Ionization Mechanisms in UV-MALDI
Previous: Molecular Dynamics Table of Contents Next: Literature
6. Surface Phenomena
Metal Substrates: photoelectron emission
Metal Substrates: enhanced MALDI signal
Structured Substrates
6a. Thin Samples on Metal Substrates
MALDI samples are frequently relatively thick, in the sense that the laser is absorbed in the matrix, no light reaches the substrate. On the other hand, the sample may crystallize inhomogenously, leaving areas of thin or no sample. Or a very small amount of sample might be used, leaving a few small scattered crystals. It is also possible to intentionally create very thin samples, so that the laser reaches the substrate, or to ablate the sample until the substrate is exposed. In all these cases, it becomes important to know if and how the substrate may affect the MALDI result.
See the following papers for more complete discussion:
Anal. Chem., vol. 76, pp. 3179-3184 (2004)
J. Am. Soc. Mass Spectrom., vol. 17, pp. 737-745 (2006)
J. Phys. Chem. A, Vol. 110, No. 47, 2006, final manuscript
Interfacial electron photoemission
If the substrate is metal, the first phenomenon that one might think of in UV MALDI is one-photon photoelectric emission. However, the 337 (nitrogen laser) and 355 nm (3rd harmonic Nd:YAG) photons are too low in energy (3.7 and 3.5 eV, respectively) to cause photoemission from the usual substrate materials (work function of steel: 4.5 eV, gold 5.1 eV). While dirty surfaces have lower work functions, reductions large enough for single photon emission in MALDI have never been observed.
Some two-photon electron emission from the metal surface could be expected, but this is relatively inefficient at MALDI laser intensities. Instead, the evidence points to an interaction of the matrix with the metal. Whenever two materials are placed into physical contact, their chemical potentials determine how electronic states are affected at the interface. In the case of UV MALDI matrixes and metal subtrates, the situation is as shown in the following figure:
Matrix-surface interactions. If the matrix LUMO (after splitting due to metal interaction) is below the Fermi level, some charge transfer occurs. From the now occupied LUMO-derived state, photoionization is a two-photon process with the usual MALDI lasers. DHB = 2,5 dihydroxybenzoic acid, SA = sinapinic acid, HCCA = hydroxy-cyanocinnamic acid.J. Phys. Chem. A, Vol. 110, No. 47, 2006, final submitted manuscript
If the metal-matrix interface has the right Fermi level and LUMO combination, two-photon ionization from the interfacial matrix layer becomes possible. Note that for DHB this is expected on a steel surface, but not on a gold one. It is easy to modify the rate equation model to handle the situation of DHB on steel. It is simpler than the DHB-only case, because the Sn upper excited state is no longer relevant, as is the plume expansion (photoelectron emission will only happen during the laser pulse) See Anal. Chem., vol. 76, pp. 3179-3184 (2004). As the next figure shows, the predicted photoelectron emission vs. laser fluence is described extremely well by the matrix-metal interface model.
Electron emission data from DHB dried-drop samples on stainless steel, compared with the two-photon model prediction. Also shown is a linear fit, which would be appropriate if the emission was single-photon. Emission data from Anal. Chem. vol. 75, pp. 6063-6067 (2003) and J. Phys. Chem. A vol. 110, pp. 926-930 (2006)
Enhanced MALDI signal at the metal-matrix interface
Why does this matter? We dont care about electrons in MALDI. But we do care about matrix ionization, and that is the difference between a process which depends on matrix excitation and one which does not. In the model presented above, two-photon photoionization of matrix occurs, not two-photon electron photoemission from the metal.
The pleasant consequence of this for MALDI is that we can have substantially enhanced signal from thin samples, or thin regions of a sample. This has been directly observed by MALDI imaging, here is an example of positive ions:
Electrosprayed MALDI spots of reserpine and substance P in DHB. Spray times are indicated on the top row, which is an optical image of the spots before MS acquisition. The second image shows the sample after the MALDI measurement. The upper traces in the lower panel show the substance P signals summed over the five columns. The data for the 50-shot columns are below. The traces with crosses are the measured values. The solid line is the average value of the most uniform central region of each spot. J. Am. Soc. Mass Spectrom., vol. 17, pp. 737-745 (2006).
As is very evident, the thinner edges of the electrosprayed spots give much stronger analyte signals. The thinner the central spot, the stronger the signal as well.
But this signal enhancement should only be observed for the appropriate metal-matrix combinations. This has also been verified by MALDI imaging:
Electrosprayed DHB MALDI spots on a stainless steel substrate which has been half-coated with 35 nm of gold. The upper image is an optical micrograph before MALDI imaging. The thinnest spot is on the right; the thickest on the left (approximately 200-3000 nm mean thickness). The lower images are for the indicated protonated positive ions. Darker pixels indicate stronger signal. The substantial signal enhancement of thin sample regions on steel is apparent, as is the much weaker effect on gold.
Here is an example of the same effect in negative mode, with a different matrix:
Negative ion ((M-H)-) MALDI images of an electrosprayed SA track on a stainless steel and gold- coated steel substrate. Near the metal boundary, the spray was held stationary for 10 s to create a thicker region. The signals are considerably stonger on the thinner regions of the steel half than on gold. The thick central spot gave the weakest signal.
Finally, as predicted in the first figure in the section, HCCA matrix shows enhancement on both steel and gold:
MALDI image (MH+) of an electrosprayed HCCA track on a stainless steel substrate which has been half-coated with 40 nm gold. The thinner areas give enhanced signal on both gold and steel, by a similar amount.
6b. Structured Surfaces
In addition to simple metal sample substrates discussed above, there have been recent efforts to use structured surfaces as MALDI or LDI substrates. In addition, fine particles have long been used as MALDI and LDI substrates. This includes Tanaka's 1988 Nobel prize work using cobalt particles, as well as more recent work on silicon, graphite and other particles. Particle aggregates have a deeply modulated surface and small features. These methods will not be listed or reviewed here, rather the aim is to consider how such surfaces might affect MALDI ionization processes. Most of the following is derived from: Eur. J. Mass Spectrom. vol. 15, pp. 189–198 (2009).
Nanopillars on silicon used as a MALDI substrate. See Angew. Chem. Int. Ed. vol. 48, pp. 1669 –1672 (2009).
In the rate equation model, the plume expansion strongly modulates bimolecular processes. As collisions slow, so do reaction rates, including not only ionization and charge transfer processes, but also pooling and collisional deactivation of electronic excited states.
Because the plume expands from a small spot on the surface of the sample, it was treated as an ideal isentropic expansion (molecular beam). It is free to expand as soon as it is created, and is limited only by the background gas pressure. Note that even in atmospheric pressure MALDI, the plume is initially at much higher pressure than the background gas, so this approximation is good for all forms of MALDI, during the critical period of ion formation.
On a structured surface, the assumption of a freely expanding plume might not be valid. It is initially confined in a capillary-like space (see the above micrograph) before becoming free to expand laterally. Fortunately, the molecular beam community has long used such capillary nozzles, and their characteristics are known.
Axial Mach number vs downstream distance for ideal and capillary expansions. The horizontal axis uses the dimensionless characteristic ratio of downstream distance over the nozzle diameter.
Because the constrained expansion is forced to direct its motion along the axial direction, it accelerates faster. It is not only initially faster, it continues to accelerate more than the ideal, unconstrained, expansion, even after it has left the capillary. The consequence for the MALDI plume is that the frequency of plume collisions is different in the two cases, and evolves differently downstream. In particular, higher mach numbers in the capillary expansion mean that the plume has cooled more, and fewer collisions occur.
Matrix ion yield as a function of laser fluence, for a 5 ns, 355 nm pulse. A conventional MALDI experiment (flat planar surface, FPS) is compared with one in which the ablated plume is emitted from a deep capillary.
It is relatively straightforward to replace the conventional plume expansion in the rate equation model with the appropriate expressions for the capillary case. As seen in the figure above, the effect on matrix ions of a capillary expansion is quite positive, especially at higher laser fluence. Because the plume expands faster, the recombination losses are less, so more ions survive to be observed in the mass spectrometer.
One could hope that more surviving primary ions also leads to more secondary ions and a significantly better MALDI spectrum. The next figures show what the calculations predict:
Capillary and conventional analyte fractional ion yields for different initial analyte mole fractions, as a function of laser fluence.
This is disappointing! We hoped for more, but got distinctly less. Once again, it is a consequence of bimolecular rates. If we could make more analyte ions, they might indeed survive better in the capillary expansion. However, we need bimolecular collisions with primary ions to make them, and these are less frequent. So the net result is a loss of analyte sensitivity.
But this is for an isolated capillary. As the micrograph above illustrates, we expect to usually have dense arrays of capillaries. Do they then behave more like a normal flat sample? That depends on their spacing and diameter. If they are close, compared to their diameter, the emitted individual plumes can merge quickly. If they are far, they merge slowly. The estimated merge times are shown next:
Time to merging of adjacent expanding plumes, in nanoseconds. A merge is defined as meeting of the 50% density radii.
Already at a diameter of 3 μm, a spacing of only one diameter needs 10 ns to merge. Merge times this long, or more, mean that most MALDI ionization process have progress quite far toward their final products, so the spectrum should look much like that predicted for an isolated capillary. So one reason to use very dense surface features is to avoid sensitivity losses in the plume.
Previous: Molecular Dynamics Model Next: Literature
MALDI Ionization Tutorial © Copyright 2007-2016 Richard Knochenmuss