Convergence failure for small planet

[bmorris3]

Hey Laura!

I'm working on simulating transits of Mars-sized planets orbiting Sun-sized stars, and it seems that I'm butting up against some convergence failures by going to small planet sizes. For example, the following code:

from batman import TransitParams, TransitModel
import numpy as np
from astropy.constants import R_earth, R_sun
import astropy.units as u

params = TransitParams()
params.per = 365
params.rp = 0.6 * float(R_earth/R_sun)
params.a = float(1*u.AU/R_sun)
params.inc = 89.9

# Sing 2009: https://arxiv.org/pdf/0912.2274.pdf
params.u = [0.6283, -0.3063, 0.8965, -0.4593] 
params.limb_dark = "nonlinear"

params.ecc = 0
params.w = np.pi/2
params.t0 = 0


times = np.linspace(-0.3, 0.3, 1000)
m = TransitModel(params, times)

m.light_curve(params)

Raises the error

---------------------------------------------------------------------------
Exception                                 Traceback (most recent call last)
<ipython-input-4-5f6b6e406502> in <module>()
     20 
     21 times = np.linspace(-0.3, 0.3, 1000)
---> 22 m = TransitModel(params, times)
     23 
     24 m.light_curve(params)

/Users/bmmorris/anaconda/lib/python3.5/site-packages/batman/transitmodel.py in __init__(self, params, t, max_err, nthreads, fac, transittype, supersample_factor, exp_time)
    125 
    126                 if fac != None: self.fac = fac
--> 127                 else: self.fac = self._get_fac()
    128 
    129                 if nthreads==None or nthreads == 1: self.nthreads=1

/Users/bmmorris/anaconda/lib/python3.5/site-packages/batman/transitmodel.py in _get_fac(self)
    209                                 else: fac_lo = fac
    210                                 n += 1
--> 211                                 if n > 1e3: raise Exception("Convergence failure in calculation of scale factor for integration step size")
    212                         return fac
    213                 else: return None

Exception: Convergence failure in calculation of scale factor for integration step size

Am I doing something wrong, or should I change the hard-coded convergence limit (1e3) to a higher number?

Thanks!

[bmorris3]

It looks like if I supply the TransitModel class with a fac value (for example, 1e-6), I get transit light curves for arbitrarily small planets – is that the parameter I should be adjusting?

[lkreidberg]

Yeah, for a quick fix I'd recommend sending fac by hand. My guess is that my automated calculation fails because the starting value is too large - I'll take a look!