This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 133.30.106.11

This content was downloaded on 08/05/2014 at 07:29

Please note that terms and conditions apply.

Microscopic origin of lattice contraction and expansion in undoped rutile TiO2 nanostructures

View the table of contents for this issue, or go to the journal homepage for more

2014 J. Phys. D: Appl. Phys. 47 215302

(http://iopscience.iop.org/0022-3727/47/21/215302)

Home Search Collections Journals About Contact us My IOPscience

Journal of Physics D: Applied Physics

J. Phys. D: Appl. Phys. 47 (2014) 215302 (13pp) doi:10.1088/0022-3727/47/21/215302

Microscopic origin of lattice contractionand expansion in undoped rutile TiO2nanostructuresBatakrushna Santara1, P K Giri1,2, Kenji Imakita3 and Minoru Fujii3

1 Department of Physics, Indian Institute of Technology Guwahati, Guwahati-781039, India2 Centre for Nanotechnology, Indian Institute of Technology Guwahati, Guwahati-781039, India3 Department of Electrical and Electronic Engineering, Graduate School of Engineering, KobeUniversity, Kobe 657-8501, Japan

E-mail: pravat g@yahoo.com

Received 16 January 2014, revised 12 March 2014Accepted for publication 25 March 2014Published 7 May 2014

AbstractWe have investigated the microscopic origin of lattice expansion and contraction in undopedrutile TiO2 nanostructures by employing several structural and optical spectroscopic tools.Rutile TiO2 nanostructures with morphologies such as nanorods, nanopillars and nanoflowers,depending upon the growth conditions, are synthesized by an acid-hydrothermal process.Depending on the growth conditions and post-growth annealing, lattice contraction andexpansion are observed in the nanostructures and it is found to correlate with the nature anddensity of intrinsic defects in rutile TiO2. The change in lattice volume correlates well with theoptical bandgap energy. Irrespective of growth conditions, theTiO2 nanostructures exhibitstrong near infrared (NIR) photoluminescence (PL) at 1.43 eV and a weak visible PL, whichare attributed to the Ti interstitials and O vacancies, respectively, in rutile TiO2 nanostructures.Further, ESR study reveals the presence of singly ionized oxygen vacancy defects. It isobserved that lattice distortion depends systematically on the relative concentration and type ofdefects such as oxygen vacancies and Ti interstitials. XPS analyses revealed a downshift inenergy for both Ti 2p and O 1s core level spectra for various growth conditions, which isbelieved to arise from the lattice distortions. It is proposed that the Ti4+ interstitial and F+

oxygen vacancy defects are primarily responsible for lattice expansion, whereas theelectrostatic attraction between Ti4+ interstitial and O2− interstitial defects causes the latticecontraction in the undoped TiO2 nanostructures. The control of lattice parameters through theintrinsic defects may provide new routes to achieving novel functionalities in advancedmaterials that can be tailored for future technological applications.

Keywords: TiO2 nanostructures, defects, lattice expansion, photoluminescence

(Some figures may appear in colour only in the online journal)

1. Introduction

Titania (TiO2) is considered to be one of the most usefulsemiconducting metal oxides for applications ranging fromsensors to photonic crystals, energy storage and photocatalysis.The unubiquitous presence of intrinsic point defects such asoxygen vacancies (Ov) and Ti interstitials (Tii) in TiO2 playa key role in enhancing the electrical, chemical and opticalproperties of the materials at the nanoscale, which ultimately

improve the efficiency of nanostructures (NSs) exploitedin solar cells [1–3], photocatalytic hydrogen generation[4], UV-mediated photodetectors [5], photocatalytic pollutantdegradation [6] and spintronics devices [7] etc. To understandthe enhanced properties of the TiO2 NS, successful fabricationof a wide range of materials is significant where themorphologies can be precisely controlled with designedfunctionalities. In particular, lattice parameters are veryimportant structural characteristics and thus their variations

0022-3727/14/215302+13$33.00 1 © 2014 IOP Publishing Ltd Printed in the UK

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

directly affect the properties of nanomaterials and therelevant applications. Therefore, a large number of studieshave concentrated on the size-dependent variation of latticeparameters and subsequently have studied the linear/nonlinearexpansion/contraction of TiO2 [8–10] and other metal-oxidenanomaterials [11, 12].

It is well-established that the reduction of particle size byless than a finite size at nanoscale causes variations in latticeparameters [8–11, 13, 14]. However, the origin of size-inducedchanges in the lattice volume of TiO2 nanomaterials has beena subject of debate. Most of the metal-oxide nanocrystals[11, 12, 15] and in particular rutile TiO2 [8, 13] and anataseTiO2 [10] exhibit lattice expansion as the crystal size reducesin contrast to the lattice contraction in metal nanoparticles(NPs) [16, 17] and is considered to be a general characteristicof all metal-oxide nanomaterials. On the other hand, latticecontraction has been reported [9] in anatase TiO2 NPs asthe particle size reduces and it was more pronounced for thecrystallite sizes below ∼15 nm. From the thermodynamicpoint of view, the lattice expansion for smaller particle sizesin oxide materials is quite surprising, since surface stresseswould cause small particles to compress due to higher surfacecurvature as is observed in non-oxide nanomaterials such as Si[18], ZnS [19] and CdSe [20] etc. Therefore, the mechanism oflattice expansion/contraction in TiO2 is not understood clearly.The existing reports suggest contradicting results on the sizedependence of the lattice contraction/expansion in TiO2 andother oxide materials at the nanoscale.

In CeO2 nanocrystals, Deshpande et al [11] proposed thatan oxygen vacancy and the accompanying valence reductionfrom Ce4+ to Ce3+ might cause lattice expansion as the particlesize decreases, while Li et al [9] found no oxygen vacanciesin anatase TiO2 NPs in their size-induced changes in latticevolume. Swamy et al [10] reported a small lattice expansion inanatase TiO2 NPs which was attributed to increased Ti vacancyand lattice strain. A correlation between Ti vacancy abundanceand lattice volume increase in rutile TiO2 was suggested byKuznetsov et al [13]. A surface defect dipole model wasproposed by Li et al [8] to explain the linear lattice expansionin rutile TiO2. In the case of BaTiO3, an increased ionicity ofTi ions with particle size reduction was proposed as a possiblereason for the lattice expansion [12]. These studies indicatethat there exists substantial controversy regarding the originof lattice expansion/contraction in TiO2 and other metal oxidesemiconductors. Using ab initio calculations, Iddir et al [21]predicted lattice expansion in rutile TiO2 via Tii and Ov defects.However, there is no experimental evidence of Tii and Ov

defects in rutile TiO2 causing the lattice expansion. Hence,a thorough understanding of the microscopic origin of latticeparameter variation through experimental studies is essentialto resolve the controversy.

Most of the studies reported to date on the size-dependentlattice expansion/contraction in TiO2 are for extremely smallNps/nanocrystals: less than ∼15 nm in anatase [9] andless than ∼27 nm in rutile TiO2 [8, 13]. So, it has beenvery challenging to identify the individual contribution ofquantum confinement, strain, intrinsic defects (i.e., Ov, Tii,Ti vacancy etc.) and other surface defects towards the lattice

parameter variation in TiO2 NS. In this work, we report on thevariation of lattice parameters in hydrothermally synthesizedrutile TiO2 nanorods (NRs), nanopillars and nanoflowers ofcomparatively large sizes and find its correlation with the Ov,Tii and O interstitial (Oi) defects in TiO2 [8, 9, 13]. Latticecontraction and expansion are observed in the as-grown NS,depending upon the growth temperature, reaction durationand stirring during growth, which is due to the differentconcentrations and types of intrinsic defects associated withthe materials. The change in lattice volume is correlated withthe calculated bandgap energy. Ov-related blue visible, Oi-related red visible and Tii defect-related near infrared (NIR)photoluminescence (PL) are observed in our as-synthesizedNS which are mainly responsible for the lattice parametervariations. Further, electron spin resonance (ESR) and x-rayphotoelectron spectroscopy (XPS) studies are performed tounderstand the nature of the defects and their contribution tothe changes in lattice parameters.

2. Experimental details

2.1. Growth and processing of TiO2 NS

Titanium butoxide (Sigma-Aldrich) and 35% HCl (Merck)were used as the precursor materials. Doubly distilleddeionized (DI) water was used during all the experiments. Ina typical synthesis, 1 ml of titanium butoxide was added to60 ml of mixed DI water and 35% HCl (1 : 1) with constantstirring for 15 min. Afterwards, the mixed solution wastransferred into a Teflon-lined autoclave (Berghof, BR-100)and the temperature inside the autoclave was monitored andmaintained at 150–190 ◦C under autogenous pressure for 12–48 h. Some of the samples were grown at 150 ◦C under constantmagnetic stirring at 250 rpm during the reactions. The formedprecipitates were obtained and finally heat treated at 500 ◦C for4 h in air. For convenience of discussion, theTiO2 NS grownat 150 ◦C are categorized as ‘A’ and samples are named asA12, A18, A48 for the growth duration of 12 h, 18 h and 48 h,respectively. Samples grown at 170 and 190 ◦C after 18 h aretermed as B18 and C18, respectively. The sample grown at150 ◦C after 18 h reaction with stirring throughout the reactionis termed as A18s. The details of the sample nomenclature areshown in table 1.

2.2. Characterization techniques

The crystal structures of the TiO2 NS are characterized by x-raydiffraction (XRD) (Rigaku RINT 2500 TTRAX-III, Cu Kα

radiation). Morphologies of the samples were studied by fieldemission scanning electron microscopy (FESEM) (Sigma,Zeiss). The energy dispersive x-ray spectroscopy (EDX)was carried out with the help of an x-ray detector (OxfordInstruments, UK) attached to a scanning electron microscope(SEM, LEO 1430VP). The high magnification surfacemorphologies and structures of the samples were studiedby transmission electron microscopy (TEM), high-resolutionTEM (HRTEM) and selected area electron diffraction (SAED)patterns (JEOL-JEM 2010 operated at 200 kV). Specimens forHRTEM investigations were prepared by dispersing powder

2

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

Table 1. Details of the sample preparation conditions, morphology, lattice parameters (a, c), and bandgaps (calculated from the absorptionspectra).

Lattice parameter (Å) Volume (Å3) Bandgap (eV)Sample Growth temperature,names reaction duration Morphology a c a × a × c Direct Indirect

A12 150 ◦C, 12 h Nanorod 4.597 2.959 62.531 2.92 2.69A18 150 ◦C, 18 h Nanopillar 4.588 2.956 62.223 3.01 2.86A18s 150 ◦C, 18 h, stirring Nanorod 4.599 2.959 62.585 2.90 2.68A48 150 ◦C, 48 h Nanopillar 4.576 2.950 61.772 2.94 2.79B18 170 ◦C, 18 h Nanopillar 4.597 2.958 62.510 2.94 2.77C18 190 ◦C, 18 h Nanoflower/nanopillar 4.598 2.959 62.558 2.93 2.74

particles in ethanol and drop casting them on a carbon coatedcopper grid of 400 meshes (Pacific Grid, USA).The UV–vis–NIR absorption spectroscopy measurements were recordedusing a commercial spectrophotometer (Perkin Emler UVwin Lab).The steady state PL spectrum was recorded atroom temperature (RT) using a 405 nm diode laser (Coherent,Cube) excitation with the help of a spectrometer (focallength: 15 cm; blaze wavelength: 500 nm; groove density:150 g mm−1) equipped with a cooled charge-coupled device(Princeton Instruments, PIXIS 100B) detector. The powdersamples were put on a conductive carbon tape (black) forthe PL measurements. Each spectrum was corrected for thedetector response as a function of wavelength after backgroundsubtraction. XPS measurements were carried out with a PHIX-Tool automated photoelectron spectrometer (ULVAC-PHI,Inc.) using Al Kα x-ray beam (1486.6 eV) with a beamcurrent of 20 mA. A carbon 1s spectrum was used for thecalibration of the XPS spectra recorded for various samples.ESR measurements were done with a JEOL (JES-FA200)instrument operating in the X band.

3. Results and discussion

3.1. Structural characterization

The XRD patterns of as-synthesized samples are shown infigure 1, where all the peaks correspond to the rutile TiO2

phase. The XRD patterns of A12, A18 and A48 are shownin figure 1(a). It is clear that with the increase in reactionduration, the diffraction peaks are shifted to higher angles(2θ). The inset of figure 1(a) shows the corresponding shiftand broadening of the (1 1 0) peak. Sample A12, grown for alower reaction duration, shows a higher broadening and lowerintensity than those of A48. Figure 1(b) shows the XRDpatterns of A18, B18 and C18. As the growth temperatureincreases, the diffraction peaks are systematically shifted tolower 2θ values (figure 1(b)). The magnified view of the(1 1 0) peak shows the shift and broadening of line shapewith different growth temperatures (inset of figure 1(b)). Itis observed from the FESEM and TEM images that as thegrowth temperature increases, the size of the NS increases.According to the earlier reports [8, 13], with the increase in NSsize the lattice volume decreases in rutile TiO2. This wouldimply a higher 2θ value if strain contribution is neglected.More importantly, the size of the NS is increased with ahigher growth temperature and reaction duration. However,

we observed two different trends, i.e., lattice contraction withan increase in reaction duration and lattice expansion with anincrease in growth temperature. So, the shift and broadeningof XRD peaks in our case cannot be explained from sizeeffects alone. Figure 1(c) shows the effect of stirring on thesample during the reaction. Comparison of the XRD patternfor A18 and A18s clearly shows that the peaks are shiftedto lower 2θ for the sample grown under stirring conditions.The change in lattice parameters a, c and the cell volume withrespect to reaction duration and growth temperatures are shownin figures 2(a) and (b), respectively and the correspondingparameters are presented in table 1. Interestingly, the axialratio c/a, representing the lattice symmetry, remains almostconstant at 0.644. The lattice parameters are calculated byusing (1 1 0) and (1 0 1) peaks in the XRD pattern. The shift andbroadening of diffraction peaks are generally related to the sizeand strain effects of rutile TiO2 NS. To isolate the contributionsof strain from that of the size effects, the Williamsons–Hall(W–H) method [22, 23] is used for sample B18 and calculatesthe crystallite size and strain for the NS using the followingequation:

β cos θ

λ= 1

D+

ε sin θ

λ, (1)

where β is the full width at half maximum (fwhm), θ is thediffraction angle, λ is the x-ray wavelength, D is the averagecrystallite size and ε is the effective strain. The strain iscalculated from the slope and the crystallite size is calculatedfrom the inverse of the intercept of the linear fit (figure 2(c)).The positive slope indicates tensile strain, consistent with thelattice expansion. Note that there is a large scatter in data in theplot, which is indicative of the anisotropic and inhomogeneousstrain in the nanocrystals [24]. Since the analysis of theanisotropy factor for TiO2 is a complex function of elasticconstants and dislocation contrast factors, and these quantitiesare not known for TiO2, accurate analysis of size and strainwas not possible from such a plot. However, if we ignore theerrors due to the scatter in data, a rough estimate of the averagecrystallite size and strain can be made from figure 2(c). Theaverage crystallite size and strain thus calculated was foundto be 114 nm and 0.008 ± 0.001 from the W–H plot for B18.The tensile strain of 0.8% is likely to be caused by the largedensity of Tii and Ov defects, as evidenced from other studiesdiscussed below. We also used the Scherrer formula thatneglects the strain effect to calculate the crystallite size. Inall the cases, the Scherrer formula yields a lower size value ascompared to the size extracted from W–H plot, as expected.

3

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

Figure 1. Comparison of XRD patterns for different samples: (a)A12, A18 and A48; (b) A18, B18 and C18; (c) samples A18 andA18s. The inset in each graph shows the magnified view of thecorresponding (1 1 0) brag peak indicating a systematic shift in peakposition.

Interestingly, despite the large scatter in the data, the averagecrystallite sizes extracted from the W–H plot is very similarto the width/thickness of the nanostructures obtained from theFESEM images.

Figure 2. Variation of lattice parameters a and c with: (a) reactionduration, (b) growth temperature. The inset in each case shows thecorresponding change in lattice volume. (c) A plot of (β cosθ)/λversus (sinθ)/λ for B18.

3.2. Raman scattering studies

The Raman studies are employed to further confirm the phase,crystallinity and defects in the as-synthesized NS. All thepeaks correspond to the characteristic Raman modes of therutile TiO2 structure (figure 3). Figure 3(a) shows the Ramanmodes of samples A12, A18 and A48. The three typical first-order Raman modes B1g (weak), most intense Eg and A1g

along with a second-order vibrational mode Eg (SO) at ∼240 cm−1were observed for all samples. The most intense Eg

mode is red-shifted while A1g mode slightly blue-shifted forsamples A18 and A48 compared to A12. The blue shift of A1g

mode may be due to the compressive strain implying a latticecontraction. The red shift of Eg and A1g modes is generally due

4

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

Figure 3. Comparison of Raman spectra for: (a) A12, A18 andA48; (b) A18, B18 and C18. The inset in (a) shows a magnifiedview of the Eg mode.

to the phonon confinement effect and oxygen vacancy defects[25–27] in rutile TiO2. As the size of the NS is increased forhigher reaction duration, the red shift cannot be explained bythe phonon confinement effect. Therefore, the red shift of Eg

mode may be due to the higher concentration of defects inA18 and A48 as compared to that in A12. Figure 3(b) showsthe Raman spectra of samples A18, B18 and C18 which weregrown at different growth temperatures. The Raman intensityis clearly higher for samples grown at higher temperatures andthis is consistent with the XRD pattern shown in figure 1(b).

3.3. Morphological studies

The morphologies of the as-synthesized samples as obtainedfrom FESEM images are shown in figure 4. The samples grownafter different reaction durations, i.e., 12 h, 18 h and 48 h areshown in figures 4(a)–(c), respectively. Sample A12 revealedNR-like morphologies, while A18 and A48 revealed tetragonalnanopillar-like structures with rough surfaces on the topportion with small NR-like morphologies. The NRs of length

∼320–550 nm and diameter ∼26–74 nm is observed from theFESEM images for the sample A12 (figure 4(a)). However,samples A18 and A48 show nanopillars with tetragonal crosssections of sizes in the range ∼109–280 nm and ∼215–684 nm,respectively (figures 4(b) and (c)). Figures 4(d) and (e)show the FESEM images of the samples grown at 170 ◦Cand 190 ◦C, respectively. Sample B18 shows the nanopillar-like morphologies, whereas C18 shows the 3D nanoflowers.The tetragonal nanopillars having square cross sections of side∼390–684 nm for sample B18. In both samples B18 and C18,the top surface consists of small NR-like budding structuresand this rough surface increases the effective surface area ofthe rutile TiO2 nanostructrures which may increase the surfacereactivity. The high magnification image of C18 shows the3D nanoflowers with nanopillars as petals (figure 4(f )). TheEDX spectrum of the corresponding nanoflowers is shown inthe inset of figure 4(e). The EDX spectrum shows the onlyTi and O element, indicating the phase purity of TiO2 NS. Itis interesting to note that the change in reaction durations andgrowth temperatures lead to the alteration of the hydrolysis rateof titanium butoxide in the presence of HCl acid, which helpsin designing various architectures from NRs to nanopillars andnanoflowers.

TEM images of as-synthesized samples are shown infigure 5. The TEM image of sample A12 shows NRs ofwidth 13 nm and length 286 nm (figure 5(a)). The inset infigure 5(a) shows the SAED pattern of corresponding NRsindicating the single crystalline and tetragonal nature of rutileTiO2. The HRTEM image of A12 is shown in figure 5(b).The d-spacing of 3.25 Å corresponds to the (1 1 0) plane ofrutile TiO2. The observed lattice fringe demonstrates that theexposed surface of the NRs has a {1 1 0} facet. Figure 5(c)shows a single nanopillar in A18 and the corresponding SAEDpattern is shown in figure 5(d). The clear electron diffractionspots indicate the single crystalline and tetragonal nature of theas-grown rutile TiO2 nanopillars. Figure 5(e) shows the TEMimage of a single NR of sample A18s and the correspondinglattice fringe is shown in figure 5(f ). The d-spacing of 2.33 Åcorresponds to the (2 0 0) plane of rutile TiO2 structure. Thelattice fringes of A18s reveal that the growth of NRs is alongthe [2 0 0] direction.

3.4. Optical absorption and PL studies

Light absorption characteristics of hydrothermally synthesizedNSs are shown in figure 6. The absorption edges of the samplesshow a systematic blue shift with reaction durations from12 h to 18 h (figure 6(a)), while the absorption edges are red-shifted with the increase in growth temperature (figure 6(b)).The Tauc’s plot ((αhν)2 versus hν) shows a direct bandgapbehaviour, as shown in the insets in each case. Theindirect bandgaps of the samples are also calculated fromthe linear extrapolation of (αhν)1/2 versus hν plot (Tauc’splot). Both the direct and indirect bandgaps are decreasedfor the samples grown at higher growth temperatures, whereasthey are increased with reaction durations from 12 h to 18 h(samples A12 and A18) and then decreased for 48 h reaction(sample A48) (see table 1). The variation of bandgaps for

5

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

Figure 4. FESEM images of the morphology of TiO2 NS: (a) A12, (b) A18, (c) A48, (d) B18, (e) C18, (f ) high magnified view of C18.The inset in (e) shows the EDX spectrum of C18 indicating Ti and O elements.

samples grown under different growth conditions may have arelation to the variation in lattice constants and this may hintto the contribution of intrinsic defects in different samples.These defect states are confirmed from our PL measurementsas discussed below. Figure 6(c) shows a comparison ofabsorption spectra for A18 and A18s. The sample A18s grownunder the stirring condition shows a red shift of absorptionof edge relative to that of A18, indicating a reduction in thebandgap. The samples (A18s, B18 and C18) with latticeexpansion show higher bandgaps compared to that in A18.However, the bandgaps of A18 and A48 showing latticecontraction are also increased. Ekuma and Bagayoko [28]reported the indirect bandgap (R–�) of 2.95 eV and directbandgap of 3.05 eV in rutile TiO2 using ab initio calculations.They observed a reduction in the indirect band gap (2.73 eV)

by 1% lattice expansion and increased to 3.02 eV by 2%contraction of both lattice parameters, which is consistentwith our results discussed above. However, we observeda reduction of bandgap with lattice contraction for sampleA48 compared to that in A18. It is quite likely that theincreased concentration of both Tii and Oi defects may help inlowering the bandgap in A48 compared to A18. The calculateddirect and indirect bandgaps for the samples grown underdifferent growth conditions are listed in table 1. Earlier opticalabsorption measurements [29, 30] and bandgap calculations[31, 32] have revealed that rutile TiO2 is direct-forbidden i.e.the direct transition �3c–�1v is dipole-forbidden [33], while theindirect-allowed transition (M—� [31] or R—� [32] is nearlydegenerate with the direct-forbidden transition. The energyseparation between the direct-forbidden and indirect-allowed

6

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

Figure 5. TEM image of (a) a single nanorod in A12, inset is the corresponding SAED pattern; (b) HRTEM lattice fringe of A12; (c) asingle nanopillar in A18, (d) SAED pattern showing single crystalline tetragonal rutile TiO2 in A18, (e) a single nanorod A18s, (f ) HRTEMlattice fringe of A18s.

transition is reported to be 10–100 meV [28, 30, 34] for bulkTiO2 and the difference in our case is ∼110–230 meV. Thehigher difference in our case may be due to the size and straineffects in the NS.

In order to understand the nature of defects and theassociated localized trap states, PL studies were performedat RT on different samples. We observed a strong NIR PLemission at 1.43 eV (∼867 nm) for all samples irrespective of

the growth conditions and a weak broad visible PL emission(see figure 7). Through in situ PL studies [35], we argue thatthe origin of the NIR and visible PL in TiO2 NS is due to Tii

defects and OV defects, respectively. Oxygen vacancy-inducedvisible PL is well known in TiO2 NS [35–37]. However, thereare very few reports on the PL signature of Tii defects in TiO2.Figure 7(a) shows that the intensity of NIR PL increases withthe increase in the reaction duration, indicating that more Tii

7

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

Figure 6. UV–vis–NIR absorption spectra: (a) A12 and A18 grownafter 12 h and 18 h, respectively; (b) A18, B18 and C18 grown at150 ◦C, 170 ◦C and 190 ◦C, respectively; (c) A18, A18s grownwithout and with stirring during the growth. The inset in each caseshows the (αhν)2 versus hν plot indicating the direct bandgap forthe corresponding absorption spectrum of each sample. Bandgapenergy is calculated from the extrapolated line (dashed) fitted to therespective linear portions.

defects are formed after longer reaction durations. TiO2(B)nanoribbons also showed similar visible and NIR PL at RT[35]. This essentially means that with the increase in reactionduration, both interstitial and oxygen vacancy defects areconsiderably increased. However, in A18 and A48, the visibleband splits into two peaks, i.e., one broad peak centred at∼2.77 eV and another peak at 1.96 eV (∼633 nm) and 1.99 eV(∼623 nm) for samples A18 and A48, respectively. The newfeature that evolves at the lower energy side is observed for the

first time and it is likely to be related to O2−i defects, whose

formation mechanism is explained later.The effect of growth temperature on the PL is shown

in figure 7(b). The NIR PL increases in the orderA18<B18<C18, indicating the concentration of Tii defectsincrease in the samples grown at higher temperatures. The PLemission at 1.96 eV disappears for B18 and C18. Figure 7(c)shows the PL spectra for the samples A18 and A18s. TheNIR PL intensity increases whereas broad visible PL decreasesfor sample A18 compared to A18s. Note that red PL at 1.96is not observed for A18s. The broad asymmetric visible PLemission of A12 is deconvoluted by two symmetric Gaussianpeaks centred at 2.82 eV (peak P1, violet curve) and 2.59 eV(peak P2, blue curve), as shown in figure 7(d). Recently, theblue PL at 2.8 eV was reported by Yamada and Kanemitsu [38]in rutile TiO2 and based on their time-resolved PL analysis, theblue PL was attributed to the localized defect-induced states,instead of the band-edge delocalized states. It was argued thatbecause only a free exciton PL is observed under low densityexcitation at low temperature, the initial photocarriers exist inthe free band-edge state rather than the trapped states [38].Interestingly, there has been no report of intrinsic PL in rutileTiO2 bulk crystals at RT, except that due to the impurity ordefects. We attributed to peak P1 and P2 to the transition fromneutral Ov and singly charged Ov(F+) trap states, respectively,just below the conduction band to the valence band. Notethat the signature of F+ is confirmed from ESR spectra. Uponthe loss of an O atom in the TiO2 lattice, the two unpairedelectrons that remain trapped in the vacancy cavity give rise toan F centre. The basic assumption is that one of the electronsin the F centre tends to occupy the neighbouring Ti4+ ion andyield a Ti3+ centre and a F+ centre. A more detailed explanationand the evidence for F+ and Ti3+ trap states in reduced TiO2

NS is provided in our earlier report [36]. However, we donot observe any signature of Ti3+ in our ESR and XPS studiesfor samples used in this work. We believe that the Ti3+ ionsmigrate towards the surface to interact with the atmosphericoxygen and are converted to Ti4+ by occupying an interstitialsite. The time-resolved PL spectra of sample A12 monitoringat 2.76 eV (450 nm) with 3.06 eV (405 nm) excitation is shownin the inset of figure 7(d). The decay spectrum is well fitted bysingle exponential decay with a time constant of 1.2 ns. This istypical of PL emission from defect states in TiO2 [38]. Thoughthere are two peaks in the steady state PL spectra (figure 7(d)),the single exponential fit to the decay curve indicates that boththe defect states have identical decay time constants. Thesestates correspond to F and F+ centres in TiO2.

3.5. XPS and ESR studies

XPS is a unique tool to investigate the surface defects andchemical environments, because of the high sensitivity tosurface. Evidence of oxygen vacancy-related surface defects[36, 39], such as Ti3+ or Ti2+ core level peaks in Ti 2p spectra arenot observed in our samples (figures 8(a) and (b)). Figures 8(a)and (b) show that Ti 2p3/2 and 2p1/2 core level peaks showshifting among the samples, keeping the difference between thebinding energy of 2p3/2 and 2p1/2 (�p = 5.7 eV) constant.

8

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

Figure 7. RT PL spectra for: (a) A12, A18 and A48 grown after 12 h, 18 h and 48 h, respectively; inset shows the magnified view of PLspectra in A12 and A18 in the visible region; (b) A18, B18 and C18 grown at 150 ◦C, 170 ◦C and 190 ◦C, respectively; (c) A18, A18s grownwithout and with stirring during the growth, respectively; (d) visible PL of A12 with Gaussian peak fitting, inset shows the PL decay of A18.The insets in (b) and (c) show the corresponding magnified view of the visible PL. The PL peaks corresponding to the oxygen vacancy,oxygen interstitial and Ti interstitial defects are labelled as Ov, Oi and Tii, respectively, in the figure.

The shift in the Ti 2p peaks may be due to the lattice distortionin rutile TiO2 NS. The Ti 2p3/2 core level peak of A18 isobserved at 457.7 eV which is attributed to Ti4+. The Ti 2p3/2peaks of A18s, B18 and C18 are shifted to higher bindingenergies compared to A18 and this may be due to the latticeexpansion (figure 8(a)). However, the lower binding energyshift of the 2p3/2 core level peak of A18 compared to A12may be due to the lattice contraction (figure 8(b)). A similarshift of O 1s core level peaks are observed (figure 8(c)). Theabsence of any shoulders at a higher binding energy of O 1sspectra except A12 and A18s indicates that samples A18, B18and C18 are free of impurity bondings such as hydroxyl (OH)groups and C-O bonds [36, 40]. The lower binding energyshoulder in A12 and A18s suggests the presence of OH groupsin these samples which show lattice expansion compared toA18 and is inconsistent with the lattice contraction in anataseTiO2 due to the surface OH group as reported by Li et al [9].

ESR is an exceptionally useful technique in detectingthe spin polarized charge state of defective TiO2 NS. Thepresence of singly ionized oxygen vacancies is identified fromthe ESR spectra, as shown in figure 8(d). The samples A12,A18, A48 and B18 show g = 2.00, 2.04, 2.06 and 2.03,respectively, which are assigned to singly ionized oxygenvacancy F+ defects. Note that various groups observed thesedefects in TiO2 from ESR studies and reported that the g-valuefor singly ionized oxygen vacancies is 2.00 [41–43]. Thisimplies that our samples contain oxygen vacancy-related F+

centres. These defects play a key role for the observed latticeparameter variations in rutile TiO2 NS.

3.6. Discussion

There are several reports on the effect of the size oflattice parameter variation in metal oxides nanocrystals[8–11, 13, 14]. Recently, it was reported [44] that the oxygenvacancy formation by itself, without a cation radius change inCeO2, results in a contraction of lattice parameters on the basisof their molecular dynamic simulation. It was explained thatupon the formation of an oxygen vacancy, the cation relaxesaway from the vacancy (by approximately 0.10 Å), howeverthe anion gets closer to it (by approximately 0.16 Å) and themagnitude of the significant distortion results in a net localcontraction of the lattice around the vacancy. However, thecombined effects of the oxygen vacancy and the associatedcation radii change results in lattice expansion of CeO2 [44].Particularly, the lattice expansion and contraction of rutile andanatase TiO2 was discussed by various groups [8–10, 13]. Liet al [9] reported that particle size reduction induces latticecontraction due to the surface hydration effect in pure anataseTiO2 nanocrystals. On the other hand, Swamy et al [10]observed lattice expansion at reduced crystallite size in anataseTiO2 and explained the lattice expansion in terms of increasedTi vacancy and lattice strain with the decreasing crystallitesize. Moreover, lattice expansion with the decreased crystallitesize is reported in rutile TiO2 [8, 13]. A surface defect dipolemodel was proposed by Li et al [8] to explain the latticeexpansion of rutile TiO2 nanocrystals with the reduction insize in terms of the strong interaction among the surfacedipoles that produced an increased negative pressure. This is

9

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

Figure 8. XPS spectra: (a) Ti 2p core level spectra for A18, A18s, B18 and C18. (b) Ti 2p core level spectra for A12 and A18. Dottedvertical lines show the change in relative peak position in each case. (c) O 1s core level spectra for A12, A18, A18s, B18 and C18. Dottedvertical line shows the change in relative peak position. (d) RT ESR spectra for A12, A18, A48 and B18.

in contrast to the ionicity increase-induced lattice expansion inBaTiO3 proposed by Tsunekawa et al [12]. The size-dependentlattice expansion of rutile TiO2 was explained in terms of Tivacancy abundance by Kuznetsov et al [13]. Using ab initiocalculation, Iddir et al [21] predicted that for Ov, the significantdisplacement extended along the [1 1 0] and [0 0 1] directionsof the equatorial plane encompassing the vacancy. The threeTi nearest neighbours have the largest outward displacementof 0.31 Å and the nearest O neighbours of Ov have the largestinward displacement of 0.14 Å. It was also reported [21] that forTii, the major displacement occurred in the plane perpendicularto [0 0 1], the largest being observed in the plane passingthrough Tii. The two nearest neighbour Ti are pushed awayfrom their ideal position by 0.3 Å and the closest four O atomsrelax toward the Tii by 0.21 Å. The lattice atoms relaxed aroundthe defects to accommodate the induced stress due to thediffusion of Tii through the interstitial region. Based on theabove information, we discuss the atomistic origin of latticeexpansion and contraction of rutile TiO2 in terms of intrinsicdefects rather than size-induced lattice variations, as the size ofour nanostructure is comparatively larger than those reportedin the literature [8, 13]. The intrinsic defects include Tii, Ov

and Oi in our rutile TiO2 NS.First, we discuss the possible mechanisms of lattice

expansion in our rutile TiO2 NS. From figure 2(b) and table 1,it is seen that the lattice is expanded as the growth temperatureis increased. Generally, in the nanoscale regime, the lattice

expansion in metal oxide nanocrystals is believed to result fromthe complex interplay of defect chemistry and physics. Forexample, (i) an increase in ionic radius upon the reduction ofthe cation valence state might cause the lattice expansion, and(ii) the electrostatic repulsion between a positively chargedoxygen vacancy and the metal cations surrounding it mayalso be responsible for the lattice expansion. ESR and XPSmeasurement on our samples do not show any signature of Ti3+,which eliminates the possibility of lattice expansion due to thereduction of cationic radii. However, the ESR signal showsevidence of F+ (singly ionized oxygen vacancy) centres in oursamples. In our previous report, the pathways for the formationof F+ and Ti3+ trap centres were explained [36]. Note that weobserved Tii -related NIR PL and F+- and neutral Ov-relatedvisible PL emission in our samples. The Ti–Ti and Ti–O bondsare relaxed due to the missing O atom. The nearest-neighbourTi atoms move outward from the vacancy to strengthen thebonding with the rest of the neighbouring O lattice. Whilethe next-nearest-neighbouring O atoms move inward due tothe absence of electrostatic repulsion by the missing O atom.It was reported [45, 46] that the local distortion due to theTi atom displacement depends upon the different charge stateof the oxygen vacancy, i.e. neutral, singly charged (+1) anddoubly charged (+2) oxygen vacancies. The positive chargescause greater repulsion for Ti–Ti distance compared to neutraloxygen vacancies [45, 46]. Note that we observed a singlycharged oxygen vacancy (F+) in ESR measurement and both

10

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

neutral Ov and singly charged F+ in PL measurement. So,we expect the lattice expansion in our samples to be causedby the oxygen vacancies creating F+ and neutral Ov defects.The electrostatic repulsion between F+ and Ti3+ helps in themigration of the unsaturated Ti3+ from the bulk towards thesurface to interact with the atmospheric O2 and form Ti4+

during the course of air annealing and occupying an interstitialsite similar to the lattice Ti sites in an octahedral TiO2 crystalsystem. It was reported from both experimental and theoreticalanalyses that Tii diffusion is faster in rutile TiO2 than the Ov

diffusion when annealing in O2 atmosphere [47–49]. Thelattice Ti is strongly bound with the lattice O atoms. When theTi4+ is placed in the interstitial site, the Ti atoms are relaxedoutward due to the electrostatic repulsion between positivelycharged Ti ions. However, the lattice O atoms are slightlydistorted towards the Ti4+

i as they are strongly bonded to latticeTi atoms. As a result, the net relaxation favours the expansionof the lattice volume which is due to the presence of Ti4+

idefects. It is noticed that the Ti4+

i increases with the increasein growth temperature from the PL spectra suggesting morelattice expansion. So, the lattice expansion is observed in oursamples in the order C18>B18>A18. The sample A18 has anadditional PL peak at 1.96 eV compared to B18 and C18 whichis related to Oi and will be discussed in the following paragraph.Our experimental results regarding the lattice expansion dueto Tii and Ov are consistent with the theoretical prediction ofIddir et al [21].

Secondly, we discuss the possible mechanism behind thelattice contraction in rutile TiO2 NS. Our earlier study on TiO2

NS showed that longer the reaction duration, the higher theOv and Tii defects for the same growth temperature [35]. Thisis also reflected in the present case, as Ti4+

i PL emission isenhanced for longer reaction samples. Since the number ofoxygen vacancies is greater in these samples, it is expectedthat the atmospheric O2 may inject rapidly into the TiO2 tooccupy the oxygen vacancy positions during the course of airannealing at 500 ◦C and occupy the vacancy site; as a resultthe oxygen vacancy concentration decreased considerably afterannealing, which is reflected in our PL spectra (reductionof broad visible PL in samples A18 andA48 compared toA12). It was recently reported that adsorbed isotopic oxygen isinjected into the bulk rutile TiO2 as O2−

i which underwent drifttowards the surface due to space-charge layers and retarded thediffusion of interstitials, promoting the exchange with latticeoxygen [50, 51]. Presumably most of the Ov near the surfaceare annihilated due to diffusion of O2 and the shut-down ofthe path for further diffusion of Oi which creates separate O2−

iions which facilitates the creation of surface electronic bands.The PL emission at 1.96/1.99 eV observed in A18/A48, maybe attributed to the surface electronic bands created by O2−

iions. The electrostatic attraction between the Ti4+

i and O2−i is

dominant over the lattice expansion due to F+ and neutral Ov

which results in the lattice contraction for the samples A18 andA48 compared to A12.

Another interesting point to be noted is that the red PL at1.96 eV is absent for the samples B18 and C18 grown at highergrowth temperatures. Our earlier study showed that for growthat higher temperatures [35], PL due to Tii was much stronger

than the PL due to Ov. Due to the lower concentration of Ov, theatmospheric O2 interacts with Ti3+ and oxidizes to Ti4+ insteadof injecting into the bulk and this may be a possible reason forthe absence of an Oi-related PL peak at 1.96 eV (1.99 eV) inB18 (C18). The lattice expansion of A18s compared to A18can be explained on the basis of the higher concentration of F+

and neutral Ov and the absence of O2−i , which results in a net

electrostatic repulsion between F+ and Ti cations. The latticecontraction due to the surface hydroxyl group in pure anataseTiO2 [9] is inconsistent with the lattice expansion in our rutileTiO2 samples A12 and A18s, which are associated with theadsorbed surface OH group, as confirmed from XPS studies.Therefore, we strongly believe that the intrinsic defects, suchas Ti4+

i , F+ and O2−i are primarily responsible for the lattice

parameter variation in our rutile TiO2 NS. A schematic diagramillustrating the lattice expansion due to the presence of oxygenvacancies and Ti interstitial defects, whereas lattice contractiondue to the electrostatic attraction between the Ti4+

i and O2−i

defects is shown in figure 9. In the case of Ov-mediatedexpansion, due to the missing O atom, the Ti–Ti and Ti–Obonds are relaxed. The nearest-neighbour Ti atoms moveoutward from the vacancy to strengthen the bonding with therest of the neighbouring O lattice. Though the next-nearest-neighbouring O atoms may move inward to fill the Ov site,the net outward movement of Ti atoms is higher than the netinward movement of O atoms. This results in lattice expansion.Lattice expansion due to Ti interstitials is more natural dueto the larger size of the Ti atoms and electrostatic repulsionbetween Ti4+ ions. On the other hand, lattice contraction iscaused by the electrostatic attraction between Ti4+

i and O2−i , as

shown by the green arrow. Note that the requirement of a sizesmaller than a critical size is not imperative in this mechanism.This calls for a revisit of the earlier reported results on the size-dependent lattice expansion in TiO2 NS.

4. Conclusions

Rutile TiO2 NS with NRs, nanopillars and nanoflower-like morphologies are synthesized by an acid-hydrothermalprocess. At higher growth temperatures, lattice expansionis observed, while lattice contraction is observed for thesamples grown with over a longer time. We also observedlattice expansion for the samples grown under constant stirringduring the reaction. Our studies reveal that the latticeexpansion/contraction depends on the types, concentrationand nature of various intrinsic defects that are created underdifferent growth and processing conditions. These NS arehighly crystalline as observed from SAED patterns and latticefringes of HRTEM analyses. The direct and indirect bandgapsof the NS grown under different growth conditions are variedfrom 2.90–3.01 eV and 2.68–2.86 eV, respectively. The NIRpeak at 1.43 eV in PL spectra is attributed to the Ti4+

i defectsin rutile TiO2 NS. This peak is enhanced with higher growthtemperatures and longer reaction durations. The relative shiftin Ti 2p and O 1s core level XPS spectral positions undervarious growth conditions hints towards lattice distortions inTiO2 NS. Based on our experimental observations, we believethat the presence of Ti4+

i , F+ and neutral oxygen vacancy

11

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

Figure 9. Schematic diagram illustrating lattice expansion and lattice contraction in TiO2 lattice due to intrinsic defects. Grey (large) ballsare Ti atoms and red (small) balls are oxygen atoms. The light pink (small) ball indicates the oxygen vacancy. The green (large) and black(small) balls indicate Tii and Oi, repectively. The arrow shows the direction of relaxation of atoms due to Ov, Tii and Oi defects.

defects are mainly responsible for lattice expansion, while theelectrostatic attraction between Ti4+

i and O2−i defects cause

the lattice contraction in rutile TiO2 NS. The intrinsic defect-induced lattice parameter variations may provide new routesto achieving novel functionalities in advanced materials thatcan be tailored for many technological applications.

References

[1] Ye M, Liu H-Y, Lin C and Lin Z 2013 Hierarchical rutile TiO2

flower cluster-based high efficiency dye-sensitized solarcells via direct hydrothermal growth on conductingsubstrates Small 9 312–21

[2] Liu B and Aydil E S 2009 Growth of orientedsingle-crystalline rutile TiO2 nanorods on transparentconducting substrates for dye-sensitized solar cells J. Am.Chem. Soc. 131 3985–90

[3] Kumar A, Madaria A R and Zhou C 2010 Growth of alignedsingle-crystalline rutile TiO2 nanowires on arbitrarysubstrates and their application in dye-sensitized solar cellsJ. Phys. Chem. C 114 7787–92

[4] Yu J, Qi L and Jaroniec M 2010 Hydrogen production byphotocatalytic water splitting over Pt/TiO2 nanosheets withexposed (0 0 1) facets J. Phys. Chem. C 114 13118–25

[5] Cao C, Hu C, Wang X, Wang S, Tian Y and Zhang H 2011 UVsensor based on TiO2 nanorod arrays on FTO thin filmSensors Actuators B 156 114–9

[6] Bayati M R, Ding J, Lee Y F, Narayan R J, Narayan J, Zhou Hand Pennycook S J 2012 Defect mediated photocatalyticdecomposition of 4-chlorophenol on epitaxial rutile thin

films under visible and UV illumination J. Phys.: Condens.Matter 24 395005

[7] Wolf S A, Awschalom D D, Buhrman R A, Daughton J M,von Molnar S, Roukes M L, Chtchelkanova A Y andTreger D M 2001 Spintronics: a spin-based electronicsvision for the future Science 294 1488–95

[8] Li G, Boerio-Goates J, Woodfield B F and Li L 2004 Evidenceof linear lattice expansion and covalency enhancement inrutile TiO2 nanocrystals Appl. Phys. Lett. 85 2059–61

[9] Li G, Li L, Boerio-Goates J and Woodfield B F 2005 Highpurity anatase TiO2 nanocrystals: near room-temperaturesynthesis, grain growth kinetics, and surface hydrationchemistry J. Am. Chem. Soc. 127 8659–66

[10] Swamy V, Menzies D, Muddle B C, Kuznetsov A,Dubrovinsky L S, Dai Q and Dmitriev V 2006 Nonlinearsize dependence of anatase TiO2 lattice parameters Appl.Phys. Lett. 88 243103

[11] Deshpande S, Patil S, Kuchibhatla S V N T and Seal S 2005Size dependency variation in lattice parameter and valencystates in nanocrystalline cerium oxide Appl. Phys. Lett.87 133113

[12] Tsunekawa S, Ishikawa K, Li Z Q, Kawazoe Y and Kasuya A2000 Origin of anomalous lattice expansion in oxidenanoparticles Phys. Rev. Lett. 85 3440–3

[13] Kuznetsov A Y, Machado R, Gomes L S, Achete C A,Swamy V, Muddle B C and Prakapenka V 2009 Sizedependence of rutile TiO2 lattice parameters determined viasimultaneous size, strain, and shape modeling Appl. Phys.Lett. 94 193117

[14] Liusai Y, Liping L, Minglei Z, Chaochao F and Guangshe L2013 Is there lattice contraction in multicomponent metaloxides? Case study for GdVO 4:Eu 3+ nanoparticlesNanotechnology 24 305701

12

J. Phys. D: Appl. Phys. 47 (2014) 215302 B Santara et al

[15] Palkar V R, Ayyub P, Chattopadhyay S and Multani M 1996Size-induced structural transitions in the Cu–O and Ce–Osystems Phys. Rev. B 53 2167–70

[16] Wasserman H J and Vermaak J S 1970 On the determination ofa lattice contraction in very small silver particles Surf. Sci.22 164–72

[17] Wasserman H J and Vermaak J S 1972 On the determination ofthe surface stress of copper and platinum Surf. Sci.32 168–74

[18] Yu D K, Zhang R Q and Lee S T 2002 Structural transition innanosized silicon clusters Phys. Rev. B 65 245417

[19] Ji Y, Guo L, Xu H, Liu J, Li X, Li Y, Wu Z and Simon P 2003Synthesis and phase transition under ultra-high pressure ofZnS nanoparticles modified by sodium bis(2-ethylhexyl)sulfosuccinate Phys. Status Solidi a 198 210–4

[20] Zhang J-Y, Wang X-Y, Xiao M, Qu L and Peng X 2002 Latticecontraction in free-standing CdSe nanocrystals Appl. Phys.Lett. 81 2076–8

[21] Iddir H, Ogut S, Zapol P and Browning N D 2007 Diffusionmechanisms of native point defects in rutile TiO2: Ab initiototal-energy calculations Phys. Rev. B 75 073203

[22] Williamson G K and Hall W H 1953 X-ray line broadeningfrom filed aluminium and wolfram Acta Metall.1 22–31

[23] Qadri S B, Yang J P, Skelton E F and Ratna B R 1997Evidence of strain and lattice distortion in lead sulfidenanocrystallites Appl. Phys. Lett. 70 1020–1

[24] Giri P K 2009 Strain analysis on freestanding germaniumnanocrystals J. Phys. D: Appl. Phys. 42 245402

[25] Oh S-M and Ishigaki T 2004 Preparation of pure rutile andanatase TiO2 nanopowders using RF thermal plasma ThinSolid Films 457 186–91

[26] Swamy V 2008 Size-dependent modifications of the first-orderRaman spectra of nanostructured rutile TiO2 Phys. Rev. B77 195414

[27] Swamy V, Muddle B C and Dai Q 2006 Size-dependentmodifications of the Raman spectrum of rutile TiO2 Appl.Phys. Lett. 89 163118

[28] Ekuma C E and Bagayoko D 2011 Ab-initio electronic andstructural properties of rutile titanium dioxide Japan. J.Appl. Phys. 50 101103

[29] Pascual J, Camassel J and Mathieu H 1978 Fine structure inthe intrinsic absorption edge of TiO2 Phys. Rev. B18 5606–14

[30] Amtout A and Leonelli R 1995 Optical properties of rutilenear its fundamental band gap Phys. Rev. B 51 6842–51

[31] Kasowski R V and Tait R H 1979 Theoretical electronicproperties of TiO2 (rutile) (0 0 1) and (1 1 0) surfaces Phys.Rev. B 20 5168–77

[32] Khan M A, Kotani A and Parlebas J C 1991 Electronicstructure and core level photoemission spectra in TiO2

compounds J. Phys.: Condens. Matter 3 1763[33] Mathieu H, Pascual J and Camassel J 1978 Uniaxial stress

dependence of the direct-forbidden and indirect-allowedtransitions of TiO2 Phys. Rev. B 18 6920–9

[34] Poumellec B, Durham P J and Guo G Y 1991 Electronicstructure and X-ray absorption spectrum of rutile TiO2 J.Phys.: Condens. Matter 3 8195

[35] Santara B, Giri P K, Imakita K and Fujii M 2013 Evidence forTi interstitial induced extended visible absorption and nearinfrared photoluminescence from undoped TiO2

nanoribbons: an in situ photoluminescence study J. Phys.Chem. C 117 23402–11

[36] Santara B, Giri P K, Imakita K and Fujii M 2013 Evidence ofoxygen vacancy induced room temperature ferromagnetismin solvothermally synthesized undoped TiO2 nanoribbonsNanoscale 5 5476–88

[37] Shi J, Chen J, Feng Z, Chen T, Lian Y, Wang X and Li C 2007Photoluminescence characteristics of TiO2 and theirrelationship to the photoassisted reaction of water/methanolmixture J. Phys. Chem. C 111 693–9

[38] Yamada Y and Kanemitsu Y 2010 Blue photoluminescence ofhighly photoexcited rutile TiO2: nearly degenerateconduction-band effects Phys. Rev. B 82 113103

[39] Zhao Q, Wu P, Li B-L, Lu Z-M and Jiang E-Y 2008Room-temperature ferromagnetism in semiconductingTiO2−δ nanoparticles Chin. Phys. Lett. 25 1811

[40] Zhuang J, Weng S, Dai W, Liu P and Liu Q 2012 Effects ofinterface defects on charge transfer and photoinducedproperties of TiO2 bilayer films J. Phys. Chem. C116 25354–61

[41] Sterrer M, Fischbach E, Risse T and Freund H-J 2005Geometric characterization of a singly charged oxygenvacancy on a single-crystalline MgO (0 0 1) film by electronparamagnetic resonance spectroscopy Phys. Rev. Lett.94 186101

[42] Chong S V, Kadowaki K, Xia J and Idriss H 2008 Interestingmagnetic behavior from reduced titanium dioxide nanobeltsAppl. Phys. Lett. 92 232502–3

[43] Li L, Li G, Xu J, Zheng J, Tong W and Hu W 2010 Insightsinto the roles of organic coating in tuning the defectchemistry of monodisperse TiO2 nanocrystals for tailoredproperties Phys. Chem. Chem. Phys. 12 10857–64

[44] Marrocchelli D, Bishop S R, Tuller H L and Yildiz B 2012Understanding Chemical expansion in non-stoichiometricoxides: ceria and zirconia case studies Adv. Funct. Mater.22 1958–65

[45] Lee H-Y, Clark S J and Robertson J 2012 Calculation of pointdefects in rutile TiO2 by the screened-exchange hybridfunctional Phys. Rev. B 86 075209

[46] Janotti A, Varley J B, Rinke P, Umezawa N, Kresse G and Vande Walle C G 2010 Hybrid functional studies of the oxygenvacancy in TiO2 Phys. Rev. B 81 085212

[47] Wendt S, Sprunger P T, Lira E, Madsen G K H, Li Z,Hansen J Ø, Matthiesen J, Blekinge-Rasmussen A,Lægsgaard E, Hammer B and Besenbacher F 2008 The roleof interstitial sites in the Ti3d defect state in the band gap oftitania Science 320 1755–9

[48] Onishi H and Iwasawa Y 1996 Dynamic visualization of ametal-oxide-surface/gas-phase reaction: time-resolvedobservation by scanning tunneling microscopy at 800 KPhys. Rev. Lett. 76 791–4

[49] Bennett R A, Stone P, Price N J and Bowker M 1999 Two(1×2) reconstructions of TiO2(1 1 0): surface rearrangementand reactivity studied using elevated temperature scanningtunneling microscopy Phys. Rev. Lett. 82 3831–4

[50] Gorai P, Hollister A G and Seebauer E G 2013 Electrostaticdrift effects on near-surface defect distribution in TiO2

Appl. Phys. Lett. 103 141601[51] Hollister A G, Gorai P and Seebauer E G 2013 Surface-based

manipulation of point defects in rutile TiO2 Appl. Phys.Lett. 102 231601

13