Explorer les exoplanètes avec un supercalculateur !

Mysteries of the Universe
550 followers
Comments are disabled for this video.
Since James Webb's entry into service,
we often have the opportunity to talk on the channel
about exciting discoveries concerning exoplanets.
K2-18b would be an icy planet,
with perhaps aquatic life.
GJ98-27d would be enveloped
in an atmosphere of water vapor.
And more recently, Trappist-1e
would have its own atmosphere,
right in the habitable zone of its star.
Exciting results,
but that must always be put to the test.
It is based on a very small set of data
and several competing models
to best explain the latter.
The models in question are at the heart
of our knowledge of exoplanets,
and we don't talk about them much.
How can we say in a few transits
that Trappist-1c probably doesn't have
a thick atmosphere,
or that its companion Trappist-1e
is much more likely to have one?
Well, that's exactly what's going to interest us today.
We're going to take the time to understand
how we can, from a few data,
create increasingly complex models
of what happens on planets
located light years away from us.
It's a field in full development,
full of uncertainty,
but it's the best tool available
to explore, in a way,
all these new worlds that are revealed to us.
A model is a simplified representation
of a system we want to study.
We can model the economy of a country,
the behavior of a group of consumers,
the shape of a flame at the exit of a funnel,
or the conditions offered by the atmosphere
of an exoplanet.
In some cases, the models are very efficient,
and we'll see that in 99.99% of cases.
In other cases,
the performances are much less good.
The systems studied are so chaotic
that a good forecast is impossible.
This is for example the case
of weather models.
We will probably never see
a weather forecast at 1 year,
even with the help of a super-artificial intelligence.
The depressions, anticyclones
and other turbulence
are too sensitive to variations
of the initial conditions.
We can deploy hundreds of weather satellites
and the most powerful calculators in the world,
there is a limit to what is possible to do.
If we have trouble predicting
the time it will take on our planet
in three weeks,
you can imagine that we are not going far
when it comes to determining
if the water is at the right temperature
in the equatorial regions of K2-18P.
But what is good with the models
is that we can complexify them
or, on the contrary, simplify them to the extreme.
If they are simpler,
they will answer much broader questions.
This is a bit the approach
that is followed in the study of exoplanet atmospheres.
We start with very simple models,
the 0D models,
which only take very few parameters into account,
and we can go to very complex predictions
with the GCM models,
which are based on terrestrial climatology.
At each step, we add complexity,
we multiply assumptions,
but we can also speculate
on more and more precise phenomena.
The speed of wind,
the presence and composition of rain,
the difference between day and night phases,
it's a whole window of knowledge
that opens up to us.
To better understand what it is all about,
let's start with the simplest,
the 0D models, for zero dimensions.
In a 0D model,
we will consider the atmosphere
of an exoplanet as a uniform point,
without variation,
with the same temperature,
the same pressure,
the same composition,
the same general characteristics.
A simplification to the extreme.
Models of this type generally allow
to draw the energy balance for the planet,
and thus to form a good intuition
of their habitability.
A 0D model can thus be used
to determine a temperature of radiative equilibrium.
It is the theoretical temperature
that the planet would reach
if it absorbed all the energy
of its star directed towards it,
before radiating it into space.
We can make a fairly simple calculation
from the orbital parameters of the planet,
its albedo, and the brightness
of its star.
For Earth, we get with such a model
a temperature of equilibrium
of minus 18 degrees Celsius.
Compared to reality,
we are next to the plate,
of about 33 degrees.
The average temperature of the Earth
is actually 15 degrees Celsius.
A simple 0D radiative model
is therefore imprecise.
But it is not useless.
The error is a few tens of degrees.
It is a lot, but enough
to have a first idea of habitability.
The water remains liquid
between 0 and 100 degrees.
With an error margin of 33 degrees,
we can still go between these two edges.
And therefore quickly estimate
if a planet is surely a burning desert,
an icy ball,
or if even with an error margin,
it still has potential.
The 33 degrees of error
for the 0D modeling
of the Earth's temperature
is mainly due to the simplicity of the model.
It omits all the atmospheric effects,
including the greenhouse effect,
which explains for Earth
the difference between the modeled temperature
and the real temperature.
This is why it becomes necessary
to go beyond complexity
and to get a little closer to reality
by using 1D models.
Well, here 1D means
one dimension.
We must imagine that we extend
our modeling of the atmosphere
in verticality.
It is as if we took a line
that goes from the ground to space
and that we tried to understand
how the different effects
can vary depending on the altitude
or atmospheric pressure.
We do not worry yet
about latitude or longitude.
We consider that the atmosphere
is uniform at the poles
and at the equator, east and west.
The only dimension that interests us
is the altitude.
With such a model,
we will continue to study the effects
of lighting and radiation,
but we will also add
the chemistry of the atmosphere,
the greenhouse effect,
the thermal transport between
the different atmospheric layers
or the condensation conditions.
We are already adding a nice layer
of complexity compared to the 0D model.
This is what will allow
to understand the observations
but also to prepare them.
Indeed, these atmospheric models
have a double role.
They are necessary upstream and downstream.
Indeed, we did not expect
to discover the first exoplanets
to make such models.
One of the most famous
is the Casting model
which was published in 1993,
two years before the discovery
of the first exoplanet.
It has allowed to define
habitable areas around
different types of stars.
In fact, the models always precede
the observations.
That's how we know what to look for
and which instruments are necessary.
If today James Webb
makes his infrared observations
close and medium with a mirror
of 6.5 meters,
it is among other things
because of such models.
They have been used to predict
what chemical species would be
at what altitudes,
on what type of exoplanets
and what spectral signatures
were to be observed.
We have a kind of virtuous loop
where the models guide the observations
and where the observations
feed the models.
And sometimes it works very well,
even with relatively simple models
like the 1D.
I take an example with the planet
TRAPPIST-1c.
If you follow the channel regularly,
you must know the TRAPPIST-1 system.
It is a red dwarf
surrounded by a procession of 7 planets,
all quite similar in size
to Earth.
A perfect laboratory to study
the habitability conditions
around red dwarfs and in particular
the presence or absence of atmospheres
on these planets.
Several of these planets are
in the habitable zone of the star
and the others are either a little too close
to TRAPPIST-1 or a little too far.
In 2023, a group of researchers
wanted to use James Webb
to find out if TRAPPIST-1c,
the second planet of the system,
had an atmosphere.
They therefore began to create
a series of 1D models
with different hypotheses.
A scenario where TRAPPIST-1
would be a rocky ball
without any atmosphere,
another where it would have
a thick CO2 atmosphere
similar to Venus,
and intermediate scenarios
where the planet would have
a thinner atmosphere
mixed with oxygen.
This paper drawing is,
I assure you,
a little scary at first,
but you will see,
it simply condenses
the predictions of the different models
and compares them
to James Webb's observations.
At the bottom, on the axis of the abscissas,
we have different chemical compositions
for the atmosphere.
On the right, an atmosphere
entirely composed of CO2,
a bit like Venus,
and on the left,
an atmosphere almost without CO2,
between the two edges
of the intermediate compositions.
On the axis of the orders,
we have the atmospheric pressure
on the ground.
It goes from a very thin atmosphere
at the top to a very thick atmosphere
at the bottom.
Finally, between each box,
there is a number that indicates
the difference between the prediction
of each model and the real observation
made by James Webb.
So, the lower the number,
the more the model corresponds
to the data.
This table shows us
that the Venusian hypothesis
occupied by the scenarios
at the bottom right of the table
does not work very well.
The models that best explain
the atmosphere of TRAPPIST-1c
are this one,
an atmosphere mainly composed of CO2
but very thin,
these two, which designate
thicker but poorer atmospheres
in CO2,
or the latter, which shows
a pressure at 1 bar,
similar to what we know on Earth
but almost without CO2.
The main conclusion of the study
is that we can exclude
the hypothesis of a thick atmosphere
of CO2. The other doors
remain relatively open.
A significant result that has been made possible
by these famous models
that interest us today.
You can probably guess it,
we can go even further in complexity
with 2D models.
In the latter, we keep our altitude dimension
but we add either latitude
or longitude.
We can thus separate the phenomena
that occur at the pole and equator
or between a day phase and a night phase.
They can be interesting
to explore planets in synchronous rotation
for example, which always present
the same face to their star.
But the ultimate step is
obviously the 3D models,
often called GCM for
Global Circulation Model.
These models try to represent
the whole atmosphere of an exoplanet
in all its dimensions.
These are models inherited from meteorology
and terrestrial climatology
as well as Martian
and Venusian models that can rely
on a large number of data.
For exoplanets, however,
it is necessary to fill the holes
with unknown parameters.
When we observe an exoplanet,
we are lucky to have a mass,
a diameter, orbital parameters
and perhaps some chemical clues.
The exact composition of the surface,
the albedo, the rotation speed on itself
must be more or less guessed
or be declined in several possibilities
The problem with GCM models
is that they are very greedy
in calculations.
At the beginning of the year,
Météo France launched an offer
for the renewal of its supercalculators
precisely because global models
are atrociously greedy.
So obviously we are not on the same level
of complexity with an exoplanet model.
We don't care if it rains tomorrow
on an unknown continent.
But you still have to cut the atmosphere
into a more or less thin grid,
add the different altitudes
and calculate the impact of different effects.
For a model, it's fine,
but if you want to test
without different hypotheses,
it is better to have well-placed friends
in the computing centers.
Despite these difficulties,
GCM models are gaining ground
and are increasingly used
in specific cases.
Let's take the example of TRAPPIST-1c.
In this study, dating from 2024,
a group of researchers
wanted to make sure
that our 2023 result was good.
That is to say that the Venusian model,
a thick atmosphere of carbon dioxide,
could be definitively ruled out.
Except that this time,
they deployed a very precise GCM model.
A pressure on the ground of 92 bar,
an atmosphere of CO2,
a little sulfuric acid,
a perfect copy of Venus.
However, they integrated
a synchronous rotation
because it is the most likely case
for a planet so close to its star.
The result is the same as in 2023.
The predicted spectrum does not correspond
to the one observed by James Webb.
And the Venusian scenario for TRAPPIST-1c
is once again rejected.
However, you will notice
that the study of 2023
could be studied
a whole grid of hypotheses
thanks to simple 1D models,
while that of 2024
must be content with a single hypothesis,
Ideally, we should be able to test
hundreds of different conditions in GCM
to see what best corresponds
to real observations.
But this is where another big problem arises.
Even if we have all the computing power
necessary,
sometimes two very different models
make the same prediction.
The most famous recent case
is undoubtedly K2-18b.
This exoplanet,
noticeably larger than Earth,
knows its little moment of glory
and this, essentially, thanks to a model.
That of the Hycean planets.
A team has thus modeled
that under certain conditions,
planets could develop a vast ocean of liquid water
under a thick atmosphere of hydrogen.
These oceans could possibly
welcome life.
James Webb has made observations of K2-18b
and his results match
the predictions of the Hycean model
with the potential discovery
of a prime biosignature, the DMS.
Great!
We can unleash the champagne,
we are no longer alone in the universe.
But not so fast.
Because another model has been proposed
which also explains quite well
James Webb's observations.
And he predicts that K2-18b is covered with magma.
Obviously less exciting.
So how do we get
quite similar predictions
with such different hypotheses?
This is where we realize
that we must necessarily speculate a lot
to launch a model.
There is a huge lack of data on these distant planets.
We can reasonably guess a few,
such as the fact that a planet
very close to its star is in synchronous rotation,
but often we have to rely
on what we already know in the solar system.
It is not for nothing that Venus
often comes back on the carpet
when we mention Trappist-1c.
Despite this, the future is very promising.
Ideally, the available computing power
will continue to climb
and launch hundreds
or thousands of hypotheses
in GCM models for a single interesting exoplanet
may be possible one day.
In addition,
each new discovery allows a return of experience
which can be used to improve the models used.
Each exploration mission
of the solar system,
each specter of detected exoplanets
increases the quality of the assumptions made.
The more precise the models become,
the more incredible their predictions are.
Such a giant would have winds
circulating at more than 20,000 km per hour.
Such another would shelter sand rains.
It's kind of a way
to do exploration in the imagination
of worlds that will remain forever
out of our reach.
I secretly hope that one day
we will have such resources
that we will be able to run
some scientifically credible models
in the form of video games,
with 3D worlds that we could explore
based on the best known representation
of real exoplanets.
It does not replace the dream of interstellar travel,
but it would stretch our explorers' souls.
By putting the whole galaxy,
or almost, within a click,
without the associated dangers of space travel.
This brings us to the end of this new format.
Thank you very much
for your feedback on the first video,
I can't wait to read them for this one.
I have already seen some suggestions
of subjects in your comments
and I encourage you to continue
the proposals.
See you next week
for a DNDE and a file
and in two weeks for our Sunday format.
