University Place

-

Jim

Welcome to Space Place. Tonight, we have a guest speaker, as usual, and our guest speaker tonight is Dr. Andreas Velten, who is with the laboratory, the UW Laboratory for Optical and Computational Instrumentation. And he's been here since getting his PhD at the University of New Mexico and then working at MIT for a while. He's been in Madison now for about three years working at this lab and coming up with, at least I'm pretty sure the first on the UW campus, of this interesting technology that he's working on that he's going to tell us about tonight and the ability to take apart pulses of light and dissect them for geometrical information about interesting places. So, Dr. Velten. -

Andreas

Thank you very much for the introduction. Thank you. Thanks for coming. What I'm going to talk to you about today is exploring caves that have been, relatively recently, discovered on the moon. And this technology that we're gonna talk about is actually going to be interesting for caves all around the solar system. This photo was actually taken on a cave on Earth. And you can see this was taken with a long exposure time. There was just somebody walking around with a light, so it's just a nice artistic effect. And the team members on this, so I am the lead on this project here at UW, and I'm working on the imaging technique. But this is a collaboration with the NASA Jet Propulsion Lab, and Karl Mitchell is the lead scientist there, and Jeff Nosanov is our PI. So, we are interested in looking at these objects. There's over 200 that we have discovered now, and they are, what we think, entrances to caves on the moon. So we can't look inside. The only thing we have are pictures from orbit. So everything we know about what's going on in there is basically just speculation about what we know about the geology and how these should have formed. And today, I mean we're gonna first look at how do these caves form and why do we wanna look at them at all? Why are we interested in caves in the solar system? And then my big part in this project is how, how can we actually look into this cave without having to go down there, because that's gonna be expensive and dangerous. So, why caves? We are interested in caves on the moon and everywhere in the solar system actually because they provide interesting science. They provide a way of getting to the underground geology of this planet, learning how it formed without having to drill, and they also can teach us something about how these caves form on Earth, and we can test our theories about what we understand, how they are formed by just comparing them to caves on other planets. Another very interesting thing we want to do in these caves is they provide shelter. Even on the moon, of course, you have solar radiation and the temperatures change a lot between night and day. So you couldn't just land a ship there and have people living on the surface of the moon, even if they are protected from the vacuum. They would basically need some serious air conditioning up there and some shielding against the radiation, but in the cave, they're fine. They're under 50 meters of rock, and you could actually just camp out there if you inflate some tent with oxygen with an atmosphere that you could live in. So they provide shelter. And the same effects that provide the shelter actually also could provide protection for volatiles, volatile materials. There could be water deposits on the wall, like water vapor, or something else that, on the surface, would have been heated away by the sun, but in the cave, it may still be there. So that's another incentive to why we want to look at these. So, from all we know from looking at them from orbit, we think that these are lava tubes. And lava tubes on Earth usually form in a volcanic eruption. So in this image, you can see you have this lava flow that is coming, lava coming out of a volcano, and it basically just forms a river. And then it kind of digs this hole further down here. And then as it starts cooling, of course, it starts cooling at the surface. And it basically, you get this roof that forms over it, and then at some point, the tube just drains, and part of the inside hardens, and then you get this cave. And you can look at these on Earth. In Hawaii, you have lots of very nice lava tubes that you can go and hike through. In New Mexico, there are lava tubes that you can go through. Now, we think that that's what happened on the moon. And then what happens, of course, there are some event that collapses part of the ceiling, and that's what we're looking at. Okay, so we would like to look at the inside of these tubes to confirm that what we think is happening is actually happening and to learn all these things I have pointed out, but it's difficult. The first thing we usually do for scouting is once we have looked at the place, we send a rover. And JPL is actually working on rover missions in two caves. It's a very interesting subject, but it's not easy. First of all, you have mobility problems. How do you get this rover, this little robot to move around in the cave where there's boulders lying on the ground everywhere and you can't avoid them? It's much harder than moving around the surface of Mars, where it's basically mostly just sand that you can drive around anything that's dangerous. Exits, you still-- The entrance of the cave is steep. You have to figure out a way to get down in there somehow. This is a picture of one of the lava tubes we actually studied in New Mexico. This is Four Windows Cave, yeah, lava tube. And you can see there is a lot of clutter lying down there. It's not easy as a person to walk through there, but for a robot, it would be much harder. Instrumentation is an issue, although rover instruments we have are for surface applications. And then, of course, communications is hard. Like, you are under 50 meters of solid rock, so a radio signal won't penetrate that very well. And then you have problems controlling it. Here is a schematic of how such a rover mission, such a rover mission could look like, where you have a lander at the surface there and you have a lava tube skylight, and then you can see the little cave crawler there. And JPL is actually building these robots. They're building robots that can... like, they have like little arms that can hook onto a solid wall and can support the robot so it can climb up walls and walk along the ceiling and stuff like that, and they want to send them in the cave. Now, but the problem is you still would like to actually make sure that what you think is a cave entrance is actually a cave entrance because otherwise, you end up with a very expensive mission. You send a rover in there, and all you find is, that, well, the entrance is blocked or it's really just a hole in the ground, which would be embarrassing, at the very least. And that's where my research really comes in. What I contribute here is I have developed this method that allows us to image beyond the direct line of sight. So, this plot on the top here shows if you're just looking with a satellite, this is how far you can see. You can basically see along these red lines. But with this multi-bounce imaging that we're gonna use, you can actually use the next reflection of the light and image into the cave and see these entrance areas of the cave, make sure that they're open. If it's blocked, it's most likely blocked by something in the first collapse that blocked the entrance. So we just wanna make sure that they're opened and maybe wanna look at the initial entrance area there to see something about the cave. Okay, so the question is, how can we see around corners? And this research kind of started in my postdoc at the MIT Media Lab, where we really looked into this problem. Can we use this multiple-y scattered light in a room to see around corners? And the first thing we really did is we built, or I built in the lab from lab equipment a really fast camera that is fast enough to capture moving light. And this is one of the videos that we made. We're not affiliated with Coca-Cola. This just ended up being the best quality video, so we're still showing it. And what you're gonna see here is this is a bottle with water and a drop of milk, so it scatters a little bit, and we send a laser pulse through this bottle. And the video is in black and white, but this is what it looks like. Now you can see it hitting the cap, and then the light from the cap hits the floor underneath, and then you can see a little bit in this air bubble up there, the light is trapped a little bit and bounces around. And then you can see it hitting the back of the bottle again, there, coming back. In this video, our time resolution is up to about two picoseconds. That's light, in two picoseconds, light moves less than a millimeter. So in each frame of this video, light moves less than one millimeter. And that laser pulse we're looking at is about a million times faster than a normal bullet. In fact, if you could capture a bullet with this technique and slow it down as much as we slowed down this laser pulse, it would take several weeks or maybe a month to cross the screen. So this is much faster than normal objects that we would be interested in looking at. But with this technique, we can capture light moving through a scene in slow motion. This is something that happens in this room everywhere. You can see the light bounces around. It's just usually too fast to see. Two picoseconds per frame, which corresponds to about half a trillion frames per second, if you wanna express it like that, and the whole events that play out in this in a nanosecond or a few nanoseconds. But that allows us to analyze how the light travels through the scene by just watching it. And that's really what we're gonna be using. Here's another scene that we set up, which is basically a tomato and a roll of tape, and on the side there, we have a diffuser. And we hit our diffuser with a laser pulse, and then you can see the light coming from there, as if there is a light bulb that you turned on, and you see the flash of light going across the scene. So here's what that looks like. Ignore this. This is just scattered light. And then you see the light coming out and illuminating all these objects. You can see it creates this sphere, this sphere of light that interacts with the scene. You have the light, the laser light coming in from the side, it bounces off this diffuser, and then creates a sphere of light that moves across all these objects, and that's what you've been seeing in the video. And then there's some interesting things that happen. You see the direct light that illuminates everything, and that's really what creates a photograph of a scene. But there's also, for example, if you pay attention to the inside of this roll of tape, and I'm gonna play this video again later. If you pay attention to the inside, the inside is not illuminated directly. There is no direct paths from this diffuser where we have our light source. Let's see if I can mark this. From here, this is where the light is coming from. There's no direct path into the roll of tape. So that lights up later and indirectly, and then it stays lit up because the light bounces around in the roll of tape. So it's indirect lighting, the same true for the top of the tomato. The tomato itself has subsurface scattering, so the light actually enters the tomato It goes inside and then takes a while to come back out. So we actually see the light stays lit up. So photons bounce out, are trapped in the tomato, and then for a while, it takes them a while to come back out. You can see the, the shadow of the tape forms after the tape has been illuminated. And you can see there's a reflection of the table on the tomato. And then the important effect that we are actually using in this research is if you pay close attention, if I play this video again, but if you paid close attention to the back wall, you're gonna see two flashes of light. One is the light coming directly from this point and hitting the wall that's very bright. And then, after that, there is one that's much dimmer, and that's from light that has hit the wall, it hit the back of the tomato and come back. And that happens later, of course, because light took longer to travel. And that tells you what's behind the tomato essentially. You can use that light to image the back of the tomato. So, I'm going to play this video again. So this is just scattered light. Then we have it illuminating everything here. Now we can see this shell of light is moving out. We can see that the inside of the roll of tape is kind of glowing now because the light is trapped in there. It bounces back and forth inside the roll of tape. It starts illuminating the tomato. Now there's a first flash on the back wall, and here's the second one. I stopped it. I think you can see it there. There's a second faint flash of light. That's the light that traveled around the corner and bounced back. And that's much dimmer than the direct light, so it's technically challenging to actually filter it out and image it. And this is actually, the intensity is on a logarithmic scale, so it's much dimmer, actually. Another thing that you can see, and this still now is the direct light is just visible a little bit up there on the side, on the corner of the screen. So, all the illumination that you see in the scene now is all indirect. The tomato basically, because it trapped some of the light and it's coming back out now, the tomato is glowing and illuminates the scene. This light pulse goes like a spherical shell across the scene, and there's the second flash of light that comes back from the back of the tomato. So that's what we want to use, essentially. That effect is what we're using to see around corners. And we have a nice video that kind of demonstrates how this works. (rhythmic drum music) So this was our original experiment that we did at the MIT Media Lab, and we had this mannequin there. (rhythmic techno music) This is the camera that looks at the scene, but it can't look at the object directly because there's a wall in between. (rhythmic techno music) And this is our reconstruction, our reconstruction of the object. (rhythmic techno music) Okay, so the way it works, we fire one laser pulse, or a series of laser pulses, but for the sake of explanation, just one pulse, and it hits one point on this wall that we can see. And from there, it scatters in all directions. (rhythmic techno music) And then some of the light hits, hits this hidden object and travels back and forth. And there, the camera can see it again. (rhythmic techno music) Now of course, the trick is that this camera can't just see where on the wall this photon hit but also when. Two picoseconds time resolution. (rhythmic techno music) So the point is we know for each photon, for each light particle or for all the light that we collect, we know exactly how long that traveled, the distance it traveled through the scene. (rhythmic techno music) And then we can change the position where we shoot our laser as well and just collect more data. And then we have a reconstruction algorithm that takes all these data and uses that, change the laser position. In this case, it was 60 different laser positions. And we have an algorithm that can take these data that we collect and reconstruct an image of this hidden object. (rhythmic techno music) And this is the actual reconstruction of this mannequin. (rhythmic techno music) So, as you can see, you don't get quite as high a resolution as you would get just looking at it, actually, quite a bit lower. (rhythmic techno music) This was done at the MIT Media Lab, so I'm leaving these on. And Jeff Marsh is actually the one who made the video for us. He works at Nature. Okay, let's recap that. We have a laser that sends out short pulses and a camera. It's called a street camera that has this time resolution that it can actually have two-picosecond time resolution and see the light. We shine that at a wall, and this dashed line on the wall indicates the line along which, that's what the street camera can see. The street camera is a one-dimension, it's a line scanner camera, so it only looks at one line at the wall. And we have an object that's not visible from either of those. We send a laser pulse that hits the wall, and we look at the light that goes from the wall, hits the object, and comes back and enters the camera. This is actually some of the, an example of the data that we collect. This is what the camera sees, essentially. This is X, along this dashed line on the wall, the position versus time, like when it happened. And then we can compute from that, again, we can compute the image of the hidden object. And the way that worked, so this is not really important for the rest of the talk, but I just wanted to briefly explain how this reconstruction actually works, in case you're interested. So this is a little involved. So what we know, for each data sample we collect on our camera, for each photon, we know... where on the wall it will send, where on the wall it was detected, and how long it took to travel this distance. If you know all that information, what we actually do is we back-project this photon. Basically, we draw the surface of all the points in the hidden scene where that photon could possibly have come from. So we know it was here at a certain time, then it traveled for this distance, and then it arrived over here. And then we basically say, okay, what, where could that object possibly have been that created that, that reflected that photon? And it turns out, if you do that, if you draw that, you end up with an ellipsoid, and the foci are these two points. And what our algorithm does is basically just draws all these ellipsoids, and then this image forms. We have an additional filter step, but that's really kind of a detail that's not very important. So the way this works, if you're in this scene, you just have a single patch. The way this looks like is you draw three ellipsoids and they all intersect at the point where the patch is, almost like triangulating the location of a cell phone with three distance measurements from a cell phone tower. But then we can draw more of them, and the more we draw, the more data we get. And then we have a filter to actually reconstruct the image of this patch. And this white line indicates the size of the patch. I think it was two-centimeter in this case. And you can see that we relatively, we can reconstruct this relatively well. So this is the object, was a 20-centimeter-high mannequin. This is, again, an example of the data. This is the back-projection after drawing in all the ellipses. Then we filter it, and then you can just do some post-processing to render it in 3D like that so it looks prettier, but the data is really already here. This is an example of our experimental setup that we have now at the University of Wisconsin. Over here, we have our laser. And from the experiment that I showed you before, one thing we did is we made it smaller because if we want to send it to space, it has to be compact. And the street camera is not and probably won't ever be space ready. It's also very expensive. So you can see our laser over here is quite a bit smaller now. The detector is actually over there, and it's a single pixel detector. Most of it in that black box is actually electronics. It's a detector that's just 20 by 20 microns, so it's really small. And that is a single pixel that detects the light and has the time resolution that we need. So this is something that we can actually send up on a satellite. There is a galvo scanning system here. And the way this system works then is this point on the wall, this is where the detector looks at. So the detector looks at one single point on the wall. And then we send our laser on different points in this area there, in this square. And we have set up some targets, some objects here around the room that we would like to measure. So this is the experiments we did for this project. And then you can see, there's a letter T and there's two patches. And then, you can see here, this is the kind of data we get for two of the laser positions. I mean, this is basically the light that bounced off of one point of the wall, and then let's say the red curve is the light that bounced off one point of the wall, bounced off of all these objects, then hit the blue point, and then got back to the detector. And every peak is a reflection from one of the objects. And then the blue curve shows you the light that, where the laser points at a different point of the wall. And you can see all the pixels shifted. And that is essentially, again, this is what we used to reconstruct the position of all these points and reconstruct an image. So, the experimental results are, this is, again, a reconstruction. You can see these two patches in the scene. So these are the two patches, this one and this one. And you can see the letter T here. And then you see this big blob over here, which is actually the lens in front of the camera. So the camera, in this case, the camera actually took a selfie. It took a picture of itself because it's in the room. And in the future, so this setup, obviously, we can't... this is not directly for space applications because we're really close to the wall. So this is a three-dimensional model. Obviously, you can turn this around and look at it from other views now. So, in this case, we just had about 50 milliwatts of laser power. A laser pointer has five milliwatts, so it's not much more intensive than that. So this could actually probably be changed so that you can use it on the Earth and it's eye safe, so you can look into the laser and you won't get injured. It scans 185 points on the wall, and the reconstruction runs fairly quickly in math lab, five seconds. We actually tried to use this and image the entire room. You can see the other side of the room. In this case, that's limited by the laser, so this is just a technical limitation. So with this current light that we have, this white patch, this big white patch is the furthest out we can go. And then what we want to do with this one, and it's just another application of this technique, the idea here is let's say you're with the fire department and you're standing in front of a burning building and you wanna look into all the windows and see whether there's somebody inside that needs help. So that kind of application is something I've had in mind for this system. And what we can do, this is fairly simple to take this detector and just move it further out and then put a bigger objective on it so it could still collect the same amount of light. The laser is columnated anyways, so that won't change. You can take that 10 meters away and still focus on the same point. And then you could do this through a window into the room and get an image of the room. This is an application that we have been working on for Earth. We also have been working on other applications for surgery. Surgery often has problems with, like if you're doing minimally invasive surgery, you have this tiny opening that you have to go through, and it's really hard to do this because you don't have a good three-dimensional sense. You can't touch anything, which is what surgeons usually do a lot, and you can't look around. You don't know what's going on next to you. You only have this tiny field of view in front. So that would be interesting. It would be interesting to do it from an airplane. It's good for disaster response. So there's lots of applications beyond looking at caves on the moon that are interesting to mention. Oh, this was just published, in case you want to read more about it. So, but this, of course, is nowhere near the scale of what we would need to do this on the moon. So, what we did for this project is set up a simulation and try to see what we actually have to do or what kind of laser, what kind of satellite would we have to build in order to get enough signal back to image this and what would the reconstructions look like. Now down here, you see a solid works model, like a 3D model we made of the cave, in the cave. This was actually made, this model, this 3D model, was made by one of my students. She handmade this and basically just worked with geologists at JPL to make it to look somewhat like what we think the inside of the cave should look like. Nobody has seen inside, so there's no telling what is in there. But we, from our understanding how these lava tubes form on Earth, we can actually tell quite a bit of what it should look like, and that's what this is. So we made a model of what the scene should look like. Then we, basically, in our simulation program, place our imaging system with one kilowatt of laser power. So as I have said, a laser pointer has five milliwatts. The laser we used in our experiment before had 50 milliwatts. This has 1,000 milliwatts, and we actually now are thinking of going to 10,000. I was surprised when I started working on this that you can actually make such short laser pulses with that much power and then put them on a spaceship, but we're getting there. This is really something where there is a lot of progress. Then instead of a single pixel detector, we actually want to use a 64 by 64 detector array. For these fast spot detectors that we're using, that hasn't been built yet either. So that's something that needs to be done. People have done 32 by 32 pixels. 64 by 64 pixels doesn't exist yet, especially for the special GADR detector we're using. And then we are in an orbit that's just 10 kilometers. We want to be just 10 kilometers above the, above the surface of the moon, which means you have to be in a very close orbit, which means you're gonna be very fast. So if we just look down and don't tilt around, the time when we have the cave in view is actually just 100 milliseconds, 0.1 seconds, because we're moving so fast at such a low orbit. These caves on the moon are actually a lot larger than the lava tubes you find on Earth. The diameter of a lava tube on Earth is typically about 10 to maybe tens of meters, so 30 feet. And the reason why they are so much bigger on the moon is that on the moon, you don't have an atmosphere and you have much lower gravity, and it's mostly the gravity that makes them larger, but also, there's no atmosphere, so the cooling process happens different. So we actually have some, there's simulations and estimates of what happens if you have lower gravity and if you don't have an atmosphere, how does the cave look different. And that's actually something that's very interesting to us and, I mean, one thing, again, we were gonna learn when we look into these caves is are these models correct and do they make sense. So, we actually, so we did this, we simulated this in the computer, what happens if we send this light pulse in. The simulation programs we use are very similar to what you use, like it's essentially 3D rendering, what Pixar uses to make their 3D movies or what we use in games, except that we have to be a little more realistic. The rendering engines in games are very optimized and work very well to create images that look real, but they tend to cut corners on effects like this multi-bounce reflections if they don't actually influence the visual appearance of the image. So this the result of the simulation. You can see on the side here, I can draw again, so on the side here, this is just a schematic of the cave. The images there are results from the simulation for the 64 by 64 pixel array. So these are 64 by 64 pixels. All these blacks are 64 by 64 pixels. That's time in nanoseconds there. And what you can see is like how things unfold if you shoot the laser up there. Actually, it's down here. If you shoot the laser where this red spot is, then you get this top row of images. Actually, it's the spot on the bottom. This is an arrow. The spot should be here. But what you can see is that for a long while here, for the first three frames, for the first 40 nanoseconds, you don't see anything, and that's because there is an opening there. I mean, we're not looking at the direct light. The light comes back at direct light, we filter out and we're looking at this multi-bounce light, the light that traveled away from here into the cave and came back. And there's nothing there because while there is no wall here that would reflect it, the light actually goes deep into the cave and it takes a while until it reflects back from the ceiling or from other objects. So that's why we're not seeing anything here. And then, at some point, you start seeing it come back. And this is probably returned from the ceiling of the cave. And then you see it further here. Let's go to the next one. This is the one where you hit the side of the cave here, over here. And you can see here the light returns much sooner because on the side of the cave, there's a wall. You send the laser there where that blue dot is into the cave, and it bounces off the wall and comes directly back into the field where the camera can see it, and that's what you see here. And then it kind of spreads out afterwards. If you send the light pulse right in the middle of this opening of the cave, you see light coming back simultaneously from both walls because it bounces off both of the walls and then hits back. And then, over here, you can see the little bit faint over there is one coming from the ceilings from the back. So, without any computation, without any reconstruction, just from the data we collected, we can actually already tell whether there's an opening or not. And, again, this is basically, we just, we send the laser pulse in and we just look at this area at the bottom of the cave and see where does the light bounce back from. After collecting these data, we actually did run our reconstruction algorithm. And this is just a view of our model. This is what we are supposed to be seeing. This is our, we basically rendered this. The camera is sitting inside the computer model of a cave and looking down one of the, one of the cave entrances. And we place this boulder in the back there, about 100 meters into the cave that's blocking the entrance. And here is our first reconstruction. As you can see, we can see these side walls on the side of the cave here. You see these nicely. We can see the boulder in the back. We can't see these walls here that face away from us, and that's kind of to be expected. This is the same, this is an effect that's similar to what you have when you do flash photography. Whenever your light source that you illuminate your scene with is at the same location as your camera, you get this intense Lambertian... It's called Lambertian shading. Like that surfaces that are facing away from you appear much dimmer because they're illuminated much less bright. The same as the sun in winter, if it comes at an angle, it doesn't create as much light. So if you look at this in a photo that you take with a flash in a dark room, you see there's all the surfaces that are kind of oblique, that are not facing you like this. They all appear darker than they should. And this is what we see here. We can't reconstruct this because of this Lambertian shading, yet. So one of the big things that we're working on right now is improving those computational algorithms to take all of these things into account and improve the reconstruction quality of this image. But even if we can show that that is all we can get, that is already enough for the purpose of the mission because all we want to know is, is there an opening and is it big enough to get our rover through? What we have been working on, so this was a preliminary reconstruction. This is like a computer model in solid works that we handmade of the cave. So the next steps we're actually working on for the first simulation is, and this is a really fun one, we are making an improved model of the cave. So our computer model is pretty crude because we just made it by hand. So the next step is we actually went and scanned a real lava tube on Earth, and we just say, okay, now we have a computer model of that and it's fairly realistic, very high resolution, so let's say what could we image into that cave with our model. And this is the cave we scanned. This is Big Skylight Cave in El Malpais National Monument in New Mexico. So we went there for a while and did some caving, and we used a LiDAR scanner to get the geometry of the cave. And this is a result there shown on the site. And here is another one. You can see in this one there's an opening. This is basically this area of the cave. And the way this LiDAR works, actually, it gives us a complete three-dimensional model, not just an image of the cave, not just image. We actually have the complete 3D location of all these points with a very high resolution. And the way it does that is it actually does use the time of light. It sends a laser pulse, and rather looking, than looking, rather than looking at the multi-bounce light that we look at, this LiDAR scanner actually just measures the time it takes for the direct light of the laser bounce to come back. So for each point in this scene, there's a LiDAR scanner that's set up over here, it sends a laser pulse to the wall, and it just measures how long does it take to come back, and from that, it knows how far that surface was away. And then it does that, like it basically scans this entire scene so we get a 3D model. And those are systems that you can actually buy or rent. So our team at JPL actually went out to the cave, and these are the people that did this. We had a lot of volunteers because there's a bunch of heavy equipment that we had to carry out into the desert until we reach the cave. This scan is a system that we can use to get this three-dimensional model of the cave. Now, if you want to make a realistic computer model, besides the geometry of the cave, besides knowing where all these surfaces are, we also need to know how they reflect light. There are some surfaces that are very specular, like glass, that reflect light only in one direction, and some surfaces that are very diffuse, like these walls, for example, in this room, or most walls are actually painted to be diffuse because it's kind of disturbing if they were specular. (laughs) And they reflect light in all directions. So in order to make a realistic model of the cave, we need to include that, and that actually becomes a little problematic. Down here, like this collapsed part of the cave, that is probably just lunar regolith, which is basically this material on the surface of the moon, and we know what that looks like. The Apollo missions actually took samples of those lunar regolith back, and that was actually measured. The reflectance properties of that lunar regolith, we have them on Earth. There are people that measured them. Some of them were at JPL in this paper. So we have data on that. The bigger problem is this stuff in here because nobody has ever looked at that. So how are we gonna know what its reflectance properties are? And what we think they are, they should be very close to unweathered basalt on Earth. So if you go to Hawaii, whether you have volcanic eruptions going on all the time and you go in a lava tube that just formed that didn't have time to oxidize and to weather yet, you should get a surface that's very close to what we would expect the surface of the tube on the moon to be. So what we're actually doing now is we're looking at papers to see whether we can find a reflectometry data, and if we can't find anything, we have to measure it. We have to just go and get a rock from Hawaii and measure the reflectance properties for our model. But it's basically just a guess of that's what it should look like in there. And basalt is actually, it's very interesting. Basalt is essentially glass with a bunch of impurities. So, if you get the cooling conditions right, you can actually get a very specular glassy surface. So maybe that's what it looks like in there, at least in some places. But that's a very interesting future challenge to see what the different materials will be. And then, with the simulation, once we have this set up, we can come up with all kinds of scenarios. We can start adding data. We don't have to capture all our data in one overpass. We could do it in several orbits. We can add data together from multiple orbits. And we can simulate all that with this model. Okay, so besides my part of simulating what we would be able to do and how much laser power and kind of estimating how much laser power do we need to do this, we have been doing other simulations. So the mission, the orbit design is actually something kind of challenging, as I already started mentioning, because you have to get very close to the moon. There was a NASA mission that was run for a while back that, or was run actually recently, that's called GRAIL, and they did gravimetric mapping of the moon. That had to get within 15 kilometers, I think, and the way that works is a satellite basically goes, gets close to the moon and measures the gravitational field and you can try to use that to image the inside of the moon. That wasn't high enough resolution to see the lava tubes properly, but it does give us a very good idea of what the gravitational field of the moon is so we can make our orbit even a little more challenging than theirs. And we think that we can go down to about five kilometers over some of those caves, not over all of them. As you can see, these orbits, these orbits are highly elliptical here. You're close to the moon on one side, and then you get very far away on the other side. So we have to basically design a trajectory that goes over as many of these caves as possible, very closely, so we can image all of them before we crash eventually. And so the spacecraft itself would basically-- And remember, I told you earlier that we are only over the cave for 100 milliseconds. So the entire laser is only on for a tiny fraction of a second, and then it has time to recharge. And that brings us to a different thing, a different component is-- Oh, here's actually an image, a sketch of all the different skylights that have been discovered that could be imaged. But you can't possibly come up with a trajectory that would go over all of them in one mission, I think. Maybe they can. I can't. (laughs) So the other question, of course, is batteries. The laser that I've been talking about, I said it's one kilowatt at least, probably 10 kilowatt of optical light that comes out of the laser. The one kilowatt laser is something that you can buy. I was surprised. You can actually buy one. You have to have spent close to a million dollars, but you can actually buy that. And it's getting space certified so it's a rugged system. It's a semiconductor seed with a slab amplifier, if anybody's interested. So you can buy these, but a one-kilowatt laser, you can use it to drill a hole in the wall. So that's a powerful system. And it's down to a picosecond of time resolution, so more than we need. It's getting space certified, so that's already ready. Since our goal for this NIAC study is really a concept, it's something that we can do 10 years from now. So the idea is we do a concept now, we simulate everything, we see whether this mission is possible, and then if NASA decides to pick it up, they'll actually develop it further. Hopefully I'll be involved in that (laughs). And then 10 years from now, it will be flown as a reconnaissance mission. And one problem that you, of course, have is how do you power it. If the laser needs even just one kilowatt, the efficiency of the laser itself is maybe 15%. That means you need a lot more than one kilowatt to actually power it. And if you wanna do 10 kilowatts, it's even worse. But since we only need that for 0.1 seconds and our orbit actually takes several hours to recharge, the energy you need over that entire time actually is not more than what would fit in a AA battery. (audience laughs) Right? So, basically, if you take the AA battery and charge a super compacitor over several hours and then dump all that energy over 0.1 seconds into the laser, you could get enough light. I don't think we're actually gonna put AA batteries on there. (audience laughs) I think that was kind of a, kind of daring estimate, but it's not... The storage itself is not such a challenging problem. You can store all the power and then, and then, like, basically charge a supercompacitor and release all of that. With that, I think I'd like to, yeah, conclude the talk. So this is, as I said, it's a mission study. Hopefully we'll be able to fly this in, with the funding, in a year or two from now. With this funding project that we're working on, the goal is to know two years from now how well it will work and if it's possible and then recommend it and develop it further. I am definitely gonna work on these systems for caves on Earth. And then, if we're lucky, this will actually be picked up, and at some point, we're gonna launch in the future. And I would like to... Everybody closes their talks with some eye candy, some nice pictures. I actually have some actual picture candy. So this is a piece of rock candy. It's five centimeters wide, like two inches. And this was actually the smallest object that we ever used with our trillion frame per second camera. Just making a video of this is what it looks like when you hit it with a laser pulse. So we just hit it from the side with a femtosecond laser pulse and captured a video, and this is what it looks like. And then I don't know why it glows so long. We have been thinking there may be some fluorescence in the eye candy. Okay, with that, I'd like to acknowledge my collaborators here. Jeff Nosanov is actually, he used to work at JPL. He now has his own consulting company in Washington, DC, but he's the main PI on this project. And Karl is the lead at JPL. He's a geologist. He knows a lot about how these caves formed and how they work. And then also, I like thanking too Penny Boston, who's an expert in caves on Earth, and I guess everyone in the solar system who helped us getting into these caves in New Mexico and starting to image them. Okay, so thank you very much for your attention and for coming. (audience applauds)