Journal ofMaterials Chemistry C

PAPER

Publ

ishe

d on

18

Sept

embe

r 20

14. D

ownl

oade

d by

Tez

pur

Uni

vers

ity o

n 04

/10/

2014

13:

29:0

8.

View Article OnlineView Journal

Room temperatu

Department of Physics, Tezpur University,

biswajit@tezu.ernet.in; Tel: +91 940131317

† Electronic supplementary informa10.1039/c4tc01070a

Cite this: DOI: 10.1039/c4tc01070a

Received 22nd May 2014Accepted 12th September 2014

DOI: 10.1039/c4tc01070a

www.rsc.org/MaterialsC

This journal is © The Royal Society of

re ferromagnetism in SnO2

nanoparticles: an experimental and densityfunctional study†

Pawan Chetri, Biswajit Choudhury* and Amarjyoti Choudhury

Room temperature ferromagnetism is observed in pristine SnO2 nanoparticles prepared via sol–gel

method. Microstructural studies with XRD and Raman reveal the coexistence of SnO and SnO2 phases in

vacuum annealed SnO2. Electron paramagnetic resonance (EPR) results confirm the presence of

paramagnetic trapped hole in lattice oxygen atoms and superoxide radical species in air and vacuum

annealed SnO2. It is observed that growth conditions have significant influence on the magnetism of the

nanoparticle. Air and vacuum annealed SnO2 display clear ferromagnetic behavior at room temperature.

However, the saturation magnetization decreases in vacuum annealed SnO2. We have compared the

experimental results with ab initio density functional calculations and observed that the Sn vacancies are

the main reason of magnetism in SnO2. The pattern of theoretical formation energy values shows that it

is easier to form Sn vacancies in SnO2 under O rich limit (similar to air annealing). This is in conformity

with the experimentally determined higher saturation magnetization values in air annealed SnO2. The

spin polarization of lattice oxygen atoms by the presence of Sn vacancy gives rise to magnetism.

1. Introduction

In the eld of spintronics it is important to design semi-conductors with ferromagnetically polarized carriers at roomtemperature (RT).1 The understanding of spin-based devices forquantum information depends upon the improvement ofvarious nanoscale semiconductor architectures that allowperfect isolation and coherent control of the quantum connedspin states.2 Room temperature ferromagnetism in transitionmetal doped semiconductor oxides has been widely reported.2–5

However, the formation of extrinsic magnetic impurity clustersis an obstacle in the practical applications of these materials.Hence, the search is for RT ferromagnetism in a semiconductormaterial with appreciable saturation magnetization. Undopedoxide materials seem to be better candidates for spintronics, asthe possibility of extrinsic magnetic phase separation becauseof magnetic impurities could be avoided. There are reports offerromagnetism in undoped SnO2 nanosystems.6–8 A fewauthors7–10 have claimed that the observed ferromagnetism inpristine SnO2 is because of the presence of oxygen vacancies. Onthe other hand, a few authors11,12 have claimed on theoreticalgrounds that the main reason behind the observed ferromag-netism in the SnO2 system is the presence of Sn4+ vacancy.Espinosa et al.12 conrmed that Sn4+ vacancies are the main

Napaam, Tezpur-784028, India. E-mail:

7

tion (ESI) available. See DOI:

Chemistry 2014

source of magnetism in SnO2 thin lms. Similarly, Rusydiet al.13 and Qi et al.14 experimentally and theoretically justiedthat the cationic (Ti) vacancy mediated ferromagnetism in Tadoped TiO2. However, Ghosh et al. reported diamagnetism inpure SnO2 nanoparticles.15 Therefore, the exact origin of ferro-magnetism in these nanoparticles is not clearly explained. Inthis article we have performed both experimental and theoret-ical studies to establish the dependence of growth conditions(oxygen rich or oxygen decient) on the magnetism of SnO2

nanoparticles. Initial characterization results of XRD andRaman show the coexistence of SnO and SnO2 phases invacuum annealed SnO2. We have used density functional (DFT)calculation using MedeA VASP (Vienna ab initio simulationpackage) 5.2 for the investigation of the intrinsic causes ofmagnetism in pristine SnO2 prepared under both air andvacuum conditions. Finally, we have veried on experimentaland theoretical grounds the inuence of cationic Sn4+ vacancyand anionic oxygen vacancy on ferromagnetism in SnO2.

2. Synthesis of SnO2 nanoparticles

The synthesis of SnO2 followed a sol–gel procedure.16 Thereagents were of AR grade. The reagents used were SnCl2$2H2O(98% Merck), double distilled water, ethanol (Merck), HCl(Merck), and ammonia solution (Merck). 1 g of SnCl2$2H2O wasmixed with 11 ml of double distilled water and 5 ml of ethanol.Then, the above solution was stirred for 15 minutes and a fewdrops of HCl were added. Ammonia solution was added drop-wise till a pH ¼ 10 was attained. The prepared gel was divided

J. Mater. Chem. C

Journal of Materials Chemistry C Paper

Publ

ishe

d on

18

Sept

embe

r 20

14. D

ownl

oade

d by

Tez

pur

Uni

vers

ity o

n 04

/10/

2014

13:

29:0

8.

View Article Online

into two parts. One part was dried at 60 �C and annealed at200 �C under air atmosphere. The other part was dried andannealed under vacuum at 60 �C and 200 �C, respectively. Thesample annealed under air was named TO200, while the sampleannealed under vacuum is quoted in this manuscript asTOVac200.

Fig. 1 XRD pattern of TO200 and TOVac200.

3. Experimental and computationaldetails

The X-ray diffraction pattern of the nanoparticles was moni-tored in a Bruker AXS D8 focus X-ray diffractometer (XRD)equipped with intense Cu Ka radiation (l ¼ 0.154 nm). Thescanning speed was 0.5 s per step with an increment of 0.0057�.High resolution transmission electron microscope (HRTEM)images of the prepared nanoparticles were obtained with a JEOLJEM 2010 transmission electron microscope operating at avoltage of 200 kV. Elemental analysis of the as-prepared SnO2

nanopowder was studied with inductively coupled plasmaoptical emission spectroscopy (ICP-OES) in Perkin-ElmerOptima 2100 DV by completely digesting the sample in a smallvolume of conc. HNO3. Raman spectra of the samples wereobtained in Renishaw inVia Raman spectrometer with an Ar+

ion laser of 514.5 nm wavelength with spectral resolution of 0.3cm�1. X-ray photoelectron spectra (XPS) were recorded on aKratos Axis 165 XPS spectrometer equipped with dualaluminum–magnesium anodes using Mg Ka radiation (hy ¼1253.6 eV). The dc magnetization measurements for the as-prepared sample were performed with a Quantum DesignMPMS superconducting quantum interference device (SQUID)VSM Ever Cool system at room temperature (RT). The ZFC (zeroeld cooling) and FC (eld cooling) temperature dependentmagnetization (M–T) measurements were performed in thetemperature range of 5–350 K. Electron paramagnetic reso-nance (EPR) was performed using a JES-FA200 ESR Spectrom-eter in X band, sensitivity �7 � 109 spins per 0.1 mT and aresolution �2.35(micro) T or better. We performed densityfunctional calculations using generalized gradient approxima-tion (GGA) functional with (PBE) to describe the electron–elec-tron exchange and correlation effects. The density functionaltheory (DFT) equations were solved via a projector augmentedwave (PAW) method using plane wave basis set as implementedin Vienna ab initio Simulation Package (VASP) and interfacedwith MedeA technology platform. All the systems were calcu-lated on a real space grid dened with a plane-wave energy cut-off of 480 eV. A spacing of 0.3 A�1 was used for K point samplingin both supercell and surface calculations. All the atomicpositions were fully relaxed until all atomic forces were smallerthan 0.01 eV A�1. A 2 � 2 � 2 supercell was built with the unitcell of SnO2 in rutile form and with a 10 A vacuum spacebetween two supercells. The space group of pure SnO2 wastaken as P42/mnm, and the supercell contained 48 atoms(Sn16O32). The effect of Sn vacancy was investigated by removinga Sn atom from the centre position of the supercell. A similarprocess was performed for O vacancy, but for Sn interstitial anextra Sn was added in the above mentioned supercell (Sn16O32).

J. Mater. Chem. C

In all the cases, spin polarized structural optimization wasperformed such that the systems obtained their minimumenergy conguration by relaxing all the atoms in all directionsand calculated the magnetic moments. For the surface (110) wealso considered 48 atoms and relaxed the structure. A 10 Avacuum slab was considered between two surfaces.

4. Results and discussion4.1 Structural characterizations

X-ray diffraction patterns of the prepared SnO2 nanoparticlesare shown in Fig. 1. All the diffraction peaks were indexed to arutile (JCPDS-41-1445) phase. XRD graph indicated the presenceof SnO (101), (200) and (112) characteristic peaks along withSnO2 phase in TOVac200, whereas these peaks are missing inTO200. Surface oxygen vacancies are easily formed in the as-prepared SnO2 under vacuum annealing. SnO2 compensatesthese oxygen vacancies by transforming itself to SnO.17 Crys-tallite sizes of all the prepared systems were calculated using theScherrer's formula in the EVA soware from the XRD instru-ment. The determined values are 5.44 nm for TO200 and 12.5nm for TOVac200. We believe that larger crystallites in TOVac200are most likely because of the coexistence of SnO2 and SnOcrystalline phases. We further studied the morphology, size andshape of SnO2 with high resolution transmission electronmicroscope (HRTEM). Fig. 2a and b displays the HRTEM imagesof TOVac200 and TO200, respectively. These images show thatparticles are of varied sizes and shapes. Particles in TOVac200are larger and lesser agglomerated than in TO200. The sizedistribution plot shows that the particles of 13 nm and 6 nm arehighly populated in TOVac200 and TO200. This compliments thecrystallite sizes obtained from XRD. Raman spectroscopy is anefficient tool to explore the vibrational modes within thesystem. The Raman spectra of both the prepared systems areshown in Fig. 3. A normal mode can be described by the 3n (n isthe number of atoms in the primitive cell) coordinates giving

This journal is © The Royal Society of Chemistry 2014

Fig. 2 High resolution TEM images of TOVac200 (a and b) and TO200(c and d). Histogram in the inset of (a) and (c) shows the size distri-bution of the nanoparticles.

Fig. 3 Room temperature Raman spectra of TOVac200 and TO200.The position of A1g peak is shifted towards higher energy in TOVac200.

Fig. 4 Room temperature electron paramagnetic resonance (EPR)spectra of TOVac200 and TO200.

Paper Journal of Materials Chemistry C

Publ

ishe

d on

18

Sept

embe

r 20

14. D

ownl

oade

d by

Tez

pur

Uni

vers

ity o

n 04

/10/

2014

13:

29:0

8.

View Article Online

the displacements of the atoms from their equilibrium posi-tions. Rutile SnO2, having a 6 atoms unit cell results in 18possible vibrations. The normal lattice vibration at the G pointof the Brillouin zone is as follows,18,19

G ¼ G1+(1A1g) + G2

+(1A2g) + G3+(1B1g) + G4

+(1B2g) + G5�(1Eg)

+ G1�(1A2u) + 2G4

�(B1u) + 4G5+(Eu) (1)

Two modes (A2u and triply degenerate Eu) are infrared active.Eu has two acoustic and one optic mode, while A2u has one optic

This journal is © The Royal Society of Chemistry 2014

and one acoustic mode. Four modes, viz., A1g, B1g, B2g anddoubly degenerate Eg are found to be Raman active, and two(A2g, B1u) modes are silent. Raman active modes for rutile SnO2

(bulk) are expected at 123, 476, 634 and 778 cm�1 for B1g, Eg, A1g

and B2g, respectively.20 In this study, only the A1g peak isprominent in both the samples. Both the samples exhibit ahump at 574 cm�1, which is refereed as the A1 mode and is dueto surface phonon in SnO2 nanocrystallites.21 One small peak isobserved at 210 cm�1 inTOVac200, which is otherwise absent inTO200. This peak appears owing to the presence of a smallfraction of SnO phase alongside the SnO2 phase.22,23 The Ramanactive modes of SnO2 (or SnO) are because of the polarizationoccurring due to the displacements of Sn and oxygen in Sn andoxygen sub-lattice in SnO2. Any changes in the stoichiometry (bythe presence of oxygen vacancies) in the basic SnO6 unit (inSnO2) and in SnO4 (in SnO) may signicantly reduce thepolarization of Sn–O bond; thus, might inuence the positionand intensity of Raman modes. We have observed blue shi inthe position of the A1g, as well as surface A1 phonon Ramanmode in TOVac200 as compared with TO200. The blue shi canbe associated with the presence of oxygen vacancies in theTOVac200. The presence of oxygen vacancies reduces thenumber of polarized Sn–O bonds in TOVac200, resulting a shiin the position of Raman peaks. Similar type of oxygen defectassisted blue shi of Raman peaks is reported in vacuumannealed SnO2 as compared with its air annealed counterpartby Shek et al.21 Room temperature EPR spectra of TOVac200 andTO200 are shown in Fig. 4. The g value is calculated using therelation

g ¼ hn/mBB (2)

Where h is the Planck's constant, n is the applied microwavefrequency (9.878 GHz), mB is the Bohr magneton and B is theresonance magnetic eld. TO200 contains an intense EPRsignal at g ¼ 2.04. This signal can be attributed to the hole

J. Mater. Chem. C

Fig. 6 X-ray photoelectron spectra (XPS) of TO200.

Journal of Materials Chemistry C Paper

Publ

ishe

d on

18

Sept

embe

r 20

14. D

ownl

oade

d by

Tez

pur

Uni

vers

ity o

n 04

/10/

2014

13:

29:0

8.

View Article Online

trapped at lattice oxygen site (O� center).24,25 TOVac200 containsa broad EPR signal at g � 2.17. This EPR signals resemble the g-value corresponding to superoxide radical.26,27 The superoxideradicals (O2

�) are formed by the transfer of trapped electronspresent in oxygen vacancies to the surface adsorbed O2 mole-cule in SnO2. These superoxide radicals are strongly attached toSn4+ or Sn2+ species on the surface. Surfaces in TOVac200 con-taining high density of oxygen vacancies with trapped electronsare active sites for easy adsorption of the O2 molecule. There-fore, TO200 contains an EPR signal because of the hole trappedat oxygen lattice sites, but does not contain any signal corre-sponding to oxygen vacancies or superoxide species. On theother hand, TOVac200 contains an EPR signal corresponding tooxygen vacancies only, but no signals of hole trapped oxygen.The relative number of spins participating in the broadeningand intensity of the EPR signal are calculated by the relation28

Ns f I(DH)2 (3)

where I and (DH) are the intensity and FWHM of the intensepeak, respectively, calculated from the integrated spectrum ofeach sample. The relative number of spins (Ns) due to electrontrapped in oxygen vacancies is higher than that of hole trappedat oxygen lattice sites, viz., Ns in TOVac200 > TO200. We havecarried out a XPS analysis of TOVac200 (Fig. 5) and TO200(Fig. 6). The standard 3d5/2 spectrum for Sn4+ and Sn2+ arefound to be at 486.72 and 487.4 eV, respectively.29,30 The sepa-ration between Sn 3d5/2 and Sn 3d3/2 should be around 8.5 eV.30

In TO200, the peaks obtained for Sn4+ are at 486.9 and 495.4 eV,

Fig. 5 X-ray photoelectron spectra (XPS) of TOVac200.

J. Mater. Chem. C

while in TOVac200 the peaks for Sn4+ are obtained at 486.7 and495.2 eV. The peaks of Sn2+ in TOVac200 are found to be at 488.1and 496.6 eV. We have not observed any Sn2+ peak in TO200,which conrms the absence of SnO in TO200. The separationbetween 3d5/2 and 3d3/2 spectra of Sn4+ and Sn2+ is 8.5 eV,although some peaks are shied to higher energies incomparison to the standard spectrum. The peaks of Sn2+ arefound to be shied to a higher energy by 0.7 eV with respect tothe standard sample value (given above). The O 1s spectra ofboth the samples are deconvoluted into two peaks. The peak atlower energy can be ascribed to lattice oxygen (OL), while thehigher energy one is because of the adsorbed oxygen at thesurface (OH).29,31 The percentage of concentrations of OL and OH

is found to be 59.5 and 40.5 for TO200, respectively. While it is53.6 and 46.4 for TOVac200, respectively. The amount of adsor-bed oxygen in TOVac200 increases than in TO200. Actually, inTOVac200, the number of oxygen vacancies increases whensurface becomes more reactive; hence, adsorbs a large numbersof free oxygen molecules.

4.2 Magnetic properties study

Prior to the magnetic measurements, an ICP-OES elementalanalysis for the magnetic impurities Fe, Co and Ni was carriedout. The concentrations of Fe, Co and Ni are 0.088, 0.003 and0.085 ppm, respectively. The form of these impurities can be inpure metal phase or in metal oxide phase. FeO and a-Fe2O3 areanti ferromagnetic, whereas g-Fe2O3 is ferromagnetic. Fe3O4

can easily oxidize to g-Fe2O3. Therefore, in our case, we have to

This journal is © The Royal Society of Chemistry 2014

Table 1 Crystallite size, saturation magnetization (Ms), coercivity (Hc)and remanence (Mr) of TO200 and TOVac200

SystemCrystallite size (nm) fromScherrer formula

Ms

(emu per g) Hc (Oe)Mr

(emu per g)

TO200 5.44 0.16 63 0.005TOVac200 12.50 0.14 200 0.011

Paper Journal of Materials Chemistry C

Publ

ishe

d on

18

Sept

embe

r 20

14. D

ownl

oade

d by

Tez

pur

Uni

vers

ity o

n 04

/10/

2014

13:

29:0

8.

View Article Online

consider the magnetization of Fe as metal or as g-Fe2O3. Themagnetization of Fe and g-Fe2O3 for 0.088 ppm is in the 10�3

order. Similarly, metallic Co and Ni are ferromagnetic, whereasCo3O4 and NiO are antiferromagnetic. The maximum contri-bution from both the phases of Co is in the order of 10�4 emuper g, and from Ni is 10�3 emu per g. In this study, we haveobtained the magnetization of SnO2 in the order of 10�1 emuper g. Therefore, we conclude that the magnetization valuesreported in this work are not because of impurities, but it isintrinsic to SnO2. Fig. 7(a) shows room temperature (300 K)hysteresis loops of TOVac200 and TO200. Even though both thesamples show RT ferromagnetism, the saturation magnetiza-tion (Ms) appears to be higher in TO200 than in TOVac200. Thevalues of Ms and Hc (coercivity) and Mr (remanence) of both thesamples are shown in Table 1.Ms and Hc for TO200 are found tobe 0.16 emu per g and 63 Oe, respectively. As Table 1 shows, Hc

decreases when Ms increases, which is also consistent with theequation32 Hc ¼ 2K1/Ms, where K1 is the anisotropy constant.Arrott–Belov–Kouvel (ABK)28 plot of TOVac200 and TO200 ispresented in Fig. 7(b). In the ABK plot,M2 is plotted againstH/Mand the spontaneous magnetization is measured from theintercept of the linear tting of the curve on the Y-axis. ABK plotsignies the presence of ferromagnetic interactions andmeasures the spontaneous magnetization in the sample. Thespontaneous magnetizations of TO200 and TOVac200 are 0.03and 0.02 emu per g, respectively. The increase in spontaneousmagnetization of TO200 indicates the presence of long rangeferromagnetic ordering. For a better understanding of themagnetic properties in the system, we have also carried out zeroeld cooling (ZFC) and eld cooling (FC) measurements ofTO200 in the temperature range of 4 to 350 K at an applied eldof 100 Oe, as shown in Fig. 8. The bifurcation between FC andZFC up to 350 K indicates the presence of ferromagnetic inter-actions in the system. Several factors might affect this bifurca-tion, including magnetic anisotropy in the system, coercive eldand the manner by which FC and ZFC measurements are beingcarried out. Absence of a characteristic sharp peak or hump inthe ZFC curve signies that the observed ferromagnetism isintrinsic in nature to the sample and there is no blocking of

Fig. 7 (a) Room temperature M–H curves of TO200 and TOVac200. (b)magnetization in TO200 and TOVac200.

This journal is © The Royal Society of Chemistry 2014

spins. Both M–H and M–T measurements indicate the presenceof room temperature ferromagnetism in the system. The highvalue of the Ms of TO200 as compared to the vacuum annealedsample indicates that vacuum annealing do affect the magneticproperties. It is known that the annealing of SnO2 nanoparticlesunder vacuum removes the loosely bound oxygen atoms on thesurface and generates oxygen vacancies. The paramagneticoxygen species that may contribute towards magnetism are (i)oxygen vacancies with single trapped electron8 (F+) and (ii)single electron trapped oxygen molecules (O2

�). Although wehave not found direct conrmation of the presence of F+ centerfrom EPR, the presence of O2� signal might indicate that F+

centers are present on the surface. The single trapped electronsmight be easily transformed to adsorbed O2molecules and formO2

� species. We could believe that vacuum annealing hasincreased the density of oxygen vacancies; thus, the density ofparamagnetic superoxide species. Because of the presence ofmore electron spins in TOVac200, one could expect a highmagnetization in TOVac200 as compared with TO200. Instead,the lowering of both saturation and spontaneous magnetizationin TOVac200 might indicate that the oxygen vacancies (or tomention superoxide species) have negative impact on the totalmagnetization. Therefore, it is likely that cationic Sn4+ vacan-cies might be affecting the ferromagnetism. Earlier reportssummarized that cationic vacancies could produce magnetismin different oxide systems.12–14,33,34 Initially prepared SnO2

nanocrystallites may contain considerable numbers of Sn4+ andoxygen vacancies on the surface. On air annealing of the as-prepared SnO2 powder at 200 �C, most of the oxygen vacanciesdisappeared and thus, only Sn4+ vacancies were le. These Sn4+

Arrott–Belov–Kouvel (ABK) plot for the determination of spontaneous

J. Mater. Chem. C

Fig. 8 Zero field cooling (ZFC) and field cooling (FC) temperaturedependent magnetization (M–T) curves of TO200. Inset of the figureshows the extended part of the curve in the temperature range of200–350 K to display the complete separation of ZFC and FCupto 350 K.

Journal of Materials Chemistry C Paper

Publ

ishe

d on

18

Sept

embe

r 20

14. D

ownl

oade

d by

Tez

pur

Uni

vers

ity o

n 04

/10/

2014

13:

29:0

8.

View Article Online

vacancies should contribute to magnetism, but there is ambi-guity in the literature reports on the observed magnetism inSnO2.7,9,35 Our experimental results indicate that oxygen andSn4+ vacancies have negative and positive contribution to theobserved magnetism, respectively. In order to establish theroles of oxygen and Sn4+ vacancies in ferromagnetism, we alsoperformed theoretical calculations based on DFT. Thus, we havecalculated electronic properties of several point defects startingfrom vacancies to interstitial in two different model structuresof SnO2 {Surface (110) and Supercell}.

4.3 Electronic structure and magnetic calculation

We have computed both surface and supercell structures andperformed a detail DFT based investigation on different types ofvacancies to attain a model that explains the observed magne-tism in a pristine SnO2 nanosystem annealed in vacuum andair. We have considered SnO2 (110) surface having 48 atoms andrelaxed the structure to settle for minimum energy congura-tion. As we know, the surface to volume ratio in the nano-regimeis very high. The contribution from the surface is very prom-inent and its effect onmagnetism has to be considered. We havechosen the surface calculation because most of the Sn atomsand oxygen atoms are not adequately compensated with bonds.The magnetism arises in the system because of the 2p state of O(a). Both these oxygens are single bonded and are notcompensated. Fig. 9 shows the total density of states (TDOS)plot for SnO2, where the partial density of states (PDOS) of O (a)split up near the Fermi level (0 eV). The TDOS of SnO2 (110)shows the hybridization of Sn 5s and Sn 5p states with O 2pstates at the top of the valence band, where Sn 5s state isdominant over Sn 5p, which indicates the formation of freeholes.36 The asymmetric distribution of spin density around theFermi level indicates the presence of magnetism in the system.

J. Mater. Chem. C

If we introduce any further Sn4+ vacancy or oxygen vacancy or aSn interstitial, the magnetism appears to be decreased. Thisshows that the resultant magnetic moment of 1.2mB is becauseof the reduction of SnO2 from bulk to surface, and that anyfurther impurity does not increase the magnetic moment. ThePDOS of oxygen proves the fact that these are the spin polarizedoxygens that induce magnetism in pure layer of SnO2. We havealso investigated the supercell of SnO2 containing a total of 48atoms. We studied the effect of intrinsic defects on the structureand the magnetism. Pure SnO2 supercell Sn16O32 has nomagnetism at all, and this is quite expected as there are nodangling bonds of Sn or oxygen. We initially introduced oxygenvacancies, next Sn vacancies, and then Sn interstitial. The defectformation energy Ef (D) is calculated using.

Ef (D) ¼ Ed � Ep + 1/2 nOE (O2) + nSnE (Sn) + nOmO + nSnmSn(4)

where Ed, Ep, E (O2) and E (Sn) are the total energy of thesupercell with defect, pure supercell, oxygen molecule andsingle Sn atom. nO and nSn are the number of O and Sn vacancy,while mO and mSn are the chemical potentials of O and Sn,respectively. The values of mO and mSn are dependent on theexternal environment i.e., on the oxygen partial pressure. Thetwo extreme conditions are known as O poor and O richcondition.17 The value of formation energy of Sn15O32 (single Snvacancy), Sn16O31 (single O vacancy) and Sn17O32 (single Sninterstitial) are shown in Table 2. We have observed thatforming an oxygen vacancy under the O poor conditions andforming a Sn vacancy under the O rich conditions is energeti-cally favorable. The formation of Sn interstitial does not appearto be energetically favorable neither in O rich nor O poorconditions. This implies that the formation of oxygen vacanciesin TOVac200 and the formation of Sn vacancies in TO200 areeasier to achieve. The formation energy values shown in Table 2cannot be directly compared to the required experimentalformation energy values in TOVac200 and TO200, but still thepattern of formation of defects can be understood. Themagnetic moment values determined based on DFT results arelisted in Table 2. We examined it for different positions foroxygen vacancies, but it did not result in any magnetism. FromTable 2 it is clear that in an ordered SnO2, we get maximumvalue of magnetization when Sn4+ vacancies are introduced.From our calculations, we have found that the major contri-bution in Sn15O32 comes from the surrounding oxygens, namedas O (a–f), as shown in magnetization density mapping ofFig. 10. Further O (a) has the same contribution as that of O (c),similarly O (b) matches that of O (d) and O (e) matches that of O(f). We have checked the splitting of oxygen near the Fermi levelby plotting PDOS of oxygen, as shown in Fig. 11. The asym-metric distribution of spin up and spin down states for oxygen(a–f) near and above the Fermi level conrms the magneticnature of the system. We have also calculated the formationenergy of oxygen and Sn vacancy on the surface (110) plane ofSnO2 and inserted the values in Table 2. We have found that theformation energy of oxygen vacancies in O decient conditionsis very low, while its formation in O rich conditions is relatively

This journal is © The Royal Society of Chemistry 2014

Fig. 9 Total density of states (TDOS) plot of SnO2 (110) and PDOS plot of O (a), Sn 5 s and Sn 5p. Red lines indicate p state while black lineindicated s state. The Fermi level is shown by blue line (at 0 eV).

Table 2 Different systems, defect formation energy (eV) in O excess and O poor condition and total Magnetic moment

System

Defect formation energy Ef (D) (eV)Total magneticmoment (mB)O excess condition O poor condition

Sn16O32 — — 0.0Sn15O32 1.3 7.5 3.9Sn16O31 3.5 0.4 0.0Sn17O32 14.3 8.2 0.0SnO2(110) — — 1.2SnO2(110) with Sn vacancy 0.9 7.1SnO2(110) with O vacancy 1.6 �1.4

Paper Journal of Materials Chemistry C

Publ

ishe

d on

18

Sept

embe

r 20

14. D

ownl

oade

d by

Tez

pur

Uni

vers

ity o

n 04

/10/

2014

13:

29:0

8.

View Article Online

higher (with high formation energy). The formation energy ofSn vacancies in O decient conditions is very high, while thecorresponding value is very low in O rich conditions. In the bulkcalculations, the formation energy of oxygen vacancies and Snvacancies in the two different conditions are found to presentsame patterns though values are higher in this case, ascompared to surface ones (110). Our claim that Sn vacancies arethe source of magnetism can be veried by the result offormation of Sn vacancies in SnO2. Both in bulk and surface Snvacancies formations are easier to achieve in O rich conditionsthan in O decient conditions. On the other hand, oxygenvacancies instantaneously form under O decient conditions inSnO2 (110) surface and at slightly high energy cost in bulk SnO2.

This journal is © The Royal Society of Chemistry 2014

4.4 Correlation between DFT and the experimental results

From the experimental and theoretical investigations, it seems thatthe origin of magnetism in SnO2 is not because of oxygen vacan-cies, but to Sn vacancies. Another possibility for this magnetism isto be holemediated. FromDFT calculations, it is observed that thevalence band splitting and spin polarization is observed near theFermi level in Sn15O32. This splitting is because of the formation offour holes (Sn4+ vacancies). The holes are distributed on the latticeoxygen nearest to the Sn vacancies and the surrounded oxygen isspin polarized. However, in TOVac200 the number of lattice oxygensurrounding Sn vacancies will be diminished. Therefore, the lossof polarized oxygen around the Sn vacancies leads to the reductionin the average magnetization in TOVac200.12 Although our EPR

J. Mater. Chem. C

Fig. 10 (a) Magnetization density in SnO2 surface (110). (b) 2 � 2 � 2supercell of SnO2 (right) and magnetization density in the system dueto Sn vacancies. Yellow atom denotes Sn, red denotes oxygen andwhite denotes Sn vacancy.

Journal of Materials Chemistry C Paper

Publ

ishe

d on

18

Sept

embe

r 20

14. D

ownl

oade

d by

Tez

pur

Uni

vers

ity o

n 04

/10/

2014

13:

29:0

8.

View Article Online

results indicate that the relative numbers of unpaired spins on theoxygen vacancies are higher in TOVac200, these unpaired spinshave no positive inuence on the total magnetization. Zhang et al.reported that point defects with local moments may not neces-sarily contribute towards ferromagnetism.37 In the present study,anionic point defects (oxygen vacancies) have no role in ferro-magnetism, whereas cationic defects (Sn vacancies) certainly

Fig. 11 TDOS plot of Sn15O32 supercell (with one Sn vacancy) and PDOSstate, whereas the Fermi level is shown by the blue line (at 0 eV).

J. Mater. Chem. C

provide the major contribution to ferromagnetism. From ourresults, we have found that the formation of Sn vacancies isfavorable in TO200 and the formation of oxygen vacancies isfavorable in TOVac200. Sn vacancies polarize the nearest latticeoxygens and induce magnetic moments in them, whereas thepresence of oxygen vacancies reduces the number of polarizedoxygen and reduces the magnetic moment. Apart from Sn4+

vacancies, the smaller size of TO200 compared with TOVac200might be another possible reason of the superior magnetization ofTO200 over TOVac200. The rst effect (magnetization due to Sn4+

vacancies) is described by the supercell approach, while the secondeffect (reduction in size) is understood by considering surface (110)calculation. In the present study, the adsorbed O2 molecule has anegative impact on the magnetization. The numbers of theseadsorbed O2 molecules are expected to be very high on the surfaceof TOVac200, when compared with TO200. We believe that becauseof the presence of high density of oxygen vacancy sites on thesurface of TOVac200, the O2molecule can be easily adsorbed on thesurface oxygen vacancies. However, in TO200 the numbers ofsurface oxygen vacancies are comparatively less; thus, theadsorption sites for O2 are expected to be very few. Unlike TO200,where hole trapped in lattice oxygen atoms, surrounding Snvacancies, become advantageous for high magnetization and theadditional surface adsorbed O2 molecule with trappedelectron (or O2

�) in TOVac200 has no contribution to the totalmagnetization.

of O (a, b and e). Red lines indicate p state while black line indicates s

This journal is © The Royal Society of Chemistry 2014

Paper Journal of Materials Chemistry C

Publ

ishe

d on

18

Sept

embe

r 20

14. D

ownl

oade

d by

Tez

pur

Uni

vers

ity o

n 04

/10/

2014

13:

29:0

8.

View Article Online

5. Conclusions

A detailed study on structural and magnetic properties is per-formed on the SnO2 nanosystems annealed in vacuum and airat 200 �C. XRD and Raman study reveals that TOVac200 haveboth SnO and SnO2 phase in the system. Room temperatureferromagnetism is because of the presence of Sn vacancies. Theformation of Sn vacancies releases holes that are trapped on thenearest lattice oxygens, and on application of a magnetic eldthe spins in 2p orbital of oxygen are polarized with an inducedmagnetic moment. Vacuum annealing results in the removal ofpolarized lattice oxygens surrounding Sn vacancies and thus, areduced magnetic moment is produced. DFT spin polarizedstructural optimization on SnO2 (110) and SnO2 supercellsupports the experimental results and predicts that magnetismin this system might arise because of the presence of Sn4+

vacancies. Oxygen vacancies have no contribution to ferro-magnetism in SnO2. Defect formation energy values shows thatthe formation of Sn vacancies in an O rich system is energeti-cally favorable compared to other point defects. The DFTprediction seems to justify the observed higher saturationmagnetization of TO200, as compared with oxygen decientTOVac200.

Acknowledgements

Pawan Chetri likes to acknowledge DST for providing Inspirefellowship. Biswajit Choudhury and Amarjyoti Choudhuryacknowledge the nancial support provided by Department ofScience and Technology (DST), India, to the project SR/NM/NS-98/2010 (G). We want to thank SAIF IIT Bombay for helping usin carrying out the EPR measurement and IIT Kharagpur forcarrying out SQUID measurement.

References

1 D. D. Awschalom, L. C. Bassett, A. S. Dzurak, E. L. Hu andJ. R. Petta, Science, 2013, 339, 1174.

2 T. Sabergharesou, T. Wang, L. Ju and P. V. Radovanovic, Appl.Phys. Lett., 2013, 103, 012401.

3 J. M. D. Coey, A. P. Douvalis, C. B. Fitzgerald andM. Venkatesan, Appl. Phys. Lett., 2004, 84, 1332.

4 Y. Li, R. Deng, Y. Tian, B. Yao and T. Wu, Appl. Phys. Lett.,2012, 100, 172402.

5 B. Y. Zhang, B. Yao, Y. F. Li, A. M. Liu, Z. Z. Zhang, B. H. Li,G. Z. Xing, T. Wu, X. B. Qin, D. X. Zhao, C. X. Shan andD. Z. Shen, Appl. Phys. Lett., 2011, 99, 182503.

6 S. B. Ogale, R. J. Choudhary, J. P. Buban, S. E. Loand,S. R. Shinde, S. N. Kale, V. N. Kulkarni, J. Higgins,C. Lanci, J. R. Simpson, N. D. Browning, S. D. Sarma,H. D. Drew, R. L. Greene and T. Venkatesan, Phys. Rev.Lett., 2003, 91, 077205.

7 A. Sundaresan, R. Bhargavi, N. Rangarajan, U. Siddesh andC. N. R. Rao, Phys. Rev. B: Condens. Matter Mater. Phys.,2006, 74, 161306.

This journal is © The Royal Society of Chemistry 2014

8 V. B. Kamble, S. V. Bhat and A. M. Umarji, J. Appl. Phys., 2013,113, 244307.

9 N. H. Hong, N. Poirot and J. Sakai, Phys. Rev. B: Condens.Matter Mater. Phys., 2008, 77, 033205.

10 L. Zhang, S. Ge, Y. Zuo, B. Zhang and L. Xi, J. Phys. Chem. C,2010, 114, 7541.

11 K. C. Zhang, Y. Liu, Y. F. Li and Y. Zhu, J. Appl. Phys., 2013,113, 053713.

12 A. Espinosa, N. Sanchez, J. Sanchez-Marcos, A. D. Andresand M. C. Munoz, J. Phys. Chem. C, 2011, 115, 24054.

13 A. Rusydi, et.al., Philos. Trans. R. Soc., A, 2012, 370, 4927.14 D. C. Qi, Phys. Rev. B: Condens. Matter Mater. Phys., 2013, 87,

245201.15 S. Ghosh, M. Mandal and K. Mandal, J. Magn. Magn. Mater.,

2011, 323, 1083.16 P. Chetri and A. Choudhury, Phys. E, 2013, 47, 257.17 K. G. Godinho, A. Walsh and G. W. Watson, J. Phys. Chem. C,

2009, 113, 439.18 R. S. Katiyar, J. Phys. C: Solid State Phys., 1970, 3, 1087.19 J. G. Traylor, H. G. Smith, R. M. Nicklow and

M. K. Wilkinson, Phys. Rev. B: Solid State, 1971, 3, 3457.20 P. Chetri, B. Saikia and A. Choudhury, J. Appl. Phys., 2013,

113, 233514.21 C. H. Shek, G. M. Lin and J. K. L. Lai, Nanostruct. Mater.,

1999, 11, 831.22 J. Geurts, S. Rau, W. Richter and F. J. Schmitte, Thin Solid

Films, 1984, 121, 217.23 V. G. Kravets, Opt. Spectrosc., 2007, 103, 766.24 D. Zwingel, Phys.Status Solidi B, 1976, 77, 171.25 G. Yang, Z. Jiang, H. Shi, T. Xiao and Z. Yan, J. Mater. Chem.,

2010, 20, 5301.26 A. Gurlo, ChemPhysChem, 2006, 7, 2041.27 B. M. Maoz, E. Tirosh, M. B. Sadan and G. Markovich, Phys.

Rev. B: Condens. Matter Mater. Phys., 2011, 83, 161201.28 B. Choudhury and A. Choudhury, J. Appl. Phys., 2013, 114,

203906.29 S. K. Song, J. S. Cho, W. K. Choi, H. J. Jung, D. Choi, J. Y. Lee,

H. K. Baik and S. K. Koh, Sens. Actuators, B, 1998, 46, 42.30 I. S. Mulla, V. J. Rao, H. S. Soni, S. Badrinarayanan and

A. P. B. Sinha, Surf. Coat. Technol., 1987, 31, 77.31 P. K. Biswas, A. De, L. K. Dua and L. Chkoda, Bull. Mater. Sci.,

2006, 29, 323.32 P. S. A. Kumar, P. A. Joy and S. K. Date, Bull. Mater. Sci., 2000,

23, 97.33 H. Ren, G. Xiang, G. Gu, X. Zhang, W. Wang, P. Zhang,

B. Wang and X. Cao, J. Nanomater., 2012, 2012, 295358.34 S. Ghosh, G. G. Khan, B. Das and K. Mandal, J. Appl. Phys.,

2011, 109, 123927.35 G. Rahman, V. M. Garcia-Suarez and S. C. Hong, Phys. Rev. B:

Condens. Matter Mater. Phys., 2008, 78, 184404.36 P. Barquinha, R. Martins, L. Pereira and E. Fortunato,

Transparent Oxide Electronics from Materials to Devices,Willey, West Sussex, 2012.

37 C. W. Zheng, H. Kao and J. M. Dong, Phys. Lett. A, 2009, 373,2592.

J. Mater. Chem. C