Simulating zodi with a telescope’s orbit#

The zodi_rate and zodi_readout entry points in coronagraphoto compute the local zodiacal-light count rate on the detector for one epoch and one line of sight. To generate a time series of the kind a paper figure or a mission yield calculation needs, the per-frame geometry has to be threaded in from an observatory model. This page shows the recommended composition using orbix.observatory.ObservatoryL2Halo for the observatory and skyscapes.background.LeinertZodi for the brightness model, and ends with a short validation checklist.

The composition#

A single-epoch zodi simulation pulls together four pieces, each responsible for a distinct part of the calculation. The ObservatoryL2Halo instance provides the heliocentric position of the telescope and the per-target sky-geometry angles needed for the Leinert lookup, the LeinertZodi instance evaluates the surface brightness from those angles, the optical path threads the brightness through the coronagraph’s sky_trans map and the detector resampling, and zodi_rate returns the per-pixel count rate. Adding Poisson shot noise on top is the job of zodi_readout, which composes a detector readout step around zodi_rate.

The two Leinert inputs are read off the geometry by the observatory helpers as described in the skyscapes Local zodi + telescope geometry doc, and then handed to zodi_rate:

from orbix.observatory import ObservatoryL2Halo
from skyscapes.background import LeinertZodi
from coronagraphoto import zodi_rate

obs = ObservatoryL2Halo.from_default()
zodi = LeinertZodi(reference_mag_arcsec2=22.0)

mjd = 60575.25
start_time_jd = mjd + 2_400_000.5
ecl_lat = float(obs.ecliptic_latitude_deg(mjd, ra_rad, dec_rad))
helio_lon = float(obs.helio_ecliptic_longitude_deg(mjd, ra_rad, dec_rad))

rate = zodi_rate(
    zodi,
    optical_path,
    start_time_jd=start_time_jd,
    wavelength_nm=550.0,
    bin_width_nm=50.0,
    ecliptic_lat_deg=ecl_lat,
    solar_lon_deg=helio_lon,
)

The returned rate is an (ny, nx) array of electrons per second on the detector. Multiplying by an exposure time and adding Poisson shot noise yields a realistic electron image, and zodi_readout packages both steps:

Year-long simulations#

A year-long animation is a simple matter of looping the single-epoch recipe over a sequence of MJD samples and stepping the geometry along the L2 halo trajectory. Each frame draws an independent PRNG key so the Poisson realisations are uncorrelated, and any frame whose Leinert lookup falls in the out-of-range region returns NaN, which downstream code uses as the “target unobservable this epoch” gate:

import jax
import jax.numpy as jnp
import numpy as np
from coronagraphoto import zodi_readout

obs = ObservatoryL2Halo.from_default(equinox_mjd=60575.25)
zodi = LeinertZodi(reference_mag_arcsec2=22.0)
prng_keys = jax.random.split(jax.random.PRNGKey(0), n_frames)
mjds = 60575.25 + np.linspace(0.0, 365.25, n_frames)

ra_rad = jnp.deg2rad(target_ra_deg)
dec_rad = jnp.deg2rad(target_dec_deg)

for i, mjd in enumerate(mjds):
    el = float(obs.ecliptic_latitude_deg(float(mjd), ra_rad, dec_rad))
    sl = float(obs.helio_ecliptic_longitude_deg(float(mjd), ra_rad, dec_rad))
    image = zodi_readout(
        zodi,
        optical_path,
        prng_keys[i],
        start_time_jd=float(mjd) + 2_400_000.5,
        exposure_time_s=exposure_s,
        wavelength_nm=wavelength_nm,
        bin_width_nm=bin_width_nm,
        ecliptic_lat_deg=el,
        solar_lon_deg=sl,
    )

A useful pattern is to run a single target through a full year, with the L2-halo trajectory and the telescope-to-target pointing vector overlaid for visual context, and a four-target grid version for visual validation against the expected phase relationships across the ecliptic.

What changes through the year#

The spatial pattern of the zodi image on the detector is constant up to a scalar, because zodi_rate multiplies a uniform sky brightness by the coronagraph’s sky_trans map and only the scalar Leinert factor varies with epoch. The dark-hole region of the detector, where sky_trans is approximately zero, stays dark all year, and the off-axis regions where sky_trans is approximately one all brighten and dim together as Δλ_⊙ sweeps. Per-pixel shot noise scales as sqrt(rate * t_exp) of whatever the scalar is at that epoch, so the relative noise structure is also constant up to the same scalar.

This means the natural “annual quantity” to plot is the scalar integrated count rate, not any per-pixel metric, because the per-pixel contrast structure on the detector is fixed.

Validation checklist#

When wiring up a new survey, mission-yield, or paper figure, sanity checks against the geometry catch the most common mistakes. The recommended check picks three on-ecliptic targets at ecliptic longitudes 0°, 90°, and 180°, which in equatorial coordinates are (RA=0°, Dec=0°), (RA=90°, Dec=+23.44°), and (RA=180°, Dec=0°). Compensating for obliquity is what dictates the middle target’s non-zero declination, as explained in the Local zodi + telescope geometry doc. A year-long zodi_readout loop over each of those targets and the argmax day of the integrated count rate will show three peaks separated by roughly 92 days, all of comparable amplitude, because each target undergoes solar conjunction on a different date but with similar zodi brightness at conjunction.

A fourth target placed at high ecliptic latitude, for example (RA=0°, Dec=+60°), serves as the flat-baseline control. The peak integrated count rate is roughly 50 to 100 times smaller than the on-ecliptic peaks, and no frame falls into the NaN unobservable window because the line of sight never approaches the Sun.

Simulating zodi with a telescope’s orbit

Contents

Simulating zodi with a telescope’s orbit#

The composition#

Year-long simulations#

What changes through the year#

Validation checklist#

See also#