Giters

rosemberg-al / pyaneti

A multi-planet Radial Velocity and Transit fit software

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

pyaneti

Written by Barragán O., Gandolfi D. & Antoniciello G.

email: oscaribv@gmail.com
Updated July 24, 2018

ascl:1707.003

Introduction

  • Pianeti is the Italian word for planets.
  • Multi-planet fitting of radial velocity and transit data!
  • It uses Markov chain Monte Carlo (MCMC) methods with a Bayesian approach.
  • Ensemble sampler with affine invariance algorithm (Godman & Weare, 2010).
  • Python does the nice things: plots, call functions, printing, in/output files.
  • Fortran does the hard work: MCMC evolution, likelihood calculation, ensemble sampler evolution.
  • Open-source code (GPL v 3.0).
  • Free and fast code with the robustness of Fortran and the versatility of Python.

Power of pyaneti

  • Multiple independent Markov chains to sample the parameter space.
  • Easy-to-use: it runs by providing only one input_fit.py file.
  • Parallel computing with OpenMP.
  • Automatic creation of posteriors, correlations, and ready-to-publish plots.
  • Circular and eccentric orbits.
  • Multi-planet fitting.
  • Inclusion of RV and photometry jitter.
  • Systemic velocities for multiple instruments.
  • Stellar limb darkening (Mandel & Agol, 2002).
  • Correct treatment of short and long cadence data (Kipping, 2010).
  • Single joint RV + transit fitting.

Dependencies

  • gfortran (or your favorite fortran compiler)
  • OpenMP
  • numpy
  • matplotlib
  • seaborn (optional)

We tested this code using Linux, including Fedora and Ubuntu. This code has also be being built on MAC computers.

Start to use it now!

Just clone or download pyaneti.

git clone https://github.com/oscaribv/pyaneti

The advantage about cloning the repository is the possibility to follow the changes to this package easily with git pull (learn more about git at https://git-scm.com/).

The next step is to enter the pyaneti folder and see what we can find inside it

cd pyaneti
ls
  LICENSE  src inpy  makefile  pyaneti.py  README.md

Now compile the code typing make

make

If you have all the dependencies installed, the make procces should end without error.

Now you are ready to run the code for the first time! Just type

./pyaneti.py test

or

python pyaneti.py test

The program will start. You will see something like:

==============================
------------------------------
    INITIAL CONFIGURATION
------------------------------
Star           = test
No. planets    = 1
------------------------------
iter max       = 100000000
thin factor    = 1
nconv          = 500
nwalkers       = 100
------------------------------
fit RV         = [True]
fit Transit    = [True]
------------------------------
LC data        = free
cadence time   =  30.000 min
n rebinning    = 1
Stellar priors = [True]
------------------------------
  PLANET testb
------------------------------
        PRIOR RANGES
------------------------------
T0 = u[ 2448285.0500 , 2448285.1500 ]
P  = u[ 365.2060 , 365.3060 ]
ew1= f[ 0.0000 , 1.0000 ]
ew2= f[ 0.0000 , 1.0000 ]
b  = f[ 0.0000 , 1.0000 ]
a  = g[ 200.0000 , 250.0000 ]
rp = u[ 0.0000 , 0.1000 ]
K  = u[ 0.0000 , 0.0010 ]
------------------------------
 Other parameter priors
------------------------------
q1 = g[ 0.3464 , 0.0500 ]
q2 = g[ 0.2839 , 0.0500 ]
Super_telescope = u[ 21.0719 , 23.0721 ]
==============================
 CREATING RANDOM SEED
 CREATING CHAINS

 STARTING MCMC CALCULATION
 RV datapoints  =           36
 TR datapoints  =          112
 No. parameters =            8
 dof            =          140
 ==================================
      Chain statistics
 ==================================
 chain |  reduced chi^2
 best  :   64667.93
 worst : **********
 mean  : **********
 ==================================
 STARTING INFINITE LOOP!
 ==================================
      Chain statistics
 ==================================
 chain |  reduced chi^2
 best  :       0.95
 worst :       1.05
 mean  :       0.99
 ==================================
 ==================================
      PERFOMING GELMAN-RUBIN
       TEST FOR CONVERGENCE
 ==================================
 ==================================
 CHAINS HAVE NOT CONVERGED YET!
         500  ITERATIONS MORE!
 ==================================
 ==================================
      Chain statistics
 ==================================
 chain |  reduced chi^2
 best  :       0.95
 worst :       1.09
 mean  :       0.99
 ==================================
 ==================================
      PERFOMING GELMAN-RUBIN
       TEST FOR CONVERGENCE
 ==================================
 ==================================
       CHAINS HAVE CONVERGED
 ==================================
    CREATING OUTPUT DATA FILES
 ==================================
STARTING CHAIN CLUSTERING
Initial number of chains: 100
Final number of chains: 100

--------------------------------------------------------------
Summary:
N_chains         =      100
N_iter           =      500
thin_factor      =        1
N_rv_data        =       36
N_tr_data        =      112
N_data           =      148
N_pars           =        8
chi2_rv          = 7.2765
chi2_tr          = 125.9964
chi2             = 133.2729
dof              =      140
chi2/dof         = 0.9519
ln likelihood_rv = 377.7453
ln likelihood_tr = 1131.1755
ln likelihood    = 1508.9208
BIC              = -2977.8640
AIC              = -3001.8417
--------------------------------------------------------------
             INPUT STELLAR PARAMETERS
--------------------------------------------------------------
M_*     = 1.0000000 - 0.1000000 + 0.1000000 solar masses
R_*     = 1.0000000 - 0.1000000 + 0.1000000 solar radii
T_*     = 5772.0000000 - 100.0000000 + 100.0000000 K
--------------------------------------------------------------
                   Parameters test b
-------------------------Fitted-------------------------------
        T0 = 2448285.0955436 - 0.0031192 + 0.0031409     days
         P = 365.2508520 - 0.0049167 + 0.0048356     days
      ew 1 = 0.0000000 - 0.0000000 + 0.0000000
      ew 2 = 0.0000000 - 0.0000000 + 0.0000000
         b = 0.0000000 - 0.0000000 + 0.0000000
      a/R* = 215.4640517 - 2.3708930 + 2.5122541
     rp/R* = 0.0093235 - 0.0000712 + 0.0000775
         K = 0.0881821 - 0.0024223 + 0.0023856      m/s
-------------------------Derived------------------------------
        Mp = 0.9841032 - 0.0704355 + 0.0718640  M_earth
        Rp = 1.0173252 - 0.1028264 + 0.1013333  R_earth
         e = 0.0000000 - 0.0000000 + 0.0000000
         w = 0.0000000 - 0.0000000 + 0.0000000      deg
         i = 90.0000000 - 0.0000000 + 0.0000000      deg
         a = 1.0019439 - 0.1014392 + 0.1005043       AU
Insolation = 0.9967687 - 0.0707513 + 0.0741733  F_Earth
      rho* = 1.4183711 - 0.0463035 + 0.0502306 g/cm^3 (transit)
      rho* = 1.4065628 - 0.3676615 + 0.5586840 g/cm^3 (stellar paramters)
     rho_p = 5.1298925 - 1.3074050 + 1.9887038   g/cm^3
       g_p = 936.8970215 - 30.5861395 + 32.3670771 cm/s^2 (K and Rp/R*)
       g_p = 933.8572371 - 172.6787114 + 234.1233896 cm/s^2 (planet parameters)
     Tperi = 2448193.7827742 - 0.0039813 + 0.0041039     days
       Teq = 278.1026529 - 5.0720559 + 5.0352570 K (albedo=0)
     T_tot = 13.0713846 - 0.1507021 + 0.1448215    hours
    T_full = 12.8294707 - 0.1483685 + 0.1432300    hours
--------------------------------------------------------------
--------------------  Other parameters -----------------------
        q1 = 0.3760168 - 0.0418577 + 0.0442539
        q2 = 0.3157777 - 0.0460379 + 0.0468480
        u1 = 0.3855446 - 0.0574350 + 0.0598140
        u2 = 0.2246861 - 0.0583548 + 0.0612362
Sys. vel. Super_telescope = 22.0719865 - 0.0000015 + 0.0000016      m/s
--------------------------------------------------------------

Creating  outpy/test_out/test_posterior.pdf
Creating  outpy/test_out/test_correlations.pdf
Creating  outpy/test_out/testb_tr.pdf
Creating  outpy/test_out/test_rv_all.pdf
Creating  outpy/test_out/testb_rv.pdf

If you see this output it means that pyaneti ended succesfully!

Now let us check the plots.

evince outpy/test_out/testb_tr.pdf outpy/test_out/testb_rv.pdf

You will see some nice plots that look like this

Let me explain you briefly what this test fit was about:

If you were an advanced alien civilization with really high technology, and "lucky" enough to see an Earth-like planet crossing in front of a Sun-like star, this is how the Earth would look like to you.

Look at those well-known parameters:

  • 1 Earth Mass
  • 1 Earth radii
  • Period of 365 days
  • 1 AU semi-major axis
  • Density of ~5.5 g/cm^2,
  • Gravity of ~10 m/s^2.

Of course you would need a spectograph with a precision of a few cm/s and also a very nice photometer.

If you are at this point, you learned two things. First, with good data you can obtain really nice planet parameters and second, you learned how to run pyaneti.

Documentation

Play with test - Joint radial velocity and transit light curve fitting.

  • There is a directory called inpy, inside this folder you will find a second directory called test.

  • This directory constains the input and data files to perform the test fit.

  • You can create an input director inside inpy for each of your systems!

  • We encorauge you to start to play with the code.

Create your own test directory and copy all the files from inpy/test folder.

mkdir inpy/my_test
cp inpy/test/* inpy/my_test

Now you are ready to run my_test

./pyaneti.py my_test

You will see an output similar to that the test case. Now the output files are inside outpy/my_test_out. You will notice that inside this directory you will find a extra file with the posterior distribution and correlation plots of the fitted parameters.

Now open the file inpy/my_test/input_fit.py and start to play with it. The file is comented. Let us change the priors for some parameters. Uncomment lines 56, 91 and 92 to fit for the scaled semi-major axis and impact factor. Save the changes and re-run the code.

./pyaneti.py my_test

Now you can see that the fitted parameters are different in comparison with the values given by test.

If you have some RV and/or transit data you only have to put the name of your data files, change the prior ranges, and start to fit your data!

Parallel run

Run the code in parallel is really easy.

Just compile the code in parallel (you need openMP installed).

make para

if you have all the libraries installed, the compilation should finish without any problem. Now you only need to run the code.

./pyaneti.py test

This option will run the code with all the processors available in your computer. If you want to specify the number of CPUs to be use by pyaneti, you have to run the env OMP_NUM_THREADS=N option, where N is the number of CPUs.

env OMP_NUM_THREADS=2 ./pyaneti.py test

More documentation will come soon!

Science with pyaneti

  • Prieto-Arranz et al., 2018, Mass determination of the 1:3:5 near-resonant planets transiting GJ 9827 (K2-135), A&A, submitted
  • Barragán et al., 2018, K2-141 b: A 5-M_Earth super-Earth transiting a K7 V star every 6.7 hours A&A, 612, A95
  • Niraula et al., 2017, Three Small Super-Earths Transiting the nearby star GJ 9827, AJ, in press..
  • Gandolfi et al., 2017, The transiting multi-planet system HD3167: a 5.7 MEarth Super-Earth and a 8.3 MEarth mini-Neptune, AJ, 154, 123.
  • Guenther et al., 2017, K2-106, a system containing a metal rich planet and a planet of lower density, A&A, in press..
  • Fridlund et al., 2017, EPIC 210894022b - A short period super-Earth transiting a metal poor, evolved old star, A&A, 604, A16.
  • Barragán et al., 2017, EPIC 218916923 b: a low-mass warm Jupiter on a 29-day orbit transiting an active K0V star, MNRAS, submitted.
  • Nespral et al., 2017, Mass determination of K2-19b and K2-19c from radial velocities and transit timing variations, A&A, 601A, 128..
  • Nowak et al., 2017, EPIC 219388192 b - an inhabitant of the brown dwarf desert in the Ruprecht 147 open cluster, AJ, 153, 131.
  • Eigmüller et al., 2017, K2-60b and K2-107b. A sub-jovian and a jovian planet from the k2 mission, AJ, 153, 130.
  • Barragán et al, 2016, K2-98b: A 32-M⊕ Neptune-sized planet in a 10-day orbit transiting an F8 star, AJ, 152, 6.

See the list of citations to the code here

Citing

If you use pyaneti in your research, please cite it as

Barragán, Gandolfi, & Antoniciello, 2017, Astrophysics Source Code Library, ascl:1707.003

you can use the bibTeX entry

@MISC{2017ascl.soft07003B,
   author = {{Barrag{\'a}n}, O. and {Gandolfi}, D. and {Antoniciello}, G.},
    title = "{pyaneti: Multi-planet radial velocity and transit fitting}",
 keywords = {Software },
howpublished = {Astrophysics Source Code Library},
     year = 2017,
    month = jul,
archivePrefix = "ascl",
   eprint = {1707.003},
   adsurl = {http://adsabs.harvard.edu/abs/2017ascl.soft07003B},
  adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

What will come next?

  • Gaussian process.
  • TTV.
  • Multiband transit photometry fitting.
  • Graphical User Interface.

If you have any comments, requests, suggestions or just need any help, please don't think twice, just contact us!

Warning: This code is under developement and it may contain bugs. If you find something please contact us at oscaribv@gmail.com

Acknowledgements

  • Hannu Parviainen, thank you for helping us to interpret the first result of the PDF of the MCMC chains. We learned a lot from you!
  • Salvador Curiel, thank you for suggestions to parallelize the code.
  • Mabel Valerdi, thank you for being the first pyaneti user, for spotting typos and errors in this document. And thank you much for the awesome idea for pyaneti's logo.
  • Lauren Flor, thank you for testing the code before release.

THANKS A LOT!

About

A multi-planet Radial Velocity and Transit fit software

License:GNU General Public License v3.0

Languages

Language:Python 61.8%Language:Fortran 37.5%Language:Makefile 0.7%