oxygen structures on fe(1 1 0

10
Oxygen structures on Fe(1 1 0) J. Weissenrieder * , M. Gothelid, M. M ansson, H. von Schenck, O. Tjernberg, U.O. Karlsson Laboratory of Materials and Semiconductor Physics, Royal Institute of Technology, Electrum 229, SE-164 40 Kista, Sweden Received 11 September 2002; accepted for publication 7 January 2003 Abstract The adsorption of oxygen on a Fe(1 1 0) single crystal has been studied by means of high resolution photoelectron spectroscopy (HRPES) and scanning tunneling microscopy (STM). Core level spectra were analyzed in detail on both clean and adsorbate covered surfaces. A shoulder on the high binding energy side of the Fe 2p core level indicates a structure comprising multiple components interpreted as an exchange split of the final state due to interaction between the 2p and 3d electrons. After adsorption of oxygen, (2 5), (2 2) and (3 1) reconstructions were observed with atomically resolved STM. The iron surface was further exposed to gradually higher doses of oxygen. Deconvolution of the O 1s HRPES spectra revealed two components separated approximately by 0.4 eV. The component at lower binding energy dominates at low coverage, while the high binding energy component increases in intensity with increasing O coverage. The formation of oxides was observed in the Fe 2p spectrum in the region between 708 and 710 eV. Further, well ordered iron oxides were grown by exposure to oxygen at 250 °C. The O 1s core level contained a single component with a binding energy similar to that of the high energy component in the just discussed O 1s spectrum. Low energy electron diffraction and STM images of this structure showed a large moir e pattern with a 22:1 A 30:9 A unit cell. Ó 2003 Elsevier Science B.V. All rights reserved. Keywords: Iron; Oxygen; Photoelectron spectroscopy; Scanning tunneling microscopy 1. Introduction Investigations of clean and adsorbate covered iron surfaces are of fundamental interest to im- prove the understanding of the initial interaction between Fe and gas. Since Fe is a commonly used and versatile material it is important to under- stand its initial oxidation. Fe(1 1 0), the most close packed low index surface, is a suitable model surface for well defined investigations of the properties of iron surfaces. A number of photoelectron spectroscopy (PES) investigations of the Fe 2p core level have been performed with standard laboratory X-ray sources [1–3]. These investigations lack the required reso- lution to resolve the complex intrinsic structures of this core level. Until recently the intensities needed to do high resolution experiments of Fe 2p at synchrotron facilities have not been available. Development of third generation sources using undulators and wigglers has made these investi- gations possible. Previous studies of the Fe 2p have investigated single crystal thin film samples * Corresponding author. Tel.: +46-8-7904161; fax: +46-8- 249131/208284. E-mail address: [email protected] (J. Weissenrieder). 0039-6028/03/$ - see front matter Ó 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0039-6028(03)00018-9 Surface Science 527 (2003) 163–172 www.elsevier.com/locate/susc

Upload: kth

Post on 01-Feb-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

Oxygen structures on Fe(1 1 0)

J. Weissenrieder *, M. G€oothelid, M. M�aansson, H. von Schenck,O. Tjernberg, U.O. Karlsson

Laboratory of Materials and Semiconductor Physics, Royal Institute of Technology, Electrum 229, SE-164 40 Kista, Sweden

Received 11 September 2002; accepted for publication 7 January 2003

Abstract

The adsorption of oxygen on a Fe(1 1 0) single crystal has been studied by means of high resolution photoelectron

spectroscopy (HRPES) and scanning tunneling microscopy (STM). Core level spectra were analyzed in detail on both

clean and adsorbate covered surfaces. A shoulder on the high binding energy side of the Fe 2p core level indicates a

structure comprising multiple components interpreted as an exchange split of the final state due to interaction between

the 2p and 3d electrons. After adsorption of oxygen, (2� 5), (2� 2) and (3� 1) reconstructions were observed with

atomically resolved STM. The iron surface was further exposed to gradually higher doses of oxygen. Deconvolution of

the O 1s HRPES spectra revealed two components separated approximately by 0.4 eV. The component at lower binding

energy dominates at low coverage, while the high binding energy component increases in intensity with increasing O

coverage. The formation of oxides was observed in the Fe 2p spectrum in the region between 708 and 710 eV. Further,

well ordered iron oxides were grown by exposure to oxygen at 250 �C. The O 1s core level contained a single component

with a binding energy similar to that of the high energy component in the just discussed O 1s spectrum. Low energy

electron diffraction and STM images of this structure showed a large moir�ee pattern with a 22:1 �AA� 30:9 �AA unit cell.

� 2003 Elsevier Science B.V. All rights reserved.

Keywords: Iron; Oxygen; Photoelectron spectroscopy; Scanning tunneling microscopy

1. Introduction

Investigations of clean and adsorbate covered

iron surfaces are of fundamental interest to im-

prove the understanding of the initial interaction

between Fe and gas. Since Fe is a commonly used

and versatile material it is important to under-

stand its initial oxidation. Fe(1 1 0), the most close

packed low index surface, is a suitable model

surface for well defined investigations of the

properties of iron surfaces.A number of photoelectron spectroscopy (PES)

investigations of the Fe 2p core level have been

performed with standard laboratory X-ray sources

[1–3]. These investigations lack the required reso-

lution to resolve the complex intrinsic structures of

this core level. Until recently the intensities needed

to do high resolution experiments of Fe 2p at

synchrotron facilities have not been available.Development of third generation sources using

undulators and wigglers has made these investi-

gations possible. Previous studies of the Fe 2p

have investigated single crystal thin film samples

*Corresponding author. Tel.: +46-8-7904161; fax: +46-8-

249131/208284.

E-mail address: [email protected] (J. Weissenrieder).

0039-6028/03/$ - see front matter � 2003 Elsevier Science B.V. All rights reserved.

doi:10.1016/S0039-6028(03)00018-9

Surface Science 527 (2003) 163–172

www.elsevier.com/locate/susc

[4,5], Fe(1 0 0) bulk crystals [6,7] and polycrystal-

line samples [8], but to the authors knowledge the

Fe(1 1 0) bulk single crystal surface has not previ-

ously been investigated with high resolution pho-

toelectron spectroscopy (HRPES).

Previous studies of the Fe 2p core level havefocused on the magnetic interaction and exchange

between the 2p core hole and the 3d electrons. In

line with this approach unpolarized, linear and

circular magnetic dichroism in the angular distri-

bution (MDAD) has been performed [4–9]. Some

studies have performed spin-resolved measure-

ments in order to elucidate the spin configuration

of the core level [4,5], others have focused on thechirality of the incident light [6]. Despite their

different approaches all previous studies have used

the same Zeeman like model [10] in their analysis

of the 2p core level.

The adsorption and chemisorption of oxygen

on iron and the growth of oxide layers have

been extensively studied in the past by a wide va-

riety of techniques: low energy electron diffraction(LEED) [11], scanning tunneling microscopy

(STM) [12,13], XPS [1–3], spin and/or angle re-

solved valence band photoemission spectroscopy

[14–17], X-ray absorption spectroscopy (XAS)

[18], magnetic dichroism [18,19] and high resolu-

tion electron energy loss spectroscopy (HREELS)

[20,21]. To the authors knowledge, up to now, no

HRPES study of the O 1s core level has beenperformed in order to examine the O2 adsorption

on Fe(1 1 0).

This work is focused on using HRPES and

STM to investigate the clean and oxygen covered

Fe(1 1 0) surface. HRPES from the clean Fe 2p

core level revealed a multiplet structure and de-

tailed measurements of the O 1s at different cov-

erages have been performed. Atomically resolvedSTM images of the Fe(1 1 0)(2� 5)-O, Fe(1 1 0)-

(2� 2)-O and Fe(1 1 0)(3� 1)-O superstructures

are presented as well as HRPES and STM images

of a grown iron oxide.

2. Experimental

The PES experiments were performed at

beamline I511 at MAX-lab, Sweden [22]. This

beamline is a third generation undulator based

VUV and soft X-ray beamline aimed at high res-

olution X-ray photoelectron spectroscopy (XPS),

XAS and X-ray emission spectroscopy (XES) and

is using a modified SX-700 monochromator. The

photoelectron spectra were recorded at 300 K witha rotatable Scienta SES200 electron spectrometer

[23].

Connected to the analysis chamber is a prepa-

ration chamber equipped with LEED, sputtering

and annealing facilities as well as leak valves for

gas dosing. Temperatures were measured using a

chromel–alumel thermocouple spot-welded to the

sample. The base pressures in the two chamberswere lower than 10�10 Torr.

The PES data were collected with the incident

photons at a grazing angle of �10� and the elec-

tron analyzer was positioned at approximately

normal emission. The polarization of the light was

in a plane approximately perpendicular to the

sample surface.

The STM experiments were performed in asystem that has been described previously else-

where [24]. In brief it consists of a two-chamber

ultra high vacuum system with a base pressure

<10�10 mbar. The measurements were performedwith an Omicron VT-STM at room temperature.

All STM data were acquired in the constant cur-

rent mode and are presented as top view gray scale

images with darker colors corresponding to lowerlevels. Most of the images have been processed to

remove a linear background in the x- and y-di-rections. Connected to the STM chamber is a re-

action chamber with either a LEED or an Auger

electron spectrometer (AES). The reaction cham-

ber is also equipped with sputtering and annealing

facilities, precision leak valves for various gases

and a mass spectrometer. The annealing temper-atures of the sample were measured by a pyro-

meter. Since the STM and PES experiments were

not performed in the same chamber the required

dose to create a super structure was somewhat

different in the two chambers.

A Fe(1 1 0) single crystal (FOM Instituut voor

Atoom- en Molecuulfysica) sample, with dimen-

sions 1� 3� 9 mm3 was used in the present study.The misalignment of the crystal from the (1 1 0)

orientation is <0.5�. The Fe(1 1 0) crystal was

164 J. Weissenrieder et al. / Surface Science 527 (2003) 163–172

cleaned by cycles of argon ion sputtering and an-

nealing up to 750 K. In order to remove residual

carbon the surface was oxygen treated at 5� 10�8

Torr during annealing at 650 K. After the clean-

ing procedure the sample displayed an excellent

(1� 1) LEED pattern and no impurities were de-tected with either valence band spectroscopy, de-

tailed scans in the C 1s, N 1s, O 1s, and S 2p

regions or with wide scan PES.

3. Results and discussion

All core-level spectra obtained were fitted using

a Voigt function with a Doniach–Sunjic line pro-

file. The fitting parameters determining the shape

of the different components are the lifetime width

for the core hole or the Lorentzian width CL, theGaussian width CG and the asymmetry parameter,

or singularity, a. CG was obtained from the total

system resolution relevant for each spectrum. awas allowed to vary with oxygen coverage, since

the density of states at the Fermi level changes. All

binding energies were measured relative to the

Fermi level of the Fe crystal.

Only few studies have investigated Fe 2p bymeans of linearly polarized photoemission from a

third generation synchrotron source [4,6–8]. Other

investigations have used circularly polarized light

[5,9,25], but most of these studies have lacked the

required experimental resolution to clearly resolve

the intrinsic structure of the Fe 2p core level. The

easy axis of magnetization of bulk Fe is along the

[1 0 0] directions [26]. Following this, the preferredmagnetization directions of the magnetic domains

is oriented in the plane and 45� out of the plane ofthe Fe(1 1 0) bulk single crystal surface. Since the

magnetic anisotropy energy of Fe is minimized

when the magnetization of a domain is oriented

along the easy axis, it can be assumed that the

majority of the magnetic domains are oriented in

the [1 0 0] directions. The magnetic anisotropy is avery sensitive parameter that is strongly dependent

on among other things stress and film thickness.

Previous HRPES studies of the Fe(1 1 0) surface

have investigated iron thin films, �70 �AA thick, and

not bulk single crystal samples [4]. They have

therefore not been measured on an exact similar

magnetization configuration as the present study.

The Fe 2p photoelectron spectra are, as ex-

pected from earlier PES studies, dominated by the

large, �13 eV, spin–orbit splitting of the 2p level.

The two main photoelectron lines are assigned tothe 2p1=2 line at higher binding energy and the

2p3=2 line at lower binding energy. Fig. 1 contains

Fe 2p3=2 spectra taken at 795 eV photon energy.

The shoulder on the high binding energy side of

spectra taken from a clean surface indicates

that the peak consists of at least two intrinsic

components with different binding energies and

intensities. During the initial deconvolution pro-cedure the spectrum was assumed to contain only

two peaks, A1 and A2 (Fig. 1a). Their respective

binding energies were found from curve fitting to

be 706.1 and 706.9 eV. During the deconvolution

CL, CG and a were set equal for A1 and A2 and a

reasonably good fit was generated with CL ¼ 0:44eV, CG ¼ 0:27 eV and a ¼ 0:33. These fitting pa-

rameters are in agreement with what have been

Fig. 1. Photoelectron spectra of Fe 2p3=2 taken at 795 eV

photon energy. Spectrum (a) is a fit with two components

shifted 0.8 eV (CG ¼ 0:27 eV, CL ¼ 0:44 eV, a ¼ 0:33), (b) is a

four component fit with an equidistant energy spacing of 0.35

eV (CG ¼ 0:27 eV, CL ¼ 0:40 eV, a ¼ 0:25).

J. Weissenrieder et al. / Surface Science 527 (2003) 163–172 165

used in previous investigations of this core level

[4,5,7]. The relative intensity of A2 is 0.2 when

normalizing to the intensity of A1 and the relative

intensity seem not to change with oxygen cover-

age. The spectra in Fig. 2b and c are taken after anexposure of 25 and 50 L oxygen and the shoulder

can still be observed in the spectra. This indicates

that the shoulder can be attributed to bulk prop-

erties and is not a surface component.

Since the measured shift in the Fe 2p3=2 was

insensitive to surface properties, our experimental

configuration can be compared to studies investi-

gating other Fe single crystals. Studies per-

formed on Fe(1 0 0) reported a shoulder shifted

�0.6 eV to higher binding energy [6,7]. Further,

the previously mentioned iron thin film study

performed spin-resolved measurements and found

a shift of �0.8 eV in binding energy between ma-jority and minority electrons in the Fe 2p3=2 [4].

The sum of the minority and majority spec-

trum show similarities with our spectrum. Thus the

present shift between A1 and A2 is comparable

with previous studies.

A previous investigation claims that the influ-

ence of exchange interaction between the valence

3d electrons and the 2p electrons only can bestudied by spin-resolved photoemission or mag-

netic dichroism [4]. It has also been shown that the

properties of the Fe 2p line shape can be explained

by a magnetic dichroism effect connected to the

alignment of the core holes along with the mag-

netization axis, which is independent of the mag-

netization of the surface and only depend on the

change of chirality of the incident light [6]. Inour study we can assume to have a multi-domain

sample, which includes areas with opposite or dif-

ferent magnetization directions. According to this

the spectra can be interpreted as magnetic field

integrated and this is in line with that Fe 2p

spectra collected from different parts of the surface

were similar.

Following this it should be possible to performa Zeeman like analysis of the splitting in the 2p

core level [4,5,7,8,10]. In this analysis the Fe 2p3=2is assumed to split into four sublevels for magne-

tized samples, mj ¼ �3=2, �1=2, 1=2 and 3=2. It isthe exchange interaction between the magnetic 3d

states and the 2p core hole that is assumed to lift

the degeneracy of the core hole state with j ¼ 3=2,like the magnetic field in the Zeeman effect. TheZeeman like description of the Fe 2p is fairly well

adapted since the exchange splitting is <10% of the

spin–orbit splitting. Therefore the Fe 2p3=2 is

supposed to be composed of at least four compo-

nents. The positions of the mj sublevels are ex-

pected to be in line with Hund�s rule, i.e., lowerexcitation energy when the spin of the remaining

2p electrons is parallel to the spin of the 3d elec-trons. This has also been shown with spin-resolved

photoemission where our spectrum is similar to the

Fig. 2. Photoelectron spectra of Fe 2p3=2 taken at 795 eV

photon energy. The spectra are taken after (a) 10 L, (b) 25 L

and (c) 50 L oxygen exposure. The different oxide peaks are

indicated with arrows.

166 J. Weissenrieder et al. / Surface Science 527 (2003) 163–172

sum of the majority and minority spectra showing

the exchange interaction [4].

In previous analysis, the four different mj

sublevels have been assumed to have equidistant

energy spacing between 0.3 and 1.0 eV and

asymmetries between 0.025 and 0.4 have been used[4,6–8]. The best fit to our data was found with an

equidistant energy spacing of 0.35 eV. This is in

excellent agreement with the results from studies of

Fe(1 0 0) [7]. The fitting parameters was set to be

equal and CL ¼ 0:40 eV, CG ¼ 0:27 eV and

a ¼ 0:25 (Fig. 1b). If assuming a weak-field Zee-

man effect in conjunction with the mean field

approximation it is possible to make a rough es-timation of the equidistant energy spread of the mj

sublevels. Using an effective field value in iron in

the order of 103 T [26] we end up with an energy

separation between sublevels in the order of 0.1

eV. Thus by means of very simple calculations it is

possible get an order of magnitude estimate. A

previous study have used a fully relativistic treat-

ment and calculated, in accordance to our results,an equidistant spread of 0.37 eV in Fe 2p3=2 [10].

A spin-resolved magnetic circular dichroism

study claims evidence for a not equidistant distri-

bution of the mj sublevels in Fe 2p3=2 [5]. They also

claim a minimum of four lines is needed to de-

scribe the Fe 2p3=2 core level and that the lifetime

broadening and binding energy of the majority

and minority states are similar and only theintensities are changing with the different spin

channels or reversal of the magnetism. In coordi-

nation with this study we released the sublevel

spacing energy, but kept the widths and asymme-

try constant. The new deconvolution did not result

in a significantly improved fit to the spectrum in-

dicating that the previous fit using an equidistant

energy shift describes our data reasonably well.Since the energy split of the different sublevels are

reasonably small and almost equal to the experi-

mental resolution the deconvolution cannot be

claimed to be very certain. A non-uniform splitting

can be explained if the spin field is comparable to

the spin–orbit interaction [27], but the spin–orbit

interaction for the 2p shell is far too large for this

explanation to hold within a one-electron model.The Fe core level spectrum is affected by the

valence electrons due to the screening of the core

hole in the final state [4,9,25,28,29]. In PES it is

generally possible to distinguish between screened

and unscreened peaks in the analysis of photo-

electron spectra. The screened peaks usually cor-

respond to the main peaks and the unscreened

peaks to some satellite structure. Thus in freeelectron like metals the binding energy of the core

state will be lowered by the collective screening of

the valence electrons [28,29]. In materials with a

more correlated electronic structure the spectrum

is expected to contain not only well-screened core

hole states at lower binding energy but also peaks

at higher binding energy resulting from many body

effects. This effect is well known from gas phasePES [30] and might show up as strong satellite

structures, which are absent in a one-particle

model where Coulomb and exchange interaction

are only taken into account by the effective po-

tential [25,28]. In metals where many different

screening channels introduce an overall broaden-

ing and asymmetry of the photoelectron lines this

effect is generally less pronounced.In some of the previous studies concerning the

Fe 2p3=2 peak the deconvolution results in the

conclusion that the four mj sublevels cannot ac-

count for the extended tail of the spectrum, not

even by assuming large asymmetry factors for the

individual mj peaks. Therefore they come to the

conclusion that the high binding energy tail in-

tensity must hide unresolved satellites [5–7]. In oneof these studies the authors introduces a replica of

the spectrum, shifted in energy and intensity, to

deal with the asymmetry of the peak [7]. In our

study we do not encounter similar problems with

satellites in our curve fitting, but on the other hand

some of the mentioned studies use lower asym-

metries in their deconvolution [6,7]. Therefore the

existence of satellites cannot be definitely ruled outeven though our deconvolution fit the data rea-

sonably well.

In Fig. 3 the corresponding Fe 2p1=2 spectrum is

shown. In the fitting procedure we include a larger

Lorentzian lifetime broadening of the 2p1=2 than

that of the 2p3=2 and keep the other parameters

constant, which is in good agreement with previ-

ous experimental and theoretical studies [9,10].The larger width of the 2p1=2 level is caused by

rapid L2L2M45 Coster–Kronig processes [31]. The

J. Weissenrieder et al. / Surface Science 527 (2003) 163–172 167

2p1=2 peak rides on a larger background and

therefore it is harder to observe the asymmetry and

the deconvolution of the peak is more uncertain.

Guided by the four components in the 2p3=2 peak,

the 2p1=2 was also fitted according to the one-electron model with two components mj ¼ 1=2 and�1=2 (B1 and B2). The best fit was achieved at an

energy split of �0.4 eV, similar to the energy

spacing found between sublevels in Fe 2p3=2. Their

respective binding energy was 718.9 and 719.3 eV

and the relative intensity of the higher binding

energy peak (B2) was 1.3 when normalized to the

lower energy peak (B1). A smaller spread in the2p1=2 peak has been reported previously [4,7,9,10]

and this result is in good agreement with the the-

oretically determined exchange induced spread of

the mj eigenvalues over 0.36 eV for the 2p1=2 and

1.11 eV for the 2p3=2 [10].

With the aim to investigate the initial interac-

tion of iron with oxygen the clean Fe(1 1 0) surface

was exposed to O2 doses between 0.5 and 50 L. At0.5 L the O 1s spectrum was found to contain

three components (Fig. 4). The dominating peak at

�529 eV binding energy has a shoulder on the high

binding energy side and, well separated from the

main peak, is another small component at 531.0

eV binding energy. The last component, C3, is at-

tributed to OH [1]. The origin of this component is

most likely due to hydrogen in the residual gas orwater vapor in the O2 reaction gas. Deconvolution

of the main peak revealed that it consists of two

components, C1 and C2, with binding energies of

528.9 and 529.3 eV respectively. When increasing

the oxygen coverage of the surface C2 increases in

intensity and grows larger than C1. The entire

main peak has previously been attributed to O2�

[1] and the two different components suggest that

there exist two different chemical bonds of the

Fig. 3. Photoelectron spectra of Fe 2p1=2 taken at 795 eV

photon energy. The fit is made with two components shifted 0.4

eV (CG ¼ 0:27 eV, CL ¼ 0:90 eV, a ¼ 0:33).

Fig. 4. Photoelectron spectra of O 1s taken at 640 eV photon

energy and with increasing O coverage. The doses are indicated

in the figure.

168 J. Weissenrieder et al. / Surface Science 527 (2003) 163–172

oxygen atoms. The iron oxide formation, obtained

from PES of Fe 2p, grows approximately as C2.

The oxide structure in Fe 2p is broad and initially

appears at �708 eV. An oxide shift of Fe 2p in the

order of 2 eV have been reported in XPS investi-

gations of bulk FeO crystals [1–3]. An exampleof this is shown in Fig. 2 where the oxide struc-

tures are indicated with arrows. In the spectra the

structures are still quite weak even though the

surface has been exposed to doses up to 50 L of

O2. At doses in the order of 10 L and higher, an-

other peak appears at �710 eV (Fig. 2a). This peak

with a shift in the order of 4 eV probably corre-

sponds to Fe2O3 [1–3]. Iron oxides are often mixedvalent systems with both Fe2þ and F3þ and these

ions can occupy many possible sites [18]. This will

give rise to a large width in the Fe 2p oxide spec-

trum. A previous LEED, MCD and XAS study

investigating the Fe(1 1 0) surface exposed to O2

found an initial formation of a FeO(1 1 1) surface

layer followed by a gradual transition into Fe3O4

at higher oxygen doses [18]. Since the Fe3O4 is acompound with both Fe2þ and Fe3þ this is in good

correlation with our study.

In order to grow a better defined iron oxide the

Fe(1 1 0) surface was exposed to oxygen at 10�7

Torr during annealing at 250 �C. The integrated

dose was in the order of 200 L and the oxides

showed a moir�ee LEED pattern. PES data of the O

1s revealed only one component as shown in thetop spectrum of Fig. 4. The peak has a binding

energy of 529.3 eV, in good agreement with C2 in

the previous O 1s spectrum. This, together with the

fairly good correlation between C2 and oxide

growth observed in Fe 2p indicates that C2 cor-

responds to an oxide while the C1 peak would

correspond to chemisorbed oxygen. The O 1s

intensity of the oxide spectrum is approximatelysix times the corresponding intensity of the Fe-

(1 1 0)(2� 2)-O bottom spectrum (see below

detailed comments on the ordered (2� 2) super-

structure). Since the oxygen coverage of the

Fe(1 1 0)(2� 2)-O structure is 0.25 monolayer the

oxygen coverage of the oxide film is determined to

be �1.5 monolayers, as seen from the relative O 1s

intensity. If the oxide composition is assumed tobe similar to FeO the thickness of the oxide would

be in the order of 3 monolayers.

After exposure to �0.5 L O2 and subsequent

annealing to 150 �C the first ordered structure was

observed with STM and LEED (Fig. 5a). This

structure correspond to Fe(1 1 0)(2� 5)-O and the

[0 0 1] and [�11 1 1] directions are indicated in the

figure. This is to our knowledge the first time thatthis structure has been reported. The structure was

found in small areas growing out from the bottom

of steps and is most likely stabilized in the vicin-

ity of the steps. The oxygen atoms are viewed as

protrusions in the image and form equidistant

rows oriented along the [�11 1 1] direction. The rowsgrow in small zigzag formations indicating a re-

construction of the iron lattice. The measuredcorrugation was 0.1 �AA along the rows and 0.3 �AAacross the rows. The STM image shows that the

coverage in the structure is slightly higher than the

nominal 0.1 monolayer coverage expected for a

(2� 5) reconstruction. Additional oxygen atoms

can be seen to occupy available positions between

the rows in the [0 0 1] directions. Since the peri-

odicity of the superstructure is five times thesubstrate lattice in the [0 0 1] direction half the

distance between the rows correspond to another

site than that occupied by the oxygen atoms in the

rows. The additional oxygen will prefer to occupy

a site similar to the ones occupied by the oxygen

atoms in the rows. Following this, they were found

to occupy one of two possible sites next to the site

halfway between the rows and thereby starting toform a super structure with two times the substrate

periodicity.

At exposures between 1 and 3 L O2 and

after annealing a Fe(1 1 0)p(2� 2)-O structure is

formed. Fig. 5b is a typical image of the (2� 2)

phase and with the [0 0 1] and [1 �11 1] directionsindicated in the image. Extended regions with

(2� 2) reconstruction are well resolved with is-lands typically several hundred �AA across. The su-

perstructure shows the expected periodicity along

the [1 �11 1] and [0 0 1] directions, 5.0 and 5.7 �AA re-

spectively, with a corrugation of �0.15 �AA. TheSTM image is consistent with the structure previ-

ously proposed for this overlayer from LEED and

HREELS measurements [20,21,32] with the O at-

oms sitting in the fourfold sites in a p(2� 2) array.Addition of more oxygen to the surface results

in the formation of a Fe(1 1 0)(3� 1)-O structure

J. Weissenrieder et al. / Surface Science 527 (2003) 163–172 169

as seen in Fig. 5c. This structure can be observed

at O doses between �3 and 15 L, but has to the

authors knowledge not previously been shown

with atomically resolved STM. A previous STM

study by Wight et al. [12] have reported a structure

of equidistant parallel rows with three times the

substrate periodicity, but not resolved the atoms

within the rows. The [1 �11 1] and [0 0 1] directionsare indicated in the image (Fig. 5c). The domains

with the same periodicity are very small and

elongated in the [�11 1 1] direction. The corrugationof the structure was �0.1 �AA. At higher exposuresthe atomic resolution is lost and the STM shows

images of equidistant rows, 7.5 �AA apart, stretching

over the surface. This structure corresponds most

likely also to a (3� 1) structure. At exposuresabove 15 L small domains (�20� 20 �AA2) with

(2� 1) and (1� 1) periodicity were observed.

Fig. 5d is a STM image of a grown iron oxide

displaying a moir�ee LEED pattern. The oxide was

prepared by exposure to �200 L oxygen during

annealing at 250 �C. The surface structure consistsof an ordered array of protrusions with a period-

icity of 22:1 �AA� 30:9 �AA, which is indicated in the

figure. Two additional protrusions are located

along one of the diagonals of the unit cell and sit�0.2 �AA deeper than the protrusions in the corners

of the unit cells. The distance between the two

protrusions located inside the unit cell was found

to be almost twice the distance between a corner

protrusion and the closest protrusion inside the

unit cell, thus leaving some sort of vacancy in the

middle of the unit cell. This structure was found

over the entire examined surface and the heightdifference between the highest and lowest site on a

terrace was found to be �0.8 �AA.

Fig. 5. STM images of oxygen covered iron surfaces: (a) Fe(1 1 0)(2� 5)-O, 100� 100 �AA2, )1.2 V, 1.8 nA, (b) Fe(1 1 0)(2� 2)-O,

100� 100 �AA2, )0.75 V, 3.0 nA, (c) Fe(1 1 0)(3� 1)-O, 135:2� 133:6 �AA2, )0.4 V, 3.7 nA and (d) ordered iron oxide with moir�ee LEED

pattern, 325� 316 �AA2, )0.2 V, 1.9 nA.

170 J. Weissenrieder et al. / Surface Science 527 (2003) 163–172

STM measurements of the clean Fe(1 1 0)

surface and of the oxygen covered surfaces shows

that the density of state around the Fermi level

is reduced on oxygen covered surfaces. In Fig. 6ðdI=dV Þ=ðI=V Þ is plotted as a function of bias

voltage for a Fe(1 1 0)(2� 2)-O surface. This mea-

sure is roughly proportional to the density of states

[33]. As seen in the figure the (2� 2) structure

posses a band gap of �0.3 eV. This is within the

wide range of previously reported band gap mea-

surements of iron oxides between 0.1 and 2.3 eV

[34].

4. Conclusions

We have performed HRPES and atomically

resolved STM in order to investigate clean and

oxygen covered iron surfaces. The clean Fe(1 1 0)

surface has a Fe 2p core level comprising multi-ple components. This is evidenced as a shoulder on

the high binding energy side in the spectrum. The

shoulder is not a surface peak; instead it is inter-

preted as an exchange split of the mj sublevels since

its relative intensity to the main component is in-

dependent of adsorption on the surface. It was

shown that the 2p core level could be interpreted

with a Zeeman like analysis with the mj sublevelsshifted 0.35 eV. The same analysis was also found

to be valid for Fe 2p1=2 where a 0.4 eV split was

estimated. Further, the adsorption of oxygen on

the surface was monitored with PES and STM.

The O 1s core level contained three separated

components that are interpreted as chemisorbed

oxygen, oxide and hydroxide. STM revealed a

number of different, some previously not reported,

oxygen induced reconstructions; (2� 5), (2� 2),

(3� 1), (2� 1) and (1� 1). At higher doses of

oxygen and subsequent annealing the iron surface

formed a moir�ee structure that was observed both

by LEED and STM.

Acknowledgements

The Swedish Natural Science Research Council

(NFR), the Swedish Research Council (VR), the

Axel Hultgren foundation, the G€ooran Gustafsson

foundation, Knut and Alice Wallenberg Founda-tion are acknowledged for funding. The MAX-lab

staff is also kindly acknowledged.

References

[1] T. Lin, G. Seshadri, J.A. Kelber, Appl. Surf. Sci. 119

(1997) 83–92.

[2] P. Graat, M. Somers, Appl. Surf. Sci. 100 (1996) 36–40.

[3] S.J. Roosendaal, I. Giebels, A.M. Vredenberg, F. Habra-

ken, Surf. Interf. Anal. 26 (1998) 758–765.

[4] F.U. Hillebrecht, Ch. Roth, H.B. Rose, W.G. Park, E.

Kisker, N.A. Cherepkov, Phys. Rev. B. 53 (1996) 12182.

[5] C. Bethke, N. Weber, F.U. Hillebrecht, ESRF Newsletter

33 (1999) 20–22.

[6] G. Paolucci, F. Sirotti, S. Lizzit, A. Baraldi, G. Panacci-

one, N.A. Cherepkov, G. Rossi, Surf. Sci. 377–379 (1997)

440–444.

[7] G. Rossi, G. Panaccione, F. Sirotti, S. Lizzit, A. Baraldi,

G. Paolucci, Phys. Rev. B 55 (1997) 11488–11495.

[8] S.H. Baker, K.W. Edmonds, A.M. Keen, S.C. Thornton,

C. Norris, C. Binns, Phys. Rev. B 61 (2000) 5026–5032.

[9] L. Baumgarten, C.M. Schneider, H. Petersen, F. Sch€aafers,

J. Kirschner, Phys. Rev. Lett. 65 (1990) 492.

[10] H. Ebert, L. Baumgarten, C.M. Schneider, J. Kirschner,

Phys. Rev. B 44 (1991) 4406–4409.

[11] C. Leygraf, S. Ekelund, Surf. Sci. 40 (1973) 609–635.

[12] A. Wight, N.G. Condon, F.M. Leibsle, G. Worthy, A.

Hodgson, Surf. Sci. 331–333 (1995) 133–137.

[13] Y. Joseph, C. Kuhrs, W. Ranke, M. Ritter, W. Weiss,

Chem. Phys. Lett. 314 (1999) 195.

[14] S. Masuda, Y. Harada, H. Kato, K. Yagi, T. Komeda, T.

Miyano, M. Onchi, Y. Sakisaka, Phys. Rev. B 37 (1988)

8088–8095.

[15] Y. Sakisaka, T. Komeda, T. Miyano, M. Onchi, S.

Masuda, Y. Harada, K. Yagi, H. Kato, Surf. Sci. 164

(1985) 220–234.

[16] H.-J. Kim, E. Vescovo, Phys. Rev. B 58 (1998) 14047–

14050.

Fig. 6. STM of Fe(1 1 0)(2� 2)-O.

J. Weissenrieder et al. / Surface Science 527 (2003) 163–172 171

[17] E. Vescovo, C. Carbone, W. Eberhardt, O. Rader, T.

Kachel, W. Gudat, Phys. Rev. B 48 (1993) 285–288.

[18] H.-J. Kim, J.-H. Park, E. Vescovo, Phys. Rev. B 61 (2000)

15284–15287.

[19] H.-J. Kim, J.-H. Park, E. Vescovo, Phys. Rev. B 61 (2000)

15288–15293.

[20] W. Erley, H. Ibach, Solid State Commun. 37 (1981) 937–

942.

[21] T. Miyano, Y. Sakisaka, T. Komeda, M. Onchi, Surf. Sci.

169 (1986) 197–215.

[22] R. Denecke, P. V€aaterlein, M. B€aassler, N. Wassdahl, S.

Butorin, A. Nilsson, J.-E. Rubensson, J. Nordgren, N.

M�aartensson, R. Nyholm, J. Electron Spectrosc. Relat.

Phenom. 101 (1999) 971.

[23] N. M�aartensson, P. Baltzer, P. Br€uuwiler, J.-O. Forsell, A.

Nilsson, A. Stenborg, B. Wannberg, J. Electron Spectrosc.

Relat. Phenom. 70 (1994) 117.

[24] J. Weissenrieder, M. G€oothelid, G. Le Lay, U.O. Karlsson,Surf. Sci. 515 (2002) 135–142.

[25] H.A. D€uurr, G. Van der Laan, D. Spanke, F.U. Hillebrecht,

N.B. Brookes, Europhys. Lett. 40 (1997) 171–175.

[26] C. Kittel, Introduction to Solid State Physics, seventh ed.,

Wiley, USA, 1996.

[27] G. Van der Laan, Phys. Rev. B 51 (1995) 240–249.

[28] G. Van der Laan, Appl. Phys. A 73 (2001) 673–678.

[29] H. Tillborg, A. Nilsson, N. M�aartensson, J. Electron

Spectrosc. Relat. Phenom. 62 (1993) 73.

[30] Ph. Wernet, J. Schulz, B. Sonntag, K. Godehusen, P.

Zimmermann, M. Martins, C. Bethke, F.U. Hillebrecht,

Phys. Rev. B 62 (2000) 14331–14336.

[31] R. Nyholm, N. M�aartensson, A. Lebugle, U. Axelsson, J.

Phys. F 11 (1981) 1727.

[32] D. Cappus, M. Hassel, E. Neuhaus, M. Heber, F. Rohr,

H.-J. Freund, Surf. Sci. 337 (1995) 268–277.

[33] R.M. Feenstra, J.A. Stroscio, Phys. Scripta T 19 (1987)

55.

[34] R.M. Cornell, U. Schwertmann, The Iron Oxides, VCH

Verlagsgesellschaft mbH, Weinheim, 1996.

172 J. Weissenrieder et al. / Surface Science 527 (2003) 163–172