Original papers

THE IMPACT OF BUBBLE ARRAYS ON GLASS FLOW

AND REDOX STATE OF GLASS DURING GLASS MELTING
LUBOMÍR NĚMEC, MILOŠ JIŘIČKA, JOSEF MATYÁŠ, ALEXANDER FRANĚK*

Laboratory of Inorganic Materials of the Institute of Inorganic chemistry ASCR and the Institute of Chemical Technology,
Technická 5, 166 28 Prague, Czech Republic
E-mail: [email protected]

*Glass Service, Rokytnice 60, 755 01 Vsetín, Czech Republic

Submitted March 27, 2001; accepted September 7, 2001.

The bubble arrays present in the glass melting spaces may extract gases from the melt and thus decrease its refining ability,
as well as influence glass flow by changing density of the mixture glass melt-bubbles. Both phenomena were examined using
the 3D and 2D models. The examination of the bubble extraction ability on elected glass pathways in the 3D model has shown
that refining gases were evidently removed from the reduced glass melt (amber) while the concentration drop of dissolved
refining gases in an oxidized glass was negligible. The distinct bubble stirring effect in the 2D melting space was observed
when inputting about 3 vol.% of small bubbles (a = 2.5 × 10-4 m) through a part of bottom. The effect of bubble concentration
gradients on the intensity of glass convection was analogical to the impact of temperature gradients on glass flow in case of
temperature convection. Different behavior was observed, however, when examining the impact of geometrical shape on the
melt flow. Both described phenomena should be considered in advanced mathematical models of glass melting spaces.

Keywords: Bubble, Glass flow, Redox state

INTRODUCTION models of processes may be used here but one must be

aware of their limited ability when transferring results
The results of simplified mathematical models of to different conditions.
glass melting show that bubble concentrations in some This work aims at the separate examination of
regions of a glass melting space, namely below the mutual interference between bubble arrays and glass
batch blanket and in the area of maximum temperature, flow, as well as redox state of glass. The partial models
may achieve up to tens of volume percents [1]. Such developed by authors of this article and by their
high bubble densities may influence not only glass co-workers were applied to find the solution of the
properties as viscosity, density or effective heat problems.
conductivity but also the fundamental melting
processes. The bubble arrays are able to evoke the THEORETICAL
macroscopic glass flow as a consequence of bubble
concentration gradients, to extract gases from the melt To study the impact of bubbles on the redox state
into atmosphere and to enhance corrosion of materials of glass, the 3D Glass Model was used [3], making it
in contact with the melt. The contemporary possible to acquire the pure glass melt temperature and
mathematical models of the refining process start from velocity field in a melting furnace under given boundary
the temperature and glass velocity fields valid for the conditions. Both value fields served as a base for the ex
pure melt and ignore therefore the probable mutual post calculation of the redox component distribution
interference of the melt and bubbles [1, 2]. This fact inside the space [4]. The bubble density distribution in
makes the model results partially unreliable and their the melting space was also obtained by applying the ex
subtle applications e.g. for glass furnace control very post calculation procedure too [1, 5]. The mentioned
problematic. The further model improvement leads to procedure used the temperature, velocity and redox
parallel consideration of mentioned effects during component concentration fields of the pure glass melt to
calculations, however, such models appear to be rather calculate the pathways of bubble representatives
complicated for operational applications. The process through the model melting space and to provide
complexity and the need of additional data also subsequently the bubble concentrations in the steady
discourage from applying this way. The alternative state. When examining the bubble influence on the glass
approach offers the separate investigation of various flow, the 2D model was constructed [6] to describe the
effects and subsequent estimation of their importance mutual impact of monodisperse particles and a glass
under real conditions. Only significant phenomena melt under isothermal conditions.
should be then included into the models. The simplified

Ceramics − Silikáty 45 (4) 129-136 (2001) 129

 L. Němec, M. Jiřička, J. Matyáš, A. Franěk

Calculation conditions RESULTS OF CALCULATIONS

In the 3D modeling, the rectangular melting space A typical picture of extraction of the dissolved SO2
with throat was 16.5 m long and 4 m wide. The glass from the flat glass melt containing sodium sulfate as the
melt layer thickness was 1 m. The temperature fining agent along 4 glass trajectories is plotted in figure
maximum 1500 °C was at X = 5.5 - 6.5 m and the throat 1a, the similar dependence valid for the same glass
was 10.5 m located from the front wall. When studying containing sulfides (amber glass) is presented in figure
the bubble and glass flow interference, the 2D model 1b. The concentration drops in figure 1a corresponds to
melting space 1 × 1 m was chosen for most calculations. short residence times of glass in the high temperature
The bubbles entered the space by the central part of the region near the glass surface. The repeated con-
bottom. The glass was the soda-lime-silica glass centration steps express the glass circulation in the
containing 74 wt.% SiO2, 16 wt.% Na2O and 10 wt.% melting region. The concentration of the dissolved SO2
CaO. The glass melt was refined by the sodium sulfate. decreased along all pathways examined, nevertheless,
The glass temperature as well as velocity fields and the maximum concentration drop did not exceed
subsequently, the redox distribution and bubble 3 mol m-3. In the reduced glass, on the contrary, the
distribution in the space were calculated using the above bubble extraction ability is much higher attaining up to
mentioned partial models. The bubble impact on the tens of mols of SO2 per m3 as shown in figure 1b.
redox state of glass was examined along glass melt
pathways through the melting space. The pathways have
started under the glass batch blanket and finished in the
throat. The local values of the bubble number density,
gas concentrations in the melt and in bubbles were used
to calculate the local mass flow of refining gases
between the bubbles present and the melt (see equation
(1) in [5]). The sum of concentration changes of
refining gases along elected pathways gave the
approximate concentration decrease of the refining
agent concentration in the melt owing to bubbles. The
accuracy of results was limited by the fact that the local
mass flow did not take into account the previous effect
of bubbles. The typical values of volume fractions of
bubbles in the melt were obtained also for usage in the
2D model. In the latter 2D model calculations, the bub-
bles of constant radii 2.5 × 10-4, 5 × 10-4 and 1 × 10-3 m
and in volume concentrations 0.2, 1, 3, 10, 20 and
30 vol.%, respectively, were used to calculate the a)
resulting bubble and glass velocity distribution in the
space. Bubbles entered the central part of the bottom
space and the fraction of bottom occupied by inputting
bubbles moved from 0.02 to 1, but in most cases was
fixed at 0.02.
The following calculations were executed to
elucidate the outlined problems:

- the decrease of the refining agent concentration

along forty typical pathways of the melt through
the 3D melting space
- the dependence between the characteristic velocity
of the melt flow in the 2D space and:
- the initial volume concentration of bubbles in
the melt, ϑo
- the fraction of bottom occupied by inputting
bubbles, X/Xo b)
- the bubble radius, a
- the shape of the space, Yo/Xo Figure 1. The concentration drop of the dissolved SO2 along
- the density difference of bubbles and melt, four typical glass pathways through the melting region. The
ρgl - ρbub concentration drop was brought about by the extraction effect
- the kinematic glass viscosity, υ of bubbles.
a) oxidized glass containing sodium sulfate, b) amber glass.
1 - 4: bubble trajectories.

130 Ceramics − Silikáty 45 (4) 129-136 (2001)

 The impact of bubble arrays on glass flow and redox state of glass during glass melting

The values of the glass volume occupied by probable. The appropriate dependencies are plotted in
bubbles in the 3D model may give an approximate figure 2.
information about bubble concentration gradients in The picture of a typical glass velocity distribution
the space. The concentration differences are the in the 2D space is presented in figure 3a, b. The bubbles
driving force of convection due to bubbles and help us entered the central part of the bottom, the fraction of the
to provide the 2D model with the appropriate input bottom occupied by inputted bubbles was X/Xo where Xo
values. The independent calculations of glass flow and was the space width. As is obvious from both figures,
bubble density distribution in the 3D model, however, the glass with bubbles rose through a relatively narrow
make the results only qualitative. The example of central vertical column to the glass level and then
the bubble distribution on two transversal straight lines reversed back to the bottom with a much lower velocity.
through the space show nevertheless both high The fraction of the bottom occupied by bubbles only
bubble concentrations and bubble concentration slightly influenced the character of the flow as the
gradients, particularly in the region of the temperature bottom backflow “swept” the inputting bubbles into the
maximum close to the glass surface. The impact of narrow rising column of glass at higher values of X/Xo
bubbles on glass flow appears therefore highly (see figure 3b). The characteristic glass velocity and the

a) a)

b) b)

Figure 2. The dependence between the volume concentration of Figure 3. The pictures of two typical glass velocity distribu-
bubbles, ϑ, and Y coordinate in the tranversal sections through tions in the 2D melting space 1 × 1 m. The length of arrows cor-
the 3D melting space; a) X = 1.5 m (under the batch blanket), responds to the distance run in the course of 10 s. At 1300 °C,
1: 0.2 m under the boundary glass batch - glass melt, 2: 0.5 m the initial bubble volume fraction, ϑo = 3 vol.%, the bubble
under the boundary glass batch - glass melt; b) X = 6.5 m (in the radius a = 5 × 10-4 m, τ = 50 hrs; a) the fraction of the bottom
temperature maximum), 1: 0.2 m under the glass level, 2: 0.5 m occupied by inputting bubbles, X/Xo = 0.02; b) the fraction of
under the glass level. the bottom occupied by inputting bubbles, X/Xo = 0.5.

Ceramics − Silikáty 45 (4) 129-136 (2001) 131

 L. Němec, M. Jiřička, J. Matyáš, A. Franěk

examined, the velocity grew to the maximum value

reached after approximately 10 - 20 hours and sub-
sequently slowly decreased approaching the steady
value. However, the steady value was not attained even
after 100 hours and similar character of the de-
pendencies between the characteristic glass velocity and
time, and the characteristic concentration gradient and
time, indicated the approximate proportionality between
both characteristic values as seen from figure 6.
The influence of the initial bubble volume
concentration on the average characteristic glass
velocity and on the average bubble concentration
gradient (5 - 50 hours) is demonstrated in figure 7. The
initial steep increase of both characteristic values slows
down at higher bubble concentrations and in case of the
Figure 4. The course of the glass velocity and bubble concentration gradient achieves the almost constant
concentration in a typical case on the line Y = 0.5 m and after value at ϑo › 20 vol.%, i.e. the driving force of the glass
50 hrs, at 1300 °C, ϑo = 3 vol.%, X/Xo = 0.02, a = 5 × 10-4 m. flow does not grow at higher bubble concentrations. The
The values of v and dϑ/dX in the central part of the curve define
the characteristic glass velocity and the characteristic bubble
concentration gradient, respectively.
! - velocity, ! - volume concentration of bubbles

Figure 6. The time development of the characteristic glass

velocity and the characteristic bubble concentration gradient at
1300 °C, ϑo = 3 vol.%, X/Xo = 0.02, a = 5 × 10-4 m.
Figure 5. The time development of the characteristic glass # - velocity, ! - concentration gradient
velocity for the different values of the initial bubble volume
fraction, ϑo, at 1300 oC, X/Xo = 0.02, a = 5 × 10-4 m.
! - ϑo = 0.2 vol.%, " - ϑo = 1 vol.%, ! - ϑo = 3 vol.%,

* - ϑo = 10 vol.%, × - ϑo = 20 vol.%, # - ϑo = 30 vol.%

characteristic bubble concentration gradient were found
to evaluate the results of calculations. The average glass
velocity in the central rising glass column and at Y = 0.5
m was determined as the characteristic glass velocity.
This is apparent from figure 4, where the velocity
distribution is plotted on the line Y = 0.5 m. The charac-
teristic bubble concentration gradient was read from the
steepest central part of bubble concentration curve at
identical Y coordinate (see also figure 4). The almost
constant concentration of bubbles close to the borders
of modeled space indicates the considerable stirring
ability of bubbles. This also illustrates the increase of Figure 7. The dependence of the average characteristic glass
the average bubble concentration with time (see below). velocity and the average bubble concentration gradient
The development of the characteristic glass (5 - 50 hrs) on ϑo, at 1300 °C, X/Xo = 0.02, a = 5 × 10-4 m.
velocity with time is shown in figure 5. In all cases - velocity, ! - dϑ / dX

132 Ceramics − Silikáty 45 (4) 129-136 (2001)

 The impact of bubble arrays on glass flow and redox state of glass during glass melting

a = 5 × 10-4 m, then the values −

v decrease. The reason of
this decline is not clear but the values of the
characteristic bubble concentration gradient of higher a
values probably do not represent the driving force of
glass flow reliably.
Two relations between the characteristic glass
velocity and the space shape were examined, namely
the dependence on the space height, Yo, for the constant
Xo = 1 m, and on the Yo/Xo ratio for the constant space
volume (XoYo = 1 m2). Both dependencies are plotted in
figure 10. To complete the parametrical study, the effect
of the density difference between glass and bubbles,

Figure 8. The dependence of the average characteristic glass

velocity on the part of bottom occupied by bubbles, X/Xo, at
1300 oC, ϑo = 3 vol.%, a = 5 × 10-4 m, τ = 100 hrs.

Figure 10. The dependence of the average characteristic

glass velocity on the geometrical shape of the space, at
1300 °C, ϑo = 3 vol.%, a = 5 × 10-4 m.
! - the dependence on the Yo at constant Xo = 1 m, " - the
dependence on the Yo/Xo at constant XoYo = 1 m2.

Figure 9. The dependence of the average characteristic glass

velocity and the recalculated characteristic glass velocity
(5 - 50 hrs) on the bubble radius, a, at 1300 °C, ϑo = 3 vol.%,
X/Xo = 0.02.

dependence between the average characteristic glass

velocity and the fraction of the bottom occupied by
inputting bubbles (figure 8) shows, on the contrary, the
maximum at about X/Xo = 0.5. The initial velocity
growth is brought about by the increasing amount of
inputting bubbles, the velocity fall at high occupation
values is caused by the changing character of glass flow
(disintegration of circulation flows).
Figure 9 presents the dependence of the average
characteristic glass velocity on the bubble radius, a.
Figure 9 shows also the values of the velocity, Figure 11. The dependence of the average characteristic glass
recalculated to the standard bubble concentration velocity on the density difference, ρgl – ρbub, and the reciprocal
gradient at a = 5 × 10-4 m. The anticipated independence value of kinematical viscosity, 1/υ. # - the dependence on the
of −v on the recalculated standard value of bubble density difference, $ - the dependence on the reciprocal value
concentration gradient is preserved up to about of glass kinematical viscosity.

Ceramics − Silikáty 45 (4) 129-136 (2001) 133

 L. Němec, M. Jiřička, J. Matyáš, A. Franěk

ρgl - ρbub, and the reciprocal value of kinematical therefore influence the resulting redox state of glass
viscosity of glass, 1/υ, on the average characteristic only very slightly. The reciprocal temperature de-
glass velocity are plotted in figure 11. pendence of sulfides stability, on the contrary, shifts the
production of SO2 and consequently, the SO2 extraction
DISCUSSION by bubbles, towards lower temperatures. That is why
the bubble extraction ability along predominantly low
The temperature stability of sulfates and sulfides in temperature pathways is high. The concentration drop
the glass is responsible for the different bubble distribution for 40 trajectories examined exhibits the
extraction ability in the oxidized glass melts with the maximum frequency at ∆c ∈ 〈40;45〉 (mol m-3) (see
sodium sulfate and in the reduced glass containing figure 12b). As the concentration of the refining agent
sulfides. As sulfate ions decompose apparently only in the melt is of the same order of magnitude as ∆c, the
above 1400 °C, the small part of glass trajectories is distinct influence of bubbles on the redox state of glass
exposed to gas diffusion into extracting bubbles and in reduced glasses containing sulfides can be expected.
consequently, the concentration decrease of sulfates is The glass flow intensity and bubble concentration
low. Figure 12 a presenting the concentration difference distribution acquired from the 2D model show that even
distribution for the examined 40 trajectories shows the a small amount of bubbles entering the space stirs the
maximum frequency at ∆c ∈ 〈0;0.3〉 (mol m-3). The melt vigorously (see figures 3 and 4). It appears to be
bubble clouds in glasses containing sulfates should useful to compare the stirring effect of bubbles with an
ideal mixer in the same 2D space.
The bubble concentration change brought about by
the inputting bubbles is given by:

X
dϑ+ = uϑo dτ (1)
Xo

where u is the bubble rising velocity. The bubble

concentration decrease by bubbles leaving the space
through the level describes the equation:
dϑ− = uϑ dτ (2)

As dϑ = dϑ+ + dϑ−, the resulting bubble concentration

change can be expressed by the equation:

X
a) ϑ+ = ϑo [1 - exp(-uτ)] (3)
Xo

Equation (3) shows that under condition of ideal

mixing, the increase in the relative bubble concentration
depends only from the bubble rising velocity (i.e. from
the bubble radius). The final bubble concentration in the
space attains the value X/Xoϑo (τ → ∞). The almost ideal
mixing of the melt is confirmed by the fact that courses
of equation (3) for two values of bubble rising velocities
almost coincide with the bubble concentration increase,
calculated from the 2D model. The results of this
comparison are presented in figure 13.
The characteristic concentration gradient of
bubbles is considered the driving force of glass
convection flow. Figure 14 illustrates the linear
dependence between the two average characteristic
values. The character of this dependence is similar to
b) the relation between the horizontal glass velocity and
temperature gradient in a rectangular space, presented
Figure 12. The concentration difference distribution for 40 by Cooper [7]:
examined glass trajectories in the glass melt containing sulfate
ions or in the amber glass containing sulfide ions; αg dT
a) soda-lime-silica glass containing sulfate ions, b) amber glass vx = Yo3 f(Y) (4)
containing sulfide ions. υ dX

134 Ceramics − Silikáty 45 (4) 129-136 (2001)

 The impact of bubble arrays on glass flow and redox state of glass during glass melting

velocity and the reciprocal value of the kinematic

viscosity. Both dependences are presented in figure 15.
Different behavior is, however, apparent from figure 16
where the approximately linear dependence between the
characteristic glass velocity, recalculated to the standard
concentration gradient (ϑo = 3 vol.%, dϑ/dX = 0.165),
and Yo1/2 (Xo = 1 m), respectively (Yo/Xo)1/4 (XoYo = 1 m2)
can be found. The cubic character of the dependence
between the glass velocity and the height of glass layer

Figure 13. The increase of the average bubble concentration, ϑ,

in the space as a function of time at 1300 °C, ϑo = 3 vol.%,
X/Xo = 0.02.
- - - - - c (final value); ! - v = 1.91 × 10-5 m s-1, a = 5 × 10-4 m;
. . . . v = 1.91 × 10-5 m s-1, equation (3); # - v = 4.78 × 10-6 m s-1,
a = 2.5 × 10-4 m; – – – – v = 4.78 × 10-6 m s-1, equation (3)

Figure 15. The dependence between the average characteristic

glass velocity, recalculated to the standard concentration
gradient (ϑo = 3 vol.%, dϑ/dX = 0.165), and the glass and gas
density difference, as well as the reciprocal value of the
kinematical viscosity of glass. # - 1/υ, $ - ρgl - ρbubl

Figure 14. The dependence between the average characteristic

glass velocity and the average bubble concentration gradient at
1300 °C, X/Xo = 0.02, a = 5 × 10-4 m.

where α is the volume thermal expansion coefficient of

glass, T is temperature, Yo is the height of the glass
layer, υ is the kinematic viscosity of glass and f(Y) is a
simple function of the vertical coordinate. The
corresponding values of the temperature gradients,
dT/dX are also presented in figure 14, showing that
driving forces caused by bubbles in the 2D model
correspond approximately to those in commercial glass
melting furnaces. This fact indicates that temperature
and concentration convection have similar impact on
the glass flow in glass melting furnaces.
The similarity between temperature and concentra-
tion convections is further illustrated by linear
dependence between the average characteristic glass
velocity, recalculated to the standard concentration Figure 16. The dependence between the average characteristic
gradient (ϑo = 3 vol.%, dϑ/dX = 0.165), and the glass glass velocity, recalculated to the standard concentration
and gas density difference (which corresponds to α in gradient (ϑo = 3 vol.%, dϑ/dX = 0.165), and the value Yo1/2
equation (4)), as well as between the characteristic glass ($ - Xo = 1 m), as well as (Yo/Xo)1/4 (# - XoYo = 1 m2).

Ceramics − Silikáty 45 (4) 129-136 (2001) 135

 L. Němec, M. Jiřička, J. Matyáš, A. Franěk

is, on the contrary, characteristic for the temperature Acknowledgement

convection (see equation (4)).
Taking into account the revealed relations between This work was supplied with the subvention by Grant
the characteristic glass velocity and examined initial Agency of AS CR, Project No. S4032103.
and boundary parameters, the following semiempirical
relation, based on results of presented 2D model may be References
written for the given 2D space with constant value of Xo:
1. Matyáš J., Němec L., Jiřička M., Franěk A.: Proc. Int.
Conference Glass in the New Millenium: Challenges and
−v = const ∆ρg dϑ
Yo1/2 (5) Break-through Technologies, Amsterdam, May 15-19,
υ dX 2000.
2. Němec L., Kloužek J., Schill P., Ullrich J.: Proc. Int.
In agreement with temperature convection, the Conference on Fundamentals of Glass Science and
intensity of glass flow is directly proportional to the Technology, p. 340, Växjö, June 9 - 12, 1997.
density differences and horizontal convection gradient, 3. Schill, P.: Proc. 2nd International Seminar on
and indirectly proportional to the glass viscosity. A dis- Mathematical Simulations in the Glass Melting, p.102,
similarity of both cases was found when calculating the Vsetín, Czech Republic, May 15 - 19, 1993.
significance of the height of glass layer, possibly the 4. Němec L., Raková M.: Ceramics - Silikáty 42, 1 (1998).
impact of geometrical shape on the glass flow. The 5. Kloužek J., Matyáš J., Němec L., Trochta M., Ullrich J.:
significance of the height of glass layer was found much Proc. 5th ESG Conference Glass Science and Technology
lower for the concentration convection being expressed for the 21st Century, p.A2, Prague, Czech Republic, June
by a proportionality between the characteristic glass 19 - 24, 1999.
velocity and the square root of the height (see equation 6. Kloužek J., Franěk A.: Ceramics-Silikáty 45, 70 (2001).
(5)), whereas the same dependence for the temperature 7. Cooper A. R.: A Collected Papers of the XIV Int.
convection was cubic (see equation (4)). One of reasons Congress on Glass, p.1, New Delhi 1986.
of this behavior may be differences in nature of both
phenomena: the space walls exhibit heat losses but no
particles could penetrate through the walls. A relatively
complicated relation was found between mixture flow VLIV SOUBORŮ BUBLIN NA PROUDĚNÍ SKELNÉ
and the portion of the bottom occupied by inputting TAVENINY A JEJÍ OXIDAČNĚ REDUKČNÍ STAV BĚHEM
bubbles (see figure 8). It is apparent from the nature of TAVICÍHO PROCESU
the model that the concentration convection should be
absent when X/Xo = 0 and should have only low LUBOMÍR NĚMEC, MILOŠ JIŘIČKA, JOSEF MATYÁŠ,
intensity when X/Xo = 1, i.e. when the flow is evoked ALEXANDER FRANĚK*
only by the requirement of flow continuity. Figure 8
confirms this assumption, however, no simple relation Laboratoř anorganických materiálů, společné pracoviště
between the average characteristic glass velocity and AV ČR a VŠCHT v Praze, Technická 5, 166 28 Praha
parameter X/Xo was revealed.
*Glass Service, Rokytnice 60, 755 01 Vsetín
CONCLUSION
Soubory bublin přítomné v tavicím prostoru během
Two phenomena seem to influence the results of tavicího procesu skel extrahují a vynášejí k hladině čeřící plyny
contemporary mathematical models of glass melting rozpuštěné v tavenině a snižují schopnost taveniny odstraňovat
spaces: refining gas extraction from the melt by rising bubliny. Snižují rovněž průměrnou hodnotu hustoty směsi
bubbles and glass melt flow, caused by horizontal tavenina – bubliny a mohou vyvolat proudění skloviny. Oba
bubble concentration gradients. Refining gases may be jevy byly zkoumány s použitím 2D a 3D modelů. Sledování
extracted by bubbles in the regions of high bubble extrakce plynů bublinami na vybraných trajektoriích skelné
densities, this influence being significant especially for taveniny v 3D tavicím prostoru ukázaly, že rozpuštěné plyny
refining agents working at lower and medium tempera- byly zřetelně extrahovány bublinami v redukovaném skle
tures (SO2 extracted by bubbles from amber glass). The obsahujícím sirníkové ionty, zatímco stejný efekt u skla
horizontal bubble concentrations gradients, present in obsahujícího síranové ionty byl zanedbatelný. S použitím 2D
commercial glass furnaces, evoke the glass flow having modelu byl prokázán výrazný efekt malého množství bublin
at least the same intensity as flow caused by tempera- (3 obj.% malých bublin vstupujících částí dna prostoru) na
ture convection. Analogy between temperature convec- proudění v izotermním prostoru. Byla zjištěna analogie mezi
tion and concentration convection in the 2D melting vlivem horizontálních teplotních a koncentračních gradientů na
space was found as for the effect of bubble concentra- přirozenou konvekci skelné taveniny. Vliv výšky tavicího
tion and temperature gradients. Further confirmation of prostoru na intenzitu proudění taveniny byl podstatně méně
the latter bubble effect by using 3D model is needed. výrazný v případě koncentrační konvekce. Oba vlivy bublin na
Both phenomena should be considered in the appro- chování taveniny je třeba uvažovat při matematickém
priate mathematical models of glass melting spaces. modelování tavicích částí sklářských pecí.

136 Ceramics − Silikáty 45 (4) 129-136 (2001)