 Great. So today I would like to talk about the Zwicky Transient Facility. And I know that all of you are committed to the CTA, which is an upcoming high energy astronomical facility. And you're very interested in transients. And you may want to regard ZTF as a Zwicky Transient Facility as a precursor to the next big survey that's happening, which is the LSSD undertaken by the Rubin Observatory. And in fact, in large part, the National Science Foundation funded or co-funded ZTF for us to precisely do that, which is to as a precursor, a facility precursor survey and understand and develop, help develop not just ourselves, but the community as you'll soon see. Okay. Great. So let's see how did the, as you know, the universe was actually very simple when it began. If you look at one part in the universe, so here's the redshift at the top. So it's a two dimensional representation. Yeah, there's one part of the universe to another part in the beginning. And the matter density and the energy density was the same to one part in 10 to the five. And it matters it was one part in 10 to the five, and not a zero because it is that small fluctuations which gave rise to, and thanks to long range of gravity, gave rise to gradual condensation of the baryonic and matter, which then led to stars, groups of stars to galaxies, clusters of galaxies, groups of galaxy clusters, and so on so forth. But the universe was actually even simple in other ways, the universe is chemically simple. In fact, the chemistry in the young universe is very simple. There's hydrogen and helium, and almost nothing else. It turns out it can actually form a little molecule out of hydrogen which was only recently found, not in the universe, but in the lab. But the rest of the periodic table was empty. And so I would say, in the last century, the great achievement of astronomers and physicists working together was to understand how did we go from a such a simple universe chemically speaking hydrogen helium to filling up the periodic table. And we understood it qualitatively, we understand, you know, how this, but then in this century we hope to finish this problem which is not how do we get to some amount of chromium but why do we have the amount of chromium we have, let's say related to hydrogen. There is a goal right now is to determine the filling of the, you know, the fact there numerically the abundance of the periodic table. Okay. And for this you need to understand three things for the light elements here, you know, up to a CNO you can get stars like the sun, then the, then the slightly more massive stars can proceed up to alpha elements here. But by the time you start reaching very high Z elements in particular to reach the sun it's iron itself, you need supernova and to certainly go beyond super iron the peak you need exotic supernovae. Okay, so that's where the supernovae come in. And the history of supernovae research, I would say really begins with Fritz Wickey, who was a Swiss astronomer or physicist hired by Caltech on the eve of Caltech getting money from the Rockefeller Foundation to build the 200 inch telescope. Okay, so it came to Caltech in the late 20s. He was the only astrophysicist we had because we didn't have an astronomy program then. And in fact the astronomy program only got founded after, you know, about 1950 or something by Jesse Greenstein. So, here's Wickey, circa 1936 maybe, and working with Bardet, he did pioneering research with in Noviens in basically explosive transits and the two coined a new word at that point called super dash nova, which are a group of objects that are brighter than no well, in fact that and the four times brighter. And now of course we accept that word without a hyphen, as Wickey appreciated that the newly invented Schmidt telescope so this is the first is the second Schmidt telescope in the world maybe the first one actually got the, the, the, the corrector lens was in fact round by Schmidt himself was a great optician. And this is an 18 inch Schmidt that was put together and started operations at Paloma, about 36 I think, and he began a pioneering study of doing systematic investigation of the sky for supernovae. And it's in his honor, when I had an opportunity to name a new facility, I would name it as a Zouki transient facility. Okay, was supernovae one part but to go to very high the elements we know we need even more exotic explosions. Just a second. So, we need even more exotic explosions. Just a minute. Okay, and that we now know is in fact, the co lessons of neutron stars, which produces the very high Z elements, so called our process elements. Okay, so I thought I'll lay down the intellectual foundation for what is called the, what we now call a ZTF but this program began, you know, in, in the first light was a chip of PTF which I call this phase one so right now we're in phase four. So it was called a Paloma transit factory run for 2009 to 2012, then intermediate Paloma factory transit factory for IPTF 2013 to 2016. Then we took a year off because we're construct then 20 or more like two years with then we started the Zouki transient facility which is really phase three of this project. And that started in 2017 finished in 2020 and starting last year we are in phase four which is a phase two. Okay, let me. So one of the things I sort of regard, you know, I have three, or should I say astronomers and look up to in the sense I say what these are like amazing I just don't know how these people even thought about what and from Pasadena, it's with Zouki, then I have, you know, and in Western Europe and, and Zeldovic in Russia. So, Zouki was pretty amazing guy, it was a really, I would say kind of a contrary and I like that. And he said I soon became convinced that all the theorizing would be empty brain exercise and therefore waste of time, unless one first a certain what the population of the university really consists of. Okay, so I know that many of you have come to a stormy from physics background. So I have to give you my views on now physics is a wonderful subject in fact it's sort of a key subject in science. The point of physics is to, it's an absolutely reductionist approach. So you look at diverse observations, see if their patterns usually it's discovered by astronomers and so on. And then you say, can I come up with a law that encapsulate this patterns. So, you know that planets that go around there's periods and so on, then you come up, and you have capitalist laws. That's fine that's a patterns. But then it can I come with a physical model which is some approximation to reality. And that's Newtonian mechanics. So you get you immediately explain very nicely with so it one equation you can now simplify so many observations. In contrast, in contrast, astronomy is the opposite. The point of an astronomer is not to try to out imagine the universe, it would be hopeless. It has 10 to the 11 galaxies, each galaxies 10 to the 11 stars, that means that 10 to the 22 systems, and the universe has 13 billion years to think and work and make changes. None, our imagination would simply not match that. And this is where I'm afraid that I sometimes tell my physics friends that they place too much emphasis on their ideas when in fact the real idea is out there for us to go and discover. To this I had a small amount I have a very strong thesis that discover is really a function of technology and algorithms. In my opinion. If you know you could be a great astronomer by simply being at the right time at the right place. Now of course it helps if you made the right time happen and the right place happen. So the what you can discover is all the things you can do with your instrument that's all that is to it, you can sit around and imagine as much as you want. But to make this career you need gizmos. Okay, even if you build stuff, I would say, why waste your time you should automate. Okay, and even that's been the principle driving goal for PDF through ZTF. And at the risk of annoying some of my colleagues up in a proper care at Princeton I said the best way to do the stones to get the stoners out of the dome. Because humans tend to slow things down they have too many ideas usually not so great. Let the machines do the work that they usually end up doing a better job. Okay, so the, the. Oops. So the idea for ZTF what became eventually become ZTF was due to a course I was teaching right now. And that's the one things about nice thing will bring a professors you're forced to teach, which means you're to actually understand what you're doing. It's only when you teach you know how much you know and what you don't know. So as a fun exercise for my high energy astrophysicist class, I considered a toy module where I said okay you have two neutron stars that collide. And we know of course it's a very complicated thing but as a Fermi energy of that neutrons is lifted, maybe in the outer portion some of the neutrons can decay freely. So it's a nice and analytical because the neutron decays simply for standard radioactivity so I can write down a simple formula for the heat supplied by neutron decay. And I was surprised that even 10 to the minus four solomance of neutron decay would in fact power a source that be visible in the sky, at about minus 13 to minus 14 magnitude and lasting maybe a few days. Okay, which I, at that time I called it as a macro Nova, and then I said well astronomers are looking at supernovae many of them, a few are studying no we why not build a machine to go after this area. And that was the start of PDF. Okay, so the idea of PDF was, it's a factory it's very I chose a name very deliberately. It's at Palomar, it studies transients, and it is designed to be a high target throughput machine. Okay. So, and this is to systematically study explosion not just supernovae, like here, not just know we like here but anything in this space space. Now we all took hardware which is necessary then we'll astronomers learned the bad way that software actually matters and we all now know we pay more for software than hardware. But if you're doing a project like this, you also need need what I'm adding this. I call it as gray where which is gray matter which is an English term for brain. So it's like brain power you really need that so PDF. We went from concept to first light in 26 months. I do tell you that I have this very strong view that anything worthwhile doing in life, in my opinion, at least for me, you have to do it in a fire in a very short time. It's almost impossible for me to imagine working for 20 years in my life to get a project. So this went from 26 months and we are on and so many young people. We finished his PhD from Cambridge UK. Quimby had finished his PhD from Texas, Austin, Iran of it from Tel Aviv. All these guys basically they're made. He was a project scientist for this whole project and this guy's software lead which doesn't sound great job but it's the worst job because it actually integrate everyone's software. And this was a robotic aspects manager machine learning was introduced to us first time by professor who is now who's become professor bloom. And the image difference in pipeline or engineer Richard the Kenya led the hardware and the photometric pipeline was led by Jason Suarez. Okay, so let's see what are the just a second. Okay, so let's I'm going to use a diagram here, which I am going to call this wiki diagram. And that is the time scale of the event on one axis, one day and days one year and luminosity peak luminosity in the visual band absolute luminosity on the on the y axis. I had to make my slight joke about our physics friends, you know, astronomy is a very rich subject, which is why many physicists want to come and work in astronomy. So, and, you know, physics are pretty bright guys so we thought very hard and so the main thing is we use. First of all, we use cgs and that makes physicists very confused at times, but then when we use magnitudes which is log and then you multiply by 2.5 and so on. At this point, the physicists just give up they say oh no we can handle this. It's just a joke. Okay, so you have a face space here the thermonuclear one a supernovae here. And if you collapse these things produce black holes or neutrons or sometimes they actually simply explode. And then of course the classical movie which are explosions on the surfaces of white walls on the disks of the creation this okay in general, the whole movie phenomena. Okay, so once, once you have a machine like this, as I told you the job of the astronomer is not to sit around and actually say, you know, what they could be is to in fact go find what there is. So, nature rewards we build machines and it's not just to just don't want to say that the population population populating this diagram is not just ptf but many surveys that came in at the same time as one little assassin. And now of course we have Atlas tf and so on. Okay, well, wow, look at this. So many new things are popping up luminous supernovae super luminous supernovae so these are 10 times brighter than supernovae they live 10 times longer so they have the total and a volumetric energy releases 10 times more than thermonuclear supernovae. Then we have supernovae which are so rich in calcium okay don't understand why only so much calcium. Then we have a whole new category here and which initially we called as luminous red movie. Yes, they're bright, they're red at peak and they're new. Okay, but it turns out there's two groups this group here that I'm circling is so called ILRT Intimidate Luminosity Red Transients. And then this group is really now only reserved for luminous red movie we believe that this is due to co lessons not co lessons merger of two stars. Stellar merger happens quite frequently it's once one every 10 years or so in the military for instance, and this we speculate is due to electron capture supernovae. And there are some things we don't understand that we give a name but not very sure that's correct. Okay, so I don't want to go into detail because you know PDF has been running 2009 to 20 now to 2021. And we have been averaging, you know paper every two weeks since we began. Maybe we're speeding up a bit now. So there's a lot of papers have happened so I, and I have to give you so my summary is mainly by the way methodology and that is the spirit that I told you my view is with the machine, the astronomy straightforward and writing the papers you may support. In fact, if you have good data. It writes the paper itself. There are times when in order to days I finished a paper. And so as some of you know I have, I do like to send papers to nature short suite, and then you're out with it. So in two days. If you have good data you can just bash to the thing, but building the machine. No that's not today's that can take many years so let me focus only on the technology. So in phase one of, of, of, which we call Paloma trends factory, we, at that time it was novel, and thanks to Josh bloom, he convinced me we should really introduce machine learning and really really work. We have rid of a lot of manual labor from the, particularly to distinguish genuine transence from bad subtractions. A lot of the science in those days was you find the supernova your candidate you report. A couple days later someone is an telescope to get a spectrum all this sort of stuff. Okay, but we went into same night classification, you know, it's easy now but those days a little hard as usual things get better with time usually. In phase two, we introduced mixed cadence observing because for phase one I said let's stick to one color will do one kind of survey and that's it, like a finite survey so we put harvest a lot of supernovae. We introduced multi band. Of course, if you know more bands it's, there's more information, but there's more complexity and not all the time complexity gives you the information. The most important thing we did in phase two was a robotic field in a spectrograph and demonstrated so let me explain this. When for those of you who are not optical astronomers that the party or to understand is the in when you have to get a spectrum you really need the traditional approach you have a slit and it's very fine, it's usually not second or so. And so you acquire the field. Okay, that's a star and usually you argue with the colleague whether it's really the star with it north is up and so on, then you use a joystick put the star in the slit, then you expose. So that takes time and also takes a person. Okay, so the. In, in robotic spectroscopy what we do is basically it's a it's a it's a system where you can about 30 by 30 seconds, you slow the telescope and every pixel in that 30 by 30 second, we get a spectrum. So one is there's no arguing with your friend whether north is up and it's a star because all in, and then you don't have to be there, you don't have to be aware get all any telescope points pretty well so 30 by 30 seconds so it so that's a part that's I would say great and innovation that we did. Okay. So, then in preparation for LIGO we did and what I call the needle in haystack search, which is, can we really look at at that time 100 square degrees and demonstrate that we can actually find a fading transient. It's not so easy, because many. This is a mistake many people not new to or new time domain don't get it. Let's say in your own program CTA you're looking for some object, you have found an object and you want to find the optical counterpart. Naturally as an experiment in CTA you'll be you'll be thinking what that kind of a looks like, you'll have some ideas, maybe they're right around doesn't matter you have some ideas. And that is that you know what you're looking for approximately. However, the sky doesn't care the sky is going to supply a very large number of false positives. And the trick in this business is not to set up a filter to find what you want is to set up a filter to reject what you don't want, because that is much larger than what you want. Okay, and that's a part we did and we are not we can now do routinely a search of 1000 square degrees and given the certain features you say this is the kind of transient I'm looking for we can isolate that and reject all the others at a very high level of confidence. Okay, phase three the ZTF and we finished it was so successful that NSF pointed us for an extended operation. And here we introduced alert distribution. So here's an example of why it's not so easy to find what you want, but we did that in IPTF error. Okay, so what we were able to demonstrate which was a novelty then is, as you know gamma ray burst, they produce. Many of them produce a afterglow the burst itself may be a few seconds, but the afterglow can last hours or days. Okay, but mostly you use the gamma ray trigger and then you go and look over the gamma ray burst as long as it pointed out well we are now in a position we actually can find GRB afterglows without even a trigger so here it is. There's a reference image, which is a standard reference image and then we have we are taking consecutive pictures, nothing here. Okay, February 26.38 something is coming up 26.43 this is an hours. Okay, it's slowly fitting away 26.50. And then we can do so many things here. Okay, see what's happening here, you know, this flagged by software spectrum from Keck, swift slews, and then it turned out, we knew this is an afterglow because it smells like an afterglow. And then a couple days later when the data from the Mars mission came in, the actually parent GRB was found. Okay, but now we find this routinely we can actually now find GRBs without GRB triggers. Okay, so as I told you I was somewhat of a gizmo approach I said why don't we automate the discovery of the universe. So here's my vision for that which is. Can we can we find something which is, you know, a machine that I'll come in in the morning and I'll say machine. So all the interesting things you're done, because it is observed, then I look at the thing then I sit down, work hard through the paper, I write a paper, then go in the evening play a game, have nice wine, come back the next and keep doing this that sounds so glorious view. Of course, the next phase is, I come in and the papers written. Okay, so this is the goal of ZTF which is how much more automation can be introduced. It's a very large project, it's got big support from NSF and worldwide partnership, we are a large number of partners. Last I counted we have 14 partners around the world it starts from Taiwan and ends in California. Okay. Okay, so we built. So our approach was to build a super large telescope it's got 47 square degrees. And this I don't think it's really to scale this is what this is for the sense squared it is this little bit looks a bit too large to be four times that. But anyway, on this is the largest field of view on the on the smallest telescope here because this is a 1.2 meter Schmidt telescope. So these 10 square degrees behind a six meter equivalent telescope past us is behind a 1.8 meter Subaru is behind an eight meter and so on so we went deliberately what I call as shallow. It's in fact we unit the not exposed for long time it's only 30 seconds, because when you want to make a discovery in my experience you really want to find the brightest you want to go faint enough to make a discovery that this volume, is as bright as bright as possible so that you can do follow up going faint is not necessarily helpful a lot of people, you know, and I have to give you my, not so much complaints and observation. Many people who are in order and LSS T will say LSS T will discover gets it in supernovae well that's not true. You can actually discover something you have to know a bit more than just finding it okay. It's very calm, it's very time dependent statement. You can, if you take two images you can set something new as popped up. That's a candidate, it could be anything. Okay, and then you can even do a probabilistic statement if that new thing is next to a fuzzy thing I a galaxy but it's not discovery. Okay, so yes, you can find many transient so what it doesn't mean much because there's a supernova every second in the universe. And you can find many of them and 100 years ago the first supernova valuable but today, you know, ZTF not only we are total capacity for discovery or finding candidates maybe 1012,000 a year, and of which we only pursue about 2000 because we're not can't pursue all of them. So I hope you that you appreciate that just because you can find lots of faint things. It doesn't mean you're finding your advanced acknowledges saying that there has to be a plan to pursue those faint things in a systematic fashion. Okay, what are the achievements of ZTF phase one. We went beyond machine learning went to deep learning or fans fully called AI. And now we have an AI assesses discovery algorithm for comments, we can routinely find them and we are finding them. We made discovery of asteroids within Earth's orbit, and we had to discover if the an asteroid within the orbit of Venus, the first one actually we found massive, the most massive white wolf unknown to date. We had some speculation that we wait long enough to actually blow, it will collapse and explode. We have doubled the population of this peculiar calcium rich supernovae. In fact, we find low rates of supernovae so routinely sort of become a nuisance now. We can find supernovae within hours of short breakout. In the past, it would be that if you wanted to do young supernova research, you would, you know, you find something that you call friends and build frantic activities. But now we'll say, yeah, you know, I got this party coming along when we just do this, you know, next Thursday, you know, the machine is there will be assured of finding something that you. So now this is instead of being dramatic drama it is not routine. I have already spoke a bit about grbs without jrb trigger, maybe we found a class of cosmological relativistic explosions that, in fact, do not produce jrbs but otherwise relativistic is yet to be proven but it looks interesting. One of the things with ZTF that is new and in fact we're the only facility. We actually supplied transants in real time so let me explain that we supply transants and these are called brokers there's a there's group in Berlin, Enboro, and Santa Barbara, and then of course, and in Tucson, Santiago and also to these days I think to Seattle. So when we take two images, take a reference image and new image comes in let's say right now, we do sophisticated image subtraction machine learning to take all the glitches out and then we get genuine candidates. Many of them are things that most of you are not interested in dwarf layers, I think very few people are interested in that. The interesting in their own right dwarf no we okay. Eclipse in binaries. Then finally you know, moving things because when you move you think it's a transit. So, following LSS TV define an event as something which is a five sigma change in right ascension that my nation or the flux in that band. If there we call it an event, and we put out a very information rich packet to the people who are receiving this okay, and these brokers then serve other communities so for example you could be talking with your colleagues in Germany. You could receive this alerts or for the southern hemisphere could be talking to your colleagues in Santa Cruz not necessarily to talk to North or South I'm just saying that each broker is not specializing in things. Okay, so that's what we do and LSS TV will of course, will have a rate which is maybe 30 times more than our rate, but it's the same, the industrialization has happened. And one example I you know there's no way just way I can do justice to explain all the achievements of PDF through ZTF in a talk like this. So double degenerates. Okay. So, the double degenerates are systems are of great value to astronomy. Of course, many of you are famous with the most the most famous ones are double neutron stars for the degenerate the co less amazing things happen. Okay. And But let me talk of double white walls. Okay, so here's a picture taken at high speed, not with ZTF with another camera. And what do you see here is you know there's a star but a very short amount it flickers and goes away for briefly. Now it turns out this system is, in fact, a seven minute binary 649, and that's a deep eclipse that you're seeing it actually has even second images. And this is a beautiful hypercam light curves obtained. I think from GTC. So, we, so these are two white walls one is hotter than the other, which is why you see the under when the part one is eclipsed by the cool one it's almost complete the cool one is bigger. And then the cool one itself is irradiated by the hot one and that's why there's a get the second big clips. Okay, these systems are so relativistic that even in in our own data set. Okay, that is from PTF to ZTF we can actually see the orbital decay. So this is the, you know, assuming if there's no decay just like the hull stellar system with the famous type and we see the orbit decay. The, the because of the orbital decay the eclipse doesn't happen that exactly the period is slowly drifting. And over this time in, we can actually see the drift is 2,500 seconds. You can do general relativity with your wristwatch because most wristwatches keep time to better than a second, even over 10 years. Okay. So anything fancy at all to see a GR at work with these eclipsing systems. And what's shown here. Sorry. Why the systems are so we anyway this is Kevin Burge is a thesis and he's now finished his PhD and gone up to MIT as a popular fellow, he's found so many of them he's practically doubled triple the population of eclipsing their systems which are already in gravitational waves. Where will these be these these systems are the prime. The first thing when Lisa goes up, they will see the systems in fact this particular source will be the brightest source for Lisa. So when I go expose the band, you know, centered around let's say within 10 hertz and maybe a kilohertz Lisa expose the band centered around, you know, something like a millihertz to about 100 millihertz. And this is where the double degenerates come in. Okay, so what's shown here is that get on the y axis is a gravitational strain. And on the x axis is a gravitational wave frequency a millihertz is here. 10 millihertz is over here. And these are the systems which are eclipsing which means you know the geometry very well. So we can actually predict model because we know m1 m2 we can in fact predict the gravitational strain in both sense of polarization and this would be really amazing for Lisa in by the way of calibration. Okay, so the last bit of my talk is in Zika phase two, which started just last September, we're almost we've done now a year. And now there's a new partnership here. Some of the old ones came in but the new ones so just our interest to you. And so let me tell you the partners we have National Center in Taiwan, Weizmann Institute in Israel, then we have Oscar Klein Center in Sweden. IN2P3 in France. Desi and Humboldt University in Germany. Work University in UK. Trinity College Dublin in Ireland. So you can see we have a huge European component then to University of Maryland, the College Park University of Wisconsin. Milwaukee Northwestern University. And then Lawrence Livermore National Lab, IPAC, and the first of Caltech so it's a pretty large operation here. Okay, so the goals of Zika phase two is, you know, we want to really want to like to focus on cosmological relative strengths, especially without gamma ray emission there's so called dirty fireballs I if they exist that'd be really fun to find. And of course, but constructions, they won't produce gamma rays. And the investigation of double degenerates has been spectacularly successful, you know, Kevin's Burgess thesis has been revolutionary. So we want to go there. And of course, there's the usual stuff, you know, of improving our computations. One of the new projects we have begun is TD ease a nuclear black holes and this is in partnership with the Russian German exclamation sg. We are now conducting the largest systematic supernova program, I would say in history. And that is something like not, as I said we can quote uncovered discover supernova candidates maybe 1000 a month, but we only look at the brightest 100 or so per month and classify them. That means to tell you these are supernova and what kind it is. Okay, and this is led by two young people Daniel pearly who is in Liverpool in the UK, Professor Fremling was a postdoc at Caltech. And so we classify objects which are our goal is to be completed 18.5 mag peak and so we sort of go a bit below 19 just to be on the same side. So this has been a fairly large substantial program. And this what happens. Our peak maybe is six or seven classifications at night. So we are now reached. So here are, let's see, blue is co collapse, red is one a thermonuclear and black means galactic interloper, usually a dwarf novel. So here's what's happening, you know, the first supernova that astronomers understood is is something unusual is SN Romeda 1860. Then, you know, Zuki, the things they finally gave the name Zuki began his program, which are continued to about 50% and then it's been steady for a long was then when the electronic the big service came in exponentially rising. But the number of follow up telescopes is not so the deficit in what you can discover and what you can understand is growing dramatically. Okay, so our goal has been to classify anything brighter than 18.5 so we're trying to hit to something like 1200 supernovae a year. So we actually do a bit better. And for that we use this robotic spectrograph that I mentioned with the IFU spectrograph. It's by the way this whole system is is not just robotic meaning you know we don't go there and all that. It's also self sequenced for the whole month we don't sit around program every day. The dome opens up pipelines run candidates are found. And there's no one actually operating this whole system at all, and high value candidates are chosen by an algorithm and send to this which then does the observations pipelines run spectra extracted the spectra then fitted again no humans to a to a library of supernovae and classifications done and an ATL is sent or saying yeah it's a one and here's the thing. Okay, so we are getting close to my vision of automated discovery. Okay, so here it is we get 10 to the five alerts that's typical sometimes we can reach about 500,000, but a few hundred thousand per night is the is a real is a real that there and there we already thrown out the moving objects from this okay. And then we use a whole bunch of the surface of classification to actually get to the ones we think are likely supernova that's a lot of reduction 10 to the five to 10. Oh, it's easy to reduce and what you don't want to do is throw the baby out of the bathwater that is if you also get rid of the real stuff that's not good. So we actually have a false positive and false negative rate metric and it's called the ROC and we have our ROC is excellent actually that is a rejection of true events is extremely small. Okay. And an acceptance of false events is very small, maybe less than one in 100. Okay, so some of this is now getting automatically classified through after the spectra done decision is taken and sent to the transit main web server. Okay, well these are statistics I don't want to dwell too much but I hope you're just impressed with the large number you're saying you're not looking at 10 or 20 we're looking at 1000s and 1000s supernova okay. And look at this collage I just want the point here so impressive it's like, it's like a machine, each of these supernova, look at that this. Then we can put it back in the Zouki diagram and this is done in real time so if any if you want to look at these things. Next day we get up there will be a few more points. Okay. Just just Google for bright transient surveys with ZTF. Okay, so just from this space space you know things are interesting so it is. Look at how rich the sky is. And if you just take this data we have okay and you plot the magnitude versus duration and this is called the Phillips relation this is the one which this is the key to the discovery of dark dark energy through supernovae. And Mark Phillips, you know, is the real hero of the story. This is what allows one is supernova to be standardized. Apparently there's an equivalent Philips relation but with a different tilt for for the co collapse supernovae. And then you find all this outliers 18 cow many of you have heard of that and now 20. There's another 18 college object 20 cxt. Sorry, no, no, there's another 20 CND it's it's it's in better than the car with the papers and coming up. So the NSF very kindly gave us an older telescope which we refurbished roboticized, and now we are, we basically have a lease on that at no cost. And grad student but yes we share my she's building a spec the one of these robotic I fuse for this so we should be able to double the supernova classification and do and do very fast photometry, the sort of describe starting coming January. So let me end on a couple of things that might interest you. Many of you know for ice cube is now producing routinely very high energy neutrinos and it's very intriguing. And of course, you know, as usual people have ideas, but ideas are, in my opinion cheap. But only when these ideas are actually real then becomes valuable. So we have. In this case, maybe, you know, as I said between 100 to 500,000 objects per night. Meanwhile, ice is not producing that was the forest here so that heavily but it's producing now and then. And the question is, we here we know many things and here it's really new territory. Can we, how do we go from this one or two rare events from ice cube to all this stuff that's happening in the sky. Okay, so you just have to get pretty good with a jet asteroid stars planets correlate. We classify and then it's hard work, you know it's very. It's very definite until you start seeing pattern again and again. The many possibilities. Okay, how you very good this high energy neutrinos gamma rays, gamma ray burst so called failed gamma ray burst type to supernovae which have very the, the, the densities are very high. It's natural to produce a lot of very strong collisions here. AGM maybe title destruction events, there are many possibilities. So, you know, you shouldn't take it's a thing where you simply correlate, I mean that's the best you can do, and if it keeps repeating every time it repeats, then slowly your confidence grows. So with each ice cube event, you know, there's a whole group here in CTF, they follow up and frankly this is a bit of a crap shoot right because you don't know just because after an ice cube event you found something. Okay, the sky so rich many things are happening so you shouldn't, you know, you should do this as a fun activity but never take it seriously, until 100 times over it's done, the same thing happens. Okay, so you know it turns out, we believe that maybe, you know, there's some hints that some of these ice cube events maybe correlated with TD, but these sorts of probability calculations are as you know historically very tough. Do you understand the look elsewhere if I correctly do you understand the sky itself correctly, and you have freedom, you know, here's the ice cube event happened here why didn't we can say, hmm, that's why didn't it make it here why didn't it make it there well we don't know but we, we have enough of a machine that over time we hope this will, this will show. So, there's some hints of a neutrinos from PD is our ice cube event colleagues excited. They're also excited about looking at dust echoes for as an indicator of PD is and then which in which in this hypothesis are related to high energy neutrinos. It's early days. So the next one is a stellar black hole so as you know, that LIGO has been finding the core, core lessons of black holes and the peculiar thing here is that the black holes that are being inferred the masses that really massive you know there's like one which is I think 7080 solo masses or something. 5050 solo masses are like routine now. Well that's not obvious at all because in our galaxy no stellar black hole is 8050 maybe highest is 30. We don't seem to know how to make that. So one idea is in fact that these are not stellar black holes made in one shot but they're made in two shots. Okay, so let's go into the central regions of galaxies you have stellar masses super massive super massive black hole in the center there's an accretion disk and then just because of a mass aggregation the world of stellar black holes hanging around. And then they go hierarchical merging and produce this very massive black holes. And then when when the. So that's fine that that seems like a good, you know it seems like a reasonable picture. Okay without invoking very exotic wisdom at very massive stellar black holes, but now let's say that not all super massive black holes have a accretion disk on them but one in 10 do and let's say there's a there's a binary system in the disk because that's where they'll be hanging around you're not messing this button in our regions. It create coalescence which means the masses reduce so now it sets off a disturbance in the accretion disk okay. So this accretion. So, due to this disturbance. Okay, you could then get an event. Okay, as as. And you could then get further on the higher accretion, and therefore, maybe there's an electromagnetic signal following coalescence, if it happens in the accretion disk of a galaxy. Now you say what there's so many ifs and maybe is yeah, what else do you do this is frontier territory and we have to, you know, be a little imaginative and keep your eyes open that's that's what's happening here so. Here by Matthew Graham or project scientist. The group looked at 21 BBH triggers, you know, three, and then we just went through the city of database in that approximate region of the sky because the localization was for the black holes is not so bad the signals are very strong. And okay here's here's sort of where you know the signature is this where the black hole manager happens, and then apparently we're getting a pulse of explanation. So, yeah, seems plausible, and this is a particularly fantastic when it's 8566 producing 142 that this event Okay, these are ZTF like those in three colors G band are better than I've. Okay, then you will say, Oh, of course, but there's only false positive right. There are lots of agent in the sky. Okay. And if you're finding these blips and claiming evidence we should also make sure that when we find a blip and if you know, and like I was not looking how do we know this all true so, you know, so we, they did a problem, as best as they could I suppose look elsewhere effect. And the idea is it looks reasonable that there's something happening here but I think what is the final thing so let's see I think they're 0.5% right now. Okay, of a chance chance coincidence for this particular event. Okay, so there's some predictions, you know, if, if, if this eventually if this newly formed black hole. Eventually supposed to as a black, black hole returns to the disk assuming that typical cake, it will, there'll be a flare one and a half years later. So, I wouldn't want to convince you that we have this. It's interesting. That's what it can do if you have a large machine and the proof is very simple with this has to happen. Many, many, many times sciences all about repeatability and with repeats and repeats and repeats, then we have confidence increases appropriately. Okay, let me wrap up. We're coming on the hour. So, you have a very large project. Perhaps more organized as a high energy physics project. This is more as an astronomy project is really basically a lot of young people who come and go graduate students and postdocs and the grads, and some collaborators so here's a group at Caltech. The young people are, the very young people are grad students, slightly youngish people are postdocs, and the old ones are the management which is on the top. And I just want to say that ZTF has been fun, it's been producing a lot of science results. But it's also, as I use that as a recruiting tool for my students, it allows you, it could make you rich. It's making it's going to make you famous likely, but can also make you rich so there's some young people like me hi she's a data scientists who's joined many of our young people are doing a lot of data science, because that's a natural counterpart to ZTF with all the data we're getting. And so here's an example. So I recall that I mentioned Josh Bloom who put in machine learning in 2009 for our candidate identification after image subtraction. Well, after PTF finished here quickly from Berkeley, and founded a company called wise dot IO, which then few years later, it was, and he had a prospect is said that we use machine here that he has experience using machine learning to solve astronomical problems every night. And so that that means they could solve early problems every day. Anyway, the company then they sold it to G and Josh is doing very well it's a very nice house in the hills of Berkeley. So I gave a talk at MIT in 2016, just before ZTF was going to start. And there was a young man in the audience he came and told me that the stock was on the automated discovery of the universe. And that led him to set up a company in Boston and Bangalore called Soroko. And this is about automation and software. Yeah, great. But when you go public, you know, maybe send a few percent of those stocks and fund ZTF. Okay, so let me stop here and take questions.