 Allright, I think we can start. I think all of you knows about the great experimental effort that is put into looking for premord, the gravitational waves, In ne sem bila, da nekaj se vse kompleme, da si se vse občajno bil, da je občaj, da je toga skupnja na zelo virum. Zelo, kako smo priješli, da je temozaj, da se vzelo priješljene prvnoj, prvnoj, prvnoj. Vse smo zelo nače všeči, čaulyn Kuo z Stemforda, weze, one of the principal investigator of this collaboration and his title is search for inflational regetational waste so we'd be modularization, device program and beyond, thank you. Thank you, Paula. So the key point here is not only the gravitational waves as you know have been discovered by LIGO recently but we knew that it existed even before that. here is is we trying to look for the primordial gravitational wave that came from the beginning. So it's kind of interesting that if you look at the edge of space which is black hole, you use gravitational wave, if you look for edge of time, you look for gravitational wave as well which is the thing that I am talking about. So this is just a coincidence maybe not Ne, ne možemo. OK, tako, as you learn from the summer school that cosmic micro background is this interface between opaque and ionized and hot, dense epoch of the universe to the transparent neutral universe and that interface looks like a firewall in all direction that's opaque to photons. So it has a typical 2.7 Kelvin black body radiation and there's tiny temperature fluctuation on it in different directions and it's great. You can learn lots of information from it but it's also a firewall that you can't penetrate. You can't look beyond that wall just using photons because it's opaque beyond that point or the universe was opaque earlier than that epoch. You don't see any of that, OK? Except you need a messenger that interacts very weakly with all that hot, dense medium, which is gravitational waves, which of course interacts very weakly because gravity is so weak compared to other forces. So in principle, by using gravitational waves you can study the beginning of everything and that beginning lots of theories have this interesting feature of rapid expansion which you will hear about next week known as the inflationary period. So we don't directly try to detect the gravitational waves like the LIGO guys do. Instead we look for electromagnetic radiation from the surface of last gathering, the cosmic microwave background radiation. So we look for an imprint of gravitational wave that cosmic microwave background polarization, OK? So we look for microwave linear polarization, we construct a pattern and that pattern could be an imprint of primordial gravitational waves. So this is a microwave experiment looking for gravitational wave in particular linear polarization, OK? The pattern, it's not circular polarization or anything, it's linear polarization creating a pattern of a curl on the sky. So you're going to hear more about this but inflation is a process that inflates space but also the quantum fluctuation in space itself for any fields that you have around that time. So for the infiltration field itself that was responsible for inflation, that fluctuation was inflated into density perturbation that you now see in microwave background or galaxy structures. So that part is verified to extremely high precision and it looks just like inflation, very similar to inflation. So the quanta of gravity, it's not all zero and empty everywhere because everything has to be quantum mechanical, is inflated by inflation to create a background of stochastic gravitational waves. That's what we're trying to study here. So the two processes are very similar. In fact, they're described by very similar equations. They describe the amplitude of this perturbation or amplitude of tensor perturbation, very similar. And that solution looks, for a particular Fourier mode, that solution looks like this. It has a decaying Fourier wave thing and then the amplitude is frozen at horizon exit when the wavelength of that mode is comparable to the Hubble length at that time. So beyond that point, the wave becomes frozen, the amplitude becomes frozen until the universe expands, until inflation ends and the universe expands to include that wavelength, naturally to include that wavelength. Next week, you're going to hear more about this. But for a given Fourier mode, for a fixed wavelength, co-moving wavelength, so that mode exits the horizon during inflation, the amplitude is frozen until it enters into the horizon. So it leaves first and enters, leaves late and enters early. So that picture here, as the universe expands. So this is different wavelength, different Fourier modes. So as you recall, Cosmic background is a snapshot at a time. So you're just seeing all these Fourier modes. So some of these Fourier modes are beyond the horizon, some of these Fourier modes are already inside of the horizon. So if you flip this picture this way, you see something like this, and this is that interface, this is the horizon scale. Beyond that, it's everything, all the modes are outside of the horizon, and this is all inside of the horizon, already processed by ordinary classical physics. This is an actual measurement made by Planck experiment of ESA. This is a temperature power spectrum plotted against theory. So this must be one of the most successful physics experiment out there. That verifies a theory. So this is an actual measurement, and the spectrum looks like this, just like the theory predicted. So we're getting serious about this picture. This whole mechanism, generation of perturbation and the processes that involves the perturbation. And the tensor part, or the gravitational wave part, is a natural byproduct of all of this. So this is all real. So now we just don't know what the amplitude is of that tensor component or the gravitational wave component. So I'd used the two words interchangeably, primordial gravitational waves or tensor. So the amplitude of the gravitational wave tells you the energy scale of inflation because it tells you the exit of horizon, that scale that I was telling you about. I'm not going to derive all of this next week, somebody else will. So inflation field range during inflation, the field range during inflation, whether it's greater or smaller than the Planck scale, depends on whether r is greater or smaller than roughly 0.01, plus or minus an order of magnitude, have an order of magnitude. So if you measure r to be 0.05, the infotain field moves more than a Planck scale during inflation. This is just from the integral of the inflation equation, assuming a single field is slower. So you need to have quantum gravity in order to have any perturbation. So gravity has to be quantized. So if you see classical gravitational waves stochastic on the sky, just like when you see this perturbation came from quantum mechanics, this has to come from quantum gravity. And if you see a scale invariant tensor, you're just measuring the Hubble scale during inflation, so it'll be very hard to not say that the Hubble scale or the scales that drive the expansion of the universe, the energy density, remains constant during inflation, so that's almost the definition of inflation, if you will. So that's why the search for this B mode is a big deal because of all these theoretical reasons. Not because we want to discover a very gravitational wave yet over again. So we don't try to directly measure the metric, spacetime metric. Instead we serve for a pattern in cosmic microwave background polarization. So it was due to these guys who proposed that you just look for a curl component in the linear polarization in the microwave background. So this mode looks like a radial pattern or circular pattern. It's called E mode, or gradient mode. And this curly pattern is known as the curl mode, or the B mode, analogous to electromagnetism. So polarization pattern, as you know, blue sky is polarized by radii scattering. So radii scattering is nothing but just, you know, E and M wave moving the, you know, charge in the molecule, air molecule around. So it's very similar to Thomson's scattering which is just free electron version of that same thing. So if you take a polarizer and map that blue sky, what would you get? So the only source, active source is the sun, which is unpolarized largely. So the sky is polarized as you know whoever had sunglasses. But if you just go one step what would you see? You're going to see this. You're going to see a circular pattern around the sun, because there's no other way. You can't create this curl pattern. There's nothing that breaks that left-right symmetry. So that unpolarized light gets scattered by the molecules there, has to be polarized perpendicular to that direction. There's no way you can get this. So this is in fact the polarization that you won't see caused by ready-scaring. So this is a nice high school science project. If you have cousins to supervise tell them to do this. Like looking for b-modes in the blue sky. So just for this summer school we created a little app that demonstrated that. So you have a source and you just there's a ring of free electrons during the scattering and you can linearly superpose these and you get different you can even move the circle that's too much. My undergrad just got too excited about Java programming. But anyway, you see what I'm saying. So no matter what you do you can't create a curl. You just keep adding sources you just add it in third dimension you can't create a curl unless you have gravitational waves or lensing or dust. Somebody is gonna ask me a question about that later. But that's my theoretical introduction of why we're doing this and why is this related to gravitational waves and why this related to inflation. So now how are you doing this? Look, this is an experiment that proves big bang or proves the origin of of someone, everybody should be doing it. Why is it so hard? Like why is it so competitive? Because even though that nice pattern is unique, the degree of polarization defined by the linear polarization power difference is one part in 30 million. So it's usually called an polarized light, not polarization. So if you measure this light it's highly polarized compared to this. I guarantee you. So this is not just a matter of getting the sensitivity to see sub-microkelement temperature differences but also make sure you're not measuring the polarization of the instrument itself to better than one part in 30 million. OK? Usually that's impossible but fortunately we learn a few tricks and each trick can get you down factor of 20, 30 and you multiply these together all of a sudden you're getting there. It's a little hard to just write down in a textbook how you're going to do this from first principles. That's why it's also very fun and if you're sharp in physics, if you have sharp physics intuition you can add another technique to bring this down by another order of magnitude. That sometimes it's kind of opaque in textbooks. So I've been doing this in this place for the past 15 years this is South Pole and how far away of this from the South Pole itself is directly at the South Pole we walk by the pole of the South Pole every day to get lunch out in the cold. So this is a great place to do cosmic micro background because it's actually a desert so South Pole is the driest place on earth not the Sahara, not the Mojave desert here because even if you have 100% humidity there's no water because the cold temperature just can't sustain any water vapor in the atmosphere. Also it only had one day and one night. That's how we trick our winter over to observe for us. We say you observe for one night only one night. But then he's stuck there for six months. There is great logistical support from the United States NSF Office of Polar Program. So the flights are military flights I had to breathe to the commander once to tell her it was a woman what we were doing it was the commander of the entire US Air Force. I had to breathe to her what we were doing at the South Pole using her fleet once. So this is the collaboration in a collaboration meeting about a year ago. It's getting bigger as we're building more and more receivers and we're running into foregrounds and other issues in the analysis. So we're including more and more people in this collaboration but it's not LHC size yet hopefully. So this is kind of the evolution of the project since we even skipped bicep one so since bicep one was 2005 we started doing this at the same place using a similar technique. From bicep one to bicep two we implemented from this traditional horn coupled detectors like the Planck satellite has to follow the graphic superconductive detectors and then we copied that receiver five times over to formal array we call kek array and still observing now bicep two is done by the way bicep two has been replaced by bicep three in the year 2005 they're all very similar in fact they're very different because bicep three is several times bigger so that's the challenge the big step that you have to go so bicep three has been observing since 2015 just this past few months we fully populated the focal plane and then we're moving to an array of bicep three like telescopes starting hopefully in two years we call bicep array there's an extra three in there so these are all refractive telescopes used similar to the ones used by Kepler long time ago, not far from here or Galileo so these are microwave telescopes using ceramic or plastic lenses two lenses and the detectors are operating at 0.25 Kelvin low temperature detectors it's similar to a typical tabletop the size is also similar to a tabletop condensed matter experiment except we have to open up the window otherwise you don't see anything and 100 watts of radiation is just going in and you have to reject all of that before you can cool the detectors down very different sets of challenges just from there so the detectors are polarization sensitive like I said earlier we're measuring linear polarization which is defined to be the E square the power difference between this polarization versus that that's Q rotate the whole thing by 45 degrees that's U polarization so we have two sets of antennas one for each polarization and we have two detectors that monitors the power entering those antenna and you take the differences the telescope scans so you're constantly taking the difference between two polarization as the telescope scans and you spin the telescope around you move it up and down to form a map so the actual detection of power is done through a superconductive thermometer it's a superconductive film bias at the transition from normal to superconductivity so that makes it a very sensitive thermometer so that monitors the temperature of the CMV light coming from the antenna getting dumped into this resistor heating up the island so you're just measuring the temperature of that island to give you an idea what's the power coming through from the antenna one for each polarization and you take the two the difference of the two some glamour shot of these detectors and then you just go through map making to make a polarization map where at each pixel you have a length and a direction that's the definition of linear polarization and all of a sudden, so this is bicep 2 by the way all of a sudden, and this is the actual data all of a sudden you see this is mostly emode already so remember that either circular and indeed, if you just go through linear algebra and project out the emodes as though you get all we measure is mostly emode already that's a huge success for that nice picture of CMV filter and scattering light has to generate emode but we're doing something harder we want to blow up the scale and see whether there's b-mode left and that's what we did and now the scale is at 0.3 Michael Kelvin on top of 3 Kelvin 3 Kelvin this is Michael Kelvin 3 Kelvin microwave background and on top of 10 Kelvin emission from the atmosphere and another 10 from the instrument so you're trying to measure the temperature difference of 0.3 Michael Kelvin over 30 Kelvin very little difference so that's why you need very sensitive detectors and that integration took 3 years so you have to just integrate out the noise to get the map so by so to saw a lot of b-mode polarization you've heard about that and we threw in kek data taken through the end of 2014 and this and you just turned that into a power spectrum power spectrum is amplitude is a function of wavelength you're going to hear more about or maybe you have heard about it already so this is multiple moment or just wave number this is about 1 degree and this line is the prediction from gravitational lensing which comes from second order of scalar basically scalar perturbation lens by logical structure can generate a b-mode polarization and nowadays you know exactly what that level would be and this is what b-7 kek measured at 150 gear hertz so this is way beyond what the lensing predicted so people got excited, we got excited so now whether this was actually gravitational wave which had a spectrum that had a peak at around L of 100 by the way, you have to look at other frequencies so in year 2014 after the initial announcement of the detection of this thing we started working with the Planck collaboration so Planck is a salamation that produced that beautiful temperature polarization map it has also polarization measurements across a wide frequency band so the instantaneous sensitivity at each direction each small patch, each pixel is not very great but the nice thing is it has very wide sky coverage all whole sky actually and it has coverage all the way out to 850 gigahertz we're in particular interested in the 353 gigahertz polarization measurement because that has the largest sensitivity to collected dust which is also polarized so we started working with them also in the meantime around 2014 right after we saw something in the signal in the maps around mid early 2013 actually we started to put in two receivers at lower frequency so combining all of that, combining Planck and this low frequency observations which is less sensitive to dust in fact you can already see that this is the galaxy and at high frequency you see nothing but dust from the galaxy even at high collected latitude at lower frequencies like 100 you see less dust polarization and at lower, even yet lower frequency you start to see a lot of synchrotron radiation we don't look in that direction obviously our field is around here we're combining all of that this is still 150 only so this is 95 only so 95 has less sensitivity and dust and if this signal is coming from primordial it will have the same amplitude or it will scatter around the red point will scatter around that higher point, green point you can also form cross correlation between 95 and 150 GHz and this is the result and if you just go through statistical analysis you can no longer say there is an access from for example the 95 GHz and 150 above the gravitational lensing if all of this came from primordial then all of these points will scatter around the same region but instead they all came now so going through the statistical analysis one has to conclude that we've detected a lot of dust because this is the amplitude of dust zero is ruled out to high significance that means we see dust at high significance now there is a lot of dust but we can't say there is a tensor anymore so zero is not ruled out at all after you throw in Planck 353 and Keck 95 GHz data but we started to do interesting inflationary science already so at r of 0.12 0.13 there is the famous 5-square model which just means the infiltraton potential is a parabola so that up till this point can explain everything in the universe so that simple model so infiltraton field has a parabolic potential it produces everything we see today so that's a nice model now that nice model is being disfavored by this result and only by this result in the power spectrum space this is what we got so this is that lensing that one always get and this is the inflationary or tensor spectrum that we are looking for and this is the latest result so in everything so of course lensing has been detected by other experiments as well but looks like we've also detected a lot of lensing at this intermediate angular scale but we're most interested in this low L point which are now upper limits only I should say if you look at the what are these colors brown, orange points these are the most sensitive measurements on primordial beam mode they're not exactly primordial they're crossed between 95 and 150 so it has some dust contribution at 95 as well to first approximation these are the primordial measurements and the size of the air bar is larger than the size of the air bar from bicep kek at 150 GHz so look at how tiny the air bars are for the green points and that's coming from the subtraction of Planck's dust template so we're now limited by our knowledge to to dust but you can get more leverage on dust by going to higher frequency so the next atmospheric window that you can observe is about 220 GHz so we've already collected two seasons worth of 220 GHz data and the map sensitivity is already better than Planck's sensitivity to dust at 353 so pretty soon when we throw in our own 220 GHz data the air bar will start to shrink again so it doesn't make sense to observe at 150 more because we've already seen the highest noise of something so right now we need to minimize the uncertainty introduced by doing the subtraction or doing the component separation which currently came from Planck and we also have rather limited sensitivity as you can tell at 95 are much larger than 150 so we want to add a lot more at 95 and improve on dust template sensitivity to dust template so a theorist will just take a look at this plot and just walk out from the talk you're laughing but this is the bottom line ENSABAS versus our plot everything we've got so far Planck temperature low L polarization oh I guess it's being updated from W map I think for this case lensing and then bisapkac 14 direct beam of measurement so far bisapkac is the only experiment directly pushing down in this direction and that phi square model is already this is 2 sigma so the cyan is 2 sigma so it's at almost 3 sigma and also I was reminded that we've already disfavor most of the convex models which some people like better than the plateau type models so this is where we are throwing everything so we just put out another paper yesterday and that was about lensing so like I said this part between L of 150 and 350 we've detected a lot of b-modes that's definitely not dust and that looks like lensing it follows the lensing power spectrum and it has the right amplitude and we know that lensing should be there so this should be lensing but we went ahead and do the analysis so there is a method that they can combine e-mode polarization this is the e-mode polarization measured by bisapkac in green and the b-mode also measured by bisapkac in blue dash line compared to simulations ok so from the e-mode and b-mode so those few points are now known to be dust dominated but we just cut it off at L of 150 and use this region of b-mode and all of this e-mode to do lensing reconstruction and you're going to hear about that next week or you heard about it a long time ago you can reconstruct the potential field using e-mode and b-mode or if you just throw in higher order statistics all of the cmb measurements and you can create a power spectrum for the convergence it's very similar to weak lensing in fact a weak lensing and you can calculate the power spectrum this is a many segment detection and the interesting thing is you can now derive the expected lensing amplitude solely from either the power spectrum the b-mode power spectrum which just tells you about the amplitude of v-mode it has nothing to do with lensing or anything you just guess it's lensing so you can derive an amplitude from bk bb the auto power and you get a value that's close to lmcdm within uncertainties and you can go ahead and do the lensing reconstruction and you can calculate the lensing derive lensing amplitude directly and it also agrees with lmcdm so I think it's lensing but it's not news to you perhaps that this is indeed lensing but the interesting thing is if you want to have a theory that can predict b-mode power at the beginning to the room is getting smaller because the stuff that we measured not only had a lensing power spectrum but also had lensing statistics so the good agreement between these two numbers tells you how much room you still have if you want to squeeze in an alternative model of b-modes like string generated kizer step in effect type thing so where are we going with this so obviously we want to add more and more receivers to get more and more sensitivity I already show you this so this is that big jump we have to made from bicep 2 to kek if you don't put things with the appropriate scale you don't know what we have to go through over the past few years so and that incur a lot of challenges infrared loading we have to develop all new infrared filters that are completely transparent in our band but completely opaque hopefully in the infrared we have to build bigger lenses entire reflection coat them so that they will still work at 4K and we have to completely redesign the focal plane so that the receivers now take modular focal modules in pictures taken over the past 2 years I guess this is only a few months ago and finally we have a fully populated focal plane all at 95 gigahertz so if you go through the math you realize in terms of power at 95 gigahertz it has 6 times less power sensitivity to dust if you normalize the C and B unit so whatever bicep 2 c's if the dust component will automatically reduce by effect of 6 at 95 gigahertz and we have already confirmed the sensitivity of bicep 3 is at roughly the Keck array level so it's running at 7 to 9 so that means we are completely on track so this is our sensitivity at 95 we had no sensitivity before we added 95 gigahertz receiver we were just integrating away cility at 150 gigahertz and then it would say we have to add 95 and we started adding that 2 years ago and now bicep 3 is on track to follow this slope down to catch up with our sensitivity at 150 gigahertz with much less contamination with dust and we also added 220 gigahertz which is harder on the ground because the atmosphere is bright at 220 gigahertz is a lot of water vapor but we did that at the end of 2014 and you started to see the line coming down and we are here, midway through here so this is helping this effort because there is enhanced sensitivity to dust template and like I said even at this point our sensitivity to dust is already better than Planck's sensitivity to dust at 353 gigahertz just a brief summary here and this is my fighting chart in a review article by Kamikowski and Kovetz that came out a few months ago so now very roughly r is limited to below 0.1 0.07 we like that 30% but as a theorist you say it is less than 0.1 so it was published in a series of papers in PRL over the past two years the latest limit is 0.07 there is two sigma so one sigma is around 0.035 or 0.03 we got a little bit higher than one sigma then two times one sigma and all of that sensitivity that uncertainty is dominated by a lot of that is dominated by our mediocre sensitivity at 95 and also dust template our sensitivity to dust template from Planck so between 1 in 5 years we are going to keep observing with bicep 3 and Keck receiver at 220 gigahertz and then we want to add more receivers the size of bicep 3 so over the next 5 1 to 5 years r will be like 2 sigma 0.01 0.02 will be easily achievable with the demonstrated technique and then if it is not found there are two big efforts one is a satellite mission called LIBORD it is a Japanese space agency mission with US contributions to detectors and there is also so called ground based CMB S4 effort led by the US department of energy office of science it becomes a lab project like a high energy experiment project experimental project both missions are targeting 10 to the minus 3 level 1 sigma r 10 to the minus 3 so one is a whole sky survey with 15 bands in space but because it is a space mission and also a cheap space mission it does not have a lot of resolution angular resolution so it can only see up to this L meaning it can't use information from lensing to d lens instead you are trying to take advantage of this little bump at very low L, very large scale and in combination you get and also the fact that you can get access to many many frequencies in space that you can't on the ground so stage 4 CMB can't get access to that very low L thing and it can't do a whole sky survey so in terms of tensor it will focus on a smaller patch go very deep and with arc minute resolution to do d lensing so if you can reconstruct the lens you can predict what the beam mode you will see at a degree scale just simply by lens the email polarization now measure to hundreds of sigma so now this is a map to extremely well sensitivity you just take your lens you lens that E mode and it becomes a template for B mode that it can in principle subtract from the directly measure B mode, this is called the d lensing technique these two complementary approaches in principle should both reach one sigma r of 10 to the minus 3 by mid 2025 about 10 years from now so let's look at this fun finding chart so there is inflation also affects the scalar perturbation and the spectral index for that perturbation so under simple single field slow row inflationary models a constraint on NS will lead to a constraint on r for a given model and these are all the very interesting very well motivated models and this current constraint on NS translates into three sigma bounds for each of the models on r so for example there is a phi square phi square our two sigma is 0.07 so this is this is really getting ruled out at this point so this is two sigma getting disfavor, heavily disfavor so the next ones to go well I don't imply that they are not great but if they are not the reality the next ones to go are these ones these will be heavily tested very soon and then there is a class of model that predicts a few times there is a central value best fit or lower limit so if we can get to that range with these guys then you might be able to detect this even if you do a little better than this target or if this value is a little higher but if you want a 5.7 detection of this particular Steribinski type model it may be hard but you can pretty much rule out the rest even before this experiment turns on ok so this is a 15 bands in space planned by Lieberg with one sigma r of 10 to the minus 3 including all the systematics in foreground uncertainties so 2025 ok so CNBS4 involves bigger telescopes they will map that lensing tail really well 500,000 detectors on the ground again mid 2020s so in addition to r it will measure the number of species of relativistic particles like neutrino-like particles and absolute mass or some of the total mass of neutrinos to 16 mille EV 0.016 roughly ok, i'll just put up my conclusion here and i'll take questions thank you