Rules for Handling
Large Borosilicate Honeycomb Mirrors
J. M. Hill, J. R. P. Angel, B. Olbert
Steward Observatory Mirror Lab
Columbus Project Technical Memo
UA-92-05
http://medusa.as.arizona.edu/lbtwww/tech/ua9205.htm
November 23, 1992
Introduction
This memo is intended to provide a summary of the proper strategy for
safely handling large borosilicate honeycomb mirrors. Specific load cases
should always be studied before handling a particular mirror, but the
general rules for borosilicate honeycomb are stated here. Many of
these rules also apply to other optics as well. The rules stated here
are based on a combination of analytical analysis, finite element
analysis and practical experience. This memo should be considered as
a sensible set of guidelines and not an infallible authority.
We have set a goal for maximum local concentration of tensile stress
in the blank due to support or handling of 100 psi (0.7 MPa). This
number is well below the practical breaking strength of processed
optics (1000 -- 2000 psi). Note that there is typically an order of
magnitude difference between the average stress level and the maximum
local concentration in the honeycomb blank. Based on 100 psi maximum
stress, detailed finite element and engineerical analysis tell us that
no location should carry a load of more than 2600 Newtons for each
honeycomb rib intersection. The same stress threshold tells us that
we should not allow temperature gradients larger than 5 C° across
the blank. The appendices provide some of the background information
for these rules.
At the risk of the stating the obvious, the answer for risk reduction
is simple: BE CAREFUL!
Fundamental Rule
- Never exceed 100 psi (0.7 MPa) for the maximum local concentration
of tensile stress in the glass. This will lead to a long and healthy
life for your blank. (See Appendix B for a lengthy explanation.) In
practice, we have been allowing 100 psi stress levels for each of
annealing, temperature gradients and mechanical loading separately.
Subsidiary Rules for SOML Borosilicate Honeycomb
Mechanical Rules
- Don't apply loads to the honeycomb ribs directly.
They are the weakest part of the mirror.
- Don't apply localized radial loads to the edge walls of the blank.
- Only apply axial loads to the faceplates near rib intersections.
- A compression load of no greater than 2600 N (584 pounds) may be applied
to the faceplate above each rib intersection. This load must be
spread over an area of 44 square centimeters (7.5 cm diameter). Within
limits, a larger area may be used to distribute a larger load. A load of
3600 N (809 pounds) may be applied over an area of 78 square centimeters
(10 cm diameter).
- A tension load of no greater than 1200 N (270 pounds) may be applied
to the faceplate above each rib intersection. This load must be
spread over an area of 44 square centimeters (7.5 cm diameter).
Thermal Rules
- Do not exceed a temperature gradient of 5 C ° peak-valley
in the glass structure. Some geometries of the temperature gradient
allow 10 C ° peak-valley.
- Do not use heat to remove glue from the glass.
- Do not change the mirror temperature faster than 20 C ° per hour with
ventilation or 1 C ° per hour without ventilation. (Avoid exposing
the mirror to desert diurnal temperature variations in an uncontrolled
enclosure.)
- Do not ever store the unaluminized mirror in direct sunlight.
(The ``greenhouse effect'' can generate significant temperature gradients.)
- Be careful when cleaning the mirror with volatile liquids.
(Evaporative cooling from the mirror surface can cause significant
temperature gradients.)
- The mirror may be stored indefinitely at any uniform temperature between
--270 C ° and 350 C °. However, metal fixtures or adhesives in
contact with the glass may restrict this temperature range.
Rules For Different Diameters
- 1.8 meter diameter and smaller mirrors may be safely supported at any
three points with pads or blocks which spread the gravity load
(``1g'') over an area of 20 square cm per pad.
- A 3.5 meter mirror may be supported at three points with 50 square
centimeter pads, but considerable care is required in selecting the
pad location. A six or nine point support is recommended.
- A honeycomb mirror larger than 3.5 meters may NOT be supported
solely by a three point support at any time. Mirrors larger than 6 meters
may not be supported with a continuous edge support. Distributed support
is required.
Rules Applied to All Mirrors
- Cracks larger than 1 cm in length should be treated to relieve
their shape factor by acid etching and/or diamond grinding.
- Do not store the mirror in high humidity environments for
extended periods. (Humidity aids crack propagation.)
- Devices or fixtures to be glued to the mirror must be matched
to the coefficient of thermal expansion and the expected temperature
range. (The thermal expansion coefficient of E6 is 2.9x10-6 per C °.)
- Do not allow liquids to pool and evaporate on the polished surface.
(This could stain or etch the surface.)
Follow approved cleaning procedures.
Incredibly Obvious Rules
- Pay attention whenever you are near the mirror.
- Do not place loose objects where they might fall or be bumped onto the
mirror (i.e. wrenches, screwdrivers, keys, broom handles).
- Always cover the mirror when working above or near it.
- Do not allow other objects to impact the mirror.
- Do not allow the mirror to impact other objects.
- Never apply loads to the mirror at corners or sharp edges.
- Never pour hot or cold liquids on the mirror.
Appendix A: Expected Internal Stress Levels
We expect the stress level in the finished blank to vary from
about 0 to 20 nm/cm of birefringence. The residual stress is a
combination of variations in the original E6 blocks and stress induced
during annealing. The residual annealing stress puts the faceplates
in compression and the ribs in tension. The prediction is based on
annealing calculations and experience with the 3.5 meter blanks made
by the same process. The stress is measured quantitatively with
a Babinet compensator to measure the strain induced birefringence via
the relative retardation technique. Measurements are made normal
to the surface of the blank looking into the ribs. The peak measured
stress in the three 3.5 meter mirrors was 73 psi, 87 psi and 65 psi
respectively. The average stress values were typically half these
values.
Appendix B: Failure Mechanisms
Failure occurs by extension of pre-existing flaws to a size critical
for the applied stress field in the region around the flaw. The goal
here is to determine the flaw size distribution and the stresses
acting on the flaw, and then to set a fracture condition. The method
is probabilistic unless you have a complete map of flaw sizes in your
blank.
In most cases, failure probabilities are estimated from small sample
strength tests. The Mirror Lab has done a substantial number of
measurements on sample castings. Once the strengths are known, there
are two ways to do mirror failure estimates. In the first method, the
small sample strength distribution is scaled to the size of the
mirror, and the failure probability is calculated at various stress
levels. In the second method, the flaw size distribution is extracted
from the strength data, and the probability of finding a critical flaw
at various stress levels is estimated. The calculated strength of the
mirror at low failure probabilities (10-6) can be several times
smaller than the small sample strengths with no slow crack growth, and
smaller still when allowances for slow crack growth are made.
The difficulty here is with the estimation procedure. The flaw
distribution measured in small sample tests may not exactly represent
the flaw distribution in the mirror blank. In cases where the sample
strengths are several times higher than the mirror blank stresses, we
must extrapolate far beyond the data range on the strength or flaw
distribution curves. In spite of the problems, most people try to
meet the conditions imposed by extrapolation.
Another design approach, called damage tolerance analysis, is
sometimes used. Here we pick a mirror stress level, calculate the
critical flaw size, and then try to hunt for flaws of this size or
larger and get rid of them. We can also assume the existence of large
flaws (like the observed rib cracks), calculate the critical stress
level, and keep casting and handling stresses below this level. In
our case (100 psi stress), the critical flaws would be large and
hopefully easy to find. Calculations like this are important because
flaws will extend with time by slow crack growth (in environments
where water is present), so a flaw smaller than critical today may
become critical 10 or 20 years from now.
Appendix C: Damage Control
Small cracks internal to the mirror blank would be treated by some
combination of diamond grinding or acid etching to prevent propagation
from stresses at the crack tip. This has been done successfully on
small cracks in each of the 3.5 meter blanks. Surface defects up to a
few centimeters or so could be repaired with a matching piece of E6
ground to fit and epoxied in place. This plug allows polishing over
the defect without any problems. This technique was used on a 1 cm
bubble in the WIYN 3.5 meter mirror.