 Hello, my name is Jessica Smith. I'm the convener of the Astronomy Mini Conf. Piota, everybody, thank you. For those of you who weren't here for the morning sessions, I'll just run very quickly through some of the schedule stuff. We've already done that bit. So this afternoon between now and afternoon tea, Dr. JJ Eldridge, who is sitting just here, is gonna be talking about open modeling stars and galaxies. And then after that, we're gonna have Nicholas Erditi, who is not in the room at the moment. He'll be here shortly, talking about the SKA, hacking the Big Bang. We will then have afternoon tea. After that, obviously, I encourage you all to come back for the afternoon session, where we have three sessions talking about using data from the MOA project. We have three students of Dr. Rattenbury's talking about their work there, after which we will have a lightning talk session. So please have a think about what little five or 10-minute presentation you might be able to give related to your own astronomy work or hobby or photography or anything that you think is really cool. Give a demo or a piece of software. Anything that works for you works for us. The schedule is there if you wanna see the timing. That's linked from the conference website as well, obviously. Housekeeping stuff. We did have a few BP devices during the earlier sessions. Please switch your phones, tablets, other devices, laptop notifications, choose silent. Please dim the screen so that you're not blinding people behind you. If you have to take a call, please just discreetly exit until you have completed it. Think about the asking questions thing at the end for the protocol. I thought that was a really handy little graphic in yesterday's session. And we will be passing the microphone around for questions. There will be a short question time after each presentation. Because we are recording, we wanna capture your questions as well as the answers. So someone will come around pretty quickly with a microphone. Just hang on for that and we will be able to hear what you're asking. If you care to tweet Google+, possibly not Facebook, any other social medias, here are the hashtags or things related to LCA 2015 and Astrid Conf. And now on to Dr. JJ Eldridge's presentation on Open Modelling of Stars and Galaxies. Thank you. We have one screen up too. Great. Thank you very much for your introduction, Jessica. And thank you for giving me the opportunity to speak. So, really good speech talk this morning. I'm going to be going along in a similar vein, but more or less talking about basically almost in a similar story, my personal history and journey for astronomy and studying and how I've used different software packages, what is open, what you can use if you really want to go and play with something and how do you can model stars. And also maybe just give you an overview of astronomical data from different data sets because a lot of the data I use is actually archived data that already exists. So this work is not, of course, done in isolation. It never is as a scientist. There are my two very good collaborators who sort of like helped me with observations of binary stars and binary galaxies as Elizabeth Stanway and Monica Rolanio. If it wasn't for Elizabeth wanting to write a paper with me so we can stay in contact after our PhD days, I wouldn't be here giving this talk because she said, ah, you know, you've looked at the stars when they die, what about when they're alive? You know, in galaxies, they kind of have stars. And so the work she was working was very apt and I can make predictions on that. Then there are my three PhD students, Lynn Shaw and Liam McKellan, who are currently working at the University of Auckland and my one student who's now become a data scientist for, if you're from the UK, for Tesco's. That's Joe Wormswell. I'm along with some other observers who do the supernova stuff, Morgan, Justin, and Stephen and my PhD supervisor, Christopher Taut. I did this slide because I think it's interesting that you put his very unstandard pathway in. Mine is slightly more standard. I did a physics degree in Cambridge. I did an astrophysics PhD. Then did a postdoc in Paris, France, which, well, yeah. Postdoc in Belfast, which I absolutely loved because Belfast is a wonderful city. And then I went back to Cambridge to do another postdoc for four years. And then of course you're trying to find that permanent position. So I applied for lots of lectureships, ended up applying to a couple in Taiwan, being offered two in Taiwan and being offered this one here at the University of Auckland and was trying to make the decision between either moving to Taiwan or to Auckland, Auckland one. And I've been here for 38.5 months so far. Not that anyone's counting. Of course, though, it's important to realize that even though we're scientists and we're standing up here and saying, look, these are the studies and these are the things we've done. You know, life goes on in the background. You know, I've got a black belt in Taekwondo. I've got married. I've got babies. All the other things that goes on in life trying to work out where you're going to. An important question that I love always answering is why do I do astrophysics? And well, now there are two answers, okay? So if you want the serious answer, I was born in the same month that Voyager 2 and Voyager 1 were launched. Okay, it was just only a few weeks before and after when I was born. And so when I was growing up, we had all these images coming back from the Voyager space probe of Jupiter to Saturn, Uranus and Neptune. Okay, so that really does inspire you to go and study space. But it's also really, let's face it, the year that Star Wars released, okay, 1977. And as I was growing up, you know, especially when I got to a certain age, when there was a certain doctor around, you end up getting hooked on science fiction. Okay, so if you actually go and read my thesis, which, you know, you can find at MastroPH on archive.org, you can read that I actually have a paragraph thanking all the science fiction authors and Doctor Who. Okay, so Sylvester McCoy was my doctor. Although, you know, when these guys were here in Auckland last year, I went and had a photo taken with all of them. Okay, hopefully not too much envy there. How do I do science? So why do I do astronomy? Why do I do astronomy the way I do? Well, Nick showed his first computer, so I thought it was only fair to show my first computer, which is the spectrum. I got this when I was five. And so whenever I applied my CV, applying for a job, I always changed the line saying so many years computer programming experience. Only one person has ever noticed that, right? Because they sort of saw my age, went, okay, he's like 30-odd something. Yeah, he's got 30-odd years of programming experience. You know, and so I put those two together, you know, and then you explain, look, I got this when I was five or six, and that's enough time to have 33 years programming experience already. Okay, because it's just like Nick, he had basic, and again, basic is probably the best programming language in the world. And with ZX Spectrum, you had to program the second you turn on. Did you get tired? Hmm? Sorry? Okay, sorry. Anyway, so let's get on to astronomy and let's start with a sense of scale because Nick's been talking about planets this morning. And the good thing about astrophysics in Auckland is we study stars, we study planets, we study the universe at large. Okay, so we actually study all the scales that you can get. So I stole this side from my friend Elizabeth. And you can see, so we have, we're here on the earth, okay, you're here in Auckland, in case you didn't know, okay? But this is Linux Australia, which is over here, okay? Or the West Island, as some of us like to call it. But, you know, we're in another state of Australia, which Australians probably say, so it's fair each way. You're in the side of solar system, so you have this sun and the sun, and then you've got all the planets. Okay, but the sun is just another star in the two to 300 billion stars in our galaxy. Okay, but then there are lots and lots of galaxies in the universe and as you zoom out, you can see there's even more galaxies. But then you know that most of the visible matter in the universe isn't what we actually see, there's actually this stuff here which is from a cosmological simulation of dark matter. Okay, so something like 95% of the universe, the matter in the universe is stuff we can't see. And that's when you go back to the cosmic microwave background radiation, which is, I think, going to be the topic of the next talk. So we're not going to talk about this, we're not talking about dark matter, we're going to be talking about this stuff, okay? The stuff we can actually see, because we don't really still even understand how stars work. We understand them on the general order of magnitude, physics, but when we try and get down to real details, it can be very, very difficult. So what do I do? Well, I use computer models, okay? So I've got all these stars in a galaxy. That's my model of a galaxy. Okay, there's a few more stars in my model galaxies than that. But you've got stars of different colors, you know, you've got red ones, yellow ones, white ones, blue ones, very dark blue ones, purple ones. I just like having green stars because it's a bit different, okay? And green stars in the universe because the bit of the spectrum is too narrow, so a star can't actually be green. But you can have green gas. But outside these stars, you know, a galaxy is not just stars, there's gas, which is what this gray band is meant to be. And so if you actually want to model a galaxy, not only have you got to model the stars, not only have you got to model the emission light coming from their surface, you've got to model how that light from those stars interacts with the other gas and dust in the galaxy. And then also, see if you're looking at a distant galaxy, not like our own, you've actually got to look at other gas clouds in between you and this other galaxy that don't have stars, but there's stuff there and you can actually pick that up. And so you've got to model all of this to try and understand the universe, okay? And so you need a code to model the structure and evolutions of stars. You need a code to model the atmosphere, which is what they look like, because that's different to the structure and evolution. And then you need to actually combine all these individual models to make a population of stars. You then need a code for the gas, okay? So that's to actually work out what this gas does when you shine a light on it, okay? It's like when you see a sunset and you see all the light coming through, refracted off clouds, it's basically a similar sort of physics, but it's a bit easier because there's not water making your life difficult. And that applies to the same code out here for the gas and the intergalactic medium space between galaxies. And then you want to actually end up for this output of stuff that you can then compare to observations, okay? So it's light comes from the stars, goes through gas, gets to this person, and you know, you go and take a picture, knows an extent, there's all this data we're getting and you want to be able to actually give them numbers, tables that people can compare to observations or do it yourself. So the codes are gonna go through and take it on a case by case basis and take you through some of the codes that we use and say whether they're open, whether they're not open. I know, and there's been research done, that if you make your codes feel available where everyone's download, your citation rate is higher than if you are very secretive and keep all your codes to yourself and make it difficult for people to get your models, okay? So I know at the moment I've made a mistake because while if somebody asks me for a code or a number, I'll give it to them. I haven't yet made this number three code down here available, but it's gonna be getting there soon. But we can take the first example of the STARS code which is called the Cambridge STARS code. One problem if this is, there's multiple different versions of it, okay? But there is a website, you can go and you can get the current Cambridge one but it's the person who created it, Peter Egleton, back in 1971. He actually wrote it in the 1960s. So it's written in Fortran, not Fortran 77, one even earlier. I like Fortran 77. But then you have to, so that tells you what's going on inside the star. I'll show you an example in a minute. But then that only gives you a luminosity and a temperature on the surface of the star and you want to actually know what they look like if you was to like take a picture of them or take a spectrum of the surface. And there are certain codes that lots of people have written because they're quite difficult to model. Some of the STARS that are not open. So one that I use here, which is quite unique is the Potsdam Wolfrett Spectra. That's not open. They'll give you the results for free even though you have to hack their website to get all the data off because it's a really annoying form. And if you want to get there like a few hundred spectra, you actually have to click on it each one, then go download. It's really annoying. So I got someone to actually write scripts to download them all. I haven't told them that yet. I hope they don't watch this talk. But there are some others that are open. So there's one called CMF Gen by John Hillier who's an Australian now living and working in the States, which you can get and you can hack yourself. Okay, so if there's some different physics you want to do, you can make it. Which is what's brilliant about open astronomy codes because it means you can take the code. If you don't like the physics or you want to find out what weird things could happen, if you change physics in your star or your atmosphere, then you can make a very quick, you don't have to deal with the basics and write a code from scratch. So the first science I'm going to try and illustrate some of these things with science as we go through is exploding stars. So this is cool stuff that was being done just as my thesis was being done in 2001. A guy called Stephen Smart had come to Cambridge on a postdoctoral fellowship and was doing pre-explosion imaging of supernovae. So supernovae go off, we now detect because of robotic searches in the same sort of searches that Nick talks about, searching for planets and there's similar ones searching for supernovae. So you just take a picture of a galaxy one night, compare it to a previous image and see if there's a suddenly more light coming from there. And so we detect thousands of these supernovae now a month. It's getting a bit difficult because we can't follow them all up. So you have to guess which ones are interesting. But very few happen nearby. Because nearby you actually stand a chance of seeing the star that exploded out to about 60 million light-years. And all Steve did and somebody else, Skylar Van Dyke in America, realized that if you see a supernova going off, if you go and look in the archives of this thing, the Hubble Space Telescope, of that galaxy, you stand a good chance of actually seeing the star that exploded. And it wasn't until about early 2000s that the archive was actually big enough, enough pictures have been taken of Hubble that you can actually stand a chance of doing this. Because you just have to have the data before the supernova explodes to actually try and pick something up. I'll show you an example of exactly how that's done in a second. But the Hubble Space Telescope is awesome because most of the data is now public. Some of the stuff is proprietary and so you have to wait a month for the person who wanted the data taken to actually have a look at it and publish if they want to, otherwise it then becomes open access. And actually the interesting thing is Steve was always getting the time on this to get the follow-up observations because you have to take a picture after the supernova to work out where it was to compare it to the pre-explosion image. But he was presumptu because somebody else said, look, we'll make the data free. And so someone else won the time, which is great for them. And they did it by this idea of saying, okay, we should let everyone, everyone's interested in the same data, everyone wants the same pre-explosions image. So that was a change in the way the science was done. So what we're talking about for supernovae, we're talking about these sort of things. So you see here, you have a spiral galaxy, much like our own, it's M51. Okay, I always get that confused. They're never sure if it's M51 or M52 and somebody should correct me if it's not. And so you see these things. Okay, so there's one supernovae and there's another. You know, this is actually unusual. We haven't seen a supernovae in our galaxy for, well, since 1604 because we're in the disk, so it's difficult to find them. But if we actually zoom in with these of the Hubble Space Telescope, in both of these cases, we were able to resolve the stellar that exploded. Not very well and I'll show you the best case in a second, but these are the sort of things we're looking at. Okay, and as you can see, this later image in 2011, that thing has faded away. So these supernovae only last for a few months, okay? And basically these things are creating all the elements in the universe heavier than iron. Okay, so all the gold in your rings and jewelry or any other heavy elements are coming from here. Okay, not the iron in your blood, that comes from a different type of supernovae that cosmologists will tell you about, I went. So how do I model these stars while I use this code, the Cambridge Stars code as I've said, there's many different codes. The really cool thing about stars code though is it's really, really tiny. It's only 5,000 lines of Fortran code. Is that a large amount of code or not? I don't know. Can somebody tell me in the audience or not? No? 5,000 lines? Okay, thank you. I just wanted to make sure of that. Because one of the codes that I'll talk about later has hundreds of thousands of lines of code and it's in different languages as well, because it's been written in Switzerland. So it's got German, French and English comments. But on that hand, they're actually much more detailed so they can do stuff that we can't with the Cambridge Stars code. But this is quite fast. So if you want to see what a star looks like when you run this code, it looks like this. Isn't that wonderful? Numbers. So this is actually towards the end of a model. We've got things here like the mass, the helium core mass, the composition in the center. And as you can see, you've got no hydrogen, helium, a little bit of carbon, no nitrogen. It's all basically oxygen and neon and it's getting towards core collapse. And the central temperature is something like, as I look over here, nearly a billion Kelvin, which is quite hot. That looks quite boring. So I then worked out how to make a movie. And so this is a model of a star exploding, sorry, a star evolving towards explosion. So actually here, you've got the outputs on the surface, which is why we need atmosphere codes, just telling you how hot it is and its temperature. And in here, you can actually see the interior structure. So already, the star has converted all its core into helium, so there's no hydrogen in the core and this is the hydrogen. And then in a second, you'll see the carbon and nitrogen come up because you'll start burning the helium to carbon and oxygen because in stars there's this progression. You've also gotten things near the density, so already the density in the center is approaching about a thousand grams per centimeter cubed. In case that's quite dense. The temperature is getting up to 100 million Kelvin, so you'll burn helium soon and the radius is growing and there's also the luminosity. So you can actually see at this point, most of the luminosity of the star is coming out from the shell here because this hydrogen at the edge of the helium core is being burnt. But what you're actually solving is you put all the physics into a code and you're just solving a bunch of nonlinear differential equations, which is quite nice. You don't have to worry about it, but it just gives you results. Here you can see all the helium in the core has been converted to carbon and oxygen. And in a second, you'll see some of this carbon being burnt and you'll get more oxygen and the star becoming red giant. But all the time, the core is becoming hotter. And when this model gets to the end, I can't actually get close to core collapse. You can see there it's now burned away some of the carbon and some of the oxygen. So now you're actually getting neon or magnesium and this is getting really close to core collapse. So this model, at that point, was a few hundred years maybe, so a very short time, astrophysically speaking, a human lifetime, before it would actually explode in the supernova. If you want the Fisher Price or the kindergarten version of stellar evolution, you've got a main sequence star that's all hydrogen and helium. And we saw that the helium core was formed in the middle and the star becomes a red giant, big puffy. We don't know why. I can say that here and you won't say, ah, but I learned in my stellar evolution course back at undergraduate days that we know why. We don't. We just put the numbers into a code and it becomes a red giant. We see red giants occur, but there's no simple reason that would take less than 4,000 lines of code to explain just by writing it out that tells you stars become red giants. But the evolution goes on, you then take your element helium and you convert it into the next stable, heavy elements and this goes on all the way into your form and ion core. And when you form an ion core, it's the most stable element in the universe. Ion isn't actually, it's one of the cobalt or nickel or nucleic, I can't remember, but it's these ion group elements, cobalt, nickel and ion that are really stable. You can't get out any more energy by nuclear fusion, smashing your nuclei together. And so what happens is that core collapses down to the size of a neutron star, becomes one big giant atomic nucleus. You release lots of neutrinos, which normally wouldn't interact with matter, but this stuff is so dense that the neutrinos transfer matter and the star explodes. And it was actually last year in 2014, the first simulation's actually started exploding stars because the physics is just evil trying to model supernovae. Stars are easy, supernovae are even more difficult. And as my really complicated simulation as I'll show you next, what happens next is a star explodes. So the envelope's thrown off, you're left with a neutron star or a black hole in the center that's really difficult to see, but that's what we've always thought was happening. And then in 1987, we actually saw a supernova go off on a nearby galaxy, the LMC, but didn't follow this rule as it was something different. But in 2003, we saw the first red supergiant exploding, but the best case we have for a red supergiant exploding is this one, because we have multicolour. So this is a Hubble Space Telescope image. So this is zoomed in on the galaxy. So this is actually similar data to the image I showed you of M51 earlier on, but it's zoomed right in and you can actually in a galaxy resolve individual stars, only a few light years or a fraction of a light year apart. So here you can actually see there, there's a red supergiant, it's all these other stars around. And so what you've actually done is you've taken this image where you've got the supernovae here, you identify these stars of these stars, these stars of these stars, these stars of these stars and you sort of like do a bit of trigonometry, you work out where the centre of the supernova is and you put a crosshair down and oh look, great, there's a star. Of course, you know, just saying that star is there, that's not proof it's this star that exploded. So a few years after that, you go and take another post explosion image a couple of years afterwards and you actually find out that that star is now disappeared. Okay, so what I've just told you is true. And down here, there is somewhere a neutron star or a black hole, but it's difficult to detect these things at this distance, so we don't know which is. It probably should be a neutron star. In case you're wondering why are these so blurry because these are Hubble Space Telescope images and this is so crisp in the centre, these are from the Hubble Space Telescope, two and a half meter telescopes and this is an eight meter ground-based telescope with the VLT in Chile in the Atacama Desert and it's just the bigger telescope gives you much higher resolution. So this star you think is one, you can actually resolve is actually two stars just very close together. Okay, so a lot of this can actually be done from the ground which is why it's such a growth industry. But you need the HST data because you can't do this sort of, these observations in the optical from the ground. Okay, so trying to model these is difficult and you know this is great for red supergiants, we always see something. Well, we know more massive stars don't die as red supergiants. If you go above about 20 times the mass of the sun, stars when they become red supergiants have really strong rings, okay? And you actually lose all this hydrogen envelope and you get this naked helium star. It's actually called a Wolf-Rayet star after the two. Well, it's one Frenchman and one German I think who discovered these stars because they've got really weird spectra which is why the codes tend to be quite secretive and not be made public. But the evolution in the center still goes the same way and so it still explodes. And so all we should have to do is just wait to see one of these supernovae because they don't have any hydrogen in. You can actually see they're different. So you can actually tell something about the star that exploded just by looking at what the ejector that comes off is made of. But you don't see anything, it's not a problem. Okay, because this is what we've seen when you look at about 13 of these supernovae. Now, one supernova missing something is not a problem, okay? And there probably was a star, no, sorry, there must have been a star there at that side of that supernova but it was just too faint to have been seen in that image. But when you start doing this 13 times you realize something about what you're predicting about these stars being really luminous is wrong because we do predict these stars to be luminous. So this is some work I did with Steve and his collaborators back in 2013 that, you know, here are these progenitive detections over here have read supergiants and other stars somewhere in the middle away from red supergiants, the yellow supergiants exploding but these hydrogen-free supernovae should be up here and these are really much more luminous, okay? Because this is the HR diagram so those of you who know astronomy will recognize it. This is temperature along the X axis so it's cool over here and it's hot over here and it's dim down here and it's luminous up here. So these stars are more luminous than red supergiants. So they should be really easy, I put in quotation marks indicating that it's really complicated but it should be easier to detect these things than red supergiants, but we didn't see anything. Well, that's because most people, whenever they consider stars and it still goes on even though myself and people like Danny Van Beveren and other people, Rob Bizarre and many others have been bashing away with the binary stick or ignoring, okay? So during my thesis it became apparent that single stars didn't match observations so you have to include binary stars. So we need to include binary stars and this actually is a computational problem, okay? So I was hated at the Institute of Astronomy in Cambridge back in 2000 because I went around and used everyone's machines and people would hunt me down saying, John, can you cancel that job because you've made my machine run slow? Bearing in mind this code only needed something like 128 megabytes of RAM. That tells you how shit the computers were in that department back then. They're much better now because then after the beginning of my thesis, people really hated me. By 2003 it was like nobody noticed anymore, nobody cared because the computers were that much faster. But yeah, I had something like 30 years of computer time even though I had, no, it's 70 years of computer time even though I was only doing a three year PhD because I was using so many different computers at the same time. But back in the day when it wasn't sexy to try and do this again, it's only a task farm. So it's not anything, it's not only difficult CPUing or GPU stuff, but it was a computational problem for myself just how to make this code that had never been run automatically before, run automatically along across a number of different grids. And actually now I'm really spoilt being at the University of Auckland because you've got the Nessie have their pan cluster here and as Nick said, he uses it, I use it, my students use it and it's just thousands of processes just to throw models at. And they come back five minutes later. Gone are the days of when it used to take a week to actually evolve these models back in the 60s and 70s. Awesome. Why do we care about binary stars? Isn't that other evidence? Well, yes. Now is this supernova back in 1993? This is why we need the Hubble Space Telescope because I'm not even sure which star exploded here but people tell me that one of these stars exploded. Okay. I think it's actually this blob here. Okay, so this is ground-based data from 1993 compared to Hubble Space Telescope imaging. This is why the Hubble was so cool. But what we did do with the Hubble is we went back after this supernova had faded in 2004 and if you zoom in on the galaxy, right, there's your galaxy, you zoom in to this patch and then you zoom in to where 1993 J exploded and you look at the position of the supernova and you're actually seeing it's still a star there. It's that little star there. Okay. That star when it exploded, there was a red star there and people thought when they analyzed the SED that there may be a blue star companion but it wasn't until the supernova faded you could actually go back and prove that that star that exploded was a binary star. There were two stars all putting around each other and actually it fits with the models because the supernova was not a standard red supergiant. It was a stripped red supergiant and so everything suddenly fitted together but it made people realize that binary stars have to be accounted for. So why is binary star physics different? Well, let's have these two stars in a binary. Evolution goes on as normal, nothing really changes but then one star becomes a red supergiant, okay? And you can't fit both of those things into the same space, okay? So if that red supergiant gets big enough, the gravity of the other star will actually strip off this hydrogen and so you get one of these helium stars from a lower mass system, okay? Sometimes you don't get a fully stripped star and you get like a various bit of hydrogen which is what that 93 J supernova was. But here you've dumped all the material onto here so this star becomes more massive and you get a hydrogen-free supernova where you might not have got one before, okay? Because this normally happens in systems where normally the star wouldn't have lost its hydrogen envelope. Why is this a problem though? Well, because, you know, stars get to very different radii so this is to scout a star, a 15-cellar mass star when it's born, that little mini circle in the middle, that bigger circle outside, at the end of core healing burning, it's actually this big and this final circle is actually off-center to fill it on slide but that's how big a star can get at the end of its lifetime. It's like a thousand factors of thousand or three orders of magnitude in difference in size. So you actually have to not create one model per mass of star, you have to put a model where you have another binary star here, here, here, here, here, here. You have to vary the mass of that of the system so rather than just making a few tens of models you have to actually make thousands which is the computational problem and what's worse is it really makes evolution unstable so you have to come up with all sorts of new novel techniques to, you know, I don't want to waste time for anyone for looking at these models, I want to automatically be able to say, ah, now it's a problem, no, it worked, it's fine, include it in the grid. So now it's all these things to try and fix and match up and the nice thing is when you do the rates and you compare the rates of these hydrogen rich to hydrogen free supernovae, it's about 70 to 30% and for single stars you don't get enough stripping and so you actually end up with hydrogen envelopes and so you get too many of these red super giant progenitors, the binaries you get too much and if you mix one single star for every binary, so three stars in every two are binaries, then you actually end up with the observed ratio of hydrogen rich to hydrogen free supernovae which is nice, but people don't like that prediction because there's another prediction we can make that if you look on the HR diagram where these hydrogen free supernovae should occur, you suddenly predict that they should go off at much lower luminosities, which is great. Okay, so in 2013 I published this luckily before a supernova was observed where we actually saw the progenitor and it's this one, IPTF13BVN and here's this really faint fuzzy blue blob, but when you put that on that diagram it's exactly where you'd expect it to be for a binary star. Okay, that's what you call lucky. Submitting and publishing that paper just a few weeks before you get the observational proof. Yes. Yeah, I was lucky. It would have been even worse if it happened afterwards, but anyway, let's not go there. I was lucky. Right, so that's what you can do with the stars. Then after my thesis, my friend Elizabeth pointed out well you've got with the rest of those tracks, okay. Look, you're not just looking at where the endpoints are, you've got the tracks over the rest of their entire evolution. And so that sort of asks you the question is what else can we predict? What else can we do? So now you have to start looking at galaxies. So, because the idea is if I'm saying more binary stars are like losing these hydrogen envelopes, you should see those stars in the galaxy. And we do, we actually see fewer red super giants and more wolf-rayant stars, so stars that have lost their hydrogen envelope, they're not from the synthetic binary population. And you can actually prove that and I've done that in some of the work. But proving that is not really straightforward and it's why I need to have this new code. So when we look at this, we've now got these stars code for making stars which you can download freely. You've got the atmosphere codes, one which is free, the rest you can get the results, but B-PASS is something where it combines these two to make some of these predictions down here about what things look like and what we can actually observe, especially in the most distant galaxies. And this will be free because I want people to tell me where I'm wrong. Because I've made mistakes, right? Everyone makes mistakes in codes. And the one thing I found is really useful is letting other people have my results. They find problems and tell me, they go, oh, I've made a mistake, which is okay, because that's how progress is made. So there's a website, bpass.alkland.ac.nz. You can also find bpass.alk.uk because I started writing this back in the UK. And it's still very much in development. I'm actually coming up for second generation of models right now. And you can download loads of stuff. And this talk will soon be linked from here as well, which is quite nice, it's being recorded. Interestingly, it's never been supported directly by any funding agencies. I've just done this as part of my day job as normal postdocs. So actually all the universities I've worked at, Institute de Astrophysique de Paris, Quincy University Belfast, University of Cambridge, the University of Auckland, just by employing me, have meant that this work has been done. And it's actually really being useful. And people really want these results because it's one of the only codes which have really had the results freely available that make predictions on what happens when you have binary interacting binaries in your stellar populations. Everyone else ignores it because it's too hard. I was a bit too keen though, because initially with Elizabeth, we wanted to look at high redshift galaxies which are really far away and we don't really understand them. And so when the referee read the paper, he said, I don't believe your code because you haven't looked at nearby galaxies. So he had to write another paper before that one just saying, look, it matches nearby galaxies, which we did. And then it started making very strange claims about high redshift galaxies which people are still sort of, I don't even know if I believe them. But it tells you something, there's something funny going on in those galaxies. Because in those galaxies, we can only see the light and we can't detect anything else. The real problem was though, stupid you used IDL. Okay, so this was about 10 years ago. Python wasn't as big as it is now. And I now know that I need to switch from IDL to Python, actually even to Fortran because it's just so much quicker than scripting languages. But Perl is really useful as well. And I quite like Perl. I'm old fashioned maybe, I don't know. So just in case you've never heard of a spectrum, I've already mentioned it. We're gonna be looking next to high redshift galaxies. So forget the fact that we can actually resolve individual stars in nearby galaxies. Those right back at the edge of the observable universe, they're so far away, we can't resolve them. We can only collect the photons we get together. So you're just basically looking at the colors of the galaxy to work out something about stars inside. But that's okay. Because if the atmosphere models are good enough, you're not looking at just like whether it's red, green or blue. You're looking for these dark and light bands of these emission lines. Okay, emission lines just like in the fluorescent lights above us in those old sodium street lamps. You know, if you see an orange glow, you know that's a sodium street lamp. If you were to take a spectrum of these fluorescent lights, you would see the emission lines which tell you what sort of elements are inside those fluorescent lights. The reason they appear white is because they've got powder around them that takes those photons and scatters them into other wavelengths. But you basically have a spectroscopy of taking a source, putting it through a prism and making a rainbow. Okay, we have all seen rainbows, right? Yes, I have in astronomy, first year of lectures, asked people if they ever seen a rainbow and some people say no. They live in New Zealand, we have so much rain. How can you not? Okay, so we're gonna look at redshift free spectra. These are really close to the, thank you. Oh, really? Okay, I shouldn't say that. I thought I had less time, okay. Thanks, Jessica. All right, mental recalculation. So for redshift free galaxies, these are, redshift is how far away something is. Okay, so the universe is expanding, cosmologists will tell you that, why that is. And so the further away things are, the faster they're moving away from us. So at certain distance, we actually measure how far away things are by how fast they're moving away from us. And of course, that's this Doppler effect that Nick talked about, that if you hear the police car siren coming towards you, the pitch of the siren sounds higher and then when it's moving away from you, it sounds lower. The same thing happens with light, but these things have to move faster, okay. But this thing is not really moving, it's because the universe is expanding and that always confuses students. So don't worry if you're confused. Now these are really faint. Well, they were back in 2003, when Alice Shapley did her study of these galaxies. And so if these things are so faint, they couldn't individually get a spectrum of each one. So they did the next best thing. They got a spectrum of around about a thousand and added those spectra together, okay. So you're actually looking at all the galaxies at a certain point in cosmic time, you're adding them together and you're looking for what the average galaxy at that time looks like, okay. And actually if you were to take the average galaxy today, it would look very different to the average galaxy back. This is like less than a billion years after the beginning of the universe. The universe is now 13.7 billion years old. And when we look at emission lines, there's these nebula emission lines due to just the gas around the stars as they're forming, which you have to model. And I'll get onto how I model that in a minute. But there's also the stellar components. So you can actually tell what stars there are not by looking and resolving individual stars but just looking for a lines of stars in these galaxies. And so here's something what we have. Don't worry too much about this slide. I'll actually show you on the next slide the individual lines, but here you can see these are actually in the ultraviolet. So these are at angstroms, okay. If you want to get nanometers, you divide by 10. And your eyes see between about 3,000 angstroms to about 6,000 angstroms. So these things are pretty far in the ultraviolet, which is awesome because there you're looking at the hottest stars in the universe because these are way in the ultraviolet and these will give you so much sunburn, then, you know, they'll burn you to a crisp. And there are two lines we're looking at. One's from carbon. So we're actually gonna measure the amount of carbon in these galaxies. And the other one is from helium, okay. And that's, so we're actually gonna measure the amount of helium. And, you know, this is just showing you that if you look at the different samples, the observational samples, this is before even showing, throwing models at the issue, there are lots of different sort of similarities. So these lines look quite similar, but otherwise there are things that are very different. Okay. Well, we did find there were two groups. And the two groups can be shown into this group, which is what is like a normal galaxy. If you looked at a galaxy, sorry, a low mass dwarf galaxy, like the LMC or the SMC, which you can see in the night sky from, oh, New Zealand or Australia. These are the sort of spectral lines we'd expect. So here you see this blob here. Okay, so the blue line is from observations. The black and the red line are the single star and binary star models. Okay, so you actually make the same prediction because you've got to convince binaries are cool. So you need to predict single stars as well. So here you can see practically actually no helium two line, okay. You'll see one in a minute and then you'll go, oh, okay, that's what he was actually talking about. And there's very little flux here. So there's actually practically no wolf-ray at stars. So they're very different. Here you've got this P-signal profile, which is the line wind. So these are actually from main sequence stars, main sequence stars like our Sun, but very much more massive. And this actually tells you something about the wind speed. And you can actually see here this red line and the black line almost follow that completely correctly, okay. This line here is actually just because of some cold gas in the front of the galaxy. So that's actually from a nebular absorption component. So we're not trying to model that. We're not trying to model this. We're not trying to model that, okay. That's from a cold, boring, gust cloud that's not doing anything. But here this tells us that there are these stars in the galaxy. We can actually measure the amount of stars just by looking at that line, right. Things get more interesting when you look at this because if you look at this helium two line, right, this one goes from being practically nothing to being a matter of big peak. There's actually even an extra nebular very narrow component on top, okay. And to actually get that, I need to change my model to get a very much bigger peak to be able to reproduce that. Really, the weird thing is that's not binary stars as I thought it would be. It's actually because you're getting stars spun up so fast they're about to fly apart in a binary star. So the physics is even weirder than I thought it was going to be. My friend Elizabeth said, let's go and model this. It must be binaries that does this problem and reproduces this bump. We were both wrong. It's much weirder. But we still have to try and find the evidence to show that this sort of evolution can occur. With the carbon four line, you can see there's actually a much shallower thing. We've still got this absorption component, which is to do further stuff. But trying to model that is actually quite difficult, okay. But it's getting there. And actually that just tells you there's actually less carbon in there than we think there should be. So you're actually not just measuring the amount of stars. You're working out how much carbon there is in that galaxy. And it's less than we expect there should be, which is interesting. Okay. Now I've already sort of indicated here that there's other stuff not in stars, okay. I've talked about nebular emission components, those very narrow absorption lines and those very narrow emission lines, okay, compared to those broader stellar features. So here we need to actually probably use one of the most famous, if you go to any astronomer, they all know this code, okay. Because it is one of the most famous astronomy codes. It was one of those that really got into the open movement quite early. So you actually download the code. You can modify it if you want. You don't really want to, because it's that big. But it will do all the physics for photo ionization. So if you just give it an input spectrum, you say, look, I have a gas cloud of this composition, this mass here. What do I see on the other side of it? Okay, so this is the perfect thing to complete that diagram. I've got my stars, okay. So they're the broad components I've just been showing. But then you've got these nebular features, these very narrow features that you get from these gas clouds and these gas clouds outside the galaxy. Okay, because they're not doing anything. So you actually need to include this thing called cloudy, okay. This is probably the best name to ever call an astronomy. Okay, it does clouds and it's called cloudy. Back when I was doing a PhD, you could go to their website, you could download their code, you could also download the little MP3 of Paul Simon and Simon Garfunkel, singing cloudy, okay. You can't anymore, I wonder why. Probably because of some, they probably got an FBI slap on the wrist or something. They should at least do a YouTube link, because it would be so cool. But it does all these physics and so you can actually do it with unresolved stellar populations, okay. Because trying to do nebular emission in three dimensions is hard and very computationally intensive. And the code is there called cloudy, this is really good fun. And the manual is like one of the best manuals because it's like so many thousand pages long and you can get lost in trying to understand it. But all you really want to do is just get one number out. But it's really, really cool. And Gary Ferland actually goes around now and this is, you know, does workshops on how to use cloudy. So this is actually where open astronomy software is going. It's not just writing in code, not just making it open. You need to train people how to use it. So you can actually see here, cloudy workshop. Where is that? Oh, it's going to be in India. And there's another one in Belfast. Cool. What does this mean you can do though? So my friend Elizabeth took my results and wrote this paper without actually me needing to do anything, which is really perfect. Because then she found something. Okay, I said other people find mistakes when they use my models. The actual really cool thing is, normally they find out that my models are correct without me having to fudge anything. Okay, which is perfect. Because if you make predictions, other people use them, they find it matches observations better and you didn't fudge the numbers. People when they accuse me of fudging my numbers have no evidence because I didn't fudge these at all. The problem is that you can look at these emission lines. So you can actually measure how wide and how much flux there are in those narrow emission lines. Okay, which tells you something about the composition but also tells you something about the light from the stars shining on that gas. Okay, so if it's really hot, you get some very extreme sort of line ratios. If it's very cool stars, then that changes. So here in these gray points and these red points, you've got some of those ratios. This is actually measuring oxygen to a hydrogen line versus that amount of hydrogen H beta emission line. And you can see when you look at the observations, okay, they're all in practically a straight line or when you try and predict them from single-star models, you find that the single-star models don't reproduce the observed galaxies. Oops, okay, and you can't really fudge them. Okay, so Elizabeth said, okay, let's check the binary star models. And then suddenly they go right through all those observed points. Okay, without any fudging or saying like, oh, we need to have this many binaries or this many single stars. It just went straight through. What's even better is that stars actually spend or the galaxies, synthetic galaxies, spend more time over here. So you'd expect to see galaxies in this part of the diagram and not really over in this part of the diagram where they don't spend more time. So I was very happy about that. But that's where you've now got this open-code stars, predicting stars, I've got B pass, which will be open. I've just got to work out how to make terabytes of data available. Like that's in the middle, making these populations and you're feeding it into another open-code to make a prediction that you can compare to observations that everyone can take. And there's many people trying to do these sort of observations and get these numbers to compare because they're telling you something about then the galaxy and also about the stars. Okay, but you can't understand the galaxy unless you can understand the stars. So then there's this horrible loop that everything depends on each other and it can keep you up at night if you have a conscience. So just to finish up in the last few minutes, I've told you about looking at things, predicting what galaxies look like and what stars look like. But a really cool thing is that binaries do more than just that. Single stars tend to just sit there. They may move a little bit as their birth cluster disperses but single stars don't really do too much. The cool thing about B-passes, I've been able to actually take it further and work out what other things can we see. Well, as I'm going through, I'm trying to write a second version at the moment. I've got a student. All my students are actually writing little bits to go into different parts, to do different chunks of the work because that's quite distributed then. But also they're doing science when they're doing this. So I've got one master student who's actually writing a Python code to actually go in along with all the FORTRAN stuff. So the synthesis stuff is being rewritten in FORTRAN because there's terabytes of data to try and combine to predict what a galaxy looks like. Okay, I need to get that terabytes down because otherwise, as I've said, how do I make that distributed without having to spend money somewhere on a server and to hold up the entire internet as people download it? And so I've got another student running this Python code to actually look at what happens when you have a binary star and one of the stars dies, okay? Because I haven't mentioned that. Oh, I've got our sequence. Okay, I'll come back to that in a second. So if you've got two stars in a supernova and say this one goes bang first, the system can become unbound. So we know when we look at these neutron stars when they're formed, most of them are actually moving at hundreds of kilometers a second. They're cosmic cannibals going through the galaxy and you can actually measure their velocities and try and reproduce that. But you also see lots of stars running around the universe. Okay, these are stars that are just moving along, again, at tens or hundreds of kilometers a second. So they can actually move quite some distance from their birthplace. And so, you know, here, we're not predicting what stars look like. We're trying to predict, you know, how stars move around the galaxy, which is kind of useful if you want to make that prediction. And here you can see some contours just where most of the stars should be and their velocities and most of them are actually quite slow. And these points are from some different observational surveys just showing that most stars are actually not runaway stars, what they normally call, but actually just walkaways because they're only going at tens of kilometers a second rather than hundreds of kilometers a second. Okay, this is the velocity here and this is actually the mass of the star. If you're wondering what is no stars down here, it's just, well, if they're not moving, they're really difficult to tell they're not. Okay, you're all moving through this. They're rotating, right? And the sun's going around the sun and the sun's going around the center of the galaxy. So you're all moving even though you're stationary, which is the problem, which is why it's difficult to actually work out what the zero velocity should be. Okay, now I'm out of sequence, I'll go back to say where B-pass is going in the future. So the moment B-pass is based on 15,000 detailed stellar evolution models and that's not enough and the reason why it's going to terabytes of data is because it's like over 100,000 models, which is a right pain because it's stupid amounts of data and it's my own fault for trying to be cool and beating everyone else. I have reduced people to tiers who do other stellar evolution models because when they say you're doing thousands of models, how can you? It's because we fudge and make approximations that make the code run faster with less issues. Other people make more detailed models with more correct physics and so actually it's more difficult to evolve them because the equations are just really nasty and no amount of computing power can actually help you. It then becomes a human power problem and it's not a human power problem that you can't fix in an easy way like you can with public science. So actually this is the real problem. Fortunately, my student has found a way of reducing some of the data needs of HDF files but I still need to get my head around that. And so yes, that was going to say what science are we going to address with the next one and this would hopefully all be done by 2015 with all the codes and everything released as well at that time because I want to get citations. I want people to use my work and so there's no point hiding it. So I have to make it with comments and publicly available. Oh and the other thing about Runaways is you have to think about this. I've said here what happens, what happens when the binary is unbound. There's also the other case when the system remains bound. So you can look at the neutron stars and stars running around the universe but then also you want to match these systems because these are cool. These are where the black holes are so you want to actually work out. Can we predict the black hole masses for stars in the universe? Which would be interesting. So there is the entire scheme. We've got a free code cloudy. We've got a free code stars, semi-free but at least you get the results for atmospheres but then this is actually really difficult so I'm not going to tell them off. B pass, okay I know I've been lax but I'm going to make it free but then you can get all this stuff out and it's all about looking for those not so obvious details down the bottom that you really need the binary stars for. The other thing to do is also switch to the most open code I have ever known in astronomy which is Mesa, right? So if you actually want to make a stellar model I'm going to be really tough now and say don't go to the Cambridge stars site. Mesa is the way forward. They actually have a community, they have a work group, they have problems trying to run these and trying to evolve models, they have tutorials. It's amazing what they do. The only problem is the code is about 100 times longer or 10 to 100 times longer than the stars, okay? But if you don't want to worry about that and you just want to make model stars of a certain mass or model the sun's evolution then this is the way to go forwards, okay? And you know they've actually even worked how to get it all parallel and that's because there's a guy called Bill Paxson who used to be a computer scientist, got the stars code and has adapted it for use and used all his knowledge over his entire lifetime to actually make almost the perfect, beautiful stellar evolution code. So I'll leave my summary up there, just talking about open science and using Linux, okay, because I didn't mention it but all of this has Linux under the background, okay? When you turn up an astronomy department you use Linux or Unix, okay? You don't use anything else, you don't use Windows apart from, okay look the only reason I use Windows is to play computer games, I'll be honest, okay? Because let's face it, you know, anyway I might go and talk computer games, that would be another talk and I could talk about the science of science fiction computer games, that's not actually the story. So yes, and the idea is that, you know, the more interesting thing for me and my motivation for now trying to make all this open is as I've said, to find out where I've gone wrong but also find out where I've gone right without trying to be right because that's just really, really cool. So yes, thank you and I'll be willing to answer your questions. Do we have the other question? That's fine, I'll do it. Sorry Nick. Okay, thank you so much Dr. Eldridge for that fantastic presentation. If we do have time for a couple of questions, if you're happy to answer them, just while Nicholas is setting up his laptop. Who has a question? Anybody, anybody, anybody, just a minute. So in CERN they're now using the simulation. So what's the breakdown in languages in astrophysics, it's still all Fortran or what's the situation there? It depends what you're using. So there is a lot of Fortran, there is a lot of C++. I've actually had arguments for people about C and Fortran and it's always an interesting one. Bill Paxson actually kept stars in Fortran because he wrote this entire one or two page article about why Fortran is just as good as C++. But the more modern codes you write a C and other languages, Fortran is big because of your right, because we've been using that for ages. Python is the really interesting one because that's really taking off. I mean, a lot of people have been using IDL but it's not free, it's like thousands of dollars to own it. And so now the Python's taking off, that's really cool. Just wondering what was prompting your decision to move from IDL or PDL with Perl apart from obviously the age of the languages themselves. And do you have any problem transitioning your associated tools along with it like the viewers and everything that come, that are dependent on these languages? So because I'm a theorist, I've just got these models. So I've just got text files. And so I can redo really quickly in whatever new language, how to edit and load in the numbers and text files and play with them and now put them. So the real reason to go away from IDL is because it's expensive and it's low. So I wanted to go to Fortran or at least C++ because it's quick. Have you ever tried PDL which is the Perl implementation of IDL? No, I haven't. I didn't know that existed. So again, this is why I think it's good to come here because it's PDL, okay. I'll do a quick search and sit down. Thank you. Any other questions? Okay, so what's the next bit of hardware that has you drooling? Oh, next bit. It's not SCAR. It's actually the EELT. Okay, so this is the extremely large telescope that's 42 meters across because there we can actually start getting these spectra but for even earlier galaxies. Because you know, you should see big differences in how those spectra look as you go further back. So you should see more of that really extreme helium-2 line in those earlier galaxies which tells you something about the stars. So yeah, but there's so many different things going on with astronomy and all these instruments. It's all just awesome. One more question down the front, did you know? One last question down the front. I was just wondering what, if you've considered developing in the open and what might stop you from doing that rather than quite until you've got a sort of a complete finished working chunk of code. Because I want to know it's working to a stage because there's an issue that if I make everything available now and it's not working, then that people will still be using that old version. I want to get a version that's mostly right and then I'll take that step to making it open and developing it in the open. So it's getting, it's being the brave point of where to put that, it's almost right and putting it in the open. But yeah, but then all the stuff that I need to go and then learn how to do this sort of open development using, was it a GitHub and other things. I know the words, I just haven't got into them yet. So yeah, but then yeah, just because I know I will find mistakes and I'll find also really good things, so yeah. Okay, thank you, we'll have to wrap it up there. Thanks again, JJ, that was fantastic.