Charge transport and trapping in BaTiO3 thin ®lms ¯ash evaporated on Siand SiO2/Si

R.E. Avilaa,*, J.V. Caballerob, V.M. Fuenzalidac, I. Eiseled

aComisioÂn Chilena de EnergõÂa Nuclear, Cas. 188-D Santiago, ChilebDepto. de FõÂsica, U. de Santiago de Chile, Cas. 307-2 Santiago, Chile

cDepto. de FõÂsica, FCFM, U. de Chile Santiago, ChiledUniv. der Bundeswehr, W. Heisenberg Weg 39, D85577 Neubiberg, Germany

Received 12 May 1998; accepted 10 December 1998

Abstract

BaTiO3 (BT) thin ®lms were prepared by ¯ash evaporation onto p-Si and SiO2/p-Si from a Re boat at 18008C in ultra high vacuum. The

®lms are amorphous and remain so after a thickness reduction by 10±20% upon annealing at 5008C for 3 min in O2 atmosphere. Annealing

raises, also, the index of refraction by some 5% and the extinction coef®cient by a factor of 2. Electron injection at the Al to BT interface of

Al/BT/Si capacitors is enhanced by the Schottky effect, yielding a value of 11.2 for the dielectric constant of BT. Modeling current±time

measurements yields a trap density of 1024 m23, 0.82 eV below the conduction band. Capacitance±voltage curve shifts due to bias stress on

Al/BT/SiO2/Si capacitors are interpreted as caused by electron injection and trapping in the BT ®lms. Starting deposition at 1708C or post-

deposition annealing reduces the trap density and increases the capacitance±voltage curve shifts by bias stress, from 0.3 to over 14 V at bias

stress of 2 10 V. q 1999 Elsevier Science S.A. All rights reserved.

Keywords: Barium titanate; Annealing; Dielectric properties; Electrical properties and measurements

1. Introduction

Perovskites such as BaTiO3 (BT), SrTiO3, and

Pb(Zr,Ti)O3 exhibit a high dielectric constant (e r), and are

ferroelectric at room or lower temperature. In addition to

their interesting physical properties these materials may ®nd

applications in piezoelectric transducers, pyroelectric and

optical devices, non-volatile random access memories

[1,2], and sensors. These applications require thin ®lm struc-

tures.

Oxygen vacancies in BT are active as traps, leading to

poor reproducibility of the electrical properties of MOS

structures [3] and current degradation under sustained bias

[4,5]. Studying the nature and energy distribution of these

traps is dif®cult when charge transport due to long-term

trapping activity is superimposed on dc leakage through

the ®lms. An approach to avoid this dif®culty, is to deposit

the thin ®lms on a thin SiO2 layer on Si.

The Al/BT/SiO2/Si structure may operate as a thin SiO2

capacitor under sustained dc bias, while presenting the high

permittivity BT ®lm at high frequency. Bare Si oxidizes

during deposition of ferroelectric ®lms, as observed by

Matsubara et al. [6] during deposition of SrTiO3 on bare

Si. The resulting SiO2 interlayer degrades the capacitance

of the overall structure. A similar observation is made by

Nam and Kim [7], who obtained a maximum relative dielec-

tric permittivity of 86 by sputtering in a 30% O2 in Ar atmo-

sphere from a target with excess SrO, and annealing at

6008C. Annealing at higher temperature, however,

decreased the effective permittivity, a result attributed to

the formation of an SiO2 layer at the SrTiO3/Si interface.

The amorphous BT ®lms exhibit a lower dielectric constant

(10±20), but higher resistivity than crystalline [8] BaTiO3.

The interest in contact of ferroelectric or high permittivity

materials to semiconductors is exempli®ed by the work of

Nagatomo et al. [9] who have built an MOS ®eld-effect

transistor on a 300 nm thick Si ®lm deposited over a

2.5 mm thick BT ®lm. The high permittivity of 100 of the

latter is considered an advantage over lower permittivity

insulators, like SiO2 or sapphire.

From the considerations above, this work focuses on the

electrical and optical characteristics of ¯ash evaporated

BaTiO3 thin ®lms on Si and on oxidized Si. Deposition on

bare Si is not intended for technological applications, but

Thin Solid Films 348 (1999) 44±48

0040-6090/99/$ - see front matter q 1999 Elsevier Science S.A. All rights reserved.

PII: S0040-6090(99)00007-3

* Corresponding author. Tel.: 1 56-2-364-6106; fax; 1 56-2-364-6255.

E-mail address: ravila@gopher.cchen.cl (R.E. Avila)

will be used to monitor charge transport and trapping in the

BT ®lm.

2. Experimental

The BaTiO3 (BT) thin ®lms were prepared as detailed in

an earlier publication [10]. In summary, the substrates are p

type, (100) oriented silicon, 10±20 V cm, with a standard

31 nm gate oxide. The BT thin ®lms were ¯ash evaporated

from a Re source at 18008C in ultra high vacuum. The

substrate was either at room temperature and allowed to

heat to approximately 708C (type I) or evaporation started

with the substrate at 1708C, which was allowed to cool

down to 708C (type II) during evaporation.

X ray diffraction (XRD) and X ray photoelectron spectro-

scopy (XPS) were used to probe for crystallinity and

elemental distribution in the thin ®lms.

Spectroscopic ellipsometry in the photon energy range of

1.65 to 2.95 eV was used to model the thickness and optical

parameters, using the Woollam Co., M44 system, which

takes the ellipsometric C and D angles at 44 photon energy

values in the given range. Data were collected at three

angles of incidence, mostly in the range of 60±808 seeking

to include data near the principal angle (corresponding to

D � 908) over most of the energy range. The use of three

scans increases data redundancy but the correlation between

the scans is high, as observed by the similarity of the respec-

tive effective n, k spectra.

Al was deposited for electrical contacts by evaporation

through a shadow mask. Current±voltage (I±V), and differ-

ential capacitance±voltage (C±V) measurements were made

at high frequency (HF). A Keithley 590 was used for HF

measurements at 100 kHz and 1 MHz. A Keithley 617 was

used for higher voltage (up to 100 V) I±V measurements.

Pulsed HF C±V measurements were performed for obser-

ving the shift of the C±V characteristics along the gate bias

axis, as a result of applying a steady dc bias. In this mode,

the bias is kept at a `base bias', allowing the capacitor to

settle. Then, short bias pulses are applied, during which the

HF capacitance is measured. Between pulses the bias

returns to base bias for a lapse at least three times the dura-

tion of the pulse.

3. Results

XRD analysis shows the ®lms to be amorphous, and

remain so after annealing at 5008C for 3 min in O2 atmo-

sphere. From XPS depth pro®les [10], the signals expected

from BaTiO3 are observed. Interdiffusion is observed at the

BT/SiO2 ®lm interface, Si and Ba being the most mobile

species, resulting in a drop by one half of the Ba signal over

approximately 8 nm.

3.1. Optical parameters

The spectroscopic ellipsometry data were ®tted with a

model which includes the ®lm thickness and optical para-

meters. The model was built on the system library data for

the Si substrate, and direct measurement of the SiO2 and BT

thin ®lms, both of which were ®tted with the Cauchy model

[11], in which

n � A 1 B=l2 1 C=l4

k � aexp 1240b l21 2 g21� �� �

A, B, C, a and b being adjustable parameters, the wave-

length, l , is expressed in nm, and g � 400 nm.

The ®t follows the experimental data closely, with values

of the mean squared error (MSE), weighted by the experi-

mental uncertainties, of the order of unity, where the devia-

tions may be due to inhomogeneities in the ®lm thickness or

optical parameters. That the Cauchy model itself is appro-

priate for this type of material was tested by ®tting data

generated out of the library dispersion relation [12] of a

60 nm BaTiO3 thin ®lm on Si. Using comparable uncertain-

ties, the Cauchy model ®ts the library generated data with an

R.E. Avila et al. / Thin Solid Films 348 (1999) 44±48 45

Table 1

Properties of BT ®lms: thickness, n and k from spectroscopic ellipsometry modeling (thickness values in parentheses from direct pro®lometer readings); n, k

are reported at 2.3 eV, the center of the instrument energy range, [*] at 2 8 V, [u] unreliable due to strong dependence on geometry, [e] erratic dependence on

the stress applied

Type I Type II

SiO2/Si SiO2/Si Si Si SiO2/Si SiO2/Si

As deposited Annealed As deposited Annealed As deposited Annealed

Thickness (nm) 65.8 51.8 87.0 76.7 93.8 82.5

(Pro®lometer) (64:3 ^ 1:1) (50:9 ^ 1:2)

n 1.85 1.93 1.88 1.90 1.88 1.93

k 1:7 £ 1023 5 £ 1023 9 £ 1024 7:8 £ 1023 1:5 £ 1023 4 £ 1023

SE e r 18.1 11.2

CV e r 20.8 16 3 9.9 [u] 20.6

DVhw at 2 10 V 1 0.3 1 2.5 1 14[*] [e] 1 2 1 4

Nt (m23) at Et 2 Ec (eV) 20 £ 1023 at 2 0.85 9 £ 1023 at ±0.82

MSE lower than that of the experimental data by factors

between 3 and 20. That the global minimum was found

was tested by starting the ®tting process at various sets of

values of the parameters. On rare occasions a local mini-

mum was found, but with a much larger MSE, than the

accepted global minimum.

The results of these analysis are reported in Table 1.

There, the ®lm thickness values carry a typical uncertainty

of ^ 0.5 nm, and compares favorably with the direct pro®l-

ometer measurements included (in parentheses) in the table.

The refraction index and extinction coef®cient increase

over the energy range (1.65 to 2.95 eV) used. n increases

typically by 5%, and k roughly by a factor of 2. The values

of n and k at the center of this energy range are included in

Table 1.

Similar values of the refractive index were found [13] in

BT thin ®lms deposited by reactive partially ionized beam

deposition (RPIB). A value of 1.90 is reported for sotoichio-

metric BaTiO3, and higher values for Ba rich ®lms, n reach-

ing the value of 2.20 for a Ba/Ti � 0.3 ®lm.

The thickness reduction with the sintering temperature is

accompanied by an increase of the refractive index and

extinction coef®cient. This correlation, on a much larger

scale, was pointed out by Trolier-McKinstry et al. [14] in

studies of PT and PZT thin ®lms deposited by the sol-gel

technique.

3.2. I±V analysis

I±V measurements on type II Al/BT/Si capacitors, are

rectifying, with higher current in the negative-top-contact

direction. The leakage current of the Al/BT/SiO2/Si capaci-

tors is below 0.1 pA at 10 V. Because of uncertainty in the

limiting process for current transport, gate bias will not be

converted to average ®elds.

Upon ®rst contact, I±V measurements on as-deposited

type II Al/BT/Si capacitors display nearly ohmic behavior

in the 0.3 to 5 V range, followed, in some cases, as shown in

Fig. 1, by a sudden increase (soft breakdown), in the 5 to

10 V range. This is attributed to local breakdown of a

complex, probably non-stoichiometric and heavily contami-

nated SiOx ®lm, at the BT/Si interface.

Following soft breakdown, the I±V curve is smooth, and

sharply rectifying with 3 orders of magnitude larger at

210 V than at 1 10 V. Over this bias range, the Si to BT

contact appears blocking for charge injection, assuming

electronic [15] transport across the BT ®lm,. On the

contrary, the Al to BT contact appears injecting, leading

to a resemblance of space-charge limited [16] (SCL) beha-

vior in the 0 to 210 V range. There, the current goes

through a series of J � J0Vn regimes, starting with an

ohmic region down to 21 V. Then, a sharp increase, in

the 21.2 to 24 V region is interpreted to be due to ®lling

a close to exponentially decaying (away from the conduc-

tion band) electron trap band. In the 24 to 210 V range, the

conduction stays close to the ®lled-trap regime (n � 2).

Further bias stress produces destructive breakdown at

bias between 170 and 1100 V, interpreted to occur at the

BT ®lm. This breakdown does not occur in annealed

samples, which may sustain the maximum stress of our

instrument (1100 V, or 22 mA).

Annealing increases the conductivity of type II thin ®lms.

As shown in Fig. 1, in the 215 to 120 V range, the Al/BT/

Si capacitors present sharply rectifying I±V characteristics,

with the absolute value of the current some 20 to 2000 times

higher at 21 V than at 11 V. The current is essentially

R.E. Avila et al. / Thin Solid Films 348 (1999) 44±4846

Fig. 1. SCL analysis of I±V curves of as-deposited and annealed BT ®lms in

type II Al/BT/Si capacitors. The lower (upper) branch of each pair corre-

sponds to V . 0 (V , 0). (a) Annealed, (b) as deposited, (c) anealed, after

soft breakdown.

Fig. 2. Schottky enhanced analysis of the data shown in Fig. 1, using the

same notation.

ohmic in the 20.5 to 10.5 V range, then, above 1 V it

saturates. Sweeping from 230 to 170 V reduces the current

in the negative branch, deviating from the SCL behavior.

After stabilizing, a Schottky enhanced [17] (SE) model (Fig.

2) provides a close ®t over the 21 to 230 V range.

In this model, the injection ef®ciency depends on the

polarization of the dielectric during the electron injection

process. This is represented by the dielectric constant, e r, at

high frequency. From this analysis of the data in Fig. 2, the

value of 11.2 obtains for e r in BT. Values of e r calculated

from the negative bias branch of the other curves are shown,

also, in Fig. 2

Other current enhancement models, like Poole±Frenkel

[18] detrapping, or Fowler±Nordheim [19] tunneling, do

not ®t the data.

3.3. C±V analysis

HF C±V characteristics of all BT capacitors, on Si and on

SiO2/Si, display the expected behavior, including accumu-

lation, depletion and inversion regions; except in the Al/BT/

Si capacitors which may go into deep depletion instead of

inversion. The Si surface is normally inverted (at zero bias).

From the analysis of HF C±V curves of Al/SiO2/Si capa-

citors, the p-Si substrate has a doping level of 2 £ 1021 m23.

The Si surface is inverted at zero bias; the (wet) SiO2/Si

interface presents a ®xed oxide charge density of 1:7 £1016 m22. Bias stress of ^ 20 V for 5 min displaces the

pulsed HF C±V measurements by less than 0.1 V along the

gate bias axis.

From the value of capacitance in accumulation, corrected

for series resistance, the dielectric constants of the BT ®lms

are given in Table 1. The value of 3 for the as-deposited ®lm

on Si is unexpectedly low. The other values are within the

range found in other experiments [20,21].

HF C±V measurements focused on displacements of the

C±V curve by bias stress. The customary procedure [22] of

bias-temperature-stress in MOS electronics involves heating

the capacitor to the 200 to 4008C range for several minutes,

cooling under sustained bias, and measuring the C±V curve

back at room temperature. The overall displacements of the

C±V curve along the voltage axis provide a measure of the

ionic charge moving in the oxide under the applied bias.

In this work, all measurements were carried out at room

temperature, so that ionic displacements are unlikely. This

was veri®ed by slow current±voltage sweeps which do not

show a current peak at zero bias. Also, the C±V curve shifts

follow, albeit with a little hysteresis, the stress bias, even

without a polarity reversal. If the shifts were due to mobile

ionic charge, the shifts would be essentially stable until a

change of polarity of the stress bias. Thus, electron injection

from the Al/BT interface is assumed, which settle at traps

(in the BT ®lm) and at the BT/SiO2 interface.

The C±V curve shifts (DVhw), at a capacitance halfway

between the accumulation and inversion levels, were

observed by pulsed HF C±V measurements after allowing

the curves to settle (5 to 30 min). As shown in Table 1,

annealing increases the C±V curve shifts, a tendency consis-

tent with the larger shifts exhibited by type II ®lms than

those of type I ®lms. The immediate suggestion is that the

shifts are caused by the diffuse interface of BT to Si or to

SiO2. Noticeably, starting the deposition at 1708C is

approximately as effective, in terms of C±V curve shifts,

as annealing at 5008C for 3 min.

3.4. Transient current analysis

The time dependence of current following a voltage step

was observed as a measure of trap activity in ®lms deposited

on SiO2/Si. The capacitors have been stressed and settled at

2 10 V, followed by a step to 2 2 V. The latter value is

chosen to keep the Si surface in accumulation, avoiding

transients due to minority carrier generation or trapping

activity in the Si, and to maintain a certain electric ®eld to

sweep out the carriers ejected from traps (or to sweep in the

carriers to be trapped). Admittedly, this electric ®eld proce-

dure introduces a minor variation (aÁ la Poole±Frenkel) of the

effective trap depth.

Analysis of the current transient according to the theory

of Simmons and Tam [23] yields an estimate of the trap

density distribution across the BT energy band gap. This

calculation requires a value for the attempt-to-escape

frequency, usually of the order of 1011 to 1012 s21. A

value of 1012 s21 has been assumed to determine the energy

scale in this work (a factor of 10 error in this value intro-

duces an error of 2.3kBT, some 0.06 eV, in energy).

Transients in annealed type I ®lms (not shown) yield trap

densities one order of magnitude below the values shown in

Fig. 3.

R.E. Avila et al. / Thin Solid Films 348 (1999) 44±48 47

Fig. 3. Trap density distribution calculated from discharge I±t measure-

ments of type II Al/BT/SiO2/Si capacitors.

4. Discussion

The values of e r from the SE analysis (er � 11:2, 13) are

expected to be close to, but lower than, values determined

from the HF capacitance in accumulation (er � 10 to 20), as

the process of electron emission into the dielectric is

assumed to occur over a time interval substantially lower

than a period of the HF (100 kHz) measurement signal. This

relation is partly satis®ed, especially so, considering the

values determined from capacitors on SiO2/Si. Moreover,

the values calculated from capacitance measurements

re¯ect the effective permittivity which may be lowered by

an SiOx layer at the BT/Si interface. The Schottky effect

analysis is free of such interference as it probes the thin

layer at the Al/BT interface where the barrier lowering

occurs. However, a calculation of the thickness of an

assumed SiOx layer is unreliable as it depends on the differ-

ence between uncertain quantities.

The SCL and SE analysis must be extended to determine

their consistency with the dependence of the current on

temperature and ®lm thickness.

The calculation of trap densities using current transients

has probed deep traps only. Measurements at earlier times,

or lower temperature will be necessary to probe traps closer

to the conduction (or valence) band. However, a distinct

rise, towards the conduction band, appears in the as-depos-

ited sample, which is reduced by annealing. The density and

energy location of these distributions coincide with our own

preliminary measurements on Ag/BT/Ti capacitors, where

the BT thin ®lms were grown by an electrochemical hydro-

thermal process.

5. Conclusions

The characterization of BT thin ®lms deposited by ¯ash

evaporation on Si and on SiO2/Si sheds light on charge

injection and trapping in the BT ®lm. The interfaces of

both Al and Si to BT are blocking for electron injection,

especially after soft breakdown, which appear to burn ohmic

injection paths. The injection from Si to BT being a few

orders of magnitude less ef®cient than from Al to BT. As the

Al to BT interface becomes more blocking, the current is

enhanced by Schottky effect. Current±time measurements

and shifts of the capacitance±voltage curves due to bias

stress on Al/BT/SiO2/Si capacitors are interpreted as caused

by electron injection and trapping in the BT ®lms. Starting

the deposition at a moderate temperature (1708C) plays a

role similar to annealing at 5008C for 3 min, in terms of C±V

curve shifts.

The analysis of current and capacitance characteristics of

structures including an SiO2 layer between BT and Si

provides much insight into the time and voltage dependence

of charge transport in the BT thin ®lm.

Acknowledgements

This work was partially funded by the Commission of the

European Communities under grant CI1CT910875, Fonde-

cyt grant 1980285 and FundacioÂn Andes contract C12510.

J.C. acknowledges Fondecyt grant 2950007.

References

[1] J. Yu, Z. Hong, W. Zhou, G. Cao, J. Xie, X. Li, Appl. Phys. Lett. 70

(1997) 490.

[2] T.B. Wu, C.M. Wu, M.L. Chen, Appl. Phys. Lett. 69 (1997) 2659.

[3] B-S. Chiou, M-C. Lin, Thin Solid Films 248 (1994) 247.

[4] R. Waser, T. Baiatu, K.H. HaÈrdtl, Mater. Sci. Eng. A109 (1989) 171.

[5] S.B. Desu, I.K. Yoo, J. Electrochem. Soc. 140 (1993) 9.

[6] S. Matsubara, T. Sakuma, S. Yamamichi, H. Yamaguchi, Y. Miyasak,

in: E.R. Myers, A.I. Kingon (Eds.), Ferroelectric Thin Films, (Mater.

Res. Soc. Symp. Proc. 200 (1990) 243.

[7] S-H. Nam, H-G. Kim, Ferroelectrics 152 (1994) 79.

[8] Q.X. Jia, Z.Q. Shi, W.A. Anderson, Thin Solid Films 209 (1992) 230.

[9] T. Nagatomo, K. Tanak, A. Yamanashi, O. Omoto, Ferroelectrics 152

(1994) 133.

[10] J.V. Caballero, V.M. Fuenzalida, R.E. Avila, I. Eisele, Mater. Res.

Soc. Symp. Proc. 403 (1996) 189.

[11] H.G. Tompkins, A User's Guide to Ellipsometry, Academic Press,

1993, and J.A. Woollam Co. M-44 System Documentation.

[12] E.D. Palik, Handbook of Optical Constants of Solids, Vol 2,

Academic Press, 1991, p. 797

[13] W.T. Liu, S. Cochrane, S.T. Lakshmikumar, D.B. Knorr, E.J. Rymas-

zewski, J.M. Borrego, T.M. Lu, Electron Devices Lett. 14 (1993) 320.

[14] S. Trolier-McKinstry, P. Chindaumom, K. Vedam, R.E. Newnham,

Characterization of ferroelectric ®lms by spectroscopic ellipsometry.

In: K. Vedam (Ed.), Physics of Thin Films, Vol. 19, Academic Press,

1994, p. 249.

[15] H.Y. Lee, L.C. Burton, IEEE Trans. Components, Hybrids, Manuf.

Technol. 9 (1987) 469.

[16] M.A. Lampert, P. Mark, Current Injection in Solids, Academic Press,

New York, 1970, p. 28.

[17] S.M. Sze, Pysics of Semiconductor Devices, Wiley, 1981, 2nd edn.,

Section 734.

[18] S.R. Dhariwal, P.T. Landsberg, J. Phys. Chem. Solids 50 (1989) 363.

[19] G. Groeseneken, H.E. Maes, IEEE Trans. Electron Devices 33 (1986)

1028.

[20] J.R. Slack, J.C. Burfoot, Phys C Solid State Phys. 4 (1971) 898.

[21] Z.Q. Shi, Q.X. Jia, W.A. Anderson, J. Vac. Sci. Technol. A 10 (1992)

733.

[22] E.H. Nicollian, J.R. Brews, MOS (Metal-Oxide-Semiconductor)

Physics and Technology, Wiley, New York, 1982.

[23] J.G. Simmons, M.C. Tam, Phys. Rev. B 7 (1973) 3706.

R.E. Avila et al. / Thin Solid Films 348 (1999) 44±4848