, 406, 11−35. (38) Juraschek, R.; Dulcks, T.; Karas, M. Nanoelectrospray - more than just a minimized-flow electrospray ionization source. J. Am. Soc. Mass Spectrom. 1999, 10, 300−308. (39) Wang, G. D.; Cole, R. B. Solvation energy and gas-phase stability influences on alkali metal cluster ion formation in electrospray ionization mass spectrometry. Anal. Chem. 1998, 70, 873−881. (40) Zhou, S.; Hamburger, M. Formation of sodium cluster ions in electrospray mass spectrometry. Rapid Commun. Mass Spectrom. 1996, 10, 797−800. (41) Doye, J. P. K.; Wales, D. E. Structural transitions and global minima of sodium chloride clusters. Phys. Rev. B 1999, 2292−2300.

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected] (L.K.). Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS Funding for this work was provided by the Natural Sciences and Engineering Research Council of Canada (NSERC). We thank David van der Spoel, Elio A. Cino, and Elias Ahadi for helpful discussions.



REFERENCES

(1) Fenn, J. B. Electrospray wings for molecular elephants (nobel lecture). Angew. Chem., Int. Ed. 2003, 42, 3871−3894. (2) Kaltashov, I. A.; Bobst, C. E.; Abzalimov, R. R. Mass spectrometry-based methods to study protein architecture and dynamics. Protein Sci. 2013, 22, 530−544. (3) Nilsson, T.; Mann, M.; Aebersold, R.; Yates, J. R.; Bairoch, A.; Bergeron, J. J. M. Mass spectrometry in high-throughput proteomics: Ready for the big time. Nat. Methods 2010, 7, 681−685. (4) Van Berkel, G. J.; Kertesz, V. Using the electrochemistry of the electrospray ion source. Anal. Chem. 2007, 79, 5511−5520. (5) Kebarle, P.; Verkerk, U. H. Electrospray: From ions in solutions to ions in the gas phase, what we know now. Mass Spectrom. Rev. 2009, 28, 898−917. (6) Covey, T. R.; Thomson, B. A.; Schneider, B. B. Atmospheric pressure ion sources. Mass Spectrom. Rev. 2009, 28, 870−897. (7) Cech, N. B.; Enke, C. G. Practical implication of some recent studies in electrospray ionization fundamentals. Mass Spectrom. Rev. 2001, 20, 362−387. (8) Fenn, J. B. Mass spectrometric implications of high-pressure ion sources. Int. J. Mass Spectrom. 2000, 200, 459−478. (9) Gabelica, V.; De Pauw, E. Internal energy and fragmentation of ions produced in electrospray sources. Mass Spectrom. Rev. 2005, 24, 566−587. (10) Xian, F.; Hendrickson, C. L.; Marshall, A. G. High resolution mass spectrometry. Anal. Chem. 2012, 84, 708−719. (11) Iribarne, J. V.; Thomson, B. A. On the evaporation of small ions from charged droplets. J. Chem. Phys. 1976, 64, 2287−2294. (12) Labowsky, M.; Fenn, J. B.; Fernandez de la Mora, J. A continuum model for ion evaporation from a drop: Effect of curvature and charge on ion solvation energy. Anal. Chim. Acta 2000, 406, 105− 118. (13) Konermann, L.; Ahadi, E.; Rodriguez, A. D.; Vahidi, S. Unraveling the mechanism of electrospray ionization. Anal. Chem. 2013, 85, 2−9. (14) McQuinn, K.; Hof, F.; McIndoe, J. S. Direct observation of ion evaporation from a triply charged nanodroplet. Chem. Commun. 2007, 2007, 4099−4101. (15) Dole, M.; Mack, L. L.; Hines, R. L.; Mobley, R. C.; Ferguson, L. D.; Alice, M. B. Molecular beams of macroions. J. Chem. Phys. 1968, 49, 2240−2249. (16) de la Mora, F. J. Electrospray ionization of large multiply charged species proceeds via Dole’s charged residue mechanism. Anal. Chim. Acta 2000, 406, 93−104. (17) Heck, A. J. R.; Van den Heuvel, R. H. H. Investigation of intact protein complexes by mass spectrometry. Mass Spectrom. Rev. 2004, 23, 368−389. (18) Kaltashov, I. A.; Mohimen, A. Estimates of protein surface area in solution by electrospray ionization mass spectrometry. Anal. Chem. 2005, 77, 5370−5379. (19) Iavarone, A. T.; Williams, E. R. Mechanism of charging and supercharging molecules in electrospray ionization. J. Am. Chem. Soc. 2003, 125, 2319−2327. 12032

dx.doi.org/10.1021/jp507635y | J. Phys. Chem. B 2014, 118, 12025−12033

The Journal of Physical Chemistry B

Article

(42) Gamero-Castano, M.; de la Mora, J. F. Modulations in the abundance of salt clusters in electrosprays. Anal. Chem. 2000, 72, 1426−1429. (43) Wang, G.; Cole, R. B. Charged residue versus ion evaporation for formation of alkali metal halide clusters ions in esi. Anal. Chim. Acta 2000, 406, 53−65. (44) Spencer, E. A. C.; Ly, T.; Julian, R. K. Formation of the serine octamer: Ion evaporation or charge residue? Int. J. Mass Spectrom. 2008, 270, 166−172. (45) Nanita, S. C.; Cooks, R. G. Serine octamers: Cluster formation, reactions, and implications for biomolecule homochirality. Angew. Chem., Int. Ed. 2006, 45, 554−569. (46) Zook, D. R.; Bruins, A. P. On cluster ions, ion transmission, and linear dynamic range limitations in electrospray (ionspray) mass spectrometry. Int. J. Mass Spectrom. Ion Processes 1997, 162, 129−147. (47) Znamenskiy, V.; Marginean, I.; Vertes, A. Solvated ion evaporation from charged water droplets. J. Phys. Chem. A 2003, 107, 7406−7412. (48) Marginean, I.; Znamenskiy, V.; Vertes, A. Charge reduction in electrosprays: Slender nanojets as intermediates. J. Phys. Chem. B 2006, 110, 6397−6404. (49) Caleman, C.; Hub, J. S.; van Maaren, P. J.; van der Spoel, D. Atomistic simulation of ion solvation in water explains surface preference of halides. Proc. Natl. Acad. Sci. U. S. A. 2011, 108, 6838−6842. (50) Steinberg, M. Z.; Breuker, K.; Elber, R.; Gerber, R. B. The dynamics of water evaporation from partially solvated cytochrome c in the gas phase. Phys. Chem. Chem. Phys. 2007, 9, 4690−4697. (51) Caleman, C.; van der Spoel, D. Temperature and structural changes of water clusters in vacuum due to evaporation. J. Chem. Phys. 2006, 125, 1545081−1545089. (52) Ichiki, K.; Consta, S. Disintegration mechanisms of charged aqueous nanodroplets studied by simulations and analytical models. J. Phys. Chem. B 2006, 110, 19168−19175. (53) Luedtke, W. D.; Landmann, U.; Chiu, Y.-H.; Levandier, D. J.; Dressler, R. A.; Sok, S.; Gordon, M. S. Nanojets, electrospray and ion field evaporation: Molecular dynamics simulations and laboratory experiments. J. Phys. Chem. A 2008, 112, 9628−9649. (54) Daub, C. D.; Cann, N. M. How are completely desolvated ions produced in electrospray ionization: Insights from molecular dynamics simulations. Anal. Chem. 2011, 83, 8372−8376. (55) Fegan, S. K.; Thachuk, M. A charge moving algorithm for molecular dynamics simulations of gas-phase proteins. J. Chem. Theory Comput. 2013, 9, 2531−2539. (56) Knochenmuss, R.; Zhigilei, L. V. Molecular dynamics simulations of maldi: Laser fluence and pulse width dependence of plume characteristics and consequences for matrix and analyte ionization. J. Mass Spectrom. 2010, 45, 333−346. (57) Larriba, C.; Fernandez de la Mora, J. The gas phase structure of coulombically stretched polyethylene glycol ions. J. Phys. Chem. B 2011, 116, 593−598. (58) Houriez, C.; Meot-Ner, M.; Masella, M. Simulated solvation of organic ions: Protonated methylamines in water nanodroplets. Convergence toward bulk properties and the absolute proton solvation enthalpy. J. Phys. Chem. B 2014, 118, 6222−6233. (59) Jungwirth, P.; Tobias, D. J. Specfic ion effects at the air/water interface. Chem. Rev. 2006, 106, 1259−1281. (60) Iyengar, S. S.; Day, T. J. F.; Voth, G. A. On the amphiphilic behavior of the hydrated proton: An ab initio molecular dynamics study. Int. J. Mass Spectrom. 2005, 241, 197−204. (61) Gibson, S. C.; Feigerle, C. S.; Cook, K. D. Fluorometric measurement and modeling of droplet temperature changes in an electrospray plume. Anal. Chem. 2014, 86, 464−472. (62) Wedekind, J.; Reguera, D.; Strey, R. Influence of thermostats and carrier gas on simulations of nucleation. J. Chem. Phys. 2007, 127, 0645011−06450112. (63) Meyer, R.; Gafner, J. J.; Gafner, S. L.; Stappert, S.; Rellinghaus, B.; Entel, P. Computer simulations of the condensation of nanoparticles from the gas phase. Phase Transitions 2005, 78, 35−46.

(64) Hess, B.; Kutzner, C.; van der Spoel, D.; Lindahl, E. Gromacs 4: Algorithms for highly efficient, load-balanced, and scalable molecular simulation. J. Chem. Theory Comput. 2008, 4, 435−447. (65) Van der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A. E.; Berendsen, H. J. C. Gromacs: Fast, flexible, and free. J. Comput. Chem. 2005, 26, 1701−1718. (66) Lindorff-Larsen, K.; Piana, S.; Palmo, K.; Maragakis, P.; Klepeis, J. L.; Dror, R. O.; Shaw, D. E. Improved side-chain torsion potentials for the amber ff99sb protein force field. Proteins 2010, 78, 1950−1958. (67) Kaminski, G.; Duffy, E. M.; Matsui, T.; Jorgensen, W. L. Free energies of hydration and pure liquid properties of hydrocarbons from the opls all-atom model. J. Phys. Chem. 1994, 98, 13077−13082. (68) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. Comparison of simple potential functions for simulating liquid water. J. Chem. Phys. 1983, 79, 926−935. (69) Caleman, C.; van der Spoel, D. Evaporation from water clusters containing singly charged ions. Phys. Chem. Chem. Phys. 2007, 9, 5105−5111. (70) Miyamoto, S.; Kollman, P. A. Settle: An analytical version of the shake and rattle algorithm for rigid water models. J. Comput. Chem. 1992, 13, 952−962. (71) Bussi, G.; Donadio, D.; Parrinello, M. Canonical sampling through velocity rescaling. J. Chem. Phys. 2007, 126, 0141011− 0141017. (72) Berendsen, H. J. C.; Postma, J. P. M.; Vangunsteren, W. F.; Dinola, A.; Haak, J. R. Molecular dynamics with coupling to an external bath. J. Chem. Phys. 1984, 81, 3684−3690. (73) Hoover, W. G. Canonical dynamics: Equilibrium phase-space distributions. Phys. Rev. A 1985, 31, 1695−1697. (74) Andersen, H. C. Molecular dynamics simulations at constant pressure and/or temperature. J. Chem. Phys. 1980, 72, 2384−2393. (75) Merenbloom, S. I.; Flick, T. G.; Williams, E. R. How hot are your ions in twave ion mobility spectrometry? J. Am. Soc. Mass Spectrom. 2012, 23, 553−562. (76) Kebarle, P.; Tang, L. From ions in solution to ions in the gas phase: The mechanism of electrospray mass spectrometry. Anal. Chem. 1993, 65, 972A−986A. (77) Rayleigh, L. On the equilibrium of liquid conducting masses charged with electricity. Philos. Mag. 1882, 14, 184−186. (78) Lide, D. R. CRC Handbook of Chemistry and Physics, 82nd ed.; CRC Press: Boca Raton, FL, 2001. (79) Smith, J. N.; Flagan, R. C.; Beauchamp, J. L. Droplet evaporation and discharge dynamics in electrospray ionization. J. Phys. Chem. A 2002, 106, 9957−9967. (80) Taflin, D. C.; Ward, T. L.; Davis, E. J. Electrified droplet fission and the Rayleigh limit. Langmuir 1989, 5, 376−384. (81) Gomez, A.; Tang, K. Charge and fission of droplets in electrostatic sprays. Phys. Fluids 1994, 6, 404−414. (82) Van Berkel, G. J.; De La Mora, J. F.; Enke, C. G.; Cole, R. B.; Martinez-Sanchez, M.; Fenn, J. B. Electrochemical processes in electrospray ionization mass spectrometry. J. Mass Spectrom. 2000, 35, 939−952. (83) Humphrey, W.; Dalke, A.; Schulten, K. Vmd: Visual molecular dynamics. J. Mol. Graphics 1996, 14, 33−38. (84) Zakharov, V. V.; Brodskaya, E. N.; Laaksonen, A. Surface properties of water clusters: A molecular dynamics study. Mol. Phys. 1998, 95, 203−209. (85) Gamero-Castaño, M.; de la Mora, F. J. Mechanisms of electrospray ionization of singly and multiply charged salt clusters. Anal. Chim. Acta 2000, 406, 67−91.

12033

dx.doi.org/10.1021/jp507635y | J. Phys. Chem. B 2014, 118, 12025−12033