Open exowanderer opened 5 years ago
@exowanderer I think you want to change w
(the longitude of pericenter, which rotates the ellipse of the orbit about the orbital plane) instead of Omega
(the argument of ascending node, which rotates the orbit on the plane of the sky, assuming it is viewed ~edge on). Changing Omega
should have no observational effect for a single-planet system, since it just rotates everything on the plane of the sky.
Excellent. I didn't see that parameter!
I tried this out, keeping planet.Omega = 90
; but varying planet.w
between [0,90,180,270].
The eclipse location definitely varies with respect to eccentricity and omega; but the transit center location also moves. I can imagine why this is true; but I have not seen this behaviour before. Note that I did not modify planet.tref
for any of these plots.
Could you confirm that the follow plots make sense when modifying omega (as per the legend).
Because the transit center is changing with planet.w
, is it possible that the observer
is being rotated around the star, instead of the planetary orbit?
[@kevin218 and I had to draw orbits on the white board to figure that out]
@exowanderer @kevin218 The behavior in your plot is correct. The issue is in your interpretation of tref
, which is not the time of transit t0
. They only coincide for a circular orbit. Instead, tref
is the time at which lambda0
is specified. Take a look at the last section of this tutorial. If you want to hold the time of transit fixed, you'll need to solve Kepler's equation to figure out what lambda0
, the initial mean longitude, is at the time of transit.
I could code up a utility to do this for you, if you think that would be useful. Just let me know (or open an issue).
That makes sense. Also, this is the third time that the answer was "well, [Jonathan's] interpretation is at fault". Thank you for your help.
Would it be sufficient to marginalize over lambda0
along with eccentricity? I would hope that the statistics work out fine; and that I can avoid inflating the error bars. I'm sure solving Kepler's equations is more direct; but marginalizing is trivial.
@exowanderer That should work as long as you're not conditioning on a given observed value of the transit time, t0
. You'll need to solve Kepler's equation to figure out what t0
is at every step in the chain if you want to (say) weight the likelihood by some prior on t0
. I'm going to re-open this issue, since I should implement a get_t0()
function (or something like that) anyway.
I assume that you are offering to develop the get_t0
as a means to translate from lambda0
and tref
into transit_time
by solving Kepler's equations; but I wonder if this is not the same as marginalizing over t0
itself.
In the similar behaviour to our wrapping of the day2night
contrast and phase_offset
from before, would it be possible to input the transit time
and eccentricity
, instead of the reference time?
I don't mean to up end your existing infrastructure; but instead to give the user an option of one or the other. I assume that this would take the same amount of time as developing a get_t0
method; but would allow observers to introduce more observation driven parameters (similar to above).
@exowanderer Thanks, that's a better way to think about it. I'll implement the general conversion between t0
and tref, lambda0
so the user can provide either one or the other.
I think that STARRY might not be computing the eclipse location relative to eccentricity of the planetary orbit correctly. I may be attributing eccentricity or Omega incorrectly; please let me know if my code snippet below is correct.
Here is a plot of four different starry phase curves with different combinations of eccentricity and Omega (see legend)
This second figure below is identical to the one above; but zoomed in on the eclipse location, which has not changed; but was expected to.
The vertical dashed line shows the location of phase = 0.5. The shape of the eclipse bottom changes; it begins to look like a grazing eclipse at some versions of eccentricity and omega.
My understanding is that if the eccentricity is > 0.0 and the Omega ("The longitude of ascending node in degrees") is > 0.0, then the eclipse location should change (+/-) by up to several tenths in phase (0.3 - 0.7)
As described here:
This script returns the following phase locations for the values of
eccentricity
andomega
in the script below and figure aboveCode Snippet