The JWST is known as the successor of the Hubble Space Telescope. It
is an infrared telescope operating from 0.6 to 28m; diffraction
limited at 2m. JWST will be constructed using deployed structures
since the maximum size of the rocket shrouds will not accommodate the
full mirror or sunshade diameter. That is why the primary mirror is
not a circular one but a 6.5m-segmented hexagonal mirror. JWST launch
is planed for late 2011.
NIRSpec (Near Infrared Spectrograph) is one
of the 3 instruments onboard JWST. This instrument will be the
multi-object spectrograph of JWST and is under ESA responsibility.
NIRSpec will be able to observe simultaneous spectra of more than 100
objects in a 9 square arcminute field of view in the 0.6 to 5m
band, thanks to an array of micro-shutters (MSA). Users will be able
to select the objects of which they want to get spectra by closing
or opening each micro-shutter.
We first need to assess the combined impact of diffraction and aberrations on the instrument performance. That is why we need to compute the PSFs (point spread functions) on the micro-shutters array and on the detector, and the image on the grating. In order to be able to use NIRSpec as a photometric instrument, we also have to assess the diffraction losses at the level of both the micro-shutters and the gratings. Simulated NIRSpec science and calibration exposures will also be very useful to validate calibration and operation procedures and to develop the data reduction software.
Fourier optics and more especially Fresnel Approximation is used to model diffraction effects. Fast Fourier transformations (FFTs) are used to progress from the telescope pupil plane through the MSA plane, the gratings plane and the detector image plane (See Figure 1).
Apertures are simulated using binary matrices called intensity masks (1 when the point is in the aperture, 0 otherwise). In our model, we do not have to assume that we have perfect optics because we take geometrical aberrations effects into account by introducing phase masks before computing FFTs. These phase masks correspond to the wavefront errors produced by the real instrument. Low frequency wavefront errors due to optical aberrations are described using Zernike polynomials. Roughness of the optics introduces high frequency wavefront errors, which are described by a random phase mask (Born and Wolf 1989). To obtain good sampling of both the pupils and PSFs we use zero padding which involves performing FFTs on a quite large matrix (above 6000*6000).
We have used Zemax (a commercial program that can model and analyze optical systems) to extract all the optical parameters (e.g. focals, aperture or diameters) and Zernike coefficients needed to simulate NIRSpec. We have written some user-friendly macros in the Zemax Programming Language in order to put all these parameters into a formatted ASCII file in a very simple way. By running these macros, one gets an ASCII file describing the system per wavelength and per point in the field. Only the format of these files really matters and they could have been obtained another way.
Zemax only supports some diffraction calculations and is not usable alone to assess the impact of diffraction in NIRSpec. To simulate propagation from pupil to image plane our code running under GNU/Linux takes ASCII files computed by Zemax as input and performs masks and FFT calculations (see figure 2). This code has been designed with a modular approach: It is a Fourier Optics library easy to reuse as a black box in other simulation software because there no hardcoded parameters describing the optical system and because one can simulate all of the most common shapes of optical apertures. Complex FITS images store the amplitude and phase repartition. Our simulation code is written in C and makes use of the Euro3D development environment and I/O libraries developed by Arlette Pécontal-Rousset (Pécontal-Rousset 2004). To compile and use it you only need the standard GNU tools and two libraries. The first one is called GSL (GNU Scientific Library) which is a numerical library written in ANSI C. The other one is FFTW which is a very efficient C library for computing the Discrete Fourier Transform (DFT) in one or more dimensions (and more especially in two dimensions in our simulation code), of arbitrary input size, and of both real and complex data.
Using this simulation C code, we have computed PSFs on the MSA, images on the gratings plane and PSFs at the detector level. PSFs are, however, only raw results and that is why we have developed tools to assess the instrument optical performances. These ``slit tools'' are written in C. Given an output FITS image computed by the simulator, it is for example straightforward to plot the encircled (or ensquared) energy using slit tools. If we assume we have a random distribution of point sources whose PSFs centers are in a given rectangular area inside an open micro-shutter (180m in spatial direction times 80m in spectral one), the slits tools are able compute the average slit losses at this level. This study led us to define an available micro-shutter area inside which the losses are not too great. This area will play a leading role in defining the operation scenarios.
We have also proven that the diffraction losses at the plane of the gratings slightly depend on the way the PSFs are centered in the open micro-shutter (See Figure 3). This fact proves that we do need an accurate model of NIRSpec to be able to extract photometric well calibrated data from the exposures.
Simulated PSFs at the detector level have enabled us to compute the spectral and spatial resolution of NIRSpec assuming for now that the detector is perfect.
Pécontal-Rousset, A. 2004, this volume, 491
Wyant, J. C., & Wasserman, L. H. 1992, in Applied Optics and Optical engineering, XI,chap 1
Born, M., & Wolf, E., in Principles of Optics: Electromagnetic Theory of Propagation, Interference, and Diffraction of Light, 6th ed, New York: Pergamon Press, pp. 459-490