In beam-based ionization methods, the substrate plays an important role on the desorption mechanism of molecules from surfaces. Both the specific orientation that a molecule adopts at a surface and the strength of the molecule-surface interaction can greatly influence desorption processes, which in turn will affect the ion yield and the degree of in-source fragmentation of a molecule. In the beam-based method of secondary ion mass spectrometry (SIMS), in-source fragmentation can be significant and molecule specific due to the hard ionization method of using a primary ion beam for molecule desorption. To investigate the role of the substrate on orientation and in-source fragmentation, we have used atomistic simulations—molecular dynamics in combination with density functional theory calculations—to explore the desorption of a sphingolipid (palmitoylsphingomyelin) from a model surface (gold). We then compare SIMS data from this model system to our modeling findings. Using this approach, we found that the combined adsorption and binding energy of certain bonds associated with the headgroup fragments (C3H8N+, C5H12N+, C5H14NO+, and C5H15PNO4+) was a good predictor for fragment intensities (as indicated by relative ion yields). This is the first example where atomistic simulations have been applied in beam-based ionization of lipids, and it presents a new approach to study biointerfacial lipid ordering effects on SIMS imaging.

In-source fragmentation (ISF) in mass spectrometry occurs when ionization creates an energetically unstable molecule that results in intramolecular bond breakage. As such, ISF results in the generation of fragment species from an intact parent ion. ISF is a significant problem in beam-based ionization methods, where an analytical probe is used for in situ molecule desorption and ionization, with the most severe ISF occurring in “hard” ionization beam-based methods, including secondary ion mass spectrometry (SIMS).1 In SIMS, a highly focused primary ion beam is used to eject secondary ions, which are rarely desorbed intact due to the high degree of ISF. ISF results in spectral complexity due to the large proportion of fragment to intact molecular ions.2 In these cases, caution must be taken in spectral annotation, as fragments can be interpreted as intact species in untargeted analyses that result in errors in structural assignments of endogenous molecules.3 Fragmentation can be minimized in SIMS through methods that include using polyatomic primary ion sources4 and dynamic reactive ionization;5 however, ISF can still be a significant problem even when using these approaches. Techniques to understand the ISF that occurs in SIMS process are, therefore, necessary. Frequently, in silico approaches are used in beam-based ionization to determine the fragmentation that occurs in a molecule.6 However, these databases are largely based on electron ionization (EI)7 and are, therefore, limited in their applicability toward ion beam-based ionization due to differences in the ionization mechanism. More importantly, they do not consider the influence of the substrate for in situ analyses.8 

In beam-based analyses, such as SIMS, molecules are desorbed from a substrate (i.e., solid support). It is given that molecules will generally arrange themselves at an interface by orientating to the lowest energy state to minimize interfacial energy.9 This orientation can greatly affect the ability for a molecule to desorb from a surface, the degree of ISF that a molecule experiences, and the resultant fragmentation pattern. For example, Baio et al. found that changes in fragmentation patterns of protein GB1 on gold surfaces occurred as a result of orientation, with the most exposed portions of the protein to the incoming primary ion beam experienced greater ISF.10 Changes in fragmentation patterns related to orientation have also been observed in analysis of antibodies,11 polymer films,12 and self-assembled polypeptides.13 

There are multiple experimental approaches to determine molecular orientations at surfaces, some of which involve second harmonic generation,14 sum frequency generation,15 and angle-resolved photoacoustic spectroscopy.16 Observed molecule configurations derived from experimental data are often complemented by computational methods, such as molecular dynamics (MD), to provide an atomic-level resolution into the assembly at interfaces.17 To date, there has been few studies exploring the effect of molecular orientation on lipid fragmentation in beam-based ionization.18 MD, when combined with density functional theory (DFT), can be applied to understand the mechanisms of fragmentation in mass spectrometry (MS) as an alternative approach to heuristic calculators.19 DFT calculations can give information about relative bond strengths and the propensity for specific bonds to break during an ionization process.10 However, in beam-based in situ approaches, bond strength can be heavily influenced by substrate interactions. For example, Torrisi et al. found, using a combined MD/DFT approach, that fluoroalkyl compounds adopt specific orientations on gold and calcarenite surfaces, which alter fragmentation patterns in SIMS.20 

To explore the influence of the substrate in SIMS analyses, we have applied atomistic simulations—MD simulations and DFT calculations—to explore the interaction of a sphingolipid, palmitoylsphingomyelin (PSM), on an Au (111) surface. MD simulations allow for the orientation of PSM to be determined at maximum packing on a surface. We then analyzed the deposited lipid by SIMS to determine the ISF that occurs and identify major species produced from fragmentation of the lipid. These identified fragments are then used to inform DFT calculations to understand the effect of the Au (111) surface on specific bonds in PSM and determine fragment adsorption energies. This methodology is the first example of correlating MD, DFT, with SIMS data for understanding primary ion beam ionization of lipids.

The N-palmitoyl-sphingomyelin (PSM) molecular structure was obtained from the CHARMM-GUI individual lipid library website.21 CHARMM36 parameters for PSM were also obtained from the CHARMM-GUI web interface.22 The Au (111) surface was obtained from the supporting information of the article by Heinz et al. along with the gold core and “electronic” shell polarized force field parameters.23 The initial Au (111) surface lattice cell from Heinz et al. was expanded to a larger 49.948 × 49.023 × 12.273 Å3 supercell using the materials studio software.24 Random two-dimensional packing of 30 PSM molecules to give full monolayer coverage on the Au (111) surface was performed with the packmol software program.25 The PSM molecules were packed with the PSM head group initially interfacing the Au (111) surface. Packing of PSM was performed in a 100 Å3 vacuum layer over the surface to allow for a sufficient distance between periodic images above and below the PSM-covered surface.

MD simulations for the PSM-gold model system were carried out with the LAMMPS molecular dynamics package.26 The particle-particle particle mesh (PPPM) method was used to treat the long-range Coulombic interactions.27 The Lorentz–Berthelot mixing rules for nonbonding interactions were applied. Each model was initially optimized with the conjugate gradient method and then equilibrated with NVT [constant number (N), volume (V), and temperature (T)] dynamics at 303 K for 1 ns. Following this initial NVT equilibration, the system was equilibrated with NVT dynamics for another 10 ns at 303 K. Following equilibration, the production NVT MD simulations at 303 K were carried out for 50 ns. The SHAKE algorithm was used to constrain the hydrogen-bearing bonds to allow longer time steps.28 Consequently, the time step adopted was 1 fs. Analysis and visualization of the MD trajectories were performed using the vmd29 and materials studio24 software packages.

For the gas phase PSM system, all bond energy calculations were performed with the NWChem quantum chemistry code.30 For these calculations, the PSM molecule and each fragment geometry from bond breaking were optimized with the M06-2X Minnesota density functional of Zhao and Truhlar31 and the 6-31G* basis set of Pople and co-workers.32,33

The vesta software was used to build the periodic system composed of a PSM on the Au (111) surface.34 Two major orientations and conformations of PSM on the Au (111) surface observed in the MD simulations were used as initial configurations of the calculations. The Au (111) surface was modeled by a four-layer thick slab in both cases. A cell parameter of 10.22 × 8.85 × 50 Å3 and 15.33 × 8.85 × 50 Å3 were used for the two different orientations, respectively. With these dimensions, approximately 30 PSM molecules with the most populated orientation are packed in a 50 × 50 Å2 Au (111) surface, which is equivalent to the density of lipid on the surface observed from the result of MD simulations (Fig. 1). There are approximately 24 PSM molecules with the second most populated orientation in the same dimension. We added enough vacuum space on the z axis (50 Å) of the system in order to minimize interactions across the periodic boundary on the z axis, such as electrostatic interactions between the tail of the lipid molecule and the bottom of the Au (111) surface. Details on the PSM with two representative configurations on Au (111) surfaces are provided in Fig. S1.53 

FIG. 1.

(a) Side view of a MD snapshot of 30 PSM molecules packed and bound to a 5 × 5 nm2 Au (111) surface and (b) a top view of the same MD snapshot. (c) Pie chart showing molecule specific orientations across the population of PSMs adsorbed to the Au (111) surface. (d) Most common orientation of a PSM molecule with the head group interacting with the Au (111) surface. (e) Second most common orientation of PSM with amide and hydroxyl groups, along with the phosphate/ammonium head group bound to the Au (111) surface. Structure of PSM is also shown (f) with the headgroup and tail group labels. For (a), (b), (d), and (e), yellow are gold atoms, cyan are carbon atoms, brown are phosphorous atoms, red are oxygen atoms, blue are nitrogen atoms, and white are hydrogen atoms.

FIG. 1.

(a) Side view of a MD snapshot of 30 PSM molecules packed and bound to a 5 × 5 nm2 Au (111) surface and (b) a top view of the same MD snapshot. (c) Pie chart showing molecule specific orientations across the population of PSMs adsorbed to the Au (111) surface. (d) Most common orientation of a PSM molecule with the head group interacting with the Au (111) surface. (e) Second most common orientation of PSM with amide and hydroxyl groups, along with the phosphate/ammonium head group bound to the Au (111) surface. Structure of PSM is also shown (f) with the headgroup and tail group labels. For (a), (b), (d), and (e), yellow are gold atoms, cyan are carbon atoms, brown are phosphorous atoms, red are oxygen atoms, blue are nitrogen atoms, and white are hydrogen atoms.

Close modal

All the calculations for the adsorption energies were performed using the quantum espresso version 6.5 code.35,36 For the geometry optimizations followed by electronic energy calculations of the Au surface (surf.), the fragments of interest (frag.), and the entire complex (surf. + frag.), and the exchange-correlation effects were applied by using the Perdew–Burke–Ernzerhof (PBE) functional37,38 with Grimme's D3 dispersion force corrections.39,40 The atomic core electrons were described by an optimized norm-conserving Vanderbilt pseudopotential (ONCVPSP) for the gold atoms41 and by projector-augmented wave (PAW)42 pseudopotentials from the ps library for hydrogen, carbon, nitrogen, oxygen, and phosphorus atoms.43 After convergence tests, a plane-wave function cutoff energy of 60 Ry along with the kinetic energy cutoff for a charge density of 480 Ry was determined for the calculations. A (3 × 3 × 1) Monkhorst–Pack k-point mesh was used for integration over the Brillouin zone of the system.44 Details on adsorption energies calculations, including the snapshots of the initial system and optimizations of energy cutoffs and k-points, are illustrated in Fig. S1 of the supplementary material.53 

22 × 22 mm2 gold coated (50 nm thick) glass coverslips (Tedpella, Redding, CA) were cleaned by sonication in methanol, acetone, and ultrapure de-ionized water (18.2 MΩ cm) for 10 min each, followed by drying with nitrogen gas. Separately, a polydimethylsiloxane (PDMS; Dow Corning, Midland, MI) layer was casted by mixing a 40 ml PDMS base and curing polymers at a 20:1 ratio and then pouring the mixture into a 140 mm diameter Petri dish. The PDMS was then cured at 70 °C for 1 h. A PDMS ring was created using a 5 mm hole punch. The PDMS ring polymer was then plasma cleaned (PX250, Nordson March, Concord, CA) and placed on the center of the cleaned gold coated coverslip and allowed to bond for 5 min. The MD simulations were used to calculate the concentration of PSM deposited in the O-ring to establish a lipid monolayer. 10 μl of PSM (>99% purity, Avanti polar lipids; Birmingham, AL) at a concentration of 3.9 nM in MeOH was the drop cast in the O-ring and allowed to dry. The O-ring was then removed from the coverslip, and samples were stored under an inert gas for surface analysis.

XPS measurements were performed using a Thermo Fisher NEXSA spectrometer with a 125 mm mean radius, a full 180° hemispherical analyzer, and a 128-channel detector. This system uses a focused monochromatic Al Kα x-ray (1486.7 eV) source for excitation and an electron emission angle of 60°. The narrow scan spectra were collected using a pass-energy of 50 eV with a step size of 0.1 eV. For the Ag 3d5/2 line, these conditions produced a FWHM of 0.84 ± 0.02 eV. The binding energy (BE) scale is calibrated using the Cu 2p3/2 feature at 932.62 ± 0.05 eV and Au 4f7/2 at 83.96 ± 0.05 eV. XPS measurements were taken from three areas (400 × 400 μm2) across the lipid spot. Overlayer calculations were performed based on a thickogram method, which relies on conversion of substrate (Au4f) and overlayer peak intensities (P2p).45 

TOF-SIMS analysis was performed using a TOF-SIMS V (IONTOF GmbH, Münster, Germany) instrument using a Bi3+ liquid metal ion gun in a high current bunched mode (HCB) with an energy of 25 keV. The target current of the analysis beam was set to 0.5 pA with a total primary ion dose density of 4.95 × 109 ions/cm2. An 8 × 8 mm2 area was imaged containing the drop cast lipid. Each 500 × 500 μm2 area was rastered in a random pattern across 512 × 512 pixels, equating to a resolution of ≈1 μm/pixel. The total scan time for imaging the lipid drop was recorded as 25 min. An electron flood gun was used to neutralize any charge buildup. Data processing was performed using surfacelab 7 (IONTOF GmbH, Münster, Germany). Positive and negative ion spectra were collected over the mass range of m/z 0–900. Regions of interest were extracted based on thresholding of the phosphocholine headgroup peak (C5H15PNO4+, positive ion mode) and the dimethyl-ethanolamine phosphate peak (C4H11PNO4+, negative ion mode). Positive ion spectra were mass calibrated using the C+, CH+, CH2+, CH3+, and C2H3+ peaks, and negative ion mode spectra were mass calibrated using the CH, OH, C2H, and C4H peaks. For a positive ion mode, the mass resolution of the C2H3+ peak (m/z 27) was >4000. For a negative ion mode, the mass resolution of the C2H peak (m/z 25) was >3000. Image tiles were exported into matlab in an ASCII text format to convert relative abundance to molar concentration in images.

Palmitoylsphingomyelin (PSM) was chosen to model in this study, as this lipid is a major constituent of cell membranes and is frequently annotated in mass spectrometry imaging (MSI) studies.46,47 Au served as a model conductive, ion-beam compatible surface to investigate the substrate effects on PSM.

Figure 1 details the MD model built of PSM on Au (111), with details on the periodic system shown in Fig. S1.53 To simulate conditions of PSM adsorption on Au (111), adsorption under maximum packing conditions was modeled. Figures 1(a) and 1(b) show the formation of a stable PSM layer from different views, where the majority of the molecules are oriented with the headgroup directly interacting with the Au (111) surface, while the ceramide tail groups are oriented away from the surface. In total, 30 PSM molecules were able to be incorporated into the 5 × 5 nm2 Au (111) surface, with a packing density of 1.2 molecules/nm2. In its lowest energy configuration, a PSM has an average diameter of 2.97 nm; therefore, a 5 × 5 nm2 surface supported up to 30 PSM indicates close packing. A graphical representation of all orientations in the PSM film is shown in Fig. 1(c). Within this packed surface, the majority of PSM molecules (60%) are absorbed via the entire headgroup, with a second most common orientation (10%) of the headgroup plus the amide group adsorbed to the surface. The remaining subpopulation of PSM molecules orient in a variety of other different ways, where the polar amide and the hydroxyl group are further up from the head group at the Au (111) surface, along with the zwitterionic phosphate/ammonium head group or with an orientation of one or both of the alkyl tails fully interacting with the surface.

When PSM interacts with the Au (111) substrate, it primarily adopts two configurations: a full headgroup adsorption [orientation 1, Fig. 1(d)] equating to 10.22 Å of surface coverage and a full headgroup with amide/hydroxyl adsorption [orientation 2, Fig. 1(e)] equating to 15.33 Å of surface coverage. In both orientations, the hydrophobic ceramide is oriented away from the surface. Sphingolipids form bilayer structures in lipid microdomains with the amphiphilic headgroup in contact with the aqueous fluid.48 Previous computational studies have demonstrated that the polar head group of the lipid molecules tends to interact with the Au (111) substrate, while hydrophobic tails are oriented away from the Au (111) in a similar way to bilayer formation.49 These two observed conformations suggest a strong interaction between the polar phosphocholine headgroup, and the Au (111) substrate could be the key contributor for the adsorption process of PSM on Au (111).

Figure 2 shows SIMS data from the corresponding analysis of the O-ring deposited lipid film on gold. Figure S2 shows TOF-SIMS ion images from positive ion mode analysis of the spotted lipid.53 Lipid adsorption was constrained within the O-ring, shown in Figs. S2(a)–S2(f),53 where the characteristic PC headgroup species (m/z 104, m/z 184, m/z 198) colocalize in the O-ring region. Region of interest extraction (ROI) was required to minimize the influence of extraction height across the imaged O-ring region on mass resolution. Depositing PSM at a concentration to match a surface density of 1.2 PSM/nm2, informed by modeling, allowed for the conversion of peak intensity to surface concentration using peak intensity values (Fig. S3).53 XPS analysis was performed to assess overlayer thickness. Averaging the per pixel lipid concentration gave a value of 2.6 × 10−14 mol/pixel. XPS data acquired from the lipid spot (Table S1)53 indicated that the film thickness was 1.6 nm +/−0.5. The P/N (phosphorus/nitrogen) ratio was calculated at 0.5 +/−0.2, matching the theoretical P/N ratio of PSM, indicating monolayer formation. However, SIMS imaging [Fig. S3 (Ref. 53)] indicates that the lipid film was nonuniform indicating and PSM aggregated in some areas of the Au surface.

FIG. 2.

Representative spectra from SIMS analysis of PSM on Au (111). (a) Positive ion mode spectra (left) for m/z 0–100 and (right) m/z 100–200, (b) structure of PSM showing fragments observed in SIMS, and (c) negative ion mode spectra (left) for m/z 0–500 and (right) m/z 600–700. The colored dots in spectra (a) and (c) corresponding to bonds broken are shown in structure (b). Abundance in spectra is represented as background subtracted peak areas.

FIG. 2.

Representative spectra from SIMS analysis of PSM on Au (111). (a) Positive ion mode spectra (left) for m/z 0–100 and (right) m/z 100–200, (b) structure of PSM showing fragments observed in SIMS, and (c) negative ion mode spectra (left) for m/z 0–500 and (right) m/z 600–700. The colored dots in spectra (a) and (c) corresponding to bonds broken are shown in structure (b). Abundance in spectra is represented as background subtracted peak areas.

Close modal

Figure 2 shows SIMS spectra and the proposed bond breakages with resultant fragment species in PSM. A breakdown of all detected fragments is also shown in Table S2 (Ref. 53) (positive ion mode) and Table S3 (Ref. 53) (negative ion mode). Positive ion mode analysis produced abundant low molecular weight species corresponding to the fragmented phosphocholine headgroup (C5H14NO+, C5H15PNO4+, C6H17PNO4+) and nonheadgroup specific species (C5H12N+, C3H8N+) [Fig. 2(a) and Fig. S4 (Ref. 53)]. No salt adducts of the molecular species (M + H, M + Na, M + K) were observed in a positive ion mode, suggesting that PSM is unstable. This is unsurprising as headgroup fragmentation occurs across most lipid classes. In a negative ion mode [Fig. 2(c) and Fig. S5 (Ref. 53)], high molecular weight species were detected (m/z 600–800) that corresponded to partial lipid headgroup fragmentation at m/z 616.47 (M – C5H12N), m/z 642.49 (M – C3H10N), and m/z 687.55 (M–CH3). Each of the fragments corresponding to the loss peaks (C5H12N+, C3H10N+, CH3+) was observed in a positive ion mode. The loss of the terminal methyl unit has been seen in electrospray ionization of PSM, indicating that this group is labile under both SIMS and electrospray ionization.50 In a negative ion mode, a species at m/z 447.29 (C22H44N2O5P) was observed indicative of the loss of C17H35O2, which was observed in a negative ion mode (Table S3).53 The formation of a C17 fragment likely arises from fragmentation of the C18:1 ceramide tail. The dimethyl-ethanolamine phosphate headgroup fragment at m/z 168.04 (C4H11PNO4) also forms in a negative ion mode [Table S3 (Ref. 53)], which has also been observed in electrospray ionization of PSM.50 We did not observe the ceramide fragment arising from loss of the PC headgroup in either polarity, indicating that the ceramide is either unstable and undergoes ISF, or that ceramide is a neutral loss peak and, therefore, undetectable.

To calculate why specific bonds might weaken in PSM and become abundantly detected fragments because of the molecule's surface interactions with Au (111), DFT was used to determine bond and desorption energies. Specifically, bonds associated with major fragments observed in a positive ion mode, the PC headgroup (m/z 58, m/z 86, m/z 104, m/z 184) species, and major fragments observed in a negative ion mode (m/z 687, m/z 642, m/z 616) were modeled using DFT.

Desorption energies and bond dissociation energies for these fragments were then compared against their relative ion intensities, in an effort to determine if these energies can be used to predict the propensity of specific fragments to form during ISF of PSM. The calculated desorption energies for positive and negative ion mode data are shown in Table S4.53 Both full headgroup (orientation 1) and full headgroup plus amide/hydroxyl (orientation 2) showed no correlation of desorption energies and major fragment intensities in either ionization polarity (Figs. S6 and S7).53 A positive ion mode (Fig. S6)53 showed no correlation with R2 = 0.21 and R2 ∼ 0.00 for the first and second orientation, respectively, where R2 shows how well the data fit the regression model as a coefficient of determination. There was no correlation between desorption energies and the relative intensities of fragments observed in a negative ion mode (Fig. S7)53 with R2 = 0.26 and R2 = 0.22 for the first and second orientation, respectively. Bond dissociation energy calculations were then performed (Fig. S8),53 where a correlation between the relative intensity of positively PC fragments and the energy of certain bonds was observed. The P-O bond forming the lowest abundant fragment in the modeled fragments, choline at m/z 104, has the largest bond energy at 186 kcal/mol, whereas the C-C bond forming the abundant phosphocholine at m/z 184 has a bond energy of 101 kcal/mol. Figure S8 (Ref. 53) shows that there is a good correlation (R2 = 0.78) of bond energy and relative intensity of certain positively charged species from headgroup fragmentation.

We compared the summation of bond dissociation energy and desorption energy with fragment abundance (Fig. 3), as bond strength is a key factor on the propensity of a molecule to undergo ISF. Combining the bond dissociation energy and desorption energy shows good correlation with the relative fragment intensity (R2 = 0.94) for both the first and second orientations [Fig. 3(c)]. These correlations clearly indicate that bond dissociation energy could play a major role in fragmentations of PSM. However, given that the best correlations were observed when both the bond dissociation energy and the desorption energy from the Au (111) surface are considered, it suggests that the desorption energy also contributes to the intensity of fragment species of PSM. We did not find any relationship between the relative intensity of negatively charged fragment species (m/z 687, m/z 642, m/z 616, m/z 447, m/z 97, m/z 79) and their combined bond dissociation and desorption energies in negative ion mode data. This may be explained by either changes in ionization behavior in a negative ion mode related to charge competition and/or contribution of the phosphate species at m/z 97 and m/z 79 as these peaks are nonspecific to lipids and may arise from a variety of surface contaminants.

FIG. 3.

(a) Bond dissociation energies for the formation of PSM fragments of interest in the gas phase. (b) Representative snapshots with two orientations that explain how desorption energy (E) between the fragments of interest and the Au (111) surface was obtained. Cyan, red, white, and yellow atoms represent the carbon, oxygen, hydrogen, and gold, respectively. Black atoms are the ones with two identical periodic images whose primary unit cell is marked with a blue box. The fragment m/z 184 is used as an exemplary case for the illustration. (c) Correlations between the relative intensities of fragment species and calculated energies that require the bond dissociation and desorption of the given fragment species from the surface where their geometries are based on the first orientation (left panel) or the second orientation (right panel). A trendline with the R2 value for the linear fit (a red dashed line) is shown as a guidance in both orientations.

FIG. 3.

(a) Bond dissociation energies for the formation of PSM fragments of interest in the gas phase. (b) Representative snapshots with two orientations that explain how desorption energy (E) between the fragments of interest and the Au (111) surface was obtained. Cyan, red, white, and yellow atoms represent the carbon, oxygen, hydrogen, and gold, respectively. Black atoms are the ones with two identical periodic images whose primary unit cell is marked with a blue box. The fragment m/z 184 is used as an exemplary case for the illustration. (c) Correlations between the relative intensities of fragment species and calculated energies that require the bond dissociation and desorption of the given fragment species from the surface where their geometries are based on the first orientation (left panel) or the second orientation (right panel). A trendline with the R2 value for the linear fit (a red dashed line) is shown as a guidance in both orientations.

Close modal

In this study, we chose to focus on lipid orientation, relative bond strength, and the adsorption energy/energy of interaction of fragments. However, it is worth noting that other components are not accounted for in this model. The mechanism of SIMS involves a collision cascade, which is a process related to the fractional surface concentration θm defined in the fundamental SIMS equation (1):

Im=IpYmαθmη,
(1)

where Im is the secondary ion current of species m, Ip is the primary ion flux, Ym is the sputter yield, α is the ionization probability, θm is the fractional concentration of m in the surface layer, and η is the transmission of the analysis system.

The impact of a primary ion on a deposited lipid results in fragment generation throughout the film, but fragments are only detected if they escape from the surface. Fragment attenuation can occur related to the structure of the substrate altering relative ion intensities.51 The model we use does not consider this, primarily because we were considering a molecular monolayer of PSM. Second, ion competition can occur between charged species, which may alter fragmentation patterns. Finally, charging effects from the surface may alter the relative abundance of fragment species observed.52 However, we observed that the combination of bond energy and adsorption energy can be used to predict the relative abundance of positively charged PC headgroup fragments in SIMS analysis of PSM on gold.

This study is the first example where molecular dynamics, in combination with density functional theory, have been applied to understand substrate effects on the in-source fragmentation that occurs in secondary ion mass spectrometry of lipids. Using this approach, we found that bond energy calculations alone cannot be used to determine the intensity of fragment species in PSM. However, combining desorption energy with bond energy can be used to predict the relative abundance, and, therefore, ionization probability, of positively charged PSM fragments in SIMS analysis—notably the characteristic PC headgroup fragments C3H8N, C5H12N, C5H14NO, and C5H15PNO4. Our intention is to build on this proof-of-concept study by examining if atomistic modeling can be used to predict the ionization potential in more complex samples with increased biological relevance, such as mixed molecular lipid and organic acid films, and with the addition of salt.

This work described in this paper is part of the m/q initiative at PNNL. It was conducted under the Laboratory Directed Research and Development Program at PNNL, a multiprogram national laboratory operated by Battelle for the U.S. DOE. A portion of this research was performed on a project award (https://doi.org/10.46936/staf.proj.2020.51782/60000315) from the Environmental Molecular Sciences Laboratory (EMSL), a DOE Office of Science User Facility sponsored by the Biological and Environmental Research program under Contract No. DE-AC05-76RL01830. Also, a portion of research was performed using the Molecular Sciences Computing Facility (MSCF) in EMSL and using resources available through Research Computing at PNNL.

The authors have no conflicts to disclose.

Ethics approval is not required.

Michael J. Taylor: Data curation (equal); Investigation (lead); Writing – original draft (lead). Hoshin Kim: Data curation (equal); Formal analysis (equal); Methodology (equal); Software (lead); Writing – review & editing (equal). William Kew: Methodology (equal); Validation (equal); Writing – review & editing (equal). Amity Andersen: Conceptualization (equal); Methodology (equal); Software (equal); Writing – review & editing (equal). Arunima Bhattacharjee: Formal analysis (equal); Methodology (equal); Writing – review & editing (supporting). Mark H. Engelhard: Formal analysis (equal); Methodology (equal); Writing – review & editing (supporting). Christopher R. Anderton: Conceptualization (lead); Funding acquisition (lead); Supervision (lead); Writing – original draft (supporting); Writing – review & editing (lead).

The raw and processed data that support the findings of this study are openly available in figshare.

1.
A. R.
Buchberger
,
K.
DeLaney
,
J.
Johnson
, and
L.
Li
,
Anal. Chem.
90
,
240
(
2018
).
2.
Z.
Postawa
,
M.
Kanski
,
D.
Maciazek
,
R. J.
Paruch
, and
B. J.
Garrison
,
Surf. Interface Anal.
46
,
3
(
2014
).
3.
T. B.
Angerer
 et al,
Anal. Chem.
91
,
15073
(
2019
).
4.
H.
Tian
,
D.
Maciążek
,
Z.
Postawa
,
B. J.
Garrison
, and
N.
Winograd
,
J. Am. Soc. Mass Spectrom.
27
,
1476
(
2016
).
5.
H.
Tian
,
A.
Wucher
, and
N.
Winograd
,
J. Am. Soc. Mass Spectrom.
27
,
2014
(
2016
).
6.
L.
Ridder
,
J. J. J.
van der Hooft
, and
S.
Verhoeven
,
Mass Spectrom.
3
,
S0033
(
2014
).
7.
I.
Blaženović
 et al,
J. Cheminform.
9
,
32
(
2017
).
8.
Y.-P.
Kim
 et al,
Anal. Chem.
79
,
1377
(
2007
).
9.
R. A.
Latour
,
Colloids Surf. B: Biointerfaces
124
,
25
(
2014
).
10.
J. E.
Baio
 et al,
Langmuir
28
,
2107
(
2012
).
11.
I.-H.
Cho
,
J.-W.
Park
,
T. G.
Lee
,
H.
Lee
, and
S.-H.
Paek
,
Analyst
136
,
1412
(
2011
).
12.
Y.-T. R.
Lau
,
L.-T.
Weng
,
K.-M.
Ng
, and
C.-M.
Chan
,
Appl. Surf. Sci.
255
,
1001
(
2008
).
13.
K.
Leufgen
,
M.
Mutter
,
H.
Vogel
, and
W.
Szymczak
,
J. Am. Chem. Soc.
125
,
8911
(
2003
).
14.
Y.
Matsuoka
,
D.
Taguchi
,
T.
Manaka
, and
M.
Iwamoto
,
Thin Solid Films
554
,
8
(
2014
).
15.
A. P.
Fellows
,
M. T. L.
Casford
, and
P. B.
Davies
,
AIP Adv.
11
,
045119
(
2021
).
16.
G. J.
Simpson
,
S. G.
Westerbuhr
, and
K. L.
Rowlen
,
Anal. Chem.
72
,
887
(
2000
).
17.
T.
Weidner
and
D. G.
Castner
,
Phys. Chem. Chem. Phys.
15
,
12516
(
2013
).
18.
P.
Joyce
,
I.
Kempson
, and
C. A.
Prestidge
,
Langmuir
31
,
10198
(
2015
).
19.
P.
Wright
,
A.
Alex
,
T.
Nyaruwata
,
T.
Parsons
, and
F.
Pullen
,
Rapid Commun. Mass Spectrom.
24
,
1025
(
2010
).
20.
A.
Torrisi
,
N.
Tuccitto
, and
A.
Licciardello
,
Int. J. Mass Spectrom.
281
,
157
(
2009
).
21.
S.
Jo
,
T.
Kim
,
V. G.
Iyer
, and
W.
Im
,
J. Comput. Chem.
29
,
1859
(
2008
).
22.
J. B.
Klauda
 et al,
J. Phys. Chem. B
114
,
7830
(
2010
).
23.
I. L.
Geada
,
H.
Ramezani-Dakhel
,
T.
Jamil
,
M.
Sulpizi
, and
H.
Heinz
,
Nat. Commun.
9
,
716
(
2018
).
24.
materials studio version 5.5.3, Accelrys, Inc., 2010.
25.
L.
Martínez
,
R.
Andrade
,
E. G.
Birgin
, and
J. M.
Martínez
,
J. Comput. Chem.
30
,
2157
(
2009
).
26.
S.
Plimpton
,
J. Comput. Phys.
117
,
1
(
1995
).
27.
S.
Plimpton
,
S.
Plimpton
,
R.
Pollock
, and
M.
Stevens
, “
Particle-mesh Ewald and rRESPA for parallel molecular dynamics simulations
,”
Proceedings of the Eighth SIAM Conference on Parallel Processing for Scientific Computing
, Minneapolis, MN, 14–17 March 1997 (
1997
).
28.
J.-P.
Ryckaert
,
G.
Ciccotti
, and
H. J. C.
Berendsen
,
J. Comput. Phys.
23
,
327
(
1977
).
29.
W.
Humphrey
,
A.
Dalke
, and
K.
Schulten
,
J. Mol. Graph.
14
,
33
(
1996
).
30.
E.
Aprà
 et al,
J. Chem. Phys.
152
,
184102
(
2020
).
31.
Y.
Zhao
and
D. G.
Truhlar
,
Theor. Chem. Acc.
120
,
215
(
2008
).
32.
W. J.
Hehre
,
R.
Ditchfield
, and
J. A.
Pople
,
J. Chem. Phys.
56
,
2257
(
1972
).
33.
P. C.
Hariharan
and
J. A.
Pople
,
Theor. Chim. Acta
28
,
213
(
1973
).
34.
K.
Momma
and
F.
Izumi
,
J. Appl. Crystallogr.
44
,
1272
(
2011
).
35.
P.
Giannozzi
 et al,
J. Chem. Phys.
152
,
154105
(
2020
).
36.
P.
Giannozzi
 et al,
J. Phys.: Condens. Matter
21
,
395502
(
2009
).
37.
J. P.
Perdew
,
K.
Burke
, and
M.
Ernzerhof
,
Phys. Rev. Lett.
80
,
891
(
1998
).
38.
J. P.
Perdew
,
K.
Burke
, and
M.
Ernzerhof
,
Phys. Rev. Lett.
77
,
3865
(
1996
).
39.
S.
Grimme
,
S.
Ehrlich
, and
L.
Goerigk
,
J. Comput. Chem.
32
,
1456
(
2011
).
40.
S.
Grimme
,
J. Comput. Chem.
25
,
1463
(
2004
).
41.
D. R.
Hamann
,
Phys. Rev. B
95
,
239906
(
2017
).
42.
P. E.
Blöchl
,
Phys. Rev. B
50
,
17953
(
1994
).
43.
A.
Dal Corso
,
Comput. Mater. Sci.
95
,
337
(
2014
).
44.
H. J.
Monkhorst
and
J. D.
Pack
,
Phys. Rev. B
13
,
5188
(
1976
).
46.
J.
Hansen
 et al,
Sci. Adv.
8
,
4965
(
2022
).
47.
M. K.
Passarelli
and
N.
Winograd
,
Biochim. Biophys. Acta Mol. Cell Biol. Lipids
1811
,
976
(
2011
).
48.
M. L.
Kraft
,
Front. Cell Dev. Biol.
4
,
154
(
2017
).
49.
50.
H.
Horai
 et al,
J. Mass Spectrom.
45
,
703
(
2010
).
51.
J. L. S.
Lee
,
I. S.
Gilmore
,
M. P.
Seah
, and
I. W.
Fletcher
,
J. Am. Soc. Mass Spectrom.
22
,
1718
(
2011
).
52.
Z.
Wang
,
K.
Jin
,
Y.
Zhang
,
F.
Wang
, and
Z.
Zhu
,
Surf. Interface Anal.
46
,
257
(
2014
).
53.
See the supplementary material at https://www.scitation.org/doi/suppl/10.1116/6.0002298 for additional tables and figures.

Supplementary Material