Journal of Non-Crystalline Solids 356 (2010) 13951399

Contents lists available at ScienceDirect

Journal of Non-Crystalline Solids

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j n o n c r y s o l

Electrical conduction mechanism in polycrystalline titanium oxide thin lms

Diana Mardare , G.I. Rusu
"Al.I.Cuza" University, Faculty of Physics, 11 Carol I Blvd., R-700506, Iasi, Romania

a r t i c l e

i n f o

a b s t r a c t
Titanium oxide thin lms were deposited onto glass substrates by a d.c. reactive sputtering technique. The temperature dependence of the electrical conductivity was studied using surface type cells, in a wide temperature range (120 K570 K). The experiments showed that, after a heat treatment within the high temperature range 300 K570 K, the temperature dependence of the electrical conductivity became reversible. The structural analysis of the heat-treated lms indicated a polycrystalline anatase or/and rutile structure. The mechanism of the electronic transport in the studied samples was explained by applying models elaborated for lms with polycrystalline (discrete) structure. Some characteristic parameters of these models were calculated: the energy barrier, EB = (0.0460.082) eV and the constant interface-state distribution, NSS = (3.02 10121.23 1013) cm 2 eV 1. 2010 Elsevier B.V. All rights reserved.

Article history: Received 9 July 2009 Received in revised form 12 May 2010 Available online 11 June 2010 Keywords: Titanium oxide; Optical band gap; Polycrystalline lms; Electrical conductivity

1. Introduction Titanium dioxide (TiO2) has been regarded as materials with many attractive properties such as large band gap, high electrical resistivity, high dielectric constant, high oxidative power etc. [1,2]. Consequently, this oxide offers an important number of applications (optical thin lm devices, capacitors in microelectronic devices, high-density dynamic-memory applications, gas sensors, etc.) [36]. There are different advantageous methods used to obtain TiO2 thin lm with different desired properties: reactive sputtering, solgel method, spray pyrolysis, etc. [17]. Reactive sputtering is indicated to obtain uniform and dense lms, with well-controlled stoichiometry [2,6,8 13], desired in electronic applications. Understanding the electronic properties of the transition-metal oxides is a precondition required in electronic device application. Many efforts have been performed in this direction [814], but the picture of the electrical transport mechanisms of these materials is not very clear yet. TiO2 thin lms, deposited by d.c. reactive sputtered, have been chosen to be studied in this paper. Their electrical conduction mechanism is explained in the temperature range 120 K570 K, by applying some models elaborated for lms with polycrystalline (discrete) structure. 2. Experimental The studied titanium oxide lms, were deposited onto heated glass substrates by a reactive d.c. sputtering technique. Argon, used as

Corresponding author. Tel.: + 40 232 201169; fax: + 40 232 201150. E-mail address: dianam@uaic.ro (D. Mardare). 0022-3093/$ see front matter 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.jnoncrysol.2010.05.049

sputtering gas, has the partial pressure of 1.4 10 3 mbar, while the reactive gas (water vapors) has the partial pressure of 0.6 10 3 mbar. The target to substrate distance was about 15 cm. The deposition rate was 0.03 nm/s. Details on other deposition parameters are indicated in Table 1. Thickness measurements, performed with a Surface Proler, revealed very close values for the samples under study, as presented in Table 1. To obtain a clear image on the phase composition of the studied TiO2 thin lms, we have chosen the Grazing Incidence Angle geometry in a computer-controlled diffractometer (Cu K radiation, = 1.54 ). The small angle (here 5) made by the incident beam with the surface sample, determines an increase in the path length of the X-ray beam through the lm, having as an effect the increase in the intensity peaks from the lm and a decrease in the diffracted signal from the substrate. In a 2 conventional geometry, some of the peaks, like the most intense peak of anatase, A(101), could be hidden by the amorphous shoulder of the glass substrate. For electrical resistance measurements, gold rectangular electrodes, placed parallel one to each other at a distance of 0.5 mm, have been deposited onto titanium oxide lms, by thermal evaporation under vacuum. The electrical measurements were performed at high temperatures (300 K570 K) with a Keithley 617 electrometer and at low temperatures (120 K320 K) with a Keithley 196 multimeter. For the high temperature measurements, the samples were heattreated in air. The heat treatment consisted in two successive heatings and coolings (20 K/min) in the temperature range T = (300 K570 K), resting 5 min at the highest temperature. For low temperature measurements, the samples were cooled in a continuous He ow cryostat (Cti-Cryogenics-Helix Technology Corporation).

1396

D. Mardare, G.I. Rusu / Journal of Non-Crystalline Solids 356 (2010) 13951399

Table 1 Thickness of the lms (d), substrate temperature (TS), weight percentage of the anatase phase (WA), average crystallite sizes for the anatase phase (DA) and for the rutile phase (DR), the temperature range in which the energy barrier height has been determined (T1), the energy barrier height (EB) in the low temperature range, relative permittivity (r), impurity concentration calculated from Eq. (6) (NG), correlation coefcient R and standard deviation SD of the straight-line ts. Film SA SR SA.R d (nm) 280 245 260 TS (K) 523 723 543 WA (%) 100 0 46 DA (nm) 13.9 19.6 DR (nm) 20.3 14.5 T1 (K) 120290 133300 150285 EB (eV) 0.046 0.082 0.058 r 31 100 66 NG (1018 cm 3) 3.27 8.82 6.80 R 0.9972 0.9996 0.9992 SD 0.067 0.040 0.047

3. Results The structural analysis of the heat-treated lms indicates a polycrystalline anatase or/and rutile structure (Fig. 1). The studied samples were labeled: SA, SR and SA.R according to its content of anatase and rutile phases, respectively. The weight percentage of the anatase phase, WA, was determined using the equation [15]: WA = 1 = 1 + 1:265IR = IA 1

For wavelengths close to values where loss scattering are dominated by the fundamental absorption of light, the absorption coefcient can be calculated using the expression [2]: =d
1

ln 1 = T

where IR is the intensity of the strongest rutile peak R(110), and IA represents the intensity of the strongest anatase peak A(101). The average crystallite size, D, was obtained, for each phase, from the diffraction peaks A(101), R(110) using the DebyeScherrer formula [16]: D= k B2 cos 2

where d is the thickness of the lm and T is the optical transmittance at wavelength . In the vicinity of fundamental absorption, we may consider that the indirect allowed transition dominates over the optical absorption, according to the relation [19]:   1 hv2 = Ai hvEg0 4

where k is the Scherrer's constant (k = 0.9 [16]), is the X-ray wavelength corresponding to CuK, B2 denotes the full-width at halfmaximum of the peak, and is the Bragg angle. Some structural parameters, obtained after the heat treatment of the investigated samples, are listed in Table 1. It is known that the thermal energy band gap of TiO2 crystals is about 3.65 eV and has been calculated at high temperatures (over 1200 K) [17,18]. At room temperature, values of about 3.0 eV3.2 eV can be obtained (depending on the different phases content) using the optical transmittance spectra [2,9,18]. Fig. 2 illustrates the typical transmittance spectrum for one of the studied titanium oxide lms (SA), the one which contains only the anatase phase. It can be observed that the lm is transparent in the visible region (one of the most important properties of titanium oxide) and its transparency exhibits a sharp decrease in the UV region, at a wavelength which corresponds to the forbidden energy band gap of the studied oxide.

where h is the photon energy, Ai is a characteristic parameter independent of photon energy for respective transitions and Eg0 is the optical band gap. In Fig. 3, a typical (h)1/2 = f(h) dependence, for the same sample SA, is presented. The values of Eg0 for the studied samples, determined by extrapolating the linear part of (h)1/2 = f(h) dependence to (h)1/2 = 0, ranged from 3.15 eV to 3.22 eV. These values are in good agreement with those reported in literature (3.11 eV for the rutile phase [20] and 3.2 eV for the anatase phase [2,9]). Generally, the effects of the crystallite boundaries on the optical absorption are small [19] and this was the reason why we have obtained the energy band gap from optical measurements. Some important parameters of the semiconducting thin lms can be determined by using the temperature dependence of the transport coefcients (electrical conductivity, Seebeck coefcients, etc.). The mechanism of the electrical conduction was explained by studying the temperature dependence of the electrical conductivity. By heat treatment, the structural characteristics of TiO2 lms may modify and, consequently, the electronic transport properties of the respective lms (particularly, electrical conductivities) may vary too. To obtain lms with stable structure, all samples were subjected to a heat treatment consisting of two successive heating/cooling cycles

Fig. 1. XRD patterns of the investigated samples (CuK radiation, = 1.54178 ).

Fig. 2. Transmittance spectrum of the anatase TiO2 sample, SA.

D. Mardare, G.I. Rusu / Journal of Non-Crystalline Solids 356 (2010) 13951399

1397

of a great concentration of trapping states at the crystallite-boundary, able to capture free charge carriers) [2628]. The rst models do not explain the temperature dependence of the electrical conductivity corresponding to the TiO2 polycrystalline lms. Seto [19,26] has elaborated a theory of electronic transport properties in polycrystalline lms, taking into consideration both the crystalliteboundary trapping model and thermionic-emission mechanism through crystallite-boundary potential barriers. The following assumptions were made in Seto's model: the crystallites have similar size and shape; only one type of monovalent trapping states (having a concentration Nt and an energy Et with

Fig. 3. Optical absorption spectrum of the anatase TiO2 sample, SA. Eg0 is obtained by extrapolating the linear part of (h)1/2 = f(h) dependence to (h)1/2 = 0 (the correlation coefcient of the linear t is 0.999).

within certain temperature ranges. The experiments have shown that, as a result of this treatment, the temperature dependence of the electrical conductivity became reversible. We supposed that this fact indicates the stabilization of the lm structure and composition in respective temperature range [1013,21]. Fig. 4 shows the dependence ln = f(103/T) plotted for each studied sample during the heat treatment. A decrease in the conductivity is observed, after rst heating, when the studied lms rest 5 min at the highest temperature (dot curved arrows in Fig. 4a,b,c). The decrease may be associated with an oxidation process. For all heat-treated samples, the temperature dependence of the electrical conductivity can be described by the exponential law [11,19,21]:   E = 0 exp a kB T 5

where Ea is the thermal activation energy of electrical conduction, 0 is a parameter depending on the sample nature and kB is Boltzmann's constant. Consequently, for a large number of heat-treated samples, the shapes of ln = f(103/T) curves were examined in the investigated temperature range. It has been found that, generally, the ln = f(103/T) dependences present almost three linear portions. In the lower temperature range (120 K b T b 290 K), the mentioned curves are characterized by a smaller slope (the activation energy Ea changes for the investigated samples, from 0.046 eV to 0.082 eV). The next portions have larger slopes: 0.1070.134 eV, for 300 K b T b 390 K and, nally, values of 0.0470.073 eV, were obtained for the domain 395 K b T b 570 K (Figs.4 and 5). 4. Discussion We consider that different electronic transport mechanisms are predominant in each of the mentioned temperature ranges. Evidently, in polycrystalline semiconducting lms, the conduction mechanism is strongly inuenced by the inherent intercrystalline boundaries [19,22,23]. Therefore, electronic transport can be explained by applying the models elaborated for lms with discrete polycrystalline structure: Volger [24], Petritz [25], Seto [26], etc. Usually, these models can be divided into two groups: segregation models (which supposed that impurity atoms tend to diffuse at the crystallite boundaries, where they become electrically inactive) and the crystallite-boundary trapping models (which consider the presence

Fig. 4. Ln versus 103/T plotted for: a) sample SA, b) sample SR and c) sample SA.R, during the heat treatment. The dot arrows show the evolution during the heat treatment, while the solid vertical arrows mark the chosen temperature domains.

1398

D. Mardare, G.I. Rusu / Journal of Non-Crystalline Solids 356 (2010) 13951399

with
2 e2 NSS Eg 4NG

!1
2

1+

11

and s kB T vc = 2m4

12

Fig. 5. Dependences ln = f(103/T) plotted in the range 120 K320 K for all the studied samples. The solid vertical arrows mark the domain T1 for each sample. The solid lines represent the tting straight lines for each sample, for T1 domain.

respect to the Fermi level at the interface) is present and is uniformly distributed within crystallites; the traps are initially neutral and become charged by trapping a free carrier; the thickness of crystallite boundaries is negligible with respect to crystallite size. Some modications of the Seto's model have been proposed by Baccarani et al. [28]. They apply Seto's model for the intermediate range of impurity concentrations and consider two trapping states: monovalent trapping states, and continuous distribution of trapping states within the energy band gap. To explain the temperature dependence of the electrical conductivity of the investigated polycrystalline TiO2 lms, the last case (continuous distribution of trapping states) was considered: it was supposed that the acceptor interface states are continuous uniformly distributed in the upper half of the band gap. If the crystallites are completely depleted, for an impurity concentration, NG, smaller than a certain value, N* (NG b N* ), G G according to the theory of thermionic-emission, the energy barrier height is [28]: EB = e2 L2 NG 8 6

The electrical conductivity of the lms is given by: F1 =   e2 Lvc Nc E exp a1 kB T kB T 7

where thermal activation energy is expressed by: Ea1 = Eg = 2LNG = NSS 8

In Eqs. (6)(12) the following notations have been used: e, electron charge; L, mean crystallites size; , low frequency permittivity of crystallites ( = 0r, where 0 is the permittivity of free space and r is the static relative permittivity of crystallite); Nc, effective state density for conduction band; kB, Boltzmann's constant; m, scalar effective mass of charge carriers; Eg, energy of forbidden band (Eg is considered to be equal to optical energy band gap Eg0); NSS, constant interface-state distribution. The values for the static relative permittivity, r, (see Table 1) were obtained using a method described in ref. [29]. We supposed that, in the lower temperature range T1 (Table 1), the activation energy of electrical conduction indicates the value of the potential barrier height, EB. Consequently, by plotting ln versus 103/T in this temperature range (Fig. 5), for each sample, and tting the dependences with straight lines (correlation coefcients given in Table 1), the values of EB were determined (Table 1). Then, from Eq. (6), the impurity concentrations, NG, were calculated, their values being introduced in Table 1. In the temperature ranges T2 (Table 2), the temperature dependence of the electrical conductivity is described by Eq. (7). From the slope of the linear ts of ln versus 103/T (for the second cooling), the thermal activation energy Ea1 was obtained. Then, knowing Ea1, the values of NSS were derived from Eq. (8). The values of Ea1 and NSS are given in Table 2. Going up to higher temperature domains T3 (see Table 3) we may consider that the crystallites become partially depleted, trapping the charge carriers. In fact, this assumption agrees with the crystalliteboundary model [26,28], which suggests (for higher temperatures) that a decrease of the carrier concentration with increasing temperature takes place, due to the crystallite contribution to the electrical conductivity. Two methods were used to obtain the thermal activation energy Ea2: an experimental method, from the slope of the linear ts of the dependences ln as a function of 103/T, in the temperature ranges T3 (for the second cooling) (Table 3), and a theoretical method, using Eqs. (6)(12). In the last case, the impurity concentration, NG, was calculated from relation (6). Then, Eq. (8) was used for the evaluation of the constant interface-state distribution, NSS. In this relation, Ea1 was experimentally calculated, using the dependences ln as a function of 103/T in the temperature range T2 (for the second cooling), indicated in Table 2 for each sample, and Eg was determined from the absorption spectra. The parameter was calculated from Eq. (11) and Ea2 was determined using Eq. (10). From Table 3, one can
Table 2 The mean crystallite size (L), optical energy band gap (Eg0), the temperature range in which Ea1 was determined (T2), thermal activation energy of the electrical conduction (Ea1), constant interface-state distribution (NSS), correlation coefcient R and standard deviation SD of the straight-line ts. Film SA SR SA.R L (nm) 13.9 20.3 16.8 Eg0 (eV) 3.22 3.15 3.18 T2 (K) 300365 315385 305390 Ea1 (eV) 0.107 0.120 0.134 NSS (cm 2 eV 1) 3.02 1012 1.23 1013 7.98 1012 R 0.9990 0.9999 0.9992 SD 0.008 0.003 0.013

* If the crystallites are partially depleted, NG N NG (the depletion width is smaller than L/2), the electrical conductivity of the lm is given by:

F2 = where Ea2

    2 e Lvc Nc NC E exp a2 kB T NG kB T

  2 2 = Eg = 2 4NG = e NSS 1

10

D. Mardare, G.I. Rusu / Journal of Non-Crystalline Solids 356 (2010) 13951399 Table 3 Temperature range in which the activation energy Ea2exp has been experimentally determined (T3), parameter () obtained from Eq. (11), thermal activation energy: theoretical, obtained from Eq. (10), (Ea2) and experimental (Ea2exp), correlation coefcient R and standard deviation SD of the straight-line ts. Film SA SR SA.R T3 (K) 400570 395525 410540 1.063 1.115 1.097 Ea2 (eV) 0.063 0.091 0.078 Ea2exp (eV) 0.047 0.073 0.066 R 0.9879 0.9974 0.9975 SD 0.015 0.018 0.010

1399

References
[1] X. Zhang, M. Jin, Z. Liu, D.A. Tryk, S. Nishimoto, T. Murakami, A. Fujishima, J. Phys. Chem. C 111 (39) (2007) 14521. [2] D. Mardare, M. Tasca, M. Delibas, G.I. Rusu, Appl. Surf. Sci. 156 (2000) 200. [3] M. Epifani, A. Helwig, J. Arbiol, R. Diaz, L. Francioso, P. Siciliano, G. Mueller, J.R. Morante, Sens. Actuators B 130 (2) (2008) 599. [4] Y. Sawada, Y. Taga, Thin Solid Films 116 (1984) L55. [5] A.M. More, J.L. Gunjakar, C.D. Lokhande, Sens. Actuators B: Chem. 129 (2) (2008) 671. [6] D. Mardare, N. Iftimie, D. Luca, J. Non-Cryst. Solids 354 (2008) 4396. [7] M. Crisan, A. Braileanu, M. Raileanu, D. Crisan, V.S. Teodorescu, R. Birjega, V.E. Marinescu, J. Madarsz, G. Pokol, J. Therm. Anal. Cal. 88 (2007) 171. [8] A. Yildiz, S.B. Lisesivdin, M. Kasap, D. Mardare, J. Non-Cryst. Solids 354 (2008) 4944. [9] H. Tang, K. Prasad, R. Sanjins, P.E. Schmid, F. Lvy, J. Appl. Phys. 75 (1994) 2042. [10] D. Mardare, G.I. Rusu, Phys. Low-Dim. Struct. 11/12 (1999) 69. [11] D. Mardare, G.I. Rusu, Mater. Sci. Eng. B 75 (1) (2000) 68. [12] D. Mardare, C. Baban, R. Gavrila, M. Modreanu, G.I. Rusu, Surf. Sci. 507510 (2002) 468. [13] D. Mardare, N. Cornei, G.I. Rusu, Superlattices Microstruct. 46 (2009) 209. [14] B.V. Kumar, T. Sankarappa, M.P. Kumar, S. Kumar, J. Non-Cryst. Solids 355 (229) (2009) 229. [15] R.A. Spurr, H. Myers, Anal. Chem. 29 (1957) 760. [16] B.D. Cullity, Elements of X-Ray Diffraction2nd Edition, Addison-Wesley, Reading, MA, 1978. [17] P.Baranski, V. Klotchkov, I. Potykevich, Electronique des Semiconducteurs, Mir, Moscou, 1978. [18] N. Tsuda, K. Nasu, A. Fujimori, K. Siratori, Electronic Conduction in Oxides, Springer, Berlin, 2000. [19] L.L. Kazmerski, Polycrystalline and Amorphous Thin Films and Devices, Academic Press, New York, 1980. [20] N. Daude, N.C. Gout, C. Jouanin, Phys. Rev. B 15 (1977) 3229. [21] G.I. Rusu, A. Airinei, M. Rusu, P. Prepelita, L. Marin, V. Cozan, I.I. Rusu, Acta Materialia 55 (2007) 433. [22] A.F. Mayadas, M. Shatzkes, Phys. Rev. B 1 (1970) 1382. [23] A.L. Fripp, J. Appl. Phys. 46 (1975) 1240. [24] J. Volger, Phys. Rev. 9 (1950) 1203. [25] R.L. Petritz, Phys. Rev. 104 (1956) 1508. [26] J.Z.W. Seto, J. Appl. Phys. 46 (1975) 5247. [27] C.H. Seager, T.G. Gasner, J. Appl. Phys. 49 (1978) 3879. [28] C. Baccarani, B. Ricco, G. Spadini, J. Appl. Phys. 49 (1978) 5565. [29] D. Mardare, G.I. Rusu, J. Optoelectron. Adv. Mater. 6 (1) (2004) 333.

see that there is a good agreement between the experimental results and the theoretical ones. 5. Conclusions The mechanism of electrical conduction in polycrystalline-d.c. sputtered TiO2 thin lms was explained in the temperature domain 120 K570 K, by applying a model proposed by Baccarani et al., based on the crystallite-boundary scattering theory of Seto. Impurity concentration (NG) and the constant interface-state distribution (NSS) were calculated using this model. We have observed that ln as a function of 103/T presents almost three linear portions in the temperature domain 120 K570 K. Experimental results (e.g. activation energies on different range temperatures) t well with the theoretical ones, indicating that the proposed model can satisfactorily explain the experimental data. Acknowledgements We would like to thank Prof. F. Lvy from the Institute of Applied Physics, Polytechnic Federal School of Lausanne, Switzerland for providing the necessary laboratory facilities to carry out a part of this investigation.