Simulations by the IPAC/STScI/Goddard Group

WFI Grism Simulations in Support of the Galaxy Redshift Survey


These simulations demonstrate the usefulness of the Roman Space Telescope for grism observations of high-redshift galaxies, specifically in mapping the distribution of emission-line galaxies, allowing the measurement of cosmic expansion history through the use of baryon acoustic oscillations. The telescope, instrument, and other parameters assumed for these simulations are given in the accompanying table.

Generating the Sky Scene

Input lists of galaxies with different fluxes, sizes, sky coordinates, and spectral energy distributions are created together with emission lines. These lists are generated both by converting grism observations taken by the Hubble Space Telescope WFC3 camera and also by generating galaxies using emission line luminosity functions. These lists are then processed through aXe and aXeSIM, applying the predicted pixel scale, throughput, wavelength resolution, and spectral sensitivity of the Roman Space Telescope grism. The same fields can be simulated at severaldifferent suggested roll angles, allowing for the examination of the potential issues of contamination and spectral overlap in the Roman Space Telescope grism data. As input one needs the Roman Space Telescope effective area and sensitivity as a function of wavelength and filter.

Simulated images from the Roman Space Telescope Wide Field Instrument. The first image shows the direct image (F110W) of a randomized list of objects taken from real detections in the HST/WFC3 images of the WISP (WFC3 Infrared Spectroscopic Parallel; Atek et al. 2010) survey. The second image shows the simulated grism image of the same field. The predicted image distortion, wavelength dispersion, and alignment relative to the direct image have been applied. The image is roughly 25% of a single full 4088x4088 detector and contains approximately 1000 galaxies down to H(AB)=25.

Telescope and Instrument Parameters Used By Grism Simulations
Mirror diameter (m) 2.4
Plate scale (arc-sec/pix) 0.11
Dispersion (Å /pix) 10.4-11.4 (position dependent)
Background flux (e/s/pix) 0.4-0.8
Exposure time (sec) 347 (1 exposure)
2100-2500 (combined exposures)

Work is underway at IPAC/Caltech to create a more advanced simulation pipeline with the ability to generate random objects with realistic morphologies, luminosity functions, and redshift distribution. STScI is working on alternate analysis tools based on linear and forward modeling and on other inputs to the sky scenes, based on existing HST grism programs that cover a cross-section of the different GO science ideas. NASA/Goddard Space Flight Center is producing a semi-automated emission line measurement tool. Below we give some examples of the ongoing simulations.

Realistic Sky Simulations

We present a simulation done for the SCA1 detector intended to be a realistic depiction of the sky. It includes emission line objects inserted based on observed Hα luminosity functions (from Colbert et al. (2013) and known distributions of equivalent width, stars based on stellar luminosity functions and a model (thin/thick disk+halo) of stellar density (the galactic stellar density model is based on model of Juric et al. 2008, mostly using the parameters of Chang et al. 2010, and the stellar luminosity function comes from Just et al. 2015 and background galaxies based on measured near-infrared number counts. All the significant galaxy parameters – major axis, axial a/b ratio, J-H color – have been chosen using distributions derived from HST WFC3 observations.

Each image is roughly half of a full Roman Space Telescope detector, 7.5 × 3.75 arc-minutes in size, and contain about 3,300 individual objects, although most are too faint to produce a detectable spectrum in the grism images. The simulated exposure times are 400s for the direct image and 2100s for the grism images. The grism integration time was chosen to represent a combination of six shorter 350s integration images, as presently proposed for individual exposures in the High Latitude Survey (HLS). Similarly, the noise is based on an assumed typical near-infrared background for the HLS. The Galactic coordinates used for this simulation and its star count model were latitude=-49 and longitude=149 (R.A. = 2h, Dec. = +10.0 deg). The star images are generated from a PSF created using STScI's WebbPSF. The galaxies are simple elliptical gaussian models.

The purpose of this set of simulations is to test the effects of crowding and contamination, in particular as a function of the grism dispersion. The top image is a simulated H-band direct image, but the next two images are the simulated grisms assuming 10.8 Å/pix (middle) and 8 Å/pix (bottom) dispersion. Both use the same spectral coverage: 1 to 1.9 microns.

Galaxy Parameter Simulations

The grism simulations are also being used to test ranges of specific galaxy parameters. In this simulation we test a range of Hα emission line fluxes (5×10-17 – 5×10-16 ergs/s/cm2) at different redshifts (z=0.55 – 1.9), while keeping the equivalent width constant (100 Å, rest wavelength). The galaxy size (major axis) is varied using known size distributions for emission line galaxies derived from HST WFC3 observations, but the same simple gaussian elliptical shape and orientation is used for all simulated objects. For these galaxy parameter tests, we lay the emission line galaxies out in a non-overlapping grid to prevent the issue of contamination from confusing the analysis. We ran these simulations at two different potential dispersions that are being considered (8 and 10.8 Å/pix) and then measured the emission lines for the new simulated fields. Below we present some of the resulting analysis: line-centroiding accuracy (in terms of measured redshift) and emission line recovery as a function of signal-to-noise for each of the two dispersions.

Top: This is the direct image, with all the emission line galaxies laid out in a grid intended to avoid spectral overlap. Each image is roughly 7.5x3.75 arc-minutes in size (half the size of a full Roman Space Telescope detector). The full image contains roughly 550 emission line galaxies. There are no "background" or non-emission line galaxies included. The simulated exposure time is 400 seconds. Middle: The dispersed grism image at the nominal 10.8 Å/pixel wavelength resolution, running from roughly 1 to 1.9 microns. The exposure time is 2100s, intended to approximate the typical depth of the High Latitude Survey after the combination of multiple shorter individual exposures. The main emission line visible is Hα, although a weaker set of [OIII] lines (fixed at half the flux of Hα) can also be seen. Bottom: The exact same set of emission line galaxies at the exact same locations on the detector as above, but done for the 8 Å/pixel wavelength resolution.

Below are two sets of analysis plots made from the nominal (top) and high dispersion (below) simulations. Each set presents the line centroiding accuracy in terms of redshift ([zmeasured-zinput]/[1+zinput]) and the recovery percentage as functions of Hα line flux. The empty circles are all the simulated Hα emission lines, while the red, green and blue points are the same emission lines after S/N = 3, 5 and 7 cuts, respectively. The dotted lines represent the redshift accuracy goal of 0.1% that has been discussed for the main redshift survey. Recovery percentage is the number of objects at each Hα line flux after the signal-to-noise cut has been made, divided by the total number or input emission line objects at that flux. While similar to a completeness, the recovery percentage only takes into account the signal-to-noise of the line and does not (at this point) include the chance the emission line might be missed altogether, due to contamination or automated line finding failure. Nevertheless, it gives a good indication of the general depth the survey can reach. We note that the sky background assumed is for a typical HLS field, so some HLS fields will reach fainter emission lines than presented here.


aXeSIM was adopted from HST/WFC3 spectral simulations to Roman/WFI. The simulation package uses the same components as used by aXe for the extraction of slitless spectra and thus aims at spectrophotometric integrity - useful for observation design but essential for the quantitative assessment of slitless data.

Work is underway at IPAC and STScI to add more advanced instrumental and observational effects, such as intrapixel capacitance, charge diffusion, cosmic rays, jitter, and others to aXeSIM, and to include more flexibility in aXe/aXeSIM, required by the large field of view and complex focal plane of the Roman Space Telescope.

For more information about aXe see Kümmel et al. (2009).

Output From the Simulations

The output from aXeSIM simulations will always consist of a simulated 2D slitless dispersed image with spectra of the simulated objects. For grism images, the simulations can include several dispersed orders. aXeSIM will also produce the direct image associated with the slitless image. Sky background and random noise (readout and photon noise from background and objects) can be added to the output images.