by: John M. Hill
In general the honeycomb rib structure looks very good. There are few
bubbles and the rib structure appears to be completely intact.
Probably this is the best set of honeycomb ribs that we have ever made
at the Mirror Lab. This is very good because it means that all of the
repair and recovery techniques can be concentrated on the upper
surface of the glass without worrying about additional leaks or
modifications to the mold structure.
The faceplate of the mirror on average is 26 mm thick. It ranges from
31 mm thick near the center of the mirror to about 20 mm thick near
the outer edge. The desired finished faceplate thickness after
grinding and polishing is 28 mm, and we could probably polish a mirror
with a 24 mm faceplate without much extra effort. The problem is that
near azimuth 60 degrees at the outer edge of the mirror the largest
leak drained away the faceplate to almost nothing over an area of
about 2 square meters. This area represents less than 5% of the total
area of the mirror surface, so the mirror blank probably could be
turned into a useful primary in its present condition. However, it
seemed to us to be imprudent and inelegant to invest the money to
polish the world's largest monolithic primary mirror and then have
this big notch out of one side.
After heating the entire furnace and mold slowly up to 650 °C where the
glass has softened so the stress relaxation time is short, we will use
the heaters on the lid of the furnace to heat the upper surface of the
glass quickly up to 1180 °C. Naturally the furnace will be spinning
during the high temperature phase to preserve the parabolic
shape. This flash heating will melt the additional glass blocks and
cause them to flow across the surface. Because we are only applying
heat from the top, the lower portions of the honeycomb structure will
remain relatively cool and viscous at a temperature of around 700 °C.
The viscous glass would not flow out through any leaks and would not
cause any of the honeycomb cores to float.
When heating the mold from the top, we must take care to prevent
radiation coupling to the sides and base of the mold and the walls of
the furnace. The radiation at high temperature would be lost to
heating the extra mass and it would generate undesirable temperature
gradients in the sides of the mold. To isolate the upper surface
thermally, we have installed a 2-inch insulating blanket of Kaowool
fiber between the top of the mold walls and the joint between the
conical lid and walls of the furnace. The lid heaters can apply
approximately 400 kW to the melting of the glass after we have
accounted for losses and duty-cycle limitations.
On June 10, at 4PM, we will start spinning the furnace up to the
nominal rotation speed of 6.81 rpm. At the same time we will start
the flash heating process with the lid heaters. After 6 hours or so,
the surface should have reached the casting temperature of 1180 °C. We
will hold at 1180 °C for about an hour and then cool back down to 650
°C overnight. The furnace will probably spin for 1.5 days total.
We also have verified that the new heating cycle does not produce
conditions where crystals can grow rapidly in the bulk of the glass
honeycomb during heating or cooling. These crystals could be a
problem in those parts of the blank that do not go above the
"liquidus" temperature which is about 1130 °C. Experiments have shown
that the heating cycle that we will use does not cause devitrification
in E6 even at the microscopic level.
The viscosity of the E6 borosilicate glass at 1180 °C is such that it
has the consistency of cold honey. There was never any danger that
molten glass would leak or fly from the rotating furnace. While that
would have made some exciting video, I'm afraid it only happens in
volcanic eruptions where the system is pressurized.
This glass will have to be excavated away before the mirror is lifted
off the furnace. The spilled glass cannot be reused in a future
casting because it has been contaminated by the mold material and
silicon carbide particles.
Our analysis shows that hydrostatic pressure was able to move the wall
because the margin of applied compression force relative to the
hydrostatic pressure was lower compared to the 6.5 m castings. This
we knew before the casting. The tension in the Inconel bands was
calculated to account for the pressure of the liquid glass, and for
the centripetal forces from the rotation, and for friction between the
bands and the silicon carbide tub walls. What we didn't anticipate was
a further reduction in the compression actually felt by the mold, due
to increased friction arising from the additional tensioning bands
used for the LBT. Because the force per band was limited to the same
value as used for the MMT/Magellan 6.5m mirrors (to avoid creep of the
Inconel at high temperature), more bands were needed. The extra
friction between one band and the next at the crossing points where
the bands are interleaved and rest on each other was not adequately
allowed for. The net effect was that at four points on the
circumference where the band tension was most reduced by friction, the
glass was able to push the wall sections apart. These are the places
where we see the greatest leak. The friction effects are reversed
upon cooling when the bands shrink faster than the tub walls. The
walls were quickly drawn in and the leak stopped when we started to
cool. Clearly the 8.4 m LBT casting was right on the borderline of
having adequate force. The problem for the second 8.4 m casting will
be corrected relatively simply by increasing the band force and
slightly lowering peak temperature to control band creep. In short
there is no mystery about the origin of the leak and we are confident
that the problem will not occur again.
There have also been repairs to the signal sliprings and lubrication
of the main drive gear. We have adjusted a servo tuning problem that
we noticed while spinning during the casting in January. These and
other items are just part of the routine operation of the furnace.
Executive Summary:
Two tons of glass have been added to the furnace.
The 8.4 meter mirror remelting is underway. High temperature is scheduled
to occur on June 10.INSPECTION RESULTS
Since getting our first look at the cool 8.4 meter honeycomb mirror blank
on April 3, we've had a chance to conduct several weeks of careful
measurements and inspections. These inspections confirm what we observed
after first opening the furnace -- about two tons of glass leaked outside
the mold wall during the high temperature phase of the casting.PLAN FOR RECOVERY
The plan for recovery is fairly simple. We are going to remelt two
tons of additional glass onto the faceplate of the existing honeycomb
blank before removing it from the furnace. This remelting differs
from the original casting because we only need to heat the upper
surface of the blank up to the melting temperature of 1180 °C. In an
analogy to cooking, we are making a very large creme brulee --- that
is cooking the upper surface without overheating the custard below.
There is also a certain precedent in Italy regarding the "biscotti"
--- twice-baked cookies.EXTRA GLASS LOADING
On the morning of April 25, 1997, we placed 2.1 tons of additional E6
glass blocks on the faceplate when the furnace was cold. The glass
blocks were stacked loosely in a single layer near the mid-radius of
the mirror. This is enough glass to bring the faceplate back to a
thickness of about 37 mm. The lid of the furnace was replaced later
that day.SCHEDULE OF REHEATING
We started reheating the furnace on the morning of April 30, 1997.
Heating will continue at a slow rate until the temperature reaches 650 °
C on June 10. While we are below the annealing temperature, the
heating schedule is just the opposite of the cooling schedule that we
followed down in March. This is to avoid inducing any temperature
gradients larger than about 8 °C which could possibly crack the
honeycomb structure. Normally we heat quite rapidly at the beginning
of the casting process, because the glass is all in the form of broken
blocks. Because the glass honeycomb is a stiff and intact structure,
it cannot withstand large temperature gradients. Don't worry about
your borosilicate cookware --- it can stand much larger gradients because
it is made in flexible shapes.WILL IT WORK?
Since January when we knew there had been some kind of leak in the
mold, we have been developing techniques to remelt additional glass on
the surface of an existing mirror blank. The results of these
analyses and experiments on scales up to 0.6 meters suggest that
sticking additional glass to the faceplate is actually quite easy once
you work out the correct heating recipe. There is little reason to doubt
that the process will be a complete success. I've tried to answer the
most commonly asked questions below.Heating?
Can we really heat the mirror from the upper side like this? From the
very beginning the furnace was designed to have different regions
heat following different temperature schedules. Now for the first
time we are really taking advantage of that part of the control system.
We also have enough surplus heating capacity to melt the entire
mass of the mirror and mold using only the lid heaters. Both computer
modelling and empirical tests show us that we can generate the proper
heating profile and heat transfer rates to melt the upper surface while
the lower surface remains around 700 °C.Sticking?
Many people have expressed doubt that the new layer of glass will
stick onto the existing faceplate. I speculate that this is because
they don't find glass to be a sticky substance in their everyday experience.
At 1180 °C the borosilicate glass has a viscosity similar to that of cold
honey. Let me assure you that the molten glass is every bit as sticky
as honey. The problem with the glass is stopping it from sticking to
things that you don't want it stuck to. There is no doubt
that the new layer of glass will be permanently joined to the
glass in the existing faceplate. The blocks of new glass will be
already starting to stick by the time we have reached 700 °C.Leaking?
Why won't the new glass leak out of the mold? As mentioned below, the
gaps in the tub wall have already closed during the cooling process.
Because we are not heating the lower portion of the honeycomb structure,
the glass will not reach a viscosity that would let it flow out of the
mold even if the gaps were to reopen. Just for good measure, we will
also increase the Inconel band tension to 110% of the previous value
to be sure that the wall sections don't move. Because the bands will
only be heated to 700 °C, they have more than enough strength to handle
the extra tension.Flotation?
The classical casting failure mode is the floatation of the hexagonal
core molds in the liquid glass. Because this set of cores and silicon
carbide bolts has been (will be) used an extra time, it is logical to
ask if they are at risk of failure. Because the glass in the lower
portion of the honeycomb structure will be at low temperature, the
timescale for a core to float is many days even if the restraining
bolt and crosspins were not present. For this reason, we believe that
floatation will not be a problem during this procedure.Devitrification?
During the normal course of casting a borosilicate glass mirror blank,
a frosty looking "scum" forms on much of the surface. This "scum" is
a thin layer of devitrified (ceased to be glass by crystalizing)
borosilicate that has crystallized because the alkali component of the
glass has evaporated from the surface. We questioned whether this
layer might reduce the structural integrity of the faceplate or cause
problems during polishing. Experiments have shown that this thin
devitrified layer re-dissolves into the bulk glass after returning to
1180 °C for an hour. Just to be sure that we had a good boundary we
used some air-powered diamond grinders to remove the devitrification
from the existing surface before loading the new glass.Geometry?
Will the honeycomb geometry remain stable? The honeycomb ribs now
appear to be in excellent geometrical order. Because the body of the
blank will not be liquid, we expect them to retain that desired
geometry. The edge wall of the honeycomb structure has some regions
that are thicker or thinner than the nominal 15 mm, because of the
motions of the tub walls. We have done a few repairs to improve the
outer wall of the mold in the area where the edge wall is thinnest.
Some excess glass (~8 mm) will need to be generated off the thick
areas while the mirror is on the polishing machine next year.Other failure modes?
There are many possible failure modes which could disrupt the casting
of an 8.4m mirror. These include lightning, power failures, flooding,
psychological breakdown of the oven pilot, and other assorted acts of God.
None of the possible failures should have anything to do with the
remelting process, but are just the normal risks associated with
melting and annealing 20 tons of glass in a complex shape. Probably
the biggest risk during the remelt is whether all the furnace
electrical systems can survive the high (40 °C) ambient temperatures on
a June day in Tucson. We will have the swamp coolers running. Bring
your own hot dogs.WHAT ABOUT THE LEAKED GLASS?
The two tons of glass that leaked out of the large pot that forms
the outside boundary of the mold appears rather lava-like on the
floor of the furnace outside the mold. When the borosilicate glass
comes in contact with the silicon carbide tub walls, it generates carbon
dioxide bubbles that cause the glass to foam. The texture of the glass
varies from a lightweight pumice-like foam to heavier glass that appears
to have been carbonated like soda water. This glass did some mechanical
damage to half a dozen thermocouples in the floor of the furnace
during the cooling phase, but is otherwise lying harmlessly on the hearth
between the mold wall and the heaters that form the outer wall of the
furnace.WHAT CAUSED THE LEAKAGE OF GLASS?
The evidence is now clear that the leakage occurred between the sections
of the circular outer wall which is constructed like a barrel. The
wall sections were pushed apart slightly by the hydrostatic pressure
from the liquid glass, but closed again as the compression bands
contracted on cooling. The tub wall sections were able to move because
the glass pressure inside the tub, were just a bit larger than the
restraining forces outside the tub. Most of the leaked glass flowed
out at the base of just a few wall sections where a gap opened between
the bottom of the outer tub wall and the circular edge of the hearth
tiles.OTHER REPAIRS AND MODIFICATIONS
We have added a fifth video camera to the oven so that we will have a
telephoto view of the faceplate at the outer edge as well as the inner
edge. It is always fun to see if the new camera will survive its
first "trial by fire". We have also upgraded the framegrabber for the
oven video system to improve the signal-to-noise in the flashed video images.SCHEDULE FOR THE REMELT
The oven is now at 245 °C and is presently heating on a schedule which
will bring it to 650 °C at 4PM on Tuesday June 10. That evening we will
spin and flash heat the glass onto the faceplate. Then we will follow
the normal 8.4 meter annealing and cooling profile. The mirror should
be back to room temperature in the second week of September.IMPACT ON TELESCOPE SCHEDULE
Because there was about a 4 month slack period between the casting
schedule and the polishing schedule, it appears that this extra
cooking will not produce any noticable delay in the completion of the
primary mirror. At the moment, the casting crew is able to cast
mirrors faster than the polishing crew can figure the highly aspheric
surfaces. The remelting has significantly reduced the amount of
fishing and four-wheeling that I was planning to do over the summer.
Back to LBT Casting Updates