Pilkington Float Glass

Review Lecture.
The Float Glass Process

Author(s): L. A. B. Pilkington
Source: Proceedings of the Royal Society of London. Series A, Mathematical and Physical
Sciences, Vol. 314, No. 1516 (Dec. 16, 1969), pp. 1-25
Published by: Royal Society
Stable URL: http://www.jstor.org/stable/2416528
Accessed: 25-07-2016 22:34 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Royal Society is collaborating with JSTOR to digitize, preserve and extend access to Proceedings of the
Royal Society of London. Series A, Mathematical and Physical Sciences
This content downloaded from 128.111.121.42 on Mon, 25 Jul 2016 22:34:46 UTC
All use subject to http://about.jstor.org/terms
Proc. Roy. Soc. Lond. A. 314, 1-25 (1969)
Printed in Great Britain
REVIEW LECTURE
The float glass process
BY L. A. B. PILKINGTONt
Pilkington Brothers Limited, St Helens, Lancashire
(Delivered 13 February 1969-Received 14 April 1969)
[Plate 1]
My subject is the float process for making flat glass. I would like, first of all, to
put the float process into perspective by describing briefly, and in simple terms, the
methods used for making flat glass before and at the time of the invention of
the float process and then to describe the development of the process itself and
the position it occupies in the flat glass industry today. Finally, I would like to
describe in as much depth as time allows, three of the main problems which had
to be tackled in developing this process.
The Egyptians seem to have been the first people to realize what could be done
with glass when it is hot and plastic, and they made vessels for cosmetics and per-
fumes by, it is assumed, trailing molten glass around a shaped core. By Roman
times glass was being blown and moulded, cut and engraved, painted and gilded,
and the Romans had mastered the plastic character of heat softened glass so funda-
mental to today's processes.
In the Middle Ages, the glass makers' main achievements were coloured glass
windows, and while the decorative application of glass progressed fairly rapidly
under the Venetians, and glass men like Ravenscroft in the seventeenth century,
it is only in the last 50 years that there have been any appreciable advances in the
development of flat glass for windows.
Since it is with flat glass that my lecture is concerned, I hope you will therefore
forgive me for skipping 3000 years of history!
The importance of the float process can be placed in perspective by taking a
look at the processes used for making flat glass before and at the time of its invention.
There have, through the ages, been two basic methods of forming flat glass:
the window glass processes and the plate glass processes.
Window glass processes have all depended on forming a sheet by stretching a
lump of molten glass. They all have the characteristic of brilliant fire finish. Three
processes, crown, cylinder and finally drawn sheet, have been used to make windows.
Up to the middle of the nineteenth century the crown process (figure 1, plate 1)
was the most commonly used and Henry Chance writing in 1888 tells us how the
t Elected F.R.S. 20 March 1969.
I 1 ] Vol. 3I4. A. (i6 December I969)
2 L. A. B. Pilkington
crown or disk was spun after the initial blowing and shaping stages on the end of
an iron rod. He says: 'A man ... with a veil before his face, stands in front of a huge
circle of flame, into which he thrusts his piece rapidly, meanwhile revolving his
ponty (the iron rod). The action of heat and centrifugal force combined is soon
visible. The nose of the piece expands, the parts around cannot resist the tendency ...
the next moment, before the eyes of the spectator, is whirling a thin transparent
circular plate of glass which but a few minutes before was lying in the glass pot.'
FIGuRE 2. Square panes were cut from a round crown, involving wastage. The numbers
give the sizes in inches.
Disks of 1.8 m diameter could be produced but the usual size was about 1.4 m
(figure 2). This restriction in size and the fact that each crown had a bullion in the
centre obviously dictated the scope for the product in windows. Another dis-
advantage was that square panes were cut from a round disk, involving wastage.
The next big advance in glass manufacture, which removed the limitations on
window size imposed by the Crown process, came in the mid-nineteenth century
with the introduction of the cylinder process which was capable of making much
bigger panes. The process involved blowing a large cylinder with was allowed to
cool before being split and flattened (figure 3). The process was mechanized some-
what until cylinders up to 4 m long and 0.6 m in diameter could be blown.
Pilkington Proc. Roy. Soc. Lond. A, volume 314, plate 1
FIGURE 1. The crown process spun glass into a disk.
FIGURE 4. In the sheet process the edge of the glass ribbon is gripped between rollers
to prevent 'waisting in'.
(Facing p. 2)
The float glass process 3
Sheet or window glass production was first mechanized on a large scale in the early
part of this century. In 1903, the American Window Glass Company developed a
method for the mechanical blowing of cylinders many times larger than hand blown
cylinders. This drawn cylinder process drew cylinders up to 13.4 m long and 1 m in
diameter from a molten pool of glass. The quality of the glass, however, was incon-
sistent.
FIGTTRE 3. Glass made by the cylinder process had to be split and flattened for windows.
The problem with the cylinder process was that it was discontinuous and the
cylinders, however large, had to be split and flattened which was costly and harmful
to the surface. The logical evolution of this process was to draw a flat continuous
sheet of glass from the molten pool.
The modern sheet glass process was first developed by Fourcault in about 1914.
The glass is drawn vertically in a ribbon from a bath of molten glass. The main
problem in drawing such a sheet is to prevent the 'waisting in'. This is achieved by
passing the edges of the ribbon between cooled rollers which stiffen them (see figure
4, plate 1).
The surface of glass made in this way has what is called 'fire-finish' (figure 5).
The fire finish surface is achieved by letting the glass cool down on its own without
touching anything solid while it is still soft.
However, the solidified ribbon of glass has a certain amount of distortion which
cannot be avoided because of small differences in viscosity due to chemical and
1-2
thermal inhomogeneities. The thickness of the ribbon of glass drawn from the bath
is controlled by the viscosity, so that the effect of even small inhomogeneities is
variations in thickness of the finished sheet. The machine needed to produce this
sheet glass is relatively simple and the glass is therefore inexpensive.
FIGURE 5. Sheet glass has a perfect fire finish. Only the edges gripped by rollers are marked.
The major problem with all the glass made by these window processes was that
it was bedevilled by distortion. All the manufacturing methods involved stretching
the molten glass whether by spinning, blowing or pulling it and this stretching con-
verted inhomogeneities into distortion. The processes also made only a com-
paratively thin glass.
Neither of these properties was acceptable for mirrors or coach windows, or, later,
for car windows or for the large distortion free shop windows which came into vogue
from 1850 onwards.
The plate process was developed to meet these requirements (figure 6). To create
distortion free glass the molten metal was cast on to a table, rolled into a plate and,
after annealing, ground flat and then polished. Grinding involved several stages
using finer and finer sand and polishing was done with rouge.
The principles and, as we shall see later, the disadvantages of this process remained
with the industry until the 1950s. Financial requirements were, from the beginning,
quite enormous. The ordinary small glasshouse for making crown or cylinder glass
could be set up fairly cheaply. But much more was required for plate manufacture.
It needed a big melting pot furnace, a casting table, cuvettes or cisterns, cranes,
grinding and polishing machinery, and extensive warehousing facilities. All this
required unprecedented organization and greatly increased capital.
FIGURE 6. In early plate processes glass was cast on a table and rolled into a plate.
Some indication of the scale of the process is given by the factory which made
Britain's first plate glass in 1773. Built by British Plate Glass Company at Raven-
head, St Helens, the casting hall was 103 m long by 46 m wide, the largest industrial
building of the period.
From 1688 until the First World War, plate glass manufacture moved ahead
only very slowly. Progress was keyed to the logical development of greater mechani-
zation with the process always embodying the principles already described. It
was in the 1920s that this classic example of evolutionary progress accelerated in
leaps and bounds mainly due to the brilliance of development engineers.
Just after the First World War the Bicheroux process was introduced. Glass was
still melted in pots but it was then rolled into a sheet between mechanical rollers
rather than being cast on to a table and then rolled. This made a smoother sheet
with a consequent saving in time and material in the grinding process.
Simultaneously, however, the search was on for a method of making the process
continuous.
The first breakthrough came from Ford in America, where it was shown that
glass could be rolled continuously.
As a result of Pilkington cooperation with Ford this discovery was successfully

exploited. And after Ford's initial work the major developments in plate glass
manufacture came from Britain.
We first developed a process which successfully combined a continuous melting
furnace with the continuous rolling of a ribbon of glass. This was then sold back to
America.
FIGURE 7. Continuous grinding and polishing of plate glass came in 1923.
In 1923, the next advance towards continuity of plate production was made when
Pilkington introduced the first continuous grinding and polishing machine
(figure 7). It took cut plates of glass onto a series of tables which moved through the
grinders and polishers. At the end of the process the table dropped into a tunnel
and returned to accept another plate of glass.
The next logical step was a machine which could grind the ribbon of glass on both
sides simultaneously as it came out of the annealing 'lehr' (a long chamber through
which a continuous ribbon passes while being cooled) and before it was cut into
plates. Pilkington developed such a machine during the early 1930s and put it
into service at their Doncaster works in 1935.
The machine, called the twin grinder, is acknowledged as the final and most re-
markable development in the long history of plate glass manufacture. It must rank
as one of the finest examples of large scale precision engineering and it gave Britain
world technological leadership in the manufacture of high quality flat glass.
In the machine a continuous ribbon of glass about 300 m long was ground on both
surfaces at the same time with enormous grinding wheels fed with progressively
finer sand (figure 8). The machine was driven by 1.5 MW and this power was ex-
pended in grinding a slender ribbon of brittle material. Even more remarkable was
the fact that the bottom grinding wheels were kept perfectly flat and level while
they were actually wearing away.
FIGuRE 8. Both surfaces of a plate glass ribbon were ground by the Pilkington twin grinder.
The polished plate process was used in Britain until the 1960s. It met all the de-
mand for thick and thin distortion free windows. Why then has it been superseded?
Glass wastage amounting to 20% of production and high capital and operating
costs were its drawbacks. Its sister process, sheet, could not replace it because it
could not be used to make the high quality products free from distortion which were
required by the market. It could, however, make glass which retained its natural
brilliance without the need for grinding and polishing.
Many minds had dreamed of combining the best features of both processes. They
wanted to make glass with the fire polish and inexpensiveness of sheet and with the
distortion free quality of polished plate. The first time this was successfully achieved
was in 1959 when the development of float was announced.
In the float process, a continuous ribbon of glass moves out of the melting fur-
nace and floats along the surface of an enclosed bath of molten tin (figure 9). The
ribbon is held in a chemically controlled atmosphere at a high enough temperature
for a long enough time for the irregularities to melt out and for the surfaces to become
flat and parallel. Because the surface of the molten tin is dead flat, the glass also
becomes flat.
controlled
atmosphere
<9 ~~molten tin 1N,\ \A

furnace float bath lehr
FIGURE 9. Diagram of the float process.
FIGURE, 1 0. Finished glass made by the float process. Grinding 'and polishing is eliminated.
The ribbon is then cooled down while still advancing across the molteni tin until
the surfaces are hard enough for it to be taken out of the bath without the rollers
marking the bottom surface; so a ribbon is produced with uniform thickness and
bright fire polished surfaces without any need for grinding and polishing (figure 10).
The principles of the process perhaps sound very simple, but there is a saying-
I do not know who coined it-that the price of progress is trouble, and it is very true.
How complex the development work was is illustrated by the fact that it took us
seven years and four million pounds to make any saleable glass and a total of
seven million pounds before the process could achieve the set objective of replacing
plate.
When we started we knew that what we were attempting was going to be one of
the most exciting things in the history of the flat glass industry if we could pull it
off. Within weeks of thinking of the idea in 1952, tests had proved promising.
Soon a ?25 000 pilot plant was built and here the night and day ordeal began. When
I look back at some of the glass we made at that time, it seems unimpressive, al-
though we were all tremendously excited and enthusiastic because for the first
time we had got glass to float on tin, and the surfaces were fire polished.
The rash of immediate problems which had to be solved included controlling the
chemical atmosphere in the float bath where the glass floated on molten tin. The
molten tin reacted readily with any oxygen in the system and this resulted in
damage to the ribbon surface through physical and chemical interaction at the glass/
tin interface. It took us many years to achieve our target of controlling the levels
of oxygen and other impurities in the bath.
The decision to go ahead on a large enough scale to see what further problems were
involved demanded big money and big effort. I have always said that to be a good
development man you must be a born optimist. That may have been all right for
those of us in the development team, but the Board were only interested in a cold
blooded, objective analysis of the project and its progress so I had to be both opti-
mistic and objective.
In 1954, the Pilkington Board decided to give the project the highest possible
priority, so that success or failure would be decided as early as possible.
We knew that under certain conditions, glass and metal could live together.
This was known even before the idea of float occurred. However, at the time when
the Board's decision was taken it seemed quite likely that glass coming into contact
with metal at the forming stage would always be spoiled too much to make a
saleable product.
The development team also had to remain optimistic in the face of all the problems
which arose.
In 1954 then, we had our pilot plant with a flow of glass coming at us continuously.
The analysis of what was happening and our next steps had to keep up with this
flow of glass.
No single fault or trouble failed to yield to our attack, but collectively they were
enormous. Controlling glass flow, mastering ribbon formation and ensuring the
whole chemistry was balanced and right were three major development areas
which I shall refer to later. There were others like these and at times we felt it was
almost going to be too difficult to hold everything together to produce saleable
glass.
We received little help from outside sources. First, because we were breaking
new ground and were working beyond the frontiers of existing knowledge, and
secondly because we could not tell people too much because of security.
The critical moment came and this was the most important decision that we made
in the whole project, when we decided to step up from the experimental stage to
the construction and operation of a full scale production plant.
We actually took the decision to build a production plant when we were still
experimenting, and we thought we were very much more knowledgeable than we
turned out to be when in 1957 we tried to make a competitive saleable product.
Once having set up a production plant, the consequences of failure could have been
enormous, and in retrospect we were woefully unaware of the magnitude of the
problems we were going to face when we reached a mass production scale. It
was a terrific fight to turn the experiment into commercial success.
We found it was one thing to make glass which looked good on a pilot plant and
quite another to make glass which could be sold as an equivalent of twin ground
plate and at the same price.
On our first production plant we made unusable glass for one year and two months.
I had to report regularly to the Board, and every month put in a requisition to
justify another month's expenditure of ?100000. It was a tremendous credit to
the Board that they gave unwavering support throughout.
After the most concentrated effort we at last managed to make the first piece of
glass which was really good.
It was easy to make glass better than sheet but our target was to match the
fantastically high quality of twin ground plate. We had at last managed to do this
and at last we were freed from our greatest fear: that the process was interesting
enough to justify further work and expenditure but that the product might not be
good enough to replace plate and, therefore, not commercially viable.
It has been said that float approaches close to a theoretically perfect process.
It certainly has many built in advantages.
First, in such a floating system the liquid surfaces become naturally flat and
parallel. While the glass floats on the surface of the molten metal, and is held at a
high enough temperature, the forces of gravity and surface tension combine to
produce a good planimetry. There is little tendency for the forces which form the
ribbon to convert any thermal or chemical inhomogeneity into distortion.
Secondly, we have not yet found a limitation on the output from the float bath.
The speed at which the ribbon is formed is determined by the melting capacity of
the furnace. We have already run at speeds in excess of 14 m a minute to make 3 mm
thick glass and there seems to be no reason why we should not increase this speed.
Thirdly, the variable costs for converting molten glass into a finished ribbon are
remarkably low and only a small fraction of the equivalent part of the polished
plate process.
Fourthly, the width of the ribbon can be easily altered at any time within the
limits of the float bath and lehr and even the dimensions of the bath itself can be
The float glass process 1 1
changed without much difficulty. This is important in reducing warehouse loss by

matching the ribbon to the orders being cut.
Fifthly, the process produces glass with a very bright finish and surface damage
is much lower than with any polished plate glass we ever made.
Sixthly, because it is a horizontal process, annealing the glass is relatively simple.
Seventhly, it produces a finished ribbon at the end of the annealing lehr which
encourages the use of an automatic warehouse.
And finally, the process lends itself to long trouble free runs. We have, in fact,
managed to run a plant for as long as 24 months without a major stop.
After the announcement in 1959 that we could make 6.5 mm float glass, we aimed
to make it thicker and thinner. The same forces which made the glass flat were
trying to determine the overall thickness of the ribbon at approximately 6 mm.
This was fortunate since 50 % of our market for polished plate glass had been for this
thickness.
But we had the other 50% to think about. The full commercial potential of float
could not have been realized without mastery of ribbon thickness. Just two years
after float was announced we had learned to make a product half the thickness of
our original float. The principle was to stretch the glass but in a gentle and controlled
way so that none of the distortions arising in sheet glass processes could occur.
In the next three years we managed to make float thicker. The principle is simple,
in that we arrest the spread of molten glass in the float bath and allow it to build
up into thicknesses which could be as great as 5 cm. The commercial range of thick-
nesses now available is 3 to 15 mm (8 to 5 in).
Another problem we had been conscious of during all stages of development was
the need to produce a warehousing facility capable of handling the flow of good glass
that the process promised to produce and cutting it into the myriad sizes our cus-
tomers require. Now we have developed fully mechanized warehouses for cutting
and handling this glass. This year they will cut and handle 16 000 km of glass ribbon.
The warehouses make full use of computers and stand by themselves as a major
advance in flat glass technology. The computers are used to match orders for a
very complex range of glass sizes to the continuous ribbon of glass. They, in fact
instruct the cutting machines in such a way that optimum usage of glass is obtained.
In addition to warehousing advances, the float process has opened the way to
other radical developments in flat glass manufacture. One such example is the
electro-float process, announced in 1967, which in itself is a major invention.
The process enables the surface of the clear float ribbon to be modified while it
passes over the bath of tin. It exploits the unique features of the float process to
effect rapid ion replacement between metals and glass. Although the full potential
of this process remains to be realized it has already proved capable of producing,
economically, small quantities of heat rejecting glass on a mass production line.
This first product, called 'Spectrafloat', is already being used in buildings and in
prototype glass roofed cars.
What I have said so far will, I hope, have put the float process into perspective.
It will, I hope, be clear why glassmakers had always wanted to develop a process
capable of making a fire polished glass with parallel surfaces and without expensive
finishing processes.
The basic float process has now been licensed to fifteen overseas glass manufac-
turers and over twenty plants are in operation or under construction. Together
they are turning out about 100 km2 of glass a year. The Americans are using it
and the Russians are using it. So are Japan, Canada, France, Germany, Belgium,
Italy, Spain, Czechoslovakia and Mexico. In fact every major flat glass manufac-
turer in the world has a licence.
Against this background it becomes clear that we had an enormous stimulus to
overcome all the problems standing in the way of successful development. I should,
then, like to outline the extent of, and solution to, just three of the fundamental
problems we met in the early development years.
Perhaps the most fundamental was the control of glass flow. We had a melting
furnace and a bath of molten tin. The problem was how to flow the glass from one
unit to the other.
When experimental work started on the float process, Pilkington already had
very great expertise in the techniques of forming a primary ribbon between water
cooled steel rollers. You will remember that the polished plate process used rollers
to form a continuous ribbon of glass which, being marked by the rollers, was then
ground and polished.
water cooled
rollers heaters atmosphere
tin
FIGURE 11. In experiments with the float process glass was rolled on to the molten tin.
It was natural enough, then, that we should begin our experiments by rolling a
ribbon from the melting furnace on to the molten tin (figure 11). There, it was
reheated and the imperfections introduced by the cooled casting rollers were melted
out under the influence of the forces of surface tension and gravity to produce a
ribbon with flat and parallel fire finished surfaces. This was gradually cooled while
on the tin and passed out of the float bath without damage to its surface.
We thought too that varying the thickness of the primary rolled ribbon would
determine the thickness of the finished ribbon. This, as we shall see later, was not
to be the case, and in turn meant that the function of the rollers was little more
than a metering device for the flow of glass.
It also soon became clear that we faced serious problems as a result of the con-
densation of tin and tin compounds on the water cooled rollers. Some of these
deposits were imprinted on to the surfaces of the glass and provided nucleating
points for the growth of crystalline faults on the glass/tin interface when the ribbon
passed through the heating and cooling schedules in the float bath.
The practical work on the pilot plants in identifying these problems and the
theoretical analysis of the physics of the process both generated a profound change
in the development of the process. We dispensed with the rolling machine and fed
molten glass straight from the furnace down a refractory spout on to the bath of
molten tin where a ribbon of glass was formed. This is known as the 'direct pour'
float process because the ribbon does not have to be preformed between casting
rollers before entering the bath.
In the early experiments, the refractory spout dipped into the tin (figure 12).
And here we found more problems. Molten glass is chemically active and dissolves
all refractory materials to some extent but it was soon found that this rate of wear
was greatly accelerated at the tin/glass/refractory interface where the spout
dipped into the tin. Glass that had been in contact with the refractory spout and
become contaminated formed the bottom surface of the ribbon. And the change in
composition of the bottom surface of the ribbon was sufficient to cause optical
distortion of objects viewed through the finished glass.
heaters atmosphere
tin
FIGURE 12. Direct pour of the glass replaced the roller method.
The spout dipped into the tin.
Because of its fine parallel line nature, the defect is known in the industry as
'music lines'.
The first move to cure this defect was to fit a spout, which did not dip into the
tin, thus allowing a free fall of glass (figure 13). This eliminated the heavy local
refractory contamination of the glass at the tin/glass/refractory interface. It was
found, however, that even the solution of the refractory spout into the flowing
glass was sufficient to cause an unacceptable level of 'music lines' in the final ribbon.
To sheath the refractory in a metal such as platinum was no solution-a reducing
atmosphere is essential to the float process and under these conditions platinum is
rapidly converted to platinum silicide.
heaters atmosphere
tin
FIGuRE 13. Finally a spout was fitted which did not dip into the tin. It allowed a
free fall of glass.
The problem was solved finally by ensuring that glass which had been in contact
with the refractory spout did not form the bottom surface of the final ribbon.
Although the first full scale float unit had begun to operate in May 1957 no sale-
able glass was produced on it until July 1958 because of our struggles to solve these
'music lines' and many other problems. Because of heavy wear and tear in solving
these difficult problems, the spout arrangement was in very poor condition. In
particular, the spout had broken its back so that it sagged badly in the middle.
Even so it continued to make good saleable glass for several months from July 1958
and on into 1959 when the public announcement of the float process was made. The
plant was then shut down for overdue maintenance and what we thought was an
exact copy of the old spout arrangement was constructed-except, of course, a
sagging spout with a broken back was not installed.
Imagine our disappointment, however, when we started up again and made
unusable glass. It was clear that conditions were still not right in this critical area
where the glass flowing off the spout formed the ribbon in the float bath. We had
no problem with 'music lines' but for some reason the glass in critical regions of
the spout area was not flowing as freely in the new arrangement as it had in the old
and crystals and bubbles were forming in the ribbon. Glass is essentially a super
cooled liquid that has failed to crystallize but, as molten glass is cooled, it passes
through a temperature range in which nucleation and growth of crystals can occur.
If the glass is held within this range for too long, the crystal growth is on a large
scale and the glass becomes opaque and glassmakers say that the glass has 'devi-
trified'. With the new spout arrangement, the more static flow conditions encouraged
the formation of devitrification crystals and these were carried steadily into the
body of the ribbon of glass to make it unsaleable.
The use of models in which dynamic similarity between full scale and model scale
was achieved proved a powerful weapon in the three months of urgent investigation
following our unsuccessful start up. In our model, silicone oil represented the glass,
and lead nitrate solution, the tin. Using this model we were able to identify those
factors of the broken spout arrangement wherein lay the key to success.
With our new knowledge of float conditions we were able to design an arrangement
which achieved not only the good characteristics of the broken backed spout, but
also fed any contaminated glass into the edges of the ribbon. From that point
float achieved stability of operation and was established as a production process.
Mastery of glass flow illustrates how difficult problems in new areas of technology
can yield to a coordinated attack by people working together in process and pilot
plant work, in theoretical analysis of the problems and by model techniques.
The same resources had to be marshalled to conquer another fundamental
problem-ribbon formation.
I mentioned earlier our experiment with rollers and our assumption that glass
of any desired thickness could be made by preforming the ribbon with water cooled
rollers and then eliminating imperfections in the float bath.
To our surprise, however, we found that whatever thickness of ribbon was rolled
into the float bath, we obtained a ribbon of final thickness close to 6 mm if the glass
were held at a high enough temperature for a long enough time to melt out irregu-
larities.
We had, in fact, discovered a classic example of a non-spreading system in which
equilibrium is established between gravitational and surface tensional forces when
the central thickness of a large pool or ribbon of molten glass has reached a definite
value.
atmosphere equilibrium thickness, T
/S9
r molten glass 1\ St
\ ~~(densitypg) I -
Sgt
molten tin, density pt
FIGURE 14. Vertical section of a pool of molten glass floating on molten tin.
T2 = (Sg +Sgt-St) gp(

Ygpg(Pt -Pg)
As indicated in figure 14-which shows a vertical section of a pool of molten

glass floating on molten tin-this thickness is determined by the values of glass/
atmosphere surface tension Sg, tin/atmosphere surface tension St, and glass/tin
surface tension Sgt and of glass density, pg and molten tin density pt. For glass of
the normal soda-lime-silica composition supported on molten tin this 'equilibrium'
thickness T is approximately 7 mm. This value is insensitive to limited changes

in chemical composition of glass, metal, or atmosphere.
The practical advantage of this phenomenon was that it immediately suggested
the possibility of a forming process in which molten glass could be allowed to spread
freely over the surface of the molten metal, the edges of the glass being absolutely
free of any shearing forces which could distort the glass surfaces. The equilibrium
thickness was also near to a commercially desirable value of 6 mm.
The embodiment of this idea into a ribbon forming process is illustrated in figure
15. A stream of molten glass several centimetres thick is supplied continuously on
to the surface of a bath of molten tin, and under the influence of the gravitational,
surface tensional, and applied tractive forces, shown here schematically, a con-
tinuous ribbon of glass is produced. During this forming process the temperature
of the glass is sufficiently high to ensure that surface imperfections or lateral thick-
ness variations are removed and a 'fire polish' is imparted to the surfaces.
suirface tension molten tin
gravity tractive 2 final ribbon

1050?C-104P BD force 2 6000C 10"P
FIGURE 15. Plan view of ribbon formation.

1 P = 10-1N S m-2.
For a normal soda-lime-silica glass it is necessary to maintain a temperature of

1050 'C-equivalent to a viscosity of 1 kNs im-2 (104 P) for approximately 1 min to
produce surfaces of the required degree of perfection. The order of magnitude of
the operative forces is indicated by the fact that a maximum combined gravita-
tional and surface tensional force of about 5 N m-2 is available to remove an undula-
tion in the top surface of the glass with an amplitude of 0.1 mm and a wavelength
of 2.5 cm. As the ribbon is advanced along the bath it is progressively cooled. At a
temperature of 600 ?C the viscosity is sufficiently high (10 GN s m-2; 1011 P) for
the ribbon to be removed mechanically from the bath without surface damage.
While the existence of an equilibrium thickness of approximately 7 mm offers the
advantages I have described in the simple ribbon forming process, it also proved to
be a practical disadvantage in the production of thicker and thinner glass. The net
surface tensional or gravitational force available to change the thickness of a ribbon
which is substantially thinner or thicker than the equilibrium value is of the order
of 100 N m-2. This is at least an order of magnitude greater than the force available
to remove surface distortion and it is therefore apparent that the viscosity-time
regime necessary to produce a ribbon of high surface quality inevitably establishes a
ribbon of overall thickness near to the equilibrium value.
A further consequence is that any attempt to change ribbon thickness by varying
the applied tractive force, i.e. by changing the speed of the rollers which convey the
ribbon from the float bath, produces relatively large changes in width but small
changes in thickness, since these applied tractive forces are of the order of 100 N m-2
or comparable with the forces which determine ribbon thickness. This is illustrated
in figure 16 where the dashed line describes the changes in width and thickness of a
ribbon initially 2.5 m wide and 6 mm thick, represented by point 0, as speed is
increased or decreased. In particular the attempt to produce thinner glass by merely
increasing speed results in an unacceptable loss in width.
op
5.0- 0cJ O)
w 2.5 IFI abO

I I I$0n
I I I
0 4 8~~~~~~
-Q\\
0 4 8
thickness/mm
FIGURE 16. Relation between ribbon width and thickness.
For example, by boosting the speed on an early production unit, a reduction in

ribbon thickness from 6 mm to about 4 mm was accompanied by a change in ribbon
width from 2.5 to about 0.75 m, represented by a change from point 0 to point Q.
It was realized that the process required further development to produce a range of
thicknesses in acceptable widths.
The physical basis of the method adopted to produce the range of thicknesses
down to 2 mm is also illustrated in the diagram. A wide primary ribbon is formed at
a speed lower than that required for 6 mm production, say a ribbon of width greater
than 5 m, represented by the point P. This primary ribbon is then stretched on the
float bath at a temperature lower than that required to form the primary ribbon,
for example at 85000C corresponding to a viscosity of about 0.1I MN sM-2 (106 P).
Under these conditions the tractive force used to stretch the ribbon, which is pro-
portional to its viscosity, is of the order of 10 kN m-2 and is thus far greater than the
net surface tensional force of about 0.1 kN m-2 which acts to increase the thickness
2 Vol. 3I4. A.
as the ribbon is stretched. Since the glass behaves as an isotropic incompressible

viscous fluid at these temperatures the reductions in width and thickness are prac-
tically proportional. The solid line (figure 16) thus represents the changes in width
and thickness produced by stretching the primary ribbon P at relatively low tem-
peratures. A final ribbon of 3 mm thickness and 2.5 m width, represented by point
F, can be produced by this method.
edge rolls
molten tin
primary )cooled reheat

ribbon ribbon 8500C final ribbon
105000C 7000C _4o6P 60000C
104P J ol0p
OT'X
FIGURE 17. System for the production of thin float glass.

IP = 10-1Nsm-2.
The original practical system used is shown in figure 17. The width and thick-
ness of the primary ribbon are controlled by pairs of driven edge rolls which grip
the edges of the cooled ribbon and define its speed. Downstream of the edge rolls,
the ribbon is then reheated to approximately 850 ?C and stretched. The formation
of the primary ribbon under conditions of low tractive force is separated from the
stretching process requiring a high tractive force by the relatively stiff section of the
cooled ribbon. The necessary force reaction is thus provided by the edge rolls which
grip this stiffer section.
The subjecting of a high quality primary ribbon to a stretching process which
does not exploit variations in viscosity as variations in thickness produces a high
quality final ribbon with acceptable width and thickness.
In contrast to the complexity of the process for producing ribbons thinner than
the equilibrium value, thicker glass can be produced by a basically simple process.
The initial outward spread of the molten glass is restricted by non-wetted parallel
longitudinal guides (figure 18). Under these conditions the thickness of the ribbon is
determined by the mass flow and the speed. The temperature is maintained at
approximately 1050?C for a time sufficient to produce a ribbon of high surface
quality and uniform thickness. As indicated earlier, any shearing forces applied at
the edges of the ribbon may produce surface distortion, hence successful operation
of this process has required the development of techniques for reducing the friction
between the molten glass and the guides to a low value. The temperature of the
T-he float glass process 19
formed ribbon is progressively reduced and guides extend longitudinally a distance

sufficient to ensure that the ribbon is dimensionally stable. After careful develop-
ment work, high quality glass up to a thickness of 15 mm is now produced by this
process.
non-wetted guide
formed 6000C
1050?C104P ribbon ^1011P
molten tin
FIGURE 18. System for the production of thick float glass.

1P = 10-1NSM-2.
In concluding this description of ribbon formation, may I draw your attention

once again to the essence of the float process. In such a floating system, the natural
forces of gravity and surface tension combine to produce a ribbon of high surface
quality and uniformity of thickness. However, these same forces also combine to
produce the phenomenon of an 'equilibrium' thickness, and thus create the problem
of producing ribbons in a range of desirable thicknesses and the necessity for
developing special techniques to overcome this limitation.
Other intrinsic and vital components of the float process also yielded problems
which development teams had to solve and here I would like to pass on the chemistry
of float.
I have already mentioned that difficulties arose from the affinity of tin for oxygen.
By comparison with other flat glass forming processes, float is highly vulnerable to
chemically produced effects. The ribbon is formed in contact with materials which
may react with it-on one side a molten metal, on the other an atmosphere of nitro-
gen and hydrogen which by glassmaking standards is highly reducing. Its surfaces
can, therefore, be expected to have somewhat unusual properties. In addition, since
the ribbon is molten and deformable along almost the whole length of the bath,
any specks falling from the bath atmosphere or any bubbles rising through the tin
will permanently damage the surface.
The chemist's objective is to produce a ribbon which has no physical imperfec-
tions caused by chemical reactions in the bath and has a surface composition which
gives no trouble in subsequent processing or in service.
The nature of the chemical problems we had to overcome turns on the choice of
2-2
the supporting metal (table 1). There is, in fact, little room for choice. The metal
must, first of all, be liquid over the approximate range 600 to 1050?C in order
that the ribbon may be formed satisfactorily. The table lists the metals that meet
this requirement-bismuth, gallium, indium, lithium, lead, thallium and tin.
TABLE 1. CRITERIA DETERMINING THE CHOICE OF A SUPPORT METAL FOR

THE FLOAT BATH
melting point boiling point estimated vapour

?C ?C density at pressure at
10500C 10270C
g cm-3 Torr
required < 600 > 1050 > 2.5 <0.1
bismuth 271 1680 9.1 27

gallium 30 2420 5.5 7.6x 10-3
indium 156 2075 6.5 7.9 x 10-2
lithium 179 1329 0.5 55
lead 328 1740 9.8 1.9
thallium 303 1460 10.9 16
tin 232 2623 6.5 1.9x 10-4
In addition, the density must be high enough to support the glass-this eliminates
lithium as a possibility. As might be expected from the boiling points, bismuth,
lithium and thallium have appreciable vapour pressures at the upper end of float
temperature range. It turns out that even the relatively low vapour pressure of lead
is unacceptable for float operation because of the effects of condensation. This leaves
gallium, indium and tin, of which tin shows the least interaction with glass in a
float system and is the most readily available. Various alloys have also been tested
but without showing clear advantages over tin.
TABLE 2. VOLATILIZATION FROM DILUTE SOLUTIONS OF OXYGEN

AND SULPHUR IN TIN
tin in saturated
impurity in tin main component of vapour at 10270C
of vapour mg mr3
none tin 0.3
oxygen 10 parts/106 by weight stannous oxide 3
sulphur 10 parts/106 by weight stannous sulphide 100
Oxide volatilization derived from Belford & Alcock (I964) and Colin & Drowart (I965);
sulphide volatilization derived from Sudo (95I), Espelund (I966) and Klushin & Chemykh
(I960) .
Pure tin, at float bath temperatures, has little interaction with glass and gives a
negligible quantity of condensable vapour. The situation changes if the tin con-
tains only tens of parts per million of oxygen and sulphur. Stannous sulphide and
stannous oxide are then evolved into the bath atmosphere. Subsequently, through
condensation and chemical reduction of these compounds, small but troublesome
specks of tin can be produced on the top surface of the ribbon. The speck shown in
figure 19, though only 0.12 mm in diameter, is a large one by our quality control
standards. Table 2 compares volatilization from oxygen and sulphur solutions in
tin. Since concentrations of the two contaminants in the bath tin are of the same
order of magnitude, it can be seen that sulphur is the more important one in the
production of specks.
FIGURE 19. Top speck on the float glass ribbon.
TABLE 3. SOLUBILITY OF OXYGEN IN TIN
temperature/0C 1000 800 600

solubility (parts/106 by weight) 630 95 5.4
After Belford & Alcock (I964), 8000C and 10000C values extrapolated.
A second source of trouble appears at the cooler end of the bath where the oxygen
content of the tin can quite easily reach its saturation value with respect to stannic
oxide. The solubilities shown in table 3 show that there is a tendency for the oxide
to be thrown out .of solution as a solid as the temperature falls, if the oxygen content
of the tin is above 5 parts per million. The resulting stannic oxide dross can cause
mechanical damage to the lower surface of the ribbon.
A third problem is that as the oxygen content of the tin rises, the ribbon absorbs
increasing amounts of stannous tin. Subsequent heat treatment at about 6000C
or above, as in the toughening of glass for car windscreens, brings about oxidation
of stannous tin to stannic tin with absorption of oxygen from the furnace atmosphere.
This causes a bluish haze, which we call 'bloom', on the surface of the glass (figure
20). Bloom is a microscopic wrinkling of the surface arising from expansion of the
surface layers as the oxygen is absorbed. A float bath must be operated to produce
glass of such a low tin content that bloom is not formed during commercial toughen-
ing or bending.
It seemed obvious from an early stage that the best way to control these faults
was to reduce the concentrations of sulphur and oxygen impurity in the bath.
FIGuRE 20. Bloom on the surface of float glass.
If there were no impurities, there would be no faults. However, total elimination

of impurities is not an attainable target, for though better and better sealing of the
bath structure progressively reduces oxygen ingress through air leaks, the glass
ribbon itself is a source of contamination. Glass from the melting furnace contains
dissolved sulphur and oxygen compounds which are evolved as water and hydrogen
sulphide at the top surface of the ribbon, while at the bottom surface they react to
give dilute solutions of oxygen and sulphur in the bath tin. Figure 21 shows the cycle
of oxygen contamination within the bath. In bad conditions, the oxygen evolved
into the tin from the ribbon can appear as stannic oxide dross, or be reabsorbed into
the glass together with tin, nominally as stannous oxide, or it can be volatilized as
stannous oxide. All these processes are, of course, undesirable. The alternative means
of escape for oxygen from the tin is by reaction with atmospheric hydrogen, and
this takes place when the hydrogen to water ratio in the atmosphere is sufficiently
high in relation to the oxygen concentration of the tin. The reverse reaction, that is
oxidation of tin by water, takes place if the hydrogen to water ratio is too low,
and the variation of the equilibrium with temperature is such that it is possible
for the atmosphere to cleanse the tin at one end of the bath and contaminate it
at the other. The final results of oxygen contamination in terms of oxygen in
tin and tin in glass are further complicated by the influence of tin flows on the
distribution of contaminant and by the relation of this distribution to the tem-
perature schedule of the ribbon.
The cleansing action of the atmosphere can be enhanced by using a higher per-
centage of hydrogen, but there are practical limits set by the cost of hydrogen and
its tendency to give rise to bubbles under the ribbon, due to hydrogen coming out
of solution.
For a long time the question underlying float chemical work was whether or not
the built-in rate of contamination associated with the ribbon, even assuming
perfect bath sealing, would be low enough to permit consistent operation at a
satisfactory standard of glass quality.
ai laks|
FIGIJIE 21. The cycle of oxygen contamination within the bath.
As regards oxygen contamination, the answer came through the steady improve-
ment in techniques of bath operation until it became obvious that good sealing of
the structure and the improved purity of the atmosphere supplied, could keep the
oxygen concentration in the tin at an acceptable value. In terms of chemical
measurements, the improvement amounted to a reduction from about 10 to less
than 5 parts per million of oxygen in tin and from about 50 to 30 hg of tin in each
square centimetre of glass. We are fortunate that the glass viscosity characteris-
tics do not require a bath temperature of much below 600 ou. At lower tempera-
tures, even with very good sealing, formation of solid oxide would be almost
impossible to prevent because of the very low solubility of oxygen in tin and the
relatively slow reduction by hydrogen.
It is possible to improve the condition of the bath by removing oxygen directly
from the tin. We developed a technique of oxygen removal which depended on the
electrolytic addition of calcium and barium to the tin. These reacted to form oxides
which were either absorbed by the ribbon or removed as dross. The technique was
useful in production at a time when control of oxygen contamination had nearly

been achieved, but the presence of dross was always a disadvantage. This is one of
the many large scale experiments which were undertaken, only to be superseded by a
better answer.
In contrast to oxygen, sulphur does not promote the absorption of tin by the
ribbon, nor does its concentration in the bath tin ever approach the solubility limit.
condensate
H2S SnS vapour
+H2 ~~~~ribbon
S dissolved in Sn
FIGURE 22. The cycle of sulphur contamination.
The sulphur cycle (figure 22) is similar to the oxygen cycle, but there is only oile
important source of contamination (the ribbon) and only one important side effect,
that is, volatilization and condensation of stannous sulphide, leading to formation of
'top speck'-that is specks of tin-on the ribbon. Control of this fault has come
chiefly through a direct attack on the flows and condensation patterns of stannous
sulphide vapour in the bath atmosphere and not through any reduction of the
sulphur introduced into the glass in the furnace.
Success in controlling the chemistry of the process can be measured by the fact
that the incidence of faults produced by the bath is generally less than one per
hundred square metres of glass.
I hope that this brief description of these three problems helps explain why it
took us seven years to get float off the ground.
We have now lived with this process for nearly seventeen years and the scale of
research and development currently being undertaken to exploit the process further
is as large as that in the formative years.
The process is able to make, more economically, 9900 of the demand previously
met by the polished plate process.
Our targets now are twofold. First, to exploit the components of the float process
to make glass with special properties. One example is the electro-float process.
Secondly, the process is now showing signs of becoming the universal process
for manufacturing flat glass. It is our target to exploit this potential with further
research and development.
Our confidence in the investments aimed at achieving these targets is enhanced

by the knowledge that none of the problems we have met in the float process-
however stubborn initially-has refused to yield to research and development.
REFERENCES
Belford, T. N. & Alcock, C. B. I964 Trans. Faraday Soc. 60, 443.

Colin, R. & Drowart, J. I965 Trans. Faraday Soc. 61, 1364.
Espelund, A. W. I966 Acta chem. scand. 20, 1994.
Klushin, D. N. & Chernykh, V. Ya. I960 Russ. J. inorg. Chem. 5, 685.
Sudo, K. I95I Sci. Rep. Res. Inst. Tokohu Univ. 3, 356.

Pilkington Float Glass

Uploaded by

Copyright:

Available Formats

Pilkington Float Glass

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Pilkington Float Glass

Uploaded by

Copyright:

Available Formats

Review Lecture.

The Float Glass Process

Printed in Great Britain

The float glass process

(Delivered 13 February 1969-Received 14 April 1969)

I 1 ] Vol. 3I4. A. (i6 December I969)

FIGURE 1. The crown process spun glass into a disk.

As a result of Pilkington cooperation with Ford this discovery was successfully

FIGURE 7. Continuous grinding and polishing of plate glass came in 1923.

<9 ~~molten tin 1N,\ \A

FIGURE 9. Diagram of the float process.

changed without much difficulty. This is important in reducing warehouse loss by

T2 = (Sg +Sgt-St) gp(

As indicated in figure 14-which shows a vertical section of a pool of molten

thickness T is approximately 7 mm. This value is insensitive to limited changes

suirface tension molten tin

gravity tractive 2 final ribbon

FIGURE 15. Plan view of ribbon formation.

For a normal soda-lime-silica glass it is necessary to maintain a temperature of

w 2.5 IFI abO

For example, by boosting the speed on an early production unit, a reduction in

as the ribbon is stretched. Since the glass behaves as an isotropic incompressible

primary )cooled reheat

FIGURE 17. System for the production of thin float glass.

formed ribbon is progressively reduced and guides extend longitudinally a distance

FIGURE 18. System for the production of thick float glass.

In concluding this description of ribbon formation, may I draw your attention

TABLE 1. CRITERIA DETERMINING THE CHOICE OF A SUPPORT METAL FOR

melting point boiling point estimated vapour

bismuth 271 1680 9.1 27

TABLE 2. VOLATILIZATION FROM DILUTE SOLUTIONS OF OXYGEN

FIGURE 19. Top speck on the float glass ribbon.

TABLE 3. SOLUBILITY OF OXYGEN IN TIN

temperature/0C 1000 800 600

FIGuRE 20. Bloom on the surface of float glass.

If there were no impurities, there would be no faults. However, total elimination

FIGIJIE 21. The cycle of oxygen contamination within the bath.

useful in production at a time when control of oxygen contamination had nearly

H2S SnS vapour

FIGURE 22. The cycle of sulphur contamination.

Our confidence in the investments aimed at achieving these targets is enhanced

Belford, T. N. & Alcock, C. B. I964 Trans. Faraday Soc. 60, 443.

You might also like