...

Growth and Characterization of Ti-Si-N Hard Coatings Axel Flink

by user

on
Category: Documents
4

views

Report

Comments

Transcript

Growth and Characterization of Ti-Si-N Hard Coatings Axel Flink
Linköping Studies in Science and Technology
Licentiate Thesis No. 1270
Growth and Characterization of
Ti-Si-N Hard Coatings
Axel Flink
LiU-TEK-LIC-2006:51
Thin Film Physics Division
Department of Physics, Chemistry, and Biology (IFM)
Linköping University, 581 83 Linköping, Sweden
ISBN: 91-85643-85-8
ISSN: 0280-7971
Printed by LiU-Tryck, Linköping, Sweden, 2006
Abstract
Metastable (Ti,Si)N alloy and TiN/SiNx multilayer thin solid films as well as SiNx/TiN
surfaces have been explored. Cubic Ti1-xSixN (0 x 0.14) films deposited onto cemented
carbide (WC-Co) substrates by arc evaporation exhibited a competitive columnar growth
mode where the structure transforms to a feather-like nanostructure with increasing Si
content as revealed by x-ray diffraction and transmission electron microscopy. X-ray
photoelectron spectroscopy revealed the presence of Ti-N and Si-N bonding, but no
amorphous Si3N4. Band structure calculations showed that phase separation of NaClstructure Ti1-xSixN solid solution into cubic SiN and TiN phases is energetically
favorable. The metastable microstructure, however, was maintained for the Ti0.86Si0.14N
film annealed at 900 °C, while recrystallization in the cubic state took place at 1100 °C
annealing during 2h. The Si content influenced the film hardness close to linearly, by
combination of solid-solution hardening in the cubic state and defect hardening. For x=0
and x=0.14, nanoindentation gave a hardness of 29.9±3.4 GPa and 44.7±1.9 GPa,
respectively. The hardness was retained during annealing at 900 °C.
Nanostructured materials, e.g., nanocomposites and nanolaminates, are defined by
internal interfaces, of which the nature is still under debate. In this work two-phase model
systems were explored by depositing SiNx/TiN nanolaminate films, including
superlattices containing cubic SiNx, by dual target reactive magnetron sputtering. It is
demonstrated that the interfacial phase of SiNx onto TiN(001) and TiN(111) can be
crystalline, and even epitaxial with complex surface reconstructions. Using in situ
structural analyses combined with ab initio calculations, it is found that SiNx layers grow
epitaxially, giving rise to strong interfacial bonding, on both TiN(001) and TiN(111)
surfaces. In addition, TiN overlayers grow epitaxially on SiNx/TiN(001) bilayers in
nanolaminate structures. These results provide insight into the development of design
rules for novel nanostructured materials.
Preface
This Licentiate Thesis is based on my research carried out with the Thin Film Physics
Division at Linköping University in collaboration with SECO Tools AB in Fagersta, the
Materials Science Department at University of Illinois at Urbana-Champaign, and the
Department of Materials Chemistry at Uppsala University. The work is supported by
SECO Tools AB, the Swedish Research Council (VR), and the Swedish Foundation for
Strategic Research (SSF).
Included Papers
Paper I
Influence of Si on the Microstructure of Arc Evaporated (Ti,Si)N Thin
Films; Evidence for Cubic Solid Solutions and their Thermal Stability
A. Flink, T. Larsson, J. Sjölén, L. Karlsson, L. Hultman,
Surf. Coat. Technol. 200 (2005) 1535-1542
Paper II
Toward Understanding Interface Structure in Superhard TiN-SiN
Nanolaminates and Nanocomposites
L. Hultman, J. Bareno, A. Flink, H. Söderberg, K. Larsson, V. Petrova,
M. Odén, J. E. Greene, I. Petrov,
Submitted for publication
Other Papers by the Author
Paper III
Deposition of Ti2AlN Thin Films by Reactive Magnetron Sputtering
T. Joelsson, A. Flink, J. Birch, L. Hultman,
Manuscript in final preparation
Paper IV
MAX-Phase Ti2AlN Coatings by Arc Deposition
A.Flink, J. Sjölén, L. Karlsson, L. Hultman,
In manuscript
Paper V
Growth and characterization of single crystalline TiN/SiNx superlattice
films
H. Söderberg, A. Flink, J. Birch, L. Hultman, M. Odén,
In manuscript
Paper VI
Role of Carbon in Boron Suboxide Thin Films
D. Music, V. M. Kugler, Zs. Czigany, A. Flink, O. Werner,
J. M. Schneider, L. Hultman, U. Helmersson,
J. Vac. Sci. Technol. A21 (2003) 1355
Acknowledgements
Many persons have contributed to my work and I would especially like to thank:
Lars Hultman, my supervisor. Thank you for giving me the opportunity to do both my
diploma work and PhD in the Thin Film Group. I am very grateful for all support you
have given me so far.
Lennart Karlsson, Jacob Sjölén, and Tommy Larsson at SECO Tools AB for fruitful
collaboration. I always enjoy going to Fagersta and SECO for meetings and work.
Javier Bareno, Vania Petrova, and Ivan Petrov from University of Illinois at UrbanaChampaign for taking care of me during my visits in Chambana.
Hans Söderberg and Magnus Odén from Luleå University of Technology. Hasse, for nice
collaboration and discussions. Magnus, for your efforts to always answer my endless list
of questions regarding nanoindentation.
Karin Larsson for introducing me and Javier to CASTEP.
Per Persson, for helping me learning TEM, both by practice and discussions.
Jens Birch for giving fast and accurate answers, and for increasing the already good spirit
in the group.
Hans Högberg, for always taking time to answer my questions, and for encouragement.
Karl-Olof Brolin, Inger Eriksson, and Thomas Lingefelt for your endless helping spirit.
Anders Hörling, for sharing your thoughts, both about science and life in general.
Per Eklund, my old neighbor in Ryd, for our interesting football discussions, and for
recommending me to apply for a diploma work at the Thin Film Physics Division.
Anders E, Johan, and Timo for our unforgettable golf/poker trips.
Erik, Fredrik, Martina, and Naureen for a fantastic week on Iceland!
All friends and colleagues, both past and present, in the Thin Film and Plasma groups. I
really have a lot of fun both at work and beside work with you!
My family, of course, for all your support throughout these years (and the years before I
started in graduate school).
Table of Contents
1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1
2
Hard Coatings for Cutting Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
The Ti-Si-N System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1
2.2
2.3
2.4
2.5
2.6
3
Phase Diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Titanium Nitride . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Silicon Nitride . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
TiN/SiNx Nanocomposites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
TiN/SiNx Multilayers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ternary Solid Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Thin Film Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1
3.2
3.3
3.4
3.4.1
3.4.2
4
Physical Vapor Deposition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Arc Evaporation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
DC Magnetron Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Growth of Metastable Solid Solution Films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Low-temperature Synthesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Ion-induced Recoil Implantation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Theoretical Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.1 Density Functional Theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.1 Approximations for Many-body Interactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.2 Pseudo Potentials and Plane Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.3 Linear Muffin-Tin Orbital. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5
5.1
5.2
5.3
5.4
5.5
6
6.1
6.2
3
3
4
4
5
5
15
16
16
16
Thin Film Characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
X-ray Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Electron Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Nanoindentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Scanning Tunneling Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
X-ray Photoelectron Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
20
21
22
23
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Ti1-xSixN Alloy Films. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
TiN/SiNx Nanolaminate Films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Paper I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Paper II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
1
Introduction
Ceramics is an interesting class of material in the sense of heat resistance, high hardness,
thermal shock resistance, and producabililty as thin films. Ceramics can be defined as
…“solid compounds that are formed by the application of heat, and sometimes heat and
pressure, comprising at least two elements provided one of them is a non-metal or a nonmetallic solid. The other element(s) may be a metal(s) or another non-metallic solid(s).”1
The properties of the ceramics make them attractive for thin film applications within the
metal cutting industry. Many industries are dependent of tools for metal cutting
applications; this has created a strong interest for developing new methods and materials
for making the tools more efficient and cheaper.
1.1
Hard Coatings for Cutting Tools
During the 1970’s replaceable cutting inserts together with hard coatings were
introduced. One of the first wear-resistant coating materials in the modern cutting tool
industry used was TiN.2 It exhibits high hardness and stiffness which makes it suitable as
a cutting tool coating. The main shortcoming is, however, limited stability due to
oxidation at temperatures above 500 °C.3 Today the work temperatures of tools are
typically between 800-1200 °C. Therefore, there is a growing interest concerning coating
materials with improved thermal stability. The present design concept has been to employ
ternary compounds, e.g. (Ti,Al)N4,5. (Ti,Al)N offers improved oxidation resistance, better
thermal stability, defect (compressive residual stress) hardening, as well as newly
discovered age hardening6. Together, these factors improve the life time of the tool, and
provide the possibility to work at higher cutting speed. (Ti,Al)N has been used as a
protective coating for cutting tools since the early 1990’s and is still a work horse for
hard coatings.7 Today, (Ti,Al)N used in metal cutting industry are examples of
metastable hard coatings that can be synthesized by arc evaporation at low temperature.
Further on, the coating technology development and research on ternary compounds
have expanded to cover a range of compounds based on the Ti-Al-N, Cr-Al-N, and most
recently, Ti-Si-N systems (see Paper I). One example from the latter are the TiN/SiNx
nanocomposites8, which exhibit thermal stability and very promising mechanical
properties including superhardness. These nanocomposites consist of TiN
nanocrystallites embedded in what is assumed to be amorphous SiNx. However, there is
still the question about the nature of this tissue phase9 (see Paper II). Also, superhard
2
1
TiN/SiNx multilayers, or nanolaminates, have been synthesized.10,11 Furthermore, recent
work points to the possibility of fabricating (Ti,Si)N solid solutions12,13,14 (see Paper I) as
well as epitaxially stabilized cubic-SiNx (see Paper II). The objective of this thesis is to
explore the synthesis, structure, and properties of the novel (Ti,Si)N alloy and SiNx/TiN
nanolaminate thin films.
1
M. W. Barsoum, Fundamentals of Ceramics, McGraw-Hill (1997)
P. O. Snell, Jernkontorets Anm. 154 (1970) 413
3
H. Ichimura, A. Kawana, J. Mat. Res. 8 5 (1993) 1093
4
O. Knotek, W. Bosch, T. Leyendecker, Proc. 7th Int. Conf. Vacuum Metallurgy, Linz, Austria 1985
5
W. –D. Münz, J. Göbel, Proc. 7th Int Conf. Vacuum Metallurgy, Linz, Austria 1985
6
A. Hörling, PhD Thesis (Linköping Studies in Science and Technology, dissertation no. 922, Linköping
University, Sweden 2005
7
S. PalDey, S. C. Deevi, Mat. Sci. And Eng. A342 (2003) 58
8
S. Veprek, S. Reiprich, Thin Solid Films 268 (1995) 64
9
S. Hao, B. Delley, C. Stampfl, Phys. Rev. B 74 (2006) 035402
10
H. Söderberg, M. Odén, J. M. Molina-Aldereguia, L. Hultman, J. Appl. Phys. 97 (2005) 114327
11
X. Hu, H. Zhang, J. Dai, G. Li, M. Gu, J. Vac. Sci. Technol. A23 (2005) 114
12
F. Vaz, L. Rebouta, B. Almeida, P. Goudeau, J. Pacaud, J. P. Riviere, J. Bessa Sousa, Surf. Coat.
Technol. 120-121 (1999) 166
13
J. L. He, C. K. Chen, M. H. Hon, Mater. Chem. Phys. 44 (1996) 9
14
G. Pezzotti, I. Tanaka, Y. Ikuhara, M. Sakai, T. Nishida, Scr. Metall. Mater. 31 (1994) 403
2
3
2
2
The Ti-Si-N System
In Paper I cubic Ti1-xSixN metastable solid solution coatings deposited by arc evaporation
were studied. In Paper II TiN/SiNx multilayers were deposited by reactive DC sputtering
with different thicknesses of the SiNx layers.
2.1
Phase Diagram
Fig. 2.1 The ternary phase diagram for Ti-Si-N at 1000 °C.1
The phase diagram at 1000 °C in Fig. 2.1 shows that Si3N4 is the only stable Si-N
compound. TiN is stable over a wide stoichiometry range. Furthermore, there is no
ternary phase present.
2.2
Titanium Nitride
TiN is a ceramic which is used in a wide field of thin film applications, from diffusion
barriers to wear-resistant coatings to decorative coatings. TiN has a rocksalt structure
(NaCl) with a unit cell consisting of 8 atoms; 4 Ti and 4 N. The lattice parameter is 4.24
Å.2 TiN exhibits high hardness, 20 GPa3 as single crystal thin film, and 26 GPa4 as
polycrystalline, both grown by reactive DC magnetron sputtering.
4
3
N
Ti
Fig. 2.2 Image illustrating the rocksalt, or NaCl, structure.
2.3
Silicon Nitride
Silicon nitride as a thin film material is mostly used within electronics. It exists as Si3N4
in three different polytypes, two hexagonal, - and , and one amorphous, a.5 There is
also a high pressure, high temperature cubic phase.6 In a-Si3N4 the average binding
distance is 1.74 Å.7
2.4
TiN/SiNx Nanocomposites
A nanocomposite can be defined as a composite structure whose characteristic
dimensions are found at the nanoscale.8 The superhard nanocomposites9 of nc-TiN/SiNx10
exhibit relatively good thermal and chemical stability. In the idealized case, this
nanocomposite consists of crystalline TiN grains which are embedded in a tissue phase of
amorphous Si3N4. A prerequisite to synthesize nanocomposites is a strong segregation
tendency between the constituents in order to get a strong interface between the
nanocrystals and the Si3N4 phase. This is the case for TiN and Si3N4, which have
essentially no solid solubility; see the pseudo-binary phase diagram in Fig. 2.3. For the
nanocomposites with the highest hardness, the grain sizes should be below 10 nm, and
the tissue phase that separates the nanocrystallites on the order of 1-2 monolayers (ML)
thick.11 The hardness enhancement is then explained by small crystallite sizes of TiN,
which gives grain boundary hardening, together with inhibited grain boundary sliding and
crack propagation from the Si3N4 phase.
5
4
2.5
TiN/SiNx Multilayers
A multilayer thin film consists of alternating layers of two or more materials. The sum of
two consecutive layers in a bilayer system is called multilayer period ( ). A multilayer
containing epitaxial layers is called superlattice. Also, multilayered structures with
characteristic dimensions on the nanoscale are referred to as nanolaminates. Recently,
publications regarding superhard TiN/SiNx multilayers have been published,2,12,13 where
metastable c-SiNx have been epitaxially strained between TiN layers in an artificial
superlattice structure. These hardness correlates to the thickness of the SiNx layer, and the
hardness is highest for the case of superlattice with SiNx layer thickness of 1-2 ML. The
high hardness is explained by hindering of dislocation motion.12
2.6
Ternary Solid Solutions
A solid solution can be defined as follows…“A solid solution is a solid-state solution of
one or more solutes in a solvent. Such a mixture is considered a solution rather than a
compound when the crystal structure of the solvent remains unchanged by addition of the
solutes, and when the mixture remains in a single homogeneous phase. The solute may
incorporate into the solvent crystal lattice substitutionally, by replacing a solvent particle
in the lattice, or interstitially, by fitting into the space between solvent particles. Both of
these types of solid solution affect the properties of the material by distorting the crystal
lattice and disrupting the physical and electrical homogeneity of the solvent material.”14
It was stated in section 2.1 that there exist no thermodynamically stable ternary Ti-Si-N
compounds. However, based on the use of our method for growth of metastable thin films
in section 3.4, metastable (Ti,Si)N cubic solid solutions can in fact be realized. Figure 2.3
shows the pseudo-binary phase diagram for TiN and SiN and indicates a strong phase
separation tendency from a solid solution into the binary phases. This implies, that for a
metastable (Ti,Si)N solid solution, a phase separation into the binary phases may be
expected during annealing. The region of ´+ ´´ is the miscibility gap, where the cubic
binary phases are preferred.
The chemical spinodal is indicated by a dashed curve in Fig. 2.3. Within this curve
the eventual decomposition is spinodal and outside of which towards the binodal (solid
curve), phase separation by nucleation and growth would take place. Spinodal
6
5
decomposition can briefly be described as ‘up-hill’ diffusion, in which atoms diffuse
towards high-concentration regions.15
4000
3500
δ-Ti1-xSixN
Temperature (K)
3000
2500
2500
2000
2000
1500
1500
δ' + δ''
1000
1000
500
500
0
0,00
TiN
3000
Temperature (oC)
3500
0
0,25
0,50
x, Ti1-xSixN
0,75
1,00
SiN
Fig. 2.3 Pseudo-binary phase diagram for TiN-SiN together with the
chemical spinodal (dashed line), calculated down to 1727 °C (Liquid state
not considered). From Paper I.
The pseudo-binary phase diagram was calculated in the following way. Gibbs free
energy, G, of a system is defined by
G = H − TS
Eq. 1
where H, T, and S are the system’s enthalpy, temperature, and entropy, respectively. The
total energies were calculated by ab initio density functional theory (DFT) in Paper I for
the Ti1-xSixN system at 0 K. This gives G = H in Eq. 1. To include the temperature and
entropy dependence of G an ideal solution16 is assumed.
1
S. Sambasivan, W. T. Petuskey, J. Mater. Res. 9 (1994) 2362
Powder Diffraction Files, JCPDS International Center for Powder Diffraction Data, Swarthmore, 1989,
card 6-642
3
H. Ljungcrantz, M. Odén, L. Hultman, J. E. Greene, J. –E. Sundgren, J. Appl. Phys. 80 (1996) 6725
4
H. Ljungcrantz, C. Engström, M. Olsson, X. Chu, M. S. Wong, W. D. Sproul, L. Hultman, J. Vac. Sci.
Technol. A16 (1998) 3104
5
I. Tomaszkiewicz, J. Thermal. Anal. Cal. 65 (2001) 425
2
7
6
6
A. Zerr, G. Miehe, G. Serghiou, M. Schwarz, E. Kroke, R. Riedel, H. Fuess, P. Kroll, R. Boehler, Nature
400 (1999) 340.
7
Aylward, A & Findlay, T. in Store (ed.) Si Chemical Data, 3rd Edition, Wiley & Sons Milton, 1994
8
http://www.uspto.gov/go/classification/uspc977/defs977.htm
9
S. Veprek, M. G. J. Veprek-Heijman, P. Karvankova, J. Prochazka, Thin Solid Films 476 (2005) 1
10
S. Veprek, S. Reiprich, Thin Solid Films 268 (1995) 64
11
A. Niederhofer, T. Bolom, P. Nesladek, K. Moto, C. Eggs, D. S. Patil, S. Veprek, Surf. Coat. Technol.
146/147 (2001) 183
12
H. Söderberg, J. M. Molina-Aldereguia, L. Hultman, M. Odén, J. Appl. Phys. 97 (2005) 114327
13
X. Hu, H. Zhang, J. Dai, G. Li, M. Gu, J. Vac. Sci. Technol., A 23 (2005) 114
14
http://en.wikipedia.org/wiki/Solid_solution
15
A. Hörling, PhD Thesis (Linköping Studies in Science and Technology, dissertation no. 922, Linköping
University, Sweden 2005
16
D. A. Porter, K. E. Easterling, Phase Transfromations in Metals and Alloys, Chapman & Hall, 2nd ed.
(1992) 14-15
8
7
9
8
3
Thin Film Deposition
3.1
Physical Vapor Deposition
Thin solid films can be synthesized by physical vapor deposition (PVD) techniques.
Generally, the coating material is vaporized in vacuum from a solid material, target. The
vapor will eventually condense onto a substrate surface. Next follows descriptions of arc
evaporation and DC magnetron sputtering, the PVD methods employed in this work.
3.2
Arc Evaporation
Arc evaporation has been widely used because of its promise of an efficient source of
highly ionized material for producing dense, adherent coatings having a wide range of
compositions.1
A cathodic arc can be described as a low voltage, high current plasma discharge
between two electrodes. The evaporation process of target material is a consequence of
the very high local surface temperature in an arc spot. This creates a molten pool from
which evaporation of the cathode (or target) material and electron emission occurs. The
electrons are then attracted by an electric field and will collide and ionize evaporated
atoms; this is called the ionization zone, see Fig. 3.1. The ions are transported to the
substrate surface where they condensate and react with a reactive gas (if present) from the
surrounding. In Paper I, N2 was utilized as reactive gas. However, the molten pool also
emits macro particles.
To synthesize high quality thin films, the importance of plasma ionization2,3,4 should
be emphasized. Arc evaporation, in contrast to sputtering, provides highly ionized
plasmas and can therefore be manipulated with electric and magnetic fields5. Other ion
induced effects are acceleration of the nucleation stage,6 enhanced adhesion,7
modification of crystal structure,3 film stress,8 densification, and in the case of deposition
from a compound target, changes in stoichiometry.
10
9
Fig. 3.1 Schematic illustration of particle flux at the arc spot.9
The changes in stoichiometry appear when arc evaporation is applied to a compound
target where the different materials cause different degree of ionization. This gives the
ions different acceleration towards a negatively biased substrate. Therefore, ions with
higher degree of ionization will impinge on the surface with higher energy and thus
penetrate deeper into the film compared to an element with lower degree. This will cause
preferential resputtering of the surface near material and the film will contain a higher
concentration of the material with higher degree of ionization.1 This phenomenon was
apparent in Paper I, where a slightly higher Ti:Si ratio was observed in the film compared
to the target. The average charge state during arc evaporation for Ti and Si is typically
+2.1 and +1.4, respectively.10 However, during reactive arc evaporation, the average
charge will be somewhat less.
11
10
Substrates
Trigger
Plasma
Target, cathode
Fig. 3.2 Image of an arc evaporation deposition chamber at SECO Tools AB.
3.3
DC Magnetron Sputtering
The process of sputtering starts by introducing a sputtering gas, preferably inert, into a
vacuum chamber. A high voltage is applied to the target; this creates a visible glow
discharge, often referred to as plasma, by ionization of the inert gas. The gas ions will be
accelerated towards, and eventually collide, with the negatively charged target. If the
kinetic energy of the incoming ions is higher than the binding energy of the target surface
atoms, the atoms will be ejected, sputtered. The ejected target material will vaporize and
travel through the plasma to the substrate. Depending on the kinetic energy of the
incoming coating material and the temperature of the substrate, ad-atoms may or may not
migrate on the surface until they occupy an energetically favorable position. As the ions
collide with the target they will also cause emission of secondary electrons. Since the
electrons are negatively charged they will be repelled from the target and instead collide
with other atoms and ions to free electrons. This will create positively charged ions to
maintain the process.
In Paper II reactive DC magnetron sputtering was used to deposit TiN/SiNx
multilayers from two separate targets, where the reactive N2 gas was mixed with the inert
Ar gas. In this process the vaporized target species in the plasma mix and react with N2
before they migrate on the substrate. In Fig. 3.3 a schematic view of the interior of the
sputtering chamber is shown. The magnetrons were of unbalanced types which together
12
11
with magnetic coupling induced by a coil around the substrate holder enhances the
ionization near the growing film. The ions will then be accelerated towards the negatively
biased substrate, in this case -50 V.
Magnetrons
Shutter
Ar inlet
Substrate
N2 inlet
Coil
Fig. 3.3 Schematic interior view of the deposition chamber.
3.4
Growth of Metastable Solid Solution Films
Energy
In Paper I metastable (Fig. 3.4) ternary solid solutions was synthesized by arc
evaporation. Two mechanisms are described below for depositing metastable solid
solutions.
Ea
I
II
Atomic
arrangement
Fig. 3.4 Schematic illustration of metastable (I) and
thermodynamically stable (II) states. Energy Ea is needed
to activate transformation from state I − to − II.
13
12
3.4.1
Low-temperature Synthesis
Low temperature deposition is preferable from an industrial point of view, since it allows
a wider range of substrate materials with respect to their thermal stability and cost.
Utilizing lower temperatures also decrease the production time. Low temperatures can be
at or below 500 °C; a typical process condition in arc evaporation. In addition, lowtemperature film synthesis (being far from the thermodynamical equilibrium of the
deposition material) induces kinetic limitations which, e.g., allow for synthesis of
metastable phases.11 This is the first mechanism for the metastable ternary solid solution
synthesized in Paper I.
3.4.2
Ion-induced Recoil Implantation
When a negatively biased target (or substrate in the thin film deposition case) is
bombarded by the energetic ion beam, atoms will be redistributed in the target.12 These
collisions also cause atomic recoils, which more precise are energetic ions traversing the
target and goes on to generate a cascade of secondary atoms. The collisions induced by
the ion bombardment cause recoil implantation, wherein target atoms are knocked
downstream by collisions and ion-mixing. This process resembles thermal diffusion, but
is driven by collisions rather than by thermal motions.13 The ion-induced recoil mixing is
proposed as a second mechanism to form the metastable ternary solid solutions in Paper
I.
1
R. L. Boxman, D. M. Sanders, P. J. Martin, J. M. Laferty, Handbook of Vacuum Arc Science,
Fundamentals and Applications, Noyes Publications, New Jersey, 1995
2
D. Dobrev, Thin Solid Films 92 (1982) 41
3
J. M. E. Harper, J. J. Cuomo, H. T. G. Hentzell, J. Appl. Phys. 58 (1985) 550
4
E. Key, F. Parmigiani, W. Parrish, J. Vac. Sci. Technol. A6 (1988) 3074
5
A. Anders, Surf. Coat. Technol. 120-121 (1999) 319
6
M. Marinov, Thin Solid Films, 46 (1977) 267
7
J. E. Griffith, Y. Oiu, T. A. Tombrello, Nucl. Instrum. Methods, 198 (1982) 349
8
J. Cuomo, J. M. E. Harper, C. R. Guarnieri, D. S. Yee, L. J. Attanasio, J. Angilello, C. T. Wu, R. H.
Hammond, J. Vac. Sci. Technol. 20 (1982) 349
9
R. L. Boxman, S. Goldsmith, Surf. Coat. Technol. 52 (1992) 39
10
I. G. Brown, IEEE transactions on plasma science 19 (1991) 713
11
I. Petrov, P. B. Barna, L. Hultman, J. E. Greene, J. Vac. Technol. A21(5) (2003) 117
12
S. M. Myers, Nucl. Instrum. Methods 168, 265 (1980)
13
I. Manning, Phys. Rev. E, B42 16 (1990) 9853
14
13
15
14
4
Theoretical Modeling
To receive further understanding in why the materials of the present thesis behave as they
do, theoretical modeling is used to investigate phase stability as well as structural and
elastic properties. Modeling is an area that has expanded tremendously during the last
decade due to increased computer power and more efficient program codes. However, in
order to provide reliable results with physical meaning, a deep understanding and
knowledge is necessary. In this thesis density functional theory was used to determine
lattice parameters and total energy for Ti1-xSixN solid solutions (0 x 1) and for the
determination of surface reconstructions for SiNx onto TiN.
4.1
Density Functional Theory
Today the density functional theory (DFT) formalism is the most used ab initio method in
computational material science and solid-state physics. The reason for this is the high
computational efficiency combined with high accuracy. Ab initio is a Latin term that
means first principles. This implies that the calculation relies on basic and established
laws of nature without additional assumptions or special models.
In the 1960s Hohenberg and Kohn1 presented and proved two theorems, which
became the fundament for DFT. The first theorem states that the external potential in
which the electrons move, is a unique functional of the ground state electron density. This
means that the systems are fully determined by the electron density. Hence, the total
energy of the system can be expressed as a functional of the density. The second theorem
states that the ground state electron density minimizes the total electronic energy of the
system.
The theory was then further developed by Kohn and Sham2 who used these
theorems to derive the Kohn-Sham equations:
−
2m
∇2 + e2
) dr '+ v (r ) + v (r ) ψ (r ) = E ψ (r )
n (r '
ext
xc
n
n n
r − r'
(Eq. 2)
where the external potential, vext(n(r)), is determined from the electronic density, n(r),
instead of from the electron wave functions as for the general Schrödinger equation. (r)
is the time-independent wave function.
16
15
4.1.1
Approximations for Many-body Interactions
The expression for the energy contribution from many-body interactions, which are
captured by the exchange-correlation energy vxc(r), is unknown. However, different
approximations have been developed; the generalized gradient approximation (GGA) and
the local density approximation (LDA) are perhaps the most commonly used ones. In
LDA the exchange-correlation energy is taken from known results of the many-body
interactions in a uniform electron gas. This apparently easy approximation works
surprisingly well for most applications3 and requires relatively short computational time.
However, for more rapid changes in the electron gas, LDA seems coarse and great effort
has been made to find better approximations. To overcome this problem gradient
correction were included to the exchange-correlation potential. This, together with
constraints on the exchange-correlation functions led to the implement of GGA.
In a comparison between LDA and GGA, the latter tends to improve total energies4,
atomization energies2,5,6, energy barriers, and structural energy differences7,8. GGA also
expands and softens bonds6, an effect that sometimes corrects9 and sometimes
overcorrects10 the LDA prediction.
4.1.2
Pseudo Potentials and Plane Waves
Cambridge serial total energy package (CASTEP) is a powerful simulation package from
Accelrys Inc.11 In this package either GGA or LDA can be utilized.
CASTEP uses pseudo potentials, which are approximations where the core
electrons are treated as frozen. Since the computational time is heavily dependent on the
number of electrons, this decreases the computational time dramatically. One type of
pseudo potential method is the ultra soft pseudo potentials12, these were used in Paper II.
4.1.3
Linear Muffin-Tin Orbital
The full-potential linear muffin-tin orbital (FP-LMTO) method within LDA was used for
the calculation in Paper I. Here, the unit cell is divided into non-overlapping muffin-tin
spheres around the atoms. The potentials and charge densities in the crystal can have any,
and not necessarily spherical, shape.
1
2
P. Hohenberg, W. Kohn, Phys. Rev. B, 136 (1964) B864
W. Kohn, J. Sham, Phys. Rev. A, 140 (1965) A1133
17
16
3
W. Kohn, Rev. Modern. Phys. 71 5 (1999) 1253
J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson, M. R. Pederson, D. J. Singh, C. Fiolhais, Phys.
Rev. B 46 (1992) 6671; 48 (1993) 4978
5
A. D. Becke, J. Chem. Phys. 96 (192) 2155
6
E. I. Proynov, E. Ruiz, A. Vela, D. R. Salahub, Int. J. Quantum Chem. S29 (1995) 61
7
B. Hammer. K. W. Jacobsen, J. K Norskov, Phys. Rev. Lett. 70 (1995) 3487
8
D. R. Hamann, Phys. Rev. Lett. 76 (1996) 660
9
V. Ozolins, M. Körling, Phys. Rev. B 48 (1993) 18304
10
C. Filippi, D. J. Singh, C. Umrigar, Phys. Rev. B 50 (1994) 14947
11
M. D. Segall, P. J. D. Lindan, M. J. Probert, C. J. Pickard, P. J. Hasnip, S. J. Clark, M. C. Payne, “First
principles simulation: ideas, illustrations and the CASTEP code”, J. Phys. Cond. Matt. 14(11) (2002)
2717-2743
12
D. Vanderbilt, Phys. Rev. B 41 (1990) 7892
4
18
17
19
18
5
Thin Film Characterization
To obtain necessary information about material properties, several characterization
techniques are needed. In this work, investigation about composition, morphology,
microstructure, thermal stability and mechanical properties using the techniques
described below were performed.
5.1
X-ray Diffraction
In x-ray diffraction (XRD) an x-ray beam is scattered, or diffracted, by atoms in the
investigated material. X-ray diffractograms are in principle created from the condition of
constructive interference from Bragg’s law
2d·sin =n
(Eq. 3)
which requires that the path difference between the traveled x-rays is equal to an integer
number of wavelengths. In Eq. 3, d is the atomic plane spacing of the investigated crystal,
the wavelength of the x-rays and 2 is the angle of diffraction.
X-ray
source
d
Detector
2
Fig. 5.1 Schematic of diffraction according to Bragg’s law.
XRD is a powerful analytical technique for microstructural investigations and can be
used for characterization of crystal structure, phase transformations, residual stress,
thickness measurement etc.
20
19
5.2
Electron Microscopy
Transmission electron microscopy (TEM) is an invaluable analysis technique due to its
ability of getting both a physical image and an electron diffraction pattern. In this thesis
an FEI analytical TEM, Technai G2 UT FEG operating at 200 keV, equipped with
scanning transmission electron microscope (STEM), energy loss electron spectroscopy
(EELS), and energy dispersive x-ray spectroscopy (EDX) has been used for high
resolution imaging and analytical analysis. A Philips EM 400T (120 keV) and a Philips
CM 20 UT (200 keV) were used for the overview images and electron diffraction.
A TEM is in many ways similar to a light optical microscope. Both types are built
up with an illumination and an image part, where the first illuminates the sample and the
second creates the image. However, instead of photons, electrons are irradiating the
sample in TEM; the sample has to be viewed in vacuum in order to increase the mean
free path of electrons. Also different lenses are used, instead of ordinary optical lenses,
electromagnetic lenses are employed. Since electrons are accelerated with, in this thesis,
120 and 200 keV, respectively, their wavelength is on the 10-12 m scale, which can be
compared to wavelength of photons, 10-7 m scale. The output is better resolution.
In STEM the electron beam been focused to a small probe, i.e. a convergent beam,
which scans over an area of the sample. Some imaging modes in STEM supply
information that cannot be obtained in a conventional TEM, e.g. micrographs containing
mainly z-contrast (described below). Due to the small probe size, chemical analysis can
be performed in different fashions, e.g., EDX line-scan.
TEM and STEM requires careful sample preparation in order to obtain an electron
transparent thin area (<100 nm thick). Depending on the purpose of the analysis the
sample are studied in cross-section or plan-view. The cross-sectional samples in this
thesis were prepared by gluing two small slices cut out from the sample face-to-face and
mount them in a titanium grid. This was continued by mechanically grinding until a
thickness of about 50-60 µm were achieved. Finally, the sample was etched by a 2-5 keV
Ar+ ion beam at a 5° incident angle in a precision ion polish sputter (PIPS) until electron
transparency was achieved.
TEM and STEM combined with the analytical techniques, EELS and EDX, give
possibilities to measure chemical composition on the nanoscale with a point-to-point
resolution in the subnanometer range.
21
20
a)
b)
50nm
Figure 5.2 a) Cross-sectional scanning transmission electron micrograph, and b) transmission electron
micrograph, of Ti0.86Si0.14N deposited onto WC(Co) and annealed at 1100 °C. From Paper I.
The STEM image in Fig. 5.2 a) is obtained by collecting high-angle scattered electrons
collected with a high angle annular dark field detector (HAADF)1 at a low camera length.
This configuration provides mainly z-contrast (thus, very little diffraction contrast) which
gives bright contrast from heavy elements. In Fig. 5.2 a) the heavy elements at the grain
boundaries correspond to W and Co according to a line-scan measurement with
EDX/STEM. Compare also with the bright field TEM image in Fig. 5.2 b), were Zcontrast and diffraction contrast are present. In TEM the heavy elements have dark
atomic number contrast. Note that some grains appear black due to diffraction contrast.
5.3
Nanoindentation
In nanoindentation an indenter deforms a material on a very small scale. During the
indent displacement and indent load are continuously recorded and by using the Oliver
and Pharr2 method, hardness – a material’s ability to resist deformation upon a load3 –,
and Young’s modulus can be evaluated. To avoid influence from substrate the
penetration depth should not exceed ~10%4 of the film thickness.
It is important to take into account that the acquired hardness data is affected by the
material’s microstructure (grain size and boundaries, voids etc.), loading orientation (for
non-isotropic materials), and environment, i.e. temperature, humidity etc. The hardness is
22
21
thus a system response and property that should be evaluated using a statistical approach,
i.e., each samples hardness should be evaluated from multiple indents.
5.4
Scanning Tunneling Microscopy
Scanning tunneling microscopy (STM) delivers images of a solid surface by moving a
sharp conductive tip in a very precise and controlled manner across the sample surface
and recording the electron tunneling current between the tip and sample as a function of
position. The tip edge ideally consists of only one atom in order to provide atomic
resolution.
Tunneling is a quantum mechanical effect in which electrons from one conductor
penetrate through a classically impenetrable potential barrier (for STM, vacuum) into a
second conductor. The phenomenon arises from the leaking of the respective wave
functions into the vacuum and their overlap within classically forbidden regions. This
overlap is significant only over atomic-scale distances and the tunnel current depends
exponentially on the distance between the conductors.
Since an image of the surface is obtained in STM, the analysis has to be carefully
performed, keeping in mind that there is always a risk for unwanted features, like for
example double tip image, vibrations, and electrical noise. A double tip image forms
when the tip picks up contamination from the sample surface or by other structural
modification with features to which tunneling can take place. The image created from a
double tip will contain doublets of the true surface features. Fig. 5.3 shows a micrograph
created from a double tip.
The Material Research Laboratory at the University of Illinois at UrbanaChampaign has a deposition system, equipped with one sputtering target and one
sublimation* target, together with variable temperature scanning tunneling microscope
(VT-STM) and low-energy electron diffraction (LEED)†; all in the same ultra high
vacuum (UHV) system. Hence, new possibilities of in situ surface studies where
depositions of the order of sub-monolayer directly can be probed by means of STM and
LEED. This system was used in Paper II.
*
Sublimation is a PVD process where the target is heated to make the target material evaporate from the
surface.
†
LEED works in principle as ordinary electron diffraction, but is utilized at lower energies which makes it
very surface sensitive and therefore can provide information from the topmost atomic layers.
23
22
20nm
Fig. 5.3 Scanning tunneling micrograph from SiNx deposited on
TiN scanned by a double tip. Note that all features appear twice
in the image.
5.5
X-ray Photoelectron Spectroscopy
X-ray photoelectron spectroscopy (XPS) is a surface sensitive technique, which can
provide information about elemental composition and chemical bonding.
X-ray photons, typically Al or Mg K , travels incident on a surface and eject
valence or core electrons from a depth of 0-10 nm. The kinetic energy, Ekin, from the
ejected photoelectrons are detected and the electron binding energy, EB, is calculated
from
Ekin = h −EB−
(Eq. 4)
where h is the incident photon energy, and a work function that corresponds to the xray source and the sample. The binding energy of a core electron does not only depend on
the element but also on the surrounding atoms. The change in binding energy is often
referred to as chemical shift. This information is of great interest since it could provide
information about the surrounding atoms, e.g. if a Si atom is located in an octahedral (like
atoms in a NaCl lattice) or in a tetrahedral (like atoms in a ZnS lattice) position. The
24
23
energy resolution of the XPS also plays a central role to resolve these shifts. In Paper I
XPS was used to determine the Si bond character.
1
E. M. James and N. D. Browning, Ultramicroscopy 78 (1999) 125
W. C. Oliver, G. M. Pharr, J. Mater. Res. 7 (1992) 1564
3
C. –M. Sung, M. Sung, Mater. Chem. Phys. 43 (1996) 1
4
R. Sburlati, Journal of Mechanics of Materials and Structures 1 (2006) 554
2
25
24
6
Results
6.1
Ti1-xSixN Alloy Films
In Paper I Ti1-xSixN (0 x 0.14) thin films were deposited onto cemented carbide (WCCo) substrates by arc evaporation at ~500 °C. Cross-sectional scanning electron
microscopy showed that all films exhibited a columnar structure. A closer investigation
using XTEM revealed that within each column for the x=0.14 film, a feather-like domain
structure could be revealed. This is an effect originating from point defects and
dislocations yielding low-angle grain boundaries. TEM together with XRD showed that
the obtained films phases were of cubic NaCl type with a lattice parameter close to the
reference value of -TiN at 4.24 Å. Calculations by FP LMTO gives that the lattice
parameter of NaCl Ti1-xSixN (x=0, 0.25, 0.5, 0.75, 1) are very similar (within 1%) of each
other. This implies that it can be hard to distinguish SiNx from TiN in nanocomposites or
multilayers or even from a (Ti,Si)N solid solution using XRD. Furthermore, the
calculations in Paper I suggests that a NaCl-lattice is more favorable than a ZnS-lattice
for Si-content x<0.67. The synthesis of the (Ti,Si)N alloy films is made possible by the
kinetic limitations for atom mobilities at the chosen low substrate temperature (~500 °C)
combined with the employed high-flux low-energy metal ion bombardment, which
induces recoil mixing. Effects of N content on the phase stability and properties of SiNx
polytypes, however, were not investigated.
The microstructure of as-deposited solid solution films with a Si content of x=0.14
was retained up to an annealing temperature of at least 900 °C/120 min (see Paper I).
This absence of phase separation −despite the deep miscibility gap− is likely due to a
limited driving force for the nucleation of a Si3N4 phase due to molar volume mismatch
(Si3N4 has a larger unit cell than the NaCl-structured solid solution). When annealing at
1100 °C/120 min, interdiffusion of W and Co occurs in the film grain boundaries which
transforms the film into a nanocrystalline cellular microstructure. Mechanical properties
of the films were investigated with nanoindentation. The measured hardness increased
close to linearly with increasing Si content and is retained up to 900 °C. At 1100 °C the
hardness decreased to below 30 GPa due to the Co and W diffusion, which weakens the
grain boundaries.
Also, initial deposition experiments were performed for the Ti1-xSixN system in the
wider composition range (0 x 0.22) onto cemented carbide (WC-Co) substrates by arc
evaporation. The as-deposited Ti0.78Si0.22N film exhibited a dense fine grained and
26
25
columnar two-phase structure consisting of a defect-rich crystalline cubic phase and a
nanocrystalline − to − amorphous structure as seen by cross-sectional and plan-view
transmission electron micrographs, see Fig. 6.1. The crystalline structure is a saturated
(Ti,Si)N solid solution similar to the Ti1-xSixN x 0.14 alloys. Due to the supersaturation
of N2 gas in the deposition, the amorphous phase is likely to by rich in nitrogen as for the
thermodynamically stable a-Si3N4. However, no z-contrast was obtained for the structure
using XSTEM at low camera length (90 mm) which indicates that the two phases have a
similar composition, i.e., a-Si3N4:Ti and c-(Ti,Si)N, respectively. The maximum hardness
was reached for Si content of 0.135 x 0.175 and was of the same order as for the
Ti0.86Si0.14N film in Paper I. Higher Si contents resulted in a leveling out or slight
reduction of the hardness.
a)
b)
a-Si3N4
2 nm
200 nm
Fig. 6.1 Transmission electron micrographs from a Ti0.78Si0.22N film deposited by arc-evaporation, a) planview containing amorphous area of Si3N4 between defect-rich (Ti,Si)N crystallites, b) cross-sectional
bright-field overview image showing diffraction contrast indicating location of amorphous and crystalline
phases, respectively. A. Flink, T. Larsson, J. Sjölén, L. Karlsson, L. Hultman, unpublished.
27
26
6.2
TiN/SiNx Nanolaminate Films
The nc-TiN/a-Si3N4 nanocomposites can exhibit extraordinary mechanical properties and
thermal stability. The interface between TiN and SiNx is of great importance to
understand since the atoms involved occupy a relatively large volume fraction of the
nanocomposite. However, there is little knowledge about the bonding at the interface and
for any tissue phase forming or effects of any segregated contamination. In Paper II, an
interface study was performed with three different approaches. 1) Investigation of the
interface between TiN and SiNx by using in situ STM to probe sub atomic layer
coverages of SiNx deposited onto TiN(001)/MgO(001) and TiN(111)/MgO(111), 2)
multilayers of TiN/SiNx deposited onto SiOx and MgO(001) which reduces the interface
to a two dimensional, instead of a three dimensional, problem, and 3) surface calculations
of different SiNx coverages placed onto TiN surfaces by using ab initio DFT methods.
In the in situ UHV STM and LEED study, epitaxial TiN(001)/MgO(001) was
deposited by magnetron sputtering at 700 °C. These substrates were then inserted in the
vacuum deposition system equipped with LEED and VT-STM. The deposition system
contains one Ti magnetron sputtering source and one Si evaporation source and N2 and
Ar gas supplies. Here, a TiN(001) template layer containing atomistically flat terraces
was sputtered. Then, different SiNx surface coverages, SiNx, was deposited onto the TiN
surface by sublimation of Si followed by annealing in N2 atmosphere during 12 h at
temperatures ranging from 600 to 800 °C. Several different crystalline reconstructions
were found for different SiNx, containing rows in the <110> directions.
The multilayers in Paper II were deposited by reactive magnetron sputtering at 500
°C onto MgO(001) substrates. The TiN layers were kept at fixed thickness, 40 Å, while
the SiNx layer thicknesses (lSiNx) were varied between 3 and 25 Å. For a SiNx layer
thickness up to 5 Å, both the TiN and SiNx layers were epitaxial, forming a superlattice.
Increasing the SiNx thickness to 13 Å yielded a polycrystalline TiN layer structure
alternating with SiNx layers that are initially crystalline to a thickness of 5-13 Å before
becoming amorphous. The epitaxial stabilization of c-SiNx is explained by the
minimization of surface area energy at the early stages of layer nucleation, i.e.,
instead of forming a high energy crystalline/amorphous interface a low-energy
crystalline/crystalline interface is formed. However, when the interfacial strain energy,
which depends on the thickness of the SiNx layer, is sufficiently large the epitaxial
growth breaks down, and an amorphous Si3N4 layer forms.
Figure 6.2 shows results from a high-resolution scanning transmission electron
microscope study from the lSiNx = 13 Å TiN/SiNx multilayer sample from Paper II of
28
27
several areas exhibit local epitaxy as in a superlattice. A HAADF detector was used with
a camera length of 300 mm for the purpose of increased image intensity.
Correspondingly, both diffraction and z-contrast are present in the images, which
promote the TiN crystals to be more or less pronounced. The dark and bright contrast
TiN
SiN
10 Å
TiN
SiNx
20 Å
Figure 6.2 High resolution scanning transmission electron micrographs of a TiN/SiNx multilayer with
lSiNx=13 Å. The inset image shows local epitaxy of c-SiNx and TiN. A. Flink, H. Söderberg, P. O. Å.
Persson, L. Hultman, M. Odén, unpublished.
correspond to SiNx and TiN, respectively. The higher-magnification inset shows that
cubic SiNx is stabilized between the TiN layers.
In Paper II also the hardness of the multilayers was determined by nanoindentation.
Both multilayer series deposited on MgO(001) and SiO2 substrates had their maximum
hardness (34±2 GPa) at lSiNx close to the epitaxial breakdown limit in the respective
series. The SiNx/TiN superlattice films exhibit Koehler hardening1, in which dislocation
generation and glide across interfaces is hindered for phases with a large difference in
shear moduli. However, the maximum hardening observed in the present multilayer
samples is significantly larger than expected from Koehler hardening. Thus, additional
hardening mechanism, including strengthening due to the coherency strain between the
29
28
two epitaxial layers, (SiNx and TiN(001)) must be present. Effects of lattice strain can be
seen in Fig. 6.2 with the meandering trace of {200} crystallographic planes in the
successive TiN and SiNx layers.
Finally, ab initio DFT calculations were used to calculate different surface
formations. Three different coverages were tested; 0.4 ML with a Si to N ratio of 1, 0.9
ML with a Si to N ratio of 5/6, and 1 ML with a Si to N ratio of 1. The DFT calculations
suggest that for increasing Si content a tetrahedral binding configuration as in hexagonal
or amorphous Si3N4 is preferred over an octahedral as in NaCl-structure SiN.
1
Koehler, J. S., Attempt to Design a Strong Solid. Phys. Rev. B 2 (1970) 547
30
29
30
PAPER I
31
32
Surface & Coatings Technology 200 (2005) 1535 – 1542
www.elsevier.com/locate/surfcoat
Influence of Si on the microstructure of arc evaporated (Ti,Si)N thin films;
evidence for cubic solid solutions and their thermal stability
A. Flink a,*, T. Larsson b, J. Sjölén c, L. Karlsson c, L. Hultman a
a
Thin Film Physics Division, Department of Physics, IFM, Linköping University, SE-581 83 Linköping, Sweden
b
Ombenning by 14, SE-737 90 Ängelsberg, Sweden
c
SECO Tools AB, SE-737 82 Fagersta, Sweden
Available online 19 September 2005
Abstract
Ti1 x Six N (0 x 0.14) thin solid films were deposited onto cemented carbide (WC-Co) substrates by arc evaporation. X-ray diffraction
and transmission electron microscopy showed that all films were of NaCl-structure type phase. The as-deposited films exhibited a
competitive columnar growth mode where the structure transits to a feather-like nanostructure with increasing Si content. Films with
0 x 0.01 had a b111 crystallographic preferred orientation which changed to an exclusive b200 texture for 0.05 x 0.14. X-ray
photoelectron spectroscopy revealed the presence of Si – N bonding, but no amorphous Si3N4. Band structure calculations performed using a
full potential linear muffin tin orbital method showed that for a given NaCl-structure Ti1 x Six N solid solution, a phase separation into cubic
SiN and TiN is energetically favorable. The microstructure was maintained for the Ti0.86Si0.14N film annealed at 900 -C, while
recrystallization in the cubic state took place at 1100 -C annealing during 2 h. The Si content influenced the film hardness close to linearly, by
combination of solid-solution hardening in the cubic state and defect hardening. For x = 0 and x = 0.14, nanoindentation gave a hardness of
31.3 T 1.3 GPa and 44.7 T 1.9 GPa, respectively. The hardness was retained after annealing at 900 -C, while it decreased to below 30 GPa for
1100 -C following recrystallization and W and Co interdiffusion.
D 2005 Elsevier B.V. All rights reserved.
Keywords: Nitrides; Arc evaporation; Transmission electron microscopy (TEM); Thin films; Solid solution; Microstructure
an annealing temperature of 800 -C for a duration of 150
min. Also, improvement on mechanical properties has been
realized with super hardness, H > 45 GPa, for specific Si
contents [12,13] in TiN –Si3N4 nanocomposites.
The equilibrium phase diagram for Ti – Si – N does not
contain any stable ternary phases [14]. However, Vaz et al.
found phases originating from a possible (Ti,Si)N solid
solution [15]. The maximum Si content in metastable
supersaturated cubic solution Ti1 x Six N is 10 –15 at.%
[14]. Furthermore, Procházka et al. [16] conclude that a
totally segregated amorphous Si3N4 only can occur when
the nitrogen activity is larger than about 10 6. This suggests
that PVD is a preferred technique to suppress nitrogen
segregation by virtue of the lower substrate temperature
employed. Thus, during deposition by reactive magnetron
sputtering Söderberg et al. [17], first, and Hu et al. [18],
more recently, stabilized sub-nm thick layers of cubic SiN in
TiN/SiNx multilayers by epitaxial growth.
1. Introduction
Advanced surface engineering of transition metal nitride
wear-resistant coatings by the introduction of alloying
elements is a growing field of research. TiN has been
widely used as hard coating on cutting tools, but the poor
oxidation resistance at temperatures above 500 -C [1,2] has
created an interest in ternary compounds, e.g., on Ti – Al– N
[3 – 5] and Cr –Al – N [6 –8] and also the more complex
quaternaries, e.g., Ti – Al – Si – N [9,10]. These coating
materials show a much improved oxidation behavior at
high temperatures and are now used by market leaders
within metal cutting tools. Recently, Choi et al. [11] showed
for Ti –Si (11 at.%) – N compounds that Si forms SiO2 on
the surface which acts as an oxygen diffusion barrier up to
* Corresponding author.
E-mail address: [email protected] (A. Flink).
0257-8972/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.surfcoat.2005.08.096
33
1536
A. Flink et al. / Surface & Coatings Technology 200 (2005) 1535 – 1542
In this work, we show that as-deposited Ti1 x Six N thin
films with a Si content up to x = 0.14 can be prepared by arc
deposition. Kinetic limitations during film deposition combined with ion bombardment induced collisional mixing are
proposed as conditions for the growth of metastable
Ti1 x Six N films instead of the thermodynamically stable –
and much more studied– TiN/Si3N4 system. Interestingly,
film hardness was found to increase close to linearly with the
Si content. The as-deposited solid solutions exhibited thermal
stability above 900 -C for 2 h. Residual stress recovery and
recrystallization, however, resulted in transformation to a
nanocrystalline structure after annealing at 1100 -C. Band
structure calculations performed using a full potential linear
muffin tin orbital method showed that for a given NaClstructure, a phase separation into cubic SiN and TiN phases is
energetically more favorable than a Ti1 x Six N solid solution.
photoelectron spectroscopy, XPS, using a VG Microlab
310F system. The XPS was equipped with a non-monochromated Al Ka at 1486.6 eV X-ray source and a
hemispherical electron energy analyzer. To compensate for
eventual sample charging, the peak position of the
adventitious carbon was recorded before Ar-etching,
thereby setting the peak position to 284.75 eV. The samples
were then Ar-etched and survey scans of the binding energy
0 – 1100 eV was recorded with a step size of 1 eV for each
sample. For accurate determination of the exact peak
positions of the Si2p and C1s peaks, local region scans
were recorded with step size of 0.1 eV. To suppress the
background noise, each scan was recorded 10 times.
After mechanical polishing of the surface, nanoindentation analysis of films was performed using a Nano Instruments NanoIndenter II with a Berkovich diamond tip. The
maximum indent load was 25 mN. The indentation
procedure is described in more detail by Hörling et al. [5].
Ten indents in each sample were made to obtain statistically
reliable results. Indents in a bulk fused silica reference
sample were made with an indent load of 8 mN, yielding a
similar penetration depth, < 200 nm, as in the investigated
coatings. The average hardness and Young’s modulus with
standard deviations was determined [19]. Poisson’s ratio, t,
for TiN was set to t = 0.22, as used, e.g., by Sue [20]. The
average hardness and Young’s modulus for the reference
sample, SiO 2, was measured to 9.65 T 0.4 GPa and
72.31 T1.32 GPa, respectively.
2. Experimental details
Cemented carbide WC-Co (6 wt.%) 12 12 4 mm3
plates were used as substrates. Before the deposition the
substrates were ground and polished to a mirror-like finish,
R a å 0.01 Am, and cleaned in an ultrasonic alkaline
degreasing agent.
The films were deposited by a commercial arc evaporation system. Three cathodes of composition Ti, Ti90Si10,
and Ti80Si20, respectively, located on top of each other in the
chamber were used to produce Ti1 x Six N films of varying
composition from one batch. Substrates with a bias of 50
V were kept at a temperature of 500 -C in an Ar/N2
atmosphere with N2-flow of 300 sccm.
Isothermal annealing of samples were performed in a
Sintevac Furnace from GCA Vacuum Industries. The
samples were annealed for 2 h at 900 -C and 1100 -C,
respectively. The annealing experiments were performed in
an Ar flow at atmospheric pressure to prevent oxidation of
the sample surfaces.
The microstructure of the coatings was studied with Xray diffraction (XRD), cross-sectional transmission electron
microscopy (XTEM), and scanning electron microscopy,
(SEM). X-ray diffractometry was performed using a Philips
PW 1820 powder diffractometer with a line-focused Cu Ka
X-ray source. h –2h scans were recorded in the 2h range of
5- to 90-. A Philips EM 400T microscope operating at 120
kV was used for the overview imaging and an FEI Technai
G2 UT FEG microscope equipped with an electron energy
loss spectrometer, EELS, and an energy-dispersive X-ray
analysis spectrometer, EDX, operating at 200 kV was used
for the high-resolution imaging and the EELS analysis.
Chemical analysis of the film compositions was performed using an Oxford Link ISIS EDX equipment,
operating at 20 kV, in connection with a LEO 1550 SEM.
Elemental mapping by EDX was measured for 30 min on a
plan-view sample magnified by 2.5k. The chemical bonding
structure in the near-surface region was analyzed by X-ray
3. Computational details
Ab initio calculations on the TiN – SiN system were
carried out using the full potential linear muffin tin orbital
method (FP-LMTO) [21,22] within the local density
approximation (LDA) of density functional theory (DFT).
The exchange correlation function used is described by
Hedin and Lundqvist [23]. Starting with the TiN structure,
and by exchanging Si for Ti atoms in the NaCl and
zincblende, ZnS, structure, respectively, a total of five
different compositions, x = 0, 0.25, 0.5, 0.75, and 1, were
investigated. The unit cell dimensions for all structures were
varied uniformly and the theoretical size of the unit cells
was obtained from the minimum in total energy. Energy
convergence was reached for all compositions with respect
to the number of k points used.
4. Results and discussion
EDX analysis of as-deposited films from the different
positions within the deposition chamber showed a continuous
composition range between 0 x 0.14 Si for samples
positioned between Ti and Ti0.9Si0.1 or Ti0.8Si0.2 targets.
The film closest to the Ti0.8Si0.2 target was expected, due to
setup geometry, to have higher Si content than the actual
34
A. Flink et al. / Surface & Coatings Technology 200 (2005) 1535 – 1542
a)
WC (100)
TiN (111)
Fig. 1a shows X-ray diffractograms from the as-deposited
Ti1 x Six N films. The diffractograms for all compositions
revealed a NaCl-structure compound with a lattice parameter
very close to TiN, 0.424 nm. The preferential film growth
orientation changed from mixed b111 to exclusive b200
with increasing Si content. For x 0.07, however, the (200)
peak broadening increased substantially.
Cross-sectional TEM micrographs from the as-deposited
samples are presented in Fig. 2a (TiN), Fig. 2b
(Ti0.92Si0.08N), and Fig. 3 (Ti0.86Si0.14N), respectively. The
films exhibited a dense columnar structure where the
column width ranged from 100 to 400 nm. Interestingly,
for all films, the top surface correlated directly with the
substrate topography. This implies that 3D-island growth
and eventual faceting was effectively suppressed during the
deposition. From Figs. 2 and 3, it is evident that the Si
content also affected the structure and increased the defect
density. The as-deposited Ti0.86Si0.14N, see Fig. 3, exhibited
within each column a feather-like structure. Higher magnification imaging revealed nm-structure of feathers (elongated crystalline grains) with large strain contrast and moiré
fringes from overlapping features. The high-resolution
electron micrograph (HREM) in Fig. 3b shows a typical
appearance of three feather features of the cubic (Ti,Si)N
phase with high defect density of dislocations. The
observations in Fig. 3 show an interesting growth mode
with a rotating lattice by branching into subgrains over each
column. Branching begins at the column boundaries and the
subgrains merge at the apparent stem of the columns. This
takes place to maintain the (002) growth surface. The
selected area electron diffraction pattern (SAED) in Fig. 3a
confirms the texture seen in XRD. No volumes of any
amorphous phase were found by the TEM analysis.
Fractured cross-sections from as-deposited Ti1 x Six N,
x = 0, x = 0.05, x = 0.08, x = 0.14, presented in Fig. 4, were
investigated by scanning electron microscopy. The micrographs showed dense columnar microstructure with macro
particles incorporated to a similar density as for (Ti,Al)N
coatings. The thickness of the (Ti,Si)N coating ranged
between 1.6 and 2.0 Am. As-deposited Ti0.86Si0.14N
WC (101)
TiN (200)
Co
x:
0.14
Intensity (arb. units)
0.13
0.11
0.08
0.07
0.06
0.05
0.03
0.01
0
34
b)
36
38
40
42
44
46
48
50
2 Theta (degrees)
TiN (111)
WC (100)
WC (101)
TiN (200)
Co
x:
0.14
Intensity (arb. units)
0.13
0.11
0.08
0.07
0.06
0.05
0.03
0.01
0
34
c)
36
38
40
42
44
46
48
50
2 Theta (degrees)
WC (101)
TiN (200)
TiN (111)
WC (100)
Co
Intensity (arb. units)
x:
0.14
0.13
0.11
0.08
0.07
0.06
0.05
0.03
0.01
0
34
36
38
40
42
44
46
48
1537
50
a)
2 Theta (degrees)
b)
Fig. 1. X-ray diffractograms from Ti1 x Six N films in (a) as-deposited, (b)
annealed at 900 -C, and (c) annealed at 1100 -C states. The Si content of
each film is indicated.
x = 0.14. However, the apparent loss of Si during arc
evaporating can be explained by the stronger impact from
Ti atoms than Si on the sample surface. Since the Ti atoms
have a higher grade of ionization, they will retrieve higher
acceleration towards the negatively biased substrate. This
will distribute the Ti atoms deeper into the structure than for
Si. Such a mechanism was shown to operate in the (Ti,Al)N
system [24]. In our case, preferential sputtering of the lighter
Si component is also possible.
Fig. 2. Cross-sectional transmission electron micrographs of (a) TiN and (b)
Ti0.92Si0.08N films on WC-Co substrates.
35
1538
A. Flink et al. / Surface & Coatings Technology 200 (2005) 1535 – 1542
a)
lattice parameter for TiN. Underestimation of the unit cell size
is normal for the local density approximation (LDA) within
density functional theory (DFT) as seen for TiN here. The
relatively constant a values are consistent with our experimental results, c.f. Figs. 1a, 3, and 6). Also, Söderberg et al.
[17] was not able to distinguish the y-SiN and y-TiN phases
by XRD due to the small difference in lattice parameters. The
present findings, further, implies that the peak with a lattice
parameter of 0.429 nm observed in an alleged (Ti,Si)N solid
solution by Vaz et al. [15] is not from a NaCl-structure phase.
Results from XRD of Ti1 x Six N films isothermally
annealed at 900 -C and 1100 -C are presented in Fig. 1b
and c. The texture of the as-deposited samples was
maintained. However, the peak width decreases as annealing
temperature increases.
Fig. 6 shows a cross-sectional TEM micrograph from the
Ti0.86Si0.14N sample annealed at 1100 -C. The film now
exhibits a cellular structure of elongated ¨ 10-nm-wide
grains with a texture that is reminiscent from the columnar
Column boundaries
Feather
features
a)
b)
I
II
III
b)
(002)
8°
(002)
Fig. 3. Cross-sectional TEM micrographs from an as-deposited
Ti0.86Si0.14N thin film on WC-Co substrate in, (a) overview with selected
area electron diffraction pattern and (b) HREM image. The columnar
microstructure with internal branching of subgrains of (002) crystallographic orientation is indicated in (a). The trace of (200) and (002) planes in
neighboring subgrains I – III with zone axes [010], [hk0], and [110],
respectively, after mutual rotation around [001] is shown in (b).
c)
behaved as a fine structure when fractured in agreement
with the nanostructure seen by XTEM.
In Fig. 5a, the total energy of a relaxed NaCl type unit cell
of Ti1 x Six N is plotted as a function of Si content. As a
comparison the total energy of Ti1 x Alx N, Ti1 z Zrz N, and
TiC1 x Nx are included in the figure. The calculations show
that NaCl-structure (Ti,Si)N is metastable with respect to
phase separation into NaCl-structure SiN and TiN. Comparison of the energy for ZnS and NaCl structure Ti1 x Six N, Fig.
5b, however, suggests that a NaCl-structure (Ti,Si)N solid
solution is energetically more favorable than a cubic ZnSstructure only up to x¨ 0.67. The calculations of lattice
parameters for NaCl-structure Ti1 x Six N gave aTi0.75Si0.25N =
0.985a 0, a TiN = 0.986a 0, and a SiN = 0.99a 0, where a 0 is the
d)
Fig. 4. Fractured cross-sectional SEM micrographs from Ti1 x Six N, (a)
x = 0, (b) x = 0.05, (c) x = 0.08, and (d) x = 0.14 on WC-Co substrates.
36
A. Flink et al. / Surface & Coatings Technology 200 (2005) 1535 – 1542
a)
NaCl
0,12
0,10
80
Ti1-xSixN
40
0,02
0
0,00
-0,02
0,00
TiC1-xNx
-40
0,25
0,50
0,75
1,00
Composition, x
b)
500
NaCl (δ)
ZnS-zb
0,28
0,24
E (Ry/cell)
a)
400
Ti 1-xSi xN
0,20
300
0,16
0,12
200
0,08
E (meV/atom)
E (Ry/cell)
120
Ti1-xZrxN
0,06
E (meV/atom)
160
0,08
0,04
Elemental mapping by EDX in combination with SEM
from the Ti0.86Si0.14N sample annealed at 1100 -C, Fig. 7
revealed that Co and W is leaking through the film. The
SEM/EDX and TEM/EELS/EDX analyses shows that such
high-temperature annealing results in W and Co diffusion
200
Ti1-xAlxN
1539
100
0,04
0,00
0,00
c)
0,25
0,50
0,75
0
1,00
Composition, x
4000
3500
-Ti1-xSi xN
2500
2500
2000
2000
1500
1500
1000
,
+
1000
,,
b)
3000
3000
Temperature (oC)
Temperature (K)
3500
500
500
0
0
0,00
0,25
0,50
0,75
1,00
Ti1-xSixN
c)
Fig. 5. Ab-initio calculations showing (a) the total energy of one unit cell as
a function of composition for the Ti1 x Alx N, Ti1 z Zrz N, Ti1 x Six N, and
TiC1 x Nx systems in NaCl-structure, (b) comparison of which structure of
NaCl and ZnS that is most energetically favorable for different Si contents
and (c) phase diagram for TiN – SiN as calculated from the graph in (a),
together with the chemical spinodal, calculated down to 1727 -C. (Liquid
state not considered.)
structure with feather-like features of the as-deposited state.
This nanocrystalline structure can be seen from the higher
magnification insert image, Fig. 5b and c, to consist of
recrystallized grains, but no amorphous phase. We find that
the boundaries were formed to relax the as-deposited
structure that contained a high density of line and point
defects. It is a further characteristic observation that the
structure exhibited cell-walls between the grains, see Fig. 6b
and c. The cell-walls that have dark atomic number contrast
were rich in Co and W as explained below.
Fig. 6. Cross-sectional TEM micrographs of a Ti0.86Si0.14N film on WC-Co
substrate annealed at 1100 -C for 2 h showing (a) an overview, (b) highermagnification view with a selected are diffraction pattern and (c) HREM
image of a recrystallized cubic-phase structure without amorphous material.
37
1540
A. Flink et al. / Surface & Coatings Technology 200 (2005) 1535 – 1542
states can be sustained. In the case of CVD deposition
[13,30] and other processes of relatively high surface
mobility of the elements, the strong segregation tendency
of the Ti and Si on the substrate surface during deposition
results in the direct formation of TiN and Si3N4 as in the
nc-TiN/a-Si3N4 nanocomposite thin films.
Fig. 8 shows how the Si content influences the hardness
and Young_s modulus of the as-deposited films and those
annealed at 900 -C and 1100 -C. The addition of Si
increased the hardness throughout the whole gradient series
for the as-deposited films from 31.3 T 1.3 GPa up to
44.7 T 1.9 GPa for Ti0.86Si0.14N. From XRD and TEM, it
is evident that the Ti0.86Si0.14N coating contains the highest
defect density. Thus, both solid solution and defect hardening may be active. The 900 -C annealed samples retained
their hardness with a similar trend of Si content coupled to
hardness. At 1100 -C, however, the hardness has decreased
to below 30 GPa for all compositions. Veprek et al. [31]
suggested that an O content of as little as 1 –1.5 at.% causes
a decrease of the hardness in nc-TiN/a-Si3N4 nanocomposites if segregated to the grain boundaries. For the present
case of metastable cubic phase (Ti,Si)N, our analysis
showed that Co and W had diffused from the substrate into
SEM
a)
b)
N Kα
c)
Ti Kα
d)
Co Kα
e)
W Lα
Fig. 7. (a) Scanning electron micrograph and (b – e) elemental maps of
different elements by EDX from an area containing the film, Ti0.86Si0.14N
annealed at 1100 -C, right part, and substrate WC-Co, left part, surface
(after local flaking). Bright contrast correlates to high concentration.
towards the film top surface via the grain boundaries formed
during the recrystallization. For comparison, Hörling et al.
[3] reported Co interdiffusion in (Ti,Al)N films on similar
substrates annealed at 1250 -C for 2 h.
Similar to arc evaporated TiN [25], Ti(C,N) [25],
(Ti,Al)N [3], and (Ti,Zr)N [26], Ti1 x Six N solid solution
films undergo recovery from a growth-induced compressive stress state during annealing above the deposition
temperature. From Fig. 1, the recovery is evident from a
decreasing peak broadening. The underlying processes can
be understood from the onset of N-related defect
annihilation with activation energy around 2 eV [27]
and defect-assisted diffusion on the Ti side with activation
energy comparable to that for surface diffusion of 3.5 eV
[28]. Based on its smaller atomic radius, however, Si
diffusion should not be the limiting factor for the
segregations.
Next we consider the tendency for phase separation of
the as-deposited Ti1 x Six N films. As was found by the ab
initio calculations, Fig. 5, the system exhibits a large
miscibility gap with a possible initial transformation path
to y-TiN and y-SiN or ZnS-structured SiN. The calculated
spinodal in Fig. 5 indicates that spinodal decomposition
can be preferred instead of nucleation and growth as the
phase separation mechanism for a wide range of compositions including 10– 12 at.% Si (the percolation threshold
in nc-TiN/a-SiN) for most processing temperatures. That
state is of course in turn apt to transform to the equilibrium
phases y-TiN and Si3N4. From the present results, we find
that annealing temperatures in excess of 1100 -C would be
required to effectively reach closer to equilibrium. Considering the larger molar volume of the Si3N4 phases
compared to its cubic polytypes, the metastable cubic
a)
48
46
as-dep.
900
1100
44
Hardness (GPa)
42
40
38
36
34
32
30
28
26
24
22
0,00
0,02
0,04
0,06
0,08
0,10
0,12
0,14
0,16
Composition, x
b)
740
as -dep.
90 0
11 00
Young's modulus (GPa)
720
700
680
660
640
620
600
580
560
540
520
0,00
0,02
0,04
0,06
0,08
0,10
0,12
0,14
0,16
Composition, x
Fig. 8. (a) Hardness and (b) Young’s modulus of as-deposited and annealed
Ti1 x Six N films at 900 -C and 1100 -C for 2 h.
38
A. Flink et al. / Surface & Coatings Technology 200 (2005) 1535 – 1542
1541
5. Conclusions
Counts (arb. units)
Si2p
Ti1 x Six N solid solution films of NaCl-structure containing at least x = 0.14 Si can be synthesized by arcevaporation at a substrate temperature of 500 -C. For the
highest Si contents, however, XRD and TEM reveal a
defect-rich lattice whereas XPS profiles do not support the
presence of amorphous Si3N4. The findings are supported
by ab initio calculations which show that the pseudobinary TiN –SiN system with metastable solid solutions
exhibits a wide miscibility gap with respect to y-TiN and
NaCl-structured SiN or ZnS-structured SiN depending on
composition.
A new characteristic nanostructured feather-like growth
mode of films was observed for the Ti0.86Si0.14N samples as
the columnar microstructure exhibited internal branching of
subgrains with more or less continuous rotation to align
growth to a common (002) direction.
The as-deposited solid solutions exhibited thermal
stability up to 900 -C for 2 h where the hardness was
retained. Film hardness was close to a linear function of Si
content (from 31 GPa in TiN to 45 GPa in Ti0.86Si0.14N).
Residual stress recovery and recrystallization, however,
was induced by 1100 -C annealing. It resulted in
transformation to a nanocrystalline cellular structure that
was void of amorphous phases, but with concomitant
diffusion of W and Co from the substrate through the film
via the grain boundaries to form a tissue phase. This
resulted in grain boundary weakening which decreased the
film hardness.
Finally, our results have impact for the interpretation of
nanocomposite or multilayer (superlattice) TiN/Si3N4 films
as the presence of cubic (Ti,Si)N compounds must now be
considered.
4500
4000
120
110
100
90
Binding energy (eV)
Fig. 9. X-ray photoelectron spectroscopy of the Si2p-peak from the asdeposited Ti0.86Si0.14N film.
the grain boundaries. The observed softening can then be
explained by grain boundary weakening by tissue phases,
e.g., CoSi, TiO2 or SiO2, the latter two assuming that
oxygen is segregated from within grains to the boundaries.
These are equilibrium phases expected from phase diagram
considerations all of which are soft in comparison to TiN
and SiNx compounds. In the form of a tissue phase, they can
be expected to weaken the matrix material.
The hardness of the as-deposited TiN film with 31.2 GPa
is significantly higher than of TiN (002) single-crystal films
of 20 GPa [32]. Defect hardening effects in arc evaporated
TiN, Ti(C,N) and TiC have, however, been reported earlier
[33]. It is due to small grain size and lattice point defects,
where the latter yields compressive residual stress state.
The Young’s modulus of the films, see Fig. 8, increased
slightly with increasing Si content for the as-deposited films
from 670 GPa, for x = 0, to 700 GPa, for x = 0.047. For
higher amount of Si, however, the Young’s modulus
decreased to ¨600 GPa. It should be noted that nanoindentation of nitrides under compressive stress systematically gives high values for the moduli. Here, we are
concerned with the relative changes and conclude that the
hardening by Si substitution in Ti1 x Six N films is not
correlated to the strength of SiN per se.
X-ray photoelectron spectroscopy of the as-deposited
Ti0.86Si0.14N showed the presence of the elements Ti, Si, N,
as well as a very small amount of O after Ar-etching. In
addition, a pronounced sample charging was detected,
suggesting the presence of an insulating phase on the film
surface. This is in contrast to the as-deposited TiN sample,
which did not exhibit sample charging before or after Aretching. Analysis of the Si2p peak, shown in Fig. 9, showed
a binding energy 100.9 eV for the Ti0.86Si0.14N film, which
suggests Si –N bonding, close to the reported value of aSi3N4 at 100.6 eV [34]. Amorphous silicon nitride, a-Si3N4,
is usually positioned at 101.8 eV [29]. No Ti – Si bonding
was seen in the present XPS operating with a ¨ 1.0 eV
resolution.
Acknowledgements
The present work was performed under the Swedish
Foundation for Strategic Research (SSF) Low-Temperature
Thin Film Synthesis program with additional support from
SECO Tools AB. Dr. Per Persson and Dr. Hans Högberg are
acknowledged for assistance with electron microscopy and
XPS analyses, respectively.
References
[1] W.-D. Münz, J. Vac. Sci. Technol., A 4 (1986) 2717.
[2] H.G. Tompkins, J. Appl. Phys. 70 (1991) 3876.
[3] A. Hörling, J. Sjölén, L. Karlsson, M. Odén, L. Hultman, J. Vac. Sci.
Technol., A 20 (2002) 1815.
[4] A. Hörling PhD Thesis (Linköping Studies in Science and Technology, dissertation no. 565, Linköping University, Sweden 1999).
[5] A. Hörling, L. Hultman, M. Odén, J. Sjölén, L. Karlsson, Surf. Coat.
Technol. 191 (2005) 384.
[6] O. Banakh, P.E. Schmid, R. Sanjinés, F. Lévy, Surf. Coat. Technol.
163 – 164 (2003) 57.
39
1542
A. Flink et al. / Surface & Coatings Technology 200 (2005) 1535 – 1542
[7] O. Knotek, F. Löffler, H.-J. Scholl, Surf. Coat. Technol. 45 (1991) 53.
[8] A. Sugashima, H. Kajioka, Y. Makino, Surf. Coat. Technol. 97 (1997)
590.
[9] P. Holubar, M. Jilek, M. Sima, Surf. Coat. Technol. 120 – 121 (1999)
184.
[10] S. Veprek, H.-D. Männling, M. Jilek, P. Holubar, Mater. Sci. Eng., A
Struct. Mater.: Prop. Microstruct. Process. 366 (2004) 202.
[11] J.B. Choi, K. Cho, M.-H. Lee, K.H. Kim, Thin Solid Films 447 – 448
(2004) 365.
[12] L. Shizhi, S. Yulong, P. Hongrui, Plasma Chem. Plasma Process. 12
(1992) 287.
[13] S. Veprek, S. Reiprich, L. Shizhi, Appl. Phys. Lett. 66 (1995) 2640.
[14] E.V. Shalaeva, S.V. Borisov, O.F. Denisov, M.V. Kuznetsov, Thin
Solid Films 339 (1999) 129.
[15] F. Vaz, L. Rebouta, B. Almeida, P. Goudeau, J. Pacaud, J.P. Riviére,
J. Bessa Sousa, Surf. Coat. Technol. 120 – 121 (1999) 166.
[16] J. Procházka, P. Karvánková, M.G.J. Veprek-Heijman, S. Veprek,
Mater. Sci. Eng., A Struct. Mater.: Prop. Microstruct. Process. 384
(2004) 102.
[17] H. Söderberg, M. Odén, J.M. Molina-Aldareguia, L. Hultman, J. Appl.
Phys. 97 (2005) 114327.
[18] X. Hu, H. Zhang, J. Dai, G. Li, M. Gu, J. Vac. Sci. Technol., A 23
(2005) 114.
[19] W.C. Oliver, G.M. Pharr, J. Mater. Res. 7 (6) (1992) 1564.
[20] J.A. Sue, Surf. Coat. Technol. 54 – 55 (1992) 154.
[21] J.M. Wills (unpublished).
[22] H. Skriver, The LMTO Method, Solid State Sciences, Springer Verlag,
1984.
[23] L. Hedin, B.I. Lundqvist, J. Phys., C 4 (1971) 2064.
[24] R.L. Boxman, D.M. Sanders, P.J. Martin, J.M. Lafferty, Handbook of
Vacuum Arc Science, Fundamentals and Applications, Noyes Publications, New Jersey, 1995.
[25] L. Karlsson, PhD Thesis (Linköping Studies in Science and
Technology, dissertation no. 565, Linköping University, Sweden,
1999.
[26] A. Hörling PhD Thesis (Linköping Studies in Science and Technology, dissertation no. 922, Linköping University, Sweden, 2005).
[27] A.J. Perry, J. Vac. Sci. Technol., A 6 (1988) 2140.
[28] S. Kodambaka, V. Petrova, A. Vailionis, P. Desjardins, D.G. Cahill,
I. Petrov, J.E. Greene, Thin Solid Films 392 (2001) 164.
[29] F. Kaufmann, G. Dehm, V. Shier, A. Schattke, T. Beck, S. Lang, E.
Arzt, Thin Solid Films 473 (2005) 114.
[30] E.-A. Lee, K.H. Kim, Thin Solid Films 420 – 421 (2002) 371.
[31] S. Veprek, H.-D. Männling, A. Niederhofer, D. Ma, S. Mukherjee,
J. Vac. Sci. Technol., B 22 (2) (2004) L5.
[32] H. Ljungcrantz, M. Odén, L. Hultman, J.E. Greene, J.-E. Sundgren,
J. Appl. Phys. 80 (1996) 6725.
[33] L. Karlsson, L. Hultman, J.-E. Sundgren, Thin Solid Films 371 (2002)
167.
[34] H. Du, R.E. Tressler, K.E. Spear, C.G. Pantano, J. Electrochem. Soc.
136 (1989) 1527.
40
Erratum Paper I
p. 1541: reference [4] reads:
should read:
“A Hörling PhD Thesis...”
“W. –D. Münz, J. Göbel, Proc. 7th Int Conf.
Vacuum Metallurgy, Linz, Austria 1985”
41
42
PAPER II
43
44
Toward understanding interface structure in
superhard TiN-SiN nanolaminates and nanocomposites
Lars Hultman 1,*, Javier Bareño2, Axel Flink1, Hans Söderberg3, Karin Larsson4, Vania Petrova2,
Magnus Odén3, J. E. Greene2, and Ivan Petrov2
1
Department of Physics, Chemistry, and Biology (IFM), Linköping University, SE-581 83 Linköping, Sweden
Frederick Seitz Materials Research Laboratory and the Materials Science Department, University of Illinois at
Urbana-Champaign, Urbana, Illinois 61801, USA
3
Engineering Materials, Luleå University of Technology, SE 971 87 Luleå, Sweden
4
Department of Materials Chemistry, The Ångström Laboratory, Uppsala University, P.O. Box 538, SE-751 21
Uppsala, Sweden
2
* Communicating Author: LH; e-mail: [email protected]
Nanostructured materials – the subject of much of contemporary materials
research – are defined by internal interfaces, the nature of which are largely unknown.
Yet, the interfaces determine the properties of nanocomposites and nanolaminates. An
example is nanocomposites with extreme hardness 70-90 GPa [1,2], which is of the
order of, or higher than, diamond. The Ti-Si-N system, in particular, is attracting
attention for the synthesis of such superhard materials. In this case, the
nanocomposite structure consists of TiN nanocrystallites encapsulated in a fully
percolated SiNx “tissue phase” (1-2 monolayers thick) that is assumed to be
amorphous [1-3]. Here, we show that the interfacial tissue phase can be crystalline,
and even epitaxial with complex surface reconstructions. Using in situ structural
analyses combined with ab initio calculations, we find that SiNx layers grow
epitaxially, giving rise to strong interfacial bonding, on both TiN(001) and TiN(111)
surfaces. In addition, TiN overlayers grow epitaxially on SiNx/TiN(001) bilayers in
nanolaminate structures. These results provide insight into the development of design
rules for novel nanostructured materials.
boundary sliding [4]. Together, these effects
provide qualitative explanation for the observed
superhardness of the nanocomposites. The
pseudobinary TiN-SiNx system, which presently
serves as an archetype in the quest for superhard
nanocomposite materials [1-3, 5-8], exhibits
strong surface segregation (TiN and Si3N4 have
essentially no solid solubility), a prerequisite for
self-organized nanocomposite formation during
vapor phase deposition. Recently, the growth of
superhard SiNx/TiN nanolaminates has also been
reported [5-7].
Despite the critical dependence of
nanocomposite properties on the nature of the
An intense research area is the synthesis of
superhard (hardness H
40 GPa [1]) nanocomposites for use as wear-resistant coatings on
tools and mechanical components as well as
scratch-resistant coatings on optics. The
nanocomposites are composed of nanocrystallites ( 10 nm) of transition metal nitrides,
carbides or borides encapsulated by 1-2
monolayers (ML) of a covalent nitride (e.g.,
Si3N4, BN, CNx, or diamond-like C) interfacial
layer. Due to the small dimensions across the
nanograins, nucleation and glide of dislocations
is impeded, while the high cohesive strength of
the thin intergranular tissue phase inhibits grain1
45
interfaces between constituent phases, there is a
complete lack of basic knowledge regarding the
three-dimensional structure of the interfacial
phase. The nanocurvature of such interfaces
presents an extreme challenge to the use of
transmission electron microscopy (TEM) and
other standard analytical methods. We approach
the problem of isolating and probing SiNx/TiN
interface chemistry and structure by preparing
planar interfaces in the form of bilayers
(heterostructures), trilayers, and multilayers
(superlattices or nanolaminates) starting with
well-defined TiN(001) and TiN(111) surfaces. In
order to minimize contamination effects, film
growth experiments are performed in ultra-highvacuum (UHV). In situ UHV variabletemperature scanning tunneling microscopy
(VT-STM), low-energy electron diffraction
(LEED), high-resolution cross sectional electron
microscopy
(HR-XTEM),
nanoindentation
measurements, and ab initio density functional
theory (DFT) calculations are used to provide
atomistic information regarding bonding and
crystallographic order at SiNx/TiN interfaces.
As a first step, we determine the growth mode
of SiNx layers in SiNx/TiN nanolaminates
deposited on MgO(001) at 500 °C by reactive
dual magnetron sputtering of high-purity Si
(99.99%) and Ti (99.97%) targets in mixed
Ar/N2 atmospheres (pAr = 4.0 mTorr, pN2 = 0.5
mTorr; 99.9997%). The thickness of the TiN
layers is maintained constant at 40 Å (~19 ML)
while varying the SiNx layer thickness (lSiNx)
from 3 Å (~1.5 ML) to 25 Å (~12 ML). Figure
1(a) is a typical HR-XTEM image from a
nanolaminate with lSiNx = 5 Å (~2.5 ML). Both
the SiNx(001) and TiN(001) layers grow
epitaxially to form a superlattice. HR-XTEM
images reveal continuous lattice fringes across
successive layers with no indication of the
presence of an amorphous or polycrystalline
SiNx phase. In fact, all nanolaminate samples
with lSiNx
5 Å are epitaxial SiNx/TiN(001)
superlattices. X-ray reflectivity (XRR) scans of
films with lSiNx = 5 Å and lSiNx = 13 Å exhibit
FIG. 1. Cross-sectional high-resolution electron
microscopy images from TiN/SiNx multilayers grown at
500 °C on MgO(001), with a substrate bias of -50 V, by
reactive dual magnetron sputtering of Si and Ti targets in
an Ar/N2 discharge. The TiN layers are 40 Å thick, while
the SiNx layer thickness is (a) 5 Å and (b) 13 Å. (c) A
higher-magnification image of the sample in (b). (d) Image
from a region in sample (b) exhibiting local epitaxial
growth across the entire the 13 Å thick SiN layer.
superstructure reflections which have a
sharpness and periodicity that denote welldefined continuous layers.
Elastic recoil detection analyses (ERDA) of
our SiNx/TiN(001) multilayers reveal, in all
cases, that the N/(Ti+Si) ratio is ~1.00. Thus,
x=1 for the SiNx layers since the TiN(001) layers
are grown under conditions known to provide
stoichiometry [9]. In addition, we have
established that the N/Ti ratio of thick TiN
layers is 1.00±0.03 using a combination of x-ray
diffraction (XRD) and Rutherford backscattering
spectroscopy. θ-2θ XRD scans of SiN/TiN(001)
superlattices exhibit only single peaks with no
2
46
is >10%. Parallel SiNx/TiN growth experiments
using SiO2 substrates (oxidized Si(001); Ts =
500 °C) with lSiNx = 3 Å (~1.5 ML) result in
polycrystalline nanolaminates with local
epitaxial growth of SiN on individual TiN
grains.
In order to probe the dependence of the
hardness H of our SiNx/TiN multilayers on lSiNx,
we carried out nanoindentation measurements on
two sets of 0.5-µm-thick samples grown at Ts =
500 °C with lTiN maintained constant at 40 Å. A
Nanoindenter II system with a Berkovich
diamond indenter was used with loads of 2-5
mN. The first sample set (Series 1) was grown
on MgO(001) substrates (the initial layer is
epitaxial TiN(001)), while the second set (Series
2) was deposited on SiO2 (the initial TiN layer is
polycrystalline). H(lSiNx) data are plotted in Fig.
2. For both sample series, the maximum
hardness (34±2 GPa) is obtained at lSiN ~ l*SiNx.
splitting at 00l positions, consistent with our
DFT calculations showing that the lattice
parameter misfit between NaCl-structure TiN
and isostructural metastable SiN is ~0.5%. The
only impurities detected by ERDA are O
( 0.1 at.%), C (~0.01 at.%), and H (~0.01 at.%).
Increasing lSiNx to 13 Å, however, leads to a
transition in SiNx layer growth from epitaxial to
amorphous (see Figs. 1(b) and 1(c)). The
combination of XTEM, HR-XTEM, XRR, and
XRD analyses shows that the film consists of
polycrystalline TiN layers, with strong 002
preferred orientation, alternating with continuous
SiNx interlayers which are initially crystalline
(and exhibit local epitaxy with underlying TiN
grains) to a thickness of 5-13 Å before becoming
amorphous. We observe a similar transition from
epitaxial to amorphous SiNx growth in
multilayers deposited on MgO(001) at 300 °C
and 700 °C. We assume that the amorphous SiNx
phase has the composition Si3N4 since growth
was carried out in a highly N2-rich ambient. The
HREM images in Figs. 1(c) and 1(d) show that
local epitaxy of cubic-SiN occurs on some
TiN(001) grains over the complete 13 Å layer
thickness.
Our results show that there is an epitaxial
breakdown thickness (see Bratland et al. [10] for
discussion of breakdown mechanisms during
low-temperature strained layer growth [10]),
l*SiNx ~2.5-6.5 ML (5-13 Å), beyond which
The lower l*SiNx value observed for the
SiNx/TiN nanolaminates compared to the
SiN/TiN(001) superlattices is due to the higher
crystalline quality and lateral homogeneity of the
epitaxial TiN interlayer.
A similar hardness dependence on lSiNx has
been reported for TiN-SiNx nanocomposite films
[1,2]. The nanocomposites reach their maximum
hardness over a narrow SiNx thickness range
near the percolation limit at ~1-2 ML. Figure 2
shows that this thickness is well within the
pseudomorphic SiN regime for SiNx/TiN
nanolaminates containing polycrystalline TiN
layers grown on SiO2.
Our SiN/TiN(001) superlattice films exhibit
Koehler hardening [12,13] in which dislocation
glide is hindered across interfaces between
phases with large differences in shear moduli
(G). In this case, GSiN = 42 GPa (calculated
using DFT), while GTiN = 217-278 GPa [14,15].
For glide along maximum shear directions
<110> oriented 45˚ to the [001] loading direction
with a Taylor factor of 0.3, we estimate,
growing SiNx layers become amorphous. We
attribute the epitaxial stabilization of metastable
cubic SiN on TiN(001) to pseudomorphic forces.
However, the corresponding interfacial strain
increases linearly with lSiNx [11] and as lSiNx >
l*SiNx, the strain energy becomes sufficiently
large that the epitaxial growth front breaks
down. We infer that strain energy minimization
is responsible for the reduced N content in
metastable cubic-SiN epitaxial layers compared
to the equilibrium phase, hexagonal-structure βSi3N4, whose lattice constant (co = 2.90 Å; ao =
7.59 Å) mismatch with that of TiN (ao = 4.24 Å)
3
47
intensity of the c-3×3 reconstruction spots
decrease and the 1×5 pattern dominates. At SiN
= 0.5 ML, the 1×5 spots become streaks along
(110) reflections.
Figure 3(a) shows STM and LEED results for
SiN = 0.32±0.05 ML. The LEED pattern exhibits
a 90º-rotated two-domain 1×5 reconstruction
along orthogonal <110> directions. The
corresponding STM image is composed of two
domains, each with a characteristic “corn-cob”
atomic-scale morphology exhibiting a height
modulation of ~1 Å peak-to-peak. The width of
the five-row repeat structure, four “corn rows”
and a missing row, along both [110] and [1-10]
is 14.4±1.1 Å, which is in agreement with the
TiN interplanar spacing of ~3 Å. SiN deposition
experiments on TiN(111) with SiN values
ranging from 0.1 to 0.3 ML also lead to epitaxial
growth of NaCl-structure SiN. In this case,
LEED patterns exhibit a 2×2 reconstruction,
with weakly spotted streaks along (100), and
STM images also show SiN row formation.
FIG. 2.
Nanoindentation hardness of 0.5-µm-thick
SiN/TiN(001) superlattices grown on MgO(001) at 500 °C
(Series 1) and SiNx/TiN nanolaminates deposited on
Si(001) at 500 °C (Series 2) as a function of the SiNx layer
thickness lSiNx. The TiN layer thickness is 40 Å in both
cases. For comparison, the hardnesses of TiN(001) and
amorphous Si3N4 films are 20 and 24 GPa, respectively.
following the analysis in [16], a hardness
increase of 8 to 8.7 GPa for SiN/TiN(001)
superlattices compared to TiN(001) layers (20
GPa [17]). The maximum hardening observed in
our samples is, however, significantly larger
than that predicted by the Koehler effect. Thus,
additional hardening mechanisms, including
strengthening due to the coherency strains
between the two epitaxial layers, SiN and
TiN(001), must be present.
For probing SiNx/TiN interface formation at
the atomic level, we use in situ UHV STM and
LEED to follow SiNx structural evolution during
the early stages of growth on TiN(001). In these
experiments, high-quality epitaxial TiN(001)
template layers exhibiting 1×1 LEED patterns
are grown on MgO(001) following the
procedures developed in [5,18]. Next, Si is
deposited by thermal evaporation on TiN(001) at
room temperature and subsequently nitrided by
exposure to N2 at pN2 ~2 × 10-8 Torr for up to 12
h at temperatures T ranging from 600 to 800 ºC.
LEED analysis of layers with SiNx coverages
SiN < 0.30 ML reveals the coexistence of c-3×3
and 1×5 surface phases, oriented along <110>
directions. As SiN approaches 0.30 ML, the
FIG. 3. In situ STM images and corresponding 100 eV
LEED patterns from samples consisting of (a) 0.32 ML
SiN on TiN(001) exhibiting a two-domain 1×5
reconstruction and (b) 3 ML TiN on 0.84 ML SiN on
TiN(001) exhibiting a 1×1 surface. STM images were
obtained using an Omicron VT/STM operated in constant
current mode (1.5 V, 0.12 nA).
4
48
[001]
(a)
[-110]
To further investigate the interfacial selfassembly of TiN-SiNx nanocomposites, we
deposited 3 ML of TiN, following the same
procedure as with the initial TiN(001) layer,
onto SiN/TiN(001) bilayers with SiN = 0.85
ML. Figure 3(b) is an STM image from such a
sample. The TiN overlayer grows epitaxially on
SiN(001) and the surface exhibits the same 1×1
LEED pattern observed for the initial TiN(001)
template layer. Moreover, the TiN(001) islands
have their equilibrium shape; nearly square with
<110> edge facets and <100> corner facets [21].
Furthermore, the measured <110>/<100> island
step-edge anisotropy,
= 0.83, is in good
agreement
with
previous
results
for
TiN/TiN(001) homoepitaxy [19].
To provide additional insight into the atomic
structure of the SiN/TiN(001) interface and the
row-like SiN surface reconstruction, we employ
ab-initio DFT to determine minimum energy
structures as a function of SiN. The calculations,
with periodic boundary conditions, were carried
out using the generalized gradient approximation
of Perdew, Burke, and Ernzerhof [20] for
electronic exchange and correlation. Figure 4
shows the atomic-scale geometry of 0.9 ML SiN
on TiN(001), following relaxation of the
interfacial region, in (a) cross-section and (b)
plan view. The results were obtained by placing
Si and N on TiN lattice positions (Si on top of
NTiN and N on top of TiTiN) and with every 5th Si
[-110] row empty. The relaxed structure
geometry is comparable to the STM results (see
Fig. 3). Moreover, the SiN overlayer assumes a
cubic SiN structure. Si binds to N supplied
during nitridation as well as to the surface,
forming a surface reconstruction with a repeat
distance of five TiN unit cells along <110>, in
agreement with the experimental results. There
is an outward relaxation of the N atoms, in both
the SiN overlayer and the upper TiN planes,
similar to that observed in vacuum-terminated
TiN(001) surfaces [21]. N atoms in the SiN layer
move closer to Si-Si bridge positions, with
characteristic Si-N-Si bond angles of 111°, 127°,
(b)
[110]
[110]
FIG. 4. DFT (a) cross-sectional and (b) plan views
showing relaxed interfacial atomic positions for 0.9 ML
SiN on TiN(001). Ti: grey; N: blue; Si: yellow. Bond
lengths and angles in the bottom TiN layer were fixed. A
vacuum slab of 3 ML in the <001> direction was used.
Atoms in the SiN overlayer are enlarged for clarity.
and 143°, and in-plane bond lengths between
1.66 and 1.87 Å. Out-of-plane Si-N bond lengths
are 1.84 ± 0.01 Å, while Ti-N bonds are
somewhat longer than in bulk TiN (2.12 Å);
between 2.40 and 2.43 Å. The original TiN(001)
surface also reconstructs (see Fig. 4(a)).
Ab initio DFT calculations were used to
determine the minimum-energy configuration
with SiN = 1 ML. Two-domain, 90°-rotated,
SiN row structures along <110> directions were
observed. Eighty percent of the Si atoms retain
cubic lattice positions almost perfectly above the
surface Ti atoms, similar to the SiN = 0.9 ML
case, but with additional pronounced
5
49
cubic and hexagonal phases exhibit a lattice
mismatch (i.e., misfit strain) with TiN(001),
while for SiNx there is no strain associated with
amorphous layer growth. Thus, the growth
transition from epitaxial cubic SiN to amorphous
l*SiNx in
Si3N4 phase takes place at lSiNx
response to a change in both composition and
bonding coordination during elastic strain
relaxation.
The results presented here suggest that the
interfacial structure of superhard TiN-SiNx
nanocomposites and nanolaminates are much
more complex and play a more important role in
controlling film properties than previously
thought. Figure 5(a) is a schematic illustration of
the originally proposed phase structure, adapted
from Ref. [1-3], of superhard TiN-Si3N4
nanocomposites. It consists of nm-sized TiN
crystallites surrounded by a 1-1.5 ML thick
amorphous Si3N4 tissue phase at the percolation
threshold (~10 vol.%) [1,2]. Figure 5(b) includes
our results with epitaxial cubic SiN layers
growing on low-index facets of TiN crystallites
to a maximum thickness of 2.5-6.5 ML (5-13 Å)
before the transition to amorphous Si3N4. The
epitaxial breakdown thickness may be extended
if the SiN tissue phase is between two TiN
crystallites with which it forms coherent
interfaces. This situation corresponds to the
TiN/SiN(001) multilayers discussed above, for
which maximum hardness is reached at a SiN
thickness of 2-5 ML, well above the percolation
limit.
In summary, we report the first results on the
structure and chemistry of the interfaces which
control the properties of TiN-SiNx nanocomposites and nanolaminates. Using direct
observations supported by ab initio calculations,
we show that the SiNx/TiN interfaces are far
more complex than previously considered,
including the epitaxial stabilization of a
metastable cubic SiN phase on low index TiN
crystal surfaces. A variety of surface
reconstructions are also observed as a function
of SiN coverage. For SiN growth on TiN(001), a
reconstruction due to Si atom pairs forming a
dimer row above the SiN(001) surface. Si atoms
in the upper (non-dimer row) are covalently
bonded to their four nearest N neighbors (3 NSiN
3
+ 1 NTiN) in a distorted tetrahedral sp
configuration.
The calculated structures indicate that SiN
surface coverages 1 ML result in tetrahedral
environments, while lower SiN values favors
octahedral environments. In addition, for SiN =
1 ML, the range of Si and N bond lengths with
SiN = 1ML is 1.60-1.87 Å, close to the reported
value for Si-N compounds, 1.74 Å [22]. Si-NTiN
bonds have a strong covalent character and a
high local electron density (between 0.45 and
1.01 electrons/bond), comparable to diamond (1
electron/bond).
Epitaxial stabilization of non-equilibrium
phases in thin films and multilayers has been
well documented for a wide variety of materials
systems [23] including cubic AlN (whose
equilibrium structure is hexagonal wurtzite) on
NaCl-structure TiN(001) [24]. Another example
is the growth of 9-12 ML of cubic AlN in
AlN/TiN(001) superlattices (lattice mismatch =
-3.84%) before transformation to the equilibrium
structure [25]. The stabilization is due to a
reduction in interfacial energy during nucleation
of the metastable heterolayer and a decrease in
the overall strain energy during subsequent
island coalescence and layer growth. In the
present case of c-SiN/TiN(001), the misfit is
much smaller (0.5%) than for AlN/TiN(001), yet
the epitaxial breakdown thickness is only 2-3
ML.
There are two primary differences between cAlN/TiN(001) and c-SiN/TiN(001). (1) For both
the equilibrium and metastable AlN phases, the
stoichiometric ratio is 1, while for SiNx, N/Si =
1 for the epitaxial cubic phase and 1.33 for the
amorphous phase. As a consequence, our DFT
calculations show that the Si bonding
coordination changes from octahedral (6-fold
with N and/or Ti) for c-SiN to tetrahedral (4-fold
with N) for Si3N4. (2) In the AlN case, both the
50
the 1×1 LEED pattern observed for the initial
TiN(001) template. Continued heteroepitaxial
growth can be maintained in SiN/TiN(001)
superlattices wih SiN layer thicknesses of up to a
few ML giving rise to greatly enhanced
hardness. A transition to amorphous Si3N4 layer
growth follows with increasing lSiNx > l*SiNx due
to bonding strain in the SiN layer as Si atoms
prefer to be tetrahedrally coordinated with N.
These results suggest new design strategies
for nanoscale materials in which nanocomposite
systems and processing conditions are
purposefully selected in order to obtain tissue
phases which exhibit local epitaxy with the
encapsulated crystallites. This provides higher
interfacial bond strength, which further reduces
the probability of grain boundary sliding and
thus yields enhanced materials’ strength. One
can envision a new class of super-to-ultra hard
all-crystalline ceramic nanocomposites formed
during primary (film deposition) or secondary
(age processing) phase transformations in which
both the interfacial structure and the overall
preferential crystallographic orientation is
controlled.
ACKNOWLEDGEMENTS
LH, KL, and MO acknowledge support from
the Swedish Research Council and the Swedish
Foundation for Strategic Research. JB, VP, JEG,
and IP gratefully acknowledge the financial
support of the Department of Energy, Division
of Materials Science, and the use of the facilities
of the Center for Microanalysis of Materials,
University of Illinois, which is partially
supported by the U.S. Department of Energy
under grant DEFG02/91/ER45439. Prof. Stan
Veprek and Dr. Andreas Bergmaier are
acknowledged for kind assistance with ERDA
measurements. Dr. Jens Birch is acknowledged
for useful discussions. The DFT calculations
were performed using the CASTEP package
[26], developed by Accelrys Inc., San Diego,
CA.
Fig. 5. Cross-sectional schematic diagrams of superhard
TiN-SiNx
nanocomposites
consisting
of
TiN
nanocrystallites encapsulated by a SiNx tissue phase with
volume fraction just above the percolation limit. (a) Twophase model adapted from Refs. [1-3] in which SiNx is
amorphous Si3N4. (b) Three-phase model in which the
SiNx layer adjacent to low-index TiN planes is either
epitaxial cubic SiN (lSiNx 2-3 ML) or a bilayer consisting
of cubic SiN and amorphous Si3N4 (lSiNx > 3 ML). Ti:
grey; N: blue; Si: yellow.
c-3×3 surface reconstruction is obtained with
0.5 ML
0 < SiN < 0.3 ML, while 0.3
SiN
corresponds to a 90°-rotated two-domain 1×5
reconstruction along <110>. On TiN(111), SiN
forms a 2×2 reconstruction at coverages between
0.1 and 0.3 ML. We also show that TiN grows
epitaxially on SiN/TiN(001) bilayers, recovering
51
12. Koehler, J. S. Attempt to Design a Strong Solid. Phys.
Rev. B 2, 547-551 (1970).
REFERENCES
13. Barnett, A. in M.H. Francombe, M.H., and Vossen,
J.L. (eds.) Physics of Thin Films Vol. 17 (Academic Press,
San Diego, 1993).
1. Veprek, S., Veprek-Heijman, M. G. J., Karvankova, P.
& Prochazka, J. Different approaches to superhard
coatings and nanocomposites. Thin Solid Films 476, 1-29
(2005).
14. Ahuja, R., Eriksson, O., Wills, J. M. & Johansson, B.
Structural, elastic, and high-pressure properties of cubic
TiC, TiN, and TiO. Phys. Rev. B 53, 3072-3079 (1996).
2. Niederhofer, A. et al. The role of percolation threshold
for the control of the hardness and thermal stability of
super- and ultrahard nanocomposites. Surf. Coat. Tech.
146/147, 183-188 (2001).
15. Marlo, M. & Milman, V. Density-functional study of
bulk and surface properties of titanium nitride using
different exchange-correlation functionals. Phys. Rev. B
62, 2899-2907 (2000).
3. Veprek, S. & Reiprich, S. Thin Solid Films 268, 64-71
(1995).
16. Xu, J et al. Superhardness effects of heterostructure
NbN/TaN nanostructured multilayers. J. Appl. Phys. 89,
3674-3678 (2001).
4. Schiøtz, J., Di Tolla, F. D. & Jacobsen, K. W. Softening
of nanocrystalline metals at very small grain sizes. Nature
391, 561-563 (1998).
17. Ljungcrantz, H., Odén, M., Hultman, L., Greene, J.E.
& Sundgren., J.-E. Nanoindentation Studies of SingleCrystal (001), (011), and (111) Oriented TiN Layers on
MgO. J. Appl. Phys. 80, 6725-6733 (1996).
5. Söderberg, H., Molina-Aldareguia, J., Hultman, L. &
Odén, M. Nanostructure formation during deposition of
TiN/SiNx nanomultilayer films by reactive dual magnetron
sputtering. J. Appl. Phys. 97, 114327-1-8 (2005).
18. Kodambaka, S., Petrova, V., Vailionis, A., Petrov, I. &
Greene, J. E. In situ high-temperature scanning tunneling
microscopy studies of two-dimensional TiN island
coarsening kinetics on TiN(001). Surf. Sci. 526, 85-96
(2003)
6. Chen, Y.-H., Lee, K. W., Chiou, W.-A., Chung, Y.-W.
& Keer, L. M. Synthesis and structure of smooth,
superhard TiN/SiNx multilayer coatings with an equiaxed
microstructure. Surf. Coat. Tech. 146/147, 209-214
(2001).
19. S. Kodambaka et al. Absolute orientation-dependent
TiN(001) step energies from two-dimensional equilibrium
island shape and coarsening measurements on epitaxial
TiN(001) layers. Surf. Sci. 513, 468-474 (2002).
7. Hu, X., Zhang, H., Dai, J., Li, G. & Gu, M. Study on the
superhardness mechanism of Ti-Si-N nanocomposite
films: Influence of the thickness of the Si3N4 interfacial
phase. J. Vac. Sci. Technol. A 23, 114-117 (2005).
20. Perdew, J. P., Burke, K. & Ehnzerhof, M. Generalized
gradient approximation made simple. Phys. Rev. Lett. 77,
3865-3868 (1996); Perdew, J. P., Burke, K. & Ehnzerhof,
M. Errata: Generalized gradient approximation made
simple. Phys. Rev. Lett. 78, 1396 (1997).
8. Flink, A., Larsson, T., Sjölén, J., Karlsson, L. &
Hultman, L. Influence of Si on the microstructure of arc
evaporated (Ti,Si)N thin films; evidence for cubic solid
solutions and their thermal stability. Surf. Coat. Technol.
200, 1535-1542 (2005).
21. Gall, D., et al. Pathways of atomistic processes on
TiN(001) and (111) surfaces during film growth: An ab
initio study. J. Appl. Phys. 93, 9086-9094 (2003).
9. Wall, M.A., Cahill, D.G., Petrov, I., Gall, D. & Greene,
J.E. Nucleation Kinetics versus Nitrogen Partial Pressure
during Homoepitaxial Growth of Stoichiometric TiN(001):
a Scanning Tunneling Microscopy Study. Surf. Sci. 581,
L122-L127 (2005).
22. Aylward, A & Findlay, T. in P. Store (ed.) Si Chemical
Data, 3rd Edition, (Wiley & Sons, Milton, 1994)
10. Bratland, K.A., et al., J. Appl. Phys. 97, 44904 –
44910 (2005).
23. Yashar, P. C. & Sproul, W. D. Nanometer scale
multilayered hard coatings. Vacuum 55, 179-190 (1999).
11. P.Tu, K.N., Mayer, J.C. & Feldman, L.C. Electronic
Thin Film Science for Electrical Engineers and Materials
Scientists (Macmillan Pub., N.Y. 1992)
24. Petrov, I. et al. Synthesis of Metastable Epitaxial
Zincblende Structure AlN by Solid State Reaction. Appl.
Phys. Lett. 60, 2491-2493 (1992).
8
52
25. Kim, I. W., Li, Q., Marks, L. D. & Barnett, S. A.
Critical thickness for transformation of epitaxially
stabilized cubic AlN in superlattices. Appl. Phys. Lett. 78,
892-894 (2001).
26. Segall, M. D. et al. First-principles simulation: ideas,
illustrations and the CASTEP code. J. Phys: Cond. Matt.
14, 2717-2743 (2002).
9
53
Fly UP