Helmholtz-Zentrum Dresden-Rossendorf (HZDR) Interatomic coulombic decay-the mechanism for rapid deexcitation of hollow atoms

Richard A. Wilhelm, Elisabeth Gruber, Janine Schwestka, Roland Kozubek, Teresa I. Madeira, José P. Marques, Jacek Kobus, Arkady V. Krasheninnikov, Marika Schleberger, and Friedrich Aumayr TU Wien, Institute of Applied Physics, Wiedner Hauptstrasse 8-10, 1040 Vienna, Austria, EU Helmholtz-Zentrum Dresden-Rossendorf, Institute of Ion Beam Physics and Materials Research, Bautzner Landstrasse 400, 01328 Dresden, Germany, EU University Duisburg-Essen, Faculty of Physics and CENIDE, Lotharstrasse 1, 47048 Duisburg, Germany, EU BioISI—Biosystems & Integrative Sciences Institute, Faculdade de Ciéncias da Universidade de Lisboa, 1749-016 Lisbon, Portugal, EU Nicolaus Copernicus University, Faculty of Physics, Astronomy and Informatics, Institute of Physics, Grudziądzka 5, 87-100 Toruń, Poland, EU (Received 5 May 2017; published 8 September 2017)

The interaction of ions with solid surfaces involves a variety of different physical processes, as, e.g., elastic scattering and the formation of a scattering cascade or inelastic scattering and associated electronic excitations. Both elastic and inelastic scattering of ions may lead to sputtering [1], nanomelting [2], interface mixing [3], and many more observable target modifications [4]. Depending on ion velocity, one of the scattering mechanisms dominates; i.e., for slow ions with v ≪ v 0 (v 0 ¼ Bohr velocity) elastic scattering is the dominant energy loss mechanism. However, in special cases the picture may change. Especially highly charged ions (HCIs) have a large inelastic scattering cross section [5], and they deposit their potential energy (total ionization potential) at a shallow layer at the surface. A significant amount of this energy-up to some tens of keVs is transferred to the target electronic system even when the HCI has a low velocity [6]. For more than 20 years there has existed a generally accepted model for neutralization and deexcitation of highly charged ions, which is explained in detail, e.g., by Arnau et al. [6]. Our findings presented in this Letter indicate that the current model of hollow atom formation and deexcitation has to be refined. Experimental indications that the modeled deexcitation cascade to the ground state is too slow (bottleneck problem) existed before [7][8][9][10][11], but were misinterpreted by ad hoc assumptions of enhanced autoionization rates or so-called sidefeeding processes. In our work we deliberately use a very asymmetric projectile-target combination, namely, Xe-C to exclude sidefeeding and corresponding molecular orbital formation. We calculate atomic autoionization rates for hollow Xe atoms formed during neutralization and find no enhancement. We use a 2D solid (single layer of graphene) to pinpoint the actual interaction time to a few femtoseconds only and thus exclude the possibility of hollow atom deexcitation in deeper layers of a 3D solid. We conclude that Interatomic Coulombic Decay (ICD), a process that was not considered so far, is responsible for the observed ultrafast neutralization and deexcitation [12]. Studies of ICD are of great importance to understand biological tissue damage, e.g., under energetic particle irradiation, because the origin of tissue damage lies mainly in slow electron production caused by ICD and their ability to efficiently cleave molecular bonds. ICD is sometimes also called direct Auger deexcitation. The idea was first discussed in Refs. [13,14] and also brought up by Cederbaum et al. in 1997 [12]. The first experimental observation was done by Marburger et al. and Jahnke et al. in noble gas clusters and dimers, respectively [15,16]. We show that ICD can explain the ultrafast deexcitation of HCIs and thus that ICD at a solid surface can be probed with HCIs.
In the beginning of the neutralization process an approaching HCI extracts electrons from the surface already at a distance of several angstroms by classicalover-barrier transport [6,17] [see Fig. 1(a)]. The ion is then almost neutralized, but highly excited, giving rise to the formation of a hollow atom with a total excitation energy of still a few tens of keVs [18,19]. After transmission through a thin solid film [7,9,20] or after scattering under grazing angles [8] the projectiles are measured in very low or neutral charge states [21]. The excitation energy must PRL 119, 103401 (2017) P H Y S I C A L R E V I E W L E T T E R S week ending 8 SEPTEMBER 2017 therefore be released by other processes than autoionization alone.
In the present work we address the deexcitation dynamics of a hollow atom in the vicinity of a surface. As presented recently by Gruber et al. [21] and supplemented by new data here (see Fig. 2) a highly charged Xe ion captures and stabilizes (does not lose electrons due to autoionization) up to 27 electrons in less than 4 fs. Measurements with highly charged Ar ions with L-shell holes also exhibit a similarly fast deexcitation [see Fig. 2(a)]. We present in the following a discussion on the suppression of the ion's autoionization by energy release to neighboring target atoms, i.e., ICD [see Fig. 1 Slow highly charged ions are produced in an electron beam ion trap from DREEBIT, Germany. Ions are extracted by means of electrostatic fields and a charge state is selected by an analyzing magnet. The ions are decelerated by an electrostatic retarding potential in front of a target chamber to energies between 10 and 150 keV. The pressure in the ion source, beam line, and target chamber is kept well below 5 × 10 −9 mbar to avoid charge exchange of the ions with residual gas. Graphene samples are grown by standard CVD on Cu foils and transferred without use of a polymer coating. The graphene layer is put on a TEM grid with a Quantifoil support. Ions are transmitted through graphene or are stopped within the Quantifoil. Thus, we only measure ions that have interacted with graphene or those that go through uncovered Quantifoil holes. Ion detection takes place about 20 cm (time of flight > 400 ns) behind the sample in an electrostatic analyzer. The analyzer allows charge state and energy measurements with a relative energy resolution of about 5 × 10 −3 .
Experimental results are compared to predictions from a model assuming a purely atomic deexcitation cascade of a hollow Xe atom and from a model including electronic level interaction during the ion collision. The details of the calculations can be found in the Supplemental Material [22]. Additionally, we calculated the electronic structure and subsequently the binding energies of ionic levels for strongly ionized Xe by means of the full-potential electronic structure calculation method [23] as well as the evolution of molecular orbitals for a Xe-C pair based on the density functional theory (DFT) code X2DHF [24].
Our ion-target system offers important features such as very asymmetric scattering partners, i.e., a heavy ion and a light target, a dense atomic environment of the target atoms, a high projectile charge, i.e., deep lying core holes, and finally a low ion velocity enabling ICD processes to take place.
Charge state distributions of Ar 16þ and Xe 30þ ions are depicted in Fig. 2(a) for ions transmitted through graphene. The number of stabilized electrons is determined by the difference between the incident charge state and the exit charge states. For the two ion incident charge states shown in Fig. 2(a) we also used two different kinetic energies and thus varied the interaction time with the graphene sheet between 1.4 and 3.6 fs (see the Supplemental Material [22]).
For all ions the mean number of electrons stabilized in the ion projectile is large, especially in light of the short interaction time. We calculate a neutralization time constant τ n for charge states investigated here, confirming an exponential dependence of the number of captured electrons on the interaction time [21]. The charge state dependence of τ n is shown in Fig. 2(b). The projectile is entirely neutralized after about 3-7 fs. The capture of a large number of electrons is certainly not surprising and was shown in many studies of the last two decades [8][9][10][11]. However, the data in Fig. 2(a) can only be explained if the charge capture as well as the deexcitation of the captured electrons happens within the interaction time due to the absence of autoionization.
In general four types of processes can lead to ion deexcitation within the short interaction time: (i) Auger neutralization [14], (ii) sidefeeding [8,25], (iii) resonant capture and enhanced atomic Auger decay [26,27], and (iv) ICD [28,29]  the first three processes are associated with rates of 10 11 -10 12 s −1 (time constant of 1-10 ps), whereas our experimental results can only be explained by ICD with a rate in the order of 10 15 s −1 or above (time constant of 1 fs). These processes are described below.
(i) Auger neutralization (also known as electron transfer mediated decay [30,31]) is a process by which an electron is transferred from the target material into a deeply bound projectile state [14]. The excess energy is released by emission of a target electron [see Fig. 3(b)]. This process is very similar to the exchange transition [28] by which a target electron is captured into a deeply bound projectile state and a projectile electron is emitted (see the Supplemental Material [22]). Core holes discussed here are states in a highly charged ion that are some hundreds of eVs to keVs deep in binding energy and their spatial extent is small. In fact, the wave function overlap between an outer target electron and the projectile's core orbital is small. The same argument holds for other kinds of interatomic Auger processes [30][31][32]. The corresponding rates of Auger neutralization and exchange transition are at least 2-3 orders of magnitude too small, i.e., in the range of only 10 11 -10 12 s −1 [29][30][31].
(ii) Sidefeeding is a charge transfer concept that is similar to resonant charge transfer (described above in the case of above-surface transport) and occurs at close target-projectile encounters [6,33]. Here, electrons are transferred from deep target levels resonantly into deep ion levels [see Fig. 3(d)]. In this case the ion deexcites, loses its outer electrons (peel-off), and leaves the target with core holes. In our asymmetric target-projectile system this process is ineffective, because valence electrons are too high in energy and carbon core electrons are only two per atom and still too high in energy (285 eV). Binding energies for n ¼ 6-13 levels of xenon ions in charge states up to 20 are shown in Fig. 4(a), whereas even lower levels with n ¼ 3-5 must be occupied from captured electrons when reaching the atomic ground state. To enhance the efficiency of this process one can think of molecular orbitals formed during the collision [see Fig. 3(d)]. They would promote deep projectile levels and demote high target levels; thus, sidefeeding may be possible again (quasiresonance) [25,34]. To check this we calculated molecular orbitals for xenon and carbon at several interatomic distances, shown in Fig 4(b). Only carbon valence states mix with outer xenon levels. Deep xenon levels are essentially not affected by the collision. Especially in a highly charged xenon ion where the shells are even more strongly bound [see Fig. 4(a)] the effect of molecular orbital formation will be completely absent.
(iii) Resonant capture and enhanced atomic Auger rates would still explain the number of stabilized electrons from the graphene valence electrons occupying ion states with binding energies of 5-25 eV [17]. These electrons may deexcite in the ion radiatively or nonradiatively [6]. Radiative decay is associated with decay times in the nanosecond range [31,35] (only for Kand L-shell filling rates of 10 14 s −1 may be possible [35]). The nonradiative Auger-type autoionization will therefore account for a significant part of the deexcitation sequence [6,14]. The electrons lost by this process must be refilled from graphene as sketched in Fig. 3(c); otherwise, only a few (1)(2)(3)(4)(5) electrons remain stabilized in the ion subsequent to the entire cascade [11]. Resupply of electrons during the short interaction time must involve the frequently made assumption of enhanced atomic Auger rates in hollow atoms [9,10]. To check for this enhancement we used a state-of-the-art relativistic atomic structure code MCDFGME [36,37] and calculated Auger transition rates for specific hollow atom configurations with occupied principal quantum numbers up to n ¼ 13. We find rates in the order of 10 8 -10 12 s −1 that are not enhanced (see Fig. 5). Only for inner shell transitions, i.e., at smaller n ¼ 1, 2, or 3, rates increase to 10 14 -10 15 s −1 [35,38,39]. In our case rates in the order of 10 15 -10 17 s −1 for n > 10 are necessary [21] for the entire cascade to succeed within the interaction time. Because of this large discrepancy we exclude also enhanced atomic Auger rates as the origin for the observed fast electronic decay.
(iv) Finally, an ultrafast electronic decay process allowing energy transfer to the target without electron transport is the only channel left. ICD is a deexcitation mechanism that involves the filling of a core hole by a valence electron of the same atom and the promotion of a valence electron of a neighboring atom into the continuum [see Fig. 3(e)] [12,28]. Commonly, ICD is discussed only in weakly bound systems, such as van der Waals systems. This is due to necessary electron-electron interaction of the outer electrons (of neighboring atoms), which is only strong enough if the electronic orbitals have large extent. In our case the projectile is a hollow atom with many occupied Rydberg orbitals (n > 20) and its distance of closest approach (0.2-1.4 Å) to a target atom is much smaller than typical equilibrium distances in a van der Waals system (≳3 Å). Hence, ICD is very well active in our collisional system. ICD (also known as direct Auger deexcitation) has gained attention due to its importance in photoionization processes of molecules embedded in a liquid environment [12,28,40,41]. ICD still describes Coulomb scattering of target and projectile electrons, which makes it hard to estimate the lifetime dependence on the interatomic distance R ¼ j ⃗ R 1 − ⃗ R 2 j. Both target and (Rydberg-like) projectile electrons have large spatial extent and thus may interact over large distances (1=R dependence). Since we assume not only that the nearest carbon neighbors of the impacting xenon and argon ion participate in the ICD, contributions of next-nearest neighbors should be taken into account. For their contribution Santra et al. derived a 1=R 6 dependence [30] in the so-called virtual photon model, which makes ICD with next-nearest neighbors a local process only present at the small impact parameters of our ions. A more sophisticated treatment of the distance dependence based on the Green's function method includes effects of the electron wave function overlap at close distances [31]. This method revealed an even stronger enhancement of the decay rate for small distances. In fact, an extrapolation of the calculated rates or decay width by Averbukh and Cederbaum for asymmetric MgNe and CaNe dimers shows a decay time constant of about 3 fs at R ¼ 1-2 Å [31]. Recently, we showed that at these small impact parameters a new charge exchange mechanism distinctly different from cases (i)-(iii) is present [11].
In addition it was shown that the rate of ICD depends strongly on the number of nearest neighbors or the cluster size [42,43]. For neon clusters a strong lifetime (1/rate) reduction well below 1 fs was calculated for clusters containing up to 13 atoms [30]. It may even be smaller than 1 fs in a macromolecule such as graphene. Since in our case the entire deexcitation sequence by ICD including most of the captured electrons [up to 30 depending on the initial charge state, see Fig. 2(a)] is finished within 7 fs [21] it is fair to assume a large contribution of next-nearest and even farther neighbors. Yet, ICD within a solid surface is not infinitely fast, since we see a clear time dependence of the neutralization dynamics when varying the ion kinetic energy [see Fig. 2(a)]. By doing so we experimentally determine a time constant τ n for both resonant charge transfer (step 1) and ICD (step 2) depicted in Figs. 1(a) and 1(b) to take place according toN stab ≡ Q in −Q out ¼ Q in ð1 − e −t=τ n Þ with the mean number of stabilized elec-tronsN stab and the incident and mean exit charge states Q in andQ out , respectively, as well as the time t. The time constant depends on the ion charge state Q in with values of a few femtoseconds [see Fig. 2(b)] well in agreement with distance and cluster size dependent values reported in literature [29][30][31]42] and discussed above. This value is an upper bound for the ICD lifetime, because it is averaged over ion-target distances of ∼1 Å (distance of closest approach) to ∼10 Å (distance of first charge transfer).
ICD also predicts slow electron emission from the neighboring atoms, which then carry part of the excitation energy. Electron emission from surfaces triggered by projectile deexcitation was previously measured directly showing that the yield γ is about 2-3 times the incident charge state [γ ¼ 200 electrons=ion for Th 79þ on clean Au(111) [44,45] ], the mean of the energy distribution is below 20 eV [46,47], and emitted electrons can even be correlated [48]. These facts are well in agreement with an ICD process. For 263 keV Xe 30þ ions we have now directly measured the electron yield from graphene in coincidence with ions transmitted in charge states Q < 30, i.e., for ions that have passed the graphene layer. As a result we find γ graphene ≈ 21 electrons=ion with an estimated collection efficiency of 28% AE 5%. This yield (corrected by efficiency γ corr ≈ ð75 AE 20Þ electrons=ion) is remarkably high considering only a single layer of carbon atoms.
In conclusion we determined experimentally neutralization and deexcitation time constants of highly charged ions of about 1-3 fs. We find evidence that HCIs neutralize by hollow atom formation and the subsequent deexcitation is mainly driven by ICD. Further evidence for ICD is presented by the measured emission of > 20 electrons per impacting ion.
Financial support by the Deutsche Forschungsgemeinschaft (Grant No. WI 4691/1-1), the Austrian Science Fund (Grant No. I1114-N20), and the TU-D doctoral college of TU Wien is gratefully acknowledged. R. K. and M. S. are grateful for support from SFB 1242 "Non-Equilibrium Dynamics of Condensed Matter in the Time Domain" and SPP 1459 "Graphene."