Thermal annealing behavior of ion-implanted silica glass

Journal of Non-Crystalline Solids 163 (1993) 59-64

jOTRNA ~ or

NON-CRYSTALLI~SOLIDS

North-Holland

Thermal annealing behavior of ion-implanted silica glass Kohei Fukumi, Akiyoshi Chayahara, Hiroshi Yamanaka, and Mamoru Satou

K a n e n a g a Fujii, J u n j i H a y a k a w a

Gor'ernment Industrial Research Institute, Osaka, 1-8-3l, Midorigaoka, Ikeda, Osaka 563, Japan

Received 17 August 1992 Revised manuscript received 14 May 1993

Infrared reflection spectra have been measured in silica glasses which are initially damaged by 2 MeV O +- and 1.5 MeV C+-ion implantations then isothermally annealed at 200, 300, 400, 500 and 600°C. Thermal annealing effects on Si-O-Si bond angles have been deduced from the peak wavenumber shift of the 1100 cm- l band in the infrared reflection spectra. The relaxation of Si-O-Si bond angles obeys the stretched exponential decay function with/3 ranging from 0.28 to 0.15. It is deduced that the relaxation of Si-O-Si bond angles has a distribution of activation energies. The major relaxation mechanism of Si-O-Si bond angles has an activation energy of 118 kJ/mol. It is inferred that the relaxation of Si-O-Si bond angle is due to a small movement of SiO4 tetrahedra without the disruption of Si-O bonds.

1. Introduction Ion implantation is a useful m e t h o d for modifying the optical, chemical and mechanical properties of glass surfaces [1]. T h e modification of properties is caused by the introduction of ions into the surfaces of glasses and the d a m a g e of these glass surfaces, although these h a p p e n simultaneously during ion implantation. T h e implanted ions disperse in the surface region of glasses [2,3] or gather t o g e t h e r to form colloid particles in some cases [4-10]. T h e d a m a g e comprises defect formation [11,12], c h a n g e of glass structure [13-15] and ejection of ions in glass substrates [16] by ion implantation. Since the d a m a g e sometimes causes unfavorable modification in the surface properties as well as favorable ones, it is necessary to remove the d a m a g e which brings about the unfavorable modification in glass surfaces. Since the d a m a g e can be r e m o v e d by thermal annealing [17], it is important to study

the annealing behavior of damage. The structure of silica glass d a m a g e d by ion implantation has b e e n studied by X-ray diffraction [13], R a m a n microscopic spectral [14], infrared reflection spectra [14] and infrared a t t e n u a t e d total reflection spectral [15] measurements. It was found that with ion implantation there were a decrease in an average S i - O - S i b o n d angle between S i O 4 t e t r a h e d r a [13-15], the formation of small m e m b e r e d rings such as t h r e e - m e m b e r e d rings [14] and a reduction of the middle range ordering in silica [13]. In these structural features, the S i - O - S i b o n d angle between SiO 4 tetrahedra is one of the principal p a r a m e t e r s describing the network structure. In this study, the relaxation of S i - O - S i bond angle with thermal annealing has been studied by the infrared reflection spectra measurements. T h e relaxation mechanism of S i - O - S i b o n d angle is also discussed.

2. Experimental Correspondence to: Dr K. Fukumi, Government Industrial

Research Institute, Osaka, 1-8-31, Midorigaoka, Ikeda, Osaka 563, Japan. Tel: 481-727 51 8351. Telefax: 481-727 51 9629.

O ÷ and C + ions were implanted in optically flat silica glass plates (Yamaguchi Nippon Silica

0022-3093/93/$06.00 © 1993 - Elsevier Science Publishers B.V. All rights reserved

60

/( Fukumi et al. / Ion-implanted silica glass

Glass Co. Ltd; O H content is about 1000 wt ppm) at room t e m p e r a t u r e with an NT-1500 tandem type accelerator (Nisshin High Voltage Co Ltd). O+-ion implantation and C+-ion implantation were performed in acceleration energies of 2 MeV and 1.5 MeV, doses of 1 × 1016 i o n s / c m 2 and 0.75 × 1016 i o n s / c m 2 and ion currents of 0.64 ~ A / c m 2 and 1.25 p~A/cm 2, respectively. The ion-implanted glass plates were cut into several samples. Each sample had an ion-implanted surface of 5 x 12 m m 2. Isothermal annealing of five samples was carried out, one at each temperature, 200, 300, 400, 500 and 600°C, in an ambient atmosphere. In the process of the isothermal annealing, the sample was put on a platinum foil in an alumina box in a p r e h e a t e d furnace, heated at a fixed temperature, taken out from the furnace and allowed to cool in air. Infrared reflection spectra of the ion-implanted surfaces of the annealed samples were measured at an angle of 20 ° off normal incidence with a Nicolet 60SXR Fourier-transform infrared spectrometer equipped with an apparatus for reflection spectra. The resolution of infrared m e a s u r e m e n t was 4 c m - ] . The m e a s u r e m e n t of the p e a k wavenumber was repeatable within about 2 c m - i .

3. Results

Figure 1 shows the infrared reflection spectra of the O+-ion implanted glasses along with the spectrum of the unimplanted glass. Reflection bands are observed at about 1100, 800 and 470 c m - 1 in the unimplanted glass. It is seen that the band at 1100 c m - 1 shifts toward lower wavenumber by 26 c m - 1 and decreases in intensity with O+-ion implantation. In addition, O +-ion implantation causes an increase in intensity of the band at 800 c m - 1 and a decrease in intensity of the band at 470 cm -]. New bands a p p e a r at about 1000 and 600 cm -1 after the O+-ion implantation. C+-ion implantation caused changes in spectra similar to those for O+-ion implantation. The annealing changes the spectral profile of the O+-ion implanted glass back toward that in the unimplanted glass. The band at 1100 cm-1 in the O+-ion implanted glass shifts towards higher

¢M ~A -i. ;,14

.d

/'

ill :il

E

~!

.,,s

i.,'i

.-":!l

l~i

/i"

rr

1600

1200

i 800

400

Wavenu tuber (cm-~) Fig. 1. Infrared reflection spectra of O+-ion implanted and unimplanted silica glasses. Solid, dashed, dash-dotted and dotted lines show as-implanted glass, glass annealed at 400°C for 7.2×103 s, glass annealed at 500°C for 3.6×103 s and unimplanted glass, respectively.

wavenumber and increases in intensity after annealing. In addition, the bands at 1000, 800 and 600 c m - I decrease in intensity and the band at 470 cm -1 increases in intensity after annealing. The annealings at 600°C for 7200 s and at 500°C for 92700 s make the spectral profile of O+-ion implanted glass almost the same as that of the unimplanted one. The spectra of the C+-ion implanted glass had annealing behavior similar to that of the O+-ion implanted glass. Figure 2 shows the relationship between the peak wavenumber of the 1100 cm -1 band of the O+-ion implanted and C+-ion implanted glasses and the annealing time. For simplicity of representation, the changes in peak wavenumber are normalized according to

nw(t) =

(w(t)

(1)

where w(0), w(o0) and w ( t ) represent the peak wavenumber of the 1100 cm - I band in the as-implanted glass, the unimplanted glass and the glasses annealed for time, t, respectively. The right hand of the ordinate in fig. 2 represents the normalized peak wavenumber change, Rw(t). It can be seen that the 1100 cm 1 band in the C+-ion implanted glass has annealing behavior similar to that in the O +-ion implanted glass. The band shifts toward higher wavenumber by the annealing at the initial stage. The band, however, shows little changes in peak wavenumber by pro-

K. Fukumi et al. / Ion-implanted silica glass 1095

-E

1100 ,

~ 1110 E

~ ~

0.5 ~

' 1~20 11250~

I

I

1.0

2.0

0

2.5

Heating time (105sec) Fig. 2. Relaxation behavior of the 1100 cm -1 band in infrared reflection spectra. Circles, triangles, squares, diamonds and inverted triangles represent the annealing temperatures of 600, 500, 400, 300 and 200°C, respectively. The hexagon and arrow show the as-implanted and the unimplanted glasses. Open and closed marks show O+-ion implanted and C+-ion implanted silica glasses, respectively. Solid lines show the stretched exponential decay function curves at each annealing temperature.

longed annealing and does not recover to the peak wavenumber of the unimplanted glass within the timescale of this experiment. A change in peak wavenumber is observed in the glass annealed at low temperature such as 200°C and the annealing at higher temperature causes larger changes in peak wavenumber.

4. Discussion

The infrared reflection spectra of O +- and C+-ion implanted glasses have the bands at about 1100, 1000, 800, 600 and 470 cm -1, as shown in fig. 1. The assignment of these bands has been discussed in detail elsewhere [14,18,19]. Since the reflection intensities of bands are only semiquantitative, the peak wavenumber of the 1100 c m - 1 band due to S i - O - S i stretching vibration is discussed in this study. Although the ion-implanted glass consists of the damaged surface region and the undamaged inner region, the spectra around 1100 cm -1 show only the structure of the damaged surface, because of large extinction coefficients (2.6-1.5 [20]) in the wavenumber region of 1090-1130 cm -~ and thick damaged regions (2.6 ~ m for the O+-ion implantation according to the reflection spectrum measurement

61

and T R I M code calculation [13] and 2.0 ~m for the C+-ion implantation according to the TRIM code calculation). The 1100 cm - l band shifts toward lower wavenumber by 26 cm -~ with O ÷- and C+-ion implantations, indicating that the average S i - O Si bond angle decreases with O +- and C+-ion implantations from the studies on the infrared reflection, infrared attenuated total reflection, Raman scattering and X-ray diffraction measurements [13-15]. The decrease of the peak wavenumber by 26 cm i corresponds to a decrease in S i - O - S i bond angle by 7" according to the central force ideal continuous random network model (CF-ICRN model) [19,21], shown in to 2 = ( a / m o ) ( 1

-

cos 0) +

(4a/3msi),

(2)

where w, m, a and 0 are the angular frequency of the 1100 cm - l band, the mass of atoms, the force constant and the S i - O - S i bond angle, respectively. On the other hand, the band shifts back toward higher wavenumber during annealing, showing that the average S i - O - S i bond angle of the O +- and C+-ion implanted glasses increases toward that of the unimplanted glass during annealing. The peak wavenumber of the 1100 c m - l band is proportional to the square root of the linear function of cos 0 according to the CF-ICRN model, as seen in eq. (2). The normalized peak wavenumber change, R w ( t ) , is, however, nearly equal to the normalized change in S i - O - S i bond angle, ( O ( t ) - 0(~))/(0(0) - 0(~)). This relation is valid on condition that 0 ( 0 ) - 0 ( ~ ) and w ( 0 ) w(~) are small as in the present study, where 0(0), 0(~) and O ( t ) represent the S i - O - S i bond angles of the as-implanted glass, the unimplanted glass and the glass annealed for time, t, respectively. The difference between Rw(t) and the normalized change in S i - O - S i bond angle is 0.02 at most (see appendix). This value shows that R w ( t ) represents the bond angle change during annealing, i.e., the relaxation of bond angle during annealing. The relaxation behavior of various physical quantities has been extensively studied in glasses [22-24]. It was shown that the relaxation behavior in glasses obeys the stretched exponential decay

62

I( Fukumi et al. / Ion-implanted silica glass

function form (eq. (3b)), but not the simple exponential decay function form (eq. (3a)): tb(t)

=

e(-t/').

(3a)

~ ( t ) = e (-`/%")~,

(3b)

where ~b(t), 7, reff and /3 represent the normalized change of a physical quantity, the relaxation time, the effective relaxation time and the parameter (0
Table 1 Parameters in the stretched exponential function and the distribution maximum of log ~--1((log~--1)max) Annealing temperature (°c) 200 300 400 500 600

/3

%ff (s)

(log 'l" l)ma x

S2

0.28 2.5X 10 7 7.0x 10-4 0.28 1.4x 106 3.5 x 10 4 0.25 4 . 9 x 10 4 6.2x 10-4 0.20 2.3x 103 8.8×10 4 0.15 8.3x 101 8.7x 10 4

-17.4 -14.5 - 10.9 -7.4 -3.2

s 2, mean square residuals of the least square fitting. the distribution function of r -1, in order to survey the distribution of the activation energies. It is easily derived from eq. (4) that the distribution function of log r - t possesses a distribution maxim u m , (log Z - 1 ) . . . . at l o g ( 2 1 / n r o t ) . T h e (log r--t)max are shown in table 1. Figure 3 shows the relationship between the (log 7-t)ma x and the annealing temperature. It is seen that the (log r-X)m~x obeys the Arrhenius' relation, indicating that the relaxation mechanism around (log r-1)m~x is unchanged by the variation of annealing temperature. The activation energy obtained from fig. 3 is 118 k J / m o l . Since the activation energy is derived from the distribution maxiT(*C) -2

600 i

400 i

i

200 i

i

o

-5

E

~ -10

7

g(r -l)

FITO

( r - i t 0 ) (-1

-n/2)

e[-(~ 1~0) "1,

0

(2rr/3) |/2

(4)

-15

n =/3/(1 -/3), T0 = reff [ / 3 ( 1 --/3)(1--/3)/fl ] - I

Suppose that the relaxation time of S i - O - S i bond angle obeys the Arrhenius' relation, the distribution function of the inverse relaxation time in logarithmic scale, log 7-1, is more suitable than

-20 lIT (103 K') Fig. 3. The relationship between the distribution maximum of log z-1((log ~'-1)m~x) and the annealing temperature. The fit of the data to the function (log ~" 1)max = A + ( B / T ) is shown by the solid line.

K. Fukumi et al. / Ion-implanted silica glass

mum (log ~'-1) . . . . it is deduced that the major relaxation mechanism of S i - O - S i bond angle has an activation energy of 118 k J / m o l . Full width at half maximum of distribution of log r-~ in the annealing at 400°C is about 10.4, which corresponds to 58 k J / m o l assuming that the pre-exponential factor of Arrhenius' relation is independent of the relaxation mechanism having different activation energies. The activation energy obtained above is much smaller than that of the viscous flow (500-700 k J / m o l between 900 and 1400°C [26]), showing that the relaxation of S i - O - S i bond angle is not due to the structural re-arrangement accompanied by the disruption of S i - O bonds. As shown in fig. 2, the O +- and C+-ion implanted glasses have similar annealing behavior. This similarity indicates that the relaxation of S i - O - S i bond is not due to the reaction between the implanted ions and the damaged glass, since these ions would have different chemical interaction with silicon ions in the glass. This result is reasonable in the present study, because the regions where the implanted O + and C + ions are present occupy only 16% of the damaged regions (the full widths at half maximum of the O + and C + ions distribution are about 0.4 and 0.3 p.m according to T R I M code calculation, respectively). An activation energy, of the order of 100 k J / m o l , was observed in the diffusion of inert gases in silica glass [27,28]. It was shown that the activation energy of the diffusion increases proportionally with an increase in a square of radius of the gas atom. For example, He, Ne, Ar and Kr gas diffusions have activation energies of 22, 38, 105 and 184 k J / m o l , respectively [27]. The activation energy of gas diffusion corresponds to the energy of elastic distortion of glass network to pass these gases in glass network [27]. Larger distortion of glass network in the gas diffusion requires larger activation energy. Since it is plausible that the distortion of glass network in the gas diffusion is caused by the change in S i - O - S i bond angle, larger activation energy would be required for the larger change in S i - O - S i bond angle in the relaxation of ion-implanted silica glass. Therefore, we deduce that the thermal relaxation of S i - O - S i bond angle in ion-im-

63

planted silica glasses is due to the slight movement of SiO 4 tetrahedra without the disruption of S i - O bonds. During implantation, silica glass is subjected to high pressure and high temperature [29], which changes structure and densities silica glass. On the other hand, it is known that silica glass is densified by the application of high pressure, which is accompanied with a decrease in S i - O - S i bond angles [30]. The pressure-densified silica glass decreases in density with heat treatment not described by a simple exponential decay [31,32]. Since the peak wavenumber of the Raman band is proportional to the density in the pressuredensified glass [33], it is reasonable to compare the relaxation behavior of the density in the pressure-densified glass with that of the peak wavenumber of the infrared reflection band in the ion-implanted glass. The relaxation behavior of the density does not obey the stretched exponential decay function involving unique 13 and ~'eff in the glass densified at high temperature and high pressure [34], contrary to that of the S i - O - S i bond angle in the ion-implanted glasses. We suggest that the ion-implanted silica glass has a structure different from that of the pressuredensified silica glass.

5. Conclusion It is found that the relaxation of S i - O - S i bond angle obeys the stretched exponential decay function with /3 = 0.28-0.15 in O + and C+-ion implanted silica glasses. It is deduced that the relaxation of S i - O - S i bond angles has a wide distribution of activation energies. The major relaxation mechanism of S i - O - S i has an activation energy of 118 k J / m o l . These effects can be explained by the relaxation of S i - O - S i bond angles by a slight movement of SiO 4 tetrahedra without disruption of S i - O bonds.

Appendix Equation (2) can be expanded in a power series of w - 0 by using the Taylor expansion, as

1~ Fukumi et al. / Ion-implanted silicaglass

64

shown in

w = ( a / m o ) l / 2 [ A + B A 2 + CA4],

(5)

A = D 1/2, B = ( - l / 4 ) D -1/2, C = ( - 3 / 9 6 ) D -3/2 + ( 1 / 4 8 ) D - t / 2 , D = 2 + ( 4 / 3 ) ( m o / m s i ) a n d A = "rr - 0. T h e t e r m s h i g h e r where

than fourth order are negligibly small. When (5) is s u b s t i t u t e d f o r oJ i n eq. (1), w e o b t a i n Rw(t)=(EA0 /(E

eq.

t+FAO2t + G AO3t + H A O t 4) A0 o +F

A0oz + G A 3 + H

A0~),

(6) where E = 2A(~)B + 4A(~)3C, F = B + 6A(oo)2C, G = 4 A ( ~ ) C , H = C, A0 t = 0(t) - 0(~), A00 = 0(0) - 0(~) a n d A ( ~ ) = rr - 0(~). Since t h e t h i r d a n d f o u r t h t e r m s in t h e d e n o m i n a t o r a n d t h e n u m e r a t o r a r e small e n o u g h to b e n e g l e c t e d in eq. (6), R w ( t ) c a n be w r i t t e n as Rw(t ) = (A0t/A00)

+

(AOt//AOo)[F(AO t - AO0)

/ ( E + F A00) ] .

(7)

T h e s e c o n d t e r m in eq. (7) c o r r e s p o n d s to t h e d i f f e r e n c e b e t w e e n R w ( t ) a n d A 0 t / A 0 0 . T h e diff e r e n c e has a m a x i m u m at A0 t = ( 1 / 2 ) A00. T h e m a x i m u m d i f f e r e n c e is a b o u t 0.02 in case o f t h e p r e s e n t study.

References [1] P. Mazzoldi, J. Non-Cryst. Solids 120 (1990) 223. [2] A. Polman, A. Lidgard, D.C. Jacobson, P.C. Becker, R.C. Kistler, G.E. Blonder and J.M. Poate, Appl. Phys. Lett. 57 (1990) 2859. [3] A. Garnera, P. Mazzoldi, A. Boscolo-Boscoletto, F. Caccavale, R. Bertoncello, G. Granozzi, I. Spagnol and G. Battaglin, J. Non-Cryst. Solids 125 (1990) 293. [4] G.W. Arnold and J.A. Borders, J. Appl. Phys. 48 (1977) 1488. [5] A. Perez, M, TreiUeux, T. Capra and D.L. Griscom, J. Mater. Res. 2 (1987) 910. [6] K. Fukumi, A. Chayahara, K. Kadono, T. Sakaguchi, Y. Horino, M. Miya, J. Hayakawa and M. Satou, Jpn. J. Appl. Phys. 30 (1991) L742. [7] H. Hosono, Y. Suzuki, Y. Abe, K. Oyoshi and S. Tanaka, J. Non-Cryst. Solids 142 (1992) 287.

[8] H. Hosono, H. Fukushima, Y. Abe, R.A. Weeks and R.A. Zuhr, J. Non-Cryst. Solids 143 (1992) 157. [9] R.H. Magruder IIl, R.A. Weeks, R.A. Zuhr and G. Whichard, J. Non-Cryst. Solids 129 (1991) 46. [10] R.H. Magruder III, R.A. Zuhr and R.A. Weeks, Nucl. Instrum. Methods B59&60 (1991) 1308. [11] G.W. Arnold, IEEE Trans. Nucl. Sci. 20 (1973) 220. [12] H. Hosono and R.A. Weeks, Phys. Rev. B40 (1989) 10543. [13] K. Fukumi, A. Chayahara, J. Hayakawa and M. Satou, J. Non-Cryst. Solids 128 (1991) 126. [14] K. Fukumi, A. Chayahara, M. Satou, J. Hayakawa, M. Hangyo and S. Nakashima, Jpn. J. Appl. Phys. 29 (1990) 905. [15] H. Hosono, J. Appl. Phys. 69 (1991) 8079. [16] P. Polato, P. Mazzoldi and A. Boscolo-Boscoletto, J. Am. Ceram. Soc. 70 (1987) 775. [17] G.W. Arnold, Rad. Eft. 47 (1980) 15. [18] G.W. Arnold and P. Mazzoldi, in: Beam Modification of Materials 2, Ion Beam Modification of Insulators, ed. P. Mazzoldi and G.W. Arnold (Elsevier, Amsterdam, 1987) p. 195. [19] F.L. Galeener, Phys. Rev. B19 (1979) 4292. [20] H.R. Philipp, in: Handbook of Optical Constants of Solids, ed. E.D. Palik (Academic Press, Orlando, FL, 1985) p. 749. [21] A.E. Geissberger and F.L. Galeener, Phys. Rev. B28 (1983) 3266. [22] K.L. Ngai, in: Non-Debye Relaxation in Condensed Matter, ed. T.V. Ramakrishnan and M. Raj Lakashmi (World Scientific, Singapore, 1987) p. 23. [23] S. Brawer, Relaxation in Viscous Liquid and Glasses (American Ceramic Society, Westerville, OH, 1985) p. 43. [24] J. Wong and C.A. Angell, in: Glass Structure by Spectroscopy (Dekker, New York, 1976) p. 707. [25] R. Saito and K. Murayama, Solid State Commun. 63 (1987) 625. [26] G. Hetherington, K. Jack and J. Kennedy, Phys. Chem. Glasses 5 (1964) 130. [27] W.G. Perkins and D.R. Begeal, J. Phys. Chem. 54 (1971) 1683. [28] D.K. McElfresh and D.G. Howitt, J. Am. Ceram. Soc. 69 (1986) C237. [29] W. Primak, E. Edward, D. Keifer and H. Szymanski, Phys. Rev. 113 (1964) 531. [30] F. Seifert, B. Mysen and D. Virgo, Phys. Chem. Glasses 24 (1983) 141. [31] J.D. Mackenzie, J. Am. Ceram. Soc. 46 (1963) 470. [32] J.D. Mackenzie, J. Am. Ceram. Soc. 47 (1964) 76. [33] G.E. Walrafen, Y.C. Chu and M.S. Hokmabadi, J. Chem. Phys. 92 (1990) 6987. [34] J. Arndt, R.A.B. Devine and A.G. Revesz, J. Non-Cryst. Solids 131 (1991) 1206.