Thermal expansion of vitreous silica: Correspondence between dilatation

curve and phase transitions in crystalline silica

Yoshikazu Kikuchi and Hajime Sudo

Nippon Silica Glass Co. Ltd. 2-4-3 Kyobashi, Chuo-Ku, Tokyo 104, Japan

Nobu Kuzuu a )

Department of Applied Physics, Fukui University, 3-9-1 Bunkyo. Fukui 910, Japan

(Received 10 October 1996; accepted for publication 9 July 1997)

The linear thermal expansion coefficients of fused quartz containing 3 ppm of OH produced by
electrical melting of natural quartz powder were measured using an interferometric dilatometer in
the temperature range 300-970 K. The data obtained show several maxima and minima, some of
which correspond to the phase transitions in crystalline silica. A sharp maximum is observed at 500
K, which corresponds to the Q' to f3 transition temperature of cristobalite. Minima at 850 and 580 K
correspond to low- to high-temperature phase transitions of crystalline quartz and tridymite,
respectively. This correspondence was not observed in synthetic fused silica containing about 800
ppm of OH. © 1997 American Institute of Physics. [S0021-8979(97)00220-X]

Since vitreous silica is a low-thermal-expansion mate­
rial, its shape and size remain constant with changing tem­
perature, unlike most glasses and inorganic materials. There­
fore, it is used as a standard material, e.g., standard flats for
measurement, photomasks for largescale integration (LSI)
manufacturing processes, and thin-film-transistor (TFT) sub­
strates. To measure the dilatation of vitreous silica, precise
measurement at extremely low thermal expansions is re­
quired. The linear thermal expansion coefficient of vitreous
silica had been reported as early as 1934. I It has also been
reported2 that the nature of the dilatation of vitreous silica is
strongly dependent on the type of vitreous silica and its' 'fic­
tive temperature." i.e., the temperature at which the equilib­
rium structure is frozen in. It is known that the linear thermal
expansion coefficient shows a negative value in the low­
temperature range 2- 5 of about 140-210 K. These observa­
tions suggest that the dilatation of vitreous silica reflects the
structure of the Si02 glass network.
Recent study on the thermal expansion of vitreous silica
has been restricted to synthetic fused silica. 6 However, a
comparison between fused quartz produced by melting natu­
ral quartz powder and synthetic fused silica will provide use­
ful information on the glass structure because there should be
a structural difference between these two types of vitreous
silica; the former has optically observable "grain
structure," 2 but the latter does not have such a structure, at
least, not an optically observable one. However, the grain
structure does not directly reflect the trace of the interface of
fusing the quartz particles, because the optically observable
"grain" size (=0.3-0.5 mm) is much larger than the diam­
eter of the quartz powder (= 2 - 3 ,um).
In this work, we compared the thermal expansion of
these two kinds of vitreous silica produced by different
methods; one is fused quartz 2 produced by electrical melting
of natural quartz powder (type-I silica glass), and the other is
synthetic fused silica produced by hydrogen-oxygen flame
hydrolysis of SiCl4 (type-III silica glass).
Two types of vitreous silica were used: type-I fused
quartz produced by melting natural quartz powder, and type­
III fused silica synthesized directly by flame hydrolysis of
SiCl4 in hydrogen-oxygen flame. These samples are com­
mercially available HR (Type I) and ES (Type III) grades of
silica glasses from Nippon Silica Glass Co. Ltd} containing
about 3 and 800 ppm of OH, respectively. These samples
were fabricated into cylinders (7 X 14 mm) ground to a
spherical shape at each end. A Shinku-Riko LIX laser inter­
ferometric dilatometer was used. The thermal expansion was
measured at temperatures between 298 and 973 K, raising
the temperature at a rate of 3 Klmin.
The results are shown in Fig. 1. The type-I silica shows
a sharp peak at 500 K, and minima at 580 and 850 K. In the
temperature range of 650-750 K, several small peaks and
valleys also appeared.
Data 8 for NIST's standard reference material SRM739
are also plotted and the interpolated curve 8 is drawn for com­
parison. Although, the data for type-III silica show consider­
able dispersion. they are in good agreement with those for
SRM739. These data do not show as complex a curve as that
for type-I silica glass. In the previous studies, no such peaks
for type-lor type-II fused quartz were observed,2 perhaps
because of the low resolution of the measurement.
The temperature dependence of the linear thermal ex­
pansion of SMR739 and type-III fused silica are essentially
the same; the data collected in this study are scattered ran­
domly about the line fitted to the data for SMR739 with an
uncertaintl of ::i:: 0.3 X 10- 7 K- 1. Since SMR739 is a type­
III fused silica,8 this confirms the reliability of the present
measurement. The dispersion of the data for the present mea­
surement is of the same order of magnitude as the recent
measurement by Okaji and Birch;6 the uncertainty of mea­
surement for SMR739 is about ::i::0.2X 10- 6 K-'. For type-I
silica. the scatter of the data seems to be smaller than that for
type-Ill silica. A possible explanation is that the viscosity of
type-III silica is considerably lower than that of type-I silica;
for example, the viscosity of type-III silica is about two or­
ders of magnitude lower than that of type-I silica9 at
~ 1000 C. The lower viscosity of the sample could explain
0

the scatter of the data, because the sample will deform during
a)Electronic mail: kuzuu@polymer.apphyJukui-u.ac.jp the measurement process.

J. Appl. Phys. 82 (8), 15 October 1997 0021-8979/97/82(8)/4121/3/$10.00 © 1997 American Institute of Physics 4121
 Iii
~
8 present result may provide a supplemental insight into the
co
(,
-:::
-+
"/.
;< , sized from a liquid, SiCI 4, no trace of the crystalline struc­
- ture remains. The other explanation is that there is some
II
200 400 600 800 1000
Temperature / K
FIG I Linear thermal expansion curve of two types of vitreous silica,
t\pe-I (- 1 and t\pe-Ill (L). Data for SRM739 are also plotted (e and solid
(un e I. \crlJ('al lines indicate the phase transition points of silica crystals:
solid line 1 CI, cmtobalite: dashed line (Qi, quartz: and dotted line (T l ,
T:- Te, T 1 and T,I: trydimites,
Type-I fused quartz exhibits a different temperature de­
pendence than does type-III fused silica. Since the type-I
silica is formed by melting crystalline quartz powder. some
trace of crystalline structure might remain.
The crvstalline forms of silica which are stable at room
temperature. are quartz, 10 . II cnsto
' b a I'!te, P- an d tn'd ymlte.
. 1011
.­
Crystalline quartz exhibits an a to (3 phase transition at 846
K. ]0.1] This temperature corresponds to a minimum in the
expansion curve of type-I silica (indicated by the dashed line
Q in Fig. I). Cristobalite exhibits a low- to high-temperature
phase transition at 493 K.] 2 This temperature corresponds to
the maximum of the dilatation curve of type-I silica (solid
line C). The phase transition of tridymite is more
complicated; 10.13 several phases, named MX, PO-IO, OP,
OS, Oc, and HP, are observed in tridymite. The OC to HP
phase transition occurs gradually between 583 and 653
K. 12-1--1 The lowest limit of this range, 583 K (dotted line T 4 ),
corresponds to the minimum value on the thermal expansion
curve, and the highest limit of this range, 653 K (T s ), cor­
responds to a small maximum. The other transition points in
tridymite, OS to OC at 464 K (T 3), OP to OS at 423 K (To),
and MX to OP transition at 383 K (T]),]O do not correspo'nd
to either maxima or minima of the dilatation curve. Other
weak peaks and valleys also appeared in the dilatation curve.
They might cOITespond to phase transition points of some
type of crystalline silica, but their origin is not known at the
present stage. At any rate, these findings suggest that some
trace of the crystalline structure that is stable at room tem­
perature could remain in the fused quartz.
For the structure of vitreous silica, Zachariasen] s pro­
posed a random network model as early as 1932, supported
by x-ray diffraction data. 16 However, some alternate models
for the structure of vitreous silica have been proposed, in
which very small crystalline regions are connected by disor­
dered regions. 17 Goodman]S proposed a strained mixed clus­
ter model. This model suggests that vitreous silica forms as a
random network of aggregated but strained clusters based on
cristobalite, trydimite, and quartz with dimensions of a few
nanometers; the other phases are assumed not to exist be­
cause they are unstable at room temperature in atmospheric
4122 J. Appl. Phys., Vol. 82, No.8, 15 October 1997
pressure. These structures exhibit short-range order which is
difficult to detect by diffraction methods. Therefore, the
structure of vitreous silica.
In type-III fused silica, such a structure is not observed.
There are two possible explanations for this. One is related to
the starting material; since this type of fused silica is synthe­
trace of crystalline structure or short-range order that contrib­
utes the phase transitions, but a considerable amount of OH
(= 800 ppm) in the type-III silica could reduce the stress in
the glass network. Such a "soft" glass network could absorb
the abrupt change of the short-range order at the phase tran­
sition point. Material with a large amount of OH has a
shorter relaxation time for structural change. Such relaxation
is also observed in the "defect line" in the Raman spectra;
the relaxation time of a middle-range order structure is in­
creased by increasing the OH content. 19 ,20
The absolute value of the thermal expansion coefficient
of type-! silica is higher than that of type-III silica at tem- ~
peratures higher than about 400 K. This can also be inter­
preted in terms of the effect of OH. The silica glass network
with high OH content should be softer than that with lower
OH content and more easily deformed. Thus, the effect of
lateral vibration of the glass network21 reduces the thermal
expansion. as in the case of the negative thermal expansion
of (3-quartz. This effect \\ ould be stronger in silica with high
OH content. because the ~ofter the glass network, the higher
the amplitude of the lateral \ibration of the glass network.
The effect of OH on the softness of the glass network is
supported by the fact that the elastic modulus of the higher
OH content (= 1200 ppm) SIlica is several percent higher
than that of lower OH content ( = 40 ppm) silica. 22 This ef­
fect is also supported by the fact that higher OH content
silica has lower viscosity at high temperature. 9 ,2.1 Another
example of the effect of OH on the "softness" of the glass
network is evident in the resistance to excimer laser
irradiation. 24 - 27 The higher the OH content, the lower the
excimer laser induced absorption. High OH content material __
is resistant to bond breakage by excimer laser irradation be­
cause OH reduces the amount of strained Si-O-Si bonds.
Even if the bond breaks. it can easily rebind after stopping
the irradiation because of the "softness" of the silica glass
network. Yet another example is the OH content dependence
of fractoluminescence. 28 The higher the OH content, the
stronger the intensity of the 2.7 e V emission band observed
from an oxygen-deficient center (ODC), which might be due
to the brittleness of the glass network. It is noted that the
scatter of the expansion data increases with increasing tem­
perature, as seen in Fig. I. This may be an effect of "soft­
ness" on the glass network.
To determine the effect of OH, the dilatation in OH-free
synthetic fused silica, which is produced from liquid mate­
rial, must be measured. This fused silica can be produced by
the soot remelting method such as vapor-phase axial deposi­
tion (V AD).2129 It should be noted that some OH-free soot­
remelted silica would contain a considerable amount of CI,
which would have a similar effect to OH on the softness of
Kikuchi et al.
the glass network 24 ,26 Consequently, a fused silica contain­
ing no OH and no Cl 23 should be used. Further study on such
materials is desirable,
Linear thermal expansion coefficients of a fused quartz
containing 3 ppm of OH (type I) and synthetic fused silica
containing 800 ppm of OH (type III) were measured using an
interferometric dilatometer in the temperature range 300­
970K.
The dilatation curve of type-I silica has several peaks
and valleys which correspond to the phase transition tem­
peratures of crystalline silica" i.e., quartz (853 K), cristo­
balite (493 K), and tridymite (580 K). According to these
observations, type-I fused quartz may have some traces of
the short-range order of the crystalline structure.
In type-III fused silica, on the other hand, the peaks cor­
responding to the phase transitions in crystalline silica were
not observed. There are two possible interpretations for this
observation:
(i) since this silica is synthesized from liquid material,
SiCI 4 , there may be no remaining trace of the short-range
order of the crystalline structure;
(ii) even if the short-range order of the crystalline structure
exists in type-III silica, the abrupt change of the short-range
order at the "phase transition" point could be absorbed
and/or relaxed.
To confirm these hypotheses, further study of the thermal
expansion using OH- and Cl-free synthetic silica glass is
required.
J. Appl. Phys., Vol. 82, No.8, 15 October 1997
I W. H. Keeson and D. W. Doborzynski. Physica (Amsterdam) 1. 1058
(1934).
'R. Bruckner, J. Non-Cryst. Solids 5, 123 (1970).
3R. Briickner, Glastech. Bef. 37, 500 (1946).
4R. Bruckner, Glastech. Bef. 37, 495 (1946),
sR. Bruckner, Glastech. Bef. 38, 153 (1946).
6M. Okaji and D. W Birch, Metrologia 28,27 (1991),
7 Production Catalogue, Nippon Silica Glass Co. Ltd.
8J. P. Cali. Standard Reference Material 739: Fused Silica Thermal Ex­
pansion, Certificate Analysis (National Bureau of Standards, Gaithersburg,
MD. 1971).
9G, Hetherington and K. H. Jack, Phys, Chern. Glasses 5, 130 (1964).
10 A. Nukui. Taikabutsu, 44. 596 (1992) (in Japanese),
II A, Nukui. Ceramics 20, 266 (1985).
12 F. E, Wagstaff, Phys. Chern. Glasses 10,50 (1969).
13 A. Nukui. Am. Mineral. 63, 1252 (1978).
14A. B, Thompson and M. Wennemer, Am, Mineral. 64,1018 (1979),
15W. H. Zachariasen, J. Am. Chern. Soc. 54, 3841 (1932).
16B. E, Warren and 1. Biscoe, J. Am. Ceram. Soc. 21,49 (1938),
17R. H. Doremus, Class Science (Wiley, New York, 1973).
18e. H. L. Goodman, Phys, Chern, Glasses 26. 1 (1985).
19 A. E. Geisseberger and F. L. Galeener, Phys, Rev. B 28, 3266 (1983).
,oF. L. Galeener, J. Non-Cryst. Solids 7, 373 (1985).
211. Yasui and H, Kawazoe, Kokinousei Carasu (High Functional Class)
(University of Tokyo Press. Tokyo. 1985) (in Japanese).
22 H. Kobayasi, 1'. Kosugi. and Y. Kogure, Physica B 219&220, 276 (1996).
23 N. Kuzuu, Sekiei Carasu no Sekai (Introduction to Silica Class Science
and Technology) (Kogyo Chosakai, Tokyo 1995) (in Japanese).
24N. Kuzuu, Proc. SPIE 2714.41 (1996).
25 N. Kuzuu, Proc. SPIE 2714, 71 (1996).
26N. Kuzuu, 1'. Taga, and N. Kamisugi, J. Appl. Phys. 80, 8011 (1997).
,7 N. Kuzuu, Laser Kenkyu (Rev. Laser Eng.) 23, 396 (1995) (in Japanese).
28y, Kawaguchi, Phys, Rev, B 54, 9721 (1996).
29T, Izawa, S. Kobayashi, S, Sudo, and F. Hanawa, Trans, Inst. Electron.
Commun. Eng, Jpn. E62, 779 (J 979).
Kikuchi et ai, 4123