Okay, We only have a few people in. Today is October 25th. This is the Astro seminar, which we're doing is a hybrid form today with both in-person and people in Zoom. And so I'll let Sallie introduce the speaker. We do have another talk next week. I will send will send out notice on that later this later this week. Okay, Sally, go ahead and introduce in book. Okay. So I'm glad to have inbox Yacc giving us a seminar today. He has associate's degree in chemistry from Delaware Technical Community College and a bachelor's and master's degree in physics from this department. And he is about to finish his PhD that will be coming up in less than a year. And so this is getting ready for going on the job market and showcasing the really cool results. But he's gotten. So go ahead. Thank you. Sally. I'll be your switch, a slide advancer. Just tell him, okay. Okay. Alright. So this positional kicking everyone, we can probably do it. Unless you, if you want to advance the slides yourself, then you can just talk more directly into the mic. I'm just wondering if everyone can hear me. Can everybody hear online right now? Everybody hears. Okay. Okay. I think it's fine. Okay. Okay. Alright. So hello, as cell introduce and my book here and today I'm giving a talk on the GPi observation of HD 15115. Now before we go ahead and let's just define this a little bit, right? Gpi stands for the Gemini Planet images. So he said ground-based observatory designed to observe planets and degree this. And we'll go more into that later on. Okay. Secondly, here is HD 15115. So that is an F star. And the reason why we are concerned about HD 15115 is because it has something called the debris. And that's what this would be a bar for the rest of the slide. Let's see. It doesn't seem to be advancing just a second. So first of all, I put that at the time, the British. So some of you guys might be wondering why the previous guess. Well, at the desk, basically, a ring of dust around the star. Dust particles can be a wide range of size or magnitude range. Ranges of size from millimeter all the way down to like Michael, Michael. And basically they're composed of silicate material, but as you get farther and farther away from the respective start, they can start. Ice becomes a more and more important part of the component. And so here you can see a picture of a degree this, the 2D grid is actually one. On the left, is called Omaha, right? This is scattered image. Picture of that the blood is taken by Hubble Space Telescope. Now, down on the right says FOMO, how can it be? Just sort of ignore that because I think subsequent studies have shown that it's fading away in a, maybe it's caused by some correlation. But the reason I just wanted to show this image is because of how striking it is, just like it's really there. I saw dust, but it does scatters sunlight for the viewer, but it also takes a lot of processing in order to observe that image, as we'll see later on the one on the right. So dipole does, so that is our inner solar system. So this is a picture of the debris dust of our own solar system. So our own solar system tab degree this also. So that represents like the warm component of our solar system degree days. So diabetes can have warm components, so components that are closer to this time, they're warm and then all the components which would represent us around the kiper belt. Services. Next slide, please. Excuse me, probably going is that is that a optical image or infrared space? So it should be optical, optical. It's going on here. Okay, So there are two ways to be a deep way. This one is direct imaging, which we saw with the former whole object beforehand. And this is what I will be going for a steel 15115. Well, it has already been done more like characterizing, but I'm going to use direct imaging in order to find out things about about HD 154 months bias degree this. But there's another way of telling whether that staff has degree this. And you do that by finding out what Something called photometry axis, right? So what you have here on the plot is the model SEB. So as it is a spectral energy distribution is the flux of this star alone, its various wavelengths, right? And so the green points there are the ethical and near infrared data which I fit, which you fit into a stars, the spectra, the models spectral energy distribution, right? And so depends. Then on the right you have the infrared for mid infrared and far infrared data. As well as you can see. Even though there's an excess above what is supposed to be, right? Because what happens is the dusk at those wavelengths absorb and emit the light at those wavelengths. You see x is above what is normal. And so you can detect that with the with infrared data. And then maybe you can model them based on based on some SAD, maybe like a black body or some models like bad black body type of thing. So let's go a little bit on. So first of all, there's the, what is the main driver of the predicting evolution. So first of all, is thus generated by collision. Okay, So if, if they just weren't generated by collision, the dust would basically be removed really fast. Okay. Because of point pointing Robertson drag and radiative pressure, which I'll explain that later, right? So you need that collision to happen or to generate all this dust. So that explains what we're observing, which is that there's still dust after however, under millions of years, right? And usually the way it happens is like volitional cascade where two objects at each other and which passes perfect collision and collision which runs down the size of the of the material. And then the second aspect is that does, is removed, right? So that either blown away by radiative pressure because as you know, radiation has momentum and that momentum is transferred onto the dust, which then turned, blows it away. And so that pressure is determined by basically the gravitational force own the DSO. So basically how closely the two star and how strong that radiation pressure is. And then the other thing is pointing Robertson's drag, which is this one is a little strange, but it's what happens when, because let's imagine that does going around the star, right? And some star-like hits the bus, right? And because of the fact that the dust is moving perpendicularly, right? When absorbs incoming starlight, its momentum shifts. But wearing it, what do you call it, emits the starlight. It does so isotropically. So it does it, even the momentum exchange takes place and it slows down the dust and in the tangential direction. And so it starts spiraling inwards. Here on the bottom you see why it's 2008 chairs plot of a further access, right? And so as you can see, the model starts with infrared excess. Amount goes lower and lower as a function of age. The other aspect we are, we are concerned about is thus distribution, right? So we have something called the I taught before previously about collisional cascade here, right? So first of all, the important relationship here is that the number density with respect to the size of the dust is proportional to the size of the dust to the power of two negative of some power, law, power basically. And so, and so here on the bottom you see, in the middle you see something is called benign 1969. One of the first calculation attempt made to determine this does distribution and key basic and Dohnanyi paper basically assumes, well empirically derived that relationship because they basically take a bunch of observations of asteroids and being said in terms of magnitude and does a histogram and sort of assumes that these magnitudes are correlated with the mask on the linear part of the histograms as okay, we do a linear fit and we know that the The magnitude is in log scale, so it's a power law of sorts, right? And they, and that paper does a bunch of calculations derived at a value of 3.5. Now, what is the empirical value of 3.5? Value? Bueller? No. Is it pretty close? It is pretty close. Yes. As you can see, in my Greg or 2016 does a bunch of observations on a variety of stars. So on the left you can see the name of the stars that they did observations on. And doing that, they found what the Q value is for each of those stars, of the debris, this of those stars like. So as you can see, it tends, the average is around, is, tends to be lower than 3.5, closer to 3.3. And that is closer to the left side of the models. So there you see the range of models, the cues that various models, modelling schemes give. And those models tend to include. Those models tend to include like very like tensile strength of an object. The gravitational interaction between the dust and things like that they take to it. There are more complicated models and David, numerical calculations. The models on the right-hand side tend to be more on the side of basically making objects for velocity depending on the mass of the dusk. So that's, so on average, we say that, okay, the first distribute number, does number distribution is such that a to the power of minus three points. Okay, thanks. So now we want to know what the optical property is, right? So first of all, there's the forward scatter aspect of the startup, of the star-like concert, the bus and then interacts with it, that does interacts with this thought and redirect, star light and redirect the starlight. And basically the redirection tends to be forward, right? But not perfectly. Of course, there's also some back scattering, but in general debris this scientists like to model it in two ways. One is the sampler is called the HENI green stack function. You don't have to know the equation, but what it does is basically give this like ovals, forward scattering, sort of shade to the scattering, right? But usually it's not super great. So I've seen papers that use some linear combination of two n equals sine function. Then the second way the modelling goals is b me scattering, which is basically you apply electromagnetic theory to a spherical particle and you do all of that. Then the second part is the absorption and emission of the starlight, which happens around mid and far, far infrared spectra, right? And so what happens is at a certain distance, the particle absorbed the dust, but it, but it has to, but it needs to release the wavelength which is proportional to its size. So it starts heating up right into close to the center of blackbody distribution. And then it releases the startling, right? But that said, it's an inefficient term I mentioned. And so that's why it's not a perfect black body. So what they usually do is, is modelling by something called the modified black body. So in the modified black body, after a certain point, right after the peak, as it goes towards the red end, they change the slope of the of the Rayleigh Jeans part of the function, the red part. The red part is the part that's going down so that it's steeper, basically. Down here. You can see what the various spectra or STDs can look like depending on where did, where did this case. And so when you are closer to the star, you start to see the silicate line light emitting and around ten microns, e.g. and then as you go, and if it does, it's farther and farther away, starts looking closer to a kiper belt or a blackbody type of administrator. Quick question. So is the steeper than rarely genes? Is that because the dust is so small that if you get to long wavelengths, they don't want officially admit that those wavelengths usually that's more like on the light. Sure. So what, what wavelengths do they steepen it faster? The early genes. What are we talking about? Micron. Microns or that's okay if you don't know that, I just was curious. If that was a reason why it was inefficient. That's okay. If you can't do it now, we can talk about it after. Okay. Next slide, please. Okay. Okay, So now the main point here is that debris, this comes with various structures and asymmetries, right? And so here on the left-hand side, we see this, this where hello, range of features like some have narrow rigs, some have worked, some others are swept back like a butterfly way. And so this is the main reason why the British are interesting. They are signposts for some of the various processes behind them. And so the question is, what are some of those processes? So e.g. 11 theory is that e.g. the one with a wing swept back or some waste that is happening because of interaction with the interstellar stellar medium, basically gassing other space. And so as the status, proper motion, a collage of some gas or somewhere and that causes the interacts with the dust and that gas passes a dust to get swept back. There's also called self-service. So what that means is that Pluto's size objects, right? I get formed. And then they cost objects to get sped up, right? And change orbit, which causes collisions. And that usually in the modeling tends to make the grade they sort of widen, right? Instead of being a narrow ring, It's like a fatter, more wider type of brain. And of course, radiation is important because they, like I said previously, they are part of the evolution of the debri, this, but the main thing is planets. Planets causes these gaps. Planets, planets can cause the work. So e.g. here on E, that's called beta victory. They first saw the war. They thought, okay, maybe there's a planet and sure enough there's a planet pictorial that they took a picture of, direct imaging of it. But here's the thing, right? Class can actually explain every feature here positive. And so the question is, really, is it possible to know at all what is the exact cost of all these pictures considering that different models can cause the same thing. And the thing is there is probably a combination of things rather than just one thing and the War e.g. it's hard to explain with just that one, tenant and paste things like okay, maybe there's some other plan is a skipping. Those things stable. But e.g. alright. So same thing with the brightness is asymmetry. You can explain that by ISM. Ism interaction rich lowers the brightness on one side because basically the proper motion of the star is in the duration of the debris. This planet can explain that. So who knows? But at the very least, we have to catalog days and study them further and see whether ng-model them and see what we can extrapolate and possibly detect planets based on those cycles. We go on today. So I've kind of lost track of the scale. So are these typically a hundreds of AU across these pictures? Depends on the stock. Typically around 100 a year. Okay. Yeah, that's right. Okay, so now we go back to my object. Why HD one byte, one byte, right? So that was all of a sudden for me to introduce this problem. Okay, so first of all, HTML5 qualified is an edge on this. As you can see, if you can imagine the ring and you, you know, you place the ring light right in your line of sight. You can barely see the back part of the desk. One thing we have found is of course, that is interesting is that the British containers asymmetry, right? It's not just a perfect circle of brain. And of course we saw that it's not unique toys be 15115, but if they want to buy for my purposes, certain specific types of asymmetry. So first of all, we have the observation with this telescope, which as you can see, there's this bifurcation on the desk. And so that's very interesting. Then on the right-hand side, we have a Nikki observation with Nicky. Does infrared, right. So what you see is that the right-hand side, you see a brighter and you see this brightness for longer than on the left-hand side. Well, I say right hand side and that's what they call the west side. Okay. So from now is called the right hand side, the west side and the left hand side. That is, I know it's confusing, but that is the actual direction. So here you see that the west side has stretches longer, brighter, the left side stretch is short. Sorry, is psi stretches shorter? Now, what we're going to do with my observation is look even deeper inside the inside the the debri, this because as you can see, the coronagraph blocks along the disk. So you can't see a lot within the one arcsecond area. And so g pi, this task I've been talking about, can look with things. And so that's one interests that I want to investigate. Okay. Next. Then secondly, there's my grandma all 2019 the observation, right. So millimeter observation on the degree of this. And basically what they found is that Turing model or a Gaussian got model does a better fit than a single ring model. So as you can see here, so my garden, god, I mean, it's a single desk, but there is a gap in-between that is shaped like a Gaussian curve. And then two races, of course, to debris that's basically one within the other. And so as you can see on the wandering model, you have plenty of residuals left. But with a dozen gap or touring model, there is no residual now, whichever one is, it's true we don't know, but at the very least, those fit better, better fit than the one race. So at the very least, I also want to investigate a second component on my degree this, okay. So now let's go to gemini. Gemini Planet Imager is an observatory that is basically made for observing, like I explained earlier, observing planets and the breeders. And so why has is called at that, there is something called adaptive optics, right? And so what it does is a force that it fits the shape of the mirror based on the, based on the turbulence of the atmosphere. And so one way to do this, of course, we shoot lasers, which creates a wage exercise that some atoms in the sky, which sort of creates like a faint star that they observe. And based on how those data reaction shifts, you change the shape of the mirror. It also has a coronagraph, which allows us to log star life, which is very important because of course, otherwise it washes out the debris. There's some degree these are not very bright, so bearing also does have polarimetry. Now, we're not gonna get too deeply into that because frankly, I don't see the debris this end the poll in image, the stokes image, I make the pulpal, the polarization, but I don't know how to explain it. But you can at least Today's in the total intensity of the polarimetry. I'm gonna get that. But polarimetry is important because if you can do something called scattering phase function, which allows us to better characterize the components. The dust of the debate is like dust distribution or maybe it does size. Then. And then finally, Angular Differential Imaging. So basically it's going to do with this is something, well, let me explain this later, but that's a very important aspect of subtracting noise from the image. So this is something that you go like once the observations take. Let's proceed. So GPI has something called the integrated field spectrograph. So this here represents the inner edge of G. So first of all, do you see the adaptive optics system, starlight coming in? And what we have then it goes to the coronagraph like the appetizer and the mask, right, which blocks the starlight and the central starlight which allows the eye, the rest, the ones around the edge to come through. Basically. Then afterwards it reaches into the integral yields spectrograph, right? And so this is what happens if the light goes through something called a lancelet array. So each of those lines is like one pixel in the image. And then they'd call it made the optics so it spreads out the h of those light from each lane set landslide into a spectral sort of sorts. If you have the polarimeter, polarimeter on, it, polarizes each of those spread out spike train two orthogonally polarized data, right? And so from that you have, for each pixel, you have like 22 spectra next to each other. I always sort of hard to explain, but they're like, there's like a mini spectra for each pixel or two. Or if you're talking about pull, pull, pull mode, polarimeter mode, to Spectra. Okay. So once you have that, that goes through the data pipeline, that is the data reduction pipeline, which, which turns the raw images into usable data cube basically, and you get the actual image. So what does the GPi reduction pipeline do? It does a few things. First of all, it does start subtracting which as you know, that's, uh, as, as, as if you're an astronomer, you know, that basically the response of the telescope, if everything is close down, the telescope itself sprayed the noise and we have to subtract that. They calibrate the wavelength of the dispersion, right? So you have to know what wavelength each part of the dispersion is. Humboldt. You also come over to this person. You were to smooth that smooth spectra and then you do bad pixel interpolation because all the pixels are no good. So you have to give it some valid value. And so if you have to go to the next to each other, you can calculate some value. And then finally, a big important part is obtained. A soft spot that the location of the soft spot and the, the flux of the SaaS port. So what is successful? So they are called satellite spots. And basically what happens is in the US, on the jet diameter, you have this in front of the appetizer. You have this rectangular mesh. And this rectangular match the fracks, the starlight coming into it, and the fluxing into these four different spots. Okay, So something called first-order diffraction. And so on the right-hand side that's before processing. But then that allows us to do is centered the star. So that allows us to say, okay, the star is locked. If you locate each of those soft spots, you can find where the star is. Okay. That's centers in and that makes everything convene. And then the other aspect of this, as for this, you can find the value of the flux, basically calibrate the flux of the image because the data that the image case, they seem some digital analog value. I know a lot of digital unit, something like that. But that's not a physical value, that's just an electronic data. You want an actual physical value. So the soft spots allows us to calibrate that. Next slide, please. So now we go to the meat of my projects. So what I'm trying to do is called angular differential entity. So angular difference differential imaging is the method for speckled noise subtract. So basically what happens is when starlight goes into the instrument, interests with the optics of the instrument that could be better creates barriers noise, and not only that. So the noise also depends on the condition of the night sky. So you can't just have some first, you'll need to have observations of that nice guy in order to know what that noise this, right? So how do we confront this specific issue of subtracting the noise from the images? Here we have a scheme for what we call classical angular differential imaging. So my classical, I mean like the first method original, which is not something that will be used. I used to do it, but I think it's illustrated, right. On the left-hand side with diets labeled a, right is h of the images that has been rotated in order to afford the noise for line. But as you, as you can see on the red dot, the planet is not a line. So as you can see, the planet itself does not add too much to the nodes. So you can take the median of those objects, right? And create a map of the nodes. Once you create the median of the noise, you can see that's supposed to represent the noise of the images. You subtract the media to each of those data pictures that you took. So here you see that C equals a minus b, right? So the image is minus the median. As you can see, it's theoretically it should subtract out all the noise. And then afterwards you'd be rotate the images so the planet or the degree this course are on top of each other. And then you take the meeting of that. Now you get the characteristic of the object with the noise subtracted out of it. And so now you have more signal and noise. Right? Now, the method I'm using is more sophisticated, is called the cartoon and love image projection. And this is based on something called the principal component analysis. And so here you see a set of formula. So let me explain those four. Okay? So z represents the cartoon law basis, right? So is the noise, the noise that is decomposed into a set of orthogonal images, basically an orthogonal basis images. So how do you find that? Well, you do or some of the eigenvector times the reference images. So the eigenvector is C and R is the reference image. So the reference image represents each of those rotated image. Okay. So first of all, how? Now the next question is, of course, how do we find the eigenvector? So you think the reference images, you can find the error covariance matrix, right? Which is basically like the on the diagonal is the error. And then the diet on the off-diagonal parts is like the, you know, the the inverse correlation with each other. And then the third one, you diagonals. You do a diagonalization problems. So you'll find the Eigen. So you do the eigenvector eigenvalue problem, which gets you the C matrix. And the C matrix allows us to guess as the eigenvector, which allows us to get the card catalog basis. Once you obtain the basis, you do a projection. You do a projection of the basis on the, on the target image, right? So then you get, then you get the the sort of like how do I say it? The image clean. Clean? Yeah, the clean image. So you'd subtract that way the original image and two, and we get the cleaned up image with is big, right? And on the right you see that's a gift of basically how many, what happens if you use 11 basis versus 20 versus 50 versus 100. So you can see us more basis. You use, the more noise you subtract off. But there's a price for that, of course, which is that something called self subtracting happens. So what happens then is you erase, you subtract paraffin signal, you erase some of that noise. And so that's not great. So we want to limit the number of KL modes you gives the optimal, is it 50 or 100 or 1500s? Probably too much, honestly, I use five, mainly because if I do ten, e.g. my computer runs out of memory because it's very expensive memory wise if you do a lot of KL modes. But it's so five for me, strike the right balance. Now, what are the results of this angular differential image? Okay, so here we see the ADI applied to spec mode TE one band and pull mode k1 bands. Remember this poll mode is saying total intensity and not in any of the The Stokes image. Okay? So first of all, on the right hand side, you see that the color bar, the colors represent signal to noise ratio. And so I outlined where signal-to-noise is equal to one. And so why I say it's okay, anything above that is the debris this. So I highlight that with green. And then you also see two interesting things. First of all, on the east side of the desk, you see this bump on the K1 band now, but it does not have a full in full mode. So I personally don't think it's real. And I heard from the g Pi t tends to happen with JPEG images. So gotta be careful. Notice it when it comes to noticing that then k1, but it also has the log of the top left. I'm not exactly sure what it is. It could be an extra galactic source, but also I don't see the whole image, so I don't know. But anyways, so but as you can see, I've drawn this and we think are one arcseconds, which is great. Oh, and one more thing. You can see cytosol and subtracting at the edge, right? Because it's darker at where near the viscous. So there's this whole sort of negative values, whole of negative values in that next to the degree this anyways, proceed. First thing I did with those image is fine. What the surface brightnesses. For surface brightness, I just took what the maximum value along each column of the debris disc. I hear I personally normalize the floods. Now I'm supposed to have I need to do flux calibration. So but so far has not worked out. I didn't look good. And so my calibrated the GPi, Rho d Phi, d pi cubes, the image cubes. And so the way I want to do is get the raw data and do the redo the pipeline stuff in order to get the flux calibration. In order to determining any sort of asymmetry, what I did is add up all the all the flux on the right-hand side or the ITO or the west side. And then by me, West right-hand side is west side. And then I add up all the brightness on the east side and say if they were close to each other within the error bar and they were within the error bars. So the way I said I see is that there's no significant asymmetry with being the one part second range of the hip. The second thing I did, sorry about that. That's probably a software issue because we're doing this 100 computer, but what I did was spine fitting. So what, why do we spend fitting is for a median or average of two columns. On the right. You see fit Gaussian vertical profile, right? So the Gaussian, the Gaussian, the calcium profile, of course. Q fit closely with the bumpy signal, right? Presents the debris. This where the top of the thing is, is where I say okay, that is the location of the debris, this in that vertical column. I do that for the average of two columns for h as i 0 along the debris. This and so on the left-hand side is the result of that spine failure. Now, first of all, for this image, I use k1 spec image K14 H spec an H plus so H is around 1.6 micron, K1 is around 2.1, something like that. So so then I also have, you can see, I also have another set of observations That's called Nicky. So that is the muscle you're 2014 data that my soul your himself was very kind enough to give me. And also recently angular came back to me and said that she had the spine feeling also. So I plan to see how that fits into the overall picture. But as you can see at the very least, that g pi observations align closely with the Nikki observations. So that's good. Then. Then with this, we can roughly, very roughly estimate the geometry of the BigQuery. Extremely rough, which is that you find the inclination and then you find the semi-major axis of the degree this course. And I use the Nagy with the GPi together. And I get an inclination of 8,686.4, which is slightly higher than all the other ones. And around 98 AU for the semi-major axis. But I would hold off on saying that that value because it really depends on how much of the leaky data you include. I saw, I saw have to plan out how to how to determine how much of a naked data, including automate the elliptical. So that dark black line is the unless that I use to determine that Rob geometry of the best. And yeah, Next slide, please, Before we go to next slide. So it looks to me like the data, the black line. Yeah, there's more data above the right and more data below on the left. So that strikes me as being some asymmetry that's not significant or just because the error bars are so large, but it happens in all the pixels. Pretty unusual. Yeah, yeah. I'll say I agree with that assessment. So there is some asymmetries there, yes. Like that? Yeah. Yeah. Okay. Now, we're finally getting into the modelling aspect, right? So this is the middle of the project or the meat of the mean of the project. Basically, it's the biggest thing I meant to do, which is model some sort of geometry of the tastes and see what, see what are the, you know, at the very least, they, the characteristic of this degree, this. So what I want to do something called forward, what former modelling S is, we're trying to simulate the noise of the telescope. That all the speckle noise onto the modal image. Here on the left, you see that is the modal image I created and I'll go over the model on the next slide. Then when you then by using the disk m program, that program uses the cargo load factors that was saved by doing the Angular Differential Imaging. And basically as it on top of that model. And that creates the image on the right hand side. So that right-hand side is basically a rough model on what this looked like from the point of view of the slide. What happened there? Okay. It looks like you're missing an image. Yeah. So I use something called a forward modeling. Okay. Sorry, sorry, I explained that. Now, before the left, the left image, you solve the model. I use something called MC4R. So why is it boss? This is radiative transfer model that the weight models of this is that they say, okay, there's this power law in terms of the dust density in the radial direction. And then an exponential, sort of like a Gaussian in the vertical mode. So that is explained the two equations below as you can see. Those. So the R or the part with the r to the minus two alpha egg and then alpha out. That dot represents the radial aspect. Now alpha, alpha a is a negative value. And so that part is actually alpha is the parameter with negative values. So that handles the upward swing and then, and then the positive value. And so that explains the negative, the part that goes slopes downwards. Then. And then below the big HR does the scale height of the, so basically roughly estimates like the vertical extent of the desk. The gamma there you say it's like a flaring value. So how might he extends vertically, layers outward? But for the grid, this usually is supposed to one protoplanetary disk. You might have higher values of that. I said I said gamma equal to one for this. And so what are the parameters we're interested in? We are interested in positioning angle. So what is the angle of the debris disc on the image? We are interested in the inclination, so that determines whether it's Agile or face, arm and mine is closer to edge on force. So that's important. The mass in solar mass, k alpha, alpha, which is the power law, the radio, radio power law are zero is basically sort of like where did this is, the ring is located because the model is basically a ring. Are in an era, determines the cutoff value of the dusk. So we say, Okay. Before, below this value, there's no dust, and above that value there's no dust. And then there's supposed to be a scale height somewhere in there for some reasons being cut off. Okay. So with that, with the modelling achieved and we'll forward modeling, why would we want to do a gets explored the parameter space by using MCMC, Markov Chain Monte Carlo. And so here. What happens is for each point in the parameter space, of course, you compare the model of that, those specific parameters with the data image. You subtract and look at the residuals. And based on that, the, the algorithm decides where the next step is. An NSAID was as though the waters go around the parameter space. It tries to go and find the most probable parameter space. And so here we have the corner plot of the Markov chain Monte Carlo. And so as you can see, it's what I do is of course, find the distribution of the parameter. Alright? And then find the median value, which I determined as the value of the parameter. And the 25th and 75th percentile represents the error bar of that parameter. So they're on the right hand side represents what value I got out of this run. Now, when I tried to do a trade to actually look at what happened. What happens is when I look at the residual. So by residual, I mean I subtract the best model image with the original, with the data image. I do not. I still have an issue with getting a model that gets rid of the dust, the best signal that properly. So for now, I say this is a working progress. Okay. Next slide please. Okay. This is unfortunate, but based on the left-hand side, while the image on the left hand side we present is supposed to be a residual, although in this case is residual, I did buy the manual failing, right? So I told the parameter based on my, based on intuition and we'll watch. My eyesight tells me. So this is not a what you call it a proper feeling. Okay. This is just an example I gave to show you what happens, right? So what happened here? What I did was, okay, I did this rough estimation of the model. I subtracted with the image. And so what you see here is, okay, with my, with my manually fitted object, you see some signal of the day. So blue outline represent a one signal to noise ratio of the residual. While the black outline represents the one signal to noise ratio, 11 signal to noise ratio of the original disk. And so I compare. So if you see, you can see that it subtracted a lot of the signal, but there's still a bunch of residual left. And so my thinking is could this be a sign of a second desk? And I tried to do some sort of preliminary preliminary spine spine feeling. That preliminary spine feeling is resulted the result of wages on the right-hand side. Now, we want something. Now we of course want to compare with the Alma's value, which is of the second S, which is 44-50 AU. But as you can see, you can't lift Spirit gives you here is not quite proper and a lot more work has to be done in order to see if I can characterize this at all. Next slide. So in summary, we use the g part investigate asymmetric, entertain us with being one arcsecond of the HCI 151155 fitting enables one to tease out features and roughly constraint this geometry. Surface brightness measurement reveals noise of nodes. Brightness asymmetry for modelling or losses to replicate telescope noise on physical models. And we find the best fit model using MCMC. And we were going to try and find the residual and then try and characterize the residual. Let's see if we can figure out anything about the second biscuit all. Does it? There's one more slide, I hope just blank. Okay. It's not supposed to be blind. Okay. Go actually, so Stan couldn't get the questions, but okay. Okay. So any questions for the people in the room? So go ahead. So this is the only kind of like sars and mers we want to do. Because there is there like a best group of pest types. People do. Well, we want to do this. Well, there's some big chunk of database of GPI that needs four bonds. I'm just happened to be working on one right there, good worker space. And they have a lot of data. And they need people to help them out. Because they're like characteristics where you see a car and then it's like an evidenced based on list or we say no, we not know. Oh, okay. So I mean, these observations, were all these observations done with g Pi. We already knew that had the ugliness. That's important. All those others have been on study, have been examined. So this is another, this is just another dataset that we're trying to do and try and examine these objects to that. In this case, like I said, we try and do something with big one. So I'm a little bit surprised that one would put so much effort into modeling a case where you're seeing the diskette John, because it seems to me you're losing a lot of information from the integration along wanting to cite it much better to try to focus on those that are more face-on. Is there a reason why? There's just so much there? Well, I guess it's the reason why you focus on this particular object that we had drawn? Well, this is not a particular focus necessarily. Like I said, like I told him, this is an object. So they're on the GPU note that the team leader assigned specific objects with different people. You just happen to take that off. Yeah, it turns out this is a top dataset because of the issue you mentioned. Because because there's not a lot of information you can glean, as you said, from a jar, right? Especially with regards to like especially wearing it sort of cuts off, cuts off at the edge. So it's also hard to see why the answer of the desk is. So the answer, of course, where they sort of myths. What do you call it a middle. Okay. All right. Yep. So that's answer is just that you got the luck of the draw for this object. You're trying to see how much information you can extract out of this case, which is John. Yeah. Like I said, there's actually there's a sauce observation from sphere. Sphere. Actually it's a lot better because you can actually see the back scattering part of the day. You know, the, the, the, the back part a says the, you know, when you tell the donor, you can see the bicarb. So yeah, like GPA is not the best thing to observe this desk with, but it's an addition. Okay. Okay. Is there anybody on the online people who have any questions? Okay. Anybody else? So it gets gym and shuck on and I are the ones here. All right. Well, I think then it's 05:00. And so unless there are more questions, Let's think back again for a nice thought. Thank you. Okay. Take care. Bye. Yeah, Thank you very much. My head. Yeah. That's always the way it is. Yeah, I know.

Astroseminar 25Oct22 - Inbok Yea

From Stanley Owocki October 25, 2022

9 plays 0 comments You unliked the media.

Zoom Recording ID: 7874089078 UUID: cblDCy3GS8Sam1g12lODzg== Meeting Time: 2022-10-25 07:59:11pm

Tags: astronomyastrophysics

Department Name: Physics and Astronomy
Date Established: October 25, 2022
Appears In: Astroseminar

Comments

Add a comment