Structural and electrical properties of electron beam gun evaporated

advertisement
JOURNAL OF APPLIED PHYSICS
VOLUME 95, NUMBER 2
15 JANUARY 2004
Structural and electrical properties of electron beam gun evaporated Er2 O3
insulator thin films
V. Mikhelashvili and G. Eisensteina)
Department of Electrical Engineering, Technion-Israel Institute of Technology, Haifa 3200, Israel
F. Edelman and R. Brener
Institute of the Solid state Physics, Technion-Israel Institute of Technology, Haifa 3200, Israel
N. Zakharov and P. Werner
Max-Plank-Institut fur Mikrostructurphysik, Weinberg 2, D-06120, Halle, Germany
共Received 7 August 2003; accepted 21 October 2003兲
We present a detailed study of the evolution with annealing temperature 共in an oxygen environment兲
of the morphological and structural properties of thin erbium oxide (Er2 O3 ) films evaporated in an
electron beam gun system. The electrical characteristics of metal-oxide-semiconductor structures
are also described. Atomic force microscope and x-ray difractometry were used to map out the
morphology and crystalline nature of films ranging in thickness from 4.5 to 100 nm. High-resolution
cross-sectional transmission electron microscopy imaging and Auger electron spectroscopy reveal
three sublayers: an outer dense nanocrystalline Er2 O3 layer, a middle transition layer and amorphous
SiO2 film placed close to the Si substrate. The effective dielectric constant depends on the thickness
and the annealing temperature. A 1–2.8 nm interfacial SiO2 layer as well as an ErO inclusion with
low polarizability are formed during the deposition and the annealing process has a profound effect
on the dielectric constant and the leakages. The minimum effective oxide thickness is 2.4 –2.8 nm
and in the thinnest films we obtained a leakage current density as low as 1 – 5⫻10⫺8 A/cm2 at an
electric field of 1 MV/cm. We observe a shift of the flatband voltage to the positive side and
significant lowering of the positive charge down to ⬃1⫻1010 cm⫺2 . For a 4.5 nm film, the
maximum total breakdown electric field was approximately 1⫻107 V/cm. © 2004 American
Institute of Physics. 关DOI: 10.1063/1.1633342兴
I. INTRODUCTION
fuse to the Si surface forming an interfacial silicon oxide
layer, which lowers the overall effective dielectric constant.
In addition, a high temperature initiates the formation of silicide layers. This causes rugged surfaces,1,9 which are the
sources of ‘‘weak’’ points where local electric fields are enhanced causing increased leakages, degradation and breakdown of the metal-oxide-semiconductor 共MOS兲 structure.7,10
A different class of high-k materials are the rare earth
metal oxides. Y2 O3 共with k⫽12– 18) has shown promising
properties with a low leakage current, and good capacitance–
voltage (C – V) characteristics.11–15 Gd2 O3 , has been reported as a gate insulator for Si and GaAs. Its dielectric
properties (k⫽12– 14) and thermodynamic stability with
silicon that prevents the formation of silicides and roughed
surfaces14 –16 make it an attractive material. The rare-earthmetal oxides are also interesting due to their large conduction band offset over 2 eV8 and large band gaps ⬃5.4 eV.
Another rare-earth-metal oxide, Er2 O3 , has been extensively studied in the context of various optical devices, however its electrical properties as a gate dielectric have only
been addressed in a few published investigations.17–19 It was
shown in Ref. 20 that Er2 O3 reacts poorly with Si during
annealing compared to the other rare-earth-metal oxides such
as La2 O3 and Gd2 O3 . Indeed, no erbium silicide was detected by x-ray diffraction 共XRD兲17 measurements of films
annealed at up to 700 °C.
The thickness of SiO2 gate dielectrics in sub 0.1 ␮m
metal-oxide-semiconductor 共MOS兲 devices is approaching the quantum-tunneling limit of 1.5–2.5 nm. Potential
substitute insulators with large dielectric constants 共high-k
dielectrics兲 based on Ta2 O5 (k⫽23– 30) 1–3 and TiO2 (k
⫽25– 60) 4 –7 have been intensively studied in the past few
years. The physical thickness of these high-k films can be
increased so they enable the required effective oxide thickness 共EOT兲 while exhibiting low leakage currents. These alternative metal oxides have, however, low band gaps 共3.1
and 4.5 eV, for TiO2 and Ta2 O5 , respectively兲 and conduction band offsets on Si 共0 and 0.36 eV, respectively, for TiO2
and Ta2 O5 ).8 For comparison, SiO2 has a band gap of ⬃9 eV
and a conduction band offset of 3.5 eV. The high-k films may
therefore introduce a different class of leakages due to
Schottky emission of electrons into the band states or donortype oxygen vacancies, which enhance the carrier concentration and consequently the leakage current density.
Achieving a correct stoichiometry of the metal oxides
requires high temperature annealing in oxygen ambient. Annealing replaces vacancies by oxygen atoms thereby reducing the leakage current density. However oxygen atoms difa兲
Author to whom correspondence should be addressed; electronic mail:
gad@ee.technion.ac.il
0021-8979/2004/95(2)/613/8/$22.00
613
© 2004 American Institute of Physics
Downloaded 23 Apr 2006 to 132.68.56.217. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
614
J. Appl. Phys., Vol. 95, No. 2, 15 January 2004
Mikhelashvili et al.
In the present work we present a detailed study of thin
Er2 O3 layers evaporated on Si using electron beam gun
共EBG兲 evaporation. We used film thicknesses of 4.5–100 nm
for which we examined the dependencies on thickness and
annealing temperature of the crystalline structure and surface
morphology, as well as the electrical characteristics of MOS
capacitors.
II. EXPERIMENTAL PROCEDURE
P-type 共100兲 silicon substrates 共␳⫽7–10 ⍀ cm兲 were
dipped in a dilute HF solution to eliminate the surface native
oxide and then placed in the evaporation system, which had
a background pressure of 3 – 5⫻10⫺8 Torr. The evaporation
was carried out on unheated substrates without additional
oxygen at a pressure of 3⫻10⫺6 Torr. The growth rate was
0.02 nm/s for films thinner than 10 nm and 0.05 nm/s for
thicker films. Following the deposition, the films were annealed in an oxygen ambient at temperatures ranging from
350 to 750 °C for 60 min. Au contacts with an area of 5
⫻10⫺4 cm2 were evaporated on the oxide surface as gate
electrodes and Al was used for the back contacts. The deposited film thickness (d Er 2 O 3 ) was determined by a ␣-step stylus profilometer as well as by ellipsometry, which also
yielded the refractive index 共⬃1.78 –1.83兲. The surface morphology was characterized by an atomic force microscope
共AFM兲 operating in tapping mode. Film structures were
characterized by XRD and cross-sectional as well as plain
view transmission electron microscopy 共TEM兲. Auger electron spectroscopy 共AES兲 depth profiling was used to study
the chemical composition of the MOS structure. High frequency 共1 MHz兲 capacitance–voltage (C – V) and dc
current–voltage (J – V) measurements were performed using
standard techniques and instruments. All electrical measurements were carried out at room temperature.
III. RESULTS
A. Structural and morphological properties
1. AFM imaging
The characterization of the surface morphology of the
Er2 O3 layers was done for films of different thicknesses and
as a function of the annealing temperature. The films were
characterized by two parameters: the roughness root-meansquare (R q ) value and the difference between the highest and
lowest points in the scan range (R max). For the thinnest 共⬍11
nm兲 as deposited films we observed very smooth surfaces as
seen for a 4.5 nm film in Fig. 1共a兲. We obtain R q in the range
of 0.1–0.15 nm and R max on the order of 1–1.35 nm. These
values are comparable to those of the underlying silicon substrate. The values of R q and R max for these films were found
not to change with annealing temperature. The thicker, 50
and 100 nm as deposited films reveal higher R q and R max
values, 0.82, 1.1 and 5.96, 7.3 nm, respectively 关see Fig. 1共b兲
for 50 nm film兴. For those thicker layers we observed a small
increase of the roughness with annealing temperature with
R q ⫽1.0, 1.2 and R max⫽7, 8 nm 关see Fig. 1共c兲 for images of
FIG. 1. AFM images of Er2 O3 films of thickness 共a兲 d Er2 O3 ⫽4.5 nm—as
deposited, 共b兲 d Er2 O3 ⫽50 nm—as deposited, 共c兲 d Er2 O3 ⫽50 nm—annealed at
750 °C.
the morphology of 50 nm film兴. These results differ from the
case of TiO2 where a drastic change of surface morphology
was observed including formation of cracks at annealing
temperature above 550 °C.7
Downloaded 23 Apr 2006 to 132.68.56.217. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
Mikhelashvili et al.
J. Appl. Phys., Vol. 95, No. 2, 15 January 2004
615
FIG. 3. X-ray diffraction peak intensity for an 11 nm Er2 O3 film. Curve
1—as deposited, curve 2—T ann⫽550 °C and curve 3—T ann⫽750 °C.
FIG. 2. Plain view TEM image 共I兲 and x-ray diffraction pattern 共II兲 for a 50
nm Er2 O3 film. 共a兲 As deposited. 共b兲 Annealed at 750 °C.
2. X-ray diffraction, cross-sectional TEM
and AES profiles
Crystalline erbium oxide usually has cubic lattices. Experimental and calculated data are given in Ref. 21. In some
cases, however, hexagonal structures were observed.22 All
the known powder cubic erbium oxides in the diffraction
angle 共2␪兲, ranged in 20–50° interval, exhibit strong reflections for the 共222兲, 共400兲 and 共440兲 planes with 100%, 30%–
40% and 30%– 45% relative intensities, respectively.
The measurement of the XRD spectra of our Er2 O3 films
shows that they have similar textures with a 具111典 preferential orientation. In the as deposited state the films were found
to contain a mixture of amorphous and polycrystalline cubic
phases21 关see the plain view micrograph and the electron
diffraction pattern of an as deposited 50-nm-thick film in Fig.
2共a兲兴. This is consistent with data obtained for an Er2 O3 film
produced by oxidation of an erbium metal layer deposited on
Si17 and also with related materials: Gd2 O3 14 and Y2 O3 . 15
However, our findings are in some contradiction with the
data of Ref. 20, where the 具100典 Er2 O3 texture was found to
be the strongest among the rare-earth-metal oxide films on a
共100兲 Si substrate.
After annealing in an oxygen environment at 750 °C, the
structural texture does not change much with temperature but
the relative amount of the crystalline phase increases with
annealing temperature 关see Fig. 2共b兲兴. The diffraction peak
intensity was found to decrease slightly at high 共⬎550 °C兲
annealing temperature as seen in Fig. 3 for an 11 nm film.
Annealing at 750 °C reduces the amount of the crystalline
phase by 10%–15%. One reason for the XRD intensity decrease during high temperature annealing is a decrease in the
amount of Er2 O3 material due to crystalline oxide consumption during a reaction with the Si substrate and formation of
an amorphous erbium silicate glass layer and an interfacial
SiO2 or SiOx film. Similar results were obtained for La2 O3
and Er2 O3 films deposited on Si and annealed in nitrogen
ambient at 800 °C.20 The difference is that the reduction of
the crystalline material is detectable here at lower annealing
temperatures, about 550 °C, most likely due to the different
evaporation technique, film thickness and annealing conditions.
The AES depth profiles presented in Fig. 4 were obtained with a sputtering rate of ⬃0.02 nm/s. They show the
distribution of Er, O and Si with depth for as deposited 关Fig.
4共a兲兴 and annealed at 750 °C 关Fig. 4共b兲兴 structures. The depth
profile of the as deposited film includes only one plateau,
which reflects the distribution of Er and O atoms in the film.
For the annealed sample, an additional plateau is apparent
with Er, O and Si atoms together. It is seen from Fig. 4共b兲
and from the cross-sectional TEM image, Fig. 5, that a second three-component layer partially grows at the expense of
the crystallized Er2 O3 film near the silicon substrate. This is
the reason for the slight reduction of the XRD 共222兲 peak
intensity with annealing temperature seen in Fig. 3. The
cross-sectional TEM image shown in Fig. 5 is of a 6.5 nm
film annealed at 750 °C. Two areas are distinguishable above
the crystalline Si substrate: a highly transparent, ⬃3.5-nmthick amorphous region and a dark polycrystalline layer of
Er2 O3 . The influence of this interfacial layer on the electrical
characteristics of the MOS structure is discussed below.
B. Electrical characteristics
1. C – V characteristics
A series of C – V measurements was performed in order
to determine the effect of the Er2 O3 film thickness as well as
Downloaded 23 Apr 2006 to 132.68.56.217. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
616
J. Appl. Phys., Vol. 95, No. 2, 15 January 2004
Mikhelashvili et al.
FIG. 4. Atomic concentration of Er, O2 and Si measured using AES for a 6.5
nm Er2 O3 film. 共a兲 As deposited and 共b兲 annealing cases. T ann⫽750 °C.
the annealing temperature. First we use a moderate annealing
temperature ⫺550 °C and a varying film thickness; these results are shown in Fig. 6共a兲. Next we examined the effect of
annealing temperature on a 6.5-nm-thick film; these C – V
curves are shown in Fig. 6共b兲. From the C – V characteristics,
we extract the accumulation capacitance, which yields the
effective relative dielectric constant k eff . Extracted k eff values as a function of film thickness are shown in Fig. 7: as
deposited films are illustrated by squares and structures annealed at 750 °C by solid dots. For the as deposited case, k eff
increases from 6.8 共for the thinnest film兲 to 14. At ⬃50 nm,
k eff reaches a constant value. The maximum value of k eff is
comparable to that observed for other rare-earth-metal oxide
films on silicon.14,15,17,23–25 The annealing process lowers k eff
in all but the 100 nm films.
The film thickness and its k eff determine an EOT which
is described in Fig. 8共a兲 as a function of the physical thickness for as deposited and annealed at 750 °C structures. The
EOT was determined using the simple relation: t ox
FIG. 5. High-resolution cross-sectional TEM image of a 6.5-nm-thick Er2 O3
film. A magnified picture of the crystalline Er2 O3 phase is circled. T ann
⫽750 °C.
FIG. 6. C – V characteristics 共a兲 T ann⫽550 °C, curve 1—d Er2O3⫽4.5 nm,
curve 2—d Er2 O3 ⫽6.5 nm, curve 3—d Er2 O3 ⫽11 nm, curve 4—d Er2 O3
⫽50 nm, curve 5—d Er2 O3 ⫽100 nm; 共b兲 d Er2 O3 ⫽6.5, curve 1—as deposited,
curve 2—t ann⫽350 °C, curve 3—T ann⫽450 °C, curve 4—T ann⫽750 °C.
⫽k SiO2 d Er2 O3 /k eff . Figure 8共b兲 shows k eff and EOT as a function of the annealing temperature for a 4.5-nm-thick Er2 O3
film. We note that k eff decreases from 6.8 to 5.5 and, consequently, the EOT increases from 2.4 to 3.2 nm.
The various C – V characteristics suggest that the structures contain, in addition to the Er2 O3 film, some interfacial
FIG. 7. The term k eff as a function of film thickness. Experimental data: as
deposited films—squares, structures annealed at 750 °C—solid dots. Two
series capacitor model: curve 1—d SiO2 ⫽1 nm, curve 2—d SiO2 ⫽2.8 nm,
curve 3—d SiO2 ⫽5.5 nm.
Downloaded 23 Apr 2006 to 132.68.56.217. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
Mikhelashvili et al.
J. Appl. Phys., Vol. 95, No. 2, 15 January 2004
FIG. 8. 共a兲 Thickness dependence of EOT. 共b兲 Change of k eff and EOT with
annealing temperature for a 4.5-nm-thick Er2 O3 film.
layer with a lower dielectric constant, which acts as a series
capacitor reducing k eff . This interfacial layer affects not only
the k eff but also the J – V characteristics and mainly the leakage currents, described below.
FIG. 9. J – V characteristics of MOS structure. 共a兲 d Er2 O3 ⫽4.5 nm, 共b兲
d Er2 O3 ⫽11 nm. Curve 1—as deposited, curve 2—T ann⫽350 °C, curve 3—
T ann⫽450 °C, curve 4—T ann⫽750 °C. The circled areas show the breakdown regions.
617
FIG. 10. Dependence of leakage current density at E⫽1 MV/cm, curve 1,
and total E BD , curve 2, on 共a兲 annealing temperature for a 4.5 nm film and
共b兲 films thickness for T ann⫽750 °C. Curve 3—contribution of the SiO2
layer, curve 4-contribution of the Er2 O3 layer.
2. J – V characteristics
Measured J – V curves of an Er2 O3 MOS structure operating in the accumulation mode 共with a negatively biased top
electrode兲 are shown in Figs. 9共a兲 and 9共b兲, respectively, for
4.5-nm- and 11-nm-thick films. Curve 1 represents as deposited conditions and curves 2– 4 are for annealing temperatures of 350, 450 and 750 °C, respectively. We note the vast
leakage reduction in the annealed films, in particular above
450 °C. Also, every curve exhibits an abrupt change in current representing breakdown conditions.
Figure 10 summarizes two important properties: leakage
current at fixed electric field of 1 MV/cm 共curve 1兲 and
breakdown electric field E BD 共curve 2兲. Figure 10共a兲 describes the dependence on annealing temperature for a 4.5
nm film. The dependence on thickness for films annealed at
550 °C is shown in Fig. 10共b兲. The obtained leakage current
density strongly depends on the film thickness reaching a
value of 1⫻10⫺8 – 5⫻10⫺8 A/cm2 in the ultrathin films.
This is to be compared with the values achieved for 50 and
100 nm films, which are three to five orders of magnitude
larger. Similar dependencies were observed for Ta2 O5 , TiO2 ,
and Y2 O3 films, which were annealed in an oxygen
environment.1–3,7,11 As for E BD , we observe a large reduction, from 15 to 2 MV/cm when the film thickness increases
from 4.5 to 50 nm but beyond 50 nm, the changes become
insignificant.
IV. DISCUSSION
The reduction of EOT with film thickness 关see Fig. 8共a兲兴
together with the observed dependence of k eff and EOT in the
Downloaded 23 Apr 2006 to 132.68.56.217. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
618
J. Appl. Phys., Vol. 95, No. 2, 15 January 2004
4.5 nm films on annealing temperature 关see Fig. 8共b兲兴 suggest the existence of a SiO2 layer, the thickness of which can
be estimated by extrapolating Fig. 8共a兲 to zero thickness. The
predicted thickness, so obtained, is 1 and 2.8 –3 nm, for the
as deposited and annealed at 750 °C films, respectively.
These values are in the range estimated from the AES and
TEM data of the ultrathin annealed films, Figs. 4 and 5. The
thickness for as deposited films 共⬃1 nm兲 equals that of the
natural oxide formed on Si after reduced calcium aluminate
cleaning. The formation of the SiO2 film is due to the diffusion of oxygen atoms through the Er2 O3 film towards the Si
substrate. Obviously, higher temperatures and thinner films
result in a thicker SiO2 layer having a larger influence on the
EOT.
The annealing process also initiates a reaction between
Er2 O3 and the Si substrate20 forming an erbium–silicate
glass, mainly at high temperatures and for short annealing
times. The dielectric constant of this silicate is rather large,
10.8 –11.8 共assumed to be similar to yttrium–silicate11,26兲
and therefore it cannot be responsible for the reduction in
k eff .
A simple model which assumes two series capacitors,
formed by the erbium–silicate glass and the SiO2 layer was
used to quantify the influence of the interfacial layers on k eff
and EOT. The model assumed a silicate thickness of 1 nm
with k silicate⫽11.8, a dielectric constant for the Er2 O3 of
k Er2 O3 ⫽14 and a 2.8 nm SiO2 (k SiO2 ⫽3.82) layer. In the as
deposited case, the SiO2 layer was assumed to be 1 nm thick.
The calculated dependence on film thickness is shown as
solid lines in Fig. 7. Curve 1 describing the as deposited case
fits the data for the thickest films only while curve 2, representing structures annealed at 750 °C, fits nowhere. We also
used thicker SiO2 films 共2.8 nm for the as deposited case and
5.5 nm for the annealed film兲. These unrealistically thick
values 共which are more than twice the interfacial thickness
estimated from AES and TEM analysis兲 result in curve 3,
which fits the data in the thinnest films only.
We conclude, therefore, that the simple picture of two
series capacitors is insufficient to explain the reduction of k eff
and EOT with thickness and annealing temperature. Moreover, a low k eff value 共9 and 7, respectively, before and after
gas annealing at 400 °C兲 was reported in Refs. 14 and 15 for
ultrathin 共3.4 – 4.5 nm兲 Y2 O3 and Gd2 O3 films, deposited in
an ultrahigh vacuum system, which avoids the formation of
any interfacial SiO2 layer. It is likely, therefore, that a different mechanism is responsible for the low k eff values in thin
dielectric layers.
We postulate that an additional important issue, related
to the lattice buildup in the initial stages of the deposition,
plays a key role in determining the film properties. The initially grown layers may comprise some material in the ErO
共erbium monoxide兲 phase12 or any other nonstoichiometric
phase with lower polarizability than that of Er2 O3 . The annealing process transforms such layers to stoichiometric
Er2 O3 and at the same time it enhances the interfacial SiO2
film. These two processes affect k eff in opposite manners
yielding the annealing temperature dependence of k eff observed in Figs. 7 and 8共b兲. For 50 and 100 nm films, there is
no temperature dependence of k eff and EOT because the
Mikhelashvili et al.
thicknesses of both the SiO2 and the nonstoichiometric layers are negligible in comparison to the Er2 O3 thickness. Note
that no trace of ErO or any other crystalline phase was detected in the XRD measurement 共consistent with Ref. 12兲.
This may be due to the extremely small quantity of material
present in the films, which yields an XRD signal whose intensity is below the sensitivity of the present XRD system.
The expected shift of the flatband voltage V FB to more
positive values and the decrease in total charge density Q t as
the film thickness increases are observed in Fig. 6共a兲. At
d Er2 O3 ⫽100 nm, V FB reaches the value of ⫺0.75 V, ⫺0.5 V
and Q t ⫽1⫻1011 cm⫺2 , 5⫻1011 cm⫺2 . A monotonic shift in
V FB towards positive voltages and a reduction of the charge
density is also observed for 4.5–11 nm films as the annealing
temperature increases 关see Fig. 6共b兲兴. The values of V FB and
Q t change from ⫺1, ⫺2 V to ⫺0.55 V, ⫺0.7 V and Q t ⫽8
⫻1013 – 1⫻1010 cm⫺2 , respectively, for as deposited and annealed at 750 °C films. For thick films a relatively weak
change of the V FB and Q t occurs at ⫺0.9 V, ⫺0.8 V and 1
⫻1011 cm⫺2 , 3⫻1011 cm⫺2 , respectively.
The negative value of the flatband voltage denotes the
existence of the positive total charge in as deposited layers
for all thicknesses. Increases in thickness and annealing temperature cause lowering of their absolute values. These observations, which are consistent with Ref. 27, suggest that a
compensation mechanism between negative charges localized in the Er2 O3 and positive charges trapped in or near the
interfacial SiO2 film are taking place. The two charge densities depend on the respective thicknesses and the net charge
is accumulated in the Er2 O3 /SiO2 interface which as discussed below plays a major role in the drastic decrease of
leakages in annealed films.
Next we address the leakage reduction with decreasing
thickness and increased annealing temperature seen in Figs.
9 and 10. Here, the key contribution is from the interfacial
SiO2 film acting as an effective series insulator. As already
stated, the formation of this film is enhanced with annealing
temperature and a thickness reduction of the Er2 O3 film
through which the oxygen atoms can easily diffuse.
The dependence of E BD on film thickness 关curve 2 in
Fig. 10共b兲兴 can be correlated with the changes in film morphology described in Fig. 1. In the thick films, the breakdown process is not bulk limited but rather it depends on the
surface roughness, which increases with film thickness.
Similarly to what is discussed in Refs. 7, 11, 28, and 29 for
TiO2 , Ta2 O5 and Y2 O3 films, the nonuniform and rough surfaces are the main source for the so-called ‘‘weak points,’’
which act as local regions of high electric field where breakdown occurs. Moreover, these weak points enhance carrier
injection and further increase leakages in the thicker films.
The breakdown electric field decreases with increased film
thickness, but for every thickness it increases with annealing
temperature similarly to the 4.5 nm case shown in curve 2 of
Fig. 10共a兲. The increased temperature improves the film stoichiometry due to oxygen diffusion, an effect that balances
and even dominates over the increase of roughness, yielding
a small but consistent improvement in E BD with temperature.
The increase in E BD for the ultrathin films with annealing temperature 关see curve 2 in Fig. 10共b兲兴 is attributed to the
Downloaded 23 Apr 2006 to 132.68.56.217. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
Mikhelashvili et al.
J. Appl. Phys., Vol. 95, No. 2, 15 January 2004
high quality morphology and to the interfacial SiO2 film.29
To explain the results we calculated the distribution of the
applied electric field between Er2 O3 and the interfacial layers
using the charge balance equation7,29,30 for the Er2 O3 /SiO2
structure and using the data of curves 2 in Figs. 10共a兲 and
10共b兲. The calculation assumed a 1–2.8 nm interfacial layer
and k Er2 O3 ⫽14. Curves 3 and 4 in Figs. 10共a兲 and 10共b兲 show
the annealing temperature and thickness dependencies of the
contribution to E BD of the interfacial and Er2 O3 layers, respectively. The fact that curves 4 and 2 in Fig. 10共b兲 coincide
for large thickness values proves that the role of the interfacial layer in this range is negligible. Similarly, curves 3 and 2
in Fig. 10共b兲 coincide for the thinnest films from which we
conclude that the increase of E BD in annealed ultrathin films
is due to the formation of the interfacial layer and is only
remotely related to the Er2 O3 properties.
The series capacitance model7,28,30 can qualitatively explain the experimental observations but quantitative reasoning is lacking. According to curves 3 in Figs. 10共a兲 and
10共b兲, the 1–2.8 nm interfacial layers withstand an electric
field, which is as high as 2⫻107 V/cm. This is an unreasonable result since direct or Fowler–Nordheim31,32 tunneling is
likely to occur which is inconsistent with the low leakage
current density we have measured. At this level of the electric fields, experimental and calculated leakage current densities for 2–3-nm-thick SiO2 films are in the range of 1
⫻10⫺3 – 1⫻10⫺1 A/cm2 . 30–36 Curve 1 in Fig. 10 shows that
our leakage current density at a bias of 2–3 V is smaller than
10⫺5 A/cm2 共see also curves 4 and 5 in Fig. 9兲.
In order to explain this inconsistency we propose to invoke the carrier transport concept demonstrated for a
Ta2 O5 /SiO2 stack in Refs. 29 and 30 and also for a
TiO2 /SiO2 double layer combination in Refs. 7 and 37. The
leakage current in the ultrathin Er2 O3 and SiO2 films causes
a charge carrier accumulation at their interface limiting the
electric field across the SiO2 layer to a value lower than the
critical breakdown field. This charge barrier shifts the inflection points on the J – V characteristics of annealed ultrathin
films away from zero voltage as seen in Fig. 9. The influence
of the thin interfacial layer on the leakage current of thick
films is negligible and its reduction with annealing temperature is only due to the reduction of the oxygen vacancies.
V. CONCLUSION
In summary, we have described MOS structures based
on EBG evaporated erbium oxide as the gate dielectric. The
effects of annealing temperature on morphological, structural
and electrical characteristics of different thickness Er2 O3 layers were investigated. Ultrathin 共11 nm兲 films were found to
be smooth in contrast to thick 共larger than 50 nm兲 films,
which exhibit significant roughness, independent of annealing temperature. All as deposited and annealed film have a
similar texture with 具111典 as the dominant orientation, which
does not change with annealing conditions.
The relative dielectric constant for ultrathin films depends on annealing temperature and varies in the range of
5.5–9, while for thick films it is approximately 14. The main
reason for the moderately low k value is the formation during
619
annealing of a 1–2.8-nm-thick interfacial SiO2 film as well
as the presence of a thin region comprising low polarizability
inclusions of ErO in as deposited and low temperature annealed films. The minimum value of EOT we obtained was
2.4 –2.7 nm. This moderate value is higher than several published high k dielectrics but achieving an ultralow EOT was
not the major purpose of this study. C – V curves for ultrathin
film yield a flatband voltage shift to more positive values
together with a significant lowering of the total positive
charge down to the 1⫻1010 cm⫺2 . An increase in leakage
current density with Er2 O3 thickness and its reduction with
annealing temperature was established. The lowest leakage
current density we obtained for ultrathin films was 1
⫻10⫺8 – 5⫻10⫺8 A/cm2 at E⫽1 MV/cm and the maximum
total breakdown electric fields were ⬃15–20 MV/cm. The
dependence of the leakage current density and breakdown
electric field on annealing temperature is qualitatively explained by the reduction of the oxygen vacancies, the
buildup of an interfacial SiO2 film and formation of a barrier
for carriers at the Er2 O3 /SiO2 boundary.
1
C. Chaneliere, J. L. Autran, R. A. B. Devine, and B. Balland, Mater. Sci.
Eng., R. 22, 269 共1998兲.
2
P. C. Joshi and M. V. Cole, J. Appl. Phys. 86, 871 共1999兲.
3
V. Mikhelashvili and G. Eisenstein, Appl. Phys. Lett. 75, 2836 共1999兲.
4
H. Tang, K. Prasad, R. Sanjines, P. G. Schmid, and F. Levy, J. Appl. Phys.
75, 2042 共1994兲.
5
J. Yan, D. C. Gilmer, S. A. Campbel, W. L. Gladfelter, and P. G. Schmid,
J. Vac. Sci. Technol. B 14, 1706 共1996兲.
6
H. S. Kim, S. A. Campbel, D. S. Gilmer, V. Kaushik, J. Conner, L. Prabhu,
and A. Anderson, J. Appl. Phys. 85, 3278 共1999兲.
7
V. Mikhelashvili and G. Eisenstein, J. Appl. Phys. 89, 3257 共2001兲.
8
J. Robertson, J. Vac. Sci. Technol. B 18, 1785 共2000兲.
9
H. Ono, Y. Hokosawa, T. Ikarashi, K. Shinoda, N. Ikarashi, K. Koyanag,
and H. Yamaguchi, J. Appl. Phys. 89, 995 共2001兲.
10
Y. S. Kim, Y. H. Lee, K. M. Lim, and Y. Sung, Appl. Phys. Lett. 74, 2800
共1999兲.
11
A. C. Rastogi and N. Sharma, J. Appl. Phys. 71, 5041 共1992兲.
12
C. H. Ling, J. Bhaskaran, W. K. Choi, and L. K. Ah, J. Appl. Phys. 77,
6350 共1995兲.
13
V. Mikhelashvili, Y. Betzer, I. Prudnikov, M. Orenshtein, D. Ritter, and G.
Eisenstein, J. Appl. Phys. 84, 6747 共1998兲.
14
J. Kwo, M. Hong, A. R. Kortan, K. L. Queeney, Y. J. Chabal, P. Mannaerts, T. Boone, J. J. Krajewski, A. M. Sergent, and J. M. Rosamilia,
Appl. Phys. Lett. 77, 130 共2000兲.
15
J. Kwo, M. Hong, A. R. Kortan, K. L. Queeney, Y. J. Chabal, R. L. Opila,
Jr., D. A. Muller, S. N. G. Chu, B. J. Sapjeta, T. S. Lay, J. P. Mannaerts, T.
Boone, H. W. Krautter, J. J. Krajewski, A. M. Sergent, and J. M. Rosamilia, J. Appl. Phys. 89, 3920 共2001兲.
16
M. Hong, A. R. Kortan, K. L. J. Kwo, P. Mannaerts, J. J. Krajewski, Z. H.
Lu, K. C. Hseih, and K. J. Cheng, J. Vac. Sci. Technol. B 18, 1688 共2000兲.
17
T. S. Kalkur and Y. C. Lu, Thin Solid Films 188, 203 共1990兲.
18
V. Mikhelashvili, G. Eisenstein, and F. Edelmann, J. Appl. Phys. 90, 5447
共2001兲.
19
V. Mikhelashvili, G. Eisenstein, and F. Edelmann, Appl. Phys. Lett. 80,
2156 共2002兲.
20
H. Ono and T. Katsumata, Appl. Phys. Lett. 78, 1831 共2001兲.
21
Powder Diffraction Files: 0.8-0050, 26-0604, 43-1007, 74-1830, 74-1983,
76-0159, 77-0459/0464, 77-0777.
22
Powder Diffraction File 19-0452.
23
U. Saxena and Srivastava, Thin Solid Films 33, 185 共1976兲.
24
N. W. Grimes and R. W. Grimes, J. Phys.: Condens. Matter 10, 3029
共1998兲.
25
D. Xue, K. Betzler, and H. Hesse, J. Phys.: Condens. Matter 12, 3113
共2000兲.
26
M. Gurvitch, L. Manchanda, and J. M. Gibson, Appl. Phys. Lett. 51, 919
共1987兲.
Downloaded 23 Apr 2006 to 132.68.56.217. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
620
27
Mikhelashvili et al.
J. Appl. Phys., Vol. 95, No. 2, 15 January 2004
S. Seki, T. Unagami, and B. Tsujiyama, J. Electrochem. Soc. 132, 199
共1985兲.
28
C. F. Chen, C. Y. Wu, M. K. Lee, and C. N. Chen, IEEE Trans. Electron
Devices 34, 1540 共1987兲.
29
Y. Nishioka, S. Kimura, H. Shinriki, and K. Mukai, J. Electrochem. Soc.
134, 410 共1987兲.
30
Y. Nishioka, H. Shinriki, and K. Mukai, J. Appl. Phys. 61, 2335 共1987兲.
31
J. G. Simmons, J. Appl. Phys. 34, 1793 共1963兲.
32
M. Depas, B. Vermeire, P. W. Mertens, R. L. Van Meirhaeghe, and M. M.
Heyns, Solid-State Electron. 38, 1465 共1995兲.
N. Yang and J. J. Wortman, Microelectron. Reliab. 41, 37 共2001兲.
34
J. Zhang, J. S. Yuan, Y. Ma, and A. S. Oates, Solid-State Electron. 44,
2165 共2000兲.
35
N. Yang, W. Kirklen, J. R. Hauser, and J. J. Wortman, IEEE Trans. Electron Devices 46, 1464 共1999兲.
36
E. Rosenbaum and J. Wu, Microelectron. Reliab. 41, 625 共2001兲.
37
B. H. Lee, Y. Jeon, K. Zavadzki, W. Qi, and J. Lee, Appl. Phys. Lett. 74,
3143 共1999兲.
33
Downloaded 23 Apr 2006 to 132.68.56.217. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
Download