A Microstructure and Mechanical Property Investigation on Thermally Sprayed Nanostructured Ceramic Coatings Before and After a Sintering Treatment

Abstract

Coatings have been deposited by air plasma spraying of alumina powders in the form of conventional particles
(C), nanostructured agglomerates (N) and sintered–nanostructured agglomerates (S). Sintering alleviated the
stresses introduced in the nanopowder by the manufacturing process (high energy ball milling). The coating
porosity is a direct consequence of the powder melting degree, which is related to the feedstock porosity.
The mechanical performance of the coatings is also closely associated with the powder melting degree. The
N coatings present the highest surface roughness due to the lowest melting degree. The slightly higher
hardness values of the N and S coatings, as compared to the C coatings, are attributed to the higher
percentages of α-Al2O3 and the presence of nanostructure. The S coatings exhibit superior adhesion
strength, relative fracture toughness and wear resistance, due to sintering consequences (intraparticle
cohesion, strain relief, tough splat boundaries), random dispersion of coherent nanozones and stress
dissipation at nanograin boundaries.

Full text

A microstructure and mechanical property investigation on thermally sprayed
nanostructured ceramic coatings before and after a sintering treatment
D. Zois
a,b
, A. Lekatou
b,
⁎, M. Vardavoulias
a
a
Pyrogenesis SA, Technological Park of Lavrion, 19500, Lavrion, Greece
b
Department of Materials Science and Engineering, University of Ioannina, 45111 Ioannina, Greece
abstractarticle info
Article history:
Received 13 January 2009
Accepted in revised form 7 June 2009
Available online 13 June 2009
Keywords:
Plasma spraying
Sintering
X-ray diffraction
Mechanical properties
Aluminium oxide coatings
Nanostructure
Coatings have been deposited by air plasma spraying of alumina powders in the form of conventional particles
(C), nanostructured agglomerates (N) and sintered–nanostructured agglomerates (S). Sintering alleviated the
stresses introduced in the nanopowder by the manufacturing process (high energy ball milling). The coating
porosity is a direct consequence of the powder melting degree, which is related to the feedstock porosity.
The mechanical performance of the coatings is also closely associated with the powder melting degree. The
N coatings present the highest surface roughness due to the lowest melting degree. The slightly higher
hardness values of the N and S coatings, as compared to the C coatings, are attributed to the higher
percentages of α-Al
2
O
3
and the presence of nanostructure. The S coatings exhibit superior adhesion
strength, relative fracture toughness and wear resistance, due to sintering consequences (intraparticle
cohesion, strain relief, tough splat boundaries), random dispersion of coherent nanozones and stress
dissipation at nanograin boundaries.
© 2009 Elsevier B.V. All rights reserved.
1. Introduction
Nanosized particles and grains are responsible for a substantial
increasein bulk material mechanical properties, such as tensile strength,
fracture toughness, hardness and wear resistance [1–4].
The novel fabrication techniques for nanopowders of spraying
applications are versatile and include a wide range of vapour, liquid
and solid state processing routes [5–9]. Of the existing methods,
mechanical milling is commonly used to produce large quantities of
nanopowders. During mechanical milling, raw powder particles of
several micron sizes are subjected to consecutive cold welding and
fracturing cycles, thus undergoing severe plastic deformation. The
impact force plastically deforms the powder particles, leading to work
hardening and fracture. The newly created surfaces enable agglomera-
tion of the particles, resulting in an increase in the powder particle size.
At the late stages of milling, fracture predominates over agglomera-
tion; therefore, the particle size is significantly reduced. High energy
ball milling processes are, by far, more effective than conventional ball
milling processes in reducing particle size; however, they introduce
strains to the milled particles. Different types of high energy milling
equipment for producing nanopowders are utilized [10].
Over the last years, thermal spraying has been employed to prepare
nanostructured coatings. Various types of spraying methods and
powder feedstocks have been used [11–13]. For instance, deposition of
WC/Co nanostructured coatings by High Velocity Oxygen Fuel (HVOF)
spraying and deposition of Al
2
O
3
–ZrO
2
nanostructured coatings by
Vacuum Plasma Spraying (VPS) have been mentioned by Kear and
Skandan [11]. Nanostructured and mixed structure (nanosized/
microsized) WC-12Co coatings have also been HVOF deposited by
Guilemany et al. [12].
Ceramic coatings using nanostructured powders as raw materials
are commonly being deposited by plasma spraying, since the plasma
torch can reach temperatures of the range 15,000–20,000 °C, capable
of sufficiently melting all the known ceramics [14]. For example, Al
2
O
3
nanostructured coatings have been air plasma sprayed [15]. Nano-
structured titania and alumina–titania coatings have been obtained by
vacuum plasma spraying (VPS) [16] and air plasma spraying (APS)
[17,18]. Nanostructured zirconia has also been APS deposited [19].
According to Lima and Marple [20], the manufacturing method of
the nanostructured powder and the deposition method are the two
main factors influencing the mechanical properties of the coatings.
Several studies have reported that the nanostructured coatings are
harder than their conventional counterparts: APS Al
2
O
3
–13 wt.%TiO
2
nanostructured coatings were found harder and of higher fracture
resistance than their conventional counterparts [21]. APS YSZ (ZrO
2
)
nanostructured coatings were reported harder and more resistant to
sliding wear or abrasion than the conventional ones [22]. Chen et al.
[23] ascribed the superior properties of APS YSZ nanostructured
coatings to their higher hardness and different wear mechanism, as
compared to the respective conventional coatings. HVOF Al
2
O
3
nanostructured coatings have exhibited higher hardness than that of
the conventional counterparts [24].
Surface & Coatings Technology 204 (2009) 15–27
⁎Corresponding author. Tel.: +30 26510 97309; fax: +30 26510 97034.
E-mail address: alekatou@cc.uoi.gr (A. Lekatou).
0257-8972/$ –see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.surfcoat.2009.06.013
Contents lists available at ScienceDirect
Surface & Coatings Technology
journal homepage: www.elsevier.com/locate/surfcoat

On the other hand, quite a few works have reported lower or similar
hardness values for the nanostructured coatings, as compared to the
respective conventional coatings. Despite that, their wear performance
has been found better than that of the conventional counterparts. Lin
et al. [25] found that APS nanostructured Al
2
O
3
–13 wt.%TiO
2
coatings
exhibited better wear performance than their conventional counter-
parts, despite the similar hardness values. They attributed this to the
better adhesion and higher crack extension force, as compared to their
conventional counterparts. Lima and Marple [26] also noticed similar
hardness values for HVOF nano- and conventional TiO
2
coatings. APS
Al
2
O
3
–13 wt.% TiO
2
nanostructured coatings had higher abrasion and
sliding wear resistance, despite exhibiting lower hardness than the
respective conventional coatings [17]. The lower hardness (but higher
wear resistance) of APS nanostructured Al
2
O
3
–13 wt.% TiO
2
coatings,
as compared to their conventional counterparts, was related to the
higher fracture toughness and the coating structural change during
wear [27]. Gell et al. [28] and Jordan et al. [29] claimed that the
microstructure and properties of nanostructured Al
2
O
3
–13 wt.% TiO
2
coatings can be related to a critical process spray parameter, defined as
the gun power divided by the primary gas flow rate. Intermediate
values of this parameter led to limited powder melting and retained
nanostructure in the coatings. The resulted mixed (nanosized/
microsized) structure was considered responsible for the superior
indentation crack resistance, spallation resistance and wear resistance
of the coatings despite the lower hardness, in comparison with the
conventional counterparts.
Liu et al. [30] did not observe any improvement of the wear
behaviour of nanostructured Al
2
O
3
–13 wt.% TiO
2
coatings over their
conventional counterparts, when both were deposited by the same
spraying method (HVOF and APS). They suggested that the wear
performance was more likely depended on the spray method itself.
It is conceivable that, despite handling the same fundamental
materials (ceramic oxide nanoparticles), quite different values for the
same properties may be obtained. Parameters involved in the
nanoparticle manufacture and treatment, as well as in the spraying
procedure, are primarily affecting the coating properties.
In the present study, a contribution to the plasma spray literature is
made concerning the effects of powder type and gun power on the
coating microstructure. At first, the critical spraying parameters,
under which the nanostructure is retained and integrated in the
deposited coatings, are examined. However, the integration of
nanostructure is not the only issue, since this should be combined
with toughness and cohesive–adhesive properties superior than those
of the conventional coatings. Secondly, the coating properties are
related to the morphology of the feedstock powder. Conventional
compact solid particles, nanostructured porous agglomerates and
sintered–nanostructured agglomerates are examined regarding their
behaviour in the plasma jet and the quality of the produced coatings.
Third, the effect of the application of a pre-spraying sintering stage on
retaining or enhancing the beneficial properties of nanostructure is
investigated. It should be emphasized that the raw material in this
study is a nanostructured powder manufactured by High Energy Ball
Milling without any post-treatment by the manufacturer. Commonly,
High Energy Ball Milling is followed by agglomeration processes, such
as spray drying [19]. In this work, a single sintering stage has been
applied, which can be conducted by the thermal shops, thus reducing
the cost of the raw material. Consequently, to the knowledge of the
authors, there is a lack of spherical investigations regarding the quality
of the obtained coatings.
This research effortextensively uses different X-ray diffraction-based
methods, aiming not only at estimating crystal sizes of feedstocks/
coatings and phase compositions of coatings, but also at examining the
dependence of their precision capabilities on theinvolved materials and
their manufacture method.
The present investigation is application targeted, since it was
induced by the Pyrogenesis SA need to replace conventional coatings
for fibre and thread wear-resistant applications at high temperatures.
Up to now, the raw material for these coatings has been conventional
Al
2
O
3
powder.
2. Experimental
2.1. Feedstock preparation and coating deposition
The nanostructured agglomerated powder employed in this work,
was manufactured by MBN
®
, in a laboratory high energy ball mill
functioning in the kinematic characteristics of a Spex
®
8000 model
[31].
After milling, the powder was centrifugally classified, to exclude
particles finer than 10 µm. Rounder particle shapes were obtained
by low energy milling (Jet Mill). Finally, an appropriate particle distri-
bution for spraying was attained by mechanical sieving. The obtained
particle size was (−53+10) µm. A portion of the nanostructured
powder was subjected to sintering in a Nabertherm box furnace
(air,1100 °C, 2 h). The conventional powder was Sulzer Metco's 105NS
©
of nominal particle size (−45+ 15) µm.
Plasma spraying was performed by an Atmospheric MiniGun torch
(Pyrogenesis Inc.) upon 50×50 × 5 mm 304 stainless steel coupons.
The main spraying parameters are given in Table 1. Three different
current values and, consequently, three different torch power values
were applied, in order to investigate the effect of the torch power on
the microstructure and the mechanical properties of the coatings.
Three types of coatings were produced: coatings deposited by
conventional Al
2
O
3
powder (C), nanostructured Al
2
O
3
powder (N)
and nanostructured Al
2
O
3
powder after sintering (S). Three series of
spraying experiments were conducted for each type of coating. Each
series was characterized bya different torch power, whereas the other
spraying parameters were kept the same for all the series. The spraying
power was increased by astep of 7 kW per series, as shown in Table 1.
Each experimental series included three identical runs. In other words,
three spraying runs were performed per coating type and torch power,
to produce three samples sprayed under identical conditions. The
produced coatings are denoted as C
i
,N
i
and S
i
, where i=1,2,3
represents the torch power; for example i=1 represents the lowest
torch power and i=3 represents the highest torch power.
2.2. X-ray diffraction
X-ray diffraction patterns (Bruker D8 Advance, Ni-filtered CuKa
radiation, Standard Slit) were received from the feedstocks and
coatings, employing a step size of 0.02°. The width of the peaks,
acquired by FWHM (Full Width at Half Maximum), was used to
estimate the average crystal size of the feedstocks and coatings [32].
Besides crystallite size, peak shapes and widths can be used to
determine non-uniform strain or defects, as uniform strain is related
to peak shifting and not peak broadening [32]. The crystal size and
lattice strain in the coatings can be determined using the X-ray
broadening techniques, since X-ray diffraction peaks are broadened
Table 1
The main spraying parameters.
Powder C C
1
C
2
C
3
Powder N N
1
N
2
N
3
Powder S S
1
S
2
S
3
Power (kW) 25 32 39
Current (A) 400 500 600
Ar (slpm) 28 28 28
H
2
(slpm) 14 14 14
Powder supply (g/min) 25 25 25
Spray distance (mm) 7 0 70 70
16 D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27

due to three factors: (a) instrumental effects, (b) small crystal size and
(c) lattice strain in the material [10]. The instrumental effects were
measured and subtracted by diffracting a SiO
2
Standard. Crystal sizes
were determined by the Scherrerequation, the Williamson Hall (WH)
plot and the Rietveld method.
The well-known Scherrer equation may be employed to calculate
the crystal size, without taking into consideration any lattice strains
[32,33]:
D=Kλ
βsizecosθð1Þ
In Eq. (1),Dis the crystal size (Å); Kequals 0.9; λis the wavelength of
the X-ray radiation (Å); β
size
is the peak width (rad); θis the Bragg
angle.
However, High Energy Ball Milling, as a method of powder
manufacture, induces high stresses into the particles that produce
high strains. Hence, in this work, the WH equation and the Rietveld
method were additionally employed, not only to estimate the crystal
size but also to examine whether the Scherrer equation is capable of
accurately calculating the crystal size of high energy milled powder
particles.
In the WH equation, crystal size and lattice strain can be separated,
so that crystal size may be determined with greater accuracy [32]:
D=Kλ
βsizecosθ+2esin θð2Þ
In Eq. (2),εis the strain (ε=Δd/d, where dis the lattice spacing and
Δdis the lattice strain).
The calculation of crystal size by the Rietveld method is based on a
different approach than that of the Scherrer and WH methods. The
Scherrer and WH techniques employ deconvolution of the observed
profile to distinguish between the phase-owing broadened profile and
the instrument-owing broadened profile. In the Rietveld method, the
whole profile is built and adjusted to the observed pattern by applying
a least-square fitting.
It should be noted that both methods are considered quite accurate
for crystals of size less than 100 nm. When this limit is exceeded, the
calculation accuracy gradually decreases with size increasing [10,34].
The Rietveld method is an XRD-based full pattern analysis method,
in which all factors contributing to the peak intensity (and measured
for each 2θstep), may be refined by a least-square procedure, until the
difference between the observed and calculated pattern is minimized.
The quantitative phase analysis performed by this method is of much
higher accuracy, as compared to the other peak intensity-based
methods [35].The X-ray Rietveld analysis was applied to determine
the phase percentages in the initial powders and their coatings by
refining the crystalline parameters and scale factors. The refinement
was carried out using PowderCell v2.4. In the present effort, the lattice
parameters used for the α-Al
2
O
3
(rhombohedral structure) percentage
determination were taken from the work of Ishizawa et al. [36],whereas
the lattice parameters used for the γ-Al
2
O
3
(defective spinel structure)
percentage determination were taken from the work of Paglia et al.
[37].Therefined crystalline parameters for α-andγ-alumina included
the scale factor, the lattice parameters, the aluminium and oxygen
occupancies, the aluminium and oxygen coordinates, the W, U, V factors
of the instrumental broadening and the likely preferential orientation.
The Pseudo-Voigt deconvolution was employed to determine the XRD
peak characteristics.
For comparison purposes, the Reference Intensity Ratio (RIR)
method was additionally applied, as a quick resource of quantitative
determinations [38,39]:
Xα
Xγ
=K1
Iα
Iγ
ð3Þ
where, X
α
and X
γ
are the weight fractions of phases αand γ,
respectively; I
α
and I
γ
are the intensities of the primary peaks of
phases αand γ;K
1
is the reference intensity ratio for a 1:1 mixture of
γ-Al
2
O
3
and α-Al
2
O
3
; in the case of using corundum, like the present
one, K
1
is given by the following ratio, which is reported in most JCPDS
cards:
K1=I100γ
I100α
ð4Þ
where, I
100 γ
denotes the intensity of the main peak of γ-Al
2
Ο
3
;I
100 α
denotes the intensity of the main peak of corundum (α-Al
2
O
3
) in a 1:1
mixture of γ-Al
2
Ο
3
and α-Al
2
O
3
.
From Eq. (3), the ratio X
α
/X
γ
can be calculated. Then, the weight
percentages of the phases are estimated according to Eq. (5):
Xα+Xγ=1:ð5Þ
In this work, the phase percentage calculation by the RIR method,
was conducted by comparing the intensity of the (400) peak for the
γ-phase with the intensity of the (103) peak for the α-phase.
Actually, the RIR method usually falls short because of problems
associated with preferred orientation and variable crystallinity [39].
This could be of high importance when conducting phase analysis
in nanostructured coatings. Wang et al. [27] noticed that alumina/
titania nanostructured coatings exhibited different preferential
growth than the conventional ones, which, in turn, seemed to
affect the wear properties. However, in the present study, the RIR
quantitative phase analysis is only introduced as a quick quantitative
analysis tool and its accuracy is compared to the Rietveld analysis
results; the latter are much more accurate but also much longer time
demanding. It should also be noted that the Rietveld analysis is
sensitive to the presence of any unidentified or amorphous phases.
Therefore, it has to be used with caution, in order to avoid erroneous
results [39].
2.3. Microscopy and mechanical property evaluation
The porosity and microstructures of the obtained coatings were
evaluated by optical microscopy (Leica 020-520.007 DMLM) and SEM-
EDX (Philips XL 40 SFEG and JEOL JSM 5600). The microhardness was
determined by a Shimadzu tester (14 measurements on the cross-
section of each sample, average of 42 measurements per torch power).
The indentation load used was 0.1 kg applied for 10 s. The surface
roughness of the coatings was measured by a TR200 Roughness
tester (four measurements per sample, average of 12 measurements
per torch power). The adhesion strength of the coatings (average of
three measurements per coating and torch power) was measured
by a portable elcometer (110 P.A.T.T.I.
®
) according to ASTM C633-01
[40].
Crack propagation resistance by indentation is considered a
useful technique to evaluate the relative fracture toughness of
coatings that exhibit similar porosity and thickness levels. It provides
a basis for qualitatively determining the relative effect of the
powder morphology on crack propagation characteristics and wear
performance [20]. Qualitative observations regarding the crack
propagation resistance of the coatings were based on the micro-
scopical examination (optical and SEM) of indentation impressions.
Cross-sections of the coatings were intended by a Vickers intender
(5 measurements per sample, average of 15 measurements at the
power of 39 kW). The indentation load used was 1 kg applied for 15 s.
The indenter was aligned, so that one of its diagonals was parallel to
the substrate surface. The total length (tip-to-tip) of the major crack
(2c) parallel to the substrate surface located at or in the vicinity of
the corners of the indentation impression was measured. Then, the
crack extension force (G
c
) was extracted; in plasma sprayed coat ings,
17D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27

the relative fracture toughness may conveniently be expressed by the
crack extension force (G
c
), which is given by Eq. (6) [41]:
Gc=6:115T10 −4α2P
c3ð6Þ
where, G
c
: the crack extension force (J/m
2
); α: the impression half-
diagonal (m); P: the indentation load (N).
Specimens polished to Rab1 µm were subjected to friction and
sliding wear testing at room temperature. A C.S.E.M pin-on-disk
tribometer was employed. The wear tests were performed under a
normal load of 10 N at a sliding speed of 0.1 m/s. A 6 mm diameter SiC
ball was used as a counter body. During testing, the relative humidity
of the environment was kept at 50%. The cycle radius was 3 mm. The
tests were completed after conducting 200,000 cycles that covered a
distance of approximately 3.2 km. The depth of the wear tracks was
determined by stylus profilometry. The wear tests were conducted
according to the ASTM G99-05 Standard [42]. The surface roughness
of the wear tracks was also measured by a TR200 Roughness tester
(10 measurements per wear track).
3. Results and discussion
3.1. Feedstock characterization
Fig. 1a, b and c shows the morphologies of the conventional (CP),
nanostructured (NP) and sintered (SP) nanostructured powders,
respectively. The C powder consists of solid particles of somewhat
irregular/angular shape (fused and crushed). The N and S powders
consist of agglomerates of nanoparticles of smoother shapes (a typical
morphological result of the milling process [10]). Macroscopically,
both sintered and non-treated nanopowders present similar particle
morphologies and size. It seems that the sintering step did not
instigate any micro-particle size increase.
Fig. 2 illustrates the surface of the agglomerated nanoparticles
in the untreated (Fig. 2a) and treated (Fig. 2b) state. The higher
magnifications, in relation with the magnifications of Fig. 1, reveal
slight microstuctural differences between the two treatment states.
Sintering has resulted in smoothing of the nanoparticle surface and a
decrease in the finer nanoparticle population. The latter have been
either consolidated or embodied to the coarser ones (Fig. 2b). The
finer nanoparticles present a higher area-to-volume ratio than the
coarser ones; hence, their spontaneity to reduce their surface energy
Fig. 1. The feedstock: (a) conventional crushed and fused powder; (b) high energy ball
milled nanopowder; (c) the nanopowder after sintering (110 0 °C, 2 h).
Fig. 2. Agglomerates of nanoparticles, (a) as received and (b) after sintering; the arrows
point at nanoparticle coalescence by necking.
18 D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27

is higher, causing their consolidation. Coalescence occurs by necking
(pointed by arrows in Fig. 2b), which is typical for alumina [43].
Fig. 3a shows that the main phase in the three powders is α-Al
2
O
3
.
The N powder additionally contains Al and α-Fe. The stoichiometry
used in the powder production is the main reason for the aluminium
presence [31]. The iron presence indicates powder contamination by
the tool steel balls and vials of the ball mill. The percentage of iron was
estimated by EDS analysis and Rietveld analysis as approximately
0.74 wt.% and 1.14 wt.%, respectively. The aluminium percentage was
estimated (by Rietveld analysis) as approximately 1.1 wt.%. Powder N
presents the widest peaks amongst the feedstocks, suggesting
nanocrystalline dimensions and possible non-uniform strain exertion.
Powder S presents narrower peaks than those of powder N, though
wider than those of powder C, indicating considerable presence of
submicron crystals and possible residual stress relief, as discussed in
the next paragraph. The detection of traces of Fe
2
O
3
and the absence of
peaks from powder S that correspond to Fe suggest that sintering has
caused oxidation of Fe to Fe
2
O
3
, in free form or in solution in Al
2
O
3
.The
latter is inferred by the α-phase peak displacement at lower angles, as
compared to the spectra of powders C and N (Fig. 3b). On the other
hand, the slight displacement of the ‘S’peaks at lower angles, as
related to the ‘N’peaks, could be a consequence of uniform strain in
relation to powder N [32].
Regarding the crystal size of powder C, the Scherrer, Williamson Hall
and Rietveld methods did not provide any significant information about
it, except that it is greater than 100 nm. As mentioned in Section 2.2,
the Scherrer and Williamson Hallmethods are considered quiteaccurate
only for crystal sizes less than 100 nm.
The crystal size of powder N was found 24 nm, 80 nm and 82 nm
when calculated by the Scherrer equation, Williamson Hall plot and
Rietveld analysis, respectively. The quite different results regarding
the Scherrer equation, are possibly owing to the production process of
the nanopowder: Fig. 4 presents the (β
size
⁎cosθ) vs. sinθplots for
powders N and S, according to the Williamson Hall Eq. (2). The line
slope equals the double of the lattice strain, according to Eq. (2);
hence, a strain-free specimen would demonstrate a (β
size
⁎cosθ) vs.
sinθline parallel to the x-axis. The line slope corresponding to powder
N suggests a strained specimen; the consecutive high energy impacts
during milling induced high strain and distortion in the grains, which
were imprinted on the XRD peak widths of powder N. The slope of
the (β
size
⁎cosθ) vs. sinθline corresponding to the sintered powder is
much lower, suggesting that the sintering stage has also served as
an annealing stage, causing stress relief. Indeed, the lattice strain of
powder N was estimated as 1.1%, by the WH equation, and 0.7%, by
the Rietveld method; the strain of powder S was estimated as 0.2%
by the WH equation, and 0.1%, by the Rietveld method. The high
strain in powder N, determined by both the Williamson Hall and
Rietveld method, suggests that the Scherrer equation is a rather
inaccurate method for nanocrystal size calculation of high energy
milled powders, since it does not consider any lattice strains.
The crystal size of powder S was found 118 nm, when calculated by
the Scherrer equation, and greater than 100 nm, when calculated by
the WH and Rietveld methods. The ‘Scherrer’crystal size of the S
feedstock should also be taken with caution, due to the questionable
accuracy of the method for crystal sizes greater than 100 nm.
The differences between the values obtained by the WH and
Rietveld methods might be attributed to two reasons: (a) The WH plots
were built on the basis of the seven strongest peaks, whereas the
Rietveld analysis used the thirteen strongest peaks. (b) In WH plots,
the FWHM peak was measured by employing the WinFit software.
However, in a certain peak pair ((214) and (800)), there was peak
overlapping, which reduced the accuracy of the FWHM measurements.
All the same, both microscopical and XRD examinations support a
limited crystal size augmentation after sintering, which more properly
classifies the crystal size of the sintered powder between the nano-
and submicron range.
3.2. Coating characterization
A major issue in spraying nanoparticles is the retention of part of
the original nanostructure in the coating, in order to maintain its
Fig. 3. (a)The XRD patterns of the three powders and their coatings. C
i
,N
i
,S
i
:
conventional, nanostructured and nanostructured/sintered coatings, respectively.
i=1,2,3: coatings deposited at 25, 32 and 39 kW, respectively. α:α-Al
2
O
3
,γ:γ-
Al
2
O
3
.(b) In the SP spectrum, the peaks of the α-phase have been displaced to slightly
lower angles, as compared to the N and C powders.
Fig. 4. The Williamson Hall Plot for the nanostructured powder (NP) and the
nanostructured/sintered powder (SP). The numbers in parentheses refer to the
crystallographic planes.
19D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27

beneficial effect on the properties. On the other hand, coatings must
exhibit integrity, cohesive and adhesive properties, which can only be
achieved by a sufficient degree of particle melting. However, complete
particle melting would practically eliminate the initial nanostructure.
Cross-sections of the conventional, nanostructured and nanostruc-
tured/sintered coatings obtained at 39 kW are shown in Fig. 5a, b and c,
respectively. The surface morphologies of the C
2
,N
2
and S
2
coatings are
illustrated in Fig. 6a, b and c, respectively. These figures are commented
in the following sections.
3.2.1. Phase analysis
3.2.1.1. Qualitative phase analysis. XRD analysis in Fig. 3a, shows that
the deposited coatings consist of α-Al
2
O
3
and γ-Al
2
O
3
,incompatibility
with previous works [15,21,44,45]. The high cooling rates of the
thermally sprayed particles (10
6
–10
7
K/s) can cause stabilization of
originally metastable phases. Thus, the formation of γ-Al
2
O
3
in the
coatings is the result of rapid quench from molten Al
2
O
3.
α-Al
2
O
3
in the
coating stems from the unmelted α-Al
2
O
3
fraction in the pristine
powders [46–48].
Crystal size calculations for the α-phase in the nanocoatings gave
crystal sizes much larger than 100 nm. As the penetration distance
of the X-ray beam in the sample can be of several micrometers,
depending on the characteristics of the sample [32,49], it is most
possible that this thermal spray-caused crystal size increase concerns
the bulk of the coating. Thus, it is considered that the energy of the
plasma torch has incited nanoparticle consolidation. As nanoparticles
have a high surface to volume ratio, they are characterized by high
reactivity and strong tendency towards consolidation [50]. As it will be
shown in the following sections, part of the initial single nanoparticles
has been retained in the coating; however, their contribution to peak
broadening (Fig. 3a) is insignificant due to their low content.
Fig. 5. Cross-sections of the (a) C
3
coating, (b) N
3
coating and (c) S
3
coating.
Fig. 6. The surface states of the (a) C
2
coating, (b) N
2
coating and (c) S
2
coating. The
arrows point at clusters of nanoparticles.
20 D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27

Regarding the γ-Al
2
O
3
phase, the Scherrer Eq. (1) suggests the
presence of nanoparticles of size 25–42 nm for the three types of
coatings, whilst the Williamson Hall Eq. (2) suggests the presence of
nanoparticles of size 23–40 nm (Table 2). It should be noted that the
Scherrer Eq. (1) only considers equiaxed particles, whereas the
Williamson Hall Eq. (2) considers both equiaxed and unequiaxed
particles. SEM investigation of the fracture surface of the N coatings
reveals the formation of nanocolumns (Fig. 7). Therefore, the crystal
size values estimated by the Williamson Hall Eq. (2) are considered
more accurate than those estimated by the Scherrer Eq. (1). The
nanocolumns have grown to the opposite direction of the heat flow.
The column formation is induced by the slow heat loss by conduction
through the previously deposited splats: When the temperature rise of
the liquid droplet (caused by the release of fusion heat during rapid
crystal growth) is higher than the temperature fall (caused by the heat
loss by conduction), further nucleation is suppressed and columnar
shapes are favoured [46]. As alumina is a relatively poor thermal
conductor, (the conductivity of alumina is approximately 5.9 W/mK at
100 0 °C [51]), columnar structure is expected to appear. Here, it should
be noted that the Williamson Hall method is only capable of accurately
measuring the width of the column base (being the only one lower
than 100 nm). Williamson Hall did not indicate any strain effect;
however, as, there is peak overlapping between the α- and γ-phases,
there should be some uncertainties regarding the stress effects.
No significant preferential growth in the 440 direction of γ-Al
2
O
3
in the N and S coatings was noticed, as was observed in the case of APS
deposited Al
2
O
3
/13 wt.%TiO
2
coatings [27]. The intensities of the 400
and 440 peaks remained constant for all the coatings, indicating no
specific surface texture.
3.2.1.2. Quantitative phase analysis. The percentages of the present
phases in the feedstock powders and their coatings are plotted against
the torch power in Fig. 8a (Rietveld analysis) and Fig. 8b (RIR analysis).
The phase percentages for the same torch power, as estimated by the
two methods, do not exhibit any significant differences. Small
differences might be attributed to two reasons: (a) The RIR analysis
does not apply any parameter refinement, in contrast to the Rietveld
analysis. (b) The deposited coatings always present a slightly prefer-
ential orientation due to solidification towards a specific direction.
The inabilit yo f the RIR method to include th e higher intensities of the
γ-phase, due to orientation, is responsible for calculating generally
increased fractions of α-phase (Fig. 8b), as compared to the Rietveld
analysis (Fig. 8a).
At low and medium spraying powers (25 and 32 kW), the α-Al
2
O
3
content of the nanocoatings is notably higher than that of the
conventional and sintered counterparts. Fig. 8a reveals similar
percentages of α-Αl
2
O
3
for the C and S coatings and, consequently,
similar melting degrees. Two reasons may account for the similar
melting states of coatings C and S: (a) Compact particles or sintered
agglomerates exhibit higher thermal conductivity than their loose
porous counterparts [52,53]; and/or (b) The sintering stage increases
the strength and density of the agglomerated nanoparticles, allowing
Fig. 7. The fracture surface of the nanocoating (N
2
) revealing columnar structure.
Table 2
Average crystal size and strain of the γ-phase in the pristine powders and the resultant
coatings.
Material Crystal size—Scherrer
(nm)
Crystal size—Williamson
Hall (nm)
Strain—Williamson
Hall (%)
Powder C –– –
Powder N –– –
C
1
25 26 (h00) 0.0 (h00)
C
2
26 39 (h00) 0.1 (h00)
C
3
28 23 (h00) 0.0 (h00)
N
1
37 35 (h00) 0.0 (h00)
N
2
42 40 (h00) 0.2 (h00)
N
3
39 35 (h00) 0.1 (h00)
S
1
30 27 (h00) 0.0 (h00)
S
2
32 24 (h00) 0.0 (h00)
S
3
35 29 (h00) 0.0 (h00)
C
i
: conventional coatings; N
i
: nanocoatings; S
i
: sintered/nanostructured coatings. The
indices i=1,2,3 correspond to spraying power of 25, 32 and 39 kW, respectively.
Fig. 8. Phase percentages in the powders and their coatings, estimated by the
(a) Rietveld and (b) RIR method.
21D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27

them to enter the plasma jet hot zone intact and melt more sufficiently
[51].
3.2.2. Porosity and melting degree
Table 3 summarizes the measured propertiesof the coatings. It can
be seen that the coatings fabricated by the three feedstocks present
differences regarding their porosity, surface roughness,microhardness
and adhesion values.
The coatingporosity (estimated by image analysis and SEM at ×500,
on cross-sections) as a function of the torch power is shown in Fig. 9.
It should be emphasized that the employed image analysis techniques
can only identify the coarser pores of thermalspray coatings and give an
overall view of the coating porosity.
Fig. 9 shows a decrease in porosity with an increase in torch power.
Porosity is higher at the lower spraying energies, where the semi-
molten particles are less deformed. At the higher spraying energies,
molten material of high diffusivity has filled the asperities and gaps of
the previously deposited layers, leading to lower porosities.
In Fig. 9 it is also observed that, for each torch power, the N coatings
exhibit the highest porosity. Three reasons may account for this:
(a) The inherent porosity of the nanoparticle agglomerates [15] results
in lower thermal conductivity (as compared to the compact conven-
tional particles and, even, the slightly denser sintered nanoparticles)
and, thus, a lower melting degree; semi-molten material has less
deformation ability leading to higher coating porosities. (b) The semi-
molten nanoparticle agglomerates have been deposited in the form of
hollow spheres that retain part of their initial nanostructure in the core
[15]. Conventional fused and crashed particles, being compact units,
are not capable of air entrapping in their interior, which is responsible
for the semi-molten hollow sphere phenomenon [54]. (c) The compact
particles do not present any risk of dissociation when inserted in the
plasma jet, retaining their momentum during their flight in the plasma
hot zone, which promotes deeper melting [51]. Sintered nanoparticles
also present less risk of dissociation than the untreated nanoparticles.
Although the melting degrees of the C and S coatings are similar, –
as reflected by the similar α-fractions in Fig. 8 –, the latter exhibit
slightly higher porosities than the former. The slightly higher porosity
of the S coatings is possibly due to the porosity of the powder
agglomerates, since sintering of the nano-feedstock has not led to a
particularly dense structure, but has caused a limited coalescence, as
fore-mentioned in Section 3.1.
At the highest spraying power, the porosities of the S and C coatings
tend to converge (Fig. 9), in compatibility with the same trend of the
α-Al
2
O
3
fractions in these coatings (Fig. 8). The high power (39 kW)
has also caused a significant reduction of the gap between the
porosity of the N coating and the porosities of the S and C coatings.
Thesameappliesfortheα-phase percentages in the coatings (Fig. 8).
It seems that a high torch power can reduce the third resultant for
porosity augmentation, namely the hollow sphere formation. Previous
effort [15] showed that a semi-molten agglomerate of nanoparticles
in an Al
2
O
3
nanocoating consistsof four concentric zones that reflect the
thermalhistory of the agglomerateduring its trajectory inthe plasma jet.
From the exterior to the core of a semi-molten agglomerate, the four
zones include a molten surface, a liquid phase sintering zone, a solid
state sintering zone and a sintering-free zone retaining its initial
nanostructure. The latter envelops a void core. It is postulated that, in
the case of the S coatings, the higher spraying energy has caused
intensification of the melting and sintering processes, leading to a
reduction of the unmelted materialand an increase in theparticle plastic
deformation; the central void core has diminished. Thus, the particles
have been deposited as splats rather than hollow spheres. Fig.10 shows
a semi-molten part of the S
3
surface. Porosity is observed in the part
that retains the nanostructure (right).
3.2.3. Surface roughness
The coating surface roughness values are given in Table 3.Fig. 6
illustrates typical surface morphologies of coatings C (Fig. 6a), N
(Fig. 6b) and S (Fig. 6).
Fig. 6b shows that the N coatings possess an irregular/rough
morphology indicating that the agglomerates have retained an
unmelted core. Moreover, clusters of individual nano/submicron
particles are discerned on the surface (pointed by arrows). Therefore,
the pile-up of the splats is not dense due to a large fraction of deposited
semi-molten particles, in good correlation with the high values of
porosity. Similarly, the decreasing trend of roughness with torch power
increasing (Table 3) is compatible with the decrease in the fraction of
unmelted deposited particles and the consequent decrease in the
coating porosity. Furthermore, for each torch power, the coating
surface roughness follows the order CbSbN, in compatibility with the
melting degree order (C NSNN) and the porosity order (CbSbN). Here,
Fig. 10. Part of the surface of coating S
3
. Porosity is observed in the part that retains its
nanostructure.
Fig. 9. The coating porosity as a function of the torch power.
Table 3
Characteristic properties of the coatings.
Designation Porosity
(%)
Surface roughness
Ra (µm)
Hardness
HV
100g/10s
Adhesion
(MPa)
C
1
6.6±0.6 7.4± 1.0 1045± 141 17.6± 0.9
C
2
6.1± 0.7 5.9± 0.8 1063± 139 18.1± 1.5
C
3
5.5±0.8 5.8± 0.7 1102± 112 19.8±1.6
N
1
14.2± 3.7 10.2±2.4 975± 126 10.5± 0.8
N
2
12.0± 2.3 10.8±0.7 1081± 166 11.4±1.3
N
3
9.2±1.7 8.4± 1.4 1149± 125 17.7± 2.4
S
1
7.3± 1.5 9.2± 2.0 1070± 215 24.1± 2.1
S
2
7.7± 2.6 8.6± 1.6 1086± 206 23.4 ± 1.8
S
3
5.9±0.5 8.4± 2.2 1126± 143 26.3±2.2
22 D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27

it should be noted that the spherical particles discerned in Fig. 6a and b
were produced by splashing of fully melted particles.
3.2.4. Microhardness
The N coatings exhibit similar hardness values with the C and S
counterparts (Table 3). Since nanostructured coatings exhibit approxi-
mately the double porosity of the respective conventional ones, it is
expected that they would have lower hardness, as the carrying load
solid area is reduced [47]. This contradictive result might be elaborated
by two facts: First, the relatively low indentation load leads to
microhardness values more depending on the material intrinsic
properties and less depending on the material extrinsic properties,
such as coarse pores, splat boundaries etc. [55]. The N coatings contain
notably higher percentages of α-phase than the C and S coatings and,
consequently, lower percentages of γ-phase. α-Al
2
O
3
is known to be
considerably harder than γ-Al
2
O
3
[56]. Second, as fore-mentioned, the
N coatings contain a higher percentage of unmelted materialthan the C
and S coatings, which has retained its nano or submicron grain size.
High hardness values in nanostructured materials are closely related to
their low grain size, as suggested by the Hall–Petch relationship [57].
Therefore, the negative effect of the high porosity on the hardness
of the N coatings is counterbalanced by the positive effects of the high
α-Al
2
O
3
fraction and low grain size. As a result, coatings N do not lack
in hardness, in relation to coatings C and S.
The similar or even slightly higher hardness of the S coatings, as
compared to the C and N coatings, at the low and medium power
levels, despite the low content of the hard α-phase (higher, though,
than that of the C coatings) may be attributed to the relatively low
porosity (lower than N but higher than C) and the retention of nano/
submicron α-grains (lower than N).
Table 3 indicates that a decrease in torch power leads to a decrease
in hardness. As previously discussed, the hardness of the coating
mostly depends on three factors: porosity, phase composition and
nanostructure. A decrease in torch power causes an increase in
porosity, the α-phase percentage and the retained nano/submicron
α-grain percentage. An increase in porosity has a weakening effect on
the coating hardness; an increase in the α-phase and the retained
nano/submicron α-grains has a strengthening effect on the coating
hardness. The slight reduction of hardness with torch power decreas-
ing indicates a slightly predominant effect of the porosity on the
coating hardness over the other two factors. However, this observation
should be used with caution, due to the statistical uncertainties in the
hardness values, as shown in Table 3.
3.2.5. Adhesion
Table 3 shows that coatings S present notably higher adhesion
strength than coatings N and C. Furthermore, at the lower torch powers
(25 and 32 kW) they exhibit approximately double values than the
ones of the N coatings. The superior adhesive properties of the S
coatings can be ascribed to the: (a) consequences of sintering, such as
increased intraparticle cohesion and strain relief; (b) existence of
randomly dispersed strong adherent zones of retained nanostructure –
nanozones–in the bulk of the coating and at the coating/substrate
interface, that may act as crack arresters or crack deflectors.
Fig. 11a and b presents cross-sections by the coating/substrate
interface of the N
3
and S
3
coatings, respectively. The N
3
coating shows
an inhomogeneous structure. The splat boundaries seem rather weak.
Microcracks have been developed in the bulk of the coating and at the
substrate/coating interface. Nanozones are discerned that act as crack
arresters (numbered 1 in Fig. 11a) or crack deflectors (numbered 2
in Fig. 11a). However, a loose nanozone by the coating/substrate
Fig. 11. Cross-sections at the coating/substrate interface of the (a) N
3
coating and (b) S
3
coating. (a) 1: crack arresting nanozone, 2: crack deflecting nanozone, 3: loose nanozone
participating in the microcracking mechanism; (b) the arrows point at strong nanozones; (c) a crack arresting nanozone in N
3
; (d) a loose nanozone in N
3
.
23D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27

interface is observed, that participates in the microcracking mecha-
nism (numbered 3 in Fig. 11a).The S
3
coating presents a totally
different image in the vicinity of the coating/substrate interface. The
structure is more homogeneous and the splat boundaries are hardly
discerned; coherent nanozones are randomly dispersed in the coating
(pointed at by arrows in Fig. 11b). Moreover, microcracks have not
been developed in the bulk of the coating and at the coating/substrate
interface. The formation of nanozones-crack arresters in the bulk of
the coating has been noticed by Lima and Marple [20,26], when they
sprayed nanostructured titania by HVOF. Bansal et al. [58] observed
microcracks at the interfaces of APS Al
2
O
3
–13 wt.%TiO
2
conventional
coatings/steel substrate, whereas they did not find any cracks at the
interfaces of the respective nanostructured coatings/substrate. They
claimed that the microstructure of the “nano”coating (consisting of
regions of fully molten splats interspersed with partially molten
rounded microstructural features), together with the presence of a
TiO
2
-rich amorphous phase at the partially molten features/steel
interface, were responsible for the significantly improved interfacial
toughness of the “nano”coating. Typical crack arresting and loose
nanozones are more clearly illustrated in Fig. 11 c and d.
To conclude, the inferior adhesive properties of the N coatings, as
compared to the C and S coatings can be attributed to the: (a) higher
porosity of the N coatings, (b) lower cohesion of the unmelted particles
due to less anchoring points; and (c) formation of nanozones that,
to a major extent, are not coherent enough to act as effective crack
arresters.
3.2.6. Fracture resistance
Fig. 12 illustrates indentation impressions on cross-sections of
the C
3
(Fig. 12a), N
3
(Fig. 12b) and S
3
(Fig. 12c) coatings. The selected
coatings exhibited porosities of 6.0%, 7.5%, and 6.4%, respectively. The
main criterion for the selection of these coatings was their, as similar
as possible, porosity values, since porosity is a paramount property
when measuring or analyzing material toughness values and
characteristics [26].
The well-shaped indentation impressions on the S coatings (an
almost perfect rhomb), as compared with the indentation impressions
on the C and N coatings (Fig. 12), indicate that the S coatings exhibit a
more ductile behaviour on indentation than the C and N coatings. In
Fig. 12a and b, cracks are observed within the area of the indentation
mark, also indicating that the C and N coatings are more brittle than
the S coating. The superior ductility of the S coatings can more clearly
be seen by comparing the toughness-to-hardness ratio of the coatings,
in Table 4. This ratio is considered proportional to the ductility of hard
materials [59]. The S
3
coating presents the highest G
c
-to-HV ratio,
suggesting a more ductile behaviour in relation to the C
3
and N
3
coatings.
For all coatings, cracks tend to propagate along splat borders,
parallel to the coating–substrate interface. Such an anisotropic
behaviour is related to accumulation of micropores between the
lamellae, decreasing the real area of contact. In an APS coating, the
preferential accumulation of micropores at splat boundaries sets the
latter as the most susceptible areas to fracture [46].
In some indentation cases, the N
3
coating has produced multiple
cracking by the indentation impression (Fig. 13). These have also been
observed in other works and have been attributed to the high porosity
Table 4
Fracture toughness and wear properties of the coatings.
Coating G
c
(J/m
2
) Wear rate
(×10
−4
mm
3
/(Nm))
Track width
(μm)
G
c
/HV
(μm)
a
Roughness of
wear tracks (μm)
C
3
7.1± 2.0 20 1378 0.66 3.25±0.39
N
3
5.3±1.7 14 1192 0.47 2.69 ± 0.28
S
3
10.9± 2.3 7 939 0.99 2.12± 0.32
a
HV number has been converted to GPa units.
Fig. 12. Indentation impressions on cross-sections of: (a) C
3
, (b) N
3
, and (c) S
3
coatings. Fig. 13. Multiple cracking by the indentation impression in coating N
3
.
24 D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27

[47]. An additional reason for multiple cracking indents could be the
insufficient melting, which produces weak intersplat bonding as
shown in Fig. 11a.
The crack extension force values, extracted from Eq. (6) are given
in Table 4.FromTable 4, it can be seen that the S coatings present the
highest fracture toughness, whereas the N coatings presentthe lowest
toughness. The fracture toughness of the S
3
coatings is more than
double the toughness of the N
3
coatings, despite the presence of more
nanozones in the N coatings (the latter is indicated by the higher
fraction of unmelted material, as shown in Fig. 8). Two are the
principal reasons for the inferior toughness of coatings N as compared
to coatings S and C: (a) The lower degree of melting has caused a
reduction of the anchoring points of the deposited splats [23,60],as
shown in Fig. 11a, reducing the bond strength and, consequently,
providing a path for cracks to propagate along the weak splat
boundaries. (b) The subsequent increased porosity has reduced
fracture toughness by reducing the solid area that carries load [47].
These two reasons may outweigh the benefit of embedded nanozones
and result in low fracture toughness. The S coating, on the other hand,
has tougher splat boundaries than the N coating, due to deeper
melting. Furthermore, the retained nanozones are adequately coher-
ent (by coalescence due to sintering) and randomly dispersed in the
coating, so that they can act as crack arresters.
The mean fracture resistance of the S
3
coating is approximately
55% higher than the fracture resistance of the C
3
coating, despite the
similar melting degree (Fig. 8) and higher porosity (Fig. 9). Hence, the
significance of strong nanozone formation in the coating fracture
toughness is inferred.
3.2.7. Wear resistance
Table 4 includes the results of the wear tests for the C
3
,N
3
and S
3
coatings, whereas Fig. 14 shows the friction evolution. The wear rate
was determined as the volume loss of coating per unit of sliding
distance and normal load applied. The S
3
coating exhibits the lowest
wear rate. Its wear rate equals approximately 50% of the N
3
coating
and 35% of the C
3
coating. The wear rates are (somewhat)
proportional to the wear track width values, also given in Table 4.
The lower wear rate and the higher friction coefficient of the S
3
coating, compared to the other two, can be justified by its higher
fracture toughness and bond strength. The lower wear rate of the N
3
coating, as compared to the C
3
coating, despite its lower fracture
toughness, is not compatible with other works, which have proposed
that fracture toughness and wear resistance are two analogous
properties [21,25,26]. On the other hand, the N
3
coatings present the
highest hardness among the three investigated types; however,
hardness is not the paramount property affecting the wear rate, but
fracture toughness is, according to previous studies [17,28,29].
Fig. 15 shows the wear tracks of the C
3
(Fig. 15a), N
3
(Fig. 15b) and
S
3
(Fig. 15c) coatings. In Fig.15a, brittle fracture is observed on the left
side of the photo, whereas limited plastic deformation is observed on
the right. EDX on the right side of the photo has revealed the presence
of Al, O and Si, suggesting material transfer from the counter body. The
left side of the photo has been found free of Si.
In Fig. 15b, a discontinuous compact layer has covered the coating
surface. A limited brittle fracture is noticed. EDX has revealed the
presence of Si in the layer, as well. Several cracks exist on the worn
surface. Micro-cracking on the coating surface is attributed to fatigue
fracture [61].
Fig. 14. Friction coefficient vs. distance for the three types of coatings.
Fig. 15. Wear tracks on the (a) C
3
, (b) N
3
, and (c) S
3
coatings; the arrows point at
compact overlayer delaminations (down centre) and crater formation (down right).
25D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27

In Fig. 15c, the Si enriched compact layer has also been formed. In
all coatings, it is postulated that the silicon was oxidized and
transferred to the worn surface of the coatings [62]. Compared with
alumina, silica is softer [63] and more effective in distributing contact
stress. Thus, the presence of the layer on the worn surface seems to
have prevented the coatings from undergoing a fully brittle fracture.
Furthermore, the existence of nanograins in the N and S coatings has
allowed the coatings to exhibit a more ductile behaviour [26,64].
Fig. 15c reveals areas of brittle and ductile fracture. Arrows point at
areas of compact layer delaminations (left) and crater appearance
(right). The craters are considered to develop inside splat boundaries,
because the latter are the only zones where strong adhesion between
transferred layer and underlying material is not likely [63].
Two reasons are considered responsible for the higher wear
resistance of the N coatings, in comparison with that of the C coatings,
despite their lower fracture toughness:
(a) The retained nanostructure: Several studies have emphasized
the importance of grain size on the wear behaviour of
conventional and nano ceramics [23,63,65]. The wear resistance
of ceramics increases with decreasing grain size [65]. The
smaller the grain size, the higher the external stress required to
induce grain boundary cracking and grain pulling out is. The
reason is that stresses accumulate at the grain boundaries;
coarse grain materials, characterized byfewer boundaries, tend
to accumulate more stresses. So, it is thought that the presence
of nanosized grains favours the spread of the stresses to
multiple boundaries. Fig. 16 shows the fracture surfaces of at
least two deposited semi-molten N
3
particles or, more precisely,
agglomerates of nanoparticles. The outer zones of the particles
have undergone melting, liquid phase sintering and solid
state sintering, according to the thermal history reported in
Section 3.2.2. The cores of the particles have retained their initial
nanostructure. It is evident that the imposed stresses on a par-
ticle may dissipate along the abundant nanograin boundaries.
(b) The Si enriched compact layer: The harder material (N
3
)is
expected to remove more counter body material, which
accounts for the extensive formation of the protective SiO
2
overlayer.
Fig. 15 confirms the observations in Section 3.2.6. regarding the
higher ductility of the S coatings. The S
3
coating exhibits a smoother
wear scar with less formation of wear “scales”(indicating the wear
direction), as compared to the C
3
and N
3
coatings. Moreover, the “S
3
”
wear track has lower roughness than the “C
3
”and “N
3
”wear tracks
(Table 4), reinforcing the claim about the higher ductility of the S
3
coating, in relation to the C
3
and N
3
coatings.
At this point, it is worthwhile to note that the friction coefficient vs.
distance curves of the C
3
and N
3
coatings are similar, in terms of values
and shape (Fig. 14); furthermore, their wear tracks present similar
morphologies (Fig. 15). The friction coefficient vs. distance curve of
the S
3
coating looks different, presenting less variation in the range
500 to 3000 m; also, the friction coefficient is notably higher than
those of the C
3
and N
3
coatings. The similar wear behaviour of the C
3
and N
3
coatings is currently under a more detailed investigation.
The S coatings seem to combine the best properties of both the
other rival coatings. In especial, the C coatings underwent the highest
melting degree, whilst the N coatings presented the highest ability to
maintain their low grain size. The S coatings did manage to maintain
nanostructure to some extent; at the same time, they were sufficiently
melted, so that they gained adequate bond strength and toughness.
The nano/submicron grains inhibited high stress build-up by disper-
sing stresses along numerous grain boundaries. In addition, they
induced a more ductile behaviour by building a protective layer. The
high bond strength prevented splat de-lamination.
4. Conclusions
Plasma sprayed coatings obtained by conventional Al
2
O
3
(C),
high energy ball milled nanostructured Al
2
O
3
(N) and sintered/
nanostructured Al
2
O
3
(S) powder feedstocks have been evaluated and
compared:
1. Discrepancies in the average crystal size values of the N powder
are observed when calculated by the Scherrer Equation and
Williamson Hall Plot. The Scherer Equation has failed to accurately
calculate crystal size, due to its inability to distinguish between
actual size and strain contribution to the overall crystal size.
2. Sintering of powder N has led to a limited coalescence of the finer
nanoparticles by necking and stress relief by annealing.
3. The coatings consist of γ-Al
2
O
3
solidified from the melt and α-
Al
2
O
3
originating from the unmelted part of the powder. γ-Al
2
O
3
has adopted a columnar morphology. The applied XRD techniques
were unable to accuratelycalculate the crystal size of the α-phase.
The width of the γ-column base could only be accurately
determined by the Williamson Hall technique.
4. Both, the RIR (quick determination) and Rietveld (tedious
determination), XRD-based methods of quantitative phase analy-
sis, were found capable of accurately calculating the constituent
phases of the coatings. The absence of significant preferential
orientation allowed a rather accurate phase analysis by RIR.
5. The XRD-based techniques revealed similar α-Al
2
O
3
percentages
and, consequently, similar melting degrees of the C and S powders
and a lower melting degree of the N powder. The lower melting
degree of the N powder is mainly attributed to its lower thermal
conductivity, as compared to the C powder. The higher melting
degree of the S powder (similar to that of C) is mostly owing to the
sintering treatment that increased the strength of the particles
entering the plasma jet hot zone and allowed them to melt more
efficiently.
6. Increasing torch power has led to an increase in the powder
melting degree and a consequent decrease in the coating porosity;
as a result, the coating surface roughness has decreased, whilst
the coating hardness and adhesion strength have increased.
7. The coatings obtained by the nano-feedstock (N) present greater
microporosity and surface roughness than their counterparts,
whilst they show approximately similar or even higher micro-
hardness, when sprayed under the same parameters.
8. The porosity of the coatings follows the order CbSbN. The higher
porosity of the S coatings, as compared to that of the C coatings, is
Fig. 16. Thefracturesurfacesoftwodepositedsemi-moltenagglomeratesof
nanoparticles in the nanocoating (N
3
), revealing concentric zones of different thermal
history.
26 D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27

attributed to their lower degree of melting, due to the inherent
porosity of the powder agglomerates; the pre-spray sintering
treatment has not caused a significant decrease in porosity but
only a limited coalescence of nanoparticles.
9. The coating surface roughness is also closely linked with the
melting degree of the powder feedstocks. The surface roughness
follows the order CbSbN.
10. Despite their higher porosities, the N and S coatings exhibit
similar or even slightly higher hardness values, as compared to
the ones of the C coatings. This is explained by the strengthening
contributions of the α-phase and α-nanostructure, which coun-
terbalance the weakening contribution of porosity.
11. The adhesion strength of the coatings follows the order N bCbS.
The superior adhesion of the S coatings is ascribed to: (a) sintering
consequences (e.g., intraparticle cohesion and strain relief) and
(b) presence of randomly dispersed nanozones, acting as crack
arresters or deflectors. The inferior adhesion strength of the N
coatings is ascribed to the: (a) higher porosity, (b) lower cohesion
of the unmelted particles and (c) formation of nanozones, which
(in their majority) are not strong enough to act as effective crack
arresters.
12. The relative fracture toughness of the coatings follows the order
NbCbS. The higher fracture toughness of the S coatings is
attributed to: (a) tougher splat boundaries than those of the N
coatings, owing to a higher melting degree, and (b) crack arresting
nanozones that do not exist in the C coatings or exist in the N
coatings but, to a major degree, they are weak and non-adherent.
Two are the main reasons for the inferior fracture toughness of the
N coatings: (a) inadequate melting resulting in deficiency of
anchoring points between the deposited splats, and (b) subse-
quent high porosity that reduces the solid area that carries load.
13. The wear resistance of the coatings follows the order: CbNbS. The
S and N coatings present lower wear loss than the C ones, owing
to the retained α-nanostructure. Delamination in the N and S
coatings has been restricted by favouring stress dissipation at the
nanograin boundaries instead of stress concentration at the splat
boundaries. The wear resistance of the S coatings is further
enhanced by the high bond strength and the subsequent high
fracture toughness.
Acknowledgements
D. Zois is grateful for the financial support by the Hellenic State
Scholarship Foundation. The authors would like to thank Prof. P.
Psyllaki and A. Vazdirvanidis for their help in the wear test conduction
and SEM observations, respectively.
References
[1] K. Jia, T.E. Fischer, Wear 200 (1996) 206.
[2] H. Gleiter, Nanostruct. Mater. (1992) 1.
[3] R.W. Siegel, Mater. Sci. Forum 235–238 (1997) 851.
[4] J.R. Weertman, D. Farkas, K. Hemker, H. Kung, M. Mayo, R. Mitra, H. Van
Swygenhoven, MRS Bull. 24 (2) (1999) 44.
[5] S. Seal, S. Shukla, Nanostruct. Mater. 11 (8) (1999) 1181.
[6] G. Skandan, Nanostruct. Mater. 5 (2) (1995) 111.
[7] Y. Xie, J. Am. Ceram. Soc. 82 (3) (1999) 768.
[8] J. Karthikeyan, C.C. Berndt, J. Tikkanen, J.Y. Wang, A.H. King, H. Herman,
Nanostruct. Mater. 8 (1) (1997) 61.
[9] H.H. Nersisyan, H.I. Won, C.W. Won, J.H. Lee, Mater. Chem. Phys. 94 (2005) 153.
[10] C. Suryanarayana, Prog. Mater. Sci. 46 (2001) 1.
[11] B.H. Kear, G. Skandan, Nanostruct. Mater. 8 (6) (1997) 765.
[12] J.M. Guilemany, S. Dosta, J.R. Miguel, Surf. Coat. Technol. 201 (2006) 1180.
[13] J. Karthikeyan, C.C. Berndt, J. Tikkanen, J.Y. Wang, A.H. King, H. Herman,
Nanostruct. Mater. 9 (1997) 137.
[14] J.R. Davis, Handbook of Thermal Spray Technology, ASM International, Materials
Park, OH, 2004.
[15] D. Zois, A. Lekatou, M. Vardavoulias, I. Panagiotopoulos, A. Vazdirvanidis, J. Therm.
Spray Technol. (2008), doi:10.1007/s11666-008-9268-z.
[16] Y. Zhu, M. Huang, J. Huang, C. Ding, J. Therm. Spray Technol. 8 (2) (1999) 219.
[17] L.L. Shaw, D. Goberman, R. Ren, M. Gell, S. Jing, Y. Wang, T.D. Xiao, P.R. Strutt, Surf.
Coat. Technol. 130 (2000) 1.
[18] X. Lin, Yi. Zeng, X. Zheng, C. Ding, Surf. Coat. Technol. 195 (2005) 85.
[19] H. Chen, C.X. Ding, Surf. Coat. Technol. 1 (2002) 31.
[20] R.S. Lima, B.R. Marple, Surf. Coat. Technol. 200 (2006) 3428.
[21] J. Zhang, J. He, Y. Dong, X. Li, D. Yan, Rare Met. 26 (4) (2007) 391.
[22] J.F. Li, H. Liao, X.Y. Wang, B. Normand, V. Ji, C.X. Ding, C. Coddet, Tribol. Int. 37 (2004)
77.
[23] H. Chen, Y. Zhang, C. Ding, Wear 253 (2002) 885.
[24] E. Turunen, T. Varis, T.E. Gustafsson, J. Keskinen, T. Falt, S.-P. Hannula, Surf. Coat.
Technol. 200 (2006) 4987.
[25] X. Lin, Y. Zeng, S.W. Lee, C. Ding, J. Eur. Ceram. Soc. 24 (2004) 627.
[26] R.S. Lima, B.R. Marple, Mater. Sci. Eng. A395 (2005) 269.
[27] Y. Wang, S. Jiang, M. Wang, S.T.D. Wang Xiao, P.R. Strutt, Wear 237 (2000) 176.
[28] M. Gell, E.H. Jordan, Y.H. Sohn, D. Goberman, L. Shaw, T.D. Xiao, Surf. Coat. Technol.
146 (147) (2001) 48.
[29] E.H . Jordan, M. Gell, Y.H. Sohn , D. Goberman, L. Shaw, S. J iang, M. Wang, T.D. Xiao,
Y. Wang, P. Strutt, Mater. Sci. Eng. A301 (2001) 80.
[30] Y. Liu, T.E. Fischer, A. Dent, Surf. Coat. Technol. 167 (2003) 68.
[31] D. Basset, P. Matteazzi, F. Miani, Mater. Sci. Eng. A174 (1994) 71.
[32] H.P. Klug, L.E. Alexander, X-ray Diffraction Procedures, 2nd ed., Wiley, New York,
1974, p . 618 .
[33] B.D. Cullity, Elements of X-ray Diffraction, Addison-Wesley Inc., 1956.
[34] G.K. Williamson, W.H. Hall, Acta Metall. 1 (1953) 22.
[35] M.L. Gualtieri, M. Prudenziati, A.F. Gualtieri, Surf. Coat. Technol. 201 (2006) 2984.
[36] N. Ishizawa, T. Miyata, I. Minatio, F. Marumo, S. Iwai, Acta Crystallogr.,B B36 (1980)
228.
[37] G. Paglia, C.E. Buckley, A.L. Rohl, B.A. Hunter, R.D. Hart, J.V. Hanna, L.T. Byrne, Phys.
Rev. B68 (2003) 144110/1.
[38] F.H. Chung, J. Appl. Crystallogr. 7 (1974) 519.
[39] R.L. Snyder, D. Bish, in: D.L. Bish, J.E. Post (Eds.), Modern Powder Diffraction
(Reviews in Mineralogy Vol.20), Mineralogical Society of America, Washington D.C.,
1989 , p. 10 1.
[40] ASTM C 633, 01, Standard test method for adhesion or cohesion strength of
thermal spray coatings, Annual Book of ASTM Standards, 20 08.
[41] P. Ostojic, R. McPherson, Mater. Forum 10 (1987) 247.
[42] ASTM G99-05: Standard test method for wear testing with a pin-on-disk
apparatus, Annual Book of ASTM Standards.
[43] H. Schaper, E.B.M. Doesburg, P.H.M. De Korte, L.L. Van Reijen, Solid State Ionics 16
(1985) 261.
[44] R. McPherson, J. Mater. Sci. 15 (1980) 3141.
[45] R. McPherson, J. Mater. Sci. 8 (1973) 851.
[46] R. McPherson, Surf. Coat. Technol. 39/40 (1989) 173.
[47] H. Luo, D. Goberman, L. Shaw, M. Gell, Mater. Sci. Eng. A346 (2003) 237.
[48] M. Wang, L.L. Shaw, Surf. Coat. Technol. 202 (2007) 34.
[49] R.S. Lima, A. Kucuk, C.C. Berndt, Mater. Sci. Eng. A313 (2001) 75.
[50] S.C. Tjong, H. Chen, Mater. Sci. Eng. R 45 (2004) 1.
[51] A. Argawal, T. McKechnie, S. Seal, J. Therm. Spray Technol. 12 (2003) 350.
[52] R. Zabbarov, J. Eng. Phys. 13 (3) (1967) 373.
[53] V.V. Skorokhod, Powder Metall. Met. Ceram. 34 (1–2) (1995) 48.
[54] Z. Karoly, J. Szepvolgyi, Powder Technol. 132 (2003) 211.
[55] R.S. Lima, A. Kucuk, C.C. Berndt, Mater. Sci. Eng. A327 (2002) 224.
[56] T. Kohara, H. Tamagaki, Y. Ikari, H. Fujii, Surf. Coat. Technol. 185 (2004) 166.
[57] R.S. Lima, B.R. Marple, J. Therm. Spray Technol. 16 (1) (20 07) 40.
[58] P. Bansal, N.P. Padture, A. Vasiliev, Acta Mater. 51 (2003) 2959.
[59] B.R. Lawn, D.B. Marshall, J. Am. Ceram. Soc. 62 (7–8) (1979) 347.
[60] R.S. Lima, A. Kucuk, C.C. Berndt, Surf. Coat. Technol. 135 (2001) 166.
[61] Y. He, L. Winnubst, A.J. Burggraaf, H. Verweij, P.G.Th. Vander Varst, B. de With,
J. Am. Ceram. Soc. 80 (2) (1997) 377.
[62] J.K. Lancaster, Tribol. Int. 23 (1990) 371.
[63] X. Lin, Y. Zeng, C. Ding, P. Zhang, Wear 256 (2004) 1018.
[64] M.J. Mayo, Nanostruct. Mater. 9 (1997) 717.
[65] K. Zum Gahr, W. Bundschuh, B. Zimmerlin, Wear 162–164 (1993) 269.
27D. Zois et al. / Surface & Coatings Technology 204 (2009) 15–27