Some new large telescopes in the near future, on Earth and in space, will use multiple mirrors. The individual mirror pieces will be hexagonal in shape. I wonder why this shape is prefered over the circle sector shape (like pie slices). I would think that corners and joints between mirrors cause practical problems. And a circle sector slice has half as many corners and two rather than three or six neighboring pieces, and shorter interfacing edges with neighboring pieces overall. So shy is the hexagonal shape prefered?

  • 2
    $\begingroup$ I don't have enough information off the top of my head to provide a good answer right now, but I believe a large part of the reason is in how the mirrors are given their precise curvature. It is easier, practically, to do this for a hexagonal mirror (because it is close to circular) than for a pie slice. $\endgroup$
    – zephyr
    Commented Dec 5, 2017 at 14:27
  • $\begingroup$ Reasons given in ESO website and TMT website are the support for the mirrors, the fabrication, testing and transport. Even though each segment has to be asymmetrical (to form together a parabol), it seems easier to work with symmetrical shapes as hexagons to "mass-produce" the mirrors. $\endgroup$
    – MBR
    Commented Dec 5, 2017 at 15:06
  • $\begingroup$ A an arc of a circle may not seem like it has corners, but it is continuously changing and is harder to work with than a straight line. In a sense an arc has an infinite number of corners, not none. $\endgroup$ Commented Dec 5, 2017 at 15:21
  • $\begingroup$ Pie shape will give you different rates of thermal expansion at different radii. Be tough to cast, and hard to grind to correct curve. Likely make adaptive optic hardware hard to design. $\endgroup$ Commented Dec 5, 2017 at 15:58

2 Answers 2


The ideal shape for the mirror is round. It's the easiest to make. It's the best-behaved while in use.

The hex tiles are already harder. The mirror is a revolution surface generated by a conic curve (circle, parabola, hyperbola, ellipse), which needs to be machined with a precision greater than 0.1 microns. That's extremely difficult already with a round mirror, where the surface can be easily distorted near the edge in the process of manufacturing.

With the hex tiles, maintaining that precision near the corners is very hard - corners are like the edge but exponentially harder. Unless you take special precautions and unless you perform extra work, you're guaranteed to end up with surface errors near the corners (the surface is either too raised, or too low, or has a more complex distortion). Based on my experience making round mirrors, I would guess the surface near the corners would tend to be turned down (too low relative to the rest), which is a defect that's extremely hard to correct - one of the hardest, in fact.

With a pie slice, all problems mentioned above increase greatly in magnitude. The blunt angle of the hex tile gives you some support in machining the surface, but the sharp angle of the pie slice gives you almost none. I can't even begin to imagine the difficulties one would encounter while making a high precision surface on a pie slice. The sharp end of the slice would be almost guaranteed to be worse quality than the rest.

What most people don't realize is that maintaining the precise shape of the mirror while in use is difficult too. You can't just lay it on a hard surface - the uneven support would distort the precise shape. Mirrors are supported on the back by a precise mechanism, with a number of support elements precisely sharing the weight in equal amounts (or, if not equal, then deduced from finite element analysis), and located in places carefully calculated.

With a round or hex tile this is, if not easy, then at least a well-understood and manageable procedure. A pie slice would be very difficult to support in the same even manner near the sharp tip. You can't rely on the symmetry of the shape to provide the even load on the support points. The support force would have to be specially adjusted there. It would probably be doable, but harder.

This is only an issue with terrestrial telescopes, which operate under gravity. Space telescopes don't suffer from this issue.

Alternatives to hex tiles could also be square tiles, or triangle tiles. These are worse than hex tiles for all the reasons shown above - the corners have sharper angles which make everything more difficult. The blunt angles of the hex tiles alleviate these issues somewhat.

TLDR: Non-round mirrors are hard to make anyway, but hex tiles are less hard compared to the alternatives.

  • $\begingroup$ Important correction: the accuracy has to be sub-wavelength, not just the precision $\endgroup$ Commented Dec 6, 2017 at 13:57

For one thing, hexagons tile the plane -- meaning you can build as large as you want with one shape. Granted in practice this won't work because there are different curvatures, though again all hexagons at a given radius will have the same curvature.

In addition, the likelihood of damage at the tip of a "pie-slice" is great, the aspect ratio (length-to-width) sucks, and remember that minimizing the longest dimension is critical to controlling cost and fab time for any mirror section.

Add most of the info in the initial comments as well.

Further, if you had pie slices it'd be just about impossible to 'refocus' the system, while local hexagons are relatively easy to deal with.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .