Investigation of the Desorption Process in UV Matrix-Assisted Laser

Publication Date (Web): December 7, 2009 ... For a more comprehensive list of citations to this article, users are encouraged to perform a search inSc...
1 downloads 0 Views 5MB Size
J. Phys. Chem. C 2010, 114, 5367–5381

5367

Investigation of the Desorption Process in UV Matrix-Assisted Laser Desorption/Ionization with a Liquid 3-Nitrobenzyl Alcohol Matrix by Photoacoustic Analysis, Fast-Flash Imaging, and UV-Laser Postionization† Andreas Rohlfing,§ Arne Leisner,‡ Franz Hillenkamp, and Klaus Dreisewerd* Westfa¨lische Wilhelms-UniVersita¨t Mu¨nster, Institute of Medical Physics and Biophysics, Robert-Koch-Str. 31, 48149 Mu¨nster, Germany ReceiVed: June 4, 2009; ReVised Manuscript ReceiVed: NoVember 6, 2009

Ultraviolet matrix-assisted laser desorption/ionization mass spectrometry (UV-MALDI-MS) is a widely used method for the analysis of a variety of biomolecules. The technique relies on the codesorption of matrix and intact analyte molecules from the condensed to the gaseous phase and subsequent analyte ionization. In spite of numerous studies, the MALDI processes are not yet fully understood. Here, we used 3-nitrobenzyl alcohol as a liquid matrix and employed three complementary analytical techniques to monitor the phase transition and development of the ejected material plume. Each of these methods provided a time resolution in the nanosecond range. (i) Photoacoustic stress waves, generated by the recoil momentum of the ejected material, were recorded simultaneously with the corresponding MALDI-generated analyte ion signals. (ii) The MALDI plume expansion was monitored by using a second illumination laser for imaging the plume in dark-field and in scattered-light geometries, allowing discrimination between ejected droplets and gaseous plume domains. (iii) Time- and space-resolved UV-laser postionization of desorbed matrix molecules complemented these studies. The photoacoustic data indicate the transition from a primarily molecular ejection according to a quasithermal model at low laser fluences to a (layer-by-layer) phase explosion above a fluence threshold of about 170 J/m2. Notably, this threshold coincides well with the onset of sizable ion generation. A few distinct droplets are detected over the whole of the investigated fluence range. Introduction Since its discovery in the 1980s, matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS1-3) has evolved into a widely used technique for the analysis of a variety of biomolecular compounds, including peptides and proteins, oligosaccharides, nucleic acids, and lipids. Lasers emitting in the ultraviolet (UV)3,4 and infrared (IR)5-7 spectral range are both employed for MALDI-MS, with the UV-MALDI variant used much more widely than its IR counterpart. MALDI ion sources can be coupled to virtually any type of mass analyzer. However, most commonly, time-of-flight (TOF) analyzers are employed since these instruments make efficient use of the pulsed nature of ion generation and allow an entire mass spectrum to be recorded for each single-laser exposure. Despite the widespread use of the MALDI-MS methodology, the physicochemical mechanisms that lead to the “soft” desorption and ionization of thermolabile biomolecules are still not entirely understood. (Although “desorption” in any more strict sense would solely represent the ejection of single molecules from a surface (layer), it has become common to use this expression to relate to the general process of material removal under MALDI conditions, which often proceeds through mechanisms of evaporation or ablation. Unless otherwise noted, we will here use the term “desorption” as equivalent to “material †

Part of the “Barbara J. Garrison Festschrift”. * Corresponding author. E-mail: [email protected]. Tel.: +49-251-8356726, Fax: +49-251-8355121. ‡ Current address: Olympus Soft Imaging Solutions GmbH, JohannKrane-Weg 39, 48149 Mu¨nster, Germany. § Current address: Sequenom GmbH, Mendelssohnstr. 15d, 22761 Hamburg, Germany.

ejection” throughout this paper.) It is generally accepted that the matrix provides several key features, including the absorption of the laser energy and the codesorption and ionization of the embedded analyte compounds. An extensive collection of experimental data has been compiled over the years,8-13 and various models describing the desorption/ionization processes have been proposed.14-19 Considering that a large variety of sample and laser conditions can be used for a successful MALDI-MS analysis, it is now assumed by most researchers that different pathways exist rather than one universal MALDI mechanism.20,21 For example, UV-MALDI uses electronic excitation of chromophoric groups, while IR-MALDI works through vibrational excitation; yet, these different mechanistic routes may produce an almost indistinguishable mass spectrum of the analyte.22 Although abundant matrix compounds have been identified over the years, a selection of solid state matrices (e.g., 2,5dihydroxybenzoic acid23 or cinnamic acid derivatives like R-cyano-4-hydroxy-cinnamic acid and sinapic acid4) is still most commonly used for UV-MALDI-MS. A few liquid compounds have also been identified as useful UV-MALDI matrices, with 3-nitrobenzyl alcohol (NBA) historically being the most prominent example.24,25 Notably, NBA is also a useful matrix for fast atom bombardment (FAB) mass spectrometry.26 The development of composite liquids27,28 as well as ionic liquid matrices29,30 has moreover found wide attention in the recent past. Liquid matrices offer the advantage of providing an inherently homogeneous sample, facilitating, e.g., a quantitative MS analysis. A disadvantage of the NBA matrix is its negligible optical absorption at the standard UV-MALDI laser wavelengths of 337 (N2 laser) and 355 nm (frequency-tripled Nd:YAG). The

10.1021/jp905251r  2010 American Chemical Society Published on Web 12/07/2009

5368

J. Phys. Chem. C, Vol. 114, No. 12, 2010

wavelength of the frequency-quadrupled Nd:YAG laser (λ ) 266 nm), however, matches well with a high optical absorption of NBA of R ≈ 1.4 × 107 m-1. This value is comparable to those exhibited by standard matrix compounds at the 337 nm excitation wavelength of the N2 laser. Since the values of other relevant physicochemical parameters like molecular weight (MW) and heat of vaporization are also comparable by order of magnitude, it may be speculated that the desorption mechanisms underlying UV-MALDI with NBA and with solid state matrices bear similar aspects. The main advantage of using liquid NBA for fundamental studies on the MALDI processes is that it substantially facilitates some experimental approaches, like the photoacoustic detection and fast-flash imaging applied in the present study. Liquid matrices are more common in IRMALDI applications. Here, glycerol has been found to provide particularly soft desorption/ionization conditions for the MS analysis of a variety of compounds.31-36 Unfortunately, in combination with TOF mass spectrometers exhibiting an axial geometry, the use of liquid matrices typically comes along with a reduced mass resolution and mass accuracy of the TOF-MS analysis in both UV- and IR-MALDI applications.37 This loss in performance can be attributed to a more pronounced ejection of a mixture of gaseous components and small droplets.38,39 Some of these problems can be solved if mass spectrometers are employed that decouple the desorption/ ionization process from the mass analysis. In this regard, orthogonal (o-)TOF instruments have been found to be particularly useful.34-36,40 To further elucidate details on the physicochemical nature of the MALDI process, the dependencies of matrix and analyte ion signal intensities on the different laser parameters, like fluence (i.e., laser pulse energy per irradiated area), pulse duration, and spot size, as well as on sample preparation-related factors have been recorded in numerous studies.8-13 Although informative and straightforward, one shortcoming of this approach is that the formation of gas phase ions is always the result of the combined desorption and ionization processes. Moreover, ionic species constitute only a minor fraction of the ejected MALDI plume. Details specific to either of the two processes cannot be discerned. Some studies have established experimental techniques that allow monitoring of the laser desorption event “separately” and, hence, “decouple” the analysis of desorption versus ionization processes. We have recently described a photoacoustic (PA) analysis scheme to study material ablation in IR-MALDI with a glycerol matrix, which is based on the detection of lasergenerated stress waves induced by the rapid heating of the matrix sample and the recoil momentum of ejected material.41 Ion signals were recorded simultaneously with the PA stress waves by preparing the analyte-matrix samples directly on the photoacoustic detector (a piezoelectric disk) and embedding the detector into a MALDI sample plate. In a closely related study, we have recently described a two-laser setup for fast-flash imaging (FFI) of the expanding glycerol IR-MALDI plume.39 This setup allows monitoring of the time course of single desorption events with nanosecond time resolution. A differentiation between the ejection of droplets and gaseous plume domains became possible by applying both dark-field and scattered-light illumination geometries. In a third study, we applied IR laser postionization (PI) to further characterize the glycerol IR-MALDI plume expansion dynamics.42 FFI and PI are two powerful methods for studying the temporal and spatial plume development. Neither the PA nor FFI setups have thus far been applied to the investigation of the UV-MALDI process.

Rohlfing et al. The present article describes the application of PA, FFI, and UV-laser PI to the investigation of the UV-MALDI process with the liquid NBA matrix. PA and FFI data have been recorded as a function of the desorption laser fluence and, in the case of the FFI and PI studies, as a function of the delay between the desorption and the imaging or postionization laser pulses. The data are discussed in the context of existing models of the MALDI process and allow one to speculate about a qualitative model for the UV-laser induced desorption of NBA. Experimental Section General Setup. Liquid 3-nitrobenzyl alcohol (NBA; 98%, Sigma-Aldrich, Taufkirchen, Germany; MW, 153.14) was used as the matrix in all experiments. MALDI samples were prepared by mixing equal volumes (1-3 µL) of NBA and a 1 × 10-4 M aqueous solution of the peptide neurotensin (Sigma-Aldrich; MW, 1672.92). Depending on the distinct experiment, the samples were either prepared on a metal sample plate or on the metallized surface of a piezoelectric detector (see below). Residual water was evaporated by placing the samples at ∼10-2 mbar for 10 min, yielding an approximately hemispherical droplet of ∼1-2 mm height. A frequency-quadrupled Nd:YAG laser (Speser 600, Spektrum Laser GmbH, Berlin, Germany) emitting pulses of 5 ns duration (fwhm) at a wavelength of 266 nm was employed for desorption/ionization in all experiments. Care was taken to always irradiate the very apex of the droplet. The laser pulse energy was adjusted by means of a variable dielectric beam attenuator. The shot-to-shot pulse energy stability was about 3% (single standard deviation). During the experiments, the long-term stability of the laser output energy was frequently checked with a laser energy meter (Rk3230, Laser Precision Corp., Yorkville, NY). Photoacoustic and postionization experiments were carried out with an in-house-built axial TOF mass spectrometer equipped with a single-stage ion mirror (reflectron) of 3.5 m equivalent drift length. Only data recorded in positive ion mode are presented here. The mass spectrometer had also been employed in the previous PA and PI studies addressing IR-MALDI.41,42 The back pressure in the ion source is about 2 × 10-6 mbar. The desorption laser beam is coupled into the ion source through a fused-silica window and focused onto the sample at an angle of incidence of 45° with a fused-silica lens (f ) 100 mm). The focal spot size was estimated by taking burn-patterns on thermosensitive paper to about 135 µm × 95 µm (1/e2 definition) with an elliptic shape, yielding an irradiated area of about 1 × 10-8 m2. A CCD camera serves for sample observation. Ions were accelerated with a two-stage Wiley/ McLaren-type continuous extraction ion source with distances of the planar grids (∼90% ion transmission each) from the sample plate surface of 6.0 and 18.5 mm, respectively. For the PA experiments, a total acceleration voltage of 6 kV was used; for the PI experiments, a value of 18 kV was applied. An Einzel lens, mounted behind the second extraction grid, serves to focus the ion beam. Extraction grid, Einzel lens, and reflectron voltages were tuned for maximum ion transmission and signal resolution. Ions were detected with a Venetian-blind type secondary electron multiplier (SEM) equipped with a conversion dynode mounted ∼10 mm in front of the first dynode of the SEM. In the PA experiments, the conversion dynode voltage was held fixed at its maximum value of -20 kV to obtain the best possible ion detection efficiency. For the PI experiments, a lower voltage of -3.5 kV was applied to increase the mass resolution (see below). Signals were amplified by a custombuilt fast amplifier, digitized with a digital oscilloscope (LeCroy

Desorption Process in UV-MALDI with a Liquid Matrix

J. Phys. Chem. C, Vol. 114, No. 12, 2010 5369

9345A, Chestnut Ridge, NY) and transferred to a PC for subsequent evaluation. Photoacoustic Analysis. Short Theory of Photoacoustic Signal Generation. Pulsed laser irradiation of an absorbing sample results in the formation of photoacoustic (PA) stress waves. A detailed description of stress wave generation, propagation, and detection under MALDI conditions has been provided in ref 41. The main features are summarized here. Generally, PA stress waves may contain (bipolar) thermoelastic as well as (unipolar) material ejection-related recoil components. Thermoelastic signals are generated by rapid laser heating of the absorbing material, leading to thermal expansion and the formation of a bipolar, symmetric compression/tension wave.43 If the intensity of the incident laser light is sufficient to provoke material ejection, the recoil momentum of the ejecta acts on the remaining sample surface, leading to the formation of an additional, purely compressive stress wave. To a first-order approximation, thermoelastic and recoil-induced stress components superimpose linearly, yielding the overall PA stress wave signal. The amplitude of a thermoelastic stress wave largely depends on the ratio of the laser pulse duration τL to the transit time of an acoustic wave through the excitation volume τac. The latter is commonly estimated by

τac )

1 Rc

(1)

where R ) δ-1 denotes the optical absorption coefficient at the given laser wavelength; c is the speed of sound inside the sample; δ is the optical penetration depth. For τL , τac the stress wave has not yet propagated notably out of the excitation volume during the laser pulse, and the deposited laser energy can, hence, contribute very efficiently to the build-up ofseventually very highsstress amplitudes. The inequality τL , τac is, therefore, often referred to as the “stress confinement” condition.44 In contrast, for τL . τac, the thermoelastic stress already dissipates from the excitation volume while radiant energy is still being deposited. This considerably reduces the amplitude of the thermoelastic stress wave, which may, eventually, even become negligible in comparison to any ejection-related stress wave component. UV-MALDI with nanosecond lasers always takes place outside stress confinement conditions (see below), whereas IR-MALDI with its larger laser penetration depths may or may not take place under stress confinement, depending on the exact values of the laser pulse duration and optical absorption of the matrix.41 Due to the large laser penetration depths in the 1 to 10 mm range, thermal confinement conditions are always effective in IR-MALDI with nanosecond laser pulses. In contrast, thermal dissipation times for the UV-MALDI case are on the order of ten to a few ten nanoseconds and, therefore, only slightly exceed typical laser pulse durations of 3-5 ns.20 Energy dissipation by heat conduction may, thus, play a role in the overall pathways.8,45 A value for the speed of sound in NBA could not be found in the literature. On the basis of data for similar organic liquids,46 the speed of sound of NBA can, however, be estimated to be on the order of c ≈ 1200 ms-1, yielding an acoustic transit time according to eq 1 of τac ∼ 60 ps. Since this is much shorter than the laser pulse duration of τL ) 5 ns, the stress confinement condition is clearly not fulfilled. The amplitude of the thermoelastic stress wave will, hence, be negligible, and the overall PA signal will be dominated by the ejection-related recoil component. Modeling the ejection as a one-dimensional process

and assuming an approximately constant and fluence-independent average velocity of the ejecta,20 one can derive the ejectionrelated stress wave component σejec to41

σejec(τ) )

Vj m ˙ (τ) A

(2)

where Vj denotes the average ejection velocity; A is the area from which the ejection takes place; m ˙ ) dm/dt is the material ejection rate; and τ ) t - z/c is the reduced time (i.e., time t after the instant of the laser pulse minus the propagation time of the stress wave from the surface to the depth z where, e.g., the acoustic detector may be situated). Integration over the stress wave signal yields

∫ σejec(τ)dτ ) AVj mtotal

(3)

which is proportional to the overall mass, mtotal, ejected per single laser exposure. Note that the dependence of σejec on the area A in eqs 2 and 3 is eliminated if the stress wave is recorded with a piezoelectric detector exhibiting a surface much larger than the cross section of the wave (like the one used in this study) since the incoming signal is integrated over the entire detector area. Stress WaWe Detection. PA stress waves were detected with a piezoelectric disk of 5 mm thickness and 10 mm diameter made of FPM202 (marco, Hermsdorf, Germany; http://www. marco.de/E/D/pb/001.html). Both sides of the disk had been metallized with 10 µm of CuNi for electrical contact. The disk was embedded into a modified MALDI sample plate as shown in Figure 1(a). Samples were prepared directly on the CuNi surface as approximately hemispherical droplets of ∼1 mm radius. To allow a simultaneous recording of PA and MALDI ion signals, the entire sample plate needs to be on high electrical potential. Therefore, a custom-built electronic circuit, outlined in Figure 1(b), was integrated into the sample plate. This allows applying an acceleration potential of 6 kV to the plate and both sides of the piezoelectric disk via two 1 MΩ resistors, while the high-frequency PA signal is decoupled from high voltage via two 1 nF HV capacitors and amplified by a custom-built differential amplifier (input impedance, 50 Ω; amplification, ∼10; bandwidth, ∼130 MHz). PA signals were recorded with the same digital oscilloscope storing the MS data and transferred to a PC for further data analysis with laboratory-developed software. As was demonstrated in our previous study, a time resolution of ∼2 ns is achieved for the PA stress wave detection with this setup.41 This high time resolution is realized by operating the piezoelectric disk in an approximated “short-circuit” mode,47 i.e., by using an amplifier with a low input impedance (50 Ω) and, hence, measuring the piezo’s discharge current rather than its idling voltage. In contrast to the “open-circuit” mode, which is often applied to register PA signals, the time resolution is not governed by the thickness of the piezoelectric element but by the RC time constant of the transducer capacitance C and the amplifier’s load resistance R. The value of R ) 50 Ω was selected to match the electromagnetic impedance of the coaxial cables connecting the amplifier and piezoelectric disk. Together with the overall capacitance of the piezo and the coupling capacitors of C ) 150 pF, the theoretical time resolution is in the low nanosecond range, corresponding well with the experimentally determined value. Data Acquisition and EWaluation. PA and ion signals were recorded as a function of the desorption laser fluence (i.e., laser pulse energy divided by the irradiated area as estimated with

5370

J. Phys. Chem. C, Vol. 114, No. 12, 2010

Rohlfing et al.

Figure 2. Fast-flash imaging of the UV-MALDI plume. Two imaging geometries were employed: (a) 90°-scattered light detection: Only light that is scattered/reflected at an angle of 90° by single particles/droplets contributes to the image. (b) Dark-field illumination: Only light that is diffracted at small angles in the forward direction is imaged onto the camera chip. See text for details.

Figure 1. Photoacoustic analysis: experimental setup for simultaneous recording of laser-induced photoacoustic stress transients and MALDI ion signals. (a) A piezoelectric transducer disk of 5 mm thickness and 10 mm diameter, mounted into a standard MALDI sample plate, was used for time-resolved detection of laser-induced photoacoustic stress waves. Samples were prepared directly on the metallized transducer surface. (b) Electrical circuit to read out the high-frequency PA signal. During the measurements the whole sample plate was held at the TOF acceleration potential of 6 kV. (c) The UV-MALDI reflectron-TOF mass spectrometer employed in this study.

the burn-pattern method described above) and averaged over typically 100-1000 single laser exposures to obtain a sufficient signal-to-noise (S/N) ratio. Electromagnetic noise, stemming from the operation of the electro-optical Pockels cell of the UV laser, was eliminated from each PA signal by subtraction of a reference signal recorded with a blocked laser beam. PA and ion signals were numerically integrated by laboratory-developed software. According to eq 3, the integral values of the PA signals constitute a measure for the overall mass of ejected material, mtotal. For the ion signals, integral peak areas serve as a measure for the number of generated ions. For both signal types, the individual “threshold fluence” is defined as the fluence above which the S/N ratio of the averaged integrated signals exceeds a value of 2. Fast-Flash Imaging. FFI experiments were carried out with a separate setup (Figure 2), essentially identical to the one used in our previous IR-MALDI study.38 Samples were prepared on

a black-anodized aluminum sample plate which was mounted in a home-built vacuum chamber (p ∼ 10-5 mbar). The desorption laser beam was coupled into the chamber via a fused silica window and focused by a fused silica lens (f ) 50 mm) in a slightly defocused manner, to obtain an irradiated area of about 110 µm in diameter (as estimated by taking burn patterns on thermosensitive paper). This produced an irradiated area of ∼9.5 × 10-9 m2, well comparable to the spot size in the TOFMS-based experiments. Different from the MS experiments, the laser beam hit the sample under an angle of incidence of 90°. A frequency-doubled Nd:YAG laser (YG660S, Quantel Int., Santa Clara, CA), emitting pulses of 8 ns duration (fwhm) at a wavelength of 532 nm, served as the illumination laser. The delay between the desorption and the illumination laser pulses was adjusted between 0 and 22 µs by means of a laboratorydesigned trigger unit. The exact time difference was monitored for each recorded image using two high-speed photodiodes and a fast digital oscilloscope (LeCroy LC564, Chestnut Ridge, NY) and is defined here as the time between the intensity maxima of the two photodiode signals. Images of the expanding MALDI plume were recorded with a CMOS camera (PixeLink PL-A633, Vitana Corp., Ontario, Canada; 1280 × 1024 pixels, 10 bit/ pixel gray scale dynamics) which was synchronized with the illumination laser pulse by the trigger unit. Its exposure time is defined by the 8 ns duration of the illumination laser pulse rather than by the 1 ms electronic shutter time of the camera. A 532 nm narrow band-pass filter mounted in front of the camera chip eliminated all ambient light other than that of the illumination laser. Images were transferred to a PC via an IEEE 1394a interface and were processed and analyzed with public-license graphics software (ImageJ, version 1.33u, http://rsb.info.nih.gov/ ij/, National Institutes of Health, MD), including a conversion of the 10-bit gray scale to false-color images. Two different illumination geometries were employed to preferentially image particulate and gaseous plume components, respectively. Particles (i.e., small droplets) are best detected by their 90°-scattered light [Figure 2(a)]. In this arrangement, the illumination laser beam is focused into the center of the expanding plume by a cylindrical lens (f ) 200 mm) to a vertical line of 130 µm width and 7 mm height at 90° to the observation

Desorption Process in UV-MALDI with a Liquid Matrix direction. This line illumination limits the depth of the illuminated zone and, hence, generates sharper images. Light scattered under 90° is imaged by an anastigmatic objective lens (DF Plan 1X-2, Olympus Europe, Hamburg, Germany; diameter 49 mm, feff ) 110 mm) onto the CMOS camera chip with a 13-fold magnification. The lateral resolution of the optical system is ∼4 µm, as determined with a graticule. Plume domains containing predominantly gaseous compounds and small clusters modulate the index of refraction and are imaged best by a dark-field illumination setup [Figure 2(b)]. Here, the laser beam intensity profile is first homogenized by passing the beam through an optical fiber (FG-200-LAT, 3 M Corp., CT; numerical aperture, 0.16) of 57 m length. After exiting the fiber, the beam illuminates an optical slit of 65 µm width and 2 cm height. The laser light exiting the slit is collimated by an anastigmatic lens 1 (DF Plan 1X-2, Olympus Europe, Hamburg, Germany; diameter 49 mm, feff ) 110 mm) to a parallel beam of 40 mm diameter, which traverses the expanding plume. The objective lens 2 (the same as in the scattered-light setup) focuses the parallel beam onto a thin metal wire of 80 µm diameter. By imaging the optical slit onto the wire with the two-lens arrangement, no direct laser light can reach the camera. Only diffracted or scattered light can pass the wire and is collected by the objective lens onto the camera chip with a 13-fold magnification. Both the dark-field and the scattered-light images exhibit some artifacts caused by interferences and reflections. Therefore, care has to be taken to avoid a misinterpretation of the images. Especially in the dark-field images interference fringes are omnipresent. Their magnitude depends on the orientation of slit and wire relative to the sample surface. Images recorded with horizontal orientation generally exhibited more pronounced interference fringes. Since the same information is contained in the images taken in the two geometries, only images recorded with vertical orientation are presented in this article. UV-Laser Postionization. PI experiments were performed on the same TOF-MS instrument as used in the PA measurements. The amended setup for the two-laser experiments is shown in Figure 3(a). The same instrument and a comparable PI configuration were also used in our previous IR-MALDI study in which a second IR-laser served for postionization.42 For the PI experiments, neurotensin/NBA samples were prepared on a stainless steel sample plate. After inserting the plate into the ion source of the mass spectrometer, the desorption laser fluence was adjusted to a value just below the MALDI ion detection threshold fluence. Each desorption laser pulse, hence, generated a plume of neutral species, while direct MALDI ion signals were not detected. A second, frequency-quadrupled Nd:YAG laser (System 2000, JK Lasers Ltd., Rugby, UK), emitting pulses of ∼14 ns (fwhm) duration at a wavelength of 266 nm, was employed for postionization of molecules out of the plume. The PI laser energy was controlled by means of a dielectric attenuator. The delay between the desorption and postionization laser pulses was adjusted between 0 and 22 µs using a custommade trigger box and was monitored with two fast photodiodes and a digital oscilloscope (LeCroy 9450A, Chestnut Ridge, NY). The PI laser beam was coupled into the ion source of the mass spectrometer “from the side” using a fused-silica window and the optical arrangement shown in Figure 3(b). Lens 1 (f ) 50 mm) and lens 2 (f ) 200 mm) of this setup form a telescope which first expands the beam by a ratio of 4:1 to reduce the beam divergence. The following lens 3 (f ) 300 mm) is mounted on an xyz-translation stage (setting accuracy, 5 µm) and serves for focusing and steering of the PI beam. The focal diameter of

J. Phys. Chem. C, Vol. 114, No. 12, 2010 5371

Figure 3. UV-laser postionization: (a) The employed reflectron-TOF mass spectrometer with the desorption and postionization laser beams. An xyz-micrometer translation stage is used to adjust the position of the postionization laser beam focus relative to the primary laser spot. (b) Close-up view of the two-beam geometry. ∆z denotes the distance between the apex of the NBA drop and the “edge” of the PI laser beam waist.

the PI beam was estimated by taking burn patterns on thermosensitive paper to ∼80 µm (1/e2 definition). The zero position (∆z ) 0) with respect to the sample surface was experimentally determined by blocking the primary desorption laser and then moving the PI laser beam waist toward the sample surface at the highest possible PI laser fluence of ∼3500 J/m2 until MALDI ion signals were generated by the PI laser alone. The PI beam was then retracted gradually until no ion signals were detected anymore. The reproducibility of this method in defining the zero position was found to be about ( 5 µm. For the PI experiments, the sample plate potential was set to a value of +18 kV. To achieve an optimal mass resolution, the potential difference between the conversion dynode and the first dynode of the SEM was set to -3.5 kV. In this “electron-only mode”, only secondary electrons produced at the conversion dynode of the ion detector are amplified by accelerating them onto the first dynode of the SEM, but not the slower secondary ions.48,49 Postionization mass spectra were recorded as a function of the delay ∆t between desorption and PI laser pulses and the distance ∆z of the PI laser beam waist from the sample surface. To enhance the signal-to-noise ratio, 25 single-shot spectra were averaged for each setting of ∆t and ∆z. Results and Discussion Photoacoustic Analysis. Time-resolved PA signals generated in the NBA/neurotensin sample by irradiation with the Nd:YAG laser at four different fluences are shown in Figure 4. As expected, the stress wave signals are by far dominated by the unipolar recoil-induced component. The time of 700-750 ns, at which the PA signals are detected relative to the instant of the laser pulse (t ) 0), corresponds to the signal propagation time through the NBA drop.

5372

J. Phys. Chem. C, Vol. 114, No. 12, 2010

Figure 4. Photoacoustic transients generated in NBA by irradiation with the Nd:YAG laser (λ ) 266 nm; τL ) 5 ns) for four different laser fluences between 210 J/m2 (slightly above the ion detection threshold) and 350 J/m2. The signals represent the average of 100 single laser exposures, each. The slight variation in the times at which the signals are detected reflects differences in the sample height (e.g., owing to the shrinkage of the droplet due to previous material ejection and slow equilibrium evaporation in the ion source).

Each signal trace represents the average of 100 single laser exposures at a given fluence. At the lowest of the four fluences of 210 J/m2, slightly above the ion detection threshold, a peak representing the compressive stress wave induced by the recoil momentum of ejected material can just be differentiated by visual inspection. The amplitude of this signal increases strongly with laser fluence. At ∼350 J/m2, which forms about the center of the useful MALDI-TOF-MS fluence range, a clear and approximately symmetric signal is observed that exhibits a width of ∼20 ns (fwhm). Within the experimental error, the width of this compressive PA peak is independent of the fluence over the full investigated range. The peak width of 20 ns sets an upper limit for the duration of the instantaneous phase of material ejection. Taking into account that the acoustic wave signal will be broadened by dispersion during the propagation in the NBA and that a spherical rather than a plane wave is impinging on the piezoelectric detector,50 it appears reasonable to assume that the material ejection process is completed on a time scale close to the pulse duration of the excitation laser of 5 ns. For fluences below ∼200 J/m2, the compressive component is not discernible by eye from the residual background noise. Albeit, upon integration of the PA signal, even these weak signal traces can be meaningfully evaluated, allowing an extension of the measurement range down to ∼70 J/m2. However, the low fluence data are associated with a large scatter and, accordingly, large experimental error. The integral PA signal intensities are plotted in Figure 5 as a function of the laser fluence, along with the simultaneously recorded integral signal intensities of the molecular neurotensin ion, [M + H]+. To first order, the integral PA signal intensity constitutes a (relative) measure of the overall amount of material, mtotal, ejected per single laser pulse. In contrast, the intensity of the ion signal corresponds to the number of generated peptide ions. A correlation between the two data sets is complicated by the fact that the ion detection threshold typically depends on the number of accumulated mass spectra, due to the steep rise of molecular ion yield with fluence and the occurrence of the MALDI-typical chemical background noise. Increasing the

Rohlfing et al.

Figure 5. Integral photoacoustic (open blue diamonds) and molecular ion signal intensities [M + H]+ of neurotensin (open and solid black circles) as a function of laser fluence. Ion data represented by open symbols are the average of 100 single laser exposures. Solid symbols represent data accumulated over 1000 shots to improve the signal-tonoise ratio and to extend the measurement range. Data have been normalized by division by the laser shot number. For a few selected fluence values, error bars indicate the variation, estimated by evaluating the standard deviation from 10 averaged signals for each fluence. Solid and dashed lines through the PA data are best fits to three model functions. Black solid line: fit curve according to the quasithermal model (eq 5) with T0 ) 293 K; Ea ) 0.37 eV; η ) 13.9 K m2 J-1. Pink dashed line: best fit to a photoablation model (eq 7) with HThr ) 200 J m-2 superimposed with the quasithermal model. Red solid line: fit curve according to the moving-heat-source model (eq 8) with HThr ) 162 J m-2 superimposed with the quasithermal model. Green solid line: fit curve according to a power law (eq 4) with n ) 8. See text for details.

number of laser shots from 100 to 1000, the threshold as derived from the integral ion intensity values drops from ∼230 J/m2 to about 170 J/m2. In line with a previous study by Westmacott et al.,51 in this fluence range there is no indication of a true “physical threshold” below which no ions are generated. Within the experimental accuracy, the same threshold fluence was found for analyte and matrix ions. The fluence dependence of the matrix ion signals was not evaluated in the scope of the present study. Above the ion detection threshold, the ion signal intensity YIon increases steeply with laser fluence H. An empirical power law of the form

YIon ∝ Hn

(4)

provides an excellent fit with an n ≈ 8 (green line in Figure 5). Both the general form of the yield-fluence dependence and the magnitude of the coefficient n are in good agreement with the ion yield fluence relationship in UV-MALDI with solid state matrices.8,51,52 Saturation effects noticeable above a fluence of ∼400 J/m2 can be attributed to a more pronounced level of analyte fragmentation. From a comparison of the PA and the ion data, it is apparent that material ejection commences already at fluences considerably below the ion detection threshold. Previous UV-MALDI PI studies revealed a similar scenario. It was shown that the abundance of ejected matrix molecules as a function of fluence can be well described by a quasithermal desorption model,8,10 for which a proportionality of the form

N ∝ e-Ea/[k · (T0+ηH)]

(5)

holds, where N denotes the number of desorbed molecules; Ea is activation energy; k is the Boltzmann constant; T0 is the initial sample temperature; H is the laser fluence; and η is a conversion

Desorption Process in UV-MALDI with a Liquid Matrix

J. Phys. Chem. C, Vol. 114, No. 12, 2010 5373

coefficient describing the increase of sample temperature with laser fluence. For the latter, a proportionality

η∝

R F · CH

(6)

holds to first order, where R is the optical absorption coefficient; F is the density; and CH is the heat capacity. Using equilibrium values for F, CH, and R as listed in Table S1 (Supporting Information), an η of 7.7 K m2 J-1 is derived for NBA. However, fitting eq 5 to the low-fluence PA data with this value results in a relatively poor fit, regardless of the used Ea (not shown). In the previous photoionization work, it was found that under the nonequilibrium conditions of MALDI laser excitation η typically deviates from the equilibrium value by a factor of about 1.8.8,53 This finding was attributed to the nonequilibrium conditions and a preferential excitation of lattice vibrations. Using the correction factor of 1.8 as determined in ref 8 for the extensively studied 2,5-dihydroxybenzoic acid (DHB) matrix, a corrected η of 13.9 K m2 J-1 is derived for NBA. Using this value, a best fit of eq 5 to the low-fluence PA data is achieved for Ea ) 0.37 eV. Given the above correction, the small fluence range covered, and the large scatter in the low-fluence data, and further taking into account that nonequilibrium effects may modify the overall process,11 the fit and the derived value for Ea should not be overinterpreted. The assumption of molecular ejection from the liquid surface according to the quasithermal model is, however, well-corroborated by the FFI data presented in the next chapter and also supported by molecular dynamics (MD) simulations.44 It seems therefore reasonable to assume that molecular ejection according to the quasithermal model dominates the MALDI desorption process in the low-fluence regime. A kink in the PA signal-fluence relation is notable at a laser fluence of about 170 J/m2, indicating a sudden change in the material ejection mechanisms. For NBA, the boiling temperature TB is tabulated for a pressure of 4 mbar as 450 K.54 Using this value as a first-order approximation (and thus neglecting the high-vacuum conditions in the MALDI ion source), TB would be reached in the surface region and the central part of the laser beam already at laser fluences in the low 10 J/m2 range, thus substantially below the threshold fluence of ∼170 J/m2. Given the high heating rates on the order of 1010 K/s, a substantial degree of overheating of the material is likely to occur, resulting in a metastable liquid state. Explosive boiling is typically assumed to be triggered in a narrow temperature range of about 0.9 of the critical temperature Tcr.57,55 Unfortunately, Tcr of NBA was not available. Typical values for the critical temperature documented for chemically similar compounds, e.g., phenylalkanols, may serve to estimate the order of magnitude. For instance, Tcr for 2-phenyl-1-ethanol has been determined to be ∼700 K (at a critical pressure of about 4 MPa).56 In the higher fluence regime, such temperatures must be expected to be reached, and the occurrence of phase explosion57 must thus be regarded as the primary ablation mechanisms; in the context of laser excitation, phase explosion is often referred to as photoablation.58 Phase explosion results in rapid homogeneous nucleation and disintegration of entire volume elements into a mixture of vapor, clusters, and droplets. MD modeling for simulated, non-stress confinement MALDI conditions also predicts the change from a quasithermal molecular ejection to material ablation by phase explosion.44 For a laser-induced phase explosion mechanism, a dependence of the amount of ejected material VAbl on the laser fluence H according to a photoablation model would be expected.59 In the

case of a Gaussian spatial excitation profile, this model would yield a fluence dependence of the form,41

[ ( )]

VAbl(H) ∝ ln

H HThr

2

(7)

where HThr denotes the threshold fluence for the process. Such a relation was recently found to describe the material ablation in IR-MALDI with a glycerol matrix with excellent precision.41 However, attempts to fit the high-fluence UV-MALDI PA data to a photoablation model, superimposed on the quasithermal relation to account for material ejection taking place from the fraction of the irradiated area receiving local fluence values below the threshold HThr (i.e., in the outer rim of the Gaussian beam profile), did not result in a meaningful outcome. The best fit using a HThr of 200 J/m2 is shown in Figure 5 as a pink dashed line. The clear deviation between the fit and the experimental data provides evidence that material ablation under UV-MALDI conditions is unlikely to proceed via a simple photoablation process according to eq 7. In contrast, the experimental PA signal intensity-fluence dependence is reproduced in good approximation by a superposition of the quasithermal relation of eq 5 and a function of the form

(

YPA ∝ H - HThr 1 + ln

H HThr

)

(8)

with a threshold fluence HThr of 162 J/m2 providing a best fit (red line in Figure 5). Equation 8 essentially reflects a linear increase of material ablation with fluence that is, however, convoluted with the excitation profile of the employed Gaussian laser beam (see Supporting Information for a detailed derivation). Theoretically, a linear increase of material ejection with deposited energy would be obtained if, once the threshold is reached, all additional energy would merely be consumed by the latent heat of vaporization. The simplest process yielding such a relationship would be “normal boiling”, which mechanistically produces vapor bubbles in the heated liquid. However, with respect to the above arguments, this process cannot reasonably form the primary material ejection pathway under the current experimental conditions. Nevertheless, below we will provide evidence that a heterogeneous nucleation (boiling) process likely forms a third additional mechanism for the overall material ejection. Notably, a relationship between material ejection and laser fluence according to eq 8 would also be predicted by a movingheat-source model.60,61 This model was originally developed to describe the ablation of metals with millisecond-long ruby laser pulses. It assumes a layer-by-layer ablation of the sample after laser heating of the surface to values provoking material ejection. A more detailed derivation of the moving-heat-source model and its adaptation to the experimental conditions of this work is provided in the Supporting Information. A discussion of its potential relevance in the context of the MALDI mechanisms is resumed below after presentation of all other experimental data. Fast-Flash Imaging. On the basis of the results of the PA study, two values for the desorption laser fluence H were chosen for the fast-flash imaging experiments, one slightly below (H ) 130 J/m2) and one well above (H ) 350 J/m2) the ion detection threshold. Images were taken in dark-field illumination and in 90° scattered-light detection configuration, to obtain an impression of the distribution of gaseous as well as particulate plume constituents. In both geometries, between 150 and 250

5374

J. Phys. Chem. C, Vol. 114, No. 12, 2010

Rohlfing et al.

Figure 6. Dark-field images of the NBA plume expansion for (a) a desorption laser fluence of H ) 130 J/m2 (below the ion detection threshold) and (b) a desorption laser fluence of H ) 350 J/m2 (well above the ion threshold). The delay between the desorption and illumination laser pulses is indicated. The surface of the NBA drop is evident by its optical contrast. The optical slit and wire of the dark-field setup are oriented vertically with respect to the sample plate surface. A linear 10-bit gray scale to false color conversion is used for presentation of all images. An animated video clip of this image series is available in the Supporting Information. Note that the horizontal lines that are visible in some of the images are caused by an optical interference effect, and that they do not reflect a spatial modulation of the gas density.

images were recorded at various delays between the desorption and illumination laser pulses. Representative images are shown in Figures 6 and 7 to display the temporal-spatial evolution of the MALDI plume. Dark-Field Illumination. Figure 6 displays the images recorded with the dark-field setup at (a) H ) 130 J/m2 and (b) H ) 350 J/m2, respectively. The surface of the NBA drop is visible in each image as a diffuse horizontal line at the bottom. For the lower fluence, first effects of the desorption laser impact become noticeable at a delay between the two laser pulses of about 4 ns. With ongoing time, an initially very dense plume evolves, and the maximum gas density (i.e., image intensity) is reached at a delay time of ∼14 ns, suggesting that the rapid material ejection phase is completed at about this time. This

observation is in agreement with the findings made in the PA study. Plume images taken at longer delay times display a steady expansion of the desorbed gas into the surrounding vacuum. A gaseous plume is still visible for delay times of up to 150 ns. For longer times, the plume has thinned out too much and becomes indiscernible from the background. Images recorded in dark-field geometry at H ) 350 J/m2 are shown in Figure 6(b). For this elevated laser fluence, first effects of the irradiation are already visible at a delay time of ∼0.5 ns (i.e., while the desorption laser is still depositing energy into the sample; note that the peak amplitudes of the two laser pulses are used for determination of the delay times). The maximum gas density, as derived from the image intensity, is now found

Desorption Process in UV-MALDI with a Liquid Matrix

J. Phys. Chem. C, Vol. 114, No. 12, 2010 5375

Figure 7. 90°-scattered light images of the NBA plume expansion for (a) a desorption laser fluence of H ) 130 J/m2 (below the ion threshold) and (b) a desorption laser fluence of H ) 350 J/m2 (well above the ion threshold). The delay time between the desorption and illumination laser pulses is indicated. The surface of the NBA drop has been made visible by illumination with a faint light source. A linear 10-bit gray scale to false color conversion is used for presentation of all images. The images recorded at low laser fluence have been contrast-enhanced a by a factor of 4.

at ∼22.5 ns, and the expansion of the gas into the surrounding vacuum can be followed for delay times as long as 200 ns (data not shown). The area from which sizable material ejection takes place increases notably with fluence, which can be attributed to the employment of the Gaussian beam profile. At 130 J/m2, a dense plume is only generated in the very center of the profile from an area of about 25 µm in diameter. At the higher fluence of 350 J/m2, sizable material ejection appears to take place from an area exhibiting a diameter of ∼60 µm, about 2.5 times as large. Apart from the larger amount of material being ablated at the larger fluence and the increasing area from which sizable ejection takes place, no further significant differences regarding the plume dynamics can be deduced from the dark-field images. In particular, there is no indication of a change in the ejection mechanisms at the threshold fluence, as was deduced from the PA data.

90° Scattered-Light Detection. Images that were recorded in the 90° scattered-light detection geometry are shown in Figure 7. For fluences below the ion detection threshold, first distortions in the sample integrity become visible in the scattered light images at a delay of 3.1 ns, well in accordance with the darkfield experiments. Granular structures become first apparent at ∼33 ns, indicating a disintegration of the initially dense plume. Single particles (droplets) become clearly discernible for times exceeding ∼100 ns. Single particles are sporadically spotted for delay times of up to 4 µs (data not shown). For fluences above the ion threshold, sizable surface distortions appear to set in at the very beginning of the 5 ns long desorption laser pulse, while granular structures are already visible at its end. The “raylike” structures visible in some of the images originate from an optical artifact and are not related to the plume dynamics or formation of gas or particle jets.39,62 Like in the low fluence case, single particles become discernible

5376

J. Phys. Chem. C, Vol. 114, No. 12, 2010

at ∼100 ns after the excitation laser pulse, and these droplets are found for up to 5 µs (data not shown). The scattered-light images also reveal that, with rising fluence, material ablation takes place from an increasing area, exhibiting diameters of ∼20 and ∼50 µm for the two investigated fluences of 130 J/m2 and 350 J/m2, respectively. The images suggest that during the early phase of the plume development more droplets are ejected at the higher fluence. However, the number of discrete droplets that survived further disintegration appears to become similar at higher delay times for both laser fluences. For example, from a delay time of ∼500 ns onward, only a few single particles are observed in both cases. In the 90° scattered-light geometry, all discrete particles produced images with an Airy-function-like shape, exhibiting a width at the 1/e2 intensity level of ∼3.3 µm (Figure S1, Supporting Information). Since the optical resolution of the imaging system is about 4 µm,39 it is evident that the measured width is determined by the optical resolution of the detection system rather than by the actual size of the particles. Given that the optical penetration depth into the matrix at 266 nm is only ∼70 nm, the particle diameters must in fact be expected to fall at most into the 10-100 nm range, even if the diameter of the irradiated area is about 3 orders of magnitude higher. Fluence Dependencies. The detailed fluence dependence of gas and particle expansion was evaluated for both illumination geometries by recording a large set of data points for fluences between ∼50 and 850 J/m2. To capture the early phase of the gas plume formation, a fixed delay of 30 ns between excitation and illumination laser pulses was chosen for the dark-field images. A delay of 500 ns was selected for the scattered-light experiments, thus probing the late phase of the particulate plume expansion. Images obtained in dark-field geometry were evaluated by calculating the average image intensity for each fluence value, hence yielding a rough measure of the amount of ejected gaseous material. The intensity values were baseline-corrected by subtracting the average intensity of a blank reference image, recorded with a blocked desorption laser beam. In the scatteredlight experiment, the number of discrete visible particles was determined for each image. The amount of gas present in the plume increases strongly with fluence [Figure 8(a)]. Notably, the intensity-fluence data can be fitted well by the quasithermal desorption model (eq 5; blue solid line graph). As for the low-fluence PA data, values of η ) 13.9 K m2 J-1 and T0 ) 293 K were used as fixed parameters for the fitting procedure, yielding, in this case, an activation energy of Ea ) 0.49 eV for the best fit; this is only slightly higher than the value of Ea ) 0.37 eV obtained from a best fit to the PA data (Figure 5). The apparent signal saturation for fluences above ∼600 J/m2 may be related either to a saturation of the CMOS camera chip or to a screening of gaseous domains within the rear side of the plume by front-side domains. Notably, first signals above the noise level are obtained at about the same fluence (H ) 50 J/m2) at which also the first photoacoustic signals are derived (H ∼60 J/m2). In both cases, this is about a factor of 3-4 below the ion detection threshold, clear evidence for the high sensitivity of the two employed techniques. The number of discrete particles that are observed in the scattered-light images is displayed in Figure 8(b). A linear regression analysis of these data [red solid line in Figure 8(b)] reveals that the average number of particles increases only moderately with fluence. Moreover, the emission of droplets appears to happen in a stochastic manner, and the number of particles identified in the scattered-light images varies consider-

Rohlfing et al.

Figure 8. Fluence dependences of gas and particle ejection. (a) Average intensity of the dark-field images as a function of laser fluence, recorded 30 ns after the desorption laser pulse (early phase of the plume development). The solid blue line is a fit according to the quasithermal model (eq 5) with Ea ) 0.49 eV, η ) 13.9 K m2 J-1, and T0 ) 293 K. Each data point represents one single desorption event. (b) Number of distinct particles observed in the scattered-light images as a function of laser fluence, recorded 500 ns after the desorption laser pulse (late phase of the plume development). Each data point represents one single desorption event. The solid red line displays the result of a linear regression analysis, yielding a correlation coefficient of R ) 0.4.

ably. On average, 8 ( 5 (standard deviation) particles are detected per single exposure. UV-Laser Postionization. Figure 9 displays two examples of postionization mass spectra obtained from the neurotensin/ NBA sample. The spectra were recorded with a delay time between the desorption and the postionization laser pulse of ∆t ) 500 ns and a distance between the sample surface and the PI laser beam waist of ∆z ) 50 µm. The spectrum shown in Figure 9(a,b) was recorded at a PI laser fluence of HPI ) 1200 J/m2; the desorption laser fluence was adjusted to ∼150 J/m2, below the threshold for direct ion generation by single-laser MALDI. Molecular neurotensin ions are not detected by PI. These negative results also hold for all other tested PI laser fluence and delay time/distance combinations (∆t ) 0-10 µs; ∆z ) 10-500 µm; HPI ) 50-3500 J/m2). A few further analytes (e.g., insulin, cytochrome c, bovine serum albumin, a monoclonal antibody, and a DNA 20mer) were also tested but did not produce any discernible molecular ion signals either. The low mass range, displayed in Figure 9(b), is dominated by a variety of species that can largely be assigned to fragment ions of the NBA matrix; some tentative assignments are indicated. Neurotensin-derived fragment ions are not evident in this mass range, although small fragment species may be obscured by the intense matrix background.

Desorption Process in UV-MALDI with a Liquid Matrix

Figure 9. Postionization mass spectra of the neurotensin/NBA sample. (a) Mass spectrum acquired with a postionization laser fluence of HPI ) 1200 J m-2; the arrow indicates the position at which protonated neurotensin ions would be detected. (b) Expanded view of the matrix ion region; presumable chemical identities are indicated for major NBAderived ion species. (c) Mass spectrum taken at a postionization laser fluence of HPI ) 220 J m-2. Mass spectra were recorded with a distance ∆z between the NBA sample surface and the postionization laser beam waist of 50 µm and a delay time between the two laser pulses of 500 ns. The desorption laser fluence was adjusted to ∼150 J m-2 in both cases. Each spectrum represents the average of 25 single laser exposures.

The spectrum shown in Figure 9(c) was recorded at a lower postionization laser fluence of 220 J/m2. The extensive fragmentation of the NBA molecules into (partially) small fragment species [Figure 9(b)] is substantially reduced under these conditions. The base peak at 124 Da can be attributed to the [M - NO + H]+ species, resulting from a photochemical cleavage of nitric oxide from the NBA matrix molecule.63 At the higher PI fluence, protonated NBA ions [M + H]+ are detected in some abundance [Figure 9(b)]; remarkably, this ion species is not observed at the lower PI fluence. While postionized molecular neurotensin species were in no case detected, a few neurotensin-derived fragment ion species can be differentiated in the m/z range between ∼180 and 300 [Figure 9(c)] when the plume is irradiated at low PI fluences; these ion signals are not detected if pure NBA is used as sample (data not shown). Further identification of these signals and their PI-fluence dependence was beyond the scope of the present work. The signal of the [M - NO + H]+ ion was generally detected with

J. Phys. Chem. C, Vol. 114, No. 12, 2010 5377 a very low shot-to-shot variation, suggesting that it essentially originates from the well-reproducible gas plume and not from the few statistically distributed particles. Notably, a direct evidence of a cleavage of NO from the NBA matrix molecules was not found in the single-laser MALDI mass spectra of NBA. Plume Dynamics. Making use of the prominent [M - NO + H]+ PI ion signal, the postionization technique can readily be applied for probing the expansion dynamics of the gaseous fraction of the NBA plume. Figure 10(a) displays the intensity of the [M - NO + H]+ postionization signal, serving as a measure of the local gas density as a function of ∆z (0-100 µm) and for delays ∆t between 50 and 400 ns. For this set of experiments, the desorption laser fluence was adjusted to a value of ∼150 J/m2, just below the MALDI ion detection threshold, and the PI laser fluence was held fixed at 220 J/m2. Expectedly, for the lowest ∆t of 50 ns, the plume is well-localized within 10-20 µm above the sample surface. With time, the plume moves away from the surface. The obvious broadening of the curves with increasing delay time reflects the velocity distribution of the molecules resulting in a thinning out of the plume with time. The dependence of the PI signal on the delay time ∆t is shown in Figure 10(b) for values of ∆t ranging from 0 to 1000 ns and for selected distances ∆z between the sample surface and the postionization laser beam waist ranging from 10 to 75 µm. For the shortest ∆z of 10 µm, the ion signal commences almost instantaneously, and small ion signal intensities are already detected after a few nanoseconds. The PI signal increases strongly with increasing delay time and reaches a maximum at ∆t ∼ 100 ns. When the PI laser is moved to probe the plume at larger distances ∆z, both the rising edge and the maximum of the curves are shifted toward longer delay time. A broad shoulder toward longer delay times is generally found for all probed distances ∆z until the signals approach the detection limit at about 1000 ns for all probed distances ∆z. Notably, for the ion intensity-delay time data recorded at the shortest distance ∆z of 10 µm, this fine feature even evolves into an additional second broad maximum at ∼300 ns as indicated by the arrow in Figure 10(b). Plume Velocities. The expansion velocity of the plume front can be estimated from the FFI and the PI data. Using all available FFI images and a subsequent linear regression analysis, an average value of the plume front expansion velocity of ∼1100 m/s is obtained. Almost identical values of ∼1000 m/s are derived if the PI data are evaluated. Interestingly, within the limits of the experimental accuracy, the values for the velocity of the plume front are identical for the three applied desorption laser fluences of 130, 150, and 350 J/m2. This result is consistent with most previous studies addressing the fluence dependence of the velocities of MALDI-generated molecules and ions.20 Those studies also showed that analyte ions typically exhibit higher average velocities than neutral matrix molecules. While the latter exhibit most probable values close to 500 m/s, most probable ion velocities reach values close to 1000 m/s.64-66 It may, therefore, be speculated that the detected MALDI ions are essentially stemming from the front part of the expanding MALDI plume.67 The bimodal velocity characteristic renders it problematic to derive average velocity values. However, the temporal development of the maximum in the PI data of Figure 10(b) may be used as a rough indicator for the center-of-mass velocity of the gaseous fraction of the plume. Derived to ∼750 m/s, this value lies within the most probable velocities for desorbed (neutral) matrix molecules and analyte ions as determined in previous studies.20 Assuming that there is no

5378

J. Phys. Chem. C, Vol. 114, No. 12, 2010

Rohlfing et al.

Figure 10. (a) Intensity of the postionization signal [M - NO + H]+ of NBA at m/z 124 as a function of the distance ∆z between the surface of the NBA drop and the postionization laser beam waist and for different delay times between the two laser pulses. (b) PI signal intensity as a function of the delay times between the desorption and the postionization laser pulse and for a set of distances ∆z between the surface of the NBA drop and the postionization laser beam waist. The PI laser fluence was adjusted to 220 J/m2 in both cases, while the desorption laser fluence was set to ∼150 J/m2, just below the MALDI ion detection threshold. Each data point represents the average of 25 single-shot spectra. To indicate the signal variation from shot to shot, an error bar is included in (a), representing the standard deviation as derived from 10 single irradiations.

significantly delayed particle emission with respect to the laser pulse, which appears reasonable given the fast cooling of the excitation volume due to heat conduction, the velocity of the distinct droplets can be calculated using the distance from the surface at which these droplets are detected in the scatteredlight images. Below the ion detection threshold, the calculated particle velocities range from a few meters/second to a maximum of ∼200 m/s, while above the threshold these velocities are somewhat higher and reach values of up to ∼300 m/s. Comparison with Molecular Dynamics (MD) Simulations. The observation of two distinct fluence regimes that display a notably different dependence of the material ejection on laser fluence is a very interesting feature of the PA data. A division into two regimes has been predicted in various MD work carried out for simulated laser excitation of a “MALDI matrix sample” under both stress and non-stress confinement conditions.44,59 These simulations predicted a transition from a molecular evaporation (i.e., an ejection of single molecules from the surface of the liquid) to a logarithmic fluence dependence reflecting bulk material ablation by phase explosion according to the photoablation model of eq 7.44,68 Notably, the amount of the overall ejected material was found to increase strongly at the transition threshold, while the number of ejected monomeric molecules does not show any sudden increase, but rather exhibits the quasithermal fluence dependence in both regimes. The dependence of NBA material ejection on laser fluence shows fine features that contrast with the results of the MD simulations. On the one hand, while an ejection of a few small droplets is already observed in the lower fluence regime, the material ejection-fluence relation otherwise shows clear characteristics of the quasithermal model of eq 5, in line with the MD simulations. On the other hand, for the upper fluence range, a clear deviation between the experimental data and the phase explosion relation of eq 7 is found. Furthermore, no jump in the amount of ejected material is found at the transition threshold. Apart from other simplifications that are unavoidable for the necessary large-scale simulations, one possible reason for these discrepancies could be the differences in the excitation pulse duration, as well as their spatial and temporal shapes, used in the experiments (τL ) 5 ns, temporal profile: near-Gaussian; spatial profile: near-Gaussian) and the simulations (square

excitation pulses with a duration of 150 ps exhibiting a flat-top spatial profile). Photochemical fragmentation of molecules has also been included more recently in related MD work.69,70 These studies showed a decrease in the fluence necessary to evoke material ablation if molecular dissociation processes occur within the excited volume, a finding which underlines the potentially relevant role of a photochemical cleavage of NO from the NBA molecules (see below). Velocities as predicted by the MD simulations show values up to above 1000 m/s for the plume front, the latter being essentially comprised of gaseous particles. Larger ejected clusters were found to exhibit velocities ranging from a few to a few hundred meters/second. These findings are compatible with the experimental data. Comparison with IR-MALDI. A division into two distinct fluence regimes was also revealed in our previous photoacoustic study on IR-MALDI with a glycerol matrix.41 In that work, an approximately linear dependence of the material ejection on laser fluence was found in the low-fluence range. Above the IRMALDI threshold, the PA signal-fluence relationship changed into a superposition of the linear relation and material ejection according to the photoablation model of eq 7. Moreover, in the case of (approximated) stress confinement conditions, realized by employment of an IR-laser with a pulse duration of 6 ns, the IR-MALDI ion detection threshold coincided strikingly well with the change in the material ejection mechanism.41 One is thus tempted to speculate that for the given matrix systems ion formation might in both wavelength regimes be supported by the sudden rupture of material. In this regard, it is interesting to note that, chemically, both the glycerol and NBA matrix differ from many common crystalline UV-MALDI matrices which are aromatic carboxylic acid compounds that carry an additional hydroxyl group. Although other functional matrices exist, an interplay between the hydroxyl and the carboxyl group has been shown to support analyte ionization by proton transfer.71 Interestingly, the two liquid matrices also seem to require a slightly higher deposited energy density to provoke ion generation than crystalline matrices with similar optical absorption properties.8,9 Possibly, the large density of nucleation sites available in the case of the crystalline matrix may account for these differences.

Desorption Process in UV-MALDI with a Liquid Matrix As for the UV-MALDI case, FFI snapshots displayed a coejection of gaseous glycerol and distinct particles: Due to the larger laser penetration depth, more and sizably larger particles, exhibiting diameters up to a few micrometers, are ablated in the IR-MALDI case.39 This large size of the IR-laser generated glycerol droplets allowed for postionization of intact analyte molecules out of the droplets by a second IR-laser (presumably by a second “normal” MALDI event).42 Several aerosol mass spectrometry studies also demonstrated the generation of intact analyte ions (peptides) from single airborne particles, that were coated with a classical UV-MALDI matrix and irradiated with a typical UV-MALDI laser.72-76 In some of these studies, NBA was used as matrix material.74,75 However, the particles analyzed in these studies were considerably larger (∼1 µm in diameter) and exhibited much larger molar analyte-to-matrix (A:M) ratios in the 1:10-1:100 range. It seems, therefore, reasonable to assume that the failure to produce postionization mass spectra of the UV-MALDI-generated NBA droplets may be related to a too low analytical sensitivity, caused by their small size and the low analyte concentration (A:M, ∼1:104 in the initial sample). UV-photoionization of gaseous peptides species with a second laser, on the other hand, seems to result in direct photodissociation. This result is in agreement with previous attempts to establish the combination of UV-MALD (i.e., the generation of neutral analyte molecules) with UV-laser postionization.20 Only when peptides are ionized by single photons using vacuumUV light,77 or are sufficiently cooled prior to resonanceenhanced multiphoton ionization (REMPI)78 (e.g., by introducing them into a supersonic gas beam), can their intact analysis eventually become possible. Two-Step Desorption Model. The data obtained in the three complementary experiments provide valuable insight into the MALDI mechanisms and can be used to test and/or develop theoretical models of the MALDI process, even if the proposed model does not provide a fully conclusive picture. For the low-fluence regime, molecular ejection from the heated NBA surface appears to dominate the MALDI process. The corresponding quasithermal desorption model (eq 5) is particularly well-reproduced by the dark-field FFI image data [Figure 8(a)]. For the elevated fluences in the upper regime it seems reasonable to assume that the NBA system will at some point during the laser pulse become exceedingly unstable and homogeneous nucleation (i.e., phase explosion) will occur, leading to spontaneous relaxation of the superheated liquid into a mixture of vapor, small clusters, and particles.55,79 Due to the exponential decay of the absorbed laser energy with the sample depth, this phase explosion would affect the top (i.e., hottest) sample layer first during the time course of the laser pulse. Lasting deposition of energy in the remainder of the laser pulse would lead to additional heating of the ejecta and cause further material disintegration. Successive overheating of the remaining surface layer would result in further phase explosions. This layer-by-layer ablation would last as long as laser energy is being deposited (i.e., for the duration of the laser pulse). An interesting finding of this study is the observation of small droplets in the FFI data throughout the tested fluence range and the only weak increase of their number with fluence. Since a molecular ejection process would produce exclusively gaseous ejecta, the observation of discrete droplets even at low laser fluences indicates that a second process could be effective. The weak but monotonous increase (to a first order) with fluence indicates that the same process is likely to be responsible for droplet formation in both fluence regimes. Material ejection by

J. Phys. Chem. C, Vol. 114, No. 12, 2010 5379 phase explosion would explain the formation of droplets. However, the PA data do not corrobate the occurrence of phase explosion in the low fluence regime, and the FFI data do not reveal the expected dependences for this material ejection pathway in the two fluence regimes. As discussed above, the average size of droplets formed by the layer-by-layer phase explosion may simply be too small to allow for their detection. Droplets, large enough to be visualized by FFI, could, in principle, be formed by extensive recondensation of ejected material in the expanding material plume. However, because the expansion takes place in a vacuum, this assumption does not seem reasonable. A possible additional material ejection pathway would be “heterogeneous boiling”. Immediately after the laser pulse, temperatures in the surface elements of the remaining sample must in fact be expected to exceed the boiling temperature in both fluence regimes, before heat dissipation leads to a relaxation of the metastable material, not long after the laser pulse. Even if “normal boiling” is often regarded as unlikely for nanosecond excitation of a superheated, shallow layer with a thickness of less than 100 nm,80,81 such processes may be boosted by the photochemical cleavage of nitric oxide from excited matrix molecules,19,82 as is indicated by the postionization mass spectra. The generated gas could provide a multitude of artificial nucleation sites for the boiling process. Since boiling must then be expected to occur not only in surface-near volume elements but also in a certain depth (“subsurface boiling”), the overlying sample layers may eventually be expelled by the expanding vapor bubbles and finally form the observed distinct droplets. A critical point that argues for further refinement of the model considerations is that, although eq 8 provides an excellent fit to the PA data, it neglects optical absorption by the material which already underwent phase explosion. In fact, the original movingheat-source model60,61 assumes that the ablated layers move out of the laser beam path fast enough to not absorb substantial fractions of the further incoming energy. This assumption may be justified for millisecond long excitation pulses, yet is clearly not fulfilled for the nanosecond pulses used here. During the 5 ns of the Nd:YAG laser pulse, the ejected material will travel only a few micrometers and is, hence, still very well-localized in the direct vicinity of the remaining sample surface for the entire pulse duration. One possible reason for a dynamically reduced optical absorption could be photobleaching. In fact, at a fluence of 170 J/m2, about each second surface matrix molecule will absorb a photon in the center of the laser focus. Unfortunately, the S1 singlet lifetime of NBA is not known. For typical crystalline MALDI matrices, S1 lifetimes in between a few tenths of one nanosecond to about 10 ns have been determined.83 Dynamic absorption effects could thus possibly modify the material ejection-fluence relationship. Notably, as was argued by Knochenmuss and Zhigilei, the onset of phase explosion might be delayed by energy storage in excited matrix molecules, and the material will, thus, eventually only be fully heated once all electronic energy is converted into heat.84 A detailed investigation of these effects was beyond the scope of the present work. Altogether, while the above “two-step” model considerations can explain relevant aspects of the experimental findings, a fully consistent model picture cannot, as of yet, be derived. Conclusion Utilizing three complementary methods for a time- and spatially resolved investigation of the UV-MALDI process with a liquid NBA matrix, we obtained valuable novel insights into

5380

J. Phys. Chem. C, Vol. 114, No. 12, 2010

the MALDI mechanisms. To our knowledge, this constitutes the first example of a study in which a PA detection scheme has been successfully employed to record the recoil momentum of material ejection under UV-MALDI conditions with nanosecond time resolution. The photoacoustic data revealed a sudden change in the ejection mechanism at a threshold fluence that coincides strikingly well with the MALDI ion detection threshold. Fast-flash imaging experiments illustrated that the MALDI plume consists of a mixture of gaseous components and a few small droplets throughout the whole of the tested fluence range. Emission of the dominating gaseous fraction in the low fluence regime is described well by a quasithermal model. Above the threshold, the amount of ejected material increases with a different fluence dependence that can be reproduced by a layer-by-layer ejection mechanism according to a moving-heat-source model. Normal heterogeneous boiling of superheated volume elements presumably adds a third mechanism of material emission and is likely to be boosted by a photocleavage of NO from NBA. Ion signals exhibit an exponential increase with fluence, in agreement with former MALDI results. The findings derived in this work improve our understanding of the UV-MALDI mechanisms. However, the possibility that some features may be unique to the investigated liquid matrix system cannot be excluded. Some caution should therefore be taken in generalizing our findings to more general MALDI systems. Acknowledgment. This work has been carried out in partial fulfillment of the requirements for the Ph.D. degree of AR and AL at the University of Mu¨nster. AR is grateful to the University of Mu¨nster for a Ph.D. grant. We would like to thank Vasan Venugopalan, Leonid Zhigilei, Richard Knochenmuss, and Barbara Garrison for numerous helpful discussions, Alfred Vogel for expert assistance with the optical setup for the imaging experiments, Ulrich Ro¨hling for technical support with the electronics, Rebekka Sto¨ckel for help with the data evaluation, and Joanne Y. Yew for critical proofreading of the manuscript. Supporting Information Available: Two animated video clips of the dark-field FFI image series below and above the ion detection threshold fluence (as shown in Figure 6), an additional figure with an intensity line scan through one single particle imaged in scattered-light geometry, a table containing various physicochemical properties of NBA and related compounds, and a derivation of the moving-heat-source model. This material is available free of charge via the Internet at http:// pubs.acs.org. References and Notes (1) Karas, M.; Bachmann, D.; Hillenkamp, F. Anal. Chem. 1985, 57, 2935. (2) Karas, M.; Bachmann, D.; Bahr, U.; Hillenkamp, F. Int. J. Mass Spectrom. Ion Proc. 1987, 78, 53. (3) Karas, M.; Hillenkamp, F. Anal. Chem. 1988, 60, 2299. (4) Beavis, R. C.; Chait, B. T. Rapid Commun. Mass Spectrom. 1989, 3, 233. (5) Overberg, A.; Karas, M.; Bahr, U.; Kaufmann, R.; Hillenkamp, F. Rapid Commun. Mass Spectrom. 1990, 4, 293. (6) Cramer, R.; Haglund, R. F.; Hillenkamp, F. Int. J. Mass Spectrom. Ion Proc. 1997, 169/170, 51. (7) Caldwell, K. L.; McGarity, D. R.; Murray, K. K. J. Mass Spectrom. 1997, 32, 1374. (8) Dreisewerd, K.; Schu¨renberg, M.; Karas, M.; Hillenkamp, F. Int. J. Mass Spectrom. Ion Proc. 1995, 141, 127. (9) Feldhaus, D.; Menzel, C.; Berkenkamp, S.; Hillenkamp, F.; Dreisewerd, K. J. Mass Spectrom. 2000, 35, 1320. (10) Dreisewerd, K.; Schu¨renberg, M.; Karas, M.; Hillenkamp, F. Int. J. Mass Spectrom. Ion Proc. 1996, 154, 171.

Rohlfing et al. (11) Menzel, C.; Dreisewerd, K.; Berkenkamp, S.; Hillenkamp, F. J. Am. Soc. Mass Spectrom. 2002, 13, 975. (12) Chen, X.; Carroll, J. A.; Beavis, R. V. J. Am. Soc. Mass Spectrom. 1998, 9, 885. (13) Menzel, C.; Dreisewerd, K.; Berkenkamp, S.; Hillenkamp, F. Int. J. Mass Spectrom. 2001, 207, 73. (14) Ehring, H.; Karas, M.; Hillenkamp, F. Org. Mass Spectrom. 1992, 27, 472. (15) Ehring, H.; Sundqvist, B. U. R. J. Mass Spectrom. 1995, 30, 1303. (16) Knochenmuss, R. J. Mass Spectrom. 2002, 37, 867. (17) Karas, M.; Glu¨ckmann, M.; Scha¨fer, J. J. Mass Spectrom. 2000, 35, 1. (18) Zenobi, R.; Knochenmuss, R. Mass Spectrom. ReV. 1998, 17, 337. (19) Knochenmuss, R. Analyst 2006, 131, 966. (20) Dreisewerd, K. Chem. ReV. 2003, 103, 395. (21) Dreisewerd, K.; Berkenkamp, S.; Leisner, A.; Rohlfing, A.; Menzel, C. Int. J. Mass Spectrom. 2003, 226, 189. (22) Niu, S. F.; Zhang, W. Z.; Chait, B. T. J. Am. Soc. Mass Spectrom. 1998, 9, 1. (23) Strupat, K.; Karas, M.; Hillenkamp, F. Int. J. Mass Spectrom. Ion Proc. 1991, 111, 89. (24) Zhao, S.; Somayajula, K. V.; Sharkey, A. G.; Hercules, D. M.; Hillenkamp, F.; Karas, M.; Ingendoh, A. Anal. Chem. 1991, 63, 450. (25) Yau, P. Y.; Chan, T. W. D.; Cullis, P. G.; Colburn, A. W.; Derrick, P. J. Chem. Phys. Lett. 1993, 202, 93. (26) Gower, J. L. Biomed. Mass Spectrom. 1985, 12, 191. (27) Sze, E. T. P.; Chan, T. W. D.; Wang, G. J. Am. Soc. Mass Spectrom. 1998, 9, 166. (28) Cramer, R.; Corless, S. Proteomics 2005, 5, 360. (29) Armstrong, D. W.; Zhang, L. K.; He, L. F.; Gross, M. L. Anal. Chem. 2001, 73, 3679. (30) Tholey, A.; Heinzle, E. Anal. Bioanal. Chem. 2006, 386, 24. (31) Berkenkamp, S.; Menzel, C.; Karas, M.; Hillenkamp, F. Rapid Commun. Mass Spectrom. 1997, 11, 1399. (32) Dreisewerd, K.; Rohlfing, A.; Spottke, B.; Urbanke, C.; Henkel, W. Anal. Chem. 2004, 76, 3482. (33) Berkenkamp, S.; Kirpekar, F.; Hillenkamp, F. Science 1998, 281, 260. (34) Dreisewerd, K.; Mu¨thing, J.; Rohlfing, A.; Meisen, I.; Vukeliæ, Z.; Peter-Kataliniæ, J.; Hillenkamp, F.; Berkenkamp, S. Anal. Chem. 2005, 77, 4098. (35) Rohlfing, A.; Mu¨thing, J.; Pohlentz, G.; Distler, U.; Peter-Kataliniæ, J.; Berkenkamp, S.; Dreisewerd, K. Anal. Chem. 2007, 79, 5793. (36) Meisen, I.; Distler, U.; Mu¨thing, J.; Berkenkamp, S.; Dreisewerd, K.; Mathys, W.; Karch, H.; Mormann, M. Anal. Chem. 2009, 81, 3858. (37) Menzel, C.; Dreisewerd, K.; Berkenkamp, S.; Hillenkamp, F. Int. J. Mass Spectrom. 2001, 207, 73. (38) Cramer, R.; Burlingame, A. L. Rapid Commun. Mass Spectrom. 2000, 14, 53. (39) Leisner, A.; Rohlfing, A.; Ro¨hling, U.; Dreisewerd, K.; Hillenkamp, F. J. Phys. Chem. B 2005, 109, 11661. (40) Krutchinsky, A. N.; Loboda, A. V.; Spicer, V. L.; Dworschak, R.; Ens, W.; Standing, K. G. Rapid Commun. Mass Spectrom. 1998, 12, 508. (41) Rohlfing, A.; Menzel, C.; Kukreja, L. M.; Hillenkamp, F.; Dreisewerd, K. J. Phys. Chem. B 2003, 107, 12275. (42) Leisner, A.; Rohlfing, A.; Berkenkamp, S.; Hillenkamp, F.; Dreisewerd, K. J. Am. Soc. Mass Spectrom. 2004, 15, 934. (43) Gournay, L. S. J. Acoust. Soc. Am. 1966, 40, 1322. (44) Zhigilei, L. V.; Garrison, B. J. J. Appl. Phys. 2000, 88, 1281. (45) Vertes, A.; Gijbels, R.; Levine, R. D. Rapid Commun. Mass Spectrom. 1990, 4, 228. (46) Lu, Z.; Daridon, J. L.; Lagourette, B.; Ye, S. Meas. Sci. Technol. 1998, 9, 1699. (47) Oraevsky, A. A.; Karabutov, A. A. Proc. SPIE 2000, 3916, 228. (48) Kaufmann, R.; Kirsch, D.; Rood, H. A.; Spengler, B. Rapid Commun. Mass Spectrom. 1992, 6, 98. (49) Schu¨renberg, M.; Schulz, T.; Dreisewerd, K.; Hillenkamp, F. Rapid Commun. Mass Spectrom. 1996, 10, 1873. (50) Sigrist, M. W. J. Appl. Phys. 1986, 60, R83. (51) Westmacott, G.; Ens, W.; Hillenkamp, F.; Dreisewerd, K.; Schu¨renberg, M. Int. J. Mass Spectrom. 2002, 221, 67. (52) Ens, W.; Mao, Y.; Mayer, F.; Standing, K. G. Rapid Commun. Mass Spectrom. 1991, 5, 117. (53) Schu¨renberg, M.; Dreisewerd, K.; Kamanabrou, S.; Hillenkamp, F. Int. J. Mass Spectrom Ion Process. 1998, 172, 89. (54) Weast, R.; Grasselli, J. G., Eds. CRC Handbook of Data on Organic compounds, 2nd ed.; CRC Prss: Boca Raton, 1989. (55) Garrison, B. J.; Itina, B. J.; Zhigilei, L. V. Phys. ReV. E 2003, 68, 041501-1. (56) Nikitin, E. D.; Popov, A. P.; Yatluk, Y. G. J. Chem. Eng. Data 2007, 52, 315. (57) Martynyuk, M. M. Combust., Explos., Shock WaVes 1977, 13, 178. (58) Vogel, A.; Venugopalan, V. Chem. ReV. 2003, 103, 577.

Desorption Process in UV-MALDI with a Liquid Matrix (59) Zhigilei, L. V.; Garrison, B. J. Appl. Phys. A: Mater. Sci. Process. 1999, 69, S75. (60) Paek, U. C.; Gagliano, F. P. IEEE J. Quantum Electron. 1972, QE8, 112. (61) Bar-Isaac, C.; Korn, U. Appl. Phys. 1974, 3, 45. (62) Leisner, A. Ph.D. Dissertation, University of Mu¨nster: Germany, 2003. (63) Madhusudanan, K. P. J. Mass Spectrom. 1996, 31, 649. (64) Zhou, J.; Ens, W.; Standing, K. G.; Verentchikov, A. Rapid Commun. Mass Spectrom. 1992, 6, 671. (65) Puretzky, A. A.; Geohegan, D. B.; Hurst, G. B.; Buchanan, M. V. Phys. ReV. Lett. 1999, 83, 444. (66) Berkenkamp, S.; Menzel, C.; Hillenkamp, F.; Dreisewerd, K. J. Am. Soc. Mass Spectrom. 2002, 13, 209. (67) Puretzky, A. A.; Geohegan, D. B. Chem. Phys. Lett. 1998, 286, 425. (68) Srinivasan, R.; Braren, B.; Seeger, D. E.; Dreyfus, R. W. Macromolecules 1986, 19, 916. (69) Yingling, Y. G.; Garrison, B. J. Chem. Phys. Lett. 2002, 364, 237. (70) Prasad, M.; Conforti, P. F.; Garrison, B. J. J. Appl. Phys. 2007, 101, 103113. (71) Krause, J.; Sto¨ckli, M.; Schlunegger, U. P. Rapid Commun. Mass Spectrom. 1996, 10, 1927. (72) Mansoori, B. A.; Johnston, M. V.; Wexler, A. S. Anal. Chem. 1996, 68, 3595.

J. Phys. Chem. C, Vol. 114, No. 12, 2010 5381 (73) McJimpsey, E. L.; Jackson, W. M.; Lebrilla, C. B.; Tobias, H.; Bogan, M. J.; Gard, E. E.; Frank, M.; Steele, P. T. J. Am. Soc. Mass Spectrom. 2008, 19, 315. (74) Fan, X.; Murray, K. K. Appl. Surf. Sci. 2009, 255, 6297. (75) Stowers, M. A.; van Wuijckhuijse, A. L.; Marijnissen, J. C. M.; Scarlett, B.; van Baar, B. L. M.; Kientz, C. E. Rapid Commun. Mass Spectrom. 2000, 14, 829. (76) Jackson, S. N.; Mishra, S.; Murray, K. K. Rapid Commun. Mass Spectrom. 2004, 18, 2041. (77) Edirisinghe, P. D.; Moore, J. F.; Calaway, W. F.; Veryovkin, I. V.; Pellin, M. J.; Hanley, L. Anal. Chem. 2006, 78, 5876. (78) Ledingham, K. W. D.; Singhal, R. P. Int. J. Mass Spectrom. Ion Proc. 1997, 163, 149. (79) Miotello, A.; Kelly, R. Appl. Phys. A: Mater. Sci. Process. 1999, 69, S67. (80) Kelly, R.; Miotello, A. Appl. Surf. Sci. 1996, 205, 96–98. (81) Kelly, R.; Miotello, A. J. Appl. Phys. 2000, 87, 3177. (82) Talroze, V. L.; Jacob, R. J.; Burlingame, A. L.; Baldwin, M. A. AdV. Mass Spectrom. 2001, 15, 481. (83) Lu¨demann, H. C.; Redmond, R. W.; Hillenkamp, F. Rapid Commun. Mass Spectrom. 2002, 13, 1287. (84) Knochenmuss, R.; Zhigilei, L. V. J. Phys. Chem. B 2005, 109, 22947.

JP905251R