Vous êtes sur la page 1sur 5

Sensors and Actuators B 126 (2007) 204208

Ab initio DFT computation of SnO2 and WO3 slabs and gassurface interactions
Michel Levy, Thierry Pagnier
Laboratoire dElectrochimie et de Physicochimie des Materiaux et des Interfaces, ENSEEG 1130 rue de la Piscine, BP75 F-38402 Saint Martin dH` eresm, France Available online 3 January 2007

Abstract Slabs of SnO2 and WO3 were computed by ab initio DFT technique, in order to model nanoribbons. Atomic positions, electron density and electronic density of states were calculated for perfect, vacancy free materials. In these conditions, oxygen is shown to adsorb as a neutral species, but this adsorption creates acceptor states in the gap which can trap the electrons due to O vacancies in real materials. 2006 Elsevier B.V. All rights reserved.
Keywords: Ab initio; DFT; SnO2 ; WO3 ; O adsorption; DOS

1. Introduction SnO2 and WO3 are both n-type semiconductors with a wide bandgap. The n-type conduction is due to the presence of oxygen vacancies, which create electrons in the conduction band through:
1 O2 + VO + 2e OO 2

oxide surfaces with gases and their inuence on the electrical properties. In this paper, we present the rst results obtained on the computation of oxide slabs and on the interactions of the calculated surfaces with adsorbed oxygen atoms. 2. Computation procedure The code used is ABINIT [2], which is distributed under the GNU General Public Licence. The software allows calculations of the ground state of periodic systems using the density functional theory (DFT). Our computations were carried out within the local density approximation (LDA) with the HartwigsenGoedeckerHutter pseudopotentials found on the Abinit WEB site [3]. These pseudopotentials include four valence electrons for Sn, and six for O. Most of the calculations were carried out on Apple workstations (PowerPC processor working at 2.5 or 2.7 GHz, 4 GB of memory). The base structure of WO3 was calculated at IDRIS on the ZAHIR machine. Crystal structures of rutile SnO2 and monoclinic WO3 were computed in order to check the consistency of the parameters used. Table 1 gives a comparison between the results obtained and the experimental data. As customary, calculations within the LDA approximation give a forbidden gap narrower than experimentally measured.

(1)

where the Kr oger notation is used. In SnO2 , these donor levels are found very close to the top of the conduction band, at 0.15 and 0.03 eV from its bottom [1]. It is generally accepted that oxygen atoms adsorb as O , thus creating an electron-depleted area at the surface of oxide grains. As a consequence, the electrical resistance of the oxide increases. When oxidizing (e.g. NO2 ) or reducing (e.g. CO) molecules co-adsorb, the depleted area is modulated. This is the reason for the very high gas sensing properties of SnO2 , and to a lesser extent of WO3 . Despite the wide acceptance of this model, there are very few experimental evidences. Ab initio computations, which are becoming powerful enough for the study of large atom assemblies, could be a way to better understand the interactions of

Corresponding author. E-mail address: Thierry.Pagnier@lepmi.inpg.fr (T. Pagnier).

0925-4005/$ see front matter 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.snb.2006.11.047

M. Levy, T. Pagnier / Sensors and Actuators B 126 (2007) 204208 Table 1 Experimental and calculated data for SnO2 and WO3 crystals Parameter WO3 a b c Gap SnO2 a c x Gap Experimental 7.306 7.540 7.692 90.88 2.62 4.737 3.188 0.307 3.60 Calculated

205

however, there is no symmetry except identity and the C1 space group was used. 3. Slab calculation and oxygen adsorption
7.459 7.455 7.778 90.895 2.045 4.712 3.174 0.306 3.167

Experimental data are from ref. [8] for WO3 and ref. [9] for SnO2 .

Slabs were calculated with surfaces parallel to the (0 0 1) plane for WO3 and (1 1 0) plane for SnO2 . In both cases, the lattice parameters of the planes parallel to the surfaces were kept constant and equal to their values in the crystal. The z-axis was chosen to be perpendicular to the ab plane in WO3 , and parallel to the (1 1 0) direction in SnO2 . Because of the slight monoclinic distortion of the WO3 cell, the z-axis is not strictly perpendicular to the surfaces of the slab, but this has no inuence on the comparisons made between free surfaces and surfaces having adsorbed atoms. The origin of the z-axis was taken in the center of the slab and the atoms of the corresponding plane were xed to their position in the crystal during the relaxation of the structure. Relaxed structures were considered to be obtained when 1 . Once forces on free atoms were lower than 5 102 eV A equilibrium positions were obtained, the electron density was calculated. In SnO2 slab, a symmetry plane remains (z = 0) and calculations were carried out in the P11m space group. For WO3

Fig. 1 shows the structure of the slabs. In both cases, in order to keep the slab electroneutrality, one surface oxygen over two have been removed. The remaining oxygen atoms form lines at the surface, the difference being that these oxygens are connected to one tungsten atom or two tin atoms. In all cases, ve-fold exposed cations have a tendency to shift towards the interior of the slab. Adsorption of one oxygen atom per unit cell is allowed on each SnO2 surface. In order to keep the plane symmetry of the slab, oxygen atoms were simultaneously added on each side of the slab. Two positions were tested: the adsorbed oxygen on top of the ve-coordinated Sn atom (called apical from now), or bridging between two Sn atoms (Fig. 2). For WO3 , there are two positions per side of the slab. One or two oxygen atoms were added in apical position. A there was no symmetry to keep, oxygen atoms were added only on one side of the slab. One of the standards outputs of ABINIT is the electron density along the z-axis direction z (z), obtained by integrating the local electron density (x, y, z): z (z) = (x, y, z) dx dy (2)

Fig. 3 shows z (z) for the SnO2 slab without and with an apical oxygen added. The difference between both curves is shown in Fig. 4. A positive difference indicates an excess of electrons when the O atom is absorbed, a negative value a lack of electrons (depleted area). The number of electrons involved can be obtained by a numerical integration of the difference in each

Fig. 1. View of the SnO2 (left) and WO3 (right) slabs with three metal planes. (Large spheres) O atoms. (Small spheres) Metal atoms. (Bottom) View of each slab surface (rotated for WO3 ) showing the alignment of surface O atoms and metal-exposed atoms.

206

M. Levy, T. Pagnier / Sensors and Actuators B 126 (2007) 204208

Fig. 2. Slab surfaces showing adsorbed apical O on WO3 (a), apical O on SnO2 (b) and bridging O on SnO2 (c). Adsorbed O atoms are darker.

(1). Donor levels due to oxygen vacancies were observed very close to the bottom of the conduction band [4,5]. As acceptor levels due to the adsorbed oxygens are localized at the surface of SnO2 , only the vacancies near this surface are involved. The order of magnitude of the depth at which a vacancy can interact with adsorbed oxygen is given by the Debye length. As adsorbed oxygen compensates oxygen vacancies, it is interesting to determine which of these defects is predominant. Temperature programmed desorption (TPD) experiments [6] have shown that O coverage is almost unity. For the slabs studied here, it means that there is one adsorbed oxygen per unit cell, or
Fig. 3. Electron density along the z-axis (perpendicular to the slab surface) for the SnO2 slab (thin line) and after adsorption of one apical oxygen (thick line). The origin of the z-axis is at the center of the slab. Maxima correspond to the planes occupied by nuclei.

zone. When an O atom is absorbed, a depleted zone is observed from the slab center, and involves 0.34 electron. This about 4 A local charge is compensated by two excess zones from z = 2.2 and for z > 4.4 A, each one carrying a negative charge to 3.4 A of 0.17 electron. At the position of the adsorbed oxygen (the the electron excess is solely due nucleus is located at z = 5.4 A), to the six valence electrons brought by the O atom. A similar observation can be made for WO3 slab. These results strongly suggest that there is no signicant electron transfer from the bulk of the slab to the adsorbed oxygen atom: O adsorbs as a neutral species. 4. Electron density of states Fig. 5 shows a schematic view of the electron density of states (DOS), calculated by the tetrahedron method with 8k points in the Brilloin zone for the SnO2 slab, apical O adsorbed and bridging O adsorbed. The forbidden gap decreases from 3.17 eV (bulk SnO2 ) to 1.8 eV (slab). When O is adsorbed, half lled states appear in the gap. For bridging oxygen, these states appear as one lled band and one empty band. For apical oxygen, there is only one band half lled. In both cases, the empty states can accommodate two electrons per adsorbed oxygen, thus suggesting that oxygen charge can go up to 2. For apical oxygen, the defect band is located 0.54 eV from the top of the valence band. These states cannot be lled with electrons originating from the slab, except through thermal population for T > 0. But they can trap the electrons coming from the oxygen vacancies through Eq.

Fig. 4. (Left) Difference between electron density along the z-axis for O adsorbed SnO2 slab and without O adsorbed. (Right) Integral of the electron difference showing that there are as many electrons from the slab center to z = 0.46 nm in both cases, suggesting no transfer from the slab to the SnO2 surface.

M. Levy, T. Pagnier / Sensors and Actuators B 126 (2007) 204208

207

Fig. 5. Schematic electron density of states (DOS) for SnO2 slab (A), with bridging adsorbed oxygen (B) and with apical adsorbed oxygen (C). Stars denote the extra bands coming from adsorbed oxygen. In both cases, the levels added in the gap are: (i) located below those of donor states coming from Eq. (1) and (ii) half lled.

4.7 1014 O cm2 . The number of oxygen vacancies, estimated either by TPD [6] or by electrical measurements [7], falls within 1016 to 1017 cm3 . For a Debye length of 10 nm, there is only 2 104 oxygen vacancy per surface unit cell. This result indicates that only a small fraction of adsorbed oxygen is negatively charged (roughly 1 over 5000), and that, in the region close to the surface, all free carriers are trapped by adsorbed oxygen. 5. Effect of foreign gases and conduction model The model that comes from the present calculations is based on: (i) the presence of a large number of unoccupied acceptor states in the gap and (ii) a region close to the slab surface virtually free of charge carriers due to oxygen vacancies. Within this model, it is hardly understandable why gases such as NO2 or CO can strongly affect the electrical properties, especially at the part per million level. Actually, air remains the major constituent of the gas phase and the oxygen coverage of the surface will remain high. We can therefore estimate that numerous acceptor levels remain. Addition of new acceptor molecules (e.g. NO2 ) will not change this situation. Nor will the presence of donor molecules (e.g. CO), because adsorbed neutral oxygen will trap these new electrons. We therefore suggest a defect conduction in a region close to the surface of SnO2 . For example, electrons could jump from an adsorbed O to a neutral adsorbed O, with a mechanism similar to polaronic hopping. In this case, the number of charge carriers could be the number of adsorbed O species, and the effect of donor (resp. acceptor) molecules will be to increase (resp. decrease) this number. As the fraction of charged oxygen is about 104 , even gases at the ppm ratio will have a signicant inuence on this number.

6. Conclusion By using a DFT ab initio calculation of SnO2 and WO3 slabs, and by adding adsorbed oxygen atoms to the slab, we have shown that O adsorbs as neutral on perfect (vacancy free) oxide. The decrease of the electrical conductivity observed experimentally when O is adsorbed to the oxide surface is more likely due to the presence of acceptor states in the bottom of the gap when O adsorbs. We have shown that only a small fraction of the adsorbed oxygen atoms are charged and that a region close to the surface is virtually free of charge carriers, thus becoming electrically insulating. We suggest that the conduction is due to surface (or near-surface) defect transport. We propose that the number of charge carriers could be assimilated to the number of charged oxygen adsorbed atoms. In this case, the effect of donor or acceptor co-adsorbed molecules would be to modify the number of charge carriers. Acknowledgements This work was carried out in the framework of the Nanostructured Solid-State Gas Sensors with Superior Performance (NANOS4) Project (No. 001528), funded by the European Community through the Sixth Framework Program. Part of the calculations were performed at IDRIS within the NANOVIB Project. References
[1] J. Maier, W. G opel, Investigations of the bulk defect chemistry of polycrystalline tin(IV) oxide, J. Solid State Chem. 72 (1988) 293302. [2] X. Gonze, J.-M. Beuken, R. Caracas, F. Detraux, M. Fuchs, G.M. Rignanese, L. Sindic, M. Verstraete, G. Zerah, F. Jollet, M. Torrent, A. Roy,

208

M. Levy, T. Pagnier / Sensors and Actuators B 126 (2007) 204208 [6] J. Mizusaki, H. Koinuma, J.I. Shimoyama, M. Kawasaki, K. Fueki, High temperature gravimetric study on nonstoichiometry and oxygen adsorption of SnO2 , J. Solid State Chem. 88 (1990) 443450. [7] H. Ogawa, M. Nishikawa, A. Abe, Hall measurement studies and an electrical conduction model of tin oxide ultrane particle lms, J. Appl. Phys. 53 (1982) 44484455. [8] B.O. Loopstra, H.M. Rietveld, Further renement of the structure of WO3 , Acta Crystallogr. B 25 (1969) 14201421. [9] A. Svane, E. Antoncik, Electronic structure of rutile SnO2 , GeO2 and TeO2 , J. Phys. Chem. Solids 48 (1987) 171180.

M. Mikami, P. Ghosez, J.Y. Raty, D.C. Allan, First-principles computation of material properties: the ABINIT software project, Comp. Mater. Sci. 25 (2002) 478492 (Abinit is a common project of the Universit e Catholique de Louvain, Corning Incorporated, and other contributors). [3] www.abinit.org. [4] S. Samson, C.G. Fonstad, Defect structure and electronic donor levels in stannic oxide crystals, J. Appl. Phys. 44 (1973) 46184621. [5] H. Teterycz, R. Klimkiewicz, M. Laniecki, Study on physico-chemical properties of tin dioxide based gas sensitive materials used in condensation reactions of n-butanol, Appl. Catal. A 274 (2004) 4960.

Vous aimerez peut-être aussi