Determination of energy barrier profiles for high-k dielectric materials utilizing bias-dependent internal photoemission

Article abstract of DOI:10.1063/1.1812831

We utilize bias-dependent internal photoemission spectroscopy to determine the metal/dielectric/silicon energy barrier profiles for Au/HfO₂/Si and Au/Al₂O₃/Si structures. The results indicate that the applied voltage plays a large role in determining the effective barrier height and we attribute much of the variation in this case to image potential barrier lowering in measurements of single layers. By measuring current at both positive and negative voltages, we are able to measure the band offsets from Si and also to determine the flatband voltage and the barrier asymmetry at 0 V. Our SiO₂ calibration sample yielded a conduction band offset value of 3.03±0.1 eV. Measurements on HfO₂ give a conduction band offset value of 2.7±0.2 eV (at 1.0 V) and Al₂O₃ gives an offset of 3.3±0.1 (at 1.0 V). We believe that interfacial SiO₂ layers may dominate the electron transport from silicon for these films. The Au/HfO₂ barrier height was found to be 3.6±0.1 eV while the Au/Al₂O₃ barrier is 3.5±0.1 eV.

Full text of DOI:10.1063/1.1812831

Determination of energy barrier profiles for high- k dielectric materials utilizing bias-
dependent internal photoemission
Julie Casperson Brewer, Robert J. Walters, L. Douglas Bell, Damon B. Farmer, Roy G. Gordon, and Harry A.
Atwater
Citation: Applied Physics Letters 85, 4133 (2004); doi: 10.1063/1.1812831
View online: http://dx.doi.org/10.1063/1.1812831
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/85/18?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Graphene as transparent electrode for direct observation of hole photoemission from silicon to oxide
Appl. Phys. Lett. 102, 123106 (2013); 10.1063/1.4796169
Energy barriers at interfaces between (100) In x Ga 1 − x As ( 0 ≤ x ≤ 0.53 ) and atomic-layer deposited Al 2 O 3
and HfO 2
Appl. Phys. Lett. 94, 202110 (2009); 10.1063/1.3137187
Oxygen defects and Fermi level location in metal-hafnium oxide-silicon structures
Appl. Phys. Lett. 87, 072902 (2005); 10.1063/1.2011791
Band alignment at the interface of ( 100 ) Si with Hf x Ta 1 − x O y high- κ dielectric layers
Appl. Phys. Lett. 86, 072108 (2005); 10.1063/1.1866640
Electrical conduction and band offsets in Si/Hf x Ti 1−x O 2 / metal structures
J. Appl. Phys. 95, 7936 (2004); 10.1063/1.1703821
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 141.210.2.78
On: Tue, 25 Nov 2014 01:58:57

APPLIED PHYSICS LETTERS
VOLUME 85, NUMBER 18
1 NOVEMBER 2004
Determination of energy barrier profiles for high-k dielectric materials
utilizing bias-dependent internal photoemission
Julie Casperson Brewer and Robert J. Walters
Thomas J. Watson Laboratory of Applied Physics, California Institute of Technology,
Pasadena, California 91125
L. Douglas Bella)
Jet Propulsion Laboratory, Pasadena, California 91109
Damon B. Farmer and Roy G. Gordon
Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02318
Harry A. Atwater
Thomas J. Watson Laboratory of Applied Physics, California Institute of Technology,
Pasadena, California 91125
(Received 15 April 2004; accepted 14 September 2004)
We utilize bias-dependent internal photoemission spectroscopy to determine the metal/dielectric/
silicon energy barrier profiles for Au/HfO2/Si and Au/Al2O3/Si structures. The results indicate that
the applied voltage plays a large role in determining the effective barrier height and we attribute
much of the variation in this case to image potential barrier lowering in measurements of single
layers. By measuring current at both positive and negative voltages, we are able to measure the band
offsets from Si and also to determine the flatband voltage and the barrier asymmetry at 0 V. Our
SiO2 calibration sample yielded a conduction band offset value of 3.03±0.1 eV. Measurements on
HfO2 give a conduction band offset value of 2.7±0.2 eV (at 1.0 V) and Al2O3 gives an offset of
3.3±0.1 (at 1.0 V). We believe that interfacial SiO2 layers may dominate the electron transport from
silicon for these films. The Au/HfO2 barrier height was found to be 3.6±0.1 eV while the
Au/Al2O3 barrier is 3.5±0.1 eV. © 2004 American Institute of Physics. [DOI: 10.1063/1.1812831]
The study of high-k dielectrics as a replacement for SiO2
in complementary metal-oxide-semiconductor devices has
become a field of enormous interest.1–3 In this letter, we in-
vestigate the band-offset characteristics of high-k dielectrics
on silicon. We utilize internal photoemission spectroscopy, a
simple optical method developed in the 1960s,4,5 which has
seen recent renewed interest in order to gain information
about barrier heights, trap states and interface dipoles in
high-k dielectrics.6 In this technique, a bias is applied across
a dielectric structure, while tunable monochromatic light
shines on the sample. At a threshold photon energy, electrons
from the substrate (or metal gate) are excited by internal
photoemission over the dielectric barrier.7 This threshold en-
ergy corresponds to the barrier height of the dielectric. Using
bias-dependent internal photoemission spectroscopy, we
have determined a barrier height profile as a function of volt-
age. By measuring the barrier height at both positive and
negative voltages, band offsets with respect to silicon (and
also the metal gate) can be determined in addition to the
flatband voltage and barrier asymmetry at 0 V.
In our experimental system, we utilize a 1000 W Hg–Xe
lamp with a monochromator as our light source. We use a
voltage source/femtoammeter to apply the bias across the
sample and to measure the current at each bias. The system is
computer controlled by LabView so that the light can be
scanned between 1 and 6 eV at any bias and photon energy
step size. A multifunction optical meter is used to determine
the lamp output spectrum to normalize the photoemission
yield. Fused silica lenses are used to focus the light onto the
top gold contact of the sample, which is held vertically.
The dielectric samples are grown on degenerately phos-
phorous doped n-type silicon to minimize the voltage drop
across the depletion region in the silicon, and enhance the
accuracy of our measurement. The dielectrics presented in
this letter are HfO2 and Al2O3 grown by atomic layer
deposition.8,9 Before deposition, samples were dipped in a
5% HF solution for 30 s followed by a 3 min UV/ozone
cleaning. Al2O3 films were grown using de-ionized water (DI
H2O) and trimethylaluminum ͑Al͓CH3͔3͒, while HfO2 films
were grown using DI H2O and tetrakis(diethylami-
do)hafnium ͑Hf͓NEt2͔4͒. Nitrogen was used as the carrier
gas, and the deposition temperature was 225 °C. The top
electrode is a 12 nm layer of evaporated gold, which is suf-
ficiently transparent so that the light source can photoexcite
carriers in the silicon. The back contact is indium.
As a calibration for our experimental system, we ana-
lyzed a 15 nm thermally grown SiO2 film deposited an n
-Si (1–10 ⍀ cm doping). The photocurrent was measured as
the photon energy was scanned at many voltages between
−10 and 10 V. Figure 1 shows the raw photocurrent versus
photon energy spectra for a variety of voltages (from
−0.5 to +3.0 V). It is evident from these plots that the sign
of the photocurrent depends strongly on the applied voltage,
and the peaks in photocurrent correspond with the peaks in
Hg–Xe lamp output (see Fig. 1 inset). For voltages from
−10 V up to +0.7 V, negative currents were observed. At
+0.7 V, the photocurrent switched signs and was positive for
all higher positive voltages. We suggest that for positive pho-
tocurrents, collected electrons originate in the silicon, while
the negative photocurrents indicate that electrons are mainly
being generated in the metal gate contact. We determined
a)Author to whom correspondence should be addressed; electronic mail:
lbell@pop.jpl.nasa.gov
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 141.210.2.78
0003-6951/2004/85(18)/4133/3/$22.00 4133 © 2004 American Institute of Physics
On: Tue, 25 Nov 2014 01:58:57

4134
Appl. Phys. Lett., Vol. 85, No. 18, 1 November 2004
Brewer et al.
FIG. 3. Barrier height profile as function of voltage for 15 nm SiO2 on
n-Si. The dotted line indicates the voltage at which the current switches
sign. The square dots on the left of the dotted line are extracted from the
yield1/2 curves and indicate the barrier for electrons from the metal. The
triangles on the right are the barrier heights extracted from the yield2/5
curves. The circles on the right are from the yield1/3 curves. The triangles
and circles indicate the barrier for electrons coming from the silicon sub-
strate. This simulation takes into account image potential barrier lowering.
FIG. 1. Current through 15 nm SiO2 film as a function of incident photon
energy. Inset: output spectrum of Hg–Xe light source.
0.7 V to be the voltage where the currents from the metal
matched those coming from the semiconductor. We expect
this voltage to be close to flatband for single dielectric layers.
The next analysis step is to calculate the photocurrent
yield
lyzing photocurrent from the valence band of
a
semiconductor.14 In order to most thoroughly report the rel-
evant results, we have computed offsets based on models
assuming both the 2/5 power and the cube root of the yield
for these situations.15
I · ប␻
Y =
,
͑1͒
P
After extracting band offsets for each voltage, we obtain
a barrier height profile as a function of voltage, as can be
seen in Fig. 3 for 15 nm SiO2. The points on the left of the
vertical dotted line are for electrons emitted from the metal,
the points on the right are for electrons emitted from the
semiconductor. As can be seen, the band offset varies greatly
with applied voltage, and this illustrates that it is of the ut-
most importance to report a corresponding bias voltage as-
sociated with a measured band offset. We report our band
offsets as the point nearest to flat band, where electrons are
coming from the semiconductor, or in the case of our SiO2
film, 4.13±0.1 eV. When we subtract the 1.1 eV SiO2 band
gap, we find that the Si/SiO2 conduction band offset is
3.03±0.1 eV. Ultimately, our results for SiO2 fit well to what
is expected for SiO2 films with image potential barrier
lowering.11 The image potential barrier lowering simulations
are shown by the black lines and can be represented by the
equation (in energy units)
where I is measured current in amperes, P is the absorbed
light power in watts, ប␻ is the photon energy in eV, and Y is
the yield in electrons/photon. Each individual current versus
photon energy curve is divided by the incident photon energy
spectrum. The square root, cube root, or 2/5 power of the
yield is then plotted versus photon energy as is shown in Fig.
2 for the 15 nm SiO2 sample. The x intercept is then ex-
tracted and is reported as the band offset relative to the va-
lence band of silicon. We assume that most electrons are
emitted from the valence band since the number of filled
initial states is much higher than in the conduction band. The
literature is in general agreement that the square root is the
appropriate power for intercept extraction when considering
electrons emitted from a metal.10,11 The case where electrons
are emitted from the semiconductor is not well resolved.
Semiclassical calculations suggest that taking a 2/5 power of
the yield is correct12 and a quantum mechanical correction to
the theory predicts that the cube root is correct13 when ana-
Vi = − 1.15␭s2/x͑s − x͒,
where
␭ = e2 ln 2/8␲␧s,
͑2͒
͑3͒
e is the electron charge, and s and ␧ are the thickness and
dielectric constant of the insulating layer. The distance of the
electron from the first (source) electrode is x. The method for
approximating image barrier lowering is described in Ref.
16. The image force was approximated as that of an insulator
between two metallic electrodes. In general, the silicon sub-
strates used in our experiments are highly doped, so this is a
reasonable approximation although the thin depletion layer
in the Si will somewhat modify the result. For SiO2, we
assumed an optical dielectric constant of 2.5.
Similar analyses were completed for Al2O3 and HfO2
FIG. 2. Yield to the 1/3, 2/5, and 1/2 power. A linear fit to these curves is
extrapolated to the x axis to provide a numerical value for the barrier height.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 141.210.2.78
films on n-Si ͑0.0001 ⍀ cm͒. The resulting barrier height
On: Tue, 25 Nov 2014 01:58:57

Appl. Phys. Lett., Vol. 85, No. 18, 1 November 2004
Brewer et al.
4135
Similar analysis can be done for HfO2 (open symbols in
Fig. 4), but because of the degraded data quality attributed to
leakage from the substrate, it is more difficult to verify our
data by simulation. The dashed curve indicates a barrier low-
ering simulation for the HfO2 barrier ͑HfO2 ␬=22,SiOx ␬
=3.9͒. The data quality is good when the electrons originate
from the metal (from −0.5 to 0.9 eV) and we can determine
a Au/HfO2 barrier height of 3.6±0.1 eV. Using the same
reasoning as for Al2O3, based on our data, our best approxi-
mation for the Si/HfO2 barrier height (from valence band) is
3.8 eV±0.2 eV. This corresponds to a conduction band off-
set with respect to Si of 2.7±0.2 eV. The experimental lit-
erature reports conduction band offsets of 2.0 and ϳ1.2 eV
for HfO2, while theoretical calculations predict 1.5 eV.18–20
Further effects of barrier lowering in HfO2/Al2O3 hetero-
structures will be addressed in a paper to be published soon
(see Ref. 21 for background).
The high-␬ field has been plagued by the growth of un-
wanted interfacial layers at the dielectric/silicon interface.
Progress is being made to decrease these layers by careful
cleaning and passivation of the silicon surface before growth,
but eliminating these layers completely remains a great
challenge.22 Although interfacial layers will in general
modify the measurements, by careful use of our barrier pro-
file technique, and by selecting the barrier height at specific
voltages (and thus specific band alignment) we can gain a
very good estimate of the band offset of any barrier. We have
shown that profiling the barrier heights of dielectrics on sili-
con as a function of applied bias is a very viable and valu-
able technique for determining the effective band offset at
any particular voltage.
FIG. 4. Barrier height profile for HfO2 and Al2O3 on n+-Si. The data and
barrier height simulation curves on the left hand side of the vertical dashed
line represent negative photocurrents. The data and simulation curves on the
right hand side represent positive photocurrents. The open symbols and
dashed lines are for HfO2 while the solid symbols and solid lines are for
Al2O3. Squares indicate data extracted from yield1/2 vs energy curves. Tri-
angles correspond to yield2/5 data while circles correspond to yield1/3 data.
This simulation considers the maximum barrier height at each voltage and
the depletion region in the silicon substrate.
profiles and barrier height simulations are shown in Fig. 4.
The barrier height profiles for these two materials are not
nearly as clear as for SiO2. This is particularly true for
HfO2—a result of leakage through the barrier and greater
difficulty in extracting band offsets from the yield curves.
Transmission electron microscopy analysis revealed that in-
terfacial SiO2 layers 2.2 nm thick were present between the
HfO2 and Al2O3 dielectric and semiconductor. These interfa-
cial layers could be attributed to the UV/ozone clean during
the substrate preparation or a postdeposition 600 °C anneal
in Ar+2000 ppm O2, and could account for the higher mea-
sured band offsets compared with literature values, though
the electrical characteristics of these layers are unknown.
Because of the presence of interfacial layers in the Al2O3
and HfO2 samples, the analysis for these films is slightly
more complicated than for SiO2. The measured barrier height
(from the Si valence band) for 16.1 nm Al2O3 is 4.6±0.1 eV
(at 1.0 V). After subtracting the Si band gap ͑1.1 eV͒, the
Al2O3 conduction band offset is found to be 3.5±0.1 eV (at
1.0 V), but this probably corresponds most directly to the
interfacial layer than to the Al2O3 layer itself. For this rea-
son, another quantity of interest is the Au/Al2O3 barrier,
which is observed to be 3.5±0.1 eV. There is a discontinuity
at the point at which the originating carrier electrode changes
from the metal to the silicon (at the vertical dotted line).
Using a consistent set of parameters ͑Al2O3 ␬=9,SiOx ␬
=3.9͒, this asymmetry can be understood to first order if we
consider the charge that is generated in the metal compared
with those originating in the silicon for the
Au/Al2O3/SiOx/Si barrier. The results for this simulation
(accounting for Si depletion) are shown by the solid line in
Fig. 4. The absence of the slope between 0 and 0.9 V is not
well understood, but the fact that we can simulate the general
shape of the profile and accurately approximate the barrier
heights for electrons coming from each electrode is very en-
couraging. The experimental literature reports conduction
1J. Kwo, M. Hong, B. Busch, D. A. Muller, Y. J. Chabal, A. R. Kortan, J.
P. Mannaerts, B. Yang, P. Ye, H. Gossmann, A. M. Sergent, K. K. Ng, J.
Bude, W. H. Schulte, E. Garfunkel, and T. Gustafsson, J. Cryst. Growth
251, 645 (2003).
2G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys. 89, 5243
(2001).
3H. R. Huff, A. Hou, C. Lim, Y. Kim, J. Barnett, G. Bersuker, G. A. Brown,
C. D. Young, P. M. Zeitzoff, J. Gutt, P. Lysaght, M. I. Gardner, and R. W.
Murto, Microelectron. Eng. 69, 152 (2003).
4R. J. Powell, J. Appl. Phys. 41, 2424 (1970).
5E. O. Kane, Phys. Rev. 127, 131 (1962).
6V. V. Afanas’ev, M. Houssa, A. Stesmans, and M. M. Heyns, Appl. Phys.
Lett. 78, 3073 (2001).
7V. K. Adamchuk and V. V. Afanas’ev, Prog. Surf. Sci. 41, 111 (1992).
8D. M. Hausmann, E. Kim, J. Becker, and R. G. Gordon, Chem. Mater. 14,
4350 (2002).
9M. D. Groner, J. W. Elam, F. H. Fabreguette, and S. M. George, Thin
Solid Films 413, 186 (2002).
10R. H. Fowler, Phys. Rev. 38, 45 (1931).
11S. M. Sze, Physics of Semiconductor Devices (Wiley, New York, 1981).
12E. O. Kane, Phys. Rev. 127, 131 (1962).
13I.-S. Chen, T. N. Jackson, and C. R. Wronski, J. Appl. Phys. 79, 8470
(1996).
14The cube root and 2/5 power laws do not apply to emission from the
conduction band.
15V. V. Afanas’ev, M. Houssa, A. Stesmans, and M. M. Heyns, J. Appl.
Phys. 91, 3079 (2002).
16J. G. Simmons, J. Appl. Phys. 34, 1793 (1963).
17R. Ludeke, M. T. Cuberes, and E. Cartier, Appl. Phys. Lett. 76, 2886
(2000).
18V. V. Afanas’ev, A. Stesmans, F. Chen, X. Shi, and S. A. Campbell, Appl.
Phys. Lett. 81, 1053 (2002).
19S. Sayan, E. Garfunkel, and S. Suzer, Appl. Phys. Lett. 80, 2135 (2002).
20J. Robertson, J. Vac. Sci. Technol. B 18, 1785 (2000).
21J. D. Casperson, L. D. Bell, and H. A. Atwater, J. Appl. Phys. 92, 261
(2002).
22K. Choi, H. Harris, S. Gangopadhyay, and H. Temkin, J. Vac. Sci. Tech-
band offsets of 2.78 and 2.15 eV for Al2O3, while theoretical
6,17
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 141.210.2.78
calculations predict 2.8 eV.
nol. A 21, 718 (2003).
On: Tue, 25 Nov 2014 01:58:57