Incredible overview of making mirrors for the world's largest telescope

Originally published at:


It’s very reflective.


Sure beats doing them by hand. Back when I was much poorer, I read some books on hand grinding your own eight inch mirrors and building simple dobsonians. It’s fascinating how similar the process is on a vastly larger scale, but with the addition of robots to do all the grinding and polishing. Yeah future!


My brainiac friend is grinding his own giant telescope mirror. I need to ask him where is at on it… been awhile.

ETA - its 44"




I like this one better:


Robots are necessary so as to achieve the required precision.

1 Like

I appreciate that your video mentioned how it becomes a mirror. This is just a glass blank. A high-tech glass blank, but still not a mirror.

1 Like

Yes, much better.


Roger that.

Also, where does the warning label go?


Here’s what I don’t understand. Why do all 18 tons of glass need to be hyper pure shit, purer than the purest water from the deepest lake on the highest mountain? It’s only the surface that needs to be smooth. The inside can go fuck itself with its impurity. Why not use “reasonably pure” glass, then deposit a layer of “hyper-pure-fuckin-shit” on top of that? Or sputter a thicker layer of hyper pure amazing-ass aluminum on top and polish that? Then they can surface at the site with other aluminum, business as usual. This just seems to me a process that is overkilled with purity. It’s really only the very top surface that is in question, and maybe a touch deeper so that impurities within the material don’t bubble to the surface with temperature changes, vibrations, age, etc.

Also, in one vid, they said it takes a year to polish the back. WTF??? The back of the thing doesn’t do anything.

Telescope gods, explain this shit to me. I don’t get it.

I’m sort of surprised that glass still gets the nod at all; since it’s a mirror rather than a lens and vastly flatter than “we use glass because the float process makes it pretty flat” grade.Were I asked to guess I would have imagined that there was something either easier to work with or with better mechanical properties for when you need a large sheet of it to not droop at all when suspended.

That said, I imagine that if you are going to the trouble of building a mirror that nice you Do Not Want even slight variations in coefficient of thermal expansion, stiffness, weight per unit area making it droop, or any of the other variables that might affect the flatness of the important surface once it’s hung on a support structure and placed in a necessarily open-roofed environment.


That’s an amazing process. What an engineering feat.

I imagine any impurity could lead to a compromise in structural integrity. All I know is everything with these fancy telescopes requires precision, or you get shit that is blurry or distorted, which is bad when you are looking at a point of light many light years away. This is because the changes they detect for things like planet hunting is such a small change. Like the back is how you mount the thing, so you need perfectly flat to mate with the flat mount.

I remember awhile ago there was worry we were basically at the maximum for how good terrestrial telescopes were because of the wiggle of light going through the atmosphere. But thanks to new digital techniques, we can compensate for that.


Those sorts of questions are interesting to me, and make me wish that I had taken at least one Astronomy class in college…

I did come across this reference-- which, for our purpose may as well be one of many.

(downloads a pdf)

Glass is the legacy material for mirror substrates, and there are several reasons why it performs so well: the material is thermally stable; it can be engineered into a stiff structure with minimal residual stress; and the face sheet can be polished to a high-quality optical surface.

While glass has been the traditional substrate of choice, other materials have been used. One example is the Spitzer Space Telescope (SST), which has an 0.85 meter beryllium primary mirror. Launched in 2003, the SST is a general-purpose observatory that is diffraction limited at 5.5 microns and operates at 4K. Beryllium offers several advantages when used as a mirror substrate. One of the most significant advantages is that Be has a large specific stiffness (or stiffness-to-mass ratio): this ratio is five times greater than ULE and 6 times greater than aluminum [Yoder 1993]. Beryllium’s superior rigidity means that it can be used in mirror substrates that take up less volume than a ULE substrate designed for the same mission. In addition to a high specific stiffness, Be has a near-zero coefficient of thermal expansion (CTE) when used below 100 K which makes it an ideal material for cryogenic mirrors.

the “why polish the back” is a interesting question. still can’t nail down an authoritative answer. May have something to do with making the mirror reflective enough for interferometry, may have something to do with removing material without damaging the structural integrity of the blank, may have something to do with proving that the glass is structurally sound before the much more delicate and expensive process of carving the front into an optically perfect surface.


Glass can be very stable at scale. But there are other materials proven to be stable at scale, cheaper than optically pure glass: concrete, ceramics, resins, some alloyed metals like chrome steel, chrome-moly steel, maybe exotics that are now common like carbon fiber in resin, known for its stiffness and rigidity.

I have also heard of laser arrays across the circumference of the face of the disc: each laser measures the distance to its receiver, thus providing real time feedback on the deformation of the lens at that time. Image minus deformation (minus atmospheric and other adjustments) = true image. That idea.

I’m trying still to understand the wisdom of 18 tons of pure silicon dioxide + 36 months of processing, compared to alternatives.

1 Like

I work down the street from this lab, and two of my good friends work in it, running the big polishing machines. Got the VIP tour last fall when a mirror was being spun and melted. It’s quite a place.

They use very good glass because stress is a killer for such big mirrors. The glass is so good, and the annealing so good, that they have successfully repaired several surface blemishes, bubbles and cracks in these mirrors. Also, glass polishes way better than any other material.

My father helped build a couple aluminum 60" mirrors in the late sixties for telescopes used for infrared photometry (no imaging needed), but they were eventually replaced with glass mirrors to gain versatility. The glass used also has the great quality of a nearly zero temperature coefficient of expansion (TCE), so it’s a lot easier to maintain nanometer-level dimensional stability through the diurnal temperature cycle.

The back is polished because the mirror blank has to be attached very firmly to the polishing machine turntable, so a flat back surface is needed.


Most of the materials you mentioned have lots of problems with residual stresses. Even if you machine them to given tolerances, they will again be out of spec in a year or earlier. This problem also important for high precision CNC machines (and the typical precision of such machine is measured in micrometers, not nanometers). This is why the machine tables are mostly made from cast iron (not steel) or granite - both can have very low residual stress, but granite is way better - it’s also used in best computer controlled measurement machines.

Residual stress is especially problematic in chemically cured materials like concrete or epoxy and polyester resins - these materials typically shrink during curing. If used as a part of composite material like steel reinforced concrete or carbon fiber reinforced plastic, residual stresses result form reinforcing material not shrinking with matrix. At work I set up measurement equipment for experiments with various composites, and from what I’ve seen, I wouldn’t expect much dimensional stability from them.

As @jerwin said, beryllium mirror would also be a good choice, but I suspect that cooling times would be similar to glass, and machining costs would be far greater - beryllium is extremely toxic, especially when ground into fine particles.

Edit: The processing time of 36 months is also not that much compared to largest lathes and mills. For such machines, casting and machining main parts can easily take 10 years.