Hafnium silicide formation on Si(001)

Article abstract of DOI:10.1103/PhysRevB.69.235322

The solid-state reaction of thick (～50 nm) and thin (～monolayer) films of Hf with cleaned and oxidized Si(001) substrates was investigated. Upon annealing to 1000 °C, films of HfSi₂ were formed after reaction times that depended upon the surfac

Full text of DOI:10.1103/PhysRevB.69.235322

PHYSICAL REVIEW B 69, 235322 (2004)
Hafnium silicide formation on Si„001…
H. T. Johnson-Steigelman, A. V. Brinck, S. S. Parihar, and P. F. Lyman
*
Laboratory for Surface Studies and Physics Department, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin 53211, USA
(Received 24 December 2003; published 25 June 2004)
The solid-state reaction of thick ͑ϳ50 nm͒ and thin ͑ϳmonolayer͒ films of Hf with cleaned and oxidized
Si͑001͒ substrates was investigated. Upon annealing to 1000 °C, films of HfSi2 were formed after reaction
times that depended upon the surface condition of the substrate before deposition. The chemical state of the
reacted surfaces was characterized using x-ray photoelectron spectroscopy, and the shifts in binding energy
upon silicide formation were recorded. Even for thick films, low-energy electron diffraction (LEED) revealed
that the ͑2ϫ1͒ pattern of the Si substrate emerged, suggesting that three-dimensional islanding of the HfSi2
film had occurred. The islanding behavior was investigated for both thick and thin films using LEED, atomic
force microscopy, and scanning electron microscopy. Streaking in the LEED patterns for the thick films suggest
that the island morphology is influenced by the underlying Si substrate.
DOI: 10.1103/PhysRevB.69.235322
PACS number(s): 79.60.Dp, 68.35.Fx, 81.05.Je
INTRODUCTION
The Hf/Si system has been investigated previously in the
hope of forming silicide device interconnects having low
sheet resistance and contact resistivity.8–10. In these works,
Hf was deposited on Si substrates, and the kinetics of the
reaction to form HfSi and then HfSi2 were monitored. No-
table findings were that HfSi and HfSi2 formed readily at
temperatures as low as 600 and 765 °C, respectively. Si was
found to be the dominant diffusing species. For a film of Hf
on crystalline Si, HfSi would form in a layerwise fashion
between the Si substrate and the unreacted Hf, but the for-
mation of HfSi2 would nucleate heterogeneously9 unless ex-
cess Si was supplied in the form of an amorphous Si capping
layer.10 Later work focused on the thin ͑ϳ12 nm͒ amorphous
hafnium silicide layer that is formed at the Hf/Si interface at
annealing temperatures of 400–500 °C.11–13
In the present work, we report results on the formation of
HfSi2 by the solid-state reaction of metallic Hf with Si͑001͒
substrates. We also investigated the effect on this reaction of
a layer of SiO2 at the interface between the deposited Hf film
and the substrate. The chemical state and compound forma-
tion was probed using x-ray photoelectron spectroscopy
(XPS) and powder x-ray diffraction (XRD), while important
inferences about the morphology of the reacted film were
derived from low-energy electron diffraction (LEED), scan-
ning electron microscopy (SEM), and atomic force micros-
copy (AFM). Because our primary interest is on the eventual
production of hafnium silicate films with a small Hf/Si ratio,
this study emphasized the formation of HfSi2 over the more
Hf-rich phases formed at lower temperatures.
The continued miniaturization of circuit elements in
ultralarge-scale integrated (ULSI) circuits has required that
the SiO2 layer that forms the gate oxide of field-effect tran-
sistors be made thinner and thinner. Eventually, a fundamen-
tal limit on the thickness of an effective insulating SiO2 bar-
rier will be reached when the layer is so thin that electron
tunneling provides a current leakage path that is unaccept-
ably high for low-power devices. This limit is expected to be
reached within a few years. As a result, an intensive search
has been launched for a material having a higher dielectric
constant than SiO2 that could serve as the gate oxide. Such a
“high-␬” insulator would allow the gate oxide to be physi-
cally thicker (for a given capacitance), dramatically reducing
tunneling. However, to serve as an effective gate insulator,
any replacement for SiO2 must possess the many favorable
electrical and structural properties of the SiO2/Si interface.
In particular, it is thought that oxides and silicates of Hf and
Zr may exhibit an interface with Si that is stable against
deleterious SiO2 formation, but would possess many of the
desirable interface attributes of SiO2.1,2
Hf forms stable silicates HfSixOy with a range of Hf/Si
ratios.2 To provide the smallest deviation from the mature
techniques established for device construction, it makes
sense to produce hafnium silicate films with a small Hf/Si
ratio if they can meet the electrical and physical require-
ments of a gate insulator. It has been asserted that a relatively
small amount of Hf can significantly increase the dielectric
constant of SiO2,3,4 although, for the related case of Zr, this
finding has been called into question.5,6 Structurally, it has
been found that hafnium silicates with a small Hf/Si ratio
readily form stable amorphous compounds,7 which are desir-
able in a gate oxide to reduce electrical leakage and dopant
diffusion along grain boundaries. The present paper reports
our investigation of the formation of hafnium silicide films,
which could then be oxidized to form a hafnium silicate.8 To
minimize the Hf/Si ratio, and in light of the fact that any
HfSix phase in the vicinity of the Si substrate will form the
disilicide upon annealing, we have concentrated on the for-
mation of the most Si-rich stable silicide phase, HfSi2.
EXPERIMENT
Hafnium silicide formation was studied in two regimes:
thick films ͑ϳ50 nm͒ and thin films (ϳmonolayer, ML). For
all experiments, the substrates consisted of commercial
Si͑001͒ wafers (p type, 5 ⍀ cm). Hf films were deposited in
either high vacuum (HV) (for the thick films) or ultrahigh
vacuum (UHV) (for the thin films), and were subsequently
annealed to various temperatures to initiate the silicidation
reaction. Because our primary interest is on the eventual pro-
0163-1829/2004/69(23)/235322(6)/$22.50
69 235322-1
©2004 The American Physical Society

JOHNSON-STEIGELMAN et al.
PHYSICAL REVIEW B 69, 235322 (2004)
duction of hafnium silicate films with a small Hf/Si ratio,
this study emphasized the formation of HfSi2 over the more
Hf-rich phases formed at lower temperatures.
(from the native oxidation of the deposited Hf films).16 Al-
though it had been reported long ago9 that Hf and Si react at
temperatures as low as 600 °C to form HfSi and at 765 °C
to form HfSi2, little change in any of our samples, either
visually or by XPS, was observed upon annealing to tem-
peratures less than 950 °C. (However, XPS is only sensitive
to the outermost several nm.) The only change observed at
these temperatures was a reduction of the HfO2 thickness in
some samples evidenced by the introduction a metallic fea-
ture in the Hf 4f spectra. No spots appeared in the LEED
pattern from these samples, indicating that the deposited sur-
face layer lacked long-range order.
Upon annealing to 1000 °C, however, the visual appear-
ance of all samples changed dramatically, from dark metallic
silver to a matte gray. Importantly, however, the amount of
time that must elapse at this annealing temperature before
this visual change occurs varied greatly with substrate sur-
face oxidation condition. Specifically, the HF-etched samples
reacted in ϳ10 mins, the samples having a native oxide took
ϳ1–2 hs, and the thermally oxidized samples took from 3 to
12 hs at 1000 °C. Thus, SiO2 appears to form a barrier that
slows the interdiffusion necessary for this reaction to pro-
ceed.
Samples were analyzed by XPS and low-energy electron
diffraction (LEED) in a VG Mk. II ESCALab UHV chamber
with a base pressure of less than 1ϫ10−10 torr. Unmono-
chromated Al K␣ radiation was used to excite the photoelec-
trons, and spectra were acquired with near-normal electron
emission. In addition to a survey spectrum, detailed scans
(instrumental resolution ഛ0.95 eV) of the O 1s, C 1s, Si 2p,
and Hf 4f regions were acquired for each surface condition.
Binding energies were referenced to the Ag 3d5/2 core level,
assumed to lie at 368.27 eV, and were determined with an
estimated overall precision of ±0.05 eV.
After XPS and LEED analysis, some samples were trans-
ported in air to an ABT-32 scanning SEM, where images
were acquired and energy-dispersive x-ray spectroscopy
(EDX) scans were performed. Other samples were scanned
using an atmospheric AFM. SEM, EDX, and AFM measure-
ments took place within days of sample production. Al-
though the surface is expected to have oxidized during tran-
sit, morphological features should remain on the surface.
For all of the samples, the condition after heating to
1000 °C was similar. After annealing (with the concomitant
change in visual appearance), all samples gave rise to essen-
tially the same XP spectra; specifically, the Hf 4f region con-
tained only one sharp spin-orbit doublet, and Si was appar-
ent, as shown in Fig 1. The Hf 4f doublet was extremely
narrow, with a linewidth of ഛ0.95 eV, and the binding en-
ergy (BE) of the 4f7/2 feature was 14.65 eV. For comparison,
we determined the BE of metallic Hf to be 14.35 eV, in good
agreement with published reports of 14.31 eV (Ref. 16) and
14.3 eV (Ref. 17). Metallic Hf shows a pronounced asym-
metry in the XPS line shape;16 upon conversion to HfSi2, the
asymmetry of the Hf 4f line shape was greatly reduced. (The
Doniach-Sunjic asymmetry parameter18 decreased from 0.32
to 0.14 upon silicidation.) The Si 2p3/2 line exhibited a
prominent feature at 99.45 to 99.50 eV [see Fig. 1(b)]. This
peak is essentially indistinguishable from that of substrate Si,
which we determined to have a Si 2p3/2 BE of 99.5 eV.19
Notable by its absence in XPS, even for samples initially
containing a 50-nm SiO2 layer, was any feature in the O 1s
region.
Thick films
For the thick-film studies, Si͑001͒ wafers with three dif-
ferent surface oxidation conditions were prepared prior to Hf
deposition. One set of samples was placed in a box furnace at
900 °C for 6 h to form a thermal oxide, calculated to be
ϳ50 nm using the Deal-Grove model.14 Another set of
samples was used as-received, with a ϳ1.5-nm native oxide.
The third set of samples was stripped of its native oxide
using dilute HF acid immediately before insertion into a HV
electron-beam evaporator. Hf was deposited at ϳ10−5 torr on
the three types of substrate samples, which were held at
room temperature (RT). Stylus profilometry indicated that
the deposited Hf layers were 50 to 80 nm thick. These
samples were then transferred through air to the UHV cham-
ber for annealing and analysis. During prolonged annealing
of samples to 1000 °C, the pressure of the UHV chamber
rose from its base pressure to ϳ2ϫ10−9 torr.
Thin films
For the thin-film studies, Si͑001͒ samples were inserted
into the UHV chamber as received, were degassed, and then
Structural order in the film was probed with LEED and
x-ray diffraction (XRD). Remarkably, samples annealed at
1000 °C for several hours gave rise to a ͑2ϫ1͒ LEED pat-
tern reminiscent of the clean surface of Si͑001͒ (see Fig. 2).
Subsequent powder XRD confirmed that annealed samples
were completely or dominantly composed of HfSi2. Thus, on
the basis of diffraction techniques, we conclude that the film
is composed of crystalline or polycrystalline HfSi2, and that
the surface of the specimen displays long-range order.
It is possible to explain these findings by the formation of
an epitaxial HfSi2 layer; this layer could have a ͑2ϫ1͒ re-
construction explaining its resemblance to the clean Si͑001͒
pattern. However, these findings are also consistent with the
formation of a HfSi2 film that has undergone three-
dimensional islanding. The formation of islands will expose
portions of the substrate to the vacuum, allowing the emer-
resistively heated to 1000 °C, rendering
a
clean
͑2ϫ1͒-reconstructed surface. Submonolayer to ML amounts
of 99.9% pure Hf (excluding ϳ3%Zr) were deposited on the
substrates using an electrostatic electron beam evaporator.15
The samples were held at RT, and the pressure typically rose
to ϳ5ϫ10−9 torr during deposition. As a control, one sample
was prepared identically, including operation of the Hf
evaporator, but was not exposed to Hf vapor. XPS and LEED
measurements were carried out immediately after deposition
and then again after subsequent anneals.
RESULTS
Thick films
Before annealing, no Si features appeared in the XPS
spectrum. The only Hf feature evident stemmed from HfO2
235322-2

HAFNIUM SILICIDE FORMATION ON Si͑001͒
PHYSICAL REVIEW B 69, 235322 (2004)
FIG. 2. LEED pattern acquired at 70 eV from thick ͑ϳ50 nm͒
HfSi2 film on HF-etched Si͑001͒ after annealing at 1000 °C for
3 h. For clarity, the video image has been inverted.
bright features as compared to the darker areas between
them.
Figure 5 depicts a typical AFM image of a thick-film
sample after annealing. The mottled surface suggests the for-
mation of islands. Typical sizes of these features are
150 nmϫ220 nm. The average feature height above the low-
est point in the image is 52 nm±8 nm. The areas between
the features are presumably exposed substrate.
Thin films
Upon RT deposition of ~1 ML Hf onto a clean Si͑001͒
-͑2ϫ1͒ surface, no spots were visible in the LEED pattern,
indicating a complete destruction of the surface’s long-range
order by the reactive metal film. The Hf 4f XP spectrum
(Fig. 6) showed a single broad feature centered near 16.0 eV
BE (which includes contributions from the 4f7/2 and 4f5/2
lines). The feature lacked the asymmetric tail characteristic
of metallic Hf. It is not possible to uniquely decompose this
peak, but it can be accounted for either as one Hf 4f silicide
doublet with twice the usual width, or as a combination of
several doublets each having the silicide line shape and
slightly displaced BE. The result is consistent with an inho-
mogeneous layer of Hf atoms bound to Si, but not fully
reacted. The Si 2p spectra did not reveal any surface com-
ponents attributable to Hf bonding. However, as demon-
strated with the thick films, little BE shift is expected upon
reaction with Hf, and moreover the data were not acquired in
a surface-sensitive mode.
Upon annealing to 750 °C for 5 min, the Hf 4f spectral
features coalesced into a sharp doublet characteristic of the
silicide phase, as shown in Fig. 6. The BE of the 4f7/2 peak
was 14.6 eV, which is about 0.6 eV shallower than prior to
annealing. Moreover, the LEED pattern changed back to a
͑2ϫ1͒ reminiscent of the clean surface (albeit with higher
background and broadened features).
This process was repeated several times, with additional
Hf metal being deposited at RT, which removed the LEED
pattern and broadened the XPS features; upon reannealing,
the ͑2ϫ1͒ pattern reemerged each time and the XPS features
FIG. 1. XP spectra from HfSi2 sample formed on HF-etched
Si͑001͒ and annealed to 1000 °C for 50 min. Reference spectra
from bulk Hf and Si are shown for comparison. Spectra are scaled
to the same height and are offset for clarity. (a) Hf 4f XPS spectrum
from thick ͑ϳ50 nm͒ HfSi2 sample (bottom) and clean metal Hf
foil (top). (b) Si 2p spectrum from thick HfSi2 sample (bottom) and
clean Si͑001͒-͑2ϫ1͒ (top).
gence of a LEED pattern from the substrate. Other evidence,
discussed below, strongly favors this latter interpretation.
For many of the samples that underwent prolonged an-
nealing and resultant islanding, a Si 2p3/2 feature at 101.2 eV
BE and a C 1s feature near 284.1 eV BE, both attributable to
SiC, became evident in XP spectra, as shown in Fig. 3. The
island interpretation of the LEED pattern is supported by
SEM imaging of the annealed samples. (No preanneal imag-
ing was conducted.) Figure 4 reveals typical surface mor-
phology of one of the thick-film samples. The features con-
sist of roughly circular islands with average diameters
ϳ400 nm to ϳ500 nm, but in many cases, two or more of
these islands merge into longer features. EDX spectra (not
shown) reveal a noticeably higher concentration of Hf in the
235322-3

JOHNSON-STEIGELMAN et al.
PHYSICAL REVIEW B 69, 235322 (2004)
FIG. 4. SEM image from thick ͑ϳ50 nm͒ HfSi2 sample on ther-
mally oxidized Si͑001͒ after annealing at 1000 °C for several
hours.
50 nmϫ70 nm, with typical heights of 1–2 nm. A control
sample, which was subjected to the same processing condi-
tions with the exception that no hafnium was deposited, ex-
hibited only the large features.
Comparison of these samples suggests that the small fea-
tures in Fig. 7 are related to the formation of Hf or HfSi2
islands, while the large features may be attributed to some
extraneous contamination inherent to the sample preparation.
Possibilities include contamination occurring before sample
loading, or SiC cluster formation20 either during sample
cleaning or due to the elevated pressure during operation of
the Hf evaporator. Hafnium-covered samples that were not
annealed (not shown) were similar to those that were an-
nealed without hafnium, i.e., they lacked the small features
seen on the hafnium-covered and annealed samples. SEM of
the thin-film samples did not reveal any visible features to
the limit of the instrument. This observation is consistent
with the feature sizes found in AFM. The concentration of Hf
was below the detection limits for EDX analysis.
DISCUSSION
Evidently, Hf and the Si substrate are able to react at
1000 °C despite the presence of a SiO2 layer at the interface
with a thickness of up to 50 nm, and no O is observed after
this annealing treatment. Recent studies suggest that Hf
metal1 or even substoichiometric HfOx films21 are able to
reduce SiO2 layers. In the present study, the O that was origi-
nally at the interface may have diffused into the bulk, or it
FIG. 3. XP spectra acquired from thick ͑ϳ50 nm͒ HfSi2 sample.
Hf was deposited on thermally oxidized Si͑001͒ after annealing at
1000 °C for 12 h. Fits are shown by solid lines. A Shirley back-
ground was subtracted before fitting. (a) Si 2p spectrum. The long
(short) dashed lines correspond to the HfSi2 (SiC) components. (b)
C 1s spectrum from same sample. The peak position is consistent
with SiC formation.
sharpened into a silicidelike line shape with a 4f7/2 compo-
nent at 14.6–14.7 eV BE. These observations suggest that
the Hf, upon annealing to 750 °C, gathered into HfSi2 is-
lands on the surface, exposing the substrate beneath.
An AFM image of a sample prepared with Hf as above is
shown in Fig. 7. Features can be classified into two size
categories, large and small. Large features had typical lateral
dimensions of 110 nmϫ150 nm, with typical heights of
Ϸ10 nm. Small features had typical lateral dimensions of
FIG. 5. AFM image (contact mode) from thick ͑ϳ50 nm͒ HfSi2
sample on Si͑001͒. The sample was annealed for 5 h at 1020 °C.
235322-4

HAFNIUM SILICIDE FORMATION ON Si͑001͒
PHYSICAL REVIEW B 69, 235322 (2004)
(i.e., those with no oxide at the Hf/Si interface) reacted and
changed their visual appearance in ϳ10 min. These samples
did not display an ordered LEED pattern after this brief an-
neal, suggesting that an epitaxial layer was not formed.
Moreover, for samples that underwent prolonged annealing,
a second feature attributable to SiC appeared at 101.2 eV BE
in the Si 2p XP spectra [see Fig. 3(a)]. We propose that SiC
forms at the bare patches of Si exposed to the residual gases,
which is not unexpected for these annealing times, tempera-
tures, and background pressures. Note that since SiC is
known to form clusters on Si surfaces,20 the presence of a
LEED pattern would not be significantly affected by the SiC
formation.
For the thick films, we may infer a little of the morphol-
ogy of the reacted layer by examining the LEED pattern
shown in Fig. 2. Streaking along the ͑00͒ to ͕01͖-like and
͕10͖ to ͕11͖-like directions indicates that the exposed patches
of substrate are long (compared to the electron coherence
length) in one dimension, and short in the other. Thus, the
exposed substrate patches must be long and narrow, and must
be influenced by the substrate crystallographic template. This
is supported by the AFM data.
For thin Hf films, the clean Si͑001͒-͑2ϫ1͒ pattern
emerged after short anneals to only 750 °C; no modification
(e.g., streaking) of the bare substrate LEED pattern was ob-
served. However, the “bare patches” in the thin film experi-
ment would be expected to be much larger than in the thick-
film experiment, likely larger than the LEED electron
coherence length. No streaking would be expected under
those conditions.
Many other transition metal silicides form three-
dimensionsal clusters on Si surfaces.23 Although the island-
ing behavior seen here for hafnium silicides is not surprising
in that light, it has not been previously reported to our
knowledge. Previous studies8,10–13,24 have made no mention
of islanding. In particular, Galakhov et al.24 interrogated
similar samples (Hf thicknesses of 50–200 nm) with depth-
dependent, electron-induced x-ray emission spectroscopy
(XES). Upon annealing to 780 °C for 30 min, the Hf film
was seen to convert to HfSi2 (as expected), but no signal
from the Si substrate was evident (for low electron excitation
energies). This does not conflict with our findings, as we do
not see a LEED pattern emerge (from thick-film samples)
until after annealing to 1000 °C. We do, however, have evi-
dence for islanding at lower temperatures ͑750 °C͒ in the
case of thin ͑ϳML͒ films. Thus, the picture that emerges is
that a thermodynamic driving force exists for three-
dimensional clustering of HfSi2 films, but kinetic barriers
associated with mass transport have prevented its previous
observation. The barrier may be overcome by sufficient an-
nealing temperatures (in the case of thick films) or may be
lowered by eliminating the need for so much mass transport
(in the case of ML films). Moreover, we have shown that the
presence of an interfacial oxide layer significantly slows the
kinetics of Hf/Si interdiffusion.
FIG. 6. Hf 4f XP spectra acquired from thin ͑ϳML͒ Hf depos-
ited at RT on clean Si͑001͒-͑2ϫ1͒ before and after annealing to
750 °C for 5 min. The lines are a guide to the eye.
may have diffused to the surface and desorbed. Although we
cannot distinguish these outcomes, it seems more likely that
O was desorbed, probably as SiO as known for the thermal
breakdown of SiO2 films. This supposition can explain why
the time required to completely react the surface increased
with the thickness of the substrate oxide layer. (Of course, it
is also possible that a silicate layer containing O remains at
the interface; however, in other studies conducted in our
laboratory,1,22 hafnium silicate layers were observed to break
down to HfSi2 under vacuum annealing to 1000 °C.)
After formation of the silicide phase, there are a number
of observations that imply that three-dimensional clusters or
islands occur on the surface of both the thin and the thick
samples. While it is possible that an epitaxial HfSi2 layer
formed, the simplest explanation supported by the experi-
mental evidence is island formation. For thick films annealed
for several hours in UHV, a LEED pattern characteristic of
the clean surface appears. The appearance of this pattern
from a ϳ50–80-nm film is difficult to explain except if the
HfSi2 film forms three-dimensional structures, thus allowing
a fraction of the bare surface to be exposed. A few samples
Unfortunately, the change in surface morphology (i.e., is-
landing) observed upon high-temperature reaction is unfa-
vorable for the intended eventual formation of a uniform
hafnium silicate layer having a small Hf/Si ratio by oxida-
FIG. 7. AFM image from thin ͑ϳML͒ HfSi2 sample on clean
Si͑001͒ after a 5-min anneal at 750 °C.
235322-5

JOHNSON-STEIGELMAN et al.
PHYSICAL REVIEW B 69, 235322 (2004)
CONCLUSIONS
tion of a HfSi2 layer. However, no LEED spots were ob-
served from the samples that completely reacted in 10 min,
so it may be possible to limit the kinetics of the island for-
mation using, e.g., rapid thermal processing.
Thick Hf films react with Si͑001͒ to form HfSi2 at
1000 °C, regardless of any initial SiO2 layer on the Si sub-
strate. An interposing SiO2 layer, however, will considerably
slow the kinetics of the reaction. The resultant HfSi2 exhibits
a sharp, fairly symmetric XPS line shape with BE of the 4f7/2
core level of 14.65 eV, while the Si 2p levels exhibit negli-
gible BE shift upon reaction. The resultant film exhibits a
pronounced tendency to form three-dimensional islands,
leaving patches of the substrate denuded of Hf. The shape of
the bare patches is long and narrow, indicating that the un-
derlying crystal structure influences the formation of the is-
lands. For thin ͑ϳML͒ Hf films on clean Si͑001͒, the reac-
tion and HfSi2 island formation takes places at temperatures
as low as 750 °C, consistent with a lower degree of mass
transport required.
The BE of the Hf and Si core levels change little upon
reaction to form HfSi2. It is tempting to try to interpret BE
shifts in XPS as arising (primarily) from charge transfer be-
tween atoms. In the present case, the Hf 4f7/2 line shifts
0.35 eV to deeper BE, while the Si 2p3/2 line barely shifts at
all. If one attempts to interpret these shifts as stemming from
charge transfer, it is difficult to understand why so little BE
shifting occurs. One expects that Hf, which has a Pauling
electronegativity of 1.3, would donate charge to Si, which
has a Pauling electronegativity of 1.9; this should result in a
substantial shift to deeper BE for Hf and a shift to shallower
BE for Si. However, this simple model is not reliable. In
particular, it has been shown that it is unable to account for
BE shifts upon alloy and compound formation.25 Our results
for BE shifts are similar overall to the findings of Zaima et
al.,12 but differ in detail. They report a Hf 4f7/2 BE shift from
13.8 eV (for the deposited Hf metal) to 14.1 eV upon forma-
tion of HfSi2, which (except for a 0.5 eV absolute discrep-
ancy) is consistent with our reported shift. However, they
report a Si 2p3/2 BE shift from 99.3 eV for bulk Si to
99.1 eV after reaction, presumably to HfSi2. We do not ob-
serve this 0.2 eV shift to shallower BE for the Si core level.
ACKNOWLEDGMENTS
This material is based on work supported by the National
Science Foundation under Grant No. DMR-9984442 and by
a REU. P.L. received additional support from the Research
Corporation. The authors are grateful to Daniel Shillinglaw
for assistance in data collection and to Michael Weinert for
useful discussions.
Author to whom correspondence should be addressed. Electronic
*
14 T. Hori, Gate Dielectrics and MOS ULSIs (Springer, New York,
1997), pp. 150–153.
address: plyman@uwm.edu
1 H. T. Johnson-Steigelman, A. V. Brinck, and P. F. Lyman, cond-
mat/0202328 (unpublished).
15 B. T. Jonker, J. Vac. Sci. Technol. A 8, 3883 (1990).
16 C. Morant, L. Galán, and J. M. Sanz, Surf. Interface Anal. 16,
304 (1990).
2 G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys. 89,
5243 (2001).
17 G. R. Gruzalski and D. M. Zehner, Phys. Rev. B 42, 2768 (1990).
18 S. Doniach and M. Sunjic, J. Phys. C 3, 285 (1970).
19 To fit the unresolved Si 2p3/2 and 2p1/2 spin-orbit components,
the two components were constrained to have a spin-orbit split-
ting of 0.60 eV, an intensity ratio of 2:1, and equal linewidths,
as suggested by F. J. Himpsel, B. S. Meyerson, F. R. McFeely, J.
F. Morar, A. Taleb-Ibrahimi, and J. A. Yarmoff, in Photoemis-
sion and Absorption Spectroscopy of Solids and Interfaces with
Synchrotron Radiation, edited by M. Campagna and R. Rosei
(North-Holland, Amsterdam, 1990), pp. 203–236.
20 L. Li, C. Tindall, O. Takaoka, Y. Hasegawa, and T. Sakurai, Phys.
Rev. B 56, 4648 (1997).
3 G. D. Wilk and R. M. Wallace, Appl. Phys. Lett. 74, 2854 (1999).
4 G. Lucovsky and G. B. Rayner, Jr., Appl. Phys. Lett. 77, 2912
(2000).
5 W.-J. Qi, R. Nieh, E. Dharmaarajan, B. Y. Lee, Y. Jeon, L. Kang,
K. Onishi, and J. C. Lee, Appl. Phys. Lett. 77, 1704 (2000).
6 G.-M. Rignanese, F. Detrauz, X. Gonze, A. Bongiorno, and A.
Pasquarello, Phys. Rev. Lett. 89, 117601 (2002).
7 M. A. Russack, C. V. Jahnes, and E. P. Katz, J. Vac. Sci. Technol.
A 7, 1248 (1989).
8 S. P. Murarka and C. C. Chang, Appl. Phys. Lett. 37, 639 (1980).
9 C. J. Kircher, J. W. Mayer, K. N. Tu, and J. F. Ziegler, Appl.
Phys. Lett. 22, 81 (1973); J. F. Ziegler, J. W. Mayer, C. J.
Kircher, and K. N. Tu, J. Appl. Phys. 44, 3851 (1973).
10 F. C. T. So, C.-D. Lien, and M.-A. Nicolet, J. Vac. Sci. Technol. A
3, 2284 (1985).
11 J. Y. Cheng and L. J. Chen, J. Appl. Phys. 68, 4002 (1990).
12 S. Zaima, N. Wakai, T. Yamauchi, and Y. Yasuda, J. Appl. Phys.
74, 6703 (1993).
13 S. Shinkai, H. Yanagisawa, K. Sasaki, and Y. Abe, Jpn. J. Appl.
Phys., Part 1 37, 643 (1998).
21 M. Copel and M. C. Reuter, Appl. Phys. Lett. 83, 3398 (2003).
22 H. T. Johnson-Steigelman, A. V. Brinck, and P. F. Lyman (unpub-
lished).
23 For a review of transition metal silicide behavior, see P. A. Ben-
nett and H. von Känel, J. Phys. D 32, R71 (1999).
24 V. R. Galakhov, E. Z. Kurmaev, S. N. Shamin, V. V. Fedorenko,
L. V. Elokhina, J. C. Pivin, S. Zamia, and J. Kojima, Thin Solid
Films 350, 143 (1999).
25 M. Weinert and R. E. Watson, Phys. Rev. B 51, 17 168 (1995).
235322-6