Uncertainty quantification of partial wave transmission coefficients

This notebook extends the uncertainty-propagation workflow to angular reaction observables. It shows how the same model choices that affect transmission coefficients flow through to angle-dependent reaction outputs.

import numpy as np
from IPython.display import Math, display
from periodictable import elements

from matplotlib import pyplot as plt

Compare mass-model inputs to Koning-Delaroche

A, Z = (24, 12)

name_core = str(elements[Z].symbol)
display(Math(f"^{{{A}}} \\rm{{{name_core}}}"))

\[\displaystyle ^{24} \rm{Mg}\]

Compute transmission coefficients

Let’s use jitr to calculate UQ’ed transmission coefficients using KDUQ Fermi energies and those from a variety of mass models

from tqdm import tqdm

import jitr

neutron = (1, 0)
proton = (1, 1)
projectile = proton
target = (A, Z)

# we have 416 samples from the KDUQ posterior
kduq_omp_samples = jitr.optical_potentials.kduq.get_samples(proton)

# com_energy_grid = np.logspace(-1, 1.3, 100)
lab_energy_grid = np.array([65, 200])
range_fm = 15
lmax = 20

reaction = jitr.reactions.Reaction(target=target, projectile=projectile, process="EL")

def set_up_grid(core, lab_energy_grid):
    solvers = []
    for _i, Elab in enumerate(tqdm(lab_energy_grid)):
        kinematics = reaction.kinematics(Elab)
        a = range_fm * kinematics.k + np.pi / 2
        N = jitr.utils.suggested_basis_size(a)
        solvers.append(
            jitr.xs.elastic.IntegralWorkspace(
                reaction=reaction,
                kinematics=kinematics,
                channel_radius_fm=a / kinematics.k,
                solver=jitr.rmatrix.Solver(N),
                lmax=lmax,
                smatrix_abs_tol=0,
            )
        )
    return solvers

solvers = set_up_grid(target, lab_energy_grid)

100%|██████████| 2/2 [00:10<00:00,  5.11s/it]

Run the uncertainty propagation

KDUQ

tcoeff_kduq = np.zeros((lab_energy_grid.size, kduq_omp_samples.shape[0], 2, lmax + 1))
for j, sample in enumerate(tqdm(kduq_omp_samples)):
    for i, _Elab in enumerate(lab_energy_grid):
        rgrid = solvers[i].radial_grid()
        central_params, spin_orbit_params, coulomb_params = (
            jitr.optical_potentials.kduq.calculate_params(
                projectile,
                target,
                solvers[i].kinematics.Elab,
                *sample,
            )
        )

        tplus, tminus = solvers[i].transmission_coefficients(
            jitr.optical_potentials.kduq.central(rgrid, *central_params),
            jitr.optical_potentials.kduq.spin_orbit(rgrid, *spin_orbit_params),
            jitr.optical_potentials.kduq.coulomb_charged_sphere(rgrid, *coulomb_params),
        )
        tcoeff_kduq[i, j, 0, :] = tplus
        tcoeff_kduq[i, j, 1, :] = tminus

 96%|█████████▋| 401/416 [00:19<00:00, 33.49it/s]/home/kyle/umich/jitr/src/jitr/optical_potentials/kduq.py:470: RuntimeWarning: overflow encountered in exp
  d2 = d2_0 + d2_A / (1 + np.exp((A - d2_A3) / d2_A2))
100%|██████████| 416/416 [00:19<00:00, 21.27it/s]

data = [np.zeros(lmax + 1)] * 8
names = [
    "E=65 MeV, j = l + 1/2",
    "E=65 MeV, j = l - 1/2",
    "E=65 MeV, j = l + 1/2, err",
    "E=65 MeV, j = l - 1/2, err",
    "E=200 MeV, j = l + 1/2",
    "E=200 MeV, j = l - 1/2",
    "E=200 MeV, j = l + 1/2, err",
    "E=200 MeV, j = l - 1/2, err",
]

from pandas import DataFrame as df

data = df.from_dict(dict(zip(names, data, strict=False)))

plus_color = "tab:blue"
minus_color = "tab:orange"

ci_plus = np.percentile(tcoeff_kduq[0, :, 0, :], 50, axis=0)
ci_plus_errs = np.percentile(tcoeff_kduq[0, :, 0, :], 84, axis=0) - np.percentile(
    tcoeff_kduq[0, :, 0, :], 16, axis=0
)

ci_minus = np.percentile(tcoeff_kduq[0, :, 1, :], 50, axis=0)
ci_minus_errs = np.percentile(tcoeff_kduq[0, :, 1, :], 84, axis=0) - np.percentile(
    tcoeff_kduq[0, :, 1, :], 16, axis=0
)

ci_minus[0] = None
ci_minus_errs[0] = None

data["E=65 MeV, j = l + 1/2"] = ci_plus
data["E=65 MeV, j = l + 1/2, err"] = ci_plus_errs
data["E=65 MeV, j = l - 1/2"] = ci_minus
data["E=65 MeV, j = l - 1/2, err"] = ci_minus_errs

fig = plt.figure(figsize=(6, 4))
ls = np.arange(lmax + 1)
plt.errorbar(
    ls,
    ci_plus * (2 * ls + 1),
    ci_plus_errs * (2 * ls + 1),
    linestyle="none",
    marker="^",
    label="$j = l + 1/2$",
)

plt.errorbar(
    ls[1:],
    ci_minus[1:] * (2 * ls[1:] + 1),
    ci_minus_errs[1:] * (2 * ls[1:] + 1),
    linestyle="none",
    marker="v",
    label="$j = l - 1/2$",
)

# plt.yscale("log")
plt.xticks([0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20])
plt.xlabel(r"$\ell$")
plt.ylabel(r"$(2  \ell +1) \mathcal{T}_{lj}$")
plt.title(r"$^{24}$Mg(p,p) 65 MeV", fontsize=14)
plt.legend()
plt.tight_layout()

ci_plus = np.percentile(tcoeff_kduq[1, :, 0, :], 50, axis=0)
ci_plus_errs = np.percentile(tcoeff_kduq[1, :, 0, :], 84, axis=0) - np.percentile(
    tcoeff_kduq[1, :, 0, :], 16, axis=0
)

ci_minus = np.percentile(tcoeff_kduq[1, :, 1, :], 50, axis=0)
ci_minus_errs = np.percentile(tcoeff_kduq[1, :, 1, :], 84, axis=0) - np.percentile(
    tcoeff_kduq[1, :, 1, :], 16, axis=0
)
ci_minus[0] = None
ci_minus_errs[0] = None

data["E=200 MeV, j = l + 1/2"] = ci_plus
data["E=200 MeV, j = l + 1/2, err"] = ci_plus_errs
data["E=200 MeV, j = l - 1/2"] = ci_minus
data["E=200 MeV, j = l - 1/2, err"] = ci_minus_errs

fig = plt.figure(figsize=(6, 4))
plt.errorbar(
    ls,
    ci_plus * (2 * ls + 1),
    ci_plus_errs * (2 * ls + 1),
    linestyle="none",
    marker="^",
    label="$j = l + 1/2$",
)

plt.errorbar(
    ls[1:],
    ci_minus[1:] * (2 * ls[1:] + 1),
    ci_minus_errs[1:] * (2 * ls[1:] + 1),
    linestyle="none",
    marker="v",
    label="$j = l - 1/2$",
)

# plt.yscale("log")
plt.xticks([0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20])
plt.xlabel(r"$\ell$")
plt.ylabel(r"$(2  \ell +1) \mathcal{T}_{lj}$")
plt.title(r"$^{24}$Mg(p,p) 200 MeV", fontsize=14)
plt.legend()
plt.tight_layout()

# NBVAL_CHECK_OUTPUT
data.head()

	E=65 MeV, j = l + 1/2	E=65 MeV, j = l - 1/2	E=65 MeV, j = l + 1/2, err	E=65 MeV, j = l - 1/2, err	E=200 MeV, j = l + 1/2	E=200 MeV, j = l - 1/2	E=200 MeV, j = l + 1/2, err	E=200 MeV, j = l - 1/2, err
0	0.742337	NaN	0.102470	NaN	0.821403	NaN	0.079719	NaN
1	0.724359	0.743781	0.105309	0.100490	0.809859	0.834128	0.082819	0.079568
2	0.717748	0.750458	0.100080	0.093144	0.794527	0.836448	0.085787	0.082877
3	0.690068	0.740712	0.099991	0.087372	0.771663	0.836196	0.088131	0.089072
4	0.666408	0.735901	0.095884	0.075141	0.740452	0.831666	0.099101	0.094705