Conference-ISSS-7-Ways to Spin Resolved Core-Hole-Clock Measurements

We describe two different ways to spin resolved measurements of charge delocalization times by resonant photoemission. The first approach is based on spin selective population of the core-to-bound resonance, and the second on spin resolved detection of the emitted decay electrons. Whereas the first method requires distinct electronic properties of the material under investigation, no such limitations exist for the second approach. In particular, it can be used for studies of organic materials containing the light elements of the second row of the periodic table.


I. INTRODUCTION: CORE-HOLE-CLOCK BASICS
The core-hole-clock (CHC) method is an approach to measure delocalization times of electrons utilizing the lifetime of inner-shell vacancies-the core holes-as an internal time standard [1][2][3][4].It has been used for investigations of charge transfer times from adsorbates to their substrates, e.g. for physisorbates on metals [5][6][7] and semiconductors [8], but also for chemisorbates on metals [1][2][3][4][9][10][11][12], including strongly chemisorbed radicals with charge transfer times in the sub-femtosecond range [10][11][12].It was also used to determine charge transit times through thin dielectric layers [13] and through monolayers of graphene on different substrates [14], and it was extensively applied in studies of charge transport dynamics in hybrid systems containing larger organic molecules [3,4,15].For self-assembled monolayers, charge transfer times from distinct ligands through the aromatic or aliphatic backbones to the substrate have been explored [16][17][18][19][20].By varying the length of the backbone, it was possible to untangle the transit times through the backbone, and across the anchor group at the interface [17].In addition, the CHC method has resolved ultrafast moleculeto-substrate charge transfer of less than 3fs between biisonicotinic acid and its TiO 2 substrate, a model system for the dye-to-substrate contact in dye-sensitized solar cells [21]; and CHC proved to be useful for studies of intermolecular and molecule-to-interface charge transfer dynamics in materials that are expected to be relevant for molecular electronics applications (see [15] and references therein).
The basics of the method are illustrated in Fig. 1.Core holes with not too large binding energies-which means not larger than that of the Z ∼ 30 K-shell-decay predominantly non-radiatively by a Coulombic or Auger process [22].An electron from a shell with lower binding energy fills the primary vacancy and a second electron carrying the excess energy is emitted.Core ions decay into two hole (2h) final states (Fig. 1a, left side); resonantly excited core states of isolated atoms or molecules into one hole (1h), or two hole-one electron states (2h1e), depending if the resonantly excited electron takes part in the decay process (participator) or not (spectator; see the cartoons in Fig. 1a).The participator states correspond to 1h states in non-resonant photoemission and the spectator states to shake-up satellites.
Fig. 1a displays exemplary 2h (black), and 2h1e plus 1h (red) decay spectra [23].For most particles, 2h and 2h1e spectra resemble each other, but with a blue-shift (spectator shift) of the 2h1e part on the kinetic energy scale because the additional electron screens the two holes.Due to energy conservation, the kinetic energy of the decay electron disperses for the resonant case with the primary excitation energy-commonly that of a narrow bandwidth photon from a synchrotron radiation source; decay electrons for non-resonant ionization appear at constant kinetic energy.For atoms, linear dispersion is observed [1-8, 13, 14, 24] whereas for molecules linear dispersion may be obscured by experimentally unresolved vibrational progressions of the core-excited and the final states, and by the evolution of the bond geometry during the lifetime of the core hole [25].We note that energy conservation applies also for core ionization followed by a normal Auger decay (Fig. 1a).Here, however, the total energy of the pair of Auger and photoelectron disperses with the excitation energy, an effect that is only resolved by electronelectron coincidence spectroscopy [26].
The situation is changed if the core-excited particle is in contact with a system that has affinity level at the energy position of the resonantly excited electron, e.g. the empty conduction band of a solid (Fig. 1b).The excited electron can then delocalize into these empty states; if this transfer is fast, even before the core hole decays.Depending if delocalization or core decay is faster, either 2h, or 2h1e FIG. 1. a) Exemplary core decay spectra for core ionization (core-to-continuum transitions, left), and resonant core excitation (core-to-bound transitions into empty valence or Rydberg levels, right) of an isolated particle.Core ionized states decay predominantly into two hole final states (2h, black), resonantly core excited states into one hole participator (1h) and two holeone electron spectator states (2h1e, red; see text for details).b) Situation where the resonantly excited electron can delocalize into a coupled continuum.For similar lifetimes of resonance and core hole, a mixture (blue) of 2h1e and 1h states (red), and 2h final states (black) is obtained.Decay satellites, e.g.shake-ups upon decay, are not shown; for most systems their intensities are an order of magnitude or more smaller than those of the main lines.and 1h final states result [1][2][3][4].If core decay and charge delocalization occur on similar time scales, the spectrum shows a mixture of both (Fig. 1b) [1][2][3][4].Apart from a reduced spectator shift by different extra-atomic screening of the 2h and the effective one hole states (Fig. 6 shows an example for this effect), the spectra for the scenario of Fig. 1b resemble those obtained for the isolated particle (Fig. 1a), but are not identical.Particularly the 2h contributions must be expected to differ.Compare, e.g., the core decay of [np] states that either have been created by a core-to-continuum transition or by charge transfer ionization of a [np]n's resonance.In the first case, the core electron is excited into continuum waves with s and d symmetry, in the latter exclusively into an orbital with s symmetry.The spatial alignment of the core holes in terms of np z and np x,y contributions will differ and the angular distribution of the decay electrons accordingly [27].
For many systems the 1h and 2h1e, and the 2h contributions to the combined spectrum-often labeled as "coherent" and "incoherent" contributions, because the coherence of the wave function is lost upon charge transfer-can be untangled, either because the spectator shift is large enough, or because parts with constant kinetic or constant binding energy can be discerned.With the intensities I of the different contributions to the decay spectrum and the known lifetime of the core hole τ core [28,29], the charge transfer time τ ct can be calculated [1][2][3][4]: The CHC approach has several assets.As an inner shell method, it is element selective.In addition, it is also orbital selective, either by fine-tuning of the excitation energy [19], or by exploiting different light polarizations and selection rules [12,20].The "starting point" of the delocalization process is therefore well defined.Our goal was to make CHC spin selective as well.Spin dependent charge transfer is of large interest for practical applications like molecular spintronics [30], where electron transport across interfaces is always involved [30], but also for more fundamental considerations, e.g. for an experimental answer to the question if large densities of empty affinity states or large transfer matrix elements are more important for charge transfer processes in individual systems [32].In the next chapters, we describe two different ways to spin resolved CHC, which, as we believe, are valuable complements to electrode based transport methods [33][34][35].

II. SPIN RESOLVED CORE-HOLE-CLOCK MEASUREMENTS A. Spin Selective Population of the Resonant State
Figure 2 illustrates our first approach to spin resolved CHC [36].A core electron of a particle that is in contact with a magnetized substrate is resonantly excited into an empty orbital.Using circularly polarized light and taking account of the electronic properties of the particle, in parhttp://www.sssj.org/ejssnt(J-Stage: http://www.jstage.jst.go.jp/browse/ejssnt/) e-Journal of Surface Science and Nanotechnology ticular of the spin orbit splitting of its electronic levels, electrons with spin directions equal to the minority (top row of Fig. 2) or the majority spins of the substrate (bottom row of Fig. 2) are selectively promoted.If we assume in zeroth order that delocalization is faster the larger the density of empty states in the substrate, we can expect that resonant states populated with minority spins will transfer their electrons faster than those populated with majority spin.As a result, core decay will yield comparatively more 2h final states for the first case and more 2h1e and 1h states for the second, see the scenarios depicted in Fig. 2. The advantage of this method is that, apart from circularly polarized light of appropriate photon energy and bandwidth that is available at practically every synchrotron radiation source worldwide, only a standard electron energy analyzer is required-no spin filter is necessary that inherently reduces the detection efficiency.
We have used this method to investigate the spin dependence of charge transfer from physisorbed Ar to magnetized Fe(110), Co(0001) and Ni(111) substrates [36].We have selected Ar for this proof of principle experiment because it was used in many CHC experiments before (see above), meaning that its spectroscopy is well known and that theoretical work on the electronic coupling of its core-excited state to the substrate is available [6,7].In addition, the [Ar2p 3/2 ]4s resonance can be easily spin polarized according to the scheme shown in Fig. 3 [37].One could doubt that physisorbed Ar is a good test system for a method that shall be used for organic adsorbates, i.e. molecules that encounter chemical interaction with the substrate.Ar is in its electronic ground state a closed shell atom that binds exclusively by Van der Waals-forces with its substrate.However, according to the well-known equivalent core or Z +1 approximation, Ar in the electronically excited [2p 3/2 ]4s state mimics an open shell alkali atom.The small [2p 3/2 ] hole has the same effect as an additional positive charge of the nucleus and core-excited Ar * behaves chemically like a neutral K atom in front of the surface.It interacts with sp and d levels of the substrate, similarly as organic molecules (see the theoretical results of refs.6 and 7).The applicability of Ar * on metals as a test system for charge transfer measurements from adsorbed organics is also approved by the very similar charge transfer times that are measured here (between 2 and 3.3 fs, see below) and the value of 2.3 fs obtained for charge transfer across the thiolate anchor of self-assembled aliphatic monolayers on gold [17].
Our Fe, Co and Ni substrates were thin films grown on a W(110) single crystal.The film thickness has been controlled by a quartz microbalance, and the cleanliness of the W substrate and the evaporated films by XPS.After annealing of the films, LEED showed for all materials well-ordered close-packed layers.The films were then magnetized with a magnetic field of 0.25 Tesla, and the magnetization was checked by XMCD.After this, a saturated monolayer of Ar was prepared by dosing Ar in excess and removing the multilayer fraction by thermal desorption under XPS control.After a check of the [Ar2p 3/2 ]4s energy position by NEXAFS, the resonance was excited with circularly polarized light of plus and minus helicity under grazing incidence, and decay electrons emitted in normal direction were recorded with a standard hemispherical electron energy analyzer (Specs Phoibos 100). Figure 4 shows the geometry used in our experiments [36].
The obtained decay spectra differed for Ar on Fe(110) and Co(0001), depending whether the [Ar2p 3/2 ]4s resonance was populated with minority or majority spins, but not for Ni(111) (Fig. 5).In Fig. 6 we exemplarily show how these decay spectra were disentangled into 2h1e, 2h, and satellite contributions (the 1h participator is negligible for the [Ar2p 3/2 ]4s excitation [39]).For Co and Fe, the 2h1e part of the spectra at higher kinetic energy was larger for excitation of majority than for minority spins, indicating faster delocalization of the latter into the substrate.
A careful evaluation of several sets of measurements with all possible combinations of substrate magnetization and light helicity was performed.Untangling the 2h and 2h1e contributions to the spectra as shown in Fig. 6 and applying the CHC formula (Section I) yielded charge transfer times for majority/minority spins of 2.67/2.08fs for Ar/Fe(110), 3.24/2.63fs for Ar/Co(0001), and 3.12/3.12fs for Ar/Ni(111), with errors of ±0.15 fs for all values [36].Decay satellites (see Fig. 6) and the incomplete polarization of the resonance according to Fig. 3 110), but only a small difference for Ni(111) [40].Our results are, therefore, in line with the assumption illustrated in Fig. 2 that the empty density of states (DOS) governs the charge transfer rate.We note, however, that a theoretical study which for core-excited Cs on Fe(110) investigated the spin dependent charge transfer from the Cs 6s orbital to the substrate came to a different result.For this system, which chemically is very close to ours, note the Z + 1 approximation mentioned above, preferred transfer of majority spins was calculated [32].This study showed that, apart from the empty density of states, the matrix element of the transfer process is an important quantity.Because of the more far-reaching wavefunctions of Fe sp-bands, the Cs 6s electron was found to delocalize predominantly into empty sp-states, despite the smaller sp-density of states [32].Because d-and sp-bands are differently spin polarized for Fe(110) [32,40], the calculation yielded faster transfer of majority electrons for this adsorbate/substrate combination [32].As yet, this interesting discrepancy between the Ar and Cs results is not resolved, but further theoretical and experimental studies are under way.
We believe that spin resolved CHC is perfectly suited for investigations of this delicate balance of DOS and matrix element effects for different systems, because it enables direct access to the spin orientation of the excited electron.Methods based on conduction require always contact with two interfaces (see refs.33-35 for an overview); in particular, if identical materials are used for the two electrodes, the above mentioned DOS and matrix element effects may be inseparable.In addition, these methods allow only indirect access to the charge transfer time via conductance data whereas CHC delivers this quantity directly.We also have tried to extract spin dependent charge transfer times from the widths of the [Ar2p 3/2 ]4s maxima seen in XMCD for excitation of minority and majority spins.We found slightly broader maxima (by 5%) if the resonant state was populated with minority spin, indicating a faster transfer to the substrate http://www.sssj.org/ejssnt(J-Stage: http://www.jstage.jst.go.jp/browse/ejssnt/) e-Journal of Surface Science and Nanotechnology in qualitative agreement with our CHC results (see Supplemental Material of ref. 36).Due to adsorption-induced inhomogeneous broadening of the resonance and other broadening effects, a reliable quantitative evaluation was not possible.For molecules additional broadening by vibrational progressions will make such an evaluation even more difficult, and for materials composed of elements from the second row elements of the periodic table XMCD will show no spin contrast because of lacking sufficiently large spin orbit splitting, see below.

B. Random Spin Polarization of the Resonance Combined with Spin Resolved Detection of the Decay Electrons
The way to spin resolved CHC described in the last chapter works well; the detection efficiency is very good because no spin filters are needed for the decay electrons.It necessitates, however, appropriate l, m/l ′ , m ′ levels for the ground state, and the intermediate core-excited state, in order to enable spin polarized population of the resonance.In practice, well resolved spin orbit multiplets are required, typically with a spacing of the individual maxima of 1 eV or more; particularly for molecules with broad vibrational progressions even 1 eV might not suffice.Appropriate levels with a spacing of 1 eV or more exist for phosphorus and heavier elements, but not for the light elements of the second row of the periodic table like C, N, O and F that are of large interest for all applications comprising organic compounds for which CHC has proved to be a very powerful method, see above.For these atoms and for molecules containing these elements we suggest a different approach to add spin selectivity to the existing assets of CHC.It has been sketched in ref. 36 and will be explained here in more detail, see Fig. 7.
Opposed to our first approach, the resonance is populated equally by minority and majority spin, e.g. by excitation with linearly polarized light.As for the method described in Section II.A, we assume that minority spins delocalize faster because of the larger empty density of states in the coupled continuum.In other words, core decay into 2h final states will predominantly occur for core holes with minority spin orientation, see Fig. 7.Because the spin is preserved upon core decay and elastic charge transfer, the decay electrons will carry information on the orientation of the core hole, and therefore on the spin of the electron that has been transferred into the continuum as well, see Fig. 7.For final states with two holes of anti-parallel spins, electrons with majority spin orientation will be emitted, whereas for final states with two holes exhibiting parallel spins, decay electrons with minority spin orientation are ejected (Fig. 7).The former will contribute to the singlet peaks of the spectrum, whereas the latter appear in the triplet maxima.For the coherent peaks similar considerations hold.In summary, the spin polarization-averaged intensity ratio of the (2h1e + 1h) and 2h contributions in combination with the lifetime of the core hole enables the retrieval of the charge transfer time, whereas the evaluation of the spin polarization of the incoherent or/and the coherent part of the decay spectrum yields the spin contrast.Particularly for KLL decays of the elements of the second row of the periodic table, the evaluation is further simplified.For these atoms, triplet Auger final states are practically absent [41].As a result, it is not necessary to treat individual maxima of the incoherent decay spectrum differently as indicated by Fig. 7.
We have tested this second approach to spin resolved CHC with Ar/Co(0001), an adsorbate/substrate system that was extensively investigated with our method outlined in chapter 2.1.[36].The experiments have been performed at the UE56-2-PGM-2 beamline of BESSY operating in the single bunch mode.A time-of-flight spectrometer coupled to a Mott polarimeter was used for the detection of the decay electrons.Within experimental errors, the results of methods 1 and 2 agree well [26,36,42].A more refined discussion of these results accounting also for effects of dynamical and transferred spin polarization [43] will be published elsewhere [44].

III. SUMMARY AND OUTLOOK
In summary, we have shown that spin resolved CHC is feasible via two different ways, either by spin selective excitation of the resonance or spin resolved detection of the decay electrons.Particularly for the second method, we foresee a wide range of applications, especially for organic systems on surfaces.For these adsorbates, the method will benefit from the large excitation cross sections commonly observed for π-resonances of organic molecules and from novel spin filters with large transmission probability.The CHC method is a "single electrode" method; cancellation of electrode effects, which are possible for, e.g., the mechanical break junction approach and other contact techniques [33,34] or spin resolved scanning tunneling microscopy [35] are avoided.We believe that spin resolved CHC will provide interesting details on electronic coupling in future experiments, particularly on the importance of the matrix elements related to the charge transfer processes [32].
We therefore suggest to apply this method to investigations of DOS-dependent excitation and tunneling effects; studies of low-lying core-resonances of chemisorbates will be of large interest, cf. the Cs/Fe(110) case [32].Within this context, the spin polarization already during the excitation step by promotion of the core electron into spin polarized empty states [45] will be of great interest.In addition, spin resolved charge transport through thin layers, including layers of chiral molecules and intercalated thin metal layers, is an appealing topic.Finally, with low temperature equipment as described in [46], we plan to extend our experiments to layers of molecular magnets.

FIG. 2 .
FIG. 2. Spin resolved CHC measurement by spin selective core-to-bound excitation of a particle in contact with a magnetized substrate.Red arrows denote minority, blue arrows majority spins; thick/thin black arrows indicate fast/slow processes.If the empty density of states governs the delocalization rate, minority spins (top) will delocalize faster than majority spins (bottom), resulting in different ratios of 2h compared to 2h1e and 1h final states for different spin polarizations of the resonantly excited electron with respect to the substrate magnetization.A level diagram for spin selective core excitation with circularly polarized light is shown in Fig. 3.No spin filter is required at the detector side.

FIG. 3 .
FIG.3.Spin selective population of the [Ar2p 3/2 ]4s resonance with circularly polarized light.The numbers in the circles denote the transition probabilities[37].A spin-population ratio of 3:1 is possible for the core-excited resonance.

FIG. 5 .
FIG. 5. Decay spectra obtained for Ar monolayers on Fe(110), Co(0001) and Ni(111) films as a function of the spin polarization of the resonance (after ref. 36).For Fe and Co, the spectator contributions at high kinetic energy (cf.Fig. 6) are larger for excitation of majority spins than for excitation of minority spins, indicating faster delocalization of the latter.

FIG. 6 .
FIG. 6. Decomposition of the decay spectrum of Ar/Co(0001) into [3p 2 ]4s 2h1e states (brown), [3p 2 ] 2h states (blue) and [3p 2 ]5s 2h1e satellite states (fawn, after ref.36).For population of the resonance with minority spins which delocalize faster (not shown) the coherent 2h1e part would be decreased and the incoherent 2h part increased.Compared to decay spectra from isolated Ar atoms[38], the energetic ordering of the [3p 2 ] 2h states and the [3p 2 ]5s 2h1e satellite lines is reversed.For adsorbed Ar, the doubly charged 2h states are more strongly blue-shifted by screening through the substrate than the effective 1h states of the satellites and appear at a larger kinetic energy than the latter.

FIG. 7 .
FIG. 7. Spin selective CHC measurements by non-spin selective excitation and spin resolved detection of the decay electrons.As for the scenario depicted in Fig. 2, minority electrons (top) are assumed to delocalize faster, leading to a net spin orientation of the core holes for these intermediate states.Because no spin flip occurs upon core decay, the emitted electrons are spin polarized, reflecting the spin orientation of the core hole as shown here for the 2h final states.The sizes of the cartoons and arrows indicate the intensities of the different channels.Different spin orientations are expected for electrons detected for final singlet and triplet states.For the final states obtained for core de-cay before charge delocalization (2h1e and 1h) similar considerations apply.
have been accounted for; the lifetime of the core hole was taken from ref. 29 (see ref. 36 and its Supplemental Material for further details of the data evaluation).Literature data on empty d-state densities at the energy position of the [Ar2p 3/2 ]4s resonance show larger values for minority spins for Co(0001) and Fe(