- Our next speaker is Kellen Lawson.

- Hi, can you hear me?

- Yeah, can you share your screen?

- Can you see that all right?

- Yeah, and Kellen will talk about high contrast polarimetry as a compliment for total intensity circumstellar disk imaging.

- Thanks so much. So yeah, just jumping right into this. This is of course work that I did not do all by my lonesome. There are a lot of people helping out with this, but I'd like to, sorry, I'm getting an annotation request. Sorry. So this is work that I did with my advisor John Wisniewski, Thayne Currie at NASA Ames and with a whole lot of other folks. So to give some background quickly, a circumstellar disk are disks of gas and dust around stars, which provide the material for planet formation, right? So by imaging these disks, we can identify morphological features such as gaps or spirals, which can provide clues about the presence and location of yet unseen substellar companions. So with cutting-edge observatories XtremeIO systems, et cetera, we can now even recover the signal of the planets that are inducing these features as in the famous case that PDS 70 system. So by combining these capabilities with integral field spectrographs, which provide images in an array of wavelengths simultaneously, we now have the equipment to learn a great deal about planetary systems. So the key challenge for direct imaging is the elimination of the much brighter pattern of diffracted starlight. Also known as these stellar point spread function or PSF. So available techniques such as Reference Star Differential Imaging or RDI ultimately cause a portion of the circumstellar signal to be a erroneously removed with the starlight. And so in order to be sure that they detected point like features indeed a planet rather than a clump of disk material, we might imagine using IFS data to compare the spectrum for the candidate, with the spectrum at other nearby locations on the disk. Unfortunately, signal attenuation is neither respectfully nor spatially uniform. And so while this signal loss can be sort of crudely approximated with forward modeling for geometrically simple, usually older disks, there are new such techniques for young disks with complicated morphologies. So Polarmetric Differential Imaging, or PDI provides an excellent compliment for disk focused studies by providing unattenuated polarized intensity imagery of disks, as opposed to the total intensity imagery of techniques like RDI. However, since self-luminous planets are only weakly polarized dust, PI isn't generally a great option for a dedicated planet searches. And so currently the capabilities of our observatories are suppressed by limitations of our post-processing techniques, loss of signal, and total intensity, post-processing prevents us from identifying and setting the youngest planets and also from conducting detailed characterization of disks as well. So to get a better idea of what's going on here and causing us attenuation, let's consider a simple case of a target image containing a stellar and circumstellar signal. So the four points here just to be clear are intentionally induced artifacts for calibration, but within this complicated pattern of diffracted starlight here, you can see the pattern of our disk. So the goal of PSF subtraction is to remove the starlight and just be left with this disk signal. So in RDI, we image a second star that lacks circumstellar signal in order to isolate the pattern of the extracted starlight. Since the diffraction pattern changes over time, it's common to collect multiple reference images and to then combine them in some fashion to optimally reproduce the pattern in our target image. So we'll call this sequence of reference images R. In conventional RDIPSF subtraction, we'll construct M our model of the starlight and our target image I, by combining the reference frames in some way to best match our target image I. So looking at our resulting model and then the final residuals, we didn't see an issue. So M is too bright because we're fitting starlight to start clutch disk light. And so ultimately this leads to loss of our circumstellar signal and significantly negative background fluxes in blue here. And so this is typically called over subtraction. So I've also highlighted a very questionable point source that emerges along the edge of the disk here. And so to solve this attenuation problem, I've been working on a technique that I call a constraint RDI in which we use available information to try and mitigate this attenuation via over subtraction. So in polarimetry constraint, RDI or PCRDI, the idea is to leverage the unattenuated polarized intensity of the disk from PDI data to make an estimate of its total intensity counterpart in our RDI data. So to do this, we just need to approximate the shape of the scattering surface and its orientation for the disk. And so this can be based on literature. So sort of system information and literature regarding the disk or you can also directly optimize for this scattering surface. And basically using a scattering surface, we can get an estimate of the fractional polarization F pol not as it's indicated here, which is just the ratio of polarized flow intensity, and then derive our total and intensity estimate from that. And from there, we proceed as for the standard PSS subtraction with one exception, which is rather than computing the starlight model by comparing our reference images to our target image, we compare a reference images to our target image from which this disk estimate has been subtracted. And so then subtracting the resulting model from the target image as sorry as above, we end up with our residuals here. And so compared with standard RDI, we end up with a much higher quality image, and the question will point like a planet, a candidate has been correctly removed. So to provide a more complete and probably interesting example for PCRDI, a simulated an IFS data sequence with a spiral arms disk and an embedded planet as well. And perform PSS subtraction using both a standard RDI and optimized PCRDI, as shown here in the images, which have been averaged over the wavelength axis. Sorry, I should say this is a synthetic IFS data. So comparing the results with the input disk at each of the 22 near-infrared wavelengths gives us this where for each wavelength image, as indicated on the y-axis the distribution of the change in the signal from the input model is shown. So for RDI, we can see that it varies wildly from wavelength to wavelength and throughout the images while PCRDI stays nice and centered about zero with just a one to 2% change typically. And so this is quite exciting using PCRDI this way we can effectively eliminate attenuation across all wavelengths for IFS data. And so effectively a PCRDI lets us do some interesting things even for complicated disks that we can't forward model. So we can assess the validity of any detective planet candidates using spectral information, helping us to better understand the true distribution of exoplanets and also study some very young and very interesting exoplanets. And then also by having a higher fidelity, multi wavelength imaging of disks, we can study local disk properties. And also since we can recover the very faintest extensive disks, we can better assess the presence and structure of exozodiacal dust and how it might impact planet detection. So very quickly here, so to help verify a real embedded planet candidate, we also applied this to observations of a young disk system from the Subaru Observatories CHARIS IFS. So here is a comparison of the results with RDI on the top and PCRDI at the bottom, again, average or for the wavelength axis. And so the improvement with PCRDI is immediately quite dramatic and in both results, we can see the candidate as a slightly extended, not of a mission to the south of the images center. So by combining CHARIS channels produce near I, R, J, H and K band images with the PCIRDI result, we can make a qualitative assessment of the difference in the spectral energy distribution at the candidates location. So here, you can see hopefully, that it appears that the candidate stands out quite a bit more at bluer wavelengths, which is consistent with a more detailed spectrophotometric analysis conducted as part of this study. So PCRDI also enabled analysis of fractional polarization maps, where we see a deficit at the candidates position indicative also of an admitting source of some sort. So combined with a whole lot of other observations analysis led by Thane Currie, PCRDI helped us to confirm this candidate as one of a very small number of imaged embedded protoplanets. And excitingly, this result provides some of the first definitive evidence of gas giant planet formation at such wide separations around 100 au rather than formation and subsequent migration outwards. And then also based on the young age, the wide separation and the presence of spiral arms, this may also provide evidence of giant planet formation by disk instability which is equally exciting. So more recently I also stimulated some upcoming JWST cycle one observations of the debris disk system HD1A647. So here I'm using a variant of a constrained RDI called a model constrained RDI where the PI based disk estimate is replaced with a purely synthetic model. And so this is shown at the bottom here, this constrained RDI result. And so compared to both nominal options plan for use with JWST data, so classical RDI and clip RDI as implemented in the pipeline. Constrained RDI seems to perform quite a bit better and especially at small separations. So the results are fairly comparable further out, at least in the case of classical RDI, but constrained RDI really uniquely lets us get a look inside of the inner edge of the ring like disk here.

- Kellen you have two minutes.

- Got it, thank you. So further, thanks to the prolific PI disk surveys, PCRDI can also be used for mini JWST disk targets using archival PI imagery. And so this is actually relevant to both of the early release science disk targets that I'm aware of, especially for one of them, I think this will be a really cool application. And then of course this will also be applicable to data from the Roman CGI. So since Roman has an onboard fuller metric imager, PCRDI will be a really interesting application here. And so with that, I will conclude, and I'll just leave up some summary of the big sort of science questions this technique can help us with. And also any animation here at the bottom showing how PCRDI optimize sorry, how it looks as we proceed through the optimization. And thank you very much.

- Thank you very much, Kellen. Let me see if we have questions yeah. Working, I read about the PCRDI, can you provide some reference.

- Yeah. Great question. So I'm working on it. We haven't published it just yet. We're hoping to get a letter submitted in the next couple of weeks here. But feel free to reach out if you have any specific questions or, you know, want to know more.

- And we have time for another question. Do you require that the bandpass of the polar metric images matches that of the conventional images?

- Right, so it certainly makes it easier. So typically I also let the sort of the peak polarization be a free parameter. And again, this is just intended to get an approximation. We're using a very simple, smooth scattering surface approximation, which is clearly not perfect for spiral arms disks, but kind of the crux of it is as long as we can sort of roughly approximate the total intensity of the disk within some region of interest for the purpose of sort of purifying the target data for building the PSF model, then it works well enough. So as long as there's not giant morphological differences between your two wavelengths, you should be okay.

- Okay, thank you very much. And it's time to move on to our last talk of the session. Before we do that, let me remind you of the poll for the extra pack 26 state in location. The link for the poll appears on the chat and the web is stat.