Modelling Annealing Lehrs for Flat Glass

ROBERT GARDON*
Research Staff, Ford Motor Company, Dearborn, Michigan 48 121

Computer simulationof heat transfer in annealing lehrs for flat residual midplane tension (or surface compression) “frozen” into
glass is combined with a method developed earlier for the the glass. No account is taken of membrane stresses.*
calculation of stresses in glass. With this combined model one
can predict stresses produced by annealing schedules that are 11. Development of Lehr Model
readily realized in practice. Implications for the design, opera- (1) Description of Lehr
tion, and energy economy of lehrs are discussed.
Annealing lehrs for flat glass are divided lengthwise into zones
and the design of each zone is tailored to local temperatures of the
I. Introduction glass. A zone in the “annealing section” of a lehr consists of an
insulated shell, through which the glass ribbon is transported on
T HE advantage of annealing glass by cooling it at progressively
higher rates has long been known, ‘ even though annealing
stresses could be predicted analytically only if glass was cooled
rollers. The particular design3 considered in this paper is shown
schematically in Fig. 1. It was chosen for its simplicity, which
allows essential features of lehr operation to be illustrated with a
through the entire annealing range at one constant rate.* Annealing
minimum number of design and operating parameters. In zones of
lehrs have therefore been designed largely on the basis of experi-
this type, the glass ribbon is cooled by countercurrent streams of
ence, and rather conservatively. To modulate cooling rates -or
air flowing through two banks of tubes, one above and one below
even only to control them more closely -annealing lehrs for flat
the ribbon. The air (coolant) may pass through a zone just once, or
glass are divided into “zones,” each of which is, in effect, a
it may be recirculated in order for its inlet temperature to be raised
separate heat exchanger. In the conventional view, cooling rates in
above room temperature at a minimum cost in energy.
each zone are regarded as constant. However, it has also become
recognized that cooling rates can change not only between zones (2) Heat Transfer in Lehr
but also within zones.3 At temperatures in the annealing range the dominant mecha-
The starting point of the present considerations was the develop- nisms of heat transfer involved in cooling the glass are radiation
ment of methods to calculate stresses produced in glass subjected from the glass to the tubes and forced convection from the tubes to
to an arbitrary heat treatmer~t.~-~ In particular, Ref. 6 showed that, the cooling air. Heat transfer by natural convection within the
given the temperature-time history of a small glass plate, the re- “tunnel,” i.e. from the glass to the air in the tunnel and from the
sulting annealing stresses can be predicted with a high measure of air to the tubes and walls of the lehr, is relatively small and could
reliability. To apply these laboratory findings to annealing on an be neglected without affecting the conclusions of this study. (It
industrial scale entails two tasks: One is to analyze heat transfer in becomes important in cooling the glass at lower temperatures.)
an annealing lehr with a view to determining the temperature-time The roof and floor of the lehr are insulated and essentially
history of the glass as a function of lehr design and operating adiabatic. Radiation from the glass to these surfaces is therefore
conditions. The second is to predict annealing stresses from this reradiated to the nearby tubes, thus augmenting the direct radiant
temperature-time history and the relevant physical properties of exchange between the glass and the tubes.” Heat losses through
the glass. side walls and along roller journals are not negligible and they are
A computer program LEHR was written to do that, and this largely responsible for the preferential cooling of the edges of the
paper reports some of the findings made in using it to study the glass ribbon and, hence, for membrane stresses. However, in line
design and operation of annealing lehrs for float glass. For the with the intention of leaving such stresses out of account for the
present we will be concerned only with those aspects of lehrs that present, these sideways heat losses are not considered, and the
affect the generation of ordinary annealing stresses, i.e. of the temperature distribution in the glass ribbon is treated as one-
dimensional, i.e. as varying primarily along its length. (The con-
Presented at the 83rd Annual Meeting, The American Ceramic Society, Washing- comitant temperature variation through the thickness of the glass
ton, D . C . , May 5 , 1981 (Glass Division No. 21-G-81). Received September 25,
1981; revised copy received February 4, 1982 is, of course, vital to the generation of annealing stresses, but it has
*Member, the American Ceramic Society. virtually no effect on heat transfer in the lehr.)
With these provisions, the principal design parameters of a lehr

Return
TC Air
” Intake
’-..
e are the number of zones into which it is divided, the lengths of the
zones, and the diameter and spacing of the cooling tubes used in
Duct txnaust the various zones. Neglecting edge effects and considering only the
central region of a wide ribbon, the distance between the ribbon
and the plane of the tubes is not a significant variable. The same
Fan
*Membrane stresses are stresses in the plane of the glass that may be superimposed

I ‘I

Glass 7 - Cooling Tubes

on ordinary annealing stresses. While the latter vary through the thickness of the glass
and are balanced over that thickness, membrane stresses vary across the width of the
ribbon and must balance over this width. They are caused by lateral temperature
variations in the glass ribbon, which are not inherent in the cooling process, so that
membrane stresses-unlike ordinary annealing stresses-can, in principle, be avoided
altogether (Ref. 7 ) . Membrane stresses can also cause a variety of problems tn
production (Refs. 8 and 9) and their control (or avoidance) is often more important for
good operation of a flat glass line than the attainment of some desired low level of
ordinary annealing stress.
Unlike ordinary annealing stresses, which are preserved in plates cut from the glass
ribbon, membrane stresses are greatly diminished as the ribbon is cut up. Thus
residual membrane stresses in plates are less likely to be troublesome than the maxi-
mum transient membrane stresses existing in the ribbon before it is cut. If one is
interested in membrane stresses, it is not sufficient, therefore, to consider only re-
sidual membrane stresses in the ribbon, which are determined in the annealing section
of the lehr. One must also consider temporary membrane stresses throughout the lehr,
and therefore lateral as well as longitudinal temperature distributions in the entire glass
ribbon. This greatly increases the complexity of the problem and the number of
Fig. 1. Longitudinal section through one zone of an annealing lehr oper- variables involved, especially since heat transfer conditions and the design of zones
ated with air recirculation. in the cold end of a lehr are very different from those in the hot end.

372
August 1982 Modelling Annealing Lehrs for Flat Glass 373
goes for the wall thickness of the tubes. The important operating 600
parameters include, first, the thickness and velocity of the glass TAC SCMR PWR
ribbon to be annealed. They determine the mass flow rate of the OC n m / c m kW/m
glass and remain the same throughout the lehr. All other operating Linear 214 -
parameters must be defined for each zone. They include the desired 25 195 2.3
glass temperatures at the ends of each zone and the mass flow rates 200 2t3 127
and inlet and outlet temperatures of the cooling air for each zone.? 300 28 2 95
Altogether, the thermal model of the lehr is a relatively simple
one, intended mainly to highlight effects of the most important
variables. Secondary heat transfer mechanisms may also have to be
taken into account in calculations intended for actually designing
a lehr. EQ 500
(3) Computer Simulation of a Lehr
The computer program LEHR calculates glass temperatures
from operating conditions of the lehr and annealing stresses in the
glass from these temperatures, taking into account both stress and
structural relaxation in the glass.4 It may be used in several differ-
ent ways.
In the “design” mode it will calculate the lengths of the zones
required to attain desired interzonal glass and coolant temperatures
for specified glass and coolant flows, using cooling tubes in a
specified arrangement. 400
More generally, the design of a lehr and all parameters pertaining 0 10 20 30 40
to the glass ribbon will be specified. Of the three parameters Position Along Lehr, X [rn]
defining the coolant stream in each zone, one may be chosen more
or less at will. The other two must then be determined by the Fig. 2. Longitudinal temperature distributions in a glass ribbon
simultaneous solution of the heat balance for the zone and the annealed in one zone with cooling air at various temperatures
equations governing heat transfer. The program LEHR can there- (cf. example 1).
fore also be operated in several “running” modes, depending on
which operating parameters one wishes to vary at will.
and how this, in turn, affects the annealing stresses they produce
There is also a ‘‘linear’’ mode to simulate the performance of
and the energy economy of their operation. Individual results and
lehrs as they are commonly conceived. In this mode, heat transfer
their preliminary discussion are presented in this section. More
in the lehr is not considered, and LEHR simply calculates an-
discussion, pulling together results from several examples, follows
nealing stresses in a glass ribbon on which a cooling schedule in Section IV.
consisting of zone-wise constant cooling rates is impressed.
All examples deal only with the “hot end” of lehrs, i.e. with
In the examples illustrating this paper, coolant streams are speci-
cooling glass through the annealing range and a little below it. The
fied by the desired air inlet temperature to each zone. Calculations
number of zones involved will be varied, as will their lengths.
must therefore begin with a determination of the coolant flow rates
However, all will have cooling elements constructed of 7.5-cm-
required to operate the lehr. The sequence of calculations is thus:
diameter tubes with a center-to-center spacing of 15 cm. All
( 1 ) For each zone, LEHR finds that combination of coolant flow
will be used to anneal a 3-mm-thick glass ribbon that is being
rate and coolant outlet temperature which, together with the speci-
produced at a speed of 21 cm/s.* Terminal temperatures of the
fied coolant inlet temperature, will simultaneously satisfy the heat
glass and flow rates and terminal temperatures of the coolant will
balance for the specified temperature drop of the glass and the
be varied with a view to exploring their influence on the per-
conditions for the requisite heat transfer. formance of lehrs.
(2) LEHR then calculates the longitudinal temperature distribu-
tion in the glass ribbon, which is equivalent to determining how (1) Annealing in Only One Zone
cooling rates of the glass vary along the zone. The examples of this section serve two purposes. They consti-
(3) Steps 1 and 2 are repeated for all zones in the annealing tute a first look at heat transfer in a lehr and at how cooling rates
section of the lehr, taking the calculated outlet temperature of the vary -or can be made to vary -within a single zone. Second, by
glass from one zone as the inlet temperature to the next. cooling the glass ribbon through the entire annealing range (from,
(4) From the temperature distribution in the ribbon throughout say, 600” to 400°C) in just one zone, one can also explore the
the annealing section, LEHR calculates transient stresses in the extent to which modulation of cooling rates in one zone can affect
glass ribbon as it is being cooled and the residual (annealing) stress the state of anneal produced.
frozen into the ribbon as it passes out of the annealing range. Example I : The first such single annealing zone to be consid-
( 5 ) Finally, to facilitate energy considerations, LEHR also calcu- ered is 40-m long and operated with various coolant inlet tem-
lates the enthalpy of the air inlet and outlet streams, the heat peratures, TAC.$ Figure 2 shows that different coolant inlet tem-
extracted from the glass (by difference), and the power dissipated peratures produce quite different thermal histories in the same glass
by the flow of the cooling air through the lehr. ribbon, cooled through the same temperature interval. For a low
TAC, such as 25”C, the temperature distribution along the glass
111. Results of Computer Simulations of Lehrs ribbon is “convex up,” which means that the cooling rate of the
glass increases as the glass passes through the lehr. The opposite
Considering the great variety of lehrs and the large number of holds for a high TAC, such as 300°C. The change in shape of these
variables involved in their design and operation, generalizations temperature distributions can readily be explained. (Cf. below and
about them may be both hard to come by and hazardous to use. The Section IV(l).) The important point to note is that the increasing
approach taken here is to present the results of computer simu- cooling rate that goes with a convex temperature distribution is
lations of a series of comparable-and progressively more what one usually seeks to achieve in successive zones of a con-
complicated -lehrs, and to examine their thermal performance ventional, multizone lehr, and here we see that it can also be
’To control the lateral uniformity of glass temperatures and thus avoid membrane obtained in a single zone.
stresses, different cooling tubes-or groups of tubes-may, in practice, have separate
air supplies. The upper and lower tube banks may also be operated differently to ‘In designing a lehr, one would also have to consider its performance on glass
compensate for the interference of rollers with heat transfer. In the present analysis this ribbons of different thicknesses, notably on the thickest ribbon to be processed at the
interference is neglected and all tubes in a zone are taken to be supplied with air at hi hest applicable rate.
the same temperature and at the same flow rate. %or definitions of symbols, see footnote to Table 1.
 374 Journal of the American Ceramic Society-Gardon Vol. 65, No. 8
Table I. Performance of a Single Zone (40 meters Long) Table 11. Performanceof a Single Zone (60 meters Long)
Cooling Glass Ribbon From 600 to 400°C (Example 1)* (Examples 2 and 3)
TAC A-G TAH RH RC RH-RC SCMR PWR TAC A-G TAH RH RC RH-RC SCMR PWR
("0 ("C) ("C/S) ("C/S) (nm/cm) (kW/m) ("C) ("C) ( W S ) (T/s) (nm/cm) (kW/m)
-Linearcooling- 1.05 1.05 1.00 21.4 - Glass cooled from 600"to 400°C [Example 2)
25 0.38 550 0.40 1.42 0.28 19.5' 2.3 -Linear cooling- 0.70 0.70 1.00 14.3 -
50 0.40 545 0.45 1.38 0.33 19.5 2.8 25 0.36 596 0.09 1.41 0.064 15.3. 2.8
100 0.46 537 0.55 1.28 0.43 19.6 4.3 100 0.42 583 0.14 1.26 0.11 14.2 5.0
200 0.64 510 0.89 1.03 0.86 20.9 12.8 200 0.54 569 0.29 1.00 0.29 13.2* 12.1
220 0.71 503 0.99* 0.97' 1.02 21.4 17 292 0.80 541 0.65+ 0.67t 0.98 14.3 40
300 1.28 456 1.70 0.65 2.59 28.2 95 300 0.85 535 0.72 0.62 1.17 14.7 48
~ ~ ~~~ ~

*Definition of column headings: Glass cooled from 600" to 300°C (Example 3)

TAC=cold end air temperature, i.e. coolant inlet temperature; -Linear cooling- 1.05 1.05 1.00 21.4 -
TAH=hot end air temperature, i.e. coolant outlet temperature; 25 0.62 506 0.93$ 0.77'
A/G=ratio of heat capacities of air and glass streams; 1.20 22.2 11.7
RH=cooling rate of glass, hot end of zone; 100 0.81 470 1.39 0.67 2.08 26.8 26
RC=cooling rate of glass, cold end of zone; 200 1.75 372 2.71 0.44 6.21 42.9 229
SCMR=stress, ealulated midplane residual;
PWR=nominal power requirement for cooling air flow. *Minimurn value. %is temperature distribution is close to linear, but the cooling
'Minimum value. rate varies slightly and nonmonotonically; it has a maximum value of
'This temperature distribution is close to linear, but the cooling rate varies slightly 0.72"C/s. %e cooling rate vanes nonmonotonically and has a maximum value of
and nonmonotonically. It has a maximum value of l.O9"C/s. 1.17"C/s at a glass temperature of =520"C.

The average cooling rate of the glass is I.O5"C/s. Cooling rate, so that the power required to operate this zone with linear
the glass at this rate throughout the zone, i.e. "linear" cooling, cooling would now become much higher (40 kW/m width of lehr).
would produce a residual midplane tension ("annealing stress") of However, as before, there is really no advantage in operating the
21.4 nm/cm.n However, one can also do better than that. Thus, for lehr in this manner. The same residual stress can also be obtained
example, using a coolant inlet temperature of 25"C, one can reduce with a TAC of IOO'C, at a cost in power of only 5 kW/m. Opera-
the residual stress to 19.5 nm/cm. tion with a coolant inlet temperature of 200°C would yield the
Table I illuminates the performance of this "single-zone lehr" in lowest residual stress obtainable with this zone, though it is ques-
greater detail. It shows that a virtually constant cooling rate can be tionable whether the slight reduction in stress to 13.2 nm/cm is
achieved in practice with an air inlet temperature of 220°C and an worth the considerable increase in power consumption to
air rate corresponding to which the ratio (A/G) of the heat capaci- 12.1 kW/m.
ties of the air and glass streams is 0.71. It is interesting to note that In this example the lowest stress is not obtained with the lowest
this is significantly less than the air-rate corresponding to coolant inlet temperature, as it was in the shorter zone. Operating
A/G= 1 .O, which would be required for linear cooling in a heat the 60-m zone with a coolant inlet temperature of 25°C still leads
exchanger in which a constant overall heat transfer coefficient was to the lowest power requirement, but yields a somewhat higher
applicable. (In annealing lehrs the "effective" overall heat transfer residual stress than the quasi-optimal operation with a TAC of
coefficient can vary considerably along the length of a zone on 200"C, or even linear cooling. To be sure, the longer zone pro-
account of the strong temperature dependence of the rate of radiant duces lower stresses under all operating conditions, but that
heat transfer. To maintain a constant cooling rate in spite of this is as expected, and it is obtained at a considerable increase in
variation, the higher heat transfer coefficient at the hot end of a capital cost.
zone must be offset by a locally smaller overall temperature differ- The conclusion to be drawn here is that cooling, even within a
ence between the glass and coolant, which, in turn, requires that single zone, will not, in general, be linear. Departure from linearity
A/G in a countercurrent heat exchanger be less than unity.) may help or hurt, for it may decrease or increase the annealing
The lower annealing stress of 19.5 nm/cm, produced with a stress produced in a given length of lehr. The important point is that
TAC of 25"C, is obtained with a cooling schedule in which the cooling rates can be varied in a constructive manner even within a
cooling rate increases from O.4O0C/s at the hot end of the zone to single zone and there is no intrinsic merit to cooling glass at a
1.42 at the cold end, i.e. by a factor of 1/0.28. The required constant rate. While the residual stress produced for prescribed
coolant flow rate corresponds to an A/G ratio of only 0.38. Con- conditions of glass production depends primarily on the length of
versely, with a coolant inlet temperature of 300°C the residual lehr available to cool the glass through the annealing range, modu-
stress would go up to 28.2 nm/cm. lation of the cooling rate--even within a single zone-can also
The advantage of operating with a low coolant inlet temperature have a marked effect. .
becomes even more pronounced when one considers the power Example 3: For the last example of this series, a 60-m-long
dissipated by the flow of the cooling air through the lehr. This is zone is used to cool the glass ribbon of the preceding examples
only 2.3 kW/m width of lehr for an air inlet temperature of 25", from 600" to 300"C, i.e. through a temperature interval 1.5 times
as against 95 kW/m for an air inlet temperature of 300°C. The as large as before. The average cooling rate through the 60-m zone
lower power requirement for the lower coolant inlet temperature thus becomes the same as it was in the 40-m zone of the first
arises principally from the fact that less coolant is required. example, so that linear cooling through the entire zone should
Example 2: The lowest coolant inlet temperature and most again produce an annealing stress of 21.4 nm/cm. (Cf. lower part
strongly convex temperature distribution is not always the most of Table 11.)
favorable one. This is shown in the upper part of Table 11, which It turns out that-taking 25°C to be the lowest practicable
summarizes the results of calculations similar to the above, but for coolant inlet temperature -it is impossible in practice to attain this
a zone that is 60-m long. stress with this lehr. The problem with attempting to cool the glass
In line with the lower average cooling rate, linear cooling of the through so wide a temperature interval in a single zone is that the
glass in this longer lehr would produce a residual midplane tension coolant must pick up a large amount of heat before it reaches that
of only 14.3 nm/cm. Whereas in the previous example nearly part of the zone in which the glass is cooled through the annealing
linear cooling could be achieved in practice with a coolant inlet range. The conditions for picking up this heat, at the cold end of
temperature of 220°, now a coolant inlet temperature of 292°C the zone, would be unfavorable because the rate of radiant heat
would be required. Offsetting the advantage of a longer lehr, this transfer from the glass to the tubes is sharply lower at low tem-
higher coolant temperature would require a higher coolant flow peratures. To offset this, the system would call for higher air rates,
which produce higher convective heat transfer coefficients in the
"This is really a stress-induced birefringence, which corresponds to a stress of tubes and maintain larger temperature differences between the
7.9 kg/crn ( I 13 psi). Since the only way to measure such internally balanced stresses
is via the birefringence they produce, the term ''stress'' will be used, for brevity, even glass and coolant. These factors would also enhance heat transfer
where measures of stress-induced birefringence are cited. at the hot end of the zone, so that the heat transfer rate there would
 August 1982 Modelling Annealing Lehrs for Flat Glass 375
Table 111. Performance of a “Two-Zone”Lehr (40 meters Long) Cooling Glass Ribbon From 600”to 4OO0C*
Example Zone TAC A-G TAH RH RC RH-RC SCMR PWR
No. No. (“C) (“C) (“CIS) (“C/s) (nmlcm) (kwlm)
Zone-wise linear cooling
4.L 1 0.61 0.61 1 .oo
2 1.71 1.71 1.oo
End result/Total 16.6 -

4.a 1 200 0.19 575 0.13 1.56 0.09 0.3

2 25 0.52 276 1.83 I .46 1.25 L!5
End result/Total 23.8 1.9
For Examples 4.b-f operating conditions (cols. 3-8) are given for zone 1 only, conditions in zone 2 remain the same as in Example 4.a,
and the PWR shown is the total for both zones
4.b (1) 300 0.26 570 0.20 1.30 0.15 21.9 2.4
4.c (1) 400 0.44 560 0.34 0.95 0.36 19.4 5.8
4.d (1) 450 0.70 550 0.51 0.71 0.73 17.9 19
4.e (1) 475 1.03 543 ’ 0.68 0.55 I .22 16St 55
4.f (1) 500 2.23 53 1 0.98 0.37 2.67 15.5 487
*Zones 1 and 2 are 24 and 16 m long, respectively, and the glass passes from one to the other at TB=530”C. ‘Matches stress produced by Lone-wise linear cooling.

become very high. Thus, relatively little heat would be transferred 600
in the major part of the zone near its cold end and a great deal in TAC 1
a short region near the hot end, so that cooling curves would be O C

markedly “concave” (as reflected by values of RH/RC> 1) for all

coolant inlet temperatures. The average cooling rate of the glass
through the annealing range proper would thus be higher than the 475
average rate through the zone as a whole. 500
(2) Annealing in Two Zones
The preceding example shows why multizone lehrs are always
used in practice: Dividing a lehr into zones, each having its sepa-
rate coolant supply, gives one closer control over glass tem-
peratures and thus more leeway to modulate cooling rates at will.
It also enables one simply to maintain near-linear cooling over an
extended section of a lehr3 where, as in the preceding example, this
would not otherwise be possible.
To maintain comparability with Example 1, the same glass rib-
bon will be cooled through the same temperature interval in a lehr
of the same design, except that now the cooling load is divided 25 O C
between two zones having a combined length of 40 m. Going to
such a “two-zone 1ehr”-or, more precisely, a lehr having a two-
-
zone annealing section-gives one a free choice of the length of
400 I
1 I I I I I
one of the two zones, the “break temperature” (TB) at which the
glass passes from the first zone to the second, and of two separate 0 10 20 30 40
coolant systems.
Example 4 : The choice of optimal cooling rates and corre- Position along Lehr, X Lrn]
sponding break temperatures for lehrs in which the cooling rate is
constant within each zone was discussed in Ref. 6. For the present Fig. 3. Longitudinal temperature distributions in a glass ribbon
annealed in two zones with cooling air at various temperatures
case, a division of the lehr into two zones of 24 and 16 m and a TB (cf. example 4).
of 530°C would appear desirable. This corresponds to average
(linear) cooling rates of 0.61 and I .71 OC/s in the two zones. (Cf.
Ex. 4.L, Table 111.) thus negating one of the reasons for going to a two-zone lehr, with
Figure 3 shows the ideal cooling schedule with zone-wise con- a “break” in cooling rates, in the first place. The reason for this is
stant cooling rates (dashed lines) and the cooling schedules actually that, as TACl is reduced, the negative (“convex up”) curvature of
realizable in the given lehr with various coolant inlet temperatures the temperature distribution in zone 1 becomes more pronounced.
(solid curves). In zone 2, a reasonably close approximation to Thus the cooling rate in the most sensitive region of the annealing
linear cooling can be realized with a coolant inlet temperature range is increased at the cost of time uselessly spent in cooling the
(TAC2) of 25”C, and this is the only TAC considered for this zone. glass more slowly near the upper limit of the annealing range.
For zone 1, values of TACl from 200“ to 500°C are considered. Another way of looking at this is that, with the cooling rate at the
Details of the performance of this lehr under various operating end of the first zone (RCI) increased and approaching that at the
conditions are shown in Table 111. A cooling schedule consisting of beginning of the second (RH2= 1.83 “C/s), the desired “break” at
two zone-wise linear segments would produce a residual stress of 530°C -which had been chosen for two zone-wise constant cool-
16.6 nm/cm. This can be achieved in practice with a TACl of ing rates-is, in effect, weakened and the “effective break” is
475°C. However, the corresponding, relatively small temperature moved upstream from the junction of the two zones.
difference between the glass and coolant must be compensated for Example 5: These examples serve to explore the vicinity of
by a rather high air rate (A/G=1.03), so that the power con- the optimal operating point in more detail. In them, the lengths of
sumption for maintaining the coolant flow is high, possibly too the two zones remain the same as before, but now TB -the “break
high to be practical. Reducing TACl lowers the power con- temperature” at which the glass passes from one zone to the
sumption but also produces progressively higher residual stresses. other-will be varied as well as coolant flow rates and tem-
Indeed, as TACl is reduced to below ==320”C, the residual stress peratures. The results are shown in Fig. 4.
becomes higher even than the 21.4 nm/cm that would be obtained We have seen that, if the glass were cooled at a single, constant
with linear cooling through the entire 40-m length of the two zones, rate throughout the entire 40-m length of both zones, the annealing
 -5
376
24 TAC I oc
Journal of the American Ceramic Society-Gardon Vol. 65, No. 8

-
350 TAC2 SCMR PWR
OC n m / c m kW/n
300 15.0 3.3

-0

(3
400 14.5 7.9

F
L

a,
L
-
1
500-
0)
a
E
-
100
? -
I I I u)

470 500 530 51

-
u)
0 -
Break T e m p e r a t u r e , T B [ OC ] (3
-
Fig. 4. Annealing stresses obtainable in a given two-zone lehr by varying
the coolant inlet temperature to zone 1 and the “break temperature” at
which the glass passes between zones (cf. example 5 ) .
0 10 20 30 40
Position Along Lehr, X [rn]
stress would be 21.4 nm/cm. We will regard this stress level as
“par” and seek to do better by producing annealing stresses lower Fig. 5. Near-optimal temperature distributions and the stresses produced
by them in a glass ribbon annealed in three zones (cf. example 6).
than par. The bold, dotted curve shows annealing stresses as a
function of TB for cooling schedules consisting of two zone-wise
linear segments. The curve is roughly parabolic, with a well-
defined stress minimum and stresses lower than par over a fairly stress obtainable in a 40-m length of lehr. As Fig. 5 shows, this
wide range of TB’s. favorable cooling schedule entails cooling fast, then slowly, then
The thin curves of Fig. 4 show the annealing stresses produced fast again. The figure also shows temperature distributions in this
by real-life cooling schedules. The curves are labelled with the lehr-and the annealing stresses produced by them-both for the
coolant inlet temperatures to zone 1 (TAC1) for which they apply. ideal case of three linear segments and for the real options of
(TAC2 was again kept at 25°C.) The point symbols on the curves cooling with the coolant inlet temperatures indicated. The an-
correspond to power requirements (for “once-through’’ flow) of nealing stress expected with three linear segments can be attained
5, 10, 20, 40, and 100 kW/m width of lehr, and those parts of with coolant inlet temperatures to the three zones of 300”, 400°,
the curves for which power requirements exceed 100 kW/m are and 25”C, respectively, even though the corresponding temperature
dashed. These curves resemble the dotted curve for two zone- distributions in two of the zones are clearly not linear. This again
wise constant cooling rates, only that the range of TB’s for which shows that there is no intrinsic merit to linear temperature distribu-
annealing stresses are below par is narrower and the stress minima tions even within a zone. From the point of view of both annealing
are generally not as low as the minimum expected with zone-wise stress and power consumption, this three-zone lehr is preferable to
linear cooling. Only for the rather high TACl of 475°C is the any of the two-zone lehrs surveyed in Table I11 and Fig. 4.
lowest stress obtainable (16.0 nm/cm) not only 25% lower than
par, but 4% lower even than the best attainable with two zone-wise (4) Another Two-Zone Lehr
constant cooling rates. However, this seemingly favorable result is The cooling schedules considered so far were all practically
of questionable practical value because its attainment requires a realizable variants of zoqe-wise linear schedules selected for what-
power consumption > 100 kW/m. The reason for this high power ever merit the latter might offer. However, it is not really necessary
requirement has already been touched on. Operation with a TACl to make segmentally linear cooling schedules one’s point of de-
of 450°C and a TB of 520°C would produce almost as low an parture. Thus, seeking to profit from our new-found ability to
annealing stress with 25% less air and at less than one-half the cost modulate cooling rates even within a zone, we next attempted to re-
in power for circulating the air. produce in only two zones the advantageous fastislowlfast cooling
Some additional calculations, in which the relative lengths of the schedule demonstrated in the preceding example in three zones.
two zones were also varied, showed that moving away from the After some trial and error, costing no more than a few minutes of
seemingly optimal 24: 16 ratio increases the residual stresses ex- computer time, the schedule shown in Fig. 6 was developed.
pected with zone-wise linear cooling. However, partly offsetting Example 7: In this schedule there is a well-defined break in
this trend, the lowest stresses actually obtainable with nonlinear cooling rates at 570”C, at the end of a short first zone. In the second
cooling in each zone turn out to be lower than those corresponding zone, a single continuous curve takes the place of the two curve
to zone-wise linear cooling over wider ranges of TACl and TB. segments in zones 2 and 3 of the three-zone lehr. The residual stress
All these examples suggest that, for lehrs in which cooling rates of 15.9 nm/cm, obtained in this manner in only two zones,
within a zone vary significantly, comparisons with cooling sched- is somewhat higher than the best obtained with three zones, but
ules consisting of linear segments having the same lengths as the the power consumption is clearly less. While it remains to be
zones in question are meaningless. This has important implications established whether such a “two-zone” lehr is really practical, on
for how one sets about to design a lehr, a point to be discussed first glance it appears to be an interesting alternative to the more
further in Section IV( 1). conventional three-zone lehr. It certainly is better than any of the
two-zone lehrs of the same length (cf. Example 4) that were de-
(3) Annealing in Three Zones rived from what would be a near-optimal lehr if one were restricted
Example 6: Using Ref. 11 as a guide, the lengths of the three to linear cooling within a zone. On the other hand, consideration
zones (3,24, and 13 m) and the two break temperatures (575” and of zone-wise linear cooling would hardly have led to the selection
510°C) were chosen to yield close to the lowest possible annealing of this lehr.
August 1982 Modelling Annealing Lehrs for Flat Glass 377
600 Table IV. Annealingin a 40-meterLength of Lehr*
Coolant inlet temp. (T) Annealing power
Net
Example stress
No. Zone 1 Zone 2 Zone 3 (nrn/cm) (kW/rn)
TAC2 SCMR PWR Ideal, segmentally linear cooling schedules
OC n m / c m kW/m 1 One constant cooling rate 21.4
25 15.9 1.7 4 Two linear segments' 16.6
Linear 22.7 - 6
7
Three linear segments*
Two linear segments'
14.6
22.7
Annealing in one Lone
I 25 19.Sn 2.3
100
500 -
19.6 4.3
220 21.4" 17
300 28.2 95
- 4 200
Annealing
25
in two zonest
23.8 1.9
- 300
400
25
25
21.9
19.4
2.4
5.8
450 25 17.9 19
- 475
500
25
25
16.5"
15.5
54
487
- 6 300
Annealing
300
in three zonesr
25 15.0 3.3
300 400 25 14.5" 7.9
4001 ' I I I I I I Annealing in two zonesP
0 10 20 30 40 7 200 25 15.9 17
*Cooling glass ribbon (3 mm thick at 21 cm/s) from 600" to 400°C. 'Lengths 24
Position Along Lehr, X [rn] and 16 m, TB=53O"C. 'Lengths 3, 24, and 13 m; TBI =5754 TB2=51O0C. "Lengths
3 and 37 m, TB=57O0C. "Minimum value. "Matches stress produced by zone-wise
Fig. 6. Temperature distributions and stresses produced by them in a linear cooling.
glass ribbon annealed in two zones (cf. example 7).
oped, it is advisable also to consider the performance of the lehr as
it might actually be operated with real-life coolant streams. It may
IV. Discussion then be found necessary to modify one or more of the assumptions
underlying one's first design. Indeed, as Example 7 shows, some
(1) The Shape of Cooling Curves and the Generation of An- advantageous designs can be developed without any reference to
nealing Stresses
segmentally linear cooling schedules.
The examples of Section 111 show that the longitudinal tem- To put this into perspective-and, incidentally, also to explain
perature distribution in a glass ribbon can take a variety of shapes, the particular effectiveness of fasffslowifast cooling schedules-
even within a single zone and between two fixed end points. The an annealing schedule can be expected to be the better the closer
positive or negative curvature of temperature distributions-and instantaneous cooling rates are to being inversely proportional to
hence cooling schedules-is determined by the interplay between the temperature- and cooling-rate-dependent weighting function
the temperature dependence of radiant heat transfer and the flow W(T) introduced by Narayanaswamy" in his optimization analy-
rate of the coolant, which largely determines how temperature sis. In turn, cooling rates can the more readily be tailored to meet
differences between the glass and coolant vary along the length of this condition the more freely one is allowed to vary operating
a zone. and/or design parameters of a lehr. Operating an existing lehr, one
The examples also show that there is no intrinsic merit to cooling may wish to vary not only coolant flows and temperatures-and
glass at a constant rate: certainly not, as has long been known, hence the shape of cooling curves within zones-but also the
through an entire lehr, and not even through individual zones of a temperatures at which the glass is passed from one zone to the next.
lehr. While departure from linearity may help or hurt, the important When designing a lehr, one can, of course, also vary the number
point is that one can now predict the effect of any changes in and lengths of zones in the annealing section.
cooling rate and, if desired, vary that rate in a constructive manner.
Thus, even if one anneals glass in a single zone-which, con- (2) Comparison of Various Annealing Schedules
ventionally, would be thought of as involving cooling at a constant Table IV pulls together the performance characteristics of some
rate-one can significantly affect the annealing stress produced of the lehrs considered in Section 111, all of which have annealing
and the power required for operation of the lehr by varying the inlet sections 40-m long. The first part of the table helps to compare the
temperature and flow rate of the coolant and hence the shape of the performance of "ideal" lehrs, as they are commonly conceived,
cooling curve. which are supposed to operate with linear cooling in each of the
The same considerations also apply to individual zones in multi- zones. The two-zone lehr is seen to be better than the one-zone
zone annealing sections. In multizone lehrs modulation of cooling lehr, and the three-zone lehr better than either of the other two.
rates has long been practiced. However, as has been shown here, The second part of Table IV shows the performance of the same
the common assumption of linear cooling in each zone is not really lehrs when operated with cooling schedules that can actually be
necessary, for one can readily calculate the cooling schedules actu- realized in practice. In addition to zone lengths and interzonal glass
ally obtained with different modes of operation of the lehr; and now temperatures, these schedules are characterized by the coolant inlet
one can also calculate the stresses produced in the glass by any of temperatures required to produce them, and their performance by
these cooling schedules-whether they be wholly linear, consist of the annealing stresses produced and the energy cost of operation.
linear segments, or be continuously curved. The single-zone lehr is the simplest, and the stress obtainable is
It was shown, notably in Examples 5 and 7, that comparisons lower than might have been expected on the basis of initial consid-
with cooling schedules consisting of zone-wise linear segments erations involving linear cooling, 19.5 nm/cm vs the 21.4 nm/cm
have little meaning for lehrs in which cooling rates vary signifi- expected. However, it should again be noted that, depending on the
cantly within a zone. In view of this, and until better guidelines for coolant flow, the performance of a lehr can be better or worse than
the design of lehrs are developed, one may well start the devel- that corresponding to linear operation.
opment of a new design by considering zone-wise linear cooling Two-zone lehrs-in which glass is intended to be cooled slowly
schedules. However, once a preliminary design has been devel- in the first zone, faster in the second-are usually designed for
 378 Journal of the American Ceramic Society-Gardon Vol. 65, No. 8
what appear to be desirable zone-wise linear cooling schedules. provision of additional degrees of freedom in operating a plant,
However, their actual performance, when operated with realistic where such flexibility is not really required, may be a handicap
cooling schedules, is likely to be less good than expected. They rather than an advantage.)
can, indeed, produce annealing stresses as low or lower than those The preferred way to control air inlet temperatures, and hence
expected with zone-wise linear cooling, but only at a considerable heat transfer in the lehr, with minimal waste of energy is to re-
(or even unacceptable) cost in energy. When operated with lower circulate the coolant. In this system, part of the air stream leaving
coolant inlet temperatures to save energy, they produce annealing the hot end of a zone is pumped back to the cold end, mixed with
stresses higher even than the stress that would be produced by room air to attain the desired air inlet temperature, and recirculated
linear cooling in a single zone. It is interesting to note that a through the zone, as indicated in Fig. 1. This way the only heat lost
reduction of annealing stresses in the two-zone lehr of Example 4 to the atmosphere is that extracted from the glass, and the “preheat”
requires higher coolant inlet temperatures and, with that, more is obtained free, from what would otherwise be waste heat.
power for coolant circulation. The opposite is the case for the To illustrate the effectiveness of recirculation as an energy-
one-zone lehr of Example 1, in which lower stresses go with lower saving system, we will consider operation of zone I of the two-
power requirements. (See, however, also Example 2, Table 11.) zone lehr of Example 4 in various ways. This is the lehr in which
The three-zone lehr is representative of the best current practice annealing stresses can be reduced by raising TAC 1, the inlet tem-
in that it produces a low annealing stress and requires relatively perature of the coolant to zone 1 (cf. Table 111). For a benchmark,
little power for operation. the amount of heat to be extracted from the glass in this zone is
The last of these examples was intended to explore the idea of 147 kW/m width of lehr. The nominal fan power required to
producing a desirable fastislowlfast cooling schedule in only two operate the zone is 0.8 kW/m if TAC1=30OoC and 17 kW/m if
zones. The performance of the rather unconventional two-zone lehr TACl =450°C.
of Example 7 does indeed come close to that of the three-zone lehr Operating the zone without recirculation, the energy required to
in point of residual stress, and its power consumption is lower. It is bring room air to the desired coolant inlet temperatures would be
also the best of the two-zone lehrs considered, in spite of the fact 153 kW/m width of lehr for a TACl of 300°, and 633 kW/m for
that with zone-wise linear cooling (cf. upper part of Table IV) it a TAC 1 of 450°C. This is from I to 5 times the heat extracted from
would appear to be a rather poor choice, less effective even than the glass and orders of magnitude larger than the power required
linear cooling in a single-zone lehr. to blow air through the cooling tubes in the lehr.
Finally, Table IV also shows that, for a lehr of given length, Calculations on the recirculating system itself (as distinct from
good multizone operation can reduce annealing stresses by =30%. those on a lehr operated with recirculation) require additional data
Conversely, the stresses produced in a single-zone lehr can be on the piping external to the lehr. Here we assume that groups of
matched by a multizone lehr 30% shorter. Considerable scope also seven of the 7.5-cm-diameter cooling tubes in the lehr have a
exists for reducing annealing stresses by modulating cooling rates common external return duct of 20-cm diameter, which makes for
within zones. The effect of such modulation can be especially about the same air velocity in the return ducts as in the cooling
marked in zones in which the glass is cooled through all (or a large tubes. For this representative case it was found that a TACl of
part of) the annealing range, as in Examples 1 and 7. 300” C can easily be maintained by recirculation, even with an
uninsulated return duct, since, as Table I11 shows, the exhaust
(3) Energy Economy of Lehrs temperature (TAHl) of 570°C is much higher than the desired inlet
Annealing entails the controlled cooling of a freshly formed temperature. On the other hand, the exhaust temperature of 550°C
glass ribbon. Why should this process require any energy input that goes with a TACl of 450” is only 100°C higher. Thus it is not
other than the heat already in the glass and the modest power surprising that maintenance of a TACl of 450°C would require a
required to transport the glass? small addition of extra heat to the coolant supply or insulation of
One could conceive an enclosure--only moderately insulated, the return ducts.
with the amount of insulation varying along its length-which acts These results suggest that with recirculation any desired coolant
as a kind of radiation shield around the hot glass ribbon, such that inlet temperature can be maintained without the need for additional
the glass “naturally” cools at the desired rate (or rates). However, heat input. High coolant inlet temperatures need not, therefore, be
even if this approach worked, it would afford one no control of the precluded on the basis of the energy cost of maintaining them.
annealing process in the face of changing ambient conditions or However, the energy cost of circulating the large volumes of cool-
desired changes in the width, thickness, or rate of production of the ant that are required with high inlet temperatures can become
glass ribbon. Hence the lehrs as we have come to know them, prohibitive (cf. Table 111). Low coolant inlet temperatures, on the
which are well-insulated enclosures in which heat is abstracted other hand, go with small coolant flows, which require less power
from the glass by a coolant. Power is, of course, required to for circulation, and with high coolant outlet temperatures, which
circulate the coolant. make maintenance of the desired coolant inlet temperatures easier.
The examples of Section 111have shown that it is often necessary Other things being equal, one should, therefore, opt for the lowest
to operate lehrs with elevated coolant inlet temperatures. But does practicable coolant inlet temperatures compatible with the condi-
this also necessitate an expenditure of energy, as is often practiced? tion of the glass and the design of the zone in question.
(A) Control of Coolant Temperatures: The simplest way to ( B ) Fan Power for Recirculation: To keep it simple, the
cool a lehr would be to draw room air through the cooling tubes and computer program LEHR calculates power requirements only for
to discharge it at the hot end of each zone. The heat extracted from “once-through” flow of the cooling air. The power required for a
the glass would thus be lost to the atmosphere. Since it represents recirculating system will usually exceed that for once-through flow
waste heat at a relatively low temperature, such a loss is generally by a factor no greater than 2, depending on the design of the piping
accepted, even these days. Two more serious drawbacks of this external to the lehr. In the preceding example, in which the air
arrangement would be that, for some cooling tasks, room- velocity in the return duct is the same as that in the seven smaller
temperature air is too cold a coolant and that--even where this is diameter cooling tubes it serves, the nominal power requirement
not the case-one would have no control over the inlet temperature for a recirculating system is -- 1.3 times that for straight-through
of the coolant in the face of changes of ambient temperature. flow. The actual power requirement will, of course, also have to
One way to overcome these drawbacks, still using “once- take into account motor and fan losses, and losses due to throttling
through’ flow, is to preheat the air. This approach is significantly in the flow control system.
more wasteful of energy, however, for now the “preheat” is lost to The question of what power consumption is economically ac-
the atmosphere as well as the heat extracted from the glass, and the ceptable is beyond the scope of this paper. Nominal power require-
former is often many times larger than the latter. The alternative of ments in the examples ranged from 2 to 10 kW/m width of lehr for
supplying heat directly into the tunnel of the lehr, using radiative a 40-m long section of lehr, operated with relatively low coolant
or convective heaters, is only slightly less wasteful of energy. It inlet temperatures, to over 100 kW/m for high-temperature oper-
also complicates lehr operation, not least by raising questions ation of some none too practical looking cooling schedules. Pro-
about how the extra heat should be distributed along the lehr. (The rated for the same length of lehr, power consumptions of 5 to
 August 1982 Temperature Dependence of the Iron Acceptor Level in Aluminum Oxide 379
10 kW/m seem to be current practice. Somewhat higher power by changes in the shape of temperature distributions in one or more
requirements may be warranted under some circumstances, zones. While, in these cases, there appears to be less scope for
especially in return for benefits-by way of capital savings or reducing annealing stresses below those produced by zone-wise
reductions in annealing stress-that may be realized from new linear cooling, poor choices of operating parameters can grievously
approaches to lehr design and operation. impair the performance of lehrs by increasing annealing stresses
and/or the power required for operation. Computer simulation can
V. Summary and Conclusions help by warning of detrimental trends.
It appears advantageous to use relatively low coolant inlet tem-
The computer program LEHR allows one to predict annealing peratures even for zones in the hot end of an annealing lehr, and to
stresses produced in a flat glass ribbon on the basis of its actual foster this trend by appropriate lehr designs. Lower coolant inlet
thermal history in an annealing lehr. Not only does this approach temperatures always make for lower power requirements for circu-
allow a rigorous analysis of the performance of lehrs, it also re- lating the coolant. Sometimes they also lead to the lowest
moves the limitation to zone-wise linear annealing schedules, on annealing stresses. In other cases lower annealing stresses can be
which the traditional, empirical method of designing lehrs has achieved with higher coolant inlet temperatures, but only at the
been based. cost of higher power consumption.
Computer experiments run with LEHR established that cooling In addition to power for circulating the coolant, some lehrs also
rates, even within a zone, are generally not constant and that use energy to maintain elevated coolant inlet temperatures, even
departures from linear cooling may improve or impair the per- though it seems irrational to put heat into a lehr, the function of
formance of a zone. Thus there is no intrinsic merit to operating a which is to extract heat from the glass. It is suggested that-except
lehr with zone-wise constant cooling rates. In some zones linear during start-up or, possibly, to control membrane stresses-this is
cooling may be readily practicable and as effective as any other. not really necessary, and that with recirculation via an insulated
More often, however, judiciously chosen nonlinear cooling sched- return duct any desired coolant inlet temperature can be maintained
ules are easier to engineer, more economical to operate, and more without any additional heat input into the lehr.
effective in that they produce a lower annealing stress in a lehr of
given length or allow one to obtain a desired annealing stress in a References
IL. H. Adams and E. D. Williamson, “Annealing of Glass,” J . Franklin Inst., 190,
shorter lehr. 597-631, 835-70 (1920).
That nonlinear annealing may be advantageous has long been 2E. B . Shand, Glass Engineering Handbook, Ch. 4 . McGraw-Hill, New
known and is one of the reasons for using multizone lehrs. What York, 1958.
)I. P. Garcelon, “Modern Processes for Flat Glass Annealing and Quality Criteria
has been demonstrated here is that the cooling curve can be shaped of Glass Produced (in Engl.), Verres Refract., 34, 3 4 3 4 9 (1980).
to advantage even within a zone. The effect of this on annealing 4R. Gardon and 0. S. Narayanaswamy, “Stress and Volume Relaxation in An-
nealing Flat Glass,” J. Am. Cerum. Soc., 53 [7] 380-85 (1970).
stresses is likely to be most significant for cases in which the glass 5 0 .S. Narayanaswamy, “Stress and Stmctural Relaxation in Tempering Glass,”
is cooled through all (or a large part) of the annealing range in one J. Am. Cerum. SOC., 61 [3-4] 146-52 (1978).
6R. Gardon, “Nonlinear Annealing of Glass,” J. A m . Cerum. SOC., 64 [ 2 ]
zone. Thus it has been shown that one can anneal glass in a single 114-19 (1981).
zone and, by modulating cooling rates within that zone, still ’G. K . Chui, “Heat Transfer and Temperature Control in an Annealing Lehr for
Float Glass,” J. Am. Cerum. SOC., 60 [11-12] 477-84 (1977).
achieve an annealing stress lower than would be expected on the *R. Gardon and P. A. Velonis, “On-Line Measurement of Membrane Stresses in
basis of linear cooling. It has also been shown that the near-optimal Float Glass,” Truns..lSA, 17 [2] 3-8 (1978).
performance of a three-zone lehr can be approximated by a simpler 9V. Caimann, “Uber den flachenhaften Eigenspannungszustand gerogener
Tafelglaser” (“Plate-Wise Balanced Internal Stresses in Drawn Sheet Glass),
two-zone lehr of the same length, in which the longitudinal tem- Glustech. Ber., 41, 498-505 (1968).
perature distribution in the second zone is shaped to advantage. ]OH. C. Hottel, “Radiant Heat Transmission”, Ch. 4 in W. H. McAdams, Heat
Transmission. McGraw-Hill, New York. 1954.
The performance of lehrs in which individual zones span only a ‘ l o . S. Narayanaswamy. ”Optimum Schedule for Annealing Flat Glass,” J. Am.
small part of the annealing range can also be significantly affected Cerum. SOC., 64 [2] 109-14 (1981).

Temperature Dependence of the Iron Acceptor Level

in Aluminum Oxide
J. B. BLUM,”.” H. L. TULLER,* and R. L. COBLE*
Ceramics Division, Department of Materials Science and Engineering, Massachusetts Institute of Technology,
Cambridge, Massachusetts 02 139

Optical absorption measurements were performed on undoped I. Introduction

and iron-doped A1203single-crystal samples cut from the same
wafer. Measurements were made in the ultraviolet spectral
range at T=25” to 500°C. A shoulder observed at 4.8 eV in the T HE electronic structure of a material can be determined by
measuring the optical absorption as a function of the incident
photon energy. This type of measurement is effective for studying
absorption spectrum of the iron-doped sample at 25°C was
crystal-field transitions and for identifying energy levels within the
attributed to a valence band-to-iron acceptor level transition.
This structure shifted to lower energies as the temperature band gap of the material. The optical absorption spectrum of alumi-
num oxide has been the subject of many investigations over the
increased at a rate of 5 . 6 ~ 1 0 eV/”C.
-~
past 20 years. Investigations have generally been performed at
room temperature and below in the areas of radiation damage and
Presented at the 79th Annual Meeting. The American Ceramlc Society. Chicago, impurity spectra. Investigations on radiation-induced absorption
Illinois, April 27, 1977 (Basic Science Division No. 16GB-77). Received luly 6,
1981; revised copy received January 29, 1982. were discussed by Yen and Coble.’
Supported by the Energy Research and Developinent Administration under Con- The impurity spectra of many transition-metal-ion-doped
tra;‘ No. E(ll-lt2390. samples of A1,0, have been studied.’-” Two mechanisms are re-
Member, the American Ceramic Society.
*Now with Glenn N. Howatt Laboratory, Department of Ceramics. Rutgers Uni- sponsible for the observed absorptions: crystal field transitions or
versity, Piscataway. New Jersey 08854. band-to-impurity-level transitions.