diff --git a/tests/integration/comparison/twostream/comparison_petitRADTRANS_CIA.py b/tests/integration/comparison/twostream/comparison_petitRADTRANS_CIA.py
new file mode 100644
index 000000000..3b2c465e0
--- /dev/null
+++ b/tests/integration/comparison/twostream/comparison_petitRADTRANS_CIA.py
@@ -0,0 +1,218 @@
+from jax import config 
+config.update("jax_enable_x64", True)
+
+import numpy as np
+import jax.numpy as jnp
+import math
+import os
+
+from exojax.atm.atmprof import pressure_layer_logspace
+from exojax.utils.grids import wavenumber_grid
+from exojax.spec.multimol import MultiMol
+from exojax.spec import contdb
+from exojax.spec.layeropacity import layer_optical_depth, layer_optical_depth_CIA
+from exojax.spec import planck
+from exojax.spec.rtransfer import rtrun_emis_pure_absorption
+
+from petitRADTRANS import Radtrans
+from exojax.spec import molinfo
+import petitRADTRANS.nat_cst as nc
+
+from exojax.utils.instfunc import resolution_to_gaussian_std
+from exojax.utils.grids import velocity_grid
+from exojax.spec import response
+
+
+
+def run_exojax(path_data, ld_min, ld_max, mols, db, T0, alpha, logg, logvmr):
+    Parr, dParr, k = pressure_layer_logspace(log_pressure_top=-3., nlayer=200)
+    ONEARR = np.ones_like(Parr)
+
+
+    R = 900000.
+    ld_min = ld_min - 5.
+    ld_max = ld_max + 5.
+    nu_min = 1.0e8 / ld_max
+    nu_max = 1.0e8 / ld_min
+    Nx = math.ceil(R * np.log(nu_max / nu_min)) + 1 # ueki
+    Nx = math.ceil(Nx / 2.) * 2 # make even
+
+    nus, wav, res = wavenumber_grid(ld_min, ld_max, Nx, unit="AA", xsmode="premodit")
+    nus = [nus]
+    wav = [wav]
+    res = [res]
+
+
+    mul = MultiMol(molmulti=[mols], dbmulti=[db], database_root_path=path_data)
+    multimdb = mul.multimdb(nus, crit=1.e-30, Ttyp=1000.)
+    multiopa = mul.multiopa_premodit(multimdb, nus, auto_trange=[500.,1500.], dit_grid_resolution=1.0)
+
+    cdbH2H2 = []
+    cdbH2He = []
+    for k in range(len(nus)):
+        cdbH2H2.append(contdb.CdbCIA(os.path.join(path_data, 'H2-H2_2011.cia'), nus[k]))
+        cdbH2He.append(contdb.CdbCIA(os.path.join(path_data, 'H2-He_2011.cia'), nus[k]))
+
+    molmass_list, molmassH2, molmassHe = mul.molmass()
+
+
+    def frun(T0, alpha, logg, logvmr):
+        Tarr = T0 * (Parr)**alpha
+        Tarr = np.clip(Tarr, 500, None)
+
+        g = 10.**logg # cgs
+
+        vmr = jnp.power(10., jnp.array(logvmr))
+        vmrH2 = (1. - jnp.sum(vmr)) * 6./7.
+        vmrHe = (1. - jnp.sum(vmr)) * 1./7.
+        mmw = jnp.sum(vmr*jnp.array(molmass_list)) + vmrH2*molmassH2 + vmrHe*molmassHe
+        mmr = jnp.multiply(vmr, jnp.array(molmass_list)) / mmw
+
+        mu = []
+        for k in range(len(nus)):
+            dtaum = []
+            for i in range(len(mul.masked_molmulti[k])):
+                xsm = multiopa[k][i].xsmatrix(Tarr, Parr)
+                xsm = jnp.abs(xsm)
+                dtaum.append(layer_optical_depth(dParr, xsm, mmr[mul.mols_num[k][i]]*ONEARR, molmass_list[mul.mols_num[k][i]], g))
+
+            dtau = sum(dtaum)
+
+            if(len(cdbH2H2[k].nucia) > 0):
+                dtaucH2H2 = layer_optical_depth_CIA(nus[k], Tarr, Parr, dParr, vmrH2, vmrH2, mmw, g, cdbH2H2[k].nucia, cdbH2H2[k].tcia, cdbH2H2[k].logac)
+                dtau = dtau + dtaucH2H2
+            if(len(cdbH2He[k].nucia) > 0):
+                dtaucH2He = layer_optical_depth_CIA(nus[k], Tarr, Parr, dParr, vmrH2, vmrHe, mmw, g, cdbH2He[k].nucia, cdbH2He[k].tcia, cdbH2He[k].logac)
+                dtau = dtau + dtaucH2He
+
+            sourcef = planck.piBarr(Tarr, nus[k])
+            F0 = rtrun_emis_pure_absorption(dtau, sourcef)
+
+            mu.append(F0)
+
+        return mu
+
+    return nus[0], frun(T0, alpha, logg, logvmr)[0]
+
+
+
+def run_petit(ld_min, ld_max, mols, mols_exojax, T0, alpha, logg, logvmr):
+    atmosphere = Radtrans(line_species = mols,
+                          rayleigh_species = ['H2', 'He'],
+                          continuum_opacities = ['H2-H2', 'H2-He'],
+                          wlen_bords_micron = [ld_min*1e-4, ld_max*1e-4],
+                          mode = 'lbl')
+
+    pressures = np.logspace(-10, 2, 130)
+    atmosphere.setup_opa_structure(pressures)
+    temperature = T0*(pressures)**alpha
+    temperature = np.clip(temperature, 500, None)
+
+
+    molmass_list = []
+    for i in range(len(mols_exojax)):
+        molmass_list.append(molinfo.molmass(mols_exojax[i]))
+    molmassH2=molinfo.molmass("H2")
+    molmassHe=molinfo.molmass("He", db_HIT=False)
+
+    vmr = jnp.power(10., jnp.array(logvmr))
+    vmrH2 = (1. - jnp.sum(vmr)) * 6./7.
+    vmrHe = (1. - jnp.sum(vmr)) * 1./7.
+    mmw = jnp.sum(vmr*jnp.array(molmass_list)) + vmrH2*molmassH2 + vmrHe*molmassHe
+    mmr = jnp.multiply(vmr, jnp.array(molmass_list)) / mmw
+    mmrH2 = vmrH2 * molmassH2 / mmw
+    mmrHe = vmrHe * molmassHe / mmw
+
+    mass_fractions = {}
+    mass_fractions['H2'] = mmrH2 * np.ones_like(temperature)
+    mass_fractions['He'] = mmrHe * np.ones_like(temperature)
+    for i in range(len(mols)):
+        mass_fractions[mols[i]] = mmr[i] * np.ones_like(temperature)
+
+
+    MMW = mmw * np.ones_like(temperature)
+    gravity = 1e1**(logg)
+
+    atmosphere.calc_flux(temperature, mass_fractions, gravity, MMW)
+
+    ld = nc.c/atmosphere.freq/1e-4 # [um]
+    nus = 1.0e4 / ld # [cm^{-1}]
+    f = atmosphere.flux #[erg cm^{-2} s^{-1} Hz^{-1}]
+    f = f * nc.c #[erg/s/cm^2/cm^{-1}]
+
+    nus = nus[::-1]
+    f = f[::-1]
+
+    return nus, f
+
+
+
+path_data = "/home/kawashima/database"
+ld_min = 23000.
+ld_max = 24000.
+mols_exojax = ['CO']
+db_exojax = ['ExoMol']
+mols_petit = ['CO_all_iso']
+
+T0 = 995.56
+alpha = 0.09
+logg = 5.01
+logvmr = [-3.06]
+
+nus1, f1 = run_exojax(path_data, ld_min, ld_max, mols_exojax, db_exojax, T0, alpha, logg, logvmr)
+nus2, f2 = run_petit(ld_min, ld_max, mols_petit, mols_exojax, T0, alpha, logg, logvmr)
+
+
+
+Rinst = 70000.
+nu_min = 1.0e8 / ld_max
+nu_max = 1.0e8 / ld_min
+Nx = math.ceil(Rinst * np.log(nu_max / nu_min)) + 1 # ueki
+Nx = math.ceil(Nx / 2.) * 2 # make even
+nusd, wav, res = wavenumber_grid(ld_min, ld_max, Nx, unit="AA", xsmode="premodit")
+
+beta_inst = resolution_to_gaussian_std(Rinst)
+res_calc = 900000.
+vsini_max = 100.0
+vr_array = velocity_grid(res_calc, vsini_max)
+
+f1_obs = response.ipgauss_sampling(nusd, nus1, f1, beta_inst, 0., vr_array)
+if len(nus2) % 2 == 1:
+    nus2 = nus2[1:]
+    f2 = f2[1:]
+f2_obs = response.ipgauss_sampling(nusd, nus2, f2, beta_inst, 0., vr_array)
+
+
+
+import matplotlib.pyplot as plt
+from matplotlib.ticker import (MultipleLocator, FormatStrFormatter, AutoMinorLocator)
+fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(40,8), gridspec_kw={'height_ratios': [3, 1]})
+plt.subplots_adjust(hspace=0)
+ax1.plot(1.0e8/nusd, f1_obs, alpha=0.5, label="ExoJAX")
+ax1.plot(1.0e8/nusd, f2_obs, alpha=0.5, label="petitRADTRANS")
+
+ax2.plot(1.0e8/nusd, f1_obs - f2_obs, "+", color="black", markersize=5, alpha=0.2)
+
+ax1.set_ylabel("Flux [erg/s/cm$\mathrm{^2}$/cm$\mathrm{^{-1}}$]", fontsize=15)
+ax2.set_xlabel("Wavelength [$\AA$]", fontsize=15)
+ax2.set_ylabel("Residual", fontsize=15)
+
+ax1.set_xlim(np.min(1.0e8/nusd), np.max(1.0e8/nusd))
+ax2.set_xlim(np.min(1.0e8/nusd), np.max(1.0e8/nusd))
+
+ax1.xaxis.set_ticks_position('both')
+ax1.yaxis.set_ticks_position('both')
+ax2.xaxis.set_ticks_position('both')
+ax2.yaxis.set_ticks_position('both')
+
+ax1.xaxis.set_minor_locator(AutoMinorLocator())
+ax1.yaxis.set_minor_locator(AutoMinorLocator())
+ax2.xaxis.set_minor_locator(AutoMinorLocator())
+ax2.yaxis.set_minor_locator(AutoMinorLocator())
+
+ax2.patch.set_alpha(0)
+ax1.tick_params(labelbottom=False, labeltop=True)
+
+ax1.legend()
+#plt.show()
+plt.savefig("output/CIA_R70000.png", bbox_inches='tight')
diff --git a/tests/integration/comparison/twostream/comparison_petitRADTRANS_narrow_R70000.py b/tests/integration/comparison/twostream/comparison_petitRADTRANS_narrow_R70000.py
new file mode 100644
index 000000000..9a919d26f
--- /dev/null
+++ b/tests/integration/comparison/twostream/comparison_petitRADTRANS_narrow_R70000.py
@@ -0,0 +1,218 @@
+from jax import config 
+config.update("jax_enable_x64", True)
+
+import numpy as np
+import jax.numpy as jnp
+import math
+import os
+
+from exojax.atm.atmprof import pressure_layer_logspace
+from exojax.utils.grids import wavenumber_grid
+from exojax.spec.multimol import MultiMol
+from exojax.spec import contdb
+from exojax.spec.layeropacity import layer_optical_depth, layer_optical_depth_CIA
+from exojax.spec import planck
+from exojax.spec.rtransfer import rtrun_emis_pure_absorption
+
+from petitRADTRANS import Radtrans
+from exojax.spec import molinfo
+import petitRADTRANS.nat_cst as nc
+
+from exojax.utils.instfunc import resolution_to_gaussian_std
+from exojax.utils.grids import velocity_grid
+from exojax.spec import response
+
+
+
+def run_exojax(path_data, ld_min, ld_max, mols, db, T0, alpha, logg, logvmr):
+    Parr, dParr, k = pressure_layer_logspace(log_pressure_top=-3., nlayer=200)
+    ONEARR = np.ones_like(Parr)
+
+
+    R = 900000.
+    ld_min = ld_min - 5.
+    ld_max = ld_max + 5.
+    nu_min = 1.0e8 / ld_max
+    nu_max = 1.0e8 / ld_min
+    Nx = math.ceil(R * np.log(nu_max / nu_min)) + 1 # ueki
+    Nx = math.ceil(Nx / 2.) * 2 # make even
+
+    nus, wav, res = wavenumber_grid(ld_min, ld_max, Nx, unit="AA", xsmode="premodit")
+    nus = [nus]
+    wav = [wav]
+    res = [res]
+
+
+    mul = MultiMol(molmulti=[mols], dbmulti=[db], database_root_path=path_data)
+    multimdb = mul.multimdb(nus, crit=1.e-30, Ttyp=1000.)
+    multiopa = mul.multiopa_premodit(multimdb, nus, auto_trange=[500.,1500.], dit_grid_resolution=1.0)
+
+    cdbH2H2 = []
+    cdbH2He = []
+    for k in range(len(nus)):
+        cdbH2H2.append(contdb.CdbCIA(os.path.join(path_data, 'H2-H2_2011.cia'), nus[k]))
+        cdbH2He.append(contdb.CdbCIA(os.path.join(path_data, 'H2-He_2011.cia'), nus[k]))
+
+    molmass_list, molmassH2, molmassHe = mul.molmass()
+
+
+    def frun(T0, alpha, logg, logvmr):
+        Tarr = T0 * (Parr)**alpha
+        Tarr = np.clip(Tarr, 500, None)
+
+        g = 10.**logg # cgs
+
+        vmr = jnp.power(10., jnp.array(logvmr))
+        vmrH2 = (1. - jnp.sum(vmr)) * 6./7.
+        vmrHe = (1. - jnp.sum(vmr)) * 1./7.
+        mmw = jnp.sum(vmr*jnp.array(molmass_list)) + vmrH2*molmassH2 + vmrHe*molmassHe
+        mmr = jnp.multiply(vmr, jnp.array(molmass_list)) / mmw
+
+        mu = []
+        for k in range(len(nus)):
+            dtaum = []
+            for i in range(len(mul.masked_molmulti[k])):
+                xsm = multiopa[k][i].xsmatrix(Tarr, Parr)
+                xsm = jnp.abs(xsm)
+                dtaum.append(layer_optical_depth(dParr, xsm, mmr[mul.mols_num[k][i]]*ONEARR, molmass_list[mul.mols_num[k][i]], g))
+
+            dtau = sum(dtaum)
+
+            if(len(cdbH2H2[k].nucia) > 0):
+                dtaucH2H2 = layer_optical_depth_CIA(nus[k], Tarr, Parr, dParr, vmrH2, vmrH2, mmw, g, cdbH2H2[k].nucia, cdbH2H2[k].tcia, cdbH2H2[k].logac)
+                dtau = dtau + dtaucH2H2
+            if(len(cdbH2He[k].nucia) > 0):
+                dtaucH2He = layer_optical_depth_CIA(nus[k], Tarr, Parr, dParr, vmrH2, vmrHe, mmw, g, cdbH2He[k].nucia, cdbH2He[k].tcia, cdbH2He[k].logac)
+                dtau = dtau + dtaucH2He
+
+            sourcef = planck.piBarr(Tarr, nus[k])
+            F0 = rtrun_emis_pure_absorption(dtau, sourcef)
+
+            mu.append(F0)
+
+        return mu
+
+    return nus[0], frun(T0, alpha, logg, logvmr)[0]
+
+
+
+def run_petit(ld_min, ld_max, mols, mols_exojax, T0, alpha, logg, logvmr):
+    atmosphere = Radtrans(line_species = mols,
+                          rayleigh_species = ['H2', 'He'],
+                          continuum_opacities = ['H2-H2', 'H2-He'],
+                          wlen_bords_micron = [ld_min*1e-4, ld_max*1e-4],
+                          mode = 'lbl')
+
+    pressures = np.logspace(-10, 2, 130)
+    atmosphere.setup_opa_structure(pressures)
+    temperature = T0*(pressures)**alpha
+    temperature = np.clip(temperature, 500, None)
+
+
+    molmass_list = []
+    for i in range(len(mols_exojax)):
+        molmass_list.append(molinfo.molmass(mols_exojax[i]))
+    molmassH2=molinfo.molmass("H2")
+    molmassHe=molinfo.molmass("He", db_HIT=False)
+
+    vmr = jnp.power(10., jnp.array(logvmr))
+    vmrH2 = (1. - jnp.sum(vmr)) * 6./7.
+    vmrHe = (1. - jnp.sum(vmr)) * 1./7.
+    mmw = jnp.sum(vmr*jnp.array(molmass_list)) + vmrH2*molmassH2 + vmrHe*molmassHe
+    mmr = jnp.multiply(vmr, jnp.array(molmass_list)) / mmw
+    mmrH2 = vmrH2 * molmassH2 / mmw
+    mmrHe = vmrHe * molmassHe / mmw
+
+    mass_fractions = {}
+    mass_fractions['H2'] = mmrH2 * np.ones_like(temperature)
+    mass_fractions['He'] = mmrHe * np.ones_like(temperature)
+    for i in range(len(mols)):
+        mass_fractions[mols[i]] = mmr[i] * np.ones_like(temperature)
+
+
+    MMW = mmw * np.ones_like(temperature)
+    gravity = 1e1**(logg)
+
+    atmosphere.calc_flux(temperature, mass_fractions, gravity, MMW)
+
+    ld = nc.c/atmosphere.freq/1e-4 # [um]
+    nus = 1.0e4 / ld # [cm^{-1}]
+    f = atmosphere.flux #[erg cm^{-2} s^{-1} Hz^{-1}]
+    f = f * nc.c #[erg/s/cm^2/cm^{-1}]
+
+    nus = nus[::-1]
+    f = f[::-1]
+
+    return nus, f
+
+
+
+path_data = "/home/kawashima/database"
+ld_min = 14600.
+ld_max = 14700.
+mols_exojax = ['H2O']
+db_exojax = ['ExoMol']
+mols_petit = ['H2O_main_iso']
+
+T0 = 995.56
+alpha = 0.09
+logg = 5.01
+logvmr = [-2.93]
+
+nus1, f1 = run_exojax(path_data, ld_min, ld_max, mols_exojax, db_exojax, T0, alpha, logg, logvmr)
+nus2, f2 = run_petit(ld_min, ld_max, mols_petit, mols_exojax, T0, alpha, logg, logvmr)
+
+
+
+Rinst = 70000.
+nu_min = 1.0e8 / ld_max
+nu_max = 1.0e8 / ld_min
+Nx = math.ceil(Rinst * np.log(nu_max / nu_min)) + 1 # ueki
+Nx = math.ceil(Nx / 2.) * 2 # make even
+nusd, wav, res = wavenumber_grid(ld_min, ld_max, Nx, unit="AA", xsmode="premodit")
+
+beta_inst = resolution_to_gaussian_std(Rinst)
+res_calc = 900000.
+vsini_max = 100.0
+vr_array = velocity_grid(res_calc, vsini_max)
+
+f1_obs = response.ipgauss_sampling(nusd, nus1, f1, beta_inst, 0., vr_array)
+if len(nus2) % 2 == 1:
+    nus2 = nus2[1:]
+    f2 = f2[1:]
+f2_obs = response.ipgauss_sampling(nusd, nus2, f2, beta_inst, 0., vr_array)
+
+
+
+import matplotlib.pyplot as plt
+from matplotlib.ticker import (MultipleLocator, FormatStrFormatter, AutoMinorLocator)
+fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(40,8), gridspec_kw={'height_ratios': [3, 1]})
+plt.subplots_adjust(hspace=0)
+ax1.plot(1.0e8/nusd, f1_obs, alpha=0.5, label="ExoJAX")
+ax1.plot(1.0e8/nusd, f2_obs, alpha=0.5, label="petitRADTRANS")
+
+ax2.plot(1.0e8/nusd, f1_obs - f2_obs, "+", color="black", markersize=5, alpha=0.2)
+
+ax1.set_ylabel("Flux [erg/s/cm$\mathrm{^2}$/cm$\mathrm{^{-1}}$]", fontsize=15)
+ax2.set_xlabel("Wavelength [$\AA$]", fontsize=15)
+ax2.set_ylabel("Residual", fontsize=15)
+
+ax1.set_xlim(np.min(1.0e8/nusd), np.max(1.0e8/nusd))
+ax2.set_xlim(np.min(1.0e8/nusd), np.max(1.0e8/nusd))
+
+ax1.xaxis.set_ticks_position('both')
+ax1.yaxis.set_ticks_position('both')
+ax2.xaxis.set_ticks_position('both')
+ax2.yaxis.set_ticks_position('both')
+
+ax1.xaxis.set_minor_locator(AutoMinorLocator())
+ax1.yaxis.set_minor_locator(AutoMinorLocator())
+ax2.xaxis.set_minor_locator(AutoMinorLocator())
+ax2.yaxis.set_minor_locator(AutoMinorLocator())
+
+ax2.patch.set_alpha(0)
+ax1.tick_params(labelbottom=False, labeltop=True)
+
+ax1.legend()
+#plt.show()
+plt.savefig("output/narrow_R70000.png", bbox_inches='tight')
diff --git a/tests/integration/comparison/twostream/comparison_petitRADTRANS_wide_R7000.py b/tests/integration/comparison/twostream/comparison_petitRADTRANS_wide_R7000.py
new file mode 100644
index 000000000..a9b7e8688
--- /dev/null
+++ b/tests/integration/comparison/twostream/comparison_petitRADTRANS_wide_R7000.py
@@ -0,0 +1,218 @@
+from jax import config 
+config.update("jax_enable_x64", True)
+
+import numpy as np
+import jax.numpy as jnp
+import math
+import os
+
+from exojax.atm.atmprof import pressure_layer_logspace
+from exojax.utils.grids import wavenumber_grid
+from exojax.spec.multimol import MultiMol
+from exojax.spec import contdb
+from exojax.spec.layeropacity import layer_optical_depth, layer_optical_depth_CIA
+from exojax.spec import planck
+from exojax.spec.rtransfer import rtrun_emis_pure_absorption
+
+from petitRADTRANS import Radtrans
+from exojax.spec import molinfo
+import petitRADTRANS.nat_cst as nc
+
+from exojax.utils.instfunc import resolution_to_gaussian_std
+from exojax.utils.grids import velocity_grid
+from exojax.spec import response
+
+
+
+def run_exojax(path_data, ld_min, ld_max, mols, db, T0, alpha, logg, logvmr):
+    Parr, dParr, k = pressure_layer_logspace(log_pressure_top=-3., nlayer=200)
+    ONEARR = np.ones_like(Parr)
+
+
+    R = 900000.
+    ld_min = ld_min - 5.
+    ld_max = ld_max + 5.
+    nu_min = 1.0e8 / ld_max
+    nu_max = 1.0e8 / ld_min
+    Nx = math.ceil(R * np.log(nu_max / nu_min)) + 1 # ueki
+    Nx = math.ceil(Nx / 2.) * 2 # make even
+
+    nus, wav, res = wavenumber_grid(ld_min, ld_max, Nx, unit="AA", xsmode="premodit")
+    nus = [nus]
+    wav = [wav]
+    res = [res]
+
+
+    mul = MultiMol(molmulti=[mols], dbmulti=[db], database_root_path=path_data)
+    multimdb = mul.multimdb(nus, crit=1.e-30, Ttyp=1000.)
+    multiopa = mul.multiopa_premodit(multimdb, nus, auto_trange=[500.,1500.], dit_grid_resolution=1.0)
+
+    cdbH2H2 = []
+    cdbH2He = []
+    for k in range(len(nus)):
+        cdbH2H2.append(contdb.CdbCIA(os.path.join(path_data, 'H2-H2_2011.cia'), nus[k]))
+        cdbH2He.append(contdb.CdbCIA(os.path.join(path_data, 'H2-He_2011.cia'), nus[k]))
+
+    molmass_list, molmassH2, molmassHe = mul.molmass()
+
+
+    def frun(T0, alpha, logg, logvmr):
+        Tarr = T0 * (Parr)**alpha
+        Tarr = np.clip(Tarr, 500, None)
+
+        g = 10.**logg # cgs
+
+        vmr = jnp.power(10., jnp.array(logvmr))
+        vmrH2 = (1. - jnp.sum(vmr)) * 6./7.
+        vmrHe = (1. - jnp.sum(vmr)) * 1./7.
+        mmw = jnp.sum(vmr*jnp.array(molmass_list)) + vmrH2*molmassH2 + vmrHe*molmassHe
+        mmr = jnp.multiply(vmr, jnp.array(molmass_list)) / mmw
+
+        mu = []
+        for k in range(len(nus)):
+            dtaum = []
+            for i in range(len(mul.masked_molmulti[k])):
+                xsm = multiopa[k][i].xsmatrix(Tarr, Parr)
+                xsm = jnp.abs(xsm)
+                dtaum.append(layer_optical_depth(dParr, xsm, mmr[mul.mols_num[k][i]]*ONEARR, molmass_list[mul.mols_num[k][i]], g))
+
+            dtau = sum(dtaum)
+
+            if(len(cdbH2H2[k].nucia) > 0):
+                dtaucH2H2 = layer_optical_depth_CIA(nus[k], Tarr, Parr, dParr, vmrH2, vmrH2, mmw, g, cdbH2H2[k].nucia, cdbH2H2[k].tcia, cdbH2H2[k].logac)
+                dtau = dtau + dtaucH2H2
+            if(len(cdbH2He[k].nucia) > 0):
+                dtaucH2He = layer_optical_depth_CIA(nus[k], Tarr, Parr, dParr, vmrH2, vmrHe, mmw, g, cdbH2He[k].nucia, cdbH2He[k].tcia, cdbH2He[k].logac)
+                dtau = dtau + dtaucH2He
+
+            sourcef = planck.piBarr(Tarr, nus[k])
+            F0 = rtrun_emis_pure_absorption(dtau, sourcef)
+
+            mu.append(F0)
+
+        return mu
+
+    return nus[0], frun(T0, alpha, logg, logvmr)[0]
+
+
+
+def run_petit(ld_min, ld_max, mols, mols_exojax, T0, alpha, logg, logvmr):
+    atmosphere = Radtrans(line_species = mols,
+                          rayleigh_species = ['H2', 'He'],
+                          continuum_opacities = ['H2-H2', 'H2-He'],
+                          wlen_bords_micron = [ld_min*1e-4, ld_max*1e-4],
+                          mode = 'lbl')
+
+    pressures = np.logspace(-10, 2, 130)
+    atmosphere.setup_opa_structure(pressures)
+    temperature = T0*(pressures)**alpha
+    temperature = np.clip(temperature, 500, None)
+
+
+    molmass_list = []
+    for i in range(len(mols_exojax)):
+        molmass_list.append(molinfo.molmass(mols_exojax[i]))
+    molmassH2=molinfo.molmass("H2")
+    molmassHe=molinfo.molmass("He", db_HIT=False)
+
+    vmr = jnp.power(10., jnp.array(logvmr))
+    vmrH2 = (1. - jnp.sum(vmr)) * 6./7.
+    vmrHe = (1. - jnp.sum(vmr)) * 1./7.
+    mmw = jnp.sum(vmr*jnp.array(molmass_list)) + vmrH2*molmassH2 + vmrHe*molmassHe
+    mmr = jnp.multiply(vmr, jnp.array(molmass_list)) / mmw
+    mmrH2 = vmrH2 * molmassH2 / mmw
+    mmrHe = vmrHe * molmassHe / mmw
+
+    mass_fractions = {}
+    mass_fractions['H2'] = mmrH2 * np.ones_like(temperature)
+    mass_fractions['He'] = mmrHe * np.ones_like(temperature)
+    for i in range(len(mols)):
+        mass_fractions[mols[i]] = mmr[i] * np.ones_like(temperature)
+
+
+    MMW = mmw * np.ones_like(temperature)
+    gravity = 1e1**(logg)
+
+    atmosphere.calc_flux(temperature, mass_fractions, gravity, MMW)
+
+    ld = nc.c/atmosphere.freq/1e-4 # [um]
+    nus = 1.0e4 / ld # [cm^{-1}]
+    f = atmosphere.flux #[erg cm^{-2} s^{-1} Hz^{-1}]
+    f = f * nc.c #[erg/s/cm^2/cm^{-1}]
+
+    nus = nus[::-1]
+    f = f[::-1]
+
+    return nus, f
+
+
+
+path_data = "/home/kawashima/database"
+ld_min = 14600.
+ld_max = 16500.
+mols_exojax = ['H2O', 'CH4']
+db_exojax = ['ExoMol', 'HITEMP']
+mols_petit = ['H2O_main_iso', 'CH4_Hargreaves_main_iso']
+
+T0 = 995.56
+alpha = 0.09
+logg = 5.01
+logvmr = [-2.93, -3.06]
+
+nus1, f1 = run_exojax(path_data, ld_min, ld_max, mols_exojax, db_exojax, T0, alpha, logg, logvmr)
+nus2, f2 = run_petit(ld_min, ld_max, mols_petit, mols_exojax, T0, alpha, logg, logvmr)
+
+
+
+Rinst = 7000.
+nu_min = 1.0e8 / ld_max
+nu_max = 1.0e8 / ld_min
+Nx = math.ceil(Rinst * np.log(nu_max / nu_min)) + 1 # ueki
+Nx = math.ceil(Nx / 2.) * 2 # make even
+nusd, wav, res = wavenumber_grid(ld_min, ld_max, Nx, unit="AA", xsmode="premodit")
+
+beta_inst = resolution_to_gaussian_std(Rinst)
+res_calc = 900000.
+vsini_max = 100.0
+vr_array = velocity_grid(res_calc, vsini_max)
+
+f1_obs = response.ipgauss_sampling(nusd, nus1, f1, beta_inst, 0., vr_array)
+if len(nus2) % 2 == 1:
+    nus2 = nus2[1:]
+    f2 = f2[1:]
+f2_obs = response.ipgauss_sampling(nusd, nus2, f2, beta_inst, 0., vr_array)
+
+
+
+import matplotlib.pyplot as plt
+from matplotlib.ticker import (MultipleLocator, FormatStrFormatter, AutoMinorLocator)
+fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(40,8), gridspec_kw={'height_ratios': [3, 1]})
+plt.subplots_adjust(hspace=0)
+ax1.plot(1.0e8/nusd, f1_obs, alpha=0.5, label="ExoJAX")
+ax1.plot(1.0e8/nusd, f2_obs, alpha=0.5, label="petitRADTRANS")
+
+ax2.plot(1.0e8/nusd, f1_obs - f2_obs, "+", color="black", markersize=5, alpha=0.2)
+
+ax1.set_ylabel("Flux [erg/s/cm$\mathrm{^2}$/cm$\mathrm{^{-1}}$]", fontsize=15)
+ax2.set_xlabel("Wavelength [$\AA$]", fontsize=15)
+ax2.set_ylabel("Residual", fontsize=15)
+
+ax1.set_xlim(np.min(1.0e8/nusd), np.max(1.0e8/nusd))
+ax2.set_xlim(np.min(1.0e8/nusd), np.max(1.0e8/nusd))
+
+ax1.xaxis.set_ticks_position('both')
+ax1.yaxis.set_ticks_position('both')
+ax2.xaxis.set_ticks_position('both')
+ax2.yaxis.set_ticks_position('both')
+
+ax1.xaxis.set_minor_locator(AutoMinorLocator())
+ax1.yaxis.set_minor_locator(AutoMinorLocator())
+ax2.xaxis.set_minor_locator(AutoMinorLocator())
+ax2.yaxis.set_minor_locator(AutoMinorLocator())
+
+ax2.patch.set_alpha(0)
+ax1.tick_params(labelbottom=False, labeltop=True)
+
+ax1.legend()
+#plt.show()
+plt.savefig("output/wide_R7000.png", bbox_inches='tight')