Electrical surface versus bulk properties of Fe-doped TiO2 single crystals

SOLID STATE

Solid State lonics 72 (1994) 12-18 North-Holland

IOHICS Electrical surface versus bulk properties of Fe-doped TiO2 single crystals A. Bernasik Ecole des Mines de Nancy. F-54042 Nancy, France

M. Rekas Laboratory of Electronic Materials, TU, Clausthal D-386 78, Germany

M. Sloma ] and W. Weppner 2,3 Max Planck Institute for Solid State Research, D- 70569 Stuttgart, Germany

Thermopower measurements,impedance spectroscopy and work-functionmeasurementsof Fe-doped TiO2singlecrystalswere carried out under conditions ( T, Po~ ) which correspond to the near-stoichiometriccomposition domain. Fe 3+ ions form acceptor centers: Fe~riwhich leads to a significant shift of the bulk n-p-type transition to lower oxygenpartial pressures as compared with undoped materials. It was shown that Fe-doped TiO2 is a mixed conductor within the n-p-type transition domain. The surface of iron-doped TiO2 exhibits a shift of the transition region to lower P02 compared with undoped samples. In the transition region the surfaceproperties are fully controlledby the acceptor type dopant. The increaseof the work-functionis enhanced by the strong iron segregationin oxidized condition, which was confirmed by SIMS depth profile analysis.

1. Introduction It is c o m m o n l y assumed that non-stoichiometric, u n d o p e d rutile (TiO2) is an oxygen-deficient (metal excess) n-type semiconductor. The variety of structural defects reported for TiO2 includes point defects, such as t i t a n i u m interstitials or oxygen vacancies, and planar defects, such as crystallographic shear planes. They affect the local cation valence of the oxygen-ligand configuration a n d cause the variation of d-orbital occupation which produces significant changes in the electronic properties. Though the electrical and transport properties of TiO2 were intensively studied [ 1-6 ] a consistent model of the defect structure in rutile for small deviations from stoit Present address: Academyof Mining and Metallurgy,Department of Metallurgy, A1.Mickiewicza 30, 30-059 Krak6w, Poland. 2 Present address: Christian Albrechts University Kiel, Chair for Sensorsand Solid State Ionics, Kaiserstr. 2, D-24098 Kiel, Germany. 3 To whom all correspondence should be addressed.

chiometry, e.g. in the vicinity of the n - p transition is still not available. The aim of the present work is to study the effect of incorporation of Fe 3+ ions on both surface a n d bulk electronic properties of TiO2 single crystals within a near-stoichiometric composition domain.

2. Experimental 2.1. Materials

Single crystals of u n d o p e d and 0.5 at% Fe-doped rutile (TiO2) prepared by the Verneuille method were obtained from H r a n d Djevahirdian S.A., Monthey, Switzerland. The specimens for thermopower measurements were cut into rectangular bars with d i m e n s i o n s of 3 . 5 × 3 . 5 × 10 m m 3 and 4 . 5 × 4 . 5 × 10 m m 3 for u n d o p e d a n d Fe-doped material, respectively. Both crystals were oriented along the [ 100 ] crystallographic axis, Samples for work-function ( W F ) measurement

0167-2738/94/$ 07.00 © 1994 ElsevierScienceB.V. All rights reserved.

A. Bernasik et al. / Fe-doped Ti02 single crystals

were in the shape of pellets of 10 m m diameter and 1 m m in thickness. The surfaces under investigation were oriented to provide (001 ) faces. All samples were polished with diamond paste down to 1 ~tm and cleaned in an ultrasonic bath with acetone and water. The purity of the undoped TiO2 single crystal was rather low with about 200 ppm of trivalent impurities (Fe, Al) [7]. 2.2. Methods

The thermopower of the thermocell (T1, P) (Pt)O2 (g)ITiOE(Fe) IO2(g) (Pt) (T2, P) (1) was measured within the temperature range 10301200 K and the partial oxygen pressure range 2 P a l00 kPa. The Seebeck effect was measured under a steady-state condition [8]. Microheaters were applied to impose appropriate temperature gradients in the range of max. _+ 15 K to determine the EMF. Platinum leads and PtRh 10-Pt thermocouples were used. Absolute values of the thermopower were determined based on the thermopower data for platinum Q~ [ 9 ]. Impedance spectroscopy (HP 4192 A LF impedance analyzer) was applied to determine the total conductivity of the doped sample at 1073 K. Dc four-point conductivity and thermopower were studied for undoped futile samples within the temperature range from 1000 to 1200 K and for the oxygen partial pressure range 50 Pa-100 kPa. WF measurements were carried out with a hightemperature Kelvin probe [ 10] as a function of temperature (1050-1150 K) and oxygen partial pressure (50 Pa-100 kPa). Argon-oxygen mixtures as well as a zirconia-based oxygen pump-gauge system were used to establish the required Po2. As a reference electrode a platinum sheet was applied [11]. Contact potential difference ( C P D ) changes were transformed to the WF changes of the measured sample based on the relation:

[ 11 ]. This assumption was carefully proven. The temperature dependence of the WF changes of Fedoped TiO2 is shown in fig. 1. Above 980 K reversible WF changes were obtained during heating-cooling cycles. Below 980 K hysteresis of adsorption effects are clearly visible. Consequently, above 980 K the parameter m, the reciprocal of the oxygen power exponent of the electron concentration at the surface, may be calculated. Assuming a non-degenerate distribution of electronic charge carriers at the surface, m may be expressed by the following formula: 1

1

1

1

OWF

m

ms

mb

kT

OIn Po2

--=--+--=--'

(2)

It was assumed that at high temperatures the surface is in equilibrium with the gas phase and changes of the WF component due to the surface adsorption are negligible compared with the total WF changes

(3)

where rns and mb correspond to the surface bandbending and bulk Fermi level shift contributions, respectively. A Cameca SMI 300 NBP secondary ion mass spectrometer (SIMS) was applied for depth profile analysis. Sputtering of electrically insulating samples was performed by a neutral beam of argon atoms [ 12 ] with an energy of 7 keV. The intensity signal was taken from the surface area with the diameter of 200 pm. The pressure of the residual gas in the measuring chamber was below 5 X 10 -2 Pa. The sputtering rate a n d the depth resolution of TiO2 were calibrated against the standard sample of 30 nm Ta205 on metallic Ta foil. The depth resolution for the standard specimen was equal to 1.6 nm and the ratio of TiO2 to Ta205 sputtering rate was equal to 0.59.

5o i TiO2-F~e0.5%,~mono(001) / o! ~i'~, ~ temperaturedependence / -5o1

..............

t ~" -1004

",, ',

, -2t

~

inArg~

~oozing:~aat,n~ i

\

~

cooling-heating rate 10 K / h o u r

-350----

600

n Oxygen

•

/ c°°"ng m i

-

AWFsample = CPD + A W F p t .

13

700

800

900

] "~%...,,,........ ::::~.j. . . . . . . . . 1000

1100

1200

Temperature [K] Fig. l. Temperature dependence of the WF changes of Fe-doped TiO2 single crystals. Heating-cooling cycle were done in argon and oxygen.

14

A. Bernasik / Fe-doped Ti02 single crystals

3. Defect equilibria

4. Thermopower and conductivity in the vicinity of the n-p transition

The oxygen interaction with pure Ti02 may be expressed as: x

TiO2~TiOE_x + ~

(4)

0 2 ,

where x denotes the deviation from ideal stoichiomerry. At low oxygen activity, in the n-type conductivity range, titanium interstitials or oxygen vacancies are considered as the predominant point defects, which leads to the oxygen power dependence of the electronic charge carriers concentration listed in table 1. Substitutional incorporation of aliovalent dopants may strongly affect both the ionic and electronic defect concentrations by the following incorporation reactions:

At low deviation from ideal stoichiometry both electrons and electron holes contribute to the thermopower, and - if one deals with a mixed conductor - the ionic contribution has to be taken into account. The following relation expresses the thermopower of such a mixed conductor:

Q=liQi +t.Q. +tpQp,

(9)

where ti, t. and lp are the transference numbers for ions, electrons and electron holes, respectively. Q. and Qp a r e the Seebeck coefficients for electrons and holes of a non-degenerate semiconductor as expressed by the formulas [ 13 ]: _k(1n N¢ +An), Q"=-ek [e']

Fe203 + 2TiTi + 4 0 o ~2FeTi + V o + 300 + 2TIO2

(5)

Fe203 + ½02 + 2Til-i + 4 0 0 2Fern + 2h" + 4 0 0 + 2TIO2.

(6)

Moreover within a quasi-stoichiometric composition domain, in the vicinity of the n-p-type conductivity transition, the following intrinsic defect equilibria have to be considered: nil~e'+h"

(7)

nii-~2Vo + V % .

(8)

The resulting m parameters in the case of doped materials are also listed in table 1.

Qp= ek ( ln [Nv ~ +A° ) '

where N¢ and Nv are the densities of states in the conduction and valence bands, An and Ap are the kinetic coefficients for electrons and holes, which depend on the transport mechanism. Qi represents the thermopower due to ionic conductivity and is related to the transported entropy of mobile ions in the solid electrolyte. From eqs. ( 8 ) - (10) one may see that in the case of a pure electronic conductor the Seebeck coefficient is related to the Fermi level position Ev within the band-gap: 1

Q= Tee (Ev-Ev) Table 1 Oxygen power exponent of electronic charge carrier concentration depending on predominant type of defect [e'] ~ P 5 ¢ / " .

(10)

t, -~eEg.

(l l)

The total conductivity is simply expressed as a sum of partial ionic (ai), electronic (an), and electron holes (ap) contribution:

a=ai Wa, Wap.

(12)

Major ionic defect

Electroneutrality condition

m parameter

Vo

[Vo ] = 1/2 [e'l

6

Vo

[Vo] = [e']

4

Ti;"

[TIT' ] = ~ [e']

4

Tit"

[TIT"] = ~[e'l

5

In the near-stoichiometric composition domain ai is assumed to be independent of Po2 and constant, and the electron and electron hole contributions are expressed as:

Vo, Fe~ri Vo, Fe~ri Fern Vo, Fe~-i

2 [ V o ] = [Fe~ri ] [Vo ] = [Fe~ri ] [Fe~ri ] = [h'] 2 [ V o ] + [h'] = [Fe:ril

4 2

an=O'nO

4....oo

p-l/m Oz ,

O'p = O'po P 1 / m "

(13)

A. Bernasik et al. / Fe-doped TiOe single crystals

5. Results and discussion

+o I

The thermopower and conductivity versus log (Po2) of undoped TiO2 single crystals are shown in fig. 2. The conductivity curves for both temperatures exhibit minimum values, which correspond to the change of sign of the Seebeck coefficient. Within this region an n-p-type conductivity transition takes place. The thermopower values for lower Po2 start to increase, which indicates the pure electron compensation condition. This effect was also observed by Baumard and Tani [2]. At lower deviation from stoichiometry (high Po~ range), Q as well as d Q / d In Po2 are positive. This is due to a higher concentration of electron holes but a still non-negligible electron contribution. The early data of Rudolph [ 5 ] obtained for high-purity TiO2 indicate that the presently observed hole conductivity results from incorporation of trivalent impurities, which are present at the level of 200 ppm in the undoped TiO2 sample [7]. The simultaneous conductivity and thermopower data analysis which was reported earlier [ 14 ] provides the following estimations: Eg=3.0 eV, #~=/Zp, Nc=Nv and An=Ap=0. The respective surface electrical behaviour in terms of the work-function versus log(Po2 ) measurements is shown in fig. 3. The relevant surface intrinsic region (n-p-type transition range) is shifted to lower Poz compared with the reported bulk data. In this region the WF changes more steeply (m = 2 ), while' the surface properties may be controlled by acceptortype impurities. However at lower Po2 the parameter

8O0

~

- -

"

400

~1.

~

o

L.~

,

• o lo3sr + Q 1166K

I

x

]

Q I038K

. . . . .

/

I

/

'~

,,~

/'×

/

-3.4

//~

:/

~,.3.0

ra

8OO

log( po 2 [PaD Fig. 2. Thermopower and conductivity versus log(Po2 ) of undoped TiO2 single crystals at different temperatures.

15

un--T+,s,no+

200 ,i 1507

1173 K; r n = 5 5

100

1123 K; m = 5 3 1073 K; m=5.0

,.

. 0 i

m

~

~' -

-

2

-

[ m?~gJ .:"

"" " i' .~:: " " ~' + -"

I 1

3K

~__

:+"

+° 1 + -1004

j

Crystal (001) face

~ transition

J

~,,-; ..... ,o,,.a;.,~ 2~

3

3.s

~

Iog(pO2 [Pal)

- 415

Fig. 3. WF changes versus log(Poz ) of undoped TiO2 single crystals (001 ) face, at different temperatures.

i0~5 aL% F:e -doped Ti02

Single Crystals D !

700

iJ

500

÷

LI

"~ 300,

m

1

K

i

~' -loo I -30o t

m

i

i 1149 K

=

1~:J1 K tJ [ 1039 K I

-500~

r

4

2:5

-

I

4.5

-

log( Po~ [Pal) Fig. 4. Seebeck coefficient versus log(Po2 ) of 0.5 at% Fe-doped TiO2 single crystals at different temperatures.

m assumes values in the range of the theoretical bulk value m = 5 (see table 1.) The thermopower measurement versus log(Po2 ) for Fe-doped TiO2 are presented in fig. 4. Above 1200 K Q changes sign and becomes negative below Po2 = 5 kPa, while at lower temperatures Q remains positive (within the Po2 range shown in fig. 4). This indicates that the Fermi level is shifted down towards the valence band edge as a result of incorporation of Fe 3+ and the sample acquires hole conductivity. Finally, the isothermal n - p transition range is shifted down to lower Po2 compared with undoped TiO2. Classical dc conductivity measurements were not suitable because of the observed high electrode po-

16

A. Bernasik /Fe-doped Ti02 single crysta£"

larization effects [ 14 ]. This suggests that within the near-stoichiometric composition domain Fe-doped titania may exhibit mixed conduction. A significant ionic conductivity has already been mentioned by Carpentier et al. [ 3 ] for Cr-doped TiO> despite the mixed-valence chromium model suggested by Tani and Baumard [ 6 ]. Consequently, the total electrical conductivity was measured by means of impedance spectroscopy. Both the conductivity and the Seebeck coefficient at T-- 1220 K and in the partial oxygen pressure range 1 Pa-100 kPa are presented in fig. 5. The fitted solid line is calculated based on eqs. (12) and (13). The parameter m was fixed to m = 4, following the iron incorporation reaction (5), and assuming doubly ionized oxygen vacancies are the predominant ionic defect. The resulting partial conductivity contributions are shown in fig. 6. The following parameters were determined: ai=2"08X10-4 [ Q - ~ c m - l ] , O'no=2.68X10 -3 and apo=2.91X 10 -5. The obtained data allow one to calculate relevant partial transference numbers, which are presented in fig. 7. It becomes visible that Fe-doped TiO2 behaves as a mixed conductor within the n-p-type transition domain with an ionic contribution which reaches about 20%. The theoretical value of the thermopower may be calculated by application of eq. ( 11 ) and the conductivities an and ap obtained. NLSF (nonlinear least-squares fitting) procedure results in Qc,~ values which are shown in fig. 8. The fitted parameter Eg was found to be equal

Single Crystals

totfittedline

,~ •7,

iooo

2.9

500

E

o

T=1220K

3

.=.

g~"I ~

I000

-6 32

t

~

~-

lot

n

p

ion

--,e 2 ,ooo

J/

%. ~.,!,,St_

RI 3~ 38

4

./~oo

/

Js

:~'-

p=Op0Po2 TM

~

2.5

3

3.s

Iog(po2[Pa])

4

4~

Fig. 6. Brouwerdiagram of Fe-doped TiOz at 1220 K.

I

~

0.5at%Fe-dopedTiO b Single Crystals

ea 0 8

E

"~-.

C 06

t

8

\,\\.

~ I) 4 j

J, ~ \

J

~02 I i. 15

2

25

3

35

4

45

Iog(Po2 [Pal) Fig. 7. Partial transference numbers relevant to the Brouwer diagram in fig. 6 as a function oflog(Po2 ).

500

f

0 . 5 a t % F e - d o p e d T i 02-I Single Crystals

I

~' 0 > -==-500 0 -1ooo

4,,

Q 1220K"]

~

jJ /

/ /

~r t

~--~ ~ / ~

3.1

(0.5at%Fe-doped TiO 2 I SinQle Crystals I-~

3

1000

0.5at%Fe-doped TiO 2 L 2s

2s ~ J

, , , , ~ , , , , t , , , , t ~ l , , , , t ~

].5

2

2.s

3

35

4 4.s

5

1500

Iog(po2 [Pal) Fig. 5. Thermopower and conductivity versus log(Po2) of Fedoped TiO2 single crystals at 1220 K. Conductivity data were obtained from impedance spectra taken in the frequency range 5 Hz-13 MHz, and the solid line results from fitting to eqs. (12) and (13).

-1500

1.5 2

2.5 3 3.5 4 Iog(Po2 [Pall

4.5

Fig. 8. Experimental and calculated values of Seebeck coefficient versus log(Po2 ) for Fe-doped TiO2 sample at 1220 K.

to 3.1 1 eV. Eq. ( 11 ) was developed for a pure electronic conductor, hence the observed deviation between Q and Qc,lc in fig. 8 may represent the ionic

A. Bernasik et al. / Fe-doped Ti02 single crystals

contribution to the Seebeck coefficient. The relative Fermi level position of the observed EF(Q) (recalculated from Q measurement, tl = 0, eq. ( 11 ) ) and predicted E F ~ values (based on the assumption [e'] , ~ p ~ l / 4 ) are shown in fig. 9. The deviation between the two lines is even more pronounced then in the case of the thermopower. This confirms that the Seebeck coefficient cannot be analyzed quantitatively by a simple model, neglecting the ionic contribution to the transported entropy in the Pt-(TiOEFe) system. Isothermal changes of the WF versus log(Po2 ) for Fe-doped rutile are shown in fig. 10. At lower temperatures where the bulk exhibits clearly p-type conductivity, the WF changes are almost linear, with the parameter m equal to 5.55. While the temperature increases, a sharp increase in WF is observed within

TiO21

0.5at%Fe-doped Single Crystals

-3000~

I

~- 32 0 0 -3400 ~e

-3600

\ r" " L---~

5 -3800 -4000

//

\

EF(Q) ") EFCalC

\~.

// ~ ~ • /"

J

~ i'.5'' '~''' 2'.5'' 3

3'.S' 4

4'.5' ' s

Iog(po2 [Pa])

Fig. 9. The position to the Fermi level relative of the observed Er(Q) (calculated from Q measurement, for ti=0) and predicted EF~c values (based on the assumption [e'] ~e~21/4 ).

O,

t

---i

0.Sa~ Fe doped Ti02

17

the n-p-type transition range. At the temperature T = 1140 K one may distinguish three domains: low Po: (up to 800 Pa), intermediate Po2 range (800 P a l0 kPa) and high Po2 range (above 10 kPa). These correspond to the n-type conductivity region with m = 5 , the transition range and the p-type conductivity region with m = 4.6, respectively. The surface of iron-doped TiO2 exhibits a shift of the transition region to lower Po2 compared with undoped samples. However, only a slight shift compared with the bulk value for the doped material was observed. In the transition region the surface properties are fully controlled by acceptor type dopants. Depth profile analysis was performed on differently annealed and subsequently cooled (cooling rate 10 K / m i n ) sample. Annealing was done at T = 1400 K for 20 h in oxygen or in a reducing atmosphere. Fig. I 1 represents the F e / T i ratio as a function of depth for oxidized and reduced sample respectively. It is clearly visible that the iron surface concentration significantly increases (by at least one order of magnitude) during oxidation. The Fermi level of TiO2 may drop considerably in the band-gap as the result of incorporation of trivalent ions. Therefore a small increase in iron concentration during oxidation within the near-stoichiometric composition domain may result in a strong continuous surface potential shift within the transition region. However, n-type and p-type linear WF ranges at low Po: and high Po2, respectively, may be distinguished.

i

I-"

i S,n,,oOr..,.(oO.,a.

100

-150! " ................. .~ -200 (jI -2SO U.

~

corresponding m-p. . . . . ler at high p02

~ ~

A

÷

"?'"':'"-'"" ~"7:~ ......

i

1.5

m= 5.55+ 0.03 m=4.92+0.18 m=4.72+0.10 m=4.79+0.09

2

25

3

3.5

4

4.5

5

log(p02 [Pa]) i " 973K ~ lo23K, 10z3K ; + 1113K ~ 114QK __

i

Fig. 10. WF changes versus log(Po= ) of 0.5 at% re-doped TiO 2

single crystals (001 ) face, at different temperatures.

TiO 2 + 0.5 at.% Fe I

O

ox

.~ ._N t--~ ~ E~ .-~ E~ U_ o.1

. . . . . . . .

0,1

i

I

....... i'o depth [nm]

lOO

1000

Fig. 11. SIMS depth profile analysis of Fe-doped single crystals. Iron segregationincreases for oxidized sample.

18

A. Bernasik / Fe-doped Ti02 single crystals

6. Conclusions

The experimental results indicate that the Fe 3+ ions incorporate substitutionally for Ti 4+ ions into the futile lattice. They form acceptor centers: Feq-i and this leads to a significant shift of the bulk n - p transition (change of the thermopower sign) to lower oxygen partial pressures compared with the u n d o p e d materials. It is clearly visible that Fe-doped TiO2 is a mixed conductor within the n - p - t y p e transition domain, and the ionic c o n t r i b u t i o n reaches about 20%. However, it is not yet possible to interpret quantitatively the thermopower data and estimate the transported entropy of oxygen ions in the Fedoped TiO2. The surface intrinsic region ( n - p - t y p e transition range) for u n d o p e d titania is shifted to lower Po~ compared with the reported bulk data. The near-surface hole concentration may be enhanced as a result of segregation of acceptor-type impurities. The surface of iron-doped TiO2 exhibits a shift of the transition region to lower P02 compared with u n d o p e d samples. Three d o m a i n s corresponding to the n-type region, transition range and p-type region are distinguished. In the transition region the surface properties are fully controlled by the acceptor-type dopant. The W F increase is enhanced by the strong iron segregation in oxidized condition, which was confirmed by depth profile analysis.

Acknowledgements Financial support by the European C o m m u n i t y , G r a n t No. BREU-0144C is greatly acknowledged.

References [ 1] N. Ait-Younes, F. Millot and P. Gerdanian, Solid State Ionics 12 (1984) 431. [2] J.F. Baumardand E. Tani, Phys. Status Solidi (a) 39 (1977) 373. [ 3 ] J.-L. Carpentier, A. Lebrun and F. Pedru, J. Phys. Chem. Solids 50 (1989) 145. [4] F. Millot, M.-G. Blanchin, R. Trtot, J.-F. Marucco, B. Poumellec, C. Picard and B. Touzelin, Prog. Solid State Chem. 17 (1987) 263. [ 5 ] J. Rudolph, Z. Naturf. 14a ( 1959) 727. [6] E. Tani and J.F. Baumard, J. Solid State Chem. 32 (1980) 105. [7 ] W. Kernler, PhD Thesis (Ttibingen, 1990). [8 ] C. Wagner,Prog. Solid State Chem. 7 ( 1972 ) 1. [9 ] J.P. Moore and R.S. Graves,J. Appl. Phys. 44 ( 1973) 1174. [ 10 ] J. Nowotny,M. Slomaand W. Weppner,J. Am. Ceram. Soc. 72 (1989) 564. [ 11 ] J. Nowotny and M. Sloma, in: Surface and Near-Surface Chemistry of Oxide Materials (Elsevier,Amsterdam, 1988) ch. 7, p. 281. [ 12] G. Borchardt, H. Scherrer, S. Weberand S. Scherrer, Int. J. Mass Spectr. Ion Phys. 34 (1980) 361. [13]T.C. Harman and J.M. Honig, Thermoelectric and Thermomagnetic Effects and Applications (McGraw-Hill, New York, 1967)p. 109. [ 14] A. Bernasik, M. Rekas, M. Sloma and W. Weppner, Proc. 14th Riso Intern. Symp. on Materials Science, in press.