NbSi2 coating on niobium using molten salt

Article abstract of DOI:10.1016/S0925-8388(01)01879-5

For obtaining better oxidation resistance of niobium in air, a niobium silicide was non-electrolytically deposited onto niobium from the molten salt, where a disproportional reaction occurs between Na₂SiF₆, SiO₂, and Si. A single phase of NbSi₂ was formed with a homogeneous thickness of about 10 μm above 1073 K. The oxidation resistance of pure niobium with this coating layer was improved. During the oxidation at the high temperatures, Nb₅Si₃ was formed at the interface between the Nb substrate and the NbSi₂ layer. Due to this intermediate layer formation, the oxidation resistance became better than for pure NbSi₂.

Full text of DOI:10.1016/S0925-8388(01)01879-5

Journal of Alloys and Compounds 336 (2002) 280–285
L
www.elsevier.com/locate/jallcom
NbSi2 coating on niobium using molten salt
1
*
Ryosuke O. Suzuki , Masayori Ishikawa , Katsutoshi Ono
Department of Energy Science and Technology, Kyoto University, Yoshida-Honmachi, Sakyo-ku, Kyoto 606-8501, Japan
Received 3 September 2001; accepted 24 September 2001
Abstract
For obtaining better oxidation resistance of niobium in air, a niobium silicide was non-electrolytically deposited onto niobium from the
molten salt, where a disproportional reaction occurs between Na2SiF6, SiO2, and Si. A single phase of NbSi2 was formed with a
homogeneous thickness of about 10 mm above 1073 K. The oxidation resistance of pure niobium with this coating layer was improved.
During the oxidation at the high temperatures, Nb5Si3 was formed at the interface between the Nb substrate and the NbSi2 layer. Due to
this intermediate layer formation, the oxidation resistance became better than for pure NbSi2.
reserved.
2002 Elsevier Science B.V. All rights
Keywords: Transition metal compounds; Coating materials; Corrosion
1. Introduction
method if the Nb sample could be dipped in the liquid
silicon. The lowest melting temperature in the Nb–Si
system is as high as 1668 K [7], so that it is practically
difficult to deposit directly niobium silicide from the Si
melt. Coatings of Cr–Si or Ti–Si binary silicide on the Nb
substrate, therefore, were applied using their lower eutectic
melting [8]. Other methods such as vacuum deposition, or
chemical vapor deposition have been applied to deposit
silicon onto the Nb substrate. These attempts need elevated
temperatures of over 1273 K and prolonged times for
silicon deposition and diffusion.
The pack-cementation method has been extensively
studied as low cost, good adhesion coating [4,5,9–13]. It is
one of the chemical vapor deposition techniques and, as
shown in Fig. 1a, it uses three kinds of materials: solid Si
or a silicon compound such as SiC as silicon source, an
alkali halide such as NaF to form a volatile silicon halide
gases such as SiFx, and an inert filler such as Al2O3. The
deposit on Nb was reported to be NbSi2 or the multi-
layered niobium silicides, depending on the constituents of
the substrate alloy and the source materials. Because this
process is based on the deposition from a halide gas, it
requires encapsulation and a high vapor pressure at tem-
peratures as high as 1273–1473 K.
Niobium is one of the promising metals for aerospace
applications, because its sublimation is significantly low
even at temperatures higher than 2300 K in vacuum [1].
However, it is easily oxidized even at 600 K in air, called
pesting, and its usage in hot air has been severely
restricted. This is because the niobium oxides are not
oxidation protective. Many intermetallic compounds based
on niobium have been developed for high temperature use,
and it is claimed that Al2O3 or SiO2 layers on the surface
of the metallic aluminide or silicide can protect oxygen
penetration [2–5].
This work aimed at depositing niobium silicide on a
niobium substrate, and it was expected that the silicide
form a protective SiO2 layer. Pure niobium was taken as
substrate to demonstrate the feasibility of silicide coating
technique. The coating should be tightly adherent and
uniformly protect the substance.
Kumon et al. applied hot dipping of a Nb substrate into
molten Al and subsequent anodic oxidation of the Al films
[6]. This dip coating seems to be the simplest coating
*
Corresponding author. Tel.: 181-75-753-5453; fax: 181-75-753-
Another coating method using molten salt [14–19] may
be applicable to the coating of refractory metals. It can be
used to deposit the iron silicide on pure iron and mild steel
[15–18]. This is an electroless plating operable in open air
at 973–1173 K, lower than the pack cementation process-
4745.
E-mail address: suzuki@energy.kyoto-u.ac.jp (R.O. Suzuki).
1Present address: Research Institute for Radiation Biology and Medi-
cine, Hiroshima University, 1-2-3 Kasumi, Minami-ku, Hiroshima 734-
8553, Japan.
0925-8388/02/$ – see front matter
PII: S0925-8388(01)01879-5
2002 Elsevier Science B.V. All rights reserved.

R.O. Suzuki et al. / Journal of Alloys and Compounds 336 (2002) 280–285
281
Fig. 1. Illustrations for niobium silicide coatings.
ing temperature. As shown in Fig. 1b, silicon deposits on
the metal surface in the salt react with the substrate metal
and form the metal silicide. It was successively applied to
other refractory metals such as nickel [14] or molybdenum
[19].
coating onto the pure niobium surface using molten salt,
and to check the fundamental applicability of this electro-
less plating technique to Nb-based alloys.
The principle of this deposition in the molten salt has
been reported in Refs. [14–19]. When Na2SiF6 and Si
powder are added in the supporting salt of NaCl–KCl, a
disproportional reaction between Si and Si41 ions deposits
a siliconized layer on the metallic substrates, M, as,
2. Experimental
The purity of the material used for the molten salt, the
experimental apparatus, and the handling procedures for
coating were identical with those described previously
[17–19]. A brief outline will be summarized here. The salt
was composed of 36.58 mol% NaCl – 36.58 mol% KCl –
21.95 mol% NaF – 4.89 mol% Na2SiF6. About 150–200 g
salt mixture and silicon powder (21.85 mol% for salt
mixture) was filled into the Al2O3 crucible, and melted at
about 900 K. A transparent SiO2 bar was brought into the
salt during coating in order to add Si41 to the salt as an
alternative source [17].
A sheet of pure niobium (99.85% in mass content) was
cold-rolled to 3-mm-thick, and cut into pieces of 2537
mm. The sample surface was cleaned chemically using an
aqueous solution of HF and HNO3. Four samples were
simultaneously suspended by molybdenum wires, and
immersed into the molten salt in open air. The samples
were removed from the molten salt and cooled rapidly in
air. The adhering salt was easily removed in water. After
drying, the samples was weighed and examined by optical
microscopy and in a scanning electron microscope (SEM)
equipped with an energy dispersive X-ray (EDX) analyzer.
The phases on the sample surface were identified by X-ray
diffraction (XRD). The oxidation resistance was measured
by thermogravimetry (TG), in which the sample was
heated in air at a constant rate of 2 K/min. Because the
niobium samples reacted with the platinum wire during
heating, a short Al2O3 tube was inserted into a hole
opened in the sample, and Pt wire was passed through this
tube.
Si (source particles in the molten salt) 1 6 F2
1 SiF262 → 2 SiF462
(1)
2SiF462 1 M (substrate) → M–Si (siliconized layer)
1 6 F2 1 SiF262
(2)
The deposited Si from Si21 penetrates into the metal
surface, and forms an alloy (M–Si) with the metal
substrate. The total reaction,
Si (particle in the molten salt) 1 M (substrate) →
M–Si (siliconized layer),
(3)
shows that the reaction is driven by the difference between
the thermodynamic activity of pure silicon and that of the
siliconized layer. In this thermodynamic sense of the
disproportional reaction, as illustrated in Fig. 1, the
deposition reactions from the silicon sources are fun-
damentally the same both for the pack cementation in the
gaseous state (Fig. 1a) and for the electroless plating in the
molten salt (Fig. 1b). The former uses gaseous species as
silicon carrier, and the latter the ionic SiF462. The amount
of silicon atoms in the media, however, is much larger in
the molten salt than in the gaseous phase. It can enhance
silicide formation even at lower temperatures, if the rate
determining step is not diffusion in the solid.
For comparison, an NbSi2 ingot was prepared using an
The purpose of this work is to propose niobium silicide

282
R.O. Suzuki et al. / Journal of Alloys and Compounds 336 (2002) 280–285
Ar arc furnace and the ingot was cut into pieces for
oxidation measurement.
layers [17,18]. This was due to extremely faster formation
of the Fe3Si layers.
Four niobium silicides, Nb3Si, a-Nb5Si3, b-Nb5Si3 and
NbSi2, were known in the binary system [7]. Although
only the two silicides, a-Nb5Si3 and NbSi2, are stable at
the studied temperatures, the formation of a-Nb5Si3 was
never detected. Note that the formation of NbSi2 was
reported in the pack cementation [10,11]. Similarly the
highest silicide, MoSi2, was deposited on Mo substrate
using the same salt [19], while the lowest silicide, Fe3Si,
was deposited on iron from the same salt [17,18], and the
intermediate silicides were reported on a pure Ni substrate
[14]. In Fe–Si binary system, however, it could be
kinetically argued that the deposited silicon formed the
Fe3Si phase preferentially, because both iron and silicon
diffusion in Fe3Si were the fastest [17,18]. Although the
diffusion coefficients in the Nb–Si system were not
analyzed, the preferential formation of NbSi2 might be
related to the slower diffusion in the a-Nb5Si3 phase.
Fig. 3 shows the mass gain per sample surface area due
to silicide formation. For the evaluation, account was taken
of the surface area of the hole used for the suspension. The
growth rate of the NbSi2 layer at 1173 K was as high as
that of the MoSi2 layer [19], while it decreased at the
lower temperatures. The NbSi2 layer grew quite slowly at
973 K, and its layer thickness was not uniform at 973 K
and 1073 K. The layer became fairly flat at 1173 K, as
shown in Fig. 2. The thickness evaluated from Fig. 3 and
the density of NbSi2 was, for example, about 18.7 mm for
the sample obtained at 1173 K for 7.2 ks, while Fig. 2
shows about 15 mm. Although the behavior of mass gain at
1173 K in Fig. 3 is parabolic, a precise kinetic analysis
should be based on flat homogenous growth.
3. Results and discussion
3.1. Formation of NbSi2 layer
The layer formed on the Nb surface showed the metallic
luster. The XRD patterns from the sample surface were
identified as single phase of hexagonal NbSi2. Although
the coating was conducted in open air, no niobium oxides
were detected by XRD. Fig. 2 shows the cross-sectional
SEM view of the layer and the corresponding EDX profiles
for Nb and Si. The coated layer was flat and it had a sharp
interface with the substrate. The Si concentration in the
layer analyzed by EDX was constant as 6761.5 mol% Si,
which is consistent with stoichiometric NbSi2. Neither
silicon powder nor salt components were observed in the
NbSi2 layer, although they were frequently found in Fe3Si
Fig. 2. SEM image of the cross-section for the Nb sample siliconized at
1173 K for 7.2 ks. Nb and Si concentration were analyzed along the white
line in the SEM image.
Fig. 3. Mass gain of the siliconized Nb sample as a function of immersed
time.

R.O. Suzuki et al. / Journal of Alloys and Compounds 336 (2002) 280–285
283
In the case of MoSi2 formation on a Mo substrate, an
3.2. Oxidation
irregular growth at the sample edges was reported for
MoSi2 coating [19]. However, such abnormal growth was
not found in the NbSi2 coating. This means that the Nb
and Si supply through the thin layer was well balanced
under our experimental conditions. For practical applica-
tions, a thicker layer might be required. Note a Fe3Si layer
as thick as 200 mm was successfully formed using
identical conditions [17]. This means that the Si supply
from the molten salt was high enough, but that the
diffusion through the solid layer was slower in the case of
the Nb substrate.
The thermogravimetric (TG) analysis of pure Nb sample
in air is shown in Fig. 4. The mass of the pure Nb sample
increased from 658 K due to the formation of Nb2O5. This
mass gain is characteristic for Nb because the formed
Nb2O5 is not oxidation protective. When the deposition
time was 3.6 ks at 1173 K, the pesting oxidation occurred
at 750 K, because NbSi2 could not cover the whole
surface. NbSi2 coating over 10.8 ks at 1173 K could
suppress the oxidation. Therefore, the silicide coating on
niobium could form a good oxidation protective layer
when the NbSi2 layer was thick enough.
A tiny amount of molten salt was observed to have
evaporated and to be deposited at cooler portions of the
apparatus. XRD analysis identified it with a mixture of
NaCl, KCl and SiCl4. In a previous paper the possibility
was mentioned that Si41 ions necessary for the dispropor-
tional reaction might be lost because of the high vapor
pressure of SiF4 or SiCl4 [17,19]. Because the loss of Si41
ions delays the deposition rate, the evaporation of SiCl4
should be minimized in our process, for example, by
decreasing the operation temperature to below 1173 K. In
the pack cementation method, however, the gaseous SiCl4
and SiF4 are the main carriers of silicon. Their vapor
pressures should be kept high by heating to higher
temperatures, and the encapsulation is needed to avoid gas
leakage (Fig. 1a). The molten salt containing Si41 ion
plays the same role to carry Si onto the substrate, however,
the ions in the salt reacted at 200 K lower than the gaseous
species in the pack cementation method. The coating in the
molten salt does not need the operation in a closed system,
but it can operate even in open air.
The kinetics of isothermal oxidation of a silicide layer rf
sputtered on the SiO2 substrate revealed that the oxidation
in air between 823 and 1123 K did not obey a parabolic
law [3]. It was concluded that the silicide layer was not
oxidation protective, but it may be due to thin thickness of
the film, 0.4 mm. For comparison, the oxidation behavior
of the arc-melted NbSi2 was measured as shown in Fig. 4.
Fig. 5. SEM image of the cross-section for the Nb sample siliconized at
1173 K for 7.2 ks. The sample was heated up to 1523 K in air. Nb and Si
concentration were analyzed along the white line in SEM image.
Fig. 4. Mass change for Nb plate, the arc melted NbSi2 and the coated
samples. They were heated at the rate of 0.033 K/s in air.

284
R.O. Suzuki et al. / Journal of Alloys and Compounds 336 (2002) 280–285
surface due to the existence of Nb2O5 crystals, the
deposited NbSi2 layer may be subsequently oxidized above
1100 K. The morphology control of these SiO2 crystals is
a key to improve the oxidation resistance.
4. Conclusion
Niobium silicide could be deposited at 1073–1173 K on
pure niobium substrate, using the disproportional reaction
between Si41 and the Si powder in the molten salt
composed of NaCl–KCl–NaF–Na2SiF6 –SiO2. Single
phase NbSi2 was formed with the thickness of about 10
mm at 1173 K. The growth rate of the NbSi2 layer was
quite low below 1073 K. A protective layer on the
siliconized sample was effective for avoiding pest oxida-
tion that usually occurs for pure Nb at low temperatures.
During the oxidation at the high temperature, Nb5Si3 was
formed at the interface between the Nb substrate and
NbSi2 layer.
Acknowledgements
The authors thank Mr T. Nishibata for his advice
throughout the experiments, Mr T. Unesaki and Mr I.
Nakagawa for SEM–EDX analysis, and Mr M. Hamura for
XRD measurements. This study was supported in part by
Grants-in-Aid for Scientific Research, under Contract No.
10555256.
Fig. 6. SEM image of the surface for the Nb sample siliconized at 1173
K for 18 ks. The sample was heated up to 1523 K in air.
The SiO2 layer generated from pure NbSi2 could not
protect the air oxidation above 1000 K in our experimental
conditions. The mass gains of the samples coated for
longer than 10.8 ks, for example, were less than that of the
single phase NbSi2. There exists another mechanism in the
oxidation protection for our coating.
References
[1] A.V. Byalobzheskii, M.S. Tsirlin, B.I. Krasilov, High-Temperature
Corrosion and Protection of Superrefractory Metals, Atomizdat,
Moscow, 1977.
Figs. 5 and 6 show the cross-sectional and surface view
for the Nb samples which were coated and subsequently
oxidized in air up to 1523 K using TG. The coated layer
split into three layers. The silicon concentrations of the
inner two layers were 67.2 and 37.8 mol% by EDX
quantitative analysis. They can be identified as NbSi2 and
Nb5Si3, respectively. The top surface was identified by
XRD as a mixture of SiO2 and Nb2O5. The rod-like
crystals shown in Fig. 6 were SiO2 as verified by EDX.
The formation of the lower silicide, Mo5Si3, was also
reported on a MoSi2 coating [19]. It is due to mutual
diffusion between the substrate and the higher silicides.
The inner layer of Nb5Si3 may prevent crack initiation
and propagation in the layer or at the interface between the
SiO2 protective layer and the matrix. It is known that the
adhesive strength between two silicides or between Nb and
Nb5Si3 is stronger than that between the oxide and Nb
[20]. The formation of Nb5Si3 should be one of the
reasons that the coated niobium was better than single
phase NbSi2. Because SiO2 could not cover the whole
[2] C.S. Giggins, F.S. Pettit, J. Electrochem. Soc. 118 (1971) 1782.
[3] T.P. Chow, K. Hamzeh, A.J. Steckl, J. Appl. Phys. 54 (5) (1983)
2716.
[4] R. Bianco, M.A. Harper, R.A. Rapp, J. Miner., Metals & Mater.
Soc. (J.O.M.) 11 (1991) 68.
[5] Y. Li, W. Soboyejo, R.A. Rapp, Met. Mater. Trans. B 30B (3)
(1999) 495.
[6] T.F. Kumon, R.O. Suzuki, K. Ono, J. Jpn. Inst. Met. 59 (9) (1995)
967, (in Japanese).
[7] M.E. Schlesinger, H. Okamoto, A.B. Gokhale, R. Abbaschian, J.
Phase Equil. 14 (4) (1993) 502.
[8] S. Priceman, L. Sama, Electrochem. Tech. 6 (9–10) (1968) 315.
[9] L.L. Seigle, in: R. Kossowsky, S.C. Singhal (Eds.), Surface En-
gineering: Surface Modification of Materials, Martinus Nijhoff
Publishing, Dordrecht, The Netherlands, 1984, pp. 345–369.
[10] R.W. Barlett, R.P. Gage, Trans. Metall. Soc. AIME 233 (1965) 968.
[11] J. Guille, L. Matini, J. Mater. Sci. Lett. 7 (1988) 952.
[12] B.V. Cockeram, Surf. Coat. Techn. 76–77 (1995) 20.
[13] M.Vilsasi, M. Francois, R. Podor, J. Steinmetz, J. Alloys Comp. 264
(1998) 244.
[14] A.J. Gay, J. Quakernaat, J. Less-Common Met. 40 (1975) 21.
[15] T. Oki, J. Tanikawa, J. Met. Finish. Soc. (Kinzoku Hyohmen
Gijutsu) 31 (10) (1980) 561, (in Japanese).

R.O. Suzuki et al. / Journal of Alloys and Compounds 336 (2002) 280–285
285
[16] T. Oki, J. Tanikawa, Molten Salt (Yoyuen) 25 (2) (1982) 115, (in
[19] R.O. Suzuki, M. Ishikawa, K. Ono, J. Alloys Comp. 306 (2000)
Japanese).
285.
[17] R.O. Suzuki, T. Nishibata, K. Nakanishi, M. Ishikawa, K. Ono, Steel
Res. 71 (4) (2000) 130.
[20] L. Shaw, D. Miracle, R. Abbaschian, Acta Metall. Mater. 43 (12)
(1995) 4267.
[18] R.O. Suzuki, Y. Nagaso, K. Nakanishi, K. Ono, Steel Res. 71 (4)
(2000) 138.