Project Topics
Get Complete Project Material File(s) Now! »
The NIRSpec Instrument Performance Simulator
Early in the development of NIRSpec, the need for an instrument simulator was realized, given the inherent complexity of a multi-object spectrograph, with all other operation modes on top. In the frame of the project, the Centre de Recherche Astrophysique de Lyon (CRAL) has developed the NIRSpec Instrument Performance Simulator (IPS) software (Gnata, 2007; Piquéras et al., 2008, 2010). Its primary functions are to assess the instrument specifications, verify the performance, and do end-to-end simulations of calibration and scientific exposures. Besides, it serves to create realistic input data for processing tools, as the Instrument Quick Look Analysis and Calibration software (IQLAC, Gerssen et al., 2008) or the final NIRSpec data reduction pipeline.
To simulate the propagation of light, the instrument is divided into optical mod-ules, mostly defined by the single TMAs (COM + FORE, COL, CAM, CAL, OTE, IFU FORE, IFU POST), and the functional parts as filters, slits, and dispersers. The IPS uses a novel approach combining Fourier optics for the diffractive effects, geo-metrical coordinate transforms between the key optical planes, and simple efficiency calculations for the radiometry. These elements produce noiseless electron rates as a first main simulation product. The data contains the number of electron per second in each detector pixel, without any photon or readout noise, separate for each disperser order.
Slit tilt implementation
The previously illustrated recipe is capable of producing very accurate coordinate transforms for the OTE and inside NIRSpec. In all modules except the FORE optics, there is no or very little rotation, and no chromatic dependence of the coefficients. However, in combination with the dispersing element in the spectrograph, another effect occurs: the tilt of the slit image in the spectra.
In imaging mode, the spectrograph transform from slits to the detector can locally be approximated by a paraxial system. A slit itself may be rotated in the MSA plane by ϑrot, so that the spatial axis along the slit yslit and the MSA yMSA are not collinear. The slit image will then be rotated by the angle ϑFPA ` ϑrot. Assuming paraxial systems, ϑFPA is the sum of the rotations in the collimator and camera ϑFPA “ ϑCOL ` ϑCAM, and can be derived from the tilt of the MSA x-axis with ϑFPA “ ϑx “ arctan ˆ BxMSA { BxMSA ˙ , yFPA xFPA B B which is the same angle for the MSA y-axis xFPA yFPA ϑy “ arctan ˆ ByMSA { ByMSA ˙ .
The local orientations are more complex in spectrographic mode, where the off-plane design of NIRSpec causes a curvature of the spectral lines (Schroeder, 2000, chapter 14). The curvature radius is very large compared to the length of a single slit, therefore it can be approximated with a slit tilt as shown in Figure 2.3. The single transforms for COL and CAM remain valid, but when including the disperser, the partial derivatives are not orthogonal any more, i.e. ϑx ‰ ϑy. The slit tilt is a shear along the projected MSA x-axis by an angle ϑslit, which can be calculated as the difference between the tilts of the y- and x-axes to ϑslit “ ϑy ´ ϑx. On the FPA, the image of the slit is finally tilted by ϑy ` ϑrot.
In the IPS this has consequences for all simulation types described in section 1.4, especially for continuum spectra. By collapsing the PSF in the FPA x-direction, all position information along this axis is lost and the resulting spectra would be straight in the FPA y-direction. The rotations of the distortion ϑx and the slits ϑrot are typically much smaller than the slit tilt. Therefore the collapsed vector is rebinned along the x-direction with a shear rate of ΔxFPA “ ΔyFPA sinpϑy ` ϑrotq, where ΔyFPA is measured from the slit center trace location.
Implementation for NIRSpec
The Fourier algorithm uses the system of optical modules in the IPS (section 1.4). They are characterized with their pupil diameter D, and nominal entrance and exit focal lengths for both axes. These can be calculated with the scheme presented in subsection 2.2.2.
To define the orientations at the principal optical planes, we used the official NIRSpec optical model, where the reference frames in pupils and focal planes are given (Figure 2.6). We then put together the steps for a FFT-based propagation from one plane to the next, including the coordinate flips and rotations. We also noted the orientation of the z-axis, and how a wavefront with positive phase is oriented locally. This is important to correctly add the single WFE maps at different steps. We tested the newly assembled Fourier module using pupil masks with unique features to detect the orientations of pupils and PSFs. Comparing with a Python-based implementation, we could successfully reproduce the results in the IPS.
To verify the scheme with real instrument data, we used measurements taken during the NIRSpec Demonstration Model (DM) test campaign (Böker et al., 2010). With a pinhole mask at the NIRSpec field stop location, and the detector at the slit plane, the broadband PSFs of the FORE optics were recorded. In this configuration, the FWA stop defines the system pupil. In the DM, the stop is an oversized shape of the OTE primary boundaries, similar to the mask shown in Figure 2.6(b) without the three spider arms.
Model verification and transformation to cold
To a certain extent, the model is intrinsically verified in the alignment process, when the subsystems were moved to match the measured distortion with the model output. Apart from the geometry, the optimization included a minimization of the average measured wavefront error in the field of view. However, that influence is not reflected in the model, as the wavefront errors were not used for the alignment prediction. To verify this particular model output we therefore compared the simulated wavefront errors with the measurements in the finally determined focal planes.
Relation between wavefronts and coordinate transforms
In the end, the verification of coordinate transforms will be done with centroids from spatially resolved images of pinholes or slits. This means that the PSFs from the Fourier propagation need to have a centered centroid, and the data from the raytracing needs to refer to the PSF centroid. To first order, this can be achieved by removing the wavefront tilt, and to create the coordinate transforms from spot diagram centroids. While the first is done routinely and fast, the second would need extensive raytracing, which is generally slow in the complex as-built models.
To enable a faster creation of the distortion grids, we compared raytracing results of single chief rays and full spot diagrams, with and without mirror surfaces. The test was done with an as-built model of the NIRSpec demonstration model and the respective mirror data, but is also applicable to the flight model. We created distortion data at the MSA plane on a 11×11 grid covering the full FOV with N “ 121 points. The spot centroid position with mirror surface maps served as the reference positions pxr , yrq, and the difference to each of the three other positions was characterized with the mean radial error 1 N 1 N Δr “ Δri “ bpxi ´ xr iq2 ` pyi ´ yr iq2 N i“1 N i“1 ÿ ÿ. and its standard deviation g g σr “ 1 N Δri2 1 N xi xr i 2 1 N yi yr i 2. N i“1 “ N i“1p ´ ` N i“1p ´ q f f q f ÿ f ÿ ÿ e e. The results are listed in Table 3.1. At the MSA, 2 μm correspond approximately to 5 mas, which is the required accuracy of the coordinate transforms. The mirror surface interferograms hardly influence the spot diagram position (Δr “ 1.76 μm), but rather enlarge the spot sizes by 50–100% (not shown). This is in line with the design of the as-built model, where the low-frequency surface shapes are put in the mirror description, while the surface maps contain the high-frequency components which mostly cause straylight and an enlargement of the geometrical spot. On the other hand, when excluding the maps and tracing only the chief ray, the difference
Table of contents :
1 Introduction
1.1 The James Webb Space Telescope
1.2 Overview of the NIRSpec instrument
1.3 Optical layout of JWST and NIRSpec
1.4 The NIRSpec Instrument Performance Simulator
1.5 Goal and structure of this thesis
2 IPS software verification
2.1 Introduction
2.2 Revision of coordinate transforms
2.2.1 General formalism
2.2.2 Derivation of transform parameters
2.2.3 Slit tilt implementation
2.3 Revision of Fourier propagation
2.3.1 General application
2.3.2 Geometrical orientation
2.3.3 Single propagation steps
2.3.4 Implementation for NIRSpec
2.3.5 Sampling of PSFs and wavefront errors
2.3.6 Required sampling for NIRSpec
3 NIRSpec model description
3.1 Model data overview
3.2 Subsystem and telescope data
3.3 NIRSpec as-built optical model
3.3.1 Motivation
3.3.2 Model description
3.3.3 Model verification and transformation to cold
3.3.4 Relation between wavefronts and coordinate transforms
3.3.5 Data extraction for the IPS model
4 Science software tools
4.1 Science data interface
4.1.1 Motivation
4.1.2 Object positioning
4.1.3 Object input file types
4.1.4 Object separation criteria
4.1.5 Technical implementation
4.2 Spectrum extraction pipeline
4.2.1 Purpose and scope
4.2.2 Software implementation
4.2.3 Spectrum extraction operations
5 NIRSpec model verification
5.1 Motivation
5.2 Instrument geometry
5.2.1 Initial data and manual tuning
5.2.2 Model optimization
5.2.3 GWA tilt sensor integration in extraction
5.3 Test of IPS with tuned instrument model
5.4 Calibration Light Source spectra
5.5 Instrument efficiency
5.5.1 Filter transmissions
5.5.2 Grating efficiencies
5.5.3 Overall instrument throughput
5.5.4 IFU throughput
5.6 Limitations of simulations
6 NIRSpec science simulations
6.1 Multi-object deep field
6.1.1 Introduction
6.1.2 Observation scene creation
6.1.3 Galaxy shapes
6.1.4 Galaxy spectra
6.1.5 Exposure simulation
6.1.6 Spectrum extraction
6.1.7 Results and discussion
6.2 Exoplanetary transits
6.2.1 Introduction
6.2.2 Host star brightness limits
6.2.3 Noise from pointing jitter
6.2.4 Effective integration times
6.2.5 Signals and noise
6.2.6 Simulations of HD189733b
6.2.7 Simulations of GJ1214b
6.2.8 Simulations of an Earth-sized planet in the habitable zone
7 Conclusion and outlook
A Additional images
A.1 Verification of the as-built optical model
A.1.1 Measured wavefront errors at the FPA
A.1.2 Simulated wavefront errors at the FPA
A.1.3 Wavefront error residuals at the FPA
A.2 NIRSpec model data
A.2.1 Wavefront error maps
A.2.2 Efficiencies
B Publications
B.1 Overview
B.2 Verification of the IPS with demonstration model data
B.3 First simulation of a JWST/NIRSpec observation
Bibliography
