SciShow Space is supported by Brilliant.org.

(Intro)

This has been a really fun week for planet news.  First, thanks to artificial intelligence, researchers from NASA and Google announced last week that they found two new sneaky exoplanets.  The first comes from the star system Kepler 90.  It's around 2500 light years away and we used to think it hosted seven planets.  Surprise!  There's actually an 8th planet sneaking around in there, now called Kepler 90i.  The second is from the star Kepler 80, a star that lives only about 1100 light years away.  We had thought there were only five planets around there, but now scientists have found a 6th called Kepler 80g. 

The reason we missed these before is tied to how the Kepler Space Telescope, which we use to study these stars, looks for planets.  It uses the transit method, which measures the stars' light intensity to see if there are any dips in brightness.  If you find a regularly timed light dip, it means something is passing in front of the star, and you've probably got something orbiting in there.  This method is really good at detecting planets that are big or have a really wide orbit, or both, thanks to perspective.  Bigger planets cause bigger dips and if the planet is farther out from its star, it'll obscure more of it compared to a similarly sized planet close in, but these new planets were stealthy and harder to detect.

Kepler 90i is small, only slightly bigger than Earth, and closer to its star than Mercury is to the sun.  Meanwhile, Kepler 80g is probably lined up with and obscured by other planets in its neighborhood.  The telescope could still detect them, but their signals were too small for our older programs to flag or for a human to notice, but now, thanks to advances in artificial intelligence, computers can teach themselves how to better recognize signals and they're getting so good that they can find planets those other programs overlooked. 

The scientists who made the discoveries used machine learning programs to pick apart the stars' datasets.  Machine learning is when a computer uses algorithms called neural networks to detect patterns in sets of data and to make extrapolations based on those patterns.  This is how we learn, too, and like us, a computer can get better at this over time. 

Before analyzing the Kepler 80 and 90 data, the scientists trained their neural networks using datasets that we already knew contained transiting planets.  They wanted to see how well the computer could tell the difference between an actual planet and a false positive, based on things like how its brightness changed over time, and it turns out, it could do this really well, like almost 99% of the time, it's correct.  So then they fed the network more data from the Kepler space telescope, this time, from almost 700 systems where we'd already found planets, just in case there were more undetected ones, and by doing this, they found Kepler 90i and 80g, detecting those tiny signals that we'd missed.  

Of course this is exciting because they're new planets, and every new planet means that our vision of the universe has to get more vast, but it also means we're getting better at processing the absolute mountain of Kepler Space Telescope data that exists.  Kepler has been sending data back to us way faster than we've been able to get through it, so using neural networks to process all that data makes a lot of sense, and as they get faster and more powerful, we'll have even more amazing tools for getting to know the universe, and closer to home, we're still learning things about our best friend Mars.  

This Wednesday, in the journal Nature, scientists published a paper that might help explain where some of Mars' water went.  We've known for years that our neighboring planet doesn't have a lot of water, even though it used to.  Part of the reason Mars is so dry is because it no longer has a strong magnetic field.  Without that, Mars couldn't retain its atmosphere so the atmospheric pressure dropped low enough that water evaporated.  But that also isn't the whole story.

The magnetic field problem can account for a lot of the water loss, but not all of it, so scientists have been on the hunt for the rest of that water, or whatever made it disappear, and now they think they've found the culprit: spongy rocks.  Billions of years ago, Mars had a ton of water and active volcanoes, so the planet's crust is mostly basalt, a volcanic rock that can form when water and lava meet.  It also contains serpentinite, a type of rock that can also form after interactions with water, which all seemed pretty suspicious.  

So the scientists built a computer model of the planet's early geology and geochemistry to investigate, and they found that the basalt specifically might be responsible for the missing water.  Their model shows that when water and lava reacted to form the basalt, a lot of that water got incorporated right into the rock.  Mars' basalt has a lot more iron oxides in it than the kind on Earth, which creates different types of minerals in the rock, and that mix of minerals can hold about 25% more water, so Mars' basalt probably trapped a ton of water, like a sponge.  

That water eventually made its way down into the mantle, the region of a planet just below the crust, probably because the hydrated basalt got buried under a bunch of lava flows.  So that's where they think the water is, deep underground, chilling in Mars' mantle.  Well, not exactly chilling.  It's pretty warm down there.  

Now, it's still possible that there are other reasons for Mars' disappearing water.  Like it could have ended up in ice closer to the surface.  We'll need more observations to be sure, but we are definitely getting closer to an answer, no machine learning required. 

If you like exoplanets but machine learning isn't for you, SciShow sponsor Brilliant has lessons in their astronomy unit that cover some of the other ways we detect and measure exoplanets.  There's a lesson about gravitational wobble, which guides you through problems to show how we know so much about our nearest exoplanet, Proxima Centauri B.  I've always found this area of astronomy fascinating, so I thought I'd give it a try.

So this quiz on gravitational wobble is coming at the end in this Worlds Beyond Earth section of Brilliant.  So it talks about the Goldilocks Zone, which is really cool, and then exoplanets and transits and how we find them and then it jumps into learning about gravitational wobble, so I have to figure out how fast Jupiter orbits the sun in km/sec.  So in this section, it's actually pretty simple geometry.  It's just with huge numbers, so if I take the distance that Jupiter is from the sun and use that as my radius for its total orbit, that's just, you know, 2pi times r.  I actually worked this out with paper and pencil beforehand, because it's really big numbers and I needed a calculator and to like, just map it out and think it through, since it's been a while since I've done geometry, so I know that the answer is 13.03 and that's right, 13.03 km/sec is how fast Jupiter orbits the sun.

But then it goes on to talk about how the sun wobbles because Jupiter is affecting the sun with its gravity, and as (?~6:04) just pointed out before, you can totally view the solution to help you figure stuff out, so I'm gonna work my way through this gravitational wobble lesson so then I can get to the interstellar travel lesson.

So Brilliant.org is really fun because you're learning and kind of like, stretching those muscles that maybe you haven't stretched in a while or maybe you just want more practice doing that.  Right now, the first 200 people to sign up at Brilliant.org/SciShowSpace will get 20% off of their annual subscription and you'll be helping to support SciShow Space also, so thank you so much and thanks to Brilliant for supporting SciShow.

(Endscreen)