Welcome to the segment of the class.

Today we're going to be talking via Skype with Debra Fischer.

Debra's a professor at Yale University and she's going to talk to us about some of

her work and some of the things we've been learning about the extra solar planets.

So today's lecture we talked about radio velocities and Debra's

a pioneer in using this technique to study and discover

extrasolar planets.

So begin by just talking about your first extrasolar planet.

[CROSSTALK] >> Okay.

Good morning. Thanks so much for having me.

So I joined the field in 1997 and

the first few exoplanets had just been discovered.

And it was a incredibly exciting time when every

planet you know, is a press release and limos

came and swept people away to do documentaries and so forth.

But it certainly changed the way that we understood

planet formation because we were finding gas giant planets

very close into their host stars and of course

the solar nebula model did not predict that at all.

yeah. >> So some of

the history of this was that people actually

didn't think they'd be able to discover planets because

we thought that our solar system was typical and

in our solar system, Jupiter is pretty far out.

>> That's right.

>> And the velocity that Jupiter induces on the sun were too small.

>> That's right.

>> To be seen by the first generation of instruments.

>> That's correct.

Yes, exactly. And but if you go back Otto Strub

had a paper in 1953 where he said, you know, even then with state of the art

precision, doppler precision and I would say even then,

it was about, at best, 200 meters per second.

He speculated that if there were planets that

were more massive than Jupiter, and perhaps orbited closer

to their stars, you know, that they

would have the sensitivity to just barely detect them.

But that seemed like

a crazy idea at the time, and you know, it took almost 5 decades later,

before people were able to find the first gas giant planets and close in orbits.

And I think that you know, the expectation was

that we were in this for the long haul.

So I was, at the time that Geoff Marcy

and Paul Butler were doing their pioneering work at San

Francisco State University, I was there as working on my master's in physics and

I was studying my thesis project was

actually on the binary frequency of M dwarfs.

But I remember those early days and the

expectation that the gas giant planets which would be

just barely detectable with sort of 5 to 10

meter per second precision would be in distant orbits.

And so they were in this for the long haul and it was reflected that

bias or that attitude was reflected in the

sampling cadence that they had at the observatory.

They would take you know, observations a few times

a year and if the orbital period was going

to be 10 years, then that would be enough

to trace out the phase in the orbit.

So, it was definitely a surprise when Didier Queloz and Michael Mayor found

this planet around 51Peg a sun-like star in a 4 day orbit, roughly 4 day orbit.

So that was immediately confirmed by Jeff and Paul at the Lich Observatory

and then, it was like the blinders sort of went off, you know?

Weâre, as scientists as human

beings, we all work with these Bayesian filters, you know, where

do we expect what happen and that's what we look for.

And so then, armed with this new knowledge that

gas giant planets might, residing close in orbits it definitely, you

know, gave great impetus to trying to get the code in

place and analyze all the data it was, those were tough days.

You know, I remember Paul Butler

giving a talk saying, I have this Doppler analysis code, and the computers

of the day were very much slower and they had trouble getting money, right?

I saw a document where Jeff Morrissey was

awarded a grant for $500 to do, research on exoplanets.

So there wasn't a lot of money, and the computer power was

tough and it would take 24 hours to analyze a single spectrum.

So if you go to the observatory and you collect 50

spectra, right, then you've got a long backlog of data to analyze.

But now that they knew, now that they realized that these

planets could reside in closed in orbits, they went to work quickly,

and they were awarded a wonderful gift from Sun Micro-systems of 3

computers that allowed them to really start crunching through their data and

they quickly found not only confirmed 51Peg, but quickly found

planets around 70 Vergenas and 55 Cancri

in very short order. So it changed the way that we looked.

>> Now 55 Cancri turned out to be a really interesting system.

right, so it's hasn't got just one planet but

>> That's right >> Up to 5 planets

>> We're now

up to 5 planets exactly.

And just thinking of the history of the whole multi-planet systems the

first of course was Upsand 3 planets around ups and and Andromeda.

And Howard and Jeff had discovered the first one.

And again, in a 4 day orbit, this one needs hot, you know, roaster Jupiters and

they saw from the data, a trend that suggested that there was one more planet

in, perhaps, a 2 or 3 year orbit.

And so that's when I joined the team and I was

racing up and down the mountain to collect data on Epsilon Andromeda.

We thought that the outer planet was just closing its orbit.

And we were trying to get it, it was

beginning to set and we were losing it from behind the sun.

But we, we got enough data's to seal, you know

the outer planet orbit and then began modeling that and

failed horribly.

But when I looked at the residuals of

the 2 planet models, saw a third planet in there.

So, even then, right?

3 planets, if we just think about them, the orbits were 4 days,

250 days, and 2 and a half years, and I really struggled with this.

Was this, is it physically plausible because these were all gas giant planets.

We saw nothing like this in the solar system.

So Greg Laughlin

began modeling the system, ran it overnight and the

next morning told me, it's still running, it's still stable.

So, you know, I think about how we struggled about that one and

that with 55 Cancri, as we began finding more planets in that system.

So finally it was, 2008 was our last paper on 55 Cancri, where we announced

5 planets around this system and have run all the dynamic stability analysis.

And that was great.

And I think the surprise with 55 Cancri was that we got one of them wrong, that

the inner planet that we found was, we thought, was in something like a 2.81

day orbit and Dan Fabrycky and Rebekah Dawson then analyzed our data.

We published our velocities, of course, and

said I think you've got this one wrong.

I think it's actually a

0.7 day orbit for the inner planet and they were right.

It was absolutely fantastic.

It was verified by transit observation you know?

And what had happened in that case was it again, we had our Bayesian filters on

and all of our Doppler signals basically had

one day aliases in them because we would, that was the window function in our data.

And so we had gotten in the habit of

>> So that window function arises because we have a day

on the earth and how you observe things is structured by the earth's day.

>> That's right.

>> And when you talk about a window function, we observe this pattern with

of the radial velocities with the filter of when we can look at it.

>> Absolutely.

>> We can't look at it constantly we look at it every day.

Absolutely.

And that's. >> Or every

night, to be more precise. >> Exactly.

And super important as we go forward.

So there's 2 aliases in the data.

I mean, there's often the one day alias is just guaranteed.

There's also seasonal aliases because you can observe stars for

maybe you know, 6 months, 7 months, something like that and

then there are also aliases because, that are imposed by

the time allocation committees, that say, you're project only needs bright

time, so we'll give you bright time, you know, each month.

And the HAARPs, team suffers from this.

[CROSSTALK]

>> When astronomers observe the sky, we're affected very

much by whether or not the moon is up.

>> Yes.

>> So when the moon is up we call that bright time, because the sky is

brighter and when the moon's not up, the

sky's darker, that time's usually given to extra-galactic astronomers.

>> That's right.

>> Because they're affected by the background.

And stellar astronomers and people studying planets are effectively stellar

astronomers because they're observing the stars, not the planets themselves.

>> That's right.

>> They get bright time, so they get. >> That's right.

>> One half of the month and their

extragalactic colleagues get the other half of the month.

>> That's right.

So in any case, knowing that this one day alias was there for all of

these planets, we got tired of looking at that pick, and essentially put in a higher

pass filter on the periodic realm analysis, so that it would eliminate you

know, we didn't have, we could zoom in then on other interesting periodicities.

And so we weren't really looking for planets shorter than one day orbits.

It was already in the [UNKNOWN] that there were so many planets that

had orbits that were shorter than

Mercury's orbit, you know from my perspective.

But great, in terms of detectability

because those planets induce much larger velocity reflex signals.

So yeah, so 55 Cancri turned out to be a very rich system that we really

struggled with to try and unravel all of the aliases everything that was built in.

So it was great the theorists got to teach us a

lot about which included Jack Wisdom and and Rebecca Dawston and

Danford Brickyteach us a lot about statistial analysis and the possibility of

aliases in our data and that perhaps we haven't been doing things right.

So that was a real fantastic opportunity and then the third multi planet

system which really threw us for a loop in the early days was [UNKNOWN] at 80, 876

and that one had a

planet in a 30 day orbit and a 60 day orbit

and that 2-1 resonance of the planets made it extremely, and

plus the window functions of sampling that we had in our

data made it very difficult to pull out the 2 planets.

So it wasn't until we got to quite a bit of data for us

to unravel that and then, of course

ultimately, a third planet was found there.

From my perspective,

doing Doppler observations, where we took such exquisite care worrying about whether

or not the systems were dynamically stable, it's been such a surprise

to see all of the transiting planets detected by Kepler that are

multi-planet systems, where there would be 5 planets inside of Mercury's orbit.

And yeah, of course, the dynamic of stability analysis

is done, but these systems are just absolutely packed with planets.

Makes it tough, yeah.

>> Yeah, I mean it seems that nature sometimes chooses to

pack in almost, almost as many panel planets as possible right there.

>> Yeah, exactly.

>> And they're just on the margin of dynamical stability.

A lot of them are trapped in resonances >> That's right.

That's right. And, and that points to the scenario.

I mean, even with 55 ca, or even with Ups Anromedae, the first multi-planet system,

I think it pointed to this picture of many protoplanets forming

in the discs and it, a sort of competition so that you were weeding things out.

Hey, this system is way overfull and weeding things out

as, like a, sort of gravitational game of musical chairs

so that you end up with you know a full system

where every gravitationally stable niche seems to be filled with something.

If that's a characteristic, it certainly is

a characteristic of our own solar system.

And if it's a pervasive characteristic, it means that the

systems are even richer than we imagine right now today.

>> And so this is pointing to this picture where you just fill

up the system with as many planets as you can fit.

The ones that don't fit in get ejected.

Things move around.

>> Yes.

>> The process may have happened in our own

solar system as Saturn, Jupiter and Uranus rearranged themselves.

It may be what was what was responsible

for the late bombardment something we talked about

earlier in the course, when we talked about

the moon, and that, one of the implications

of this, and this is something we will turn

to later on when we talk about gravitational lensing.

If there should be orphan planets, things that have been

ejected from their system that are now traveling through empty space.

>> Yes. >> And these planets, without stars.

>> Yes.

>> Is one of the predictions of this model.

>> Absolutely.

>> One of the other big hints, I think, about planet formation and

the whole process came from your work with Jeff Valenti, showing this trend

that, you know, between planet frequency and properties of the star.

>> That's right.

>> So what did you find?

And when you found it, it was, well, it was 10 years ago.

We've learned so much.

In the past 10 years that picture's continued to evolve.

>> Yeah, absolutely. So, there was a hint after the first

few planets were discovered.

These of course, were all gas giant

planets, all discovered with the Doppler technique.

Most of them in, I mean I think it's

virtually all of them in short orbits right, close-in

orbits and Guillermo Gonzalez immediately said wow, isn't it

odd that all of these are fairly metal-rich stars?

And so, people right after in 1997-1998 were speculating that there

was a planet-metallicity correlation. And it seemed

right and so Jeff Valenti and I sat out to do a very complete analysis.

We had 2000 stars that were being observed at between Lick

Observatory, the Anglo Australian Telescope and the Keck Observatory.

And and we simply analyzed them all with a uniform technique,

which was very nice.

And then in each metallicity bin, we said, what fraction of those stars have planets?

And the great thing about that technique is that

even if you're observing more stars that are metal-rich because

you bought into the early suggestion and so you

bias your sample, it doesn't actually buy us the answer.

It still, you're still saying, what fraction

in each metallicity bin have planets, and what happens by adding let's say

more metal rich stars, is that you decrease the Poison air bars, say you

get a more precise answer, but you don't necessarily skew the answer as long

as you have a decent number of stars in each of the metallicity bins.

So that's what we did.

And we found a very strong correlation. It went exponentially

with this sort of a power of two. And the probability of a star.

Having a planet as a function of metallicity, so of course,

this applied to gas giant planets and that result has held upIt was

the same result at virtually the same time that Nuno Santos and the

Geneva team was carrying on an analysis using a different technique and

got virtually the same result.

And over time now with all of the Kepler planets, people have been able to

verify that, they see yes the, the short

period gas giant planets preferentially orbit metal-rich stars.

Now what's interesting is that at the time, we were

not able to really detect the, the Neptunes or the super-Earths.

These are planets that have been, that are smaller in

mass so the reflex velocity is much lower.

But there are planets that have been detected

in great abundance with the Kepler transit surveys.

We've detected some with radial velocity techniques and I would say that the

Geneva team Has succeeded at this better than we have in the US.

Their, their, the HARPS machine that they've built gets higher precision.

So already, there was a hint

that maybe this metallicity correlation didn't hold up for the lower-mass

planets and now with the Kepler results, we very clearly see

That the correlation, the slope in the correlation is definitely dropping down

and may even be flat as you go to super earths orbiting stars.

So they way I look at that is a slightly different way.

My guess is that there's a time scale that

I mean the gas and, sorry.

>> Right, so just yeah, just wanted to back up

and just remind people the students the picture in which the.

>> Yes.

>> Solar system and all these extra solar systems formed

out by first forming a disk of dust and gas.

The gas as you're about to say, is evaporates about five to 10 million years.

And then the dust and planetessimals persist longer.

Yes. >> So.

And the way that a massive Jupiter-like planet has to form

is, first assemble the rocky core, and then accrete the gas.

So this is what drives the short time scale, right?

>> Exactly.

It's kind of surprising.

I think if you know, if before you knew anything, if you would have

had to guess, you would have guessed that the bigger planets take longer to form.

Somehow, right?

You, there's more accretion. But in

fact I think what's happening is

that having high metallicity in the protoplanetary

disc actually triggers very rapid planetesimal formation

and so the accretion process happens quickly.

The gas giant planets form quickly and, you know, and the gas is

still around for them to grab their atmosphere and become gas giant planets.

If you don't have that high metallicity,

it's probably, is there some kind of a threshold.

Then the planet formation is slowed down a bit.

And by the time the planetessimals reach this

critical mass where they could have grabbed on, gravitationally,

to the gas. The gas is gone, and so

the disk is left with low with lower-mass planets.

So I think that this is really, it's a time scale argument,

that the metallicityy pushes planet formation more quickly so that the

gas giant planets can form and you know we just

don't see that with the systems that seem to not have

gas giants, at least in close orbits, but only have super Earths.

So that, correlation goes away.

So it's kind of reassuring.

It doesn't mean that you have to have high metallicity to form planets.

If you did it would mean that only a

tiny fraction of stars would be planet host.

It just means that the time scale is faster and the gas giant planets can form

around metal-rich stars but not so much around

the solar-metallicity/sub-solar metallicity stars which are more common.

So much of the work on searching for

planets is often done with big telescopes one

of the things you've been looking at is

using smaller telescopes, dedicated telescopes for this purpose.

>> Right.

>> Do you want to tell us some of the things

you've been doing and planning to do in this area.

>> Yeah.

Yeah, its it's been a game changer in my mind its unbelievable.

Another sort of blinds, blinders that's gone off and that we views typically

the three meter telescope at Lick, which no longer is a big telescope but when we

started using it was a definitely a

pretty large telescope, Keck had not come online yet.

So Keck came online in the, sort of early to mid 1990s and then with

the success of this project, we've gotten quite a bit time on the Keck telescope.

But quite a bit of times means, you know, a few nights a month.

That's incredible.

Alpha Centauri a or b. But as part of this project

>> Alpha Centauri of course, is the nearest death start to the earth.

>> That's correct.

>> That's besides the sun.

>> But then the catch is that it's also a binary star system.

With the separation of 25 au's, one au being one earth sun distance, and so

one would wonder if you have a star that's basically two stars, one where the sun

is and out, you know, out sort of where

Saturn is, can you actually form planets in between there?

So, theorists would say no, they did say no, but they

also said don't listen to us, you should go and look.

And so that's what we get.

And in, but in the process, so we built a

spectrometer to go on this telescope is now getting excellent precision.

When we look at one of

our stable stars at TOF SELDI over a couple

of months, we're getting a precision and our mass scatter.

That's less than one meter per second.

So that's really, I think, quite

phenomenal and we're very excited about that,

but what we see when we continue to look at Tau Ceti, so now

I'm telling you something that's unpublished

yet, we're working on the paper right

now, is that, you look for two months and the radial velocities are flat.

And then suddenly there's a little wiggle that starts up and has a 32

day period. Small amplitude and then it

goes away and it comes back in phase the next year and

bizarrely this is something that I saw at Lick observatory.

But now we can truly track this out.

And what I suspect is happening is that there are spots

on the surface of the star and as the star spots come and go you see the amplitude.

You know, you see the signals come and go.

So this has been, the Geneva team has said that they see nothing

around the star but I went back and looked at one of their papers.

Where they were talking about a star HD20794

and publishing three planets there exquisite, beautiful data set,

and as a standard, they showed their

tertiary data, and when I look, I see exactly

the little tiny, its very low in amplitude but

I see absolutely the little signal that we're seeing.

So the point is that with extremely

high Cain's data, we're observing almost every night.

We can see features that we would never see at Keck.

And as an example in this two

month data set where we had very low RMS, I

said, oh let me plot, over-plot the Keck data

and see what it looks like is You know, to

see how much better our data looks than the Keck data.

By the last, there was one data point in that two months from Keck.

And so it, you know, if we're going to detect these very

subtle signals, whatever they are, whether they're

stellar signals coming from the star itself,

spots and flows on the star, or whether they're very low

mass planets, we're going to desperately need very high cadence data.

So it's something we wouldn't have said before.

Again, the point is that we don't know the answer

in the back of the book when we start looking.

If we knew what we were looking for, we could tune

our observing cadence and we would start with something like a logarithmic

observing cage, where you start out with high frequency sampling to try and

pick out the things in short periods and, you know, relax your sampling.

Instead now we we do know something.

We know that nature has dealt us a pretty tough, you know, hand.

That the samples are packed with planets.

That detangling the multi-planet systems and the

aliases are tough, and that the observing cadence

that we need is much higher than we

ever would have guessed when we started this game.

>> So when do you think you'll be able to kind

of see an earth-like planet around Alpha Centauri at an earth-like orbit?

How are away are we from that?

>> We are have a big problems with Elvis and

Tori because as I said it's a binary star orbit.

Zavia Demoosk

on the heart, harp's team has announced one

planet that is truly an earth mass planet.

M sinai as one essentially.

But is in the 3.24 day orbit, so it's a roaster.

now, if there's one planet, there, we're finding

that from Kepler, there should be more.

If we run the dynamical stability analysis, we find

that out to and beyond the habitable zone

around both a and b, if planets formed, they would be stable.

Which was one surprise to me.

But they would be stable around either star.

but our problem now is that the projected orbit

of Alpha Centauri A and B put them pretty close.

They're now separated, when we started this

in 2008, the separation was 7 arcseconds.

It's now 5 arcseconds.

That means that the seeing halo from the stars

are spilling into each other.

So when we observe one star, we're also

observing small fraction of contamination from the other star.

So we're dealing with that right now, trying to

figure out if we can, not, we're start revising

our models so that we can account for a

scaled flux from the other star knowing the Doppler offset.

So that's still work to be done but

my guess is that whenever you can terminate the

data in this way, even if you come up with the better model.

It's going to set the flaw in our precision to the level that's not great.

By 2015, Alpha Centauri A and B will

be moving apart again in the projected separation.

They're actually not at periastron right now closest approach.

It's just the way that the orbit is, you know, oriented from our view.

So they'll start moving away, and then they'll

be very well separated, you know, by 2018, 2020.

So that gives us time to build the next

generation spectrograph which is what we're trying to do.

We've learned a lot in the last 2 decades

about what effects the floor of the Doppler precision.

We know that there is an air budget when

we observe and that air budget has many terms,

which include the temperature, the

variations in the spectrograph, the pressure

variations in the spectrograph, The star, signals from the star itself.

How we model the point spread function,

that is the smearing function of our instrument.

So my assessment is that right now, we're in a position

that each of the terms in the air budget have similar magnitude.

So that is a tough situation because if you

tap down on one, you don't necessarily see an improvement.

So it's going to require a leap of faith.

We're going to push down on all of the error terms in our instrument.

And we're also going to build an instrument That's

capable not just to detecting Doppler shifts, which

are actually the easier of the two problems,

but also stellar signals, very subtle signals that affect

the low in profile instead of move in the line back and forth.

And so then we, we expect to be able to

find sulfur, the stellar signals, it will be tough, but it

will leave us with fairly clean residual velocities and the

aim is to get down to 10 centimeter per second precision.

It's ambitious.

I feel really confident that, you know, right now

we're almost at a half a meter per second.

Even though with this inexpensive spectrograph that we built on a one meter

telescope, I'm sure we can get down to 20 centimeters per second.

And you know, I'm hopeful that we'll get down to 10, or a little bit better.

And so it certainly is the direction the European team is going.

And I hope

that we can keep up here in the U.S. and do the same thing.

I'm looking forward to that with great excitement.

So.

>> Yes, thank you.

>> So, thank you so much for talking to us.

This was really wonderful. >Sure.

Yeah, my pleasure. Thanks.

Interview with Debra Fischer

假如其他星球也有生命

与讲师见面