Results in Physics 12 (2019) 1574–1585

Contents lists available at ScienceDirect

Results in Physics
journal homepage: www.elsevier.com/locate/rinp

Characterisation of titanium oxide layers using Raman spectroscopy and T

optical profilometry: Influence of oxide properties
⁎
E.J. Ekoia, , A. Gowenb, R. Dorrepaalb, D.P. Dowlinga,1
a
University College Dublin, School of Mechanical and Materials Engineering, Belfield, Dublin 4, Ireland
b
University College Dublin, School of Biosystems and Food Engineering, Belfield, Dublin 4, Ireland

A R T I C LE I N FO A B S T R A C T

Keywords: This study evaluates the use of a combination of Raman spectroscopy and optical profilometry as a surface
Titanium oxidation characterisation technique for the examination of oxide layers grown on titanium metal substrates. The titanium
Raman spectroscopy oxide layers with thickness of up to 8 µm, were obtained using a low-pressure oxygen microwave plasma
Surface roughness parameters treatment of the titanium metal substrate. The effect of the microwave plasma processing conditions (input
Raman intensity
power, pressure and treatment time) on the Raman bandwidth, intensity and peak position was evaluated. Also,
Microwave plasma
Raman mapping
the effect of these processing conditions on the surface roughness parameters (Sa, Sdq, Ssk and Sku) of the oxide
layers was investigated. Analysis of the peak positions of Eg and A1g modes indicated that the effects of input
power and chamber pressure was to induce a shift towards the lower frequency with increasing input power and
pressure (1–2 kPa). The intensity of the Raman bands was found to be dependent on the morphology and surface
chemistry of the oxide layer. The intensity of Raman band (A1g), was found to be particularly influenced by the
average surface roughness (Sa) and the crystallite size. Exponential and polynomial relations were found to
correlate with these properties. A two-latent variable Partial Least Squares Regression model developed on
Raman spectral data could predict surface roughness with a coefficient of determination (R2) of approx. 0.87
when applied to the testing of an independent set of titanium oxide test coatings.

Introduction the following irreducible representation of normal vibrations:

9A1g + 9B1g + 9B2g + 9B3g + 9A1u + 8B1u + 8B2u + 8B3u [12]. 36 of
Raman spectroscopy has been found to be a sensitive characterisa- the predicted modes represented by A1g, B1g, B2g and B3g are active in
tion technique for the investigation of complex transition-metal oxides Raman whereas the rest are active in infrared except for the A1u mode
structures, such as those present at either the bulk phase or at a surface which is inactive in both Raman and infrared [12]. For anatase, there
[1]. The Raman technique is sensitive to metal-oxide vibrations, which exists 15 optical modes with the normal vibrations:
are related to the metal – oxygen bond length as well as its structural 1A1g + 1A2u + 2B1g + 1B2u + 3Eg + 2Eu [13]. Six of the modes re-
environment [1]. This paper investigates the sensitivity of Raman presented by 1A1g, 2B1g, 3Eg symmetries are Raman active [7]. In rutile,
Spectroscopy to changes in the physical features of oxide layers grown the two TiO2 units in the unit cell implies 15 vibrational modes. These
on titanium. These oxides are used in applications such as energy modes belong to the following irreducible representations [7,8]:
capture, medical devices and as catalytic surfaces [2–5]. 1A1g + 1A2g + 1A2u + 1B1g + 1B2g + 2B1u + 1Eg + 3Eu. Rutile has
Titanium dioxide (TiO2) phases are usually Raman active in the four Raman active modes with symmetries: B1g, Eg, A1g and B2g [7].
100–900 cm−1 region [6–8]. The oxide can exist in three main poly- Several authors have investigated the Raman vibration modes of ana-
morphs: anatase and rutile, which exhibit tetragonal crystal structure tase and rutile TiO2 [8,14–22]. It has been reported that molecular
and belong to the space group D419h and D414h , respectively. Brookite (or- bonds make different contributions to the Raman-active modes (by
thorhombic), the third polymorph has a lower symmetry than the influencing the intensity of the Raman modes), depending on the crystal
previous two and belongs to the space group D215h [9–11]. For brookite, planes present [8,14,22]. The Eg, B1g and the A1g Raman modes are
factor group analysis indicates the existence of 69 optical modes with mainly associated with the symmetric stretching vibration, the

⁎
Corresponding author.
E-mail addresses: emmanuel.ekoi@ucdconnect.ie (E.J. Ekoi), aoife.gowen@ucd.ie (A. Gowen), ronan.dorrepaal@ucdconnect.ie (R. Dorrepaal),
denis.dowling@ucd.ie (D.P. Dowling).
1
University College Dublin, School of Mechanical and Materials Engineering, Belfield, Dublin 4, Ireland.

https://doi.org/10.1016/j.rinp.2019.01.054
Received 24 December 2018; Received in revised form 16 January 2019; Accepted 18 January 2019
Available online 25 January 2019
2211-3797/ © 2019 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
E.J. Ekoi et al. Results in Physics 12 (2019) 1574–1585

symmetric bending vibration and the anti-symmetric bending vibration which in-turn was supported on a quartz rod. The processing conditions
of O-Ti-O, respectively [14,18]. The intensities of these Raman vibra- were systematically altered with input powers in the range 0.9 to
tion modes are dependent on the oxide’s crystallinity, which can be 2.1 kW, operating pressures between 1.0 and 5.0 kPa and treatment
altered during oxide formation due to furnace treatment parameters time between 5 and 25 min. Single test coupons were evaluated under
such as pressure and temperature [8,14]. For example, the vibrational each processing condition. The oxygen gas (99.9%) flow rate was
modes (A1g and Eg) of rutile TiO2, have been shown to exhibit sys- maintained at 100 sccm for all treatments. After each treatment, the
tematic Raman shifts, broadening and intensity reductions as the sample was left in oxygen and vacuum atmosphere for 3 and 2 min,
crystallite size decreased; effects attributed to a three-dimensional respectively, before chamber ventilation.
phonon confinement [17]. Li Bassi et al. [21] observed a correlation Temperature measurements were carried out using a LASCON
between Raman peak at 144 cm−1 (Eg mode) shift and crystal size of QP003 and LPC03 ratio pyrometers from Dr. Mergenthaler GmbH & Co.
anatase TiO2 nanoparticles (< 40 nm). The observed Raman shift to KG. The temperature generated on the titanium substrate by the dis-
lower wavenumber with increasing crystal size (particularly for sizes charge was found to be in the range 716–910 °C, depending on the
less than 10 nm) was interpreted based on a phonon quantum con- plasma processing conditions used.
finement model. Yan et al. [14] observed that the ratios between dif- The morphology of the resulting titanium oxide-layer samples was
ferent Raman vibrational modes (A1g/Eg and B1g/Eg in anatase and A1g/ characterised using a FEI Quanta 3D FEG DualBeam scanning electron
Eg in rutile) were approximately the same. However, for a mixed phase microscope (SEM) and cross-section analysis of the oxide layer thick-
structure, the ratio of the integrated Raman peak intensity of rutile peak ness obtained using a focused ion beam (FIB). Energy-dispersive X-ray
at 446 cm−1 (Eg) to that of anatase at 396 cm−1 (B1g) has been sug- spectroscopy (EDX) operating at 15 kV was used to obtain elemental
gested to yield a semi-quantitative measure of the weight ratio of rutile data at five different points on each sample. For EDX measurement
to anatase [23,24]. validation, two different EDX systems (20 mm2 Oxford Inca and 30 mm2
In this study TiO2 is grown by controllably oxidising the pure tita- Oxford AZtecOne EDX detectors) were used to obtain the elemental
nium metal using an oxygen plasma treatment. In previous studies the information. Similar results were obtained from these systems. Surface
use of both physical and chemical treatments has been investigated for roughness measurements were obtained using both a Bruker’s NT1100
this oxidation [25–31]. The physical methods include the use of and NPFLEX 3D optical profilometers in vertical scanning inter-
thermal treatment, thermal spraying, physical vapour deposition and ferometry (VSI) mode. The titanium test substrates were found to have
ion implantation techniques, while the chemical oxidation methods a roughness (Sa) of 0.01 µm.
include the use of sol–gel treatment, electrochemical treatments The phase composition and crystallinity were determined using
(anodic oxidation) and chemical vapour deposition (CVD) [2]. For ex- Siemens D500 XRD system and an inVia Micro-Raman confocal spec-
ample, Zwilling et al. [32,33] obtained oxide-layers with structure and troscopy system (Renishaw, Wotton-under-Edge, Gloucestershire, UK).
morphology either compact and thin (less than 20 nm), or porous and The XRD system operated at 40 kV and 30 mA at a wavelength of
thick (more than 50 nm depending on the electrolyte used) using 0.1541 nm (CuKα radiation). The scan wavelength for this system was
electrochemical treatments. In addition to conventional chemical in 2θ mode with a scanning range from 20° to 80° with steps of 0.02° per
treatments, there have also been a number of reports on the use of second. The average crystallite size of the samples treated with various
plasma treatments for the oxidation of titanium [34–36]. Plasmas have input power, pressure and time was estimated from the most intense
the advantage of speed of oxidation, with treatment times reduced from (1 1 0) XRD peak using the Scherrer equation [38,39]: τ = (k λ )/(β cos
hours in conventional chemical treatments to seconds or minutes. For θ), where τ is the crystallite size, k is the shape factor with a value of 0.9
example, exposure to a DC plasma yielded titanium oxide layer thick- (assuming spherical crystallites), λ is the X-ray wavelength, β is the line
nesses of 30 nm, after 5 min’ treatment [34]. Microwave plasma dis- broadening described by the full width at half-maximum (FWHM) peak
charges are usually more effective than DC plasmas, because of the height in radians (where βStructural = βObserved − βStandard), and θ is half
higher density of atomic oxygen that can be generated using this type of of the Bragg angle (in radians). Lanthanum hexaboride (LaB6) was used
discharge; and they can be operated in a wide gas pressure range [36]. as the standard for instrument broadening.
This study investigates how Raman spectra are influenced by the Raman scattering was recorded using a laser of wavelength 532 nm
physical and surface chemistry of titanium oxide layers, obtained using with a maximum laser power of 2.25 W. A 10× 0.25NA objective lens
low-pressure microwave plasma oxidation treatments under three was used to focus the laser beam on the sample surface for an exposure
process conditions (input power, pressure and treatment time). Of time of 10 s each time. Spectra were calibrated to a silicon shift at
particular interest is to investigate how the grown oxide surface 520 cm−1. The detector used was a NIR enhanced Deep Depletion CCD
roughness parameters (Sa, Sdq, Ssk & Sku), oxygen concentration and array (1024 × 256 pixels) which was Peltier cooled to −70 °C. A 1800
crystallite size influence the intensity of the associated Raman band. lines/mm grating was used for all the Raman scans. Point spectra were
obtained using the “Extended” scan option in this system, resulting in
Materials and methods spectra in the Raman shift range of 22–1806 cm−1 measured over 1015
spectral data points with an average spectral step size of 1.75 cm−1.
A microwave Circumferential Antenna Plasma (CAP) reactor oper- Baseline correction and curve fitting of the Raman spectra was carried
ating at 2.45 GHz was used for the plasma oxidation treatment of ti- out using WiRE StreamLineHR image acquisition software and
tanium substrates. The CAP system is described in detail elsewhere MATLAB. Raman maps were acquired using the same system in “static
[37]. Commercially pure (CP) titanium blocks mode” at 100% power, which resulted in a slightly different spectral
(250 mm × 150 mm × 8 mm) were faced and cut into discs (diameter range of 90–1860 cm−1 with a mean spectral size of 1.75 cm−1 over
25 mm and thickness 5 mm), using a Hurco VM3 CNC mill system. 1015 measured spectral bands. An area corresponding to 4 × 4 mm at
These discs were ground and polished to a mirror finish using silicon the centre of sample 3 was mapped. Spectra were acquired at steps of
carbide abrasive paper (240, 320, 600 and 800), diamond suspension (6 50 μm in the X and Y mapping directions. The exposure time per pixel
and 3 µm) and colloidal silica. They were subsequently cleaned in an was 1 s, and the total acquisition time was approximately 100 min.
ultrasonic bath using methanol and deionised water; then blow-dried
with air prior to plasma treatment. The microwave plasma process in- Partial least squares regression
volved a five-minute pump down to a base pressure of 8 Pa, before the
introduction of oxygen into the chamber. The pressure was allowed to Non-Linear Iterative Partial Least Squares (NIPALS) regression [40]
rise to 0.6 kPa, before igniting the microwave discharge. The titanium was used in order to predict measured surface roughness from the
disc was mounted in the plasma using a molybdenum metal holder, Raman spectra. The model was trained and calibrated using Raman

1575
E.J. Ekoi et al. Results in Physics 12 (2019) 1574–1585

spectra from five samples and then tested on an independent validation plasma electrons/photons leads to the excitation and mass transport of
set of spectra from five separate TiO2 coatings. The trained PLS re- metal ions towards the surface [42]. It is also reported that there is a
gression vector was applied to the full NIR images and compared with build-up of electrons at the metal surface such that a negative potential
the k-means results. In order to remove the baseline effects, a second is raised, and thus, this attracts positive ions to the surface to form
derivative Savitsky Golay pretreatment was applied to the Raman neutrals [43]. These oxygen neutrals can react with Ti ions to form
spectra. oxide nuclei. Depending on the input power and pressure used during
Prior to model development the point spectra and mapping data treatment, the number density of oxygen atoms in the plasma lead to
were trimmed to the Raman shift range 100–1086 cm−1 and the map- either the creation of high or low number of nucleation sites [44]. It is
ping data were interpolated to match the exact Raman shifts of the likely that the nucleation density controls both the roughness and the
point spectra. In addition, differences in Raman peak intensity due to porosity of the titanium-oxide layer formed on the titanium substrate by
the use of the “extended” scan mode in the spectral acquisition and inducing a Volmer-Weber type of growth mode [45] in which the oxide
“static” scan mode in the map acquisition were removed by scaling the nuclei grow as islands until coalescence occurs. At low input power and
Raman map data to the ratio of mean Raman spectra of static and ex- pressure, a dense oxide structure which reduces the diffusion of the
tended modes at 506 cm−1. Tin+ ions through the growing oxide layer to the oxide-plasma interface
is obtained. However, with increasing input power and pressure, a
Results and discussion change in the oxide structure; as evidenced by the dominant vertical
growth of the grains at the top regions of the oxide, is observed. It can
In this work, the influence of microwave plasma processing condi- be observed that as the thickness of the oxide layer increased, there
tions of input power, chamber pressure and treatment time on the appeared to be a critical thickness (influenced by porosity level) above
roughness and Raman spectra of the oxide layer obtained were sys- which a more porous structure is observed (Fig. 1). The difference in
tematically investigated. The input power processing condition was the grain structure of the bulk and top regions of the oxide layer may
investigated in the range of 0.9–2.1 kW at a pressure and treatment have become necessary due to the dense nature of the bulk oxide. It has
time of 4.6 kPa and 10 min, respectively. A pressure in the range of been reported by other researchers who used microwave plasma system
1–5 kPa was investigated at a constant input power and treatment time to grow diamonds that it is possible that not all the grains which were
of 1.2 kW and 10 min, respectively. Finally, the influence of treatment formed during nucleation survived to reach the surface due to a van der
time was investigated in the range of 5–25 min at constant input power Drift type of competition [44]. Only the grains which grew perpendi-
and pressure of 1.2 kW and 4.6 kPa, respectively. cular to the substrate surface would survive. Others would be buried
As shown in Fig. 1 the oxygen microwave plasma oxidised metal during the growth process further limiting diffusion. This would ac-
exhibits a highly porous morphology both at the surface and in the bulk count for the differences in the oxide layer structure observed during
of the oxide. Examination of the oxide layer revealed that it was com- microwave oxygen plasma oxidation. A typical schematic mechanism
posed of a less porous structure in the lower region and a columnar on how the oxygen microwave plasma process induced high porosity in
porous structure in the upper region. The morphology was mostly in- the oxide film is given in Fig. 2. Another observable feature of the
fluenced by high input power and pressure. For instance, at an input microwave plasma obtained oxide is the presence of well-defined
power and pressure of 0.9 kW and 1 kPa, respectively, the oxide films crystal facets. The thickness of the oxide layer shown in Fig. 1 is about
were relatively compact. This morphology is different from what would 10 µm. This thick oxide was grown in 10 min using an input power and
be obtained in a conventional furnace treatment in that it exhibited a chamber pressure of 2.1 kW and 4.6 kPa, respectively.
relatively rough structure, with large grains. This is in contrast to that
reported for furnace oxidation treatments of titanium, where the more Effect of microwave plasma processing conditions on the phase composition,
slowly grown oxide exhibits a much denser morphology, with little or Raman peak position and linewidth of oxide layers
no porosity [36]. A possible explanation is that the first stage of the
oxidation process involves the adsorption of oxygen on the titanium Fig. 3 shows the Raman spectra of TiO2 layer grown with the var-
surface as a function of the available oxygen species [41]. Compared to ious microwave plasma processing conditions discussed in Section
air furnace treatment, microwave plasma oxidation is more effective “Results and discussion”. The thickness of the oxides grown at the
because of the higher density of oxygen species that can be generated. various processing conditions is also shown. As can be observed, the
Associated with this level of oxygen species is the high rate of bom- thickness generally increased with increased input power and pressure.
bardment of the growing film by oxygen species obtainable. It has been The thickness of the oxide layers increased for treatment times of
reported previously that the energy impacted to the substrate by the 10–20 min. Above 20 min treatment period, a reduction in thickness

Oxide Layer
Thickness

Fig. 1. SEM images of a typical morphology and thickness of oxide grown in an oxygen microwave plasma. The thickness and growth temperature of the oxide layer
were 9.89 µm and 910 °C, respectively.

1576
E.J. Ekoi et al. Results in Physics 12 (2019) 1574–1585

Low input Stranski-Krastanov (SK) High input

power Volmer-Weber (VM) Mode
Mode power

Frank-van der Merwe (FM) Stranski-Krastanov (SK)

Low pressure High pressure
Mode Mode

Short treatment Stranski-Krastanov (SK) Stranski-Krastanov (SK) Long treatment

Ɵme Mode Mode Ɵme

Fig. 2. A typical schematic mechanism on how oxygen microwave plasma processing conditions induce porosity in the oxide film.

was observed. This reduction in thickness may indicate “oxide thickness a redshift (Fig. S2). The spectra obtained from titania coatings from this
saturation” [36]. It has been reported that rutile crystal exhibits work are also compared to previous results in Table 1 [9–11,36]. As
dominant peaks at 446.6 and 609.8 cm−1 [36] (This is supported by the shown in Table 1, there is good agreement with examples of previous
XRD results of Fig. 4 which confirms that the oxide phase was mostly reports in the literature.
that of rutile). These Raman peaks have been assigned to two of the four In order to investigate further the structure, phase composition and
Raman active modes of rutile single crystal (Eg and A1g respectively). transformation in the oxides as a function of processing conditions, XRD
The intense Eg peak supports the increasing crystallinity of the samples profiles of the region enclosed in rectangle in Fig. 4 is presented in
as the treatment conditions are altered. The sharp low-frequency line at Fig. 7 and depicts the influence of pressure on the oxide phase com-
143.2 cm−1 and the weak high-frequency line at 800.27 cm−1 are as- position and transformation. The observed peaks in the diffraction
signed to the remaining of the four active modes (B1g and B2g respec- profile between 37 and 41° (using JCPDS-ICDD card numbers 10-63,
tively). The peak at 800.27 cm−1 is the weakest and can only be ob- 11-217, 11-218, 11-474, 18-1401, 18-1402, 18-1403, 18-1404, 21-
served on some sample profiles. The peak became more pronounced 1272, 21-1276, 23-606 and 40-806) suggest the presence of phases
with increase in input power from 0.9 to 2.1 kW and treatment pressure different from rutile. The presence of Ti2O, Ti3O5 and some members of
from 1.0 to 5.0 kPa (Fig. 3(a) and (b), respectively. There was no sig- the homologous series TinO2n−1 (where n is a number between 4 and
nificant change observed in the intensity of this peak for samples pro- 10) [50] phases was observed. Phase transformation could be observed
cessed in the range of 5–25 min plasma treatment time. The broad peak with increasing chamber pressure (2–5 kPa). The double peak struc-
at 232.4 cm−1 may be attributed to either disorder induced scattering tures observed between 37 and 41° for samples treated at a pressure of
or second-order effect [46], which is a result of multi-phonon processes. 1.0 kPa in Fig. 7 represent mixed phases of titanium metal, Ti-O solid
It was noted that this peak became more intense for the oxide layers solution and other phases which with Rutile, Ti2O, Ti3O5 and some
deposited at higher input power and chamber pressures. This effect, members of the homologous series TinO2n−1 becoming the dominant
however, was not found to correlate as strongly with changes in plasma phase with increasing pressure in this region. These phases would in-
treatment time (Fig. 3(c)). For samples grown using input power in the fluence the level of conductivity and semiconductor-to-metal transi-
range of 0.9–2.1 kW, there was a 1-fold increase and decrease in the tions in the oxide layers [50,51]. Similar results were obtained when
linewidth of Eg and A1g bands, respectively, with increasing input input power and treatment time were varied while other parameters
power. For samples processed in the range of 5–25 min, both Eg and A1g remained constant, however, mixed phases of titanium metal and Ti-O
band linewidths decreased by 1-fold. A typical example of the effect of solid solution were absent at the lowest treatment conditions (See Figs.
processing conditions on the Eg and A1g band widths is given in Fig. 5 S3 and S4). It can be observed that in all the samples treated, the peaks
and shows the influence of pressure on the band widths. In this case, the representing the mixed phases decreased with increasing input power,
linewidths decreased and increased for Eg and A1g bands, respectively, pressure and time indicating a transition in phase from a mixed phase to
as the pressure changed from 1 kPa to 2 kPa. It has been suggested that that of purely rutile phase structure (Figs. 4, S3 and S4). The implica-
peak broadening occurs due to nanocrystallinity and phonon confine- tion of the phase transition as a function of process conditions is a
ment effects, and that there was a characteristic dependency between transition from a defect structure dominated by the Magnéli phases to
grain size and peak position and broadening in Raman analysis one mostly containing point defects such as oxygen vacancies and ti-
[47–49]. Analysis of the peak positions of Eg and A1g modes indicates tanium interstitials [52–54]. The presence of oxygen vacancies at the
that the effects of input power and chamber pressure was to induce a oxide surface would influence the adsorption of molecules [54].
shift towards the lower frequency with increasing input power and
pressure (1–2 kPa) (Figs. 6 and S1). However, as the pressure increased
to the range of 3 to 5 kPa, there was no significant redshift observed Influence of processing conditions on the oxygen concentration and intensity
(Fig. S1). Samples processed for 5–15 min induced a slight shift towards of the A1g Raman band
the higher frequency for the Eg band; whereas for the A1g band, there
was no significant effect on the Raman peak positions. For samples Fig. 8 shows a typical morphology of oxide layers grown at various
processed using more than 15 min, both the Eg and A1g bands exhibited microwave plasma processing conditions. As can be seen, grain size and
porosity progressively changed with increased input power and

1577
E.J. Ekoi et al. Results in Physics 12 (2019) 1574–1585

(a) 447 610 cm¯¹

cm¯¹ (Eg) (A1g) Oxide Layer Thickness
0.9 kW 0.77 ± 0.12 μm

Raman Intensity (a.u)

1.2 kW 2.89 ± 0.14 μm
1.5 kW 5.82 ± 0.21 μm
232 cm¯¹
nd
(2 Order) 1.8 kW 6.96 ± 0.21 μm
143 cm¯¹ 2.1 kW 8.05 ± 0.08 μm
(B1g)

0 250 500 750 1000 1250 1500 1750

Raman shi (cm-1)

(b) 447 cm¯¹ 610 cm¯¹ (A )

1g
(Eg) Oxide Layer Thickness
1.0 kPa 1.26 ± 0.18 μm
2.0 kPa 1.14 ± 0.04 μm
232 cm¯¹
nd 3.0 kPa 1.99 ± 0.10 μm
(2 Order)
Raman Intensity (a.u)

4.0 kPa 3.62 ± 0.18 μm

143 cm¯¹
(B1g) 5.0 kPa 4.63 ± 0.28 μm

0 250 500 750 1000 1250 1500 1750

Raman shi (cm-1)

(c) 447 cm¯¹ 610 cm¯¹ Oxide Layer Thickness

(Eg) (A1g)
5 minutes 3.68 ± 0.17 μm
10 minutes 3.12 ± 0.19 μm
Raman Intensity (a.u)

232 cm¯¹ 15 minutes 4.16 ± 0.24 μm

nd
(2 Order)
143 cm¯¹ 20 minutes 4.41 ± 0.20 μm
(B1g) 25 minutes 3.58 ± 0.15 μm

0 250 500 750 1000 1250 1500 1750

Raman shi (cm-1)
Fig. 3. Raman spectra of samples treated at: (a) input power of 0.9–2.1 kW; (b) pressure of 1.0–5.0 kPa; (c) treatment times of 5–25 min.

pressure. The change in the oxide structure may be related to the Fig. 9. It must be noted that although two different EDX measurement
oxygen concentration in the growing oxide layer as a function of mi- systems were used to characterise, verify and validate measured oxygen
crowave plasma processing conditions. A typical example of the influ- concentration in these oxides, the O/Ti ratio was in the range of
ence of pressure on the oxygen concentration (determined using EDX (1.92 ± 0.04–2.22 ± 0.05), (2.30 ± 0.03–2.18 ± 0.06) and
analysis) and the Raman spectra intensity of the A1g band is shown in (2.33 ± 0.03–2.37 ± 0.04), respectively, for pressure, input power

1578
E.J. Ekoi et al. Results in Physics 12 (2019) 1574–1585

(110)

(101) (202)
(211) (112)

(111)
(310)

(200)
(220) (301)
5.0 kPa (210) (002) (311)
Intensity (a.u.)

Treated Titanium
4.0 kPa

3.0 kPa

2.0 kPa

1.0 kPa

Untreated Titanium

15 25 35 45 55 65 75 85
2ࣝ (Degree)

Fig. 4. XRD diffraction patterns obtained for untreated and plasma oxidised Fig. 7. Region in the XRD profile of Fig. 4 showing the oxide phase transfor-
titanium using microwave plasma pressure of 1 kPa–5 kPa. The peaks indicate a mation with the increase in pressure from 1 kPa during plasma oxidation
Rutile phase of TiO2. treatments.

42 and treatment time processing conditions. A Possible reason could be

41 the presence of excess oxygen in the oxide structure which would result
40 in the formation of interstitial oxygen [55,56]. This could be observed
in Fig. 9 where at a pressure of 1 kPa, the O/Ti ratio was about 1.92,
Linewidth (cm-1)

39
447 cm-1 however, above this pressure, the ratio was 2.23. The explanation is
38 610 cm-1 that at higher oxygen pressure, there are more oxygen species bom-
37 barding the film; and more oxygen condensed onto the oxide film when
36 the microwave plasma source was turned off after treatment [55,56].
The effect is the formation of interstitial oxygen which could have
35
combined with the lattice oxygen to form either a substitutional or
34
0 1 2 3 4 5 6
interstitial O2 [56–58]. This would result in either oxide crystal lattice
Chamber Pressure (kPa)
strain due to the slight outward movement of the neighbouring Ti
atoms or oxygen vacancy defects [55–57]. The observed Raman in-
Fig. 5. Influence of pressure on the linewidth of the Eg (447 cm−1) and A1g tensity trend as a function of pressure may be related to the number of
(610 cm−1) bands. Ti-O molecules available for excitation during Raman scan as discussed
above. Similar results were obtained for samples investigated using
610.8 input power and treatment time conditions (Figs. S5 and S6).
610.7
A1g Mode Raman Shi (cm-1)
Eg Mode Raman Shi (cm-1)

447.6 447 cm-1 610 cm-1

447.4 610.6
610.5
447.2 Influence of crystallite size on Raman peak position and intensity
610.4
447
610.3
446.8
610.2
The crystallite size of the oxides was estimated from the XRD profile
446.6
610.1 of the samples, using the Scherrer’s equation. The 110 XRD peak of TiO2
446.4 610 was used for this estimation. The estimated crystallite sizes obtained
446.2 609.9 were plotted against the Raman intensity of the A1g band. The result is
446 609.8 presented in Fig. 10. There appears to be a correlation between the
0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 crystallite size and the intensity of the Raman spectra obtained. This
Input Power (kW) plot was carried out without consideration of the microwave plasma
Fig. 6. Effect of input power on the position of the peaks for the Eg (447 cm−1) treatment conditions used to obtain the oxides. It was observed that the
and A1g (610 cm−1) Raman bands. Raman peak positions of A1g and Eg bands exhibited a strong correla-
tion with the oxide crystallite size as demonstrated in Fig. 11. The peak
positions tend to shift towards lower wavenumbers with increasing
Table 1
crystallite size. A similar correlation between Raman peak at 144 cm−1
Comparison of the Rutile band obtained in this study with works from other
researchers. (Eg mode) shift and crystal size was observed by Li Bassi et al. [21] for
anatase TiO2 nanoparticles (< 40 nm) and interpreted based on a
This Study & Ekoi et al. [36] Ma et al. Arsov et al. Assigned Mode
phonon quantum confinement model. The red shift of the Eg rutile
[9] [10] [11]
mode has been reported to be a result of the nonstoichiometric effects
143.2 ± 0.3 140.2 143 B1g for TiO2 nanophase [59] although other authors have disagreed with
232.4 ± 0.9 235.5 236 Multi-phonon this correlation [20]. It is important to note that in the present work,
process that the estimated grain size of 76–315 nm was larger than the sizes
446.6 ± 0.3 445.8 447 Eg
609.8 ± 0.4 609.8 612 A1g
reported by these two groups which on average was not larger than
– 825.5 826 B2g 12 nm [20,59]. Thus, factors which influence the Raman shift and in-
tensity are the grain size range, the grain size of the oxide, along with
the roughness.

1579
E.J. Ekoi et al. Results in Physics 12 (2019) 1574–1585

A B

C D

Fig. 8. Typical SEM images of oxidised layers demonstrating the influence of increased input power and chamber pressure on oxide morphology and pore structure:
(A) 1.2 kW; (B) 2.1 kW; at a fixed pressure of 4.6 kPa, (C) 2.0 kPa; (D) 5.0 kPa at a fixed input power of 1.2 kW.

70 25 45
Raman Absolute Intensity (counts)

40
Raman Intensity (Counts) x103

69
20 35
Oxygen Conc. (At%)

68 30

67 15 25
x103

20
66 10 15

65 10

Oxygen Concentra on (At %) 5 5

64
Raman Intensity (Counts) 0
63 0 60 110 160 210 260 310 360
0 1 2 3 4 5 6 Crystallite Size (nm)
Chamber Pressure (kPa) Fig. 10. Plot of Raman absolute intensity against the estimated crystallite size
Fig. 9. Effect of processing pressure on the oxygen concentration (At % weight) of the treated samples.
and the intensity of the A1g Raman band.
lack of peaks. Kurtosis parameter Sku describes the probability density
Influence of microwave processing conditions on the roughness parameters sharpness of the profile and represents the peakedness degree of a
and pore structure of oxide layers surface height distribution. An Sku value of < 3 implies that the surface
has mostly few high peaks and low valleys while a value > 3 implies a
The evaluation of surface roughness is important in applications surface with many high peaks and low valleys [60].
related to friction, contact deformation, heat and electric conduction, The influence of microwave plasma processing conditions on the
contact joints, positional accuracy [60] as well as surface enhanced commonly used 3D roughness parameters of the oxide layers are pre-
scattering [61,62]. The average surface roughness (Sa) is the most sented. The effect of input power on the amplitude (Average surface
commonly used roughness parameter and provides a very good overall roughness Sa, Skewness Ssk; and Kurtosis Sku) and hybrid (Root-Mean-
height variation, although its lack of sensitivity to small changes in Square Surface Slope Sdq) parameters is shown in Fig. 12. It can be
profile and wavelength information has been reported [60,63]. The observed that as the input power increased during processing, the
skewness parameter Ssk is sensitive to occasional deep valleys or high average surface roughness amplitude (Sa) and hybrid (Sdq) parameters
peaks. A skewness value of zero implies a symmetrical height dis- increased (Fig. 12(A)). Similar trends were obtained for processing
tribution, whereas positive skewness describes a surface with filled conditions of pressure and time except for the Sdq values obtained
valleys or high peaks. A negative skewness represents a surface with during treatment time which did not show any change. These results
are presented in the Supplementary material (Fig. S7). In order to

1580
E.J. Ekoi et al. Results in Physics 12 (2019) 1574–1585

611
low and high peaks) found in the sample processed at 0.9 kW to one
A1g
610.8
Eg 447.4 whose surfaces indicated positive skewness. The positive skewness in-
Linear (A1g) dicates that the morphology of the samples processed in the range
A1g Mode Raman ShiŌ (cm -1)

Eg Mode Raman ShiŌ (cm -1)

610.6
447.2
Linear (Eg) 1.5–2.1 kW consisted of mostly high peaks and large pore structure. The
610.4
447 value of the kurtosis parameter Sku were all above 3 for all samples
610.2
446.8
treated in the range of 0.9–2.1 kW indicating that the oxide surfaces
610 contained mostly high peaks and low valleys. Similar results were ob-
446.6
609.8 tained for samples processed at 5 to 25 min using an input power and
446.4 pressure of 1.2 kW and 4.6 kPa, respectively (see Figs. S8 and S9). For
609.6
the samples treated using a pressure in the range of 1–5 kPa at 1.2 kW
609.4 446.2
over a treatment period of 10 min; the skewness transitioned from a
609.2 446 negative to positive value at a pressure range of 1–2 kPa. This gradually
50 100 150 200 250 300 350
changed back to a negative skewness value with increasing treatment
Crystallite Size (nm)
pressure indicating that the chamber pressure was the dominant factor
Fig. 11. Correlation between the positions of the rutile A1g (filled black circles) influencing the growth rate of the oxide layer (Fig. 13). The kurtosis
and Eg (open circles) peaks and TiO2 crystallite size. values indicate that the surface of the samples processed at low pressure
(1–2 kPa) contained higher percentage of sharp peaks than the ones
understand the influence of microwave plasma input power on the processed at higher pressures. Another useful indication is that low
oxide surface evolution, the values of Ssk and Sku is presented for pressure treatments resulted in surfaces with a high morphological
samples processed in the range of 0.9–2.1 kW at a pressure and time of variation compared to the oxide surfaces obtained at higher pressure
4.6 kPa and 10 min, respectively (Fig. 12(B)). By considering the where a relatively homogeneous morphology was obtained (Fig. S10).
skewness parameter Ssk, it can be observed a change in morphology The discussed effect of the morphological variation in the samples
consistent with a transformation from a combination effect (where the treated using various processing conditions can be observed on the level
surfaces were composed of a combination of small and large pores; and of intensity of the Raman spectra (Fig. 3) as the processing conditions

Average Surface Roughness Sdq (Degrees)

Average Surface Roughness Sa (nm)

800 90

700 Sa Sdq 80

600 70
60
500
50
400
40
300
30
200 20
100 10
0 0
0.9 1.2 1.5 1.8 2.1
Input Power (kW)

B 8
7
Ssk Sku
Roughness Parameter Values

6
5
4
3
2
1
0
0.9 1.2 1.5 1.8 2.1
-1
-2
Input Power (kW)
Fig. 12. Effect of microwave plasma input power on the roughness parameters (Sa, Sdq, Ssk and Sku) of the oxide layer.

1581
E.J. Ekoi et al. Results in Physics 12 (2019) 1574–1585

70 3

60 2.5

Roughness Parameter Sku

Roughness Parameter Ssk

Ssk Sku
50 2
1.5
40
1
30
0.5
20
0
10 -0.5
0 -1
1 2 3 4 5
-10 -1.5
Chamber Pressure (kPa)
Fig. 13. Influence of treatment pressure on the roughness parameters (Ssk & Sku) of oxide layer samples processed in the range of 1–5 kPa.

are altered. This will be discussed further in subsequent sections. correlation between the Raman peak intensity and the average surface
roughness (Sa). A plot of Raman peak intensity (both absolute and
baseline corrected intensities) against average surface roughness (Sa) of
Influence of roughness parameter (Sa) on Raman intensity the 10 oxide samples investigated was found to be independent of
treatment conditions (Fig. 14). It is unclear the mechanism(s) of the
It can be observed in Figs. 3 and 5 that the peak intensities and line correlation observed between Raman intensity and roughness of the
width are clearly influenced by the microwave plasma process condi- oxide layers, however, reports have shown that enhanced Raman
tions. Also, the surface roughness parameters showed a correlation with scattering in semiconductor nanostructures may be attributed to con-
the processing conditions as discussed in Section “Influence of micro- tributions from a number of resonance effects, namely, surface
wave processing conditions on the roughness parameters and pore plasmon, exciton, charge-transfer and molecular resonances
structure of oxide layers”. The dispersion of the measured intensity as a [69,71,73].
function of the focusing point on the surface and the reproducibility of In order to further investigate the correlation between average
the measured values was considered by carrying out scans at five dif- surface roughness and Raman absolute intensity, titanium samples were
ferent regions on a sample (three measurements at each region) and oxidized using input powers of 1.5 kW and 1.8 kW, with three samples
then averaging the results. Thus, the change in intensity (as process evaluated at each power. Each test sample after being examined using
conditions change) may be due to change in surface roughness and Raman Spectroscopy, was ground to reduce its roughness and then re-
related open pore morphology of the oxide samples [61,62,64,65]. It examined. Samples grown at 1.5 kW did not indicate any significant
has been reported that the physical structure of some surfaces can be change in Raman intensity when the average surface roughness was
related to the intensity of Raman scattering; and this phenomenon is reduced by grinding. However, samples grown at 1.8 kW showed a 16%
currently applied in the detection of molecules by Surface Enhanced increase in Raman intensity when the average surface roughness of the
Raman Scattering (SERS) method [66–72]. The SERS enhancements samples was reduced. Examination of roughness parameters such as the
generally depend on the features of the substrates, such as surface amplitude (other than Sa), spatial, hybrid and functional parameters,
morphology, size, surface roughness and porosity [61]. In this study, which are related to the overall height, frequency, combination of
the A1g band was used to study the effect of average surface roughness height and frequency, and function application of surface features, re-
(Sa) on the intensity of the Raman spectra obtained for the treated spectively, revealed changes in the values of these parameters after
samples. The reason for the use of the average surface roughness (Sa) as grinding. Thus, the observed changes in the Raman intensity of oxide
the representative roughness parameter lies in the fact that it the most layers after grinding was influenced by the change in the Sa values as a
commonly used roughness parameter in literature and it is easily function of other roughness parameters. This confirms the sensitivity of
comprehensive. The dependence of the Raman peak intensity on the Raman spectra intensity to roughness of oxide layer prepared by a low-
average surface roughness (Sa) is shown in Fig. 14. There is a clear pressure microwave plasma. It must be stated, however, that the
technique of manually grinding down the oxide surface did not produce
45 0
Absolute Raman Intensity a surface with reproducible roughness parameter values. Thus, the
40 -50 above stated effect of manual grinding is mostly qualitative.
2n Deriva ve Preprocessed Raman

2n Deriva ve Preprocessed
Absolute Raman Intensity x103

35 Raman Intensity -100

30 -150 Prediction of surface roughness from Raman spectra using multivariate
25 -200 partial least squares regression (PLSR)
Intensity

20 -250
15 -300 Prediction of surface roughness from point Raman spectra
10 -350 The Raman spectra intensity of the samples used in this study was
correlated to their surface roughness (Sa) as discussed in Section
5 -400
“Influence of Roughness Parameter (Sa) on Raman Intensity”. This is
0 -450
0 100 200 300 400 500 600 700
supported by the Raman spectra of representative samples which
Surface Roughness Sa (nm)
clearly show that the intensity of the main peaks increased with in-
creasing roughness (Fig. 15). The second derivative pretreated spectra
Fig. 14. The relationship between average surface roughness (Sa) and Raman are shown in Fig. S11. Clearly the baseline has been removed and the
intensity of the A1g band independent of microwave plasma treatment condi- major peaks at the Eg and A1g bands are visible as negative peaks in the
tions. second derivative spectra. In order to predict the roughness of the

1582
E.J. Ekoi et al. Results in Physics 12 (2019) 1574–1585

Fig. 15. Raman spectra of typical samples showing correlation between intensity and surface roughness Sa.

small number of pixels were predicted to have considerably large

Roughness values, i.e. 0.19% of pixels had a predicted Roughness >
300 nm. Examination of the corresponding spectra revealed spectral
features consistent with fluorescence, which may have been due to the
presence of small dust particles or other organic contamination on the
sample surface.

Conclusions

This study evaluated how the physical features of an oxide layer of

titanium metal processed using various oxygen microwave plasma
conditions (input power, treatment time and pressure) influenced the
linewidth, peak position and intensity of the Raman bands (Eg and A1g)
obtained. It was concluded that the intensity of the representative band
(A1g) of the TiO2 was influenced by the average surface roughness (Sa)
irrespective of processing condition and the oxide crystallite size. Based
on the review of the effect of the physical features of the oxide layer
processed using various microwave processing conditions such as input
power, chamber pressure and time it was concluded that:
Fig. 16. Multivariate PLS model to predict surface roughness.
– The Titanium dioxide obtained using the microwave plasma treat-
samples base on their Raman intensity, point spectra were taken at five ment was mostly rutile.
regions (C, R1, R2, R3, & R5) on each sample which resulted in 50 data – The presence of a low concentration of Ti2O, Ti3O5 and some
points. These were analysed and a PLS model built on half of the total members of the homologous series (TinO2n−1) was also observed in
number of 10 samples evaluated, then used to predict the roughness of the oxide structure.
the rest of the samples. As shown in Fig. 16, a two-latent variable model – The processing conditions influenced the concentration of oxygen in
on raw data can predict roughness with a coefficient of determination the oxide layers.
(R2) of approx. 0.87 when applied to the independent group of oxide – A 1-fold increase and decrease in the linewidth of Eg and A1g bands
test samples. This compares well with a single channel prediction using was observed, respectively, with increasing input power. For sam-
only the 608 cm−1 band (R2 approx. 0.81). When the PLS model was ples processed in the range of 5–25 min, both Eg and A1g band
constructed on second derivative spectra, the model performance was linewidths decreased by 1-fold. In this case of pressure, the line-
lower than for the models built on absolute Raman intensity, leading to widths decreased and increased for Eg and A1g bands, respectively,
a R2 of 0.82 on the prediction set (Fig. S12). For this reason, subsequent as the pressure changed from 1 kPa to 2 kPa.
model validation on the mapping data was carried out using the ab- – Analysis of the peak positions of Eg and A1g modes indicated that the
solute Raman intensity [74] effects of input power and chamber pressure was to induce a shift
towards the lower frequency with increasing input power and
pressure (1–2 kPa). However, as the pressure increased to the range
Prediction of surface roughness using Raman mapping data of 3–5 kPa, there was no significant redshift observed. Samples
The PLSR model developed in the previous section was applied to processed for 5–15 min induced a slight shift towards the higher
the Raman map obtained over an area of 4 × 4 mm on a sample which frequency for the Eg band; whereas for the A1g band, there was no
was treated using an input power of 0.9 kW and at a pressure 4.6 kPa. significant effect on the Raman peak positions. For samples pro-
The measured surface roughness of this sample based on the optical cessed using more than 15 min, both the Eg and A1g bands exhibited
profilometry technique ranged from 142 to 217 nm, while the mean and a redshift.
standard deviation of roughness predicted from the Raman map were – The crystallite size influenced the intensity and the band position of
179 and 29 respectively, indicating the accuracy of the PLSR model. the Raman spectra.
The distribution of the predicted surface roughness values over the – There was a correlation between the average surface roughness (Sa)
sampled area are shown in Fig. 17, where the prediction map and its of the oxides and the Raman intensity observed. The correlation was
corresponding histogram are visualised. The prediction map enables found to be largely independent of the microwave plasma proces-
variation in the Roughness over the sample surface to be visualised. sing conditions used to obtain the oxide.
Overall, the distribution of Roughness values is symmetric, however, a – The relationship between surface roughness and Raman scattering

1583
E.J. Ekoi et al. Results in Physics 12 (2019) 1574–1585

Fig. 17. Prediction map enables variation in the Roughness over the sample surface to be visualised (left hand side) and histogram of prediction map (right hand side)
obtained by applying PLSR model to Raman area map of the TiO2 layer (area size 4 × 4 mm with 50 µm step size).

was further explored by the development of Partial Least Square [11] Ohsaka T, Izumi F, Fujiki Y. Raman spectrum of anatase, TiO2. J Raman Spectrosc
Regression model to predict surface roughness. The resulting model 1978;7:321–4. https://doi.org/10.1002/jrs.1250070606.
[12] Tompsett GA, Bowmaker GA, Cooney RP, Metson JB, Rodgers KA, Seakins JM. The
was found to facilitate accurate prediction of roughness, with a Raman spectrum of brookite, TiO2 (Pbca, Z = 8). J Raman Spectrosc
coefficient of determination (R2) of approx. 0.87, when applied to 1995;26:57–62. https://doi.org/10.1002/jrs.1250260110.
the independent test set. The developed model, when applied to a [13] Sekiya T, Ohta S, Kamei S, Hanakawa M, Kurita S. Raman spectroscopy and phase
transition of anatase TiO2 under high pressure. J Phys Chem Solids
Raman map of the surface of a sample, provided information on the 2001;62:717–21. https://doi.org/10.1016/S0022-3697(00)00229-8.
spatial distribution of roughness over the sample surface while also [14] Yan J, Wu G, Guan N, Li L, Li Z, Cao X. Understanding the effect of surface/bulk
indicating the accuracy of the model. defects on the photocatalytic activity of TiO2: anatase versus rutile. Phys Chem
Chem Phys 2013;15:10978. https://doi.org/10.1039/c3cp50927c.
[15] Lan T, Tang X, Fultz B. Phonon anharmonicity of rutile TiO2 studied by Raman
Acknowledgments spectrometry and molecular dynamics simulations. Phys Rev B
2012;85:094305https://doi.org/10.1103/PhysRevB.85.094305.
[16] Aita CR. Raman scattering by thin film nanomosaic rutile TiO2. Appl Phys Lett
The authors would like to acknowledge the support of SFI funded I-
2007;90:213112https://doi.org/10.1063/1.2742914.
Form Advanced Manufacturing Research Centre 16/RC/3872. The [17] Swamy V, Muddle BC, Dai Q. Size-dependent modifications of the Raman spectrum
second and third author acknowledge funding from the EU FP7 under of rutile TiO2. Appl Phys Lett 2006;89:163118https://doi.org/10.1063/1.2364123.
the European Research Council Starting Grant programme (ERC-SG- [18] Zhang Y, Harris CX, Wallenmeyer P, Murowchick J, Chen X. Asymmetric lattice
vibrational characteristics of rutile TiO2 as revealed by laser power dependent
335508). Raman spectroscopy. J Phys Chem C 2013;117:24015–22. https://doi.org/10.
1021/jp406948e.
Appendix A. Supplementary data [19] Jiang X, Zhang Y, Jiang J, Rong Y, Wang Y, Wu Y, et al. Characterization of oxygen
vacancy associates within hydrogenated TiO2: a positron annihilation study. J Phys
Chem C 2012;116:22619–24. https://doi.org/10.1021/jp307573c.
Supplementary data to this article can be found online at https:// [20] Mazza T, Barborini E, Piseri P, Milani P, Cattaneo D, Li Bassi A, et al. Raman
doi.org/10.1016/j.rinp.2019.01.054. spectroscopy characterization of TiO2 rutile nanocrystals. Phys Rev B
2007;75:045416https://doi.org/10.1103/PhysRevB.75.045416.
[21] Li Bassi A, Cattaneo D, Russo V, Bottani CE, Barborini E, Mazza T, et al. Raman
References spectroscopy characterization of Titania nanoparticles produced by flame pyrolysis:
The influence of size and stoichiometry. J Appl Phys 2005;98:074305https://doi.
org/10.1063/1.2061894.
[1] Hardcastle FD, Wachs IE. Determination of vanadium-oxygen bond distances and
[22] Tian F, Zhang Y, Zhang J, Pan C. Raman spectroscopy: a new approach to measure
bond orders by Raman spectroscopy. J Phys Chem 1991;95:5031–41. https://doi.
the percentage of anatase TiO2 exposed (001) facets. J Phys Chem C
org/10.1021/j100166a025.
2012;116:7515–9. https://doi.org/10.1021/jp301256h.
[2] Liu X, Chu PK, Ding C. Surface modification of titanium, titanium alloys, and related
[23] Zhang Jing, Li Meijun, Feng Zhaochi, Chen Jun, Li C. UV Raman spectroscopic
materials for biomedical applications. Mater Sci Eng R Rep 2004;47:49–121.
study on TiO2. I. Phase Transformation at the Surface and in the Bulk. J Phys Chem
https://doi.org/10.1016/J.MSER.2004.11.001.
2005. https://doi.org/10.1021/JP0552473.
[3] Fujishima A, Honda K. Electrochemical photolysis of water at a semiconductor
[24] Hardcastle FD. Raman spectroscopy of Titania (TiO2) nanotubular water-splitting
electrode. Nature 1972;238:37–8. https://doi.org/10.1038/238037a0.
catalysts. vol. 65; 2011.
[4] Hanaor DAH, Sorrell CC. Sand supported mixed-phase TiO2 photocatalysts for water
[25] Lu G, Bernasek SL, Schwartz J. Oxidation of a polycrystalline titanium surface by
decontamination applications. Adv Eng Mater 2014;16:248–54. https://doi.org/10.
oxygen and water. Surf Sci 2000;458:80–90. https://doi.org/10.1016/S0039-
1002/adem.201300259.
6028(00)00420-9.
[5] Kuciauskas Darius, Freund Michael S, Gray Harry B, Winkler Jay R, Lewis NS.
[26] Pérez del Pino A, Serra P, Morenza J. Oxidation of titanium through Nd:YAG laser
Electron transfer dynamics in nanocrystalline titanium dioxide solar cells sensitized
irradiation. Appl Surf Sci 2002;197–198:887–90. https://doi.org/10.1016/S0169-
with ruthenium or osmium polypyridyl complexes. J Phys Chem 2000. https://doi.
4332(02)00447-6.
org/10.1021/JP002545L.
[27] Gemelli E, Camargo NHA. Oxidation kinetics of commercially pure titanium.
[6] Deo G, Turek AM, Wachs IE, Machej T, Haber J, Das N, et al. Physical and chemical
Matéria 2007;12:525–31. https://doi.org/10.1590/S1517-70762007000300014.
characterization of surface vanadium oxide supported on titania: influence of the
[28] Li L-H, Kong Y-M, Kim H-W, Kim Y-W, Kim H-E, Heo S-J, et al. Improved biological
titania phase (anatase, rutile, brookite and B). Appl Catal A Gen 1992;91:27–42.
performance of Ti implants due to surface modification by micro-arc oxidation.
https://doi.org/10.1016/0926-860X(92)85176-C.
Biomaterials 2004;25:2867–75. https://doi.org/10.1016/J.BIOMATERIALS.2003.
[7] Porto SPS, Fleury PA, Damen TC. Raman spectra of TiO2, MgF2, ZnF2, FeF2, and
09.048.
MnF2. Phys Rev 1967;154:522–6. https://doi.org/10.1103/PhysRev. 154.522.
[29] Diefenbeck M, Mückley T, Schrader C, Schmidt J, Zankovych S, Bossert J, et al. The
[8] Lukačević I, Gupta SK, Jha PK, Kirin D. Lattice dynamics and Raman spectrum of
effect of plasma chemical oxidation of titanium alloy on bone-implant contact in
rutile TiO2: the role of soft phonon modes in pressure induced phase transition.
rats. Biomaterials 2011;32:8041–7. https://doi.org/10.1016/J.BIOMATERIALS.
Mater Chem Phys 2012;137:282–9. https://doi.org/10.1016/J.MATCHEMPHYS.
2011.07.046.
2012.09.022.
[30] Kofstad P, Hauffe K, Kjollesdal H, Reilly JJ, Wiswall RH. Investigation on the oxi-
[9] Ma HL, Yang JY, Dai Y, Zhang YB, Lu B, Ma GH. Raman study of phase transfor-
dation mechanism of titanium. Acta Chem Scand 1958;12:239–66. https://doi.org/
mation of TiO2 rutile single crystal irradiated by infrared femtosecond laser. Appl
10.3891/acta.chem.scand.12-0239.
Surf Sci 2007;253:7497–500. https://doi.org/10.1016/J.APSUSC.2007.03.047.
[31] Orii Y, Masumoto H, Honda Y, Anada T, Goto T, Sasaki K, et al. Enhancement of
[10] Arsov LD, Kormann C, Plieth W. Electrochemical synthesis and in situ Raman
octacalcium phosphate deposition on a titanium surface activated by electron cy-
spectroscopy of thin films of titanium dioxide. J Raman Spectrosc 1991;22:573–5.
clotron resonance plasma oxidation. J Biomed Mater Res Part B Appl Biomater
https://doi.org/10.1002/jrs.1250221006.
2010;93B:476–83. https://doi.org/10.1002/jbm.b.31605.

1584
E.J. Ekoi et al. Results in Physics 12 (2019) 1574–1585

[32] Zwilling V, Aucouturier M, Darque-Ceretti E. Anodic oxidation of titanium and defects in rutile TiO2–x. Phys Rev B 2006;73:193202https://doi.org/10.1103/
TA6V alloy in chromic media. An electrochemical approach. Electrochim Acta PhysRevB.73.193202.
1999;45:921–9. https://doi.org/10.1016/S0013-4686(99)00283-2. [55] Leng Y, Huang N, Yang P, Chen J, Sun H, Wang J, et al. Influence of oxygen pressure
[33] Zwilling V, Darque-Ceretti E, Boutry-Forveille A, David D, Perrin MY, Aucouturier on the properties and biocompatibility of titanium oxide fabricated by metal plasma
M. Structure and physicochemistry of anodic oxide films on titanium and TA6V ion implantation and deposition. Thin Solid Films 2002;420–421:408–13. https://
alloy. Surf Interface Anal 1999;27:629–37. https://doi.org/10.1002/(SICI)1096- doi.org/10.1016/S0040-6090(02)00814-3.
9918(199907)27:7<629::AID-SIA551>3.0.CO;2-0. [56] Etacheri V, Seery MK, Hinder SJ, Pillai SC. Oxygen rich Titania: a dopant free, high
[34] Aronsson B-O, Lausmaa J, Kasemo B. Glow discharge plasma treatment for surface temperature stable, and visible-light active anatase photocatalyst. Adv Funct Mater
cleaning and modification of metallic biomaterials. J Biomed Mater Res 2011;21:3744–52. https://doi.org/10.1002/adfm.201100301.
1997;35:49–73. https://doi.org/10.1002/(SICI)1097-4636(199704)35:1<49::AID- [57] She H, Zhou H, Li L, Wang L, Huang J, Wang Q. Nickel-Doped excess oxygen defect
JBM6>3.0.CO;2-M. titanium dioxide for efficient selective photocatalytic oxidation of benzyl alcohol.
[35] Kakizaka S, Sakamoto T, Matsuura H, Akatsuka H. Titanium oxidation by micro- ACS Sustain Chem Eng 2018;6:11939–48. https://doi.org/10.1021/acssuschemeng.
wave discharge oxygen plasma and relationship with plasma parameters. J Adv 8b02217.
Oxid Technol 2007;10:253–9. https://doi.org/10.1515/jaots-2007-0205. [58] Cabrera N, Mott NF. Theory of the oxidation of metals. Reports Prog Phys
[36] Ekoi EJ, Stallard C, Reid I, Dowling DP. Tailoring oxide-layer formation on titanium 1949;12:163–84. https://doi.org/10.1088/0034-4885/12/1/308.
substrates using microwave plasma treatments. Surf Coat Technol 2017:325. [59] Parker JC, Siegel RW. Calibration of the Raman spectrum to the oxygen stoichio-
https://doi.org/10.1016/j.surfcoat.2017.06.046. metry of nanophase TiO2. Appl Phys Lett 1990;57:943–5. https://doi.org/10.1063/
[37] Ekoi E, Awais M, Dowling D. Microwave plasmas as a processing tool for tailoring 1.104274.
the surface properties of ceramic coatings. In: Basheer Al-Naib U, editor. Recent [60] Gadelmawla ES, Koura MM, Maksoud TMA, Elewa IM, Soliman HH. Roughness
Adv. Porous Ceram. IntechOpen; 2018. https://doi.org/10.5772/intechopen. parameters. J Mater Process Technol 2002;123:133–45. https://doi.org/10.1016/
71686. S0924-0136(02)00060-2.
[38] Patterson AL. The Scherrer formula for X-ray particle size determination. Phys Rev [61] Qian LH, Yan XQ, Fujita T, Inoue A, Chen MW. Surface enhanced Raman scattering
1939;56:978–82. https://doi.org/10.1103/PhysRev. 56.978. of nanoporous gold: smaller pore sizes stronger enhancements. Appl Phys Lett
[39] Langford JI, Wilson AJC. Scherrer after sixty years: a survey and some new results 2007;90:153120https://doi.org/10.1063/1.2722199.
in the determination of crystallite size. J Appl Crystallogr 1978;11:102–13. https:// [62] Macias G, Alba M, Marsal LF, Mihi A. Surface roughness boosts the SERS perfor-
doi.org/10.1107/S0021889878012844. mance of imprinted plasmonic architectures. J Mater Chem C 2016;4:3970–5.
[40] Gowen AA, Downey G, Esquerre C, O’Donnell CP. Preventing over-fitting in PLS https://doi.org/10.1039/C5TC02779A.
calibration models of near-infrared (NIR) spectroscopy data using regression coef- [63] Sedlaček M, Podgornik B, Vižintin J. Correlation between standard roughness
ficients. J Chemom 2011;25:375–81. https://doi.org/10.1002/cem.1349. parameters skewness and kurtosis and tribological behaviour of contact surfaces.
[41] Lawless KR. The oxidation of metals. Reports Prog Phys 1974;37:231–316. https:// Tribol Int 2012;48:102–12. https://doi.org/10.1016/J.TRIBOINT.2011.11.008.
doi.org/10.1088/0034-4885/37/2/002. [64] Hildebrandt P, Stockburger M. Surface-enhanced resonance Raman spectroscopy of
[42] Miles JL, Smith PH. The formation of metal oxide films using gaseous and solid Rhodamine 6G adsorbed on colloidal silver. J Phys Chem 1984;88:5935–44.
electrolytes. J Electrochem Soc 1963;110:1240. https://doi.org/10.1149/1. https://doi.org/10.1021/j150668a038.
2425633. [65] Kreibig U, Genzel L. Optical absorption of small metallic particles. Surf Sci
[43] Schroen W. Physics of preparation of Josephson barriers. J Appl Phys 1985;156:678–700. https://doi.org/10.1016/0039-6028(85)90239-0.
1968;39:2671–8. https://doi.org/10.1063/1.1656651. [66] Nie S, Emory SR. Probing single molecules and single nanoparticles by surface-
[44] Hayashi Y, Drawl W, Messier R. Temperature dependence of nucleation density of enhanced raman scattering. Science 1997;275:1102–6. https://doi.org/10.1126/
chemical vapor deposition diamond. Jpn J Appl Phys 1992;31:L193–6. https://doi. SCIENCE.275.5303.1102.
org/10.1143/JJAP.31.L193. [67] Xu H, Bjerneld EJ, Käll M, Börjesson L. Spectroscopy of single hemoglobin mole-
[45] Tang CJ, Neves AJ, Fernandes AJS. Influence of nucleation density on film quality, cules by surface enhanced raman scattering. Phys Rev Lett 1999;83:4357–60.
growth rate and morphology of thick CVD diamond films. Diam Relat Mater https://doi.org/10.1103/PhysRevLett. 83.4357.
2003;12:1488–94. https://doi.org/10.1016/S0925-9635(03)00179-1. [68] Kneipp K, Wang Y, Kneipp H, Perelman LT, Itzkan I, Dasari RR, et al. Single mo-
[46] Balachandran U, Eror NG. Raman spectra of titanium dioxide. J Solid State Chem lecule detection using surface-enhanced Raman scattering (SERS). Phys Rev Lett
1982;42:276–82. https://doi.org/10.1016/0022-4596(82)90006-8. 1997;78:1667–70. https://doi.org/10.1103/PhysRevLett. 78.1667.
[47] Yang M, Huang D, Hao P, Zhang F, Hou X, Wang X. Study of the Raman peak shift [69] Milekhin AG, Yeryukov NA, Sveshnikova LL, Duda TA, Kosolobov SS, Latyshev AV,
and the linewidth of light-emitting porous silicon. J Appl Phys 1994;75:651–3. et al. Raman scattering for probing semiconductor nanocrystal arrays with a low
https://doi.org/10.1063/1.355808. areal density. J Phys Chem C 2012;116:17164–8. https://doi.org/10.1021/
[48] Richter H, Wang ZP, Ley L. The one phonon Raman spectrum in microcrystalline jp210720v.
silicon. Solid State Commun 1981;39:625–9. https://doi.org/10.1016/0038- [70] Milekhin AG, Sveshnikova LL, Duda TA, Yeryukov NA, Rodyakina EE, Gutakovskii
1098(81)90337-9. AK, et al. Surface-enhanced Raman spectroscopy of semiconductor nanostructures.
[49] Campbell IH, Fauchet PM. The effects of microcrystal size and shape on the one Phys E Low-Dimensional Syst Nanostruct 2016;75:210–22. https://doi.org/10.
phonon Raman spectra of crystalline semiconductors. Solid State Commun 1016/J.PHYSE.2015.09.013.
1986;58:739–41. https://doi.org/10.1016/0038-1098(86)90513-2. [71] Xue X, Ji W, Mao Z, Mao H, Wang Y, Wang X, et al. Raman investigation of na-
[50] Smith JR, Walsh FC, Clarke RL. Electrodes based on Magnéli phase titanium oxides: nosized TiO2: effect of crystallite size and quantum confinement. J Phys Chem C
the properties and applications of Ebonex® materials. J Appl Electrochem 2012;116:8792–7. https://doi.org/10.1021/jp2122196.
1998;28:1021–33. https://doi.org/10.1023/A:1003469427858. [72] Bontempi N, Carletti L, De Angelis C, Alessandri I. Plasmon-free SERS detection of
[51] Liborio L, Mallia G, Harrison N. Electronic structure of the Ti4O7 Magnéli phase. environmental CO2 on TiO2 surfaces. Nanoscale 2016;8:3226–31. https://doi.org/
Phys Rev B 2009;79:245133https://doi.org/10.1103/PhysRevB.79.245133. 10.1039/C5NR08380J.
[52] Liborio L, Harrison N. Thermodynamics of oxygen defective Magnéli phases in [73] Lombardi JR, Birke RL. Theory of surface-enhanced Raman scattering in semi-
rutile: a first-principles study. Phys Rev B 2008;77:104104https://doi.org/10. conductors. J Phys Chem C 2014;118:11120–30. https://doi.org/10.1021/
1103/PhysRevB.77.104104. jp5020675.
[53] Anderson JS, Tilley RJD. Crystallographic shear in oxygen-deficient rutile: an [74] Colomban P, Slodczyk A. Raman intensity: an important tool to study the structure
electron microscope study. J Solid State Chem 1970;2:472–82. https://doi.org/10. and phase transitions of amorphous/crystalline materials. Opt Mater (Amst)
1016/0022-4596(70)90106-4. 2009;31:1759–63. https://doi.org/10.1016/J.OPTMAT.2008.12.030.
[54] Cho E, Han S, Ahn H-S, Lee K-R, Kim SK, Hwang CS. First-principles study of point

1585