Dustin Lang on big data from a big universe

Dustin Lang is a computational scientist at Perimeter Institute who develops techniques for finding needles in the cosmic haystack. He works on several large sky survey projects, tackling the statistical data analysis required to discern meaningful insights from huge datasets gathered by telescopes. In this conversation with Lauren and Colin, he shares his experiences mapping galaxies with the Dark Energy Spectroscopic Instrument (DESI), and hunting for mysterious fast radio bursts (FRBs) with the Canadian Hydrogen Intensity Mapping Experiment (CHIME). He explains how he and his colleagues employ software solutions to expand capabilities of hardware telescopes, the challenges he has encountered in radio versus optical astronomy, and the important role of chicken wire at CHIME. View the episode transcript here.

We are looking for feedback about our audience’s preferences as we chart the path forward. Please take a few minutes and fill out our podcast listener survey.

Conversations at the Perimeter is co-hosted by Perimeter Teaching Faculty member Lauren Hayward and journalist-turned-science communicator Colin Hunter. In each episode, they chat with a guest scientist about their research, the challenges they encounter, and the drive that keeps them searching for answers.

The podcast is produced by the Perimeter Institute for Theoretical Physics, a not-for-profit, charitable organization supported by a unique public-private model, including the Governments of Ontario and Canada. Perimeter Institute acknowledges that it is situated on the traditional territory of the Anishinaabe, Haudenosaunee, and Neutral peoples. Perimeter’s educational outreach initiatives, including Conversations at the Perimeter, are made possible in part by the support of donors like you. Be part of the equation: perimeterinstitute.ca/donate

Episode Transcript

(00:00):
(bright music)

(00:08):
- Hi, everyone, and welcome
to "Conversations at the Perimeter."
I'm Lauren Hayward here with Colin Hunter.
- Hello.
- Today we're excited to share
with you our discussion with Dustin Lang.
Dustin is a computational scientist
here at Perimeter Institute
who specializes inastrophysical data sets,
which means he works on software solutions
that help researchers study some

(00:30):
of the biggest openquestions in our universe.
- And my mind really reeled
when Dustin describedthe enormous quantities
of data involved in these projects
that he and his colleagues are working on.
It's literally astronomicalamounts of data
that he and his colleagueshave to sift through
looking for these faintsignatures of phenomena
that are incredibly far away.

(00:50):
- And they're far awayboth in space and in time.
Dustin tells us about his work
with an international project called DESI,
which is building maps of the universe
to look back over its history
and gain insight into dark energy.
And he explains the CanadianCHIME Project as well,
which is searching formysterious fast radio bursts

(01:10):
from deep in the cosmos.
- Dustin tells us too about his work
in both optical astronomyand radio astronomy,
which are more differentthan I had realized.
He also tells us aboutthe important roles played
by chicken wire and ametaphorical sad trombone.
Whomp-whomp.
It's a really fascinating chat.
So let's step inside thePerimeter with Dustin Lang.

(01:33):
Dustin, thank you for being here
at "Conversations at the Perimeter."
- Oh, my pleasure.
- We've been lookingforward to talking to you
for a number of reasons.
There's much that wewant to explore with you,
including a number of acronyms
of projects that you're working on
that have to do with deep space
and distant explosionsand everything else.
But before we get to that,you're a computer, No-

(01:54):
- Computational scientist.- Computational scientist.
So first I wanna get into what that means,
but I want to do so by saying
that a couple years ago Iinterviewed you for a story
and you joke that when the job posting
for a computational scientist
came online at Perimeter Institute,
that your friends basically said
"This job was written for you Dustin,
you have to get this job,"

(02:15):
because it blended big dataanalysis and astrophysics.
So can you tell us what do you do
as a computational scientist?
- Sure. So I have a kindof unusual job here.
I'm half in the IT department
helping other researchersmake use of computing
and half a researcher myself.
So I work on astronomical surveys,

(02:35):
surveys that go out andmeasure big chunks of sky,
often without preconceived notions
of what we're going to find
in order to kind of make new discoveries.
- And when you talkabout big chunks of sky,
like how big are we talking here?
- In the one project weare looking at basically
all of the sky we can seefrom the Northern hemisphere
except for the parts that are filled

(02:56):
with the Milky Way galaxy.
We care about things thatare beyond the Milky Way
for this particular project
so the Milky Way gets in the way.
There are too many stars in our own galaxy
to see the stuff behind it.
- We're getting in our ownway, in our own galaxy?
- Pretty much.
And then you can't see thesouthern part of the sky
because there's too much dirt in the way.
(Colin laughs)
- So you're looking basicallyeverywhere you can look.
- Pretty much.

(03:17):
- And why is a computational scientist
essential to doing this work?
- So my degree was in computer science.
I kind of picked up physics on the job
(both laugh)
and a lot of physicists arein the opposite position
where they know the physics
and they're suddenly facedwith ever-growing data sets
and there's just a real challengeto process some of them.

(03:39):
So having people withexpertise in both is kinda key
to making some of the advancements
that we want to do in this
kinda to push the next generation
of understanding of the universe.
- Would you say thatastronomy and cosmology
is an area in particular whereresearchers with expertise
in how to do these computationsis really necessary?
- Lots of areas of physics

(04:00):
are pushing computational boundaries.
I know that our data rates, for example,
aren't anywhere nearwhat you would encounter
at CERN, at the Large Hadron Collider,
but we're probably in the ballpark.
I know that we use a DepartmentOf Energy supercomputer
for one of my jobs andmy group uses basically
the second or third largestuser of the whole center,
which has like 1,000s of users.

(04:21):
So we're kind of up there Iguess in terms of data rates.
- Is there so much data becausethe universe is so enormous
and you're looking at so much of it?
- Pretty much.
- Like when we see images from telescopes,
we see billions of starsand billions of galaxies,
is essentially all of thatstuff out there in the universe
is data that needs to be crunched?

(04:43):
- Yep. Exactly.
Basically the sky is big
at the scales that youcan see from the ground
and that kind of sets thebasic scale of the problem.
So with the largestcamera we have right now,
it still takes 1,000s ofimages to cover the entire sky.
And we want not just oneimage but multiple images
to understand not only what'sgoing on at any instant,
but trying to understand someof the changes with time.

(05:06):
- So some of the work thatyou've done has been with DESI.
That's one of the acronyms thatwe'll be bringing up today.
I like that one 'cause it's a nice name
but it stands for morethan just a nice name.
Can you tell us what DESIis and what it's for?
- Sure, so DESI stands for
the Dark Energy Spectroscopic Instrument.
So this is an instrument,it's a device that is sitting

(05:28):
at the top of a telescope in Arizona.
Instruments on these telescopescan be either cameras
or spectrographs for the most part.
Cameras, most people arepretty familiar with.
Spectrographs are a little bit different.
This one is called amulti-object spectrograph.
So basically we can observemany galaxies at once

(05:48):
and break their lightinto spectra or rainbows
and take precise measurementsof like the brightness
at each point in the rainbow.
So the innovation with DESI
is that it can take many more at once
than previous generations of instruments.
It can observe 5,000 starsor galaxies every exposure.

(06:09):
It's really cool.
- That's like-- Part of the-
- One camera taking, well sorry,
it's not a camera, it's a spectograph.
But one instrumenttaking 5,000 observations
all at the same time.
- Yeah, that's right.
So this is the realinnovation of this instrument.
So to give you a kind of a context,
the previous generationcould take 1,000 at once.
That was the Sloan Digital Sky Survey.

(06:30):
And that project is also cool.
But basically in these projects
you have to choose ahead of time
which objects you're going to observe
because how they work is youstick a fiber optic cable
and point it directly at eachobject that you wanna observe.
The light comes from your galaxydown the fiber optic cable
to a spectrograph that actuallysplits it into the rainbow.

(06:53):
So then the challenge is, you know,
how do you point 1,000 littlefiber optics at once and-
- How do you point one atonce let alone 1,000 or 5,000?
- Well so, and the otherchallenge is you have to like,
the fibers are like this kindof the size of a human hair
and you have to point themto finer than that precision.
- At galaxies that are-- Yeah, exactly.
- Billions of gajillions of miles-

(07:14):
- And your telescope weighs many tons.
So the thing like it's really,
the engineering is really amazing.
- How do you do it?
It's not a person with tweezers, right?
(both laugh)
- Right, well...
- Or is it?
- In the Sloan Digital Sky Survey,
what they did was theychose which galaxies
they want to observe ahead of time.
They compute wherethey'll appear on the sky.

(07:36):
Oh, you have to choose a set of nights
that you're going to observe it on
and a time within that night.
And given that, you can predictwhere they're going to be,
they take an aluminum plate,
drill little precision holes in the plate,
1,000 holes for 1,000 galaxies.
Ship those plates to the mountain
and then a crew of people, by hand,
plug in fiber optic cablesinto each of those holes.

(07:58):
- Wow. That's not how Iimagine this would happen.
- Yeah, exactly, it doesn't sound
very high tech.- Right.
- So during the night they would go out
and plug one of theseplates into the telescope
and that plate steers the light.
You know, the fibers arein just the right place
to steer the light down those fibers
to be collected in the spectrographs
and make those measurementsof 1,000 galaxies at once.

(08:19):
Let me say just for a second,
'cause I was talking about thehand-plugged fibers in SDSS.
When DESI was being designed or proposed,
one of the challenges was scaling up
from 1,000 to 5,000,
doing that by hand just started
to get like to be infeasible.
So the way that DESI instrumentoperates is really cool.

(08:40):
It uses these little robots,
so 5,000 of them and each ofthem has two little motors
that allow it to rotate the fiber
to any place within its little region.
So it's sort of like yourshoulder and elbow joints.
One of the motors moves the shoulder
or like rotates the shoulder in a circle

(09:01):
and the other can rotatethe elbow in a circle.
So between that,
they can position the fiberanywhere within their reach
and then they're placedclose enough together
that they can just reachtheir, or like they have
a little bit of overlapwith their neighbor.
So no matter where a star or galaxy lands
on the focal plane of the instrument,

(09:22):
at least one of them canreach it with its fiber
and it holds out its fiber
and the light pours down andgoes into our spectrographs.
So another innovation of DESI
was that in the previous generation,
the spectrographs were boltedto the side of the telescope
and they flopped around during the night
and were subject to thesurrounding temperature.

(09:45):
So for DESI, what we do instead
is the spectrographs are put
in a nice climate-controlled cleanroom,
but then we have to get the light
from the top of the telescopedown through the telescope.
It has moving parts of course.
So there's a 50 meter run of fiber,
5,000 fibers that goesdown to this cleanroom.

(10:06):
So 500 fibers each pluginto these spectrographs,
there's 10 of them.
And the fibers come in in a big stack,
like they're lined up in a big stack
and then their light shines onto a prism,
basically, that splitstheir light into a rainbow.
And then that rainbowlands on like a sensor,

(10:27):
a CCD sensor, like a camera basically.
So what you see in the imagesare 500 like rows of rainbows.
But of course they're not,
these sensors themselves are monochrome,
like they only, they're justmeasured black and white.
So you see kind of abrighter or fainter line,

(10:48):
500 of those spacedtogether across the chip.
So brighter spots areplaces in the spectrum
that are brighter.
So during the afternoon weuse these calibration sources.
So like you know,
you can shine light of a known wavelength
and measure where itappears in the images.
So you can say, oh that little bump
is red 540 nanometers

(11:10):
and this little bump issome other wavelength.
The thing that's kind of amazing
looking at the raw data though,
is that all of themlook the same basically.
And that's because the skyis pretty bright, (chuckles)
even the night sky at the darkest times
is actually the thing
that we detect moststrongly in the images.
So it's only by subtractingout the contribution of the sky

(11:32):
that we get to see thestars and galaxies in kind.
It's not an easy way to live.
(both laugh)
- And once all thatinformation is collected
from those 1,000 or 5,000 points,
does it then go to you to figure out,
or you and your team,
to then do all of the computationalwork to understand it?
- Yeah, other peopleon my teams. (chuckles)

(11:54):
My work on DESI comes earlier actually.
I've been involved in,
remember I said you haveto choose ahead of time
which things you want toobserve, which we do from images.
So first you go out andtake an image of the sky.
in our case in like threedifferent filters or three colors,
and you measure all the stars and galaxies
and measure their brightnesses and colors
and choose some set of them

(12:15):
that are interesting for follow up.
We get to choose about 1% of them.
So when we started DESI,
there was no imaging survey that existed
that was deep enough to makethose measurements, right.
We wanted to measure thingsthat were faint enough
that they just didn't appear
in the existing generationof imaging surveys
so we had to go out anddo those imaging surveys.
So that's the part
that I was kind of mostmostly involved with.

(12:37):
- And I'm hoping you can tell us
a little bit more aboutthis idea you referred to
as splitting up theelectromagnetic spectrum.
So the electromagneticspectrum is quite wide
and only a small portion of it is visible
and then you also do some splitting up
within that visible piece.
Can you just tell us alittle bit more about that
and how different telescopes
focus on different parts of the spectrum?

(12:58):
- Sure. I call myself mostlyan optical astronomer,
which means I work in more or less
the visible part of the spectrum,
which then also now bleedsinto the infrared a little bit
because you can use the sametechnologies to do that,
to observe light thatwe can't quite observe.
So different telescopestend to be optimized
for observing differentparts of the spectrum.

(13:18):
Partly from the ground,
only parts of the spectrum
actually make it through our atmosphere.
If you go very much bluerthan we can see with our eyes,
that atmosphere just blocks everything.
Just the air absorbs all of that light.
As you go toward the infrared,
water is actually one of the annoyances.
So water vapor in the atmosphere
also emits at thosesame frequencies, so...

(13:40):
- You don't often hearwater called an annoyance.
It's also essential for life on planet.
- Some people enjoy it. Yeah.
(all laughing)
- It has its pros and cons.
- Right. As long as it would just-
- Stay outta the way.- Stay outta the upper
atmosphere or just thecouple of cubic kilometers
around our telescopes,that would be great.
And then if you gofurther into the infrared,

(14:01):
that is just heat andthen it's really hard
to observe something faint in the sky
when like your telescope andyour mirrors are all glowing,
which is basically whathappens in the infrared.
And then so there's abig chunk of the infrared
that we can't reach,
which is why peoplelaunch things into space
to observe in that frequency range.
So JWST for example, anda telescope I really love,

(14:22):
the Wide-Field InfraredSurvey Explorer, WISE,
also a NASA mission, and they go to space
because basically you can't observe
or it's very, very difficult
to observe that from the ground.
My advisor did a bunchof infrared observing
as part of his PhD andspent many, many nights
on some of the biggesttelescopes in the world
in order to make these measurements,

(14:43):
despite the fact that your telescope
is glowing at those frequencies.
And he said the Spitzer Space Telescope,
one of the first infrared missions,
totally made obsoleteall of his observations
within its first secondof observation. (laughs)
- Wow.
- Like it's really good
to observe when thesky is dark, basically.
It's not easy, basically,observing during the daytime.

(15:06):
I mean basically, the atmosphere sets
what we can do from the ground
and sets what we can do with telescopes.
And then there's anotheratmospheric window,
we call it in the radio.
So I think we'll come back to that later.
- Mm-hmm, DESI is called
the Dark Energy Spectroscopic Instrument.
You've told us a bit aboutthe spectroscopic part.

(15:26):
What is the dark energyaspect of this experiment?
- (laughs) Dark energy.
- (laughs) Big subject?
- Pretty big subject, yep.
Dark energy is one of the real mysteries
in astrophysics these days, or cosmology.
To explain that, go right back
to the beginning, to the Big Bang.
Around 100 years ago, theobservation was made by Hubble

(15:49):
that if you look atgalaxies, you can measure
whether they're movingtowards us or away from us.
And Hubble observed
that all the galaxiesare moving away from us.
And not only that,
the ones that are furtheraway are moving away faster.
So that tells you basicallythat the universe is expanding,
which then kind of leads you to the idea

(16:09):
that, oh, in the past itmust have been smaller.
What's the end point of that?
Is all of the universebeing in a very small place
and they're being kind of a big bang
that makes it expand out from there.
So if you just imagine there's a big bang,
everything starts expandingaway from everything else
and then gravity is tryingto pull it back together.

(16:29):
You might think there're kindof three possibilities there.
So one would be like theBig Bang gives it a kick,
it expands and then gravitystarts pulling it back together.
And then gravity is strong enough
to pull everything back together
and everything collapses againand there's a big crunch.
Option two is there's a big bang,
gravity is trying to pulleverything back together

(16:52):
and it's just not quite strong enough
to pull everything back together.
But everything kind of stops
or slowly drifts down to zero speed.
- So it's expanding but it's slowing down.
- Yeah.
- Until it reaches an equilibrium
and stays there?- Maybe, it's pretty hard
to hit a perfect balance like that.
So then the third optionis the big bang kick

(17:13):
is big enough that gravitycan't pull it back together.
It tries, but as you get furtherapart, gravity gets weaker.
So then it's sort of, youhit a constant drift rate
where everything's drifting further apart
at a constant speed, basically.
The mystery of dark energy,
which was discovered in the '90s
is that there's adifferent thing going on.
Not only the driftingapart at a constant speed,

(17:34):
it's drifting apart andthere's an acceleration
that's pushing it faster than that.
It's like not only was there the big bang,
there's something else that'scontinuing to give it a kick.
So there's something thatwe don't know what it is
and things that we don't knowwhat they are in astronomy,
we call them dark.
So we've got dark matter,we've got dark energy,
we dunno what they are.
And it's just making the sizeof the universe accelerate,

(17:57):
like grow larger and speedup right in its growth.
And it's a basically amystery of what it is.
When Einstein firstwrote down the equations
for general relativity
that there is a term in those equations
that Einstein put in tokeep the universe stable,
to keep the universe from collapsing again
'cause Einstein wanted theuniverse to be able to be stable.

(18:18):
And then with Hubble's findings,
Einstein called that his greatest blunder.
But then it turns outthat that same factor,
that same constant in the equations,
if you make it negative,it gives you dark energy,
it explains dark energy
or like at least appears in the equations.
That doesn't really help us
to understand what it physically is.

(18:38):
Is it something that wecan ever interact with
in any kind of real way or is it just like
a fact of the way spaceand the universe works?
There are lots of ideas
about what dark energyis or how it could work
and with DESI we're basically just trying
to go out and make the measurements
and those measurementswill help to disentangle

(19:00):
or to tell the differencebetween different models
of what dark energy might be.
So the goal of DESI is tomeasure the size of the universe
at different times in the past.
So basically we're trying to chart
that growth of the sizeof the universe over time
and different models of whatdark energy will predict,
different shapes of that curve

(19:20):
of how fast the universe grows over time.
So by just going out andmaking the measurement,
we should be able to kindof tell the difference
between different models of dark energy
and help to rule out somepossible explanations.
- When you mention over time,
you don't mean you doan observation one week
and then the next week and the next week,
you mean over like cosmic time, right?

(19:40):
You're essentially looking back
at where galaxies werebillions of years ago
versus where they were, I dunno,
another amount of billion years ago.
Is that generally fair?- Yeah, that's exactly right.
- And how can you tellhow fast they're moving?
Or if you know were they atone point and another point,
then you know the speed of acceleration?

(20:00):
- So like you said, on human time-scales,
basically the extra-galacticuniverse is static.
We can see the stars moving,they don't move very much.
But with precision instruments
you can tell that they're moving.
But the galaxies more or lessare stationary on the skies
to the precisions that we can measure.
Distances in cosmologyare really complicated.
(both laugh)

(20:21):
It's hard to just talk aboutthe distances between things
when the whole fabric thatthey're sitting on is growing.
So distances in cosmology are complicated.
So the two things we can really measure
are angles on the sky and redshifts.
So redshifts, lots of peoplehave heard explained before,

(20:41):
but basically the light from the galaxy,
if you break it into a rainbowhas a certain signature.
And what we observe is not that signature
as we'd expect to see it,
but that signature shifted.
It's sort of like theDoppler effect when you know,
when you hear the train goesfrom moving towards you,
from moving away from you,
the whistle shifts from higher to lower.

(21:03):
So if you're talking about light,
lower is redder toward the red.
So what we observe is allthe galaxies signatures
are shifted toward thered by different amounts.
So they're redshiftedby different amounts.
And that observation from Hubble was that
galaxies that are more distantare more shifted to the red.
So that's one thing we canactually measure, redshifts,

(21:24):
and that's what DESI's real thing is.
The other is angles on the sky.
Another thing that DESIis very good at doing,
because we have to knowwhere the galaxies are
to actually observe them.
So the thing that lets ustie those two things together
and measure the scale ofthe universe over time
is this nice little featurethat the universe gave us.

(21:46):
A little bit after the Big Bang
the universe was this, we kindacall it a hot soup I guess,
of plasma and photons.
Basically, everything's sohot that there aren't atoms.
There's basically just abig roil of plasma and light
and it's all exchanging energy
and it wasn't uniformly spread.
There were kind of denserand less dense spots.

(22:09):
And that soup kind of allows things
like sound waves to propagate.
So if you have like a dense spot,
you get a ring that comes out from it.
And then there's a magical point
380,000 years after the Big Bang
where the universe hasgrown and cooled enough

(22:29):
that plasma can cool downand you can form atoms.
It's not a soup anymore.
The photons kind of get liberatedand are allowed to escape.
But those rings of over densities
are frozen-in at that point.
- They're sort of imprinted for good?
- That's right. They'reimprinted for good.
We can see them by observingthe light from that time.

(22:49):
That light is now reallyredshifted into the microwave
and we can see it in all directions.
And it's called the cosmicmicrowave background.
It's currently threedegrees above absolute zero.
So it's at three Calvin.
- It's chilly.- Yep. (laughs)
And it looks like it's threedegrees in all directions,
but if you make very,very precise measurements,

(23:11):
you see that there are little variations
above and below that three degrees,
1 part in 10,000 where youcan just see the places
that were brighter and colder,
more dense and less dense at that time.
And the parts that were more dense,
remember our good old friend gravity,
pulls all of that matter together
to form stars and galaxies.
So that little ring

(23:33):
that was frozen-in at thatpoint has stuck around.
So what we get to observe
is that if you look at a single galaxy,
galaxies aren't spread uniformlyon the sky, they cluster.
Around a galaxy, you're likelyto find other galaxies nearby
and then they sort of drop offin density around the galaxy.
But then at the radius of that ring,

(23:53):
there's a little bump whereyou're a little bit more likely
to find another galaxy.
It's about 1% more likely.
It's a little bit of a subtle signal.
The universe is very kind togive us anything but it's-
- You may not wanna place money
on it being there all the time 1% off.
- Well by building DESI,
we've placed a lot ofmoney on on it being there.
But the beautiful thing about it is that

(24:14):
that scale was frozen-in,
there's kind of nothing you can do to it
to change what that scale is.
So it just basically gets stretched along
with the fabric of the universeor the fabric of spacetime.
So what we can do, finally, with DESI
is measure the angular scale
of that feature at different redshifts.

(24:34):
- Right.
- Whew.(Colin laughs)
Remember when I said distancesin cosmology are complicated?
- Yes. Yeah.
- It's a long way to go from-
- It's not how we think of,you know, driving distances.
This is, it's a verydifferent sense of distance.
- Or just taking out a ruler or something.
- (laughs) Well, so thisis called a standard ruler
because it's a thing
that we think we know the physical size of

(24:56):
and then we measure whatangular scale on the sky
it fills at different times.
If you think about thisin your everyday life,
you take a ruler and youserve it at arms length,
it fills a certain angle, right?
If you move it twice as far away,
it fills half the angle and so on.
So the weird thing about cosmology is that
that doesn't hold becausethe universe was growing

(25:18):
while all of this was going on.
That angular diameterdistance, it's called,
it's one of many differentkinds of distances in astronomy,
angular diameter distance,
gets smaller as things get further away,
but then it turns over andactually gets bigger again.
Things that are very distantare actually bigger in the sky.
You know, with DESI weget to kind of chart out
this angular size of aruler of a known size.

(25:41):
- And have you personallybeen one of the people
who pokes tiny holes in aluminum
and feeds fiber optic cables through them?
Have you been there on thesite doing this kind of work?
- So it's embarrassing.
I'm like an expert onsome of these telescopes
that I've never been to
and the Sloan telescope is one of them.
I've still not managedto get to that site.

(26:01):
So in these projects,they're large projects,
they have 100s of peopleinvolved, usually,
dozens of institutions.
So we do complicated time tracking
to keep track of like whohas actually contributed
and I'm a, what am I?
I'm an architect in the SDSS project
but I still haven't managedto go to the telescope.
It looks nice.
(Colin laughs)
I have seen the machine shopin the University of Washington

(26:24):
where they drill the holes
but that's not quite as glamorous.
- You were telling us beforethat a lot of your work
was in this pre-analysis stage
to decide where theinstrument should be pointed.
What are you doing now that
that pre-analysis, I guess, is finished?
- It's funny being involved
in these projects from the early part
because our work was mostly done

(26:44):
by the time the instrumentwas on the mountain
mounted on the telescope,taking observations.
Because we're trying to measurethese really subtle signals
where there's like a 1% more galaxies
at a certain radius than you'd expect.
It's pretty important
to understand not onlythe ones you observe
but the ones you don't observe.
So we go to a lot of effortto track all of the effects,

(27:05):
all of the statisticaleffects that can cause us
to not observe a galaxyor observe more galaxies
on a certain part of sky than uniform.
For that reason, to make thebookkeeping easier, basically,
these projects usually freeze the sample
like we choose the set ofgalaxies we want to observe
at the start of the projectand then hold that fixed.

(27:25):
Like just proceed with thatplan for the next five years
in the case of DESI.
Our work had to be done beforethe main survey started.
So one of the things I'm doing
is figuring out what weshould do with DESI next.
It was funded for a five-yearmission or five-year survey,
but at the end of that timeit's still gonna be the,
or at least one of the best instruments
in the world for this work.
So we're currently kind oftrying to devise some plans

(27:48):
of what to do with it next,
which is kind of a combination
of an interesting science case
and a feasible set of galaxies to observe.
And part of that might involve going out
and doing more imaging.
- Are you confident that
the mystery of dark energy can be solved
or maybe will be solvedthrough some of these efforts

(28:09):
and the ones that will follow?
- That is a fascinating question.
- I know it requires some optimism
and you don't have all the information
but there's a lot ofprogress being made it seems.
- Yeah, it's one of thebig mysteries in cosmology
so we're putting in a fairbit of effort toward it.

(28:30):
The thing that is a challenge
is that all of the currentobservations point to it,
are consistent with it being kind of
the simplest explanation,
which is kind of thatcosmological constant
that Einstein's equations allow.
So everything so far is consistent
with kind of the most boring explanation,

(28:51):
which is still like mind boggling
and really difficult to understand
or like to have a a reallike intuitive sense for.
We don't really havean explanation for it,
it's just kind of like,
it's just a fact of how space behaves.
That there's this weirdfluid kind of thing
that pushes space apart (laughs)

(29:13):
and when you push spaceapart you make more space
and then there's more of that stuff in it
that's pushing it apart more.
It's pretty noodle-bending.
- Yeah. I was gonna say.
(both laugh)
Yeah, I saw it described sort of like:
if you had a balloon, justa normal party balloon
and you squeezed it, the analog would be
the balloon would justkeep collapsing even after,

(29:34):
it wouldn't resume it's original shape.
But in this case, no matterwhat you do to the universe,
it seems to be acceleratingand getting bigger.
- Yeah, I guess withDESI it's possible for us
to make this nextgeneration of measurements
of like how big the universe is over time.
So for some of us that is good enough
the fact that it's there and we can do it.
And those measurementsthen kind of push theorists

(29:56):
toward coming up withdifferent explanations
or refining their explanations.
A lot of cosmology ends up beingthis kind of back and forth
between theory and observation
and computation and simulation.
So basically this is just our next step
on the observational sideis to make the measurements
and see what the theorists can do with it.
- And you mentionedobservational astronomy
being more of your bread andbutter than radio astronomy,

(30:19):
but you're also involvedin radio astronomy.
And until you told us this couple days ago
when we were chatting,
I never really made thedistinction in my head
that there's two different,or at least two different,
could you tell us sort of the difference
and then maybe tell us how you work
in radio astronomy as well?
- Yeah, it's funny,
astronomy is not that bigof a scientific field,
but we're still split into these silos

(30:41):
and part of it is justbasically technologies.
The trick with observational astronomy
is focusing and capturing the light
and the tools you need to do that
depend on the kind of lightyou're trying to gather.
So for optical astronomy, thewavelengths are really short.
So if you wanna make a mirrorthat focuses that light,
it has to be ground really precisely.

(31:01):
It takes years to makean astronomical mirror.
And when new projects get funded,
that's often the first thing they do
is book a spot in the mirror lab
to get their mirror built and polished
because that will take as long
as the rest of the project put together.
- 'Cause even if there's a tinylittle defect in the mirror
it could ruin everything right?
- As long as the whole thingis basically the right shape,

(31:23):
you can get away with smallparts of it being imperfect.
But if the whole thing is the wrong shape,
then you're just in a world of hurt.
So when Hubble was originally launched,
it had this issue and thatjust means that you want
all of the light thatcomes from a distant point
to bounce off your mirror
and hit the sensor at the same place.
And if your mirror's the wrongshape, that doesn't happen.

(31:44):
If your mirror is too rough,
then that also doesn't happen
because the wave's hittingdifferent parts of the mirror
instead of adding together,
interfere with each other and subtract.
So in optical astronomy
the mirrors have to be just beautiful.
In radio astronomy, thewavelengths are really long.
So in CHIME,
this experiment that I'm involved with,

(32:05):
the radio waves are like40 centimeters long.
So if you wanna make something
that looks like smooth to a radio wave
that's 40 centimeters long,
it doesn't have to be very smooth.
You know, it has to be
like within millimeters kind of smooth.
So radio telescopes,
the mirrors or reflectorstend to be really cheap
compared to everything else.
In CHIME they're madeoutta a kinda metal mesh.

(32:27):
But then the challenge
is collecting thatlight and processing it.
So radio astronomy'soften kinda thought of
as chicken wire and supercomputers.
- I love it.
- I do too.
- So I love how you saythat radio astronomy
is basically chickenwire and supercomputers.
What really is the roleof the chicken wire?

(32:48):
- The chicken wire is the mirror
or the equivalent of the mirror.
I'm kind of by trainingan optical astronomer
so it's really bizarre to beworking in radio astronomy
where the light acts so differentlythan what we're used to.
But as far as a radio wave is concerned,
a parabolic-shaped mesh ofwire looks like a mirror
and it can focus it

(33:09):
so it bounces right off the chicken wire.
And if your chicken wire'sshaped in just the right way,
it can focus it onto a place like onto,
in the case of CHIME, onto the antennas.
So the half-pipe shape is a parabola,
so it focuses all of the light
coming from one point on the sky
to a point onto the antenna.
- You mentioned CHIME, weshould explain a little bit.

(33:30):
It's not like anytelescope I've seen before
and when I first saw it,
I don't know if I would'veguessed telescope,
I might have guessed skateboard park.
So can you tell us what CHIME is
and why it's like the it is?
- Yeah, CHIME is wonderful.
CHIME is the Canadian HydrogenIntensity Mapping Experiment.
You can see why you just use the acronym?
And it's a radio telescope

(33:52):
at the Dominion RadioAstrophysical Observatory
near Penticton, British Columbia.
So it's a really unusual telescope design.
It doesn't focus light in two dimensions,
it only focuses light in one dimension.
So it's made out of these parabola-shaped,
like half-pipe-shaped tubes.
So it focuses light in thedirection across the tube

(34:16):
but not the direction along the tube.
So if you have light comingfrom a distant galaxy, say,
it hits the reflector andit's focused onto a line
along the middle of that half-pipe
and then CHIME has a bunchof antennas along that line
that gather all the lightand then it goes into
our handy supercomputer.

(34:38):
- Behind the chicken wire?
(both laugh)
- That's a-- That's a different part?
- That's a different challenge, so...
- Yeah, we'll get to that.
- Yep.
And so the cool thing about that
is that you can focusin that other dimension
after the fact in the supercomputer.
So if you think about a star
that's to the north of the telescope,

(34:59):
it will hit the northernpart of the half-pipe
sooner than the southern part
and all those waves will bounce up
to the antennas along that line.
In the supercomputer, then takethe northernmost telescope,
sorry, northernmost antenna
and then take that value,
the antenna just to the south of it,
and delay it a littlebit and add them together

(35:21):
and take the one just to the south of that
and delay it a little bit more.
You can add together the waves
that hit the telescope at different times
and that basically like acts
as though you tilted thetelescope by that amount
so that they would hit at the same time.
- 'Cause the telescope itself,it doesn't have moving parts.
- Yeah, the telescope is huge.

(35:42):
It's 20 meters wide.
And sorry, each half-pipeis 20 meters wide.
There are four of themand it's 100 meters long
and it's heavy and huge.
Yeah, so it has no moving parts.
We can't steer it in any direction.
It basically just sees a strip of the sky
and then the Earth conveniently rotates.
So we get to see basically half of the sky

(36:02):
or two-thirds of the sky every day.
- That's handy.
Nice of the Earth to do that for you.
- It's pretty kind.- Yeah.
- But the cool thingthen is that you know,
we can by more or less delaying the signal
from the different antennasand adding them together,
it acts like a telescope ispointed in a certain direction
but then if you just delayit by a different amount,
you can point it in another direction.

(36:23):
- And this is all done by software?
- Yeah, that's right.
It's all done in software andyou can do it all at this,
you can point it inall of those directions
at the same time.
And then with the four half-pipes
you can combine those in different ways
and point it in-software in the other,
in the east-west direction as well.
- And this is not dark energy search,
this is a different or is it related?

(36:43):
- It is related.
So the CHIME telescope was built
for doing this thing calledhydrogen intensity mapping,
the HIM part of CHIME,
And the idea there is thatas you go further away
or farther back in cosmictime or to higher redshift,
it gets harder and harderto observe galaxies
'cause they're just faint.

(37:04):
So doing this trick that we do in DESI
of trying to measure galaxies
and then measure the slightlymore likely to observe one
at that magical distance away,
that trick just gets really hard
'cause the galaxies are faint.
And the thing that's kindof frustrating about it
is that you're gonnameasure a bunch of them,
but you know that they cluster
and like you have to measurea whole bunch of them

(37:25):
to kind of map out this cosmic web.
So the idea with hydrogenintensity mapping
is let's not measure individual galaxies,
let's just measure all ofthe hydrogen collectively.
And that hydrogen isaround all the galaxies
and along the cosmic weband filaments and everything
so that to understand the growthof the universe over time.
So CHIME was built to do that experiment

(37:47):
and they're trying tomap range of redshifts
that slightly overlap DESI,
but go further than we canreally go with galaxies.
So it's looking backcloser toward the Big Bang
with this totally different technique
of mapping hydrogenwhich emits in the radio
and then gets stretched out.
So I'm not actuallyinvolved in that side of it,
the cosmology side, thehydrogen intensity mapping side.

(38:11):
And this is another kind of cool thing
about radio telescopes.
While CHIME was being kind of proposed
and built and designed,
people realized that it wouldalso be really well-suited
to uncovering anotherastrophysical mystery.
The mystery of fast radio bursts.
So fast radio bursts
were first discovered in 2007.

(38:35):
(both laugh)
- That's recent, that'snot that long ago in-
- Yep, exactly.
And they were discoveredin archival observations
or rather the first one
was discovered in archival observations
and what fast radio bursts are
or what we observe are these really brief,
they're like a millisecondlong, burst of radio light.

(38:55):
That's the (laughs) quick, they're fast,
they're in the radio, they're bursts.
- Oh, it's a good name for them. Yeah.
- Yep, and in the time
between the first one discovered in 2007
and when CHIME was being constructed,
a few more had been discovered.
So they were getting tobe not a one-off event
but something that kindaexisted in the universe

(39:16):
that we could possibly go outand try to measure a bunch of.
So the fact that CHIME can see
a huge chunk of the sky at once
and observes the whole sky once a day
thanks to the Earth rotating
makes it a really good instrument
for searching over the whole sky
for something that you don't know
where it's gonna come from.
So funding was secured

(39:38):
to build an addition tothe CHIME's telescope,
which was just a fast radioburst search part of CHIME.
So it's called CHIME/FRB.
So remember how I said in software
you can focus the telescopeat different directions.
Basically we ask that supercomputer
to do some different computations

(39:58):
and send the data to the CHIME/FRB system,
which is itself anotherlittle supercomputer
that does this real-timesearch for fast radio bursts.
So all over the sky.
- When you say a real-timesearch all over the sky,
is this where the big data comes in?
Lots and lots of data?
- Yeah, that's right.
So the CHIME correlator, that's the,

(40:20):
well one of the supercomputersinvolved in this whole thing,
focuses the light in 1,000spots in the sky for us
and breaks it into 16,000frequency channels.
So you know when you're tuning the radio
and you can choose different FM stations,
we have 16,000 stations to choose from.
Some of them are just full
of people's cell phone LTE traffic.

(40:43):
(both laugh)
Thankfully we can just ignore those ones.
Everyone has a radio stationthey don't like, right?
- Yeah. Just tune them out.
- Yep. Just skip those ones.
- But how many of them aretaken up by the cell phone?
- More and more.
- It's a noisy world withall the communication?
- It is a noisy world. Yeah, that's right.
We lose 10 or 20%.

(41:05):
It's pretty bad.
- But it's kind of a consistent range?
- For the most part.
The 4G LTE bands are just lostto us entirely. (chuckles)
And then there's some other ones
that come on and off periodically
that we have to filter out.
So anyway, the correlator sends us
1,000 places on the sky, 16,000 channels,
and the brightness in eachchannel one time per millisecond.

(41:29):
- Okay.
- So that's 1,000 times 1,000
times 16,000 per second.
And that is basicallyjust too fast for us.
It's too much data forus to write to disc.
So those signals get sentto this set of 128 computers
that are searching throughthe data in real-time
looking for the signatureof a fast radio burst.

(41:49):
So I said that they're a burst,
but they're a burst at their origin
but then they have to travel
through a bunch of space to get to us
and space isn't quite empty.
So when those radio wavesinteract with electrons,
what happens is the highfrequencies arrive first
and the lower frequencies arrive later.
It's called dispersion.
So what we observe is thatthere's kind of a sweep down

(42:13):
from high frequency to low frequency
that can be tens of secondslong or like a minute long.
So this real-time search hasto store like a minute of data
and look for kind of all thepossible different sweeps down
depending on how many electrons
were between us and the source
that determines the shape of that sweep.

(42:33):
So it's searching for allthese different sweeps
corresponding to kindof different distances
of the fast radio burst being away from us
for these 1,000 places onthe sky simultaneously.
And then basically,
if we find somethingthat looks interesting
we write down just the dataaround that place on the sky
and that little chunk oftime for later analysis.

(42:54):
- So in those cases you'll saveeverything that's coming in,
but most of the time
you'll just get rid of most of the data?
- Yeah, that's right.
So we'll save everything that comes
to the CHIME fast radio burst side
that's been reduced a lot already
from the raw data rate collected
by the first supercomputer in the chain
for things that are really bright.

(43:14):
We'll also ask that one,
it also saves a little chunk of past data
and we can ask it to alsosave a little chunk of data
around the sweep.
That one collects 800gigabytes of data per second.
So we only ask it for a 10th of a second
around where the sweep was.
- Wow. Sorry, how muchper how little time?

(43:35):
I'm trying to wrap my head around this.
Like in the sense of data,the way we understand it,
this is enormous right?- Yeah that's right.
800 gigabytes a second.
So if you go out and buy thebiggest hard drive you can,
these days, say 12 terabytes,
that fills up in like 15 seconds.
- And this is the data toCHIME or just CHIME/FRB.

(43:57):
- That's the data to CHIME. Yeah.
So that's reading all of the voltages
from all of the antennasalong the half-pipe of CHIME
that then can get addedtogether in different ways
to point the telescope indifferent directions on the sky.
- You told us the otherday when we were chatting
that just the sheer volumeof data is equivalent to,

(44:18):
or it's a portion ofthe entire data exchange
on our cell phonenetworks in North America.
- So yeah, I looked it up.
It's a moving target but if you look
at the international datatransfers on the internet,
inside the CHIME supercomputer,it's doing 1% of that.
So 1% of the world internet traffic

(44:39):
is being exchanged withinthat CHIME correlator
to do those additions of likethe pointing the telescope
at different points on the sky.
- And it's doing that over and over again.
- Just continuously.
- It's amazing.- Whoa.
- Yeah, during the dayradio telescopes don't care.
We can see the sun
but it's not the brightestthing in the sky.
Rain is a little bit of a downer.
- And you mentioned airplanesare a bit of a pain as well.

(45:00):
- Airplanes are terrible.
It's not so much the signals
that the airplanes themselves are emitting
as far as the radio waves are concerned,
they're a mirror in the sky so we can
like see over the horizondown to the noisy cities
and cell phones and other things around.
The CHIME telescope's not that far
from the Kelowna Airport.
So we see many, many airplanesand have to filter them out.

(45:23):
- The Milky Way's in ourway, waters in our way.
All these things
we take for granted.- Noisy world out there. Yeah.
- And where do youactually process this data?
- So for CHIME it's almost all on-site
just because the data rates
are too big to move anything off,
it would be way too muchtraffic to try to compute,
like to move it somewhereelse and compute there.
So all the computing isdone on-site basically.

(45:45):
- When you say on-site,
my first thought maybe wouldbe this huge bank of computers
in a sophisticated room with monitors,
but there's steel shippingcontainers on site, right?
- Yep. Steel shipping containers.
Good old 40' shipping cans or sea cans
are kind of the building ofchoice to stick these things in.
They're cheap enough to get and robust.

(46:05):
So yeah, one of the challenges
is that a big computer cluster
is itself really noisy in the radio.
It emits a lot of, it justmakes a lot of electrical noise.
So inside of the steel shipping container
we also have to build like a shielded room
that the computers can go in
so that they don't make a bunch of noise
that we then hear with the telescope.
- So there's natural challenges

(46:25):
and challenge that we create ourselves
with our technology thatwe have to get around.
- Yeah, that's right.
And the kind of fun thing
is that because the radiowaves are pretty long,
if you drill a small holein the shipping container,
the radio waves can't get through it.
So the shipping containershave all of these, you know,
basically small holes whereall of the cables and power
and cooling and everything comeinto the shipping container

(46:48):
and into the supercomputers inside.
- I'm wondering if you can also speak
maybe a little bit more broadly
to a challenge that you might face
when collecting all ofthis data in an experiment
and then having to figureout how to store it.
And maybe we can play thequestion from Dominica.
- My name is Dominica,
I'm a student at theYachay Tech University

(47:08):
and the PSI Start Program.
I was wondering if, isit a fundamental issue,
the fact that computationsdepend on the discrete
whereas the physical lawsdepend on the continuum?
- Yeah, that's a deep question.
The physical world iscontinuous as far as we observe.
Quantum theorists might argue about that,

(47:28):
but at our scales it's continuous.
But we have to do all this.
Our current computing is all discrete.
So in CHIME the antennas
are really measuringthis continuous signal.
But those come through cables
into the first supercomputer in CHIME
and basically the first thing we do
is turn them into digital signals.
So there's a resolutionproblem there basically

(47:51):
where you have to choose how many bits
to use to represent it.
So if you look at yourcomputer display, you know,
it sort of looks like itcan make all of the colors
that you can observe, right?
But modern computerdisplays use eight bits
for each of red, green, and blue.
So they can make 256 different levels

(48:11):
of red, green, and blue.
And that's enough that we
kind of can't distinguish between them.
So as far as like, you know,
we can observe with our eyes or our brains
that's fine enough that a discrete set
of levels looks continuous to us.
And it's kind of, it's a littlebit similar in the radio.
It turns out that partly becausewhile the world is so noisy
and in radio you have to add together

(48:33):
a lot of individual samples
before you actually measuresomething significant,
it turns out that it's okayto do that discretization
or conversion from analog to digital.
In CHIME actually they only use four bits.
So there's only 16 levels of the signal
and that's still enough to kinda recover
the continuous phenomenathat are observed.

(48:54):
- CHIME has been extremelysuccessful in this FRB mission.
The fast radio bursts, they'rea relatively new phenomenon
and then there was only a few detected.
And then with chicken wire andsupercomputers and ingenuity,
CHIME ramped up the game so to speak.
Can you tell us, youknow, what it's discovered
and what we're learningabout fast radio bursts?
- Sure, so when CHIME came online,

(49:16):
there were about 50fast radio bursts known
and intriguingly one ofthem was seen to repeat.
So there's not only just one boom,
but then the same one wasemitting multiple bursts,
which really threw thetheorists for a loop
because some of theirexplanations required the thing
to be destroyed to make a burst of energy.
The challenge is that fast radio bursts,

(49:37):
we've now discoveredthat they're far away,
which means that they'reintrinsically really bright.
So it's hard for theoriststo come up with ways
of kind of generatingthat much radio energy.
And if you don't get to destroythe thing in the process
then that puts even more limitson what you can contrive,
what can think of ways of explaining
what they can possibly be.
Right, so when CHIME cameonline, about 50 were known

(49:59):
and the fun thing is there was a catalog
of known fast radio burstsand there was also a catalog
of theories of what they could be
like, possible explanations
of what could produce a fast radio burst.
And there were more theories
than there were fast radio bursts.
(both laughing)
And then CHIME, in the first two months
while we were still kind ofputting the thing together,

(50:20):
the chicken wire was in place,
but the supercomputerswere still being built,
discovered 13 new onesand one new repeating one.
And then after the firstyear of observations,
our first catalog paper has 492 sources,
including 18 repeaters.
So basically just blew the lid

(50:41):
off the fast radio burst game.
But I think a lot ofthe current feelings are
that the repeaters and the one-off bursts
are different populations.
Now the theorists can still destroy
the regular fast radio bursts,
but then they still have to explain
where the repeating ones come from
through some other mechanism.
- You've mentioned a term
that I just love in ourprevious chat, sad trombone.

(51:03):
That actually has ameaning in this research.
What is a sad trombonein the CHIME effort?
- (laughs) This was one of those,
like when the term was coin,you knew it would stick.
So the repeating fast radio bursts
tend to have this structure.
They're not just a single burst,
they kind of have a burst
and then maybe a few milliseconds later
a repeat at a lower frequency

(51:25):
and then it'll often in three like,
so they'll sort of have ainitial burst lower and lower.
So it's like whomp-whomp-whomp.
- Sad trombone.- Sad trombone.
- But it's only theserepeating FRBs that do this?
- One of the things that
the CHIME data really contributed to this
is kind of understanding the diversity
of the fast radio bursts.
Like some of the non-repeatingones cover the whole band.

(51:48):
Like we see them being bright
all across the frequenciesthat we measure.
Some of them are just bright in the top,
some of them are justbright in the bottom,
some in the middle even.
Some are really briefand some are scattered,
which you get through kind of
traversing different kinds of material
between us and the source.
Part of the beauty of doingthis large-scale search,

(52:08):
observing 1,000 placeson the sky all the time
and observing the northernhalf of the sky every day,
is that we get to build upstatistics about what they are
and collect it in a kind of uniform way
so that it's much easierto try to understand
what the real population is
before whatever affects causeyou to observe some more,
like the unable to observe some or others.

(52:29):
So it looks like many of the repeaters
have the sad trombone.
So now sometimes if wesee a new burst in CHIME
and it has the sad trombone structure,
we'll say, "Oh maybe thatone's gonna come back again."
- Is there a prevailing theory or theories
about what these things actually,
what's causing these distant bursts?
Or do you need to do your cataloging

(52:50):
and tracking them firstto even come up with
an explanation of what they could be?
- One thing is just thatthey're fast, right?
So they're a millisecond long,
so it's really hard to generatesomething a millisecond long
from some astrophysical thing
that's bigger than alight millisecond in size,
just 'cause you know,
you have to emit it all at the same time

(53:11):
from all over the source.
So you know, you can'treally generate something
that's that short from something
that's like the size of the sun
'cause it just won't allarrive at the same time
so it won't be a millisecond-long burst.
So that pushes you towardthings that are small
and one of the like families of things
that could be are neutron stars.
So if you start with a star that's,

(53:33):
I forget the numbers exactly,
8 to 20ish times heavier than the sun.
It goes through its life burning hydrogen
and then burning some other things
toward the end of its desperatelife trying to stay a star
and eventually runs outta fuel
and collapses to a neutron star.
And neutron star materialis really bizarre

(53:56):
'cause you take all of like,
say something most of the size,
like bigger than the mass of the sun
and squeeze it down to10 kilometers in size.
There aren't atoms anymore.
Everything's been squeezed so far together
that it's just like abig ball of neutrons.
So it's really bizarre.
One teaspoon of neutron star material
weighs billions of tons.

(54:17):
Like it's just mind boggling.
- Right, it really doesmake the mind reel.
- Like it's a number
that you just can'treally like comprehend.
So they're pretty weird. (laughs)
But the other interesting things are that,
like when this process happens,
if the star was spinninginitially, it keeps spinning,
but now instead of you know,

(54:38):
a very stately slow rotationof something the size of a sun,
if you can picture afigure skater spinning
and then pulling in their arms
and spinning faster and faster and faster,
imagine that just continuing on to go.
Instead of spinning, you know,
once a week or once a day or something,
some of the neutronstars that are observed
will spin like 1,000times a second or more.

(54:59):
So they're the likeincredibly heavy things
that can be spinning really fast.
And similarly their magneticfields, they often keep,
So then you have somethingwith a magnetic field
that's spinning really fast.
If you're a theorist,that's good ingredients
to make something thatcan emit radio waves.
So these pulsars areknown, like neutron stars

(55:19):
that are observed to emitperiodic pulses of radio waves.
They were first discovered
in 1967 by Jocelyn BellBurnell who is amazing.
Some of the theories for whatfast radio bursts could be
are kind of exotic types ofneutron stars of some kind.
The problem is that the fast radio bursts

(55:40):
are like millions of timesbrighter than neutron stars
that we know in the Milky Way.
And you can't just make them bigger
because if you make them too big
they collapse to black holes.
So you can't just makea bigger neutron star.
There has to be kind ofsomething else going on.
We got another kind of clueor a hint maybe in 2021.
There was a fast radio burst
from a neutron star in our own galaxy,

(56:02):
a special kind called a magnetar.
So it has kind of neutron stars
with really extreme magnetic fields.
And CHIME observed that,like we caught that one,
we saw it go streaming by
and we said, "Ooh, that's interesting."
And it kind of has anenergy that's in between.
So it's a few 100 times brighter,
I think, than usual pulsars.

(56:24):
So it's kind of filling in a bit of
that factor of a million you need
to get to fast radio bursts.
So maybe they're an extreme,
kind of this extreme kind of magnetar.
So there're kind of hints and clues,
but it's still a pretty big mystery
and we keep kind of finding odd things.
Another thing discovered lastyear, or the year before,

(56:47):
by a graduate student in the CHIME group
was that one of therepeaters not only repeats
but it repeats on a clock.
She found that if shetook all of the pulses,
she was looking at all
when we had observed the fast radio bursts
and she said it looks likeit's repeating every 16 days.
So she took the signal and like folded it

(57:09):
and found that all of the bursts
come within a five-dayperiod around that 16 days.
So it's like, you know,
on for five days and then off for 11 days,
on for five days off for 11.
And most of them appear
within like a one-daywindow around the peak.
So it's like mostly on and on day one
and then it's kind of on alittle bit for the next four days

(57:31):
and then off for 11 days.
So that adds anotherelement to the mystery.
And we don't know if allof the repeaters do this,
but maybe some of them we haven't,
maybe they have different periods
and we haven't observedmost of them for long enough
to be able to notice that.
So then that maybe makes you think
that maybe there's like a neutron star

(57:51):
and something else in a binary,like orbiting each other.
And then when you have that,
you can get it so that theneutron star is spinning
and it's sort of like a lighthouse
or like a top that's wobbling
and when you're lookingstraight down on the top
you can see a burst from it.
So maybe that's what's doing it
and that, you know,wobbles once every 16 days

(58:13):
and it's when it's pointed like more at us
that we see the bursts.
So now you know,
you make the picture moreand more complicated.
Like it has to be a reallyextreme magnetar in a binary
with something else that'sgiving it this wobble.
- The mystery remains.- Yep. The mysteries remain.
- Well that's the exciting part.
There's lots for you to do. (chuckles)
- It's really, it's the first time

(58:34):
I've been involved in a project like this
that's kind of broken open anew part of observing space
and is really just like finding
all kinds of cool things there.
So it's been reallyfast-paced and really fun.
And part of the wayCanadian projects work,
there are a lot ofgraduate students involved.
So a lot of the peoplemaking these discoveries are,
you know, people who areworking on their PhDs

(58:54):
or master's degrees, you know,
they're just at theforefront of this field.
So it's really exciting,
it's really neat to see allthe things they're discovering.
- On the topic of being at the forefront.
You have told us also thatlots of the work here relies
on being at the forefrontof computational technology
and we had a question sentin on the topic of GPUs.

(59:15):
This was sent in from Craig
in the IT and AV departmenthere at Perimeter.
- Hi Dustin.
I heard it mentioned hererecently at Perimeter,
this specific piece of hardware known
as an Einstein equation code GPU,
which is the graphicsprocessor from a video card,
reprogrammed to run
the Einstein equationcode for simulations.

(59:37):
I wonder if you could explainin a little more detail
what an Einstein equation code GPU is,
how one is programmed to runthe Einstein equation code
and how successful it hasactually been in simulations.
- I'm gonna first talk a littlebit about CHIME, I guess.
I said that, you know,
it's chicken wire and supercomputers,

(59:58):
multiple supercomputers in this case.
So in CHIME the firstsupercomputer it comes into
are these custom-built computer boards
that use FPGAs,field-programmable gate arrays.
And they're these kindof really low-level,
it's sort of like a computer chip
where you get to choosewhere the wires go.
So they're really difficult to program
but really fast at what they do.

(01:00:19):
Program them once and theydo a single task very fast.
The task that first computer has to do
is simple enough that this is achievable
and then it sends all the datato the second supercomputer,
the CHIME correlator that hasto do more complicated tasks.
You can't do that
in these reallydifficult-to-program FPGAs,
but it turns out that you can use

(01:00:40):
these GPUs, graphics processing units,
to do the computations.
And GPUs are harder to programthan garden-variety CPUs
but they're way moreflexible than like FPGAs.
So the CHIME correlator hasto use these GPUs basically
to get the amount of computationout that that it has to do.

(01:01:01):
And it uses 1,024 what were at the time,
very cutting-edge GPUs.
I love the whole thing,
I love all of thetechnology involved in it.
They're water-cooled andthe water kind of comes in
and goes over each GPU in turn
and we have sensors on them
and you can kind of seethe water heating up
as it goes through each GPU and cools it.

(01:01:22):
But yeah, basically these GPUs,
although they were originally built
for doing graphics for video games,
if you think about it,graphics for video games,
a lot of the tasks arelike running something
that's going to produce, a color say,
for each pixel on your screen.
And you know,
if you have a screen that'slike 2000 by 2000 pixels,

(01:01:42):
I'm making that number up,
then you have 4 million computations to do
but you're doing kind of thesame thing for each one, right?
So GPUs are kind of specialized
for doing relatively simpletasks but in massively parallel.
And that just turns outto be a really good match
to some of the tasks that we have to do.
'Cause in radio, you know,
for the radio astronomy computations,

(01:02:02):
it's the same task done alot of times in parallel.
So say 1,000 places on thesky or 16,000 frequencies,
that computation is the same for each one.
So it's basically, you know,kind of a fairly simple process
that you just have torepeat a bunch of times.
So that really works well for GPUs.
So GPUs are really widelyused for, also now,

(01:02:23):
a bunch of machinelearning or AI applications
because a lot of those problems
can also be phrased as doinga fairly simple operation,
a lot of times in parallel.
They're kind of just a way of
accessing a lot of computing power
at the expense that youthey're harder to program
so you have to put more effort
into describing theproblem you want to solve
and especially how to solveit in massive parallel.

(01:02:45):
So this Einstein equations,
this was actually work done by people
including my boss and office mate,
Erik Schnetter at Perimeter,
they work on computer programs
that solve the Einstein'sgeneral relativity equations.
So you might have heard itsaid that in general relativity
matter tells space how to bend

(01:03:08):
and space tells matter how to move.
So you know, when there's massit changes the shape of space
and then mass moves alongstraight lines in bendy space.
So if you're a mathematician,
that sounds like differential equations.
It's, you know, there's sort of two things
and they're affecting each other.
Those are equations that you can solve.

(01:03:29):
You know, if you put a bunch of mass down,
you can compute howthis space will be bent
and then you can compute
how the mass will movearound in that bendy space.
And you only need this
when you're dealing with reallyextreme kinds of situations.
So black holes often comeup, neutron stars probably,
but in order to understandsituations like that,
basically you can either try
to understand really simple situations

(01:03:51):
with math on a blackboard
or you can do computersimulations of them.
And those computer simulations
involve doing a lot of thesame computation in parallel
so they lend themselves to GPUs.
Erik's group have made implementations
of solving the Einstein equations on GPUs.

(01:04:12):
That's the sense in whichthere's a, you know,
a graphics card that cansolve the Einstein equations.
- Right, yeah. That's fascinating.
I knew that that question was coming up.
I was looking forward to your answer
'cause that's an area thatI know very little about
and now I know somethingas opposed to nothing,
thanks to you.
We have two more questionsfrom students. Let's hear.

(01:04:33):
- Hi Dustin. I'm Summer from Waterloo.
If you could travelanywhere in the universe
to see something with yourown eyes, what would it be?
- Oh goodness.
I don't think I'd wanna putmy own eyes close enough
to a fast radio burst to see it.
- Let's say you're safe,
you're in a safe space vehicle somehow.
- Okay good with enoughshielding, (laughs)

(01:04:54):
I would love to see a fast radio burst.
'Cause what on earth are they?
You know, like I said, you have to,
the theorists really are working hard
to contrive scenarios thatcan make a fast radio burst.
So there's gonna be allsorts of wild stuff going on
around something that canmake a fast radio burst
is my guess or my hope at least.
Black holes of course, orlike the accretion disc

(01:05:15):
and like the, you know,
we don't see bendy spacein our everyday lives.
So there was a recent news article
of looking at light behind a black hole
and it's bent all the way around
or sometimes bends aroundand makes multiple laps
before it gets out and sees you.
So like we don't really experience
the fact that space is bendyso it would be pretty cool

(01:05:37):
to see bendy space around a black hole.
- I agree. (laughs)
And we have a second question
that may follow from the first.
- Hi Dustin, I'm Justina from Waterloo.
I was wondering,
what's the most fascinatingthing to you about the universe?
- Wow that's goingright to the core of it.
(all laughing)
One of the really bizarre things
is that the universe seems

(01:05:58):
to be like kind ofcomprehensible with math.
It's kind of bizarre that you can,
in cosmology you canwrite down like, you know,
a set of equations withlike five or six parameters
that kind of explain at the large scales,
like how the universe grows over time.
Like that to me is just bizarre.
The weirdest thing is that it seems to be

(01:06:20):
like comprehensible or likewithin the realm of possibility
that we could understand things
about the universe withlike basically math
and that we can like understandthings about the universe
by writing computer code
and that somehow people willpay me to do this for a job.
Like it's... (laughs)
- Yeah, I suppose you would be,
that job posting thatyour friends joked to you,

(01:06:41):
you had to go for it,Perimeter wouldn't have existed
had the universe not beensomewhat comprehensible
and that there would bemysteries for you to dive into.
- Yeah, well some people say that like,
we are like the universe'sway of understanding itself.
- Mm-hmm.
You mentioned that one ofthe downsides of your job
is you don't always getto go to the telescopes
that are doing the work
and you haven't been to CHIME

(01:07:03):
even though it's really closeto where you grew up, right?
- Yeah, it's just one mountain range away
from where I grew up inChristina Lake, British Columbia.
- It's a long way, it's over the mountain.
So yeah, you're fromBritish Columbia originally
and you still haven'tmade it to the telescope
that's one mountain range across the way.
- I know, I still have, mymom is quite upset. (laughs)

(01:07:25):
My work somehow hasn't contrived
to manage to make me go out there.
We have staff members on-siteand team members on-site.
So the goal is for the whole system
to be remotely operable.
From time to time we have to get somebody
to go and unplug somethingby hand or turn it off.
But for most of it, it's allset up for remote observation

(01:07:48):
partly because whenever people are on-site
they just, they tend to, not the staff,
the staff are very good,
but whenever we have visitors,contractors, whatever,
they never turn their cell phones off.
- And that interferes with-
- That's the loudest thing in the sky.
It's louder than anything in the sky.
So the fewer people onthe site the better,
actually for the most part.

(01:08:09):
During the building of CHIME
there was a huge amountof physical effort put in
as far as as like pullingcables, 'cause you know,
there's 2000 cables thatcome from the half-pipes
into the first supercomputer
and then hundreds of fiber optic lines
that come from that one tothe next computer and so on.
So there was a huge amount of effort,

(01:08:31):
but I thankfully came on the project
a little bit after that.
It was all in place.
It is still a huge treatto go to the telescopes.
I spent a lot of time at the DESI site
and at its twin telescope in Chile
and it's just beautiful up there.
It's a real treat too
to have the privilege toobserve from those places.
- Well, you'll have to get to CHIME
and then visit your mother or vice versa.

(01:08:53):
Your enthusiasm for this stuff,
especially the real mysteriousstuff is just infectious
and you know, I've learned so much
and my mind is reeling at some of the data
and the sizes and the scale.
So thank you so much forsharing with us today.
- Thank you. It was my pleasure.
(bright music)
- Thanks so much for listening.
Perimeter Institute is

(01:09:14):
a not-for-profit charitable organization
that shares cutting-edgeideas with the world
thanks to the ongoing support
of the governments of Ontario and Canada,
and also thanks to donors like you.
Thank you for being part of the equation.