Surface science using radioactive ions at ISOLDE: from metal surfaces to two-dimensional materials

We review the research carried out using the apparatus for surface physics and interfaces (ASPIC), at ISOLDE, CERN. We give an overview of the research highlights since 2000, focusing on magnetic and non-magnetic metallic surfaces, and introduce the scientific program that will follow the upgrade which is currently underway, focusing on two-dimensional materials. ASPIC was formerly used for the growth of ultrathin metallic films and their characterization by means of perturbed angular correlation spectroscopy. Past research has mainly focused on the determination of the magnetic hyperfine field at the probe atom located on different sites at the surface such as terraces, kinks, steps as well as on the investigation of the static magnetic polarization at the interface between ferromagnetic and paramagnetic layers. Future research on two-dimensional materials using ASPIC is foreseen to focus on the investigation of structural and electronic properties of adatoms (adsorption sites, hybridization effects, intra-atomic charge transfer, magnetic moments, etc). We emphasize, in this context, the exceptional capabilities of ASPIC in terms of broad applicability, high precision and low detection limits.


Introduction
Reduced dimensionality is a rich playground for contemporary solid state physics, made accessible by continuous progress in experimental techniques which can probe ever smaller amounts of matter.The advent of nanoscience and nanotechnology is a clear manifestation of that: materials are made so small (nanometer scale) that new physical phenomena emerge as they become nearly two-dimensional (surfaces, thin films, and interfaces), one-dimensional (nanowires) or zero-dimensional (nanoparticles).Graphene, a two-dimensional monolayer of C atoms, is one of the richest and most recent manifestation of the revolutionary new physics that emerges from the solid state in reduced dimensionality.
Basic research on solid state physics, at any dimensionality, is strongly based on the understanding of the interplay between local structure (the arrangement of atoms) and electronic functionality (electric, magneticK).However, experimentally investigating structural and electronic properties of 2D systems (surfaces, interfaces and two-dimensional materials) implies probing a small amount of material, confined within one or few atomic monolayers.This raises a number of challenges, not only because it requires detecting a small amount of material, but also distinguishing it from the macroscopic counterpart associated with the 2D system: the bulk corresponding to the surface; the materials above and below the interface; the substrate in which the two-dimensional material is supported, etc.Furthermore, investigating the link between a specific atomic configuration and the resulting electronic properties requires both high precision on the structural parameters (atomic scale) and the ability to unambiguously link different configurations with the respective electronic behaviors.
Experimental techniques based on hyperfine interactions, and in particular using radioactive isotopes, are ideally suited to tackle the challenges raised by 2D systems.First, hyperfine parameters translate to structural, electric and magnetic observables which are measured locally (at the atomic scale) and simultaneously.Second, using radioactive probes makes it possible to probe small amounts of atoms, orders of magnitude below the monolayer regime of ∼10 15 atoms per cm 2 .This low detection limit is a general consequence of the fact that, within one half-life, half of the probes emit radiation which can contribute to a spectrum.This is in contrast with experimental techniques based on external beams (of photons, electrons, ions, neutrons) which depend on (generally small) cross sections for interaction between the probed atoms and the beam particles.Third, the detected radiation originates uniquely from the radioactive probes, without background signals from rest of the material system.
The apparatus for surface physics and interfaces (ASPIC) was developed at ISOLDE, CERN, to explore this untapped potential of hyperfine techniques.It is based on (timedifferential) perturbed angular correlation (PAC) spectroscopy and was envisioned for the study of magnetic surfaces, thin films and multilayers.For that purpose, ASPIC combines PAC spectroscopy with conventional surface preparation and thin film growth techniques, all in situ, in an ultra-high vacuum (UHV) environment.The technical details of the ASPIC chamber along with the description of the research before the year 2000 have been reviewed in Bertschat et al (2000).Although similar setups have been implemented elsewhere (Schatz et al 1989, Laurens et al 1999), ASPIC is unique in the sense that it enables experiments on the widest range of material systems, since it is installed online at ISOLDE, the radioactive ion beam facility providing radioactive isotopes of the widest range of elements in the periodic table.ASPIC is currently undergoing an extensive upgrade (surface characterization instrumentation, vacuum, etc), which includes the relocation at the versatile ion-polarized techniques online (VITO) beam line.The new VITO beam line is a modification of the former UHV beam line dedicated to ASPIC, and it has been under construction since the beginning of 2014 (Ruiz et al 2015, Stachura et al 2016).It is aimed to deliver beams of either spinpolarized or non-spin polarized atoms/ions to three fully independent experimental stations (operating from 10 −10 to 50 mbar): in addition to ASPIC, a β-asymmetry station and openend station suitable for traveling experiments.VITO opens up numerous new opportunities for multidisciplinary science in the areas of solid state physics, biophysics, nuclear physics and fundamental interactions (Ruiz et al 2015, Fenta et al 2016, Stachura et al 2016).
Here we review the research carried out using ASPIC since 2000, on metal surfaces (section 2), and introduce the scientific program that will follow the upgrade which is currently underway, focusing on two-dimensional materials (section 3).

Research at ASPIC since 2000: surfaces of metals
From 2000, the experiments performed in ASPIC can be divided in three main categories: (1) magnetic hyperfine fields and anisotropies measured using 111 Cd probe atoms on ferromagnetic Ni surfaces; (2) magnetic polarization of a Pd layer induced by a Ni substrate measured using 111 Cd probe atoms; (3) temperature dependent behavior of isolated 77 Se probe atoms adsorbed on different metal surfaces.

Magnetic hyperfine fields and anisotropies of ferromagnetic Ni surfaces
This category represents the major research performed using ASPIC after the year 2000 (Potzger et al 2002, Prandolini et al 2004, Manzhur et al 2005a, Potzger et al 2005, Manzhur et al 2007).The basic idea behind these experiments is the fact that a surface offers a variety of different local structures, (terraces, steps, kinks, etc).The probe atoms located at all sites available would allow for the determination of the magnetic hyperfine field depending on the number of nearest neighbors (NN), i.e. the coordination number which are different for different surface structure, even if maximum coordinated (hollow) adsorption sites are expected as in the case of metal-metal adsorption.Magnetic properties induced by the NN of the probe atoms are expected to significantly depend on the coordination number as well as their arrangement.As we already outlined in Potzger et al (2002), the magnetic hyperfine field B hf at the nuclei of the probe atoms which are in/on ferromagnetic materials are caused by electronic spin and orbital contributions.Using the non-transition and non-magnetic impurity Cd, the polarization of the s electrons arising from the hybridization with the valence d electrons of the ferromagnetic host is probed by B hf which interacts with the nuclear magnetic dipole moment.The advantage of the application of radioactive probe atoms is the fact that the local electric field gradient (EFG) at the structure which interacts with the quadrupole moment of the probe nucleus is sensitive to local lattice symmetry.The values of the EFG components experienced by the probe nuclei at the various surface structures are known from earlier experiments on non-magnetic surfaces (Klas et al 1988, Wesche et al 1989, Hunger and Haas 1990).Since the recorded PAC spectrum contains information of both EFG and B hf , the latter can be directly assigned to a certain local symmetry arising from the coordination number experienced by the probe atoms located at a certain surface structure.The deconvolution of the EFG and the B hf was performed using the DEPACK program developed by Lindgren (1980).
The radioactive probe atoms used for the experiments are either 111m Cd or 111 In.They have been produced and mass-separated using the mass separator ISOLDE at CERN (Köster et al 2003).Both of them decay to the PAC probe 111 Cd.Details about the probes nuclear properties as well as the experimental processes, especially how the probes are positioned onto the precleaned surfaces can be found in Potzger et al (1998) and Bertschat et al (2000).As the ferromagnetic host, the authors used Ni(001) and Ni(111) single crystals.Ni exhibits a saturation magnetization of 57.5 Am −2 kg −1 at 0 K and a Curie temperature of 631 K (Bozorth 1993).The Ni(111) surface is dense packed with a coordination number of 3 for the adatom at the hollow site and a coordination number of 9 at the site in the terrace.The Ni(001) surface exhibits respective coordination numbers of 4 and 8 (figure 1).
The major outcome of this research is shown in figure 1.In earlier studies, only the modulus of the magnetic hyperfine field could be measured since no external magnetic field could be applied to the sample in ASPIC.For the experiments described in Manzhur et al (2007), the external magnetic field was realized inside the UHV by means of permanent magnets producing 0.05 T at the sample.A positive/negative magnetic hyperfine field is defined to be aligned parallel/antiparallel to the film magnetization.As outlined in Manzhur  et al (2007), 111 Cd probe atoms with 9, 7, 6 or 5 nearest Ni neighbors have negative hyperfine fields.At the hollow adatom site on Ni(001), i.e. for 4 nearest neighbors, the hyperfine field is positive.The sign of the hyperfine field at the hollow 111 Cd adatom site on Ni(111), i.e. for 3 nearest neighbors was not measured but assigned to be positive from the systematics.The origin of the sign of B hf for different coordination numbers was explained by theory (see references in Manzhur et al (2007)).B hf is mainly determined by the s-electron spin polarization.It was found in Bellini et al (2004) that it is governed by the general shape of the s-electron density, i.e. by the presence of a characteristic antiresonance dip (AR) in the s-electron density below the Fermi energy arising from the 5sp-3d bonding between impurity atom and host.Note that the s-d hybridization is different for the bonding states below the AR as compared to the antibonding states above the AR which leads to an opposite sign of the s-electron magnetization depending on the position of the Fermi level, i.e. depending on the 5sp element used as probe.Moreover, the antibonding states above the AR are split two parts when the symmetry is reduced, i.e. at surfaces.Depending on the particular value of the coordination number NN, whose reduction leads to band narrowing as well as the position of the Fermi level, the s-spin density can adopt positive or negative values.
The high sensitivity of the hyperfine field on the local lattice structure was further proven in Manzhur et al (2005a), where 111 Cd probe atoms are soft landed onto a two monolayer thick Ni film deposited on a Pd (001) single crystal.The measurement of the EFG components for the hollow-site adatom position with 4 nearest neighbors revealed that the ultra-thin Ni film is highly strained as compared to the Ni single crystal but still exhibits (001) orientation similar to the Pd substrate surface.The measured value of the magnetic hyperfine field, however, was found to be |5.2T| as compared to |7.3 T|, i.e. about 30% smaller.

Magnetic polarization of a Pd layer induced by a ferromagnetic Ni substrate
This part of the research using ASPIC is mainly motivated by the fact that the conduction electrons of a paramagnet such as Pd can be magnetically polarized by an adjacent ferromagnetic layer such as Ni.This static polarization is expected to extend only to the first layers of Pd in contact with Ni.Full-potential linearized augmented-plane-wave calculations of the spin-dependent density of states for transition-metal overlayers on Pd (Blügel 1988) showed that the magnetic moment in Pd induced by Ni amounts to 0.24 μ B in the interface layer and 0.20 μ B in the second layer.Positioning radioactive probe atoms in those layers allows for a direct measure of the magnetic polarization via the magnetic hyperfine field.A comprehensive description of the measurements and results can be found in Potzger et al (2005) and Manzhur et al (2005b).
For the preparation of the films, ASPIC has been equipped with an e-beam evaporator for Pd thin film growth.The structure of the Pd overlayers has been controlled by means of low energy electron diffraction.It was found that Pd grows in a hexagonal close packed structure which is quasi-incommensurate with respect to the substrate surface.This effect does not depend on whether Ni(001) or Ni(111) single crystals have been used as substrates.The quasi-incommensurate epitaxial relation can be quantified as c(13×13)Pd/Ni(111) and c(16×2)Pd/Ni(001).As in the case of pure Ni surfaces, again different structures at the surface such as terraces and kinks occur which give rise to different absorptions sites of the 111 Cd probe atoms but now with mixed coordination numbers since the nearest neighborhood consists of both Ni and Pd atoms, the latter with different magnetic moments depending on the distance from the Ni/Pd interface.Due to the incommensurate structure of the Pd overlayers, the same absorption site should exhibit varying coordination numbers, e.g. 6 Pd and 1..3 Ni nearest neighbors for Cd atoms located in a terrace in a Pd layer on Ni(111) and 6 Pd and 1..4 Ni nearest neighbors for Cd atoms located in a terrace in a Pd layer on Ni(111) (Manzhur et al 2005b).Therefore, it is expected to obtain overlapping hyperfine parameters leading to unclear PAC-spectra.Since this is not the case (Potzger et al 2005) the authors assumed that the 111 Cd probe atoms always adopt the lattice site with the highest coordination representing the local energy minimum arising from the metallic-like bonding.
The major outcome of the research is shown in figure 2, where the expected magnetic moment of the cluster formed with the nearest neighborhood shell (as expected from Blügel (1988)) was compared to the modulus of the magnetic hyperfine field measured at Cd probe atoms located at different surface structures with different mixed coordination numbers.It was found that for the determination of the static magnetic polarization at the Pd interface, | B hf | basically follows the trend of the calculated magnetic moment.

77 Se adatoms on metal surfaces
While the research reviewed in the previous two parts was devoted to the investigation of magnetic properties of metallic layers, this third part focuses on the understanding of the fundamental behavior of dilute, i.e. non-interacting chalcogen adsorbed atoms (adatoms) on metal surfaces, especially with respect to its temperature dependence.For that purpose, where Q N is the known nuclear quadrupole moment of |0.76 b| (Granzer et al 1996) for 77 Se, and V zz represents largest component of the unknown EFG tensor induced by the nearest neighborhood of probe adatom at the surface, linearly increases with increasing temperature for all close-packed surfaces investigated.Note that, as for B , hf only the modulus of V zz and thus Q n can be measured by PAC.The deviation from this behavior for Pd(111) observed above 500 K was explained by lattice vibrations.The quadrupole coupling constant and thus the EFG for the (001) surfaces of Pd and Ni, however, linearly decrease with decreasing temperature.The authors assumed, that on (001) surfaces, the probe atoms would occupy fourfold coordinated hollow sites being closer to the substrate as compared to threefold coordinated positions of the adatoms for the close-packed surfaces of Pd(111), Ni(111) or Co(0001), respectively.An explanation of the behavior was given by referring to theoretical calculations of the EFG as a function of the distance between the probe atom and the top layer of the surface (Lindgren 1986).Using those calculations, a negative EFG at the probe atom could be assigned to a surfaces with (001) orientation and a positive EFG to a surfaces with (111) orientation due to the different distances.Using the calculations once more, consistency with the measurements presented in figure 3 was achieved if the temperature increase is again associated with an increasing distance between the adatom and the top surface layer.

Future physics at ASPIC: adatoms on 2D materials
From the moment it was isolated as a two-dimensional material (Novoselov et al 2004), graphene has become a remarkable subject of research, exhibiting novel phenomena that extend to virtually every domain of solid state physics (Geim and Novoselov 2007, Geim 2009, Neto et al 2009, Novoselov et al 2012).The success of graphene has also triggered the birth of a whole new field of two-dimensional materials, including the group-IV analogues of graphene (silicene and germanane), MoS 2 and other metal dichalcogenides, phosphorene, among numerous other classes of materials (Novoselov et al 2012, Bhimanapati et al 2015).A particularly active domain of research deals with the interaction between the 2D lattice and atoms which are adsorbed on its surface (adatoms).For example, adatoms can be used to modify the electronic properties (e.g.electrical doping (Ren et  ).In all these contexts, ASPIC offers tremendous opportunities to locally probe structural and electronic properties (adsorption sites, hybridization effects, intra-atomic charge transfer, magnetic moments, etc).In this section, we focus on the case of magnetic adatoms on graphene as a showcase of the unrivalled capabilities of ASPIC.
Even though graphene is not intrinsically magnetic, spintronic phenomena can be induced by magnetic point defects such as vacancies (Nair et al For example, the extreme single-atom anisotropy of a magnetic adatom on a graphene sheet, combined with strong spin-orbit coupling, results in a large magnetocrystalline anisotropy energy (MAE).Thanks to the strong sensitivity of graphene to external electric fields, MAE values of up to 30 meV (i.e. with characteristic temperatures of ∼300 K) were predicted for 5d transition metal adatoms (Zhang et al 2012), i.e. much larger than ∼1 meV for 5d adatoms on metallic surfaces (Błoński et al 2010).Owing to their strong spin-orbit coupling, heavy (4d and 5d) transition metals and (4f) rare earths are expected to exhibit the highest MAE values.However, while promising theoretical work in this context has been rapidly accumulating, experimental progress has been hindered by limitations of currently available techniques.Indeed, adatom magnetism is not only a good example of the rich new physics underlying two-dimensional materials, but it is also a showcase of the tremendous challenges that experimentalists must face in the flatland.
One the biggest challenges of all is of a very fundamental nature: how to locally probe the magnetic moment of a single atom without affecting it?ASPIC allows precisely that, since it is based on measuring radiation emitted upon decay of radioactive adatoms, i.e. by completely avoiding any influence of external probes (e.g. a tip in scanning probe microscopy (Elbo et al 2013)).The hyperfine parameters measured with PAC (EFG V zz , asymmetry parameter η, and hyperfine magnetic field B hf ) can then be used to determine the position 4. Density functional theory calculations illustrating an achievable precision in adatom-graphene distance of the order of 0.04 Å for In adatoms on graphene.The calculations were performed using the Wien2k code (L/APW + lo method, PBE+DFTd3 functional), with a 2×2 supercell (18×18×1 k-points).The muffin-tin radii were set to 1.29 and 1.70 atomic units, for the carbon atoms and adatom, respectively.The number of plane waves is set by the parameter R mt K max = 5.5.The z direction separates graphene sheets by 15 Å.The calculations do not include spin polarization.The experimental lattice parameter a=b=2.46Å was used, which is close to our optimized parameters with LDA (2.45 Å) and GGA (2.47 Å).The left axis corresponds to the calculated total energy in the cell in the vicinity of the equilibrium position (3.05Å).The right axis corresponds to the calculated V zz value.A typical precision of ±0.5×10 −21 V m −2 in V zz (for PAC experiments with 117 In) corresponds to a precision in adatom-graphene distance of approximately 0.04 Å.
occupied by the adatom relative to the graphene lattice (V zz and η) as well as its magnetic moment (B hf ) by comparison with ab initio density functional theory calculations.
Other challenges relate to the multitude of observables involved (structural, electric and magnetic) and the required experimental precision when measuring then.For example, an applied electric field can affect the magnetic moment of an 3d adatom by directly changing its charge state (i.e. the d orbital occupancy), but also by changing the distance between the adatom and the graphene plane (by changing the local crystal field).In other words, local magnetism and local structure are intrinsically linked, and a complete description of the adatom-graphene system requires a precise characterization of both.ASPIC is also unique in that sense, as it allows for the assignment of magnetic properties to a specific local structure configuration; and to measure both with the high precision inherent to hyperfine interactions.For example, figure 4 illustrates how the distance between an adatom and the graphene plane can be determined via the V zz with a precision down to 0.04 Å.The low detection limits achievable in ASPIC constitutes another remarkable strength for the study of adatoms on two-dimensional materials.Characteristic of nuclear techniques relying on radioactive probes, ASPIC can be used for measurements with less than 10 11 atoms, corresponding to a coverage of 0.02% for a typical 5×5 mm 2 sample.Finally, because the measurement is based on the detection of high-energy photons (high penetration depths and insensitive to electromagnetic deflection), ASPIC is generally compatible with experiments under varying temperature and applied electric and magnetic fields.
Once ASPIC is fully upgraded, a first set of experiments is foreseen with Cd, In and Hg as adatom elements, using 111m Cd(t 1/2 =48 m), 111 In(2.7d)/ 111 Cd, 117 Cd(2.5 h)/ 117 In and 199m Hg(42 m) as probe isotopes.These are well established PAC probes, and are therefore an ideal entry point for such challenging experiments on graphene adatoms using ASPIC.These studies will focus on determining adsorption sites and investigating electronic phenomena such as adatom-graphene hybridization effects and intra-atomic charge transfer.In the context of magnetism, since Cd (4d), In (4d) and Hg (5d) are closed-shell (d 10 ) metals, they are less likely to exhibit a non-vanishing electronic magnetic moment.Future work on magnetic adatoms may include adatom elements with open shells (3d, 4d, 5d and 4f) using, for example, 61 Ni (3d 8 ), 99 Ru (4d 7 ), 181 Ta (5d 3 ), 140 Ce (4f 1 ), and 143 Pr (4f 5 ).

Summary and conclusion
We reviewed the applications of the apparatus for surface physics and interfaces (ASPIC), at ISOLDE, CERN.
In the first part, we gave an overview of the research carried out since 2000, focusing on surface magnetism, interface polarization and adatoms on metallic surfaces.PAC spectroscopy was proven to be a very sensitive technique for local measurements of magnetic and structural properties within the nearest neighbor atomic shell.It was found that Pd layers can be magnetically polarized by a ferromagnetic underlayer up to a thickness of two monolayers.A systematic study of the dependence of the magnetic hyperfine field on the local coordination number on Ni surfaces has been presented.The quadrupole coupling constant, a sensitive measure for the position of the adatom, has been measured and systematically explained for Se probe atoms on various metallic surfaces at different temperatures.
In the second part, we gave an outlook into future research using ASPIC in the field of two-dimensional materials, namely for the investigation of structural and electronic properties of adatoms (adsorption sites, hybridization effects, intra-atomic charge transfer, magnetic moments, etc).Although we used the case of magnetic adatoms on graphene to illustrate the capabilities of this approach, we note that the applicability of ASPIC is much broader: (i) any two-dimensinal material; (ii) any adatom element with a suitable PAC isotope produced at ISOLDE; (iii) any research field in which the structural and electronic properties of adatoms play a role.This includes a wide range of scientific domains associated with two-dimensional materials, from nanoelectronics and catalysis to spintronics and topological quantum matter.

Figure 1 .
Figure 1.Coordination number dependence of B hf of Cd on Ni.The symbols refer to different measurements and calculations: (a) experimental results taken from Potzger et al (2002) and Potzger et al (2005) for surface sites and from Shirley et al (1968) for the substitutional lattice site (NN=12); (b) band structure calculations using the realspace Green's function embedding method (Mavropoulos 2003); (c) band structure calculations using the supercell method (Bellini et al 2004); (d) magnitudes and signs of magnetic hyperfine fields as determined in Manzhur et al (2007).The figure was reproduced with permission from Potzger et al (2002) and Manzhur et al (2007).© EDP Sciences, Società Italiana di Fisica, Springer-Verlag 2009, reproduced with permission from The European Physical Journal, copyright 2002 The American Physical Society.

Figure 2 .
Figure 2. Magnetic hyperfine field B hf versus the calculated total magnetic moment of the cluster of nearest neighbors.The moment m total was constructed by addition of the contributions from every atom belonging to the cluster formed by the nearest neighborhood of the probe atom, e.g., 0.89 μ B from Ni atoms at the Ni/Pd interface, 0.6 μ B from the Ni bulk, 0.24 μ B from the first Pd monolayer, and 0.20 μ B from the second Pd monolayer with respect to the interface.Reprinted figure with permission from Potzger et al (2005), copyright 2005 by the American Physical Society.

Figure 3 .
Figure 3. Temperature dependence of the quadrupole coupling constant ν Q of 77 Se on various surfaces.The solid lines are linear fits to the data on surfaces where the adatoms occupy threefold hollow sites.Dashed lines are fits to the data of adatoms on fourfold hollow sites.The data for Ni(001) and Pd (001) was taken from Granzer et al (1996).The bold solid line fits the data obtained for Pd(111) where only the electric hyperfine interaction is present.Reprinted figure with permission from Weber et al (2001), copyright 2001 by the American Physical Society.
al 2010)), for nanostructuring (e.g.adatom-mediated etching (Ramasse et al 2012)) or even to induce completely new phenomena (e.g.ferromagnetism (González-Herrero et al 2016), superconductivity (Profeta et al 2012), and topological phenomena (Qiao et al 2010, Ding et al 2011, Hu et al 2012, Zhang et al 2012) 2012) and adatoms (low-Z elements such as H and F, as well as 3d transition metals such as Fe, Co and Ni (Nair et al 2012, Elbo et al 2013, González-Herrero et al 2016)).Open-shell elements such as transition metals and rare earths are especially promising for spintronics applications (Uchoa et al 2008, Qiao et al 2010, Ding et al 2011, Hu et al 2012, Zhang et al 2012), in particular considering the ability to tune the electrical/magnetic behavior via an applied electric field (gate voltage) (Uchoa et al 2008, Zhang et al 2012).