Glass Technology Vol. 46 No. 1 February 2005 www.sgt.

org 11
Glass Technol., 2005, 46 (1), 1119
The chemical composition of container glass has evolved
over many years to become almost generic, providing a
cost effective material which fits its purpose very well.
Container glass compositions are essentially the cheap-
est acceptable formulations in terms of raw materials,
but increasingly stringent environmental legislation may
soon cause changes in the economic balance. A large
body of work has been published regarding melting aids
and new raw materials in order to increase melting effi-
ciency. However there is far less published work dealing
with the specific aim of modifying the chemical compo-
sition of container glass to reduce its melting tempera-
ture, whilst at the same time maintaining its desirable
characteristics. Environmental benefits resulting from re-
formulation may include reductions in melting energy
and lower emissions of CO
2
, NO
x
, SO
x
, particulates and
heavy metals.
This paper discusses the development of a number of
reformulated container glass compositions, with the main
emphasis on the physical properties of the new glasses
and estimated reductions in thermal NO
x
generation.
Compositional changes ranged from single substitutions
to complex alterations involving several components. Com-
puter calculation of the high temperature viscosity of the
new compositions was carried out and the results agreed
very well with measured values. High temperature viscos-
ity, liquidus temperature, chemical durability, thermal
expansion coefficient, density, dilatometric softening
point, glass transition temperature and fluxline corrosion
were measured on a selection of the new glasses. The re-
sults were analysed and evaluated against a set of prop-
erty criteria based on a benchmark glass composition
representative of current UK container glass compositions.
Reformulated glass compositions exhibited reductions
in melting temperature, defined as the log 2 viscosity, of
up to 115C. Several of the new glasses also fulfilled the
benchmark criteria for chemical durability and liquidus
temperature. I t was estimated that the reduced furnace
temperatures which these compositions would enable
could cut generation of thermal NO
x
by approximately
2040%. Reductions in melting energy and emissions of
CO
2
, SO
x
, particulates and heavy metals may also re-
sult, however they were not quantified in this study as
their accurate estimation requires consideration of the
effects of raw materials and cullet. Further work is cur-
rently underway to study these effects and the behav-
iour of the glasses on a larger industrial scale.
Recent publications discussed the effects and benefits
of replacing a small amount of SiO
2
by P
2
O
5
in con-
tainer glass,
(1)
and gave a brief overview of a project to
reformulate container glass with the aim of reducing
its melting temperature and thereby decreasing melt-
ing energy and emissions.
(2)
That project, from which
the work discussed in this paper was also derived, was
the first stage of a series of studies and therefore its
scope was not comprehensive. For example, the effects
of raw materials and cullet additions have not yet been
included in development of the new compositions, yet
they are important to the glass manufacturing proc-
ess. I t is intended that further stages of the project will
study these issues. The reader is referred to the first
stage project report, available in electronic form from
the authors, for full details of the work to date.
(3)
Whilst
the choice of raw materials and cullet levels are major
factors influencing melting energy, particulates and
certain other emissions, it is the oxide composition of
the final glass which determines its high temperature
Reformulation of container glasses for environmental
benefit through lower melting
temperatures
P. A. Bingham
I mmobilisation Science Laboratory, Department of Engineering Materials,
University of Sheffield, Sheffield, S1 3J D, UK
M. Marshall
1
Glass Technology Services Ltd, 9 Churchill Way, Chapeltown, Sheffield, S35 2PY, UK
Manuscript received 5 August 2004
Revised version received 15 November 2004
Accepted 7 December 2004
1
Corresponding author. Email address: m.marshall@glass-ts.com
12 Glass Technology Vol. 46 No. 1 February 2005 www.sgt.org
viscosity. Viscosity is the main parameter governing
the potential reduction in furnace temperature which
can be achieved. Reduction in high temperature vis-
cosity was therefore the focus of the first stage of the
work.
The melting of container glass is an energy inten-
sive process which produces a number of side effects
that are environmentally undesirable. Carbon diox-
ide results from burning gas or oil and generating elec-
tricity, and also from decomposition of some raw
materials. The melting process also causes the emis-
sion of thermally created nitrogen oxides (NO
x
),
which are pollutants. Thermal NO
x
generation is very
temperature sensitive, particularly at temperatures
above approximately 1400C, such as are found in
glass furnaces.
(35)
For example it is estimated that in
a typical glass melting furnace, a reduction in cham-
ber temperature from 1600C to 1500C could reduce
thermal NO
x
generation by up to half.
(35)
This be-
haviour means that even small decreases in furnace
temperature from their current levels could lead to
substantial decreases in thermal NO
x
generation. I n
addition the melting process also generates emissions
of oxides of sulphur (SO
x
), particulates, and heavy
metals, specifically selenium.
UK and European legislation is tougher than ever
before in terms of emissions limits for CO
2
, NO
x
, SO
x
,
particulates and heavy metals.
(6)
This legislation will
affect the economic balance which has hitherto existed
for manufacturers breaches of emission limits will
incur harsh penalties. Glass manufacturers must there-
fore look to new and novel technologies in order to
meet not only the current requirements but also those
projected for the future. The current best available tech-
nologies (BATs) for emissions reduction, as described
by the integrated pollution prevention and control
(I PPC) guidance document,
(7)
consist of secondary
(abatement) and primary methods. Electrostatic
precipitators (EPs), bag filters and acid gas scrubbing
operations are technologies offered for the reduction
of particulates and SO
x
, but these methods have sub-
stantial electricity requirements and therefore gener-
ate CO
2
. They also create waste materials which must
be disposed of, although some manufacturers now re-
cycle their filter dusts. I n addition the installation and
operating costs of abatement technology can be high.
The 3R process developed by Pilkington, which sacri-
ficially burns methane to reduce NO
x
emissions, is
amongst the secondary technologies. However it car-
ries a 65% energy penalty and generates additional
CO
2
. Primary methods of emissions reduction include
a number of combustion technologies such as low-
NO
x
burners and oxy-fuel melting and each carry their
own advantages and disadvantages. Other primary
methods not discussed in the I PPC guidance docu-
ment include chemical reformulation and the use of
alternative raw materials. A sizeable body of work
has been published dealing with the use of new raw
materials to improve the properties or melting behav-
iour of glass.
(814)
These have included materials such
as spodumene, borax and blast furnace slag, the lat-
ter of which is now widely used in UK container and
flat glass manufacture.
Sodalimesilica container glass compositions have
evolved over many years to reach their current form.
They are used extensively today because they strike a
balance between the often competing factors of raw
materials availability, cost, melting and fining charac-
teristics, machine productivity, chemical durability,
optical properties and thermal shock resistance.
(15,16)
By
1932, container glass compositions showed a reason-
able resemblance to those found today.
(1720)
However,
average SiO
2
contents declined steadily from 741 wt%
in 1932 to 717% in 1960. Over the same period, the
intermediate and stabilising oxide content (Al
2
O
3
+
MgO+CaO+SrO+BaO) rose steadily from 91 to 135
wt% and alkalis and minor fluxes (Na
2
O+K
2
O+
B
2
O
3
+F
2
+SO
3
) decreased from 168 to 149 wt%. Dif-
ferences between these glasses and modern composi-
tions are relatively small, however they are not
negligible. Deliberate additions of boric oxide (B
2
O
3
)
at up to 1 wt%, barium oxide (BaO) at up to 07 wt%,
and fluorine (F
2
) at up to 02 wt% were made through-
out the period 19321960. By 1977, B
2
O
3
and BaO had
been removed from US glass compositions, although
they were still used in some European container glasses
in the late 1980s.
(20)
Over the period 1948 to 1960, glass
properties changed slightly, reflecting the changes in
composition. Gob temperatures corresponding to log
3 viscosity fell by about 10C, whilst softening tem-
peratures increased by 5C. These small changes in
composition coincided with greatly improved furnace
efficiencies which were the focus of much effort in the
1960s and 1970s.
Current UK and European environmental legisla-
tion controls the release of CO
2
, NO
x
, SO
x
, Cl, F, Cd,
Tl, As, Co, Ni, Se, Sb, Pb, Cr, Cu, Mn, V, Sn, and NH
4
into the atmosphere as described in the relevant I PPC
guidance document.
(7)
Release of Pb, Cd, Cr
6+
and Hg
into solution is strictly controlled. Clearly the use of
halogens such as fluorides and chlorides, as well as lead
and several other heavy metals and transition metals,
are prohibited by legislation from deliberate addition
at any meaningful level in container glass, and they
were therefore rejected as potential candidates for the
reformulation process.
The physical, chemical and mechanical properties
of container glass are fundamentally important to its
applications. These include:
Viscositytemperature relationship
Glass stability
Thermal properties
Optical properties
Refractory corrosion
Chemical durability
Density.
Volf
(21)
discussed a large proportion of the work
carried out on individual glass components in terms
of their effects on physical properties and behaviour
in sodalimesilica type glasses. The effects of chang-
ing the levels of the main glass components and other
additions can now be accurately predicted based on
the work of Lakatos et al.
(2225)
This was utilised dur-
ing early formulation stages to model prospective can-
P. A. BINGHAM & M. MARSHALL: REFORMULATION OF CONTAINER GLASSES FOR ENVIRONMENTAL BENEFIT
Glass Technology Vol. 46 No. 1 February 2005 www.sgt.org 13
didate compositions and review their expected perform-
ances in terms of viscosity reduction prior to labora-
tory melting trials.
Liquidus temperatures in container glasses are gen-
erally 1020C below the working point (log 4 viscos-
ity), and hence the likelihood of crystallisation during
forming is kept to a minimum. The difference between
forming and liquidus temperatures is an important
criterion in the development of new glass formulations,
and it is defined here as DT
FL
. For example, using the
phase relations in the relevant part of the SiO
2
Na
2
O
CaO system, the replacement of 1% SiO
2
by Na
2
O
would be expected to have only a small effect on
liquidus temperature within the Na
2
O.3CaO.6SiO
2
(devitrite) phase field.
(26)
Viscosity would decrease so
DT
FL
would decrease, possibly becoming negative, hence
there would be an increased risk of devitrification dur-
ing forming. Whether this would actually cause prob-
lems in reality is dependent largely on the particular
conditions in a given furnace and forming operation.
We have received anecdotal evidence that some manu-
facturers operate with negative DT
FL
quite safely. How-
ever this could not be assessed in a laboratory, so the
DT
FL
criterion was introduced to assess the potential
likelihood of devitrification problems. This criterion
states that if DT
FL
for a given glass is significantly less
than DT
FL
for glass B, the representative benchmark
composition, then the likelihood of crystallisation prob-
lems would increase. This would make such a composi-
tion less desirable in production but does not necessarily
preclude it from consideration.
At SiO
2
contents below approximately 68 wt%, the
chemical durability of flat glass decreases rapidly with
further decreases in SiO
2
content, although this can be
mitigated by increasing Al
2
O
3
levels.
(27)
That situation
is compounded if one refers to the relevant part of the
SiO
2
Na
2
OCaO phase diagram.
(26)
At SiO
2
contents
lower than about 6869 wt%, the Na
2
O.3CaO.6SiO
2
phase field gives way to the Na
2
O.2CaO.3SiO
2
phase
field. I n this phase field, a small reduction in SiO
2
con-
tent strongly increases the liquidus temperature. The
combined effects of decreasing SiO
2
below 68 wt% on
durability and liquidus temperature mean that SiO
2
contents below approximately 68 wt% are likely to be
undesirable for container glass. Reformulations in this
study were therefore limited to SiO
2
contents above this
level. The effect of composition on durability in such
glasses has been investigated.
(15, 28)
Density is a factor requiring consideration because
it affects the number of articles which can be produced
from a given weight of glass. It was thus considered
desirable for the density of candidate glasses to not be
significantly greater than the benchmark glass, glass B.
Refractory corrosion testing is regularly carried out
in the glass industry as a means of assessing both the
resistance of refractories and the aggressiveness of glass
under melting conditions. Several workers have dis-
cussed various factors affecting refractory corrosion,
(29)
the effects of temperature,
(30)
and the effects of glass
composition.
(31)
Pavlovskii & Sobolev
(30)
noted a ten-
dency for increasing activation energy with increasing
corrosion resistance of a melt cast refractory. The cor-
rosion rate displayed a linear relationship when log
(corrosion rate) was plotted against reciprocal tempera-
ture. Therefore at high temperatures, small decreases
in the melting temperature can cause large decreases
in corrosion rate. These effects were quantified.
(31)
Cor-
rosion rates, R, were found to increase with tempera-
ture roughly by factors of R
1500C
/R
1400C
=3 and R
1550C
/
R
1500C
=2. Hence any increase in refractory corrosion
resulting from a more aggressive glass composition may
be offset by the decreased furnace temperatures.
The linear coefficient of thermal expansion of con-
tainer glass is 859010
-7
/C in the range 20300C.
I t is a rule of thumb that additions which decrease
viscosity also increase thermal expansion and decrease
thermal shock resistance. This does not apply in all
cases, for example B
2
O
3
, but it can be used as a general
rule. Several workers have discussed mathematical re-
lationships to describe the effect of different constitu-
ents on the thermal expansion coefficient. These works
were discussed in depth and summarised by Volf.
(32)
The main optical property considerations for con-
tainer glass tend to be issues of colour rather than con-
sideration of refractive index or dispersion. The main
components involved in container glass colour are iron
and selenium (colourless flint glass), iron and chro-
mium (green glass) and iron, carbon and sulphur (am-
ber glass). These colours and their mechanisms and
formulations are well established, and the colourants
used are present only at low levels. Colour in glass was
covered fully by Weyl.
(33)
Significant changes in the lev-
els of colourants would dramatically alter the optical
properties of container glass, so the addition of
colourants such as most transition metals and lantha-
nide elements was eliminated from this study.
A survey of current UK container glass composi-
tions was carried out to provide a representative bench-
mark composition against which candidate glasses
could be assessed. The survey showed differences be-
tween glasses from different manufacturers, but a rep-
resentative average was chosen, namely SiO
2
720, Al
2
O
3
14, Na
2
O 134, K
2
O 04, MgO 17, CaO 109, SO
3
01,
Fe
2
O
3
01 (wt%). Slight differences between industrial
and laboratory compositions in terms of levels of SO
3
and trace contaminants were ignored on the basis that
the expected effects on the important properties would
be negligible.
Experimental
Batches were weighed to produce 500 g of glass using a
calibrated two decimal point balance, and mixed for five
minutes in a Turbula mixer, then added to the crucible
and placed in the furnace. Glasses were melted in an
electric furnace in DPH platinum crucibles at 1450C
for 1 h. They were then poured into water, dried, crushed
and remelted at 1450C for a further 5 h to ensure ho-
mogeneity. Glasses were then poured into preheated
moulds and held at 560C in an oven for 1 h to relieve
internal stresses, then cooled at 08C/min to room tem-
perature. Visual inspection with polarised light
microscopy ensured the final glass samples were ho-
mogeneous. Raw materials were analytical grade silica
sand, sodium carbonate, sodium sulphate, potassium
P. A. BINGHAM & M. MARSHALL: REFORMULATION OF CONTAINER GLASSES FOR ENVIRONMENTAL BENEFIT
14 Glass Technology Vol. 46 No. 1 February 2005 www.sgt.org
carbonate, magnesium carbonate, calcium carbonate,
lithium carbonate, and boric acid. Nominal glass com-
positions based on batch calculations and measured
compositions are shown in Table 1.
Elemental analyses were carried out on the candi-
date glasses using a J EOL J SM6400 scanning electron
microscope with energy dispersive spectroscopy (EDS)
facility. All samples were measured at an acceleration
voltage of 20 kV. Several EDS measurements were
made at different points on each sample, and these re-
sults were averaged. By comparing modelled viscosity
based on batch compositions with measured viscosity
for each glass, steady underestimation by approximately
10C in log two viscosity and 5C in log 4 viscosity
resulted in each case. These differences were attributed
to volatilisation losses during melting. I t was consid-
ered that these small but consistent losses may influ-
ence glass composition enough to slightly affect
physical properties, however these factors were taken
into account when evaluating the errors associated with
each property measurement. Owing to limitations of
the EDS system it was not possible to measure the con-
tent of light elements (Li, B, Na). Sodium content was
measured using x-ray fluorescence (XRF) spectroscopy
with a Uniquant analysis program. Both EDS and
XRF Uniquant are semi-quantitative methods and the
errors associated with these techniques are estimated
to be 1 wt% for SiO
2
, 05 wt% for other major con-
stituent oxides (CaO and Na
2
O) and 03 wt% for
minor constituent oxides (Al
2
O
3
, MgO, K
2
O, SO
3
).
Within the limits of error of these techniques, glass
compositions were generally the same as compositions
based on batch calculations. Volatilisation of small
amounts of alkali and boron could be expected to oc-
cur during melting, but within the limits of error of
the analysis techniques used it was not possible to quan-
tify losses. I n depth fully quantitative analysis would
be necessary during optimisation of any of these com-
positions in future work in order to understand the
volatilisation losses and to help minimise them. I t
should be noted that sample C60A was a powdered
sample whereas all others were solid glass, and this
may explain a slightly higher measured CaO content
in this sample, since EDS can be influenced by sample
geometry.
High temperature viscosity was measured by a ro-
tating viscometer fitted with a platinum spindle in an
electric furnace. A calibrated thermocouple was used
to measure the glass temperature. Measurements were
carried using the method of ASTM Standard C965-
81; Standard practices for measuring viscosity of glass
above the softening point. Modelled and measured
viscosities are shown in Table 2.
Liquidus temperature was measured using a tem-
perature gradient furnace. Pieces of sample glass were
placed in a platinum boat and held at temperature for
24 h in a known temperature gradient measured with
a calibrated thermocouple. Liquidus temperature was
measured by observing the sample using a polarised
light microscope and deriving the corresponding tem-
perature above which no crystals existed. Results are
shown in Table 2.
P. A. BINGHAM & M. MARSHALL: REFORMULATION OF CONTAINER GLASSES FOR ENVIRONMENTAL BENEFIT
Table 2. Measured properties of benchmark and candidate glasses
Property Glass Glass Glass Glass Glass Glass Glass Glass Glass Glass Glass
B C44 C3 C112A C60A C16B C39C C39B C112C C60B C46B
Modelled T log 2 (C) 1446 1430 1403 1406 1405 1397 1372 1375 1376 1365 1347
Modelled T log 3 (C) 1178 1163 1139 1145 1137 1126 1108 1110 1112 1099 1083
Modelled T log 4 (C) 1027 1013 996 1001 995 986 968 969 972 962 949
Measured T log 2 (C) 1458 1430 1411 1417 1408 1405 1380 1385 1384 1376 1343
Measured T log 3 (C) 1188 1175 1148 1158 1155 1134 1116 1136 1120 1129 1103
Measured T log 4 (C) 1021 1010 990 994 996 980 962 974 966 970 951
Liquidus 5 (C) 1015 1032 1010 990 955 942 975 985 965 964 950
DTFL 5 (C) +6 22 20 +4 +41 +38 -13 -11 +1 +6 +1
Density (gcm
-3
) 2507 2508 2524 2519 2518 2493 2570 2500 2524 2561 2508
Density factor - 1000 1007 1005 1004 0994 1025 0997 1007 1022 1000
a10
-7
(C) 897 913 969 1028 980 954 995 1038 1063 1033 1116
T
d
(C) 624 614 625 612 628 637 601 599 600 600 591
USP 25 acid (ml) 669 - 1010 799 758 - - 924 1035 644 878
USP classification I I I - NP I I I I I I - - NP NP I I I NP
Static corrosion 3 (%) 173 - 160 173 190 - - - - - -
Tm reduction (C) - 28 47 41 50 53 78 73 74 82 115
Estimated thermal - 20 30 25 30 30 35 35 35 38 43
NO
x
reduction 5 (%)
Table 1. Compositions in wt% based on batch calculations. Bracketed numbers are normalised combined
EDS and XRF analyses
Wt % Glass Glass Glass Glass Glass Glass Glass Glass Glass Glass Glass
B C44 C3 C112A C60A C16B C39C C39B C112C C60B C46B
SiO2 720 (725) 7175 (719) 704 (711) 694 (702) 700 (694) 700 (705) 698 (705) 692 (704) 695 (703) 690 (701) 6890 (707)
Na
2
O 134 (133) 134 (134) 15 (146) 134 (133) 134 (134) 134 (135) 146 (141) 146 (142) 145 (147) 140 (137) 145 (140)
K 2O 04 (04) 04 (03) 04 (04) 30 (25) 24 (21) 24 (20) 20 (17) 20 (18) 24 (19) 25 (21) 20 (16)
MgO 17 (17) 17 (19) 17 (16) 17 (17) 07 (06) 0 (0) 05 (07) 05 (05) 05 (06) 05 (05) 03 (03)
CaO 109 (106) 109 (107) 109 (108) 109 (107) 109 (118) 109 (108) 111 (110) 112 (107) 111 (107) 109 (108) 112 (103)
Al2O3 14 (12) 14 (12) 14 (12) 14 (11) 14 (13) 14 (12) 14 (13) 19 (17) 15 (13) 15 (12) 14 (11)
Fe2O3 01 (0) 01 (0) 01 (0) 01 (0) 01 (0) 01 (0) 01 (0) 01 (0) 01 (0) 01 (0) 01 (0)
SO
3
01 (03) 01 (03) 01 (03) 01 (03) 01 (04) 01 (03) 01 (03) 01 (03) 01 (03) 01 (03) 01 (03)
Li2O 0 025 0 0 0 0 04 04 03 04 05
B2O3 0 0 0 0 10 17 0 0 0 10 10
Glass Technology Vol. 46 No. 1 February 2005 www.sgt.org 15
Density was measured by the Archimedes method.
Samples were weighed using calibrated four decimal
point scales, once in air then again with the sample
suspended in distilled water. This allowed calculation
of glass density using
Density=[W
A
/(W
A
W
W
)]d
W
(1)
where W
A
is the weight in air, W
W
is the weight in
distilled water, and d
W
is the correction factor for the
density of water at the measurement temperature,
available from standard chemistry data. Units of den-
sity are gcm
-3
and the measurement temperature was
21C for all samples. The measured densities are given
in Table 2.
Chemical durability was measured using US Phar-
macopoeia Method No. 25 for hydrolytic resistance
of glass containers, and the results are shown in Table
2. Powdered glass samples weighing 10 g of a known
particle size distribution were held in 50 ml of pure de-
ionised water under pressure in an autoclave at 121C
using a specific heating, holding and cooling regime.
Using 002 N acid, the solution was titrated to find the
total alkali released. Values shown refer to the volume
in ml of acid required to fully neutralise the leachate
solution. USP 25 durability classifications pertinent to
this study are Type I I I and Type NP (non-parenteral).
The upper limits for 002 N acid titration are 85 ml
and 150 ml for Types I I I and NP, respectively.
Thermal expansion coefficient, dilatometric soften-
ing point and glass transition temperature measure-
ments were obtained from dilatometric analysis and
are given in Table 2. A calibrated Netzch DI L-402PC
dilatometer connected to a computer was used for these
measurements.
Static finger corrosion testing was carried out on a
limited number of samples. Fingers of a standard AZS
refractory material were suspended in molten glass
samples in DPH platinum crucibles at 1370C in an
electric furnace for 24 h, then removed, cooled and
analysed for fluxline corrosion as a percentage of the
original AZS specimen diameter. All the measured
glasses were found to have equal fluxline corrosion rates
within the errors of this measurement technique. Re-
sults are given in Table 2.
Estimated reductions in thermal NO
x
generation,
as given in Table 2, were calculated using data from
Kremer et al.
(5)
They plotted NO
x
in flue gas versus
furnace temperature in the range 11001600C, for a
number of different air ratios. All air ratios showed
very similar behaviour so we were more confident in
using this data qualitatively in our estimations. These
were calculated using the measured log 2 viscosity of
each glass. This figure was cross checked with the data
provided by Kremer et al
(5)
to give an estimated ther-
mal NO
x
reading. These estimations were then used to
calculate the percentage reduction in NO
x
for each can-
didate glass versus the benchmark glass. This method
produced what may be conservative estimates of ther-
mal NO
x
reduction as in reality the furnace tempera-
ture is substantially higher than the glass
temperature.
(34)
We stress that our estimations of NO
x
reduction are intended for use as qualitative guidelines
only, and to illustrate the large reductions in NO
x
which
may be possible using this technology.
Results and Discussion
Computer models based on the work of Lakatos et
al
(2225)
were used, where possible, to estimate the ef-
fects of compositional changes on viscosity. Model-
ling proved more difficult when considering other
properties, such as glass stability or chemical durabil-
ity, owing to the complexity of thermodynamic phase
relations and diffusion mechanisms. No simple pre-
diction methods exist for these properties that remain
accurate over wide composition ranges, although sev-
eral methods were discussed by Volf,
(32)
Bonetti,
(28,3537)
Karlsson et al
(38)
and Backman et al.
(39)
I t was found
that general rules are obeyed and phase relations and
durability can be qualitatively predicted based on the
relationships shown by Rawson.
(26)
I n this study it was decided to change the levels of
those components whose presence causes the greatest
increase in high temperature viscosity, namely SiO
2
,
Al
2
O
3
and MgO. The motivation for this was that it
would allow the smallest deviations from current com-
position in order to achieve maximum viscosity reduc-
tion, which is desirable for economic and logistical
reasons. Clearly further work in this area is justified,
for example an investigation of pro rata substitutions.
The nominal compositions and results for the most
promising glasses developed during this work are given
in Tables 1 and 2, respectively.
Glasses C44, C3 and C112A represent simple 1-for-
1 substitutions of SiO
2
by Li
2
O, Na
2
O and K
2
O re-
spectively. Different levels of each alkali oxide were
added owing to their differing effects on viscosity. Sub-
stitution of 025% SiO
2
by Li
2
O in glass C44 produced
a large decrease in viscosity. The addition of lithium
greatly affects many of the properties of sodalime
silica glasses. I t is known to have a very strong fluxing
effect
(1113,15,2125,40)
and lithium raw materials generally
improve the melting behaviour of sodalimesilica
glasses. Lithium also has lower volatilisation rates than
the other alkalis. Some workers have concluded that
the liquidus temperature of sodalimesilica glass de-
creases slightly upon addition of Li
2
O. The results of
this study show that this depends upon the compo-
nents for which it is substituted. Comparison of the
results for glasses B and C44 shows that substitution
of 025% Li
2
O for SiO
2
increased the liquidus tempera-
ture by 17C. This creates an additional complication
because DT
FL
, the temperature difference between form-
ing viscosity and liquidus temperature, changes from
+6 to 22C. As discussed previously, this does not
necessarily present a problem, however the chances of
devitrification are increased relative to the benchmark
glass. The chemical durability of this sample was not
measured. Due to the very small change in composi-
tion and the mixed alkali effect,
(41)
durability was ex-
pected to be very similar to that of glass B and almost
certainly in the same USP durability class, I I I . This
level of Li
2
O had little effect on thermal expansion or
density, however T
d
decreased by 10C, indicating a
small effect on the working range. Estimated thermal
P. A. BINGHAM & M. MARSHALL: REFORMULATION OF CONTAINER GLASSES FOR ENVIRONMENTAL BENEFIT
16 Glass Technology Vol. 46 No. 1 February 2005 www.sgt.org
NO
x
generation was decreased by approximately 20%.
These results demonstrate the reasons why lithium has
long been under discussion as an ingredient of con-
tainer glasses. Objections have thus far been based
largely on economic grounds but with this situation
potentially changing, the use of lithium may become
more acceptable in future.
Substituting 16% SiO
2
by Na
2
O in glass C3 brought
the expected changes in measured properties. The USP
durability deteriorated from 669 to 1010 ml of 002 N
acid required for neutralisation, hence the durability
class changed from Type I I I to Type NP. The log 2
viscosity was reduced by 47C, and, as with the addi-
tion of lithium, this may cause concerns due to a value
of DT
FL
of -20C. This study found a reduction of
only 5C in liquidus temperature between glasses B
and C3, whereas a reduction of about 45C was sug-
gested in studies on the substitution of similar levels
of SiO
2
by Na
2
O in similar glasses.
(35)
Observation of
the phase diagram for the Na
2
OCaOSiO
2
glass sys-
tem
(26)
does not support a decrease of 45C in liquidus
temperature if the primary phase field is
Na
2
O.3CaO.6SiO
2
, devitrite. I t may be that the higher
levels of Al
2
O
3
in the other glasses
(35)
caused the pri-
mary phase field to shift to SiO
2
, tridymite, and this
may explain such large shifts in liquidus temperature
for small replacements of SiO
2
by Na
2
O. Glass C3 gave
an estimated 30% reduction in thermal NO
x
genera-
tion. This glass would probably not be considered for
production without further modifications owing to its
relatively poor durability and negative DT
FL
.
Glass 112A completes the sub-series of glasses
studying the partial replacement of SiO
2
by alkali ox-
ides. This glass is particularly promising because it ful-
fils the criteria set out earlier for container glass. The
benefits include reductions of 41C in log two viscos-
ity, corresponding to a reduction of approximately 25%
in thermal NO
x
generation. I n addition the liquidus
temperature decreased to 990C, and this gives a value
of DT
FL
of +4C, matching that of glass B. Therefore,
the likelihood of devitrification problems is no greater
than for glass B. The chemical durability is only slightly
lower than glass B, but still USP 25 Type I I I . This is
evidence of the mixed alkali effect, which has received
considerable discussion. Early work by Peddle
(41)
dem-
onstrated the strength of this effect on the durability
of sodaleadsilica glasses containing Na
2
O and K
2
O.
A glass with 3 mol% K
2
O and 12 mol% Na
2
O was
twice as durable as one containing 15 mol% Na
2
O. This
explains the relatively high durability of glass C112A
considering its higher total alkali content. Fluxline cor-
rosion was unaffected by this increase, within the er-
rors of measurement. Thermal expansion increased to
102810
-7
/C however it is unlikely that this repre-
sents a significant change in thermal shock resistance.
Glass C112A demonstrates the potential benefits of
adding K
2
O to container glass, and utilising the mixed
alkali effect to maintain durability.
The action of the alkaline earth elements in soda
limesilica type glasses is well understood
(15,16,21,26)
so it
is possible to accurately predict their effects on prop-
erties such as viscosity,
(2225,32)
thermal expansion
(32)
and
density.
(32)
The phase relations have also been carefully
studied in these systems
(15,21,26,42)
and qualitative pre-
dictions of liquidus temperature and crystallisation rate
are possible based on phase diagrams.
(26)
I t is widely
held that the addition of MgO decreases the liquidus
temperature of sodalimesilica glass. The phase rela-
tions in sodalimesilica glasses where some of the
CaO is replaced by MgO were discussed by Morey.
(42)
For glasses with compositions in the tridymite phase
field, such changes can dramatically reduce liquidus
temperature, with a minimum at 25 wt% MgO.
(21,42)
However, the compositions of modern container glass
are such that the primary phase field is devitrite, and
the addition of more MgO to these glasses would not
produce significant reductions in liquidus temperature.
Morey
(42)
indicated that when devitrite is the primary
phase field, the addition of MgO to the parent glass
would increase liquidus temperature. The effects of
Al
2
O
3
were found to be similar to MgO,
(42)
in that when
the primary phase field is tridymite, the addition of
Al
2
O
3
strongly decreases liquidus temperature, but
when the primary phase field shifted to devitrite, the
liquidus temperature increased with increasing Al
2
O
3
levels. I ncreased levels of MgO or Al
2
O
3
could there-
fore not be justified in many of the glasses in this study.
I n addition to the lack of beneficial effects on glass
stability in the devitrite phase region they increase high
temperature viscosity, contrary to the objectives of this
study. The beneficial effects of Al
2
O
3
on chemical du-
rability are well known
(21,26,42)
so in general the level of
Al
2
O
3
was maintained, however the replacement of
MgO proved very effective in reducing liquidus tem-
perature and high temperature viscosity, as shown in
Table 2.
Boron is one of the most efficient melting accelera-
tors in silicate glasses when added at low levels.
(9,10)
Bo-
ron compounds in the batch initiate glass formation at
an early stage and boron can therefore be useful as a
subsidiary flux in sodalimesilica glass batches in
amounts up to about 15% B
2
O
3
.
(9)
Boron lowers sur-
face tension, further increasing the speed of melting
and refining, and it reduces thermal expansion. Al-
though it has not been quantified in this study, it is
clear that such fluxing action would contribute to lower
melting energies. In the 1920s and 1930s a great number
of container glass manufacturers in the UK and USA
incorporated small levels (0315%) of B
2
O
3
in their
glass.
(10,17,43)
Simplification of glass batches in the in-
tervening years led to the removal of B
2
O
3
from most
western container glasses. Other workers have indicated
that low levels of B
2
O
3
can reduce melting and refin-
ing times and decrease the liquidus temperature and
crystallisation rate.
(10,4245)
Morey
(42)
described the effects
of B
2
O
3
on crystallisation of sodalimesilica glass.
Addition of B
2
O
3
shifted the phase field boundary be-
tween Na
2
O.3CaO.6SiO
2
(devitrite) and
Na
2
O.2CaO.3SiO
2
to lower SiO
2
contents and acted to
reduce liquidus temperatures. The beneficial effects of
B
2
O
3
on viscosity and chemical durability are well-
documented.
(21,2325,28,42)
The addition of B
2
O
3
and more
K
2
O to glasses C60A and C16B reduced both viscos-
ity and liquidus temperature. This effect was probably
P. A. BINGHAM & M. MARSHALL: REFORMULATION OF CONTAINER GLASSES FOR ENVIRONMENTAL BENEFIT
Glass Technology Vol. 46 No. 1 February 2005 www.sgt.org 17
enhanced by the reduced MgO content. Replacing some
SiO
2
and MgO by K
2
O and B
2
O
3
decreased liquidus
temperatures in these glasses to approximately 950C
and reduced log 2 viscosity by about 50C. This gave
DT
FL
of approximately +40 and a chemical durability
only slightly lower than glass B, still USP 25 Class I I I .
The fact that DT
FL
is strongly positive indicates that
with further development it should be possible to in-
crease levels of low cost, high temperature fluxes such
as CaO by several tenths of one percent. I n so doing,
log 2 viscosity could be decreased further whilst not
allowing DT
FL
to fall below +6C. The mixed alkali
effect and the beneficial action of B
2
O
3
combined to
minimise the decrease in chemical durability, and only
a very small deterioration was observed. Glasses C60A
and C16B both carried estimated NO
x
reductions of
30% compared with the benchmark glass.
Anecdotal evidence received by the authors indicated
that there may be concerns over the potential effects
of container glass reformulation on furnace lifetimes.
This may manifest in fluxline corrosion, crown corro-
sion, or increased levels of particulates. Volatilisation
of glass components has been studied in some key sys-
tems.
(4647)
However in container glasses, the effects of
changes in levels and types of alkalis or small addi-
tions of lithium and/or boron require further study.
Hwang et al
(47)
studied the effects of stepwise replace-
ments of Na
2
O by K
2
O in television glasses. Their re-
sults indicated that this only led to significant increases
in weight loss when the Na
2
O/K
2
O molar ratio <1. I f
this applies to container glasses, which are broadly simi-
lar in composition to television glass, it indicates that
a higher K
2
O content would not cause a substantial
increase in volatilisation rate by itself, although in-
creased total alkali content may do. Volatilisation from
glass melts is strongly dependent upon glass melting
temperature,
(46)
so lower melting temperatures could
offset any changes to volatilisation rates. Careful meas-
urement of the glass compositions in question could
supply this information, although this in turn would
depend on raw materials and furnace conditions. Raw
materials choice will be a major factor in bringing any
new glass formulation into production. Not only the
types of raw materials
(820)
but also their interactions
during melting, as evidenced by the effects of MgO/
CaO ratio on melting and refining,
(48)
can play an im-
portant role.
The remaining five candidate glasses each contained
altered levels of several components. Glasses C39C,
C39B, C112C, C60B and C46B all had substantially
reduced log 2 viscosities compared with the benchmark.
I n particular glasses C60B and C46B had very low
melting points, around 13501360C. With the excep-
tion of glass C60B, these five glasses had poorer du-
rability than glass B. This was attributed to the higher
levels of alkali in these glasses, in particular the in-
creased Na
2
O content. The mixed alkali effect
(41)
miti-
gated some of the deterioration in durability, but not
sufficiently to completely offset it. Comparison of the
durability results for glasses C60B and C46B shows
an interesting effect. Aside from differences in Na
2
O
and K
2
O levels these glasses are almost identical in
composition. Both glasses had the same (Na
2
O+K
2
O)
contents but different ratios of one alkali to the other.
This is a good demonstration of the benefits of the
mixed alkali effect.
(41)
Thermal expansion coefficients
increased to above 10010
-7
/C, again attributable
mainly to increased alkali levels. Generally this would
decrease thermal shock resistance although it would
not be a large deterioration. We attribute the very low
liquidus temperatures in these five glasses to the re-
duced MgO levels. Total alkaline earth levels decreased,
whilst total alkali levels increased. The presence of
boron in some glasses also contributed to reductions
in liquidus temperature.
Many of the properties offered by glass C60B were
very promising; melting temperature was reduced by
over 80C, with DT
FL
equal to the benchmark, and
chemical durability higher than the benchmark. Pro-
jected reductions in thermal NO
x
generation were very
substantial, at over 35%. However, the density of this
glass was more than 2% higher than the benchmark,
and this may adversely affect the prospect of using such
a formulation on economic grounds.
Glass 46B gave the largest reduction in melting tem-
perature of any glass, but it also has the highest Li
2
O
content. A relatively small improvement in chemical
durability would be sufficient to put it in USP 25 Class
I I I durability. Computer modelling of viscosities us-
ing Lakatos factors
(2225)
indicated that replacement of
Li
2
O in this glass by CaO would still give an overall
reduction in melting temperature of about 80C. Even
if the Li
2
O was replaced by SiO
2
, the predicted reduc-
tion in melting temperature would still be nearly 70C.
The effects of such changes on liquidus temperature
and chemical durability would have to be assessed,
however it is clear that with further development this
composition could be optimised with more attention
to economic and logistic factors. As noted previously,
the addition of Li
2
O acts to increase liquidus tempera-
ture in container glass, so replacing at least some of
the Li
2
O by CaO should have a smaller impact than
would be the case if CaO replaced SiO
2
. I f CaO were
added, one option may be to replace a small amount
of SiO
2
by P
2
O
5
as a liquidus temperature suppres-
sant.
(1)
Glass 46B glass carried the highest estimated
reduction in thermal NO
x
generation of all the glasses
in this study, at over 40%.
Our reformulated glasses could reduce thermal NO
x
generation by approximately 2040%, depending upon
the composition. To put these figures into context, the
UK container glass industry generates approximately
5000 tonnes of NO
x
per annum.
(49)
The great majority
of this is thermally generated NO
x
, so the reduction in
total NO
x
could potentially be up to 2000 tonnes per
annum for the UK container glass industry.
Estimation of the reduction in melting energy and
particulate generation associated with glass production
using these new formulations was discussed in the
project report,
(3)
but this was based solely on final ox-
ide glass compositions and no consideration was given
to the effects of raw materials or cullet levels. I ts prac-
tical applicability is therefore limited. Particulate re-
ductions are governed by furnace pull rate, glass
P. A. BINGHAM & M. MARSHALL: REFORMULATION OF CONTAINER GLASSES FOR ENVIRONMENTAL BENEFIT
18 Glass Technology Vol. 46 No. 1 February 2005 www.sgt.org
composition and raw materials so quantification at this
stage would also be limited in its usefulness. Based on
temperature factors alone and assuming all other fac-
tors remain unchanged, we estimated that reductions
in particulates of up to 50% may be possible for a drop
of 100C in melting temperature.
(3)
However, any
chemical reformulation of container glass will require
different levels and/or types of raw materials and cullet
additions. This will in turn affect melting behaviour
and particulate generation. I t was therefore considered
that discussion in this paper of the potential reduc-
tions in melting energy and particulate generation
would be premature, and should only be carried out
when an investigation of the effects of raw materials
and cullet has been carried out. Likewise the potential
reductions in SO
x
and heavy metal (selenium) emis-
sions, as these are also dependent upon the final glass
batch. We believe it is likely that reductions in SO
x
and
selenium would occur using the reformulated compo-
sitions, as volatilisation of these components is strongly
temperature dependent.
(46,50)
SO
x
emissions are also
governed by glass redox, sulphate solubility, melting
rate and raw materials so these will vary from furnace
to furnace. As with melting energy and particulates,
further work will be required to quantify these reduc-
tions. Nevertheless, we feel that there is a strong likeli-
hood of achieving reductions in melting energy,
particulates, SO
x
and heavy metals by implementing
our reformulated compositions.
I n addition to its direct effect upon melting energy,
the use of cullet could also affect other aspects of the
production of reformulated glasses. Since the cullet
currently available would no longer be of the same
composition as the final target glass, it would be nec-
essary to make batch alterations to take account of
this fact. Over time the average cullet composition
would gradually change as the new glass entered the
recycling chain. This cullet would then begin to more
closely resemble the final glass in its composition. This
opens up a number of logistical and political issues
which would need to be addressed, and it depends upon
uptake of the technology and cullet and raw material
sourcing.
Although some efforts have been made towards as-
sessing the logistical and economic factors associated
with container glass reformulation,
(3)
further studies
are required in order to link all the pertinent factors
and direct development. Crystal growth rate was not
measured, however changing the glass composition
may affect this parameter.
(21,26,45)
I t is suggested that
future work should feature analysis of crystal growth
rate, volatilisation losses, furnace crown refractory cor-
rosion and the melting and refining behaviour of new
batches. Quantification of reductions in melting en-
ergy and emissions of CO
2
, NO
x
, SO
x
, particulates and
heavy metals could take place in tandem with this
optimisation process.
Conclusions
A number of reformulated container glass composi-
tions have been developed with reductions of up to
115C in melting temperature and an estimated 20
40% in thermal NO
x
generation. Reductions in melt-
ing energy and emissions of SO
x
, particulates and heavy
metals are also expected to occur but have notbeen
discussed here, as the effects of raw materials must first
be assessed.
Several candidate glass compositions met the prop-
erty criteria set out for the benchmark composition,
glass B. These criteria included high temperature vis-
cosity, liquidus temperature, DT
FL
, and USP Type 25
chemical durability. Other important measured prop-
erties were density, dilatometric softening point, static
refractory fluxline corrosion and thermal expansion
coefficient. The new compositions offer a range of
options for optimisation through further development,
from relatively simple substitutions to multiple com-
ponent changes. However, it is estimated that even rela-
tively small compositional changes can result in
significant reductions in thermal NO
x
generation.
A full optimisation process must take account of
wider issues impacting upon the industrial container
glass production process, including logistical and eco-
nomic concerns, furnace lifetimes, volatilisation and
particulate generation, raw materials and cullet selec-
tion and availability, and melting reactions. This will
allow full development of the most suitable, most ac-
ceptable new glass composition. We believe the tech-
nological foundations for such a process have been laid
in this paper.
Acknowledgements
The majority of the work described in this paper was
carried out at Glass Technology Services Ltd. P. A.
Bingham is now employed by the University of Shef-
field and wishes to thank both organisations for their
support. The authors acknowledge the funding sup-
port of the Carbon Trust for the work described in
this paper, which constitutes part of a larger project.
The statements made and views expressed are those of
the authors and should not be attributed to the Car-
bon Trust which accepts no responsibility for the con-
tents of the paper. The Carbon Trusts funding is
provided by grant from Defra (the Department for En-
vironment, Food and Rural Affairs), the Scottish Ex-
ecutive, the National Assembly for Wales and I nvest
Northern I reland. Glass Technology Services have now
secured further funding from the Carbon Trust to con-
tinue the project. This work will include industrial tri-
als involving the UK container glass industry. Enquiries
concerning the ongoing project should be directed to
M. Marshall.
References
1. Bingham, P. A. The effects of 1 wt % P2O5 addition on the properties of
container glass. Glass Technol., 2004, 45(6), 2558.
2. Bingham, P. A. Container glass formulation: a fresh look at an old
problem. Glass, 2003, 80(11), 336.
3. Bingham, P. A. Carbon Trust project reference 2002-6-32-1-3, Refor-
mulation of container glass for energy efficiency final project report,
November 2003. Available electronically upon request.
4. Good Practice Guide 127, Energy efficient environmental control in the
glass industry - revised, Energy Efficiency Best Practice Programme,
published by ETSU, Harwell, Didcot, Oxfordshire, OX11 0QJ, UK.
5. Kremer, H., May, F. & Wirtz, S. The influence of furnace design on the
NO formation in high temperature processes. Energy Cons. Man.,
2001, 42, 193752.
6. Stack emissions monitoring in the UK glass industry. Glass, 2004, 81
P. A. BINGHAM & M. MARSHALL: REFORMULATION OF CONTAINER GLASSES FOR ENVIRONMENTAL BENEFIT
Glass Technology Vol. 46 No. 1 February 2005 www.sgt.org 19
(4), 106.
7. I ntegrated pollution prevention and control (I PPC), Secretary of States
guidance for glass manufacturing activities with melting capacity more
than 20 tonnes per day, DEFRA, UK , J une 2003; http://
www.defra.gov.uk/environment/ppc/laippc/sg2.pdf
8. Simpson, W. Two decades of raw materials development. Glass, 1989,
66(12), 4757.
9. Taylor, R. Minor ingredients and their effects on glass melting and
properties. Glass Prod. Tech. I nt., 1990, 2936.
10. Stevenson, P. B2O3 in container glass. Glass, 1991, 68(12), 5003.
11. Kingsnorth, D. J. & Haigh, M. Lithium mineral boosts glass produc-
tion efficiency. Ceram. I nd., 1990, 134 (10) 336.
12. K ingsnorth, D. J . Adding lithium minerals can reduce melting
costs. Glass I nd., 1988, 69(3) 1825.
13. Meitz, H. Use of spodumene ore as a flux for glass melting. Glass
Technol., 1989, 30(1), 69.
14. Simpson, W. Calumite slag as a glassmaking raw material for the in-
crease of furnace productivity. Glass Technol., 1976, 17(1), 3540.
15. The Handbook of Glass Manufacture, Third Edition, Ed. F. V. Tooley,
Ashlee Publishing Co. I nc., New York, 1984, 318.
16. Lyle, A. K. Design and development of glasses for manufacture of
containers. Glass I nd., 1967, May, 252.
17. Loesell, R. E. & Lester, W. R. Container glass compositions 1932 to
1960, Glass I nd., 1961, 623.
18. Stadler, L. E. & Cronin, D. Container glass composition, Glass I nd.,
1977, December, 10.
19. Stadler, L. E. & Cronin, D. Container glass composition. Part 2. Glass
I nd., 1978, J anuary, 10.
20. Smrcek, A. European container glasses 1982 to 1988. Glastech. Ber.,
1990, 63(10), 30917.
21. Volf, M. B. Chemical Approach to Glass, Elsevier, 1984.
22. Lakatos, T., J ohansson, L-G. & Simmingskold, B. Viscosity tempera-
ture relations in the glass system SiO2Al2O3Na2OK2OCaOMgO
in the composition range of technical glasses, Glass Technol., 1972, 13
(3), 8895.
23. Lakatos, T., J ohansson, L.-G. & Simmingskold, B. The effect of some
components on the viscosity of glass. Glastekn. Tidskr., 1972, 27(2),
258.
24. Lakatos, T., J ohansson, L.-G. & Simmingskold, B. I nfluence of Li2O
and B2O3 on the viscosity of soda-lime-silica glass. Glastekn. Tidskr.,
1975, 30(1), 78.
25. Lakatos, T. Viscosity-temperature relations in glasses composed of SiO2
Al2O3Na2OK2OLi2OCaOMgOBaOZnOPbOB2O3. Glastekn.
Tidskr., 1976, 31(3), 514.
26. Rawson, H. I norganic glass-forming systems. Academic Press, 1967.
27. Combes, J. M. & Sachot, D. Clear glass composition intended for the
production of panes, US Patent 5,656,559, Saint-Gobain Vitrage,
August 12th 1997.
28. Bonetti, G. A statistical/mathematical and thermodynamic approach to
predicting chemical resistance of glass. Riv. Stn. Sper. Vetro, 1985, 15
(3), 11120.
29. Kasselouri, V., Kouloumbi, N. & Mendrinos, L. Effect of glass melt on
corrosion of the lining of an industrial glass furnace, Glass Technol.,
2002, 43(2), 759.
30. Pavlovskii, V. K. & Sobolev, Y. S. Effect of temperature on the corro-
sion of refractories in glass melts (review). Glass Ceram., 1991, 48
(12), 55861.
31. Technical University of Berlin, Refractory corrosion rates in various
glass melts. Glass, 1992, 69 (2), 5860.
32. Volf, M. B. Mathematical Approach to Glass, Elsevier, 1988.
33. Weyl, W. Coloured Glasses, Society of Glass Technology, 1951.
34. Hayes, R. R., Brewster, S., Webb, B. W., McQuay, M. Q. & Huber, A.
M. Crown incident radiant heat flux measurements in an industrial,
regenerative, gas-fired, flat-glass furnace. Exp. Therm. Fl. Sci., 2001,
24, 3546.
35. Bonetti, G. Considerations for ASTM C-829. Liquidus temperature
(devitrification curve). Riv. Stn. Sper. Vetro, 1994, 23(4), 15569.
36. Bonetti, G. Mathematical methods for calculation of glass properties
as a function of composition. Part 3: viscosity. Riv. Stn. Sper. Vetro,
1998, 27(4), 17590.
37. Bonetti, G. Mathematical models for calculating the hydrolytic resist-
ance of glass. Riv. Stn. Sper. Vetro, 1995, 24 (3), 10718.
38. Karlsson, K . H., Backman, R., Cable, M., Peelen, J . & Hermans,
J. Estimation of liquidus temperatures in silicate glasses. Glast. Ber.
2001, 74(7), 18791.
39. Backman, R., Karlsson, K. H., Cable, M. & Pennington, N. P. Model
for liquidus temperature of multi-component silicate glasses. Phys.
Chem. Glasses, 1997, 38(3), 1039.
40. Tang, Y. & Frischat, H. I nfluence of small additions of Li2O raw mate-
rials on glass melting. Glastech. Ber., 1995, 68(7), 213.
41. Peddle, C. J. The development of various types of glass. Part XI . The
effect of the joint presence of sodium and potassium on the solubility
of lead glasses. J . Soc. Glass Technol., 1921, 5, T195201.
42. Morey, G. W. Properties of glass. American Chemical Society, Reinhold
Publishing Corp., 1938.
43. Higginbotham, G. & Gilbert, J. L. The effect of small changes in glass
composition on melting and refining rates and sulphur retention, Part
I I I the effect of small additions of B2O3 and Li2O at 1470C. BGI RA
Research Report No. 57, 1978.
44. Allison, R. S. & Turner, W. E. S. Further investigations upon the influ-
ence of boric oxide on the rate of melting of the batch, and on the rate
of refining and of setting of commercial glasses of the soda-lime-silica
type. J . Soc. Glass Tech., 1954, 38, T297364.
45. Walker, G. E. The influence of boric oxide on the crystallisation of a
soda-potash-lime-silica glass. J . Soc. Glass Technol., 1945, 29, T38
47.
46. Beerkens, R. C. & Van Limpt, J. Evaporation in industrial glass melt
furnaces. Glastech. Ber., 2001, 74(9), 24557.
47. Hwang, J.-H., Park, J.-H., Kim, K.-D. & Choi, S.-S. I nfluence of mixed
alkali oxides on some melt properties of TV screen glass. Glastech.
Ber., 2000, 73(1), 817.
48. Kazazoglu, E., Albayrak, G. & Yaraman, A. Effect of MgO / CaO ratio
on the melting behaviour of soda-lime-silica glasses. Glastech. Ber.,
1983, 56K (1), 1924.
49. UK glass manufacture. A mass balance study. British Glass, 2003. http://
www.britglass.co.uk/Files/UKGlassManufactureAMassBalance.pdf
50. Beerkens, R. G. C. The role of gases in glass melting processes. Glastech.
Ber., 1995, 68 (12), 36980.
P. A. BINGHAM & M. MARSHALL: REFORMULATION OF CONTAINER GLASSES FOR ENVIRONMENTAL BENEFIT