Home Search Collections Journals About Contact us My IOPscience

The float glass process

This content has been downloaded from IOPscience. Please scroll down to see the full text.

1970 Phys. Bull. 21 153

([Link]

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: [Link]
This content was downloaded on 17/07/2016 at 07:07

Please note that terms and conditions apply.

The float glass process Atmosphere

Molten glass

H CHARNOCK, Pilkington Brothers Ltd

PI
Research and Development Laboratories, Ormskirk, Lanes.
Figure 1 Idealized distortion in the top surface of the glass
The development of the float process for the manufacture
of a continuous ribbon of flat glass has provided a rich
variety of novel technological and scientific problems for
the engineers and scientists involved. The context and 1050 °C is 6-5 compared with ~ 2-4 for window glass, it is
main features of this development have recently been molten over the temperature range 232 to 2623 °C, and it
described by Sir Lionel Pilkington (1969) in his lecture to shows little chemical interaction with window glass.
the Royal Society with some reference to the physics of the The special physical properties of the glass are also
process. The intention in this article is to illustrate, in exploited in using this concept as the basis of a continuous
rather more detail, the application of classical physics to forming process. Since the viscosity of glass increases
the understanding and development of this ribbon forming continuously as temperature decreases, the viscosity level
process. appropriate to any desired operation can be chosen con-
The commercial requirement is to produce a flat ribbon veniently by varying the temperature. This 'working
of glass of uniform thickness and substantially free of range' of viscosity or temperature is common to all glass
discrete faults or variations in optical thickness which can forming processes.
distort or obscure an object viewed through the glass.
For most applications, the desired ribbon thickness is REMOVAL OF SURFACE DISTORTION
between 1-5 and 15 mm and the ribbon width preferably The early experiments to establish the feasibility of the
greater than 1-5 m for economical usage. These two main basic idea were carried out using a ribbon of glass formed
physical features of a ribbon, its dimensions and its optical in the conventional way by a pair of water cooled steel
quality, are often found to be interrelated because of their rollers. This rough cast ribbon, so described because of
common dependence on the forces operative during the surface irregularities produced by contact with the rollers,
forming process. It is convenient to view this forming was then supported on a bath of clean molten tin over
process as the controlled viscous deformation of a stream which it was continuously drawn and subjected to a
of molten glass. Hence the important physical parameters heating and cooling schedule. The object of the heating
to be considered initially are the system of forces available process was to reduce the viscosity of the ribbon to a value
and the viscosity of the glass. With this background the sufficiently low so that surface irregularities and thickness
development of the process to meet the commercial require- variations would be removed as the molten glass floated on
ments can then be traced. the molten metal. The subsequent cooling process then
increased the viscosity of the glass to a value high enough
BASIC CONCEPT to prevent surface damage as the ribbon was removed
In the absence of fluid motion and of any extraneous from the bath and transported further.
forces of electrical or chemical origin, the upper and lower An idealized distortion of the top surface of the glass is
surfaces of a layer of liquid supported by a liquid of higher illustrated in figure 1, which is a lateral cross section of
density are approximately flat and parallel, provided the part of the ribbon, the corrugations being assumed to
layer is sufficiently extensive and we consider positions extend indefinitely in the longitudinal direction. With axes
remote from any boundary. In fact the surfaces of the of coordinates and origin O as shown, the distortion is
lower density liquid are surfaces of equal gravitational assumed to take the form of a displacement y = a sin
potential and thus approximate to spheres with earth's 2nx/X from the horizontal plane y = 0. It may be assumed
radius. The thickness of the layer can be considered to be that the amplitude a is considerably smaller than the
absolutely uniform and the slight curvature, corresponding wavelength X.
to a deviation of 8 /im from a tangential plane in 10 m, is If we consider a static system, there is a change in pres-
not significant in the present context. Hence a natural sure due to gravity of gpgy at any point within the molten
process for manufacture of high quality flat glass was glass which is defined by a given value of x and attributable
conceived, given the possibility of supporting molten glass to the surface distortion; g is gravitational acceleration and
upon a suitable liquid of higher density. pg the glass density. Assuming the glass-atmosphere sur-
In choosing such a liquid for window glass manufac- face tension to be Sg, there is also a change in pressure due
ture, the basic requirements are: to curvature of the surface of approximately — Sgy" or
(1) The density must be greater than that of molten glass. Sg4n2fy/X2. Hence a combined surface tensional and
(2) The melting point should be below 600 °C. gravitational pressure of magnitude given by the expression
(3) The boiling point should be sufficiently high to (gp% + Sg4n2IX2)y is produced by the surface distortion
ensure a low vapour pressure at ~ 1050 °C. which is alternately positive and negative with y and
(4) Chemical interaction with glass should be minimal. which produces viscous flow in the molten glass and so
Pure tin is unique in fulfilling all the above requirements reduces the amplitude of the distortion. It will be noted that
and also being readily available. Its specific gravity at the gravitational and surface tensional contributions are
153
 of magnitude (Sg + Sgt—St), equilibrium is established
when the central ribbon thickness has a value Te) the
'equilibrium thickness', which is given by the relation
2pt
= (Sg+Sgt-St)
£P g (Pt-Pg)
Since the magnitudes of the surface tensions may depend
on the chemical state of the system and 5 g t is not indepen-
Figure 2 Cross section of a ribbon of molten glass floating on
dently known, the value of Te must be found experimen-
molten tin tally and is discovered to be approximately 7 mm for
window glass on molten tin under a nitrogen-hydrogen
atmosphere. Since Se is known to be ~ 0-35 N m " 1 and
St to be ~ 0-5 N m " 1 , then Sgt is found also to be
equal when A is equal to (Sg4n2/gpg)*. Assuming Sg to ~ 0-5 N H I " 1 .
have a value 0-35 N m " 1 and pg to be as quoted above, It will be apparent that the upper and lower surfaces of
this value of A is about 24 mm. Gravity thus predomi- the glass and the molten tin surface are defined in cross
nates in removing surface distortion when A is greater than section by capillary curves which intersect at a common
24 mm whilst surface tension increases rapidly in impor- point at either edge of the ribbon. These curves intersect
tance as A is decreased below this value. at angles defined only by the magnitudes of the various
The order of magnitude of the forces available to remove surface tensions, but their inclination to the horizontal at
surface distortion is calculable since the incremental pres- the edge is determined by the densities of the liquids. By
sure has a maximum value when y = a. When A ~ 24 mm calculation the ribbon thickness is found to be within
this maximum pressure has a value 2gpga equal to 1 [im of the central ribbon thickness at a position 75 mm
~ 5 N m ~ 2 when a = 0-1 mm. For distortions of very from either edge. Hence theoretically a ribbon of width
long wavelength, e.g. thickness variations between edge greater than 150 mm can be considered as practically
and centre of a wide ribbon, the surface tension term is uniform in thickness over the central region. If this central
negligible and the force is half this value. It is therefore ribbon thickness is assumed greater than the equilibrium
necessary to control the viscosity of the molten glass at a value by a small amount AT, there is an unbalanced out-
relatively low value for a time sufficient to remove sub- ward spreading pressure of magnitude gpgAT (p t —pg)/pt
stantially surface distortion and thickness variations which is almost exactly equal to the pressure available to
under the action of these forces of low magnitude. remove a long-wave surface distortion of amplitude AT.
Since similar forces also operate to remove surface Hence in the absence of other forces, subjecting the glass
distortion from the lower surface of the glass it is possible to a viscosity-time schedule which is adequate to remove
to deduce an approximate time constant for the removal surface distortion or thickness variation must inevitably
of surface distortions or analogous thickness variations of establish a layer of molten glass of thickness which is very
specified wavelength. This time constant is inversely pro- close to the equilibrium value Te when the glass is free to
portional to the viscosity of the glass and is a function of spread. High quality ribbons in a range of desirable thick-
the surface tensions of the upper and lower surfaces of the nesses thus cannot be expected to be produced directly
glass, the densities of molten glass and tin as well as of the from a simple float system with a fixed value of equilibrium
glass thickness and wavelength of the distortion. thickness.
In the experimental work referred to above it was estab- Since it was obviously desirable to change the equili-
lished that a ribbon of good optical quality could be brium thickness, in particular to reduce its magnitude to,
achieved by holding a rough cast ribbon at a temperature say, 2-3 mm, investigations to this end have been carried
of ~ 1050 °C, corresponding to a viscosity of 1 k N s m " 2 out in our laboratories. From the above expression for Te
for a time of the order of one minute. it can be deduced that feasible changes in density of molten
glass or molten metal do not reduce its value substantially.
THE INFLUENCE OF SURFACE TENSION For example, increasing the specific gravity of the metal
During this experimental work, it was also discovered that from 6-5 to 20 reduces Te from 7 mm to 5-9 mm; increasing
molten glass does not spread indefinitely over the surface the specific gravity of the glass from 2-4 to 3-25 gives a
of clean molten tin. Referring to figure 2, which is a cross minimum value of 6-75 mm for Te when the specific
section normal to the length of a ribbon of glass of uniform gravity of the molten metal is 6-5. Experimental work was
width, the condition for the system to be nonspreading is therefore undertaken to investigate the possibility of
that the sum of the glass-atmosphere surface tension Sg increasing the spreading coefficient Fs in order to reduce
and the tin-glass surface tension S^t is greater than the Te. Although this work embraced a wide range of molten
value of St, the tin-atmosphere surface tension. The value metals and alloys and limited changes in chemical com-
of St — (Se + Sgt) is defined as the spreading coefficient Fs, position of the glass and the atmosphere, a useful reduction
and when this is negative the surface energy of the system in the equilibrium thickness of the glass together with a
increases as the lighter liquid spreads. bottom surface of acceptable quality was not achieved.
Following Langmuir (1933), let the ribbon be wide In general, an increased spreading coefficient requires a
enough to be considered of uniform thickness T over its reduction in the glass-metal interfacial tension relative
central portion. The net gravitational spreading force per to the metal-atmosphere surface tension which implies
unit length of the floating ribbon is given by the expression increased chemical interaction between glass and metal.
gT2ps
THE RIBBON FORMING PROCESS
A definite advantage of the nonspreading property of
where pt and pg are the densities of molten tin and molten molten glass on molten tin was that it led to the develop-
glass and g is gravitational acceleration. Since there is an ment of a ribbon forming process in which the molten
effective net inward surface tensional force per unit length glass could be poured directly on to the tin as a con-
154
tinuous stream. As described in detail by Pilkington
(1969), a number of problems were encountered when
attempting to deliver a stream of suitably high quality
glass on to the surface of the tin. One of these problems
is that the glass which has flowed from a melting unit and
which has been in direct contact with refractory material
cannot be used to form the lower surface of the ribbon
since the resulting small variations in chemical composition
produce variations in refractive index which cause optical
distortion. Using a combination of model work and pilot
plant experimental work a flow system in which contami-
nated glass is diverted into the extreme edges of the ribbon
was devised and a stream of molten glass of high quality
could be delivered on to the molten metal bath.
The lateral spread of the glass over the bath surface and
the forming of a continuous ribbon is illustrated in the Figure 3 Plan view of simple ribbon forming process
diagrammatic plan view of figure 3. The stream of molten
glass at a temperature of ~ 1050 °C, corresponding to a
viscosity of 1 k N s m " 2 , has an initial thickness several
times greater than the equilibrium value and therefore
initially spreads outwards over the surface of the molten
tin. The movement of the glass along the bath is main-
tained by a longitudinal tractive force applied by the roller
conveyor used to transport the formed ribbon. As the
average thickness of the glass approaches the equilibrium
value the rate of spreading decreases, and if the temperature
of the ribbon is maintained the width and thickness may
then decrease under the influence of the tractive force, as
shown in the diagram. It is essential to maintain the tem-
perature level for a time sufficient to produce a ribbon of
uniform thickness and free of surface distortion.
As the ribbon moves along the bath it is cooled and its
viscosity is increased, limiting further dimensional change.
Eventually it is cooled sufficiently to be removed from the
bath without incurring surface damage, the required
Figure 4 Process for thin glass manufacture
temperature level for window glass for this operation being
approximately 600 °C corresponding to a viscosity of
10 GN s m~ 2 . The variation of width, thickness and speed
of the ribbon with distance along the bath is seen to be to develop special techniques for the manufacture of rib-
determined by the interacting surface tensional, gravi- bons of a range of thickness in acceptable widths.
tational and tractive forces which viscously deform the
ribbon at a rate determined by the viscosity. Since the THIN GLASS MANUFACTURE
edges of the ribbon are not constrained there are no applied As indicated above, wide ribbons of thickness very much
shear forces which can lead to distortion. less than the equilibrium value cannot be produced directly
The magnitude of the longitudinal tractive force in this in a simple forming system for the reasons outlined.
simple forming system is determined by the relative speeds However it was recognized that acceptable ribbons of
of the molten glass stream and the final ribbon, and the thickness down to 3 mm could be made if attenuation by
viscosity regime. The establishment of a viscosity level of the longitudinal tractive force could take place almost
approximately 1 k N s m " 2 for a time sufficient to ensure independently of the influence of surface tension forces.
the production of high quality ribbon predetermines the One method of achieving this objective is to stretch a
approximate level of the longitudinal tractive force per formed ribbon on the float bath under a tractive force per
unit area in the forming region. This force may be lower unit area which is very high compared with the net surface
than or of the same order as the gravitational and surface tensional force per unit area which acts to increase ribbon
tensional forces which control the lateral spread of the thickness. Under these conditions it is expected that
ribbon and which are of approximate magnitude 100 proportional reductions in width and thickness would
N m~ 2 . Under these conditions this forming process will occur since the glass is isotropic.
produce a ribbon of thickness near to the equilibrium The practical process for producing thin ribbons using
value and of a width chosen to be commercially desirable, these principles is illustrated in figure 4. A wide primary
for example a 6 mm ribbon 2-5 m wide can be produced. ribbon of good quality is formed in the usual way under
A corollary is that any attempt to produce thinner ribbons conditions of low tractive force, say of width greater than
by increasing the speed of the formed ribbon will generally 5 m and of thickness near 7 mm. The speed of this primary
produce a proportionally greater decrease in width than ribbon is defined by pairs of driven edge rolls which grip
in thickness, since the net surface tensional force per unit the ribbon edges after it has been cooled and stiffened.
area available to maintain thickness is comparable with Downstream of the edge rolls, the ribbon is reheated and
the tractive force per unit area. Increasing the speed on stretched by applied tractive force. If the temperature of
an early production unit decreased the ribbon width from this reheated ribbon is approximately 850 °C, corres-
2-5 m to less than 1 m whilst the thickness was reduced ponding to a viscosity of 100 kN s m~ 2 , the tractive force
only from ~ 6-5 mm to ~ 4 mm. Hence it was necessary is approximately 100 times greater than its value in the
155
 Technological forecasting

PD WILMOT, MSc, FlnstP, European Technological

Forecasting Association, Duxford

"Science has entered a phase of specialization previously

Figure 5 Process for thick glass manufacture
unknown and this will forever remain the case. Not only
externally but inwardly, matters stand at a point where the
individual can acquire the sure consciousness of achieving
primary ribbon forming system and this force is high something truly perfect in the field of science, only in case
compared with the surface tension forces. Under these he is a strict specialist". This forecast by Weber, speaking
conditions, reductions in width and thickness are almost at the University of Munich in 1910, has proved far from
proportional and a final ribbon of width greater than 2 m the truth. Although greater specialization and new
and of 3 mm thickness may be produced. The section of branches of all the sciences have emerged since his day,
cooled ribbon gripped by the edge rolls provides the neces- development has been mainly influenced by the interplay
sary reaction to the applied tractive force. of quite distinct and apparently unrelated sets of dis-
coveries and skills.
THICK GLASS MANUFACTURE
Within the realm of physics, the application of wave
The principle of the method used to manufacture glass theory to corpuscles has led to quantum mechanics, and
thicker than the equilibrium value is relatively simple. the development of microwave circuits for radar has
The outward spread of the molten glass over the molten brought into being the science of radioastronomy, to name
metal is restricted by non-wetted parallel longitudinal only two cases. Also in the interplay between physics and
guides as indicated in figure 5, and the ribbon thickness is other sciences and technologies, nuclear magnetic reson-
thus determined by the volume flow and the speed. The ance has revolutionized branches of analytical chemistry,
high temperature is maintained for a time sufficient to and cryoscopy and superconductivity will make a profound
produce a ribbon of uniform thickness and free of surface impact on electrical engineering.
distortion. The ribbon temperature is then gradually
It is this interplay between disciplines, particularly the
reduced, the glass finally becoming sufficiently viscous to
effect of new knowledge in one field on another field, which
be dimensionally stable. Since any lateral shearing forces is the vital stuff of technological forecasting. Jantsch (1967)
applied to the edges of the ribbon may produce surface in the most comprehensive reference book on this subject
distortion, successful production of high quality glass has defines technological forecasting as "the probabilistic
necessitated the development of methods of reducing the assessment on a relatively high confidence level of future
friction between the guides and the molten glass to a low technology transfer". Physicists, who often require ex-
value. Glass up to 15 mm in thickness is produced by this tremely expensive equipment and who are usually members
method. of large multidisciplinary research teams, are as aware as
any other scientists of the need to analyse what paths new
CONCLUSION
knowledge can take and how it can be most effectively
In tracing the physical aspects of the development of the directed. A wrong timing can lead us into expensive by-
float process, emphasis has been placed on the advantages ways. We are also in sight of the time when the choice will
and disadvantages of exploiting the natural tendency of be not just between guns or butter, but between guns,
liquid surfaces to become flat as the basic principle of a butter and more time to look at sunsets - social and
flat glass manufacturing process. In pursuing the overall aesthetic factors will weigh more and more heavily in the
objective, a wide range of other physical problems has been choice of options.
encountered, some of them common to other glass making
processes, others unique to this process such as the heat
THE ORIGINS OF TECHNOLOGICAL FORECASTING
transfer and flow phenomena in a liquid metal. The under-
standing and control of these aspects have been equally as The exercise of technological forecasting is by no means a
important as an appreciation of the physics of ribbon scholastic activity but very much a hard practical business
formation in the successful development of the float linked to the decision making of any organization. It is an
process. aspect of corporate planning. Like market research or
Acknowledgements are obviously due to the large num- financial forecasting, it does not pretend to offer firm
ber of colleagues who have contributed to the under- predictions but gives to the management of the concern an
standing and development of this process, to Dr D S analysis of the possible future environments and the likely
Oliver, Director of Research and Development, and also outcome of decisions undertaken by that organization.
to the Directors of Pilkington Brothers Limited for per- Technological forecasting is by no means a new activity,
mission to publish this paper. for all persons concerned with decision making, in an
environment where technological innovation is of any
FURTHER READING importance, must make some kind of technological fore-
Langmuir I. 1933 /. Client. Phys. 1, 756. cast. However, about 1960 especially in the USA, many
Pilkington L. A. B. 1969 Proc. R. Soc. A./314 1 companies particularly those concerned with aircraft and
156