PDA

View Full Version : Planet hunting with Kepler



clint
2010-Jan-14, 12:07 AM
Article in The Economist (http://www.economist.com/sciencetechnology/displaystory.cfm?story_id=15211385)


Kepler is a telescope that looks simultaneously and continuously at more than 150,000 stars, recording the amount of light coming from them. It is seeking the tiny, periodic diminutions of illumination caused by planetary transits and, on January 4th, the team running it announced at a meeting of the American Astronomical Society, in Washington, DC, that five such patterns had shown up in the first six weeks of the probe’s operation.

This line caught my attention:


Hundreds of flickers that might have been caused by planets with longer orbits have been seen, but have not yet been confirmed as transits.

So five new planets seem to have been confirmed so far, but there are hundreds more potential candidates - yet unconfirmed.
I wonder how many of those (what %) can realistically be expected to be confirmed at some point...

astromark
2010-Jan-14, 10:08 AM
The Kepler is not a whole sky instrument. monitoring 150,000 stars is not all of the visable stars is it ?
The design of the Kepler instrument is to find exo-solar planets., and its working.
What % do I expect to be confirmed ? All of them. Its my assertion that planets are the norm... Kepler is finding the same. Its a random and slow process as for this detection method to be accurate the star must be passed infront of and a wobble in magnatude detected.

clint
2010-Jan-14, 01:04 PM
The Kepler is not a whole sky instrument. monitoring 150,000 stars is not all of the visable stars is it ?
The design of the Kepler instrument is to find exo-solar planets., and its working.

Indeed, great instrument to have!!


What % do I expect to be confirmed ? All of them.

I very much doubt that, not every 'flicker' found so far will turn out to be a planet.
Are you implying there will be no false positives?


Its my assertion that planets are the norm... Kepler is finding the same.

No argument here :)


Its a random and slow process as for this detection method to be accurate the star must be passed infront of and a wobble in magnatude detected.

As far as I understand this detection method, it's more a 'flicker' (slight reduction in luminescence) rather than a wobble.
The five confirmed planets have very close/fast orbits, and have already transited several times during the 2 months Kepler has been observing.

Planets farther out, obviously, will take longer to confirm, since they don't transit quite as often.

I was just amazed how many "unconfirmed positives" (flickers) Kepler has found already,
and thus wondering what approx. ratio of false positives should be expected among those...

Murphy
2010-Jan-14, 06:18 PM
I would expect that almost all the dips in brightness that Kepler is observing really are planets. What else could they be frankly?

Changes in star brightness might be caused by several different phenomena, like flare stars, but we already know a great deal about such stellar perturbations from decades of study, I think planets will be noticeably different from those types of things.

The Kepler scientists are just being conservative about how they confirm the planets, with their 3 orbits rule, but they already know they've found a lot of planets.

astromark
2010-Jan-14, 06:43 PM
Indeed, great instrument to have!!



I very much doubt that, not every 'flicker' found so far will turn out to be a planet.
Are you implying there will be no false positives?

mark ---" yes. clouds of interstelar debris could be detected as any blip in magnatude is detected." ---

No argument here :)



As far as I understand this detection method, it's more a 'flicker' (slight reduction in luminescence) rather than a wobble.
The five confirmed planets have very close/fast orbits, and have already transited several times during the 2 months Kepler has been observing.

mark --- "yes, I should not have used wobble. I was refering to a wobble in brightness into magnatude not a phisicle wobble sorry." ---

Planets farther out, obviously, will take longer to confirm, since they don't transit quite as often.

mark --- " Which leads to a problem, We are going to be waiting a long time for Earth like confermations... say three years. " ---

I was just amazed how many "unconfirmed positives" (flickers) Kepler has found already,
and thus wondering what approx. ratio of false positives should be expected among those...

mark --- " Stars have by the process of forming, have orbital material. Patients is required..." ---

ngc3314
2010-Jan-14, 07:03 PM
I would expect that almost all the dips in brightness that Kepler is observing really are planets. What else could they be frankly?


From The Kepler Followup Observation Program (http://arxiv.org/PS_cache/arxiv/pdf/1001/1001.0352v1.pdf), the false positives they see from spectroscopic followup include grazing (shallow partial) eclipses by secondary stars, genuine triple systems, and stars with blended but unrelated eclipsing binaries, all of which make the simply-derived size of the transiting object smaller than it really is. To date, they list these false positives as outnumbering confirmed planets by about 5:1, although there are still 10 times as many planet candidates with confirmation data in progress as have been reported as confirmed. Careful examination of the Kepler data alone lets them reject some kinds of image blends involving eclipsing binaries, leaving a followup sample that turns out to be about 1/4 genuine planets.

JohnD
2010-Jan-14, 07:16 PM
Mike Brown is a distinguished planet hunter and planet demoter!
His blog here: http://www.mikebrownsplanets.com/2009/04/kants-crowded-universe.html deals with the Kepler story.

John

thoth II
2010-Jan-14, 07:19 PM
From [url=http://arxiv.org/PS_cache/arxiv/pdf/1001/1001.0352v1.pdf] To date, they list these false positives as outnumbering confirmed planets by about 5:1, although there are still 10 times as many planet candidates with confirmation data in progress as have been reported as confirmed. .

That is amazing still, because with all those stars, even 5/1 ratio false positives means an amazing number of planet discoveries. If they keep going at this rate, 3 years from now, we'll have a vast catalog of new discovered planets with a wide range of sizes.

EDG
2010-Jan-14, 08:27 PM
I would expect that almost all the dips in brightness that Kepler is observing really are planets. What else could they be frankly?

Starspots? (I'm guessing there's a way to tell starspots rotating into view from planets transiting the disk, I just don't know what it is. Particularly if you don't know the rotation period of the star to begin with).

George
2010-Jan-15, 09:00 PM
Starspots? (I'm guessing there's a way to tell starspots rotating into view from planets transiting the disk, I just don't know what it is. Particularly if you don't know the rotation period of the star to begin with). I would guess it would require mulitple observations at the time the spot should be re-appearing, assuming the object to be an orbiting body.

EDG
2010-Jan-15, 10:33 PM
Well, the spot is on the star's surface, but a planet transiting across the star would look like a spot from our perspectiv wouldn't it? So multiple observations should reveal the same thing - a dip in brightness.

I'm guessing there's another way to tell them apart - maybe the duration of the dip? But again we usually don't know the star's rotation period (which will determine the time between the spot appearing and disappearing from view). Then again, spots usually aren't permanent, so I guess if they take enough observations it should disappear over time if it's a spot.

clint
2010-Jan-15, 10:37 PM
That is amazing still, because with all those stars, even 5/1 ratio false positives means an amazing number of planet discoveries. If they keep going at this rate, 3 years from now, we'll have a vast catalog of new discovered planets with a wide range of sizes.
It is amazing, indeed, because AFAIK Kepler has been looking for only a couple of months, so far.
Extrapolating this ratio to a couple of years would give us - what - thousands of flickers and hundreds of confirmed planets?

Very cool :cool:

George
2010-Jan-16, 01:03 AM
Well, the spot is on the star's surface, but a planet transiting across the star would look like a spot from our perspectiv wouldn't it? So multiple observations should reveal the same thing - a dip in brightness. The average life of a sunspot is only a few weeks. I would bet stellar spots would all have a relatively short life. They would also change in size as they moved, which might be a give-away in the light profile.


I'm guessing there's another way to tell them apart - maybe the duration of the dip? But again we usually don't know the star's rotation period (which will determine the time between the spot appearing and disappearing from view). I wonder how uncommon it is to determine these rates with today's technology. The broadening in the spectral lines can determine the rotation rate, and these can be distinguished from "pressure broadening".


Then again, spots usually aren't permanent, so I guess if they take enough observations it should disappear over time if it's a spot. Yes, but your idea of seeing the way the light profile dips may still be an alternative method. I don't know.

Hungry4info
2010-Jan-16, 05:54 AM
Well, the spot is on the star's surface, but a planet transiting across the star would look like a spot from our perspectiv wouldn't it?

Geometrically (and assuming that the stellar rotation axis is perpendicular to our line of sight), a star spot will be visible half the time, as the star will make half a rotation between the time the star spot is first visible to when the star spot is no longer visible.

For a planet to transit the star throughout half of the light curve, it's semi-major axis must be equal to the radius of the star.

If the stellar rotation axis is not perpendicular to our line of sight, than a starspot above a certain latitude would be always visible, so that would clearly not be a planet's transit. For starspots below that latitude (but still visible to Earth), they will again be present in half the light curve.

The dimming of a star will now, however, be purely geometrical. Limb darkening will affect the depth of the spot's effect on the light curve. It will still be a slow, gradual dip with no clearly defined bottom, and again affect ~half the light curve.

George
2010-Jan-16, 04:25 PM
Geometrically (and assuming that the stellar rotation axis is perpendicular to our line of sight), a star spot will be visible half the time, as the star will make half a rotation between the time the star spot is first visible to when the star spot is no longer visible. Ah ha. That's simple enough.

EDG
2010-Jan-16, 06:18 PM
So you'd rule out a planet just because of the duration of the light curve? That seems a bit hopeful - after all, a lot of planets out there are orbiting very close to the star's surface, and if you don't know the star's rotation rate beforehand then how will you be able to tell that you're not looking at a planet rather than a spot?

A planet can be close enough to orbit a star in 1-3 days, and I see no reason why a star can't rotate with a period of 1-3 days as well.

I guess the question is whether we have a way of knowing the stellar rotation period beforehand. And also whether Kepler is sensitive enough to see the limb-darkening (also, won't a planet's atmosphere affect the initial drop in the light curve too? There's that effect we see when Venus transits the sun, where the atmosphere blurs the edge of the star as it crosses the limb).

George
2010-Jan-16, 09:59 PM
So you'd rule out a planet just because of the duration of the light curve? That seems a bit hopeful - after all, a lot of planets out there are orbiting very close to the star's surface, and if you don't know the star's rotation rate beforehand then how will you be able to tell that you're not looking at a planet rather than a spot? The key is the time period of light attenuation. If this reduction is 1/2 of the period, then a planet seems very unlikely? If there were a significant margin of error in the measurements then maybe a close orbiting planet could be possible. I just doubt that the error margins are actually this great. Further, a fast orbiting planet would also have to match a fast rotation rate of the star. The odds seem very much against all these cicrumstances coming together.

JESMKS
2010-Jan-16, 10:56 PM
This is a question about Kepler. If a planet similar to earth was orbiting Proxima Centauri, the nearest star, and if the inhabitants of that planet launched a searching satellite identical to Kepler, would they be able to detect earth or any other planets in our solar System?

George
2010-Jan-16, 11:13 PM
This is a question about Kepler. If a planet similar to earth was orbiting Proxima Centauri, the nearest star, and if the inhabitants of that planet launched a searching satellite identical to Kepler, would they be able to detect earth or any other planets in our solar System? No, because, from Proxima Centauri, the Earth does not cross (transit) in front of the disk of our white Sun. Kepler is a transit hunter so the exoplanets must cross in front of their host star for it to detect them. Only a small percentage of exoplanets will align to our advantage, so Kepler observes a huge field of over 100 square degrees (about the size of your hand when your arm is fully extended) to greatly improve its chances.

cjameshuff
2010-Jan-16, 11:16 PM
This is a question about Kepler. If a planet similar to earth was orbiting Proxima Centauri, the nearest star, and if the inhabitants of that planet launched a searching satellite identical to Kepler, would they be able to detect earth or any other planets in our solar System?

Proxima Centauri is far from the orbital plane of the solar system, and no planets ever cross the sun from that point of view, so a Kepler-type experiment wouldn't detect anything in our solar system. They would have a hard time seeing the wobbling of the sun, too, since there will be little radial motion from that perspective. Microlensing could be effective, and with a very large telescope they might directly image some planets.

EDG
2010-Jan-17, 12:58 AM
The key is the time period of light attenuation. If this reduction is 1/2 of the period, then a planet seems very unlikely? If there were a significant margin of error in the measurements then maybe a close orbiting planet could be possible. I just doubt that the error margins are actually this great. Further, a fast orbiting planet would also have to match a fast rotation rate of the star. The odds seem very much against all these cicrumstances coming together.

Hrm... I'm still not picturing what you're describing... are there any light curves out there that you could link to that show the curve you'd get from a starspot and the curve you'd get from a planet? I suspect if I saw the graphs it'd be a lot more obvious to me :)

ngc3314
2010-Jan-17, 03:34 AM
This paper (http://www.aanda.org/index.php?option=article&access=standard&Itemid=129&url=/articles/aa/full/2002/47/aa2693/aa2693.right.html) shows full-phase-coverage light curves of a bunch of spotted rotating stars in the Orion Nebula (I can't get a direct link to the figures by themselves). The spot variation is almost continuous through phase (half for a small spot, all the time for a spotted hemisphere); whereas a planetary transit occupies only a few hours out of a typical hot-Jupiter period (something which can easily be hidden in plots zoomed in to show details of the transit).

Hungry4info
2010-Jan-17, 05:55 AM
While slightly less than professionally done, here's some drawings of light curves. The grey area shows over what phase the starspot/planet transit would affect the light curve.

http://i159.photobucket.com/albums/t137/CrossingStyx/LC.png

Hungry4info
2010-Jan-17, 06:01 AM
Microlensing could be effective

Assuming they're willing to wait a few eons for the sun to move past another star to within the sun's Einstein radius.

EDG
2010-Jan-17, 09:58 AM
While slightly less than professionally done, here's some drawings of light curves. The grey area shows over what phase the starspot/planet transit would affect the light curve.

http://i159.photobucket.com/albums/t137/CrossingStyx/LC.png

Ah, right. Now I understand you, thanks!

But still, given how close some of the planets are to the stars, I think it would still be a bit tricky to tell them apart from something actually on the surface of the star wouldn't it? I mean, shrink the size of the orbit right down so that it's one or two stellar radii from the star and you should see something similar to the starspot curve right?

Hungry4info
2010-Jan-17, 10:31 AM
No problem ^.^


I mean, shrink the size of the orbit right down so that it's one or two stellar radii from the star and you should see something similar to the starspot curve right?

Let's take the scenario of a planet orbiting one stellar radii from the host star's surface.
Orbit radius = 2Rstar
Orbit circumference = 4πRstar
Percent of the phase that the planet is in transit will be ~ Rstar / 4πRstar*
Thus, the duration of the transit will be ~1/4π ≈ 8% of the total light curve for a complete orbit of the planet.

Having planets at just one stellar radius from the star's surface is rare. AFAIK, none have been found. Once planets get that close, tidal forces quickly kill off that unfortunate planet.

* This takes the simplistic assumption that the distance the planet traverses in its orbit as it transits the star is equal to the radius of the star. This is not correct: Because the orbit is curved, the planet will traverse slightly more distance during the transit than simply 1/4π % of the orbit.

Hungry4info
2010-Jan-17, 02:20 PM
For kicks and giggles, here's the light curves of three CoRoT planet hosts. (CoRoT-2, 3 and 4). The transits of the planets are very easy to see (sharp teeth in the light curve). Starspot activity manifests itself as periodic shallow and wide dips in the light curve.

CoRoT-2's activity shows a mix of what appear to be pulsations(?) and starspot activity.
CoRoT-3's activity shows some longer period variability, but you can see some starspot activity around ~2800 days. This is a great example of the variability of starspot-induced activity.
CoRoT-4's starspots are readily visible throughout the duration of the light curve.

http://i159.photobucket.com/albums/t137/CrossingStyx/CoRoT-Exo-2b_Light_curve.jpg

http://i159.photobucket.com/albums/t137/CrossingStyx/CoRoT-3_LC.jpg

http://i159.photobucket.com/albums/t137/CrossingStyx/CoRoT-4b_Light_Curve.jpg

George
2010-Jan-17, 08:00 PM
Let's take the scenario of a planet orbiting one stellar radii from the host star's surface.
Orbit radius = 2Rstar
Orbit circumference = 4πRstar
Percent of the phase that the planet is in transit will be ~ Rstar / 4πRstar*
Thus, the duration of the transit will be ~1/4π ≈ 8% of the total light curve for a complete orbit of the planet. A nit... It will be the star diameter as the numerator, 2Rstar/4πRstar, or ~ 16.5%. Your point is still good one either way.

[Added: I should have looked even closer. The arc length during transit is simply 1/6th the total orbit length (if circular), or 16.7%, and not the above equation, I think.]

EDG
2010-Jan-17, 09:08 PM
OK, much clearer :). Thanks!

Hungry4info
2010-Jan-17, 10:15 PM
Ah yeah the diameter... thanks George. Someone needs more sleep.