Brans-Dicke with MCMC Hi-CLASS and Planck likelihood

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
Fabien Dournac
Posts: 21
Joined: March 27 2010
Affiliation: CERFACS
Contact:

Brans-Dicke with MCMC Hi-CLASS and Planck likelihood

Post by Fabien Dournac »

Hello,

I am faced to an issue that could seem to be simple to fix. I am looking for the posterior of Brans-Dicke parameter omega_BD.
For this, I am using autoemcee. My problem is that I can't get accepted proposal to build a MCMC chain.

Here is a part of code :

Code: Select all

path_to_planck_likelihood = './baseline/plc_3.0/hi_l/plik_lite/plik_lite_v22_TT.clik'
planck_likelihood = clik.clik(path_to_planck_likelihood)

def get_spectrum(omega_BD):

    omega_BD = float(omega_BD)
    params['parameters_smg'] = '0.7,' + str(omega_BD) + ', 1., 1.'

    # Configurez et exécutez CLASS
    cosmology = Class()
    cosmology.set(params)
    cosmology.compute()

    # Obtenez les spectres pertinents
    cl = cosmology.lensed_cl(2500)
    spectrum = cl['tt']
    #spectrum = np.append(spectrum, [0]*9)
    cosmology.struct_cleanup()
    cosmology.empty()
    return spectrum
1) As you can see, I am obliged to fill to zeros the last 10 values of spectrum because the code prints :

Code: Select all

   loglike = -0.5 * planck_likelihood(spectrum)
  File "lkl.pyx", line 82, in clik.lkl.clik.__call__
Exception: Bad shape (expecting (-1,2510) got (1,2501))
Why Hi-CLASS can't generate the 2510 spectra and only 2500 since Planck likelihood seems to also expect 2510 values ?

2) Is the filling by zero of last 10 elements correct ?


3) Have I well taken the good type of spectrum like this :

Code: Select all

spectrum = cl['tt']
??

Any help is welcome.

ps : here the file brans_dicke init file used by Hi-CLASS :

Code: Select all

Parameter file for Brans Dicke theory
 -> See Avilez+13 (1303.4330) or Bellini+17 (1709.09135)

Omega_Lambda = 0
Omega_fld = 0 
Omega_smg = -1 


1) Brans-Dicke theory:  G_2 -> (omega_BD/phi)X -V0,  
                        G_4 -> phi M_p^2/2
   initial field value sets M_*^2 (effective Planck mass)

gravity_model = brans dicke 
	               V0, omega_BD,  phi_ini,  phi'_ini
#parameters_smg =   0.7,   45000,       1.,        0
parameters_smg =   0.7,   42500,       1.,         1e-3 


2) Options for tuning of effective M_pl using phi_ini
   Note that (b,c) are very close for large omega_BD (viable models)
   
   a) No specifications uses the input_value
   
#comment M_pl_today_smg line below
   
   b) Specifying M_pl_today fixes phi_0 
      (overwrites phi_ini)

M_pl_today_smg = 1.0
   
   c) Specifying M_pl_today AND normalize_G_NR 
      sets G_eff -> 1 for non-relativistic matter
      (overwrites phi_ini AND M_pl_today_smg)
 
#normalize_G_NR = yes


3) Brans Dicke runs into some numerical problems at very early times
a_min_stability_test_smg = 1e-6



------ other parameters --------

root = output/brans_dicke_

#output = tCl,mPk 

output_background_smg = 10
write parameters = yeap
write background = yup
write thermodynamics = yeahh

input_verbose = 4
background_verbose = 4
output_verbose = 1
thermodynamics_verbose = 1
perturbations_verbose = 2
spectra_verbose = 2
Post Reply