Source code for optika.sensors.materials._ramanathan_2020._ramanathan_2020

import pathlib
import math
import random
import numpy as np
import numba
import astropy.units as u
import named_arrays as na
import optika
from .._stern_1994 import (
    _thickness_implant,
    _thickness_substrate,
    _width_pixel,
    _cce_backsurface,
)

__all__ = [
    "energy_bandgap",
    "energy_pair",
    "energy_pair_inf",
    "quantum_yield_ideal",
    "fano_factor",
    "fano_factor_inf",
    "electrons_measured",
]


def _probability_of_n_pairs_from_file(
    path: pathlib.Path,
) -> na.FunctionArray[na.CartesianNdVectorArray, na.ScalarArray]:

    a = np.loadtxt(path)
    a = na.ScalarArray(a, axes=("wavelength", "num_electron"))
    energy = a[dict(num_electron=0)] << u.eV
    pn = a[dict(num_electron=slice(1, None))]
    n = na.arange(1, pn.shape["num_electron"] + 1, axis="num_electron")

    return na.FunctionArray(
        inputs=na.CartesianNdVectorArray(
            components=dict(
                n=n,
                energy=energy,
            ),
        ),
        outputs=pn,
    )


def _probability_of_n_pairs_ramanathan() -> na.FunctionArray[
    na.CartesianNdVectorArray,
    na.ScalarArray,
]:

    directory = pathlib.Path(__file__).parent
    pn_000K = _probability_of_n_pairs_from_file(directory / "p0K.dat")
    pn_100K = _probability_of_n_pairs_from_file(directory / "p100K.dat")
    pn_300K = _probability_of_n_pairs_from_file(directory / "p300K.dat")

    probability = na.stack(
        arrays=[
            pn_000K.outputs,
            pn_100K.outputs,
            pn_300K.outputs,
        ],
        axis="temperature",
    )

    n = pn_000K.inputs.components["n"]
    energy = pn_000K.inputs.components["energy"]

    temperature = na.ScalarArray(
        ndarray=[0, 100, 300] * u.K,
        axes="temperature",
    )

    return na.FunctionArray(
        inputs=na.CartesianNdVectorArray(
            components=dict(
                energy=energy,
                temperature=temperature,
                n=n,
            )
        ),
        outputs=probability,
    )


[docs] def energy_bandgap( temperature: u.Quantity | na.ScalarArray = 300 * u.K, ) -> na.ScalarArray: r""" Bandgap energy in silicon given by :cite:t:`Ramanathan2020`. Parameters ---------- temperature The temperature of the silicon. Examples -------- Reproduce Figure 2 of :cite:t:`Ramanathan2020`, and plot the bandgap energy as a function of temperature. .. jupyter-execute:: import matplotlib.pyplot as plt import astropy.units as u import astropy.visualization import named_arrays as na import optika T = na.linspace(0, 350, axis="temperature", num=101) * u.K energy_gap = optika.sensors.energy_bandgap(T) with astropy.visualization.quantity_support(): fig, ax = plt.subplots() na.plt.plot( T, energy_gap ) ax.set_xlabel(f"temperature ({ax.get_xlabel()})") ax.set_ylabel(f"bandgap energy ({ax.get_ylabel()})") Notes ----- :cite:t:`Ramanathan2020` gives the bandgap energy as .. math:: E_g(T) = E_g(0) - \frac{a T^2}{T + b} where :math:`T` is the temperature of the silicon, :math:`E_g(0) = 1.192 \, \text{eV}`, :math:`a = 4.9 \times 10^{-4} \, \text{eV / K}`, and :math:`b = 655 \, \text{K}`. """ T = temperature.to(u.K, equivalencies=u.temperature()) energy_gap_0 = 1.1692 * u.eV a = 4.9e-4 * u.eV / u.K b = 655 * u.K return energy_gap_0 - a * np.square(T) / (T + b)
[docs] def energy_pair( wavelength: u.Quantity | na.ScalarArray, temperature: u.Quantity | na.ScalarArray = 300 * u.K, ) -> na.ScalarArray: """ Calculate the average pair-production energy in silicon given by :cite:t:`Ramanathan2020`. Parameters ---------- wavelength The vacuum wavelength of the incident photons. temperature The temperature of the silicon. Examples -------- Compute the pair-production energy as a function of incident photon energy. .. jupyter-execute:: import matplotlib.pyplot as plt import astropy.units as u import astropy.visualization import named_arrays as na import optika energy = na.geomspace(1, 100, axis="energy", num=1001) * u.eV energy_pair = optika.sensors.energy_pair(energy) with astropy.visualization.quantity_support(): fig, ax = plt.subplots() na.plt.plot( energy, energy_pair, ) ax.set_xscale("log") ax.set_xlabel(f"incident photon energy ({ax.get_xlabel()})") ax.set_ylabel(f"pair-production energy ({ax.get_ylabel()})") """ energy = wavelength.to(u.eV, equivalencies=u.spectral()) temperature = temperature.to(u.K, equivalencies=u.temperature()) pn = _probability_of_n_pairs_ramanathan() _n = pn.inputs.components["n"] _energy = pn.inputs.components["energy"] _temperature = pn.inputs.components["temperature"] _probability = pn.outputs _iqy = (_n * _probability).sum("num_electron") _energy_pair = _energy / _iqy _energy_pair_inf = energy_pair_inf(temperature) energy_pair = na.interp( x=temperature, xp=_temperature, fp=_energy_pair, ) energy_pair = na.interp( x=energy, xp=_energy, fp=energy_pair, right=_energy_pair_inf.value, ) return energy_pair
[docs] def energy_pair_inf( temperature: u.Quantity | na.ScalarArray = 300 * u.K, ) -> na.ScalarArray: r""" The asymptotic electron-hole pair production energy in silicon given by :cite:t:`Ramanathan2020`. Parameters ---------- temperature The temperature of the silicon. Notes ----- :cite:t:`Ramanathan2020` gives the mean pair production energy as .. math:: \epsilon_{eh} = 1.7 E_g + 0.084 A + 1.3, where :math:`E_g` is the bandgap energy of silicon calculated using :func:`energy_bandgap` and :math:`A = 5.2 \, \text{eV}^2`. """ A = 5.2 * u.eV**2 E_g = energy_bandgap(temperature) result = 1.7 * E_g + 0.084 * A / u.eV + 1.3 * u.eV return result
[docs] def quantum_yield_ideal( wavelength: u.Quantity | na.ScalarArray, temperature: u.Quantity | na.ScalarArray = 300 * u.K, ) -> na.ScalarArray: r""" Calculate the ideal quantum yield of a silicon detector for a given wavelength and temperature using the model given in :cite:t:`Ramanathan2020`. Parameters ---------- wavelength The vacuum wavelength of the incident photons. temperature The temperature of the silicon detector. Examples -------- Plot the ideal quantum yield vs wavelength .. jupyter-execute:: import matplotlib.pyplot as plt import astropy.units as u import named_arrays as na import optika # Define an array of wavelengths wavelength = na.geomspace(100, 100000, axis="wavelength", num=1001) << u.AA # Compute the quantum yield iqy = optika.sensors.quantum_yield_ideal(wavelength) # Plot the quantum yield vs wavelength fig, ax = plt.subplots() na.plt.plot(wavelength, iqy, ax=ax); ax.set_xscale("log"); ax.set_xlabel(f"wavelength ({wavelength.unit:latex_inline})"); ax.set_ylabel(f"quantum yield ({iqy.unit:latex_inline})"); """ energy = wavelength.to(u.eV, equivalencies=u.spectral()) W = energy_pair( wavelength=wavelength, temperature=temperature, ) iqy = energy / W return iqy * u.electron / u.photon
[docs] def fano_factor( wavelength: u.Quantity | na.ScalarArray, temperature: u.Quantity | na.ScalarArray = 300 * u.K, ) -> na.ScalarArray: r""" Calculate the Fano factor of a silicon detector for a given wavelength and temperature using the model given in :cite:t:`Ramanathan2020`. Parameters ---------- wavelength The vacuum wavelength of the incident photons. temperature The temperature of the silicon detector. Examples -------- Plot the Fano factor vs wavelength .. jupyter-execute:: import matplotlib.pyplot as plt import astropy.units as u import named_arrays as na import optika # Define an array of wavelengths wavelength = na.geomspace(100, 100000, axis="wavelength", num=1001) << u.AA # Compute the Fano factor f = optika.sensors.fano_factor(wavelength) # Plot the Fano factor vs wavelength fig, ax = plt.subplots() na.plt.plot(wavelength, f, ax=ax); ax.set_xscale("log"); ax.set_xlabel(f"wavelength ({wavelength.unit:latex_inline})"); ax.set_ylabel(f"Fano factor ({f.unit:latex_inline})"); """ energy = wavelength.to(u.eV, equivalencies=u.spectral()) temperature = temperature.to(u.K, equivalencies=u.temperature()) pn = _probability_of_n_pairs_ramanathan() _n = pn.inputs.components["n"] _energy = pn.inputs.components["energy"] _temperature = pn.inputs.components["temperature"] _probability = pn.outputs _iqy = (_n * _probability).sum("num_electron") _v = (np.square(_n) * _probability).sum("num_electron") _fano_factor = (_v - np.square(_iqy)) / _iqy _fano_factor_inf = fano_factor_inf(_temperature) fano_factor = na.interp( x=energy, xp=_energy, fp=_fano_factor, right=_fano_factor_inf.value, ) fano_factor = na.interp( x=temperature, xp=_temperature, fp=fano_factor, ) return fano_factor * u.electron / u.photon
[docs] def fano_factor_inf( temperature: u.Quantity | na.ScalarArray = 300 * u.K, ) -> na.ScalarArray: r""" The asymptotic Fano factor in silicon given by :cite:t:`Ramanathan2020`. Parameters ---------- temperature The temperature of the silicon. Notes ----- :cite:t:`Ramanathan2020` gives the mean Fano factor as .. math:: \epsilon_{eh} = -0.028 E_g + 0.0015 A + 0.14, where :math:`E_g` is the bandgap energy of silicon calculated using :func:`energy_bandgap` and :math:`A = 5.2 \, \text{eV}^2`. """ temperature = temperature.to(u.K, equivalencies=u.temperature()) A = 5.2 * u.eV**2 E_g = energy_bandgap(temperature) result = -0.028 * E_g + 0.0015 * A / u.eV + 0.14 * u.eV result = result * u.electron / u.photon / u.eV return result
def probability_of_n_pairs( wavelength: u.Quantity | na.ScalarArray, temperature: u.Quantity | na.ScalarArray = 300 * u.K, ) -> na.FunctionArray[na.ScalarArray, na.ScalarArray]: r""" Calculate the PMF of the number of electron-hole pairs generated in a silicon detector for a given wavelength and temperature using the model given in :cite:t:`Ramanathan2020`. Parameters ---------- wavelength The vacuum wavelength of the incident photons. temperature The temperature of the silicon detector. """ energy = wavelength.to(u.eV, equivalencies=u.spectral()) temperature = temperature.to(u.K, equivalencies=u.temperature()) pn = _probability_of_n_pairs_ramanathan() _n = pn.inputs.components["n"] _energy = pn.inputs.components["energy"] _temperature = pn.inputs.components["temperature"] _probability = pn.outputs probability = na.interp( x=temperature, xp=_temperature, fp=_probability, ) probability = na.interp( x=energy, xp=_energy, fp=probability, ) result = na.FunctionArray( inputs=_n, outputs=probability, ) return result
[docs] def electrons_measured( photons_absorbed: u.Quantity | na.AbstractScalar, wavelength: u.Quantity | na.ScalarArray, absorption: None | u.Quantity | na.AbstractScalar = None, thickness_implant: u.Quantity | na.AbstractScalar = _thickness_implant, thickness_depletion: None | u.Quantity | na.AbstractScalar = None, thickness_substrate: u.Quantity | na.AbstractScalar = _thickness_substrate, width_pixel: ( u.Quantity | na.AbstractScalar | na.AbstractCartesian2dVectorArray ) = _width_pixel, cce_backsurface: u.Quantity | na.AbstractScalar = _cce_backsurface, temperature: u.Quantity | na.ScalarArray = 300 * u.K, axis_xy: None | tuple[str, str] = None, wrap: bool = False, shape_random: None | dict[str, int] = None, ) -> na.AbstractScalar: r""" A random sample from the distribution of measured electrons given the number of photons absorbed by the light-sensitive layer of the sensor. This function accounts for both Fano noise and recombination noise due to partial-charge collection. Parameters ---------- photons_absorbed The number of photons absorbed by the light-sensitive layer of the sensor. wavelength The vacuum wavelength of the absorbed photons. absorption The absorption coefficient of silicon per unit perpendicular depth. For oblique incidence, supply the effective coefficient from :func:`optika.sensors.absorption_effective`, which folds in the refracted angle, so no separate angle argument is needed. thickness_implant The thickness of the implant layer, where partial-charge collection occurs. thickness_depletion The thickness of the depletion region, the region with significant electric field. If :obj:`None` (the default), this is set to the same value as `thickness_substrate`. thickness_substrate The thickness of the entire light-sensitive region of the device. width_pixel The size of a single pixel on the sensor. A scalar gives square pixels; a :class:`named_arrays.AbstractCartesian2dVectorArray` whose ``x``/``y`` components are the pixel widths along ``axis_xy[0]``/``axis_xy[1]`` gives rectangular pixels. cce_backsurface The differential charge collection efficiency on the back surface of the sensor. temperature The temperature of the silicon detector. Default is room temperature. axis_xy The two logical axes corresponding to the pixel grid of the sensor along which electrons will diffuse. If :obj:`None` (the default), there is no charge diffusion. wrap Controls how diffused charge is treated at the edges of the pixel grid. If :obj:`False` (the default), charge that diffuses past the edge of the grid is lost, as it would be at the physical edge of a sensor. If :obj:`True`, the grid is treated as periodic and the charge re-enters the opposite edge (a toroidal boundary). shape_random Additional shape used to specify the number of samples to draw. Examples -------- Plot the energy spectrum of twenty 6 keV photons emitted from an Fe-55 radioactive source. .. jupyter-execute:: import matplotlib.pyplot as plt import astropy.units as u import astropy.visualization import named_arrays as na import optika # Define the number of experiments to perform num_experiments = 100000 # Define the expected number of photons # for each experiment photons_absorbed = (20 * u.photon).astype(int) # Define the wavelength at which to sample the distribution wavelength = 5.9 * u.keV wavelength = wavelength.to(u.AA, equivalencies=u.spectral()) # Compute the actual number of electrons measured for each experiment electrons = optika.sensors.electrons_measured( photons_absorbed=photons_absorbed, wavelength=wavelength, shape_random=dict(experiment=num_experiments), ) # Define the histogram bins step = 10 bins = na.arange( electrons.value.min()-step/2, electrons.value.max()+step/2, step=step, axis="bin", ) * u.electron # Compute a histogram of resulting energy spectrum hist = na.histogram( electrons, bins=bins, axis="experiment", ) # Plot the histogram with astropy.visualization.quantity_support(): fig, ax = plt.subplots() line = na.plt.stairs( hist.inputs, hist.outputs, ax=ax, ) """ temperature = temperature.to(u.K, equivalencies=u.temperature()) if absorption is None: absorption = optika.chemicals.Chemical("Si").absorption(wavelength) if thickness_depletion is None: thickness_depletion = thickness_substrate if shape_random is None: shape_random = dict() if not isinstance(width_pixel, na.AbstractCartesian2dVectorArray): width_pixel = na.Cartesian2dVectorArray(width_pixel, width_pixel) width_pixel_x = width_pixel.x width_pixel_y = width_pixel.y shape = na.broadcast_shapes( na.shape(photons_absorbed), na.shape(wavelength), na.shape(absorption), na.shape(thickness_implant), na.shape(thickness_depletion), na.shape(thickness_substrate), na.shape(width_pixel_x), na.shape(width_pixel_y), na.shape(cce_backsurface), na.shape(temperature), shape_random, ) if axis_xy is not None: axis_x, axis_y = axis_xy shape[axis_x] = shape.pop(axis_x) shape[axis_y] = shape.pop(axis_y) else: axis_x = "__dummy_x__" axis_y = "__dummy_y__" shape[axis_x] = 1 shape[axis_y] = 1 photons_absorbed = na.broadcast_to(photons_absorbed, shape) absorption = na.broadcast_to(absorption, shape) thickness_implant = na.broadcast_to(thickness_implant, shape) thickness_depletion = na.broadcast_to(thickness_depletion, shape) thickness_substrate = na.broadcast_to(thickness_substrate, shape) width_pixel_x = na.broadcast_to(width_pixel_x, shape) width_pixel_y = na.broadcast_to(width_pixel_y, shape) cce_backsurface = na.broadcast_to(cce_backsurface, shape) if not isinstance(cce_backsurface, u.Quantity): cce_backsurface = cce_backsurface << u.dimensionless_unscaled pmf_pair = probability_of_n_pairs(wavelength, temperature) p_n = pmf_pair.outputs n = pmf_pair.inputs shape_n = na.broadcast_shapes(shape, n.shape) p_n = p_n.broadcast_to(shape_n) n = n.broadcast_to(shape_n) energy_inf = 1 * u.keV energy_pair_inf = energy_inf / quantum_yield_ideal(energy_inf, temperature).value fano_inf = fano_factor(energy_inf, temperature) wavelength = na.broadcast_to(wavelength, shape) energy_pair_inf = energy_pair_inf.broadcast_to(shape) fano_inf = fano_inf.broadcast_to(shape) result = _electrons_measured_quantity( photons_absorbed=photons_absorbed.ndarray, wavelength=wavelength.ndarray, absorption=absorption.ndarray, thickness_implant=thickness_implant.ndarray, thickness_depletion=thickness_depletion.ndarray, thickness_substrate=thickness_substrate.ndarray, width_pixel_x=width_pixel_x.ndarray, width_pixel_y=width_pixel_y.ndarray, cce_backsurface=cce_backsurface.ndarray, p_n=p_n.ndarray, n=n.ndarray, energy_pair_inf=energy_pair_inf.ndarray, fano_inf=fano_inf.ndarray, wrap=wrap, ) result = na.ScalarArray( ndarray=result, axes=tuple(shape), ) if axis_xy is None: result = result[{axis_x: 0, axis_y: 0}] return result
def _electrons_measured_quantity( photons_absorbed: u.Quantity, wavelength: u.Quantity, absorption: u.Quantity, thickness_implant: u.Quantity, thickness_depletion: u.Quantity, thickness_substrate: u.Quantity, width_pixel_x: u.Quantity, width_pixel_y: u.Quantity, cce_backsurface: u.Quantity, p_n: np.ndarray, n: np.ndarray, energy_pair_inf: u.Quantity, fano_inf: u.Quantity, wrap: bool, ) -> u.Quantity: shape = np.broadcast_shapes( photons_absorbed.shape, absorption.shape, thickness_implant.shape, thickness_depletion.shape, thickness_substrate.shape, cce_backsurface.shape, width_pixel_x.shape, width_pixel_y.shape, ) num_x = shape[~1] num_y = shape[~0] unit_length = u.mm photons_absorbed = photons_absorbed.to_value(u.photon) energy = wavelength.to(u.eV, equivalencies=u.spectral()) absorption = absorption.to_value(1 / unit_length) thickness_implant = thickness_implant.to_value(unit_length) thickness_depletion = thickness_depletion.to_value(unit_length) thickness_substrate = thickness_substrate.to_value(unit_length) width_pixel_x = width_pixel_x.to_value(unit_length) width_pixel_y = width_pixel_y.to_value(unit_length) cce_backsurface = cce_backsurface.to_value(u.dimensionless_unscaled) energy_pair_inf = energy_pair_inf.to_value(u.eV) fano_inf = fano_inf.to_value(u.electron / u.photon) result = _electrons_measured_numba( photons_absorbed=photons_absorbed.reshape(-1, num_x, num_y), energy=energy.reshape(-1, num_x, num_y), absorption=absorption.reshape(-1, num_x, num_y), thickness_implant=thickness_implant.reshape(-1, num_x, num_y), thickness_depletion=thickness_depletion.reshape(-1, num_x, num_y), thickness_substrate=thickness_substrate.reshape(-1, num_x, num_y), width_pixel_x=width_pixel_x.reshape(-1, num_x, num_y), width_pixel_y=width_pixel_y.reshape(-1, num_x, num_y), cce_backsurface=cce_backsurface.reshape(-1, num_x, num_y), p_n=p_n.reshape(-1, num_x, num_y, p_n.shape[~0]), n=n.reshape(-1, num_x, num_y, n.shape[~0]), energy_pair_inf=energy_pair_inf.reshape(-1, num_x, num_y), fano_inf=fano_inf.reshape(-1, num_x, num_y), wrap=wrap, ) result = result.reshape(shape) result = result << u.electron return result @numba.njit( cache=True, fastmath=True, parallel=True, ) def _electrons_measured_numba( # pragma: nocover photons_absorbed: np.ndarray, energy: np.ndarray, absorption: np.ndarray, thickness_implant: np.ndarray, thickness_depletion: np.ndarray, thickness_substrate: np.ndarray, width_pixel_x: np.ndarray, width_pixel_y: np.ndarray, cce_backsurface: np.ndarray, p_n: np.ndarray, n: np.ndarray, energy_pair_inf: np.ndarray, fano_inf: np.ndarray, wrap: bool, ) -> np.ndarray: num_i, num_x, num_y, num_n = p_n.shape result = np.zeros((num_i, num_x, num_y)) for i in numba.prange(num_i): for x in range(num_x): for y in range(num_y): num_photon = int(photons_absorbed[i, x, y]) energy_i = energy[i, x, y] a = absorption[i, x, y] W = thickness_implant[i, x, y] h_0 = cce_backsurface[i, x, y] cmf_i = np.cumsum(p_n[i, x, y]) n_i = n[i, x, y] energy_pair_inf_i = energy_pair_inf[i, x, y] fano_inf_i = fano_inf[i, x, y] z_substrate = thickness_substrate[i, x, y] z_ff = z_substrate - thickness_depletion[i, x, y] wp_x = width_pixel_x[i, x, y] wp_y = width_pixel_y[i, x, y] # guard against a non-absorbing medium (a == 0), where the # reciprocal absorption length is undefined d = 1 / a if a > 0 else 0.0 # every photon is absorbed within the substrate, so sample the # depth from the exponential truncated to [0, z_substrate) fraction_absorbed = 1 - math.exp(-a * z_substrate) mean_inf = energy_i / energy_pair_inf_i std_inf = math.sqrt(fano_inf_i * mean_inf) low_energy = energy_i <= 50 for j in range(num_photon): if low_energy: x_ij = random.uniform(0, 1) for k_ij in range(num_n): if cmf_i[k_ij] > x_ij: break n_ij = n_i[k_ij] else: n_ij = random.normalvariate( mu=mean_inf, sigma=std_inf, ) n_ij = round(n_ij) y_ij = random.uniform(0, 1) if a > 0: z_ij = -d * math.log(1 - y_ij * fraction_absorbed) else: # with no absorption the truncated-exponential depth # distribution reduces to a uniform one over [0, z_substrate) z_ij = y_ij * z_substrate if z_ij < W: h_ij = h_0 + (1 - h_0) * z_ij / W else: h_ij = 1 m_ij = np.random.binomial(n=n_ij, p=h_ij) u = random.uniform(-0.5, 0.5) v = random.uniform(-0.5, 0.5) for e in range(m_ij): if z_ij < z_ff and wp_x > 0 and wp_y > 0: w = z_ff * math.sqrt(1 - z_ij / z_ff) p = random.gauss(u, w / wp_x) q = random.gauss(v, w / wp_y) p = round(p) q = round(q) else: p = q = 0 x_e = x + p y_e = y + q if wrap: # toroidal boundary: charge re-enters the opposite edge result[i, x_e % num_x, y_e % num_y] += 1 elif (0 <= x_e < num_x) and (0 <= y_e < num_y): result[i, x_e, y_e] += 1 # otherwise the electron diffused off the sensor and is lost return result