The EXOplanet traNsits and rAdIal veLocity fittER (EXO-NAILER), is an easy-to-use code that allows you to efficiently fit exoplanet transit lightcurves, radial velocities (RVs) or both.
Author: Néstor Espinoza ([email protected])
If you make use of this code, please cite Espinoza et al., 2016, ApJ, 830, 43 (http://arxiv.org/abs/1601.07608)
This code makes use of seven important libraries:
- Numpy.
- Scipy.
- The Bad-Ass Transit Model cAlculatioN (batman) for transit modelling (http://astro.uchicago.edu/~kreidberg/batman/).
- The RadVel package (http://radvel.readthedocs.io/en/latest/).
- emcee for MCMC sampling (http://dan.iel.fm/emcee/current/).
- Astropy for time conversions (http://www.astropy.org).
- The GNU Scientific Library (https://www.gnu.org/software/gsl/)
All of them are open source and can be easily installed in any machine. Be
sure to install them before running the installer (see below), otherwise, it
will complain. This code also makes use of the flicker-noise module
(https://github.com/nespinoza/flicker-noise), for modelling 1/f noise. A copy of the
source code of this module is included in this repository
and will be installed automatically.
To install the code, simply run the install.py code by doing:
python install.py
After this is done, the code will be ready to use!
To use the code is very simple, and to help you understand how to use it, we have
added a synthetic dataset for the target my_target, along with this package. my_target,
of course, is a generic target name. You can put any target name you want as long as it
includes no spaces. With this decided:
1. Put your photometry under the `transit_data` folder. The file containing the photometry
has to be named `target_lc.dat`, where `target` is the name of your target. In our case,
`my_target`. Similarly, put the RVs (if you have any) under the `rv_data` folder. The file
containing the RVs has to be named `target_rvs.dat`, where `target`, again, is the name of
your target. In our case, `my_target`. These are expected to have four columns: times, data,
error and name of the instrument (which is a string); however, only the two first are mandatory:
the code will recognize that you don't have errors on your variables and if no
instrument names are given, it will assume all come from the same instrument.
The flux is expected to be normalized to 1. The RVs are expected to be in km/s.
2. Create a prior file under the `priors_data` folder. The file containing the photometry
has to be named `target_priors.dat`, where `target` is (you guessed!) the name of your target.
The code expects this file to have three columns: the parameter name, the prior
type and the hyperparameters of the prior separated by commas (see below).
If you want a parameter to be fixed, put `FIXED` on the Prior Type column
and define the value you want to keep it fixed in the hyperparameters column.
Also, if a prior other than `FIXED` is defined, a fourth column can be entered for
each parameter where you can specify the starting point of the parameter (say, one
obtained by a previous least-square fit, or a value you know to be close to the
true parameter, etc.).
As can be seen from the above, the code can handle data taken with different instruments. For RVs, this means that a different center-of-mass velocity can be fitted for each instrument in order to account for offsets between them, and if jitter is included, a different jitter term can also fitted for each instrument. For transits, this means a different photometric jitter can be fitted to each instrument, as well as different limb-darkening coefficients and different transit depths.
As previously stated, there is a synthetic dataset along with this code which is useful to understand
how to get your fit started. The lightcurves for this dataset are under the transit_data folder and is
labeled my_target_lc.dat, while the RVs are under the rv_data folder and is named my_target_rvs.dat.
Next, you can modify the options of your fit in the options_file.dat file.
The GENERAL OPTIONS are:
TARGET: The name of your target.
MODE: This defines which kind of fit you want to perform. `full` means a full
transit and radial-velocity fit. `transit` means you only will fit the
transit lightcurves and `rvs` means you will only fit the radial-velocities.
NWALKERS: This is the number of walkers on the MCMC runs (for more information on this
parameter, check out the `emcee` documentation).
NJUMPS: This is the number of jumps on the MCMC (for more information on this
parameter, check out the `emcee` documentation).
NBURNIN: This is the number of burn-in runs of the MCMC (for more information on this
parameter, check out the `emcee` documentation).
PLOT: If set to `NO`, no plots will me shown at the end. If set to `YES`, a plot at the
end of the `exonailer` run will be shown similar to the one shown above.
The PHOTOMETRY OPTIONS have to be defined for each instrument. For each one, you must define:
INSTRUMENT: The name of the instrument. These have to match the instruments in the transit
lightcurves.
PHOT_NOISE_MODEL: This parameter defines the noise model used for the photometry. If set
to 'white', it assumes the underlying noise is white-noise. If set to
'flicker', it assumes it is a white + 1/f.
PHOT_DETREND: This performs a small detrend on the photometry. If set to 'mfilter'
it will median filter and then smooth this filter with a gaussian filter.
It works pretty well for Kepler data. If you don't want to do any kind
of detrending, set this to `NO`.
WINDOW: This defines the window of the 'mfilter'. Usually way longer than your
transit event, and is defined in number of datapoints.
PHOT_GET_OUTLIERS: This automatically sigma-clips any outliers in your data if set to `YES`.
It relies on having decent priors on the ephemeris (t0 and P). If you don't want
to remove them, set this to `NO`.
NOMIT: It is a sequence of numbers, separated by commas, that lets you ommit transit in
the fitting procedure (e.g., transits with spots). Just put the number of the transits
(counted from the first event in time, with this event counted as 0) that you want
to ommit. If you don't want to ommit any transit, don't put this option.
RESAMPLING: Set this to `YES` if you want to use the selective resampling scheme of Kipping (2010, MNRAS,
408, 1758), e.g., to apply it to 30-minute cadence Kepler lightcurves (or for TESS lightcurves).
TEXP: Exposure time for the case in which RESAMPLING is YES. For Kepler/K2, this is 0.020434 days.
PHASE_MAX_RESAMPLING: This define the maximum phase at which the data will be resampled if `RESAMPLING` is set to
`YES`.
NRESAMPLING: This defines the number of instantaneous lightcurve points used to resample the lightcurve
if `RESAMPLING` is set to `YES`.
LD_LAW: Limb-darkening law to use. For all the laws but the logarithmic the
sampling is done using the transformations defined in Kipping (2013).
The logarithmic law is sampled according to Espinoza & Jordán (2015b).
TRANSIT_TIME_DEF: Defines the input and output time scales (the times are assumed to be in the
JD format, i.e., JD, BJD, MDJ, etc.) of the transit times. If input transit times
are, for example, in utc and you want results in tdb, this has to be 'utc->tdb'.
STELLARDENSITY: If you have a stellar density measurement for your object of interest, this can help
constrain a/R*, the period, or both (via Kepler's third law). If you have an estimate,
you can input it by defining the mean and error `mean,error`. This assumes the inputted
density is in kg/m^3.
The RADIAL-VELOCITY OPTIONS have to be defined for each instrument as well. For each one, you must define:
INSTRUMENT: The name of the instrument. These have to match the instruments in the radial-velocity
data.
RV_TIME_DEF: Defines the input and output time scales (the times are assumed to be in the
JD format, i.e., JD, BJD, MDJ, etc.) of the radial-velocity times. If input RV times
are, for example, in utc and you want results in tdb, this has to be 'utc->tdb'.
Once you are done with this, just run the code by doing:
python exonailer.py
The priors currently supported by the code are:
Normal: Expects that the third column in the prior file has the form
mu,sigma
where mu is the mean value and sigma the standard-deviation.
Uniform: Expects that the third column in the prior file has the form
a,b
where a is the minimum value and b is the maximum value.
Jeffreys: Expects that the third column in the prior file has the form
low,up
where low is the lower limit of the variable and up is the upper
limit of the variable.
Beta: Expects that the third column in the prior file has the form
alpha,beta
where alpha and beta are the parameters that define the beta distribution.
FIXED: This assumes you are giving the fixed value of the variable in
the third column.
The mandatory variables that must have some of the above defined priors in the case of a transit fit are::
P: The period of the orbit of the exoplanet. Same units as the time.
t0: The time of transit center. Same units as the time.
a: Semi-major axis in stellar units.
p: Planet-to-star radius ratio. If you want to define a different one for each
instrument, add a lower-dash and put the name of the instrument (e.g., `p_Telescope`).
q1: Limb-darkening parameter. This corresponds to the transformation of the uninformative
limb-darkening threatment scheme defined in Kipping et al. (2013, MNRAS, 435, 2152).
The exact transformation is defined by the limb-darkening law used. If you want to define
a different one for each instrument, add a lower-dash and put the name of the instrument
(e.g., `q1_Telescope`).
q2: Second limb-darkening parameter. Same as for `q1`.
inc: Inclination of the orbit in degrees.
sigma_w: Standard-deviation of the underlying white noise process giving rise to
the observed noise (in ppm). If you want to define a different one for each
instrument, add a lower-dash and put the name of the instrument (e.g., `sigma_w_Telescope`).
ecc: Eccentricity of the orbit.
omega: Argument of periapsis (in degrees)
Of course, e.g., for a circular fit, you might want to fix ecc (to 0) and omega (e.g., to 90). If you
define the PHOT_NOISE_MODEL as flicker, you must add an extra parameter, sigma_r (see Carter & Winn, 2009).
The variables which have to be defined in case of a rvs fit, in addition to the eccentricity, period,
time of transit-center and omega, are:
mu: Center-of-mass velocity of the RVs. If you want to define a different one for each
instrument, add a lower-dash and put the name of the instrument (e.g., `mu_Spectrograph`).
K: Radial-velocity semi-amplitude.
sigma_w_rv: Jitter term for radial-velocities (see below). If you want to define a different one for each
instrument, add a lower-dash and put the name of the instrument (e.g., `sigma_w_rv_Spectrograph`).
If you do not want to include jitter, set this to `FIXED` in the prior file, and set it to zero.
In the case of full exonailer fits, all of these parameters have to be defined.
The outputs of exonailer will be under the results folder. In this folder, you will find a folder for
each of your fits and, inside, three files:
posterior_parameters.dat: This file saves the posterior parameters for each variable in
the fit. The first column lists the variable name, the second
the median of the posterior of that parameter (50th percentile),
the third the 84th percentile of the posterior ("upper 1-sigma
error") and the fourth the 16th percentile of the posterior
("lower 1-sigma error").
posteriors.pkl: This file has the actual posterior distributions for each parameter.
priors.dat: This file saves which prior you used for the given dataset (useful
in case you are trying different priors to see how your results
change).
In addition, the data, model and residuals of the transit, radial-velocities or both will be exported as .dat files to this folder, so you can easily plot them yourself.
+ Create a tutorial explaining usage of GPs for transits (implemented, but not yet documented)
+ Add MULTI-NEST support.
+ Transit and RVs for multi-planet systems.
+ Noise models (e.g. GPs) for RVs.