Electronic structure and surface reconstruction of DyNi films grown on W(1 1 0)

Article abstract of DOI:10.1016/S0039-6028(01)00809-3

Angle-resolved photoemission and low-energy electron diffraction (LEED) were used to study thin films of DyNi(0 0 1) grown in situ under ultra-high vacuum conditions by codeposition of Dy and Ni onto a W(1 1 0) substrate and subsequent annealing. Guided by the results of electronic structure calculations an observed (3 × 1) overstructure of the LEED pattern and the appearance of a Dy 5d-derived surface state at the Fermi energy are explained by a reconstruction of the (0 0 1) surface resulting from missing Ni rows in the topmost atomic layer. Similar to metallic Gd(0 0 0 1) the d-like behavior of the obtained surface state may influence surface magnetic properties of thin films of DyNi.

Full text of DOI:10.1016/S0039-6028(01)00809-3

Surface Science 482±485 ꢀ2001) 746±751
www.elsevier.nl/locate/susc
Electronic structure and surface reconstruction of DyNi ®lms
grown on Wꢀ1 1 0)
a,
b
a
S.L. Molodtsov a,1,W. Schneider *,A.B. Preobrajenski ,J. Boysen ,
c
Manuel Richter ,T. Chass e ,C. Laubschat
a
ꢀ b,d
a
Institut fur Ober¯achenphysik und Mikrostrukturphysik, Technische Universitat Dresden, 01062 Dresden, Germany
b
W.-Ostwald-Institut fur Theoretische und Physikalische Chemie, Universitat Leipzig, Linnestr. 2, 04103 Leipzig, Germany
c
Abteilung fur Theoretische Festkorperphysik, IFW Dresden, 01171 Dresden, Germany
d
Institut fur Ober¯achenmodi®zierung, 04303 Leipzig, Germany
Abstract
Angle-resolved photoemission and low-energy electron di€raction ꢀLEED) were used to study thin ®lms of
DyNiꢀ0 0 1) grown in situ under ultra-high vacuum conditions by codeposition of Dy and Ni onto a Wꢀ1 1 0) substrate
and subsequent annealing. Guided by the results of electronic structure calculations an observed ꢀ3  1† overstructure
of the LEED pattern and the appearance of a Dy 5d-derived surface state at the Fermi energy are explained by a
reconstruction of the ꢀ0 0 1) surface resulting from missing Ni rows in the topmost atomic layer. Similar to metallic
Gdꢀ0 0 0 1) the d-like behavior of the obtained surface state may in¯uence surface magnetic properties of thin ®lms of
DyNi. Ó 2001 Elsevier Science B.V. All rights reserved.
Keywords: Metallic ®lms; Epitaxy; Surface electronic phenomena ꢀwork function,surface potential,surface states,etc.); Surface
relaxation and reconstruction
The investigation of the electronic structure
of thin ®lms and multilayers of rare-earth ꢀRE)
3d-transition-metal ꢀTM) compounds is of high
importance for a variety of technological applica-
tions like the development of ultra-high resolution
video systems or magnetic storage elements [1].
With decreasing thickness of the layers,surface
and interface e€ects become more and more sig-
ni®cant that may lead to strong deviations of the
®lm properties from the characteristics of the re-
spective bulk materials. Particularly large e€ects
may be expected for changes of the 4f occupancy
and the appearance of new electronic states at the
surfaces. Surface valence transitions have been
reported for several RE systems [2±6] that are
characterized by an increase of the 4f occupancy in
the outermost atomic surface layer and lead,e.g.,
in the case of trivalent Eu compounds to the for-
mation of a magnetic surface layer on top of a
diamagnetic bulk material [7]. Decrease of the 4f
occupancy,on the other hand,may occur at the
interfaces of REs with TMs. Valence band-derived
* Corresponding author. Tel.: +49-351-463-5647; fax: +49-
351-463-3457.
E-mail address: wolfs@physik.phy.tu-dresden.de ꢀW. Schne-
ider).
1
On leave from Institute of Physics,St. Petersburg State
University,198904 St. Petersburg,Russia.
0039-6028/01/$ - see front matter Ó 2001 Elsevier Science B.V. All rights reserved.
PII: S0039-6028ꢀ01)00809-3

S.L. Molodtsov et al. / Surface Science 482±485 92001) 746±751
747
non-dispersive surface states of d2z symmetry have
been found at the Fermi energy ꢀEF) at the close
packed surfaces of all trivalent RE metals [6,8±12].
This state is half ®lled and strongly spin polarized
in case of ferromagnetic Gd [12].
thicknesses of the deposited layers ꢀabout 100 A)
were monitored with quartz microbalances. Dur-
ing the evaporation the pressure was in the low
10À9 mbar range. ``As deposited'' polycrystalline
®lms were step-by-step annealed up to 500°C until
an intense rectangular LEED pattern with addi-
tional spots of weaker intensity,which point to a
ꢀ3  1† overstructure,was monitored. The rect-
angular LEED pattern may be attributed to the
simple orthorhombic BFe-type structure of the
DyNi crystal,which is common for RE±TM
compounds with 1:1 stoichiometry. The BFe-type
crystals consist of ``zig-zag'' coupled Dy and Ni
rows. Depending on the orientation of the crystals
di€erent kinds of rows with di€erent interrow
distances can be found in the topmost layer of the
samples. In the case of epitaxially grown thin ®lms
low-index surfaces are formed in most cases for
energetic reasons. Therefore,in the following we
consider only ꢀ0 0 1),ꢀ0 1 0),and ꢀ1 0 0) surfaces
revealing rectangular LEED patterns. To this end
we compare the tabulated [16] and LEED-derived
ratios of the two lattice parameters for each hy-
pothetical surface considered. The LEED pattern
obtained for the Wꢀ1 1 0) substrate was used for
absolute calibration of the DyNi data. By this
procedure a surface with ꢀ0 0 1) orientation was
concluded for the grown ®lms.
ꢁ
In the past spectroscopic investigations of the
surface electronic structure of RE±TM compounds
have mostly been concentrated on the properties
of polycrystalline samples,since ꢀi) single crys-
talline samples,which are necessary for angle-
resolved mapping of electronic bands,are dicult
to obtain and ꢀii) preparation of clean single
crystalline surfaces is not easy due to the high
chemical reactivity of the REs. Recently,a possi-
bility to grow thin ®lms of RE±TM compounds
in situ under ultra-high vacuum ꢀUHV) condi-
tions was reported [13±15]. The Ce±Rh,La±Rh and
Dy±Pd systems were prepared by codeposition of
Ce,La,Dy and Rh,Pd onto a Wꢀ1 1 0) substrate
with an atomic deposition ratio of 1/3 and subse-
quent annealing [13,14]. The resulting low energy
electron di€raction ꢀLEED) patterns revealed a
ꢀ2  2† overstructure with respect to the LEED
patterns of pure Rhꢀ1 1 1) and Pdꢀ1 1 1) surfaces
pointing to morphologies equivalent to the ꢀ1 1 1)
faces of the cubic ꢀRE)ꢀTM)3 phases ꢀAuCu3-
type). Changing the atomic deposition ratio to
1=1,the same method was successfully used to
synthesize thin ®lms of DyPdꢀ0 0 1) compound
ꢀCsI-type) [15].
The experiments were performed at the Berliner

Elektronenspeicherring fur Synchrotronstrahlung
In the present study thin ®lms of DyNiꢀ0 0 1)
were grown in situ under UHV conditions by co-
deposition of Dy and Ni in the stoichiometric
amount onto a Wꢀ1 1 0) substrate and measured by
angle-resolved photoemission ꢀPE) and LEED.
The obtained data are in good agreement with
results of band-structure calculations. An observed
ꢀ3  1† overstructure of the LEED pattern and the
appearance of a Dy 5d-derived surface state near
EF are assigned to a reconstruction of the ꢀ0 0 1)
surface resulting from missing Ni rows in the
topmost atomic layer. Similar to the case of a pure
Gdꢀ0 0 0 1) surface,presence of this surface state
may strongly in¯uence the surface magnetic
properties of thin DyNi ®lms.
ꢀBESSY I) using the radiation from the TGM 7
beamline. Angle-resolved PE spectra were taken
using a rotatable hemispherical electron-energy
analyzer ꢀVSW-ARIES) tuned to an energy reso-
lution of about 150 meV ꢀFWHM) and an angular
resolution of 1°. The angle of the photon incidence
was selected to be 30° relative to the sample sur-
face. The basic pressure in the experimental set-up
was better than 1 Â 10À10 mbar. The PE spectra
obtained for the epitaxially grown ®lms were
compared to data taken for single-phase poly-
crystalline bulk compounds with DyNi composi-
tion. The latter reference samples were prepared by
arc alloying in Ar atmosphere of the Dy and Ni
metallic pieces carefully weighted according to the
required chemical composition. The surfaces of the
reference samples were cleaned by scraping under
UHV conditions.
Dysprosium and nickel were thermally depos-
ꢁ
ited with rates of about 3±5 A= min onto a clean
Wꢀ1 1 0) substrate kept at room temperature. The

748
S.L. Molodtsov et al. / Surface Science 482±485 92001) 746±751
The obtained PE data were ®rst analyzed on the
basis of local-density-approximation ꢀLDA) lin-
ear-combination-of-atomic-orbitals ꢀLCAO) cal-
culations of the electronic structure of DyNi and
YNi,which were carried out with the standard
Hedin±Lundquist exchange-correlation potential
[17]. Thereby the 4f states of Dy were considered
to be fully localized and,therefore,were not in-
cluded in the basis set of valence-band wave
functions. Electronic structure at the surface of the
samples was studied by means of slab calculations
ꢀCASTEP package code,Molecular Simulations
Inc.) using a total-energy-pseudopotential method
within the density-functional theory. For the elec-
tron exchange and correlation,LDA functional
in the form of Ceperley and Adler [18] has been
used. The supercell calculations were performed
for two di€erent atomic arrangements at the
DyNiꢀ0 0 1) and YNiꢀ0 0 1) surfaces: ꢀi) for ideal
surfaces ꢀtype I) and ꢀii) for surfaces with a ꢀ3  1†
reconstruction due to one missing Ni row per three
surface unit cells ꢀtype II). The supercell of type I
has been made up from eight monolayers of DyNi
ꢀYNi) and four monolayers of vacuum. For the
type II supercell six monolayers of YNi and four
monolayers of vacuum have been used.
Fig. 1. Angle-resolved PE spectra of DyNi/Wꢀ1 1 0) taken ꢀa) in
normal emission geometry for di€erent hm and ꢀb) with
hm ˆ 35 eV at di€erent H in the direction of the surface re-
construction. Surface and bulk components of the Dy 4f signal
are marked by s and b,respectively.
Angle-resolved PE spectra taken ꢀa) with vari-
ous photon energies in normal emission geometry
and ꢀb) with surface sensitive photon energy
hm ˆ 35 eV at di€erent polar energies ꢀHs) and an
azimuthal angle in the direction of the LEED re-
construction are shown in Fig. 1. The presented
spectra reveal four main structures located at the
Fermi energy,within the ®rst 1 eV below EF,and
in the regions of approximately 2 eV and 4±5 eV
binding energies ꢀBEs). The intense structure
measured at ꢁ2 eV BE stems from Ni 3d electronic
states forming rather narrow band. The small
width of this band with respect to the one in Ni
metal can be explained by the reduced number of
nearest Ni neighbors in the compound. In contrast
to Ni metal the Ni 3d-derived states in DyNi are
shifted away from EF that can be understood by a
®lling of the Ni 3d band due to a charge transfer
from Dy atoms. The Ni 3d emission reveals dis-
persion both as a function of hm and H. Addi-
tionally,other dispersive PE features of low
intensity are seen between 2.5 and 3.5 eV BEs in
Fig. 1ꢀb) that may be assigned to bands with other
than Ni 3d character.
The structure monitored in the region of 4±5 eV
BEs can mainly be assigned to a part of the Dy 4f
®nal-state PE multiplet,although weak contribu-
tion of Dy 6s states,can also not be excluded. The
observed suppression of this structure at hm < 40
eV is due to the drastic decrease of the cross sec-
tion of 4f states at low energies of photoexcitation
[19]. The doublet lineshape of the Dy 4f signal,that
is clearly seen in the more surface sensitive spectra
taken with low photon energies,is caused by
emissions from bulk and surface Dy atoms,re-
spectively. Thereby,as expected for metals with
less than half ®lled valence bands [20],the surface
contribution is shifted to higher BEs relative to the
bulk signal. While the bulk component in the
present case is found at about the same BE posi-
tion as in Dy metal,the observed surface core-level
shift of 0.5 eV is larger than the value of 0.3 eV

S.L. Molodtsov et al. / Surface Science 482±485 92001) 746±751
749
obtained for the close-packed Dyꢀ0 0 0 1) surface
[21]. Since the surface core-level shift relates to the
di€erent atomic coordination in the bulk and at
the surface of solid,the above di€erence could
point to lower coordinated Dy atoms at the sur-
face of the grown ®lms with respect to Dyꢀ0 0 0 1).
According to results of the band-structure cal-
culations performed in the present study,5d elec-
tronic states of Dy are expected within the ®rst
1 eV below the Fermi level. Experimentally,two
spectral features are observed in this region,from
which a sharp non-dispersive Fermi-level peak
reveals particularly high intensity around the C
points in the Brillouin zone,which are probed in
the present experiment at H ˆ 0° and 34°. These
two spectral features show highest intensity be-
tween 30 and 45 eV photon energies,while at
higher hm the signals become strongly reduced.
This behavior resembles the one of the 5d emis-
sions of pure RE metals [22]. As in the present
case,PE spectra of close-packed surfaces of the
latter are characterized by a similar double peaked
structure. There,the Fermi-energy peak is due to a
non-dispersive surface state of d2z symmetry,while
the component at higher BE represents a 5d-
derived bulk emission [6,8,22]. An interpretation
of the Fermi-energy feature of the DyNi com-
pounds in terms of a similar surface state is sug-
gested by the fact that for the polycrystalline
sample it shows much lower intensity,while the
other spectral structures look similar to the ones
measured for the ordered ®lm ꢀFig. 2ꢀa) and ꢀb)).
Further support in favor of the surface-state
origin of this Fermi-energy feature is provided by
results of our LDA±LCAO calculations,which
reveal for the DyNiꢀ0 0 1) surface an energy gap
only in the region of the C points of the projected
bulk band structure in analogy to the case of the
RE metals. In contrast to the spectra taken at the
polar angles close to 0° and 34°,the Fermi-energy
PE intensity measured for other regions in the
Brillouin zone is caused by the bulk electronic
bands. In Fig. 2,PE spectra of both the ordered
®lm and a polycrystalline sample are compared to
the calculated bulk electronic density of states
ꢀDOS). Excluding the EF region there is a fair
agreement between the theoretical and experi-
mental results. In particular,the total width of the
Fig. 2. Normal emission PE spectra of ꢀa) the epitaxial ®lm and
ꢀb) a polycrystalline reference sample of DyNi ꢀsolid lines
through experimental points) in comparison to the DOS ꢀsolid
lines) calculated within the LDA±LCAO approach for DyNi ꢀc)
and the slab approach for ꢀd) DyNi [type-I supercell],ꢀe) YNi
[type-I supercell],and ꢀf) YNi [type-II supercell]. DOS close to
EF are shown by shadowed areas.
Ni 3d-derived PE bands is well reproduced by the
calculations.
To obtain information about surface e€ects on
the electronic structure of DyNiꢀ0 0 1) supercell
calculations were carried out. Note that our su-
percell approach is restricted to calculations of
angular-momentum-character integrated DOS and
charge-density distributions including contribu-
tions of the Dy 4f states. Fig. 2ꢀd) presents the
DOS obtained by the supercell calculations for an
ideal DyNiꢀ0 0 1) surface. In the region of the Ni
3d states the obtained results are in a good
agreement with the DyNi DOS calculated within
the bulk LDA±LCAO approach. The main dif-
ference between the results of these two calcula-
tions is a Dy 4f-derived doublet ꢀdashed line),
which appears at EF in the case of the super-
cell calculations. This 4f contribution,that is not

750
S.L. Molodtsov et al. / Surface Science 482±485 92001) 746±751
observed in the experimental data due to the
strongly correlated character of the 4f electrons,
appears just in the energy region of the expected
surface state making a clear identi®cation of the
latter impossible. By this reason,Dy atoms were
substituted by Y atoms in further consideration.
Yttrium is isoelectronic to dysprosium ꢀexcept the
4f states) providing a similar type of chemical
bonding. Due to their resembling atomic sizes,Dy
and Y form isostructural intermetallic compounds
with nearly the same lattice constants.
A calculated slab DOS for a YNiꢀ0 0 1) ®lm with
ideal surface geometry is shown in Fig. 2ꢀe). As
expected the obtained results are very similar to
the ones for DyNiꢀ0 0 1) except the Dy 4f-derived
structure. For the non-reconstructed YNiꢀ0 0 1)
surface no surface-related electronic state,which
could be associated with the experimentally ob-
served peak in the vicinity of the Fermi energy,is
found by the calculations. On the other hand,the
observed ꢀ3  1† reconstruction of the LEED
pattern points to structural changes at the ꢀ0 0 1)
surface,that may heavily in¯uence the surface
electronic structure. The ꢀ3  1† reconstruction of
the ꢀ0 0 1) surface can easily be understood as-
suming that each third row,either Dy or Ni,in the
topmost layer is missing. From the similarities of
the Fermi-energy feature in Fig. 1 and the surface
state observed at the close packed surfaces of the
RE metals one should expect that a model with
missing Ni rows is more favorable to describe the
results of the PE experiment.
The DOS calculated for such a surface recon-
struction is shown in the bottom of Fig. 2. It de-
viates from the results for the ideal surface in
several points: A more pronounced peak is now
observed directly at EF and in the region of the
Ni 3d-derived band the agreement with the ex-
perimental data is clearly improved. Comparing
spectrum ꢀa) with curve ꢀf) in Fig. 2 one has to
consider,however,that the angle-resolved PE
spectrum samples electronic states of a given kll
value,while curve ꢀf) represents a DOS integrated
over all k and all layers of the slab. Deviations in
shape are,therefore,natural here. Particularly,the
strong intensity di€erence of the Fermi-energy
peak in theory and experiment is caused by the fact
that this state is restricted to a narrow region
Fig. 3. Charge-density distribution for the Fermi-energy elec-
trons of YNi calculated within the slab approach for the re-
constructed ꢀ0 0 1) surface with missing Ni rows. The upper
panel represents a side view,the lower a top view of the slab.
Hatched and open circles mark the position of Y and Ni atoms,
respectively; shadowed clouds represent contours of constant
charge density. Crossed out circles denote missing Ni atoms.
around the C point of the outermost surface layer
and contributes only weakly into the DOS.
The charge-density distribution ꢀtype-II super-
cell) related to the Fermi-energy peak is shown in
Fig. 3. Here,the positions of Y and Ni atoms are
indicated by large hatched and small open circles,
respectively,and the charge density is visualized by
shadowed clouds. The ®gure shows that the states
corresponding to the Fermi-energy peak are lo-
cated in the surface area of the ®lm concentrating
around the Y atoms close to Ni vacancies ꢀcrossed
out open circles). The latter fact strengthens the
above hypothesis on the similar origin of the sur-
face state at the close packed surfaces of REs and
DyꢀY)Niꢀ0 0 1). A lack of Ni atoms as compared
to the stoichiometric value is obviously a condition
to obtain the surface state for the intermetallic
®lms under consideration.
In summary we have studied ordered ®lms of
DyNi grown in situ under UHV conditions on a
Wꢀ1 1 0) surface. The obtained LEED and PE data
evident a ꢀ0 0 1) orientation of the DyNi surface.
An observed peak at the Fermi energy is assigned
to a Dy 5d-derived surface state and related to a
ꢀ3  1† surface reconstruction resulting from
missing Ni rows in the topmost atomic layer. A

S.L. Molodtsov et al. / Surface Science 482±485 92001) 746±751
[8] R. Wu,C. Li,A.J. Freeman,C.L. Fu,Phys. Rev. B 44
ꢀ1991) 9400.
751
similar surface state was observed for the close
packed surfaces of pure RE metals.
[9] B. Kim,A.B. Andrews,J.L. Erskine,K.J. Kim,B.N.
Harmon,Phys. Rev. Lett. 68 ꢀ1992) 1931.
[10] A.V. Fedorov,C. Laubschat,K. Strake,E. Weschke,
K.-U. Barholz,G. Kaindl,Phys. Rev. Lett. 70 ꢀ1993) 1719.
Acknowledgements

ꢀ

[11] G. Kaindl,A. H ohr,E. Weschke,S. Vandr e,C. Sch uûler-
Langeheine,C. Laubschat,Phys. Rev. B 51 ꢀ1995) 7920.
[12] G.A. Mulhollan,K. Garrison,J.L. Erskine,Phys. Rev.
Lett. 69 ꢀ1992) 3240.
This work was supported by the DFG,Son-
derforschungsbereich 463,TPB4 and TPB11; and
by the BMBF,project 05-OD1SFB/4. Expert help
by the sta€ of BESSY is gratefully acknowledged.
[13] W. Schneider,S.L. Molodtsov,Th. Gantz,P. Engelmann,
C. Laubschat,M. Richter,Phys. Rev. B 57 ꢀ1998) 14930.
[14] J. Boysen,P. Segovia,S.L. Molodtsov,S.A. Gorovikov,
M. Richter,C. Laubschat,J. Alloys Comp. 275±277 ꢀ1998)
493.
References
[15] W. Schneider,Th. Gantz,S.L. Molodtsov,J. Boysen,C.
Laubschat,in preparation.
[16] Binary Alloy Phase Diagrams,in: T.B. Massalski ꢀEd.),
ASM International,Materials Park,OH,1990.
[17] L. Hedin,B.I. Lundqvist,J. Phys. C 4 ꢀ1971) 2064.
[18] D.M. Ceperley,B.J. Alder,Phys. Rev. Lett. 45 ꢀ1980) 566.
[19] J.J. Yeh,I. Lindau,Atom. Data Nucl. Data Tables 32
ꢀ1985) 1.
[1] D. Weller,W. Reim,Appl. Phys. A. 49 ꢀ1989) 599.
[2] G.K. Wertheim,G. Crecelius,Phys. Rev. B 40 ꢀ1978) 813.
[3] C. Laubschat,G. Kaindl,W.-D. Schneider,B. Reihl,N.
ꢁ
Martensson,Phys. Rev. B 33 ꢀ1986) 6675.
[4] M. Domke,C. Laubschat,M. Prietsch,T. Mandel,G.
Kaindl,W.D. Schneider,Phys. Rev. Lett. 56 ꢀ1986) 1287.
[5] C. Laubschat,E. Weschke,C. Holtz,M. Domke,O.
Strebel,G. Kaindl,Phys. Rev. Lett. 65 ꢀ1990) 1639.

ohr,K. Starke,L.
[20] E. Weschke,C. Laubschat,A.
H
Baumgarten,A.V. Fedorov,G. Kaindl,J. Electron Spec-
trosc. Relat. Phenomena 68 ꢀ1994) 515.

[6] E. Weschke,A. H ohr,G. Kaindl,S.L. Molodtsov,S.

ꢀ


Danzenbacher,M. Richter,C. Laubschat,Phys. Rev. B 58
ꢀ1998) 3682.
[21] G. Kaindl,A. H ohr,E. Weschke,S. Vandr e,C. Sch uûler-
Langeheine,C. Laubschat,Phys. Rev. B 51 ꢀ1995) 7920.

[22] M. Bodenbach,A. H ohr,C. Laubschat,G. Kaindl,M.
Methfessel,Phys. Rev. B 50 ꢀ1994) 14446.
[7] S. Wieling,S.L. Molodtsov,Th. Gantz,J.J. Hinarejos,C.
Laubschat,Phys. Rev. B 58 ꢀ1998) 13219.