def curvelens(cmb, phi, nside):
alm_cmb = hp.map2alm(cmb)
alm_phi = hp.map2alm(phi)
grad_alm_cmb = hp.alm2map_der1(alm_cmb, nside)
grad_alm_phi = hp.alm2map_der1(alm_phi, nside)
lensedmap = cmb + grad_alm_phi * grad_alm_cmb
return lensedmap
def map_derivatives(m):
"""
Compute the derivatives of a map
"""
npix = m.size
nside = hp.npix2nside(npix)
# Take the spherical Fourier transform of the map
alm = hp.map2alm(m, iter=1)
# Convert back to real space and also get the first derivatives
m_map, der_theta_map, der_phi_map = hp.alm2map_der1(alm, nside)
# Fourier transform the first derivatives
der_theta_alm = hp.map2alm(der_theta_map, iter=1)
der_phi_alm = hp.map2alm(der_phi_map, iter=1)
# Convert the first derivatives back to real space and also get the second derivatives
der_theta_map2, der2_theta2_map, der2_theta_phi_map = hp.alm2map_der1(
der_theta_alm, nside)
der_phi_map2, der2_phi_theta_map, der2_phi2_map = hp.alm2map_der1(
der_phi_alm, nside)
# return all
return [
der_theta_map, der_phi_map, der2_theta2_map, der2_phi2_map,
der2_theta_phi_map
]
def test_der1(self):
""" compare output of alm2map_der1 with the equivalent spin-1 transform using alm2map_spin """
m, dt_der1, dp_der1 = hp.alm2map_der1( self.almg, self.nside )
alm_spin = hp.almxfl( self.almg, np.array( [np.sqrt(l*(l+1.)) for l in six.moves.xrange(0,self.lmax+1)] ), inplace=False )
dt_spin, dp_spin = hp.alm2map_spin( [alm_spin, alm_spin*0.], self.nside, 1, self.lmax )
np.testing.assert_array_almost_equal( dt_der1, dt_spin, decimal=8)
np.testing.assert_array_almost_equal( dp_der1, dp_spin, decimal=8)
def evaluate_gradient(self):
"""Evaluate the 2D gradient of the CMB map.
Evaluate the 2D gradient of the original input map across
the sky. This method uses the ``healpy`` function that returns
d/dtheta and d/dphi/sin(theta).
"""
alm = hp.map2alm(self.map_in)
m, self.map_dtheta, self.map_dphi = hp.alm2map_der1(
alm, self._NSIDE_small)
del m, alm
def compute_intensity_derivatives(self, fromalm=False):
"""
Compute derivatives of the input temperature map (healpix).
Not updated for dichroic for the moment.
Parameters
----------
fromalm : bool, optional
If True, loads alm file from disk instead of fourier
transform the input map. Automatically turns True if you input
alm files. False otherwise.
Examples
----------
>>> filename = 's4cmb/data/test_data_set_lensedCls.dat'
>>> hpmap = HealpixFitsMap(input_filename=filename, fwhm_in=3.5,
... nside_in=16, compute_derivatives=True, map_seed=489237)
>>> hasattr(hpmap, 'dIdp')
True
"""
if fromalm:
alm = hp.read_alm(self.input_filename[0])
else:
alm = hp.map2alm(self.I, self.lmax)
# lmax = hp.Alm.getlmax(alm.size)
if "T1" in self.derivatives_type:
junk, self.dIdt, self.dIdp = hp.alm2map_der1(
alm, self.nside_in, self.lmax)
else:
# computes first and second derivative as derivatives of spin-1
# transform of a scalar field with _1Elm=sqrt(l(l+1))Ilm _1Blm=0
l = np.arange(self.lmax + 1)
grad = np.sqrt(l * (l + 1))
curl = np.zeros_like(alm)
dervs = alm2map_spin_der1([hp.almxfl(alm, grad), curl],
self.nside_in, 1)
self.dIdt = dervs[0][0]
self.dIdp = dervs[0][1]
self.d2Id2t = dervs[1][0]
self.d2Id2p = dervs[2][1]
self.d2Idpdt = dervs[2][0]
def gmfs(self, input_map):
"""calculates mask-edge-trimed MFs for a single field"""
_alm = hp.map2alm(input_map)
_map, _Dthe, _Dphi = hp.alm2map_der1(_alm, self._nside)
#_Dthethe = self.d11(_Dthe)
#_Dthephi = self.d12(_Dphi)
#_Dphiphi = self.d22(_Dthe,_Dphi)
#v2_num = (2.*_Dthe*_Dphi*_Dthephi-(_Dthe**2)*_Dphiphi-(_Dphi**2)*_Dthethe)
v2_denom = (_Dthe**2 + _Dphi**2)
v0 = np.zeros(self._nlv)
#v1 = np.zeros_like(v0)
#v2 = np.zeros_like(v0)
for i in range(self._nlv):
tmp_sigmoid = self.sigmoid(input_map - self._levels[i],
self._width)
v0[i] = np.sum(tmp_sigmoid[self._mask]) / self._neff
#tmp_delta = self.dirac(input_map-self._levels[i],self._width)
#v1[i] = 0.25*np.sum( (tmp_delta*np.sqrt(v2_denom))[self._mask>0] )/self._neff
#v2[i] = (0.5/np.pi)*np.sum( (tmp_delta*v2_num/v2_denom)[self._mask>0] )/self._neff
return v0 #(v0, v1, v2)
pix = hp.vec2pix(NSIDE, x_red, y_red, z_red)
# Transverse momenta for each halo in chi-bin:
theta, phi = hp.vec2ang(np.column_stack((x_red, y_red, z_red))) # in radians
v_theta = np.cos(theta)*np.cos(phi)*vx + np.cos(theta)*np.sin(phi)*vy - np.sin(theta)*vz
v_phi = -np.sin(phi)*vx + np.cos(phi)*vy
# Map of transverse momentum, velocity fields
p_theta_map = np.zeros((hp.nside2npix(NSIDE)), dtype=np.float)
np.add.at(p_theta_map, pix, M/M_avg*v_theta)
p_phi_map = np.zeros((hp.nside2npix(NSIDE)), dtype=np.float)
np.add.at(p_phi_map, pix, M/M_avg*v_phi)
# Angular gradient of transverse momentum field
dthetaPtheta = hp.alm2map_der1(hp.map2alm(p_theta_map), NSIDE)[1]
dphiPphi = hp.alm2map_der1(hp.map2alm(p_phi_map), NSIDE)[2]
dP = dthetaPtheta + dphiPphi # \grad*P = "dP"
# Inverse angular laplacian of transverse momentum ~ 1/grad * P
dPlm = hp.map2alm(dP)
lmax = hp.Alm.getlmax(dPlm.size)
ls = np.arange(lmax)
invlap = 1.0 / ls / (ls + 1.0)
invlap[0] = 0.0 # zero monopole (?)
dPlm = hp.almxfl(dPlm, invlap)
# contribution to map from each radial bin
print("Computing dT_in_bin.")
# total map is this integrated in all bins * geometric factors...
chi_bin = (bin_lower_chi+bin_upper_chi)/2.0
z_bin = zofchi(chi_bin)
pyplot.savefig("dif_pixwin.png")
#%%
"""
STEP 5: Spherical harmonic transforms: alms --> map + 1st derivatives
"""
mapa1 = hp.synfast(Cls,
512,
alm=True,
pol=False,
pixwin=False,
fwhm=np.deg2rad(0.))
alms = mapa1[1]
mapa2 = hp.alm2map_der1(alms, 512)
len(mapa2)
hp.mollview(mapa1[0], title="mapa1[0]")
pyplot.savefig("mapa1_ampl_col0.png")
hp.mollview(mapa2[0], title="mapa2[0]")
pyplot.savefig("mapa2_ampl_col0.png")
hp.mollview(mapa2[1], title="der. theta")
pyplot.savefig("mapa2_derTheta_col1.png")
hp.mollview(mapa2[2], title="der. phi")
pyplot.savefig("mapa2_derPhi_col2.png")
def gentemp(self):
"""Let's fit the pair diff map with the healpix map. """
print('Sampling off template map...')
sys.stdout.flush()
# Get regression template, which is the gnomonic projection of the input
# healpix map around every pixel center
nt = self.ra.size
for k in range(nt):
# Construct T template
#xs = [4.0, 40.0] # x size in degrees
#ys = [4.0, 40.0] # y size in degrees
#res = [0.1, 0.25] # resolution in degrees
xs = 20.0
ys = 20.0
res = 0.1
xx, yy, ra, dec, tempT = gnomproj(self.hmaptemp[0],
self.ra[k],
self.dec[k],
xs,
ys,
res,
rot=self.alpha[k],
Full=True)
tempT = np.ravel(tempT)
if k == 0:
XT = np.zeros((nt, tempT.size))
XT[k] = tempT
# Construct pairdiff predictor from Q and U template maps
tempQ = hp.get_interp_val(self.hmaptemp[1],
self.ra,
self.dec,
lonlat=True)
tempU = hp.get_interp_val(self.hmaptemp[2],
self.ra,
self.dec,
lonlat=True)
chiA = self.alpha
chiB = chiA + 90
siga = tempQ * cosd(2 * chiA) + tempU * sind(2 * chiA)
sigb = tempQ * cosd(2 * chiB) + tempU * sind(2 * chiB)
pairdiff = (siga - sigb) / 2.
pairdiff = pairdiff.reshape(nt, 1)
nQU = 1
# Deriv map templates
Tsm = hp.smoothing(self.hmaptemp[0], fwhm=30.0 / 60. * np.pi / 180)
alm = hp.map2alm(Tsm)
dum, dth, dph = hp.alm2map_der1(alm, self.tempNside)
dth_alm = hp.map2alm(dth)
dph_alm = hp.map2alm(dph)
dth, dth2, dthdph = hp.alm2map_der1(dth_alm, self.tempNside)
dum, dum, dph2 = hp.alm2map_der1(dph_alm, self.tempNside)
# Sample off deriv map templates
Tsm = hp.get_interp_val(Tsm, self.ra, self.dec, lonlat=True)
dth = hp.get_interp_val(dth, self.ra, self.dec, lonlat=True)
dph = hp.get_interp_val(dph, self.ra, self.dec, lonlat=True)
dth2 = hp.get_interp_val(dth2, self.ra, self.dec, lonlat=True)
dph2 = hp.get_interp_val(dph2, self.ra, self.dec, lonlat=True)
dthdph = hp.get_interp_val(dthdph, self.ra, self.dec, lonlat=True)
Xderiv = np.zeros((nt, 6))
Xderiv[:, 0] = Tsm # Gain
Xderiv[:, 1] = dth # Diff point x
Xderiv[:, 2] = dph # Diff point y
Xderiv[:, 3] = dph2 + dth2 # Diff bw
Xderiv[:, 4] = dph2 - dth2 # Diff plus ellip
Xderiv[:, 5] = 2 * dthdph # Cross elip
# Initialize regressor matrix
self.X = np.concatenate((XT, Xderiv, pairdiff), axis=1)
self.nT = tempT.size # Number of T regressors
self.tempxx = xx
self.tempyy = yy
