Initial stages of ZrO2 chemical vapor deposition on Si(1 0 0)-(2 × 1) from zirconium tetra-tert-butoxide

Article abstract of DOI:10.1016/j.susc.2008.03.016

The initial stages of chemical vapor deposition of ZrO₂ from zirconium tetra-tert-butoxide (ZTB) on Si(1 0 0)-(2 × 1) have been studied by scanning tunnelling microscopy (STM) and synchrotron radiation excited photoelectron spectroscopy (PES). The STM images and core level (PES) spectra indicate that the predominant surface modifications induced by ZTB are due to silicon carbonization and formation of zirconium dioxide. The carbonization reaction leads to formation of subsurface carbon and two types of reconstructions are discussed: dimer vacancies and dimer vacancies in conjunction with a rotated surface Si-dimer. Indications for the formation of small amounts of zirconium silicide are also found. No evidence for silicon oxidation can be observed with PES, in contrast to the interface properties previously found after larger exposures to ZTB.

Full text of DOI:10.1016/j.susc.2008.03.016

Surface Science 602 (2008) 1803–1809
Contents lists available at ScienceDirect
Surface Science
journal homepage: www.elsevier.com/locate/susc
Initial stages of ZrO2 chemical vapor deposition on Si(100)-(2 ꢀ 1) from
zirconium tetra-tert-butoxide
P.G. Karlsson a, E. Göthelid a,b, J.H. Richter a, A. Sandell a,
*
a Department of Physics and Materials Science, Uppsala University, P.O. Box 530, SE-75121 Uppsala, Sweden
b Department of Chemical Engineering, Mälardalen University, P.O. Box 325, SE-631 05 Eskilstuna, Sweden
a r t i c l e i n f o
a b s t r a c t
Article history:
The initial stages of chemical vapor deposition of ZrO2 from zirconium tetra-tert-butoxide (ZTB) on
Si(100)-(2 ꢀ 1) have been studied by scanning tunnelling microscopy (STM) and synchrotron radiation
excited photoelectron spectroscopy (PES). The STM images and core level (PES) spectra indicate that
the predominant surface modifications induced by ZTB are due to silicon carbonization and formation
of zirconium dioxide. The carbonization reaction leads to formation of subsurface carbon and two types
of reconstructions are discussed: dimer vacancies and dimer vacancies in conjunction with a rotated sur-
face Si-dimer. Indications for the formation of small amounts of zirconium silicide are also found. No evi-
dence for silicon oxidation can be observed with PES, in contrast to the interface properties previously
found after larger exposures to ZTB.
Received 13 November 2007
Accepted for publication 18 March 2008
Available online 23 March 2008
Keywords:
High k dielectrics
Zirconium dioxide
Silicon
Chemical vapor deposition
Semiconductor–insulator interfaces
Synchrotron radiation photoelectron
spectroscopy
Ó 2008 Elsevier B.V. All rights reserved.
Scanning tunnelling microscopy
1. Introduction
ular precursors rather than impinging atoms as in physical vapor
deposition (PVD) [5–8]. The growth process may involve adsorp-
Ultrathin films of metal oxides with high dielectric constants
have received considerable attention as a replacement to SiO2 as
gate insulator in metal–oxide semiconductors field effect transistor
(MOSFET) devices. Zirconium dioxide (ZrO2) has a high thermal
stability, low thermal conductivity and a high dielectric constant,
and is therefore a suitable candidate material as gate oxide. MOS
devices using ZrO2 as dielectric have shown most promising elec-
trical characteristics [1,2].
The chemical vapor deposition (CVD) technique is well adapted
to gate oxide deposition within a MOSFET manufacturing process.
With CVD one can achieve atomic layer control of the film thick-
ness, uniform thickness of large areas, stoichiometric control and
excellent conformal step coverage of non-planar devices [3,4].
With a proper choice of precursor it is furthermore possible to
grow films at low temperatures with a satisfactory deposition rate,
making the method economically viable.
tion, desorption, site-specific reactions and transport of precursor
fragments on the substrate surface.
In order to unravel the underlying atomistic processes in CVD of
thin films a combination of surface sensitive spectroscopic and
microscopic tools is a necessity. We have previously studied CVD
growth of ZrO2 on Si(100) and Si(111) using the zirconium tet-
ra-tert-butoxide [Zr(OC(CH3)3)4, (ZTB)] precursor [9–11]. The film
thickness range in these studies was 4–74 Å. Employing photoelec-
tron spectroscopy (PES) and X-ray absorption spectroscopy (XAS)
the interfacial properties and the band alignment were character-
ized in detail. Growth at 400 °C was found to lead to extensive Si
oxidation and hydrocarbon fragments embedded in the interface.
Scanning tunnelling microscopy (STM) is most valuable as it of-
fers the possibility to atomically resolve surface structures at ex-
tremely low coverages. There are very few examples of previous
investigations of CVD undertaken with STM. From studies of metal
deposition using metal carbonyl precursors it can be concluded
that STM can reveal significant differences in the CVD growth pro-
cess as compared to PVD [12,13].
Since the electronic properties of a MOSFET device is strongly
dependent on the interfacial properties, studies of the initial stages
of film deposition are highly motivated. Moreover, the CVD tech-
nique has a very high degree of complexity due to the use of molec-
The drawback with STM is however the difficulties in obtaining
chemical information. This information can instead be provided by
core level photoelectron spectroscopy. The binding energies of the
core levels are element specific and detailed information about the
* Corresponding author. Tel.: +46 18 471 35 48; fax: +46 18 471 35 24.
E-mail address: anders.sandell@fysik.uu.se (A. Sandell).
0039-6028/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2008.03.016

1804
P.G. Karlsson et al. / Surface Science 602 (2008) 1803–1809
local chemical surrounding is obtained by the chemical shift. The
technique is surface sensitive by nature due to the detection of
electrons. By exploiting the virtues of high photon flux and variable
photon energy at a synchrotron radiation source the surface sensi-
tivity can be dramatically enhanced as compared to standard XPS
instruments. This allows for studies of the very low surface cover-
ages needed in order to understand the initial nucleation
processes.
In this paper, we use STM and core level PES to study the initial
stages of ZrO2 chemical vapor deposition on the Si(100)-(2 ꢀ 1)
surface from the ZTB precursor. We report on the details in the
ZTB surface reactions involved in the formation of submonolayer
structures. It is possible to identify two principal surface processes:
Silicon carbonization and formation of ZrO2. We propose that the
two types of surface modifications can be distinguished in the
STM images. The carbonization process leads to characteristic sur-
face reconstructions, the structures of which are discussed. Indica-
tions for Zr–Si bond formation are found but there is no sign of
silicon oxidation. The absence of observable Si oxidation makes
the initial stages quite different from the latter stages of growth.
ple steps were also oriented 90°. Hence, the small terraces are not
primarily induced by the cleaning procedure but are an inherent
property of the wafer.
While heated to 400 °C, the sample was exposed to ZTB via a
stainless steel tube (10 mm in diameter). Some relatively thick
films were always produced to confirm that the system was work-
ing satisfactorily and to further passivate the UHV parts. The doses
are given in milliLangmuirs (mL, 1 mL = 10ꢁ9 Torr s) read as the
background pressure.
3. Results
3.1. STM images
A top view of the clean Si(100)-(2 ꢀ 1) surface is presented in
Fig. 1a. It consists in elongated terraces typically about 20 nm wide
and separated by monoatomic steps. No step bunching was ob-
served and the step density is as expected for the specified misori-
entation of the wafer. Rows of silicon dimers run along the terraces,
rotated by 90° from one to another. Two different types of defects
are observed in good agreement with previously published data
and consist in single dimer vacancy (DV) defects (also denoted type
A defects) and half missing dimer defects (type C defects). Their
density is also in agreement with the literature, with about 1 A-type
defect and 1.2 C-type defect per 5 ꢀ 5 nm2 [16,17]. Overview
images of at least 100 ꢀ 100 nm2 (not showed here) exhibit the
same low defect density. The zigzag chains are due to imaging of
‘‘frozen” buckled dimers, as observed previously [16].
2. Experimental
The experiments were performed in two different Ultra High
Vacuum (UHV) systems with
a base pressure lower than
2 ꢀ 10ꢁ10 torr, one for each technique used (STM and PES). The
STM measurements were performed with a customised variable
temperature (VT) UHV scanning probe microscope from Omicron
Vakuumphysik GmbH, Germany. All STM data were acquired in
the constant current mode at room temperature (RT), and are pre-
sented as top view grey scale images with darker colors corre-
sponding to lower levels. Both Pt/Ir and W tips were used for the
experiments. Subsequent reconditioning of the tip was undertaken
by running line scans at 10 V sample bias and 50 nA tunnelling cur-
rent. This procedure resulted in stable and reproducible images of
the surface. All images shown here correspond to occupied states
(Vt < 0), since it was very difficult to probe the unoccupied states
after ZTB dosing. The photoelectron spectroscopy (PES) measure-
ments were carried out at beam-line I311 at MAX-lab [14]. The
end station is equipped with an electron energy analyzer of type
SCIENTA SES200. The spectra are calibrated relative to the Si 2p3/
The image presented in Fig. 1b corresponds to a 20 ꢀ 20 nm2
topograph obtained after 0.1 mL ZTB exposure. It stands clear that
the defects density has increased dramatically already after this
small dose. The number of dimer vacancies is considerably higher
than before ZTB exposure, with up to 15.5 DV per 5 ꢀ 5 nm2. At
several places adjoining DVs are observed forming one-dimen-
sional ‘‘trenches” perpendicular to the dimer rows. Extended
ꢀ
grooves along the Si(100) [011] direction are also imaged (la-
belled E). The depth of these dark areas is about 0.5–1.0 Å. As their
length varies, the evaluation of their density is not reported here.
Moreover, a small number of very bright protrusions can be ob-
served, typically adjacent to the grooves E. The height of the bright
spot from the top of the dimer rows is about 1.5 Å. From the line
scan in Fig. 1 it can be inferred that this height is in good agree-
ment with the apparent step height.
peak for the clean Si(100)-(2 ꢀ 1) surface, set to 99.2 eV binding
2
energy (BE) relative to the Fermi level.
In both cases, a separate preparation chamber is connected to the
analysis chamber, equipped with low energy electron diffraction
(LEED)andsampleheatingfacilities, andonwhichthedosingsystem
was connected. The zirconium tetra-tert-butoxide (ZTB) precursor
(99.99%, Stream Chemicals) was introduced in a pyrex tube welded
to a conflat (CF)flange and swiftly connectedvia a valve to the dosing
system. The later was in its turn connected to the preparation cham-
ber via a precision leak valve. All vacuum parts were thoroughly pas-
sivated by exposure to ZTB right after baking.
Small rectangular samples were cut from n-doped (P) Si(100)
wafers (Polishing Corp. of America, 0.5°, 0.1–1.0 ohm cmꢁ1). Dust
particles were removed by blowing with dry N2. The samples were
put into the UHV systems as received. This was followed by a thor-
ough outgasing/annealing procedure [15], which involved flashes
to 1250 °C by passing direct current through the sample. This pro-
cedure resulted in clean samples according to PES and STM
standards.
Furthermore, an intriguing observation is the appearance of
small reconstruction patches (R) randomly across the surface, with
a density of about 4 per 5 ꢀ 5 nm2. In these patches, one dimer is
missing and the other is replaced by a smaller protrusion. These
patches thus cover each an area approximately corresponding to
a (3 ꢀ 2) cell. The protrusion within the R feature is centered in
the direction of the dimer rows and off centered in the direction
perpendicular to the dimer rows. In the STM images, the protru-
sions appear at intermediate height as compared to the bottom
of the grooves E and the dimer rows. They are not tip-induced arti-
facts as their position varies from defect to defect within the same
image and are independent of the scanning angle. The vast major-
ity of the defects observed at this stage can be assigned to DV, E
and R categories, either separately or in combination. However,
the R reconstructions are always observed as separate entities or
in combination with E or DV defects, that is, there is no tendency
for extended (3 ꢀ 2) island formation.
It can be noted that the tolerance for misorientation is 0.5° for
the present grade of silicon wafer. The misorientation observed
on the stepped images obtained on the clean surface is within this
value. For different samples cut from approximately the same po-
sition of the wafer but rotated 90° relative to each other, the sam-
Fig. 2 shows a 50 ꢀ 50 nm2 STM topograph after a ZTB exposure
of 0.2 mL, i.e. after approximately doubling the exposure. The sub-
strate dimer rows are still visible and the (2 ꢀ 1) areas amount to
at most 50%. The rest of the surface is basically imaged as depres-
ꢀ
sions running along the ½011ꢂ and the [011] directions. At this

P.G. Karlsson et al. / Surface Science 602 (2008) 1803–1809
1805
C
A
_
[011]
[011]
_
[011]
[011]
[nm]
0.30
0.20
0.10
0.00
20 nm
0
10
20
30
40
50 [nm]
Fig. 2. Filled states STM topographs of a Si(100)-(2 ꢀ 1) surface exposed to 0.2 mL
of ZTB at 400 °C. The topograph is measured at room temperature using a Pt/Ir-tip.
The sample bias was ꢁ1.9 V and the tunnelling current was 0.9 nA. The substrate
dimer rows are still visible. No preferential reactions are observed at the step edge.
Apart from the significantly increased area of depressions there is also a substantial
amount of very bright protrusions.
R
E
amounts to a nominal exposure of approximately 0.2 mL. This dos-
age results in no detectable ZTB-induced Zr 3d, C 1s and O 1s sig-
nals. Only very small changes in the Si 2p spectrum were observed.
Most apparent was a slight decrease of the state associated with
the ‘‘up-atoms” of the surface dimers forming the (2 ꢀ 1) recon-
struction (spectra not shown).
[011]
_
[011]
DV
The total nominal dose after the second exposure is estimated
to about 1 mL. Zr 3d, C 1s and O 1s signals are now detectable
and the spectra are shown in Fig. 3. The Zr 3d5/2 BE of 182.3 eV sug-
gests that Zr4+ species are present, see Fig. 3a [18]. It is possible
also to discern very weak structures at 180–178 eV BE. This is
indicative of a minute amount of Zr0, most likely due to silicide for-
mation [19]. There is one symmetric O 1s peak at 532 eV BE
(Fig. 3b). The C 1s spectrum of the clean surface displayed in
Fig. 3c has a broad structure at 284.4 eV BE and a weak feature
at 282.8 eV. The peak of 284.4 eV BE is attributed to spurious car-
bon in graphitic form due to surface contamination. The peak of
282.8 eV BE is assigned to carbidic carbon formed from surface
contaminations present during the annealing procedure in the
cleaning process. After 1 mL ZTB exposure, the intensity at
282.8 eV, i.e. the signature of carbide-like species, increases signif-
icantly, whereas the intensity of the peak due to graphitic carbon is
unchanged. Since no additional intensity emerges at 284–285 eV
the presence of adsorbed hydrocarbons on the surface can be ex-
cluded [10,20]. The behavior at this early stage of film nucleation
is thus different from that found for a ZTB-grown film a few Å
thick, where clear evidence for hydrocarbon species is found
[10]. The formation of carbonaceous residues including C–O bonds
is expected to give rise to C 1s states at 286–287 eV [10,20]. The C
1s spectrum does not give any clear evidence for the formation of
such species.
[nm]
0.30
0.20
0.10
0.00
0
5
10
15
20 [nm]
Fig. 1. Filled states STM topographs for (a) the clean Si(100)-(2 ꢀ 1) surface and (b)
the surface of the same sample after 0.1 mL of ZTB exposure. The sample was kept
at 400 °C during exposure to ZTB. The topographs are measured at room temper-
ature using a W-tip. The sample bias was ꢁ1.4 V in both measurements and the
tunnelling current was (a) 2.7 nA and (b) 2.0 nA. The image of the clean surface (a)
shows a low density of A-type (dimer vacancy, DV) and C-type defects. After exp-
osure (b) an increased amount of dimer vacancies (DVs) are observed, sometimes
forming trenches. Additional defects are squares or elongated areas along the S-
i(100) high symmetry directions imaged as depressions (E) and small reconstruc-
tions (R).
stage, the high defects density precludes a clear separation into the
different types previously defined (E, DV, or R). No preferential
reactions are observed at the step edge. As seen by comparing
Fig. 1b and Fig. 2, the number of very bright 1.5 Å high protrusions
has increased considerably although their overall surface coverage
remains very small. The majority of these bright protrusions ap-
pear as singular entities but there are also a few aggregates.
To obtain a rough estimate of the oxygen coverage a Si(100)
surface was oxidized with O2. An assumption was made that the
5 Å thick SiOx film produced in this way correspond to about
1 mL oxygen. By normalizing the integrated O 1s intensity the O
coverage after 1 mL ZTB exposure was estimated to be 0.1–
3.2. Core level photoemission spectra
Core level photoemission spectra are shown in Figs. 3 and 4. In
the PES experiments ZTB was dosed in two steps. The first dose

1806
P.G. Karlsson et al. / Surface Science 602 (2008) 1803–1809
the low BE shoulder followed by subtraction. Since the low BE
shoulder is associated with surface states the difference spectrum
represents the states of bulk character. The comparison at the bot-
tom of Fig. 4b shows that the changes in the bulk contribution
upon ZTB exposure are negligible. The surface contribution to the
spectrum for the clean (2 ꢀ 1) surface contains several compo-
nents. In the following only the 2p3/2 spin-orbit component of
the various species will be considered for simplicity and brevity.
The structure at 98.6 eV BE (labeled 1) is associated with the
‘‘up-atoms” of the surface dimers forming the (2 ꢀ 1) reconstruc-
tion [22]. The state at 99.30 eV BE (labeled 2) contains contribu-
tions both from the dimer down-atoms and second layer atoms
[22,23]. These have not been delineated for clarity.
Zr 3d
= 243 eV
hν
Zr0
Zr4+
1 mL ZTB
The exposure to 1 mL ZTB induces changes in the Si 2p surface
contribution as shown at the top of Fig. 4b. The relative intensity in
the BE regime 98.5–99 eV has decreased significantly and there is
no well-resolved structure. The decreased intensity of state 1 upon
ZTB exposure can be due to dimer decomposition or saturation of
the surface dangling bonds [23]. However, state 1 is not sufficient
in order to obtain a good fit in this BE regime – an additional state
of 98.9 eV BE is also needed (labeled 3). We attribute the new state
to formation of zirconium silicide. ZrSix has been found to give rise
to a peak at 98.8 eV, well in line with the BE of state 3 [19]. The for-
mation of small amounts of ZrSix is consistent with the appearance
of weak Zr0 states in the Zr 3d spectrum. With respect to the STM
images it is furthermore noteworthy that the decrease of state 1
(about 60%) suggest that the situation attained after 1 mL ZTB in
the PES system is comparable to the situation attained after
0.2 mL ZTB in the STM system.
186
184
182
Binding Energy [eV]
180
178
O 1s
= 610 eV
h
ν
1 mL ZTB
Clean (2x1)
From Fig. 4b it is evident that exposure to 1 mL of ZTB also gives
rise to a significantly increased relative intensity at about 99.5 eV.
Keeping state 2 unchanged in shape and BE position from the spec-
trum of the clean surface, an additional component located at
99.55 eV (labeled 4) is needed to reproduce the extra intensity.
Features at this BE have previously been observed upon reaction
with hydrogen and ethylene [23,24]. The deposition temperature
is however close to that of hydrogen desorption [24]. That state 4
is the result of the reaction with a carbonaceous compound is more
likely. We therefore propose that this state is related to the C 1s
peak, which then suggests the formation of a SiC-like compound.
Especially noteworthy is finally that exposure to ZTB does not
lead to the formation of oxidized Si species. The Si 2p3/2 component
of Si1+ is located at a BE 0.9–1.0 eV higher than that of the bulk
peak and higher oxidation states appear at even higher BE:s. There
is no sign of such states in the Si 2p spectrum. Thus, during the ini-
tial stages Si react with ZTB by formation of Si–Zr and Si–C bonds.
536
534
532
530
528
Binding Energy [eV]
C 1s
= 330 eV
h
ν
1 mL ZTB
Clean (2x1)
288
286
284
Binding Energy [eV]
282
280
4. Discussion
Fig. 3. Zr 3d (a), O 1s (b) and C 1s (c) core level photoemission spectra before and
after exposure to 1 mL of ZTB. The spectra are normalized to number of scans and
photon flux.
The PES results demonstrate that the initial reaction between
Si(100)-(2 ꢀ 1) and ZTB at 400 °C leads predominantly to forma-
tion of ZrO2, ZrSix and SiC. The STM images show that the interac-
tion with ZTB gives rise to dark patches following the high
symmetry directions, bright protrusions, reconstructions described
as (3 ꢀ 2) units and an increased density of dimer vacancies. In the
following, we will discuss the relation between the PES results and
the STM structures and compare the initial stages of ZTB-mediated
ZrO2 growth with our previous results on the latter stages of the
growth.
A patch of ZrO2 must have a lower density of states than the Si
surface within a few eV from the Fermi level and should therefore
appear as a darker area in the STM images. The most straightfor-
ward explanation is therefore that the dark patches (labelled E in
Fig. 1b) are associated with ZrO2. Previous STM results for the reac-
tion of O2 with Si(100)-(2 ꢀ 1) at room temperature reveal that
0.2 mL. The tabulated atomic cross sections [21] suggest that the
carbon coverage is of a similar magnitude. That is, the C:O ratio
is found to be lower than the factor of four expected from the stoi-
chiometry of the ZTB molecule (16:4). No reliable value of the
amount of Zr could be attained.
Si 2p spectra for the clean Si(100)-(2 ꢀ 1) surface and after
1 mL ZTB are shown in Fig. 4a. Two different photon energies were
used, 114 eV and 150 eV. These values provide Si 2p spectra with
low and high surface sensitivity, respectively [22]. In Fig. 4b the
photon energy dependent surface sensitivity has been used to sep-
arate the spectra into a ‘‘bulk” contribution and a ‘‘surface” contri-
bution. This was accomplished by normalizing the two spectra to

P.G. Karlsson et al. / Surface Science 602 (2008) 1803–1809
1807
2
4
Si 2p
Si 2p
Surface components
1 mL ZTB
1 mL ZTB
150 eV
3
1
Surface components
Clean (2x1)
114 eV
Clean (2x1)
150 eV
Bulk components
1 mL ZTB
Clean (2x1)
114 eV
101
100
99
98
101
100
99
98
Binding Energy [eV]
Binding Energy [eV]
Fig. 4. (a) Si 2p spectra for the Si(100)-(2 ꢀ 1) surface before and after small doses of ZTB. The spectra recorded at two different photon energies, 114 and 150 eV, of which
the latter results in a higher surface sensitivity. (b) Separation of the spectra in contributions predominantly from the bulk and surface regions, respectively. The spectra are
normalized as to have the same maximum intensity in order to facilitate comparison of relative intensities.
SiOx gives rise to dark patches following the high symmetry direc-
tions [25–27]. These patches appear as depressions at both positive
and negative bias. From this follows, that the ZrO2 areas are imaged
similar to the SiOx areas.
Moreover, the STM image recorded after the lowest ZTB expo-
sure reveals an increased density of DVs by more than one order
of magnitude. These DVs are in some cases are forming one-dimen-
sional trenches, as seen in Fig. 1b. A previous STM study supported
by ab initio calculations of surfaces with low carbon coverage
formed by hydrocarbon reactions demonstrate that incorporation
of C in the fourth layer is most likely [33]. On top of the carbon
atom a surface dimer is missing and the reconstruction is com-
monly denoted DV41 [33]. From this follows that the ZTB induced
behavior observed in Fig. 1b can be interpreted in terms of carbon
incorporation and DV41 formation [33].
There is no conclusive result that shows that the R reconstruc-
tions are induced by carbon species. However, given the amount of
carbon provided by the ZTB molecule this is plausible. A tentative
interpretation of the R structure is shown in Fig. 5. A carbon atom
has been placed in the fourth layer and the (2 ꢀ 1) dimer above the
carbon atom is missing, i.e. a DV41 defect. In addition, a rotated di-
mer (RD) has been introduced next to the DV41. A DV41+1RD de-
fect is consistent with the STM image of the R feature: the
structures are confined within a (3 ꢀ 2) cell and a protrusion that
is off-center in the direction along the (2 ꢀ 1) rows can be
expected.
The c(4 ꢀ 4) structure has been descried in terms of DV41 + 2RD
defects [31]. The model presented in Fig. 5 can in principle be
viewed as locally similar to the c(4 ꢀ 4) reconstruction. However,
the c(4 ꢀ 4) structure has a strong tendency to grow by island for-
mation, in contrast to the dispersed nature of the R features. Even
at an exposure where about half of the surface is affected by ZTB,
no sign of c(4 ꢀ 4) patches can be found. The quenching of the
c(4 ꢀ 4) island formation can be related to the complexity of the
ZTB decomposition, which involves both ZrO2 formation and sub-
sequent reaction with the hydrocarbon groups.
Apart from the formation of ZrO2 and SiC there are two other
observations in the PES spectra that are noteworthy: the absence
of oxidized Si and the finding of very small amounts of Zr0. The for-
mation of ZrO2 without oxidation of Si raises the question of how
the ZrO2 is linked to the Si substrate. In the absence of oxidized
The oxidation of Si(100) leads furthermore to the appearance of
bright protrusions, observed as such at both positive and negative
bias [26,27]. The apparent height of the bright protrusions has
been reported to be 1.4 Å [27], in good agreement with the 1.5 Å
found in the present study. It has been argued that bright sites ap-
pear in the STM image due to (neutral) Si atoms ejected to the sur-
face by the oxidation reaction [25]. In an alternative picture based
on scanning tunnelling spectroscopy, the bright sites are oxidized
Si adjacent to adsorbed oxygen [27]. In our case, the origin of the
bright protrusions is even more uncertain since there are several
different elements present. The Si 2p spectrum shows however
no sign of oxidized Si atoms. Thus, if the bright sites are associated
with Si atoms, they are either due to neutral ejected atoms or due
to oxidized atoms too few to be observable with PES.
The PES results, and to some extent the STM images, suggest
that carbon incorporation is most likely upon small exposures of
ZTB at 400 °C. The BE of the C 1s peak is identical to that reported
for the c(4 ꢀ 4) structure. The C atoms of the c(4 ꢀ 4) structure are
embedded in a subsurface layer as inferred from the BE of the C 1s
peak (282.8 eV) [28,29] and X-ray photoelectron diffraction results
[30]. Bias-dependent STM images suggest that the carbon atoms
are located in the fourth layer [31]. The Si 2p spectrum exhibits a
component shifted by +0.45 eV vs. the bulk peak. A similar shift
has previously been observed upon annealing of a C2H4 layer on
Si(100) to a temperature at which the c(4 ꢀ 4) structure is formed
[24]. This is another indication that similarities to the c(4 ꢀ 4)
structure exist in our case. The C:O atomic ratio is furthermore
lower than that of the ZTB molecule. Given the surface sensitivity
of the measurements, it is possible that the low C:O ratio reflects
different locations of the carbon and oxygen atoms, i.e. the carbon
species is located subsurface while the oxygen species is at the sur-
face. The C:O ratio can however also be decreased due to desorp-
tion of carbonaceous species (most probably isobutene [32]).

1808
P.G. Karlsson et al. / Surface Science 602 (2008) 1803–1809
: Top layer Si dimer atom
: 2nd layer Si atom
: 3rd layer Si atom
: 4th layer Si atom
: C atom
Fig. 5. The figure shows a tentative structure model of the R reconstruction observed in the STM image shown in Fig. 2b. The carbon atom is located in the fourth layer, above
which a surface dimer is missing. This is denoted a DV41 defect. Adjacent to the DV41 is a rotated dimer (RD). The DV41 + 1RD configuration is confined within a (3 ꢀ 2) cell
in line with the STM image.
Si a ZrO2–Si interface has to contain Zr suboxide species bonded to
Si atoms. We have no conclusive evidence for Zr in an intermediate
oxidation state, i.e. +1 to +3. However, the Zr 3d BE of an O3Zr–Si
B
ZrO2/ZrSix
Oy/Si(100)-(2x1)
species may be shifted by about ꢁ1 eV relative to the Zr4+ peak
assuming a 3+ oxidation state [34]. The Zr 3d peaks shown in
Fig. 4a are nearly 2 eV broad and the presence of a Zr3+ minority
species cannot be excluded. The Zr0 species is most likely associ-
ated with the formation of ZrSix as discussed above. We have how-
ever no STM structure that can be identified as due to ZrSix and we
can therefore not determine whether ZrSix is formed separately or
e.g. in connection with ZrO2.
4 Å
2
hν = 130 eV
Si1+
The results discussed above demonstrate a very different
behavior at the very beginning of the growth as compared to the
later stages of growth. We have previously studied ZTB-mediated
ZrO2 growth in the thickness regime 4–74 Å [10]. Using PES we
found that an extended interface of about 30 Å thickness is formed.
The interface can be described in terms of a ZrSixOy layer with a
composition that varies with the distance from the Si(100) sub-
strate. The Si 2p spectrum for a thick film (51 Å) gives evidence
for extensive Si oxidation. A distribution of oxidized species is ob-
served of which the +4 state clearly dominates [10]. The situation
is different for thinner films. Fig. 6 shows a Si 2p spectrum for a 4 Å
thick film on Si(100). A feature shifted by +0.3 eV relative to the
bulk peak is needed to obtain an accurate fit. This is exactly the po-
sition previously found for the state labeled 2 in Fig. 4b, which was
associated with dimer down-atoms and second layer atoms. Si
atoms bonded to hydrocarbon also give rise to a peak at this bind-
ing energy [20,24]. Since the 4 Å film contains Si bonded to hydro-
carbon fragments a contribution to this state is possible. However,
the peak still carries the label 2 for simplicity. Most important is
that oxidation of Si is evident after 4 Å deposition but in contrast
to the thick film the Si1+ state is predominant. For submonolayer
coverage no oxidized Si species is detectable with PES as shown
in Fig. 4. Consequently, oxidation of Si starts by formation of Si1+
species.
104
103
102
101
100
99
98
Binding Energy [eV]
Fig. 6. The Si 2p spectrum for a 4 Å thick ZrO2 film deposited on the Si(100)-(2 ꢀ 1)
surface. The presence of oxidized Si in the form of Si1+ is clearly observed.
In the present case the photon flux was higher than in our pre-
vious study [10], which in combination with a much lower cover-
age may increase the susceptibility towards radiation-induced
reactions. We can therefore not completely exclude radiation in-
duced hydrocarbon decomposition and subsequent SiC formation.
However, in our previous work we were able to correlate the pres-
ence of oxidized Si to the presence of hydrocarbon species [10].
This is the case already after deposition of 4 Å where the Si1+ spe-
cies is observed (Fig. 6) along with hydrocarbon species [10]. The
absence of oxidized Si in the PES spectrum at submonolayer cover-
age (Fig. 4) is not likely to be an effect of radiation damage. Pro-
vided that SiOx is needed to stabilize adsorbed CHx the absence
of CHx at submonolayer coverage is fully consistent with the ab-
sence of SiOx.
C 1s states in the BE range 283.6–285.3 eV were also observed
for the thicker films [10]. We assigned the state at 283.6 eV to car-
bidic carbon but in the light of the present findings all the C 1s
states observed at film thicknesses of 4–74 Å are rather associated
with various forms of surface hydrocarbon fragments. That is, car-
bon incorporation and formation of a SiC compound is only found
at low (submonolayer) coverage.
From this it stands clear that the reactions involving the butoxy
ligands and the Si atoms changes drastically somewhere between
submonolayer coverage and 4 Å film thickness. Carbon incorpora-
tion has to be terminated and replaced by processes that yields
surface hydrocarbon fragments and allows for Si oxidation. The

P.G. Karlsson et al. / Surface Science 602 (2008) 1803–1809
1809
[6] W.L. Gladfelter, Chem. Mater. 5 (1993) 1372.
[7] D.J. Burleson, J.T. Roberts, W.L. Gladfelter, S.A. Campbell, R.C. Smith, Chem.
Mater. 14 (2002) 1269.
complexity of the surface precludes a more detailed description of
the processes.
[8] C.J. Taylor, D.C. Gilmer, D.G. Colombo, G.D. Wilk, S.A. Campbell, J. Roberts, W.L.
Gladfelter, J. Am. Chem. Soc. 121 (1999) 5220.
[9] A. Sandell, P.G. Karlsson, J.H. Richter, J. Blomquist, P. Uvdal, T.M. Grehk, Appl.
Phys. Lett. 88 (2006) 132905.
5. Conclusions
The initial stages of chemical vapor deposition of ZrO2 from zir-
conium tetra-tert-butoxide (ZTB) on Si(100)-(2 ꢀ 1) have been
studied by STM and PES. The decomposition of the ZTB molecule
leads to carbon incorporation and formation of SiC. It is proposed
that dimer vacancy structures and (3 ꢀ 2) patches observed with
STM are carbon induced. A model consistent with a (3 ꢀ 2) struc-
ture is presented. Indications for Zr–Si bond formation are found
but there is no sign of silicon oxidation.
The results are compared to those previously attained upon
deposition of thicker films. The absence of Si oxidation and the
SiC formation makes the initial stages quite different from the lat-
ter stages of the growth. Between submonolayer coverage and 4 Å
film thickness the reactions involving the butoxy ligands and the Si
atoms has to change from processes giving rise to carbon incorpo-
ration and ZrSix to processes leading to surface hydrocarbon frag-
ments and SiOx.
[10] P.G. Karlsson, J.H. Richter, J. Blomquist, P. Uvdal, T.M. Grehk, A. Sandell, Surf.
Sci. 601 (2007) 1008.
[11] J.H. Richter, P.G. Karlsson, B. Sanyal, J. Blomquist, P. Uvdal, A. Sandell, J. Appl.
Phys. 101 (2007) 104120.
[12] D.P. Adams, L.L. Tedder, T.M. Mayer, B.S. Swartzentruber, E. Chason, Phys. Rev.
Lett. 74 (1995) 5088.
[13] Z. Song, T. Cai, J.A. Rodriguez, J. Hrbek, A.S.Y. Chan, C.M. Friend, J. Phys. Chem. B
107 (2003) 1036.
[14] R. Nyholm, J.N. Andersen, U. Johansson, B.N. Jensen, I. Lindau, Nucl. Instrum.
Meth. A 467–468 (2001) 520.
[15] B.S. Swartzentruber, Y.-W. Mo, M.B. Webb, M.G. Lagally, J. Vac. Sci. Technol. A 7
(1989) 2901.
[16] R.J. Hamers, R.M. Tromp, J.E. Demuth, Phys. Rev. B 34 (1986) 5343.
[17] R.J. Hamers, U.K. Köhler, J. Vac. Sci. Technol. A 7 (1989) 2854.
[18] R. Puthenkovilakam, J.P. Chang, Appl. Phys. Lett. 84 (2004) 1353.
[19] T.S. Jeon, J.W. White, D.L. Kwong, Appl. Phys. Lett. 78 (2001) 368.
[20] J.W. Kim, M. Carbone, M. Tallarida, J.H. Dil, K. Horn, M.P. Casaletto, R. Flammini,
M.N. Piancastelli, Surf. Sci. 559 (2004) 179.
[21] J.J. Yeh, I. Lindau, Atom. Data Nucl. Data 32 (1985) 1.
[22] E. Landemark, C.J. Karlsson, Y.-C. Chao, R.I.G. Uhrberg, Phys. Rev. Lett. 69
(1992) 1588.
[23] R.I.G. Uhrberg, E. Landemark, Y.-C. Chao, J. Electr. Spectrosc. Rel. Phenom. 75
(1995) 197.
[24] F. Rochet, F. Jolly, F. Bournel, G. Dufour, F. Sirotti, J.-L. Cantin, Phys. Rev. B 58
(1998) 11029.
Acknowledgements
We acknowledge financial support from the Swedish Science
Council (VR) and the Göran Gustafsson Foundation.
[25] Ph. Avouris, D. Cahill, Ultramicroscopy 42–44 (1992) 838.
[26] M. Udagawa, M. Niwa, I. Sumita, Jpn. J. Appl. Phys. 32 (1993) 282.
[27] H. Ikegami, K. Ohmori, H. Ikeda, H. Iwano, S. Zaima, Y. Yasuda, Jpn. J. Appl.
Phys. 35 (1996) 1593.
[28] H.C. Wang, R.F. Lin, X. Wang, Phys. Rev. B 36 (1987) 7712.
[29] L. Simon, L. Kubler, J. Groenen, J.L. Balladore, Phys. Rev. B 56 (1997) 9947.
[30] M. Stoffel, L. Simon, D. Aubel, J.L. Bischoff, L. Kubler, Surf. Sci. 454–456 (2000)
201.
References
[1] J.P. Chang, Y.-S. Lin, S. Berger, A. Kepten, R. Bloom, S. Levy, J. Vac. Sci. Technol. B
19 (2001) 2137.
[31] H. Kim, W. Kim, G. Lee, J.Y. Koo, Phys. Rev. Lett. 94 (2005) 076102.
[32] J. Kim, K. Kim, K. Yong, J. Vac. Sci. Technol. A 20 (2002) 1582.
[33] W. Kim, H. Kim, G. Lee, J.-Y. Koo, Phys. Rev. Lett. 89 (2002) 106102.
[34] T.M. Grehk, J. Engkvist, U. Bexell, J.H. Richter, P.G. Karlsson, A. Sandell, Thin
Solid Films 516 (2008) 875.
[2] B.-O. Cho, J.P. Chang, Appl. Phys. Lett. 93 (2003) 745.
[3] M. Ritala, M. Leskelä, L. Niinistö, P. Haussalo, Chem. Mater. 5 (1993) 1174.
[4] M. Ritala, K. Kukli, A. Rahtu, P.I. Räisänen, M. Leskelä, T. Sajavaara, J. Keinonen,
Science 288 (2000) 319.
[5] K.L. Choy, Prog. Mater. Sci. 48 (2003) 57.