Thermally stimulated ionic conduction in MOS (MoSiO2Si) structures

Journal of Electrostatics, 3 (1977) 203--212 203 @ Elsevier Scientific Publishing Company, Amsterdam -- Printed in The Netherlands

T H E R M A L L Y S T I M U L A T E D I O N I C C O N D U C T I O N IN MOS (Mo'-SiO2--Si) STRUCTURES

J.C. M A N I F A C I E R , P. P A R O T

and J.P. F I L L A R D

Uniuersit~ des Sciences et Techniques du Languedoc, Pl. E. Bataillon, 34060 MontpeUier (France)

Summary Results on thin films of silicon dioxide made by a low temperature process (oxidation of silane Sill4) are presented. We used a MOS structure Mo--SiO~--Si for our investigations. Both capacitance--voltage and thermostimulated current measurements were made. A high ionic contamination (Na +) of 8 × 1012 cm -2 has been found. There is some experimental evidence that the Na + ions are more strongly trapped at the SiO2--Si interface than at the Mo--SiO2 one. We als6 observed negative charges due to electrons trapped at the interface between the SiO 2 made by oxidation of Sill,, and a thin layer of thermal SiO~ formed on the silicon during etching.

1. I n t r o d u c t i o n T h e t e c h n o l o g y o f MESA j u n c t i o n s requires t h e use o f low t e m p e r a t u r e passivation processes o f t h e silicon surfaces, and thus excludes t h e r m a l oxidation. T h e results we p r e s e n t c o n c e r n silicon d i o x i d e o b t a i n e d at l o w t e m p e r a t u r e ( 3 0 0 - - 5 0 0 ° C ) b y o x i d a t i o n o f silane Sill4 [1, 2]. T h e ionic c o n t a m i n a t i o n , which is high in these SiO2 films, has b e e n s t u d i e d using a MOS s t r u c t u r e o f Mo--SiO2--Si(n) (Fig. 1). T w o m e t h o d s w e r e used: (1) C a p a c i t a n c e - - v o l t a g e ( C - - V ) m e a s u r e m e n t s ; t h e shift o f t h e C - - V curves have b e e n used f o r t h e o b s e r v a t i o n o f ion m o t i o n in t h e films [3]. (2) T h e r m a l l y s t i m u l a t e d c u r r e n t (TSC) m e a s u r e m e n t s ; t h e t h e r m a l l y s t i m u l a t e d charging and discharging c u r r e n t s allow a s t u d y o f t h e kinetics o f ion m i g r a t i o n [ 4 ] .

v///T///~/j/"/~ VG

_Molybdenum A = 0.144 cm "2

b.--- Si02

s, (n)<~: 5,~ cm) z//////////////.-

F i g . 1. M O S s t r u c t u r e .

• ,~K)O A)

~J(d" I

J . - - - A[ ( e u t e c t i c )

204

Both methods show a preponderant ionic conduction, due to mobile positive ions (most probably Na +) and to negative charges from trapped electrons at the interface between the SiO2 made by Sill4 oxidation, and a thin layer of thermal SiO2 (10--30 A), which arises from the etching of the silicon substrate. 2. Preparation of samples [2] * SiO2, obtained by Sill4 oxidation at 425°C, was deposited on a < I l l > oriented, n-type, phosphor-doped silicon substrate (p -~ 5 ~ cm) cleaned by chemical polishing. The equations of the reaction are: Sill4 + O2 -* SiO2 + 2H2,

(I)

for low temperature and a low concentration of Sill4, and Sill4 + 202 ~

SiO2 + 2H20,

(2)

for high temperature (> 600° C) and a high concentration of Sill4. The temperature of preparation, 425°C, favoursreaction (1); the deposit rate is 700 A rain-I. The use of electrochemicaltechniques for detecting localized defects, and particularly the electrolytic copper decoration, leads to the conclusionthat the density of SiO2 defects(microcracksand pinholes) is high, viz. 500-1000 cm-2. Moreover,SiO2 prepared by this method is non-stoichiometric with an excess of silicon. The contacts are made by an Al eutectic on the Si, and by vacuum evaporated Mo on the SiO2 (Fig. 1).

3. Measuring techniques The samples were placed in ~' cryostat, the temperature of which could be varied between 77 and 520 K. The C--V measurements were made at room temperature with a PAR model 410 C--V plotter; the applied gate VG voltage was varied between -100 and +100 volts. The C--V measurements were made before and after exposing the sample to different bia~temperature stress cycles (BTS). The measuring frequency v = 1 MHz is high enough to get rid of the disturbing effects-of surface states. For the TSC measurements, we used a logarithmic picoammeter Keithley 412, the heating rate being 0.35 K s-1. To correlate the shift in the C--V characteristics with ion transport in the SiO2 films, the c u r r e n t - t i m e curve of the short~circuited sample was measured at a high temperature. *We should like to thank Mr Peyre-Lavigne and Mr Dequidt (Sescosem, Aix en Provence, France) who kindly provided the samples.

205 4. C-- V Measurements Before presenting the results obtained, we should say that, qualitatively, they were reproducible from sample to sample. But due to the high density of pinholes, the steady-state conduction current was sometimes fairly high for this sort of structure. Similarly, the low breakdown field observed: E -~ 106 V cm -1, put some restraints on the measuring conditions. 4.1 Normal C--V measurements Figure 2, Curves (b), show the C--V characteristics at T = 300 K and u = 1 MHz. The samples were measured as received without any prestressing. I

I

I

I

,

T

)

i

/--Co

v

1000

f

dVFB:

o

36 V

8OO

-

--(a)

(b)

6oo A : 0.144 c m -2 I

I

I

Y = 1 MHz I

-I00

I

I

I

I

°100

0

V~ ( v o l t S )

Fig. 2. C - - V m e a s u r e m e n t s , v ffi 1 M H z a n d T = 3 0 0 K . (a) T h e o r e t i c a l C--V c u r v e ; ( b ) e x p e r i m e n t a l C--V c u r v e .

Using the theory of M O S structure given in [5], we can deduce from Co a SiO2 thickness: d = 0.495 ~m. Furthermore, the ratio (ref. 5, p. 436): C ~ F)/Co = 0.65 gives for the impurity concentration in the silicon space charge region: N D = 1.2 × 10 Is c m -3. In Fig. 2 (Curve (a)) we have also shown the theoretical curve of an ideal M O S structure [6]. W e observe: (1) that Curves (a) and (b) are not parallel,which, according to theory [5], is attributed to the presence of fast surface states, the occupancy of which being a function of surface potential at the Si--SiO2 interface; (2) that Curves (a) and (b) are shifted towards negative gate voltages, which is characteristic of a fixed positive space charge in the SiO2. W e have [7] (cf. Fig. 3): A v

=

v G

-

ex,,

=

(Q./Co)

~ms = Cm - [X + ( E o / 2 ) - ~F]

-

¢,,,,

(3)

(4)

206

where @m, is the metal--semiconductor work function difference, ~ . 4.28 eV the Mo work function [8], × = 4.05 eV the silicon electron affinity, E~ = 1.15 eV the silicon gap, and @f = Ei - EF = 0.3 eV. By substituting the given values in eqns. (3) and (4), we obtain @ms ~- -0.05 eV, which is a negligible correction. In the case of thermal SiO2 free of ionic contamination, Qs, represents the positive space-charge localized at the Si--SiO= interface, which is due to an excess of Si [9]. But in the preparation method used here, this net positive "space charge" Q® (C cm -2) probably is uniformly distributed throughout the SiO2. It could be due to an excess of Si or to an excess of alkali ions, probably Na +. The voltage shift A VFB with respect to the flat-band condition (no space charge in the semiconductor) leads to (Fig. 3): =

QG + Qo + Q .

(5)

= O,

d

f0 E d x = --AVFB = - 3 6 V .

(6)

In eqn. (5), we have introduced a negative charge due to trapped electrons, which is equal to (Sect. 4.2) Q e / - q = 2.6 x 101~ cm-2. QG is the charge on the metallic electrode, and Q~ is the positive uniform charge in the oxide.

.

~ ~ ~

WMc

~"

cations

'/.

~ ~ ~. "f" "/" /

11.

........T'I:I;' Vacuum

~

I

I

I qAVGFB I

)

(a)

Mo

SiO 2

I

I

Si --

,9(x) F/ / / / / / /

Thermal

/¢'¢'/~

SiO 2

--

X

o

(b)

-

-

QG

Q@

Qe

Fig. 3. (a) Band structure: fiat band condition; (b) charge distribution, T = 300 K.

207

Using Gauss' theorem for the electric field, we obtain: Q®/q = (2Co/q) A VFs - ( 2 Q J q ) -~ 8 × 1 0 1 2 c m -2.

(7)

4.2 C - - V measurements after a bias.temperature stress So far, we have n o t taken into account the influence of interfacial states that we have mentioned earlier, as well as the important hysteresis effect that can be observed after a heating cycle (Fig. 4). This hysteresis effect cannot be explained by ion drift and should be related to slow interface trapping. o v

lOO0 ~9

800

A = 0.144 crn-2

-10

6OO

'1,/= 1 MHz 0

*100 vG ( v o l t s )

Fig. 4. C--V measurements, v = 1 M H z , a n d T = 3 0 0 K. (a) Theoretical C--V curve; (b) experimental C--V curve (before BTS); (c) C--V curve after a -BTS: V = -50 V, T = 380K, and t = 5 rnin; (d) C--V curve after a -BTS: V ffi-50 V, T ffi380K, and t ffi 16 rain; (e) C--V curve after a +BTS: V = +50 V, T = 380K, and t = 1.5 rnin.

The results we shall present in this section can be explained by the presence of: (1) a mobile positive charge Q® due to alkali ion contamination (Na*); and (2) a much less " m o b i l e " negative charge Q~ whose origin is most probably electrons trapped at the Si--SiO~ interface (Fig. 3). It should be mentioned, however, that IR absorption measurements showed the presence of a high concentration of Si--OH bonds. Hence, the contamination by negative ions such as OH- due to the m e t h o d of preparation, cf. eqns. (1) and (2), cannot be ruled out, but these negative ions have been shown to be immobile [10] under the experimental conditions used here. We applied a positive or a negative bias for variable lengths of time at temperatures between 370 and 420 K (+BTS and -BTS). Next, the shift in the C - - V curves was measured at 300 K, cf. Fig. 4. The shift observed is characteristic of an ionic displacement. 4.2.1 +BTS After a +BTS, V = +50 V, T = 380 K, and t = 1.5 min, s~.e Curves (e) of

208

Fig. 4, we observed a very rapid displacement towards accumulation (negative voltage drift) and after a +BTS, 50 V, 380 K, and 3 min, the accumulation is complete over the range VG = - 1 0 0 to +100 V (not represented in Fig. 4). According to the analysis of Snow et al. [3], we can infer that the positive ionic charge Q. moved towards the Si--SiO2 interface, and neutralized the Q~ charge due to trapped electrons. We can derive a minimum value for Q./q. Since Q J - q = 2.6 × 1012 cm -2 (Sect. 4.2.2) and AVFB > 100 V:

(Q. + Q.)/q •

AVFBCo/q

--

4.3 × 1012 cm -2

(8)

indicates that Q./q > 6.9 × 1012.cm-2. This lower value for Q., which is attributed to the mobile Na + ions, is only slightly lower than the value given earlier, c.f. eqn. (7). This rules o u t the possibility that an important contribution to Q~ could derive from mixed positive charges such as Si + in the silica [10]. Application of - B T S can bring the C--V characteristic back to its initial position, b u t in contrast with the findings in [3 ], application of a temperature stress to a shorted sample does n o t shift the C--V curves appreciably towards their initial position (see Curves (b) of Fig. 4). This could be due to a particular charge distribution with a low internal electric field, and to a strong trapping of the Na ÷ ions due to coulombic attraction with trapped electrons.

4.2.2 - B T S After - B T S = - 5 0 V, 380 K, and 5 min, we observed a displacement towards positive voltages, see Curves (c) of Fig. 4. This displacement arises from the motion of Na ÷ ions towards the Mo electrode. It is, however, much slower than for a +BTS, cf: Curves (c) and (d) of Fig. 4, which can be attributed, as we have already shown to the stronger trapping of some Na + ions at Si--SiO2 interface. For this particular sample, the maximum A VFB shift is obtained (Curves (d) of Fig. 4) for a - B T S : VG = - 5 0 V, T = 380 K, and t = 16 min. We can evaluate the trapped electron concentration corresponding to this shift of A V F B = --60 V towards a positive voltage Q . / - q ~-- AVFBCo/-- q

=

2.6 × 1012 cm -2 .

(9)

With a temperature stress (T = 380 K) applied to a shorted sample, we can shift the C--V curve back towards its initial position, cf. Curves (b) of Fig. 4, after a - B T S (but n o t after a +BTS as we have seen). This could be due to a higher internal electric field, or to a weaker trapping of Na ÷ ions at the Mo-SiO2 side. We also measured, see Fig. 5, the isothermal depolarization current at a high temperature for a short-circuited sample in the following series of measurements: (1) C--V at T = 300 K after a - B T S ; (2) I--t at T = 370 K (Fig. 5); and (3) C--V a t T = 3 0 0 K . Measurement (2) gives ( l / q ) f I dt = AQ~/q = 4 × 1011 cm -2 which is of the

209

|

I

I

I

1

i

I

l

1 10

i

!

U

I

l

r

I

I

20 t (rnin)

Fig. 5. Isothermal " d e p o l a r i z a t i o n " current at T = 370 K, after a - B T S .

same order of magnitude as Q®/qcalculated from the C--V shift between measurements (1) and (3). We must note at this stage another feature in accordance with trapped electrons near the Si--SiO2 interface. We observed that starting from the C--V curve (d) of Fig. 4 after having applied a - B T S ( - 5 0 V, T = 380 K) for one hour, the curve starts to shift towards negative voltages, b u t at a much slower rate. Due to the negative bias applied, the Na ÷ ions are still on the Mo side; this shift can then be attributed to the release of trapped electrons (which are n o w in a non-equilibrium state, the silicon being strongly inverted) via tunneling through the thermal SiO2 into the silicon conduction band, see Fig. 6. cotions

electrons

I ~'f///ff//

Fig. 6. Energy band diagram representing the release of trapped electrons.

5. Thermally stimulated current measurements We made TSC measurements between 77 and 500 K using a heating rate of 0.35 K s -1. Only the experimental results obtained between 200 and 500 K are reported. During the entire measuring cycle (high temperature excitation -- cooling to low temperature -- TSC measurements during heating) care was taken to stay either in the accumulation, or in the inversion, m o d e of the silicon. This assures that during the TSC measurements the Fermi level is always pinned at the Si--SiO2 interface and the contribution of fast surface states is weak.

210

I(A)

,

r

I(A)

;

,

A : 0 . 1 4 4 c m ":~

d : 4950

.~

1°8 t

200

300

400

500

200

~0

L14oo , 370

~o

(TK)

Fig. 7. TSC measurements. (a) Start: complete inversion, initial stress -BTS (-50 V, 420 K, and t = 20 rain); biased sample cooled to 77 K; TSC of shorted sample. (b) Start: complete accumulation, initial stress +BTS (+50 V, 420 K, 5 rain); biased sample cooled to 77 K; TSC with a bias of -8 V. (c) Start: complete inversion (-BTS); shorted sample cooled to 77 K; TSC with a bias of -8 V. (d) Start: complete inversion (-BTS); polarized sample (-8 V) cooled to 77 K; TSC with a bias of-8 V. (1) After a -BTS excitation (-50 V, 420 K) with the silicon being inverted, the biased sample is cooled and the short~circuit TSC current is measured during heating to 400 K, see Curve (a) of Fig. 7. The high TSC current observed could arise from a transport of Na + from the Mo side towards the Si side induced by the internal field or by diffusion. An Arrhenius plot (Curve (a) of Fig. 8) gives: Ea = 0.55 eV; this activation energy is a representative of the feasibility of the Na+ ion displacement [11] from the Mo side towards the Si side. (2) The structure being in the same initial conditions as before (silicon surface inverted), we short~circuited the sample and cooled it to 77 K (Curve (c) of Fig. 7). We then measured the TSC current during heating with an applied bias of - 8 V. After the high temperature biasing, the ionic charges are slightly displaced during cooling in short~circuit. The peak observed at 370 K is due to the displacement of the Na ÷ ions towards their equilibrium position. (3) To confirm this, we started with the same initial conditions (Curve (d) of Fig. 7) with a - 8 V bias applied during cooling and heating. Since no Na ÷ ions had moved, we observed no peak. The rising current observed for T > 400 K could be due to a conduction current or thermal release of trapped electrons on the Si--SiO2 side. (4) After a +BTS (50 V, 420 K) with the silicon surface being accumulated, the biased sample is cooled to 77 K, see Curve (b) of Fig. 7, and the TSC

211

[ (A)

I0"e~

l

i

t

[

l

i

t

!

(~)5 ~ 005 eV

\

\

\

(b) 1.1 eV

\ 10-11 2

3

4

~

(K -~)

Fig. 8. A c t i v a t i o n energy p l o t o f ?,he TSC measurements.

current is measured with a negative bias of -8 V. Compared with Curve (a) of Fig. 7, the current is more than one order of magnitude lower with an activation energy Ea = 1.1 e V (Curve (b) of Fig. 8). This indicates that the N a ÷ ions are more strongly trapped on the Si--SiO~ side. This is in accord with the absence of a shift in the C--V curves after a high temperature (T = 380 K) stress when the sample is short-circuited. The high level of the T S C attained at T = 500 K (Curve (b) of Fig. 7) suggests that the density of N a + ions is very high. (The fact that the conduction current was two orders of magnitude lower -- Curves (c) and (d) of Fig. 7 -- cannot explain this.)

6. Conclusion Both the C--V and TSC methods have revealed in the SiO2 layer two types of " m o b i l e " charges: cations (Na ÷) and electrons trapped on the Si--SiO2 side. The C--V curves have enabled us to estimate the density of the positive ion charge: Q®/q ~ 8 × 10 ~2 cm -2, and from these measurements we can conclude that the Na ÷ ions are less strongly trapped on the Mo side, because the "activation energy" for the displacement of Na ÷ ions from the Mo side towards the Si side is only 0.55 eV, while it is 1.1 eV in the reverse direction.

Acknowledgments We wish to t h a n k A.K. Jonscher and J. van T u r n h o u t for m a n y valuable discussions and suggestions.

212

References 1 B.J. Baliga and S.K. Ghandhi, J. Appl. Phys., 44 (1973) 990. 2 G. Chabert, I~tude de la cin~tique de d~pSt de siliceobtenue ~ partir d'oxydation du silane, Sescosem, Aix en Provence, France. 3 E.H. Snow, A.S. Grove, B.E. Deal and C.T. Sah, J. Appl. Phys., 36 (1965) 1664. 4 J. van Turnhout and A.H. van Rheenen, Thermally stimulated charging and discharging currents in SiO2, in 1975 Annual Report of N A S - - N R C Conference on Electrical Insulation and Dielectric Phenomena. 5 S.M. Sze, Physics of Semiconductor Devices, Wiley Interscience, N e w York, 1969. 6 A. Goetzberger, Bell. Syst. Tech. J., (1966) 1097. 7 A.S. Grove, B.E. Deal, E.H. S n o w and C.T. Sah, Solid-State Electron., 8 (1965) 145. 8 A.G. Milnes and D.L. Feucht, Heterojunctions and Metal--Semiconductor Junctions, Academic Press, N e w York, 1972, p. 158. 9 B.E. Deal, M. Sklar, A.S. Grove and E.H. Snow, J. Electrochem. Soc., 114 (1967) 266. 10 B.E. Deal, J. Electrochem. Soc., 121 (1974) 198 C. 11 C. Bucci, R. Fieschi and G. Guidi, Phys. Rev., 148 (1966) 816.