The Giant Magellan Telescope Needs Engineers!

A deep dive into the engineering challenges of building the world’s largest optical observatory.

(Image courtesy of Giant Magellan Telescope –GMTO Corporation.)

(Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

In 1609, Galileo pointed his telescope at the moon and forever changed our understanding of the universe and humanity’s place within it. Astronomers have been following in his footsteps ever since, probing the hidden mysteries of the cosmos with ever more advanced optics. The telescopes of today are remarkable feats of engineering, and they’re only getting better.

The Giant Magellan Telescope (GMT) is one such example. Currently under construction and slated for completion in 2025, the GMT will utilize seven primary mirror segments, each 8.4 m (27.6 ft.) in diameter, and engineering marvels in their own right. At the time of its first light, the GMT will be the largest optical observatory in the world.

ENGINEERING.com had the opportunity to discuss the latest step in the project with Robert N. Shelton, President of the Giant Magellan Telescope Organization (GMTO) and Patrick McCarthy, VP Operations and External Relations.

Can you give us a brief overview of the Grand Magellan Telescope Organization?

Shelton: One of the main things to keep in mind is that, while we do have international partners, most of our partners are U.S. institutions and universities. So, this is an organization that’s founded very much via private partnerships. We have public universities, such as the University of Arizona—where the mirrors are made—the University of Texas and Texas A&M. We also have great private universities like Chicago and Harvard, and we have institutions like the Smithsonian and Carnegie, which is providing the site in the Chilean Andes.

In addition to these U.S. organizations, we have our international partners in South Korea and Australia through some of the universities and institutions there, Sao Paulo in Brazil and then, of course, Chile is very important, since they have the site.

How many people in total are working on the project?

Shelton: We’ve been hiring lately, so there are, what, 80 people here in Pasadena?

McCarthy: Yeah, we have a team of about 100 people working for GMTO —85 here in Pasadena and 15 in Chile—but the distributed workforce is much larger. We have about 50 people working full time in Tucson on the production of the primary mirrors. We also have teams distributed across our partner institutions developing scientific instruments and helping us with the design of some of the more challenging optomechanical subsystems. So, right now there are a few hundred people working on it. When we get to the peak of construction, there will be another 250 or so contractors working on the infrastructure.

(Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

(Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

Shelton: That’s about what you’d expect on a billion-dollar capital project: you’re going to need on the order of several thousands of person-years of effort. We’re motivated by the science and the way astronomy captures the imagination, but right now this is really an engineering project. The talent Pat referred to is coming overwhelmingly from the engineering disciplines, so it’s very appropriate that we’re discussing it on ENGINEERING.com.

This will be the largest optical observatory in the world at the time of its first light. Can you contrast the GMT with some of the other extremely large telescopes that are already operating (e.g., the Gran Telescopio Canarias in the Canary Islands or the Large Binocular Telescope in Arizona)?

McCarthy: When you look at the evolution of the telescope, the quest has always been to make the collecting area larger so you can gather more light and see objects that are fainter—either because they’re farther away or because they’re just intrinsically fainter. So, the challenge has been building a big mirror.

The Large Binocular Telescope in Arizona.

The Large Binocular Telescope in Arizona.

We started with lenses, but lenses reach their peak around one-meter diameter. When we learned to make glass mirrors with metal coatings, that opened up a whole new path of development that led to the Mount Wilson Telescope, the Palomar Telescope and eventually the largest telescopes now: the Keck, the Grand Canary Telescope and the LBT.

What we leaned in the 1980s and ‘90s is that there’s a maximum size you can make any individual mirror; in practice, it appears to be about 8 or 9 meters in diameter. Some of that is set by the physics of glass and how uniformly you can allow it to cool without building up internal strains, and some of it is practical issues, such as how can you pick up a piece of glass that big and take it to a mountaintop without it breaking just by the fact of you lifting it and it sagging under its own gravity.

So, people realized that once you get to that 8-meter scale, you have to segment the mirror—make the primary mirror in pieces—and the two telescopes that you mentioned are great examples because they illustrate different technical approaches.

The Gran Telescopio Canarias. (Image courtesy of Pachango.)

The Gran Telescopio Canarias. (Image courtesy of Pachango.)

The Keck Telescope and the GCT made their primary mirrors in relatively small pieces—hexagons about 1.8 meters across. The LBT approach was to make the largest continuous optical surface we know how to manufacture—which are the 8.4 meter mirrors made in Arizona—and then combine the light coherently to make a diffraction-limited large collecting area.

Those two paths have both been very successful: the Keck and the GTC have made great scientific discoveries, and the LBT has shown the unique power of adaptive optics, large baseline interferometry and it’s probing a whole new range of science on very small angular scales, looking for planets around nearby stars and looking at the discs that are the early stages of planetary formation.

Comparison of nominal sizes of primary mirrors of notable optical telescopes. Dotted lines show mirrors with equivalent light-gathering ability. (Image courtesy of Cmglee.)

Comparison of nominal sizes of primary mirrors of notable optical telescopes. Dotted lines show mirrors with equivalent light-gathering ability. (Image courtesy of Cmglee.)

These are great examples, and they lead right to the path we’re seeing in the next generation of large telescopes that Europe is building on the heritage of the GTC and in some sense we’re building a bit on the LBT path, but we’re taking it in a different direction.

Would you put the GMT closer in kind to the GTC or the LBT, or is it in a category all its own?

McCarthy: I think it takes the best of those two approaches because, first, a binocular telescope—as in the name—is two on-axis optical systems with a beam combiner. However, the two big mirrors in the LBT are on-axis symmetric optical systems, whereas the GMT takes a 25-meter ellipsoid and segments it into these large 8-meter segments, six of which are off-axis. But, when we align all the mirrors we make one large coherent optical surface, whereas the LBT takes two independent telescopes and combines their light together. So, we’re taking the GTC approach of one single optical surface, but the LBT approach of the largest possible segments.

Can you explain how the mirror casting process works?

McCarthy: I like to think back to the first big astronomical mirrors up at Mount Wilson, because here in Pasadena we can look up at Mount Wilson where Hubble worked and where the expansion of the universe was discovered.

For those mirrors, the glass was poured in Europe and they’re essentially one solid piece of glass that’s basically window pane or green glass. The problem is that once you make a mirror very large and thick, it’s hard to make it uniform—to keep out the bubbles—and it takes a long time for the glass to anneal so you don’t build up stresses and strains.

The completed mold for GMT mirror 5 casting. Each unique ceramic core has an identifying number. (Image courtesy of Giant Magellan Telescope - GMTO Corporation.)

The completed mold for GMT mirror 5 casting. Each unique ceramic core has an identifying number. (Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

The Palomar mirror—at 200 inches—was a real breakthrough, in that Corning realized that if they poured the glass into a mold that has a kind of waffle-iron shape that created empty spaces behind the front surface, you could make the mirror significantly lighter than it would be as a solid glass disc. In the ‘30s, they learned how to make that lightweight mirror out of Pyrex, which doesn’t change its shape much as the temperature changes. You don’t want an optical surface changing its shape all the time as the outside temperature varies, so having this low-expansion glass was important.

One of the challenges in making the Palomar mirror was that they would melt the glass in the furnace and then pick it up in a giant ladle, move it across the factory floor and pour it into the mold. Then they’d go back and get another ladleful and pour it in, but by the time they got to the third or fourth ladle, the glass from the first one was already starting to cool and solidify. It’s difficult to make a really uniform glass disc that way.

Mirror Lab staff finish placing the first layer of glass on to the mold for GMT mirror 5. From left to right: Randy Lutz and Britt Kayner. (Image courtesy of

Mirror Lab staff finish placing the first layer of glass on to the mold for GMT mirror 5. From left to right: Randy Lutz and Britt Kayner. (Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

So, the team in Arizona had a very interesting insight as they were developing the successor to this process, using low expansion glass that was super lightweight to make a mirror that’s almost 85 percent hollow space. They realized that the ladling process doesn’t work very well, but if you build the mold, and then put the glass on top of the mold and melt it in place, then all that happens is the glass flows into the mold, you ramp down the heating and now you have a lightweight mirror essentially molded in place.

That’s the basic idea behind the Arizona casting process, but the group was clever enough that they had to take it one step further.

The goal is to put a roughly parabolic shape on the front surface of the mirror. In conventional mirrors, you make a solid glass disc or a sheet with a thick top face and then grind away all the glass in the middle to make it a deep dish. Well, that’s a lot of glass that you paid for getting thrown away.

So, the Arizona team thought, “While the glass is molten on top of the mold, why don’t we just spin the whole thing?” It’s like water in a spinning bucket: it will creep up on the side and make an approximately parabolic shape. So, if you can cool the glass while it’s spinning, you can freeze that shape in. So, that’s what they do.

Mirror Lab staffer Linda Warren places the last piece of glass into the mold for GMT mirror 5.(Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

Mirror Lab staffer Linda Warren places the last piece of glass into the mold for GMT mirror 5.(Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

They have a very large furnace that they fill with a mold material that’s kind of like the heat shield on the old Space Shuttle: it’s very strong and it can withstand very high temperatures. They load 19 U.S. tons of glass on top of that material, then bring in the heating elements to heat it up to about 2100°F. As the glass starts to liquefy, they start spinning the oven so the glass flows into the empty spaces in the mold and takes on this parabolic shape. After about an hour of the glass cooking to remove the bubbles, they ramp down the temperature to freeze in that shape and then slowly cool it over several months.

It’s kind of like your soufflé, where you look in the oven and voila, it’s done and now you just have to put on the icing and you’ve got yourself a mirror.

You’ve already cast four mirrors and you’re making the fifth now, correct?

McCarthy: Yes. There’s a lot of long lead time and sole-source materials that go into making these. It’s a very specialized field, as you can imagine. We’re making mirror number five, but we have all the glass for mirror number six in a warehouse in Tucson, and we’re starting to get the glass for mirror number seven. We’re thinking of the whole production chain, from the first segment to the last—and we’re in different stages for each mirror—but mirror five is now in this critical stage where we’re changing it from a bunch of blocks of glass into to an 8.4m disc.

The lid of the furance for GMT mirror 5 positioned over the furnace, ready to be lowered into place. (Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

The lid of the furance for GMT mirror 5 positioned over the furnace, ready to be lowered into place. (Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

Mirror number one was finished some time ago, and it really was an engineering marvel because it was the most challenging large optic ever made. We think about how difficult optics are to polish by how far away they are from a sphere, because a sphere is pretty easy to polish. Difficult optics are tens or hundreds of wavelengths of light away from the nearest sphere. These mirrors are 50,000 wavelengths away, so it’s a very difficult surface to polish and it took us a long time. But, the first mirror is finished and safely in storage; we’re ready to send it to Chile once the facility is ready for it.

Mirror number two is being polished on the front surface and that will finish about 12 months from now. Mirror number three is ready for diamond machining, which will rough out the rest of that parabolic shape. Mirror number four is currently facedown while we work on the back part to attach all the hardware that will hold it in the telescope.

It’s kind of like those old cooking shows, where one just came out of the oven, and then we have one at each stage, so it’s almost an assembly line process.

So, are all the mirrors identical?

McCarthy: The six that go around the outside will be as identical as we can make them. The one in the center is on-axis, so it’s simpler. The key thing that needs to match is the focal length of each mirror, since the magnification of each mirror needs to be the same. If you tried to combine the light from seven mirrors that all had different magnifications, you’d just get a big, blurry image.

One of the challenges in the polishing and testing is making sure that each mirror has the same focal length to within about 300 microns. On a mirror that has a focal length of 20 meters or so, that’s a challenging measurement. But that segues to another interesting engineering aspect of this.

Comparison of the Hubble Space Telescope's view of the core of the galaxy M100 before (left) and after(right) corrective optics were deployed to compensate for the optical aberration in Hubble's primary mirror.

Comparison of the Hubble Space Telescope’s view of the core of galaxy M100 before (left) and after(right) corrective optics were deployed to compensate for the optical aberration in Hubble’s primary mirror.

You’re probably familiar with the Hubble Telescope, and the fact that when it went up, the astronomers discovered that the primary mirror wasn’t quite the right shape. So, what we’ve done is made sure that we don’t just have one good optical test, or two that have to agree—we have four independent tests on the shape of the mirror, and we don’t declare the mirror to be done until all four tests give the same answer. Those four tests have different sensitivities to different aspects of the mirror’s shape, so when we bring them all together we can be confident that we’ve covered all of the possible optical aberrations, not just focal length.

Can you tell us a bit more about these tests?

McCarthy: The first is a classical interferometric test: we make an optic that produces a wave front that matches the mirror shape that we want. We do that now with computer-generated holograms; in the old days, you had to make lenses that produce the right shape, now holograms do that for us.

We project that perfect shape on the mirror and then we bounce light off the mirror so those two wave fronts interfere with each other. When the mirror’s just the right shape, you get a nice, uniform pattern. If it’s the wrong shape, you get places where the light is brighter or darker than it should be. That gives us a map of the surface of the mirror compared to the surface we want to polish in. That’s the most sensitive test.

The polishing tool being used on GMT mirror 2 in August 2017. (Image courtesy of Giant Magellan Telescope - GMTO Corporation.)

The polishing tool being used on GMT mirror 2 in August 2017. (Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

We have other tests where we simply use a high-precision laser to measure the surface from the focal point, so we can get the focal length and large-scale shape accurately. We have other tests where we project a uniform line grid onto the mirror and then focus it to see if the grid is distorted—that tells us if there are slope errors on the mirror. The fourth test is a little subtler, but it also measures the low-order aberrations: the focal length and overall curvature of the mirror.

What material are the mirrors made of?

McCarthy: It’s a borosilicate glass made in Japan. It’s very exacting, with very low thermal expansion and very uniform thermal expansion from piece to piece. It’s also made in very small batches, so it takes a lot of batches to get one mirror’s worth of glass: about a year to produce one of our mirror segments.

Shelton: Early on, they had a lot more errors in the batches they were producing, but now it’s much more reliable.

McCarthy: Yeah, the rejection rate now is very low.

My impression is that this whole process is getting easier with each mirror. Is that right?

McCarthy: Sure, there’s always a learning curve and it’s always steep at the beginning. We learned a lot on the first mirror, and we’re bringing those lessons and new techniques to subsequent mirrors. The speed at which we converge has been improving, and the risk of missteps has been reduced dramatically. You learn as you go, and that’s just part of the nature of a big engineering challenge.

GMT mirror 1's transporter container lid is lifted into place at the Richard F. Caris Mirror Lab at the University of Arizona in September 2017. (Image courtesy of Giant Magellan Telescope - GMTO Corporation.)

GMT mirror 1’s transporter container lid is lifted into place at the Richard F. Caris Mirror Lab at the University of Arizona in September 2017. (Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

Shelton: Practice makes better. I was president at the University of Arizona when the first mirror was undergoing polishing, and there were some nervous times, I can tell you. Would it converge? Would these tests prove that we can get the necessary off-axis shape?

But, in fact, a whole lot of clever astronomers, scientists and engineers figured it out, and now that the algorithm has been proven, it’s going much more rapidly for each subsequent mirror. The fourth mirror, of course, should be the most straightforward because that’s the center mirror. It was wise of them to start with the off-axis mirrors, because if you can’t do those, there’s no point to making the center one.

You’ve mentioned a number of challenges already, but what would you say was the biggest engineering challenge you’ve faced with the project so far?

McCarthy: I think there are two fundamental challenges. If you consider what a telescope does, it collects light and brings it to a focus. It needs to do that not just in one place, but across a field of view. So, the first challenge is making the optics.

Our biggest concern at the beginning was whether we could make these big off-axis mirrors. Now that we’ve demonstrated that we can do that, the question is whether we can align them to a fraction of a wavelength of light, so that when the light is combined we get the power of all the mirrors together and whether we can hold that optical shape as the telescope tracks across the sky and responds to the outside elements. One of the challenges of telescopes is, even if they’re in a dome, they’re effectively outside, so you have temperature changes, wind, humidity, etc.

(Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

(Image courtesy of Giant Magellan Telescope – GMTO Corporation.)

We’re past the first challenge of manufacturing the optics. Now we’re dealing with the challenges of aligning them and holding that alignment. We’ve been working on that through the standard engineering techniques of modelling and prototyping. We’ve done multiple subscale prototypes on telescopes that we have in Chile and in the laboratory as well.

We’ve convinced ourselves that we know how to do realignment and we’ve done it on a subscale, so now we simply need to prototype some of the other elements—sensors and closed loop tracking—to convince us that we can keep the mirrors aligned under all or nearly all observing conditions. That’s the next tall pole in terms of engineering challenges. But we feel confident that we’ve got that well in hand.

Shelton: That’s one of the reasons why the very first stage will be to get on the mountain in 2023 with four mirrors, while mirrors five, six and seven are going through this multi-year process of casting, polishing and refining. If we can get on the mountain with four mirrors—which is still a huge advance—and do the engineering studies, that can then confirm in reality what we feel confident we can do in theory and with models. That’s a nice operational advantage: being able to test it out before all seven mirrors are polished and completed.

McCarthy: Besides that, for astronomers like me, if we see four mirrors sitting on the ground next to a telescope, we want to put those mirrors in the telescope and start doing some science!

What do you find most exciting about the GMT?

Shelton: For me, the physicist but non-astronomer, I get excited by all the recent discoveries that we’re seeing in cosmology and astronomy—the LIGO discoveries, the neutron star collisions—and knowing that the GMT is going to take that even further: the discovery of all these exoplanets, understanding what kinds of molecules are on those planets and in their atmospheres. To me, it’s being able to take the marvelous discoveries that are coming about in cosmological astrophysics and probing farther and deeper to answer some of the fundamental questions that we all ask about being alone and the origins of the universe.

McCarthy: There are other exciting parts too, but I’d say they’re more adrenaline-rush thrilling. You want to make sure that you’re meeting deadlines and getting contracts out. We’ve focused on the mirrors because of the fifth casting, but when we take the first four mirrors to the summit, we have to have a place to put them in the telescope mount, and we have to have a place to put that telescope mount into the dome. So, making sure these three huge components and all their subsystems march along in sync and come together at the same time, that’s a bit thrilling too, if I can use the word in a broad sense.

Is there anything we haven’t covered that you’d like to share with our readers?

Shelton: We’re always looking for talent, and the success of a project like this rests on many different aspects, but the talent and dedication of engineers is essential. We have scientists and we know what we want to achieve scientifically, and we have an organization that knows what it needs to provide in terms of resources and structure, but really this rests on engineering talent, creativity and ingenuity.

It’s about finding people who may have had a passion for astronomy in their youth, and now have the analytical and engineering skills to bring it about. Those engineers play a huge part in the scientific discoveries that come from facilities like this. They’re an integral part of the team, but they don’t always get the recognition they deserve. Without them, we couldn’t do the science.

Think you can contribute to the GMT? Visit the GMTO employment opportunities page.

Written by

Ian Wright

Ian is a senior editor at engineering.com, covering additive manufacturing and 3D printing, artificial intelligence, and advanced manufacturing. Ian holds bachelors and masters degrees in philosophy from McMaster University and spent six years pursuing a doctoral degree at York University before withdrawing in good standing.