Ions interacting with complex molecular systems: The effect of a surrounding environment

This paper highlight results from studies of keV-ion impact on complex molecules and molecular clusters, which have been carried out at the ARIBE facility in Caen (France) during the last decade. Studies of fullerenes, Polycyclic Aromatic Hydrocarbons (PAHs), and biomolecules are reviewed with focus on the effect of a surrounding environment when ions interact with weakly bound clusters of theses species. One common result is that charge and energy are rapidly shared between the individual molecules in the clusters, in contrast to e.g. weakly bound atomic clusters where the charge stay localized to a few atoms from which the electrons are removed during the collisions. Another important finding is that ion collisions may induce reactions within clusters such as e.g. proton transfer and different types of molecular growth processes. In the latter case, these processes may be driven by prompt non-statistical atom knockouts in billiard-ball like atom-atom collisions favouring highly reactive fragments. In contrast, statistical fragmentation in general yields different and less reactive fragments.


Introduction
During the last decades, there has been a strong development of different methods to produce pure macroscopic quantities of complex molecular systems and techniques to bring fragile systems such as e.g. biomolecules into the gas phase. Complex molecular systems have thus been extensively studied in the laboratory by means of various excitation and ionization tools such as e.g. photons, electrons, and energetic atoms/ions. When isolated molecules are excited in these interactions, the excess energy is typically redistributed across all internal degrees of freedom. This leads to statistical decay through electron emission, photon emission, or fragmentation processes on timescales exceeding the typical vibrational timescales of picoseconds. In such cases, the nature of excitation is unimportant -it is only the amount of excess energy which determines the fate of the molecules.
Studies of weakly bound clusters have also been subjects of strong research efforts. Such studies provide a bridge across the disciplines of physics and chemistry and aim to answer fundamental questions like: Do charge and energy stay localized to a few constituents in clusters which have been exposed to ionizing radiation? If not, how fast is the charge and energy communication? Is the nature of excitation important? Are the individual molecules protected from damage in the cluster environment? Are new fragmentation pathways observed in the cluster environment, which are not present for the isolated molecules? Is it possible to induce intracluster reactions and in that case what drives such processes?
In this paper, the first ever studies of collisions between keV-ions and weakly bound clusters of fullerenes, Polycyclic Aromatic Hydrocarbons (PAHs), and biomolecules are reviewed. These pioneering experiments were carried out at the ARIBE facility in Caen (France) and have provided new insights into the above-mentioned fundamental questions. A few examples of the molecular structures used in these studies are shown in figure 1. Fullerenes are closedcage all carbon structures where C 60 is the most prominent family member. Due to its highly symmetric structure where all atoms experience identical chemical environments, C 60 molecules serve as excellent model systems for complex molecular systems. Weakly bound clusters of fullerenes may be viewed as small pieces of fullerite -the solid state manifestation of C 60 . The bonds between neighbouring C 60 are of van der Waals type with C 60 -C 60 binding energies of only about 0.3 eV, which means that they may rotate freely and thus to a large extent keep their individual properties in fullerite and in clusters. Individual fullerenes (C 60 and C 70 ) have been identified in space due to their characteristic IR-emission features [1], and possible signatures of fullerite (or clusters) have also been reported recently [2]. There they may coexist with Polycyclic Aromatic Hydrocarbons (PAHs) [3], which typically consist of fused benzene rings (cf. figure  1). PAHs are believed to be ubiquitous in space [4] and are thought to be the building blocks of interstellar dust grains. These findings raise questions about how fullerenes and PAHs are formed, react, respond to energy transfers, and the role of the surrounding (cluster) environment in such processes. One part of the motivation for the studies presented in this review was to shed light on these intriguing issues, which then may contribute to a better understanding of aspects of the origin and evolution of complex molecules in space [5].
Biomolecular precursors have also been identified in space and, consequently, the possibilities that the building blocks of life came to Earth from e.g. meteorite impacts have been much debated. This illustrates that it is important to understand how stable they are, how they are destroyed and how their fragments may later react to form new molecules. In the present case, keV-ions are used as tools to investigate the inherent properties of biomolecules such as e.g. DNA/RNA (nucleobases) and peptide/protein (amino acids) building blocks (cf. figure  1). Such studies also aim to further our knowledge on radiation damage processes at the single molecular level and to systematically investigate the effects of embedding a molecule inside a cluster. The latter is crucial to mimic a more realistic natural environment as for instance in an aqueous solution.
The paper is organized as follows. In section 2, the experimental multi-coincidence mass spectrometry techniques are briefly described. Sections 3 and 4 are devoted to discussions of results for collisions with monomer and cluster targets, respectively. The results are summarized in section 5 and an outlook with possible future experiments and challenges is also given there.

Experimental techniques
The experiments were performed at the ARIBE facility at GANIL in Caen, France. The experimental procedure has been discussed in detail elsewhere [6] and only a brief description is given here. Ion beams from an Electron Cyclotron Resonance (ECR) ion source are accelerated to keV energies. The ion beams are then pulsed at a repetition rate of a few kHz into microsecond long beam pulses, and interact with a broad distribution [7] of neutral weakly bound molecular clusters from a liquid nitrogen cooled cluster aggregation source. The positively charged collision products are analyzed with the aid of a linear time-of-flight mass spectrometer (cf. figure 2) [8] promptly after the beam pulses have left the interaction region. The collision products hit a gold coated steel plate at the end of the spectrometer, where secondary electrons emitted from the plate are guided to a microchannel (MCP) detector by a weak magnetic field. This arrangement give high detection efficiency, which is crucial for coincidence measurements of charged fragment from the same collision events.

Collisions with monomer targets
In the present collision energy regime, the projectiles are slower than the typical target valence electrons and electron capture is the dominant target ionization process. By varying the projectile charge, velocity, and mass, it is then possible to systematically investigate the effects of charging and heating molecular systems. In the present Section we summarize results for collisions with monomer targets -i.e. for targets in which the individual molecules are isolated from each other in vacuum.   through electron capture, where small amounts of energy is deposited directly in the collisions. Still, the molecules may be internally heated as the so formed multiply charged systems are born in their neutral geometry on the subfemtosecond timescale of the collision (vertical ionization), but later relax to minima on the potential energy surfaces from which they may statistically decay. For rigid molecular structures such as e.g. PAHs and fullerenes this heating processes is negligible and highly charged metastable systems may therefore survive on the experimental microsecond timescales. This is illustrated in figure 3, which shows the mass spectrum from collisions between 300 keV Xe 20+ projectiles and coronene molecules (C 24 H 12 ) [9]. The dominant peaks correspond to intact C 24 H q+ 12 in charge states up to q=4, which are due to sequential electron capture processes at large impact parameters. A small fraction of the multiply charged systems (q=2-4) decay through statistical fragmentation processes, mainly through C 2 H 2 and H-loss -the lowest dissociation energy channels at about 5 eV [10]. Thus, the relative intact peak intensities (including H-loss channels) reflect the sequence of ionization energies, i.e. the relative ionization cross sections. These are in good agreement with the results from a novel classical over-the-barrier model where PAHs are modeled as infinitely thin metal discs [11]. In this model, the energy barriers experienced by the active electrons are calculated along straight line ion trajectories. When these barriers become lower than the Stark shifted ionization energies the so called over-the-barrier criterion is fulfilled, and electrons are captured by the projectiles. The favorable comparison with experimental results show that the model catches the essentials of the collision dependent target-polarization effects, and that PAHs may be successfully described as small metallic objects when perturbed by an external electrostatic field [11]. Interestingly, fullerene molecules also display metallic behaviors in such interactions, as was recently demonstrated through comparisons between the electrostatic interaction energy for a metal sphere in the presence of a point charge and the corresponding results from molecular structure calculations [12,13].
collisions with energetic multiply charged ions. 30 In the present work, we improved it by including, in addition to the standard exploration of the PES, ab initio molecular dynamics simulations, which are crucial to obtain a complete picture of the problem. The mass spectrum of the cationic products measured after collisions of highly charged Xe 25+ ions with glycine molecules at the energy of 387.5 keV is shown in Figure  1 (experimental details are given in the Supporting Information, SI). The intact molecule survives the interaction with a certain probability, and a corresponding peak is visible at m/q = 75 amu attributed to the NH 2 CH 2 COOH + radical cation. However, the spectrum is dominated by molecular dissociation with the main peaks at m/q = 30, 28, and 45 amu. These peaks correspond to the fragments produced by cleavage of the C carboxyl −C α bond leading to the formation of NH 2 CH 2 + , HNCH + , and COOH + , respectively. In general, these ions are the characteristic features obtained in the fragmentation of ionized α-amino acids. 31−34 Note that the survival yield for the intact molecule is higher after interaction with multiply charged ions compared to electron-impact ionization or photoionization due to a lower energy transfer. 34 Interestingly, we clearly identify the formation of doubly charged molecular cations at m/q = 28.5, 27.5, and 14.5 amu (see zoom-ins in Figure 1). emitted cationic fragments from a single collision (see details in the SI and in ref 30). Figure 2 shows a so-called "correlation or coincidence map" obtained in collisions of Xe 25+ projectiles with the glycine molecule. It displays the time-of-flight of the heavier cationic fragment (TOF2) as a function of the time-offlight of the lighter cationic fragment (TOF1), thus reflecting the fragmentation of the glycine dication into two charged fragments. The insets in Figure 2 give more details in the regions of interest. The most intense islands observed in the coincidence measurements correspond to the cleavage of the C carboxyl −C α bond (island 30 + /45 + in Figure 2b) and subsequent neutral moiety emissions from the two singly charged fragments, H or H 2 from NH 2 CH 2 + and O or OH from COOH + (islands 30 + /29 + , 30 + /28 + , and 29 + /28 + in Figure 2a). Table 1 in the SI gives the identification of the fragments produced with their relative intensities. As can be seen, in the coincidence map and in the corresponding "two stops" mass spectrum (see the SI), the doubly charged species (at 14.5, 27.5, and 28.5 amu) are absent. Thus, the analysis of the coincident mass spectra shows that these doubly charged fragments are not measured in coincidence with any other charged fragment, that is, they are formed by neutral moiety emission from the doubly ionized glycine molecule. In addition, Mass spectrum recorded for collisions between 387.5 keV Xe 25+ projectiles and glycine (NH 2 CH 2 COOH). (Reproduced with permission from [14] c 2013 ACS). The molecular structures of the singly charged parent ion and a few typical fragment ions are shown in the inset.
Smaller impact parameter collisions yield more highly charged coronene molecules (q>4), which promptly Coulomb explode into the small hydrocarbons with m/q<70 as can be seen in figure 3. This is commonly referred to as charge driven fragmentation. For smaller and less rigid systems such as e.g. biomolecules, internal heating through relaxation processes (cf. above) and Coulomb explosions become important already for moderately charged systems. This gives rise to a much richer fragmentation mass spectrum as illustrated by the example in figure 4 for collisions between 387.5 keV Xe 25+ projectiles and the amino acid glycine (m/q=75) [14]. Two of the most intense peaks in this spectrum have m/q=30 and m/q=45, which correspond to NH 2 CH + 2 and COOH + , respectively. From the coincidence analysis it becomes evident that these fragments stem from the doubly charged parent molecule, i.e. from Coulomb exploding doubly charged glycine (NH 2 CH 2 COOH 2+ ). This is consistent with results from molecular structure calculations and ab-initio Molecular Dynamics simulations, which give information about the fragmentation dynamics leading to the formation of the doubly charged fragments (cf. the peaks at mass-to-charge ratios 14.5, 27.5 and 28.5 amu in the insets in figure 4). Interestingly, these calculations show that Coulomb explosions compete with neutral fragment emission following ultrafast intramolecular hydrogen migrations (∼ 30 fs). Such ultrafast hydrogen transfers are expected to be important for other biomolecular systems [14]. This example and the recent study of γ-Aminobutyric Acid (NH 2 -(CH 2 ) 3 COOH) at ARIBE [15] clearly illustrate the synergy effects and power of combining coincidence time-of-flight mass spectrometry with ab-initio calculations and simulations. For low charge state projectiles electrons are captured at small impact parameters, typically accompanied with substantial amounts of energy transfer to the molecule. Here the energy transfers may be localized to regions close to the projectile ion trajectories and are due to interactions with the electron clouds (electronic stopping) and with the individual atomic nuclei in the molecule (nuclear stopping). For light projectiles such as e.g. He + , the excess energy is mainly due to electronic stopping processes, as the results from model calculations illustrate in figure 5 [16]. The left panel shows the pure electronic stopping for face on 11.25 keV He collisions with anthracene. This was calculated using the friction coefficients from Puska and Nieminen [17] for He traversing a free electron gas, where the molecular valance electron densities were taken from Density Functional Theory calculations. The corresponding nuclear stopping energies for binary He+C/H interactions using a screened Bohr potential is shown in the middle panel of figure 5, while the total (electronic+nuclear) stopping is shown in the right panel. The total stopping is clearly dominated by electronic stopping with energies on the order of 40 eV, which is well above the activation energies of about 5 eV for fragmentation [10].

Low charge state projectiles
After this subfemtosecond excitation process, the stopping energy is typically converted into internal heating followed by statistical decay processes [19]. This is for instance seen in the mass spectrum recorded in He + + C 14 H 10 collisions [18], which is shown in the upper panel of figure  6. The most prominent peaks corresponds to intact singly and doubly charged anthracene. On the left hand side of these there are peaks due to fragmentation of moderately heated systems, yielding fragments corresponding to losses of H atoms and loss of C 2 H x (x=2-4). High energy deposits lead to multifragmentation processes as manifested by a broad distribution of C n H + x fragments with n=1-11. In this case the total fragmentation yield is (63±1)% , i.e. significantly larger than the (39±1)% recorded using 360 keV Xe 20+ projectiles (cf. lower panel of figure  6). This clearly demonstrates that the target molecules on the average are much hotter when exposed to projectiles in low charge states compared to when they are exposed to high charge with which mass-to-charge spectra and kinetic energy release distributions were recorded [16]. The ion beams and the extraction voltages were pulsed with pulse lengths of $1 s and with the extraction switched on $0:1 s after passage of the ion pulse. Mass spectra of anthracene monomers which are ionized and fragmented in collisions with He þ or Xe 20þ are shown in Fig. 1 (left panels). For He þ , we observe singly and doubly charged intact C 14 H 10 molecules (Ant þ and Ant 2þ in the figure), losses of H atoms, loss of C 2 H x from C 14 H þ 10 and C 14 H 2þ 10 , and smaller fragment ions, C n H þ x , with n ¼ 1-11. This type of bimodal distribution is similar to the ones observed for fullerene fragmentation where evaporation and multifragmentation processes compete [17,18]. For He þ þ C 14 H 10 collisions the overall probability for fragmentation is ð63 AE 1Þ% and the C n H þ x peaks with n ¼ 2-9 are intense due to strong internal heating in small impact parameter collisions [6,19]. For Xe 20þ , the overall fragmentation probability is ð39 AE 1Þ% and the relative intensity distribution changes in favor of C n H þ x fragments with n ¼ 1-5. Further, the H-loss and C 2 H x -loss channels become very weak and the peaks for Ant 2þ=3þ and C n<14 H 2þ x become stronger in relation to the C n<14 H þ 10 peaks. The differences between He þ and Xe 20þ collisions on monomers are readily explained. For He þ the fragment spectrum is typical for thermally driven processes, while The mass-to-charge spectra for He þ or X cene clusters are shown in the middle and Fig. 1. The broad peaks in the rightmost pan doubly charged clusters stemming from fr still larger clusters. As expected the overall forming fragments below the C 14 H 10 mass in He þ decreases strongly with the cluster ta internal excitation energy is distributed wi before fragmentation. Thus, C 14 H þ 10 ions e ments from a cluster have less excitatio directly ionized molecules from the mono relative C n H þ x intensity distributions, integr as functions of n (upper inset Fig. 1) are sim monomers (open symbols) and clusters (cl Thus, although the C 14 H þ 10 fragmentation is the cluster case, this similarity strongly sug ments below 178 amu stem from therm C 14 H þ 10 also in this case. Doubly charged ions are clearly observed with the monome absent with the ½C 14 H 10 k cluster target ind charge is distributed on the clusters before The scenario that emerges for the He þ þ ½ sions is thus that the He þ projectile ionizes cluster in close, peripheral collisions. Charg energy distribute within the cluster before which often yields singly charged monomer state projectiles. In the latter case, the molecules are typically multiply charged (cf. above) leading to several small singly charged fragments in Coulomb explosions, which explains why the fragment mass distribution is shifted towards smaller fragments in the lower panel of figure  6 and in figure 3.

Collisions with cluster targets
As mentioned in the Introduction, it is possible to systematically investigate the role of a surrounding environment by embedding the molecules inside a cluster. Here, key findings from such studies are briefly summarized. The first part is devoted to charge and energy flow processes when clusters are exposed to keV-ions, while the second part deals with intracluster reactions induced in such interactions.

Charge and energy flow processes
In 2003, the first study of keV-ions interacting with weakly bound clusters of complex molecules was carried out at the ARIBE facility in Caen [20]. The total mass spectrum from this study is shown in figure7, which is due to 400 keV Xe 20+ projectiles colliding with a broad distribution of neutral clusters of fullerenes (including monomers). In this spectrum, intact singly and multiply charged fullerenes are the most dominant products. A large fraction of these are due to collisions with neutral monomers in the target, as they appear in the single-stop spectrum (the spectrum for events where only one charged ion is detected) and has a charge-state intensity distribution (C + 60 , C 2+ 60 , C 3+ 60 etc.) similar to those for a pure monomer target (cf. Ref. [20] for details). However, a substantial fraction of the singly charged intact monomers were shown to be detected in coincidence with each other or with smaller singly charged fragments. This means that they stem from fragmentation of multiply charged clusters, where the charges have been rapidly redistributed before the decay. Thus, in this study it was demonstrated that weakly bound clusters of fullerenes become highly conducting when charged, while being insulators as neutrals (i.e. like small pieces of bulk fullerite). This behavior is in strong contrast to weakly bound atomic Ar n clusters where the charge stays localized to a few atoms when ionized by highly emitted from the conversion plate are focused on a channel plate. The signals are registered with a multihit timeto-digital converter on an event by event basis. Typical aggregation and drift times for neutral clusters between formation and interaction are of the order of several ms, whereas typical lifetimes between 10 and 40 s are necessary for the ionized clusters in order to be registered as stable clusters. Figure 1 shows a typical mass spectrum obtained in collisions of Xe 20 projectiles with neutral clusters of fullerenes. On the right-hand side contributions from singly and multiply charged clusters are present; on the left-hand side multiply charged fullerene monomers dominate the spectrum.
The spectrum looks similar to that reported for Ar clusters [14], with the exception that in the latter case no multiply charged clusters have been detected. As can be seen from the abundance of doubly charged clusters, the size distribution ends at about n 16. However, at higher oven temperatures and hence higher vapor pressures in the cluster source, fullerene clusters containing up to 40 molecules have been detected. Figure 2 gives a detailed view of the mass spectrum. The complexity is due to the presence of mixed clusters of C 60 and C 70 , as a mixture of these fullerenes has been used, and to the formation of multiply charged clusters. The analysis of the measured spectra leads to the following conclusions.
(i) Clusters in charge states up to q 4 are identified as isolated peaks whereas contributions from 5 times like Ar q n and Na q n . Compared to othe appearance sizes of fullerene clusters are r the size of the constituents is larger, as wel izabilities and hence the binding energies. T petition between cohesive forces and Coul is more favorable for fullerene than for A ters. The observed difference in the appea q 2 for ionization by highly charged ion radiation is due to the lower energy transfe with highly charged ions as double ion already at very large impact parameters [1 (ii) The analysis of mixed clusters sho is no remarkable difference between C 60 a cules when forming clusters. The appeara more or less identical for C 60 q n , C 60 n C 60 nÿ2 C 70 q 2 for all observed charge thermore, the relative population of these m corresponds approximately to a statistical (iii) The mass spectra for individual char peaks with enhanced intensity reflectin shell effects. For singly charged ions these ations are not observed proving a low inter the absence of evaporation processes. For increase occurs for n 13, whereas a cle intensity is observed for n 13, 19, and and n 25 and 29 (for q 4). This indica ing heating of the clusters with the degree    [16], and (d) from [15]. charged keV-projectiles [21]. To the right hand side of the singly charged monomer peak, there are distributions of intact singly and multiply charged clusters surviving on the experimental microsecond timescales, where the smallest cluster sizes for a given charge state (appearance sizes) were shown to be as small as n=5, 10, 21, and 33 for q=2, 3, 4, and 5, respectively. the projectile to escape from the target in the experiments, which is of the order of 1 fs. 8,10,11 Thus the rapid electron transfer invoked in Refs. 10 and 11 to explain the even-odd effects observed in the experiments is confirmed by the present analysis. Waal's system. 7 The sequence of calculated vertic tion energies follows a similar ͑but less steep͒ linea the corresponding monomer values, suggesting tha ization cross section should have a smooth decreasi ior in both cases. However, in strong contrast with mer case, the recently measured ionization yield even-odd effects as functions of charge. 10 We sh this apparent contradiction is due to the experimen tions ͑collision geometry͒ and can only be expla consequence of the rapid ͑subfemtosecond͒ charge within the charged dimer systems. The multiply ch tems were found to be unstable and decayed charged intact monomers in agreement with the exp findings. 11 The calculated kinetic energy releases w to be lower than the corresponding experimental on ing at significant internal excitations in the frag processes.  Figure 8. The kinetic energy releases in the fragmentation of [C 60 ] q+ 2 → C q1+ 60 + C q2+ 60 as functions of charge, q1=q2 (even q), q1=q2+1 (odd q). (Reproduced with permission from [22] c 2009 AIP). The DFT results (solid squares) [22], the experimental results (solid circles) [23,24], and the results from an electrostatic model (open squares) [23,25].

ACKNOWLEDGMENTS
As the charge is rapidly redistributed, the detection of two charged intact fullerene monomers in coincidence can be taken as a fingerprint for Coulomb explosion of a multiply charged dimer system. In Refs. [23] and [24] this was used to extract the relative ionization yields (by counting the number of such events) and the kinetic energy releases in the fragmentation processes (from the peak widths) as functions of the dimer charge state. Interestingly, the relative ionization yields display strong even-odd effects, which can only be explained if the charge communication takes place during the collisions, i.e. on the sub-femtosecond timescales (cf. Ref. [23] for details). This was later supported by combined Density Functional Theory (DFT) molecular structure calculations and modeling work [22]. However, the kinetic energy releases predicted from theory were shown to be significantly smaller than the measured ones (cf. figure 8). The reason for this discrepancy was attributed to internal heating by Coulomb explosions, i.e. that a substantial fraction (about 50%) of the potential energy is converted into internal energy of the separating monomers. The theoretical values are for internally cold systems and thus serve as upper limits for the kinetic energy release values.
ass-to-charge spectra and kinetic energy tions were recorded [16]. The ion beams on voltages were pulsed with pulse lengths with the extraction switched on $0:1 s f the ion pulse. of anthracene monomers which are ionized in collisions with He þ or Xe 20þ are shown panels). For He þ , we observe singly and intact C 14 H 10 molecules (Ant þ and Ant 2þ sses of H atoms, loss of C 2 H x from C 14 H þ 10 d smaller fragment ions, C n H þ x , with n ¼ e of bimodal distribution is similar to the for fullerene fragmentation where evaporafragmentation processes compete [17,18]. H 10 collisions the overall probability for s ð63 AE 1Þ% and the C n H þ x peaks with n ¼ e due to strong internal heating in small ter collisions [6,19]. For Xe 20þ , the overall probability is ð39 AE 1Þ% and the relative ution changes in favor of C n H þ x fragments Further, the H-loss and C 2 H x -loss channels weak and the peaks for Ant 2þ=3þ and þ The mass-to-charge spectra for He þ or Xe 20þ on anthracene clusters are shown in the middle and right panels of Fig. 1. The broad peaks in the rightmost panels are singly or doubly charged clusters stemming from fragmentation of still larger clusters. As expected the overall probability for forming fragments below the C 14 H 10 mass in collisions with He þ decreases strongly with the cluster target where the internal excitation energy is distributed within the cluster before fragmentation. Thus, C 14 H þ 10 ions emitted as fragments from a cluster have less excitation energy than directly ionized molecules from the monomer target. The relative C n H þ x intensity distributions, integrated over x and as functions of n (upper inset Fig. 1) are similar for He þ on monomers (open symbols) and clusters (closed symbols). Thus, although the C 14 H þ 10 fragmentation is much weaker in the cluster case, this similarity strongly suggests that fragments below 178 amu stem from thermally activated C 14 H þ 10 also in this case. Doubly charged intact C 14 H 10 ions are clearly observed with the monomer target but are absent with the ½C 14 H 10 k cluster target indicating that also charge is distributed on the clusters before fragmentation. The scenario that emerges for the He þ þ ½C 14 H 10 k colliþ line). Left panels: Mass-to-charge spectra for He þ =Xe 20þ þ C 14 H 10 (monomer) collisions (molecular structure in le panels: Mass-to-charge spectra below 190 amu=e for He þ =Xe 20þ þ ½C 14 H 10 k (cluster) collisions. Insets show tensity distributions as functions of n for the cluster (filled symbols) and monomer target (open symbols) scaled by the e cluster and monomer fragmentation efficiencies. Right panels: Size-to-charge spectra for fragment clusters, duced in He þ =Xe 20þ þ ½C 14 H 10 k collisions. Insets: zoom-ins for j > 4 with logarithmic intensity scales.

(2010) P H Y S I C A L R E V I E W L E T T E R S
week ending 19 NOVEMBER 2010 Interestingly, rapid charge and energy redistribution appear to be a general feature in collisions with weakly bound molecular systems, as it has also been demonstrated for biomolecular clusters [26,27,15,28] as well as for PAH clusters [18,29]. The latter is illustrated in figure9, which shows the mass spectra recorded in collisions between He + /Xe 20+ projectiles and anthracene (C 14 H 10 ) clusters [18]. The monomer-and cluster-fragment regions of the mass spectra are shown in the left and right panels, respectively. The cluster distributions are rather similar for both projectiles, with a rapidly decreasing trend as a function of size, but the collision scenarios are different. In the He + case (upper panel), energy is mainly deposited in electronic stopping processes along the ion trajectories (cf. Sec. 3.2). This energy is redistributed and the clusters cool down by long sequences of neutral emissions of intact anthracene molecules, in most cases all the way down to the singly charged anthracene. Indeed, in this case monomer fragments stemming from clusters are significantly colder than when the anthracene monomer target is ionized by the same projectile at the same velocity (cf. figure 6). The fragmentation yield is about a factor of ten smaller, only (6±1)% in the cluster case. Thus, the monomers are  . 3.1). Surprisingly, the monomers are not protected from damage in collisions with Xe 20+ as the fragmentation yield is similar to that for the pure monomer target (cf. figure 6). The reason for this is most likely heating by Coulomb explosions of multiply charged clusters, which was shown to be important for multiply charged fullerene dimers (cf. Fig 8).
vestigation of nucleobase fragmentation within a cluster. In this context, the spectrum in Figure 2 already reveals one issue of particular importance: at masses higher than the thymine monomer, only peaks which are almost exclusively due to thy n q + can be observed. This indicates that the ion impact does not lead to polymerization but only to dissociation of one or more single molecules within the cluster. Figure 3 shows a zoom into the low-mass region of the spectrum in Figure 2. The lower masses are not displayed since this range is overwhelmed by contributions coming from the condensation gas (He).
The fragmentation in the cluster case is obviously different from what is observed in the gas phase (see Figure 3). Not only are the relative peak intensities very different, but also in particular the structures around m/q % 109 and m/q % 97, basically due to a loss of OH and NH 2 CH, respectively, are not observed at all for ion-induced fragmentation of isolated thymine molecules. For comparison, m/q % 109 was not observed either in photoionization nor in electron-impact-ionization studies of isolated thymine molecules; for m/q % 97, only traces were found [25] . However, the peaks around m/q % 109 appeared in the mass spectra from ion collisions with condensed-phase thymine. [21] Apparently the finite environment of a cluster is sufficient to allow "characteristic" fragmentation channels, which are normally only present in the condensed phase, with the OH-loss peak at m/q % 109 being one of the fingerprints. OH loss is the only channel observed that can solely be due to exocyclic bond cleavage.
This finding entails the question whether the appearance of the peaks around m/q = 109 and m/q = 97 (loss of an OH or an NH 2 CH/OCH group, respectively) depends on the parent cluster size. This issue will be discussed in greater detail in the section on coincidence experiments. tion studies, [26] we find as prominent pe (m/q = 112), C 3 H 3 NO + (m/q = 69), C 2 H 2 O HCNH + (m/q = 28). Figure 5 shows a wide-range spectrum o ter-ion distribution (ura n q + ) obtained in c O 5 + with neutral uracil clusters. The charac trum are similar to those of the thymine o a somewhat lower resolution, only clust states one and two could be identified Figure 5, the series for q = 1 and q = 2 are thymine case, for q = 1, prominent peaks c magic numbers n = 7, 10 are observed.
As for thymine, we do not observe ev zation: For n ! 1, only peaks which are alm to ura n q + can be observed. In the region lecular-fragmentation channels specific t cluster are evident. The respective zoom be found in Figure 6.
Again, fragmentation differs strongly fo ed molecules. The large fragments at ar The assignment of the peaks essentially follows the work of Deng et al. [21] (thy indicates the parent thymine molecule). The peaks around m/q = 97 were not observed by Deng et al. and the assignment from Jochims et al. [25] is used. Similar o the uracil case, this peak could also be assigned [thy-HCO] + . For comparison, the data for isolated molecules is added in grey.  [25] (ura indicates the parent uracil m are also present for isolated thymine, whereas the former two (open symbols) are cluster-specific. From Figure 12 it is obvious that the yield for cluster ions observed in coincidence with cluster-specific fragments systematically exceeds the yields for cluster ions observed in coincidence with the nonspecific fragments. Since the data is normalized, this implies that clusterspecific fragments, particularly [thy-NH 2 CH] + , become more important at larger n values. From the coincidence data, two conclusions can be drawn: 1) "Characteristic" fragmentation channels, forbidden for the isolated molecule, already open up for the smallest possible clusters (such as dimers and trimers); 2) the relative importance of these channels slightly increases with cluster size. Fragmentation channels typical for the condensed phase can thus already be observed in very small clusters; however, their relative strength becomes higher in larger systems.
This implies that, for example, the mechanism underlying [thy-OH] + formation is already active in very small clusters. We can thus try to link this mechanism to the calculated structures of thy 2 dimers. Kelly and Kantorovich [30] have shown that the most stable dimers are bound by two hydrogen bonds between an H donor and an O acceptor. The most stable dimer is shown in Figure 13.
The chemical interaction between the two thymine entities is, to a large extent, limited to the two hydrogen-bond regions, that is, in these regions the electronic structure of the involved molecules is strongly altered. For any O or H atom subject to hydrogen bonding, the presence of an additional intermolecular OH bond will always weaken the intramolecular one. It is thus straightforward that during molecular fragmentation of one thymine within the dimer, particularly these O and H atoms will be affected by the presence of the second thymine. This suggests that the characteristic peak observed in the cluster data (m/q = 109) is probably due to the loss of a single O and a single H atom involved in the hydrogen bonds. This is also supported by the results of Figure 12: The peak due to O + H loss becomes relatively more important for larger clusters. With increasing cluster size, the fraction of thymine molecules that contain O atoms that are involved in two hydrogen bonds increases. Those O atoms are subject to an even stronger weakening of the intramolecular OH bond.
The hydrogen-bond line of argumentation can also be used from another angle: Observation of OH or O + H loss indicates hydrogen bonding between cluster constituents (planar clusters) rather than dispersive-force-driven bonding (stacked clusters). To confirm this, we performed quantum-chemical calculations with the Gaussian 03 package. [33] Thymine clusters with up to five constituents were geometry optimized at the B3PW91/6-31G(d) level for a variety of different starting geometries. The stabilization energy, that is, the total energy of the optimized combined clusters minus the total energy of their optimized components, [30] has then been computed at the B3PW91/6-311 + GA C H T U N G T R E N N U N G (2df,2p) level. Figure 14 shows the geometries of the most-strongly-bound neutral thymine trimers, tetramers, and pentamers we found. Similar to the observations for dimers, the geometry optimization for these small clusters almost always leads to planar geometries. Stacked geometries do not play a role. However, with increasing cluster size, the geometries deviate more from the perfect planarity of the dimer. In any case, the calculations show that not only dimers but also the larger systems are based on hydrogen bonds between O and H atoms.
Because of the dominant role played by hydrogen bonding also in condensed-phase thymine, we can conclude-in a straightforward way-that the characteristic fragmentation  Figure 13. Geometry of the most stable thy2 dimer. [30] Hydrogen bonds are indicated by dotted lines.   Another intriguing result is that the cluster environment may open up decay pathways which are closed in interactions with the isolated molecules. This was demonstrated in collisions between 60 keV C 5+ and nucleobase monomers and clusters [26]. Figure 10 shows that the mass spectra for collisions with thymine targets display similar features, but there are important differences. The most striking effect is the appearance of new channels in the cluster case corresponding to O-and OH-loss. Using coincidence data, it was shown that this channel appears already in collisions with the smallest clusters (dimers) [26]. The most likely explanation for this new channel is that the intermolecular O-H bonds weakens the intramolecular bonds in the clusters, which favor the loss of a single O and a single H atom involved in the hydrogen bonds (cf. figure 10). This illustrates another type of effect due to the cluster environment which so far has not been observed with PAH and fullerene targets.

Intracluster reactions
It is well established that heated fullerenes emit C 2 -molecules rather than single carbon atoms in statistical fragmentation processes [31]. This simply reflects the large differences in dissociation energies, about 10 eV [32] and 15 eV [33], respectively. However, despite intense research the actual fullerene formation mechanism is still debated, as it is not fully understood why C 60 are formed so much more abundantly than other fullerenes in nature (e.g. in a fire flame). Possible explanations involve e.g. bottom-up mechanisms in which individual C-atoms or C 2 -molecules are ingested by smaller fullerenes [34], top-down mechanisms from decay of hot giant fullerenes [35], or emission of electromagnetic radiation [36]. These rely on the lower reactivity [34], more stable structure [35], and more effective radiative cooling of C 60 [36] with respect to neighboring fullerene sizes.
It has been demonstrated that larger fullerenes may be formed from smaller ones when fullerene films and fullerene clusters are exposed to various kind of photon fluxes [37,38,39]. This typically leads to broad distributions of fullerene sizes with an even number of carbon atoms, luster nsity and shift at the r low sities s below . ltiply , the onoak is ere is r low f that much sities ratio from wder highly excited C 60 + , emitted from multiply ionized ͓C 60 ͔ m clusters, which later ͑as they are hot͒ emit one or several C 2 units. 20 A key aspect here is that ͑charged͒ clusters of fullerenes, in contrast to, e.g., weakly bound argon clusters, 21 are excellent electrical conductors 20 within which the charge is distributed on subfemtosecond time scales. 22,23 Thus, a Xe 20+ ion may, e.g., pass closely outside a ͓C 60 ͔ m surface and initially removes many electrons from individual C 60 molecules and then also excites them strongly. These charges spread out before the C 60 + are emitted from the cluster ͑picosecond timescales͒ and these fullerenes may also fragment but on even longer time scales ͑nanoto microseconds typi-cally͒. For high a broad intensity distribution with peaks separated by two carbon masses appears with very small odd ͑n / q͒-peaks in between. As this distribution is absent for low there is a clear correlation between large clusters ͑m Ͼ 15͒ in the target and C n + production, showing that these molecules mainly form inside the ͓C 60 ͔ mՆ15 -clusters and not on their surfaces. In the latter case we would observe this distribution-or at least traces of it-also for low .
In Fig. 3, we show integrated C nՆ70 + intensities ͑from Gaussian fits͒ after subtracting the background tail from the C 60 + peak ͑cf. Fig. 2͒ and the C 70 + contribution due to the 4% C 70 -content in the powder. The remaining C 70 + intensity is responsible for 56Ϯ 4% of all C nՆ70 + fullerenes formed. In femtosecond laser exciton of ͓C 60 ͔ m ͑Ref. 14͒ and in laser desorptions of C 60 -films 12 there are broad maxima around n = 120, 180, 240, etc., indicating that the fused fullerenes in these experiments are due to coalescence reactions involving two, three, four, etc., fullerene monomers. In the present case, the C n + distribution is markedly different with a broad maximum ͑peaking well below n = 120 but also extending ͑ low , in the m + and FIG. 2. Intensities due to Xe 20+ + ͓C 60 ͔ m collisions for low ͑below͒ and high ͑above͒. Fullerene dimers, trimers, and pentamers are indicated with ͓C 60 ͔ 2 + , ͓C 60 ͔ 3 + , and ͓C 60 ͔ 5 2+ and fullerene monomers with C n + ͑cf. text͒.
reflecting the higher stabilities of those with respect to the odd-numbered ones. Similar mass distributions were observed in collisions between Xe 20+ projectiles and weakly bound clusters of fullerenes [30]. As shown in figure 11 these molecular fusion reactions are only seen with large cluster targets. The scenario here is that the individual molecules are destroyed along cluster-penetrating ion trajectories and that the surrounding environment (i.e. large clusters) allows the fragments to react with neighboring intact fullerenes or with other fragments before the clusters completely disintegrate. The measured kinetic energies of the fusion products in Ref. [30] suggest that both size-up and size-down processes are responsible for the mass broad distribution in figure 11. A large fraction of these fusion products corresponds to the most stable closed caged fullerenes, as concluded from the strong correlation between the measured relative peak intensities and the calculated relative stabilities as a function of fullerene size. More recently, a novel type of molecular fusion reaction has been observed in collisions between He 2+ projectiles and clusters of fullerenes [33]. This is illustrated in figure 12, where the left part shows the mass spectra for all collision events (upper panel) and events correlated with one or several intact C + 60 ions (lower panel). The total mass spectrum is rather similar to the one recorded in Xe 20+ + [C 60 ] n collisions (cf. figure 7), but there is significantly more intensity in the dimer region. A large fraction of this intensity appears in the coincidence spectrum and is due to three peaks at the positions for 120, 119, and 118 carbon masses per atomic unit of charge. Surprisingly, the most prominent peak corresponds to an odd number of carbon atoms. These are produced in prompt, nonstatistical, knockouts in Rutherford-like binary He+C scattering processes. The so formed C + 59 ions rapidly react with neighboring C 60 molecules in the clusters efficiently forming dumbbell shaped molecular fusion products while the clusters explode (cf. the right part of figure12). This scenario was supported by molecular dynamics simulations, which show that C + 59 are highly reactive and thus easily form thermodynamically stable dumbbell C + 119 systems [33]. Direct evidence for prompt knock-outs have recently been reported in PAH + + He collisions [40,41,42], at significantly lower center of mass collision energies (110 eV) where nuclear stopping is the dominant energy loss process. Thus, the molecules are only slightly heated internally (by about 4 eV) in electronic stopping processes and the fragments, which are as highly reactive as C + 59 , may therefore survive on the experimental microsecond timescales. This clearly illustrates that it is a general mechanism, and one would therefore expect to also see such non-statistical driven molecular fusion reactions for complex molecular cluster systems other than fullerenes. However, this has not yet been reported in the literature for collisions with pure PAH or mixed  Fig. 2. Small amounts of C 60À2m þ fragments are also observed and are mainly due to evaporation of m C 2 units [7] from the few hotter C 60 þ ions emitted from the clusters. The narrow ½C 60 þ n and ½C 60 2þ n peaks at large mass-tocharge ratios (labeled as n þ and n 2þ in the upper right panel in Fig. 2) are mainly due to collisions in which the clusters have been ionized but not fragmented. These peaks are observed for events in which one and only one ion is produced and are very weak in coincidence with C 60 þ ions as can be seen in the lower right panel of Fig. 2. The latter spectrum is instead dominated by peaks at the positions for 120, 119, and 118 carbon masses per atomic unit of charge, n C =e, where the intensity relation is 0:44:1:0:48 in a Gaussian fit. Taking the differences in flight times into account we find that these three peaks are almost twice as broad as (a factor of 1:7 AE 0:1 broader than) the narrow ½C 60 þ n (½C 60 þ 6 ) peaks in the upper panel of Fig. 2. This shows that the n C =e ¼ 120, 119, 118 peaks are due to cluster fragmentation processes. The peak at n C =e ¼ 120 may be due to (i) ½C 60 þ 2 dimers remaining after emissions of C 60 þ ions and C 60 molecules from larger clusters similar to what has been observed for other loosely bound cluster systems [17,18], and/or (ii) C 60 þ þ C 60 ! C 120 þ covalent bond formation. The peaks at n C =e ¼ 119 and n C =e ¼ 118 are, in contrast, most likely due to low energy FIG. 2. The measured mass-to-charge distributions following 22.5 keV He 2þ þ ½C 60 n collisions. The top and bottom panels show the distributions including all events (total spectrum) and for events correlated with one or several intact C 60 þ ions (coincidences with C 60 þ ions), respectively. Note the different scales for the left and right panels. The inset in the lower right panel shows a zoom-in on the mass region from 100 to 125 carbon masses per atomic unit of charge, n C =e.

I C A L R E V I E W L E T T E R S
week ending 3 MAY 2013 185501-2 show the distributions including all events (total spectrum) and for events correlated with one or several intact C + 60 ions, respectively. Note the different scales for the left and right panels. The inset in the lower right panel shows a zoom-in on the mass region from 100 to 125 carbon masses per atomic unit of charge. Right: Snapshot from molecular dynamics simulations where a covalently bound dumbbell C + 119 system has been formed in a C + 59 +C 60 collision when a cluster of thirteen fullerenes explodes. fullerene/PAH cluster targets. According to pilot molecular structure calculations and molecular dynamics simulations, the kinetic energy required to fuse defect PAH structures with intact PAHs/fullerenes is too high, but it should be possible with molecular systems which are less rigid and more easily bond to the defects [16]. This calls for future studies of mixed clusters. nation in the pure cluster case (see figure 4a)), we assume that protonation of adenine (b) is due to the presence of water molecules. The protonated adenine monomer peak was also observed in the case of electron impact and multiphoton ionization [34,37,38]. The electron impact studies of hydrated nucloebases suggest that the protonated nucleobase monomer is only observed when the electron energy is above the ionization potential of water [36]. Therefore, to explain such a formation, we propose the following mechanism in the case of single electron capture for a water molecule within an hydrated adenine cluster: Other types of intracluster reactions have been observed for biomolecules embedded in water clusters [28]. The left panel of figure 13 shows the singly charged adenine monomer peaks following collisions between 37.2 keV O 3+ ions and adenine monomers, pure adenine clusters, and hydrated adenine clusters. The latter is significantly broader and centered around a higher mass to charge ratio compared to the other ones. This is due to formation of protonated adenine, as revealed from the fit shown in the right panel of figure 13. Here, the formation mechanism is most likely similar to that observed in 12 MeV/u Ni 25+ impact on pure water clusters [43], where it was demonstrated that water may capture protons from nearby ionized water molecules. The proton affinity is larger for adenine than for water, which readily explains the efficient formation of protonated adenine inside the water cluster.

Summary and outlook
To summarize, the first ever studies of ion impact on various kinds of clusters, such as e.g. pure clusters of fullerenes, PAHs, nucleobases, amino acids and hydrated nucleobases, have been performed at the ARIBE facility in Caen, France. One common conclusion from these studies is that charge and energy are rapidly redistributed among the clusters constituents. In addition, different types of intracluster reactions may be induced depending on the actual collision conditions. The cluster environment plays an important role here, proton transfer reactions have for instance been observed for nucleobases in water but not for pure clusters of nucleobases. Intracluster reactions also depend on the size of the clusters and the projectile type, as e.g. certain fullerene fusion reactions are only seen when large clusters are exposed to Xe 20+ projectiles, while He 2+ ions ignite different fusion processes.
In the near future it will be possible to study collisions between keV-ions and mass-selected cluster ions with the aid of the novel PIBALE (Plateforme d'Irradiation de Biomolécules et d'Agrégats Libres et Environnés) setup at ARIBE. This will be a tremendous advantage when interpreting the experimental results, as the collision system will be well-defined in contrast to the previous studies of broad distributions of neutral cluster sizes. Another important aspect is to determine the energy being deposited in the interactions, which is an unknown parameter and hampers the interpretations of the results. This is truly challenging from both an experimental and a theoretical point of view. To realize this, common efforts will be essential, for instance within the COST Action XLIC and the European Associated Laboratory (DYNAMO) consisting of research groups from ARIBE, Stockholm University (Sweden), and Universidad Autónoma de Madrid (Spain).
In a somewhat longer perspective, the unique expertise and experience gained from studies of complex molecular systems at ARIBE could serve as an excellent platform for even more advanced experiments, possibly including electrostatic storage devices -traps and/or rings. There ions could be monitored and manipulated on longer timescales. New such facilities, including cryogenically cooled ones, are now being built and commissioned in e.g. Japan [44], Germany [45], and Sweden [46,47]. The cooling option is a way of controlling the experimental conditions as the ions are in more well-defined quantum states than in room temperature environments. Having such a storage device would further broaden the range of unique studies at ARIBE, and drive the relevant technology and science forward in this new emerging community.