NASA's Tiny Chip Could Find Alien Earths!

>> Hi everybody. My name is Rachel Morgan. I'm really excited to be here today. Um, in addition to my PhD, I was actually here for my undergrad and masters. So, it's here for nine years straight. So, it's really fun to be back in town. Um, and I was actually in Professor Miller's last capstone class when I was in undergrad um, with Reichat. So, it was exciting to see that on the screen. Um, but today I'm very excited to be talking a bit about my research at NASA as on an integrated fatonic coronagraph called Astropic. Um, so first I'm going to talk about what integrative photonics is and what coronagraphs are and why integrative phatonics are really exciting technology for them. Then I'll talk about a test bed that we've been building up at NASA as some recent results from the project and I'll finish with some next steps. Um, so who here knows about the

habitable worlds observatory? Is everyone excited? Um, so it's another big flagship. So in in addition to all the excitement around Aremis right now, another big push for NASA in the next couple of decades is the search for life on other planets and the habitable worlds observatory. This is like the next James Web Space Telescope. This is going to be a huge um observatory that is going to be sent out to space with the goal of characterizing and finding signs of life. Um and they want to be able to survey at least 100 nearby stars and see if there are bio signatures in their atmospheres. Um, so this is a diagram that shows the different kinds of exoplanets that have been discovered so far. Um, the this is honestly amazing. There's over 5,000 I think it's over 6,000 now exoplanets that have been found, which is really cool considering that the first ever exoplanet discovery didn't happen until like the '9s. Um, and so it's a very new field and it's been growing very quickly. Um, but as you can see, uh, we've discovered a lot of different exoplanets. We've learned a lot about different types of solar systems, but only 4% of the discovered planets so far have been terrestrial. And so, as we go out and search for signs of life on other planets, we're kind of looking for planets that roughly look like our own.

Um, and it's very, very difficult to find signs of an Earthlike planet, um, because the Earth is very small. It is very close to the star and it is a lot dimmer than the star. Um so this is a a a diagram that kind of shows us how we study these planets. So right now we are looking at a star. Stars as we all know are quite bright. Um and the way that we are able to study for and look for signs of planets around stars is we use something called a coronagraph. So the coronagraph is the component that blocks out the starlight very carefully and allows you to see planets that are orbiting nearby. Um, coronagraphs are a very powerful kind of optical technology, but they're very, very difficult and very complicated to get right. Um, one of the biggest challenges is wavefront errors in your system. Um, if your wavefront is off by like a level that is less than the width of your hair, your optical system could be totally ruined and you might miss your signal and you won't be able to see it. In particular, this sort of device is extremely difficult for an Earthlike planet because the Earth is actually 10 billion times dimmer than the Sun. um and it's also very very close in the sky. And so this is a really very difficult optical challenge and we still don't really know for sure uh the optical technology that we're going to use to actually study these uh exo earth worlds. Um so the motivation for this project is

to look into an advanced optical technology to see if we can make this challenge a little bit easier. So chronograph instruments have really really intense wavefront control requirements. Um and they're also extremely sensitive to mechanical and thermal effects in space and they're very very complicated to design and implement which leads to very large and very expensive uh instruments. So this is a plot of different coronagraphs that have been studied. Um and so this plots the exo earth yield which is the number of earthlike exoplanets that you can study over their mission lifetime as a function of your telescope diameter on the x-axis. Um so you want to have a small telescope and you want to have as many exoplanets studied as possible. Um because as your telescope goes larger your mission cost tends to scale with that. Um but something that is very interesting is that these Roman architectures so the Nancy Grace Roman Space Telescope is the next uh space telescope that's launching in October. Um this is actually the first time that a coronagraph is being sent to space which is really exciting. And this is a picture of the coronagraph instrument that is a technology development um instrument on this mission. Um so this architecture is going to be able to get the best imagery of exoplanets uh with a space telescope that's ever been possible before. Um but this instrument is about the size of a baby grand piano. It's very heavy. It cost a lot of money. It was very difficult to design, implement, test everything. Um and it's still not going to be able to study exorth. still not going to be able to hit the contrast requirement that we need for an Earthlike planet. And so the next type of uh coronagraph for HWO or the Habitable Worlds Observatory is going to need to be even more advanced than this. So it's going to need to be even bigger. So even bigger than a grand piano. Um it will likely be the size of like a school bus or something. These instruments are going to be extremely capable but also extremely complicated and extremely difficult. And they're also still not hitting the theoretical uh limit of what you should be able to see. So physically um the people who wrote this paper, Russ Bell, he looked at the challenge of studying exoplanets and he asked what is the best possible efficiency we could get out of an instrument. Um and it turns out to be it's similar to doing a projection. So if you have your star in the sky and then you have a planet right next to it, you can get a certain amount of information from your planet that's distinct from the information from the star. So when you get rid of the star information, you are left with the planet signal. And so that's what that uh dotted line is up top there. And as you can see, current chronograph architectures are still quite far below that line. And that has an impact on your exo Earth yield. That means that you can study fewer planets. It will take longer to reach mission goals. You get less science. Um so I'm talking about a alternative

technology which is called photonic integrated circuits or pics. Um so picss are this red line here. Um which are the only technology so far that we've studied that can seem like it can actually get closer to this theoretical limit. Uh fatonic integrated circuits apply a lot of the advancements from nanop fabrication that have happened in the semiconductor industry to materials that can route and manipulate light. So instead of having mirrors and grading and extremely complicated optical systems that look like that with the Nancy Grace Roman instrument, you can actually integrate a lot of tiny tiny nanocale components onto a chip. And so that is our demo chip which has it's not as capable as the chronograph instrument there. So it's not a fully fair comparison, but we have hundreds of optical devices on that tiny little chip there that you can hold in your hand, which is really incredible. Um, and the reason that a pick can be

used instead of a bil optical instrument is that when you build an optical instrument, every single component can be described by a matrix. So a lens can be described by a matrix, a grading can be described by a matrix, and your whole system is the product of all of those matrices. Um and another Dave Miller by chance has done a lot of research on showing that if you have uh any optical system that can be described by a matrix, there's actually a decomposition that exists that can make your optical system matrix look like the product of many small interferometers. And so a Moxender interferometer is shown in this bottom left corner here. And so this is an interferometer that has a 50-50 coupler on either side and then there are two degrees of freedom where you can control the phase. Um so that's the fi and theta phase shifter there. And so by controlling the phase of phi and theta you can actually implement any like rotation of your input signal that you want. Um and so the idea of this sort of decomposition is that with enough of these small uh interferometers you can actually implement any matrix that exists as long as it is linear. So you can emulate any type of optical device you want uh using this sort of onchip arrangement of interferometers as long as you have the right number of degrees of freedom which is pretty incredible. And so the benefits of this are that

we're not limited to what we can actually achieve with like the lenses and the optochanics that are currently available. Uh we can come up with any matrix we want and apply it onto this chip. And so this has the potential to drastically reduce the mass and power footprint uh by over a hundred times. um lighter and over 30 times smaller and also increase the robustness of our instrument because integrated photonics are created using wafer scale fabrication. So they end up being monolithic devices that it's impossible to make them misalign relative to themselves. Uh that actually is what inspired me to join into the field of integrated photonics is that I did a lot of alignment and I thought this is really difficult. I'm going to do phatonics because you don't have to align things anymore. Um and so I think that will have huge benefits for not just coronagraphs but space instruments in general but this is in particular a very good application of this technology. Another big advantage is that integrated photonics are particularly good at accessing planets that are very very close to their host star. Um so earthlike planets tend to be very close to their host star. And so this improved performance can actually help improve the risk margins of your overall mission. And so hopefully it can help lower the cost and increase the science performance of the Habitable Worlds Observatory. So now I'm going to talk about the

design of our instrument. Um so this diagram here shows the overall concept. We take the light coming from the telescope and then a microl lens array is used to concentrate the light onto a set of grading couplers. These are small periodic structures that defract the light into the waveguides. So the waveguides are like onchip optical fibers where they can find the light and have single mode routing with very low loss. And so once the light is coupled into one of these waveguides, it can get all around the chip without incurring a lot of extra losses. And then we have an arrangement of these uh interferometers and those are the devices that we tune in order to implement the coronagraph operator that we want to use. Um and then on the right there you can see our first uh prototype chip uh that we built for Astropic. Uh so this was built by a commercial foundry and this is an image of it after it was wire bonded to a host PCB that we used to control it. Um and as you can see it's very very tiny. It's about the size of a dime. Um and so the concept of operations is that we're essentially using this chip as a very very fancy type of filter. Um so when we look at the star in the sky and the planet we know that they occupy different spatial modes because they're coming from different directions in the sky and so we can use that information uh to tune this interferometer to sort out the light into its various spatial modes. So the idea is that the incoming

light is a mix of planet and starlight and then it passes through this filter and so and then you've confined the star light to the upper modes leaving the planet light in the bottom modes which you can then send to a spectrograph. Uh here's a picture of the first prototype that we fabricated. Um so we actually have two different chronographs on this very first prototype shown here. Um but I'm going to talk about the bottom one and this is a nine input 9 output full triangular mesh of interferometers. Um and so we have a 3x3 array of inputs. Um so these are like you can think of them as the pixels. So there's 3x3 pixels and then from those inputs we use all of the interferometers along the chip and can measure the outputs um in order to implement our coronagraph system. Here is a diagram of our chronograph

test bed built at NASA as the goals of this test bed were to demonstrate the performance of astropic as a coronagraph uh with a realistic telescope pupil. So to do that, we have a laser coming in. We can control the polarization. We have a monitor photo detector uh to make sure that our laser power isn't fluctuating. And then we send the light onto the pick. Uh we use a DAC to control the different phase shifters on the PIC. Sorry. Um and then we have a fiber array at the output that couples the light from each channel to its own photo detector and we can read those all out simultaneously. So here's a picture of uh the test bed built up at in California at SA as um and so it's very very tiny so I have to zoom in a lot. Um but this is a side view of the pick. Um so this is showing an input of a single optical fiber to kind of show where the input grading couplers are. Then you can see all the wire bonds or those little gold lines. And then that uh glass thing at the end is a fiber array and that is what goes to all of those yellow optical fibers which go to the photo detectors so that we can read it out. Um so now I'll talk about the most

important building block which is that Moxender interferometer. Um so we can actually tune a Moxender interferometer uh using something called a thermal phase shifter. This is a tiny little wire that is integrated right above our wavegu. And by sending a current through that wire, it's a resistive heater. And so that heat changes the refractive index of the wavegu. And that changes the phase of the light going through. And so this little gif is showing what happens when you sweep through the voltage. Um, and you can see that it actually toggles the light between the two outputs. So it's an active switch that we have on the chip. And then the plot on the right is showing the extinction ratio. So that is the amount of light in each channel relative to the total amount of light um summed up between the two channels. And you can see that even with a single Moxender interferometer, if you control it to exactly the right voltage, you can actually get over eight uh orders of magnitude extinction of your signal, which is pretty incredible. With just one tiny device that is so small you couldn't even see it with your eyes um on our chip, you can get this level of performance. So the first kind of experiment that I did with this new test bed was that I wanted to see how far we could push this. Um so I said before that we need 10 * 10 thegative -10 like 10 billion times extinction of our star signal. So I wanted to see how much extinction I could get on the chip. Um so I aligned a single optical fiber to the input. This is like the simplest kind of input we could have. Um, and I used all of the DI Mox enders on the mesh and I tuned them to try to have as little signal as possible in the bottom output. And I did that um I think this is going to go okay. Yeah. So I did that by sweeping through the voltage iteratively uh using smaller and smaller increments until I hit the voltage resolution of the DAC. And so you can see that routine in this video over here where um yeah, after finding the right voltage, we can actually get rid of all of the light from the channels that I'm videoing here and confine it to the upper channels. And this is a plot of the extinction ratio uh over time during this experiment. So, we're actually able to get to like 3.5 * 10 -10 extinction, which is not quite the 1* 10 extinction necessary for uh Earthlike uh exoplanet detection, but is still a very exciting technological achievement. Um, because we were able to get this result in an optics table in a normal lab in air without any kind of like clean room environment. Whereas to get this sort of performance out of a bulk optics chronograph, you would likely need to be in a vacuum chamber, you would need to have this very expensive complicated setup in order to get this sort of performance. Um, so this really shows the high stability offered by integrated phatonics and how that's very exciting for this sort of telescope application. And then next, I wanted to do a test

where with a more realistic telescope input. Um, so before I was just using a single optical fiber, which isn't actually what the output of a telescope looks like. So we also built a free space coupling input where we have the laser source launched in, we columnate the light, and then we reimage the light um, right on top of the chip input and that allows us to image something that looks more like a telescope pupil. So this is a picture of what is coming out of our source uh, input over there and then we align it over the pick as shown in that little inset there. Um, and then we also insert the microl lens array in order to concentrate the light onto our gradient coupler inputs. Um, and so in order to do this test, uh, this is implementing that mode sorting that I was talking about before. So to start, I take the first diagonal and I tune each interferometer in order to try to maximize the amount of light that reaches the top channel, channel number eight there. So after doing that I you can see that number eight has gotten a lot brighter and there's less light everywhere else. So then I moved on to the next diagonal and so this is the result of that uh where we have most of the light in the top two channels and less everywhere else. And then from there I sorted all the modes that I could see and then once I got to a point where the noise was too high I would um just tune the rest of the interferometers just to directly minimize that signal that I wanted to get rid of. Um so this is the result. Uh so the blue line is the signal in each of the output each of the nine output channels before I started tuning the mesh. And then the orange line is after tuning it. So you can see that we were able to confine a lot of the light and control the light uh into the upper channels. And then there's one dark channel that got an extinction of 1* 10us 7 which is very exciting.

And uh then in order to see that this was still a useful chronograph device because it's one thing to get rid of light but you actually want to be able to receive the light that you want. So having low noise is the first hard challenge but then the even harder challenge is still recovering the signal from the planet that's right next to the star that's 10 billion times brighter. And so in order to emulate that I actually moved our source position. Uh so I changed the angle of the incident light to make it look like it's a planet coming from another spot in the sky. And that's what this green line is. So you can see that when I move the source position, you can actually still get about 20% of the light going into that bottom channel where all the star light has been rejected. And so that shows us that this is a very promising technology and can actually operate as a chronograph. Um so yeah, in conclusion, uh the

Astropic project is doing some very exciting experiments showing that integrated phatonics is a powerful technology option for space telescope coronography. Um we have recent results showing very high extinction on chip that we're very excited about. Um and then moving into the future uh we have a lot of more work to do to make sure that it can actually be a useful coronagraph. Um so there's a lot of different architecture discussions that are ongoing. Um but yeah, thank you very much for your attention. I'd like to acknowledge all my teammates as well. Thank you.