Handbook of Materials for Product Design Part 11 pps

8.96 Chapter 8
refractory materials themselves. Crowns are often self-supported re-
fractory structures. The refractories must withstand or resist high
temperatures, heavy loads, abrasion, and corrosion.
Types of oxide refractories used in glass melters are as follows:
1. Clay refractories (generally used for insulation)
■
Fireclay: kaolinite (Al
2
O
3
· 2SiO
2
· 2H
2
O) plus minor compo-
nents; classified as low-, medium-, high-, or super-duty; (25 to
45% Al
2
O
3
)
■
High alumina (50 to 87.5% Al
2
O
3
)
2. Nonclay refractories (generally used for glass contact or furnace
superstructure)
■

Magnesia (MgO)
■
Silica (cristobalite and trydimite)
■
Stabilized zirconia
■
Extra-high alumina (>87.5%)
■
Mullite (3Al
2
O
3
· 2SiO
2
)
■
AZS (alumina-zirconia-silica) containing 30 to 42% zirconia
■
Zircon (ZrO
2
· SiO
2
)
■
Chrome-magnesite and magnesite-chrome (combinations of
Cr
2
O
3
and MgO)

Classifications based on methods of of refractory manufacture (in or-
der of increasing density) are as follows:
■
Bonded
■
Sintered (sometimes densified by cold isostatic pressing before fir-
ing)
■
Fusion cast (cast as blocks from arc-melted raw materials); also re-
ferred to as fused or fused cast
Many factors affect the choice of refractory for a particular applica-
tion. These include melting temperature, thermal conductivity, me-
chanical strength, creep resistance, and resistance to corrosion and
spalling, to name a few. Generally, different refractories are used in
different regions of a glass-melting tank, because the requirements
are different. However, the ultimate design consideration is resistance
to corrosion by molten glass and by the hot gas atmosphere within the
melter. Corrosion determines tank lifetimes and affects the rates at
08Seward Page 96 Wednesday, May 23, 2001 10:16 AM
Inorganic Glasses 8.97
which certain glass defects (such as stones) are generated. In selecting
a refractory the considerations are, in order,
1. glass quality
2. tank lifetime
3. initial cost
Many different corrosion mechanisms operate within a melter. The
types and severity depend on the glass composition, the composition
and microstructure of the refractory, and the temperature. Corrosion
rates tend to increase dramatically with temperature. This is an im-
portant reason why different refractories are often selected for differ-

ent portions of the melter. It is also an important reason why the
harder (higher melting temperature) glasses are generally more ex-
pensive to manufacture. Corrosion types include:
■
Front surface attack (frontal attack). This is direct attack at the
glass/refractory interface. Its mechanisms include alkali diffusion
into the refractory with consequent fluxing and dissolution of the re-
fractory crystals. Porous refractories are generally more susceptible.
■
Melt line corrosion (attack where glass surface, air and refractory
meet). Corrosion at this location is enhanced by localized convec-
tion currents and fluctuations in glass level within the melter.
■
Upward drilling. This form of corrosion occurs where bubbles form
under horizontal refractory surfaces, such as throat cover blocks or
submerged horizontal refractory joints. Corrosive vapor species con-
centrate in the bubbles. As with melt line corrosion, the greatest cor-
rosive activity is believed to take place where the vapor, refractory,
and glass touch.
■
Downward drilling. This type of corrosion results when droplets of
molten metal settle on the bottom of the tank. Sources of metal can
be contaminants in batch raw materials or cullet, chemical reduc-
tion of certain glass components (such as lead oxide), and even tools
or metal parts accidentally dropped into the tank.
Glass contact refractories. The most common glass contact refractories
include the following:
■
Fused AZS (alumina-zirconia-silica). This is the most common today
(41% ZrO

2
in high-wear areas and electrode blocks, as opposed to
less expensive 34% variety). The oxidation state is critical (the re-
fractory contains a residual glassy phase which, if produced in re-
duced condition, will oxidize in use, swell, and exude from the brick).
08Seward Page 97 Wednesday, May 23, 2001 10:16 AM
8.98 Chapter 8
■
Dense sintered zircon (ZrO
2
· SiO
2
). This is used in some low expan-
sion borosilicate melters.
■
Clay, fused alumina, bonded AZS, and dense sintered alumina.
These are used for lower-melting specialty glasses.
Typical glass contact refractories used for melting various glass types
are:
■
Container glass. Fused AZS, life 8 to 10 years; also, fused α-β alu-
mina; finer bottoms are sometimes bonded AZS, zircon, and clay.
■
Float glass. Fused AZS in melting zones; fused alumina in condi-
tioner zones, life 10 to 12 years.
■
Hard borosilicates. Fused AZS and zirconia, sometimes dense sin-
tered zircon.
■
Fiberglass wool. Highly corrosive, melted electrically in fused

chrome-AZS or fused alumina-chrome refractories (coloration due to
chromium is of little consequence in this application).
■
E-glass (textile fiber). Less corrosive, melted in dense sintered
chrome oxide.
■
Lead crystal. Tendency to electric melting in AZS.
Several refractory sidewall design considerations are based on cor-
rosion concerns. First, the thicker the wall, the lesser the heat lost and
the lesser the energy consumed. But thicker walls create a smaller
temperature gradient, allowing chemical attack to penetrate more
deeply into the refractory. Thus, the design thickness of a wall must be
a compromise between heat loss and wear. To give all sections of the
wall approximately equal lifetimes, thickness, and type of refractory
are often varied from location to location. These techniques are some-
times referred to as zoning by thickness and zoning by type.
To avoid melt infiltration of horizontal refractory seams and the con-
sequent increased opportunity for upward drilling of corrosion, glass
contact wall refractories are often large, full-height blocks arranged
adjacent to each other in a “soldier” course fashion. However, if multi-
ple courses are required, close-fitted diamond ground horizontal joints
can help minimize melt infiltration. Similarly, to avoid horizontal
joints, the paver blocks composing the top layer of the tank bottom are
butted up against the side blocks, not placed under them.
Superstructure and crown refractories. The superstructure, which in-
cludes all furnace walls above the melt line and the crown, are subject
to corrosion by aggressive vapor species such as NaOH, KOH, PbO
and HBO
2
, batch dust particles, liquid condensates and liquid reaction

08Seward Page 98 Wednesday, May 23, 2001 10:16 AM
Inorganic Glasses 8.99
products running down from refractories higher up. Superstructure
temperatures in fuel fired furnaces are often 60-100°C hotter than the
glass. Typically, walls and ports made of fused AZS from the back wall
(where the batch enters) to the hot spot; fused β-alumina is used
downstream. But these are not hard and fast rules. Crowns typically
consist of sintered silica block. Although silica crowns are attacked by
alkali vapors, the drips are homogenized into melt. Crown life is more
of a concern. This is especially true with gas-oxy firing (more aggres-
sive vapors, see Sec. 8.3.3.5), in which case more costly refractories
may be justified. Silica and alumina blocks should never be in direct
contact, for example at the joint where the crown and superstructure
walls meet (they will react); zircon is used as a buffer.
Regenerative heat exchanger refractories. Special refractory consider-
ations are needed because of the large temperature gradient (top–bot-
tom) within the checker chambers and the corrosive nature of the
exhaust gases. As the gases cool, a temperature is reached at which
the corrosive vapors condense on the refractory surfaces, enhancing
the corrosion. Fine batch particles carried over with the exhaust gases
also tend to react with and corrode the regenerator refractories. High
thermal conductivity and heat capacity are also important character-
istics. Checker construction for a typical soda-lime-silica melting tank
is: top third, bonded 95 to 98% MgO bricks; middle third, lower mag-
nesia content bricks; bottom third (where alkali vapors condense), sin-
tered chrome (chromic oxide) or magnesium-chrome bricks.
A relatively new approach to checker construction, especially in Eu-
rope, is with special interlocking shapes (cruciforms) of AZS or high-
alumina fused cast refractories.
8.3.3.4 Electric boosting and all-electric melting. Crucibles, pots, and

day tanks for glass melting can be heated in electric furnaces where
the heat is generated by resistance heating in windings or bars. In
come cases, heat is produced by the flow of electricity through the
metal crucible itself, and in others by the flow of electricity through
the molten glass, which is a moderately good ionic conductor at high
temperatures, between submerged electrodes.
These principles are applied to varying degrees in large continuous
melters as well. In some fuel-fired furnaces, electrodes are installed in
the walls below the glass line so as to provide a source of heat below
the batch layer, thus the batch is actively melted from below as well as
above. By such means, the melting rate for the tank can be increased,
or boosted, leading to the term electric boosting. Resistance heating by
external windings or bars is often used to control temperatures at the
orifice or delivery tube by which the molten glass leaves the melter.
08Seward Page 99 Wednesday, May 23, 2001 10:16 AM
8.100 Chapter 8
Sometimes, all the heat required for melting is supplied electrically
within the molten glass. In this case, electrodes are positioned as
plates at the walls of the furnace, or as water-cooled metal rods ex-
tending upward through the bottom of the furnace. The electric cur-
rent passes from electrode to electrode, through the glass, with the
amount of heat generated depending on the applied voltages, the
shape and spacing of the electrodes, and, very importantly, the electri-
cal resistivity of the molten glass. In this case of all-electric melting,
the batch components are melted solely by heat flow up from below. It
is possible and desirable to maintain a continuous layer of batch
across the top of the melt, eliminating the need for a refractory roof or
crown to contain and reflect the heat. However, in many cases, the
crown is there for other reasons, but it is cold in temperature. Hence,
all-electric melting in this manner is sometimes referred to as cold-top

or cold-crown melting.
It should be noted that the electrical resistivity of a pile of non-
melted batch, or even a glass melt at temperatures well below
1,000°C, is too great to allow for efficient heat generation. Conse-
quently, all-electric melters are started in a more or less traditional
way using fossil-fuel burners and a hot crown. Once the molten glass
has reached sufficient temperature, the burners (and sometimes the
crown) are removed.
Cold-top melting is valuable for two reasons. First, the batch layer
acts as a thermal insulating blanket (the batch blanket), which helps
reduce heat loss out the top of the melter (thus enabling it to operate
cold-top). Second, the top layers of the batch blanket, being much
cooler than the molten glass below, act to condense volatile vapor spe-
cies that might otherwise escape into the atmosphere. This is espe-
cially valuable for fiberglass and other specialty glass melting where
the compositions contain fluorides and other very volatile, and some-
times unhealthy, components.
Electrodes are made of materials such as carbon, tin oxide, molybde-
num, and platinum, with the choice depending on the temperature of
operation and the composition of the glass being melted. Melt temper-
atures, temperature gradients, and convection currents are greater at
and near the electrodes; therefore, better refractories (higher temper-
ature, more corrosion resistant) are required at these locations.
8.3.3.5 Oxygen for combustion. Over the past decade, there has been
a trend toward the use of oxygen instead of air, in combination with
natural gas, to heat fuel-fired furnaces. This is called oxy-fuel firing. It
has several advantages. First, since one is not using air with its 80%
nitrogen content, much less polluting NO
x
gases are produced. In the

08Seward Page 100 Wednesday, May 23, 2001 10:16 AM
Inorganic Glasses 8.101
face of increasingly stringent air quality legislation, this factor alone
is often sufficient to justify conversion to oxy-fuel firing. Second, since
there is not the large volume of nitrogen to heat and expel from the
furnace, much less waste heat is generated than with gas-air. Some
operations have been able to eliminate the massive, costly regenera-
tors as a consequence. Third, higher flame temperatures are possible.
Fourth, as claimed by some manufacturers, more stable furnace oper-
ation, with an associated improvement in glass quality, is achieved.
This is especially the case when regeneration, with its inherent peri-
odic reversals of gas and heat flow, is eliminated.
There are some disadvantages to oxy-fuel firing. One is the need for
liquid oxygen storage or oxygen generation on site. A second is that,
without the large volumes of air moving through the furnace, the con-
centrations of water vapor (a product of combustion) and corrosive vol-
atile species from the melt (e.g., NaOH) are much higher, in some cases
leading to increased deterioration of the refractory superstructure.
8.3.3.6 Furnaces for specific applications. Furnaces designed for spe-
cific applications include the following:
■
Container glass. Typically cross-fired regenerative; maximum melt
temperatures ~1600°C; large, up to 500T/day.
■
Float glass. Typically cross-fired regenerative; no bridge wall, but
rather an open surfaced narrow region called a waist to keep inho-
mogeneities running parallel to the surface of the glass sheet; maxi-
mum melt temperatures ~1600°C; larger, up to 800T/day.
■
Fiberglass. Smaller gas-fired recuperative or all-electric.

■
Lead crystal. Small electric boosted or all-electric.
■
Hard borosilicates. Tending to all-electric or heavily boosted regen-
erative; melt temperatures > 1600°C.
■
Aluminosilicate glass-ceramics. Regenerative gas-fired; tempera-
tures near 1700°C required for efficient fining.
■
Optical glass. Small fuel-fired or electric heated; fining and condi-
tioning often done in platinum tubes to avoid refractory contact and
resulting inclusions and inhomogeneity.
With the advent of oxy-fuel, or more specifically gas-oxygen, firing,
many of the above listed regenerative and recuperative furnaces have
been converted to use this new technology. However, as of this writing,
float glass manufacturing is just beginning to convert to gas-oxygen
firing.
08Seward Page 101 Wednesday, May 23, 2001 10:16 AM
8.102 Chapter 8
Trends. As is typical in most industries, new designs are aimed at
lower overall cost of operation (the calculation of which includes initial
cost, melter lifetime, and costs of repairs as well as the daily operating
costs) less energy consumption and less overall environmental impact.
Lower cost almost always must be achieved in combination with im-
proved glass quality and less adverse environmental impact.
8.4 Glass Making II—Glass Forming
The term forming collectively refers to all the processes of glass mak-
ing used to form a solid object or product from the molten glass. His-
torically, all glass objects were formed by hand using relatively simple
implements. Over time, the techniques were modified, automated, and

scaled up. While several glass forming methods in use today have no
precedent in early glass history, most still bear important resemblance
to their forbearers. Due to space limitations here, this section will de-
scribe only processes used in today’s manufacturing plants and, when
relevant, the early hand-forming operations. Little attention will be
paid to the many processes that have intervened. We will first discuss
processes involving molds.
8.4.1 Blowing
By far, containers (bottles and like products) account for the largest
volume of glass production. Almost all these products are manufac-
tured using some form of a blowing process.
Historically, glass containers have been blown to shape by gathering
a gob of molten glass on the end of a hollow iron pipe, the blowpipe or
blowing iron, and blowing a puff of air into the soft glass to form a
bubble, which is gradually expanded and worked into shape by the
combined effects of gravity and the forces of tools pressed against it.
Generally, the blowing iron, with the soft glass attached, is rotated to
balance the effects of gravity and provide an axial symmetry to the
product. While useful containers of remarkably repeatable shapes and
dimensions can be created in this manner, for rapid and precise pro-
duction, it is preferable to use a two-step process. First, a hollow pre-
form, called a parison, is prepared using a simple blowing process.
Second, the parison is blown to the final shape in a mold.
This process has been automated to a very high degree in modern
times, to the point where more than a dozen containers per minute
can be generated from each mold. Generally, rotating split molds are
used for shapes involving bodies of revolution whenever visible seam
lines from the molds are undesirable, such as for light bulbs or high-
quality drinkware. Stationary split molds must be used for containers
08Seward Page 102 Wednesday, May 23, 2001 10:16 AM

Inorganic Glasses 8.103
having handles, flutes, or other nonrotationally symmetric shapes.
The rotating split molds are generally paste molds, called that because
their molding surface is coated with a thin layer of cork or similarly
permeable substance, which is saturated with water after each mold-
ing cycle. When the mold surface is contacted by hot, molten glass, a
steam layer results, which provides a low-friction layer between glass
and mold, giving the product a highly polished appearance without
seam lines.
The stationary molds are generally hot iron molds. These metal
molds are operated at a temperature hot enough to keep the molten
glass from being chilled so quickly that surface cracks or checks result,
but cold enough to quickly extract heat from the glass and allow it to
become rigid before removal. Any metal mold surface defects, as well
as the mold seam lines, are transferred to the ware, but production
rates can be much faster than with paste molds. Also, on the plus side,
intentional designs such as logos can be molded or embossed into the
glass surface.
When blowing by a hand-type operation, the final product must be
separated from the blowing iron, usually by cracking it off. This leaves
a rough surface that must be properly finished by grinding or fire pol-
ishing, a process step that involves locally reheating the glass to a
point at which it will flow to a smooth surface under the influence of
surface tension. In modern automated container production, free gobs
of glass are handled in the molds, so separation from a blowing iron is
not required. Two common processes are called blow-and-blow and
press-and-blow, depending on the method used to form the parison.
Blow-and-blow is generally used for narrow-neck containers such as
beverage bottles. The parison is blown in one mold in a way that forms
the neck and then, held by the relatively cold newly formed neck, it is

transferred into a second mold to blow the body of the container. One
of the more common machines featuring these operations is Hartford
Empire’s (now Emhart Corporation’s) Individual Section (IS) ma-
chine. This mechanism may have as many as 12 sections driven in
tandem by a cam with overlapped timing or, more recently, by elec-
tronically synchronized operation. Each section operates on as many
as four gobs. Processing speeds are about 10 s per section. In addition
to speed, an advantage of the IS machine (as opposed to a rotating tur-
ret machine) is that the machine can be programmed to run the re-
maining sections while one is being repaired. The operation of a single
two-mold IS section is shown in Fig. 8.12.
Press forming of the parison before blowing to final shape is used for
wide-mouthed containers such as food jars. Press forming will be de-
scribed in the next section. For container manufacture, while pressing
of the parison is complicated by the need for an additional tool (the
08Seward Page 103 Wednesday, May 23, 2001 10:16 AM
8.104 Chapter 8
plunger), this disadvantage is offset by yielding a product of more uni-
form wall thickness, hence a more efficient utilization of glass and a
lighter-weight product than produced by blow-and-blow.
A very high-speed process for blowing light bulb envelopes and the
like, known as the ribbon machine, was developed in the 1920s by
Figure 8.12 The H.E. IS (individual section) blow-and-blow machine. The gob is
delivered into a blank mold, settled with compressed air, and then preformed with
a counter-blow. The parison or preform is then inverted and transferred into the
blow mold where it is finished by blowing.
27
08Seward Page 104 Wednesday, May 23, 2001 10:16 AM
Inorganic Glasses 8.105
Corning Glass Works (now Corning Inc.) and is still in use worldwide.

In this machine, a stream of molten glass is continuously fed between
a set of rollers, one flat and the other with pocket-like indentations.
These rollers form a ribbon of glass several inches wide, containing
regularly spaced circular mounds of glass down the centerline. The
parison for each light bulb is formed by inserting a synchronously
moving blow head (analogous to a blowpipe) into each mound of glass
and blowing it through a synchronously moving orifice plate. As the
ribbon travels horizontally along the machine, the parison is enclosed
in an also synchronously moving rotating paste mold, and the blowing
process is completed. The moving molds open and swing away to allow
the finished glass envelope to be cracked off the ribbon at the machine
exit. The operation of the ribbon machine is illustrated in Fig. 8.13. In-
candescent lamp envelopes (for example A-19, 60-W bulbs) can be
made at speeds in excess of 1,200 per minute on a single machine us-
ing this technique. Small automotive and other specialty lighting
bulbs can be made at rates exceeding 2,000 per minute.
8.4.2 Pressing
In simplest terms, pressing or press forming of glass involves placing a
gob of molten glass in a hot metal mold and pressing it into final shape
with a plunger. Sometimes a ring is used, as illustrated in Fig. 8.14, to
limit the flow of glass up the side of the mold and produce a rim of well
Figure 8.13 The “ribbon machine” used for light bulb envelope manufacture.
U.S. patent 1,790,397 (Jan. 27, 1931), W. J. Woods and D. E. Gray (to Corning
Inc.). (Courtesy of Corning Inc.)
08Seward Page 105 Wednesday, May 23, 2001 10:16 AM
8.106 Chapter 8
controlled shape. The process steps can be performed entirely by hand
or fully automated. It produces more accurate and controllable wall
thickness distributions than blowing but is generally limited to open,
moderately shallow articles such as dinnerware, cups, baking dishes,

sealed-beam headlamp lenses, and television panels and funnels, or
for solid objects. Pressing is capable of generating intricate and accu-
rate patterns in the glass surface, such as found in sealed beam spot-
Figure 8.14 Pressed glass, mold types and pressing operations. (Courtesy of McGraw-
Hill)
28
08Seward Page 106 Wednesday, May 23, 2001 10:16 AM
Inorganic Glasses 8.107
light, floodlight, and automotive headlamp lenses and in street and
traffic light refractors and lenses. Large objects, such as 27- and 35-in.
(diagonal) color television bulb panels weighing more than 25 pounds,
can be made using automatic pressing equipment.
Glass is generally pressed at a viscosity between 2000 and 3000 P
with an applied pressure of about 100 lb/in
2
of article surface area. For
large television panels, the total force on the plunger can exceed 20 T.
Temperature control of the molds and plunger is crucial; too cold leads
to brittle fracture of the glass under the pressing forces, and too hot
leads to sticking of the glass to the mold surface, requiring it to be
physically broken free. Vents within the mold body, through which
cooling air or water may flow, are often used to maintain uniform tem-
perature distribution across the mold surface.
8.4.3 Casting
Casting is a relatively little used process, found mostly in hand shops
and for the production of very large pieces of glass such as glass sculp-
tures and astronomical telescope mirrors. For the large pieces, glass is
poured into hot ceramic refractory molds (often sand with a small
amount of binder) that are slowly cooled after the mold is completely
filled. Alternatively, chunks of rigid glass may be placed in a cold mold

and raised in temperature until the glass is sufficiently fluid to flow
and fill the mold. This latter method is more susceptible to entrap-
ment of bubbles. Generally, slow cooling and long annealing times are
required. The mold can be used only once. The glass surfaces in con-
tact with the mold are generally rough.
8.4.4 Centrifugal Forming
Centrifugal forces have often been utilized by the glassmaker. A glass
bubble on the end of a blowing iron can be elongated by swinging the
iron back and forth to aid gravity in elongating the bubble to generate
the parison. A thick-walled bubble on the end of a rod can be cut open
at the point opposite to the rod, and the rod rotated to generate suffi-
cient centrifugal force to open the bubble and spin it into a relatively
flat, circular sheet of glass. This is one of the earliest flat glass manu-
facturing methods, the crown process. Glass made this way is often
found in old European churches. A droplet of very fluid glass placed at
the center of a rotating turntable will also spread under centrifugal
force, a process utilized in spin coating or spin casting. The latter is
sometimes simply called spinning.
If molten glass partially fills a rotating container such as a mold or a
crucible, the molten glass will tend to climb the walls, propelled by the
08Seward Page 107 Wednesday, May 23, 2001 10:16 AM
8.108 Chapter 8
centrifugal forces, giving the glass surface the shape of a paraboloid of
revolution. This method, called centrifugal casting, is used to form the
parabolic shapes for thin astronomical telescope mirrors. It has also
been used to spin, rather than press, large, deep television tube fun-
nels and glass-ceramic missile radomes. Six- to 8-meter diameter mir-
ror blanks are spun at about 10 rpm, a television bulb funnel at about
200 rpm. In the 1960s, Corning Glass Works used this spinning pro-
cess to make large, 56-inch diameter glass hemispheres of 1.5-inch

wall thickness for use in undersea exploration.
A continuous centrifugal process of forming tubing from very short
(steep viscosity) glasses or easily devitrifiable glasses has been de-
vised. It is somewhat analogous to the Danner tubing process (see be-
low) except that the stream of glass is fed into the open end of an
inclined rapidly rotating pipe. The very fluid entering glass is flat-
tened against the wall and the adjacent layer of previously deposited
glass and is maintained in position by centrifugal forces until it is
cooled to sufficient rigidity to be withdrawn from the end of the pipe.
8.4.5 Rod and Tube Drawing
Drawing is the term for a process in which a preshaped blank, or glass
flowing from an orifice, is elongated (stretched) in one dimension while
diminishing in orthogonal dimensions without losing its cross-sec-
tional characteristics.
The above statement is exactly true for the drawing of cane (rods) or
fiber. It is not so for tubing or sheet, where the ratios of inside to out-
side diameter or width to thickness are not the same as they were at
the root. (The solid section of the blank or the glass at the orifice is of-
ten referred to as the “root.”)
Redrawing is the specific case of drawing from a solid preform (or
blank) rather than from a melt. This involves reheating the end of the
blank to provide glass sufficiently fluid to be stretched and attenuated.
In a continuous process, the blank is replaced at the volume rate at
which it is used up by gradually feeding it into the hot zone of the re-
draw furnace.
In a steady-state process, the volume flow per unit time, the quan-
tity Q, is constant and equals A (area) × v (velocity) at any point in the
process. As we will show, A is a very important parameter in the tube
drawing process.
Tube drawing processes. In a hand process, a gob of glass is gathered

on the end of a blow pipe, a bubble is blown within the glass, and an
assistant attaches a rod to the side of the gob opposite the blow pipe
(or grabs the gob with a pair of tongs) and walks across the room to
stretch out the glass and the bubble within it. The final diameter of
08Seward Page 108 Wednesday, May 23, 2001 10:16 AM
Inorganic Glasses 8.109
the resulting tubing, and its wall thickness, depend on several factors,
including how fast the assistant walks (compared to how rapidly the
glass cools) and how much pressure the blower maintains in the bub-
ble. If faster cooling is needed, a second assistant may fan the tubing
as it is drawn out. The air pressure resists tubing collapse from the
draw forces and surface tension. After the drawing step is completed,
the hollow glass tubing is cut away from the bulky pieces at each end.
There is only about 10 to 20% glass utilization. The rest of the glass
remains on the blowpipes or is of unusable dimensions and is gener-
ally recycled as cullet. The process is highly labor intensive.
The Danner process, named after its inventor, Edward Danner, was
developed by the Libbey Glass Company. It is one of the oldest contin-
uous tubing drawing processes still in use today and is the common
method for forming fluorescent lamp tubing. The process is somewhat
unique in the manner of preparing the root of glass from which the
tubing is drawn. See Fig. 8.15. A ribbon-shaped stream of glass is fed
onto a slowly rotating (~10 rpm) hollow clay (or metal) mandrel, in-
clined downwards perhaps 15° from horizontal. The glass stream flows
onto the mandrel at a viscosity about 1,000 P, wraps around the man-
drel, and overlaps itself to form a cylinder, which is smoothed by the
forces of gravity and surface tension. The glass cools (to a viscosity of
Figure 8.15 Rotating mandrel used in Danner tube drawing.
08Seward Page 109 Wednesday, May 23, 2001 10:16 AM
8.110 Chapter 8

about 50,000 P) as it moves along the mandrel and is drawn off the
mandrel end in a horizontal direction forming a catenary. Air is fed
through the mandrel, so the latter acts somewhat like a hand blow
pipe with a continually replenished supply of glass. The tubing is
drawn (stretched to smaller dimensions) by a tractor device located
many feet from the mandrel and is cut into lengths after it passes
through the tractor.
The Danner process is capable of drawing 1/16- to 2.5-inch diameter
tubes. Because of temperature nonuniformity, coupled with gravity ef-
fects, Danner tubes often exhibit some ovalness and wall thickness
(called siding) variations. Solid rods of similar diameters can be
drawn by stopping the airflow through the mandrel or even drawing a
slight vacuum. Composition variations can produce hairpin-shaped
cord defects.
The Vello process, after inventor Sanchez Vello, was developed by
Corning Glass Works and dates back to the early part of the 20th cen-
tury. Here, the glass is delivered from the glass-melting furnace, at
about 100,000 P viscosity, through an annular orifice created by the
spacing between a conical bowl and a bell-shaped blowpipe, called the
bell, centered within the bowl. See Fig. 8.16. The drawn tubing first
extends vertically downward then turns horizontally, following a cate-
nary curve as it stretches under its own weight, then is transported on
a runway of “V” rollers often several hundred feet long as the glass
cools. As with Danner, the tubing is cut into lengths after it passes
through the tractor at the end of the runway.
The Vello process allows drawing of precision-bore tubing, such as
for thermometers and burettes. It is fast (for example, 800 52-inch
sticks/min at a 2000 lb/hr flow rate) and can draw tubing of diameters
up to 3 in. without significant oval.
Control of diameter and wall thickness is based on a mathematical

equation relating volume flow of glass, tubing velocity, and tubing
cross-sectional area, which determines how fast one must run the
tractor pulling the tubing to give the requires cross-sectional area of
the glass in the tubing. This area is given by
where Q = volume flow of glass from the melter
v = speed of the tractor
w = wall thickness
D = outer diameter of the tubing
Only in the case of a rod, where w = D/2, is the diameter uniquely de-
termined by the pulling speed. In all other cases, the diameter and
Q
v
-
A πwD w–()==
08Seward Page 110 Wednesday, May 23, 2001 10:16 AM
Inorganic Glasses 8.111
wall thickness interact, and their ratio is maintained within specifica-
tion by the pressure of the air flowing through the tubing, just as with
hand drawing.
The downdraw tubing process is essentially a vertical Vello. The
glass tubing is cut off one or more floors below the draw. It is useful for
tubing that is too large in diameter or wall thickness to be successfully
turned horizontally (i.e., without breaking or deforming from cylindri-
cal shape). For example, Corning has drawn 6-in. dia., 3/8-in. wall
borosilicate tubing for Pyrex
TM
brand pipe by this process. Molds and
dies can be added to give controlled cross-sectional shapes.
The updraw tubing process is analogous to the updraw processes,
flat or cylinder, used to manufacture sheet glass, which will be de-

scribed in the next section. The air pressure needed to keep the tube
bore open is supplied from below, through a refractory cone positioned
in the melt just beneath and on axis with the drawn tubing. The
height of the cone helps control the tubing wall thickness. Continuous
updraw of thermometer tubing with enclosed colored glass ribbons is
only slightly more complicated.
Figure 8.16 Bell and bowl arrangement used in Vello tube drawing process. From lec-
ture notes, E. H. Wellech, Corning Glass Works, 1963. (Courtesy of Corning Inc.)
08Seward Page 111 Wednesday, May 23, 2001 10:16 AM
8.112 Chapter 8
8.4.6 Sheet Drawing
Sheets of glass can be drawn either upward or downward from a bath
of molten glass by a tractive mechanism, provided a method can be
found to maintain the root of the draw in a fixed position at constant
dimensions. For a downdraw process, the root is essentially a rectan-
gular slot in the bottom of the melter. Molten glass flows from this slot
under the combined effects of hydrostatic pressure from above and
tension from below. Below the melter, the sheet is cooled to become
rigid, after which its weight is supported by pairs of horizontal rollers.
As the glass is cooled by radiation beneath the slot-shaped orifice, the
rollers stretch it to the desired final thickness. The sheet also tends to
become narrower in width during stretching, but this effect can be
minimized by the judicious use of edge coolers to help hold the edges of
the sheet out. These edgewise forces, while maintaining the desired
sheet width, also help maintain its flatness. One is essentially pulling
on a stretched membrane. The formed sheet is lowered through a
heated annealing zone (annealer) and cut into separate sheets below,
often in a subbasement or a specially excavated pit.
A disadvantage of the process is that any nonuniformities in the
slot, such as might be caused by erosion or corrosion, lead to vertical

streaks in the glass surface. In part to overcome this difficulty, and in
part to have better thickness control over the resulting sheet, Corning
Glass Works (now Corning Inco.) developed their fusion downdraw
process. In this process, molten glass is fed into one end of a slightly
inclined refractory trough at a viscosity of about 40,000 P and allowed
to overflow both sides, as shown in Fig. 8.17. (Sometimes this trough is
called the overflow pipe for obvious reasons.) The outside of the trough
tapers to a line at the bottom where the two layers of overflowing liq-
uid meet and fuse together, forming the root of the draw, hence the
name fusion. The outside surfaces of the glass are generated from
within the interior of the melt and are therefore never subsequently in
contact with other materials; thus, they are pristine and defect free. A
key element of Corning’s initial patent for this process is the mathe-
matical design of the tapering cross-sectional profile of the trough
which, in combination with the incline of the top of the trough, assures
that the volume flow of glass over the pipe is uniform along its length.
Along with a method to precisely control the temperature along the
root of the draw, this assures uniform sheet thickness across its width.
Below the root, the process is somewhat similar to a downdraw from
a slot. One notable difference is that the glass at the root is far more
viscous. Having been cooled to a viscosity of 500,000 P or more as it
descended the tapered lower refractory of the trough, it must be pulled
downward with greater tensile stress. The pulling forces are provided
08Seward Page 112 Wednesday, May 23, 2001 10:16 AM
Inorganic Glasses 8.113
by the weight of the sheet and edge rollers only, so the glass surface is
not subject to roller marking. A wide range of thicknesses may be
made, ranging from less than 1 mm to greater than 1 cm. This process
is currently used to manufacture thin, flat glass of exceptional quality
for active matrix liquid crystal displays (AM-LCDs) for flat-panel com-

puter screens and televisions.
A variety of updraw processes have also evolved. The Fourcault pro-
cess, invented in 1910 by the Belgian Emile Fourcault, uses a partially
submerged refractory block containing a long machined slot to form
the root of the draw. This block is called the debiteuse. The molten
glass that forms the root of the draw is forced up through the slot by
the buoyant force of gravity. The draw is started by lowering a metal
mesh, called the bait, to the slot and then, once it is wetted by the
glass, drawing it upward to form the sheet. Once the process has been
started, the sheet of glass is pulled upward by pairs of horizontal roll-
ers, drawn through an annealing zone, and cut into separate sheets
above. The bath of molten glass is held at temperatures providing a
viscosity of about 100,000 P. Water-cooled edge rollers, located some-
what above the melt surface, help prevent the sheet from narrowing in
width as it is pulled upward. This process is still in use throughout the
world for the manufacture of window glass. Process disadvantages in-
clude vertical streaks, called peignage or music lines, caused by ero-
sion of the slot in the debiteuse.
The Pennvernon process, developed by PPG, uses a fully submerged
draw bar to control the location and straightness of the root. Viscous
Figure 8.17 Illustration of Corning’s “fusion” overflow sheet drawing
process. (Courtesy of Corning Inc.)
29
08Seward Page 113 Wednesday, May 23, 2001 10:16 AM
8.114 Chapter 8
forces confine the drawn sheet to a position over the center of the bar.
In the Colburn-LOF process, the soft glass is bent 90° over a roller a
short distance above the melt surface. This roller, along with water-
cooled edge rollers, serves to keep steady the position where the glass
is drawn from the melt surface. In all these updraw processes, the

pulling rollers and the bending rollers, when used, tend to mar the
glass. Another disadvantage of updraw is that whenever the sheet
breaks in the rolls, the broken glass falls back into the machine and
into the melt. The updraw processes are more difficult to restart than
are the downdraw processes. With the possible exception of Fourcault,
these updraw processes are in relatively little use today, primarily be-
cause they have been superseded by float (see Sec. 8.4.8).
8.4.7 Rolling
Rolling can be used to manufacture thin or thick sheets of glass. In the
simplest form, a puddle of glass can be poured onto a metal table, and
a metal roller is used to spread it to a constant thickness. Generally,
parallel spacer bars are used to limit the thinness to which the roller
can spread the glass. In a continuous process, a stream of glass is fed
between a pair of counter-rotating water-cooled metal rollers. The
spacing between the rollers determines the thickness of the resulting
sheet; the width is controlled by the amount of glass fed to the rollers.
The sheet is generally rolled vertically downward, or at a sufficient in-
cline that a puddle of molten glass can be maintained at the entrance
to the rollers. The speed of the operation and the diameter of the rolls
must be sized to the thickness of the sheet being rolled. The glass en-
ters the rolls relatively fluid, but it must be considerably less fluid
when it leaves so as to maintain its shape. The thicker the glass, the
more heat that must be removed by the rollers. Generally, thick sheet
requires large rollers, maybe 4 to 10 ft in diameter, and forming rates
of only a few feet per minute. Very thin ribbon can be made with small
rollers, a few inches in diameter, at rates of several feet per second. Af-
ter rolling, the continuous sheet is transported horizontally through a
heated annealer.
The surface finish quality and thickness uniformity is generally in-
sufficient for mirror, automotive, and architectural applications. Prior

to the development of the float process, described below, rolled glass
sheet was ground and polished on both sides, sometimes simulta-
neously, to meet these requirements. These products were known as
plate glass. The process was inherently wasteful and expensive.
Patterned glass is made by applying texture to one or both glass sur-
faces using suitably embossed rollers. Applications include shower
doors, furniture tops, room dividers, and windows. There is even an
08Seward Page 114 Wednesday, May 23, 2001 10:16 AM
Inorganic Glasses 8.115
application for roller-applied Fourcault-type sheet texture for use in
restorations of nineteenth century homes.
As a variation, wired glass, such as used in fire doors and building
skylights, can be manufactured by continuously feeding wire mesh be-
tween the rollers along with the molten glass.
8.4.8 The Float Process
Grinding and polishing of rolled glass was very expensive, labor inten-
sive, and wasteful of materials. In the 1950s and 1960s, the Pilkington
company in England developed a much more economical process based
on floating a continuous ribbon of molten glass on a bath of molten tin
as the glass cooled and solidified. This process is illustrated in Fig.
8.18. A detailed description of the process and the difficulties that
were overcome in its development lie outside the intent of this hand-
book, but some key points should be made. The glass product, known
as float glass, has excellent surface properties, the upper surface hav-
ing flowed freely without contact with rollers or any other forming de-
Tweel
Lipstone
Glass
Tin
Reducing atmosphere

Refractory
Metal shell
Refractory
Metal shell
Glass
Canal
Tin
(b)
(a)
Figure 8.18 Float process: Pilkington-type tin bath: (a) side view and (b) top view.
30
08Seward Page 115 Wednesday, May 23, 2001 10:16 AM
8.116 Chapter 8
vices before its solidification, and the lower surface similarly having
been in contact only with a flat, smooth liquid metal surface that was
incapable of marring it. The product also has exceptionally uniform
thickness.
Regarding thickness, a freely spreading puddle of glass suspended
on molten tin (by buoyant forces in a gravity environment) will reach
an equilibrium thickness determined by the tin and glass densities
and the various surface and interfacial tensions. For soda-lime-silica
glass on tin in the Earth’s gravity, this thickness is between 6 and
7 mm, approximately that of traditional plate glass. Several tech-
niques have evolved to make thinner and thicker float glass. These es-
sentially involve pulling the glass off the bath at a rate faster or
slower than would maintain the above-defined equilibrium thickness,
and doing so in a manner that preserves the thickness uniformity.
Several techniques for “stretching” the glass in this manner have
evolved. All employ gripping the edge of the spreading glass puddle on
its top surface with knurled rollers to assist, restrict, or redirect the

glass flow. One method is illustrated in Fig. 8.19.
It should be noted that many early references to the float process de-
scribe the glass sheet as being formed by rolling between two rollers
before it is fed onto the tin bath. This approach was tried initially by
Figure 8.19 Decreasing the thickness of float glass by both lateral and longitudinal
stretching, with knurled wheels pressing on the edges of the ribbon.
31
08Seward Page 116 Wednesday, May 23, 2001 10:16 AM
Inorganic Glasses 8.117
Pilkington, but it proved unsuccessful. (It led to surface defects in the
glass.) In the commercialized processes, the molten glass is fed onto
the tin bath by flowing it over a refractory block. The Pilkington and
PPG designs differ in how this is done.
The Pilkington and PPG float processes have proven so effective and
so economical that they have virtually replaced all plate glass manu-
facture (ground and polished rolled sheet) throughout the world, and
most of the drawn sheet products as well. Most of the flat glass pro-
ducers in the world have been licensed to use the float process. Today’s
largest float glass plants can produce about 1,000 T of finished glass
per day at widths up to about 12 ft and thickness between about 2 and
25 mm. The overall length of the production line, including melter, tin
bath, and annealing lehr can exceed 700 ft, with the tin bath itself oc-
cupying between 100 and 200 of those feet.
Specially designed float lines can produce glass less than 1 mm
thick. While initially developed for the manufacture of soda-lime-silica
glass, several manufacturers have successfully applied the techniques
to borosilicate glasses. However, because of temperature limitations of
the tin bath, and its required chemically reducing atmosphere, not all
commercially useful glass compositions can be manufactured by the
float process.

8.4.9 Fritting
Techniques used for making glass frit (granules) include dry gauging
or dry gaging (drizzling or pouring a stream of molten glass into cold
water) and rolling as very thin ribbon, followed by particle attrition or
comminution (size reduction) processes. These techniques involve a
rapid quenching of the melt and can be used to vitrify (make glass
from) compositions that tend to crystallize readily. Cooling the glass
quickly, directly from the melter, creates high thermal stresses, which
shock and often break the glass into small pieces suitable for charging
into a ball mill. Dry gauging sometimes forms clinker-like pieces that
are difficult to mill and may require an intermediate process step.
Thin rolled ribbon often provides the better, more uniform mill feed.
8.4.10 Spheres, Marbles, and Microspheres
Glass spheres can range in size from a few nanometers in diameter to
a meter or more. They can be solid, porous, or hollow, in all but the
smallest sizes. The application range includes
1. The extremely small solid precursor particles used in Types III, IV,
and V fused silica manufacture
08Seward Page 117 Wednesday, May 23, 2001 10:16 AM
8.118 Chapter 8
2. The small, hollow spheres used to contain fuel in inertial confine-
ment nuclear fusion research
3. The small, <0.2 mm microspheres used in reflective signs and pro-
jection screens or as fillers in plastics and elastomeric composites
4. The ~1/2-in. marbles used in games
5. Fish net floats
6. Deep ocean submersible vessels of military interest
Because of the wide range of size and the need for solid, porous, and
hollow variations, manufacturing techniques necessarily vary. Direct-
ing a high-temperature, high-velocity flame across a vertically de-

scending stream of molten glass can generate small solid spheres. If
the glass is sufficiently fluid, strands of glass are formed that quickly
break apart into droplets. These droplets spheroidize under the forces
of surface tension and cool as they leave the flame. The resulting par-
ticle size distribution is not easily controlled. Similarly, molten drop-
lets can form and detach from an orifice at the bottom of a crucible or
glass-melting tank. If the droplets have sufficiently low viscosity and
are released from a great enough height, they will spheroidize and
cool before reaching the ground, where they are collected. Particle size
control is better.
Microspheres (<0.2 mm dia.) of controlled size and composition can
also be prepared by fritting, sieving, and injecting into a heated region
to remelt, spheroidize, and cool, somewhat as described above. The
precursor particles can be fed into the top of a tall column having a
thermal gradient decreasing downward and collected at the bottom, or
they can be injected into an upward-directed flame, whereby the mol-
ten droplets are drafted to cooler higher altitudes and collected.
To mass produce marbles (solid spheres about 1/2 in. in diameter), a
more viscous stream of glass is delivered from the melter and mechan-
ically cut (sheared) into mini-gobs having the required volume. The
soft gobs fall into the space between two counter-rotating cylinders in
which there are machined opposing spiral grooves. The gobs of glass
are simultaneously rolled into spheres and cooled to temperatures
near their annealing point as they are transported down the length of
the rolls. Streams of different colored glass can be partially mixed to-
gether before gobbing to give variegated appearance. Less spherically
perfect marbles can be generated by dropping the mini-gobs into cylin-
drical holes in vibrating molds, sometimes mounted on a conveyor
belt. Steady vibration, maintained until the gobs have been well
cooled, generates near-spherical shapes. Of course, decorative marbles

of varying sizes can be created as novelties or works of art by studio or
hand-shop techniques.
08Seward Page 118 Wednesday, May 23, 2001 10:16 AM
Inorganic Glasses 8.119
Perfectly spherical spheres of a wide variety of sizes with optical fin-
ish can be produced from near-spherical starting blocks by grinding
and lapping on optical finishing machines.
Hollow spheres of moderate size can be hand blown freely or in
molds, but the location where the blowpipe is separated after forming
is seldom perfect. Hemispheres can be pressed and then pairs fused
together. Very large, thick-walled spheres can be made from pairs of
centrifugally cast hemispheres.
At the other end of the size spectrum, very small, hollow glass
spheres such as used in inertial confinement nuclear fusion (ICF) re-
search are also prepared by a variety of techniques, all relying on gen-
erating small volumes of precursor materials that are injected into a
hot zone. The surface of the precursor particle melts or otherwise re-
acts to generate a viscous liquid layer. As the interior material heats,
it evolves gasses that serve to blow the hollow sphere. Solid precursors
can be formed as small aggregates of batch material, sometimes by
spray drying or sol-gel techniques; alternatively, the required chemi-
cal species can be dissolved in an aqueous or organic solution with pre-
cursor droplets of the required size being generated by an appropriate
means such as an ultrasonic nebulizer.
Porous spheres can be prepared by leaching one glassy phase from
a two-phase spherical product or by processes similar to those used
for hollow microspheres whereby the surface layer never forms in a
fully continuous manner; i.e., the blowing bubbles are exposed at the
surface.
8.5 Annealing and Tempering

8.5.1 Development of Permanent Stresses
in Glass
Stresses in an unconstrained elastic solid develop only if there is a
nonlinear temperature gradient across the body. Such stresses are
temporary, or transient; they exist as long as the temperature gradi-
ent exists. Liquids, on the other hand, cannot sustain shear stresses
for any finite length of time; such stresses relax by viscous flow.
Glasses behave like a liquid when heated into the liquid state; i.e., all
stresses relax due to viscous flow. However, upon cooling through the
glass transition range into the solid state, stresses are likely to de-
velop within the body, and such stresses no longer relax in the absence
of a viscous flow. The various mechanisms for such permanent stress
development are as follows:
1. Cooling from the outside results in a “frozen” temperature gradi-
ent with a higher temperature in the interior (Fig. 8.20). Inner lay-
08Seward Page 119 Wednesday, May 23, 2001 10:16 AM
8.120 Chapter 8
ers continue to relax from fluid flow while the outer layers
gradually freeze. In effect, the stress-free state is a solid with a
temperature gradient. The need to shrink the inside more relative
to the outside at room temperature and the enforcement of the
elastic compatibility criteria between the layers cause the appear-
ance of compression on the outside and tension on the inside. The
magnitude of the stresses so developed is related to the linear ex-
pansion coefficient of the solid. This mechanism is also called the
viscoelastic mechanism.
2. The outside layers cool faster than the inside layers during normal
cooling. Hence, the outside layers tend to possess a “faster-cooled”
structure having a higher volume in the free state (see Ref. 8,
p. 15). This, in principle, is a permanent structural heterogeneity.

Again, the enforcement of the elastic compatibility criteria causes
the outside layers to develop compression and the inside layers to
develop tension. The magnitude of the stresses so developed is re-
lated to the difference between the volumes of the fast-cooled and
the slow-cooled solids.
3. The fact that the various layers travel through the glass transition
range at different instants in time causes the development of a
“frozen fictive temperature gradient.” This is a transient structural
heterogeneity. As in (1), the removal of the transient fictive temper-
Figure 8.20 Simplified concept of permanent stress production in glass due to a “fro-
zen temperature gradient.” (a) Glass with no temperature gradient well above the
transition region has no stress. (b) Temperature gradient develops on cooling; how-
ever, no stress develops because of rapid relaxation while the glass remains well
above T
g
. (c) Cooling to below T
g
while maintaining the same “frozen temperature
gradient” produces no stress yet. (d) Final removal of the temperature gradient pro-
duces stresses that are now permanent.
32
08Seward Page 120 Wednesday, May 23, 2001 10:16 AM