DataPDF Available

Abstract

We review the thin film growth, chemistry, and physical properties of Group 4-6 transition-metal diboride (TMB2) thin films with AlB2-type crystal structure (Strukturbericht designation C32). Industrial applications are growing rapidly as TMB2 begin competing with conventional refractory ceramics like carbides and nitrides, including pseudo-binaries such as Ti1-xAlxN. The TMB2 crystal structure comprises graphite-like honeycombed atomic sheets of B interleaved by hexagonal close-packed TM layers. From the C32 crystal structure stems unique properties including high melting point, hardness, and corrosion resistance, yet limited oxidation resistance, combined with high electrical conductivity. We correlate the underlying chemical bonding, orbital overlap, and electronic structure to the mechanical properties, resistivity, and high-temperature properties unique to this class of materials. The review highlights the importance of avoiding contamination elements (like oxygen) and boron segregation on both the target and substrate sides during sputter deposition, for better-defined properties, regardless of the boride system investigated. This is a consequence of the strong tendency for B to segregate to TMB2 grain boundaries for boron-rich compositions of the growth flux. It is judged that sputter deposition of TMB2 films is at a tipping point towards a multitude of applications for TMB2 not solely as bulk materials, but also as protective coatings and electrically conducting high-temperature stable thin films.
Vacuum 196 (2022) 110567
Available online 3 September 2021
0042-207X/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Invited review
Review of transition-metal diboride thin lms
Martin Magnuson
*
, Lars Hultman, Hans H¨
ogberg
Thin Film Physics Division, Department of Physics, Chemistry, and Biology (IFM), Link¨
oping University, SE-581 83, Link¨
oping, Sweden
ARTICLE INFO
Keywords:
Diborides
Thin lm growth
Phase diagrams
Density functional theory
Properties
ABSTRACT
We review the thin lm growth, chemistry, and physical properties of Group 46 transition-metal diboride
(TMB
2
) thin lms with AlB
2
-type crystal structure (Strukturbericht designation C32). Industrial applications are
growing rapidly as TMB
2
begin competing with conventional refractory ceramics like carbides and nitrides,
including pseudo-binaries such as Ti
1-x
Al
x
N. The TMB
2
crystal structure comprises graphite-like honeycombed
atomic sheets of B interleaved by hexagonal close-packed TM layers. From the C32 crystal structure stems unique
properties including high melting point, hardness, and corrosion resistance, yet limited oxidation resistance,
combined with high electrical conductivity. We correlate the underlying chemical bonding, orbital overlap, and
electronic structure to the mechanical properties, resistivity, and high-temperature properties unique to this class
of materials. The review highlights the importance of avoiding contamination elements (like oxygen) and boron
segregation on both the target and substrate sides during sputter deposition, for better-dened properties,
regardless of the boride system investigated. This is a consequence of the strong tendency for B to segregate to
TMB
2
grain boundaries for boron-rich compositions of the growth ux. It is judged that sputter deposition of
TMB
2
lms is at a tipping point towards a multitude of applications for TMB
2
not solely as bulk materials, but
also as protective coatings and electrically conducting high-temperature stable thin lms.
1. Background
1.1. Introduction
Due to their technological importance, refractory and electrically
conductive diborides (TMB
2
) formed by Group 46 transition metals
(TM): Ti, Zr, Hf, V, Nb, Ta, Cr, Mo, and W are attracting increasing
research interest. This critical review is focused on Group 46 TMB
2
thin
lms with a predominant AlB
2
-type crystal structure (Strukturbericht
designation C32). A historical survey gives that the synthesis and
characterization work on borides date back to the late 19th century,
which resulted in the materials class known as TMB
2
. The family of
boron-containing compounds, thin lms and bulk, is larger than TMB
2
as
boron interacts with many of the elements in the periodic table to form a
variety of compounds with different properties. Examples are mono-
borides and covalently bonded borides with high hardness, like
aluminum magnesium boride (BAM) with the composition AlMgB
14
, and
superconducting MgB
2
.
In section 2, the atomic sizes and the electronegativities of B and the
group 46 TM are applied to predict the stability and chemical bonding
in the C32 crystal structure to describe how the arrangement of atoms
affects the property envelope shown by TMB
2
. From binary phase dia-
grams, the stabilities of the Group 46 TMB
2
are considered in com-
parison to other competing B-containing phases of different
stoichiometries. To understand properties, the underlying electronic
structure, chemical bonding, and band structures are evaluated in sec-
tion 3.
The progress of thin lm growth of TMB
2
is discussed in section 4
with a focus on sputtering, as a commonly employed physical vapor
deposition (PVD) technique. Other PVD techniques, such as cathodic
arc-deposition, pulsed laser deposition (PLD), and e-beam evaporation
as well as chemical vapor deposition (CVD) processing of TMB
2
lms,
are presented for reference. A survey of the eld reveals that TMB
2
has
been much less studied as thin lm materials compared to transition
metal nitrides (TMN) and transition metal carbides (TMC). Ternary and
pseudobinary boride systems are also discussed here.
In section 5, we outline difculties in controlling the level of con-
taminants, B/TM ratio, and microstructure in TMB
2
thin lms that
compromise their hardness, resistivity, and high-temperature stability
properties. Deviations from stoichiometry and chemical purity is pre-
sented as the culprit here, and this review shows that the typical sputter-
deposited Groups 46 TMB
2
lm exhibits B-rich composition when
* Corresponding author.
E-mail address: martin.magnuson@liu.se (M. Magnuson).
Contents lists available at ScienceDirect
Vacuum
journal homepage: www.elsevier.com/locate/vacuum
https://doi.org/10.1016/j.vacuum.2021.110567
Received 21 April 2021; Received in revised form 17 August 2021; Accepted 29 August 2021
Vacuum 196 (2022) 110567
2
grown from compound targets. This is due to mass differences between
boron and the Group 46 TM resulting in different scattering angles
when transported from the target to the substrate in the gas phase. In
addition, a characteristic ne-grained microstructure is frequently re-
ported. It too, can be explained by the strong driving force of B segre-
gation towards grain boundaries as well as its strong reactivity towards
oxygen and hydrogen, as common elements in the process ambient.
Section 6 summarizes the reviewed research studies. To enable
further advancement of stoichiometric sputter deposited TMB
2
lms,
and place TMB
2
thin lms in the position as the next generation of hard
refractory and conductive functional ceramic coatings, a higher level of
B-to-TM ratio control is needed. This is particularly the case for sput-
tering conditions pertaining to differences in mass of the metal and B
species and resulting anisotropic angular distribution of their ux from
the target. From the conducted research studies, we highlight the
importance of retaining a strict 2B:TM composition for the growth ux
in sputter-deposition of TMB
2
lms with well-dened properties. In
section 7, we suggest future research approaches with purer and more
dense compound targets, HiPIMS and hybrid PVD/CVD techniques.
1.2. Historical expos´
e
The earliest work on bulk synthesis of borides depended on isolating
the element boron, which was rst reported in 1808 by Gay-Lussac and
Th´
enard [1] and later Davy in 1809 [2]. The efforts were promoted by
the development of the electric furnace of the arc type by Siemens in
187879 see e.g. Ref. [3], and later seen by an operational furnace by
Readman in 1888. In 1894, Moissan [4] applied an electric furnace of
the arc type, equipped with C cathodes to study the reactions of the TM
chromium, including that to B. Moissan [4] found that the reaction
product was hard and resisted acids. In two publications 1901 and 1902,
Tucker and Moody [5,6] followed in the steps of Moissan and synthe-
sized borides from Zr, Cr, W, and Mo, using currents of 200275 A and
voltage 6075 V between the carbon cathodes. Tucker and Moody
concluded that the reaction products were crystalline, hard, of a high
specic gravity, not easily attacked by acids, and having very high
melting points. The synthesized compounds were determined to have
the formulæ Zr
3
B
4
, CrB, WB
2
, and Mo
3
B
4
. In addition, their studies
concluded that copper and bismuth exhibit no afnity to boron [5]. In a
publication from 1906, Binet du Jassoneix [7] continued by blending
MoO
2
with amorphous B in stoichiometric amounts and reducing the
oxide by H
2
. He found that compounds containing ~20% B were easily
synthesized, while higher boron contents up to 45.6% required higher
temperatures, but at the expense of BC
x
precipitations in the synthesized
material as well as formation of boric acid. The properties of the syn-
thesized compounds were shown to be highly dependent on the B con-
tent judged from hardness, scratching topaz, but not corundum (Al
2
O
3
)
and with corrosion resistance for HF and HCl, but not sulfuric acid. The
visible appearance of the synthesized product containing more than 20%
B was as described by Binet du Jassoneix [7]; bluish grey color, less
metallic, but with no visible grains (microstructural characterization by
electron microscopy was not available at the time).
Wedekind developed bulk synthesis using a vacuum oven and the
parent TM, Mn, and Cr as cathodes to strike the arc and initiate boride
formation with Mn [8] and Cr [9]. In a publication from 1913 [10], he
summarized the work for more TMs, like Zr and described zirconium
boride to be silver colored with a metallic luster and being temperature
stable, but without presenting crystallographic data. The described at-
tributes are typical for zirconium diboride, thus indicating successful
synthesis of the compound. In 1936, McKenna [11] advanced the eld
by demonstrating a synthesis route for bulk ZrB
2
from heating
zirconium-oxide and C with an excess of boron oxide at about 2000 C.
The resulting material consisted (in percent) of Zr 78.55%, B 18.15, C
1.89%, and Si 0.03%, in total 98.62% according to chemical analysis and
with a density of 5.64 g cm
3
: the density determined for bulk ZrB
2
is
6.104 g cm
3
[12]. Characterization by x-ray diffraction (XRD) revealed
a material with a hexagonal crystal structure and c =3.53 Å, a =3.15 Å
with a c/a =1.12. For ZrB
2
, c has later been determined to 3.53002(10)
Å and a =3.16870(8) Å, which gives c/a =1.1140250 [12]. In addition,
McKenna attempted to synthesize borides of tantalum and columbium
(Cb), later named niobium (Nb), but with limited description of the
synthesis products. Although McKenna concluded from XRD that ZrB
2
belongs to the family of materials with hexagonal crystal structure, no
information was provided about the symmetry of the structure [11]. In
1936, Hoffmann and J¨
aniche [13] determined the AlB
2
type structure,
often referred to by the Strukturbericht designation C32 or simply as
α
-type. This is the predominant crystal structure adopted by the TMB
2
formed by the Group 4-6 d-block elements to be discussed as thin lm
materials in this review. Consequently, in 1947, Ehrlich [14] found that
TiB
2
crystallizes in the C32 structure and where Kiessling [15] in 1949
showed that ZrB
2
adopts the C32 structure. In a publication from 1949,
Norton et al. [16] conrmed the crystal structure determination by
Ehrlich for TiB
2
[14] and Kiessling [15] for ZrB
2
as well as expanded to
the TMB
2
formed by the Group 5 TM, Nb, previously called Cb, Ta, and
V, and determined them to be of the C32 structure. In his work from
1950, Kiessling [17] showed that the Group 6 transition-metal Cr form
CrB
2
with C32 structure while Kiesslings earlier work from 1947
showed that Mo and W predominantly form borides of other composi-
tions and crystal structures [18]. The most well-known composition is
TM
2
B
5
[17], often referred to as the
ω
-phase. By 1953, Glaser et al. had
determined that the remaining Group 4 TM, Hf forms a HfB
2
with C32
crystal structure [19].
Improved synthesis methods for bulk TMB
2
and determination of its
crystal structure made it meaningful to investigate the properties of this
class of materials. Hence in 1954, Post et al. reported melting points for
both Mo
2
B
5
and W
2
B
5
and their solid solutions [20]. During the 1950
and 1960s, other researchers focused on the Hall effect and the electrical
conductivity of Group 46 TMB
2
[21], elastic constants of TiB
2
[22],
hardness as a function of temperature [23] for TiB
2
, ZrB
2
, HfB
2
, and
W
2
B
5
, and strength, fracture mode and thermal stress resistance of HfB
2
and ZrB
2
[24]. The properties of bulk TMB
2
have been comprehensively
reviewed by Fahrenholtz et al. [25].
In the 1970s, theoretical work on TMB
2
was initiated seen from
band-structure calculations on CrB
2
[26,27], TiB
2
[28], and ZrB
2
[29].
From calculations on CrB
2
, Liu et al. [27] reported difculties in
determining the density of states at the Fermi level (E
F
), which was
explained by spin uctuations. For TiB
2
and ZrB
2
, the above studies
found difculties in deciding whether a charge transfer from the TM to B
occurs. Ihara et al. [29] considered the band structure of ZrB
2
as a hybrid
structure similar to that of graphite and zirconium metal. These early
studies were pursued with insufcient accuracy in calculation of the
exchange correlation and hybridization of the orbital overlaps and
interstitial regions in the parent metal at the time [30]. Todays so-
phisticated density functional methods with higher numerical accuracy
in the potentials including more atoms and shells are better suited to
determine and predict properties, including magnetism and
charge-transfer in TMB
2
. DFT modelling can be utilized as a trend-giver
in target-oriented experimental work. For example, Moraes et al. [30]
used semi-automated DFT calculations across transition metal diborides
and showed that point defects such as vacancies inuence the phase
stability that can even reverse the preference for the AlB
2
or W
2
B
5-x
structure. Recently, they also found that V
x
W
1-x
B
2
exhibits ~40 GPa
hardness that can be useful in demanding applications [31]. Alling et al.
[32] observed that metastable Al
1-x
Ti
x
B
2
alloy could be of interest for
coherent isostructural decomposition (age hardening) in thin lms ap-
plications with a strong driving force for phase separation. Euchner and
Mayhofer [33] also found that the ternary diboride alloys Al
x
W
1-x
B
2
,
Ti
x
W
1-x
B
2
and V
x
W
1-x
B
2
represent a new class of metastable materials
that may open a large eld for further investigations.
The desire to obtain materials with better dened properties resulted
in the development of techniques for thin lm growth. Of importance for
the TMB
2
thin lms in this review is PVD and CVD. The development of
M. Magnuson et al.
Vacuum 196 (2022) 110567
3
PVD began in the 1850s with the pioneering work of Groove [34] and
later the more practical applications for sputter-deposited single and
multi-layer metal lms used as mirrors and optical coatings on telescope
lenses and eyepieces were discussed in papers published in 1877 by
Wright [35,36]. Work with CVD was initiated already in the 1600s [37],
but with more controlled processes developed in the late 1800s and in
the early 1900s for C by Sawyer and Man [38], carbon monoxide on Ni
by Mond, Langree and Quincke [39], electrical incandescing conductors
by Aylesworth [40], and for Ti in 1910 by Hunter [41].
For TMB
2
thin lms, Moers [42] developed halide-CVD in the 1930s,
while sputtering was reported in the 1970s by Wheeler and Brainard
[43]. In the 1980s, as the sputtering techniques were introduced, the
properties of the target material were improved. In 1997, Mitterer
summarized works on thin lm growth of the Group 4 TMB
2
, TiB
2
, and
ZrB
2
, including zirconium dodecaboride (ZrB
12
) coatings [44], and bo-
rides formed by the lanthanides LaB
6
, CeB
6
, SmB
6
, and YB
6
by different
sputtering techniques. Composition, structural properties, and micro-
structure of the investigated borides were thus directly connected to
their mechanical properties. Hexaborides are formed by conguration
interaction of 4f narrow-band states of heavy-fermion elements La and
Ce, LaB
6
[45] and CeB
6
[46] with charge-transfer to B. These materials
exhibit a variety of interesting properties, e.g., high melting points,
resistance to cathode poisoning and the lowest known work function
~2.5 eV, useful as electron emitters in electron microscopes, microwave
tubes, electron lithography, electron beam welding, x-ray tubes, and free
electron lasers [47]. In 2015, Andrievski [48] reviewed the Group 4
TMB
2
both as bulk and thin lms, focusing on their synthesis and the
resulting microstructure and mechanical properties.
Expanding the boride family from the TM, we note that in 1970,
Matkovich and Economy [49] synthesized and determined the structure
of a new class of ultrahard (>40 GPa) AlMg borides referred to as BAM.
In BAM, boron forms a three-dimensional (3D) network of four B
12
icosahedra in the unit cell that are stabilized by the electron-donating
metals. Property determination followed the successful synthesis of
BAM, where one outstanding property is the hardness with a reported
micro-hardness in the range 3235 GPa and with even higher hardness
values of 4046 GPa when alloyed with 30 mol% TiB
2
[50]. The prom-
ising results from bulk synthesis by Cook et al. [50], inspired thin lm
growth by sputtering, see, e.g. Refs. [51,52], for recent studies.
Boron carbide (BC
x
) discovered already in the 19th century as a by-
product of reactions involving metal borides is a very stable and
oxidation-resistant compound that is one of the hardest ceramic mate-
rials after diamond and cubic boron nitride (c-BN). In the 1930s, the
chemical composition of boron carbide was estimated as B
4
C [53]. Later,
x-ray crystallography showed that the structure of B
4
C is carbon de-
cient and highly complex, with a mixture of CBC chains and B
12
ico-
sahedra consisting of a combination of the B
12
C
3
and B
13
C
2
(B
12
CBC)
units [54]. The material is commonly used in body armor systems due to
its low weight, in reactors as neutron absorber and in stainless-steel
claddings. The different stoichiometries ranging from B
4
C to B
10.5
C
correspond to a carbon content between 20 and 8.7 at.%. Typical
properties of boron carbides are high hardness (third highest Vickers
hardness of 38 GPa after diamond and c-BN), corrosion resistance, and
high-temperature stability. Thus, covalent materials containing con-
nected B
12
icosahedras are boron-carbides B
12
C
3
[55] often referred to
as B
4
C [56]. The B
12
C
3
structure consists of eight B
12
icosahedra located
at the vertex of the rhombohedral unit cell and bonded together along
the long diagonal by a chain of three atoms.
In 2001, the discovery of superconductivity in bulk MgB
2
[5759]
further increased the interest in borides [58]. It soon turned to TMB
2
with C32 crystal structure. In 2001, Kaczorowski et al. [60] found a
critical temperature (T
c
) of 9.5 K in a powder sample of TaB
2
. Later in
2001, V. A. Gasparov et al. [61] investigated polycrystalline ZrB
2
, NbB
2
,
and TaB
2
pellets with over 6090% of the theoretical mass density. From
measurements of these materials, they determined T
c
to be 5.5 K for
ZrB
2
, while NbB
2
and TaB
2
were found to be superconducting only up to
0.37 K. Although potential superconducting materials, albeit at low T
c
, it
seems that the crystalline quality properties of the investigated TMB
2
bulk materials determines the measured T
c
, which was supported from
an earlier work published 1979 by Leyarovska and Leyarovski [62] that
showed no superconductivity above 0.42 K when conducting measure-
ments of pressed tablets of TMB
2
, where TM =Ti, Zr, Hf, V, Nb, Ta, Cr,
and Mo. Thus, it is apparent that to conclusively determine supercon-
ductivity in TMB
2
requires synthesized material of high crystal quality,
suggesting thin lm approaches.
Furthermore, the borides formed by Group 7 and 8 TM: Re, Os, and
Ru have received considerable attention due to the high hardness re-
ported for OsB
2
in 2005 [63], where a bulk piece of OsB
2
was found to
readily scratch a polished sapphire window. The interest in these TMB
2
phases originates from the properties of their parent TM, where for
instance Os, like diamond, shows a bulk modulus of 395462 GPa and
high valence electron density of 0.572 electrons/Å
3
[63]. Following the
promising results with OsB
2
as an ultra-incompressible hard material,
the attention was directed to ReB
2
in 2007 as Re share the previously
described properties of Os with bulk modulus of 360 GPa and valence
electron density of 0.4761 electrons/Å
3
[64]. For a ReB
2
pellet pressed
from Re and B powders then arc-melted to form a solid metallic ingot,
Chung et al. measured a Vickers hardness (HV) of 55.5 GPa at an applied
load of 0.49 N [64]. Although promising boride materials, it should be
stressed that the results on ultrahigh hardness in OsB
2
and ReB
2
have been
debated following a technical comment by Dubrovinskaia et al. in 2007
[65] on the data evaluation from the mechanical testing of the ReB
2
sample as well as from a study by Qin et al. [66] in 2008 who measured a
HV of only about 20 GPa for a ReB
2
compact synthesized at 5 GPa and
1600 C. Thus, OsB
2
and ReB
2
need to be conrmed as superhard thin
lm materials.
2. Transition-metal diborides
2.1. Atomic size effects
Fig. 1 shows the periodic table in which the 2nd row p-block ele-
ments B, C, and N are highlighted in yellow and the Groups 46 TM: Ti,
Zr, Hf, V, Nb, Ta, Cr, Mo, W in light blue. The atomic radii for each
element are listed above the International Union of Pure and Applied
Chemistry (IUPAC) symbol determined for each TM as well as for the p-
block elements. Atomic sizes are important when considering crystal
Fig. 1. Periodic table of the elements highlighting the second-row p-block el-
ements and Groups 46 transition metals forming diborides with C32 hexagonal
AlB
2
structure. The atomic radii and electronegativity values are from Refs. [67,
72,291,292].
M. Magnuson et al.
Vacuum 196 (2022) 110567
4
structures from a close-packing model, i.e. packing together equal-sized
spheres in 3D in a hexagonal closest packing (hcp) arrangement or in a
cubic closest packing (ccp) arrangement. In the model the size of the
atoms that form the basis determines the size of the interstitials that will
be present between the atoms in a hcp or a ccp arrangement. Conse-
quently, the size of the interstitials will also be decisive to what element,
atomic size, that can be positioned in the hcp or ccp arrangement.
In TMB
2
, the TM atoms will form close-packed layers, where B oc-
cupies trigonal prismatic sites, to be described in more detail in section
2.4. As observed from the trend in atomic size, the radii of the transition-
metal atoms decrease from left to right in the periodic table and
consequently the size of the interstitials present in the close packed
structure. This is due to a more effective nuclear charge that contract the
size of the electron cloud and hence the size of the atom.
Another general observation in Figure l is that the atomic radii in-
crease when moving down columns corresponding to a given Group in
the periodic table, but where Zr and Hf exhibit similar atomic radii due
to what is referred to as the lanthanide contraction [67]. Consequently, Zr
exhibits the largest size with a radius of l.60 Å, while Cr is the smallest
atom with a radius of l.28 Å. When B is compared to C and N, Slaters
rule for screening implies that the extension and localization of the 2p
x
orbitals increases with the lling by x =1 to 2 and 3 electrons, respec-
tively. However, by considering the increasing effective nuclear charge
with increasing atomic number, it is evident that B is the largest atom of
the three elements, exhibiting an atomic radius of 0.88 Å compared to C
(radius 0.77 Å) and N (radius 0.70 Å) [67].
In the 1930s, Gunnar H¨
agg showed that hydrides, borides, carbides,
nitrides, and oxides could be described in terms of metal atoms spheri-
cally packed with smaller H, B, C, N, and O atoms in octahedral or
tetrahedral interstices [68]. From x-ray diffraction and considering the
atomic radius ratio r
X
/r
TM
between non-metals (X) and TM, H¨
agg
empirically determined a trend in the complexity of hydride, carbide,
nitride, and oxide crystal structures. According to H¨
aggs stability rule
from 1931 [68], the structure of a non-metal/metal compound will be
either simple or more complex depending on whether the atomic radii
ratio is less than, or greater than 0.59. If r
x
/r
TM
<0.59, the phase will be
simple. This rule applies to interstitial structures of carbides, nitrides,
and borides formed by the transition metals of Group 46 in the periodic
table. Thus, H¨
aggsrule depends on the possibility of determining
atomic sizes. Here, we note that H¨
agg [68] applied a radius for B of
0.97 Å as derived from the phase Fe
2
B to compare to the smaller radius
of 0.88 Å that was later adopted by H¨
agg [67], while the radius of the C
and N atoms were identical or close to our values, cf. 0.77 Å for C and
0.71 Å for N. Following in the path of H¨
agg, Kiessling [17] developed the
stability criteria further to include more complex crystal structures
among the borides.
Using H¨
aggs empirical rule as starting point and considering the C32
structure as simple, we apply the model to explain the stability of TMB
2
.
In the C32 structure, there are trigonal prismatic interstitials (voids),
instead of octahedral interstitials that is common in other structures
such as cubic NaCl type structure (Strukturbericht designation B1). The
radius of the transition-metal atom in the C32 structure determines the
size of the void for the interstitials. For r
B
/r
TM
<0.59, the model predicts
a simple structure such as C32. However, for r
B
/r
TM
0.59, the model
predicts more complex crystal structures, decreasing stability, and other
compositions beyond that in TMB
2
with decreasing stability as discussed
in section 2.3.
Applying the atomic radii from Fig. 1, we nd that for the 3d TM:Ti,
V and Cr, the r
B
/r
TM
ratio 0.59 for all TM and increases from 0.60, to
0.66, and 0.69, and the model predicts more complex structures favored
for the VB and CrB binary systems, see section 2.3. For the 4d TM: Zr,
Nb and Mo, the r
B
/r
TM
ratio increases from 0.55, to 0.60, and further to
0.63 and the model predicts a simple structure for ZrB
2
, and where Mo
prefers a boride with Mo
2
B
5
composition (as mentioned in the intro-
duction). Finally, for the 5d TMs Hf, Ta, and W, the r
B
/r
TM
ratio increases
from 0.56, to 0.60, and 0.63 and the model predicts a simple structure
for HfB
2
and where W similar to Mo forms a boride with W
2
B
5
compo-
sition, but with a different stacking sequence compared to Mo
2
B
5
[17].
For the stabilization of metastable structures such as WB
2
or MoB
2
,
Moraes et al. [31,69] showed that defect engineering in terms of boron
and metal vacancies is also important.
From above, the H¨
aggs rule predicts that ZrB
2
is the most stable C32
TMB
2
followed by HfB
2
, while TiB
2
is the least stable phase in Group 4
[25,70]. TiB
2
, TaB
2
and NbB
2
are close to the border of fullling H¨
agg
stability, but less stable phases compared to ZrB
2
and HfB
2
. Further-
more, the determined r
B
/r
TM
ratio from Fig. 1 are identical to those
presented by Kiessling [17] for Cr, Nb (Cb), Mo, and Ta, but with 0.59
for Ti, 0.54 for Zr, 0.65 for V, and 0.62 for W; while Kiessling provided
no data for Hf.
When applying H¨
aggs rule [68] to TMC and TMN, the smaller radii
of C and N compared to B will increase the number of phases with a
simple B1 structure for both materials systems. For TMC:s Cr alone ex-
hibits a r
C
/r
TM
ratio 0.59 (0.60), while all TMN show r
N
/r
TM
<0.59.
Thus, from H¨
aggs rule transition metal borides appear to have less
preference in forming a simple crystal structure as C32 when compared
to TMC and TMN.
2.2. Electronegativity
Borides are, by denition, formed by B and an electropositive
element i.e., most metals. Thus, in a boride the B acts as electron
acceptor and the metal acts as electron donor. As an estimate on the
degree of electron transfer, the character of the chemical bonding as
metallic, covalent, or ionic, we consider the difference in electronega-
tivity value (
χ
) between the electron acceptor B and the electron
donating metal. A large difference in electronegative value (Δ
χ
) is
synonymous with a high degree of electron transfer, i.e. ionic character
in the bonding. For
χ
using the Paulings scale, there is maximum Δ
χ
of
3.19 in the periodic table for CsF, as the binary compound is formed
from the most electronegative element in the periodic table, F with
χ
=3.98 and the most electropositive and stable element, Cs with
χ
=0.79. This Δ
χ
value is said to correspond to a bonding with 100%
ionic character and where smaller Δ
χ
implies mixtures between ionic,
covalent, and metallic bonding [71]. Here, we stress that this approach
is a rough estimate of the chemical bonding in TMB
2
as the C32 crystal
structure makes it necessary to not solely consider B-TM bonds, but in
addition BB bonding and TM-TM bonding, see section 2.3-2.4. How-
ever, as electron transfer is the foundation for compound formation, we
apply Δ
χ
to consider the stabilities of the TMB
2
formed by the Group
46 TM.
Table 1 lists the difference in electronegativity between B and the
Group 46 TM, using the values presented in Fig. 1 below the IUPAC
symbol of each element. As can be seen, BHf followed by BZr have the
largest difference in Δ
χ
with 0.74 and 0.71, respectively. This is far from
a bonding with ionic character, and rather corresponding to a mixture of
metallic and covalent bonding with an estimated ionic character of less
than 20% [71]. It is a consequence of Hf and Zr being the most elec-
tropositive TM in Group 46. For the other TM, Δ
χ
decreases seen from
BTa, BTi, BNb, BV, and BCr. Particularly, the high
χ
values for Mo
with
χ
=2.20 and W with
χ
=2.36 that are higher than that of B with
χ
=2.04 [72]. By considering the degree of electron transfer from the
TM to B as an indicator of the stability of the TMB
2
, note that HfB
2
and
ZrB
2
should be the two most stable phases, while MoB
2
and WB
2
should
be less stable due to their weaker B-TM bond. This is supported from the
discussion above of H¨
aggsrule and the fact that the stronger TM-TM
interaction in TMB
2
formed by Mo and W results in other crystal
structures than C32. In section 2.4, we report difculties in sputtering
compounds like MoB
2
and WB
2
as thin lms due to their lower stability
than competing phases.
The lower
χ
of В compared to C with
χ
=2.54 and N with
χ
=3.04
will result in different mixtures between ionic, metallic, and covalent
bonding. Comparing Group 46 TMB
2
to TMC, and TMN, TMB
2
shows
M. Magnuson et al.
Vacuum 196 (2022) 110567
5
more metallic character, TMC bonds are of more covalent character,
while TMN has the highest degree of ionic contribution [71]. Section 5,
compares the hardness and resistivity of TMB
2
to those of carbides and
nitrides.
2.3. The AlB
2
crystal structure
From our previous discussion, it is clear that all Group 46 TM form
TMB
2
phases with C32 structure and where ZrB
2
and HfB
2
seems to be
the two most stable phases considering H¨
aggs rule (r
B
/r
TM
ratio) and
difference in electronegative value (Δ
χ
) while particularly Mo and W
prefer to form borides of other compositions such as TM
2
B
5
and crystal
structures. In this section, we describe the C32 structure in more detail
as it holds the key to the properties demonstrated by the TMB
2
, which
are characterized by high hardness [25]; high melting points; good
electrical and thermal conductivities, corrosion, and erosion resistance
[48]; high chemical inertness and stability; as well as high wear and
thermal-shock resistance [25].
Fig. 2a shows the unit cell of the C32 crystal structure, with metal
atoms in the basis (0,0,0) and boron atoms positioned in trigonal pris-
matic interstices at (1/3, 2/3, 1/2) and (2/3, 1/3, 1/2) belonging to the
hexagonal crystal class and space group 191, in HermannMauguin
notation P6/mmm. Fig. 2b shows the projection in reciprocal space in
hexagonal symmetry with the major symmetry directions indicated. The
symmetry points Γ, M, A, K, L, and H in the band structure give rise to
prominent peak structures in the density of states (DOS) discussed in
section 3. In Fig. 2(c) and (d), the unit cell has been translated to reveal
the symmetry of the layered structure characteristic to the C32 crystal
structure. As can be seen, it consists of honeycomb 2D graphite-like
sheets of B-atoms (borophene) alternating with hexagonal close-packed
TM-layers. The C32 structure is often described from a stacking
sequence of AHAHAH . . [17], where A corresponds to the TM layers
and the borophene sheets to H. Furthermore, each M-atom is surrounded
by six equivalent M-neighbors in metal planes and by 12 equidistant B
neighbors: six above and six below. Each B has three B neighbors in the
basal plane and six M-atoms out-of-plane: three above and three below.
The borophene sheets are characteristic host sites for vacancies as well as
common impurities such as oxygen and carbon and to some extent ni-
trogen. Vacancies and impurities within the borophene sheets may in-
uence the properties by weakening the direct BB bonding and
destabilizes the TMB
2
structure. This is different from the interstitial
TMC and TMN compounds with predominantly B1 crystal structure that
are characterized by large homogeneity ranges [73], where the isolated
C or N atom can be removed to form a vacancy as well as possible solid
solutions, where the C or N can be replaced by for instance O [74].
Table 2 lists the lattice parameters and internal bond distances in the
C32 structure for the TMB
2
formed by the Group 46 TM. As can be seen,
the longest a axis is found in ZrB
2
, followed by HfB
2
, NbB
2
, TaB
2
, MoB
2
,
TiB
2
, WB
2
, VB
2
, and nally CrB
2
. By assuming that the length of the a
axis in TMB
2
corresponds to the TM-TM distance, we note from the
atomic radii in Fig. 1 that TiB
2
deviates from this hypothesis as Nb and
Ta have identical atomic radii as Ti 1.46 Å and where the Mo atom is
slightly smaller than the Ti atom (radius of 1.39 Å). Post et al. [20]
explained this from the BB distances in the borophene sheets, where 4d
and 5d TM slightly increase the BB distances to achieve TM-TM
contraction in the a-axis direction, while 3d TM retain a normalBB
distance of 1.75 Å ((r
B
+r
B
)/2) as in TiB
2
or even slightly decreasing the
BB separation as for VB
2
and CrB
2
, see Table 1, and the BB separation
in Group 46 TMB
2
. Here, WB
2
deviates from the a axis trend postulated
by Post et al. [20] and where the authors presented no data for WB
2
. This
can be explained by the stability of WB
2
with C32 crystal structure as
discussed in section 2.4. For the length of the c axis, there is a different
Fig. 2. The C32 hexagonal AlB
2
crystal structure: (a) unit cell, (b) reciprocal
space projection for the P6/mmm group that shows the major symmetry di-
rections, (c) schematic of chemical bonds viewed from the side, and (d) from
the top.
Table 2
Lattice parameters a and b from Ref. [293] for TiB
2
[12], for ZrB
2
[294], for HfB
2
[295], for VB
2
[296], for NbB
2
[297], for TaB
2
[298], for CrB
2
[299], for MoB
2
, and
[82] for WB
2
. The B-TM and BB bond lengths and c/a ratios were determined from the lattice parameters using the structure program Visualization for Electronic and
Structural Analysis (VESTA).
Boride TM group a (Å) c (Å) c/a -B-TM (Å) BB (Å)
TiB
2
4 3.03034(8) 3.22953(14) 1.0657295 2.38085(0) 1.74957(0)
ZrB
2
4 3.16870(8) 3.53002(10) 1.1140250 2.54208(0) 1.82945(0)
HfB
2
4 3.14245(8) 3.47602(10) 1.1061472 2.51244(0) 1.81430(0)
VB
2
5 2.99761(9) 3.05620(12) 1.0195429 2.30875(0) 1.73068(0)
NbB
2
5 3.11133(13) 3.2743(2) 1.0523855 2.43045(0) 1.79633(0)
TaB
2
5 3.09803(7) 3.22660(12) 1.0415001 2.40874(0) 1.78865(0)
CrB
2
6 2.9730(13) 3.0709(2) 1.0329319 2.30301(0) 1.71646(0)
MoB
2
6 3.04 3.07 1.0098684 2.33169(0) 1.75514(0)
WB
2
6 3.02 3.05 1.00993377 2.31641(0) 1.74360(0)
Table 1
Electronegativity values for B and the Group 46 TM [72].
χ
(B)
χ
(TM) Δ
χ
2.04 Ti =1.54 0.5
2.04 Zr =1.33 0.71
2.04 Hf =1.30 0.74
2.04 V =1.63 0.41
2.04 Nb =1.60 0.44
2.04 Ta =1.50 0.54
2.04 Cr =1.66 0.38
2.04 Mo =2.20 0.16
2.04 W =2.36 0.32
M. Magnuson et al.
Vacuum 196 (2022) 110567
6
trend where the c axis of TiB
2
is slightly larger than that of TaB
2
and
much larger than that of MoB
2
, but where VB
2
and MoB
2
exhibit a
shorter c axis than CrB
2
. For the stable TMB
2
phases, the c/a ratios in
Table 1 are in the range 1.01.1. Norton et al. [16] and Kiessling [17]
noted the rather regular increase in the c/a ratio that scales with the size
of the metal atom. Consequently, ZrB
2
exhibits the largest c/a ratio and
where CrB
2
should exhibit the smallest c/a ratio given the smallest
atomic radius of 1.28 Å. Post et al. [20] determined a similar trend for
the c/a ratios in stable TMB
2
. From the work by Post et al. [20] and
Table 2 in this review, we note the deviation for MoB
2
and WB
2
that is
probably due to difculties in determining the lattice parameters for
these TMB
2
s. The trends in the a-lattice parameter and the c/a ratio as
well as the B-TM and the BB bond lengths are further developed in
section 3.1.
In addition to TMB
2
with C32 crystal structure, the Group 46 TM
form borides of other compositions and crystal structures with the B
atoms hosted as: 3D B
12
icosahedrons or B
6
octahedrons at or puckered
2D nets or fragments of nets, 1D zig-zag chains, paired atoms or nally
as individual isolated atoms depending on the stoichiometry of the
phase. Table 3 lists the most prevalent compositions and crystal struc-
tures that are used to support the discussion on the binary phase dia-
grams in section 2.4. It should be emphasized that there are other more
boron-rich borides than TMB
2
formed by the Group 46 TM such as
TMB
12
, ZrB
51
and CrB
41
as well as not listed compositions such as V
5
B
6
.
For this extension of boride phases, the reader is referred to Ref. [75].
For Group 4 to 6 borides, the crystal structure for borides with TM
2
B
5
and TMB composition is worth commenting. The former composition is
preferred by the TM:s Mo, and W, but differ in crystal class, since Mo
2
B
5
is reported to be trigonal, while W
2
B
5
is hexagonal. In his summary of
crystal structures adopted by borides from 1950, Kiessling [17]
described the stacking sequence for Mo
2
B
5
and W
2
B
5
following the
terminology for hexagonal closest packing with ABABAB and cubic
closest packing with ABCABCABC This results in a stacking sequence
of AHAKBHBKCHCKAHA in Mo
2
B
5
and AHAKBHBKAHA in W
2
B
5
and
where we previously dened the stacking sequence in TMB
2
with C32
crystal structure to be AHAHAHAH. Here, A, B, and C are close-packed
Mo layers, where layers B and C are shifted by (a/3, 2a/3) and (2a/3,
a/3), respectively, while H are planar graphite-like sheets and K are
(buckled) boron rings. Thus, both the Mo
2
B
5
and W
2
B
5
crystal structures
contain plane nets of B atoms (H) as in the C32 crystal structure, but
where these nets are interleaved by puckered sheets of B atoms (K). In
addition, some of the K sheets contain an additional B atom positioned in
the center of the mesh. To further increase the complexity, the compo-
sition in Mo
2
B
5
and W
2
B
5
has been reported to concern solely a structure
type and not the actual composition that is rather MoB
2-x
and WB
2-x
[76].
2.4. Monoborides
Similar to the TMB
2
materials with C32 crystal structure, all Group
46 form TM monoborides TMB, albeit with different crystal structures
as reviewed by Minyaev and Hoffmann in 1991 [75]. Kiessling [17]
identied at least three crystal type structures for TMB phases; CrB, FeB,
and MoB, where CrB and FeB belong to the orthorhombic crystal class,
while MoB is tetragonal. Common to these crystal structures are the B
zig-zag chains, but where the angle between the boron atoms differs
from 115 to 117in CrB, to 110-112in FeB and to 127in MoB [17],
and with possible homogeneity ranges in the MoB type structure seen
from 48.8 to 51.5 at.% B in MoB and 48.0 to 50.5 at.% B in WB [17].
Lately, the different structures of binary borides have been reviewed by,
e.g., Chen and Zou [77].
In the upcoming section on the binary phase diagrams for: BTi,
BZr, BHf, BV, BNb, BTa, BCr, BMo, and BW, the homogeneity
ranges will be discussed in TMB
2
with the C32 crystal structure as well as
the phase order presented in Table 3 in connection to their respective
phase diagrams.
To conclude, the group 46 TM borides exhibits similarities between
the crystal structures determined for the same composition, although,
differs in terms of symmetry, composition and homogeneity ranges. This
motivates continued synthesis, exploration and theoretical work.
2.5. Phase diagrams and stability
Fig. 3 shows binary phase diagrams for BZr, BNb, and BMoB. As
a general observation, the number of phases increases with increasing
Group number in the periodic table [75].
Starting with Group 4 and the ZrB system, it is clear from the phase
diagram in Fig. 3 (top panel) that Zr forms a stable TMB
2
compound with
C32 structure as adopted from Massalski et al. [78] Stable TMB
2
phases
are also found in both the TiB and HfB systems. From the ZrB phase
diagram, the phase ZrB
2
is described as a line compound with a
composition 33.33 at.% Zr and 66.67 at.% B. In a recent study, Engberg
et al. [79], discussed the possibly of a narrow homogeneity range in ZrB
2
lms deposited by direct current magnetron sputtering (DCMS) inves-
tigated by atom probe tomography (ATP) in combination with
cross-sectional TEM and Z-contrast STEM data, where grain stoichiom-
etry measurements show up to 4 at.% deviations from stoichiometry. For
TiB
2
, the deviation from stoichiometry is ~1 at.% [48] and where the
phase diagram published by Massalski et al. [78] indicates that a vari-
ation in stoichiometry is possible already at 500 C, which is a typical
temperature condition applied in PVD and CVD. The BHf phase dia-
gram presented in Ref. [78] only provides data above 1400 C and with a
distinct homogeneity range from ~65 to ~68 at.% B at ~1500 C.
Andrievski [48] supports a homogeneity range in HfB
2
above 1400 C
from a calculated phase diagram, but where HfB
2
is a line compound at
Table 3
Crystal structures adopted of borides with CrB
4
type [86], Ta
3
B
4
type [75], U
3
Si
2
type [84], and CuAl
2
type [300].
Composition
(TM
x
B
y
)
Crystal type
structure
Crystal class Comments
TMB
12
UB
12
Cubic Borides with three-dimensional B frameworks (B
12
icosahedra)
TMB
6
CaB
6
Cubic Borides with three-dimensional B frameworks (B
6
octahedra)
TMB
4
CrB
4
Orthorhombic Borides with three-dimensional framework of tetrahedrally-coordinated B
TM
2
B
5
Mo
2
B
5
Trigonal Boride with B atoms in hexagonal at nets (H) or slightly puckered sheets (K): stacking sequence
AHAKBHBKCHCKAHA
TM
2
B
5
W
2
B
5
Hexagonal Boride with B atoms in hexagonal at nets (H) or slightly puckered sheets (K): stacking sequence AHAKBHBKAHA
TMB
2
AlB
2
(C32) Hexagonal Borides with two-dimensional hexagonal nets (H) of B atoms: stacking sequence AHAHAH
TM
3
B
4
Ta
3
B
4
Orthorhombic Borides with double chain (fragments of nets)
TMB CrB Orthorhombic Borides with boron zig-zag chains
TMB FeB Orthorhombic Borides with boron zig-zag chains
TMB MoB Tetragonal Borides with boron zig-zag chains
TM
3
B
2
U
3
Si
2
Tetragonal Borides with B dumbbells
TM
2
B CuAl
2
Tetragonal Boride with isolated B atoms
M. Magnuson et al.
Vacuum 196 (2022) 110567
7
temperatures typical for PVD i.e., below 1000 C [48]. Although there
are indications of homogeneity ranges in the Group 4 TMB
2
at elevated
temperatures, it is evident that the C32 structure is stable and preferred
by ZrB
2
and HfB
2
. This is supported by our discussion in section 2.2 on
the stabilities of TMB
2
:s with C32 structure considering H¨
aggs rule and
Δ
χ
.
In the BZr system, there is a B-rich zirconium dodecaboride ZrB
12
with UB
12
-type structure [48], and reports of an ZrB monoboride with
the FeB-type structure [48] not presented in the phase diagram by
Massalski et al. [78]. Similar to Zr, Ti and Hf form monoborides, and
where Hf forms also a HfB
12
phase [48]. In contrast to Zr and Hf, Ti
forms Ti
3
B
4
with Ta
3
B
4
-type structure [75].
Moving to Group 5 with TMs characterized by decreasing atomic
radii and hence size of the interstitials in combination with increasing
electronegativity (
χ
) values promote the formation of phases with other
stoichiometries and crystal structures. As can be seen for the BNb
system in Fig. 3 (middle panel), the number of binary phases compared
to the BZr system have increased to at least four [78]. Similar as in the
Group 4 B-TM phase diagrams, the Group 5 TMs form TMB
2
phases with
C32 structure as present in the phase diagram for NbB in Fig. 3.
Differently from the Group 4 TMB
2
, the NbB
2
(CbB
2
) and TaB
2
phases
exhibit more pronounced homogeneity ranges as reported by Kiessling
[17] and later determined by Lundstr¨
om [80] to extend from 65 to 70 at.
% B in NbB
2
and 6673 at.% B in TaB
2
, but with no reports for any
homogeneity range in VB
2
, where Massalskis phase diagram [78] marks
the phase as a line compound with a melting point at 2747 C. In the
BNb phase diagram, there is a Nb
3
B
4
phase with Ta
3
B
4
type structure
[75], a NbB phase but with CrB type structure [17], and Nb
3
B
2
with
possible U
3
Si
2
type structure similar to that of V
3
B
2
[78]. The BV and
the BTa systems are similar as the BNb system with the formation of
TM
3
B
4
, TMB, and TM
3
B
2
phases of crystal structures previously dened
for the BNb system. In the BTa phase diagram there is an additional
metal-rich boride Ta
2
B with CuAl
2
type structure [17], while the BV
phase diagram musters two additional phases with the composition V
5
B
6
and V
2
B
3
[78]. The large number of vanadium borides with different
stoichiometry and crystal structures is explained by the small radius of
the V atom of 1.34 Å and the relatively high
χ
value for V of 1.63, which
results in smaller interstitials that seems to destabilize the boron sheets
and where the higher
χ
value promotes metallic bonding. This is sup-
ported by observing the even higher number of determined phases for
Vs right hand neighbor Cr [75] and where we note the atomic radius for
Cr of 1.28 Å with a
χ
value of Cr of 1.66 i.e., smaller atomic radius and
higher
χ
compared to V.
As presented in Fig. 1, the atomic radii decrease and
χ
values increase
further for the Group 6 TM. This has impact on the BMo phase diagram
that show additional phases compared to the BNb phase diagram. The
Group 6 TMs Cr and Mo form TMB
2
phases with C32 structure as seen for
the MoB system in Fig. 3 (bottom panel), but where MoB
2
is only stable
above 1500 C [78], i.e., higher than deposition temperatures applied in
PVD. For MoB
2
, Klesnar et al. [81] studied the composition of bulk
MoB
2-x
and determined a narrow homogeneity range centered at 61 at.%
B in the temperature range 16001800 C, while in the BCr phase di-
agram [78], CrB
2
is illustrated as a line compound to the melting point of
2200 C [78]. In the BW phase diagram [78], no WB
2
with C32 crystal
structure is present at temperatures above 1800 C. Instead, there is a
two-phase area shared by
α
-WB (MoB type structure) and the previously
described W
2
B
5
phase. However, WB
2
with C32 crystal structure was
prepared by Woods et al., in 1966 [82] by heating a boron wire in an
atmosphere of WCl
6
and Ar and later as a sputtered thin lm by Sobol in
2006 [83].
Returning to the BMo phase diagram, we nd the Mo
2
B
5
phase,
α
-MoB with MoB type structure, a possible Mo
3
B
2
phase (U
3
Si
2
type
structure) [84], and a Mo-rich Mo
2
B phase with CuAl
2
type structure
[85]. On the B-rich side there is a MoB
4
phase with CrB
4
-type structure
[86] and a high temperature β-MoB phase with CrB-type structure [80].
Naturally, the BCr and the BW binary phase diagrams both exhibit
similarities to the BMo phase diagram with respect to phase distribu-
tion as seen from CrB
4
and WB
4
phases with CrB
4
type structure, CrB and
high temperature β-WB with CrB type structure, and metal-rich phases
Cr
2
B and W
2
B with CuAl
2
type structure. For Cr, there are two additional
phases with Cr
3
B
4
with Ta
3
B
4
type structure and a phase with Cr
5
B
3
composition. From the Group 46 phase diagrams, there is a trend where
the number of phases increases when moving to the right in the periodic
table and where Cr seems to form the largest number of phases ranging
from B-rich CrB
41
to Cr-rich Cr
4
B [75]. Ternary boride systems and solid
solutions are presented in section 4.7.
3. Electronic structure of diborides
3.1. Chemical bonding
In section 2, we described the borides in terms of atomic size, elec-
tronegativity, and crystal structure as a basis to determine the chemical
bonding in TMB
2
. The predominant bonding consists of a mixture of (i)
covalent, (ii) metallic, and (iii) ionic [25]. Generally, the borides have
Fig. 3. Phase diagrams of the binary systems ZrB, NbB, and MoB. Reprinted
from Ref. [78] with permission of AMS International.
M. Magnuson et al.
Vacuum 196 (2022) 110567
8
less directional bonds in terms of well-dened electronic orbitals
involving covalent s-p-d hybrid congurations, compared to the stronger
p-d hybridization in the carbides and the nitrides [87]. This has impli-
cations on their mechanical properties and properties at elevated tem-
peratures as discussed in section 5.
Firstly, considering the covalent portion of B-TM bonding in TMB
2
, it
exhibits a distorted hybridization in the valence electron conguration
of B. While an isolated B-atom has a valence of 2s
2
2p
1
, a mixed sp
2
-sp
3
hybridized covalent bonding occurs when diborides are formed, which
affects the materials properties. The strength of the covalent bonding
depends on the combined effect of the energy of orbital overlaps relative
to E
F
and the electron density in the bonding orbitals.
As shown in Table 2, for Zr (the largest TM atom), the BB distance in
ZrB
2
of 1.83 Å exceeds the normalequilibrium BB distance of 1.75 Å
that has a minimum bond strain by 0.08 Å for TiB
2
. This is due to tensile
stretching in the BB bonds as the Zr atoms are in contact with each
other and give rise to the highest conductivity among the Group 46
TMB
2
[25]. In contrast, Cr and V (the smallest TM atoms) have reduced
BB distances and larger bond strengths. Among the TMB
2
, the shortest
BB bond length, 1.72 Å, occurs for CrB
2
due to compressive strain on B
from the TM atoms. An explanation to this is a decrease of the size of the
interstices, where the boron atoms are positioned in the C32 crystal
structure.
The hardness, thermal stability, and elastic properties originate from
the strong BB and B-TM bonds [88,89] that are most signicant for
Group 4 TM borides (Ti, Zr, and Hf) [25]. The strengths of the covalent
TM d B 2p bonding in TMB
2
compounds mostly connects with the bulk
modulus i.e., the resistance to compression, while the strength of the
covalent BB bonding and atomic separation in the borophene sheets,
interleaved between hexagonal close-packed (hcp) structure TM atoms,
reects the hardness and the length of the a axis in the C32 crystal
structure. As further discussed in section 5.5, the amount of covalency
and directionality in the TM-B bonds and the ionic radii can, to a rst
approximation, explain the trend in the melting points.
Secondly, the metallic contribution to bonding (near E
F
), increases
when moving to the right in the periodic table as the d-bands become
more lled by electrons [90]. However, the metallic bonding is partly
counteracted by the amount of antibonding metal states in the vicinity of
E
F
that cause weakening of the bond strength. As discussed in section
2.2, this phenomenon is a consequence of the increased difference in
electronegativity between the metal and the B atoms.
Note that the bonding in the C32 crystal structure discussed here, for
which the B atoms are positioned in 2D sheets, differs signicantly from
the class of materials with 3D B
12
icosahedra and UB
12
type structure
commonly known as covalent metals that have more itinerant metallic
electronic states at E
F
. An example is ZrB
12
, which has strong covalent
bonding between the Zr 4d orbitals and the B 2p orbitals, but 3D
icosahedral BB interactions [91,92], as there are no 2D borophene
sheets in the structure.
Thirdly, the ionic part of the bonding in the TMB
2
originates from the
electron transfer of two electrons from the metal atoms to each of the B
atoms, involving signicant charge transfer. In the simplest charge-
transfer model, the TMB
2
can be denoted as TM
2+
(B
)
2
, where the
charge transfer from the metal atoms stabilizes the BB bonding in the
2D borophene sheets of the C32 crystal structure.
It has been shown theoretically that in Group 4 the early TMB
2
compounds prefer the C32 crystal structure due to their low density-of-
states (DOS) at the E
F
compared to competing phases such as the W
2
B
5
-
structure [30]. According to the calculations by Moraes et al. [30], early
transition metal diborides (TiB
2
, VB
2
, etc.) tend to be chemically more
stable in the AlB
2
structure type, whereas late transition metal diborides
(WB
2
, ReB
2
, etc.) are preferably stabilized in the W
2
B
5x
structure type.
However, point defects such as vacancies also signicantly inuence the
phase stability and can even reverse the preference for the AlB
2
or
W
2
B
5x
structure [30]. In reality, TMB
2
compounds with higher TM
atomic numbers, in the same row, than VB
2
(3d), NbB
2
(4d), and TaB
2
(5d) exhibit higher chemical stability in the hexagonal W
2
B
5
crystal
structure than in the hexagonal C32 crystal structure. Thus, the theo-
retical calculations deviate from the observations that can be explained
by B point defects such as vacancies in the buckled borophene layers. For
HfB
2
(1012
μ
Ω cm), the bulk resistivity is higher than for TiB
2
(912
μ
Ω cm) and ZrB
2
(710
μ
Ω cm) as discussed in section 5.4.
Numerical computational methods based on the formation energy at
0 K for selected phases, including known competing phases, are often
applied to predict ground-state stabilities of crystal structures. For stable
structures, trends in bonding properties such as chemical orbital popu-
lation and charge-transfer among atoms can be systematically compared
as a function of the atomic radius of the transition-metal atom. In
particular, the size of the trigonal prismatic interstitial space decreases
when the size of the metal-atom decreases. This has the consequence
that graphite-like B sheets are distorted from a planar atomic layer to a
buckled layer. Buckling occurs when the B atoms are very close to each
other and are forced to buckle as there is not enough space for them in
the planar (graphite-like) structure [93]. The differences in the
geometrical stabilities and the total energies of initially planar and
buckled AlB
2
structures were theoretically analyzed by Pallas et al. [93].
They found that the early TM atoms (Y, Zr, Nb) prefer the planar AlB
2
structure, while the late TM atoms (Tc, Ru, Rh, and Pd) prefer the
buckled structure where the TM-TM antibonding orbitals play an
important role. In this case, MoB
2
is a limiting case that has no obvious
preference for either planar or buckled structure and therefore form
Mo
2
B
5
instead of MoB
2
. On the other hand, initially buckled structures
of YB
2
and ZrB
2
became planar after geometry optimization, while
buckled structures were not formed for these systems. This is due to the
fact that the trigonal prismatic holes become too small to form a buckled
structure. For the buckled structures, strong TM-TM antibonding or-
bitals are a decisive factor for the structure to form while for the planar
structure, charge-transfer is more important.
The W
2
B
5
-type of structure is often referred to as an archetype
complex structure in terms of stacking sequence [30] as well as phases
with the orthorhombic crystal structure, which become more stable
when moving to the right in the periodic table. The trend in complexity
has been theoretically conrmed for both the 3d [94] and the 4d [93]
transition-metals borides, where the r
B
/r
TM
radius ratio <0.59 is
correlated to the most energetically preferred planar structures. TM
borides with a radius ratio equal or above 0.59 tend to form buckled
structures [93]. However, for light atoms like B, phonon modes may also
stabilize structures that is not included in 0 K ground-state calculations.
The borophene layer in the C32 crystal structure has unique proper-
ties as it is stronger and more exible than graphene, has low resistivity,
is a superconductor, and play an important role for the properties of the
TMB
2
. In 1997, different polytypes of quasi-planar surfaces of mono-
layer sheets of B were theoretically predicted using ab initio quantum
chemical and density functional theory [95]. In 2015, borophene was
rst synthesized in UHV on Ag(111) substrates using CVD by Mannix
et al. [96] and in 2018 on Al(111) substrates [97]. In the latter case, the
increased charge-transfer from Al to B in graphene-like borophene was
found to improve the phase stability. At least four different structures of
borophene have been identied; Besides the corrugated or buckled phase,
mentioned above, with space group 2-Pmmn, the triangular lattices with
space groups b
12
, c
3
[98] and the planar graphene-like phases with
hexagonal honeycomb symmetry. Recently, the production of
free-standing borophene has been shown [99].
The strength, exibility, and resistivity properties of borophene can
be tuned by the arrangement of vacancies and the orientation of the
material [100]. We anticipate that borophene and borides with built-in
borophene layers have potential to become important in emerging ap-
plications as anode materials in metal-ion batteries, hydrogen storage,
hydrogen peroxide, catalytic reactions, and sensors.
M. Magnuson et al.
Vacuum 196 (2022) 110567
9
3.2. Density of states
As discussed in the previous section, the electronic structure of the
metal-diborides affects the properties via the chemical bonding. Fig. 4
shows calculated partial density of states (DOS) for Zr, Nb, Mo s, p, d,
and B s, p in the left panels and B p
xy
, p
z
states of ZrB
2
, MoB
2
, and NbB
2
in
the right panels. For ZrB
2
, E
F
is found at a DOS minimum between
completely occupied bonding states and unoccupied antibonding pd-
bands. This position of the E
F
located in the center of a pseudogap,
energetically stabilizes the crystal structure in a similar way as for TiC
[87]. It indicates that the isoelectronic Group 4 TiB
2
, ZrB
2
, and HfB
2
are
the most stable phases [32,101]. This observation is consistent with
H¨
aggs rule and the trend in the difference in electronegativities be-
tween the metal and B atoms discussed in section 2.1 and section 2.2,
respectively. The position of E
F
also inuences the resistivity (see section
5.4). Comparing ZrB
2
with NbB
2
and MoB
2
in Fig. 4, the E
F
for the latter
two TMB
2
:s is situated among antibonding states, which is not ener-
getically ideal and weakens the covalent bonding.
The density of metallic 3d, 4d, and 5d states at E
F
increases towards
the right in the periodic table as the bands become progressively more
lled [88,90]. The B-TM bonding is associated with two relatively weak
2p
z
-
π
orbitals oriented perpendicular to the B-planes while forming a
strong graphite-like B network involving four 2p
xy
in-plane
σ
orbitals in
the basal plane. For ZrB
2
shown in Fig. 4, the states within 6 eV of E
F
are
dominated by metallic Zr 4d states, while B 2p and B 2s states are more
localized at 34 eV and 11 eV, respectively. The right panel in Fig. 4
shows that the B 2p
xy
-
σ
states in the basal plane of ZrB
2
exhibit a
double-peak at 3.35 eV and 4.75 eV with a 1.4 eV peak splitting. The
peak at 3.35 eV is mainly due to a at band containing B 2p
σ
states in
the vicinity of the M (1/2,0,0) symmetry point [102] in the Brillouin
zone of the reciprocal hexagonal lattice as shown in Fig. 2b. In addition,
there are peaks at 7.5 eV and 10 eV near the bottom of the valence band
due to B 2sp and B 2s states, respectively. Notably, these bands only
occur for the p
xy
σ
states in the basal plane and are due to the strong BB
bonds in the B sheets. The out-of-plane B 2p
z
π
contribution also exhibits
a double peak at 2.33 eV and 3.35 eV (1 eV peak splitting) that is due to
B 2p Zr 4d interaction. The sharp ZrB
2
B 2p
z
interplanar peak located at
3.35 eV, increases to 4.74 eV for NbB
2
and 5.34 eV for MoB
2
. A larger
binding energy position of the B p
z
peak relative to E
F
signies a shorter
bond length and stronger bond in the c direction between the B sheets
and the metal layers, which affects the elasticity of the material.
For the TMB
2
compounds, the DOS consists of TM spd hybridized
states, B sp-hybridized states, and antibonding states above the pseu-
dogap. The pseudogap at E
F
in the TMB
2
compounds has primarily been
associated with the strong covalent interaction between the in-plane BB
2p-states, and less due to the covalent interplanar B-TM interaction
[103].
According to Motts law of conductivity in 1969 [104], the increasing
number of states at the E
F
from ZrB
2
(0.2933 states/eV/atom) versus
NbB
2
(1.0335 states/eV/atom) and MoB
2
(1.4241 states/eV/atom)
should [104] increase the conductivity of the TMB
2
material due to the
metal bonding and affect the stability by charge redistribution in the
valence band. This rough approximation can be compared to the
opposite trend in the experimental resistivity values of bulk materials in
Table 4 for ZrB
2
yielding a value of 710
μ
Ωcm (8.0
μ
Ωcm) [105]
compared to 12
μ
Ωcm [73] for NbB
2
and 18
μ
Ωcm [73] for Mo
2
B
5
.
However, Motts law assumes isotropic scattering and carrier mobility
(the Mott approximation), which differs from the experimental deter-
mination of anisotropy in carrier lifetimes. Furthermore, the higher re-
sistivity of Mo
2
B
5
can be attributed to the larger amount of anti-bonding
states at E
F
in comparison to ZrB
2
. Thus, ZrB
2
has the lowest resistivity:
Fig. 4. Zr, Nb, and Mo s, p, d, and B s, p partial densities of states for ZrB
2
,
NbB
2
, and MoB
2
(left panels) and B p
xy
and p
z
states (right panels) [Authors
work, unpublished]. The E
F
is represented by the vertical dashed line at zero
energy. The energy shift of the main peak for the three systems is shown by the
vertical dot/dashed line and the horizontal arrows.
Table 4
(top): Bulk hardness in GPa, and (middle) bulk resistivity in
μ
Ωcm. Melting
points C (bottom) [73]. *TMB
2
with other crystal structures than C32. TMC
and TMN with other crystal structure than B1.
Bulk hardness [GPa]
4 5 6
TiB
2
1545 VB
2
20.9 CrB
2
20.5
ZrB
2
22.523 NbB
2
23.2 *Mo
2
B
5
23.0
HfB
2
28 ref. [25] TaB
2
22.6 *W
2
B
5
26.1
TiC 2835 VC 27.2 Cr
3
C
2
1018
ZrC 25.9 NbC 19.6 Mo
2
C 15.524.5
HfC 26.1 TaC 16.7 WC 22 (0001)
TiN 1821 VN 14.2
ZrN 15.8 NbN 13.3
HfN 16.3 TaN 11.0
Bulk resistivity [
μ
Ωcm]
4 5 6
TiB
2
915 VB
2
13 CrB
2
18
ZrB
2
710 NbB
2
12 *Mo
2
B
5
18
HfB
2
1012 TaB
2
14 *W
2
B
5
19
TiC 68 VC 60 Cr
3
C
2
75
ZrC 43 NbC 35 Mo
2
C 71
HfC 37 TaC 25 WC 22 (0001)
TiN 2025 VN 85
ZrN 721 NbN 5878
HfN 33 TaN 135250
Bulk melting points [C]
4 5 6
TiB
2
2980 VB
2
2100 CrB
2
2170
ZrB
2
3040 NbB
2
3050 *Mo
2
B
5
2100
HfB
2
3250 TaB
2
3200 *W
2
B
5
2600
TiC 3067 VC 2830 Cr
3
C
2
1810
ZrC 3420 NbC 3600 Mo
2
C 2520
HfC 3928 TaC 3950 WC 2870 (0001)
TiN 2950 VN 2177
ZrN 2980 NbN ~2400
HfN 3387 TaN 3093
M. Magnuson et al.
Vacuum 196 (2022) 110567
10
ZrB
2
710
μ
Ωcm, NbB
2
12
μ
Ωcm and Mo
2
B
5
18
μ
Ωcm (see Table 4 for
bulk materials). The respective borides resistivity is discussed further in
section 5.4.
3.3. Orbital overlaps and hybridization
Fig. 5 schematically illustrates energy levels of the bonding and
antibonding orbitals involved in the chemical bonding of diborides
(middle), compared to a B sheet (left) and a transition metal (right). E
F
is
indicated by the horizontal dashed line. The energy levels of the B sheet
consist of well-dened B 2p
xy
σ
and B 2p
z
π
molecular orbitals, while the
sp and d states in the transition metal are spread out in broader bands. As
the TM and B atoms combine to form the C32 crystal structure, the
energy positions of the hybridized bands are associated with changes in
the electronic lling of the d-band and the covalent B p TM d band.
As discussed above, the valence band of TMB
2
consists of B 2p states
that reects the 2D graphite-like layers [106] and form four in-plane
2p
xy
states with
σ
bonds and two out-of-plane 2p
z
states with
π
bonds.
Foremost two in-plane B 2p
xy
σ
bands cross E
F
and contribute consid-
erably to the density of states at E
F
and, hence, to the metallic-like
properties. The out-of-plane B 2p
z
-
π
bands correspond to the weak
interlayer bonding. The electrical and magnetic properties largely
depend on the number of electrons in the valence band. Calculations
have shown that the electron-phonon coupling of the
σ
and
π
bonds is
highly anisotropic with a dominant role due to optical in-plane phonons
[107] and phonon dispersion around the Γ-symmetry point in the Bril-
louin zone of ZrB
2
[108].
Since the discovery of bulk superconductivity in magnesium dibor-
ide, MgB
2
by Nagamatsu in 2001 [59] with a transition temperature of
T
c
~ 39 K, which at that time was the highest determined for a con-
ventional (phonon mediated, non-copper-oxide) bulk superconductor,
the existence or nonexistence of superconductivity in several diborides
has been a controversial issue. For example, in 2001, Gasparov et al.
[109] claimed the discovery of a superconducting transition at 5.5 K in
ZrB
2
polycrystalline pressed powder samples while Kaczorowski et al.
[60] claimed a superconducting transition at 9.5 K in powder samples of
TaB
2
and that ZrB
2
was not superconducting. Moreover, the difference in
T
c
in the TMB
2
materials has been associated with different
electron-phonon coupling and the ratio between the B p-states versus TM
d-states at the E
F
[92]. For MgB
2
, the obtained T
c
largely depend on the
sample purity, in particular, it is sensitive to the amount of surface
oxidation into understoichiometric MgB
x
O
y
[110].
The B layer has been identied as critical for the superconducting
properties of bulk MgB
2
with T
c
39 K. The MgMg, MgB, and BB
bond lengths are 3.080, 2.502, and 1.778 Å, respectively. Raman spec-
troscopy has shown that for the B atoms, there are two orthogonal in-
plane phonon modes around the Γ-symmetry point having E
2g
symme-
try that dominates [111,112]. The E
2g
phonon mode that changes the
BB bond length is related to the
σ
-bonded B 2p
xy
in-plane orbitals.
These phonon modes have been identied as the key factor that in-
uences the electron-phonon interaction and the superconducting
properties of MgB
2
[113].
A similar superconductivity mechanism as in MgB
2
with B/phonon
interactions leading to an energy gap has been suggested [114,115] for
other borides with the C32 crystal structure. In comparison to ZrB
2
(0.2933 states/eV/atom), band-structure calculations show a higher
DOS at E
F
for MgB
2
(0.8858 states/eV/atom with E
F
located at a peak of
occupied states) while it is 1.0335 states/eV/atom for NbB
2
and 1.4241
states/eV/atom for MoB
2
. The relatively high superconducting transi-
tion temperature T
c
, ~39 K, for bulk MgB
2
has primarily been associated
with the large density of B p-states at E
F
, while for ZrB
2
, the DOS at E
F
is
dominated by metallic 4d-states. A dominating B 2p DOS for MgB
2
has
also been conrmed experimentally by x-ray absorption and emission
spectroscopies [116,117]. The electronic structures of bulk ZrB
2
and
TiB
2
have been experimentally investigated by photoemission spec-
troscopy [29,118], x-ray absorption spectroscopy [118,119], optical
spectroscopy [120122], and point-contact spectroscopy [123].
The factors responsible for the possible superconducting properties
of TMB
2
with C32 crystal structure have been theoretically analyzed
[92,102]. It was found that the main contributing factors to supercon-
ductivity in MgB
2
are the high density of 2p
xy
(in-plane) BB bonds with
σ
-symmetry consisting of partly occupied degenerate at bands in the
vicinity of the E
F
and relatively weak ionic MgB bonding. The degen-
erate and partly occupied at B
σ
-bands around E
F
in MgB
2
is associated
with an effective electron-phonon coupling between to the
σ
BB
stretching phonon mode within the B-plane and the low-frequency
anharmonic E
2g
phonon mode that is advantageous for creating
Cooper pairs across the small energy gaps across the E
F
. MgB
2
has both
σ
and
π
electrons with different energy gaps of 2.2 and 7.1 meV, at the E
F
with different coherence lengths (51 nm and 13 nm) that indicate a
mixture of type I and type II kind of superconductivity that has also been
referred to as type-1.5 superconductivity [124]. For comparison, Group
5 TMB
2
, the TM - B covalent bonding changes the energy and increases
the dispersion of the B
σ
-band away from the E
F
that is unfavorable for
the energy gaps and phonon modes of superconductivity. For TMB
2
to
the right of Group 4 in the periodic table, the B
σ
-band is completely
occupied and does not have the possibility to split the degeneracy by
electron-phonon coupling. A high density of TM d-states at E
F
with
strong covalent TM-B bonding and hybridization is disadvantageous as
the
σ
-band disperse away from the E
F
. As shown in section 3.2 (Fig. 4),
the distance of the B
σ
- and
π
-bands increase from the E
F
as the d-band is
being lled from Group 4 to 6 that also affects the bands crossing the E
F
.
Fig. 5. Energy levels of the bonding and antibonding orbitals involved in the chemical bonding of diborides [Authors work, unpublished].
M. Magnuson et al.
Vacuum 196 (2022) 110567
11
Thus, among the layered diborides, only MgB
2
is a superconductor at
normal pressure without strain.
Furthermore, it has been theoretically shown that pressure-induced
strain can reduce the covalent TM-B overlap when the lattice is com-
pressed along the a-b direction or stretched along the c direction. Both
types of strain rise the degenerate
σ
-band (p
xy
) involved in the BB bonds
and increases the band splitting so that it nestles around the E
F
while
elongation of the c axis softens the degenerate E
2g
stretching phonon
that split the B
σ
band that is necessary to produce Cooper pairing [102].
Thus, a suitable strain can be chosen to reduce the hybridization of the
TM-B bonds and increase the electron-phonon coupling. It is anticipated
that extreme pressures could increase the transition temperatures of
suitable materials, ideally even up towards room temperature.
Another way to control and improve the superconductivity than
strain is doping of TMB
2
, e.g., V-doped ZrB
2
. Renosto et al. [125] studied
the Zr
0.96
V
0.04
B
2
alloy that exhibits a bulk superconducting transition
temperature of T
c
~ 8.7 K as V induces superconductivity in the
non-superconducting material ZrB
2
. In 2017, Barbero et al. [126] also
found that V-doped ZrB
2
and HfB
2
alloys Zr
0.96
V
0.04
B
2
and Hf
0.97
V
0.03
B
2
had a T
c
of ~8 K. On the other hand, doping of MgB
2
with TiB
2
[127] and
ZrB
2
[128] can cause superconductivity by increasing the c-lattice
parameter. This is also the case with excess B, as demonstrated for
NbB
2+x
[129,130].
A small fraction of V substitution of TMB
2
has several effects, that
combined give rise to superconductivity. The E
F
moves very close to the
bottom of the d
xz
, d
yz
degenerated bands and obtains a at shape along
the Γ-A symmetry line. These factors change the Fermi surface due to the
phonon mode involved that break the d
xz
-d
yz
degeneracy, in a similar
way as the E
2g
phonon mode breaks the degeneracy of the
σ
band in
MgB
2
[102].
3.4. Bond length by x-ray spectroscopy
In the quest to understand the B-TM bonding, it is important to keep
the oxygen level as low as possible as it causes unnecessary complica-
tions associated to B enrichment at surfaces and grain boundaries in the
microstructure [131].
Experimentally, oxidation states and local short-range order atomic
coordination symmetry have been probed with the combination of x-ray
absorption near-edge spectroscopy (XANES) and extended x-ray ab-
sorption ne structure (EXAFS) spectroscopy [132] that are comple-
mentary tools to long-range order probed by x-ray diffraction (XRD).
XANES provides; (i) a quantitative measure of the average oxidation
state by the energy shift of the appropriate x-ray absorption edge; (ii) the
amount of p-d hybridization in the chemical bonds, and thus the sym-
metry of the structure; and (iii) information about the coordination
symmetry when the XANES line shape is compared to reference mate-
rials [133]. EXAFS gives quantitative information on average (i) bond
lengths, (ii) coordination number (number of nearest- and next-nearest
neighbors in different directions), and (iii) the mean-square disorder
(Debye-Waller factor that also depends on the temperature and phonon
vibrations).
For diborides, it is primarily bulk ZrB
2
[25,48] that has been studied,
with only a few reports on thin lms [131,134,135]. Chu et al. [136]
made EXAFS analysis of polycrystalline ZrB
2
samples synthesized from
Zr and B powders by the oat-zone method. Temperature-dependent
measurements showed little difference between in-plane and
out-of-plane vibrations in the ZrZr bonds [137]. B¨
osenberg et al. [138]
investigated the chemical states of ZrB
2
and Zr powders by XANES and
found that the energy of the Zr K-edge for ZrB
2
was 10 eV above that of a
Zr foil reference. High-energy shifted unoccupied Zr 4s states were also
found for e-beam co-evaporated ZrB
2
thin lms by XANES and EXAFS
applied to investigate the local chemical bonding structure and the
atomic distances [131]. However, the O content in these lms resulted
in the formation of crystalline tetragonal ZrO
2
, which yielded longer
atomic bond distances as determined by EXAFS.
Recently, a XANES and EXAFS diboride lm study by Magnuson et al.
[135] on a magnetron sputter deposited epitaxial ZrB
2
lm using the Zr
K-edge showed shorter BB bonds (1.827 Å) than in a ZrB
2
compound
target (1.833 Å) with 99.5% purity from which the lm was synthesized.
The ZrB distance was also somewhat shorter (2.539 Å) in the epitaxial
ZrB
2
thin lm than in the bulk sample (2.599 Å). This can be compared
to the reference value of 2.542 Å in Table 2 that is in between the thin
lm and bulk in Ref. [135] The ZrB bond distance was found to be
longer in bulk than in the thin lm sample due to additional super-
imposed ZrO bonds resulting from contaminants, which is consistent
with the observations in XANES, XPS, and XRD. The measured ZrB
value (2.539 Å) was in better agreement with the calculated bond dis-
tance from the lattice parameters of the relaxed equilibrium bulk liter-
ature value (ZrB =2.542 Å) [12] than that of the ZrB
2
compound
target. Furthermore, the ZrB bond length in the ZrB
2
epitaxial lm was
shorter than the 2.546 Å obtained by Stewart et al. [131,134] for e-beam
co-evaporated lms from elemental Zr and B sources of 99.5% purity.
Chu et al. [136] obtained a ZrB bond length of 2.55 Å on single-phase
polycrystalline samples prepared by the rf oating-zone method from
stoichiometric mixtures of Zr and B elements compacted to pellets and
arc melted. For comparison, Lee et al. [139] found a ZrB bond length of
2.81 Å in magnetron sputtered ZrB
2
multilayers to be used in sensors. A
longer ZrB bond is likely due to additional impurities and
non-directional bonds. The ZrB
2
lm in Ref. [135] exhibited superior
electronic structure properties in terms of chemical bonding alignment,
crystal quality and level of contaminants when compared to the ZrB
2
compound target from which it was synthesized. Thus, it can be antic-
ipated that - at present level of synthesis process optimization - thin lms
provide better means than bulk materials to investigate and determine
fundamental properties. As discussed in section 5.3, ZrB
2
lms are more
elastic than bulk ZrB
2
, by virtue of the formers high purity.
Other diborides such as TiB
2
, VB
2
, and CrB
2
have been scarcely
studied as bulk materials and even less as thin lms. This is due to dif-
culties thus far of growing thin lms with well-dened properties of
these materials. Chu et al. [137] used EXAFS to study the bond lengths of
bulk polycrystalline TiB
2
and found TiB and TiTi distances of 2.37 and
3.07 Å. They also measured the bond lengths of bulk polycrystalline VB
2
and CrB
2
[140]. Chartier et al. [141] reported EXAFS on amorphous
understoichiometric thin lms of TiB
2
deposited by dynamic ion mixing
a TiB distance of 2.37 Å and a TiTi distance of 3.03 Å in comparison to
bulk TiB
2
with distances of 2.39 Å and 3.01 Å, respectively. Analogous to
ZrB
2
, we anticipate that the bond distances in stoichiometric
single-crystal lms of TiB
2
, VB
2
, and CrB
2
will be somewhat shorter and
as a result the properties would be superior compared to polycrystalline
materials.
4. Transition-metal diboride lm synthesis with extension to
ternary systems
There are two predominant techniques for vapor-phase growth of
thin lms: CVD and PVD. This review focuses on sputter deposition with
DCMS, radio frequency (rf) sputtering, and pulsed techniques such as
high-power impulse magnetron sputtering (HiPIMS). In addition to
sputtering, PVD growth of borides has been performed with cathodic-arc
evaporation as further developed in section 4.2 and exemplied by
studies on the TiB system. In section 4.3, e-beam evaporation, and
pulsed-laser deposition (PLD) are discussed, followed by the few studies
on reactive sputtering of borides in section 4.4. In section 4.5, we bring
attention to sputter deposition of Group 46 TMB
2
layers with C32
crystal structure and where the properties of these lms are further
treated in section 5.
4.1. Chemical vapor deposition
Vapor-phase synthesis of borides by CVD was initiated in the early
1930s, where Moers [42] investigated growth from the B precursor BBr
3
,
M. Magnuson et al.
Vacuum 196 (2022) 110567
12
and the metal halides TiCl
4
, ZrCl
4
, HfCl
4
, VCl
4
, and TaCl
5
. Moers focused
on the chemistry in the CVD processes studied and provided limited
amount of information on the properties of the deposited lms. For
instance, the phase distribution of the deposited material was charac-
terized by x-ray diffraction (XRD), but with no conclusive evidence for
growth of, e.g., ZrB
2
or ZrB. However, the work [42] represented a
stepping-stone for halideCVD that was later developed mainly in the
USA starting from the 1960s [142147] and most likely inspired by the
Apollo programs (19601972) demand for new coating materials such
as TMB
2
. An advantage of halide CVD was that the process chemistry
resembled that developed in parallel for industrial growth of carbide and
nitride coatings on cutting tools used for metal machining. Thus, expe-
rience from halide CVD of carbides and nitrides could be used when
developing processes for TMB
2
. On the other hand, halide CVD of TMB
2
coatings was, like CVD of TMC and TMN coatings, limited by the high
temperatures required to activate and dissociate the precursors, typi-
cally well above 1000 C, as well as parasitic etching by the hydrochloric
acid generated from the chemical reaction. Under these conditions,
Gannon et al. [142] in 1963 deposited crystalline and 110-oriented TiB
2
coatings on graphite sleeves at 1600 C, but with no additional infor-
mation on the composition of the grown coatings. Lowering the depo-
sition temperature to 1400 C, Gebhardt and Cree [143] in their work
from 1965 grew crystalline stochiometric and non-porous TiB
2
on
graphite using thermal decomposition of gaseous reactants (a pyrolysis
process similar to CVD) of halide and boron trichloride in a graphite
resistance-heated vacuum furnace. The B and metal content in the de-
posits were determined from chemical analysis including pyrohydrolysis
and titration for elemental analysis of the boron and calorimetric pro-
cedures for the metal. ZrB
2
and HfB
2
coatings were also deposited, albeit
as phase mixtures together with rhombohedral boron.
In 1974, Takahashi and Kamiya [144] studied growth of TiB
2
coat-
ings on graphite substrates below 1200 C, and where TiB
2
was depos-
ited together with an excess of (amorphous) B. The deposition of B-rich
coatings indicates a higher reactivity of the BCl
3
precursor compared to
the applied metal halide, making it necessary to investigate alternative
boron precursors to better match the reactivity to that of the metal
halide. To lower the deposition temperature, reduce etching, and
improve properties of the deposited coatings, Pierson and Mullendore
[148] in 1980 replaced BCl
3
by diborane (B
2
H
6
) and grew TMB
2
at
500900 C. Stoichiometric TiB
2
coatings were thus deposited on
graphite at 700 C, but with B-excess at lower deposition temperatures
as determined from electron microprobe analysis and induction-coupled
plasma spectroscopy i.e., like that reported by Gebhardt and Cree [143]
for ZrB
2
and HfB
2
coatings as well as by and Takahashi and Kamiya
[144] for TiB
2
coatings. Given the undesired etching and the high
temperature processing characteristic for halide CVD described above,
there was a need for developing low temperature synthesis routes,
specically at low substrate temperatures, below 600 C. For this pro-
pose, single-source precursors such as transition-metal tetrahy-
droborates TM(BH
4
)
4
have been investigated starting in the late 1980s
[149151].
Epitaxial growth of ZrB
2
lms from the precursor Zr(BH
4
)
4
has been
demonstrated on Si(111) [152] [153], Si(001) [154], and Al
2
O
3
[155],
but at substrate temperatures of 9001000 C. These high temperatures
are favorable for decomposition of the Zr(BH
4
)
4
precursor and sufcient
to activate the H
2
for abstraction of the excess B. These conditions
promote epitaxial growth of ZrB
2
lms of high crystal qualities as
illustrated by HREM images presented by Tolle et al. [152] for a ZrB
2
lm deposited on Si(111) substrate at 900 C showing a good registry
and a sharp interface between the substrate and the lm as well as the
in-situ reection high-energy electron diffraction patterns for a ZrB
2
lm
deposited on Al
2
O
3
(0001) at 1000 C in a study by Bera et al. from 2009
[155]. The process conditions required for CVD growth of epitaxial ZrB
2
lms, in particular the choice of deposition temperature and the
ultra-high vacuum (UHV) deposition conditions, inspired the develop-
ment of DCMS of epitaxial ZrB
2
lms with growth from a ZrB
2
compound target and on the substrates Si(111) [156], 4HSiC(0001)
[156] [157], and Al
2
O
3
(0001) [158] as well as Si(001) substrates [159].
A limitation in epitaxial growth of ZrB
2
from Zr(BH
4
)
4
is low depo-
sition rates with 0.3 nm/min on Al
2
O
3
(0001) [155] and 1.2 nm/min on
Si(111) practically limiting lm thicknesses to typically <100 nm.
Higher growth rates result in the growth of amorphous lms on Si(111)
[152]. This may be explained by difculties at higher growth rates to
remove excess B from the Zr(BH
4
)
4
precursor in the growth zone by
formation and desorption of volatile B
2
H
6
molecules, resulting in the
growth of B-rich lms. This is exemplied by the study by Sung et al.
[151] demonstrating growth of lms with B/Zr ratios close to 3 in the
temperature range 250450 C [151].
In the CVD community, there is an active development of tailored
precursors for growth of specic materials systems including TMB
2
thin
lms preferably at low temperatures. Precursor development, however,
is much less explored in the PVD community. It has been shown that
reactive magnetron sputtering of tungsten target in krypton/trime-
thylboron (B(CH
3
)
3
) plasmas results in growth of W-rich 100-oriented
WC
1-x
with a potential boron solid solution [160]. Moreover, there is
an interesting process development in hybrid sputtering using typical
CVD precursors such as B
2
H
6
, pentaborane(9) (B
5
H
9
) decaborane(14)
(B
10
H
14
) as further discussed in section 7. We anticipate that a similar
approach can be developed for TMB
2
thin lms.
4.2. Cathodic arc deposition
The high melting points of TMB
2
, 3000 C generates evaporation
characteristics like metals with high melting points such as W and Ta i.e.,
the generation of solid glowing macroparticles. In 1991, Knotek et al.
[161] studied DC arc deposition of TiB
2
coatings from high-pressure
sintered TiB
2
cathodes manufactured from TiB
2
powder with small ad-
ditions <1 wt % of Al and Ni as well as C and B in Ar and Ar/N
2
plasmas.
The focus for the study was on improving the evaporation properties of
ceramic cathodes by investigating cathode materials of different purities
>99% and densities between 98.7% and 99.2% of the theoretical values,
resulting in specic electrical resistivities and thermal conductivities
between 8.1 to 25.4 and 15.640 W/m K at 20 C, respectively. A
deposition process was found. [161] that was less dependent on the
properties of the cathode material, but with emission of glowing solid
macroparticles and problems with controlling the movement of the
arc-spot on the cathode as it often remained at a single location
(immobilization) causing local overheating i.e., prevailing problems in
arc evaporation from ceramic cathodes. Despite these difculties,
0001-oriented hexagonal TiB
2
coatings with a HV 0.05 up to 4250 were
deposited on cemented carbide and high-speed steel substrates with
deposition rates of 24
μ
m/h, but with no information provided on the
composition of the coatings. The addition of N
2
during arc deposition
appeared to stabilize the arc-spot movement, but at the expense of ni-
trogen uptake to form TiBN coatings with reduced hardness. Here,
Knotek et al. noted even the formation of crystalline TiN indicating a
higher afnity of the metal Ti towards N compared to B in arc deposi-
tion. This is an important property of Ti that comes into play for this
metal as well as the other Group 46 TM when TMB
2
is co-sputtered with
N
2
.
The pioneering work by Knotek et al. [161] inspired other re-
searchers to further improve the properties of arc deposited TMB
2
coatings. In a publication from 1993, Tregilo et al. [162] investigated a
pulsed arc deposition of a TiB
2
cathode in vacuum. From pulsing of the
arc and by applying pulsed high voltage bias to the substrate, the authors
sought to minimize the emission of glowing solid macroparticles and the
problems with controlling the movement of the arc-spot previously
described in Ref. [161] The results showed that thick up to 10
μ
m,
adherent and hard 2600 to 3300 (HV 0.025) coatings with compressive
stress of 2.3 GPa were deposited on stainless steel substrates without
external heating and using pulsed high voltage bias to the substrate. The
authors provided no information on composition and the structural
M. Magnuson et al.
Vacuum 196 (2022) 110567
13
properties of their lms to support that TiB
2
with C32 crystal structure
had been deposited.
In a publication from 2015, Zhirkov et al. [163] investigated the
inuence of an external magnetic eld for controlling the movement of
the arc-spot on the cathode. From DC arc deposition of a TiB
2
cathode in
vacuum, the authors showed that deposition without the external
magnetic eld stabilizes the conditions for the arc on the cathode. This
enabled growth of smooth TiB coatings on MgO substrates with an
approximate 1:1 Ti to B ratio as determined from quantitative analysis
by XPS. The discrepancy between the composition of the cathode and
the deposited lm was explained by the spatial plasma distribution on
the ion mass. The composition of lms sputtered from TMB
2
sources is
further discussed in section 5. To deposit TMB
2
as a next generation of
hard coatings should be of benet for progress in this eld.
4.3. E-beam evaporation and pulsed-laser deposition
In a classical work from 1978, carried out under the aegis of the U.S.-
U.S.S.R. Science and Technology Cooperative Agreement, Bunshah et al.
[164] used electron-beam evaporation to deposit TiB
2
and ZrB
2
coatings
from sintered evaporation billets manufactured from TiB
2
and ZrB
2
powders. Growth was carried out on polycrystalline Mo sheets at
deposition rates ranging from 0.113
μ
m/min to 6.3
μ
m/min and tem-
peratures from 600 to 1300 C. Crystalline TiB
2
and ZrB
2
coatings were
deposited, but with TiB
2
and TiB phase mixtures at a deposition rate of
6.3
μ
m/min, and where quantitative analysis of the phases by XRD
showed TiB
2
-rich coatings at 690 C910 C and TiB-rich coatings at
higher temperatures. Due to difculties in the quantication of the
amount of B, there was no information on the composition of the coat-
ings in, e.g., Ref. [164] For deposited carbide coatings, the authors used
the lattice parameter to determine the carbon-to-metal ratio in the
coating, which is not possible for compounds with narrow homogeneity
ranges such as TiB
2
and ZrB
2
. Bunshah et al. communicated lattice pa-
rameters for the deposited TiB
2
and ZrB
2
coatings in an earlier publi-
cation [165]. The general trend for both borides is an increasing length
for the c axis with increasing growth temperature, while the length of
the a axis remained relatively unchanged and close to that determined
for TiB
2
with 3.03 Å and 3.17 Å for ZrB
2
, see Table 2. The TiB
2
coatings
were predominantly 0001-oriented while the orientation of the ZrB
2
depended on the deposition rate as well as temperature with 0001-ori-
ented coatings grown at a deposition rate of 2.14
μ
m/min and in the
range 670 C850 C. Lower deposition rate of 0.113
μ
m/min resulted in
1010-oriented coatings at temperatures from 600 C to 1300 C. The
observations made in Refs. [164,165] on lm orientation at different
deposition rates and temperatures are valuable when discussing the
results obtained from sputtered deposited lms.
Bunshah et al. [164,165], investigated the microhardness (micro--
Vickers hardness under 50 g for 10 s) of the deposited TiB
2
and ZrB
2
coatings, where the general trend was increasing hardness values from
~2200 to ~3200 kg/mm
2
for TiB
2
coatings at increasing substrate
temperatures, while the ZrB
2
coatings showed an opposite trend seen
from ~3200 to ~1900 kg/mm
2
. The hardness trend for 0001-oriented
TiB
2
coatings is clear as higher temperatures should result in larger
crystals, while the trend for coatings with a phase mixture was more
unclear as TiB
2
due to the higher B content should exhibit a higher
hardness compared to TiB. Possibly, there is a connection to stresses in
the coatings given the much higher deposition rate of 6.3
μ
m/min
applied during deposition. The trend for ZrB
2
can probably be explained
by the change in lm orientation from 0001 to 1010, which is further
discussed in section 5.3 and supported from nanoindentation measure-
ments on sputtered ZrB
2
and TiB
2
lms.
In 1997, Zergioti and Haidemenopoulos [166] reported PLD of a TiB
2
source for growth of TiB
2
lms on Si(001) at 600 C. Analysis of an
~50 nm thick lm by transmission electron microscopy (TEM) at the
zone axis <110>showed that the phase TiB
2
was deposited as columns
between 10 and 50 nm and with an epitaxial relationship to the sub-
strate i.e., similar to epitaxial ZrB
2
lms deposited by DCMS, see section
4.1.
The study in Ref. [166] on epitaxial growth of TiB
2
was followed by
Zhai et al. [167] demonstrating epitaxial TiB
2
growth on Al
2
O
3
(0001)
and SiC(0001) at 600 C and Ferrando et al. [168] with epitaxial growth
of TiB
2
on Al
2
O
3
(0001) at 720 C. Later epitaxial growth of ZrB
2
and
ScB
2
lms were demonstrated on Al
2
O
3
(0001), SiC(0001), and Si(111),
albeit at slightly higher substrate temperatures of 900 and 950 C [169].
The primary motivation for epitaxial growth of TMB
2
by PLD thus far
has been on deposit thin buffer layers <10 nm to serve as templates for
growth of superconducting MgB
2
layers given the higher stability of
TMB
2
compared to MgB
2
at elevated temperatures. While PLD appears
to be a promising technique for epitaxial growth of TMB
2
, difculties
with upscaling and low deposition rates as well as being an expensive
technique limits PLD in growth of TMB
2
coatings under industrial
conditions.
4.4. Reactive sputter deposition
Growth experiments of thin-lm borides by sputtering methods is
legio and has predominantly been conducted by non-reactive sputtering
and using compound targets, see section 4.5. There are several dif-
culties in sputter deposition of TMB
2
lms using compound sources that
are presented below in section 4.5 and 5.15.6. Comparably fewer
studies exist on reactive sputtering of TiB
2
that makes it possible to
deposit lms with different compositions. In fact, there are only two
studies from 1989 using reactive sputtering of TiB
2
, using B
2
H
6
diluted
to 6% in argon and rf diode sputtering of a Ti target by Larsson et al.
[170] and Blom et al. [171]. For lms deposited without external
heating on vitreous carbon, oxidized silicon, and Si(001) substrates,
Auger electron spectroscopy and Rutherford backscattering spectros-
copy (RBS) (carbon substrates) showed that lms with a 2:1 B to Ti ratio
can be reactively sputtered. XRD patterns recorded from lms deposited
on oxidized silicon, and Si(001) substrates showed peaks from TiB
2
for
samples heat treated for 30 min in a vacuum furnace between 400 and
900 C [171]. From their experimental set-up with rf diode sputtering,
the authors reported substantial target poisoning, resulting in sputtering
of B at low rates [170]. However, the same group showed that there is a
simple solution; known as bias sputtering [172].
Although promising, diborane is problematic as precursor since it is
both explosive and highly toxic. To handle this gas in a laboratory or an
industrial environment requires rigorous safety requirements. Such
routines are available in the semiconductor industry as diborane is the
preferred B-source for doping Si, but comes at a high cost. Thus, it seems
challenging to integrate diborane in an industrial sputtering process, but
not impossible. In addition to diborane, halide BF
3
has been investigated
for reactive sputtering of a TiB
2
target by rf sputtering [173]. Coatings
deposited on NaCl and Si substrates exhibited a 2B:1Ti ratio, and had to
be stored in a desiccator to avoid a chemical reaction with the moisture
in atmosphere, otherwise resulting in etching by HF on the surface of the
coating. Combined PVD deposition methods by growth of individual
targets RF sputtering on B targets and DC sputtering of metal has also
been shown to be effective, as discussed later.
To summarize, reactive sputtering is a promising tool for growth of
compounds such as TMB
2
. In addition to B
2
H
6
there are only a few
studies on the use of precursors in the sputtering process. Pentaborane
(9) (B
5
H
9
) and decaborane (14) (B
10
H
14
) should be attempted as they
are less poisonous than diborane, see further discussion in section 7,
Outlook.
4.5. Sputtering from compound sources
Growth from compound sources, sintered TMB
2
bodies as well as TM
and B compacts, is the predominant synthesis route for sputter deposited
TMB
2
lms with C32 crystal structure as initiated by Wheeler and
M. Magnuson et al.
Vacuum 196 (2022) 110567
14
Brainard [43] and summarized by Mitterer [44], but with a few studies
where TMB
2
lm growth have been studied by growth from elemental
sources as discussed in this section. A survey of the literature shows that
all Group 46 TMB
2
have been synthesized as by this technique. The
most investigated TiB
2
, ZrB
2
, and CrB
2
materials systems are treated as
historical markers for research on sputter deposited TMB
2
lms, see also
sections 5.3 and 5.4. In this section, we discuss important work in the
eld with a focus on the less investigated materials systems HfB
2
, VB
2
,
NbB
2
, TaB
2
, MoB
2
, and WB
2
.
In 1978, Wheeler and Brainard [43] investigated 13.56 MHz rf diode
sputtering of CrB
2
, MoSi
2
, Mo
2
C, TiC, and MoS
2
hot-pressed and
disk-shaped compacts, 15.2 cm in diameter. For CrB
2
lms deposited at
a sputtering power of 600 W and a bias voltage of 500 V, and without
external heating, no data on the structural properties was provided,
while XPS analysis of the Cr 2p and B
1s
peaks revealed large amounts of
O in the form of Cr
2
O
3
and B
2
O
3
, respectively [43]. The authors
concluded that the origin for the O was the target material as noted in
their publication: No other target shows as great a propensity to outgas as
did the CrB
2
. The properties of compound targets applied for sputtering
of Group 46 TMB
2
lms will be further discussed in section 5.1. In
1981, Padmanabhan and Sørensen [173] applied rf sputtering of
13.6 MHz of a TiB
2
compound target (99.95%) at a sputtering power of
500 W for growth of lms on silicon substrates. No information from
XRD or selected area electron diffraction (SAED) patterns on the struc-
tural properties of the lms was communicated, but where TEM images
showed the lms to be relatively smooth and RBS measurements gave a
B-decient composition in the lms with a Ti/B stoichiometric ratio of
0.66, i.e., B/Ti 1.5. In 1981, Yoshizawa et al. [174] published on lat-
tice constant measurement of 56
μ
m thick TiB
2
coatings sputtered on
Mo, Cu, graphite, Ti-coated Mo, and Al-coated Mo substrates at a sub-
strate temperature of about 200 C, using dc cylindrical post-cathode
magnetron sputtering sources. From SAED patterns, the authors
concluded that polycrystalline TiB
2
coatings with C32 crystal structure
had been deposited and where TEM and XRD showed ~5 nm ne grain
sizes. In 1983 Shappirio and Finnegan [175] applied rf diode sputtering
of 5 inch in diameter TiB
2
and ZrB
2
hot-pressed commercial compound
targets with measured densities of 80% and 92% of the bulk values,
respectively. TiB
2
and ZrB
2
lms up to 800 nm thick were deposited
without external substrate heating for growth on Si, thermal SiO
2
on Si
and, Be substrates for interconnect metallization. An XRD pattern
recorded from a ZrB
2
lm revealed broad 1011, 1010, and 1012 peaks of
low intensities and where the peak distribution indicated randomly
oriented lms. Growth of nanocrystalline lms were determined from
TEM micrographs showing an average grain size of 20 nm i.e., slightly
larger TiB
2
grains than previously observed by Yoshizawa et al. [174]
Furthermore, quantitative analysis with AES showed a B-decient
composition for both TMB
2
and with 6 wt% O in ZrB
2
and 12 wt% O in
TiB
2
, and where the O content was supported, but not the B/TM ratio
from RBS of the lms deposited on Be substrates.
Shappirio et al. [175,176] evaluated the resistivities of the TiB
2
and
ZrB
2
lms with measured values of 500
μ
Ωcm and 250
μ
Ωcm, respec-
tively, for as-deposited lms. However, the inuence from the substrate
on the resistivity was unclear. To decrease resistivity values in the TiB
2
and ZrB
2
lms, annealing by halogen lamp (rapid anneal) was per-
formed at 1050 C with the samples heated in owing argon. The
treatment was successful since the resistivity values were reduced to
150
μ
Ω cm for TiB
2
and 75
μ
Ω cm for ZrB
2
. Recrystallization of ZrB
2
in
the lms was revealed by x-ray diffractograms with sharp 1010, 1011,
and 1012 peaks from ZrB
2
by Shapiro and Finnegan [175]. Similar to
Wheeler and Brainard [43], Shappirio et al. [175,176] attributed the
impurity and density properties of the target material to the less favor-
able electrical properties of the deposited TiB
2
lms compared to ZrB
2
lms, where the TiB
2
sputtering source is said to exhibit a higher
porosity that precludes elimination of contaminants even after extensive
pre-sputter target cleaning [175]. The somewhat precarious condition of
available TiB
2
targets has steered studies on TMB
2
as contact materials
on semiconductors towards ZrB
2
. Porous targets imply trapped gas and
crystalline oxides in pure Zr sputtering target, and in the case of
contaminated targets, the base pressure is less important for the sput-
tering process. Contaminants such as H
2
O (97%) in the residual gas play
a role predominantly at low temperature growth without external
heating, while at higher temperature the H
2
O and other contaminants, C
and O, desorb from the growth surface. In several applications of
low-temperature growth, the lms strive to become two-phase systems.
The use of borides as contact materials is elaborated on in section 5.4 in
relation to the resistivities of TMB
2
lms. However, the high sensitivity
of a boride growth surface to O and OH was already observed by Mit-
terer [44] for growth of TiB
2
and ZrB
2
lms. Target contamination is
further developed in section 5.2.
In 1997, Mitterer [44] summarized work on depositing TiB
2
and ZrB
2
lms with C32 crystal structure by sputtering in terms of stability and
orientation, crystallinity as a function temperature as further developed
in section 5.2. Recently, epitaxial growth of ZrB
2
has been demonstrated
on Si(111) [156], 4HSiC(0001) [156] [157], Al
2
O
3
(0001) [158], and Si
(001) [159] with the substrates held at 900 C. The orientational re-
lationships are ZrB
2
(0001)//Si(111) out-of-plane with two in-plane
domains: ZrB
2
[1120]//Si[101] for the majority orientation and
ZrB
2
[1120]//Si [112] for the minority orientation [156].
ZrB
2
(0001)//4HSiC(0001) lms have the ZrB
2
[1100]//4HSiC[1100]
in-plane relationship [157]. For ZrB
2
(0001)//Al
2
O
3
(0001) layers, the
two equally probable in-plane domains: ZrB
2
[1010]//Al
2
O
3
[1010] and
ZrB
2
[1120]//Al
2
O
3
[1010] were observed [157]. For growth on Si(001)
there are two in-plane orientations where ZrB
2
[1010] grows
out-of-plane.
Beyond TiB
2
and ZrB
2
, CrB
2
has attracted interest as a thin lm
material, due to good corrosion resistance from the metal Cr. The pre-
viously described work by Wheeler et al. [43] published in 1978 reveled
rf sputtered lms with high content of O as determined from XPS and the
OB and OCr bonds present in the lms. In two publications from 1998,
Oder et al. [177,178] used dc sputtering of a CrB
2
target with unspeci-
ed properties in a deposition system held at a base pressure 79 x
10
7
Torr to deposit 100200 nm lms as Ohmic contacts. Similar to
Wheeler and Brainard [43] no data on the structural properties of the
lms were communicated, but with a disturbing O contribution
observed in the RBS data for as-deposited lms.
In 2006, Audronis [179] showed the effect of pulsed magnetron
sputtering on the structure and mechanical properties of CrB
2
coatings.
The substrate temperature was in the range of 110150 C during the
deposition and since CrB
2
is a line-compound, the microstructure is
probably sensitive to minor deviations from the stoichiometric compo-
sition. TEM analysis revealed two very different types of microstructures
in the coatings; coatings with negatively biased substrate were signi-
cantly affected by high-energy ion bombardment, while coatings were
not affected as the substrate was allowed to oat; coatings with negative
bias contained single crystal nanocolumns of CrB
2
with (0001) planes
oriented along the growth direction with random rotation along the axis.
On the contrary, coatings with oating bias had a randomly oriented
polycrystalline microstructure with 35 nm crystal size with a mixture of
grain shapes with a majority of different columnar shapes. Recently,
Dorri et al. [180] achieved epitaxial growth of close-to-stoichiometric
CrB
2
lms sputter-deposited from a CrB
2
target onto Al
2
O
3
with B-rich
inclusions at overstoichometric composition.
Following the observations by Wheeler and Brainard [43] on the
porosity properties of Cr targets seems to have steered most of the early
work on CrB
2
to pulsed sputtering techniques. Pulsed deposition tech-
niques such as HiPiMS are further developed in section 4.6. As a com-
plement to Ti and Zr, Oder et al. [177] studied CrB
2
using DCMS as
discussed in section 5.3.
HfB
2
is less studied for thin lm TMB
2
in Group 4. In 1988, Lee et al.
[181] deposited HfB
2
lms by rf magnetron co-sputtering from Hf and B
M. Magnuson et al.
Vacuum 196 (2022) 110567