Grinding and polishing a mirror on an 18" quartz substrate
Quartz is a wonderful material for making telescope mirrors. The thermal performance of quartz offer minimal wait time before observing sessions can begin. That same property helps me as a mirror maker in that I don't need to let the mirror aclimate as long between polishing and testing. But on the other hand, quartz offers some minor challenges during the fabrication.
Quartz is a bit harder than borosilicate glass, so it takes a little longer to grind and polish. In the case of this particular mirror, I'm starting with a plano/plano blank that hasn't been pre-generated. That means I had to quarry out many cubic inches of quartz to get down to the final curve. To do this work, I fabricated a special tile/plaster tool for the job. The tiles I use are very hard, but they are also thin, less than 1/8". With the normal single-layer tile tool, the tiles may have ground through before I got the mirror to depth and finished all the grinding steps. So my trick is to band together a group of five tiles and embed them upright in the plaster, instead of flat. That gives me a very thick tile set that won't grind through. This is far better than having flat tiles fall off then trying to glue on replacements.
The photos here show the process. I start by making up the bundles, literally held together with small rubber bands. The tiles have just been peeled off the usual sheets. Their sharp edges cause the snapping of a lot of little rubber bands, but in the end I used about 70 little bundles to make the grinding tool.
I put a sheet of sticky shelf-liner paper face up on the mirror blank, and then arrange the tile bundles on the surface. There's really no preferred pattern. I just want good coverage over the entire surface. The tiles don't even have to be in tidy rows. I have a series of circular MDF rings of various sizes that are the basis of the mold. For this mirror, I used a 14.5" diameter MDF ring, and wrapped thin plastic around it, securing it with tape. As a rule of thumb, I make my general grinding tools about 75% the size of the mirror. It's better to be a little oversize than too small. If a tool is too small, it won't cover the center of the mirror when the overhang is adjusted to the correct amount. If the tool is too big, it's much harder to change the depth of the curve.
I use a 1/2" black pipe nipple molded into the tool as a "push point" for the stainless steel drive rod in my grinding machine. The nipple is positioned on top of the tiles, the plastic ring is put in place, then I pour dental plaster to cover to the top of the nipple. I use Kerr Vel-Mix 31008 for the plaster, and mix it to about the consistency of heavy cream for this special tool. For normal tile tools it's nicer if the plaster is a little thinner, as it will run down between the tiles better. About 30 minutes after pouring the plaster the tool will be hard and I remove the plastic and unmount the tool from the mirror. I use a "surform" tool to break the edges of the tool, and I use a rotary tool and diamond bit to route little channels between the tiles. You need the channels so the abrasives distribute and work properly, and to avoid suction of the tool against the mirror.
For the heavy hogging needed on a deep mirror like this, I usually start with #40 until the curve is established and nearly at depth. Then I move up through #60 silicon carbide and #80. There are a couple of photos that demonstrate the surface roughness after the #40 abrasive does its work. As always, I use fixed-post grinding, as the photos show. In this method, the tool is positioned with overhang over the edge of the mirror, and fixed in place. As the mirror spins on the turntable, the tool is forced to spin on top with abrasive in between. To deepen the curve, you move the tool in toward the center of the mirror. To flatten the curve, you move the tool out. This is one process where having a great spherometer is a must.
In the end, this mirror came out with a great figure. The Foucault test and FigureXP rate the mirror as a little under 1/20th wave, with a 0.99 Strehl. (The usual mention here that FigureXP is optimistic, and I don't claim that the mirror actually reaches these numbers!) The surface RMS error is only 4.3 nm, and the Millies-LaCroix chart shows a good pattern with all zones easily contained within the "tornado" tolerances.