Document 17865275

advertisement

>> Desney Tan: I don't get to introduce astrophysicists very often, so this is pretty awesome, but it's my pleasure to introduce Marla Geha who is faculty in astronomy and physics at Yale

University. When I first met Marla a couple of years ago at one of the [indiscernible] events she was doing her sabbaticals. It was a year-long sabbatical and she was traveling around the world to all of the different observatories and spending time and sending pictures, and I'm sure it was an amazingly productive year, but it just looked like a wonderful trip looking at it from the outside, so I was really jealous. If you look Marla up on the web, again, I don't get to introduce astrophysicists all that much. If you look her up on the web you see words like cosmology and dark matter and Milky Way structure and dwarf galaxies and she has done some wonderful work in the area, tons of awards to back her. Popular Science's top ten brilliant minds in 2009; is that right?

>> Marla Geha: Just that year. Last year was terrible.

>> Desney Tan: Hubble Fellow, Sloan Fellow, Kavli Fellow and the list goes on and on, but rather than stand in the way of us and a stellar talk, let me hand it over to Marla.

>> Marla Geha: Awesome. Thank you. [applause]. It's a real delight to be able to talk to you today. I appreciate the opportunity. What I want to talk about today is my research looking at little galaxies around the Milky Way and trying to use them as tools to understand dark matter, to understand cosmology and the tools that we actually use our essentially computer science. I had a conversation this morning; I was staying on Bainbridge Island with friends. They have a seven-year-old and the parents were trying to get the seven-year-old into coding. They had tried the Hour of Code. He looks at me and he says do you computer program? And I was stunned because that's essentially all we do. Astronomers, unlike biology or chemistry, we are not able to go and ask, take two galaxies and throw them together and ask what happens. We just have light. We study light. All my tools are in a computer, so if I want to take two galaxies and throw them together I can do numerical simulation and so all of my tools are computer programs. We are terrible computer programmers. We're awful. We don't take computer science classes as a rule; it's mostly physics classes and so I have a problem that I want to do and I just go and solve it, so that usually becomes very inefficient, but we're okay. The things that I want to chat about today are the satellites around the Milky Way. The image that I'm showing here is not actually the Milky Way. This is our nearest neighbor, probably a galaxy that looks very much like the Milky Way. This is Andromeda. You can, in fact, see it with the naked eye. It's one of the only galaxies you can really see with the naked eye. All of the dots that you see are actually foreground stars in the Milky Way, and this will actually become important to this story that I'm going to tell. We're in the Milky Way. We're looking through the disk of our

Milky Way and so those are all these little stars that are kind of annoying in front of us, and then looking out to a galaxy nearby, the Andromeda galaxy. But this probably looks a lot like our own Milky Way and it's a pretty picture so I have it up here. What I am most interested in in this picture are actually these two things right here, so right here and right there. Those are galaxies of their own right. They are the stars in these two galaxies are bound to each other so it's a conglomeration, a group, and its orbiting around the larger object, so these are two satellite galaxies. We call them dwarf galaxies. A dwarf galaxy is defined as something roughly

a 10th of the Milky Way's size or less. The objects I'm going to talk about today, more or less, is one 10,000th of the Milky Way's mass, so very, very tiny galaxies. But they turn out to be incredibly useful tools in terms of trying to understand the underlying cosmology of the universe as well as tools to studying what dark matter is, so they are kind of fun objects to study because I can talk about basically everything in astronomy from stars up to the underlying properties of the universe. To explain why satellite galaxies are so important, we have to go to the very beginning of the universe. I'm going to show a movie in a second which starts just after the Big Bang and just allows matter in the universe to congeal and emerge to form the

Milky Way that we see today. This simulation, actually, can we dim the lights? Thank you. This is a numerical simulation. What you see starting at the very beginning of the universe, there's over densities of matter. Those over densities just from gravity coalesce and merge. Smaller objects are merging to form larger objects. At the center you'll see what becomes the Milky

Way. There's a major merger happening right there, but you will just notice all the structure.

There's just tons of smaller things merging and we call this a hierarchical merging. The underlying physics of the universe is encoded in this simulation, so if the properties of dark matter were slightly different, if the topology of the universe was slightly different, the structure that you would see would be changed just a bit. There's one hitch and that is that this simulation is entirely dark matter. There are no stars. There's no gas. There are certainly no people in this simulation and that kind of sounds like a bit of an oversight, except that we know dark matter, the matter in the simulation makes up something like 90 percent of the universe and so we can ignore the other 10 percent pretty confidently. These simulations really represent what we see in the universe, so the image that I showed you just before, this beautiful galaxy with all of these stars lives kind of right in the center of this and it is surrounded by dark matter. It is surrounded by all of these dark matter structures. Here is the problem. This simulation predicts around a galaxy like our Milky Way, that there should be something like a thousand satellites, that is a thousand of these little structures around our

Milky Way. As of 2005 we knew of 11 and in 2005 -- 11 and thousand are different numbers even in astronomy. In 2005 I would have convinced you and I would have been really, really convincing that we knew of all of the satellites around the Milky Way. We had looked really deeply across the entire sky. Maybe there was one or two that we missed, but this was the population of satellites. There was a crisis really, smartly named the missing satellite problem, and the thing is is that if we know of all the satellites around the Milky Way, yet we predict thousands, what's wrong? Either the underlying cosmology of the universe is wrong, which is really disturbing because for big galaxies like the Milky Way and larger these simulations actually represent what we see. There might be a problem with just our understanding of the universe. The alternative is maybe we have missed some of the satellites, but again, in 2005 I would have said no. Most of the research was trying to understand how maybe you could have dark matter satellites that didn't have stars, that didn't have luminous matter and so we, these dark matter structures are there. We just don't see them and only occasionally there's stars that are in these things and so basically trying to suppress luminous matter in these things. It wasn't a very satisfying solution, but that's where most of the research happened. In 2005 a couple of things happened. The first major mapping, digital mapping of the sky was released, so the Sloan Digital sky survey, and I'll talk about that a little bit, released the first homogeneous catalogs of stars across a very large region of sky. That allowed us to search for

satellites in a way that we hadn't before. As of 2014 we now know of 25 satellites around the

Milky Way. These were all found in the Sloan catalog and I'll tell you exactly how we found them. We now know that instead of 11, and I would have said 11 is definitely the number that we know, 25. I know that there should be a lot more out there because we haven't searched the full sky and we haven't searched the full depth of the sky and so, in fact, doing completeness calculations, and I'll tell you about those, we think there should be something like

400 luminous satellites around the Milky Way out there to be found which is pretty exciting.

Then I will leave you on the edge of your seats until the very end of the talk to ask whether these things solve this missing satellite problem or not. If these things are so important and they perhaps solve this problem, how did we miss these objects? This is what one of the satellites that was known before 2005 looks like, even with the lights not dim, you can kind of see it. It's an over density of stars. This is what one of the new satellites looks like, and I promise you that no amount of imagination, even though I'm sure you're going to try, can reveal the satellite that's in that image. All of the ones before 2005 were found by visual searches, just by looking at images and looking for these sort of smudges. Let's talk about why these new things were found. Here is, again, the same image; I have now reversed the colors for no particular reason. There are stars in this image that are associated together, but most of the things that you see are actually these foreground stars right in the disk of our Milky Way.

Imagine we're in the Milky Way. We are looking through a disk of stars and we're trying to find objects that are a little bit farther away. The problem is that most of the things you are seeing in this image are foreground objects and we need to get rid of them somehow. The way we're going to do that is to remove these Milky Way stars and we're going to do it understanding the physics of stars. Stars are formed out of gas that collapse and we have really good physical models to understand how bright a star is and what its color would be. For example, our Sun has a particular brightness and it’s kind of yellow. If it was a lower mass star it would be a bit more red. More massive stars are bluer, hotter. We can actually use this sequence of brightness and color to remove stars that we think are in the foreground. What we actually do algorithmically is we take that really pretty picture and we turn it into a bunch of dots in a catalog. For every dot in this catalog I have a brightness and I have a color. What I do is I filter out stars that are not in a particular window of color magnitude. This is what it looks like. So this is brightness and color and stars that are at a particular distance should line a very particular part of the space. If I take this image and I remove all the stars that don't lie in this filter, it gets rid of a lot of these foreground objects. You might, with a lot of imagination, start to see something right in the center, but I'll smoothed it so I can actually see something. What we do is we go through and we filter the catalogs and we look for over densities of stars at a particular distance. If I now take that same image and I circle all of the stars that are at the same distance, this is the galaxy, and again the stars are faint and you wouldn't pick them out just by eye. This is how all of these new galaxies have been found, by searching algorithmically these catalogs. Images show you where these new things were found. This is just an all sky map. The disk of the Milky Way is this blue thing. The disk of the Milky Way has lots of gas and dust. It's extremely irritating. We can't find things through it. We can't look particularly far through it. So this is kind of a zone of avoidance. The green are the satellites that we knew of before 2005. If you've ever been in the southern hemisphere and seeing the Magellanic clouds, which are pretty awesome, these are the two Magellanic clouds here. These are where the

new galaxies have all been found and you'll notice these are not ingeniously distributed on the sky. And the reason is is that this is the coverage of the Sloan Digital Sky Survey. This amazing mapping of the sky is in the northern hemisphere. It's the telescope that is in New Mexico. It has only map the northern part. Hopefully, in the next couple of years we'll have maps of the southern part. This is when I say we found 25 satellites. We know they're going to be more just by mapping the rest of the sky. We also know they're going to be more because the Sloan

Digital Sky Survey only goes so deep. It goes down to a magnitude that, sorry, down to a brightness limit that is good but it's not great. What we need is surveys that go even deeper to find galaxies that are beyond the distances of these objects. This is the Sloan Digital Sky Survey, again, in New Mexico. It was the first systematic digital mapping, so it really allowed us to apply these techniques. The hope is that the next generation of surveys, the Large Synoptic

Survey Telescope is a big initiative in astronomy right now. It's one of the projects that the

National Science Foundation is really supporting. It looks like it might actually go. They have just broken ground, actually, for this telescope. It's going to be in the southern hemisphere and the first light is predicted for maybe 2020 or 2022. The Sloan Survey did a digital map of the sky; LSST is going to do a digital movie, so it's going to do one Sloan Survey every three nights.

If you stack all those together we're going to go much, much deeper and so in terms of data, we're absolutely terrified of this survey because it's just going to be terabytes upon terabytes per hour and trying to search this sort of thing we need to be thinking about it now. There's a lot of people; I'm on one of the working teams trying to figure out how the heck we're actually going to be able to deal with all of these data and do the science that we want to do now for the fire hose is turned on in less than ten years. So what can we actually expect with LSST?

This is a plot. Astronomers have really goofy units, terrible units. So bright things are over there at smaller numbers and faint things are down here. This is a logarithmic scale of brightness called magnitude. It's horrible. And this is the number, the cumulative number of satellites, so right now we know of 25 satellites. That's the red. The green if we just do a full survey with the Sloan Digital Sky Survey, we would predict just based on the incompleteness, something like 50 satellites around the Milky Way. And then if we actually do really deep surveys, something like LSST, somewhere between 100 and 600 because we don't really know exactly how to do that completeness correction, but it should be something like 600 satellites around the Milky Way. Now with these satellites we can actually do science. The least luminous of these objects, the faintest of these objects we can do a lot of really interesting physics with. The first thing that we've already talked about is we can actually test cosmology by comparing the number of satellites and we see to the numerical simulations. We'll come back to that and see whether or not those two things add up. These Lewis luminosity galaxies are also really interesting. They're the least luminous galaxies that we know of and they also turn out to be the most dark matter dominated, so I will explain exactly why I think these objects are dark matter dominated. Then we can use them to study dark matter itself. We can use these as particle physics probes to actually try to detect dark matter and it's really wonderful that these small, little satellites I can talk to an enormous range of people just using these objects. Let's talk about dark matter for a second and why I think these objects are dark matter dominated. We'll come back to this image and we've now identified a galaxy or a candidate galaxy where we have stars we think are associated together, but this could just be a random collection of stars. This might be some kind of statistical fluctuation where we just

happen to see stars that are aligned up, you know, and asterism. That's one question is how can I tell that this isn't a random collection of stars. If it's not a random collection of stars what

I want to understand, the question I really want to ask is what is the mass of this object. Just from Newton's law of gravity we know that a more massive object will have, those stars will be orbiting around each other faster, so the Milky Way stars are orbiting around at about 200 kilometers per second. If it was a smaller galaxy, those things would be flung out just based on gravity. And so if I can measure the relative velocities of stars, I can actually directly turn that into the mass of the galaxy. Let's do that. The problem is that to measure the velocity of the stars, these things are super faint and so I need a particularly big telescope. The biggest telescope in the world is the one I'm going to choose. The biggest optical telescope in the world is the Kech telescopes. These are in Hawaii on the top of Mauna Kea. The Keck telescopes are ten meter mirrors so the primary mirror wouldn't actually fit in this room by a long shot. It's run by the University of California and Caltech. Yale has just bought into the Keck telescopes, so we have something like 20 nights per year on the telescope. We internally write a proposal to ask for time. I actually organize that and run the committee to decide who gets time on the telescopes, so I get time on the telescopes. There's no, I walked out of the room when that decision is being made. [laughter]. We use the Keck telescopes to measure velocities of stars in these really faint galaxies. We measure individual stars to measure their velocity. Let me take a moment to talk about how we measure velocities and many of you may know this already, but I kind of wanted to talk about it. This is the spectrum of a star. If you take, let's say sunlight and you turn it into, put a prism in front of it; you turn it into a rainbow.

If you look really carefully even at the sun spectrum, you see dark lines in front of it, and what's happening here is you have the sun itself. It's very, very hot. It's acting like a black body and there's atmosphere, the sun's atmosphere is kind of on top of the cooler stuff, and so as the light comes through the atmosphere of the sun, there's absorption from various different chemical elements. Hydrogen, helium, all of those sorts of things and you can see a very clear footprint of these absorption lines. If the star is moving, these elements just shift just slightly and just like the Doppler shift, we can measure the shift of these lines and then infer a velocity.

It's essentially the Doppler shift. This is the pretty version of that. This is what the real data actually look like, so with the Keck telescope we can measure something like 100 stars, with a particular instrument we can measure something like 100 stars at a time. What you are looking at here is each of these vertical lines is the spectrum of a star in that galaxy. The lines across it are actually lines from the atmosphere. The data reduction part of this is, again, me writing a bunch of awful code to figure out what the wavelength is and then infer the velocities from this. This is where most of my PhD work was done and I love this kind of stuff. I have these stars. I then measure their velocities, and this is what the data look like from that. Here is the velocity of a particular star that I measured. This is just a histogram. Here I'm going to plot for those stars, their brightness their colors. First, we have stars here and I have a model for what the Milky Way should look like. Stars in the Milky Way, because the stars are near us and they are moving with us, they should have a relative velocity to the sun of zero, but they'll be a big dispersion around that. These are stars that are all associated with the Milky Way. You'll also notice here if you can, it's kind of hard to see, the blue stars here are all over the place. They're kind of scattered in brightness and color. Then you'll notice that there's a peak at a different velocity, a very different velocity than the sun’s velocity. These are stars that we associate with

this particular galaxy. You'll note that those red stars are all in a fairly localized region in this color of magnitude diagram space. If I over plot a model for stars at a particular distance, stars at a single distance should lie on these tracks. It's kind of consistent. There's lots of noise and stuff, but these red stars are, in fact, associated with each other. They all have kind of the same systematic velocity and they're orbiting around the Milky Way. Now what I can do is ask given what is the width of that histogram, what is the width of the stars that are associated together, and turn that into a mass. So I can first predict what I would expect that dispersion would be, and what I can do is I can ask -- okay. I know all the stars in this galaxy. I know what the mass of a star is. I'm just going to add up the mass of all the stars and just say okay. Given that mass, what would that dispersion of this histogram be? If I do that, if the mass was just from everything I can see, everything luminous, the measurement would be half a kilometer a second. Let's compare this to what I actually measure. I actually measure something like three or so plus or minus one. I can talk about that error bar forever. That took me a year to calculate. I'm very confident in that error bar the reason it took me a year to calculate is because now I can say with real confidence that this dispersion is much, much larger than what

I expect from just the stars. I've been conservative when I figured out what this upper limit was. I included stars that I might not see that are too faint. I included some gas. I included some dust. This is a real upper limit and so I am measuring something that is much, much larger than what I infer from what I can see. Therefore, the mass is dominated by something that I cannot see and we call that dark matter at this point. Dark matter at this point is just stuff that we can't see, gravitational mass. In fact, all of the newly discovered galaxies, these small galaxies are not only dominated by dark matter, their stars make up something like a percent or less of the total inferred mass. That is, there really are the stars are really the tip of mass iceberg. And then the ultra faint, these faint galaxies are the most dark matter. These are the extreme things were the star’s gravitational mass have ratios that are 1000 to 2000.

Compare that to, let's say, the Milky Way, this image that I showed you earlier. The Milky

Way's mass is least within the sun’s radius are a little farther out, the stars make up something like 10 percent of the total mass, so it's not as dark matter dominated as the 1 percent or .1 percent of these lower mass things are dominated by. Really, velocity measured from basically every type of galaxy we inferred dark matter. And I think to me that's empirically the strongest evidence for dark matter this doesn't give us any indication for what this matter actually is. In fact, at this point, this is what dark matter is. Dark matter is dark. We can't see it. And not dark matter is stuff that we can actually see. To understand what dark matter actually is we need to do a lot of other experiments, and so we can just ask well what could this dark matter stuff actually be, write couple of columns and start crossing stuff out there I apologize. I couldn't think of any other word besides stuff. So dark matter can either be normal stuff or it could be other stuff. Under normal stuff I can write down everything that is sort of normal is made out of the electrons and protons that we might not be able to see, so things like dead stars, so stars that have run out of hydrogen and they are no longer producing their own luminosity. We wouldn't see them, but they would provide mass. Planets, they could just be free-floating planets out there and there certainly are. We wouldn't see luminosity from those things. Black holes might not, you might not intuitively put that under normal stuff, however, the only way we know how to make a black hole is from a star, from a supernova of a normal star. Therefore, I actually put this under normal matter because it's coming from the normal

matter part. You might say, okay. What's other stuff? Other stuff is a little bit more exotic, something like a neutrino particle that has mass or a particle that has mass that we don't really understand. We sweep this all under the rug and we call this a weakly interacting massive particle. So a particle that has gravity but only interacts doesn't interact electromagnetic. In this case it only interacts using the weak force. If I had three more hours and I know I don't, I could explain all of the observations that suggest that none of these things make up enough mass. There certainly are dead stars. Are certainly are black holes, but they don't even come close to making up the mass we need to explain galaxy dynamics. If I had another two hours we could get rid of neutrinos, and there's a bunch of interesting experiments to suggest that neutrinos actually don't work from a couple of different points of view. And maybe in a little, there's ways to make neutrinos work. In the last couple of weeks, actually, I was reading some papers that neutrinos me be making a comeback, whatever. The most popular idea right now for why dark matter is a weakly interacting particle, and there are actually very interesting other lines of evidence that suggests this sort of thing. So these WIMPs, and there's a great story as to why they're WIMPs, but I'm not going to tell you. We can actually use the galaxies that we just found, these ultra faint galaxies to study and test whether or not the dark matter in them is made out of WIMPs. If dark matter is made out of a WIMP, it's actually predicted occasionally that dark matter particles will collide and annihilate and actually make something observable, make a real photon. In this case dark matter is not strictly dark. Occasionally, to dark matter particles will collide and interact and they will actually create a very high-energy photon, a gamma ray photon. The particle physics in this is actually something I don't fully understand, but most of the preferred models for this right now for this particle interaction predict a gamma ray. What we need to be able to see this, this is a very, very not probable occurrence. It's extremely low probability so we need a place where there's a lot of dark matter. It's very, very dense, therefore the probability for interaction is higher, and it also needs to be nearby because it's a photon and so it'll be 1 over R squared the signal, so we need to have something that's very dark matter dense and nearby. The best place to do that is actually the middle of our Milky Way. The center of our Milky Way should have lots of dark matter and it should be very, very dense, and it's extremely nearby. It's actually closer than any of the dwarf galaxies that I showed you. The problem is that in the center of the galaxy is every astrophysical process that creates gamma rays, so if I see a gamma ray from the Milky Way, from the center of the Milky Way, it could be created from a supernova. It could be created from matter falling into the center of the black hole, the black hole that's in the center of the

Milky Way, so I can't cleanly say if I see a gamma ray that it's coming from dark matter annihilation. It could be coming from normal astrophysical processes. The dwarf galaxies are awesome because they don't have any of these astrophysical processes, so there are no, there's no ongoing star formation. There's no processes that create gamma rays from normal processes in these dwarf galaxies, so if I see gamma rays coming from these, I can say this must be coming from annihilation. To look for gamma rays, I can't do it from the ground, thankfully, otherwise, we would all be dead. You have to have a satellite. You have to be above the atmosphere. The Fermi Gamma Ray Satellite has been flying now for about three or four years and it's been doing all sky gamma ray surveys of the sky. There's been no detections of the dwarf galaxies. There's been no, if we look at all the different dwarfs, we found that there is no over densities of gamma rays, but we can actually stack them together and get some kind of

upper limit. So the fact that we don't see them actually puts some limits on this annihilation process and so for the aficionados in the room, we can actually rule out some dark matter particle models just by the fact that we don't see these dwarf galaxies. Deeper gamma ray surveys will actually be able to get into ranges that are really interesting in terms of what we expect to see. Coming back around, we've now used the dwarf galaxies to test particle physics.

We can now test cosmology as well. I've said that the number of dwarf galaxies around the

Milky Way is a test for cosmology. The number of predictions of the satellites is directly related to the underlying physics of the universe. Let's actually see whether dwarf galaxies work or not. Here is, this is now in a different units. The mass of a satellite galaxy and this is a number, the cumulative number of galaxies. The open symbols are the predictions, so that's actually taking that numerical simulation, asking what the mass is of all of the little things around it and just adding them. So that's the predictions and this is where I came up with this idea that there's about, there should be about a thousand satellites around the Milky Way. These are the numbers that we knew of before, so only about 11 satellites in 2005. So those were the observations. These are where the observations stand now assuming that we have searched the entire sky, so this is 25 to 50 number of satellites. And you can see that it definitely is getting closer, but there are still factors from 2 to 50 between this and so we can now ask look.

If I look really deep, can I actually get up to the numbers that were predicted? And for the smaller satellites, for the lowest mass things, I think we can. These are pretty generous predictions and I think we can actually get up to the numbers that we predict. The lowest luminosity satellites have actually, at least eased this missing satellite problem at the lowest mass end, and so here, you know, if it's a factor of a few, we actually feel that things are going pretty well. What's interesting and this will be sort of one of the last points in the talk is that we now, instead of having a problem at the lowest mass end where we were having problems of thousands, we're now actually very concerned with the brighter satellites, the things that we knew of since 2005 and earlier, that I know around the Milky Way that there are no more very massive bright satellites. I'm quite certain of that still, and so we're talking here of factors of ten below the simulation still. So the question here is well maybe the Milky Way, it's a sample of one. Maybe the Milky Way is just a weird, you know, outlier in this distribution and that we put a lot of weight into the Milky Way. We should probably go and look for other Milky Ways and ask what the distributions are around those satellites. And so this is something, a new project that I've just started in the last six months. This is with a collaborator at Stanford.

We're really excited about it. It's a completely insane project. Let me explain why it's insane.

Here is an analog to the Milky Way. We can find these all over the place. A Milky Way analog is something that is probably the same luminosity, maybe roughly the same mass as the Milky

Way. What we want to do is find satellites around these things. The question is, is the Milky

Way an outlier and what are the satellite populations around other galaxies. We can't get down to the incredibly faint, ultra-faint things that I showed you earlier around other galaxies.

They're beyond our capabilities, but we can get down pretty far and those two data points that

I showed you earlier we certainly can find satellites around those. Here is the problem. Here is the Milky Way. This is far away, but not all that far away. Many of my colleagues would consider this the local universe. That's a blowup of this Milky Way analog. It's kind of looks nice. We've got spiral arms. It looks very Milky Way like. We have gone and gotten spectroscopy, so we've actually gotten distance measures for a handful of objects in this region.

Actually, can we dim the lights? Does that work? While. Sorry. [laughter]. That's what I'm going to go home and tell my colleagues about [laughter]. This is a satellite galaxy. This is, again, coming from the Digital Sloan Survey. It doesn't look like much. And here is the problem. This is a bunch of background galaxies in the same region and they look pretty darn similar. The problem is that we expect something like four or five satellite galaxies, so 4 or 5 galaxies associated with that large thing. In this same region there's something like 5000 to

8000 objects that have the same colors, the same properties as the satellites. There's two things I can do to solve this problem. I can go and get spectroscopy. I can actually measure distances for absolutely everything in this field. That's a fairly expensive proposition, where expensive means I need to go to telescopes and spend nights, weeks of nights just getting spectroscopy for all of these. It's actually what we're doing right now, so in the next six months we have about 15 nights of telescope time to sort of line we get distances for everything in a couple of fields around a couple of Milky Ways. The reason we're doing this is because we're trying to build up training sets. The idea is that if I look at this image and I look at these images, there is something a little bit different. I can't actually put my finger on it, but when I see a satellite galaxy, I kind of know it looks like a satellite galaxy. If I look at that catalog properties, the sizes, the colors and I compare that to these things they're identical. I can't really break the problem there, but I do think that somehow my brain knows that there's something different and so what we want to do is try and build up training sets and then throw these into some kind of machine learning algorithm. So we're trying this now. Yeah?

>>: Does your instrument have any spatial resolution for this the spectroscopy?

>> Marla Geha: Right now we're just getting -- no is the answer. We're trying to get, we're doing fiber spectroscopy trying to get really course, really low signal-to-noise, this awful spectra, to simply measure velocities. It's expensive to get spatial, but once we find things, we're following up with spatial resolution. The hope is that if I can teach a machine what that looks like, and we need to build up samples to do that, perhaps we can at least get rid of, instead of having 5000 galaxies per Milky Way, we can get down to something that's more reasonable, maybe a few hundred and actually be able to build up a statistical sample of Milky

Ways. This has been really fun. We have so far found maybe 20 new satellites or something.

Twenty is not great in terms of trying to teach a machine anything, and so we're trying to come up with some clever ideas of how to do this, but it's been really fun, and for me it's been a real learning experience in terms of using new tools to try and get at some old problems. I went really fast, but that's cool. We have more time for questions, so I'm going to stop there. I have hopefully convinced you of a few things. First of all, it's just interesting that the number of satellites around our Milky Way has doubled just in the last, you know, five to eight years and there are new discoveries happening all of the time. Last week there was a paper, an archive, the first satellite discovered outside of the Sloan Digital Sky Survey, which is pretty awesome.

There are new imaging surveys that are coming out in sort of small patches of the sky. It won't be until LLST where we can really map the entire sky and get a sense for the entire population around the Milky Way. But all of these new things are interesting. The least luminous galaxy, they've actually had us questioning, what is a galaxy. With luminosities that are only multiples of the sun, we really have to think about what we define as a galaxy. They're the most dark

matter galaxies and so, the dark matter dominated galaxies, which means they're really great tools for studying this crazy thing called dark matter. We can test. We can use them to test cosmological models, so many of my colleagues who have basically ignored dwarf galaxies for many, many years because they are small and tiny. They don't make up much mass in the universe, have all sorts of like a herd started simulating things with dwarf galaxies because they're such cool crucial tests. There are hundreds of Milky Way satellites to be found and beyond the Milky Way there are lots of satellites to be found and around, things beyond, so in the next ten years we are really going to understand both the population of dwarf galaxies themselves, and then be able to use that to understand galaxy information and sort of underlying properties of the universe. Thanks. [applause]

>>: Are there some local galaxies that might be obscured by the disk or the center?

>> Marla Geha: It's a small percentage, so yes. Absolutely

>>: What will account for that gap?

>> Marla Geha: Yeah. So within about ten degrees of the disk of the Milky Way all bets are off.

You can go to different wavelengths, so that's in the optical. In the infrared we can pierce through the dust of the disk a little bit better and get down a little bit lower. Yes, the -- that's actually pretty easy to account for in my incompleteness, and so I know where I can search for things and so I can just account for that. But there are interesting ways to try and get around it.

>>: Is the proportion of the dark matter in satellite galaxies reasonably consistent?

>> Marla Geha: That's really interesting. So the results right now which I sort of glossed over, are that all of the low mass galaxies, no matter what their luminosities, so over three or four orders of magnitude in luminosity, they are all it seems sitting in the same mass of dark matter.

That is, so the most luminous ones have a certain amount of dark matter, but the least luminous ones have the same amount, which is actually a little bit weird and not quite up to predictions.

>>: Is it possible then that there are large, massive satellite galaxies that we can't see because there is such a high proportion of dark matter, or such a low proportion of luminous matter?

>> Marla Geha: Yes. It's possible. It's possible. If those stars are spread out, we have a detection threshold essentially. We have to see this over density and so if they are spread out a little bit larger, we won't see them. So really low surface brightness things we just don't see.

Again, with LSST as we go deeper we should be able to push those limits a lot more. Yes?

>>: Maybe this is more of a basic question. What is holding all those satellite or dwarf galaxies together? Is it the dark matter, otherwise…?

>> Marla Geha: Yeah, so it's all gravity and so they are being held together. Therefore, there's matter there.

>>: So that yeah, there's both light matter…

>> Marla Geha: So the light matter actually, so the luminous matter actually doesn't play a part in the story. Those numerical simulations that I showed were just, each particle has a mass and the, you can actually calculate like a tidal radius. So at what radius away from the center of a dwarf galaxy does a star filled gravity more from the object itself or from the outside field from the Milky Way? The radii of that radius where the transition happens is actually well beyond the luminous matter, so the stars in a dwarf galaxy only feel gravity basically from each other.

It's a much larger radius that it transitions into starting to feel the title field from the Milky Way.

We certainly see things like globular clusters, clusters of stars that don't have dark matter.

Similar luminosities, we see these things been a part and I have, another project. I love to talk about like three different projects in here. One of my graduate students is working on trying to use these clusters which are being torn apart in the gravitational field of the Milky Way using the stream and actually inferring the gravitational field of the Milky Way. So we're using those two weigh the Milky Way, which is a really cool way of trying to get at the Milky Way’s mass.

But we certainly see things been tidally shrift and shredded. In those movies you can kind of tell. There's, it's a chaotic mess is galaxy formation, which is like a different problem.

>>: So there's no radio signal at all or anything else?

>> Marla Geha: Radio waves are, how are those produced? It's usually in active regions where there is supernova going on or something like that.

>>: It's not just I itself? It just happens to be a byproduct like you said before? It just happens to be in there?

>> Marla Geha: Not yet. I have a pet project that maybe I'll talk to you about after, but basically there's nothing except for the stars and those are brightest in sort of luminosity and in the optical and I are [phonetic].

>>: So the fact that you are not seeing gamma rays coming off of the WIMP interactions, that must be influencing the models for WIMPs that…

>> Marla Geha: Yeah. There's a preferred model right now and the limits are still slightly above those preferred models, so we haven't yet made anyone extremely uncomfortable, but within a year we're going to make people extremely uncomfortable. And that's happening. It's really interesting because it's also happening from the direct dark matter detection, so that's in direct from annihilation. The direct dark matter detection are in labs waiting for a dark matter particle to knock like say a xenon atom or an atom that doesn't interact with things usually.

Those limits are also getting really interesting and have actually ruled out some of the preferred

models for what dark matter is. It will be interesting in the next couple of years to see how this field starts changing as the preferred models are ruled out.

>>: You said that a lot of the dwarf galaxies that you find have similar mass over the dark matter? Does that strike you as unlikely?

>> Marla Geha: It is interesting. I am fairly confident that the observations are right, not 100 percent, but fairly confident. There are models that explain it. There's models that say look basically at this mass of dark matter, galaxy formation becomes extremely stochastic and so if you go higher, a higher mass dark matter mass, the luminosity that's in that object is always kind of the same. It's not all that stochastic. As you get to smaller and smaller dark matter masses, the amount of luminous mass gets stochastic and so that's why you get sort of the range of luminosity in that dark matter mass. Those models predicted there are lots of dark matter, smaller dark matter masses that don't have stars. So in this model that I showed you, there should be a lot of things that don't have stars in them. How do we find those? So that's a really interesting task. One idea is to go back to these streams and see whether or not there are holes punched through these streams by dark matter things. It's a bit controversial, but we're going to try it anyway, but trying to find things you can't see is really challenging and you have to be clever. But we do it all the time.

>>: Maybe one last question. If you took a step back and sort of hypothesized that things are there that could speak to science, where would those bottlenecks be right now? Is it in the measurements, in the processing? Is it in pure people time? What is it?

>> Marla Geha: I would say most of it right now is actually pure people time and clever coding.

I would like to, so something like LLST with getting a lot more data, that would be amazing and that would be, that will, that should break open this problem. That said, really clever coding, that's how the satellites were found and, in fact, the Sloan Digital Sky Survey was available four or five years before the first discovery and it was just us kind of, you know, processing and crunching through and trying to come up with our stupid clever ways. In fact, more interactions with people who think about interesting algorithms, there's a lot there. There's a whole lot there. Probably the people and the ideas is where things are a bottleneck.

>> Desney Tan: Any other questions? Thank you.

>> Marla Geha: Thank you. [applause]

Download