def append_fhl(models, bmax,
bin_models, bin_dict, bin_distx, bin_disty, bin_radr, bin_frlog,
snr_models, snr_dict, snr_distx, snr_disty, snr_radr, snr_frlog,
isnr_models, isnr_dict, isnr_distx, isnr_disty, isnr_radr, isnr_frlog,
dist_sigma=3., sig50_thresh=3., eph_thresh=100.):
"""
Append missing models from Fermi high-energy catalog to gammalib model container
:param models: ~gammalib.GModels, gammalib model container
:param bmax: float, maximum latitude of models to include
:param dist_sigma: float, minimum distance in sigma's from pre-existing source to add as new
:param sig50_thresh: float, threshold sigma on significance above 50 GeV to apply
:param eph_thresh: float, minimum energy of detected photons required
:return: result: dictionary
"""
# filter FHL table based on latitude and hardness
# hard sources are seleted based on significance above 50 GeV and highest photon energy
# calculate significance above 50 GeV
sig50 = np.sqrt(np.sum(np.power(fhl['Sqrt_TS_Band'][:,2:], 2),axis=1))
# filter
fsources = fhl[(np.abs(fhl['GLAT']) <= bmax) & (sig50 >= sig50_thresh) & (
fhl['HEP_Energy'] >= eph_thresh)]
# prepare new gammalib model container
newpt = 0
newext = 0
n_bin_del = 0
n_snr_del = 0
n_isnr_del = 0
newmodels = gammalib.GModels()
# keep track also of artificial cutoffs
ecut_pwn = []
ecut_snr = []
ecut_unid = []
ecut_agn = []
n_ecut_pwn = 0
n_ecut_snr = 0
n_ecut_agn = 0
n_ecut_unid = 0
msg = ''
# loop over fermi sources
for fsource in fsources:
# assume source is new
new = 1
# fermi source position as gammalib.GSkyDir
fdir = gammalib.GSkyDir()
ra = np.double(fsource['RAJ2000'])
dec = np.double(fsource['DEJ2000'])
fdir.radec_deg(ra, dec)
######################################### treat pointlike sources for Fermi ###########
if fsource['Extended_Source_Name'] == '':
# case of sources pointlike for Fermi
# loop over gammalib container and determine closest neighbor
# initialize at 1000
dist_min = 1000.
for source in models:
dir = get_model_dir(source)
dist = fdir.dist_deg(dir)
dist_min = np.minimum(dist, dist_min)
if dist_min < dist_sigma * fsource['Conf_95_SemiMajor'] / 2:
# closeby source foud in container, source will not be added
new = 0
else:
# source will be added to container, set spatial model
spatial = gammalib.GModelSpatialPointSource(fdir)
newpt += 1
######################################### treat extended sources for Fermi ############
else:
# retrieve Fermi extended source radius
ext_fsource = ext_fhl[ext_fhl['Source_Name'] == fsource['Extended_Source_Name']][0]
fradius = np.double(ext_fsource['Model_SemiMajor'])
# "corrected" distance, i.e, centre distance - radius of source
# initialize at 1000
dist_corr = 1000.
for source in models:
dir = get_model_dir(source)
dist = fdir.dist_deg(dir)
# get source radius according to model used
radius = get_model_radius(source)
# use as radius the max between Fermi and model
radius = np.maximum(fradius,radius)
# subtract from distance the max radius
dist -= radius
dist_corr = np.minimum(dist,dist_corr)
if dist_corr < 0.:
# overlapping source foud in container, source will not be added
new = 0
else:
# source will be added to container, set spatial model
fradius2 = np.double(ext_fsource['Model_SemiMinor'])
fpangle = np.double(ext_fsource['Model_PosAng'])
# retrieve Fermi spatial model to set it in container
if ext_fsource['Model_Form'] == 'Disk':
if fradius2 == fradius:
spatial = gammalib.GModelSpatialRadialDisk(fdir,fradius)
else:
spatial = gammalib.GModelSpatialEllipticalDisk(fdir, fradius, fradius2, fpangle)
elif ext_fsource['Model_Form'] == '2D Gaussian':
if fradius2 == fradius:
spatial = gammalib.GModelSpatialRadialGauss(fdir,fradius)
else:
spatial = gammalib.GModelSpatialEllipticalGauss(fdir, fradius, fradius2, fpangle)
elif ext_fsource['Model_Form'] == 'Ring':
spatial = gammalib.GModelSpatialRadialShell(fdir,fradius, fradius2)
elif ext_fsource['Model_Form'] == 'Map':
print('{} modeled by spatial template, which is not implemented, skip'.format(fsource['Source_Name']))
new = 0
else:
print('{} modeled by model type {}, which is not implemented, skip'.format(fsource['Source_Name'],ext_fsource['Model_Form']))
new = 0
if new == 1:
newext +=1
######################################### spectra #####################################
if new == 1:
# spectral model
eref = gammalib.GEnergy(np.double(fsource['Pivot_Energy']), 'GeV')
norm = np.double(fsource['Flux_Density'])
norm *= 1.e-3 # ph GeV-1 -> ph MeV-1
# use power law model only if signifcance of curvature < 1
# this avoids extrapolating hard power laws not justified by the data
if fsource['Signif_Curve'] < 1:
index = np.double(fsource['PowerLaw_Index'])
if index < 2.4:
# correction for fake pevatrons
if fsource['CLASS'] == 'PWN' or fsource['CLASS'] == 'pwn':
# dummy model to obtain search radius
mod = gammalib.GModelSky(spatial, gammalib.GModelSpectralPlaw())
# search radius
rad = get_model_radius(mod) + 0.2
# set cutoff
ecut = get_cutoff(ra, dec, 'PSR', rad_search=rad)
ecut_pwn.append(ecut)
n_ecut_pwn += 1
elif fsource['CLASS'] == 'SNR' or fsource['CLASS'] == 'snr':
gname = fsource['ASSOC1']
if 'SNR' in gname:
pass
else:
gname = None
# compute cutoff
# we use default value for particle spectral index because measurements are highly uncertain
ecut = get_cutoff(ra, dec, 'SNR', name=gname)
ecut_snr.append(ecut)
n_ecut_snr += 1
elif fsource['CLASS'] == 'bll' or fsource['CLASS'] == 'bcu' or fsource['CLASS'] == 'fsrq':
ecut = get_cutoff(ra, dec, 'AGN', z = fsource['Redshift'])
ecut_agn.append(ecut)
n_ecut_agn += 1
else:
if fsource['CLASS'] == '' or fsource['CLASS'] == 'unknown':
pass
else:
# set warning if we have hard source of unexpected type
msg += 'FHL source {} of type {} has an unxepctedly hard spectrum ' \
'with index {}. We are setting a random artificial cutoff\n'.format(
fsource['Source_Name'], fsource['CLASS'], index)
print(msg)
ecut = get_cutoff(ra, dec, 'UNID')
ecut_unid.append(ecut)
n_ecut_unid += 1
ecut = gammalib.GEnergy(np.double(ecut), 'TeV')
spectral = gammalib.GModelSpectralExpPlaw(norm, -index, eref, ecut)
else:
spectral = gammalib.GModelSpectralPlaw(norm, -index, eref)
else:
index = np.double(fsource['Spectral_Index'])
curvature = np.double(fsource['beta'])
spectral = gammalib.GModelSpectralLogParabola(norm, -index, eref, curvature)
# assemble model and append to container
model = gammalib.GModelSky(spatial, spectral)
model.name(fsource['Source_Name'])
newmodels.append(model)
# delete synthetic sources as needed
if fsource['CLASS'] == 'HMB' or fsource['CLASS'] == 'hmb' or fsource['CLASS'] == 'BIN' or fsource['CLASS'] == 'bin':
rname, bin_dict, distx, disty, radr, frlog = find_source_to_delete(bin_dict, fdir.l_deg(),
fdir.b_deg(), get_model_radius(model),
flux_Crab(model,1.,1000.))
bin_models.remove(rname)
n_bin_del += 1
bin_distx.append(distx)
bin_disty.append(disty)
bin_radr.append(radr)
bin_frlog.append(frlog)
elif fsource['CLASS'] == 'PWN' or fsource['CLASS'] == 'pwn' or fsource['CLASS'] == 'spp':
# implement synthetic PWNe here
pass
elif fsource['CLASS'] == 'SNR' or fsource['CLASS'] == 'snr':
# determine if snr is interacting
assoc = snr_class[snr_class['ASSOC1'] == fsource['ASSOC1']]
if assoc[0]['is int'] == 'yes':
# interacting
rname, isnr_dict, distx, disty, radr, frlog = find_source_to_delete(isnr_dict,
fdir.l_deg(),
fdir.b_deg(), get_model_radius(model),
flux_Crab(model, 1.,1000.))
isnr_models.remove(rname)
n_isnr_del += 1
isnr_distx.append(distx)
isnr_disty.append(disty)
isnr_radr.append(radr)
isnr_frlog.append(frlog)
else:
# young
rname, snr_dict, distx, disty, radr, frlog = find_source_to_delete(snr_dict,
fdir.l_deg(),
fdir.b_deg(), get_model_radius(model),
flux_Crab(model,1.,1000.))
snr_models.remove(rname)
n_snr_del += 1
snr_distx.append(distx)
snr_disty.append(disty)
snr_radr.append(radr)
snr_frlog.append(frlog)
else:
pass
else:
# source already present, skip
pass
# add new models to original container
for model in newmodels:
models.append(model)
# assemble output in dictionary
result = { 'models' : models, 'newpt' : newpt, 'newext' : newext,
'ecut_pwn' : ecut_pwn, 'ecut_snr' : ecut_snr, 'ecut_agn' : ecut_agn,
'ecut_unid' : ecut_unid, 'n_ecut_pwn' : n_ecut_pwn, 'n_ecut_snr' : n_ecut_snr,
'n_ecut_agn' : n_ecut_agn, 'n_ecut_unid' : n_ecut_unid, 'msg' : msg,
'bin_models' : bin_models, 'bin_dict' : bin_dict, 'n_bin_del' : n_bin_del,
'bin_distx' : bin_distx, 'bin_disty' : bin_disty, 'bin_radr' : bin_radr, 'bin_frlog' : bin_frlog,
'snr_models': snr_models, 'snr_dict': snr_dict, 'n_snr_del': n_snr_del,
'snr_distx': snr_distx, 'snr_disty': snr_disty, 'snr_radr': snr_radr, 'snr_frlog': snr_frlog,
'isnr_models': isnr_models, 'isnr_dict': isnr_dict, 'n_isnr_del': n_isnr_del,
'isnr_distx': isnr_distx, 'isnr_disty': isnr_disty, 'isnr_radr': isnr_radr, 'isnr_frlog': isnr_frlog
}
return result
# Set artifical cutoffs for sources measured with hard PL spectra without measured cutoff
# Binaries and pulsars are skipped because they are addressed later
# We also leave Westerlund 1 and HESS J1641-463 as PeVatron candidates
if index < 2.4 and not source['classes'] == 'bin'\
and not source['classes'] == 'psr'\
and not source['common_name'] == 'Westerlund 1'\
and not source['common_name'] == 'HESS J1641-463':
# if source may be PWN treat as such
if 'pwn' in source['classes']:
# dummy model to obtain search radius
mod = gammalib.GModelSky(spatial,
gammalib.GModelSpectralPlaw())
# search radius
rad = get_model_radius(mod) + 0.2
# set cutoff
ecut = get_cutoff(ra, dec, 'PSR', rad_search=rad)
ecut_pwn.append(ecut)
n_ecut_pwn += 1
# otherwise consider SNR
elif 'snr' in source['classes']:
# try to get Green name
onames = source['other_names'].split(',')
gname = [
name for name in onames
if name[:5] == 'SNR G' or name[0] == 'G'
]
if len(gname) > 0:
gname = gname[0]
else:
gname = None
hess = True