rtg2.py

import numpy as np
import matplotlib as mpl
mpl.use('Agg') 
mpl.rcParams['text.usetex'] = True 
mpl.rcParams['font.family'] = 'serif'
mpl.rcParams['font.serif'] = 'cm'
mpl.rcParams['font.size'] = '22'
import matplotlib.pyplot as plt
from scipy.integrate import quad
from scipy.interpolate import RectBivariateSpline
import sys
import gammapi
from gammapi import get_gammapi_percentiles
import matplotlib.patches as mpatches

# These npz files generated by rhoqso_fit2.py.
data = np.load('e1450_18.npz')
z18 = data['z']
median18 = data['median']
up18 = data['up']
down18 = data['down']

print 'median18=', median18[:10]

data = np.load('e1450_21.npz')
z21 = data['z']
median21 = data['median']
up21 = data['up']
down21 = data['down']

print 'median21=', median21[:10]



# z18 = data['z']
# medianbright18 = data['medianbright']
# medianfaint18 = data['medianfaint']
# downbright18 = data['downbright']
# upbright18 = data['upbright']
# downfaint18 = data['downfaint']
# upfaint18 = data['upfaint']

# data = np.load('e1450_21_2.npz')
# z21 = data['z']
# medianbright21 = data['medianbright']
# medianfaint21 = data['medianfaint']
# downbright21 = data['downbright']
# upbright21 = data['upbright']
# downfaint21 = data['downfaint']
# upfaint21 = data['upfaint']

# data = np.load('e912_18_2.npz')
# z18_912 = data['z']
# median18_912 = data['median']
# down18_912 = data['down']
# up18_912 = data['up']

# data = np.load('e912_21_2.npz')
# z21_912 = data['z']
# median21_912 = data['median']
# down21_912 = data['down']
# up21_912 = data['up']

print 'data loaded'

z_HM12 = np.loadtxt('z_HM12.txt')
data_HM12 = np.loadtxt('hm12.txt')
w_HM12 = data_HM12[:,0]
e_HM12 = data_HM12[:,1:]

# kx=1 and ky=1 required here to get good interpolation at the Lyman
# break.  Also remember that you will most likely need the
# 'grid=False' option while invoking emissivity_HM12.
emissivity_HM12 = RectBivariateSpline(w_HM12, z_HM12, e_HM12, kx=1, ky=1)

jdata_HM12 = np.loadtxt('j_HM12.txt')
w_j_HM12 = jdata_HM12[:,0]

# Remove lines with same wavelength so that RectBivariateSpline can
# work.  Not sure why these lines occur in HM12 tables.
w_j_HM12, idx = np.unique(w_j_HM12, return_index=True)
j_HM12 = jdata_HM12[idx, 1:]

bkgintens_HM12 = RectBivariateSpline(w_j_HM12, z_HM12, j_HM12, kx=1, ky=1)

yrbys = 3.154e7
cmbympc = 3.24077928965e-25
c_mpcPerYr = 2.998e10*yrbys*cmbympc # Mpc/yr 
c_angPerSec = 2.998e18
nu0 = 3.288e15 # threshold freq for H I ionization; s^-1 (Hz)
hplanck = 6.626069e-27 # erg s

omega_lambda = 0.7
omega_nr = 0.3
h = 0.7

def qso_emissivity_hm12(nu, z):

    """HM12 qso comoving emissivity.

    This is given by equations (37) and (38) of HM12.

    """

    w = c_angPerSec/nu
    nu_912 = c_angPerSec/912.0
    nu_1300 = c_angPerSec/1300.0
    
    a = 10.0**24.6 * (1.0+z)**4.68 * np.exp(-0.28*z)/(np.exp(1.77*z)+26.3)
    b = a * (nu_1300/nu_912)**-1.57

    if w > 1300.0:
        e = b*(nu/nu_1300)**-0.44
    else:
        e = a*(nu/nu_912)**-1.57

    return e

vqso_emissivity_hm12 = np.vectorize(qso_emissivity_hm12, otypes=[np.float])

def qso_emissivity_mh15(nu, z):

    """HM12 qso comoving emissivity.

    This is given by equations (37) and (38) of HM12.

    """

    w = c_angPerSec/nu
    nu_912 = c_angPerSec/912.0
    nu_1300 = c_angPerSec/1300.0

    loge = 25.15*np.exp(-0.0026*z) - 1.5*np.exp(-1.3*z)
    a = 10.0**loge
    
    if w > 912.0:
        e = a*(nu/nu_912)**-0.61
    else:
        e = a*(nu/nu_912)**-1.70

    return e

vqso_emissivity_mh15 = np.vectorize(qso_emissivity_mh15, otypes=[np.float])

def H(z):

    H0 = 1.023e-10*h
    hubp = H0*np.sqrt(omega_nr*(1+z)**3+omega_lambda) # yr^-1

    return hubp

def dzdt(z):

    return -(1.0+z)*H(z) # yr^-1

def dtdz(z):

    return 1.0/dzdt(z) # yr 


def f(N_HI, z):

    """
    HI column density distribution.  This parameterisation, and the
    best-fit values of the parameters, is taken from Becker and Bolton
    2013 (MNRAS 436 1023), Equation (5).

    """

    A = 0.93
    beta_N = 1.33
    beta_z = 1.92
    N_LL = 10.0**17.2 # cm^-2 

    return (A/N_LL) * ((N_HI/N_LL)**(-beta_N)) * (((1.0+z)/4.5)**beta_z) # cm^2

def f_HM12(N_HI, z):

    """HI column density distribution from HM12.

    See their Table 1 and Section 3.

    """

    log10NHI = np.log10(N_HI)

    if z < 1.56:
        if 11.0 <= log10NHI < 15.0:
            a = 1.729816e8 
            b = 1.5
            g = 0.16
        elif 15.0 <= log10NHI < 17.5:
            a = 5.495409e15 
            b = 2.0
            g = 0.16 
        elif 17.5 <= log10NHI < 19.0:
            f17p5 = 5.495409e15 * (1+z)**0.16 * (10.0**17.5)**-2.0
            f19 = 1.279381 * (1+z)**0.16 * (10.0**19)**-1.05
            b = np.log10(f17p5/f19)/1.5
            a = f17p5 * (10.0**17.5)**b
            return a * N_HI**-b
        elif 19.0 <= log10NHI < 20.3:
            a = 1.279381
            b = 1.05 
            g = 0.16 
        elif 20.3 <= log10NHI:
            a = 2.471724e19 
            b = 2.0 
            g = 0.16 
        return a * (1+z)**g * N_HI**-b
    
    elif 1.56 <= z < 5.5:
        if 11.0 <= log10NHI < 15.0:
            a = 1.199499e7
            b = 1.5
            g = 3.0
        elif 15.0 <= log10NHI < 17.5:
            a = 3.801894e14
            b = 2.0
            g = 3.0
        elif 17.5 <= log10NHI < 19.0:
            f17p5 = 3.801894e14 * (1+z)**3.0 * (10.0**17.5)**-2.0
            f19 = 0.449780 * (1+z)**1.27 * (10.0**19)**-1.05
            b = np.log10(f17p5/f19)/1.5
            a = f17p5 * (10.0**17.5)**b
            return a * N_HI**-b
        elif 19.0 <= log10NHI < 20.3:
            a = 0.449780
            b = 1.05
            g = 1.27
        elif 20.3 <= log10NHI:
            a = 8.709636e18
            b = 2.0
            g = 1.27
        return a * (1+z)**g * N_HI**-b

    elif 5.5 <= z:
        if 11.0 <= log10NHI < 15.0:
            a = 29.512092
            b = 1.5
            g = 9.9
        elif 15.0 <= log10NHI < 17.5:
            a = 9.332543e8 
            b = 2.0
            g = 9.9
        elif 17.5 <= log10NHI < 19.0:
            f17p5 = 9.332543e8 * (1+z)**9.9 * (10.0**17.5)**-2.0
            f19 = 0.449780 * (1+z)**1.27 * (10.0**19)**-1.05
            b = np.log10(f17p5/f19)/1.5
            a = f17p5 * (10.0**17.5)**b
            return a * N_HI**-b
        elif 19.0 <= log10NHI < 20.3:
            a = 0.449780 
            b = 1.05
            g = 1.27
        elif 20.3 <= log10NHI:
            a = 8.709636e18
            b = 2.0
            g = 1.27
        return a * (1+z)**g * N_HI**-b

    return 0.0

vf_HM12 = np.vectorize(f_HM12, otypes=[np.float])

def sigma_HI(nu):

    """Calculate the HI ionization cross-section.  

    This parameterisation is taken from Osterbrock and Ferland 2006
    (Sausalito, California: University Science Books), Equation (2.4).

    """
    
    a0 = 6.3e-18 # cm^2

    if nu < nu0:
        return 0.0
    elif nu/nu0-1.0 == 0.0:
        return a0*(nu0/nu)**4
    else:
        eps = np.sqrt(nu/nu0-1.0)
        return (a0 * (nu0/nu)**4 * np.exp(4.0-4.0*np.arctan(eps)/eps) /
                (1.0-np.exp(-2.0*np.pi/eps)))

def tau_eff(nu, z):

    """Calculate the effective opacity between redshifts z0 and z.

    There are two magic numbers: N_HI_min, N_HI_max.  These should
    ideally be 0 and infinity, but I have chosen to avoid improper
    integrals here.

    """

    N_HI_min = 1.0e11
    N_HI_max = 10.0**21.55 # Limit taken in HM12 Table 1. 
    n = np.logspace(np.log(N_HI_min), np.log(N_HI_max), num=50, base=np.e)
    fn = n * vf_HM12(n, z) * (1.0-np.exp(-n*sigma_HI(nu)))
    return np.trapz(fn, x=np.log(n))

def luminosity(M):

    return 10.0**((51.60-M)/2.5) # ergs s^-1 Hz^-1 

def fnu(nu, M):
    
    L = luminosity(M)
    w = c_angPerSec / nu
    nu_912 = c_angPerSec / 912.0
    nu_1450 = c_angPerSec / 1450.0

    a = L 
    b = a * (912.0/1450.0)**0.61
    
    if w > 912.0:
        e = a*(w/1450.0)**0.61
    else:
        if M < -23.0: 
            e = b*(w/912.0)**1.70
        else:
            e = b*(w/912.0)**0.56 
            
    return e 

vfnu = np.vectorize(fnu, excluded=['M'], otypes=[np.float])

def fnu_bright(nu):

    """Only the frequency term for M < -23.

    """
    
    w = c_angPerSec / nu
    nu_912 = c_angPerSec / 912.0
    nu_1450 = c_angPerSec / 1450.0

    a = 1.0
    b = a * (912.0/1450.0)**0.61
    
    if w > 912.0:
        e = a*(w/1450.0)**0.61
    else:
        e = b*(w/912.0)**1.7
            
    return e

vfnu_bright = np.vectorize(fnu_bright, otypes=[np.float])

def fnu_faint(nu):

    """Only the frequency term for M > -23.

    """
    
    w = c_angPerSec / nu
    nu_912 = c_angPerSec / 912.0
    nu_1450 = c_angPerSec / 1450.0

    a = 1.0
    b = a * (912.0/1450.0)**0.61
    
    if w > 912.0:
        e = a*(w/1450.0)**0.61
    else:
        e = b*(w/912.0)**1.7
            
    return e

vfnu_faint = np.vectorize(fnu_faint, otypes=[np.float])

def emissivity(w, z, loglf, theta, mbright=-30, mfaint=-18):

    m = np.linspace(mbright, mfaint, num=100)
    nu = c_angPerSec/w

    farr = np.array([10.0**loglf(theta, x, z)*vfnu(nu, x) for x in m])

    return np.trapz(farr, m, axis=0) # erg s^-1 Hz^-1 Mpc^-3

def emissivity_18(w, z):

    nu = c_angPerSec/w

    e1 = np.interp(z, z18, median18)
    e = e1*vfnu_faint(nu)
    
    return e # erg s^-1 Hz^-1 Mpc^-3

def emissivity_18_down(w, z):

    nu = c_angPerSec/w

    e1 = np.interp(z, z18, down18)

    e = e1*vfnu_faint(nu)
    
    return e # erg s^-1 Hz^-1 Mpc^-3

def emissivity_18_up(w, z):

    nu = c_angPerSec/w

    e1 = np.interp(z, z18, up18)

    e = e1*vfnu_faint(nu)
    
    return e # erg s^-1 Hz^-1 Mpc^-3

def emissivity_21(w, z):

    nu = c_angPerSec/w

    e1 = np.interp(z, z21, median21)

    e = e1*vfnu_faint(nu) 
    
    return e # erg s^-1 Hz^-1 Mpc^-3

def emissivity_21_down(w, z):

    nu = c_angPerSec/w

    e1 = np.interp(z, z21, down21)

    e = e1*vfnu_faint(nu) 
    
    return e # erg s^-1 Hz^-1 Mpc^-3

def emissivity_21_up(w, z):

    nu = c_angPerSec/w

    e1 = np.interp(z, z21, up21)

    e = e1*vfnu_faint(nu) 
    
    return e # erg s^-1 Hz^-1 Mpc^-3

def g_lsa(z, loglf, theta):

    """Photoionisation rate with Local Source Approx.
    
    """

    w = 912.0 
    
    em = emissivity(w, z, loglf, theta, mbright=-30.0, mfaint=-23.0)
    alpha_EUV = -1.7
    part1 = 4.6e-13 * (em/1.0e24) * ((1.0+z)/5.0)**(-2.4) / (1.5-alpha_EUV) # s^-1 

    em = emissivity(w, z, loglf, theta, mbright=-23.0, mfaint=-18.0)    
    alpha_EUV = -0.56
    part2 = 4.6e-13 * (em/1.0e24) * ((1.0+z)/5.0)**(-2.4) / (1.5-alpha_EUV) # s^-1

    return part1 + part2

def gs_lsa(lfg, dz=0.1, zmax=7.0):

    zmin = 0.0
    n = (zmax-zmin)/dz+1
    zs = np.linspace(zmax, zmin, num=n)

    theta = np.median(lfg.samples, axis=0)
    
    gs = np.array([g_lsa(x, lfg.log10phi, theta) for x in zs])

    return zs, gs 
    
def j(emodel, loglf=None, theta=None, dz=0.1, n_ws=200, n_ws_int=100, zmax=7.0):

    """Calculate the mean specific intensity.

    Actually, directly outputs a redshift, photoionisation rate, pair. 

    Needs a model for emissivity.  From my tests, I have found that dz
    = 0.01, n_ws = 600, n_ws_int = 400, and zmax = 9 is necessary for
    convergence.

    """
    
    ws = np.logspace(0.0, 5.0, num=n_ws)
    nu = c_angPerSec/ws 

    j = np.zeros_like(nu)
    
    zmin = 0.0
    n = (zmax-zmin)/dz+1
    zs = np.linspace(zmax, zmin, num=n)
    gs = []

    for z in zs:

        if loglf is not None:
            e = emodel(ws/(1.0+z), z, loglf, theta)
        else:
            e = emodel(ws/(1.0+z), z)
        
        # [j] = erg s^-1 Hz^-1 cm^-2 sr^-1
        j = j + (e*c_mpcPerYr*np.abs(dtdz(z))*dz*cmbympc**2)/(4.0*np.pi) 

        nu_rest = c_angPerSec*(1.0+z)/ws 
        t = np.array([tau_eff(x, z) for x in nu_rest])
        j = j*np.exp(-t*dz)

        n = 4.0*np.pi*j*(1.0+z)**3/(hplanck*nu_rest)

        nu_int = np.logspace(np.log10(nu0), 18, num=n_ws_int)
        n_int = np.interp(nu_int, nu_rest[::-1], n[::-1])
        s = np.array([sigma_HI(x) for x in nu_int])
        g = np.trapz(n_int*s, x=nu_int) # s^-1 

        gs.append(g)

    gs = np.array(gs)

    return zs, gs

def em_hm12(w, z):

    """HM12 Galaxies+QSO emissivity."""

    return emissivity_HM12(w, z*np.ones_like(w), grid=False)

def em_qso_hm12(w, z):

    """HM12 QSOs-only emissivity."""

    return vqso_emissivity_hm12(c_angPerSec/w, z)

def em_qso_mh15(w, z):

    """HM12 QSOs-only emissivity."""

    return vqso_emissivity_mh15(c_angPerSec/w, z)

def calverley(ax):

    zm, gm, gm_sigma = np.loadtxt('Data/calverley.dat',unpack=True) 
    gm += 12.0

    gml = 10.0**gm
    gml_up = 10.0**(gm+gm_sigma)-10.0**gm
    gml_low = 10.0**gm - 10.0**(gm-gm_sigma)
    
    ax.scatter(zm, gml, c='k', edgecolor='None',
               label='Calverley et al.\ 2011', s=64, marker='v') 
    ax.errorbar(zm, gml, ecolor='k', capsize=5,
                elinewidth=1.5, capthick=1.5,
                yerr=np.vstack((gml_low, gml_up)),
                fmt='None', zorder=1, mfc='darkorange',
                mec='darkorange', markeredgewidth=1, ms=5)

    return


def wyithe11(ax):

    zm, gm, gm_u, gm_l = np.loadtxt('Data_new/wyithe11_gammaHI.txt',
                                    unpack=True)

    gml = gm
    gml_up = gm_u
    gml_low = gm_l
    
    ax.scatter(zm, gml, c='k', edgecolor='None',
               label='Wyithe \& Bolton 2011', s=64, marker='^')
    
    ax.errorbar(zm, gml, ecolor='k', capsize=5,
                elinewidth=1.5, capthick=1.5,
                yerr=np.vstack((gml_low, gml_up)),
                fmt='None', zorder=1, mfc='darkorange',
                mec='darkorange', markeredgewidth=1, ms=5)

    return


def daloisio18(ax):

    zmin, zmax, gm, gm_u, gm_l = np.loadtxt('Data_new/daloisio18_gammaHI.txt',
                                            unpack=True)

    zm = (zmin+zmax)/2.0

    
    gml = gm
    gml_up = gm_u
    gml_low = gm_l
    
    ax.scatter(zm, gml, c='dodgerblue', edgecolor='None',
               label="D'Aloisio et al.\ 2018", s=64, marker='h')
    
    ax.errorbar(zm, gml, ecolor='dodgerblue', capsize=5,
                elinewidth=1.5, capthick=1.5,
                yerr=np.vstack((gml_low, gml_up)),
                fmt='None', zorder=1, mfc='darkorange',
                mec='darkorange', markeredgewidth=1, ms=5)

    return


def davies17(ax):

    zmin, zmax, gm, gm_u, gm_l = np.loadtxt('Data_new/davies17_gammaHI.txt',
                                            unpack=True)
    gm += 12.0

    zmin = zmin[(gm_u!=0.0)]
    zmax = zmax[(gm_u!=0.0)]
    gm = gm[(gm_u!=0.0)]
    gm_l = gm_l[(gm_u!=0.0)]
    gm_u = gm_u[(gm_u!=0.0)]

    zm = (zmin+zmax)/2.0
    gml = 10.0**gm
    gml_up = 10.0**(gm+gm_u)-10.0**gm
    gml_low = 10.0**gm - 10.0**(gm-gm_l)
    
    ax.scatter(zm, gml, c='g', edgecolor='None',
               label="Davies et al.\ 2018", s=64, marker='s')
    
    ax.errorbar(zm, gml, ecolor='g', capsize=5,
                elinewidth=1.5, capthick=1.5,
                yerr=np.vstack((gml_low, gml_up)),
                fmt='None', zorder=1, mfc='darkorange',
                mec='darkorange', markeredgewidth=1, ms=5)

    zmin, zmax, gm, gm_u, gm_l = np.loadtxt('Data_new/davies17_gammaHI.txt',
                                            unpack=True)
    gm += 12.0

    zmin = zmin[(gm_u==0.0)]
    zmax = zmax[(gm_u==0.0)]
    gm = gm[(gm_u==0.0)]

    zm = (zmin+zmax)/2.0
    gml = 10.0**gm

    ax.scatter(zm, gml, c='g', edgecolor='None',
               s=64, marker='s')
    
    ax.errorbar(zm, gml, yerr=0.05, ecolor='g', capsize=5, uplims=True,
                elinewidth=1.5, capthick=1.5,
                fmt='None', zorder=1, mfc='darkorange',
                mec='darkorange', markeredgewidth=1, ms=5)

    return


def bb13(ax, puc=False):

    zm, gm, gm_up, gm_low = np.loadtxt('Data/BeckerBolton.dat', unpack=True)
    
    gml = 10.0**gm
    gml_up = 10.0**(gm+gm_up)-10.0**gm
    gml_low = 10.0**gm - 10.0**(gm-np.abs(gm_low))

    if puc:
        ax.scatter(1+zm, gml, c='k', edgecolor='None', marker='v',
                   label='Becker \& Bolton 2013', s=64, zorder=10)
        ax.errorbar(1+zm, gml, ecolor='k', capsize=5,
                    elinewidth=2, capthick=3,
                    yerr=np.vstack((gml_low, gml_up)),
                    fmt='None', zorder=10, mfc='k', mec='k',
                    markeredgewidth=1.5, ms=5)
        return
        
    ax.scatter(zm, gml, c='k', edgecolor='None',
               label='Becker \& Bolton 2013', s=64, zorder=10)
    ax.errorbar(zm, gml, ecolor='k', capsize=5,
                elinewidth=2, capthick=3,
                yerr=np.vstack((gml_low, gml_up)),
                fmt='None', zorder=10, mfc='k', mec='k',
                markeredgewidth=1.5, ms=5)

    return

def g_hm12_total(ax, puc=False):

    zs = np.linspace(0.0, 7.0, num=200)
    nu = np.logspace(np.log10(nu0), 18, num=100)
    g_hm12 = []
    
    for r in zs:
        j_hm12 = bkgintens_HM12(c_angPerSec/nu, r*np.ones_like(nu), grid=False)
        n = 4.0*np.pi*j_hm12/(hplanck*nu)
        s = np.array([sigma_HI(x) for x in nu])
        g_hm12.append(np.trapz(n*s, x=nu)) # s^-1

    if puc:
        ax.plot(1+zs, np.array(g_hm12)/1.0e-12, c='brown', lw=1, dashes=[1,1],
                label='Haardt \& Madau 2012', zorder=2)
        return

        
    ax.plot(zs, np.array(g_hm12)/1.0e-12, c='brown', lw=1, dashes=[1,1],
            label='Haardt \& Madau 2012')

    return

def lfis(individuals, ax):

    for x in individuals:
        get_gammapi_percentiles(x, x.z.mean()) 

    reject = [0, 1, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20]

    m = np.ones(len(individuals), dtype=bool)
    m[reject] = False
    minv = np.logical_not(m) 
    
    individuals_good = [x for i, x in enumerate(individuals)
                        if i not in set(reject)]
    individuals_bad = [x for i, x in enumerate(individuals)
                       if i in set(reject)]

    c = np.array([x.gammapi[2]+12.0 for x in individuals_good])
    u = np.array([x.gammapi[0]+12.0 for x in individuals_good])
    l = np.array([x.gammapi[1]+12.0 for x in individuals_good])

    gml = 10.0**c
    gml_up = 10.0**u-10.0**c
    gml_low = 10.0**c - 10.0**l
    
    zs = np.array([x.z.mean() for x in individuals_good])
    uz = np.array([x.z.max() for x in individuals_good])
    lz = np.array([x.z.min() for x in individuals_good])

    uz = np.array([x.zlims[0] for x in individuals_good])
    lz = np.array([x.zlims[1] for x in individuals_good])
    
    uzerr = uz-zs
    lzerr = zs-lz 

    ax.scatter(zs, gml, c='k', edgecolor='k',
               label='Individual fits (accepted; $M<-18$)',
               s=44, zorder=4, linewidths=1.5) 
    ax.errorbar(zs, gml, ecolor='k', capsize=0, fmt='None', elinewidth=1.5,
                yerr=np.vstack((gml_low,gml_up)),
                xerr=np.vstack((lzerr,uzerr)), 
                mfc='#ffffff', mec='#404040', zorder=3, mew=1,
                ms=5)

    c = np.array([x.gammapi[2]+12.0 for x in individuals_bad])
    u = np.array([x.gammapi[0]+12.0 for x in individuals_bad])
    l = np.array([x.gammapi[1]+12.0 for x in individuals_bad])

    gml = 10.0**c
    gml_up = 10.0**u-10.0**c
    gml_low = 10.0**c - 10.0**l
    
    zs = np.array([x.z.mean() for x in individuals_bad])
    uz = np.array([x.z.max() for x in individuals_bad])
    lz = np.array([x.z.min() for x in individuals_bad])

    uz = np.array([x.zlims[0] for x in individuals_bad])
    lz = np.array([x.zlims[1] for x in individuals_bad])
    
    uzerr = uz-zs
    lzerr = zs-lz 

    ax.scatter(zs, gml, c='#ffffff', edgecolor='k',
               label='Individual fits (rejected; $M<-18$)',
               s=44, zorder=4, linewidths=1.5) 
    ax.errorbar(zs, gml, ecolor='k', capsize=0, fmt='None', elinewidth=1.5,
                yerr=np.vstack((gml_low,gml_up)),
                xerr=np.vstack((lzerr,uzerr)), 
                mfc='#ffffff', mec='#404040', zorder=3, mew=1,
                ms=5)
    
    return 

def Gamma_HI(em, z):

    """Taken from Equation 11 of Lusso et al. 2015.

    """

    alpha_EUV = -0.56
    g = 4.6e-13 * (em/1.0e24) * ((1.0+z)/5.0)**(-2.4) / (1.5-alpha_EUV) # s^-1

    return g 

def gamma_HI_Manti17(z):

    """Manti et al. 2017 (MNRAS 466 1160) Equation (9).

    """

    logg = - 11.66 - 0.081*z - 0.00014*z**2 + 0.0033*z**3 - 0.0013*z**4
    
    return 10.0**logg # s^-1

def kollmeier(ax, puc=False):
    """Plot Kollmeier's Gamma_HI values.

    These are from 2014 ApJ 789 L32.  Adjusted by Gabor to our 373
    cosmology.  See his note for details.

    """

    z = 0.1
    g = 2.2e-13 # s^-1

    if puc:
        ax.scatter(1+z, g*1.0e12, c='k', edgecolor='None',
                   label='Kollmeier et al.\ 2014', s=64, marker='p')
    else:
        ax.scatter(z, g*1.0e12, c='k', edgecolor='None',
                   label='Kollmeier et al.\ 2014', s=64)

    return


def bolton17(ax):
    """Plot Sherwood Gamma_HI value.

    These are from 2017 MNRAS 464 897.  Adjusted by Gabor to our 373
    cosmology.  See his note for details.  We only show the z = 2
    value on the PUC plot. 

    """

    z = 2.0
    g = 1.87e-12 # s^-1

    ax.scatter(1+z, g*1.0e12, c='k', edgecolor='None',
               label='Bolton et al.\ 2017', s=64, marker='s')

    return


def shull(ax, puc=False):
    
    """Plot Shull's Gamma_HI values.

    These are from 2015 ApJ 811 3.  Adjusted by Gabor to our 373
    cosmology.  See his note for details.

    """

    if puc: 
        z = np.linspace(0, 0.47)
        g = 5.0e-14 * (1.0+z)**4.4 # s^-1
        ax.plot(1+z, g*1.0e12, lw=2, c='k', label='Shull et al.\ 2015',
                dashes=[7,2], zorder=5)
        return
    
    z = np.linspace(0, 0.47)
    g = 5.0e-14 * (1.0+z)**4.4 # s^-1
    ax.plot(z, g*1.0e12, lw=4, c='#8c564b', label='Shull et al.\ 2015')

    return


def khaire(ax, puc=False):
    
    """Plot Khaire's Gamma_HI values. 

    These are from 2015 MNRAS 451 L30.  They correspond to Khaire's
    fesc=0 model.  Obtained from Khaire by email. 

    """

    z, g = np.loadtxt('Data_new/khaire15_gammaHI.dat',
                      usecols=(0,1), unpack=True)

    if puc:
        ax.plot(1+z, g*1.0e12, lw=2, c='peru',
                label='Khaire \& Srianand 2015 QSOs',
                zorder=2, dashes=[5,2])
        return 
        
    ax.plot(z, g*1.0e12, lw=2, c='peru',
            label='Khaire \& Srianand 2015 QSOs',
            zorder=2, dashes=[5,2])

    return

def puchwein(ax, puc=False):
    
    """Plot Puchwein's Gamma_HI values. 
    
    These are from https://arxiv.org/src/1801.04931v1/anc/rates_fiducial.txt

    """

    z, g = np.loadtxt('Data_new/puchwein18_gammaHI.dat',
                      usecols=(0,1), unpack=True)

    if puc:

        ax.plot(1+z, g*1.0e12, lw=2, c='brown',
                label='Puchwein et al.\ 2018',
                dashes=[7,2], zorder=2)
        return
    
    ax.plot(z, g*1.0e12, lw=2, c='grey',
            label='Puchwein et al.\ 2018',
            dashes=[7,2], zorder=2)

    return


def bolton(ax):

    """Plot Bolton's Gamma_HI values. 

    These are from MNRAS 2017 464 897, as rescaled by Gabor.

    """

    z = np.array([2.0, 3.0, 4.0, 5.0])
    g = np.array([1.87e-12, 9.15e-13, 8.83e-13, 7.55e-13]) # s^-1

    ax.plot(z, g*1.0e12, lw=2, c='#9467bd', label='Bolton et al.\ 2017')

    return 


def onorbe(ax, puc=False):

    """Plot Onorbe's Gamma_HI values.

    These are from ApJ 2017 837 106, Table 3. 

    """

    log1pz, g = np.loadtxt('Data_new/onorbe17_gammaHI.dat',
                           usecols=(0,1), unpack=True)
    z = 10.0**log1pz - 1.0 

    if puc:
        ax.plot(1+z, g*1.0e12, c='grey', lw=2,
                label='O\~norbe et al.\ 2017',
                dashes=[2,2], zorder=2)
        return
        
    ax.plot(z, g*1.0e12, c='grey', lw=2,
            label='O\~norbe et al.\ 2017',
            dashes=[2,2], zorder=2)
    
    return

def viel(ax, puc=False):

    """Plot Viel's Gamma_HI values. 

    These are from MNRAS 2017 467 L86.

    """

    z = 0.1
    g = 0.071e-12 # s^-1

    if puc:

        zmin = 1.09
        emin = 0.05
        zmax = 1.11
        emax = 0.1 
        dz = zmax - zmin
        de = emax - emin
        rect = mpatches.Rectangle((zmin, emin),  dz, de,
                                  ec='k', color='#ffffff',
                                  lw=2, zorder=10,
                                  label='Viel et al.\ 2017')
        ax.add_patch(rect)
        
        return

    return

def gaikwad_a(ax):

    """Plot Gaikwad's Gamma_HI values.

    These are from his Paper I (MNRAS 2017 466 838) rescaled to our cosmology by Gabor.

    """

    z = np.array([0.1125, 0.2, 0.3, 0.4])
    zmin = np.array([0.075, 0.15, 0.25, 0.35])
    zmax = np.array([0.15, 0.25, 0.35, 0.45])

    uzerr = zmax - z
    lzerr = z - zmin

    g = np.array([7.095e-14, 1.075e-13, 1.559e-13, 2.258e-13])
    gerr = np.array([1.613e-14, 2.258e-14, 3.978e-14, 5.590e-14])

    ax.scatter(z, g*1.0e12, c='#d62728', edgecolor='None',
               label='Gaikwad et al.\ 2017a',
               s=64, zorder=10, linewidths=1.5) 

    ax.errorbar(z, g*1.0e12, ecolor='#d62728', capsize=0,
                fmt='None', elinewidth=1.5, zorder=10,
                xerr=np.vstack((lzerr, uzerr)),
                yerr=gerr*1.0e12)
    
    return 


def gaikwad_b(ax, puc=False):

    """Plot Gaikwad's Gamma_HI values.

    These are from his Paper II (MNRAS 2017 467 3172) rescaled to our
    cosmology by Gabor.

    """
    
    z = np.array([0.1125, 0.2, 0.3, 0.4])
    zmin = np.array([0.075, 0.15, 0.25, 0.35])
    zmax = np.array([0.15, 0.25, 0.35, 0.45])

    uzerr = zmax - z
    lzerr = z - zmin

    g = np.array([6.7e-14, 9.5e-14, 1.45e-13, 2.0e-13])
    gerr = np.array([5.0e-15, 5.0e-15, 1.5e-14, 2.2e-14])

    if puc: 

        ax.scatter(1+z, g*1.0e12, c='k', edgecolor='None',
                   label='Gaikwad et al.\ 2017b', 
                   s=64, zorder=10, linewidths=1.5) 

        ax.errorbar(1+z, g*1.0e12, ecolor='k', capsize=0,
                    fmt='None', elinewidth=1.5, zorder=10,
                    xerr=np.vstack((lzerr, uzerr)),
                    yerr=gerr*1.0e12, linewidths=1.5)

        return
    
    ax.scatter(z, g*1.0e12, c='k', edgecolor='None',
               label='Gaikwad et al.\ 2017b', marker='p',
               s=80, zorder=10, linewidths=1.5) 

    return


def fumagalli(ax, puc=False):

    z = 0.0
    g_up = 8.0e-2 # 10^-12 s^-1
    g_low = 6.0e-2 

    z = np.array([0.0, 0.0])+0.05
    g = np.array([6.4e-14, 7.6e-14]) # s^-1

    if puc:

        zmin = 0.99
        emin = 0.06 
        zmax = 1.01
        emax = 0.08 
        dz = zmax - zmin
        de = emax - emin
        rect = mpatches.Rectangle((zmin, emin),
                                  dz, de, ec='orange',
                                  color='None', lw=2,
                                  zorder=10,
                                  label='Fumagalli et al.\ 2017')
        ax.add_patch(rect)

    return


def draw_g(lfg, z2=None, g2=None, individuals=None):

    fig = plt.figure(figsize=(7, 7), dpi=100)
    ax = fig.add_subplot(1, 1, 1)

    plt.minorticks_on()
    ax.tick_params('both', which='major', length=7, width=1)
    ax.tick_params('both', which='minor', length=5, width=1)

    ax.set_ylabel(r'$\Gamma_\mathrm{HI}~[10^{-12} \mathrm{s}^{-1}]$')
    ax.set_xlabel('$z$')

    ax.set_yscale('log')
    ax.set_ylim(1.0e-2, 100)
    ax.set_xlim(0.,7)

    locs = (0.01, 0.1, 1, 10, 100)
    labels = ('0.01', '0.1', '1', '10', '100')
    plt.yticks(locs, labels)

    zmax = 9.7
    zmin = 0.0
    dz = 0.1 
    n = (zmax-zmin)/dz+1
    zc = np.linspace(zmax, zmin, num=n)

    nsample = 100
    rsample = lfg.samples[np.random.randint(len(lfg.samples), size=nsample)]
    nzs = len(zc) 
    g = np.zeros((nsample, nzs))

    for i, x in enumerate(rsample):
        print i 
        z, g[i] = j(emissivity, loglf=lfg.log10phi, theta=x, zmax=zmax)

    up = np.percentile(g, 15.87, axis=0)/1.0e-12
    down = np.percentile(g, 84.13, axis=0)/1.0e-12
    ax.fill_between(zc, down, y2=up, color='goldenrod', zorder=1)
        
    theta = np.median(lfg.samples, axis=0)
    z, g = j(emissivity, loglf=lfg.log10phi, theta=theta, zmax=9.7)
    ax.plot(z, g/1.0e-12, c='k', lw=2, label=r'Global model ($M<-18$)')
    
    if z2 is not None:
        ax.plot(z2, g2/1.0e-12, c='k', lw=2,
                label=r'Global model ($M<-18$) local source approximation')
    
    zs_hm12, gs_hm12 = j(em_qso_hm12)
    ax.plot(zs_hm12, gs_hm12/1.0e-12, c='dodgerblue', lw=2,
            label='Haardt \& Madau 2012 QSOs')

    g_hm12_total(ax)

    zs_mh15, gs_mh15 = j(em_qso_mh15)
    ax.plot(zs_mh15, gs_mh15/1.0e-12, c='forestgreen', lw=2,
            label=r'Madau \& Haardt 2015')
    
    bb13(ax)
    calverley(ax)

    zg, eg, zg_lerr, zg_uerr, eg_lerr, eg_uerr = np.loadtxt('Data_new/giallongo15_emissivity.txt', unpack=True)
    
    eg *= 1.0e24
    eg_lerr *= 1.0e24
    eg_uerr *= 1.0e24

    gg = Gamma_HI(eg, zg)
    gg_lerr = gg - Gamma_HI(eg-eg_lerr, zg)
    gg_uerr = gg - Gamma_HI(eg-eg_uerr, zg)

    gg *= 1.0e12
    gg_lerr *= 1.0e12
    gg_uerr *= 1.0e12
    
    ax.scatter(zg, gg, c='grey', edgecolor='None',
               label=r'Giallongo et al.\ 2015 ($M<-18$)',
               s=72, zorder=4)

    ax.errorbar(zg, gg, ecolor='grey', capsize=0, fmt='None', elinewidth=2,
                xerr=np.vstack((zg_lerr, zg_uerr)),
                yerr=np.vstack((gg_lerr, gg_uerr)), 
                zorder=3, mew=1)

    shull(ax)

    khaire(ax)

    bolton(ax)

    onorbe(ax)

    fumagalli(ax)

    if individuals is not None:
        lfis(individuals, ax)

    kollmeier(ax)

    viel(ax)

    gaikwad_b(ax)

    handles, labels = ax.get_legend_handles_labels()

    plt.legend(handles[::-1], labels[::-1], loc='upper right',
               fontsize=10, handlelength=3, frameon=False,
               framealpha=0.0, labelspacing=.1,
               handletextpad=0.1, borderpad=0.3, scatterpoints=1, ncol=2)

    
    plt.savefig('g.pdf'.format(z),bbox_inches='tight')
    plt.close('all')

    return

def draw_g_paper():

    fig = plt.figure(figsize=(11.33, 7), dpi=100)
    ax = fig.add_subplot(1, 1, 1)

    plt.minorticks_on()
    ax.tick_params('both', which='major', length=7, width=1)
    ax.tick_params('both', which='minor', length=5, width=1)

    ax.set_ylabel(r'$\Gamma_\mathrm{HI}~[10^{-12} \mathrm{s}^{-1}]$')
    ax.set_xlabel('$z$')

    ax.set_yscale('log')
    ax.set_ylim(1.0e-2, 2)
    ax.set_xlim(0.,7)

    locs = (0.01, 0.1, 1)
    labels = ('0.01', '0.1', '1')
    plt.yticks(locs, labels)

    zmax = 9.7
    zmin = 0.0
    dz = 0.1 
    n = (zmax-zmin)/dz+1
    zc = np.linspace(zmax, zmin, num=n)

    z, g = j(emissivity_18, zmax=15.0)
    z, g_down = j(emissivity_18_down, zmax=15.0)
    z, g_up = j(emissivity_18_up, zmax=15.0)
    g18f = ax.fill_between(z, g_up/1.0e-12, y2=g_down/1.0e-12,
                           color='red', zorder=5,
                           alpha=0.6, edgecolor='None')
    g18, = plt.plot(z, g/1.0e-12, lw=2, c='red', zorder=5)

    gamma18 = g
    gamma18_up = g_up
    gamma18_low = g_down 

    z, g = j(emissivity_21, zmax=15.0)
    z, g_down = j(emissivity_21_down, zmax=15.0)
    z, g_up = j(emissivity_21_up, zmax=15.0)
    g21f = ax.fill_between(z, g_up/1.0e-12, y2=g_down/1.0e-12,
                           color='blue', zorder=5,
                           alpha=0.6, edgecolor='None')
    g21, = plt.plot(z, g/1.0e-12, lw=2, c='blue', zorder=5)
    
    gamma21 = g
    gamma21_up = g_up
    gamma21_low = g_down 

    tabulate = True
    if tabulate:
        e1 = np.interp(z, z18, median18)
        e1450_18 = e1

        e1 = np.interp(z, z18, down18)
        e1450_18_low = e1

        e1 = np.interp(z, z18, up18)
        e1450_18_up = e1
        
        e1 = np.interp(z, z21, median21)
        e1450_21 = e1

        e1 = np.interp(z, z21, down21)
        e1450_21_low = e1

        e1 = np.interp(z, z21, up21)
        e1450_21_up = e1

        e912_18 = e1450_18*((912.0/1450.0)**0.61)
        e912_18_up = e1450_18_up*((912.0/1450.0)**0.61) 
        e912_18_low = e1450_18_low*((912.0/1450.0)**0.61)

        e912_21 = e1450_21*((912.0/1450.0)**0.61)
        e912_21_up = e1450_21_up*((912.0/1450.0)**0.61)
        e912_21_low = e1450_21_low*((912.0/1450.0)**0.61)

        with open('rtg2_data.txt', 'w') as f: 
            fs = r'{:.1f}  ' + '{:.3e}  {:.3e}  {:.3e}  '*6 + '\n'
            for i in range(len(gamma21)):
                f.write(fs.format(z[i],
                                e1450_18[i], e1450_18_up[i], e1450_18_low[i],
                                e1450_21[i], e1450_21_up[i], e1450_21_low[i],
                                e912_18[i], e912_18_up[i], e912_18_low[i],
                                e912_21[i], e912_21_up[i], e912_21_low[i],                            
                                gamma18[i], gamma18_up[i], gamma18_low[i],
                                gamma21[i], gamma21_up[i], gamma21_low[i]))
    
    zs_hm12, gs_hm12 = j(em_qso_hm12)
    ax.plot(zs_hm12, gs_hm12/1.0e-12, c='maroon', lw=2, zorder=4,
            label='Haardt \& Madau 2012 QSOs', dashes=[7,2])

    g_hm12_total(ax)

    zs_mh15, gs_mh15 = j(em_qso_mh15)
    ax.plot(zs_mh15, gs_mh15/1.0e-12, c='darkgreen', lw=1,
            label=r'Madau \& Haardt 2015', zorder=2, dashes=[1,1])
    
    bb13(ax)
    calverley(ax)
    wyithe11(ax)
    daloisio18(ax)
    davies17(ax)

    zg, eg, zg_lerr, zg_uerr, eg_lerr, eg_uerr = np.loadtxt('Data_new/giallongo15_emissivity.txt', unpack=True)
    
    eg *= 1.0e24
    eg_lerr *= 1.0e24
    eg_uerr *= 1.0e24

    gg = Gamma_HI(eg, zg)
    gg_lerr = gg - Gamma_HI(eg-eg_lerr, zg)
    gg_uerr = gg - Gamma_HI(eg-eg_uerr, zg)

    gg *= 1.0e12
    gg_lerr *= 1.0e12
    gg_uerr *= 1.0e12

    ax.errorbar(zg, gg, ecolor='tomato', capsize=0, fmt='None', elinewidth=1.5, 
                xerr=np.vstack((zg_lerr, zg_uerr)),
                #yerr=np.vstack((gg_lerr, gg_uerr)), 
                zorder=11, mew=1)

    ax.scatter(zg, gg, c='#ffffff', edgecolor='tomato',
               label='Giallongo et al.\ 2015 ($M_\mathrm{1450}<-18$)',
               s=55, zorder=11, linewidths=1.5)
    
    khaire(ax)

    onorbe(ax)

    puchwein(ax)

    gaikwad_b(ax)

    handles, labels = ax.get_legend_handles_labels()
    handles.append((g18f,g18))
    labels.append('Kulkarni et al.\ 2018 (this work; $M_\mathrm{1450}<-18$)')
    handles.append((g21f,g21))
    labels.append('Kulkarni et al.\ 2018 (this work; $M_\mathrm{1450}<-21$)')

    plt.legend(handles, labels, loc='center',
               fontsize=12, handlelength=3, frameon=False,
               framealpha=0.0, labelspacing=.1,
               handletextpad=0.1, borderpad=0.3,
               scatterpoints=1, bbox_to_anchor=[0.33,0.27])

    plt.savefig('g.pdf',bbox_inches='tight')
    plt.close('all')

    return

import matplotlib.ticker as ticker

def draw_g_puc():

    fig = plt.figure(figsize=(7, 7), dpi=100)
    ax = fig.add_subplot(1, 1, 1)

    plt.minorticks_on()
    ax.tick_params('both', which='major', length=7, width=1)
    ax.tick_params('both', which='minor', length=5, width=1)
    ax.tick_params('x', which='major', pad=6)

    ax.set_ylabel(r'$\Gamma_\mathrm{HI}~[10^{-12} \mathrm{s}^{-1}]$')
    ax.set_xlabel('$z$')

    ax.set_yscale('log')
    ax.set_xscale('log')
    ax.set_ylim(1.0e-2, 10)
    ax.set_xlim(0.9,4)

    locs = (0.01, 0.1, 1, 10)
    labels = ('0.01', '0.1', '1', '10')
    plt.yticks(locs, labels)

    locs = (1,2,3,4)
    labels = ('0', '1', '2', '3')
    plt.xticks(locs, labels)

    ax.xaxis.set_minor_locator(ticker.MultipleLocator(0.1))
    
    zmax = 9.7
    zmin = 0.0
    dz = 0.1 
    n = (zmax-zmin)/dz+1
    zc = np.linspace(zmax, zmin, num=n)

    z, g = j(emissivity_18, zmax=15.0)
    z, g_down = j(emissivity_18_down, zmax=15.0)
    z, g_up = j(emissivity_18_up, zmax=15.0)
    g18f = ax.fill_between(1+z, g_up/1.0e-12, y2=g_down/1.0e-12,
                           color='red', zorder=5,
                           alpha=0.6, edgecolor='None')
    g18, = plt.plot(1+z, g/1.0e-12, lw=2, c='red', zorder=5)

    z, g = j(emissivity_21, zmax=15.0)
    z, g_down = j(emissivity_21_down, zmax=15.0)
    z, g_up = j(emissivity_21_up, zmax=15.0)
    g21f = ax.fill_between(1+z, g_up/1.0e-12, y2=g_down/1.0e-12,
                           color='blue', zorder=5,
                           alpha=0.6, edgecolor='None')
    g21, = plt.plot(1+z, g/1.0e-12, lw=2, c='blue', zorder=5)
    
    zs_hm12, gs_hm12 = j(em_qso_hm12)
    ax.plot(1+zs_hm12, gs_hm12/1.0e-12, c='grey', lw=2, zorder=2,
            label='Haardt \& Madau 2012 QSOs', dashes=[7,2])

    g_hm12_total(ax, puc=True)

    zs_mh15, gs_mh15 = j(em_qso_mh15)
    ax.plot(1+zs_mh15, gs_mh15/1.0e-12, c='forestgreen', lw=1, zorder=2, dashes=[1,1],
            label=r'Madau \& Haardt 2015')
    
    bb13(ax, puc=True)

    shull(ax, puc=True)

    #khaire(ax, puc=True)

    onorbe(ax, puc=True)

    puchwein(ax, puc=True)

    fumagalli(ax, puc=True)

    kollmeier(ax, puc=True)

    viel(ax, puc=True)

    gaikwad_b(ax, puc=True)

    bolton17(ax)

    handles, labels = ax.get_legend_handles_labels()

    l1 = plt.legend(handles, labels, loc='upper right',
               fontsize=11, handlelength=3, frameon=False,
               framealpha=0.0, labelspacing=.1,
               handletextpad=0.1, borderpad=0.3, scatterpoints=1, ncol=2)

    handles, labels = [], [] 
    handles.append((g18f,g18))
    labels.append('Kulkarni et al.\ 2018 (this work; $M_\mathrm{1450}<-18$)')
    handles.append((g21f,g21))
    labels.append('Kulkarni et al.\ 2018 (this work; $M_\mathrm{1450}<-21$)')

    l2 = plt.legend(handles, labels, loc='lower right',
                    fontsize=11, handlelength=3, frameon=False,
                    framealpha=0.0, labelspacing=.1,
                    handletextpad=0.1, borderpad=0.3, scatterpoints=1, ncol=1)

    ax.add_artist(l1)
    
    plt.savefig('g_puc.pdf',bbox_inches='tight')
    plt.close('all')

    return 


def draw_g_puc_talk():

    fig = plt.figure(figsize=(7, 7), dpi=100)
    ax = fig.add_subplot(1, 1, 1)

    plt.minorticks_on()
    ax.tick_params('both', which='major', length=7, width=1)
    ax.tick_params('both', which='minor', length=5, width=1)
    ax.tick_params('x', which='major', pad=6)

    ax.set_ylabel(r'$\Gamma_\mathrm{HI}~[10^{-12} \mathrm{s}^{-1}]$')
    ax.set_xlabel('$z$')

    ax.set_yscale('log')
    ax.set_xscale('log')
    ax.set_ylim(1.0e-2, 10)
    ax.set_xlim(0.9,4)

    locs = (0.01, 0.1, 1, 10)
    labels = ('0.01', '0.1', '1', '10')
    plt.yticks(locs, labels)

    locs = (1,2,3,4)
    labels = ('0', '1', '2', '3')
    plt.xticks(locs, labels)

    ax.xaxis.set_minor_locator(ticker.MultipleLocator(0.1))
    
    zmax = 9.7
    zmin = 0.0
    dz = 0.1 
    n = (zmax-zmin)/dz+1
    zc = np.linspace(zmax, zmin, num=n)

    z, g = j(emissivity_18, zmax=15.0)
    z, g_down = j(emissivity_18_down, zmax=15.0)
    z, g_up = j(emissivity_18_up, zmax=15.0)
    g18f = ax.fill_between(1+z, g_up/1.0e-12, y2=g_down/1.0e-12,
                           color='red', zorder=5,
                           alpha=0.6, edgecolor='None')
    g18, = plt.plot(1+z, g/1.0e-12, lw=2, c='red', zorder=5)

    z, g = j(emissivity_21, zmax=15.0)
    z, g_down = j(emissivity_21_down, zmax=15.0)
    z, g_up = j(emissivity_21_up, zmax=15.0)
    g21f = ax.fill_between(1+z, g_up/1.0e-12, y2=g_down/1.0e-12,
                           color='blue', zorder=5,
                           alpha=0.6, edgecolor='None')
    g21, = plt.plot(1+z, g/1.0e-12, lw=2, c='blue', zorder=5)
    
    zs_hm12, gs_hm12 = j(em_qso_hm12)
    ax.plot(1+zs_hm12, gs_hm12/1.0e-12, c='grey', lw=2, zorder=2,
            label='Haardt \& Madau 2012 QSOs', dashes=[7,2])

    g_hm12_total(ax, puc=True)

    zs_mh15, gs_mh15 = j(em_qso_mh15)
    ax.plot(1+zs_mh15, gs_mh15/1.0e-12, c='forestgreen', lw=1, zorder=2, dashes=[1,1],
            label=r'Madau \& Haardt 2015')
    
    bb13(ax, puc=True)

    shull(ax, puc=True)

    khaire(ax, puc=True)

    onorbe(ax, puc=True)

    puchwein(ax, puc=True)

    fumagalli(ax, puc=True)

    kollmeier(ax, puc=True)

    viel(ax, puc=True)

    gaikwad_b(ax, puc=True)

    #bolton17(ax)

    handles, labels = ax.get_legend_handles_labels()

    l1 = plt.legend(handles, labels, loc='upper right',
               fontsize=11, handlelength=3, frameon=False,
               framealpha=0.0, labelspacing=.1,
               handletextpad=0.1, borderpad=0.3, scatterpoints=1, ncol=2)

    handles, labels = [], [] 
    handles.append((g18f,g18))
    labels.append('Kulkarni et al.\ 2018 (this work; $M_\mathrm{1450}<-18$)')
    handles.append((g21f,g21))
    labels.append('Kulkarni et al.\ 2018 (this work; $M_\mathrm{1450}<-21$)')

    l2 = plt.legend(handles, labels, loc='lower right',
                    fontsize=11, handlelength=3, frameon=False,
                    framealpha=0.0, labelspacing=.1,
                    handletextpad=0.1, borderpad=0.3, scatterpoints=1, ncol=1)

    ax.add_artist(l1)
    
    plt.savefig('g_puc_talk.pdf',bbox_inches='tight')
    plt.close('all')

    return 

draw_g_puc_talk()

def draw_g_talk():

    fig = plt.figure(figsize=(11.33, 7), dpi=100)
    ax = fig.add_subplot(1, 1, 1)

    plt.minorticks_on()
    ax.tick_params('both', which='major', length=7, width=1)
    ax.tick_params('both', which='minor', length=5, width=1)

    ax.set_ylabel(r'$\Gamma_\mathrm{HI}~[10^{-12} \mathrm{s}^{-1}]$')
    ax.set_xlabel('$z$')

    ax.set_yscale('log')
    ax.set_ylim(1.0e-2, 2)
    ax.set_xlim(0.,7)

    locs = (0.01, 0.1, 1)
    labels = ('0.01', '0.1', '1')
    plt.yticks(locs, labels)

    zmax = 9.7
    zmin = 0.0
    dz = 0.1 
    n = (zmax-zmin)/dz+1
    zc = np.linspace(zmax, zmin, num=n)

    z, g = j(emissivity_18, zmax=15.0)
    z, g_down = j(emissivity_18_down, zmax=15.0)
    z, g_up = j(emissivity_18_up, zmax=15.0)
    g18f = ax.fill_between(z, g_up/1.0e-12, y2=g_down/1.0e-12,
                           color='red', zorder=5,
                           alpha=0.6, edgecolor='None')
    g18, = plt.plot(z, g/1.0e-12, lw=2, c='red', zorder=5)

    gamma18 = g
    gamma18_up = g_up
    gamma18_low = g_down 

    z, g = j(emissivity_21, zmax=15.0)
    z, g_down = j(emissivity_21_down, zmax=15.0)
    z, g_up = j(emissivity_21_up, zmax=15.0)
    g21f = ax.fill_between(z, g_up/1.0e-12, y2=g_down/1.0e-12,
                           color='blue', zorder=5,
                           alpha=0.6, edgecolor='None')
    g21, = plt.plot(z, g/1.0e-12, lw=2, c='blue', zorder=5)
    
    gamma21 = g
    gamma21_up = g_up
    gamma21_low = g_down 

    zs_hm12, gs_hm12 = j(em_qso_hm12)
    ax.plot(zs_hm12, gs_hm12/1.0e-12, c='maroon', lw=2, zorder=4,
            label='Haardt \& Madau 2012 QSOs', dashes=[7,2])

    g_hm12_total(ax)

    zs_mh15, gs_mh15 = j(em_qso_mh15)
    ax.plot(zs_mh15, gs_mh15/1.0e-12, c='darkgreen', lw=1,
            label=r'Madau \& Haardt 2015', zorder=2, dashes=[1,1])
    
    bb13(ax)
    calverley(ax)
    wyithe11(ax)
    daloisio18(ax)
    davies17(ax)

    zg, eg, zg_lerr, zg_uerr, eg_lerr, eg_uerr = np.loadtxt('Data_new/giallongo15_emissivity.txt', unpack=True)
    
    eg *= 1.0e24
    eg_lerr *= 1.0e24
    eg_uerr *= 1.0e24

    gg = Gamma_HI(eg, zg)
    gg_lerr = gg - Gamma_HI(eg-eg_lerr, zg)
    gg_uerr = gg - Gamma_HI(eg-eg_uerr, zg)

    gg *= 1.0e12
    gg_lerr *= 1.0e12
    gg_uerr *= 1.0e12

    ax.errorbar(zg, gg, ecolor='tomato', capsize=0, fmt='None', elinewidth=1.5, 
                xerr=np.vstack((zg_lerr, zg_uerr)),
                #yerr=np.vstack((gg_lerr, gg_uerr)), 
                zorder=11, mew=1)

    ax.scatter(zg, gg, c='#ffffff', edgecolor='tomato',
               label='Giallongo et al.\ 2015 ($M_\mathrm{1450}<-18$)',
               s=55, zorder=11, linewidths=1.5)
    
    khaire(ax)

    onorbe(ax)

    puchwein(ax)

    gaikwad_b(ax)

    handles, labels = ax.get_legend_handles_labels()
    handles.append((g18f,g18))
    labels.append('Kulkarni et al.\ 2018 ($M_\mathrm{1450}<-18$)')
    handles.append((g21f,g21))
    labels.append('Kulkarni et al.\ 2018 ($M_\mathrm{1450}<-21$)')

    plt.legend(handles, labels, loc='center',
               fontsize=12, handlelength=3, frameon=False,
               framealpha=0.0, labelspacing=.1,
               handletextpad=0.1, borderpad=0.3,
               scatterpoints=1, bbox_to_anchor=[0.33,0.27])

    plt.savefig('g_talk.pdf',bbox_inches='tight')
    plt.close('all')

    return


draw_g_paper()