Development and characterization of a laser-induced acoustic desorption source

We characterize the desorbed molecular plume using strong-field ionization (SFI) from a focused femtosecond Ti:Sapphire laser as a universal probe Calvert et al. (2012); Teschmit et al. (2017).

The observed TOF-MS of phenylalanine (PA) is shown in Figure 2 and compared to a literature spectrum obtained using electron-impact ionization (EI) Linstrom and Mallard (2017). Both spectra are normalized to the most abundant fragment ion at mass 74 u, corresponding to loss of a benzyl-radical fragment. It is evident that both ionization schemes strongly induce fragmentation, however, we note that using SFI a significant contribution from intact PA is observed at 165 u; this could even be enhanced using shorter duration laser pulses Calvert et al. (2012). We observe no evidence for the production of larger clusters of PA, and hence attribute this channel to desorption of intact PA monomers. Furthermore, we observe an additional fragmentation peak at 28 u in the SFI data, corresponding to CNH₂, e. g., C-NH${}_{2}^{+}$ or HC=NH⁺, fragment ions, which is absent in the EI mass spectrum. These spectra clearly demonstrate the production of intact PA following desorption from the foil band. We do not observe the emission of any tantalum atoms or clusters, which would easily be ionized by the SFI probe, since the ionization potential of tantalum is lower than of PA. This indicates that the desorption laser does not penetrate through the foil band nor ablate metal from the foil by other means.

To assess the depletion of sample from the foil and determine the required moving speed for sample replenishment, we measured the parent ion yield as a function of the number of desorption laser shots onto the same spot. The resulting abundances are shown in Figure 3 a, where the solid line represent a power-law fit of the form $y=A\times{}x^{n}$, with an exponent of n = $-0.68\pm 0.03$. We observe a rapid decay of signal, reaching around 10% after 330 desorption laser shots. Similar power-law behavior has previously been observed and rationalized with the existence of several isolated desorption centers on the foil Zinovev et al. (2007). This is consistent with our observation of many large crystalline islands, see supplementary information, many of which fall within the desorption laser spot size.

During further data collection the foil band is continuously moved at 50 µm/s, corresponding to a movement to a new sample spot every $\mathord{\sim}120$ desorption laser shots. The corresponding shot-to-shot signal stability for the moving foil band is shown in Figure 3 b. The signal exhibits large fluctuations with a single shot standard deviation of 70% of the mean value. No long-term drift of the overall signal levels is observed. Averaging over 50 desorption laser shots reduces the standard deviation to below 10%, as indicated by the red markers and error bars in Figure 3. Further data points in this manuscript are typically averaged over 1200 desorption laser shots, resulting in a standard deviation of $\mathord{\sim}2.5$%.

In the following we investigate the spatial extent, density, velocity, and translational temperature of the “plume” of molecules desorbed from the foil band. We estimate absolute number densities from ion counting measurements and the known interaction volume as defined by our ionization laser. In Figure 4 a we show the measured number density of parent ions in the center of the desorbed plume as a function of distance from the foil band. We note that the shown densities are lower limits, since their calculation assumes an ionization efficiency of 1 for SFI and considers the measured intact parent ions only, such that any fragmentation induced by the SFI probe will reduce the derived density. The obtained densities exhibit approximately an inverse-square-law behavior with distance from the foil, since the expansion along the laser propagation direction is not reflected in the measurements due to the large Rayleigh length of the ionization laser ($z_{R}\approx 38$ mm). We note that the data point closest to the foil band for the measurement at 0.64 J/cm² shows a significantly lower than expect density, which we can only explain with a lower density of molecules attached on the desorption foil band for this measurement, due to some instability during the aerosolization process.

We assess the spatial extent of the plume, i. e., the transverse profile, by translating the ionization laser in height along the $y$-axis (Figure 1), across the plume of molecules. This is shown in Figure 4 b for three distances between the foil band surface and interaction point. The initial profile close to the foil band is very narrow, with a FWHM of $\mathord{\sim}0.6$ mm after 0.5 mm of free flight. The plume then rapidly spreads out, reaching a FWHM of around 2 mm after 2.5 mm propagation and within 4.5 mm of free flight the extent of the plume exceeds the spatial acceptance of the TOF ion optics (indicated by the gray shading in Figure 4 b), such that no accurate data can be measured at larger separations. This rapid diffusion of the plume in space is consistent with the fast drop in density observed as the distance between the foil band and the interaction point is increased, Figure 4 a, and indicates rapid diffusion of the molecular plume in space following desorption from a well-defined spot defined by the desorption laser profile.

To investigate the longitudinal extend and velocity of the plume of desorbed molecules we measure mass spectra as a function of delay between the desorption and ionization lasers, and at different distances from the foil band. Results for the intact-parent-ion yield following desorption with a fluence of 0.8 J/cm² are shown in Figure 5 a. Similar data for other desorption fluences are shown in the supplementary data. It is very clear that even when the interaction point is very close to the foil band a broad temporal profile is observed, lasting several tens of µs, much broader than the 8 ns duration of the desorption-laser pulse. At larger distances from the foil band these distributions widen considerable more, demonstrating that during free flight through the vacuum chamber the plume spreads out also in the longitudinal direction. We identify two physical origins for the observed profiles and their temporal evolution; (i) the desorption process itself that does not release molecules at one instant in time, but with a certain temporal and kinetic energy distribution and (ii) the propagation of molecules in free flight with a certain finite translational velocity distribution. Whereas (i) contains information about the physical desorption mechanism from the foil, the translational velocity spread from (ii) corresponds to the translational temperature in the moving frame of the molecules.

In order to accurately fit the measured data, one needs to convolute the initial desorption time distribution from the foil band with the Maxwell-Boltzmann free-flight propagation. Since so far no quantitative model is available to describe this desorption process accurately, we take the experimental data measured closest to the foil band, i. e., 0.5 mm, as a measure of the initial desorption time distribution and numerically convolute this with the Maxwell-Boltzmann model of the free-flight propagation. Details of this convolution procedure and the Maxwell-Boltzmann model are given in the supplementary information. We then perform a global fit of the data for all propagation distances $l$ simultaneously using a common temperature $T$ and offset velocity $v_{0,z}$, while we introduce only a single linear scaling parameter for the different data sets. The latter essentially accounts for the drop in intensity along the probed center-line of the plume. The results of this fit for a desorption laser fluence of 0.8 J/cm² are shown as solid lines in Figure 5 a, data for other fluences is provided in the supplementary information. The obtained translational temperatures and forward velocities are summarized in Table 1.

We observe a strong, nearly linear, dependence of the translational temperature of the molecular plume on the fluence of the desorption laser. Even at the lowest fluence used a translational temperature of nearly 600 K is obtained. In the current experimental setup using SFI we cannot measure the internal (vibrational or rotational) temperature directly. However, given the large density of states in systems such as phenylalanine, and the microsecond timescales of the desorption process, we can assume a large degree of thermalization between the different degrees of freedom. Thus the measured translational temperatures can be considered as a good indicator of the internal temperature of desorbed molecules.

Unlike the temperature, the observed forward velocity appears to be approximately constant for the different desorption laser fluences. The slightly elevated velocity for the measurement at 0.64 J/cm² could be due to instabilities in the sample preparation for this measurement, as mentioned above. Similar observations of identical forward velocity have been previously reported Zinovev et al. (2007); Shea et al. (2007). This invariability of the velocity with desorption laser fluence suggests that this might be determined by material properties of the substrate and the molecular sample.

Figure 5 b shows the yield of intact parent ions as a function of desorption laser-ionization laser delay for different desorption fluences. While the peaks of the distribution overlap in time, the distribution is significantly broader for higher fluences. These observations fully support our finding of a constant translational velocity, but increasing translational temperature as the desorption laser fluence is increased (vide supra).

In how far the observed fragmentation is due to the desorption or the SFI process is hard to assess from the mass spectra in Figure 2 alone. In order to disentangle these contributions, we collect mass spectra for different ionization and desorption laser intensities.

Figure 6 a shows the ion yield for the PA parent and the three dominant fragment ions as a function of ionization laser intensity, with all ion channels showing a steep increase with increasing laser intensity. These data were fit with a power-law dependence of the form $A\times{}x^{n}$. Figure 6 b further shows the ratio of fragment-to-parent ion abundances for the three dominant fragment ions, i. e., comparing the relative abundances of the two respective channels. We observe only a very slight increase in fragmentation as the laser intensity increases, in good agreement with previous studies suggesting that SFI induced fragmentation is very sensitive to the employed pulse duration, but not the intensity Calvert et al. (2012).

Figure 6 c shows the dependence of ion yields on the intensity of the desorption laser and Figure 6 d the corresponding fragment-to-parent ratios. The overall measured ion abundances are again well described by a power-law fit and show a steep increase for higher intensities, especially noticeable for fragment ions. This is confirmed by the fragment-to-parent ratios, which also significantly increase with laser intensity, indicating enhanced fragmentation. Thus, the desorption-laser interaction clearly induces fragmentation, either directly during the desorption process or thereafter, but prior to ionization, i. e., as molecules travel through the vacuum chamber toward the interaction point. To test the latter, we recorded mass spectra at different distances behind the foil band, changing the laser-laser delay such that we always probe the highest density part of the molecular plume, i. e., we follow the center of the plume as it travels through the vacuum chamber.

This data is shown in Figure 7 a, collected for distances of 0.5–10.5 mm between the foil band and the interaction point, which corresponds to flight times of around 0–50 µs. Over this distance we observe no significant increase in fragmentation, indicating that fragmentation occurs on much faster timescales, i. e., most likely during the desorption process itself, either while molecules are still attached to the metal substrate or very shortly after desorption into the gas-phase.

We now consider the distribution of fragments within a single plume coming from the foil band, i. e., if the fragmentation changes depending on which part of the plume is observed. This is shown in Figure 7 b, where we plot the fragment-to-parent ratio for the most abundant molecular fragment ion as a function of desorption-laser-to-ionization-laser delay for a fixed distance from the foil band, i. e., 6.5 mm. We observe an initial peak in the fragment-to-parent ratio at the onset of desorption, i. e., the “front” part of the molecular plume, which then decreases on a timescale of tens of microseconds. These timescales are consistent with thermal processes, in particular we associate the observed distribution with the rapid heating of the foil band by the nanosecond laser pulse, causing increased fragmentation, followed by slow dissipation of the thermal energy, i. e., cooling down of the front surface and, hence, reduced fragmentation. Further evidence that the fragmentation occurs during the desorption process and that it is of a thermal nature comes from the comparison of the fragment-to-parent ratios throughout the plume for different desorption laser fluences, also shown in Figure 7 b. These clearly show that the highest degree of fragmentation occurs for the most intense desorption laser pulse. This is also consistent with the higher translational temperatures derived for these conditions. Once the foil band cools down, i. e., at longer desorption-laser-to-ionization-laser delays, the fragment-to-parent ratio approaches an asymptotic value independent of initial desorption conditions.

Several possible mechanisms have been suggested in the literature for the underlying physical processes occurring in the LIAD process Lindner (1991); Golovlev et al. (1997); Zinovev et al. (2007); Goodfriend et al. (2016). It is important to note that the experimental conditions for the different published LIAD-based molecule sources are very different; pulsed Calvert et al. (2012); Zinovev et al. (2007) and continuous Calegari et al. (2014, 2015) desorption lasers are used and sample preparation methods vary greatly, from the thick sample layer used here of $\sim$500 nmol/cm² Calegari et al. (2015); Borton et al. (2013), to intermediate thicknesses of tens of nmol/cm² Shea et al. (2006, 2007), to near-monolayer coverage in other studies Zinovev et al. (2007). As such, we do not aim to provide a general model for the LIAD mechanisms, but seek to explain our observations and compare these with previous studies where applicable.

One of the suggested desorption mechanisms, and indeed the origin of the term “acoustic desorption” Lindner (1991); Golovlev et al. (1997), is the direct momentum transfer from a shock wave induced by the desorption laser in the foil band to the sample molecules. Our data firmly rules out this mechanism for our molecule source. We observe a slow rise in molecular signal on the order of $\mathord{\sim}10~\text{\textmu{s}}$, see Figure 5, which is not compatible with molecules being “shaken off” by an impulse traveling through the foil, as this should lead to a sharp sudden onset of signal as the impulse reaches the front surface, followed by an immediate drop as the impulse is reflected on the surface. Additionally one might expect to observe a periodic revival of signal as the impulses bounces back and forth within the metal foil. We observe no evidence for this behavior. Furthermore, the travel time for a mechanical wave through a 10 µm tantalum foil is approximately 2 ns Rigg et al. (2014), significantly shorter than the delay we observe between the desorption laser impacting on the foil and molecules being desorbed. A purely acoustic desorption mechanism would, furthermore, not explain the observed increase in fragmentation for increased desorption laser fluences. Similar observations have been made previously for a pulsed LIAD setup, and the “shake off” mechanism similarly discredited Zinovev et al. (2007).

The other conceptually simple mechanism is a simple thermal one; the incident laser pulse heats up the material from the backside and this thermal energy then diffuses to the front of the foil where it heats up molecules and they eventually desorb. However, the observation that the velocity and, therefore, the kinetic energy of desorbed molecules is independent of the incident desorption laser power and thus surface temperature is not compatible with a purely-thermal desorption model.

The observation that the kinetic energy of desorbed molecules is independent of desorption laser fluence indicates that this is determined by material properties of the foil substrate and/or the molecular sample. This observation, along with the increase in translational temperature in the moving frame, is consistent with a desorption model proposed by Zinovev et al. Zinovev et al. (2007). They explain the LIAD process by an introduction of surface stress between the substrate and the molecular sample – located in isolated islands on the substrate – due to the acoustic and/or thermal wave created by the desorption laser. This surface stress can lead to elastic deformation, decomposition, and cracking of sample islands on the foil band and, eventually, to desorption of molecules. In this conceptual model the kinetic energy transferred to a desorbing molecule is independent of the total incident laser power, and rather depends on the intrinsic characteristics of a given sample island and substrate. A higher laser fluence leads to the introduction of more surface stress and the formation of more cracks and deformation sites, leading to an increase in molecular signal, but does not influence the amount of kinetic energy per molecule. At the same time we note that due to thermal conductivity the higher temperature of the substrate reached for higher desorption laser fluences will also heat up deposited sample molecules due to thermal conduction, leading to internally hotter molecules, increased fragmentation as well as higher translational temperatures.

While it is difficult to theoretically model the amount of energy transferred to each desorbed molecule, Zinovev et al. provide a simple formula to estimate the energy per analyte molecule based on material properties and thermal stress theory Zinovev et al. (2007). Based on this we estimate 25–100 meV of energy per molecule for temperature differences of $\Delta{}T=100$–200 K. ¹¹1We evaluated the energy release per molecule based on the known physical constants for anthracene Bondi (1966), since data for PA was not available. The thermal expansion coefficient of the film is assumed to be $2.8\times 10^{-4}~\mathrm{K}^{-1}$. This is well within the range of the measured kinetic energy per molecule which is, based on the average velocity observed, around 50 meV. Thus, our data is fully supportive of the proposed surface stress model.