How Simulated Observations Help Future Telescopes

we're now just months away from the ver Rubin telescope coming online we're just a couple of years away from the Nan Grace Roman telescope flying to space uclid has already begun these three telescopes ground and space are going to be working together to map out to almost the very edge of the observable universe they will help us understand the concentrations of dark matter the effects of dark energy at a scale that astronomers could only have dreamed of in the past and yet this flood of information is going to be almost too much pedabytes of data that astronomers are going to have to turn through to try and make the kinds of measurements and so they need to get prepared my guest today is working to solve this problem for them he's Dr Michael Troxel an associate professor of physics at Duke University and he's the image simulation lead for open Universe now open universe is a gigantic simulation of the entire universe but what Michael and his did was take that universe and then convert it into what is expected will be visible to Reuben to Roman so that astronomers working with those telescopes will be able to get prepared for when that data comes we talk about what it took to build a simulation like this and what steps were required to make it accessible to these different instruments so that astronomers can make some real progress now we talk about that we talk about simulations in general we talk about the telescopes and then we talk of course about what questions that astronomers have that they're hoping to answer and how their understanding of the universe will grow and expand as these new Incredible telescopes come online so enjoy this conversation with Dr Michael Troxel Michael I'm fascinated by these Large Scale simulations of the Universe done on supercomputers when you run a simulation for the first time and you consume however many days of time on a supercomputer and then the simul you know you get ding simulation is complete um what's the first thing that you look for to find out if it's better yeah so when you're running these large simulations on a supercomputer the work starts 6 months before you press go things like the HST Cosmos uh data for example which is the largest space bace data set we have um at these sorts of depths uh so it's it's really an iterative process over many months where we're checking every step of the pipeline and every step of the input uh and then in categorizing what we're getting out visually numerically uh running our pipelines on it but but I wonder about emerging behavior in the universe at the largest scales I mean there's something about having everything put in I know that you know simulations got better when you included the the interaction of neutrino know they got better when you included the the interactions of the stars and the gas and the dust and the dark matter and so on and so forth and and and also at the larger scales I mean we have these large scale structures in the universe the voids the the C the walls the and so on is there's something special that emerges when the full simulation runs with both all of the inputs as well as sort of the largest scales that you're hoping to achieve yeah so it's it's it's in the way that you're you're talking about it is really hard to predict exactly what you're going to get out there are lots of technical elements to the models that we produce and there are limits to what we can validate on those models for example we can check that the number of objects as a function of how bright they are is approximately right matches what we've observed about the universe so far we can measure how correlated different properties are in the simulation and how those look like uh for the real universe but those are are lowlevel um lowle values lowle measurements in the sense that they only measure particular modes of information very basic modes like the mean or number counts uh and we we we did run into this when we produced this latest set of large simulations all of those metrics looked great uh on paper and then when you looked at the images things just looked a little off in our first test round uh we had lots more bright objects clustered together in what we call Halos which are the the Dark Matter structures that bionic um things like galaxies and stars form within uh and so things look much more spectacular interesting in the images than the real universe is because there's all this bright clustered stuff uh and we were able to to build a feedback loop in uh we we saw this thing that looked emergently like you said a little weird uh too too interesting compared to the real Universe you might say and we're able to go back and tune our numeric tests our validation metrics to catch what ultimately caused that uh for the main run and this is one of those um those integration steps or or validation Loop test what I talked about yeah um but you know I think when I talk about simulations and you know we see this in a lot of fields I mean it comes up in climate science all the time that you know that that what you just mentioned this sort of process of looking at the final result comparing it against your understanding of the universe and then going okay obviously we need to tweak the simulation because we didn't get the right universe that seems very you know I think for most people that'll feel a little unproductive but I think the so the the goal though is now you can make predictions about the universe about things that we don't know so what you know as you sort of complete this final version of of the open Universe you have the universe in front of you at a resolution you know you can zoom in to any part of the universe that you want what are now the predictions that you're making about the universe that hopefully telescopes like Ruben and Roman are going to be able to to confirm or or deny yeah so so there are kind of two two ways that we use simulations one uh utilizes bottomup simulation of the universe like you're talking about where we take our best data set and we try and produce a universe that has all of the physics we know about in it that ends up with some final Universe realization that still matches what we know what we observe the universe to be uh the utility of that is that there there's all the physics that we are una aware of or that could you can even create universes that have slightly different physics that are still consistent with what we see if you want to test a particular model for example and we use those as test beds to to understand what we can observe about the universe we can do the same sort sorts of things we would do on the real data but we have much larger volumes uh we can create the entire universe or as large as section of the universe as we want if we have a course simulation um and and those are used for developing models we can even uh for example um extend the the the regions of of measurement that we can we can study uh for example gravitational physics is well understood on on large linear scales linear meaning small perturbations and the the curvature of SpaceTime or the matter density uh but when you want to look at the small scales scales that are getting closer to the the size of a cluster of galaxies or individual galaxies there's a hugee number of um there's a huge amount of more information there in terms of what we call number of modes which are the number of fora modes you can sample and in in a field um but it's very hard to build an a priori analytical model for that information because things become very nonlinear uh but we can make measurements of that information from simulations that match the real universe and we can use that to build a model or a suite of models um over some parameter space some cosmological parameter space uh to to to use when we study the real Universe uh the the Image level simulations that I do are have a slightly different goal in that we have we can have all of this physics that ex exists and the simulations I just talked about but we are simulating also how we observe the data the instrument characteristics how the photons that we would receive in the telescope get translated through the electronic system to uh counts data uh and we use that in a slightly different way to prepare for future data or to calibrate data at a very low level so our goal with open Universe simulation Suite that we just made combining uh Ruben Observatory data and r Space Telescope data is to give the teams that are working on those experiments an opportunity to have data that has all the weirdness that their actual telescope data will have before they have that data and that allows them to build the pipelines that will produce cataloges of objects or measurements of objects that we'll ultimately chose science with and to understand how we need to better calibrate that uh before we get to actually doing it on the real data so it's so it's not just enough to simulate the universe if you could fly around and and look at any object that you wanted it's about simulating the universe in a way as it would be perceived both by Roman out in space or Reuben through the soup of the atmosphere yeah yeah my my science field we lensing gravitational lensing uh we we Ed for example to map where dark matter is in the universe the and we do this by measuring the shapes of objects uh and so if a bunch of objects far away from us their light travels to us passes through structure along the line of sight uh that structure uh the photons or or light rays are going to follow the curvature of SpaceTime SpaceTime is locally curved around that mass and that causes a deflection in the path or a distortion in the shape uh and because all of those background galaxies light travel through the same structure they have a coherent Distortion and we try and measure that Distortion by averaging over the correlation of shapes of of objects in the background um the difficulty is that it's essentially impossible to make an unbiased measurement of that uh we're trying to measure the shape of objects that are just a few pixels across that have a signal to noise of like 10 across many pixels so they're if you just look at an image they're barely visible they're just a bunch of noisy pixels that look correlated uh and and just very hard to do a real it's very hard to extract the actual shape of any individual object in an unbiased way so that's a function of the telescope it's a function of how the Optics of the telescope or the atmosphere also smear out and and and change the shapes of things through what we call point spread function this atmospheric Distortion or Optic Distortion uh it's a function of the ultimate signal to noise of the objects the number of photons that actually make it through the instruments um for any given object for some type of object or some some flux intrinsic flux of the object and we really have to get all of those things simulated accurately to understand how biased that measurement is uh and so there are ways we can kind of self-calibrate the measurement with the data uh with techniques uh one of which we call Meta calibration and that gets us most of the way but there are still things that we can't fully calibrate without reproducing the entire measurement on these simulations um so this is an example of a case where we really need these large simulations to understand the characteristics of our measurement oh it's really interesting so it's funny so like we share pictures in our news and and on the website of like gravit really nice gravitational lenses and Einstein rings but those are like the charismatic megaphon of gravitational lenses the you know they're the rhinoceroses and the Tigers of of gravitational lenses but the reality is the vast majority and the ones that you're going to be using to map the Precision of dark matter are going to be three pixels that are like a little bit distorted and it takes really like you need to look at millions of them in together to sort of measure what is their average warping of oh that's really interesting yeah um and in fact if you look at our for example the dark energy survey which is another experiment I work on if you look at the mask uh so you can characterize where which parts of the sky which data we can use and which data we can't use if you look at that mask that we use for the weak lensing measurement most of the well not most but a large number of the large blocked out areas of the survey are exactly those really bright pretty galaxies that you see in popular photos online from from HST for example the ones that are big and bright are too bright too big uh for our sensitive cameras that we're using to do these very deep sensitive weak cleansing measurements and so they're actually systematics to the galaxies that we care about which are all the very tiny low signal to noise ones uh which are most of the objects in the image I think people are people have a fairly outdated notion of how well astronomers are able to measure the concentrations of Dark Matter today uh can you give us a sense of of to like what level of precision we know where the dark matter is in relation to the galaxies that they inhabit and sort of how that sort of Narrows down the possibilities of what Dark Matter could be yeah so there are kind of three three regimes where we measure or or infer things about dark matter one is on the very largest scales the the total composition of the universe uh the total amount of Dark Matter Dark Energy standard particle physics material uh in cosmology we call them Barons even though there are non barion stuff in there uh those broad categories are extremely well measured um first with the cosmic microwave background experiments U like w map and plank um and now more recently with low red shift Universe uh probes like Galaxy clustering and weak censing uh Baron acoustic oscillations the these current low red shift surveys that are going on uh and there you're really just measuring how much of the stuff in the universe is this this type of matter for example Dark Matter uh we know that very well we we can we can we can measure that value that that fractional constituency very well then something that has only recently been possible uh with with with low red shift Galaxy surveys like the dark energy survey is that we can take these hundreds of millions of galaxies in say an eighth of the sky for the dark energy survey and we can reconstruct these distortions in their shapes map the this Distortion map in the curvature of SpaceTime from all of the mass there not just the mass we can see the Barons but also the dark pattern and we're actually producing maps on a coarse maybe half a degree resolution scale uh of where projected on the sky the 2D plane of the sky all of the dark matter is up to around half the age of the universe ago uh and we're creating these large scale maps of the dark matter and that's interesting because it tells you about how dark matter is clustered on large scales and then there's a third regime where we we look at dark matter which is like you mentioned individual Halos we know that dark matter is gravitationally forming these high density regions this is where galaxies tend to form clusters of galaxies tend to accumulate uh and for those we we're looking at much more um specific measurements of these distortions around specific Halos for example or we're taking very large structures like clusters of galaxies or uh background distortions like the Einstein Rings you mentioned and we can really learn about the shape of the the Halo of the dark matter for individual clusters of galaxies or even individual very massive Galaxies for example um and we can we can do all of that with weak lensing uh gravitational lensing which is why I find it fascinating uh but you can also use other specific types of physical processes to to infer these different things like on the largest scale the CNB um on the smallest scale Galaxy clustering or these strong lensing events even on the smallest scale you can use the Dopp measurements to track the Motions of all of the stars in one Galaxy and then use that as a way to to map out the quantities of of Dark Matter around it and so sort of at the at the smallest scales and at the largest scales we really it's funny right like we don't know what it is yeah but it's the level of precision that it's being measured to is is incredible yeah yeah tell you exactly thisy but what it is yeah who knows yeah and so you mentioned like these half degree on the sky so like the size of the moonish on the sky you've got a map a 2d projection of are you saying like of of all of the dark matter that is sort of behind it in a in a sort of conical slice all the way out to the end of the universe uh so we we usually measure distances in red shift so this is like a red shift of 1.2 which is about 6 billion years ago in Look Back Time um so we can infer that the projected distribution of the Dark Matter anywhere between us and the furthest galaxies that we can observe with a survey like dark energy survey which is that kind of 1.2 limit in red shift or six billion years of Look Back Time and so any any anywhere those photons have passed through that dark matter a the line of sight we we can learn about the the the cumulative distortions in the photons path which tell us the the the coar distribution of the dark matter so so I want to kind of Reas my question the one that I asked you earlier about you know when the simulation starts up and then you look at the data and what are you looking for but when Rubin comes online and begins producing its data when Roman launches and begins producing its data uclid is already going and doing its version and then there's Desi and there's other surveys and so they're all working together to produce as much data as possible what are you going to be looking for yeah uh so each of the surveys have their own strengths uh unique strengths and this was this is also true of the current generation of of surveys the dark energy survey observes a huge area of the sky 5,000 square degrees Reuben will be similar except 10 times better in every way it will observe half the sky and much better resolution uh depth um so Reuben is going to see half of the universe and it's going to see that over and over again every few days uh which is which is another transformative element of that survey and that you get a basically a Time sequence of how things in the universe are evolving which is really important for ways that we study the universe like through Supernova for example um then on the other end of the spectrum is is a survey like Roman uh Roman is in space so it will see everything sharper but you're limited to how much you can send up on a rocket in a single launch so the telescope is smaller uh it's going to view it's not going to be able to view half the sky it might view several thousand square degrees of the sky similar to to the size of the Dark Energy survey but it's going to see much deeper much sharper and so a lot of the things that we worry about with uh something like weak lensing where we're trying to measure the shapes of individual galaxies as you go uh as You observe deeper in deeper into the universe there are more and more things you can resolve eventually they start becoming stacked on top of each other uh or overlapping in projection which is what we call blending uh so having a telescope in Space is really critical for the sharpness of what it can see it doesn't have to look through the atmosphere so things aren't smeared out by atmospheric temperature gradients or turbulence so we can we can better distinguish individual objects and we can distinguish them much further away uh with a telescope like Roman but then we have better statistical power from all of the objects the area of the sky will see from Reuben and so one of the goals for these sorts of simul campaigns was really putting uh providing for the the community simulations of these two instruments because their true power is going to be when we combine that information when we use the Roman Space Telescope to help de Blend or or differentiate individual objects in a statistical sense in the in the ruin data uh which is going to increase its its ability to two we gravitational lensing uh or vice versa the Roman space t scope is observing in the near infrared um around 1 to two microns uh that is difficult from the ground because water in the atmosphere absorbs that sort of light um but to do Precision distance measurements with Roman we'll need the optical information that we get from the ground from the the ruin Observatory as well so there's this this huge Synergy that gets formed by the strengths of each survey they're very powerful in different ways and their own right but then the ability to learn how to combine the data which is a non-trivial thing with the pedabytes and pedabytes of Imaging we'll have uh that's I think where these simulation Suites that we've been producing really shine we we're enabling the community to start learning how to do that now uh so that we can combine the strengths of these surveys when they when they launch or when they when they start right because I I know I'm going to be all over it as soon as Ruben comes online I'm going to be asking lots of questions and I'll talk to astronomist and they be like I don't know I'm still downloading it it's huge I don't know talk to me in three years once it's done but but I'm you know I can't wait for that yeah um so I want to talk about like with the open Universe with the simulation that that you did you had nine days on this super computer will you have time to do another version a more like maybe a more powerful version a larger version before the the actual these teles clips actually come along uh it's unlikely that we'll be able to do that before Reuben starts taking data in the next year or two um Roman is scheduled to launch still in October 2026 which is the earliest launch window um it has available uh it is possible that we'll do another round before Roman's launch um mostly that depends on interest and the community um improvements to the input that we can make in that time and the avail Computing resources uh it's not um it's pretty rare to get the sort of opportunity that the Argon leadership Computing facility gave us with the theeta cluster um so I I I'm hopeful that we can do that again certainly we will need to once we have the telescope's data uh anyway because this is all an iterative process the best data we have now is not as good as the data we'll have from these next Generation observatories and so we will learn things about the univers that we'll need to fold in into a new generation of better simulations at that point as well um but the the the campaign I think was a really great uh it did a really good job of bringing the community together across the the reuin and Roman and NASA and due uh communities um we produced something that was I think very successful uh in that there are there's a huge number of projects ongoing using the simulations already um and it it really showed what we can do when we Bridge traditional expertise from the due funded Community like the ruin observatory in the NASA funded Community like Roman uh so my hope is that this is a good Benchmark or demonstration of what we can do if given the sort of resource um so I do hope in the next maybe three years or so that we'll we'll be able to do something better I I I think of it kind of like a video game when they released the sort of version 1.0 of their game and then you you have everybody play it and then everyone gives their feedback on the things that worked that didn't work and then they they release a new patch and in this case you know you've released version one and the astronomers are working with it but really you're not going to know whether you properly simulated the universe until they make their observations so what will you be looking for to do you know 1.1 when once you've actually got all that from Roman from Reuben from uid from Desi yeah so the the UN when we model the universe we kind of model the ideal average universe that could form which is not the same as the universe we actually got necessarily we call this Cosmic variant and within the universe different parts of the universe evolve quite differently they they may overall have this average value of Omega matter which is3 or something dark omeg a dark matter .25 or something but if you look in individual regions that varies uh the best large field data set we have to calibrate our simulations off of is the HST Cosmos data which is less than a square degree on the sky a very tiny part of the sky so we don't know how representative that tiny part of the sky is to the whole sky and so all of our calibrations have been trying to match the constraints from this tiny part of the sky but once Roman turns on in a matter of months we will have tens hundreds of square degrees of better than Cosmos data in some aspects uh that we can recalibrate against we'll really understand the cosmic variance of that tiny field much better uh and so even just just turning the telescope on the amount of data that Roman can take compared to something like Hubble or web uh it will really revolutionize what we're able to calibrate or compare against in terms of this space-based photometric Imaging to torture my analogy even further uh you you know you've from the game the it's gone through QA so they've had a chance to play it and they found some bugs but uh it's really till it actually meets the the full Cosmos will you actually know those those answers do you have any uh like hunches is there is there stuff where you you wish if if you had double the compute time that you wish you could explore a couple of different V variations on the universe to find something that maybe would both sort of satisfy some questions that you might have but also maybe give some some variety to the astronomers yeah so there there are kind of two levels yes that I would say yes to one is at the underlying simulation of the Universe um I don't think that we've ever really had a truly what we call blind um and what we call blinding is is really important in cosmology because this this an aside to explain what I'm going to say um because we have one Universe we're not a particle physics experiment where we have a collider and we can Collide millions and millions of pairs of particles and see what happens we get one experiment it's already done and we can observe it once in an unbiased way and so we really worry a lot more I think about Observer bi or or or kind of subconsciously reinforcing things that we expect uh so there there sort of sort of natural models for dark energy for example uh one is that it's a cosmological constant in Einstein's field equations where it has a particular dark energy equation State uh it's really easy to expect that to be the truth because every experiment has been consistent with that and and so even without thinking about it there there are some deeply rooted expectations about things that the Universe could be doing so we we have to be very careful about stepping back and and not building in safeguards to prevent that sort of Observer Bias and I think that one thing you could do with simulations is you could create universes that look like a cosmological constant and the field equations or that look like something completely different that is entirely consistent with our current observations but could be distinguished with this next generation of experiments and I don't think anyone has really successfully done that sort of blind test on simulations where they can show that they really can infer what the universe is really doing when they don't know what the input to the simulation was there have been attempts by many collaborations I think to do this in the past I don't know that any has has actually worked out very well but that's one thing that we could do with these large simulation Suites if we were able to produce multiple of them uh the other is that when we simulate the instruments we're simulating their specifications or their requirements but we I mean when you launch a telescope into space things rattle around water evaporates gets much colder uh things change uh same thing on the ground uh over time things evolve uh and the the sorts of characteristics in the instruments were going to realize is going to be different at some level than what we've planned or what the requirements are some will be better some will be worse uh and so especially for something like the Roman Space Telescope as we're preparing the pipelines to calibrate the data uh it would be really interesting to me to simulate a version of the telescope that is a little different in ways that the people at the science operations center who are building the calibration pipeline don't know and then see if they actually calibrate properly uh see if we can actually calibrate L lensing signal properly uh if we don't know um what the actual characteristics of the instrument are which is the case will be in when we launch the telescope U so I think those are two kind of very different ways in which simulating more or multiple realizations of the data would be be quite valuable in in a scientific sense so going back to that the first version that you talked about you know you've got this imulation of the cosmological constant as an explanation for dark matter but we see with the Hubble tension we see a difference in the measurements of the expansion rate of the Universe from the cosmic microwave background compared to the more closer seied variables and if you just like took that as a given like we don't know why but but dark matter is this at the beginning and then this in the middle and then this now run that through your simulation and see what it does and you might got get a different universe that will have enough subtle differences that you could actually make some kind of predictive you know comparison when you actually put it in front of is that a good example of that yeah yeah and people definitely do make these different types of simulations um I think the the key difference here is whether you know what to expect out of the simulation in in the same way we don't know what to expect out of the universe or not um yeah and then on your your second point I mean I think ason was working with James Webb were surprised at how well they were able to align everything and I think they got a better they're getting better data than they were anticipating and so in the best possible World in fact the astronomers with with Ruben with Roman are going to need maybe something that's a little better so they can even distinguish it even even more and is that in your data like could you because I mean if you've already done the compute to actually like are you Computing their ability to observe this universe at the same time that you're calculating the universe or is that like a you know like a post-training additional layer and you could produce multiple versions of that yeah so there's there's the underlying Universe which is like the model of all the stars and galaxies and things that's one thing and that that exists before the campaign that we did at argon um that on its own is expensive and and and we for this simulation we were used existing in body simulations like gravitational physics the evolution of the Universe um so we didn't have to recompute that and then there's the telescope what happens to the photons above the atmosphere through when they get converted to electrons and red in the electronics of the sensors that's the the process that we simulated at archon in January um so that's done now depending on the instrument because we have um we have different simulation pipelines for the Reuben observatory in the Roman Space Telescope that use common tools that we've built over the years but depending on the instrument there is an element of what you talked about where we can kind of uh save intermediate steps and then reprocess things so for the Roman Space Telescope we save a version of the the images which is basically where every photon is before it goes through the instrument or before it hits the sensor for example and then we have a second processing stage where we convert those photons above the sensor through the quantum mechanical processes that happen in the sensor to what we read out as electron counts or data units uh so for some types of physical processes we can re simulate and we do res simulate um the final images from those intermediat get we call true images because they're just the the true Photon distribution um we can do that and change tweak things about the sensors for the Roman Space Telescope uh depending on how your simulation is set up that's not necessarily possible for all instruments um and some physical processes you can't really change them without re simulating everything from the start like the point spread function for example um but there we are doing some of that we've we've in this simulation we the main images for the Roman Space Telescope we produce are kind of what you would expect to get after calibrating the images there's another version of the images we can produce if people want to use them that have all of the craziness of the uncalibrated data um which is not useful for most people right now because there isn't a final Roman calibration pipeline now that people can use so they no one would be able to calibrate the data right now but uh there are aspects of that that we we can do and will do as we iterate without redoing the whole simulation I mean I mean some of the most fascinating stories to me are times when astronomers working with these telescopes realize a clever trick whether you know ligo is a good example this where they figured out they could do Quantum squeezing to get better resolution on their the Imaging of black hole mergers of detecting the gravitational waves and for groundbased telescopes obviously you can come in later and you can put in a better instrument strument you can remove additional noise from all of the instruments spend more time figuring out where things are you know could be improved space bace telescopes not so much but there you know there's some techniques people can can try to to adapt um and then I wonder if if using your simulation they can try different ideas what if we you know can we get better a better place and that's where that raw data might come in useful is they could they could try a bunch of ideas and then run it against the you know a version of the data to see what their expected outcome might be and so it sounds like you know it's another way to to improve um yeah that that is my hope uh for for what people can get out of these simulations of the actual data we got this at least for the Roman Space Telescope this true true image that they can make measurements on and then the image that goes through the sensors that they can make measurements on so you can you can iterate comparing against those and hopefully improve what we'll be able to do with the real data it's two audiences I mean one audience is the astronomers who are just going to get their picture in the pipeline they're going to look at it and they're going to measure their gravitational lenses and call it a day using the the telescope that they have and then there's the people working on the instrument itself kind of going how can we make this better and then running Sim you know running the simulations against their simulations to see what could produce a potential better outcome it's a sort of a nice virtuous circle um so Michael what are you obsessed with right now uh well so right now my main focus is preparing for these these surveys uh the the type of data that we're going to get from Roman is in in many ways simply revolutionary for for a science like we lensing uh there are lots of things you have to worry about from the ground the atmosphere smearing messing things up that just don't exist in space uh the Roman Space Telescope is in many ways an ideal instrument for weak lensing that that was one of its designs uh and so we're in a completely new regime of of weak lensing science with this Mission um but there are also things that we'll have to do for the first time that we've never had to deal with uh so there's a cost to sending things into space that are space space ready um that can survive space for 5 years where instruments don't uh don't don't enjoy their life uh in the radiation which vacuum of space um and so one of the the consequences of that is that space-based images have relatively large pixels compared to their resolution uh and so the information that we get from a telescope like the Roman Space Telescope is the data is so good in some ways psf is so small from the Optics that it's smaller than the pixel scale and so you can't actually on a single image map out what the psf is and we call this under under sampling because we can't map all the small scale features of the psf and so we'll have to combine many images together to reconstruct that sub pixel information that doesn't exist in a single image and that's not something you have to worry about with groundbased data typically uh so there's a huge regime change and how we have to deal with the data for we censing that that is opens up a whole new challenge um that's one thing uh the other is that we're going to see so much deeper uh so with with something like the Roman Space Telescope so we were thinking about dark matter or dark energy figuring out what they are uh if you have a cosmological constant Dark Energy evolves in a certain way it's energy dens density is constant per unit volume at all times uh if it's a phenomenological field or something that is not just a constant like the gravitational constant in the field equations then it's going to evolve differently its density will evolve over time and so its effect on the expansion of space is going to change over time with current data we can't differentiate that we're looking at a relatively small region in terms of time or distance in the space and so measuring that Evolution over time uh in the the expansion rate of the universe we don't have a very strong lever arm and in that we were not comparing very very far apart instances in terms of space or time but that will change completely especially for the Roman Space Telescope which is going to observe Back 9 10 maybe even 11 billion years ago uh now you have a l arm that is basically the size of the observable universe and we're going to really see for the first time how things are changing over time in a way that current telescopes just can't and that's going to give us the opportunity to completely blow open these questions because now instead of measuring how much dark energy there is we're going to be measuring how much dark energy there is as a function of time and that will really open up what we can learn about the model space the the physical features of of what really dark energy is or maybe even location yep yeah yeah um yeah I mean I think you know like science always runs up against the very limits of its ability to observe like at a certain point I think about say that like the searches for for dwarf planets out in the outer solar system like if you notice there haven't been a lot of new dwarf planets found because the current crop of telescopes are are reaching the very limits of what they can they can observe but when the next round comes out then you know we're g to within weeks of Ruben coming online we're going to learn of probably many new dwarf planets and other objects out there in the outer solar system and same thing when when these these first surveys start to come in what was all at the very limits and the edge and and the hints and the it all just now just becomes easy to access uh it's it's a pretty great I mean it's the same thing that we're seeing with with web where they're just like oh do you want the spectroscopy on a Galaxy with red shift 14 no problem right here you go and and the kinds of things that Hubble require to gravitational lens and you know and low Sigma to to accomplish so um yeah yeah it it is I know what you mean that that seeing stuff that was all you had to really squint suddenly all just becomes beautifully into Focus but then come the next questions right I mean this process never ends and so you know you're going to have this resolution on the sum but then here come all the new questions yeah and I think in some ways our our imagination about what Dark Matter could be or what Dark Energy could be is is is framed by the limitations of what we can measure I mean differentiating a cosmological constant versus not a cosmological constant is a tiny tiny thing theoretically uh there there are way more exotic things that that could exist but we've really been focused for decades now on just that one number and now we're going to turn that one number into a whole field of numbers and the I think it will spur an entire new generation of thinking about the way we view these problems fundamentally yeah yeah and it it it is funny how this is what astronomers want like I think astronomers want this to be complicated and messy and interesting and and move from just one number that you're like what are you going to do right to picking away at all of the little bits and threads and strings and stuff that you can actually sort to make some progress again on on our understanding of the cosmos and listen to what nature is trying to tell us um good luck with uh uncovering all that thank you so much for your time yeah thank you for talking I hope you enjoyed this conversation with Dr Michael Troxel I'm going to give you some of my thoughts and Impressions but first I'd like to thank our patrons thanks to Abe Kingston Adam Schaefer andream gross Bill David Gilton and David mats Dennis Liberty Dustin cable Jeremy murn Jim bur Jordan young Josh Schultz modzel Paul Roo Steven krosaki Steven fer Munley and Vlad chiplin who support us at the master of the universe level and all of our other supporters on patreon I know I've been talking about these telescopes for years and yet we are now really just months away from Ruben coming online just a couple years away from Roman coming online and all of these data are going to be dumped into the internet so astronomers can then start to work through it and start to make some questions and it was right near the end of that interview where this really struck home for me that up until this point based on the best observations that astronomers can do so far we're just getting the very slightest hints of what some of the characteristics of dark energy are we see Dark Matter well resolved in some places but the entire un Universe it's still a bit of a mystery and suddenly these instruments will be able to fill in all of the gaps we will see the influence of dark energy from recent times all the way back to almost the beginning of the universe we will see Dark Matter not in just this tiny little slice of the universe that Hubble was able to look at for a long period of time but half the sky and definitely a significant portion of the entire universe what was a fairly rough number becomes an incredibly deep nuanced number both in terms of time in terms of space terms of degrees in the sky comparison between different instruments and this will give what is a lot of arguments final understanding and yet of course will open up new questions about the universe and I can't wait to hear what happens