Abstract
ROBERT SHAWN JOHNSON. Properties of Aluminum Oxide and Aluminum Oxide Alloys and their Interfaces with Silicon and Silicon Dioxide. (Under the direction of Gerald Lucovsky.)
A remote plasma enhanced chemical vapor deposition method, RPECVD,
was utilized to deposit thin films of aluminum oxide, tantalum oxide, tantalum aluminates, and hafnium aluminates. These films were analyzed using auger electron spectroscopy, AES, Fourier transform infrared spectroscopy, FTIR, X- ray diffraction, XRD, nuclear resonance profiling, NRP, capacitance versus voltage, C-V, and current versus voltage, J-V.
FTIR indicated the alloys were homogeneous and pseudobinary in
character. Combined with XRD the crystallization temperatures for films >100 nm were measured. The alloys displayed an increased temperature stability with the crystallization points being raise by >100ºC above the end point values.
In-situ AES analysis provided a study of the initial formation of the films'
interface with the silicon substrate. For Al2O3 these results were correlated to
NRP results and indicated a thin, ~0.6 nm, interfacial layer formed during
deposition.
C-V characteristics indicated a layer of fixed negative charge associated
with Al2O3. For Ta2O5 the C-V and J-V results displayed high levels of leakage
current, due to a low conduction band offset with silicon. Both aluminates were
dominated by electron trapping states. These states were determined to be due
to (i) a network "break-up" component and (ii) localized atomic d-states of
hafnium and tantalum atoms.
PROPERTIES OF ALUMINUM OXIDE AND ALUMINUM OXIDE ALLOYS AND THEIR INTERFACES WITH SILICON AND SILICON DIOXIDE
by ROBERT SHAWN JOHNSON
A dissertation submitted to the Graduate Faculty of North Carolina State University in partial fulfillment of the requirements for the Degree of Doctor of Philosophy
PHYSICS
Raleigh, North Carolina
2001
Biography
Robert Shawn Johnson was born in Saranac Lake, NY on April 1st, 1974.
He attended the AuSable Valley Central School system and received his high school diploma in June of 1992. From there he enrolled in the State University of
New York College at Geneseo. He majored in physics with a minor in mathematics. While at Geneseo he contributed the operation and maintenance of a 2 MeV Vande Graff accelerator, on which he performed Rutherford backscattering experiments. He received his Bachelor of Arts in Physics in May of 1996.
After graduating he moved to Raleigh, NC where he attended North
Carolina State University. At NC State he studied Physics and worked as a research assistant for Dr. Gerald Lucovsky. While under the direction of Dr.
Lucovsky he studied replacement materials for silicon dioxide in CMOS applications. In December of 2001, he received his Ph.D. from NC State.
ii Acknowledgements
I would like to thank all the members of my committee, Dr. Aspnes, Dr.
Lucovsky, Dr. Misra, and Dr. Parsons, for their insightful views and comments during my tenure at NC State.
Thank you to my advisor, Dr. Lucovsky, whose experience and guidance has provided a wonderful path to this degree. In addition to his knowledge in the field of semiconductors, I would like to thank him for his in-sights into the subtle area of life.
To the members of the Lucovsky research group, Dr. Doug Stephens, Dr.,
Bruce Hinds, Dr. Dave Wolfe, Dr. Fuchao Wang, Dr. Kwangok Koh, Dr. Hanyang
Yang, Dr. Yider Wu, Dr. Bruce Claflin, Brian Solazo, Dr. Hiroaki Niimi, Bruce
Rayner, Dr. Robert Therrien, Dr. Donghun Kang, Michael Schultz, Klaus Flock,
Yi-Mu Lee, Choelhwyi Bae, Joon Goo Hong and Christopher Hinkle, it has been an interesting and enjoyable experience. Thank you all for the many times you placed your research to the side and aided in mine.
I am grateful to all the members of the cleanroom. Thank you for all the hours you spent trying to educate me in the ways of device processing.
Finally, I would like to thank my family. Especially my wife, Christine, and daughter, Irene, two best reasons to complete this degree and move forward.
And a big thank you to the New York fan base, the Johnson's and the Phillips'
(sometimes called the Case's).
iii Table of Contents
Page LIST OF TABLES. vi
LIST OF FIGURES. vii
1 INTRODUCTION. 1
1.1 The Need for High Dielectric Constant Materials. 1 1.2 General Properties of High Dielectric Constant Materials. 2 1.3 Objective. 4 1.4 Overview of the Dissertation. 4 1.5 References. 6
2 EXPERIMENTAL METHODS. 11
2.1 Materials Deposition. 11 2.1.1 Remote Plasma Enchanced Chemical Vapor Deposition, RPECVD. 11 2.1.2 Metal Organic Bubbler. 12 2.2 Auger Electron Spectroscopy, AES. 13 2.3 Nuclear Resonance Profiling, NRP. 14 2.4 Fourier Transform Infrared Spectroscopy, FTIR. 15 2.5 X-Ray Diffraction, XRD. 15 2.6 Electrical Characterization - Capacitance and Current versus Voltage, C-V and J-V. 16 2.7 References. 17
3 PROPERTIES OF RPECVD ALUMINUM OXIDE. 21
3.1 Introduction. 21 3.2 Interface with silicon and silicon dioxide. 22 3.3 Bulk properties, FTIR and XRD. 24 3.4 Electrical Characterization. 25 3.5 Conclusions. 30 3.6 References. 31
iv Page
4 PROPERTIES OF RPECVD TANTALUM OXIDE. 45
4.1 Introduction. 45 4.2 Interfacial Formation on HF-Last Silicon and Preoxidized Silicon. 45 4.3 FTIR Results. 47 4.4 Electrical Characterization. 47 4.5 Conclusions. 49 4.6 References. 50
5 PROPERTIES OF TANTALUM AND HAFNIUM ALUMINATES. 59
5.1 Introduction. 59 5.2 Bulk Composition and Properties by AES, RBS, and FTIR. 60 5.3 Interface Formation by AES. 61 5.4 Electrical Characterization. 62 5.5 Discussion. 64 5.6 Conclusions. 69 5.7 References. 71
6 SUMMARY AND FUTURE WORK. 91
6.1 Aluminum and Tantalum Oxide. 91 6.1.1 Aluminum Oxide, Al2O3.91 6.1.2 Tantalum Oxide, Ta2O5.92 6.2 Tantalum and Hafnium Aluminates. 92 6.3 Future Work. 93
v List Of Tables
Page
Table 1.1 Local electronic and physical properties of selected high-k materials. 8
Table 2.1 Remote plasma process conditions. 18
vi List Of Figures
Page
Figure 1.1 General molecular orbital diagram for oxides of transition metals. 9
Figure 1.2 Plot of oxygen coordination as a function of average bond ionicity, Ib.10
Figure 2.1 RPECVD glas flow and bubbler assembly for all depositions. 19
Figure 2.2 RPECVD chamber for depositions. 20
Figure 3.1 AES of aluminum oxide deposition on HF-last silicon and 0.6nm of silicon dioxide on HF-last silicon. 33
Figure 3.2 NRP for Al2O3 on HF-last silicon. 34
Figure 3.3 NRP for Al2O3 on 2.2nm of silicon dioxide deposited on HF-last silicon. 35
Figure 3.4 FTIR for >100nm of aluminum oxide on silicon before and after a 900ºC anneal. 36
Figure 3.5 X-Ray Diffraction, XRD, for aluminum oxide on Si(100). 37
Figure 3.6 C-V data for aluminum oxide on HF-last silicon. 38
Figure 3.7 Flatband voltage for Al2O3 on HF-last silicon as a function of the aluminum oxide EOT. 39
Figure 3.8 Flatband voltage for stacked Al2O3 and SiO2 MOSCAP’s as a function of the aluminum oxide EOT. 40
Figure 3.9 Leakage current density versus VG-VFB.41
Figure 3.10 Ideal band bending for n- and p-MOS devices. 42
Figure 3.11 Temperature Dependence of C-V for n-MOS Al2O3 and SiO2.43
vii Page
Figure 3.12 Temperature dependence of J-V for Al2O3 on HF-last silicon. 44
Figure 4.1 AES of tantalum oxide deposition on HF-last silicon and 0.6nm of silicon dioxide on HF-last silicon. 52
Figure 4.2 FTIR for >100nm of tantalum oxide on silicon before and after an 800°C anneal. 53
Figure 4.3 Room temperature C-V for Ta2O5 on pre-oxidized silicon. 54
Figure 4.4 1 MHz Temperature Dependence C-V for n-MOS Ta2O5 and SiO2.55
Figure 4.5 J-V for Ta2O5 on pre-oxidized silicon. 56
Figure 4.6 Temperature dependence of J-V for Ta2O5 on pre-oxidized silicon. 57
Figure 4.7 J vs. 1/T for n-MOS Ta2O5 device at VG-VFB = 0.4 V. 58
Figure 5.1 Differential AES spectra for the Al2O3-Ta2O5 alloy system. 73
Figure 5.2 Fourier transform infrared spectroscopy, FTIR, for as- deposited Ta-aluminates and their end-members, Al2O3 and Ta2O5.74
Figure 5.3 FTIR spectra of crystalline Al2O3-Ta2O5 alloy films. 75
Figure 5.4 Fourier transform infrared spectroscopy, FTIR, for as- deposited Hf-aluminates and their end-members, Al2O3 and HfO2.76
Figure 5.5 FTIR spectra of crystalline Al2O3-HfO2 alloy films. 77
Figure 5.6 RBS compositional calibration of AES signal levels of Al-O and Ta-O peaks. 78
Figure 5.7 RBS compositional calibration of AES signal levels of Al-O and Hf-O peaks. 79
viii Page
Figure 5.8 AES of interface formation for x=0.43, (Ta2O5)x(Al2O3)(1-x).80
Figure 5.9 AES of the initial stages of deposition of Hf-aluminates. 81
Figure 5.10. Room temperature, 1 Mhz, C-V plots for Al2O3, Ta2O5, and Ta-aluminates on (a) p-silicon and (b) n-silicon with aluminum gates. 82
Figure 5.11. Room temperature, 1 Mhz, C-V plots for Al2O3, HfO2, and Hf-aluminates on (a) p-silicon and (b) n-silicon with aluminum gates. 83
Figure 5.12. Temperature dependent C-V for SiO2, Al2O3, Ta2O5, and HfO2 on p-silicon with aluminum gates. 84
Figure 5.13. Temperature dependence of (a) C-V and (b) J-V for 43% Ta2O5 Ta-aluminate on p-silicon with aluminum gates. 85
Figure 5.14. Temperature dependence of (a) C-V and (b) J-V for 38% HfO2 Hf-aluminate on p-silicon with aluminum gates. 86
Figure 5.15. Temperature dependence of (a) C-V and (b) J-V for 48% HfO2 Hf-aluminate on n-silicon with aluminum gates. 87
Figure 5.16. J vs. 1/T for 43% Ta2O5 Ta-aluminate on p-silicon. 88
Figure 5.17. J vs. 1/T for both 38% and 48% HfO2 Hf-aluminates on p- and n-silicon, respectively. 89
Figure 5.18. Localized bonding, b., and anti-bonding, a.b.*, states in Ta- and Hf-aluminates. 90
ix 1 Introduction.
1.1 The Need for High Dielectric Constant Materials.
As device dimensions are scaled according to the 1999 International
Technology Roadmap for Semiconductors,1 the equivalent gate oxide thickness,
EOT, must decrease below about 1.5nm. At this thickness of SiO2, the direct
tunneling current for a one volt potential drop across the oxide is > 1 A•cm-2 and begins to reduce the ratio of on- to off-state current in a field effect transistor. To reduce the off-state leakage currents due to tunneling through silicon dioxide and maintain a capacitance that is equivalent to that obtained with a SiO2 dielectric
with a physical thickness of 1.5 nm and below, alternative high-k dielectrics are
being investigated.2, 3 and references therein These high-k alternative dielectrics can
provide the required levels of EOT for device scaling at larger physical thickness,
thereby providing a materials pathway for reducing the tunneling current.
In order to form ideal materials we need to consider more than just the dielectric constant. Some necessary considerations include the (i) band gap and the subsequent conduction band offset with both silicon and the gate electrode,
(ii) stability with silicon and against crystallization, and (iii) ability to form a useable interface with silicon that includes values of interface trap density (Dit)
10 -2 -1 10 -2 ~10 cm •eV and fixed charge (Qf) <5x10 cm . To this date, no one material is being portrayed as the ideal replacement for silicon dioxide.4 1.2 General Properties of High Dielectric Constant Materials.
In general as the dielectric constant increases the band gap tends to
decrease.3,4 The band gap is fundamentally related to the local electronic states associated with the composing members of the material. There are inherent differences between silicon/aluminum and the transition metals. Both silicon and aluminum's valence bands are composed of 3s and 3p states. In the case of
transition metals an additional state is introduced, nd. The principle quantum
number for this state is lower by one from the (n+1)s and (n+1)p valence states.
The energy level of the atomic nd state is therefore lower than the (n+1)s and
(n+1)p valence states. (Fig. 1.1)
The valence and conduction band edges are typically closely associated to the atomic valence states for a material.5 From molecular orbital theory, the
valence band is comprised of bonding orbitals and the conduction band of anti- bonding orbitals. It follows, since the transition-metals nd states are lower in energy than the (n+1)s and (n+1)p valence states the conduction band edges will be lower in energy than non-transition metals. (Fig. 1.1) The lower conduction band edge tends to have two effects: (i) to decrease the conduction band offset with silicon; and (ii) to decrease the band gap. (Table 1.1)
J. Robertson uses theories of Schottky barrier heights based on metal- induced gap states and charge neutrality levels to predict a dielectric's conduction band offset with silicon.2 His values were derived from the crystalline
forms of the dielectrics. By this method he predicted values of 2.8eV for Al2O3,
0.36eV for Ta2O5, and 1.5 eV for HfO2, with corresponding band gaps of 8.8 eV,
2 4.4 eV and 6.0 eV, respectively. S. Miyazaki has measured the same offset
values for amorphous forms of Ta2O5 and Al2O3 with photoemission
6 spectroscopy. He obtained values of 2.08 eV for Al2O3 and 0.28 eV for Ta2O5.
The difference between the two values for Al2O3 is consistent with the differences
in coordination between crystalline corundum or α-Al2O3, where the coordination
is 6, and the non-crystalline Al2O3, where the ratio of 4- to 6-fold coordinated Al is
3:1.7,8
Another local electronic effect is the bond ionicity. This property is derived
from the difference in Pauling's electro-negativities. For a binary system the
9 Pauling’s empirical definition of bond ionicity, Ib, is shown below. (See Table 1.1)
2 [Eq. 1.1] Ib = 1 – exp ( -0.25•(∆X) )
Bond ionicity has been demonstrated as a useful tool to help predict
important physical and electronic properties of materials.4 It can be related to the
coordination of the oxygen atom, crystallization temperature, and interface
properties.
The plot in Figure 1.2 demonstrates the correlation of the average bond
ionicity to the oxygen atom coordination.4 An increase in average bond ionicity
tends to correspond to a decrease in crystallization temperatures; i.e. SiO2, Ib =
0.45, cannot be crystallized below its melt temperature, >1600°C, Al2O3, Ib =
0.57, crystallizes at 900°C for a 30 second anneal, and Ta2O5, Ib = 0.61,
8,10,11 crystallizes at 800°C for the same anneal. Both ZrO2, Ib = 0.71, and HfO2, Ib
= 0.68, crystallize below 700°C.12 For thin films, crystallization is expected to be
3 dependent on the thickness and there may be a quantitative difference when the film thickness approaches the crystallite size.4
The bond ionicity can also be related to the amount of charge present at
the interface. If there is sufficient ionization, the metal center will display enough
effective charge that the interface between the dielectric and silicon will no longer
13 be neutral. This has been demonstrated in Al2O3 which displayed negative fixed charge, and positive fixed charge has been shown for some IIIB and IVB
14 15 16 elements; ZrO2 , Y2O3 , and (Y2O3)x(SiO2)1-x.
1.3 Objective.
This dissertation will attempt to evaluate materials of interest to the high-k
community. It will concentrate on providing a general quantitative and qualitative
description of the materials. It will demonstrate bulk, interfacial, and electrical
properties. In order to understand the measured properties a general qualitative
description of the material will be extracted.
1.4 Overview of the Dissertation.
This dissertation will attempt to demonstrate useful physical and electrical
properties and qualitative understanding of aluminum oxide, tantalum oxide,
tantalum aluminates, and hafnium aluminates. It was shown previously that it is
necessary to choose a material that satisfies requirements other than just a high
dielectric constant. Aluminum oxide whose dielectric constant is only ~9, provides
a large band offset, 2.08eV,6 and is stable on silicon. Tantalum oxide has a band
4 offset of only ~0.3eV, however, it provides an almost neutral interface. Hafnium oxide provides a positively charged interface with a band offset of 2.3eV. The
alloys of tantalum and hafnium with aluminum oxide is thought to be able to
provide a neutral interface with sufficient band offsets, and will be examined in
Chapter 5.
CHAPTER TWO provides a description of all the experimental techniques
used in the fabrication and evaluation of the materials.
In CHAPTER THREE, we provide an in depth investigation into the
properties of amorphous aluminum oxide. It includes information about the
interface, bulk, and electrical properties. We suggest a physical model that
describes the electrical properties.
In CHAPTER FOUR, there is a brief description of tantalum oxide.
Included are its interface, bulk, and electrical properties. This material has been
studied in-depth by many other investigators17 and references therein. We have felt its
brief inclusion was necessary because of its significance to the tantalum
aluminates.
In CHAPTER FIVE, we investigate the properties of tantalum and hafnium
aluminates, to determine if they form a continuous pseudo-binary alloy or simply
a compound of aluminum and tantalum or hafnium oxide. Electrical investigations
will concentrate on an apparent trapping mechanism, observable in hysteresis
and flatband shifts in temperature dependent C-V measurements.
CHAPTER SIX provides some final words and conclusions. Some future
directions for these materials are also provided.
5 1.5 References.
1 Semiconductor Industry Association, International Technology Roadmap for Semiconductors: 1999 Edition. (http://public.itrs.net)
2 J. Robertson, J. Vac. Sci. Technol. B 18, 1785 (2000).
3 G.D. Wilk, R.M. Wallace, and J.M. Anthony, Appl. Phys. Rev. 10, 5243 (2001).
4 G. Lucovsky, J. Vac. Sci. Technol. A 19, 1553 (2001).
5 J.E. Huheey, E.A. Keiter, and R.L. Keiter, Inorganic Chemistry, Principles of Structure and Reativity 4th ed., (Harper Collins College Publishers, New York, 1993).
6 S. Miyazaki and M. Hirose, AIP Conf. Proc. 550, 89 (2001).
7 G. Lucovsky, G., A. Rozaj-Brvar, and R.F. Davis, in The Structure of Non- Crystalline Materials 1982, edited by P.H. Gaskell, J.M. Parker and E.A. Davis (Taylor and Francis, London, 1983), p. 193.
8 B. Rayner, H. Niimi, R. Johnson, R. Therrien, G. Lucovsky and F.L. Galeener, in Proc. of Characterization and Metrology for USLI Technology, AIP Conf. Proc. 550, 149 (2001).
9 L. Pauling, The Nature of the Chemical Bond, 3rd ed. (Cornell University, Ithaca, NY, 1948).
10 B. Rayner et al., Mater. Res. Soc. Symp. Proc. 661, c1 3.1 (2001).
11 R. S. Johnson, J. G. Hong, and G. Lucovsky, J. Vac. Sci. Technol. B 19, 1606 (2001).
12 G. Lucovsky and J. C. Phillips, J. Non-Cryst. Solids 227, 1221 (1998).
13 R. S. Johnson, G. Lucovsky, and I. Baumvol, J. Vac. Sci. Technol. A 19, 1353 (2001).
14 S. W. Nam et. al., J. Vac. Sci. Technol. A 19, 1720 (2001).
15 R. N. Sharma and A. C. Rastogi, J. Appl. Phys. 74, 6691 (1993).
6
16 J. Chambers, Ph.D. Thesis, (North Carolina State University, Raleigh, NC, 2000). (http://www.lib.ncsu.edu/etd/)
17 C. Chaneliere, J. L. Autran, R. A. B. Devine, and B. Balland, Mater. Sci. Eng., R. 22, 269 (1998).
7 Table 1.1. Local electronic and physical properties of selected high-k materials.2,4 * estimated value
Coordination Coordination Band Gap, Eg Dielectric ∆X I b metal/silicon oxygen eV
SiO2 1.54 0.45 4 2.0 9.0
Al2O3 1.84 0.57 4 and 6 (3:1 ratio) 3.0 8.8
(ZrO2)0.5(SiO2)0.5 d-states 1.88 0.59 8 and 4 3.0 n/a
TiO2 present 1.9 0.59 6 3.0 3.05
Ta2O5 1.94 0.61 6 and 8 (1:1 ratio) 2.8 4.4
(Y2O3)1(SiO2)1 1.99 0.63 6 and 4 3.11 n/a
HfO2 2.14 0.68 8 4.0 6
ZrO2 2.22 0.71 8 4.0 5.8
Y2O3 2.22 0.71 6 4.0 6
La2O3 2.34 0.75 6 4.0 6*
8 t1u (σ∗,π∗) 6
a1g (σ∗) 2
TM (n+1)p
TM (n+1)s eg (σ∗) 4
t2g (π∗) 6
TM nd
O 2p (σ,π)
t1g + t 2u 12
t1u (σ,π) 6
t2g (π) 6
eg (σ) 4
t1u (σ,π) 6
a1g (σ) 2
Figure 1.1. General molecular orbital diagram for oxides of transition metals. The
left hand side depicts the transition metal atomic orbitals and the right hand side
is the oxygen 2p atomic orbitals. The oxygen 2s orbital energy is much lower and is not used in most bonding arrangements and is therefore not included in this diagram. The middle levels are the oxide's; the top of the valence band is
determined by the oxygen 2p non-bonding level and the lower level of the
conduction band are the levels caused by the d levels.
9
Y O 2 3 4.0
3.5
Al O 2 3 3.0
Ta O 2 5
2.5
Oxygen Atom CoordinationOxygen SiO 2 2.0
0.45 0.50 0.55 0.60 0.65 0.70 0.75 Average Bond Ionicity, I b
Figure 1.2. Plot of oxygen coordination as a function of average bond ionicity, Ib. From Table 1.1.
10 2 Experimental Methods.
2.1 Materials Deposition.
2.1.1 Remote Plasma Enhanced Chemical Vapor Deposition, RPECVD.
All the materials deposited for this dissertation were deposited using a
Remote Plasma Enhanced CVD technique, RPECVD.1 This is a technique that
has been used extensively by our research group. It provides device quality
silicon dioxide for CMOS applications and has been applied to the formation of
silicon nitride and oxynitrides.2,3
RPECVD differs from conventional or direct plasma processing in: (i) it provides a way to selectively activate gases by isolating the plasma to a certain area, gases to be excited are injected through a quartz plasma tube and the others, source gases, through either a shower head ring (Fig. 2.1) or an aluminum heated injector located in the sides of the chamber (Fig. 2.2); (ii) deposition occurs outside the glow region; and (iii) the source gases are prevented from back streaming into the plasma region by a combination of pressure differential and gas flow.1
Typical excited gases used are oxygen, nitrogen, nitrous oxide and ammonium. The source gas used depends on the material to be deposited. For
silicon dioxide, 2% silane in a balance of helium is used. For depositions of high-
k materials, the source material is a liquid and is housed in a bubbler container,
discussed in the next section.
11 The energy for deposition comes from the RF power, 30 Watts at 13.56
MHz. This allows the substrate to be held at relatively low temperatures, 200ºC to
300°C. These temperatures are generally sufficient to desorb organics that may remain on the surface after cleaning and low enough to prevent thermal oxidation of the substrate by ambient oxygen.
Typical background pressures of the system are ~10-8 Torr. During depositions the pressure is held at 0.3 Torr by the use of an absolute Baratron and throttle valve located just prior to the turbo molecular process pump. (Fig.
2.1)
2.1.2 Metal Organic Bubbler.
Films containing aluminum, tantalum or hafnium were deposited from a source gas extracted from a liquid. The source materials were housed in an
Schumacher4 BK1200SSZ, Al and Ta, or a BK500SSN, Hf, stainless steel
bubbler container. Operational theory of bubblers is contained in Ref. 5.
The purpose of the bubbler container is as a vessel to hold the liquid
source through which a carrier gas, helium in our case, is passed. (Fig. 2.1) The passage of the carrier gas through the liquid causes some of the liquid molecules
to evaporate and be transported by the gas. The pressure above the liquid will
determine the amount of vapor that can be removed. When the pressure is
increased less evaporation occurs and correspondingly the deposition rates will
decrease. A closed loop controller monitors an absolute Baratron and operates a
12 needle valve to regulate the pressure. The temperature of the liquid is controlled by a NESLAB RTE-211 oil immersion bath.6
The aluminum source is triethyldialuminum tri-sec-butoxide7,8 (TEDA-
TSB), the tantalum source is Tantalum Pentaethoxide4 (TAETO) and the hafnium
is from hafnium t-butoxide.8 The source gases are mixed upstream, for alloys,
and are injected into the chamber through two heated injector assemblies located
on opposite sides of the sample.
2.2 Auger Electron Spectroscopy, AES.
AES was performed with a Physical Electronics 11-010 5kV electron gun
control, 10-155 cylindrical mirror analyzer (CMA), 32-150 digital analyzer control,
137 PC interface board assembly, 96A V/f preamplifier, 32-100 electron multiplier
power supply, and the software interface program AES_CGA.
The auger process and analytical techniques are discussed in Ref. 9. The
auger process requires an energetic electron beam and sample damage is a
concern. We use a minimal beam technique, i.e., the current of the beam is kept
low and the detector multiplier voltage high.10 We do not use a defocusing
technique; the beam is focused to attain a maximum number of counts.
In order to provide a repeatable technique and a minimal amount of beam
current, the multiplier voltage is kept at a constant value and the filament current
is varied to attain an appropriate number of counts. The multiplier value is 4.0 on
the 32-100 power supply dial and corresponds to ~1,444 volts. For alignment
purposes, the accelerating voltage is set a 2.0 keV and the main elastic peak is
13 aligned so its peak falls on 2.0 keV and the filament current is adjusted so between 900 and 1000 kilo counts per second are obtained. This step ensures consistent results from run to run. Once this is accomplished, an AES scan at 3.0 keV can be completed. We have observed that with these settings the spectrum of Si-O, 76eV, from a thermal silicon dioxide sample will remain unchanged for
>20 minutes.
2.3 Nuclear Resonance Profiling, NRP.
The interfaces of Al2O3/SiO2 and Al2O3/Si were investigated using nuclear
resonance profiling,11 NRP. The narrow, isolated resonance at 404.9 keV12 in the
cross section curve of the 27Al(p, γ)28Si nuclear reaction was used to obtain Al
concentration depth distributions. The measured excitation curves, i.e., gamma-
ray yield versus incident proton energy, around the resonance energy were converted into concentration profiles using the SPACES program,13 which is
based on stochastic theory of energy loss of ions in matter. By measuring the
excitation curves for thin and thick aluminum films, the nuclear reaction
resonance at 404.9 keV was determined to be narrower than 40 eV. A sample
geometry with a tilt of 65º with respect to the incident beam was used to increase
depth resolution. Several factors contribute to the obtained depth resolution: (i)
an extremely narrow nuclear reaction resonance, 40 eV; (ii) a significant energy
-1 -2 loss of 405-keV protons in Al2O3, approximately 380 keV mg cm , if a density of
-3 3.98 g cm is assumed for Al2O3; (iii) a low energy spread of the proton beam,
approximately 80 eV FWHM Gaussian at 405 keV, as provided by the 500 keV
14 HVEE ion implanter in Porto Alegre; and (iv) an effective thickness magnification by a factor of 2.4 due to the tilted geometry. The ultimate resolution is limited by energy straggling and angular multiple scattering processes. In the present work the depth resolution was 0.4 - 0.5 nm near the surface, which degrades to 0.6 to
0.8 nm near the buried interface.
The silicon profiles were determined by NRP using the narrow, isolated resonance around 414 keV (∆R = 100 eV) of the 29Si(p,γ)30P nuclear reaction11,
12, 13, at 60 degrees sample tilt. Depth resolution is 0.7 nm at sample surface, which degrades to approximately 1.0 nm at a depth of 8.0 nm into the film.
2.4 Fourier Transform Infrared Spectroscopy, FTIR.
Fourier transform infrared, FTIR, measurements were performed with a
Nicolet 750 spectrometer in transmission mode with a resolution of 4 cm-1.
Background spectra were subtracted from a reference sample cut from the same
Si(100) single sided wafer; one side was chemically roughened to prevent
internal reflection errors. The film thickness were >100nm.
2.5 X-Ray Diffraction, XRD.
X-ray diffraction was performed on a Brucker Angular X-Ray
Spectroscopy, AXS. A beam energy of 30keV with a current of 20-30mA was
used. The films and substrates investigated were the same used for FTIR.
15 2.6 Electrical Characterization - Capacitance and Current versus Voltage,
C-V and J-V.
For electrical measurements, field isolation MOS capacitors were fabricated on 0.06 - 0.08 Ω-cm boron doped Si, and 0.02 - 0.05 Ω-cm phosphorous doped Si substrates. Prior to the Al gate metal contact evaporation the samples were annealed to between 700 and 900°C, depending on the
materials crystallization temperature (see appropriate chapters), for 30 seconds
using an AG Associates minipulse rapid thermal annealer. It has been shown for
SiO2 films deposited by RPECVD, that an RTA at a temperature of ~900°C is necessary to promote structural and chemical relaxation of the film.14 A 30-
minute anneal in a forming gas mixture of 10% H2 in N2 at 400°C was used to
minimize Dit. This step was performed before metallization for all the materials, to
reduce reactions between the Al metal gate and the dielectric films,15 except for
Al2O3 and SiO2 where it was the final preparation step.
All electrical measurements were performed on a Material Development
Corporation system with version 1.494 software interface program. An HP
4284A LCR meter with a frequency range of 10kHz to 1MHz was used to perform
capacitance voltage measurements. The current-voltage measurements were
performed by an HP 4140B voltage source with a pico-ammeter and voltage
ramp of dV/dt = 0.05 V/sec. The temperature of the sample was regulated
between 25ºC and 300ºC. All measurements were performed in a light tight box.
16 2.7 References.
1 T.Yasuda, Y. Ma, S. Habermehl, and G. Lucovsky, J. Vac. Sci. Technol. B 10, 1844(1992).
2 H.Y. Yang, H. Niimi, and G. Lucovksy, J. Appl. Phys. 83, 2327 (1998).
3 Y. Wu, H. Niimi, H. Yang, G. Lucovsky, and R. B. Fair, J. Vac. Sci. Technol. B 17, 1813 (1999).
4 http://www.schumacher.com
5 S. D. Hersee and J.M. Ballingall, J. Vac. Sci. Technol. A 8, 800 (1990).
6 http://www.neslab.com
7 R. G. Gordon, K. Kramer, and X. Lui, Mat. Res. Soc. Symp. Proc. 446, 383 (1997).
8 Strem Chemicals ( http://www.strem.com ).
9 L.E. Davis, N.C. MacDonald, P.W. Palnberg, G.E. Raich, and R.I. Weber, Handbook of Auger Electron Spectroscopy 2cnd. ed. (Physical Electronics Industries, Inc.) Eden Pairie, MN (1976).
10 Y. Ma, T. Yasuda, and G. Lucovsky, J. Vac. Sci. Techon B 11, 1533 (1993).
11 I.J.R. Baumvol, Atomic Transport During growth of Ultrathin Dielectric, Surface Science Reports, 36 (1999) pg. 1-166.
12 S.E. Hunt and W.M. Jones, Phys. Rev. 89, 1283 (1953).
13 I. Vickridge and G. Amsel, Nucl. Instr. Meth. B 45 , 6 (1990).
14 B. J. Hinds, F. Wang, D. M. Wolfe, C. L. Hinkle, and G. Lucovsky, J. Vac. Sci. Technol. B 16, 2171 (1998).
15 L-A. Ragnarsson, E. Aderstedt and P. Lundgren, Mat. Res. Soc. Symp. Proc. 367, 451 (1999).
17
Table 2.1. Remote plasma process conditions. Where: (i) PLASMA refers to gas injected through the quartz plasma tube and excited by the RF power (Fig. 2.1), (ii) INJECTOR refers to metal organic vapor in a He carrier gas injected through the heated injectors (Fig. 2.2); (iii) RING refers to gas injected through the shower head ring (Fig. 2.1); and (iv) Pressure in the metal organic columns refers to the pressure maintained in the bubbler containers (Fig. 2.1).
He O2 Al (TEDA-TSB) 75.0°C Ta (TEATO) 120°C Hf (t-butoxide) 15-25ºC SiH4 Process (200 sccm) (20 sccm) He (20 sccm) Pressure He (20 sccm) Pressure He (20 sccm) Pressure (10 sccm) Pressure Temperature RF Power
Oxidation of Si Substrate PLASMA PLASMA 0.3 Torr 300°C 30 Watts
SiO2 PLASMA PLASMA RING 0.3 Torr 300°C 30 Watts
Al2O3 PLASMA PLASMA INJECTOR 30.0 Torr 0.3 Torr 300°C 30 Watts
Ta2O5 PLASMA PLASMA INJECTOR 30.0 Torr 0.3 Torr 300°C 30 Watts
HfO2 PLASMA PLASMA INJECTOR 30.0 Torr 0.3 Torr 200°C 30 Watts
Ta-Aluminates 11% Ta2O5 PLASMA PLASMA INJECTOR 30.0 Torr INJECTOR 200 Torr 0.3 Torr 300°C 30 Watts
34% Ta2O5 PLASMA PLASMA INJECTOR 30.0 Torr INJECTOR 100 Torr 0.3 Torr 300°C 30 Watts
43% Ta2O5 PLASMA PLASMA INJECTOR 60.0 Torr INJECTOR 60.0 Torr 0.3 Torr 300°C 30 Watts
54% Ta2O5 PLASMA PLASMA INJECTOR 100 Torr INJECTOR 30.0 Torr 0.3 Torr 300°C 30 Watts
74% Ta2O5 PLASMA PLASMA INJECTOR 200 Torr INJECTOR 30.0 Torr 0.3 Torr 300°C 30 Watts
Hf-Aluminates The results were similar to Ta-aluminates except that there is much less source material and the bath temperature needs to increased as the 0.3 Torr 300°C 30 Watts source ages.
18 Absolute Baraton
Needle Helium/Oxygen Valve Helium
Silane Plasma Coil (RF Power) TAETO Bubbler
Substrate Mechanical Bypass Pump Light Bulbs (heating) Absolute Baraton Chamber Throttle Needle Valve Valve Helium
TMP
TEDA-TSB Bubbler
Mechanical Bypass Pump
Figure 2.1. RPECVD gas flow and bubbler assembly for all depositions. The
TAETO set-up is replaced by a hafnium t-butoxide set-up for HfO2 depositions.
19 He/O2
Metal Organics
Figure 2.2. RPECVD chamber for depositions. The shower head ring above the substrate for silane has been left out for clarity.
20 3 Properties of RPECVD Aluminum Oxide.
3.1 Introduction.
In this chapter results of RPECVD Al2O3 will be introduced and discussed.
FTIR and XRD were utilized to determine the crystallization temperature for film
thicknesses >100nm. The interfaces with (i) HF- last, H-terminated silicon, (ii)
remote plasma oxidized silicon and (iii) RPECVD silicon dioxide were
investigated by AES and NRP and correlated to C-V measurements. C-V and J-V
characteristics were measured from room temperature to 200°C.
It will be shown that with RPECVD a minimal interfacial layer, 0.6 to 0.8nm
forms during the initial stages of aluminum oxide deposition on HF last silicon.
Previous studies have indicated when aluminum oxide is deposited directly on
silicon a relatively thick, ~2.0 to 3.0nm, silicate layer forms.1,2,3,4 These results
utilized different deposition techniques; i.e., thermal CVD, rapid thermal CVD
(RTCVD), direct PECVD, and physical vapor deposition.
However, our relatively abrupt interfaces do not ensure good device
performance. C-V measurements determined that ~1012 charges•cm-2 of
negative fixed charge resides at the interface. This charge was shown to be
inherent to Al2O3 and could be moved from the silicon substrate interface by
placing a layer of deposited silicon dioxide between the substrate and the
aluminum oxide.
21 3.2 Interface with silicon and silicon dioxide.
The interface formation for the alloys was investigated with on line AES at the initial stages of film deposition. (Fig. 3.1) The deposition process was interrupted every 10 seconds, the sample was transferred to the AES chamber, and an AES scan was taken.
The AES spectra shown in Fig. 3.1 display the evolving chemical composition at the Si-interface and within the deposited Al2O3. The left-hand plot
is for Al2O3 deposited on HF-last, H-terminated Si. The right-hand plot is for
Al2O3 deposited on 0.6 nm of SiO2 that was formed by remote plasma assisted
oxidation, i.e., exposing the HF-last silicon for 30 seconds to active O-atoms and
metastable molecules extracted from a remote He/O2 plasma discharge. The
thickness of the interfacial SiO2 layer was determined by comparing the relative
amplitudes of the Si-O (76eV) and the Si-Si (91eV) AES features.5
For 10 seconds of Al2O3 deposition, the two AES spectra in Fig. 3.1 are
essentially the same. In particular the ratios of the Si-O and Si-Si features indicate that approximately 0.6nm of SiO2 was formed during the initial stages of
the Al2O3 deposition on HF-last Si. For increasing deposition times the line- shape of the Al-O feature at ~54 eV does not evolve. It is a single peak and located at an energy consistent with aluminum bonded to oxygen.6 From this we
conclude that there is no silicide formation at the interface and the deposited film
is fully oxidized.
NRP measurements were performed for relatively thin layers of Al2O3 on:
(i) HF-last Si; (ii) oxidized Si; and (iii) RPECVD SiO2. NRP independently
22 demonstrates that aluminum oxide on HF-last Si is abrupt, with the spread of Al atoms into the interfacial Si oxide layer being less than 0.6-0.8 nm, the limiting resolution of the NRP measurement. This estimate is based on the spread of the proton beam within the film. Figure 3.2 is the NRP study for the deposition of
Al2O3 directly onto HF-last Si. The lower traces with the experimental points are
the gamma ray yield data, and the upper portions of the figure display the
calculated concentrations of Al and Si as a function of depth;7,8,9 the
concentrations are normalized to Al2O3 and bulk Si.
For the deposition of Al2O3 on the HF-last Si, there is an abrupt drop in the
Al concentration followed by a rapid onset of the Si concentration. If we take the
intersection of the Al and Si concentrations as the location of the plane where the
interfacial layer begins, then the bulk Si concentration is achieved in less than 1.0 nm. There are essentially no changes in the experimental data or in the Al and
Si concentrations after an 800°C anneal. This indicated that there is no
detectable intermixing of the Al2O3 and SiO2 and the transition regime remains
abrupt.
Figure 3.3 is for deposition of Al2O3 onto an RPECVD SiO2 layer. The
shoulder in the Si concentration profile indicates the thickness of the SiO2 layer,
~2 nm. There are changes in the abruptness of the Al profile before and after the
800°C anneal indicating intermixing of the deposited Al2O3 and SiO2 films. Since
the two plots in Fig. 3.3 come from different regions of the same wafer, and since
the deposited Al2O3 was not of uniform thickness, the extent of the mixing is best
estimated from the changes in the abruptness of the interfacial profiles of both
23 the Si and Al atoms. Based on this subjective analysis of the profiles, the internal
transition layer thickness is of the order of 1 nm, or approximately three molecular layers, thick.
3.3 Bulk Properties, FTIR and XRD.
The features in the FTIR spectra are bond specific and give information
about coordination as well, whereas the XRD spectra can be used to determine
the onset of crystallization, i.e., the transition from a non-crystalline or amorphous
film to a crystalline or partially crystallized film. Based on FTIR and XRD data
shown in Figs. 3.4 and 3.5, we observe that above 800°C the coordination of the
Al atoms has changed and the film has crystallized.
Figure 3.4 displays two FTIR spectra, as-deposited and 900°C anneal.
FTIR spectra of films annealed at 600°C, 700°C and 800°C were essentially the
same as that of the as-deposited film. The features in the FTIR spectra after the
900°C anneal are in the spectral regime of features previously identified in
10 crystalline Al2O3 in the corundum structure. The broad feature in the as-
deposited film results from both Al and O atom motions, whereas the sharper
feature in the 900°C spectrum is due mostly to O-atom motion. The shift of the
spectrum to lower wave numbers is consistent with (i) an increase in the average
coordination of the Al atoms from ~ 4.5 in the non-crystalline state to 6 in the
corundum structure, and (ii) an increase in the ionic character of the bonding that
accompanies the change in average coordination of the Al and O atoms.11
24 For as-deposited films, and those annealed up to 800°C, there are no sharp features in the XRD spectrum (Fig. 3.5); however, a diffraction peak associated with crystalline Al2O3 appears at about 40° after the 900°C anneal,
10 which corresponds to the (110) plane for corundum Al2O3. The broad feature at
about 70° is due to the Si(200) substrate plane.
3.4 Electrical Characterization.
Field isolation MOS capacitors were made on n-type and p-type Si(100)
substrates with carrier concentrations in the mid 1017 cm-3 range. RPECVD
Al2O3 was used as the gate dielectric and deposited onto four different interfaces
(i) HF-last, H-terminated Si, (ii) 0.6 nm SiO2 prepared by remote plasma-assisted
oxidation via a He/O2 plasma, (iii) 1.2 nm of deposited RPECVD deposited SiO2,
and (iv) 2.2 nm of RPECVD deposited SiO2.
C-V measurements were performed at 1 MHz, and the data was analyzed
using a model that performs a least squares fit routine and yields values of flat
12 band voltage and oxide thickness normalized to the dielectric constant of SiO2.
Samples had flat band voltages that were shifted positively for increasing
thickness for both p-type and n-type Si substrates, (Fig. 3.6) i.e., in so-called n-
MOSCAP and p-MOSCAP, metal oxide semiconductor capacitor structures,
where n and p, respectively, refer to the minority carriers in the p-type and n-type
substrates (this notation is the convention for field effect transistor devices,
where the channels for carrier transport are formed by substrate inversion).
25 Flat band voltage as a function of EOT for MOS capacitors of Al2O3 on
HF-last Si is displayed in Figure 3.7. The EOT value was obtained by subtracting
0.6 nm from the total measured EOT. This subtraction was based on the AES and NRP data for the HF-last depositions, and on experimental data that indicated that the fixed charge levels at the remote plasma-processed Si-SiO2
interfaces were at most in the mid 1010 cm-2.
The flat band voltage-EOT data has been fit with straight lines consistent
with the analysis to be presented below. The y-intercept in this type of fit is the
metal-semiconductor work function difference, Φms, and the slope indicates the
13 fixed charge at the interface between the Al2O3 and the interfacial layer.
The slopes, obtained from the fits to the p-MOSCAP and n-MOSCAPs
correspond to negative fixed interfacial charge densities of –7.0±1.0×1012/cm-2.
The values of Φms are consistent with the work function difference between the
aluminum electrodes and the Fermi level position in the doped Si substrates.
In addition to HF-last interfaces, capacitors with two other interfaces were
also studied: (i) 1.2 nm of deposited RPECVD deposited SiO2 and (ii) 2.2 nm of
RPECVD deposited SiO2. The EOT and flat band voltages from these stacks are
displayed in Figure 3.8. The EOT reported is that of the Al2O3 portion of the gate
dielectric (see Eqn. 3.3 below). By varying the thickness of the SiO2 layer
deposited between the Si substrate and the Al2O3 layer, the position of the fixed
charge can be confirmed. Before discussing the data for these stacked
dielectrics, the solution to Poisson’s equation is presented.
26 By using Poisson’s equation, the potential due to the interfacial charge can be calculated, and be subtracted from the work function difference, Φms,
which is the flat band voltage in the absence of fixed charge, and the flat band
13 voltage, VFB, can be then be calculated as in Equation 3.1;
q k k SiO SiO [Eq. 3.1] VFB = ΦMS − ⋅ QSiO ⋅ dSiO + ⋅ dAlO + Q AlO ⋅ ⋅ dAlO εo ⋅kSiO k AlO k AlO
QSiO is the charge located at the Si-SiO2 interface and QAlO is the charge located
at the internal dielectric interface between the SiO2 interface layer and the
RPECVD Al2O3 layer. εo is the permitivity of free space, kSiO and kAlO are,
respectively, the dielectric constants of SiO2 and Al2O3, 3.8 and 9.0, and dSiO and
dAlO are, respectively, the physical thicknesses of the SiO2 and Al2O3 constituent
layers of the stacked dielectric. Equation 3.1 can be rearranged and written in
terms of the contributions to VFB in which dSiO and dAlO are the scaling variables,
as in Equation 3.2:
q kSiO [Eq. 3.2] VFB = ΦMS − ⋅ QSiO ⋅ dSiO + ()QSiO + Q AlO ⋅ ⋅ dAlO εo ⋅kSiO k AlO
If the flat band shift is plotted as a function of aluminum oxide thickness,
then fixed charge residing at the Si-SiO2 interface will affect both the slope and
the y-intercept value. If the charge is only at the internal SiO2-Al2O3 dielectric
interface then only the slope will be affected; i.e., the slope will be proportional to
QAlO and the intercept with ΦMS.
Figure 3.8 presents the flat band voltage-EOT plots for n-MOSCAP and p-
MOSCAPs. From n-MOS data, the slopes and y-intercepts for varying amounts
27 of SiO2 are the same to within experimental error as the respective slopes and y-
intercepts or the HF-last devices with 0.6 nm of interfacial SiO2. The p-MOS data
exhibits changes in both slope and y-intercept as compared to the HF-last
devices. This analysis places the fixed charge at the interface between the
interfacial SiO2 and the RPECVD Al2O3 layer.
The NRP results indicated a mixing of Al2O3 and the RPECVD SiO2 after
the 800°C anneal. This mixing is confirmed in the C-V results by a reduced value
of EOT for aluminum oxide on RPECVD silicon dioxide. The measured EOT of
the device should be a weighted addition of the silicon dioxide and aluminum
oxide as in Equation 3.3:
k [Eq. 3.3] EOT = Tox(SiO ) + SiO2 ∗ Tox(Al O ) . 2 k 2 3 Al2O3
Analysis of MOSCAPs on p-type and n-type Si substrates with the same
deposition times as those in Figure 3.8 resulted in gate oxides with physical
thicknesses, respectively, of 1.2 nm and 2.2 nm. For the SiO2-Al2O3 stacks the
EOT was extracted from the C-V data. The thickness of the SiO2 constituent was
then subtracted from EOT to give the contribution due to the Al2O3 layer. If dSiO
values of 1.2 nm and 2.2 nm were used to calculate the thickness of the Al2O3,
the resultant Al2O3 layer thickness was significantly thinner compared to that of
an equivalent Al2O3 deposited on the HF-last Si. Instead thicknesses of 1.0 nm
and 1.6 nm were required to attain aluminum oxide thicknesses that
corresponded well to HF-last devices. This effect is presumed to result from a
mixing of the SiO2 and Al2O3 layers that was determined from the resonant NRP
28 studies. (Fig. 3.3) When SiO2 is mixed with higher dielectric materials the
effective dielectric constant of the resulting material is always higher than that of
14 SiO2.
Figure 3.9 displays current density versus voltage, J-V, data for n-MOS
and p-MOSCAPs, for substrate accumulation. The primary features of the
leakage current in Figure 3.9, is an asymmetry in the J-V traces for equivalent
EOT and bias. This is attributed to an inherent asymmetry in the band alignment
and resulting band bending. (Fig. 3.10) For n-MOSCAPs, the current is injected
from the gate and the semiconductor bands bend down providing a large barrier
to that reduces direct tunneling. In the p-MOSCAPs, the current is injected from
the substrate; the semiconductor bands bend up providing a smaller tunneling
barrier for an equivalent bias and EOT. This inherent difference in barrier heights
results in an increased current density for p-MOSCAPs.
Temperature profiling of C-V and J-V provide a simple and effective way
to investigate charge trapping. Figures 3.11 and 3.12 display the results of C-V
and J-V, respectively, when the temperature is increased to 200°C. Figure 3.11
displays both SiO2 and Al2O3, this provides a reference to compare the Al2O3
results. Immediately observable is the positive flatband shift described
previously. In addition, at 200°C there is a divergence from the 50°C curve just
before the inversion capacitance is reached. Following the modeling approach in
Ref. 15, this shoulder can be assigned to an electronic state with an energy level
located in the silicon band gap.
29 It is difficult to extract any qualitative understanding from the J-V alone.
(Fig. 3.12) However in Chapter 5, tantalum and hafnium aluminates will be introduced and, when compared to this material, some interpretations will become more obvious. For the present it can be stated that there appears to be a systematic and uniform increase of the current density for increasing temperatures.
3.5 Conclusions.
AES and NRP have demonstrated that it is possible to deposit Al2O3 on H-
terminated silicon with only a minimal amount of interfacial silicon dioxide, 0.6 to
0.8nm. NRP further indicated that aluminum oxide on silicon dioxide is not stable
at elevated temperatures. The NRP gamma ray profiles indicated that a mixing of
the two materials occurred for anneals at 800°C. This was correlated by C-V
measurements where device EOT's were considerably smaller than expected.
This is a result of the increase in dielectric constant of silicon dioxide when it is
mixed with a material of a higher dielectric constant.
FTIR and XRD for Al2O3 indicate that for the films, >100nm, the structure is amorphous when deposited at 300°C and crystallizes after anneals at 900°C.
Typical device dimensions requires films that are considerably thinner. It is conceivable that when the film thickness approaches the crystal grain size that there will be some suppression of the crystallization.
The analysis of the C-V data, in particular the flatband shift as a function of EOT, indicates the presence of fixed negative charge. This charge is shown to
30 be inherent to the Al2O3 interface and can be removed from the substrate
interface by the deposition of the SiO2 layer prior to the Al2O3 deposition.
The fixed negative charge is consistent with a model for the local atomic
bonding of noncrystalline Al2O3 that has two different bonding environments for
the Al atoms;16 (i) a tetrahedrally coordinated Al site that has a net negative charge, and (ii) octahedrally coordinated site in which the Al has a charge of +3.
The negatively charged Al atoms can bond directly to the O atoms of the
interfacial SiO2 and is the arrangement that is responsible for the fixed negative
charge. The tetrahedral arrangement with a negatively charged Al is unique, and
essentially all of the other transition metal oxides studied to date display fixed
16 positive charge at their interfaces with Si or SiO2. Temperature dependent C-V
measurements indicate electron traps in the immediate vicinity of the Si-dielectric
interface; the trapping sites may be intrinsic and associated with the octahedrally
coordinated Al.
Finally, J-V data for substrate accumulation shows a marked asymmetry
between n-MOSCAP and p-MOSCAP devices. This has been attributed to
inherently asymmetric band bending in stacked dielectric structures.
3.6 References.
1 T. M. Klein, D. Niu, W. S. Epling, W. Li, D. M. Mayer, C. C. Hobbs, R. I. Hedge, I. J. R. Baumvol and G. N. Parsons, Appl. Phys. Lett. 75, 4001 (1999).
2 E. P. Gusev, M. Copel, E. Cartier, I. J. R. Baumvol, C. Krug and M. A. Gribelyuk, Appl. Phys. Lett. 76, 176 (2000).
31
3 A. Chin, Y.H. Wu, S.B. Chen, C.C. Liao and W.J. Chen, Tech. Dig. VLSI Symp., p.16 (2000).
4 D.A. Buchanan, E.P. Gusev, E. Cartier, H. Okorn-Schmidt, K. Rim, M.A. Gribelyuk, A. Mocuta, A. Aymera, M. Copel, S. Guha, N. Bojarczuk, A. Callegari, C. D`Emic, P. Kozlowski, K. Chan, R.J. Fleming, P.C. Jamison, J. Brown and R. Arndt, Tech. Dig. Intl. Electron Devices Meet., p.223 (2000).
5 H. Niimi, K. Koh, and G. Lucovsky, Electrochemical Society Proc. 12, 623 (1996).
6 L.E. Davis, N.C. MacDonald, P.W. Palnberg, G.E. Raich, and R.I. Weber, Handbook of Auger Electron Spectroscopy 2cnd. ed. (Physical Electronics Industries, Inc.) Eden Pairie, MN (1976).
7 I.J.R. Baumvol, Atomic Transport During growth of Ultrathin Dielectric, Surface Science Reports, 36 (1999) pg. 1-166.
8 S.E. Hunt and W.M. Jones, Phys. Rev. 89, 1283 (1953).
9 I. Vickridge and G. Amsel, Nucl. Instr. Meth. B 45, 6 (1990).
10 JCPDS-ICDD, Newton square, PA (1993).
11 B. Rayner, H. Niimi, R. Johnson, R. Therrien, G. Lucovsky and F. L. Galeener, Proc. of Characterization and Metrology for USLI Technology, (American Institute of Physics) (2000), in press.
12 J. R. Hauser and K. Ahmed, AIP Conf. Proc. 449, 235 (1998).
13 D. K. Schroder, Semiconductor Material and Device Characterization 2cnd. ed. (John Wiley and Sons, Inc.) New York (1998).
14 G. Lucovsky, G. B. Rayner, Jr., Appl. Phys. Lett. 77, 2912 (2000).
15 E.H. Nicollian and J.R. Brews, MOS (metal oxide semiconductor) Physics and Technology, New York, Wiley, (1982).
16 G. Lucovsky, J. C. Phillips, and M. F. Thorpe, Proceedings of the of Characterization and Metrology for USLI Technology (American Institute of Physics, Melville, NY, 2000) (in press).
32
x2.5 20 sec Al O 20 sec Al O 2 3 2 3
x2.5 10 sec Al O 10 sec Al O 2 3 2 3
Si-0
Al-O Al-O 30 sec O Plasma HF Last Si 2 on HF Last Si dI/dE (arbitary units) dI/dE (arbitary dI/dE (arbitrary units) (arbitrary dI/dE
Si-O
Si-Si Si-Si
40 50 60 70 80 90 100 110 40 50 60 70 80 90 100 110 Energy (eV) Energy (eV)
Figure 3.1. AES of aluminum oxide deposition on HF-last silicon and 0.6nm of silicon dioxide on HF-last silicon.
33 As Deposited 30s 800C Anneal
1.0
0.8
0.6
Al O 2 3 0.4 Si Si Concentration (%) 29 0.2 Al and Al
27 0.0
012345678910 012345678910 Depth (nm) Depth (nm)
1200 800 1200 800
1000 1000 600 600 800 800
600 400 600 400
400 400
Gamma Yield 200 200 200 200
0 0 0 0 0.0 1.0 2.0 3.0 4.0 -1.0 0.0 1.0 2.0 3.0 0.0 1.0 2.0 3.0 4.0 -1.0 0.0 1.0 2.0 3.0 E - E (keV) E - E (keV) R R
Figure 3.2. NRP for Al2O3 on HF-last silicon. Where the scale is normalized to
Al2O3 and bulk silicon.
34 As Deposited 30s 800C Anneal
1.0
0.8
0.6 Al O 2 3 Si 0.4 Si Concentration (%) 29 0.2 Al and
27 0.0
0123456789101112 0123456789101112 Depth (nm) Depth (nm)
800 1200 800 1200
1000 1000 600 600 800 800
400 600 400 600
400 400 200 200
Gamma Yield 200 200
0 0 0 0 0.0 1.5 3.0 4.5 0.0 1.5 3.0 4.5 0.0 1.5 3.0 4.5 0.0 1.5 3.0 4.5 E - E (keV) E - E (keV) R R
Figure 3.3. NRP for Al2O3 on 2.2nm of silicon dioxide deposited on HF-last silicon. Where the scale is normalized to Al2O3 and bulk silicon.
35
1.0 As Deposited 900ºC Anneal
0.8
0.6
Absorbance (a.u.) Absorbance 0.4
0.2
1100 1000 900 800 700 600 500 Wavenumber (cm-1)
Figure 3.4. FTIR for >100nm of aluminum oxide on silicon before and after a
900ºC anneal.
36
1000ºC
30s Anneal 900ºC
800ºC Si(200) Intensity (arbitrary units) (arbitrary Intensity
As Deposited
30 35 40 45 50 55 60 65 70 75 80 2Θ (Degrees)
Figure 3.5. X-Ray Diffraction, XRD, for aluminum oxide on Si(100).
37
NMOS Capacitors PMOS Capacitors 2.0 : 1.5 nm 2.0 1.2 nm : 1.7 nm 3.5 nm : 2.6 nm : 3.5 nm ) ) 2 2 1.5 1.5 F/cm F/cm µ µ
1.0
1.0
0.5 Capacitance/Area ( Capacitance/Area ( Capacitance/Area
0.5
0.0 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Vg (Volts) Vg (Volts)
Figure 3.6. C-V data for aluminum oxide on HF-last silicon.
38
2.5
2.0 p-MOS
1.5
1.0
0.5 n-MOS
0.0
-0.5 Flat Band Voltage (Volt) -1.0
-1.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Equivalent Oxide Thickness (nm)
Figure 3.7. Flatband voltage for Al2O3 on HF-last silicon as a function of the aluminum oxide EOT.
39
2.5 Interface Layer Q (cm-2) = - 11×1012 HF Last f 1.0nm SiO 2.0 2 1.6nm SiO 2 12 1.5 - (7±1)×10
1.0
p-MOS 0.5 12 - (7±1)×10
0.0
-0.5 n-MOS Flat Band Voltage (Volt) -1.0
-1.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Equivalent Oxide Thickness of Al O Layer (nm) 2 3
Figure 3.8. Flatband voltage for stacked Al2O3 and SiO2 MOSCAP's as a function of the aluminum oxide EOT.
40
10 10 NMOS PMOS 1.6nm 1.3nm 1 1 2.0nm 1.4nm 2.3nm 1.8nm 0.1 2.9nm 0.1 2.0nm
0.01 0.01
) 1E-3 1E-3 2
1E-4 1E-4 J (A/cm 1E-5 1E-5
1E-6 1E-6
1E-7 1E-7
1E-8 1E-8 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 -0.5 0.0 0.5 1.0 1.5 Vg-Vfb (Volt) Vg-Vfb (Volt)
Figure 3.9. Leakage current density versus VG-VFB.
41 n-MOS p-MOS a)
(i) (iv)
(ii) (iii)
b)
Figure 3.10. Ideal band bending for n- and p-MOS devices. Where (i) aluminum gate metal, (ii) aluminum oxide layer, (iii) silicon dioxide interfacial layer, (iv) silicon substrate. a) Flat band condition. b) Accumulation condition where both are biased with an equal magnitude of gate voltage. For n-MOS the device experiences gate injection and for p-MOS there is substrate injection.
42
0.9 50ºC 1.2 200ºC 0.8
1.0 0.7 ) 2 0.8 SiO Al O 0.6 2 2 3 F/cm
µ
0.5
C/A ( 0.6
0.4 0.4
0.3
0.2 -2 -1 0 1 V (Volt) G
Figure 3.11. Temperature Dependence of C-V for n-MOS Al2O3 and SiO2. SiO2 is included as a reference and will not be discussed.
43
1E-4 225ºC
1E-5 ) 2
J (A/cm 1E-6
1E-7 50ºC
-4.5 -4.0 -3.5 -3.0 -2.5 -2.0 V -V (Volt) G FB
Figure 3.12. Temperature dependence of J-V for Al2O3 on HF-last silicon.
44 4 Properties of RPECVD Tantalum Oxide.
4.1 Introduction.
In this chapter results, of RPECVD Ta2O5 will be briefly introduced and
discussed. If the reader would prefer a more in depth discussion please refer to
Reference 1, a comprehensive review article.
FTIR was used to determine the crystallization temperature for film
thicknesses >100nm. The interfaces with (i) HF- last, H-terminated silicon and (ii)
remote plasma oxidized silicon were investigated by AES. C-V and J-V
characteristics were measured from room temperature to 250°C.
AES shows that an interfacial layer occurs for Ta2O5 in contact with
silicon. The quantity and composition is difficult to determine with AES alone.
Other researchers have shown this layer to be a silicate in composition.2,3 It is
worthwhile to note that similar interfacial results are shown for Al2O3 deposition
methods other than RPECVD. (See chapter 3)
J-V results demonstrate a small band offset with silicon as measured4 and
calculated5 by other researchers. Temperature dependent J-V results further
indicate a thin interfacial layer with a higher band offset than Ta2O5.
4.2 Interfacial Formation on HF-Last Silicon and Preoxidized Silicon.
Similar to chapter 3, AES was gathered at 10-second intervals during the
initial stages of deposition of Ta2O5. The AES spectra shown in Figure 4.1
display the evolving chemical composition and bonding at the Si-interface and
45 within the deposited Ta2O5 thin film. The left-hand plot is Ta2O5 deposited directly
on HF-last, H-terminated silicon. The right-hand plot is Ta2O5 deposited on 0.6
nm of SiO2 that was formed by remote plasma assisted oxidation, i.e., exposing
the HF-last silicon for 30 seconds to active O-atoms and metastable molecules
extracted from a remote He/O2 plasma discharge. The thickness of the interfacial
SiO2 layer was determined by comparing the relative amplitudes of the Si-O
(76eV) and the Si-Si (91eV) AES features.6
For 10 seconds of Ta2O5 deposition the two AES spectra in Figure 4.1 are
essentially the same. In particular, the ratio of intensities of the Si-O and Si-Si features indicates that approximately 0.6nm of SiO2 was formed during the initial
stages of the Ta2O5 deposition on HF-last silicon.
For increasing deposition times the line-shapes of the Ta-O features at
~45 and ~175 eV do not evolve. They are a single shape and are located at
energies consistent with tantalum bonded to oxygen.7 From this we conclude that
there is no silicide formation at the interface and the deposited film is fully oxidized.
For further depositions the substrate, Si-Si, and the Si-O signals are attenuated. The apparent increase in the Si-O:Si-Si ratio can be related to one of two possibilities (i) an increase in the amount of the silicon dioxide or silicate layer or (ii) an increase in the amount of Ta2O5. It is unclear which effect
dominates the spectra.
If one looks to past literature, the likely interpretation would be one of an
increasing silicate layer. However, from experience with Al2O3, chapter 3, we
46 have found that the RPECVD method can successfully limit the amount of interfacial reactions that occur. Without other methods, i.e., NRP, SIMS profiling,
TEM, or medium energy ion scattering, it is difficult to determine the composition and quantity of the layer with any precision.
4.3 FTIR Results.
FTIR results for as-deposited Ta2O5 and for a 30 second rapid thermal
anneal, RTA, at 800°C are displayed in Fig 4.2. The as-deposited spectrum
indicates two distinct features at 628 cm-1, with a shoulder at ~520-550 cm-1, and
at 280 cm-1. The emergence of the sharp features after the 800°C RTA indicates
crystallization has occurred and is confirmed by XRD.8
Other researchers have combined these FTIR results with Raman
9 8 10 spectroscopy, EXAFS, and XRD and have found that as-deposited Ta2O5 is octahedrally coordinated. EXAFS further shows that there are two different bond distances for the nearest neighbor with an approximate ratio of 2:1. Since the Ta atom is in an octahedral site, the shorter bond length can be assigned to two neighbors and the longer to four.
4.4 Electrical Characterization.
Figure 4.3 is results of 0.01 to 1 MHz C-V measurements at room temperature for p- and n-type substrates, n- and p-MOS respectively. Similar to
Al2O3 at elevated temperatures, a shoulder is observed for Ta2O5 at room
temperature. The state is also observable in temperature dependent studies,
47 Figure 4.4. As with Al2O3 this shoulder can be assigned to an electronic state
with an energy level located in the silicon band gap.Ch. 3, 11
The flat band voltage for the p-MOS capacitors is approximately the
correct value considering the doping level of silicon substrate and the aluminum
gate electrodes. For n-MOS the value is slightly higher than expected. This may
be due to an increase in tunneling current due to the asymmetric band bending
between n- and p-MOS devices that results in higher leakage current for p-MOS devices.12, Ch. 3 An increase in leakage current can cause the capacitance meter
to give false readings since the equivalent circuit used to determine the
capacitance is no longer valid.13
Room temperature J-V is displayed in Figure 4.5. The values of J are
higher than those measured for similar Al2O3 devices. Since the dielectric
constant of Ta2O5 has been measured to be >20, an EOT of 2.0 nm would correspond to a physical thickness of >10.5 nm.1 If the conduction band offset
was similar to Al2O3 then the measured leakage current for this thickness would
be surprisingly high. This is not the case. The band offset is only about 0.3 eV
4,5 compared to Al2O3’s of ~2.2 eV. Therefore a higher leakage current is
expected.
Figure 4.6 displays J-V as a function of temperature. As the temperature
increases, a systematic increase in J follows. If J is plotted as a function of 1/T at
a fixed voltage it is possible to calculate a temperature dependent activation
energy, W. (Fig. 4.7) The activation energy is of the form: (See chapter 5 for
details)
48 [Eq. 4.1] βt = βo exp (-W/kT)
Where βt is a temperature dependent parameter for tunneling and βo is an
appropriate prefactor. An activation energy of 0.6 eV is calculated from the fit.
This value is approximately double the value we would expect if the
measurement was a result of only the conduction band differences between
Ta2O5 and Si.
4.5 Conclusions.
In this chapter, we have found that Ta2O5 has a crystallization temperature
between 700°C and 800°C. This result is for a relatively thick layer and as the
thickness of the layer approaches the crystallite size, there maybe a suppression
of the crystal formation.
AES indicates that an interfacial layer forms during the deposition of Ta2O5
and maybe a silicate in composition, as past researchers have discovered. Our
techniques do not allow us to resolve this but do indicate that layer is less than
2.5nm. Further, the AES indicates there is no metal-metal bonding, i.e., no metal
Ta or silicide formation.
C-V data presents a state in the silicon band gap that is observable even
at 25°C. The J-V data for room temperature indicates a higher current density
than similar Al2O3 results. The temperature dependent J-V demonstrates a 0.6
eV activation energy. This activation energy may be related to the conduction
band offsets as in tantalum aluminates.14, Ch.5 This larger than expected value
could be the result of an interfacial silicate layer.1,15
49 Since Ta2O5 has been studied in depth and it does not present itself as a
strong candidate for CMOS gate replacement. We have not delved as deeply into
these results as we would have liked. So the results of our electrical
measurements will have to stand as only preliminary work and further information should be attained through the various articles referenced.
4.6 References.
1 C. Chaneliere, J. L. Autran, R. A. B. Devine, and B. Balland, Mater. Sci. Eng., R. 22, 269 (1998).
2 G.D Wilk, R.M Wallace, and J.M. Anthony, Appl. Phys. Rev. 10, 5243 (2001).
3 G. B. Alers, D. J. Werder, Y. Chabal, H. C. Lu, E. P. Gusev, E. Garfunkel, T. Gastafsson, and R. S. Urdahl, Appl. Phys. Lett. 73, 1517 (1998).
4 S. Miyazaki, J. Vac. Sci. Technol. B 19, (2001). (in press)
5 J. Robertson, J. Vac. Sci. Technol. B 18, 1785 (2000).
6 H. Niimi, K. Koh, and G. Lucovsky, Electrochemical Society Proc. 12, 623 (1996).
7 L.E. Davis, N.C. MacDonald, P.W. Palnberg, G.E. Raich, and R.I. Weber, Handbook of Auger Electron Spectroscopy 2cnd. ed. (Physical Electronics Industries, Inc.) Eden Pairie, MN (1976).
8 B. Rayner, H. Niimi, R. Johnson, B. Therrien, G. Lucovsky, and F. L. Galeener, AIP Conf. Proc. 550, 149 (2000).
9 F. L. Galeener, W. Stutius, and G. T. McKinely, The Physics of MOS Insulators, edited by G. Lucovsky, S. T. Pantelides, and F. L. Galeener, New York: Pergamon Press, 1980, pp. 77-81.
10 N.C. Stephenson and R.S. Roth, J. Solid State Chem. 3, 145 (1971).
50
11 E.H. Nicollian and J.R. Brews, MOS (metal oxide semiconductor) Physics and Technology, (John Wiley and Sons, Inc.), New York, (1982).
12 R. S. Johnson, G. Lucovsky, and I. Baumvol, J. Vac. Sci. Technol. A 19, 1353 (2001).
13 D. K. Schroder, Semiconductor Material and Device Characterization 2cnd. ed. (John Wiley and Sons, Inc.), New York, (1998).
14 R. S. Johnson, J. G. Hong, G. Lucovksy, J. Vac. Sci. Technol. B 19, 1606 (2001).
15 S. Banerjee, B. Shen, I.Chen, J. Bohlman, G. Brown, and R. Doering, J. Appl. Phys. 65, 1140 (1989).
51 x9
30 sec Ta O x2 2 5 30 sec Ta O 2 5
x9 x2 20 sec Ta O 20 sec Ta O 2 5 2 5
x2 x4.5 10 sec Ta O 10 sec Ta O 2 5 2 5
dI/dE (arbitary units) (arbitary dI/dE 30 sec O Plasma 2 dI/dE (arbitary units) (arbitary dI/dE on HF Last Si HF Last Si
40 60 80 100 120 140 160 180 200 40 60 80 100 120 140 160 180 200 Energy (eV) Energy (eV)
Figure 4.1. AES of tantalum oxide deposition on HF-last silicon and 0.6nm of silicon dioxide on HF-last silicon.
52
1.0 As Deposited 800°C Anneal
0.8
0.6
0.4
0.2 Absorbance (a.u.)
0.0
-0.2 1200 1000 800 600 400 200 Wavenumber (cm-1)
Figure 4.2. FTIR for >100nm of tantalum oxide on silicon before and after an
800°C anneal.
53
1.4 1.4 PMOS - 2.3nm NMOS - 2.1nm V = -1.00±0.01 V V = 0.35±.01 V FB FB 1 MHz 1 MHz 1.2 100 kHz 1.2 100 kHz 10 kHz 10 kHz ) 2 1.0 1.0 F/cm µ 0.8 0.8 C/A (
0.6 0.6
0.4 0.4
-2.0 -1.5 -1.0 -0.5 0.0 0.5 -1.0 -0.5 0.0 0.5 1.0 V (Volt) V (Volt) G G
Figure 4.3. Room temperature C-V for Ta2O5 on pre-oxidized silicon.
54
1.6 50ºC Ta O 1.4 2 5 200ºC
1.2 )
2 SiO 1.0 2 F/cm
µ 0.8 C/A ( 0.6
0.4
0.2 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 V (Volt) G
Figure 4.4. 1 MHz Temperature Dependence C-V for n-MOS Ta2O5 and SiO2.
SiO2 is included as a reference and will not be discussed.
55
100
10
1
0.1
0.01 ) 2 1E-3
1E-4
J (A/cm 1E-5
1E-6 n-MOS p-MOS 1.3nm 1.2nm 1E-7 1.7nm 1.5nm 2.6nm 2.3nm 1E-8 -4 -3 -2 -1 0 1 2 V (Volt) G
Figure 4.5. J-V for Ta2O5 on pre-oxidized silicon.
56
1 250ºC
0.1
0.01 ) 2 1E-3
J (A/cm 1E-4
1E-5
25ºC 1E-6 0.0 0.2 0.4 0.6 0.8 1.0 1.2 V -V (Volt) G FB
Figure 4.6. Temperature dependence of J-V for Ta2O5 on pre-oxidized silicon.
57
0.1
0.01 ) 2
0.62eV
1E-3 J (A/cm
1E-4
2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 1000/T (K-1)
Figure 4.7. J vs. 1/T for n-MOS Ta2O5 device at VG-VFB = 0.4 V.
58 5 Properties of Tantalum and Hafnium Aluminates.
5.1 Introduction.
In chapter 1, we discussed the need for high dielectric constant materials to replace SiO2 as a gate oxide. In chapter 3, we established that a large amount of
intrinsic fixed negative charge, ~6x1012 cm-2, is present at the internal dielectric
1,2 interface of Al2O3 and SiO2. Alloying Al2O3 with an oxide that has inherent
positive interfacial charge such as Ta2O5 or HfO2 represents a possible way to
compensate the negative charge and thereby provide a charge neutral interface.
The coordination of Al-atoms and Ta- and Hf-atoms in their non-crystalline
oxides suggests that Ta- and Hf-atoms substitutionally replace the 6-fold
coordinated Al-atoms in the (Ta2O5)x(Al2O3)1-x or (HfO2)x(Al2O3)1-x alloys and form
a homogenous oxide with unique electrical and physical characteristics.3,4 This
expectation is consistent with the results presented in this chapter.
Robertson has calculated the band gaps and band offset energies with
5 respect to Si for Ta2O5, HfO2 and Al2O3. He has predicted values of 0.36 eV for
the Ta2O5-Si band offset, 1.5 eV for the HfO2-Si band offset and 2.8 eV for the
Al2O3-Si band offset. These assignments are based on crystalline forms of these
oxides. Miyazaki has determined values of 0.28 eV and 2.08 eV for these offset
energies for amorphous Ta2O5 and Al2O3, respectively, using photoemission
6 spectroscopy. The difference between the two values for Al2O3 is consistent with
the differences in coordination between crystalline corundum or α-Al2O3, where
59 the coordination is 6, and the non-crystalline Al2O3, where the ratio of 4- to 6-fold coordinated Al is 3:1.7,8
The lowest lying conduction band states of Al2O3 are associated with 3s and
3p states of Al, whereas the lowest lying conduction band states of Ta2O5 and
5 HfO2 are associated with 5d states of Ta and Hf. Additionally, the anti-bonding
d-states are localized on the Ta and Hf atoms,9 whilst, the Al 3s and 3p states
are significantly more delocalized, i.e., they are better characterized as extended
transport states than localized trapping states.
5.2 Bulk Composition and Properties by AES, RBS, and FTIR.
AES for Ta-aluminate, Figure 5.1, indicates the end-member peak locations
under go a small energy shift for the aluminates. This result is consistent with a
second neighbor effect where Ta substutionally replaces an Al atom and vice
versa. In this arrangement, the intermediary oxygen atom would shield the
individual metal atoms and cause only a slight perturbation in their electronic
structure.10,Ch.3,Ch.4
The infrared active vibrational modes of FTIR for the alloys are not simply a
superposition of the end-member compounds, but rather display features that are
indicative of new bonding arrangements expected for a homogeneous pseudo-
binary alloy (Figs. 5.2 to 5.5). Further FTIR indicates the crystallization
temperatures are increased by at least 100ºC from the end-members. The Al rich
compositions are increased from 900ºC (Al2O3) to 1000ºC and the Ta and Hf rich
11,12 compositions are increased from 800ºC (Ta2O5) and 700ºC (HfO2) to 900ºC.
60 RBS was performed by the Evans Analytical Group13 to determine the
concentration of Al, Ta and Hf in the films. Figures 5.5 and 5.6 plot the amount of
Ta2O5 and HfO2 to the ratio the AES Al-O:Ta-O and Al-O:Hf-O signals. With this calibration, it is possible to make an AES scan of any future tantalum and hafnium aluminate films and determine the concentration of Al, Ta and Hf.
5.3 Interface Formation by AES.
The interface formation for the alloys was investigated with on-line AES at the initial stages of film deposition (Figs. 5.8 and 5.9). The deposition process was interrupted every 10 seconds, the sample was transferred to the AES chamber, and an AES scan was taken. The spectrum revealed two silicon peaks, one for silicon bonded to silicon, Si-Si, at 91eV and another for silicon bonded to oxygen,
Si-O, at 76eV. The Si-Si peak was present for hydrogen-terminated silicon at the start of the deposition. The 76 eV peak appeared after 10 seconds along with decreased amplitude of the 91 eV peak. The ratio of the two is consistent with
~0.6nm of SiO2 being formed at the interface during the initial stages of
deposition.14 Additionally the Al-O (54eV), and Ta-O (40, 160 and 172 eV) or Hf-
O (40, 165, and 173 eV) peaks were also observed. After 40 seconds, the Si peaks are no longer visible; preceding this the Si-O becomes increasingly larger in relation to the Si-Si peak. The evolutions of the Si-O and Si-Si peaks are consistent with an increasing overlayer causing the deeper layers to be attenuated faster.
61 Previous results from Al2O3 indicate that this SiO2 layer remains relatively thin, 0.6 to 1.0 nm, and is not increased for increasing deposition times or during
1,Ch.3 post-deposition annealing. However, interfacial studies of Ta2O5 demonstrate the possibility of growth of a larger interfacial layer.Ch.4
The results of the tantalum aluminates differ from Ta2O5 in that the interfacial layer is buried after 30 second of depositions. Ta2O5 spectra further indicate an increase in the Si-O signal for longer times, while for the tantalum aluminates the ratio of Si-O:Si-Si is not altered significantly for increasing deposition times. We believe the Al2O3 may have stabilized the interface and only an interfacial layer of
~0.6nm was formed. This layer should be stable for longer deposition times and for rapid thermal anneals.
5.4 Electrical Characterization.
C-V data was taken at 1Mhz for the devices at room temperature, and are shown for the end-members and alloys in Figs. 5.10 and 5.11. The alloys display flat band shifts to more positive voltages in comparison with the respective end- member values. These positive increases in flat band voltage have been attributed to negative charge injected into, and trapped at the Si interface, or within the bulk oxide.15 The dependence of flat band voltage on thickness has
1,Ch.3 been used to determine the location of this charge in Al2O3 dielectrics. For
Al2O3 devices there is a linear dependence on thickness indicating that the charge is located at the interface between the Al2O3 film and the interfacial SiO2 layer.15 This assignment is also supported by the temperature dependent C-V
62 and J-V traces, where the absence of significant hysteresis indicates that the dominant contribution to the positive flat band voltage shift is due to fixed, not trapped charge.
The flat band voltage dependence has not been investigated for the alloys, where the magnitude of the shift to positive values is much larger than in equivalent end-member layers, and also includes significant hysteresis. This is consistent with increased negative charge, due to trapping of electrons that masks smaller changes in the flat band voltage. The direction of this hysteresis at room temperature is consistent with electron trapping in the oxide, or in the immediate vicinity of the interface.16
Temperature dependent C–V data for Al2O3, Ta2O5, and HfO2 devices is
displayed in Figure 5.12. At 200°C the Al2O3 displays a shoulder in the C–V curve; a similar shoulder is also visible for all temperatures in both Ta2O5 and
HfO2 devices. Following the modeling approach in Reference 15, this shoulder
has been assigned to an interfacial electron trapping state in the silicon band
gap.
Figure 5.13 shows the temperature dependence of C–V and J–V for the 43%
Ta2O5 alloy devices on p-type silicon. The hysteresis in C–V traces was found to
decrease initially with increasing temperature, and then reverse direction at the
highest temperature used in this study. Over the same temperature range, the
flat band voltage decreases as well. Similar C-V data could not be obtained for
devices on n-type silicon substrates due to significantly higher leakage currents.
63 Figure 5.14 shows the temperature dependence of C–V and J–V for the 38%
HfO2 alloy devices on p-type silicon. In this case the hysteresis increases, rather
than decreases, for temperatures above 200ºC. From 25ºC to 200ºC the flat-
band voltage increases slightly, and above these temperatures the flat-band is
mostly constant for the positive to negative gate voltage sweeps. Figure 5.15 is
the plot for 48% Hf-aluminate on n-type silicon substrate, and indicates that there
is no significant change in the flat-band voltage or the hysteresis until a
temperature of 250ºC is applied.
The temperature dependence of the J-V traces is displayed in Figure 5.13(b),
5.14(b) and 5.15(b); the x-axis is plotted as a function of the square root of the electric field across the dielectric. This dependence is expected for a Frenkle-
Poole type process. Figures 5.16 and 5.17 display J versus 1/T at a fixed field, and indicate different temperature dependent activation energies for the two aluminate alloy systems.
5.5 Discussion.
Shifts in flat band voltages are generally accepted to be due to (i) fixed charge, Qf, (ii) mobile charge, Qm, (iii) oxide trapped charge, Qot, and/or (iv)
16 interface trapped charge, Qit, sometimes called Dit. Hysteresis in C-V
measurements is normally attributed (i) to charge trapping in the dielectric, or at
the dielectric interface, (ii) mobile charged species or (iii) remnant polarization as
in ferroelectrics.16
64 For the entire temperature range investigated for the Hf-aluminate devices, the direction of hysteresis is consistent with charge trapping. Mobile charge is
generally attributed to positive ionic impurities such as Na+, Li+, and K++ and gives
a hysteresis loop in the opposite sense as shown in Figures 5.14(b) and
5.15(b).15 None of the organic precursor sources contain these metals as
constituents, or as impurities. We have followed standard SiO2 processing
guidelines that in our previous results have indicated no evidence for mobile
charge in deposited SiO2. In addition, there is no evidence for these impurities in
the respective end-member oxides.
The decrease and reversal of hysteresis for Ta-aluminate device is consistent
with a release of trapped charge, and the onset of a conduction process through
11 the extended states of the Al2O3 alloy component.
The temperature dependence of the conduction through an insulating film can
be derived from two mechanisms: (i) carrier excitation, such as trap release, and
(ii) transport.17 Each of these mechanisms can have a temperature dependence
described in the form of Equation 5.1, where W is the activation energy, and βo is
an appropriate prefactor.
[Eq. 5.1] βt = βo exp (-W/kT)
A single activation energy has been used to characterize the temperature
dependence for Ta-aluminate in Figure 5.16. We have not made any attempt to
separate the excitation and transport mechanisms because of differing
assumptions that are required for specific transport processes. Instead we have
analyzed the temperature dependent J-V in terms of a carrier activation process
65 alone. This approach implicitly assumes that over the range of temperatures explored any contributions to the activation energy from temperature dependent bulk mobility are small with respect to the trap release energy. The activation energies derived in this way are interpreted using an energy level scheme in
Figure 5.18 that is consistent with experimentally determined conduction band offset energies,5,6 and with defect states that are generated by a disruption of the
18 Al2O3 network bonding.
The data in Figure 5.16 for Ta-aluminate, is approximated by two exponential
functions with different activation energies. The value of the low temperature
activation energy is consistent with a localized Ta d-state that acts as an electron
trap at approximately 0.3 eV above the Si-conduction band edge. At low
temperatures trap filling dominates, and at high temperatures trap release
dominates. The larger activation energy at higher temperatures, of 1.4 eV is
attributed to the energy necessary for the electrons to escape from the localized
Ta d-state traps into the delocalized/extended Al2O3 conduction band states. This
release from localized traps into extended transport state also accounts for the
shift in flat band voltages back to negative values, and the decrease and reversal
in hysteresis sense for the higher temperature regime.19
The addition of these activation energies along with a calculated Schottky
barrier lowering of 0.2 eV for a field of 6×106 V/cm should then be approximately
15 equal to the conduction band offset between Al2O3 and Si. This addition of the
activation energies and the Schottky barrier lowering give a value of ~2.0 eV in
66 agreement with the measurements of Reference 6 which indicate a band offset energy of ~2.1 eV.
Using a similar argument, Hf-aluminate in the Fig. 5.17(a) can also be fit with two exponential functions. However, unlike the Ta-aluminates C-V traces, the Hf- aluminates C-V traces do not display a decrease in flat band voltage or hysteresis in the high temperature regime. For Hf-aluminates, the process characterized by the second energy produces increases in both the hysteresis and flat band voltage. This is more consistent with a two-level trapping system where in the high temperature regimes electrons are not injected in the aluminum oxide conduction band states, but instead into a different set of localized traps.
If we assume the two energies for the Hf-aluminates correspond to two different energy trap levels then the initial energy, 0.2 eV, is the barrier height for injection into the first trap, and the second, 0.9 eV, is the difference between the two states, modified by an appropriate Schottky barrier lowering. To calculate the level of the second state, we add the two activation energies, also including a
Schottky barrier lowering of 0.3 eV, which is appropriate to the higher field of
9×106 V/cm. This results in an energy of 1.4 eV for the second trapping state.
This is very close to the 1.5 eV offset of the HfO2 conduction of band with respect
to the silicon conduction band.5 In this model, the second level is assigned to a
localized Hf d-state.
The question of the origin of the first activation energy for Hf-aluminates
remains to be determined. Even though it is at approximately the same energy as
the initial trapping state energy reported for the Ta aluminate alloys, it cannot be
67 assigned to an intrinsic state of the Hf aluminate alloys. Molecular orbital theory
calculations are being performed by our group and collaborators on transition
metal silicate and aluminum alloys.18 These studies are consistent with the
bonding of metal atoms in silicate glasses.20
The addition of Hf(Zr)O2 into a SiO2 continuous random network results in two
broken Si-O bonds that then form four terminal network oxygen atoms with
negative charge, introducing a silicate bonding feature into the infrared absorption spectrum at 910-20 cm-1.21 Studies of Zr and Hf silicate alloys indicate
1- that the bonding and anti-bonding states of this terminal, SiOterminal group must
be well removed from the silicon band gap, since they are not active as trapping/defect states.
For the Hf aluminate alloys a similar disruption occurs for the addition of HfO2
into an Al2O3 host. Based on the local bonding of Al atoms in Al2O3, this alloy-
1- atom oxide insertion will also disrupt the network bonding AlO4/2 component of
1- - the Al2O3, and generate two terminal AlOterminal groups. However, the AlOterminal
does not occur naturally in any crystalline material and therefore is assumed to
1- be less stable than the SiOterminal group, which is a constituent of all silicate
glasses in the low alloy regime.20 In addition, empirical calculations based on
electronegativity models indicate that it is significantly less stable than the
1- SiOterminal group due to a higher ratio of electrons to nuclear charge. This
1- suggests that the bonding and anti-bonding states of the AlOterminal may be in, or
near the band gap of silicon and therefore be active as trapping/defect states.
For the case of Ta-aluminates, the localized Ta d-state is low in energy, and may
68 overlap with this defect state, thereby making it difficult, if not impossible, to distinguish between the defect state and the localized Ta d-state based on an analysis of J vs. T data alone. 5,11 This point is supported by the quantitative differences between low temperature trapping between the Ta and Hf aluminate alloys. The trapping in the Ta aluminate devices is larger, consistent with the trapping state being an intrinsic energy level of the alloy, rather than a defect that results from the alloy formation process.
Figure 5.17(b) is for similar Hf-aluminate results on n-silicon substrates. The temperature range is not as large as p-silicon because of a significant increase in leakage current.1 Here there is also only one activation energy present; however,
the possibility of a second higher temperature activated trapping process is
suggested by an increase in hysteresis of the C-V trace at the 250ºC
measurement temperature.
A much larger energy barrier for holes injection eliminates the possibility that
we are measuring hole trapping effects. The valence band offset energy with
6 respect to silicon, as measured by photoemission spectroscopy for Al2O3 is 3.75
5 eV and 3.25 eV for Ta2O5, and as calculated for HfO2 is 3.4 eV.
5.6 Conclusions.
From the temperature dependence of C-V and J-V traces for Al2O3-HfO2
alloys, we have concluded that there are two localized electron traps at
approximately 0.2-0.3 eV and 1.4eV above the Si conduction band edge. The
1- 0.2-0.3 eV traps are assigned to AlOterminal bonding groups that derive from a
69 breaking or rupturing of the network component of the Al2O3 host material, and the 1.4 eV traps are assigned to anti-bonding Hf atom d-states that form the
lowest conduction band states of these alloys. In the Ta aluminate alloys, two
activation energies have been determined, one for trapping into Ta d-states ~ 0.3
eV above the silicon conduction band edge, and a second for release into
11 conducting states of the Al2O3 host matrix. The release of electrons from the Ta
d-state traps into conducting states of the Al2O3 matrix is accompanied by a
decrease and reversal in hysteresis, and a decrease in the flat band voltage shift.
Thus, in contrast to the Ta-aluminates, the Hf-aluminates demonstrate a two
level trapping system, where at low temperatures there is a trapping state due to
1- the AlOterminal groups, and at higher temperatures, a release from this state into
to a localized Hf d-state. In the high temperature region, the flat band increases
because the two states are now each active as traps. At still higher temperatures,
the electrons should eventually be thermally emitted from the localized Hf d-
states into the aluminum oxide matrix conduction band, resulting in a decrease in
flat band voltage and hysteresis.
Recent experiments performed in our laboratory have used X-ray absorption
22 spectroscopy, XAS, to study Zr and Ta anti-bonding d-states in ZrO2-SiO2 and
23 Ta2O5-Al2O3 alloys with essentially the same result. Zr and Ta introduce
localized d-states whose energies with respect to deep core states, are
independent of the alloy composition and the state of crystallization. These
localized states do not mix in any significant way with the extended s-like
conduction band states of the SiO2 or Al2O3 alloy host oxides. The absence of
70 mixing is consistent with a band picture in which aluminum oxide acts as the bulk transport matrix, and the localized trapping centers are distributed throughout the bulk and have energies with respect to the Al2O3 and Si conduction bands which
are essentially independent of the alloy compositions.
5.7 References.
1 R.S. Johnson, G. Lucovsky, and Isreal Baumvol, J. Vac. Sci. Tech. A 19, 1353 (2001).
2 G. Lucovsky, J.C. Phillips and M.F. Thorpe, in Proc. of Characterization and Metrology for USLI Technology, AIP Conf. Proc. 550, 154 (2001).
3 D. Muller, W.Gessner, H.J. Behrens, and G. Scheler, Chem. Phys. Lett. 79 (1), 59 (1981).
4 N.C. Stephenson and R.S. Roth, J. Solid State Chem. 3, 145 (1971).
5 J. Robertson, J. Vac. Sci. Technol. B 18, 1785 (2000).
6 S. Miyazaki and M. Hirose, AIP Conf. Proc. 550, 89 (2001).
7 G. Lucovsky, G., A. Rozaj-Brvar, and R.F. Davis, in The Structure of Non- Crystalline Materials 1982, edited by P.H. Gaskell, J.M. Parker and E.A. Davis (Taylor and Francis, London, 1983), p. 193.
8 Rayner, H. Niimi, R. Johnson, R. Therrien, G. Lucovsky and F.L. Galeener, in Proc. of Characterization and Metrology for USLI Technology, AIP Conf. Proc. 550, 149 (2001).
9 G. Lucovsky, Y. Zhang and J.L. Whitten, Appl. Surf. Sci. (2001), in press.
10 L. E. Davis, N. C. MacDonald, P. W. Palnberg, G. E. Raich, and R. I. Weber, Handbook of Auger Electron Spectroscopy 2cnd. ed. (Physical Electronics Industries, Inc.) Eden Pairie, MN (1976).
11 R.S. Johnson, J.G. Hong, and G. Lucovsky, J. Vac. Sci. Tech. B 19, 1606 (2001).
71
12 R.S. Johnson, J.G. Hong, C. Hinkle and G. Lucovsky, J. Vac. Sci. Tech. B, (submitted).
13 http://www.cea.com
14 H. Niimi, K. Koh, and G. Lucovsky, Proc. Electrochem. Soc. 12, 623 (1996).
15 E.H. Nicollian and J.R. Brews, MOS (Metal Oxide Semiconductor) Physics and Technology, (John Wiley and Sons, Inc.), New York (1982).
16 D. K. Schroder, Semiconductor Material and Device Characterization 2cnd. ed., (John Wiley and Sons, Inc.), New York (1998).
17 W.C. Johnson, "Study of Electronic Transport and Breakdown in Thin Insulating Films," Tech. Rep. No. 7, Princeton University, 1979.
18 Y. Zhang, G.Lucovsky and J. Whitten, (unpublished).
19 T.P. Ma, (unpublished).
20 R. Zallen, "The Physics of Amorphous Solids", (John Wiley and Sons), New York, (1983).
21 Lucovsky and Rayner, G. Lucovsky and G. Rayner, Appl. Phys. Lett. 77, 2912 (2000).
22 G. Lucovsky, G.B. Rayner, Jr., G. Appel, R.S. Johnson, Y. Zhang, D.E. Sayers, H. Ade and J.L. Whitten, submitted to Appl. Phys. Lett. (2001).
23 G. Appel, (unpublished).
72
Ta O 2 5
74% Ta O 2 5
57% Ta O 2 5
43% Ta O 2 5
34% Ta O 2 5
11% Ta O 2 5
Al O
2 3 dI/dE (arbitrary units) (arbitrary dI/dE
40 60 80 100 120 140 160 180 Energy (eV)
Figure 5.1. Differential AES spectra for the Al2O3-Ta2O5 alloy system. The Al
LVV feature shifts from 68 eV in metallic Al, to 54 eV in Al2O3 and 56 eV in a
Ta2O5-rich alloy. The feature at 170 eV in metallic Ta, shifts to 163 eV in Ta2O5, and is at 160 eV in an Al2O3-rich alloy. The 1s oxygen atom feature shifts from
504 eV in Ta2O5 to 508 eV in Al2O3.
73
As Deposited
Ta O 2 5
(57% Ta O ) 2 5
(43% Ta O ) 2 5
Absorbance (a.u.) (34% Ta O ) 2 5
Al O 2 3
1000 800 600 400 200 Wavenumber (cm-1)
Figure 5.2. Fourier transform infrared spectroscopy, FTIR, for as deposited Ta- aluminates and their end-members, Al2O3 and Ta2O5. The spectrums of the Hf-
aluminates can not be attained by a linear-superposition of the two end-members
and thus represent new and unique materials.
74
Annealed
800ºC Ta O 2 5
900ºC (57% Ta O ) 2 5
1000ºC
(43% Ta O ) 2 5
1000ºC
Absorbance (a.u.) (34% Ta O ) 2 5
900ºC Al O 2 3
1000 800 600 400 200 Wavenumber (cm-1)
Figure 5.3. FTIR spectra of crystalline Al2O3-Ta2O5 alloy films.
75
As Deposited
HfO 2 70±3% HfO
2
59±3% HfO 2
9±3% HfO 2
Absorbance (a.u.) Absorbance
Al O 2 3
1200 1000 800 600 400 200 Wavenumber (cm-1)
Figure 5.4. Fourier transform infrared spectroscopy, FTIR, for as deposited Hf-
aluminates and their end-members, Al2O3 and HfO2. The spectrums of the Hf- aluminates can not be attained by a linear-superposition of the two end-members and thus represent new and unique materials.
76
Crystalline
HfO 700ºC 2
70±3% HfO 900ºC 2
59±3% HfO 1000ºC 2
9±3% HfO 1000ºC 2 Absorbance (a.u.)Absorbance
Al O 900ºC 2 3
1200 1000 800 600 400 200 Wavenumber (cm-1)
Figure 5.5. FTIR spectra of crystalline Al2O3-HfO2 alloy films.
77
10
1
AES Ratio (Al-O/Ta-O) Ratio AES Experimental Data Fit: x = (-0.126)·ln(AES Ratio / 63.09) 0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
RBS - x from: (Ta O ) (Al O ) 2 5 x 2 3 (1-x)
Figure 5.6. RBS compositional calibration of AES signal levels of Al-O and Ta-O peaks.
78
Experimental Data 10 Fit
x = (-0.181) • ln(AES Ratio / 19.88)
1 AES Ratio (Al-O/Hf-O) Ratio AES
0.1
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 RBS - x from: (Al O ) (HfO ) 2 3 (1-x) 2 x
Figure 5.7. RBS compositional calibration of AES signal levels of Al-O and Hf-O peaks.
79
30 sec. x8.7
20 sec. x8.7
10 sec. x8.7
Ta-O
HF Last Si dI/dE (a.u.) dI/dE
Al-O Si-O
Si-Si
50 75 100 125 150 175 200 225 250 Energy (eV)
Figure 5.8. AES of interface formation for x=0.43, (Ta2O5)x(Al2O3)(1-x).
80
40 sec. x5
30 sec. x5
20 sec. x5
10 sec. x3.3
dI/dE (a.u.) dI/dE
Al-O Hf-O HF Last Si
Si-O
Si-Si
50 75 100 125 150 175 200 225 250 Energy (eV)
Figure 5.9. AES of the initial stages of deposition of Hf-aluminates. These results are consistent with Al2O3 results and indicate a thin layer, 0.6-1.0nm, of silicon dioxide formation at the silicon interface.
81
a) b) Ta O Ta O 2 5 2 5 1.4 1.4
1.2 1.2 Al O 2 3 43% Ta O 2 5 ) 2 1.0 1.0
57% Ta O 2 5
F/cm Al O 2 3 µ 0.8 0.8 C/A ( 0.6 0.6
0.4 0.4
-2 0 2 4 024 V (Volts) G
Figure 5.10. Room temperature, 1 Mhz, C-V plots for Al2O3, Ta2O5, and Ta- aluminates on (a) p-silicon and (b) n-silicon with aluminum gates.
82
a) b) 1.6 1.6 HfO 2
1.4 1.4
48% HfO 1.2 1.2 2 38% HfO 2 ) 2 HfO Al O 1.0 2 Al O 1.0 2 3 2 3
F/cm
µ 0.8 0.8 C/A (
0.6 0.6
0.4 0.4
-2 -1 0 1 -1 0 1 2 V (Volts) G
Figure 5.11. Room temperature, 1 Mhz, C-V plots for Al2O3, HfO2, and Hf- aluminates on (a) p-silicon and (b) n-silicon with aluminum gates.
83
50ºC Ta O 1.4 2 5 Al O 200ºC 2 3 2.0 0.8
1.2 HfO2
1.5 )
2 1.0
0.6
F/cm µ 0.8 1.0 C/A ( 0.6
0.4 SiO 0.4 2 0.5
-2 -1 0 1 -1 0 1 -3 -2 -1 0 1
VG (Volt)
Figure 5.12. Temperature dependent C-V for SiO2, Al2O3, Ta2O5, and HfO2 on p- silicon with aluminum gates.
84
a) b) 0.01 200ºC
1E-3
200ºC
1E-4 ) 2
J (A/cm 1E-5
) 1.0 2
1E-6 0.8 F/cm
µ 0.6 25ºC 1E-7 C/A ( 0.4 25ºC
-10123 1.0 1.5 2.0 2.5 3.0
V (V) 3 G √E (x10 √V/cm)
Figure 5.13. Temperature dependence of (a) C-V and (b) J-V for 43% Ta2O5 Ta- aluminate on p-silicon with aluminum gates.
85 a) b) 1E-3
300ºC
300ºC
1E-4
1E-5 ) 2
J (A/cm
1E-6
) 1.2 2
1.0 50ºC F/cm 0.8
µ
0.6 1E-7 25ºC C/A ( C/A 0.4 -1.0 -0.5 0.0 0.5 1.0 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 V (V) 3 G √E (x10 √V/cm)
Figure 5.14. Temperature dependence of (a) C-V and (b) J-V for 38% HfO2 Hf- aluminate on p-silicon with aluminum gates.
86
a) b) 1E-3 250ºC)
250ºC
1E-4
1E-5 )
2
1.0 J (A/cm
) 1E-6 2 75ºC 0.8
F/cm µ 0.6
C/A ( 50ºC 1E-7 0.4
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.5 1.0 1.5 2.0 2.5 V (V) 3 G √E (x10 √V/cm)
Figure 5.15. Temperature dependence of (a) C-V and (b) J-V for 48% HfO2 Hf- aluminate on n-silicon with aluminum gates.
87
0.01
1E-3 1.4 eV
1E-4 ) 2
0.3 eV 1E-5 J (A/cm
1E-6
V -V = -3.0V 1E-7 G FB 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 1000/T (K-1)
Figure 5.16. J vs. 1/T for 43% Ta2O5 Ta-aluminate on p-silicon. Activation energies are from fits of exp(-W/kT).
88
a) p-Si b) n-Si 38% HfO 2 48% HfO2
1E-4 0.9 ± 0.1 eV 1E-4
0.55 ± .1 eV ) 2
1E-5 1E-5
0.2 ± 0.1 eV J (A/cm
1E-6 1E-6
V -V = -3.5V G FB VG-VFB = 1.5V 1.6 2.0 2.4 2.8 3.2 1.6 2.0 2.4 2.8 3.2 1000/T (K-1)
Figure 5.17. J vs. 1/T for both 38% and 48% HfO2 Hf-aluminates on p- and n- silicon, respectively. Activation energies are from fits of exp(-W/kT).
89 Si-conduction band - 3.15 eV SiOterminal a.b.*
2 eV Al23 O a.b.*
1.5 eV Hf(Zr) d-state a.b.*
Ta d-state a.b.* 0.2-0.3 eV - AlOterminal a.b.* Si-band gap
- AlOterminal b. Si-valence band
- SiOterminal b.
Figure 5.18. Localized bonding, b., and anti-bonding, a.b.*, states in Ta- and Hf- aluminates. The left-hand side indicates the conduction and valence bands for the silicon substrate. On the right are the proposed localized defect states, and bands for the aluminate alloys. The arrows indicate the transitions obtained through the analysis of temperature dependent C-V and J-V measurements. The dashed arrow is for a predicted high temperature transition from the localized Hf d-state (not observed) to the aluminum oxide matrix extended conduction band states.
90 6 Summary and Future Work.
6.1 Aluminum and Tantalum Oxide.
6.1.1 Aluminum Oxide, Al2O3.
For aluminum oxide deposited using a remote plasma enhanced chemical
vapor deposition, RPECVD, method it was found that a thin, 0.6 nm, thermally stable interface layer was formed for deposition on HF-last silicon. In contrast, aluminum oxide was shown to mix with RPECVD silicon dioxide upon an 800ºC rapid thermal anneal, RTA.
Electrical results of RPECVD aluminum oxide indicated a layer of fixed negative charge, 6-8×1012 charges/cm2. This was shown to be associated with
the aluminum oxide and not the silicon substrate by placing a layer of RPECVD
silicon dioxide between the aluminum oxide and the silicon substrate. The
resulting flat band voltage corresponded to a movement of the fixed charge from
the silicon interface to the aluminum oxide / silicon dioxide interface.
The fixed negative charge is consistent with a model for the local atomic
bonding of noncrystalline Al2O3 that has two different bonding environments for
the Al atoms; (i) a tetrahedrally coordinated Al site that has a net negative
charge, and (ii) octahedrally coordinated site in which the Al has a charge of +3.
The negatively charged Al atoms can bond directly to the O atoms of the
interfacial SiO2 and is the arrangement that is responsible for the fixed negative
charge. The tetrahedral arrangement with a negatively charged Al is unique, and
91 essentially all of the other transition metal oxides studied to date display fixed
positive charge at their interfaces with Si or SiO2.
Temperature dependent C-V measurements indicate electron traps in the
immediate vicinity of the Si-dielectric interface; the trapping sites may be intrinsic
and associated with the octahedrally coordinated Al.
6.1.2 Tantalum Oxide, Ta2O5.
Deposition of RPECVD tantalum oxide on HF-last silicon resulted in the
formation of an interfacial layer. The composition of this layer was not fully
characterized, however other researchers have shown it be a silicate.
Electrical characterizations were dominated by high leakage currents,
most likely due to the low conduction band offset of tantalum oxide with silicon,
~0.3 eV. Temperature dependent measurements demonstrated interfacial
trapping states with energy levels in the silicon band gap.
6.2 Tantalum and Hafnium-Aluminates.
FTIR and AES results of RPECVD tantalum and hafnium-aluminates
indicated that the films were homogeneous and pseudobinary in character. FTIR
also demonstrated that the crystallization temperatures were increased by at
least 100ºC from their respective end-members. In-situ AES showed a thin, 0.6 -
1.0 nm, interfacial layer formed during the deposition of HF-last silicon.
Despite these accomplishments an electron trapping system dominated
the electrical properties, where it was shown the traps were a result of (i) a
- network "break-up" component, AlOterminal and (ii) atomic d-states of the Ta and
92 Hf-atoms that form the bottom conduction band states of Ta2O5 and HfO2. Due to
the energy level of the Ta d-states we were not able to make separate
measurements of the network "break-up" and the d-states for the tantalum-
alumintates. For hafnium-aluminates the d-states was high enough in energy that the network "break-up" and the d-states acted at two separate trapping states.
6.3 Future Work.
The electrical results presented in this dissertation indicate that none of
the studied systems will be suitable candidates to replace silicon dioxide in
CMOS applications. However below are several suggestions to improve the
fundamental understanding of these materials and several alternative materials that may yield more desirable results.
1. Develop a chemical based modeling approach to confirm:
(i) The existence and energy level of the network "break-up"
- - components SiOterminal and AlOterminal .
(ii) The localized nature and the energy levels of the Ta and Hf d-
states in their respective oxides and in an aluminum oxide matrix.
2. Engineer an interfacial layer at the substrate and gate interfaces that will
act as an electron-tunneling barrier preventing the filling of aluminate
traps.
(i) This may not succeed since it is possible the traps are initially filled
and the subsequent release at elevated temperatures may create
an internal dipole resulting in hysteresis.
93 3. Aluminum-silicate may provide a temporary solution.
(i) The dielectric constant will be increased from silicon dioxide's
value.
(ii) The extended conduction band states of Al and Si may be close
enough in energy to form a continuous state without trapping.
4. Yttrium and lanthanum oxide have conduction band offsets with respect to
silicon that are similar to aluminum oxide and aluminates of these
materials may yield systems that form a continuous conduction band state
with no trapping sites.
94