Creation of Nanohillocks on CaF2Surfaces by Single Slow Highly Charged Ions
A.S. El-Said,1,*R. Heller,2W. Meissl,1R. Ritter,1S. Facsko,2C. Lemell,3B. Solleder,3I.C. Gebeshuber,1G. Betz,1
M. Toulemonde,4W. Mo ¨ller,2J. Burgdo ¨rfer,3and F. Aumayr1,†
1Institut fu ¨r Allgemeine Physik, Vienna University of Technology, 1040 Vienna, Austria, EU
2Forschungszentrum Dresden-Rossendorf, D-01328 Dresden, Germany, EU
3Institute for Theoretical Physics, Vienna University of Technology, 1040 Vienna, Austria, EU
4CIMAP, ENSICAEN, CEA, CNRS, University of Caen, 14070 Caen, France, EU
(Received 20 December 2007; published 10 June 2008)
Upon impact on a solid surface, the potential energy stored in slow highly charged ions is primarily
deposited into the electronic system of the target. By decelerating the projectile ions to kinetic energies as
low as 150 ? q eV, we find first unambiguous experimental evidence that potential energy alone is
sufficient to cause permanent nanosized hillocks on the (111) surface of a CaF2single crystal. Our
investigations reveal a surprisingly sharp and well-defined threshold of potential energy for hillock
formation which can be linked to a solid-liquid phase transition.
DOI: 10.1103/PhysRevLett.100.237601PACS numbers: 79.20.Rf, 34.35.+a, 61.80.Jh
High-energy photons, electrons, and ions are frequently
employed as tools to lithographically modify surfaces on
the nanometer scale. Among them, ion beams are, perhaps,
still the least developed technique in the field of nano-
lithography, in part due to the fact that many fundamental
aspects of their interactions with the surface are not yet
well understood. Three techniques, focused ion beam,
proton beam writing, and ion projection lithography,
have now breached the technologically difficult 100 nm
barrier and are capable of fabricating structures on the
nanoscale . Current research in the field of advanced
ion-beam techniques focuses on the energy and charge
state dependence of the primary beam. While the interac-
tion of intense beams of keV ions with surfaces can result
in well-ordered patterns, such as ripples or self-ordered
dots [2–4], drastic modifications to the surface topography
by individual ions are induced only if the material is
exposed to energetic ions (MeV to GeV region). At these
high impact energies, the energy deposition leads to the
creation of nanosized hillocks or craters randomly distrib-
uted on the surface. Recently, the formation of multiple,
regularly spaced nanodots on SrTiO3surfaces has been
demonstrated for single swift Xe ion impact under grazing
angles of incidence .
One major limitation for the application of swift heavy
ions to three-dimensional structure formation is the radia-
tion damage of deeper layers. The desire to confine the
energy deposition to the surface layer has stimulated the
interest in slow (eV to keV) and highly charged ions. With
increasing charge state, these ions carry an increasingly
large amount of potential energy (e.g., 14 keV for bare
Ar18?and 51 keV for Ne-like Xe44?). The potential-
energy deposition and electron depletion induced by the
neutralization sequence of slow highly charged ions (HCI)
on insulator targets is expected to be confined to a
nanometer-sized volume close to the surface and to occur
on a femtosecond time scale . Intuitively, one expects
the formation of nanosized surface craters due to Coulomb
explosion. Incontrast to theseexpectations, the firstexperi-
ments with 7q keV HCI (q denoting their incident charge
state) on Muscovite mica showed nanohillocks [7,8] when
the interaction zone was inspected by atomic force micros-
copy (AFM). For ion charge states below q ? 30, no
damage could be identified. Systematic experiments on
mica revealed that the observed structures do not represent
topographic changes of the surface but rather changes in
the surface friction leading to the observation of ‘‘hil-
locks’’ and sometimes ‘‘craters’’ depending on the scan-
ning direction of the AFM . Furthermore, the observed
structures disappeared after repeated scanning with AFM.
The volume of the ‘‘apparent’’ structures was found to be
roughly proportional to the potential energy of the imping-
ing ions [8,9] while only weakly dependent on the projec-
tile kinetic energy . Studies on other surfaces (such as
highly oriented pyrolytic graphite) found similar results
(for a recent review of this field, see ). However, due to
the high kinetic energy of the projectile ions (up to several
hundred keVas a result of 5–10 kVacceleration voltage),
kinetic effects (e.g., contributions from kinetic energy
transfer to the target cores or electrons) could not be ruled
In this Letter, we present experiments with very slow
(down to vp? 0:03 a:u: or 30 eV=amu) HCI creating
hillocklike topographic nanostructures on the surface of
CaF2single crystals which are stable in air and nonerasable
by AFM scanning. Surprisingly, these nanostructures
closely resemble those created by swift heavy ions at the
surface  while leaving deeper layers of the target
undamaged. Moreover, we find a strong dependence of
the formation on the potential energy rather than on the
kinetic energy with a sharp and well-defined threshold of
potential energy required for the onset of nanohillock
formation. Simulations of the dissipation of potential en-
ergy into the target material on the basis of an extended
PRL 100, 237601 (2008)
13 JUNE 2008
© 2008 The American Physical Society
classical over-the-barrier model have been performed to
facilitate the interpretation of the experimental findings.
Since CaF2is used as an insulator in silicon microelec-
tronic devices [12,13] epitaxially grown on semiconductor
surfaces , our findings might be of importance for high
resolution patterning of thin CaF2films on Si and for the
creation of nanostructured templates for adlayer growth
during fabrication of CaF2=Si-based epitaxial insulator-
Our experiments were performed on air-cleaved
CaF2?111? single crystal surfaces. Cleavage is known to
result in a fluorine-terminated surfacewhich is stable in air.
Contact-mode AFM in UHV revealed large atomically flat
surfaces with occasional cleavage steps separating individ-
ual terraces. Irradiation of CaF2samples (freshly cleaved
before their transfer into a vacuum chamber of pressure in
the 10?10mbar range) took place at the ion-beam center of
129Xeq?ions (q ? 24–36) were extracted from a Dresden
electron beam ion trap source  and decelerated by a
two-stage deceleration system to the desired final impact
energy before impinging onto the single crystal CaF2
surface under a normal angle of incidence. Deceleration
to a final potential difference between source and target
down to 150 V resulted in the lowest impact energies of
150q eV (150 V times projectile charge q), i.e., an impact
energy of only 28–42 eV per atomic mass unit. The time-
averaged beam flux varied between 104and 106ions=s.
After exposure to fluences of about 1010ions=cm2, the
crystal was transferred to an UHV-AFM/STM (Omicron)
and inspected by contact-mode AFM.
A typical AFM topographic image of a CaF2?111? sur-
face (Fig. 1) after irradiation with 2q keV Xe33?ions
(?500 eV=amu) displays hillocklike nanostructures pro-
truding from the surface. The AFM images were evaluated
in terms of their areal density, height, and width distribu-
tions of the hillocks. The hillocks in Fig. 1 are typically
20 nm in diameter and 0.8 nm in height. Because of the
finite radius of curvature of the AFM tip (nominally 7–
10 nm), the diameter of the hillocks is subject to a system-
atic error . Measurements of heights of structures,
however, are known to be reasonably accurate. From the
number of hillocks per unit area and the applied ion
fluence, we determine that a vast majority of projectiles
(about 80% ? 10%) produce one hillock each.
To demonstrate that the hillocks are solely due to the
deposition of potential rather than kinetic energy (in the
form of nuclear or electronic stopping) of the projectiles,
we decelerated the Xe33?ions to final impact energies as
low as 38 eV=amu (150q eV). In Fig. 2, we show the
measured mean volume of the hillocklike nanostructures
on CaF2produced by the impact of Xe33?projectile ions as
a function of their kinetic energy together with previous
results for much more energetic (10q keV) Xe33?ions
. Despite the reduction of the kinetic energy by almost
2 orders of magnitude, the measured hillock volume is
essentially unaffected and stays almost constant. The data
might even indicate a slight increase with decreasing ki-
netic energy, a trend that was also found for other charge
states (cf. Fig. 3).
In order to explore the dependence on the potential
energy of the projectiles, we employ Xeq?ions with
charge states ranging from q ? 24 to q ? 36while leaving
the potential difference between the ion source and target
surface at a constant value of 150 V. The hillock volume
was found to be strongly dependent on the potential energy
of the projectiles (Fig. 3). A remarkably well-defined sharp
threshold in potential energy (between 10.4 keV for Xe27?
and 12.0 keV for Xe28?) for hillock formation emerges.
Above this threshold, an increase of the potential energy
leads to a strong increase of hillock volume. The hillock
volume seems to increase slightly for decreasing kinetic
energy, and the threshold shifts by about 2 keV. Repeated
measurements confirmed that hillocks are produced by
slow (150q eV) Xe28?but not for fast (10q keV) Xe28?
projectiles. While above the threshold the hillock volume
increases linearly with potential energy, the shape of the
FIG. 1 (color online).
of a CaF2?111? surface after irradiation with 2q keV Xe33?ions
showing hillocklike nanostructures protruding from the surface.
Topographic contact-mode AFM image
FIG. 2 (color online).
tures on CaF2produced by the impact of Xe33?projectile ions as
a function of their kinetic energy. Solid symbols: Present results;
open symbol: result from Ref. .
Mean volume of hillocklike nanostruc-
PRL 100, 237601 (2008)
13 JUNE 2008
hillocks doesnotdepend onbeam parameters; alsothe base
diameter shows only a small dependence on the potential
energy of the projectile.
Surprisingly, the hillocks observed in our experiment
resemble the surface structures generated by swift heavy
ions. For the latter, hillock formation was observed above
an energy loss of 5 keV=nm  which results from
500 eV=nm nuclear stopping Snand 4:5 keV=nm elec-
tronic stopping Se. While Sncorresponds to direct
transfer of kinetic energy into lattice heating, i.e., excita-
tion of phonons along the projectile trajectory, Seis a
measure for the excitation of the electronic subsystem of
the target. It acts as a precursor for lattice excitation via
electron-phonon coupling . By contrast, for very slow
highly charged ions (kinetic energy of about 5 keV for
150q eV Xe33?), the total stopping power amounts to only
1:3 keV=nm with a less than 5% contribution from Se.
Deposition of the potential energy (Epot) of the highly
charged projectile, 75% of which is stored in the target
material , must therefore play a decisive role for slow
HCI.However,asthe potential-energythreshold wasfound
around 12 keV (see Fig. 3), it is obvious that only part of
Epotis effectively converted into lattice excitations.
In the following, we present simulations of the energy
transfer from the HCI to the lattice of the CaF2target
combining above and below surface electron emission
processes along the projectile trajectory  with electron
transport within the target material including the genera-
tion of secondary electrons and heating of the crystal
lattice . This sequence involves a twofold conversion
of energy: First, potential energy is converted into kinetic
energy of emitted primary electrons. In turn, electrons
deposit their energy in the crystal as heat, eventually lead-
ing to melting of the material.
Highly charged ions approaching solid surfaces undergo
a largenumber ofneutralization anddeexcitation processes
which are well described within the classical-over-barrier
model developed for metal surfaces  and its extension
for insulator targets . Electrons from the target are
transferredinto highlyexcited states oftheprojectilewhich
may decay by collisional, radiative, and Auger processes.
Transfer of electrons to the projectile leaves unbalanced
holes (F0atoms) in the surface which store part of the
potential energy carried into the collision. Upon impact of
the projectile, the target is structurally weakened.
Projectiles reach the surface far from ground state as the
time spent in front of the surface is not sufficient for a
complete relaxation. At this stage, electrons are captured
into moderately excited states by either resonant charge
transfer from the valence band or Auger neutralization
processes followed by an Auger deexcitation sequence.
Along this sequence, electrons with low to intermediate
energies up to a few hundred eVare emitted. If inner-shell
holes are to be filled (e.g., in the cases of Ar17?and Ar18?
), electrons with keV kinetic energies are released. The
potential energy stored in the incoming HCI will be de-
posited along the first few nanometers of its trajectory
below the target surface. The kinetic energy of the projec-
tile determines the depth within which the neutralization is
completed (?1 nm for 150q eV and ?4 nm for 10q keV
projectiles; see Fig. 4). It is much smaller than the total
range of the ion in the solid (?6 nm for 150q eV and
?90 nm for 10q keV projectiles ).
For an HCI with q ? 40, we estimate about 250 unbal-
anced holescreated in thecourseoftheinteraction ofa
single ion affecting the crystal structure of the target. In our
electron-transport simulation, elastic and inelastic scatter-
FIG. 4 (color online).
(a),(c) and Xe33?(b),(d) projectile ions in a CaF2 crystal.
Calculations were performed for impact energies of 150q eV
(a),(b) and 10q keV (c),(d). For details, cf. the text.
Energy density deposited by Xe28?
FIG. 3 (color online).
tures as a function of the potential energy of Xeq?projectiles.
Solid symbols correspond to measurements taken at 150q eV
impact energies, while open symbols show the results taken for
10q keV. Hillocks are found only above a potential-energy
threshold which slightly shifts with kinetic energy. Lines are
drawn to guide the eye. The error bars represent the statistical
variation of the actual hillock volume and are not due to limited
resolution of our AFM.
Mean volume of hillocklike nanostruc-
PRL 100, 237601 (2008)
13 JUNE 2008
ing processesare taken into account, leading to the creation
of secondary electrons (whose trajectories are followed as
well) and to excitations of phonons in the interaction with
the crystal atoms. Energy transfer to the lattice will even-
tually lead to heating and melting of the crystal. As a
consequence of the high-energy density required for melt-
ing of a CaF2crystal (?0:55 eV=atom), low-energy elec-
trons contribute more efficiently to the melting process
than high-energy electrons, which distribute their energy
over a much larger volume because of their larger inelastic
and elastic mean free paths . Contrary to naive expec-
tations and quite surprisingly, the decisive difference be-
tween below and above threshold charge states is not the
additional fast Auger electron but the many additional slow
electrons emitted along the deexcitation sequence resulting
from the filling of the additional inner-shell hole.
The average energy density deposited in the target along
the trajectory as a function of the distance from the pro-
jectile track features a ‘‘hot’’ core (bright yellow region in
Fig. 4) in which the critical energy density required for
melting is reached. The shape of this volume strongly
depends on the velocity of the projectile. While for slow
projectiles the volume is almost hemispherical, fast pro-
jectiles create an elongated volume resembling the shape
of a candle flame. If either the velocity is increased or the
potential energy is reduced (smaller initial charge states),
the diameter of the heated volume shrinks. While the
present electron-transport simulation assuming a structure-
less medium cannot account for effects of the crystalline
structure, important information on the spatial distribution
of energy deposition into the electronic degrees offreedom
preceding structure modification and melting can be in-
ferred: A minimum volume heated above the threshold
energy density of 0:55 eV=atom is needed for restructuring
and hillock formation. The core volume in Fig. 4(a) is
found to be about 2:5 nm3or, equivalently, about 15 unit
cells of CaF2(lattice constant of a ? 5:462?A) containing
about 102atoms. Equally important is the (smallest) linear
dimension of the hot core. Only if the diameter of the core
exceeds the size of the unit cell can the above-critical
energy density be retained for a sufficiently long time
such that the relatively slow processes of restructuring
and melting occur before cooling sets in. Hillock formation
was experimentally observed for all cases displayed in
Fig. 4 except for Fig. 4(c) (Xe28?, 10q keV) in which the
diameter of the core region is reduced to about the lattice
constant. In this case, the deposited energy is apparently
dissipated too quickly, and the melting process is sup-
pressed. Clearly, future simulations must be extended
from the present multifemtosecond to the multipicosecond
scale by employing molecular-dynamics techniques to
quantify the melting process in more detail.
In conclusion, we have shown for the first time that the
potential energy of highly charged ions can be exclusively
responsible for the production of permanent nanosized
hillocks on insulating CaF2single crystals. Accompany-
ing simulations of the energy density deposited on the
target atoms suggest a link of observable surface modifi-
cations to a solid-liquid phase transition. They are also able
to qualitatively explain the existence and shift of the
threshold charge state for hillock formation observed for
projectiles with different kinetic energies.
This work has been supported by Austrian Science
Foundation FWF (Projects No. 17449 and No. M894-
N02) and by the European Project No. RII3#026015.
Transnational access to the Rossendorf ion-beam facilities
was provided through AIM (EU Contract No. 025646).
University, 35516 Mansoura, Egypt.
†To whom correspondence should be addressed.
 F. Watt et al., Int. J. Nanosci. 4, 269 (2005).
 S. Facsko et al., Science 285, 1551 (1999).
 M. Castro et al., Phys. Rev. Lett. 94, 016102 (2005).
 B. Ziberi et al., Phys. Rev. B 72, 235310 (2005).
 E. Akcoltekin et al., Nature Nanotechnology 2, 290
 A. Arnau et al., Surf. Sci. Rep. 27, 113 (1997).
 D.H. Schneider et al., Radiat. Eff. Defects Solids 127, 113
(1993); D. Schneider et al., Surf. Sci. 294, 403 (1993).
 C. Ruehlicke, M.A. Briere, and D.H. Schneider, Nucl.
Instrum. Methods Phys. Res., Sect. B 99, 528 (1995).
 D.C. Parks et al., J. Vac. Sci. Technol. B 13, 941 (1995).
 F. Aumayr, A.S. El-Said, and W. Meissl, Nucl. Instrum.
Methods Phys. Res., Sect. B (in press).
 N. Khalfaoui et al., Nucl. Instrum. Methods Phys. Res.,
Sect. B 240, 819 (2005).
 T.P. Smith et al., Appl. Phys. Lett. 45, 907 (1984).
 L.J. Schowalter and R.W. Fathauer, J. Vac. Sci. Technol.
4, 1026 (1986).
 C.A. Lucas and D. Loretto, Appl. Phys. Lett. 60, 2071
 F. Grossmann et al., Nucl. Instrum. Methods Phys. Res.,
Sect. B 256, 565 (2007).
 A.S. El-Said et al., Nucl. Instrum. Methods Phys. Res.,
Sect. B 258, 167 (2007).
 J.F. Ziegler, J.P. Biersack, and U. Littmark, The Stopping
and Range of Ions in Solids (Pergamon Press, New York,
 M. Toulemonde, C. Dufour, and E. Paumier, Phys. Rev. B
46, 14362 (1992); A. Meftah et al., Nucl. Instrum.
Methods Phys. Res., Sect. B 237, 563 (2005).
 D. Kost et al., Phys. Rev. Lett. 98, 225503 (2007).
 L. Wirtz et al., Phys. Rev. A 67, 012903 (2003).
 C. Lemell et al., Solid-State Electron. 51, 1398 (2007).
 J. Burgdo ¨rfer, P. Lerner, and F.W. Meyer, Phys. Rev. A 44,
 W. Meissl et al., e-J. Surf. Sci. Nanotech. 6, 54 (2008).
PRL 100, 237601 (2008)
13 JUNE 2008