Chapter 1: An Introduction to MALDI Ionization Mechanisms for Users of Mass Spectrometry Imaging

Matrix-Assisted Laser Desorption/Ionization (MALDI) was a rather surprising phenomenon when it appeared in the late 1980s, prompting immediate questions about the underlying mechanisms. The topic continues to generate significant interest and controversy more than 30 years later. In particular, generation of the first (primary) ions remains hotly debated. Reviews^1,2  can be consulted for a more detailed discussion, this contribution is intended to give a brief overview, with an emphasis on those aspects which are particularly relevant for MALDI mass spectrometry imaging (MALDI-MSI) applications.

Starting with a solid sample and ending with free ions, MALDI is a convolution of a series of chemical and physical processes. This makes it considerably more complicated than some mass spectrometry ionization techniques. Compare, for example, electron impact, in which a gas phase analyte is ionized in an isolated, well-defined event. The “D” (desorption) and “I” (ionization) aspects are not independent in MALDI, it is necessary to consider them together. The laser interaction with the solid sample determines not only what ions are initially created, but sets the conditions for the subsequent sample ejection and expansion, or so-called plume. Early ions continue to react in the plume as long as collisions occur, which is orders of magnitude longer than the laser pulse. At the end of the process, the observed mass spectra rarely, if ever, reflect the true composition of the original sample. Much useful information is obtained, of course, but it is important to keep in mind the ways in which the results reflect the processes inherent in the method.

We begin with the desorption/ablation event, then consider primary and secondary ionization reactions, which include suppression phenomena with considerable importance for MS imaging.

As is apparent from the “desorption” in the MALDI name, the early picture of the role of the laser was to gently warm the surface such that some molecules would smoothly evaporate. This was supported by the observation, which made MALDI so useful, that large, fragile molecules were found in the mass spectrum. Since then, much more has been learned about this aspect of MALDI,^3,4  and it has become clear that the general term “ablation” is more appropriate, since the event is not necessarily very smooth. This has been directly visualized by, for example, Leisner et al.⁵  and others⁴  (Figure 1.1).

The characteristics of the plume vary considerably with laser pulse duration and energy. This contributes to the substantial changes in mass spectral quality and content vs laser intensity that all users are familiar with. Molecular dynamics simulations of MALDI^7–9  provide detailed insight as well as visualization of the differences between long (several nanosecond), low energy (∼10 mJ cm⁻²) pulses and short (pico- or femtosecond), powerful (100s of mJ cm⁻²) ones. The former are not so different from the original picture of soft desorption, but do not eject much analytically useful material. The latter cause large scale material disintegration (phase explosion) with creation of big pieces of condensed material (up to macroscopic) that may be too cold to fully vaporize later. Such an event is also of limited analytical value since the dense plume hinders ions from being properly collected and focused in the mass spectrometer.¹⁰  Also, one or a few shots may ablate most of the available material, but provide proportionately few ion counts.

These effects have consequences for spatial resolution in imaging. Even if the optical focus is the same, a weaker laser pulse typically ablates a spatially smaller region than a more powerful one. In some instruments this is largely due to the beam profile, which may be gaussian or similar peaked function. In that case, only the center of the illuminated region will ablate substantially. This is obviously not the case for top-hat or other more homogeneous profiles. There is also a significant mechanical component. A deeper and more violent ablation event will affect a considerable region around the illuminated region, blasting some of it away. For these reasons, lower laser fluences may be preferable for achieving the best spatial resolution.

Because considerable material is heated above the boiling or sublimation point in a few nanoseconds (for example by a Nd:YAG or nitrogen laser), the initial pressure in the ablation plume may reach tens of atmospheres. This is not gentle evaporation from the surface. The pressurized mix of gas and condensed material does not expand uniformly. The top layers are energized most by the laser and are not hindered in their expansion. They reach high velocities, around 1000 m s⁻¹. Deeper layers are cooler (the laser is attenuated by the upper layers) and may be ejected with velocities as low as 100 m s⁻¹.¹¹  This velocity pattern means that there is little downstream mixing of sample layers in the plume, as long as it can expand freely. Typical peak plume temperatures are calculated to reach over 1000 K, consistent with previous experiments.¹² 

If the ablation is performed in a background gas, rather than in a vacuum, the plume cannot expand as much before it stagnates and mixes with the background. However, given that the starting conditions are much higher in pressure and density than even ambient air, a significant portion of the expansion is essentially identical in all MALDI events.

The rather harsh initial plume conditions decay with the plume expansion, which is adiabatic (insignificant heat flow to/from the surroundings). This is an important fact, since it leads to a MALDI spot size effect: even at the same fluence (laser energy per area), changing the laser focus leads to different results.^13–16  This has recently been investigated in oversampled imaging, as well.^17,18  In particular, a higher fluence is typically needed to obtain a good signal from a smaller spot. In MALDI imaging this tradeoff can be important, since higher spatial resolution is often desired. The effect arises because the plume expands not only perpendicular to the irradiated surface, but also laterally. For a small spot, lateral expansion is somewhat faster than for a wide spot.¹⁹  This fast radial expansion means less time for ion–molecule reactions in the plume, so there are fewer analyte ions created.

The high pressures and temperatures in the early plume imply many intermolecular collisions, raising the question whether the material in the plume reaches or approaches thermal equilibrium. Some models of MALDI ionization are based on this assumption.^20–22  As noted above, the plume is in fact very inhomogeneous, especially perpendicular to the surface, and changes considerably with time. This means that at most a transient local thermal equilibrium (LTE) might be established. This is most likely during the early hot, dense period, since forward and backward reaction rates are highest. As the plume expands and cools, equilibria must shift, but reaction rates also decrease, so even if LTE is lost at some point, relative ion populations may remain near their values from earlier times. The topic of LTE in the plume is not easily explored experimentally because it is very difficult to sample the full plume, not least due to the wide velocity range noted above. In nearly all such studies, the acquired MALDI mass spectra are a narrow slice out of the whole picture, typically of the fastest, topmost ions, which are the easiest to collect.

As is not surprising from the above considerations, some data are consistent with LTE, others are not, and some show a transition between the two situations. For example, it was found in one case that the matrix : analyte ion ratios corresponded to the thermodynamically expected equilibria for a tripeptide in two different matrices.²³  However, this required a relatively high laser fluence. At low fluence other ratios were observed. This is consistent with the idea that plume LTE needs a sufficiently long period of high temperature and density, along with a large quantity of matrix primary ions vs. the quantity of analyte. Other examples of such thermodynamically predicted ratios have been reported.^24,25  As will be discussed below, models incorporating full plume physics and chemical kinetics also exist and are a superset of those that assume equilibrium.

Primary ionization is the formation (or separation, in the case of “preformed” ion pairs) of the first ions in a MALDI event. These may or may not react further in secondary reactions with neutral molecules or with other ions (especially neutralization with opposite charged ions). Despite many years of investigation, it has proven difficult to definitively elucidate primary ionization mechanisms in MALDI. It is possible or even probable that multiple mechanisms exist, and more than one may be active for a given matrix/analyte/laser combination.

A difficulty which has repeatedly arisen in studies of MALDI mechanisms is devising experiments and methods which can clearly differentiate between primary ionization mechanisms. This is largely a consequence of the fact that measured MALDI results are never purely dependent on primary ionization, but also on the complex ablation event and secondary reactions. Results which might initially seem simple and clear tend to be far less conclusive on closer inspection. In the discussion below we will briefly mention a few of the complications that occur.

An early proposal for primary ionizations is the “Lucky Survivor” preformed ion model of Karas et al.^26,27  Free ions of matrix and analyte obviously exist in the preparation solution, and the model suggests that they remain free in the dried sample. They are then simply released by ablation. Most may not be set free, but some escape – the lucky survivors. A nice demonstration of the existence of candidate ions in the solid state was provided by using pH indicator dyes as analytes in MALDI preparations. The dried matrix crystals retain the colors of the solution.²⁸ 

However, unpaired ions are unfavorable in solid materials, and ions in the dried sample are expected to be stoichiometrically coordinated with an oppositely-charged counterion. But releasing an ion pair in the ablation is not helpful, since it has no net charge. Separating these ion pairs is not trivial, they are bound by strong electrostatic forces. In addition, the model also postulates most ions are entrained in neutral chunks. Although these certainly evolve and release some ions,^29,30  simulations and data suggest that matrix clusters and chunks mostly remain intact,^6,11  so few free ions are released.

One can certainly imagine that the Lucky Survivors mechanism contributes some ions in MALDI. After all, salt crystals yield ions by laser ablation. But it makes no quantitative predictions and seems to rely on processes of very low ion yield.

Rather than separating preformed ions as in the previous model, charge separation by mechanical fracturing of matrix crystals (triboelectricity) is a possibility. This occurs due to the unequal free energies of the newly formed surfaces. Further physical separation occurs as the crystal fragments fly apart. However, triboelectricity has not been found to be necessary for an effective MALDI matrix, so this role of this mechanism seems limited in traditional MALDI. It may be much more relevant in “laserspray” or “inlet” methods, in which crystal fracture during rapid sublimation has been associated with ion production.³¹ 

Other surface-based mechanisms involve ion fractionation in ablated droplets. Ions of one polarity may be enriched on the surface compared to the interior. Final mechanical ion separation occurs by bursting of bubbles or other mechanical instabilities of the droplets.³²  The yield of such processes is quite low, and the ions are again released largely in clusters and chunks of neutrals.³³ 

Ions can be created from neutrals by heat, for example by electron emission from molecules, atoms and surfaces. Some early mass spectrometers used hot wires as ion sources. Gaseous organic molecules were ionized by thermally activated electron transfer to the wire. While a recent miniaturized hot wire source has shown some limited results in MALDI-like applications,³⁴  this method fails for many biomolecules which are analyzed by MALDI, because they fragment too easily at higher temperature. In the late 90s it was then proposed that in the hot early MALDI plume the matrix itself could be acting as a solvent to facilitate ion separation. Since most, but not all, matrix materials are polar molecules, this was called the Polar Fluid Model (PFM).^35,36  There continues to be significant interest in the concept.^20–22 

In the PFM the hot matrix fluid reduces the energetic requirements for charge separation by solvating the nascent ions or ion pairs, just as in a beaker of water. Instead of 500–700 kJ mol⁻¹ to separate matrix ion pairs (for example 2M → M⁺ and M⁻, or 2M → MH⁺ + (M–H)⁻), the energy required could be reduced to 100–200 kJ mol⁻¹. If these reactions approach equilibrium in the hot plume, this would generate enough ions to be easily measurable.

More detailed theoretical investigation shows that there are fundamental reasons why this model is very unlikely to be relevant for MALDI.³⁷  For example, at the temperatures and pressures in the early plume, organic substances have low dielectric constants.^37,38  Even the dielectric constant of water is near 1 under such conditions³⁹  vs. 80 at lab conditions. This means ion solvation is nearly negligible.

Comparison of measured MALDI results with thermal predictions has also shown poor correlation in several cases.^40–42  Other results were initially interpreted as evidence for thermal mechanisms, but later shown to be equally compatible with non-thermal models.^43,44  Again, secondary reactions in the plume complicated the interpretation of these data.

MALDI can be performed with a wide range of laser wavelengths, but here we consider only ultraviolet excitation as it is most widespread. The laser energy must be efficiently deposited, so the absorption spectrum of the matrix in the solid state is a key factor in determining MALDI efficacy.^45–48  This spectrum is typically much broader and shifted to the red compared to the spectrum in solution. This reflects interactions between neighboring matrix molecules and delocalized electronic states. The energy needed to ionize matrix molecules is on the order of a few (2–3) quanta of typical ultraviolet MALDI laser light (337 to 355 nm). These facts suggest ionization mechanisms involving a small number of electronically excited matrix ions.

A model based on electronic excitation has been proposed for some matrices, involving migration of excitations in the matrix solid and pooling reactions of two excitations.¹⁹  The model also includes secondary ionization, as well as full plume physics. It is therefore called the Coupled Chemical and Physical Dynamics (CPCD) model of MALDI.⁴⁹ This remains the only comprehensive model of MALDI to date.

There is considerable evidence for exciton-based mechanisms in some matrix materials, such as dihydroxybenzoic acids (DHB).⁵⁰  Based on the photophysical properties relevant for this mechanism, the CPCD has correctly predicted the relative MALDI performance of these isomers.⁴²  Previously there was no explanation for the fact that they do not perform equally well. These results were also not compatible with thermal models.

The CPCD model has also successfully reproduced extensive data for wavelength- and fluence-dependent performance of DHB, alpha-cyano-4-hydroxycinnamic acid (CHCA), and alpha-cyano-4-chlorocinnamic acid (ClCCA) matrices.^40,41,51,52  The laser wavelength studies led to an important result: Rather than being fluence dependent,⁵³  MALDI ion yield is actually a function of the number of laser photons deposited. At a constant wavelength, these are functionally equivalent, but the underlying mechanisms are apparently quantized. This is also reproduced by the CPCD.

A phenomenon which can be particularly relevant for MALDI imaging involves the sample support. If the sample is either sufficiently thin that some laser light reaches the support (a few 100 nm), or the sample is ablated down to the support, the support material can significantly affect the ion yield.^54,55  It was found that a stainless-steel support led to dramatically better signal than a gold one, for several matrices. This effect was proposed to involve interaction of the matrix excited states with the conduction band of the metal, leading to more efficient matrix ionization. It is useful to be aware of this effect since gold supports may otherwise be preferred due to their chemical inertness. The effect can also be taken advantage of, since the surface enhancement for a thin sample can be substantial, giving much better signal than a thicker sample. However, the sample does need to be quite thin. In the experiments noted above, this was achieved by rapidly moving an electrospray sample applicator over the substrate. There may be a physically similar enhancement when using thin (sputtered) metal over coatings on a MALDI sample (to suppress surface charging), but it appears to be less dramatic.

Regardless of the mechanism by which the initial ions are created, only a few types of ions are typically observed. These are defined by the charge carrier which is involved. If electrons are transferred, radical cations and anions are formed. Considering a sample of pure matrix:

In most common matrix molecules these are radical ions.

Alternatively, a proton disproportionation might be involved:

Proton transfer ions tend to be energetically preferred for many of the classical matrix materials (DHB, SA, CHCA, etc) which have acidic and basic sites. However, electron transfer ions can be dominant for other matrices (DCTB, etc⁵⁶ ), and both types may appear in the same spectrum. Which ones are dominant depends on the relative ionization energies, electron and proton affinities, and basicities of the species involved. For many matrices, interconversion reactions (between electron and proton transfer ion pairs) are rather low in energy, a few tens of kJ mol⁻¹, and are therefore facile in the early plume.

Metal adduct ions are common in positive ion mode, since the commercial matrix (or the sample) often contains considerable salt, and can be enhanced by addition of metal salts (for example, silver ions are particularly useful for some synthetic polymers). The larger the cation, the lower the binding energy with the matrix or analyte. Protons (H⁺) bind the strongest by far, with affinities up to 900 kJ mol⁻¹. Sodium affinities are around 150 kJ mol⁻¹, for K⁺ they are near 100 kJ mol⁻¹. Metal adducts may form directly from the corresponding matrix salt, or by interconversion reactions. Anion adducts are rare because anions are large and diffuse, leading to low binding energies with neutral molecules.

Assuming the laser energy is directly absorbed, corresponding electron or proton transfer ions of analytes can also be formed, but the matrix is typically in excess and most primary ions will be of the types mentioned above.

Reactions of the primary ions with neutrals (or other ions) in the plume are probably the most important process for MALDI imaging. Regardless of how primary ions are formed, and which ions they are, they are typically embedded in a plume composed mostly of neutral matrix molecules and clusters. Secondary reactions involving matrix species are therefore typically the most important.

One of the most straightforward methods of investigating these reactions involves changing the relative amounts of matrix and analyte. It was observed that increasing the analyte mole fraction in MALDI preparations could lead to a strong decrease, or even complete disappearance, of matrix ions in the mass spectrum.^57,58  This has been named the Matrix Suppression Effect (MSE). The MSE is not only dependent on the matrix : analyte (M : A) ratio, but also on laser fluence. At a given M : A, there is less MSE at higher fluence.⁵⁸  There is also an effect of analyte molecular weight – larger analytes cause MSE at lower M : A.⁵⁹  Finally, MSE also depends on the laser spot size – it requires less analyte for a larger spot.⁴⁹ 

The MSE was initially surprising, but is readily understood in terms of ion–molecule reactions in the expanding plume. As noted above, the number of initially generated primary ions is limited by the number of laser photons, so stronger pulses generate more. The sample is ablated, and the ion/neutral mixture begins to expand. Also, as noted above, the initial state of this fluid is dense and hot, so collisions are frequent and there is sufficient activation energy for many reactions. The reaction of matrix ions with analytes is normally energetically favorable (or else the matrix is inappropriate for the analyte), so reactions like the following begin to create analyte ions:

Similar reactions can occur for electron or cation transfer. This reaction may be kinetically limited or move toward LTE depending on the reaction activation energy, local plume conditions, and the evolution of the local plume over time.^23–25  The net ion mole fraction is not high, perhaps in the order of 10⁻⁵, maybe as low as 10⁻⁹.¹⁸  It is higher early in the plume before recombination of ion pairs reduces it,⁶⁰  but the mole fraction of analyte for MSE may be 1–10%. Therefore, the above reaction is pushed far to the right. With enough analyte, matrix ions are so depleted that they become negligible in the mass spectrum. On the other hand, if more MH⁺ is provided (higher laser fluence), more neutral analyte is needed for full depletion. Of course, matrix ion depletion and MSE occur in negative polarity as well.⁵⁹ 

The MSE is reproduced in the CPCD model because it includes the full forward and reverse kinetics of such reactions, and how they are modulated by the plume density and temperature. The reaction rates use a linear free energy approximation: the activation energy, and hence rate constant, is assumed to be inversely related to the reaction free energy (a more negative ΔG implies a lower activation energy), as is common for many organic reaction mechanisms. The extent of the reaction is affected by how the speed of the plume expansion compares to the secondary reaction rates. This in turn depends on the laser spot size, as noted above. In imaging, a small spot is often preferred since spatial resolution is so important. Unfortunately, this means faster expansion and decreased analyte ion formation. It may sometimes be possible to qualitatively evaluate this by the degree of MSE observed.

The CPCD includes ions of both polarities, which make it possible to examine how ion–molecule reactions in one polarity affect those in the other. This is an aspect which is seldom considered, since only one polarity is observed in the mass spectrum. They are connected via neutrals, especially analyte, which can also be a limiting reagent. For example, if the above reaction to form a protonated analyte is energetically highly favorable (analyte more basic than the matrix), the corresponding reaction in the opposite polarity is probably less favorable:

This means the positive matrix ions will react faster with neutral analyte and consume more of it. This reduces the negative analyte yield compared to what it would be otherwise. The ratio of positive to negative ions in the corresponding mass spectra will then not be one, for either matrix or analyte.⁶¹  To put it another way, if the analyte signal is poor, it might be useful to look at the opposite polarity. The reaction thermodynamics and kinetics might actually favor those products, even if the original polarity “should” give good results.

An important MSE observation is that all ion types are affected at the same time. For example, matrix ions of different types often appear in positive mode. Protonated species are often dominant, but radical cations and alkali adducts are frequently also strong. Adding enough analyte to induce MSE will suppress all of these ions, even if the analyte only appears as, for example, AH⁺. This is again a consequence of extensive secondary ion molecule reactions, which allow the system to proceed toward the kinetically and thermodynamically most favorable products. It implies electron transfer reactions to interconvert the ion types.

The key condition for MSE is that enough analyte is present to consume all the available primary ions. This must happen before the plume has expanded enough that reactions become too slow. This kinetic aspect is the reason why larger analytes induce MSE at lower concentration than smaller ones. An analyte of 10 000 Da is physically much larger than one of 100 Da, with a correspondingly larger collision cross section. It is therefore more efficient in reaction with primary ions.

We are now in a position to consider the situation, which is nearly always encountered in MALDI, and especially in imaging: more than one analyte. It should be clear from the discussion of the MSE that ion–molecule reactions with all analytes will occur in parallel, which causes the analyte signals to be coupled. Taking the example of proton transfer reactions again, the matrix–analyte reactions with analytes A and B are:

In addition, direct analyte–analyte reactions can occur:

Because the matrix is in excess, even under MSE conditions, A and B are most strongly coupled by the reversibility of the matrix reactions:

giving a net reaction:

As for the MSE, it should not be forgotten that the opposite polarity reactions are also taking place, affecting the results in the observed polarity. The most important consequence of these reactions is that it is difficult to obtain a mass spectrum which even approximately reflects the composition of the sample. In the extreme case, called the Analyte Suppression Effect (ASE), some analytes can completely suppress others.⁶² 

The conditions for ASE are, not surprisingly, similar to those for MSE: sufficient analyte concentration, both compared to primary ions (laser fluence) and to other analytes. The analyte which is favored is the one which forms the energetically most favored ions. In the examples above, it would be the analyte with the highest proton affinity. Of course, this assumes that the reaction kinetics of the two analytes are not very different. The only case where a general kinetic limitation might arise is for electron transfer reactions, in the so-called Marcus inverted region. At low ΔG, rates of electron transfer reactions increase with ΔG, but can slow at high ΔG due to poor overlap of vibronic wavefunctions.⁶³ 

Both ASE and MSE are illustrated in Figure 1.2 by a series of measurements of an equimolar mixture of 5 analytes.⁶⁴  The M : A ratio was changed over 4 orders of magnitude. At the highest analyte concentration (lowest M : A mole ratio, 77), the thermodynamically most favored species is dominant in the spectrum, while only very weak signals of 2 other analytes were observed. No matrix ions are visible. At higher M : A = 690, the other analytes increase in relative intensity, though the least favorable species is still not visible. Matrix ions begin to appear only at M : A = 6900, as does the last component of the mixture. Even at this dilution, relative intensities in the spectrum must be interpreted with caution.

Both MSE and ASE were investigated by MALDI imaging of dried droplet samples.⁶⁵  As seen in Figure 1.3, these results illustrate how sample inhomogeneity at the scale of the pixels, 100 μm, is reflected in the suppression effects. Even when the M : A ratio in the preparation solution is low enough, not all pixels exhibit suppression, though most do. The inverse can also occur, some pixels may exhibit suppression even if it is not typical in the sample. The best results might then be obtained by filtering pixels (or spectra summed in one pixel) based on the relative magnitudes of matrix peaks, since these give some indication of suppression. Of course, a different, more uniform, sample preparation (other than dried drop) would likely provide a far more consistent result.

Strong ASE is a common problem in MALDI imaging of lipids. Boskamp and Soltwisch investigated suppression of different glycerophospholipid classes in both positive and negative modes.⁶⁶  In positive-ion mode, phosphatidylcholines (PC) suppressed phosphatidylethanolamines (PE) or phosphatidylserines (PS), in a concentration-dependent manner. As could be expected from the considerations noted above, in negative polarity the situation was reversed, and PC was difficult to observe. Interestingly, a cross-polarity effect was also observed, the presence of PC enhanced the signal of other lipids in negative mode. This was attributed to the reactions leading to PC positive ions.

Methods to control suppression effects generally involve modified sample preparations. An example is ensuring that the matrix and analyte are not uniformly co-crystallized, by applying them separately.⁶⁷  This is roughly the situation in MALDI tissue imaging, but works against sensitivity, which is also an issue in imaging. Another option is derivatization by applying an appropriate reagent on top of the rest of the preparation.⁶⁸  This changes the chemical nature of some sample components, with corresponding modification of the suppression patterns. Alternatively, expected sample constituents can be spiked either to act as reference suppressors,⁶⁹  or as internal standards.⁷⁰ 

Tissue sections may contain considerable salt, which reduces sensitivity. Desalting⁷¹  may restore signal and thereby ameliorate nonspecific suppression effects, but does not decrease analyte–analyte suppression. In an interesting reversal of strategy, Popkova and Schiller proposed saturating the matrix solution with cesium iodide.⁷²  Most of the ions in the resulting spectra were Cs⁺ adducts, which gave better lipid coverage. ASE is much reduced because analyte affinities for large alkali cations are low. As an indirect result, there are small absolute differences in the reaction thermodynamics for cationization of the various analytes, reducing competition between analytes.

An important recent extension of MALDI is the so-called MALDI-2 technique.⁷³  Earlier efforts to increase sensitivity by “post-ionization” of neutral molecules in the MALDI plume met with limited success. The key insight in the MALDI-2 method was to introduce a few hundred Pa of buffer gas to slow and compress the plume. Crossing the compressed plume with a time-delayed second laser about 0.5 mm above the sample was found to increase signals of many analytes by up to 2 orders of magnitude. In the context of imaging, this obviously makes small ablation spot sizes and high spatial resolution much more feasible.

As shown by systematic variation of parameters,⁷⁴  much of the enhanced signal is generated in a primary + secondary reaction cascade (denoted Mechanism 1), similar, but not identical to normal MALDI. Neutral matrix molecules (not large aggregates) are 2-photon ionized, generating radical primary ions. These react in the compressed plume to form proton transfer matrix species, which continue to react with analytes as collisions allow. This mechanism requires the second laser to be of sufficiently short wavelength that the energy of two photons is above the matrix ionization potential. This is typically shorter than the wavelength of the ablation laser. A Mechanism 2 is also proposed to also be active, involving conventional MALDI of the clusters and aggregates entrained in the plume. This mechanism does not require a short wavelength second laser. Finally, some analytes may be directly 2-photon ionized by the second laser, if they have suitable chromophores.

Given that very similar ion–molecule reactions are involved, one could imagine that suppression effects are also an issue in MALDI-2, just as in MALDI. However, even though it is compressed by the background gas, the MALDI-2 plume is still much less dense than the early stages of the MALDI expansion. This has the advantage that secondary reactions have less opportunity to proceed toward LTE. Also, the method is based on greatly increasing the number of primary ions, which also prevents suppression effects in normal MALDI. Consistent with these considerations, one study has found no correlation between analyte proton affinity and MALDI-2 signal, in contrast to normal MALDI.⁷⁵  This in turn means reduced ASE, as was found in images of pharmaceuticals in tissues. It seems possible that the worst suppression effects in standard MALDI can be reduced or avoided using MALDI-2, but questions still remain about response factors and relative sensitivities.⁷⁶ 

Although experimentally very simple, MALDI is a superposition of multiple physical and chemical processes, and it is found to be rather complex when examined in detail. In spite of this, many important MALDI phenomenon, especially those involving suppression effects and relative sensitivities can be qualitatively and, in some cases quantitatively, understood in terms of reactions of primary matrix ions with analytes in the expanding plume. Many of these effects are particularly pronounced in MALDI imaging because tissue samples are complex and sample preparation options are limited. The framework presented here can guide experimental design, and especially aid in interpretation of results. The “M” in MALDI does not stand for “Magic”, even if many questions remain.